The Cosmic Century

A History of Astrophysics and Cosmology

The twentieth century witnessed the emergence of the disciplines of astrophysics and cos-
mology, from subjects which scarcely existed to two of the most exciting and demanding
areas of contemporary scientific inquiry. There has never been a century in which fundamen-
tal ideas about the nature of our Universe and its contents have changed so dramatically. This
book reviews the historical development of all the key areas of modern astrophysics, linking
the strands together to show how advances have led to the extraordinarily rich panorama of
modern astrophysics and cosmology. While many of the great discoveries were derived from
pioneering observations, the emphasis is upon the development of theoretical concepts and
how they came to be accepted. These advances have led astrophysicists and cosmologists
to ask some of the deepest questions about the nature of our Universe and to stretch our
ability to address them by advanced observation to the very limit. This is a fantastic story,
and one which would have defied the imaginations of even the greatest story-tellers.

m al c o l m longa ir completed his Ph.D. in the Radio Astronomy Group of the Cavendish
Laboratory, University of Cambridge, in 1967. From 1968 to 1969 he was a Royal Society
Exchange Visitor to the Lebedev Institute, Moscow. He has been an exchange visitor to
the USSR Space Research Institute on six subsequent occasions and has held visiting
professorships at institutes and observatories throughout the USA. From 1980 to 1990, he
held the joint posts of Astronomer Royal for Scotland, Regius Professor of Astronomy at
the University of Edinburgh and Director of the Royal Observatory, Edinburgh. He has
been Head of the Cavendish Laboratory, Cambridge since 1997, and was made a CBE in
the 2000 Millennium Honours list. Longair’s primary research interests are in the fields of
high-energy astrophysics and astrophysical cosmology. He has published 15 books and over
250 journal articles on his research work.
For
Deborah
The Cosmic Century
A History of Astrophysics and Cosmology

MALCOLM S. LONGAIR
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo
Cambridge University Press
The Edinburgh Building, Cambridge CB2 2RU, UK
Published in the United States of America by Cambridge University Press, New York

www.cambridge.org
Information on this title: www.cambridge.org/9780521474368


C Cambridge University Press 2006

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 2006

Printed in the United Kingdom at the University Press, Cambridge

A catalogue record for this publication is available from the British Library

ISBN-13 978-0-521-47436-8 hardback

ISBN-10 0-521-47436-1 hardback

Cambridge University Press has no responsibility for the persistence or accuracy of URLs for
external or third-party internet websites referred to in this publication, and does not guarantee that
any content on such websites is, or will remain, accurate or appropriate.
 Contents

Preface page x
Acknowledgements xiv

Part I Stars and stellar evolution up to the Second World War 1

1 The legacy of the nineteenth century 3

1.1 Introduction 3
1.2 From Joseph Fraunhofer to Gustav Kirchhoff 4
1.3 The first stellar parallaxes 7
1.4 The invention of photography 9
1.5 The new generation of telescopes 11
1.6 The prehistory concluded 15
Notes to Chapter 1 16

2 The classification of stellar spectra 18

2.1 William Huggins – the founder of stellar astrophysics 18
2.2 The first spectral classification systems 20
2.3 The Harvard classification of stellar spectra 21
Notes to Chapter 2 27

3 Stellar structure and evolution 30

3.1 Early theories of stellar structure and evolution 30
3.2 The origin of the Hertzsprung–Russell diagram 34
3.3 The impact of the new physics 39
3.4 Eddington and the theory of stellar structure and evolution 42
3.5 The impact of quantum mechanics and the discovery of new particles 48
Notes to Chapter 3 50
A3 Explanatory supplement to Chapter 3 51
Notes to Section A3 60

4 The end points of stellar evolution 61

4.1 The red giant problem 61
4.2 White dwarfs 63

v
vi Contents

4.3 Supernovae and neutron stars 66

Notes to Chapter 4 68
A4 Explanatory supplement to Chapter 4 69
Notes to Section A4 73

Part II The large-scale structure of the Universe, 1900–1939 75

5 The Galaxy and the nature of the spiral nebulae 77

5.1 ‘Island universes’ and the cataloguing of the nebulae 77
5.2 The structure of our Galaxy 79
5.3 The Great Debate 84
5.4 Hubble and the Universe of galaxies 87
5.5 The discovery of Galactic rotation 89
5.6 Interstellar matter and extinction by dust 91
5.7 The Galaxy as a spiral galaxy 92
Notes to Chapter 5 95
A5 Explanatory supplement to Chapter 5 96

6 The origins of astrophysical cosmology 100

6.1 Physical cosmology up to the time of Einstein 100
6.2 General relativity and Einstein’s Universe 102
6.3 De Sitter, Friedman and Lemaı̂tre 107
6.4 The recession of the nebulae 109
6.5 The Robertson–Walker metric 113
6.6 Milne–McCrea and Einstein–de Sitter 114
6.7 Eddington–Lemaı̂tre 118
6.8 The cosmological problem in 1939 118
Notes to Chapter 6 120

Part III The opening up of the electromagnetic spectrum 123

7 The opening up of the electromagnetic spectrum and the new astronomies 125
7.1 Introduction 125
7.2 The discovery of subatomic particles and cosmic rays 130
7.3 Radio astronomy 136
7.4 X-ray astronomy 142
7.5 Gamma-ray astronomy 147
7.6 Ultraviolet astronomy and the Hubble Space Telescope 150
7.7 Infrared astronomy 153
7.8 Optical astronomy in the age of the new astronomies 162
7.9 Other types of astronomy 170
Notes to Chapter 7 170
 Contents vii

Part IV The astrophysics of stars and galaxies since 1945 173

8 Stars and stellar evolution 175

8.1 Introduction 175
8.2 Nucleosynthesis and the origin of the chemical elements 175
8.3 Solar neutrinos 179
8.4 Helioseismology 183
8.5 Evolving the stars 188
8.6 The discovery of neutron stars 192
8.7 X-ray binaries and the search for black holes 197
8.8 Radio pulsars and tests of general relativity 201
8.9 The search for gravitational waves 202
8.10 Supernovae 204
Notes to Chapter 8 211
A8 Explanatory supplement to Chapter 8 212
Notes to Section A8 215

9 The physics of the interstellar medium 216

9.1 The photoionisation of the interstellar gas 216
9.2 Neutral hydrogen and molecular line astronomy 217
9.3 The multi-phase interstellar medium 222
9.4 The formation of stars 225
9.5 Extrasolar planets and brown dwarfs 232
9.6 Cosmic-ray astrophysics and the interstellar medium 236
Notes to Chapter 9 241
A9 Explanatory supplement to Chapter 9 242
Note to Section A9 243

10 The physics of galaxies and clusters of galaxies 244

10.1 The galaxies 244
10.2 Dark matter in galaxies 248
10.3 The dynamics of elliptical galaxies 253
10.4 The large-scale distribution of galaxies 253
10.5 The physics of clusters of galaxies 259
Notes to Chapter 10 266
A10 Explanatory supplement to Chapter 10 267
Notes to Section A10 268

11 High-energy astrophysics 269

11.1 Radio astronomy and high-energy astrophysics 269
11.2 The discovery of quasars and their close relatives 271
11.3 General relativity and models of active galactic nuclei 275
11.4 The spectroscopy of active galactic nuclei 279
11.5 The masses of black holes in active galactic nuclei 283
viii Contents

11.6 Non-thermal Phenomena in Active Galactic Nuclei 290

11.7 The γ -ray bursts 300
Notes to Chapter 11 304
A11 Explanatory supplement to Chapter 11 306
Notes to Section A11 316

Part V Astrophysical cosmology since 1945 317

12 Astrophysical cosmology 319

12.1 Gamow and the Big Bang 319
12.2 Steady state cosmology 323
12.3 The counts of radio sources 326
12.4 The helium problem 328
12.5 The discovery of the cosmic microwave background radiation 329
12.6 The helium problem revisited 331
Notes to Chapter 12 335
A12 Explanatory supplement to Chapter 12 336
Note to Section A12 339

13 The determination of cosmological parameters 340

13.1 Sandage and the values of H0 and q0 340
13.2 Hubble’s constant 343
13.3 The age of the Universe, T0 346
13.4 The deceleration parameter, q0 349
13.5 The density parameter, 0 357
13.6 Summary 361
Notes to Chapter 13 361
A13 Explanatory supplement to Chapter 13 362

14 The evolution of galaxies and active galaxies with cosmic epoch 365
14.1 The cosmological evolution of active galaxies 365
14.2 The counts of galaxies 376
14.3 The Lyman-α clouds 379
14.4 The abundances of elements in Lyman-α absorbers 383
14.5 The Lyman-break galaxies 384
14.6 The global star-formation rate 387
14.7 Conclusion 391
Notes to Chapter 14 391

15 The origin of galaxies and the large-scale structure of the Universe 392
15.1 Gravitational collapse and the formation of structure
in the expanding Universe 393
15.2 The thermal history of the Universe 394
 Contents ix

15.3 The development of small perturbations with cosmic epoch 398

15.4 The adiabatic and isothermal scenarios for galaxy formation 402
15.5 Hot dark matter – neutrinos with finite rest mass 405
15.6 Cold dark matter and structure formation 406
15.7 Biassing 410
15.8 Reconstructing the initial power spectrum 411
15.9 Variations on a theme of cold dark matter 413
15.10 Fluctuations in the cosmic microwave background radiation 415
15.11 The discovery of Sakharov oscillations 422
15.12 The determination of cosmological parameters 423
15.13 The post-recombination Universe 427
Notes to Chapter 15 433
A15 Explanatory supplement to Chapter 15 436
Note to Section A15 438

16 The very early Universe 439

16.1 The big problems 439
16.2 The limits of observation 442
16.3 The anthropic cosmological principle 443
16.4 The inflationary Universe and clues from particle physics 444
16.5 The origin of the spectrum of primordial perturbations 447
16.6 Baryogenesis 448
16.7 The Planck era 449
Notes to Chapter 16 451

References 453
Name index 510
Object index 517
Subject index 519
 Preface

How this book came about

The origin of this book was a request by Brian Pippard to contribute a survey of astrophysics
and cosmology in the twentieth century to the three-volume work that he edited with Laurie
Brown and the late Abraham Pais, Twentieth Century Physics (Bristol: Institute of Physics
Publishing and New York: American Institute of Physics Press, 1995). This turned out to be
a considerable undertaking, my first draft far exceeding the required page limit. By drastic
editing, I reduced the text to about half its original length and the survey appeared in that
form as Chapter 23 of the third volume.
I was reluctant to abandon all the important material which had to be excised from the
published survey and was delighted that the Institute of Physics agreed to my approaching
Cambridge University Press about publishing the full version. The Press were keen to take
on the project, with some further expansion of the text and, in particular, with a number
of explanatory supplements to chapters where a little simple mathematics can make the
arguments more convincing for the enthusiast. I have also made liberal use of references to
my other books, where I have already given treatments of topics covered in this book. The
result has been a complete rethink of the whole project and an expansion of the text by a
factor of five as compared with the original published version.
As when I was writing my book, Theoretical Concepts in Physics, 2nd edn (Cambridge:
Cambridge University Press, 2003), I have learned so much during the preparation of this
book that I wish I had known when I was learning these subjects. The historical material
provides real physical insight into the intellectual infrastructure of astrophysics and cos-
mology, and it is saddening that it is not more easily accessible to the student, researcher
and lecturer. Even worse, in many cases, the folk-tales of astrophysics and cosmology have
acquired mythical status, which do not necessarily coincide with how many of the great
insights came about.
The original sub-title of the book was to be A History of Twentieth Century Astrophysics
and Cosmology, reflecting its origin as a chapter of Twentieth Century Physics. As pointed
out by the CUP editors, the story runs right up to 2005 and furthermore astrophysics and
cosmology in their modern physics-related guises scarcely existed before 1900. Therefore,
it seemed much more appropriate to drop the words ‘Twentieth Century’ from the subtitle.

Warnings and apologies

The magnitude of the task I set myself only became apparent once I was well into the writing
of the final text. It is folly to pretend to completeness, or to hope to make reference to all

x
Preface xi

the important contributions of so many distinguished colleagues. Therefore, I have had to

be selective and am only too aware of the limitations of what is published here. Even worse,
I do not believe it is possible to write a wholly objective history of as complex a field as the
development of astrophysical and cosmological understanding over the twentieth century.
I have tried to be fair in my assessments of what is of lasting importance, but this is bound
to be a subjective process.
Equally significant is the fact that I am one of the lucky generation who began research in
the early 1960s when the whole astrophysical and cosmological landscape changed forever
from one dominated by optical astronomy to one of multi-wavelength astronomy in which
quite different types of astrophysics began to dominate much of the scene. The influx of
physicists into astrophysics from that time onwards has been one of the most important
features of this story and I write from that perspective. One of the most revealing aspects
of the story told in this book is the close link between developments in physics and their
impact upon astrophysics and cosmology, and vice versa, and the fact that this symbiosis
has been at the heart of these disciplines from the beginning. I find it revealing that the
author index of this book includes references to large numbers of physicists as well as to
astrophysicists and cosmologists.
Although the author index includes about 1000 individuals, I am aware that it omits
many who have made important contributions, sometimes simply because they were not the
first author on the paper. The problem of attributing credit to individuals has become very
much more difficult during the last few decades of the twentieth century when many of the
key papers can involve tens or hundreds of authors. This reflects the fact that many of the
large space- and ground-based projects can now involve very large numbers of individuals,
and so the credit should go to the project team rather than to individual scientists. I have
made value judgements about whom to credit in these cases, often giving up and simply
giving the detailed authorship in the bibliography. I hope my colleagues will understand the
impossibility of doing justice to everyone involved.
I am bound to repeat the disclaimer that I am not a professional historian, and far less
a philosopher, of science. My objectives in this book are astrophysical and cosmological,
specifically to track the intellectual history of the development of astrophysics and cos-
mology through what has been one of the most extraordinary centuries in the history of
scientific endeavour. Therefore, this is not a history of astronomy per se, but astronomy
viewed through the mirror of physical understanding. Numerous controversial topics will
be treated in this history, but my approach has been to concentrate upon the astrophysical
and cosmological issues rather than the more sensational aspects of the story.

Secondary literature

There is an enormous wealth of fascinating material on the history of twentieth-century

astrophysics and cosmology which I have had to condense into a modest space. In
the references, I have given complete bibliographical citations to all the original arti-
cles discussed. In preparing this book, I have found the following volumes particularly
helpful:
xii Preface

Bernstein, J. and Feinberg, G. (1986). Cosmological Constants: Papers in Modern Cos-

mology (New York: Columbia University Press). This volume includes translations of
many of the seminal papers in cosmology.
Bertotti, B., Balbinot, R., Bergia, S. and Messina, A., eds (1990). Modern Cosmology in
Retrospect (Cambridge: Cambridge University Press).
Bondi, H. (1960). Cosmology, 2nd edn (Cambridge: Cambridge University Press).
Gillespie, C. C., ed. (1981). Dictionary of Scientific Biography (New York: Charles Scrib-
ner’s Sons).
Gingerich, O., ed. (1984). The General History of Astronomy, Vol. 4. Astrophysics and
Twentieth-Century Astronomy to 1950: Part A (Cambridge: Cambridge University
Press).
Harrison, E. (2001). Cosmology: The Science of the Universe (Cambridge: Cambridge
University Press).
Hearnshaw, J. B. (1986). The Analysis of Starlight: One Hundred and Fifty Years of Astro-
nomical Spectroscopy (Cambridge: Cambridge University Press).
Hearnshaw, J. B. (1996). The Measurement of Starlight: Two Centuries of Astronomical
Photometry (Cambridge: Cambridge University Press).
Kragh, H. (1996). Cosmology and Controversy: The Historical Development of Two Theo-
ries of the Universe (Princeton: Princeton University Press).
Lang, K. R. and Gingerich, O., eds (1979). A Source Book in Astronomy and Astrophysics,
1900–1975 (Cambridge, Massachusetts: Harvard University Press). This volume con-
tains reprints of and brief historical introductions to many of the original articles pub-
lished between 1900 and 1975 referred to in this survey. All the articles are translated
into English.
Learner, R. (1981). Astronomy through the Telescope (London: Evans Brothers Limited).
Leverington, D. (1996). A History of Astronomy from 1890 to the Present (Berlin: Springer-
Verlag).
Martínez, V. J., Trimble, V. and Pons-Bordeı́a, M. J., eds (2001). Historical Development of
Modern Cosmology, ASP Conference Series, vol. 252 (San Francisco: ASP).
North, J. D. (1965). The Measure of the Universe (Oxford: Clarendon Press).
A key resource for all aspects of astrophysics and cosmology is the series entitled Annual
Review of Astronomy and Astrophysics, which first appeared in 1963. These reviews are
authoritative and represent understanding at the year of the review. The more recent volumes
include autobiographical essays by a number of the key personalities who appear in this
book. For many topics, I have given references to authoritative books and reviews in the
Notes to each chapter.
I have also assumed some familiarity with astronomical terminology. For more details
of the terminology and reviews of many areas of astronomy, the following can be recom-
mended:
Nicholson, I. (1999). Unfolding our Universe (Cambridge: Cambridge University Press).
This is an elementary text, but it includes a large amount of useful background material
on all aspects of astronomy.
Preface xiii

Maran, S. P., ed. (1992). The Astronomy and Astrophysics Encyclopedia (New York: Van
Nostrand Reinhold, and Cambridge: Cambridge University Press).
Murdin, P., ed. (2001). Encyclopaedia of Astronomy and Astrophysics (4 vols) (Bristol and
Philadelphia: Institute of Physics Publishing, and London, New York and Tokyo: Nature
Publishing Group).
 Acknowledgements

My thanks are warmly accorded to the many friends and colleagues who have helped in
numerous ways in bringing this book into being. Clearly, the first set of thanks goes to Brian
Pippard who started the whole project off. Many colleagues provided advice about my
chapter ‘Astrophysics and cosmology’ in Twentieth Century Physics. They included Tony
Hewish, David Dewhirst and the late Peter Scheuer. Once the book project was under way,
John Hearnshaw kindly read the whole long first draft of my chapter and made many key
corrections and observations about what I had assembled. I also thank him for his excellent
hospitality in Christchurch, New Zealand, where the final proofreading and finishing touches
took place. The late Sir William McCrea also kindly read the first draft of this book and
made many valuable comments about the history as he had experienced it. I have quoted
from the wonderful letter he wrote to me on 11 December 1993 in the text.
The stimulus for looking deeper into the history of the technology of modern astrophysics
and cosmology was provided by the invitation to participate in the excellent Valencia con-
ference organised by Vicent Martı́nez, Virginia Trimble and Maria-Jesus Pons-Bordeı́a
entitled Historical Development of Modern Cosmology. Michael Hoskin kindly reviewed
the contents of my paper for that meeting. The invitation by Wendy Freedman to celebrate
the centenary of the foundation of the Carnegie Observatories by providing a brief history
of twentieth-century cosmology for the symposium Measuring and Modelling the Universe
also contributed to the enrichment of the present text.
Special thanks are due to Leon Mestel and John Faulkner for their help with the history
of the understanding of stellar evolution. I have picked the brains of countless colleagues
on the contents of this book – my apologies if I cannot record them all. I also thank many
colleagues for allowing me to quote their birth years in the main text – I hope colleagues
of all ages will take encouragement from the wide range of ages at which individuals have
made seminal contributions to astrophysics and cosmology.
Special thanks are also due to David Green for his help in customising the CUP LATEX
macros to format the book just as I wanted it to appear in its published form.
Particular thanks are due to the following whose help was invaluable in tracking down
many of the obscurer references which have been consulted: Judith Andrews, for her help
in tracking down many of the old references referred to in the bibliography, and the birth
and death years of the individuals mentioned in the text; Gillian Wotherspoon and Nevenka
Huntic of the Rayleigh Library of the Cavendish Laboratory, for help in finding old books
and journals; Mark Hurn, librarian at the Institute of Astronomy, for his help in tracking
down many old astronomical journals held in the Institute’s splendid library; the librarians

xiv
Acknowledgements xv

at the Gordon and Betty Moore Library, the Cambridge University Library and the Royal
Observatory, Edinburgh, for their assistance. Judith Andrews also deserves special thanks
for acting as my secretary for the last eight years and for defending me from the excessive
demands of management of the Laboratory so that this book could be completed in a
reasonable time.
As ever, it is an enormous pleasure to dedicate this book to Deborah, whose love and
support mean so much more than can be adequately expressed in words.

Picture acknowledgements

I am most grateful to the following publishers and organisations for permission to reproduce
the diagrams and pictures which appear in this book.
Addison-Wesley Publishing Company (Fig. 15.4)
American Astronomical Society – Astronomical Journal (Figs 10.2, 10.3, 10.7, 14.9, 15.14)
American Astronomical Society – Astrophysical Journal (Figs 3.6, 5.2, 5.4, 5.6, 5.7, 5.8,
6.1(b), 7.4, 7.12, 7.14, 7.15(a) & (b), 8.10, 9.5, 9.6, 9.10, 9.15, 10.1, 10.5, 11.1(b), 11.3,
11.5, 11.7, 12.1, 12.4, 12.5, 12.6, 13.6, 13.7, 14.13, 14.17, 15.9, 15.17)
American Astronomical Society – Astrophysical Journal Supplement Series (Figs 7.8, 9.8,
15.12)
American Physical Society – Physical Review D (Fig. 15.13)
American Physical Society – Physical Review Letters (Figs 7.7, 8.4)
American Physical Society – Reviews of Modern Physics (Figs 8.1, 12.2)
American Science and Engineering (AS&E) (Fig. 7.6)
Anglo-Australian Observatory (Fig. 10.9)
Annual Review of Astronomy and Astrophysics (Figs 9.2, 9.7, 11.12, 11.20)
Annual Review of Nuclear and Particle Physics (Fig. 9.12)
Astronomical Society of Japan (Fig. 9.4)
Astronomical Society of the Pacific (Figs 7.11, 13.4)
Astronomical Society of the Pacific Conference Series (Fig. 11.14)
Astronomy and Astrophysics (Figs 7.19(a) & (b), 8.5(b), 14.5(b), 15.11)
Astrophysics and Space Science (Fig. 15.2)
AT&T (Fig. 7.17)
Birr Scientific and Heritage Foundation (Figs 1.4(a) & (b))
CalTech Submillimetre Observatory (Fig. 7.16)
Cambridge University Press (Figs 6.3, 6.4(a) & (b), 7.1(a) & (b), 8.3, 8.11, 10.11(b), 15.1,
A15.1, A15.2, 16.1, 16.2)
Deutsches Museum, Munich (Fig. 1.2)
Edition Frontières (Fig. 13.8)
Elsevier Publishers (Figs 15.15, 15.16)
European Space Agency (Figs 3.4, 8.5(a), 8.6, 8.7, 15.10)
Harvard College Observatory (Figs 2.1, 5.3)
Harvard-Smithsonian Astrophysical Observatory (Fig. 10.8)
xvi Acknowledgements

Harvard University Press (Fig. 7.9)

Huntingdon Library and the Observatories of the Carnegie Institution (Fig. 1.6)
International Gemini Observatory (Fig. 7.18)
Living Review in Relativity (Fig. 8.12)
Los Angeles Times (Fig. 4.2)
Mary Lea Shane Archives of the Lick Observatory (Fig. 1.5)
Max Planck Institute for Extraterrestrial Physics (Fig. 14.5(a))
National Academy of Sciences of the USA (Fig. 6.1(a))
National Aeronautics and Space Administration (NASA) (Figs 5.8, 7.10, 7.13, 8.15, 9.11,
11.2, 11.6, 11.19, 11.21, 12.4, 14.15(a), 15.9)
National Radio Astronomy Observatory of the USA (Figs 7.3, 10.11(a), 11.1(b), 11.13(a)
& (b))
Nature (Figs 3.1, 3.3(a) & (b), 7.5(a) & (b), 8.2, 8.14, 9.9, 10.4, 11.1(a), 11.8, 11.9(a) &
(b), 11.10, 11.11, 11.15, 11.16(a) & (b), 13.9(b), 14.8)
The Observatory (Figs 12.3, 13.1)
Physica Scripta (Fig. 10.12)
Potsdam Astrophysical Observatory (Fig. 3.2)
Royal Astronomical Society – Memoirs (Fig. 9.3)
Royal Astronomical Society – Monthly Notices (Figs 3.7, 9.1, 10.6, 11.4, 11.13(b), 11.17,
13.5, 13.10, A13.1, 14.3, 14.4, 14.6(a) & (b), 14.7, 14.10, 14.12, 15.7, 15.8)
Royal Astronomical Society – Quarterly Journal (Fig. 8.9)
Royal Society of London – Proceedings (Figs 5.1, 8.13, 15.5)
Science Museum of London/Science and Society Picture Library (Fig. 1.3)
Space Telescope Science Institute (Figs 7.13, 8.15, 9.11, 11.2, 11.6, 14.15(b))
Springer-Verlag (Figs 8.8, A9.1)
Springer-Verlag – D. Reidel Publishing Company (Figs 7.2, 9.13, 9.14, 13.9(a))
Springer-Verlag – Kluwer Academic Publishers (Figs 10.10, 14.2, 14.14)
SUSSP Publications (Fig. 15.6)
Swiss Society of Astronomy and Astrophysics (Fig. 14.1)
Tartu University Astronomical Observatory (Fig. 4.1)
W. H. Freeman and Company (Fig. 13.2)
World Scientific Publishers (Figs 14.11, 14.15(a), 14.16)
Yale University Press (Figs 5.5, 6.2)
Zeitschrift für Astrophysik (Fig. 3.5)
Part I
Stars and stellar evolution up
to the Second World War
1 The legacy of the nineteenth
century

1.1 Introduction

The great revolutions in physics of the early years of the twentieth century have their exact
counterparts in the birth of astrophysics and astrophysical cosmology – these astronomical
disciplines scarcely existed before 1900.
The history of the interaction between astronomy and fundamental physics is long and
distinguished. From the birth of modern science, astronomy has provided scientific infor-
mation on scales and under physical conditions which cannot be obtained in laboratory
or terrestrial experiments. There is no better example than the history of the discovery of
Newton’s law of gravity, which provides a model for the process by which astronomical
discovery is absorbed into the infrastructure of physics.1 The technological and manage-
rial genius of the great Danish astronomer Tycho Brahe (1546–1601) and his magnificent
achievements in positional astronomy during the period 1575 to 1595 provided the data
which led to the discovery of the three laws of planetary motion of Johannes Kepler (1571–
1630) during the first two decades of the seventeenth century. The technical skill of Galileo
Galilei (1564–1642) in telescope construction resulted in his discovery in 1610 of the satel-
lites of Jupiter, which were recognised as a scale-model for the Copernican System of the
World. Finally, in an extraordinary burst of scientific creativity, Isaac Newton (1643–1727)
used Kepler’s laws to discover the inverse square law of gravity and synthesised the laws of
mechanics and dynamics into his three laws of motion. This story is too well known to need
further comment, except to emphasise its astronomical roots – Newton had unified the laws
of celestial mechanics with those of free fall on Earth. It is difficult to top this achievement
in any branch of the physical sciences, but it illustrates beautifully the intimate relation
between the astronomical and physical sciences – this is a theme which will be emphasised
throughout this book.
Until the late nineteenth century, astrophysics as such did not exist. Astronomy meant
positional astronomy, and the techniques of accurate observation had improved steadily
since the time of Tycho Brahe. The accurate measurement of the motions of the Sun, Moon
and planets against the background of the fixed stars had a practical application as a means
of keeping track of time and of measuring position at sea. One of the early by-products
of accurate time keeping was the first reasonably accurate measurement of the speed of
light in 1676 by the Danish astronomer Ole Rømer (1644–1710), who observed that the
interval between eclipses of Jupiter’s innermost satellite, Io, by the planet was greater when

3
4 1 The legacy of the nineteenth century

the Earth moved away from the planet and was shorter when the Earth moved towards it.
Interpreting these differences as resulting from the changing distance between the Earth
and Jupiter, Rømer found a value for the speed of light of c = 225 000 km s−1 .
All observations were made by eye using telescopes as large as the astronomers could
afford. The revolution which was to take place at the beginning of the twentieth century
can be traced to three important technical developments during the nineteenth century – the
invention of astronomical spectroscopy, the first measurements of astronomical parallaxes
for nearby stars and the invention of photography. To take full advantage of these develop-
ments, telescope design and operation had to be substantially improved, and the resulting
instruments were to dominate the astronomy of the first half of the twentieth century. Let us
review briefly these technical developments, since they were to provide the observational
foundations for the great revolutions in astrophysics and cosmology in the first decades of
the twentieth century.

1.2 From Joseph Fraunhofer to Gustav Kirchhoff

The first decades of the nineteenth century marked the beginnings of quantitative experi-
mental spectroscopy. The breakthrough resulted from the pioneering experiments and theo-
retical understanding of the laws of interference and diffraction of waves by Thomas Young
(1773–1829). It is said that his ideas on the interference of light waves were stimulated
by observing the patterns of radiating ripples in the pond in the Paddock at Emmanuel
College, Cambridge, where he was a Fellow Commoner. In his Bakerian Lecture of 1801
to the Royal Society of London, ‘On the theory of light and colours’, he used the wave
theory of light of Christian Huyghens (1629–1695) to account for the results of inter-
ference experiments, such as his famous double-slit experiment (Young, 1802). In the
same lecture, Young introduced the tri-chromatic theory of colour vision in its modern
form. Among the most striking achievements of this paper was the measurement of the
wavelengths of light of different colours using a diffraction grating with 500 grooves per
inch. From this time onwards, wavelengths were used to characterise the colours in the
spectrum.
In 1802, William Wollaston (1766–1828) made spectroscopic observations of sunlight
and discovered five strong dark lines, as well as two fainter lines.2 He interpreted the dark
lines as delineating the four primary colours of sunlight, rather than the seven colours of the
rainbow of Newton or the three colours of the tri-colour theory of colour vision (Wollaston,
1802).
The full significance of these observations only began to be appreciated following the
remarkable experiments of Joseph Fraunhofer (1787–1826). Fraunhofer was the son of
a glazier and he became one of the two directors of the Benediktbraun glassworks in
Bavaria in 1814. The firm manufactured high-quality optical glass for military and survey-
ing instruments. Fraunhofer’s motivation for studying the solar spectrum was his realisa-
tion that accurate measurements of the refractive indices of glasses should be made using
monochromatic light. In his spectroscopic observations of the Sun, he rediscovered the
narrow dark lines which would provide precisely defined wavelength standards. His visual
1.2 From Joseph Fraunhofer to Gustav Kirchhoff 5

Figure 1.1: Fraunhofer’s solar spectrum of 1814 showing the vast numbers of dark absorption lines.
The colours of the various regions of the spectrum are labelled, as are the letters A, a, B, C, D, E, b,
F, G and H, indicating the most prominent absorption lines. The continuous line above the spectrum
shows the approximate solar continuum intensity, as estimated by Fraunhofer (Fraunhofer 1817a,b).

observations were made by placing a prism in front of a 25 mm aperture telescope. In his

words,

I wanted to find out whether in the colour-image (that is, spectrum) of sunlight, a similar bright stripe
was to be seen, as in the colour-image of lamplight. But instead of this, I found with the telescope
almost countless strong and weak vertical lines, which however are darker than the remaining part of
the colour-image; some seem to be completely black.

He labelled the ten strongest lines in the solar spectrum by the letters A, a, B, C, D, E, b,
F, G and H, and he recorded 574 fainter lines between the B and H lines (Figure 1.1); see
Fraunhofer (1817a,b).3 This notation is still used to describe the prominent absorption lines
in the spectra of the Sun and stars.
From the technical point of view, a major advance was the invention of the spectroscope
with which the deflection of light passing through the prism could be measured precisely. To
achieve this, Fraunhofer placed a theodolite on its side and observed the spectrum through
a telescope mounted on the rotating ring (Figure 1.2).
In a second paper, Fraunhofer measured the wavelengths of what are now referred to
as the Fraunhofer lines in the Solar spectrum using a diffraction grating, which consisted
of a large number of equally spaced thin wires (Fraunhofer, 1821) – he was one of the
early pioneers in the production of diffraction gratings. He found that the wavelengths of
these lines were stable and so provided accurate wavelength standards. In addition to his
observations of the Sun, Fraunhofer was the first to make spectroscopic observations of the
planets and the stars. In his papers of 1817, he reported the observation of Fraunhofer lines
in the spectrum of Venus, inferring that the spectrum was the same as sunlight. In the case
of the first magnitude star Sirius, he found, to his surprise,

. . . three broad bands which appear to have no connection with those of sunlight.
6 1 The legacy of the nineteenth century

Figure 1.2: A portrait of Fraunhofer with his spectroscope (Courtesy of the Deutsches Museum,
Munich). This portrait is located in the Hall of Fame of the museum.

In 1823, Fraunhofer made further observations of the spectra of the planets and the brightest
stars, anticipating by about 40 years the next serious attempts to measure the spectra of the
stars (Fraunhofer, 1823). He concluded that the stars have dark lines in their spectra similar
to those seen in the Sun, but that the lines present differ from star to star.
From the perspective of the glass industry, Fraunhofer was then able to characterise the
chromatic properties of glasses and lenses quantitatively and precisely. These developments
led to much superior glasses, as well as to much improved polishing and testing methods
for glasses and lenses. These technical improvements also resulted in the best astronomical
telescopes then available. Fraunhofer’s masterpiece was the 24-cm Dorpat Telescope built
for Wilhelm Struve at the Dorpat, now Tartu, Observatory in Estonia. In addition, he built
a heliometer for Friedrich Bessel at Königsberg, to which we will return in Section 1.3.
The understanding of the dark lines in the solar spectrum had to await developments in
laboratory spectroscopy. In his first report of the multitude of lines in the solar spectrum,
Fraunhofer had noted that the dark D lines coincided with the bright double line seen
in lamplight. In 1849, Léon Foucault (1819–1868) performed a key experiment in which
sunlight was passed through a sodium arc so that the two spectra could be compared precisely.
To his surprise, the solar spectrum displayed even darker D lines when passed through the
arc than without the arc present (Foucault, 1849). He followed up this observation with an
experiment in which the continuum spectrum of light from glowing charcoal was passed
 1.3 The first stellar parallaxes 7

through the arc, and the dark D lines of sodium were found to be imprinted on the transmitted
spectrum.4
Ten years later, the experiment was repeated by Gustav Kirchhoff (1824–1887), who
made the further crucial observation that, to observe an absorption feature, the source of
the light had to be hotter than the absorbing flame. From these considerations, Kirchhoff
concluded that sodium was present in the solar atmosphere. These results were immediately
followed up in 1859 by his understanding of the relation between the emissive and absorptive
properties of any substance, now known as Kirchhoff’s law of emission and absorption of
radiation (Kirchhoff, 1859). This states that, in thermal equilibrium, the radiant energy
emitted by a body at any frequency is precisely equal to the radiant energy absorbed at the
same wavelength. From thermodynamics arguments, he was able to show that there must be a
unique spectrum of radiation in thermal equilibrium, which depended only upon temperature
and frequency.5 This profound insight was the beginning of the long and tortuous story
which was to lead to Planck’s discovery of the formula for black-body radiation and the
inevitability of the concept of quantisation over 40 years later.
Throughout the 1850s, there was considerable effort in Europe and in the USA aimed
at identifying the emission lines produced by different substances in flame, spark and arc
spectra. The fact that different elements and compounds possessed distinctive patterns of
spectral lines was established, and attempts were made to relate these to the lines observed
in the solar spectrum. In 1859, for example, Julius Plücker (1801–1868) identified the
Fraunhofer F line with the bright Hβ line of hydrogen, and the C line was more or less
coincident with Hα, demonstrating the presence of hydrogen in the solar atmosphere. The
most important work, however, resulted from the studies of Robert Bunsen (1811–1899) and
Kirchhoff. In Kirchhoff’s great papers of 1861 to 1863 entitled ‘Investigations of the solar
spectrum and the spectra of the chemical elements’, the solar spectrum was compared with
the spark spectra of 30 elements using a four-prism arrangement with which it was possible
to view the spectrum of the element and the solar spectrum simultaneously (Kirchhoff, 1861,
1862, 1863). He concluded that the cool, outer regions of the solar atmosphere contained
iron, calcium, magnesium, sodium, nickel and chromium and probably cobalt, barium,
copper and zinc as well.

1.3 The first stellar parallaxes

From the seventeenth century onwards, most astronomers assumed that the stars were objects
similar to the Sun, but at vastly greater distances.6 The method of distance determination
used by Newton and others involved assuming that the Sun and stars have the same intrinsic
luminosities, a procedure known as the method of photometric parallaxes. Then, the inverse
square law can be used to measure the relative distances of the Sun and the stars. The major
technical problem was that the Sun is so much brighter than the brightest stars that it was
difficult to obtain good estimates of the ratio of their observed flux densities, or apparent
magnitudes. An ingenious solution was discovered in 1668 by James Gregory (1638–1675),
who used Jupiter as an intermediate luminosity calibrator, assuming that its light was entirely
composed of sunlight reflected from the disc of the planet and that its surface was a perfect
8 1 The legacy of the nineteenth century

reflector. Then, the apparent magnitudes of Jupiter and the bright star Sirius could be
compared, and the distance of Sirius from the Earth was found to be 83 190 astronomical
units (Gregory, 1668). The same method was used by John Michell (1724–1793) in 1767
to estimate a distance of 460 000 astronomical units for Vega, or α Lyrae, from the Earth
(Michell, 1767).7 This distance was about a factor of 4 smaller than that found in 1838
by Wilhelm Struve, who used the method of trigonometric parallax. The problem with this
approach is that it depends upon the assumption that the intrinsic luminosities of the Sun
and the stars are the same.
Direct evidence for the large distances of the stars came from James Bradley’s first
definitive measurements of the effects of the aberration of light caused by the Earth’s motion
about the Sun in 1728 (Bradley, 1728). Ever since the time of Copernicus (1473–1543) it
had been realised that a test of the hypothesis that the Earth moved about the Sun would
be the observation of the annual parallax of the stars. Attempts to measure these small
movements of the stars had been subject to a variety of insidious systematic errors. Instead
of the expected effect, Bradley (1693–1762) discovered the phenomenon of the aberration
of light due to the motion of the Earth, the effect amounting to about ±20 arcsec for the
star γ Draconis. A consequence of this remarkable result was that an upper limit could
be derived for the annual parallax of γ Draconis and hence a lower limit to its distance of
400 000 astronomical units,8 a figure consistent with Newton’s estimate using the method of
photometric parallax published in the same year. Bradley’s pioneering observations ushered
in a new epoch of precision astrometry.
The first definitive distance measurements were made in the 1830s by the method of
trigonometric parallax, the apparent motion of nearby stars against the background of the
distant stars due to the Earth’s motion about the Sun. Priority for the first trigonometric
parallax is accorded to Friedrich Bessel (1784–1846) at Königsberg. The instrument he
used was a 16-cm heliometer custom-built by Fraunhofer. The heliometer consisted of
a lens cut in half to form two D-shapes, after a design by John Dollond (1706–1761).
The images of separated stars could be brought together and their separation measured by
the reading on a micrometer screw. Bessel used this telescope to measure the movement of
the high proper motion star 61 Cygni relative to distant background stars, and he announced
its parallax in 1838 (Bessel, 1839). The parallax amounted to only about one-third of
an arcsec, corresponding to a distance of 10.3 light-years. Three months later, Thomas
Henderson (1798–1844) published a parallax of 1.16 arcsec for the southern star α Centauri
(α Cen) (Henderson, 1840), and almost contemporaneously Wilhelm Struve (1793–1864)
measured the parallax of α Lyrae to be 0.12 arcsec (Struve, 1840). Henderson was unlucky
not to publish the first parallax – he had measured a parallax of 1 arcsec in declination a
few years earlier, but delayed publication until he had reduced his data in right ascension as
well. These observations set the scale of the Universe of stars and showed unambiguously
that the stars are objects similar to our Sun.
One of the key programmes for the development of astrophysics in the late nineteenth
century and the early years of the twentieth century was the gradual accumulation of trigono-
metric parallaxes for nearby stars, but it was a difficult and demanding task. By 1900, less
than 100 parallaxes for nearby stars had been measured with any accuracy.9 The measure-
ment of parallaxes is still the only direct method of measuring astronomical distances for
stars and it remains one of the great challenges of observational astronomy. Matters improved
 1.4 The invention of photography 9

dramatically in the final decade of the twentieth century with the magnificent set of par-
allaxes measured by the Hipparcos satellite of the European Space Agency, which has
measured precision parallaxes for many thousands of stars (see Figure 3.3).

1.4 The invention of photography

The third major contribution to the development of astrophysics was the invention of the
photographic process by Louis-Jacques-Mandé Daguerre (1789–1851) and William Henry
Fox Talbot (1800–1877). Daguerre began life as an inland revenue official and then became
a scene painter at the opera. The search for methods of recording images by what was to
become the photographic process began with the discovery that some natural compounds
are rendered insoluble when they are exposed to light. In the course of his experiments,
Daguerre discovered that iodine-treated silver paper was also sensitive to light. By 1835,
he had made the important discovery of the latent image which was recorded on sensitised
paper, even if the light was not intense enough to darken the paper. The latent image could
then be developed by exposure to mercury vapour and fixed by a strong salt solution. The use
of the latent image meant that exposures could be reduced to 20 to 30 minutes. Interestingly,
the announcement of the discovery of what was called the daguerreotype process was made
by François Arago (1786–1853), the director of the Paris Observatory, on 7 January 1839.
A similar announcement was made almost simultaneously by Fox Talbot in England. One
of the earliest, and for me most moving, images is the picture taken in February 1839 by John
Herschel (1792–1871) of his father’s 40-foot telescope. In a two-hour exposure, the support
for the large tube of the telescope can be clearly seen – the telescope was dismantled in
the following year (Figure 1.3). John Herschel had a passionate interest in photography and
invented much of its terminology, including the terms ‘photography’, ‘positive’, ‘negative’
and so on.10
The first astronomical images were taken in the succeeding years, but the process was
slow. Isolated examples of successful daguerreotype images of astronomical objects were
reported over the following decade and included the Moon, a solar eclipse and the Sun.
Among the most significant images of these early years of photography was the first
daguerreotype spectrum of the Sun obtained by Edmond Becquerel (1820–1891) in 1842
which showed the complete spectrum of Fraunhofer lines as well as many lines in the
visually unobservable ultraviolet region of the spectrum (Becquerel, 1842). The problem
with the daguerreotype process was that, even for terrestrial objects, the typical exposure
times were about 30 minutes. This was greatly reduced by the invention of the wet collodion
process by Frederick Scott Archer (1813–1857) in 1851 (Archer, 1851). This process pro-
duced finely detailed negatives, and typical terrestrial exposures were reduced to 10 seconds.
Astronomical exposures were limited to 10 to 15 minutes because the plates had to remain
wet during exposure.11 The net result was faster, fine-grained plates which quickly super-
seded the daguerreotype process. These inventions sparked an enormous popular interest
in photography in the 1850s and many commercial photographic studios were set up. The
wet collodion process was used by Julia Margaret Cameron (1815–1879) in her spectacular
portraits of great nineteenth-century figures, including her famous images of the aged John
Herschel.
10 1 The legacy of the nineteenth century

Figure 1.3: John Herschel’s photograph of 1839 of part of the support structure of his father’s
40-foot telescope just before the telescope was dismantled. The details of the photographic process
are described in the text. (Courtesy of the Science Museum/Science and Society Picture Library.)

The story now diverges in two directions. Firstly, the wet collodion process was suffi-
ciently fast for astronomical images and spectra to be recorded, and the search for improved
photographic materials continued throughout the remaining years of the century. The boom
in photography meant that there was no lack of plates for astronomical use. Secondly, tele-
scope design had to be considerably improved. To take advantage of the use of photographic
plates, it had to be possible to track and guide the telescope with very much improved preci-
sion as compared with a telescope used visually. In the latter case, the length of the exposure
is determined by the response time of the eye, which is only about one-tenth of a second.
Let us first complete the story of the development of photographic techniques.
The development of photographic astronomy was largely in the hands of inspired ama-
teurs. Warren de la Rue (1815–1889) in England designed and built a photographic camera
for taking daily images of the Sun from Kew Gardens in London using very short exposures.
The result was a remarkably complete set of daily sunspot records for the period 1858 to
1872. The first photographic spectrum using the wet collodion process was obtained for
the bright star Vega by Henry Draper (1837–1882) in 1872 (Draper, 1879). The spectrum
showed the Hγ and Hδ lines of hydrogen, as well as the first detections of the next seven
ultraviolet lines in this hydrogen series. These ultraviolet lines were discovered by astro-
nomical spectroscopy seven years before they were measured in the laboratory. Subsequent
observations of the spectra of Vega and Sirius by William Huggins were used by the Swiss
schoolmaster Johann Jakob Balmer (1825–1898) in his remarkable papers of 1885 on the
Balmer formula, which describes the wavelengths of these lines in the spectrum of hydrogen.
 1.5 The new generation of telescopes 11

Balmer wrote the formula as follows:

m2
wavelength = h, (1.1)
m2 − 4
where m = 3, 4, 5, . . . and h = 3645 Å. Using Huggins’ spectral data, Balmer was able to
test his formula up to m = 16 (Balmer, 1885).12 This was the first quantum mechanical
formula to be discovered. Sadly, Balmer died 15 years before the deep significance of his
numerological discovery was appreciated by Niels Bohr.
A key development for astronomical photography was the invention of dry collodion
plates, which were much easier to use than wet plates. The speed of the dry plates was similar
to that of the wet plates but they allowed much longer exposure times to be used. The search
for improved materials continued and culminated in the discovery of emulsions consisting
of silver salts suspended in gelatin by Richard L. Maddox (1816–1902) and Charles Bennett
(1840–1927) in 1879. It was soon found that the speed of the gelatin emulsions could be
vastly increased by prolonged exposure to heat, or by the addition of ammonia. This was the
beginning of the dark art of hypersensitising photographic plates to increase their quantum
efficiencies.13 As a result of these developments, the typical exposure time for terrestrial
photography was reduced to about 1/15 second. Over the next few years, some superb
astronomical images were taken of star clusters and nebulae, revealing unambiguously the
remarkable power of photography for astronomy.
It is striking that the photographic pioneers developed their techniques on small tele-
scopes – the larger telescopes were still used for the traditional pursuits of astronomers, the
accurate measurement of time and stellar positions. As expressed by Richard Learner,14

The lessons of photography and spectroscopy, where astronomy of the highest class had been carried
out by observers with very modest telescopes . . . were not learned by the astronomical establishment.
To them, Urania, the muse of astronomy, was cold and distant, concerned with the smooth and silent
motions of the stars, not a grubby figure in an apron, standing at the laboratory sink and doing the
washing up.

1.5 The new generation of telescopes

The need to be able to track and guide the telescope accurately for long exposures required
major improvements in telescope design. The early pioneers of this story were Lewis
Morris Rutherfurd (1816–1892), also famous for his pioneering spectroscopic observa-
tions of bright stars, and John Draper (1811–1882), the father of Henry Draper. The key
developments concerned the tracking and guiding of the telescope, as well as the continued
improvement in the quality of the lenses and mirrors. Rutherfurd invented a clockwork drive
for his photographic telescope and, during the 1850s and 1860s, he produced some excel-
lent astronomical images. Besides obtaining photographic images of star fields, Rutherfurd
obtained detailed photographic spectra for the Sun. In his observations made in the 1870s,
the solar spectrum consisted of 28 overlapping plates totalling about 3 metres in length.
John Draper devoted huge efforts to optimising telescopes for photographic purposes.
Over a three-year period, he devised a series of seven grinding and polishing machines
12 1 The legacy of the nineteenth century

(a) (b)

Figure 1.4: (a) Lord Rosse’s 72-inch telescope at Birr Castle in Central Ireland following refurbish-
ment of the instrument in the 1990s. (b) A drawing of the nebula M51 and its nearby dusty companion
made by Lord Rosse from visual observations with the 72-inch telescope, showing clear evidence for
spiral structure in the galaxy. (Birr Scientific and Heritage Foundation, courtesy of the Earl of Rosse.)

and produced over 100 mirrors ranging up to 19 inches in diameter. In the last year of his
life, 1882, he succeeded in obtaining the spectra of 10th-magnitude stars in the region of
M42. Draper’s legacy, in a financial as well as a technical sense, was to be crucial for the
development of astrophysics over the succeeding years. These endeavours, all carried out
on small telescopes, were to pave the way for the spectacular burst of telescope construction
in the late nineteenth and the beginning of the twentieth centuries.
Refracting telescopes had been the preferred choice for astrometric applications, but
this development reached the end of the line with the completion of the 1-metre (40-
inch) refractor at the Yerkes Observatory of the University of Chicago located at Williams
Bay, Wisconsin. The refractors had outstanding capabilities for the visual determination of
parallaxes and for the detection of double stars. In the latter case, the observer simply waited
until a period of good seeing occurred and then, by direct observation, noted whether the
stellar image was single or double.
Several large reflecting telescopes had been constructed earlier in the century. The largest
of these was built by William Parsons, the third Earl of Rosse, a 1.8-metre (72-inch) reflector
known as the ‘Leviathan’ at his home at Birr Castle in Ireland (Figure 1.4(a)). Despite almost
insuperable problems, Rosse (1800–1867) was able to make good visual observations of
diffuse nebulae, perhaps his greatest achievement being the observation of spiral arms in
nebulae such as M51 (Figure 1.4(b)). It was, however, a struggle, not only against the
weather, but also with the materials of the telescope itself.
The biggest problem lay with the large reflector. Domestic flat mirrors had been produced
for many years, the reflection being produced by depositing tin compounds on the back
surface of a sheet of flat glass. The technical problem of producing parabolic mirrors,
1.5 The new generation of telescopes 13

which were silvered on the back surface, had not been solved. In consequence, telescope
builders, including Newton, used metal mirrors. In Lord Rosse’s telescope, the mirror was
made of speculum metal, an alloy of tin and copper with a pinch of arsenic, which is 50%
reflective. The problem was that speculum metal is a very brittle material and consequently
it is very difficult to work with. When the mirror tarnished, the mirror had to be repolished,
a hazardous procedure which could potentially destroy the mirror. In fact, Rosse had two
speculum mirrors so that one could be installed on the telescope while the other was being
repolished.
The solution to the problem of producing large silvered telescope mirrors was discovered
by Justus von Liebig (1803–1873) who, in 1835, showed how metallic silver could be
deposited by reducing silver nitrate chemically. At the Great Exhibition of 1851, glass-
makers had on show decorative items in which silver had been chemically deposited on
glass. The telescope builders realised that this was the solution to the problem. The film
of silver could be deposited on the front surface of the mirror and could be made thin
and uniform, with the result that, when the silver tarnished, rather than having to repolish
the mirror, the layer of silver could be removed chemically and a new surface laid down.
Foucault used this process to silver the rapidly rotating mirror in his famous speed-of-light
experiment in 1850. The first reflecting telescopes using silvered mirrors were built by
Karl Steinheil (1801–1870) in 1856, a 10-cm reflector, and by Foucault, who constructed
successively larger telescopes, culminating in his 80-cm reflector which was housed at the
Marseilles Observatory in 1864.15
The problems of constructing larger reflectors were considerable, not least because the
reflector design is much more susceptible to flexure and to vibrational and temperature
effects. The challenge was taken up by Andrew A. Common (1841–1903), the telescope
designer and astronomer, and George Calver (1834–1927), the mirror-maker. During the
1870s, they made a major effort to overcome the inherent problems of the design of reflecting
telescopes and introduced a number of innovations which were to be incorporated into the
next generation of instruments. The principal innovations involved in constructing their
91-cm reflector were to relieve the weight on the bearings, by submerging a hollow steel
float in mercury, and the introduction of an adjustable plate-holder. The result was that
the tracking and guiding of the telescope were very smooth – the adjustable plate-holder
had the great advantage that a guide-star could be selected outside the field of view of the
photographic plate and continuously monitored to ensure that precisely the same field was
exposed on the photographic plate within the limits of the seeing disc. Their 90-minute
exposure of the Orion Nebula won the Gold Medal of the Royal Astronomical Society in
1884.
The next advance came through the generosity of the English amateur astronomer Edward
Crossley (1841–1905). In 1895, he presented his 91-inch reflector, built to the design of the
Calver–Common telescope, to the Lick Observatory of the University of California at Santa
Cruz (Figure 1.5(a)). An important development was that the observatory was located on an
excellent Californian mountain site at Mount Hamilton, where the transparency and stability
of the atmosphere were very good and there was a large percentage of clear nights. The
mirror was repolished by Howard Grubb (1849–1931) and the mounting of the telescope
was stiffened by James E. Keeler (1857–1900). During the commissioning of the Crossley
14 1 The legacy of the nineteenth century

(a) (b)

Figure 1.5: (a) The Crossley 91-cm reflector at the Lick Observatory on Mount Hamilton. (b) A
photograph of the galaxy M51 taken by Keeler and his colleagues during the commissioning of the
telescope in 1900. (Courtesy of the Mary Lea Shane Archives of the Lick Observatory, University of
California at Santa Cruz.)

reflector in 1900, Keeler obtained spectacular images of spiral nebulae, including his famous
image of M51 (Figure 1.5(b)). Not only were the details of its spiral structure observed in
unprecedented detail, but there were also large numbers of fainter spiral nebulae of smaller
angular size. If these were objects similar to the Andromeda Nebula, M31, they must lie at
very great distances from our Solar System. Tragically, just as this new era of astronomy
was dawning, Keeler died of a stroke later in 1900 at the early age of 42.16
The next step in increased aperture followed the appointment of George Ellery Hale
(1868–1938) as founding Director of the Mount Wilson Observatory in 1904. He persuaded
his father to buy the 1.5-metre blank for a 60-inch reflecting telescope. The design was to
be an enlarged version of the Calver–Common design for the 91-cm reflector at the Lick
Observatory. Before the 60-inch telescope was completed, however, he persuaded John D.
Hooker (c.1838 – 1911), an elderly Los Angeles businessman with a passionate interest in
astronomy, to fund an even bigger telescope, the 100-inch telescope, to be built on Mount
Wilson. In 1906, the American philanthropist Andrew Carnegie visited the fledgling Mount
Wilson Observatory and pledged an additional $10 million to the endowment of the Carnegie
Institution, specifically requesting that the benefaction be used to enable the work of the
Observatory to proceed as rapidly as possible.
The technological challenges presented by the 100-inch were proportionally greater, the
mass of the telescope being 100 tonnes, but the basic Calver–Common design was retained.
The tracking was provided by a large 2-ton weight, very much like the mechanism of a
grandfather clock, which had to be wound up at the beginning of each night’s observing.
 1.6 The prehistory concluded 15

Figure 1.6: The 100-inch Hooker Telescope at the Mount Wilson Observatory. (Courtesy of the
Observatories of the Carnegie Institution of Washington and the Huntingdon Library, Pasadena.)

The optics were the responsibility of George Ritchey, an optical designer of genius, who
invented the ingenious optical configuration known as the Ritchey–Chrétien design, which
enabled excellent imaging to be achieved over a wide field of view. This was the telescope
which was to be at the heart of observational cosmology through the key years from 1918
until 1950 when the 200-inch telescope was commissioned.

1.6 The prehistory concluded

Thus, by the first decades of the twentieth century the tools and techniques which were
to provide the foundations for the revolutions in astrophysics and cosmology that were
about to take place were well developed. The number of professional astronomers was,
however, still very small. The ability to carry out large surveys of the sky with advanced
facilities and more complex procedures and instruments needed a new generation of pro-
fessionals. It is noteworthy that Hale had the foresight to hire Harlow Shapley and Edwin
Hubble as staff astronomers for the new Mount Wilson Observatory – they were to play
central roles in the history of astrophysics and cosmology in the first half of the twentieth
century.
16 1 The legacy of the nineteenth century

Notes to Chapter 1

1 I have given an account of the achievements of Tycho Brahe, Johannes Kepler, Galileo Galilei and
Isaac Newton in Case Study 1 of Malcolm Longair, Theoretical Concepts in Physics (Cambridge:
Cambridge University Press, 2003).
2 John Hearnshaw makes the interesting point that Isaac Newton narrowly missed discovering the
dark lines in the solar spectrum. The principal reason was that he used a small circular aperture
1/4 inch (0.64 cm) in diameter, whereas Wollaston used a slit 1/20 inch (0.13 cm) wide. Wollaston’s
discovery of the dark lines in the solar spectrum was essentially a footnote to his paper, which was
principally concerned with the refractive indices of a wide range of different substances.
3 Fraunhofer’s discoveries were first reported in lectures to the Munich Academy of Sciences in
1814 and 1815 and printed in the Denkschriften der München Akademie der Wissenschaften and
Gilbert’s Annalen der Physik in 1817. This paper was published in English in the Edinburgh
Philosophical Journal in two parts; see Fraunhofer (1917a).
4 The principal artificial sources of light for laboratory experiments during the early nineteenth
century were flame, arc and spark spectra. Flame spectra, obtained by burning gas in air as in
a Bunsen burner, had a typical temperature of about 2000 K. The hotter arc spectra could have
temperatures between about 3000 and 6000 K. The hottest sources were the spark spectra, which,
on average, had temperatures similar to arc spectra but, because of the presence of hot-spots, small
regions of very much higher temperature gas were produced. Thus, in terms of the ionisation state
of the material under investigation, spark spectra contained the lines of the highest excitation and
flame spectra the lowest.
5 I have given a simple derivation of Kirchhoff’s laws in Section 11.2 of Longair, Theoretical
Concepts in Physics. Case Study 5 of that book describes in some detail the subsequent history
which led to Planck’s and Einstein’s discoveries of quantisation and quanta.
6 An excellent review of early estimates of stellar distances is contained in M.A. Hoskin, Stellar
Astronomy (Chalfont St Giles, Buckinghamshire: Science History Publications, 1982), Section A.
7 Michell noted that Vega and Saturn have the same brightness when Saturn is in opposition, that is
in the direction away from the Sun. Therefore, since he knew the angular diameter of Saturn, he
could work out how much of the Sun’s light was intercepted by the planet and, assuming that all
Saturn’s light was reflected sunlight, he could use the inverse square law to estimate how far away
Vega must be. Specifically, the angular size of Saturn as observed from the Sun is 17 arcsec and so
the illuminated circular hemisphere of Saturn intercepts only (17/3600)2 × (π/720)2 of the Sun’s
light. If Vega has the same intrinsic luminosity as the Sun, Vega must be (3600/17) × (720/π)
= 48 500 times further away than Saturn, that is, 460 000 astronomical units. See Z. Kopal, in
Dictionary of Scientific Biography, Vol. 9, ed. C. C. Gillespie (New York: Charles Scribner’s Sons,
1981), pp. 370–371.
8 See the article entitled ‘Hooke, Bradley and aberration’ in M. A. Hoskin Stellar Astronomy.
9 A contemporary account of the state of stellar distance measurements in 1900 is contained in Chap-
ter 20 of the remarkable book by Agnes M. Clerke, The System of the Stars (London: MacMillan
and Company, 1890; 2nd edn, 1905).
10 John Herschel’s long-standing interest in photography predated Daguerre’s announcement. As a
result, within weeks of the announcement, Herschel was able to produce his own images.
11 According to Learner, the wet collodion process involved the following procedure.
r Cover a clean glass plate with a mixture of collodion (gun-cotton or cellulose nitrate) and
potassium iodide dissolved in ether.
r Allow the ether to evaporate and, while still tacky, immerse in a solution of silver nitrate, later
to be improved by mixing with silver bromide.
r The silver nitrate reacted with the potassium iodide to precipitate insoluble silver iodide.
r Expose the plate, but do not let it dry out. Once exposed, developed and dried, a permanent
negative image is created.
r Then, the positive could be printed at leisure on albumen coated paper.
Notes to Chapter 1 17

12 Balmer’s paper of 1885 published in the Annalen der Physik und Chemie was a synthesis of two
papers originally published in the Verhandlungen der Naturforschenden Gesellschaft in Basel 7,
pp. 548–560, 750–752.
13 Eventually, in the 1970s, the hypersensitised IIIaJ plates developed by the Kodak company reached
a quantum efficiency of about 1–2%, which has proved to be the effective limit for the photographic
process. These quantum efficiencies should be compared with those of current CCD detectors,
which can reach 80% or even greater in the red region of the spectrum. As a result, for most
astronomical applications, CCDs have replaced the photographic plate as the preferred detector
for astronomy, the exception being for very wide field astronomy with large Schmidt telescopes.
14 See R. Learner, Astronomy Through the Telescope (London: Evans Brothers, 1981.)
15 A splendid account of the contributions of Léon Foucault to these and many other areas of physics
and astronomy is contained in the book by William Tobin, The Life and Science of Léon Foucault
(Cambridge: Cambridge University Press, 2003).
16 An excellent biography of Keeler and his pioneering contributions to astrophysics has been writ-
ten by Donald E. Osterbrock, James E. Keeler: Pioneer American Astrophysicist (Cambridge:
Cambridge University Press, 1984).
 2 The classification of stellar
spectra

Somewhat surprisingly, Fraunhofer’s great discoveries in astronomical spectroscopy were

not followed up in any detail until 1863, almost 40 years later, when a number of independent
investigators, Giovanni Donati (1826–1873) in Florence, Rutherfurd in New York, George
Airy (1801–1892) at the Royal Greenwich Observatory, Huggins in London and Secchi in
Rome, began the systematic study of the spectra of the stars and nebulae.1

2.1 William Huggins – the founder of stellar astrophysics

William Huggins (1824–1910) was inspired to take up astronomical spectroscopy on read-

ing Kirchhoff’s great papers of 1861 to 1863 on the chemical composition of the solar
atmosphere. In his words,2

This news came to me like the coming upon a spring of water in a dry and thirsty land. Here, at last
presented itself the very order of work for which in an indefinite way I was looking for – namely, to
extend his novel methods of research upon the Sun to the other heavenly bodies.

Huggins was an inspired amateur astronomer who had no formal university training in the
sciences, but from 1856 until his death in 1910 he supported himself by his private income
and dedicated his efforts to the advance of astrophysics. Much of his early work was carried
out in collaboration with William Miller (1817–1870), who was professor of chemistry
at King’s College London and an expert on spectral analysis, as well as being his friend
and neighbour at Tulse Hill in London. Together, Huggins and Miller immediately began a
programme of stellar spectroscopy, the distinctive feature of their observations being that
they were carried out with good spectral resolution. In 1864, they published the first results
of these studies, those for the brightest stars being of particular importance (Huggins and
Miller, 1864a). For Aldebaran, for example, 70 lines were recorded, and for Betelgeuse
about 80 lines could be measured. About a dozen spectra were described in detail and
the common elements found in all of them. Sodium, magnesium and iron lines were very
common, while hydrogen lines were observed in some stars but not in others. Huggins’
conclusion is best summarised in his own words, written many years later:3

One important object of this original spectroscopic investigation of the light of the stars and other
celestial bodies, namely to discover whether the same chemical elements as those of our Earth are

18
2.1 William Huggins – the founder of stellar astrophysics 19

present throughout the Universe, was most satisfactorily settled in the affirmative; a common chem-
istry, it was shown, exists throughout the Universe.

At the time he concluded (Huggins and Miller, 1864a):

It is remarkable that the elements most widely diffused through the host of the stars are some of those
most closely connected with the constitution of living organisms on our globe, including hydrogen,
sodium, magnesium and iron. . . These forms of elementary matter, when influenced by heat light and
chemical force, all of which we have certain knowledge are radiated from the stars, afford some of the
most important conditions which we know to be indispensable to the existence of living organisms
such as those with which we are acquainted.

It is probably not coincidental that The Origin of Species by Charles Darwin (1809–1882)
was published in 1859.
In the same year, 1864, Huggins turned his attention to the nebulae, the nature of which
was uncertain. The common view was that they consisted of associations of unresolved
stars, in which case their spectra would be expected to display the common stellar absorption
features. While some nebulae displayed the expected absorption features, in eight of them
there were prominent bright emission lines, quite unlike those of any stellar spectrum
(Huggins and Miller, 1864b). The four most common emission lines were the Hβ and Hγ
lines of hydrogen and two strong unidentified lines at wavelengths of 500.7 and 495.9 nm.
Precise measurements of the wavelengths of the latter lines showed that they could not be
associated with any of the lines found in absorption in typical stellar spectra, and they became
known as the ‘nebulium’ lines. On the basis of the observations of the strong emission lines
of hydrogen observed in some nebulae, Huggins and Miller correctly concluded that these
objects were not associations of unresolved stars but rather

. . .must be regarded as enormous masses of luminous gas or vapour.

By 1868, they had observed some 70 nebulae, about one-third of them displaying strong
emission-line spectra, while the others possessed continuous, star-like spectra with promi-
nent absorption lines (Huggins, 1868).
The solutions to the problem of the nebulium lines and the nature of the star-like nebulae
had to await the 1920s. In 1927, Ira S. Bowen showed that the nebulium lines were the
forbidden lines of doubly ionised oxygen, which can be emitted by an ionised gas at low
densities (Bowen, 1927). The problem of the nature of the nebulae with essentially stellar
spectra was conclusively resolved by Hubble in 1925 when he showed that the sample of
nebulae consisted of a mixture of diffuse gas clouds belonging to our own Galaxy, star
clusters in our Galaxy and nearby galaxies, the light of which is the integrated emission of
millions of stars.
The first photographic spectra using wet collodion plates were recorded by Henry Draper
in 1873, the spectrum of Vega showing the Hγ and Hδ lines of hydrogen as well as the
next seven members of the series (Draper, 1879). Huggins took up stellar spectroscopy
again in 1876. He was the first to use the new dry collodion plates for spectroscopy, but he
was soon converted to the use of dry gelatin plates which had greater sensitivity. By 1880,
he had obtained excellent photographic spectra of about a dozen of the brightest stars. Of
particular significance for atomic spectroscopy were the spectra of the ‘white stars’ in his
20 2 The classification of stellar spectra

sample, which extended into the ultraviolet region of the spectrum. In these, he found 12
strong absorption lines of the hydrogen series extending from Hγ into the ultraviolet region
of the spectrum. He noted that these were all likely to be associated with hydrogen. For
the first four lines in the series, these identifications were confirmed by Hermann W. Vogel
(1834–1898) in his laboratory studies in Berlin. As noted in Section 1.4, Balmer used
these observations to demonstrate the accuracy of his formula for the Balmer series of
hydrogen up to transitions originating from the principal quantum number m = 16 (see
equation (1.1)).

2.2 The first spectral classification systems

Huggins’ brilliant analyses involved high-spectral-resolution studies of small numbers of the

brightest stars and demonstrated the power of spectroscopy in understanding their nature.
At the same time, much effort was devoted to the classification of the spectra of much
larger samples of stars in an attempt to bring some order to the diverse features which they
exhibited.
Although Rutherfurd had made the first attempt to place stellar spectra into different
classes, the most influential of the pioneers of stellar classification was the Italian Jesuit
priest Father Angelo Secchi (1818–1878), who founded the Roman College Observatory,
the Collegio Romano, in 1852 with the generous support of Pope Pius IX (1792–1878). At
the observatory, the principal instrument was a 24-cm refractor equipped with a direct-vision
spectroscope. Secchi was a prolific observer whose 700 publications appeared over a period
of 30 years up to the time of his death in 1878. The definitive version of his classification
system was completed by 1868 on the basis of spectroscopic observations of about 500
stars (Secchi, 1866, 1868). He placed the stars into four classes.
r Class I consisted of white or blue stars, such as Sirius, which exhibit hydrogen absorp-
tion lines. This class included those with the hydrogen lines in emission.
r Class II consisted of slightly coloured, yellow or solar-type stars, which displayed the
principal Fraunhofer lines.
r Class III were red stars with wide absorption bands, an example of which was Betel-
geuse.
r Class IV were the ‘carbon stars’, as they are now known, which have ‘luminous bands
separated by dark intervals’. These were only identified after the other classes among
samples of faint red stars.
The spectra of stars in each of these classes have similar patterns of spectral lines, for example
the Class II stars resembling the spectrum of the Sun, although there were considerable
variations within each class. Secchi continued his spectroscopic observations and, by the
time of his death, had classified over 4000 stellar spectra, including most of the stars visible
to the naked eye in the northern hemisphere.
Many of the pioneers of stellar spectroscopy and the classification of stellar spectra died
in the period 1870 to 1880, leaving Huggins as the sole survivor and the father figure of stellar
spectroscopy. Of the succeeding generation, major contributions were made by Hermann
C. Vogel (1841–1907), who improved greatly the techniques of precision spectroscopy, in
 2.3 The Harvard classification of stellar spectra 21

particular in the precise measurement of stellar velocities through their Doppler shifts. He
also devised his own system of spectral classification, which was similar to Secchi’s (Vogel,
1874). It excluded the Class IV spectra, but included the subdivision of the three classes
into subclasses.
As the techniques of photographic spectroscopy developed, the objectivity of the clas-
sification procedures improved, but the full complexities of stellar spectra also began to
be appreciated. The basic problem with the classification schemes was that there was a
lack of understanding of the physical basis for the classification procedures. It was not
clear the extent to which the colours and properties of the stars were affected by the pres-
ence of different elements in their atmospheres. There was some evidence that the blue
stars were hotter than the red stars, but it was not clear that the colour of a star was an
indicator of its temperature – Huggins argued that the redness of some stars could simply
be due to the presence of large numbers of absorption lines towards the blue end of the
spectrum.
David DeVorkin estimates that 23 different spectral classification systems had been
proposed by 1900. While a number of groups studied the problems of stellar classification,
the whole enterprise was overtaken by the mammoth surveys of stellar spectra which were
undertaken at Harvard under the direction of Edward C. Pickering. The ultimate results
of these efforts were the Harvard system of spectral classification and the Henry Draper
Catalogue. These endeavours were to lead to the physical understanding of stellar spectra
and also to new aspects of atomic physics.

2.3 The Harvard classification of stellar spectra

Henry Draper was trained in medicine but, after a visit in 1858 to William Parsons, the
third Earl of Rosse, at Parsonstown in Ireland, he devoted all his energies to pioneering the
application of photography to astronomical observation. As discussed in Section 1.4, he
obtained the first photographic spectrum using the wet collodion process in August 1872.
Following a visit to Huggins in 1879, he was converted to the use of dry photographic plates,
which were to revolutionise astronomical spectroscopy. Among his technical innovations,
he built an excellent clockwork drive for the telescope which enabled him to obtain long
exposure spectra, his longest exposure being for 140 minutes.
Following his untimely death at the age of 45 in 1882, his widow, Mrs Anna Palmer
Draper (1839–1914), established the Henry Draper Fund. She provided funds to the Harvard
Observatory for Edward Pickering and his assistants to photograph, measure and classify
the spectra of stars and to publish the resulting catalogue in the Annals of the Harvard
Observatory as a memorial to Draper.
Edward Pickering (1846–1919) entered the Engineering Department at Harvard Univer-
sity and graduated summa cum laude on his nineteenth birthday. Two years later, he was
appointed assistant professor of physics at the newly founded Massachusetts Institute of
Technology. He revolutionised the teaching of physics by instituting a carefully designed
practical course in physics, subsequently published as Elements of Physical Manipulations.4
In 1876, Pickering was appointed director of the Harvard College Observatory, a con-
troversial appointment since he was not an observational astronomer. It was, however, an
22 2 The classification of stellar spectra

appointment of great foresight since the nature of astronomical research was changing. The
introduction of spectroscopy as a tool for astronomical research was the springboard for
the new science of ‘astro-physics’. Symbolic of this new direction of astronomical research
was the foundation of the Astrophysical Journal by George Ellery Hale and James Keeler
in 1895, which included the interpretative subtitle An International Review of Spectroscopy
and Astronomical Physics.5
Another feature of the revolution which was overtaking the nature of astronomical
research was the increase in funding necessary to carry out large programmes in astron-
omy. The Henry Draper Fund provided several hundred thousand dollars over a number of
years to support the Harvard programme. The ambitious programme which Pickering was
to establish received other major benefactions. The Paine Fund donated about $400 000 in
1886 and the Boyden Fund gave $230 000 in 1887. In addition, the Bruce Fund contributed
$50 000, and Pickering himself provided more than $100 000 from his own resources. This
private sponsorship was to enable Pickering to carry out his huge programme to comple-
tion. Already, at the very birth of the science of astrophysics, astronomy was big science,
needing substantial ongoing funding for the construction of state-of-the-art telescopes and
their long-term operation.
Pickering fully realised the great scientific potential of astrophysics, as opposed to posi-
tional astronomy, which remained the principal concern of the national observatories. He
pioneered three fields: visual photometry, stellar spectroscopy and stellar photography. The
visual photometry was undertaken by means of a meridian photometer, in which the bright-
ness of a star on the meridian is compared with the brightness of the pole star, Polaris. Over
1.5 million visual photometer readings were made, most of them by Pickering himself. In
1908, these studies culminated in the publication of the ‘Revised Harvard photometry’,
which was crucial for the study of variable stars and which was to prove to be of special
importance for astrophysics and cosmology (Pickering, 1908).
Stellar spectroscopy was funded by the Henry Draper Fund. Pickering decided that the
most effective means of undertaking spectral studies of very large numbers of stars was to
equip the survey telescope with an objective prism to disperse the images of all the stars
in the region of sky under observation.6 The first experiments were carried out in May
1885, and regular observing began in October of that year. The Bache 8-inch telescope at
the Harvard College Observatory had a field of view of 10◦ and, with an objective prism of
angle 13◦ , the spectra of 6th-magnitude stars could be recorded on photographic plates in
5 minutes – for comparison, the very faintest stars visible to the unaided eye are about 5th
magnitude. The programme was in three parts: a general survey of stellar spectra for all stars
north of declination −25◦ brighter than 6th magnitude; a study of the spectra of fainter stars;
and a detailed investigation of the spectra of the brighter stars. The principal investigators
were Williamina P. Fleming (1857–1911), Annie Jump Cannon (1863–1941) and Antonia
C. Maury (1866–1952). They were supported by a large corps of women ‘computers’, the
team being jokingly referred to as ‘Pickering and his harem’ (Figure 2.1).
The first part of the observing programme was complete by January 1889 and consisted
of 633 plates containing the spectra of 10 351 stars. The tasks of examining and classifying
the spectra, as well as estimating their magnitudes, were carried out by Fleming. The Draper
Memorial Catalogue describing the spectra and properties of these stars was published in
18907 and provided by far the largest and most systematic classification completed in the
2.3 The Harvard classification of stellar spectra 23

Figure 2.1: Pickering and his team of ‘computers’ in 1913. Annie Cannon is the second to the right
of Professor Pickering. (Courtesy of Harvard College Observatory)

nineteenth century (Pickering, 1890). The spectral classification was based upon Secchi’s
four classes, but they were now divided into further subclasses. Class I was divided into
four subclasses A, B, C and D, Class II into seven subclasses, E, F, G, H, I, K, L, and
Classes III and IV were renamed M and N. Special letters were reserved for particular
classes of object. The designation O was used to describe the class of star discovered
by Charles J. F. Wolf (1827–1918) and Georges A. P. Rayet (1839–1906), which displayed
prominent very broad emission lines in the blue spectral region on a continuous background
and which are now known as Wolf–Rayet stars. The planetary nebulae were designated P
and Q for stars otherwise unclassifiable through the sequence A to P. The intention was
to provide uniformity of spectral type within each of the subclasses and continuity along
the sequence so that, for example, progressing from E, F, G, through H, the prominent
Fraunhofer lines seen in solar-type stars become more prominent, while the spectra became
weaker at wavelengths shorter than 431 nm.
The task of analysing the spectra of the bright stars was undertaken by Antonia Maury,
who graduated from Harvard in 1887 and who was a niece of Henry Draper. The observations
for this programme were made with the 11-inch Draper telescope at the Harvard College
Observatory with much higher spectral resolution than the general surveys, a maximum
dispersion of 11 Å mm−1 being available. A total of 4800 plates of the 681 stars in the
survey were analysed by Maury, and Pickering allowed her to develop her own system of
spectral classification. The classification scheme was similar to that devised by Mrs Fleming
24 2 The classification of stellar spectra

but consisted of 22 classes. One important difference was that she placed the B stars, which
are similar to the blue stars in Orion, earlier in the sequence than the A stars because of
their simpler spectra. The work was complete by 1895 and published in 1897 (Maury and
Pickering, 1897).
Maury’s most important contribution, however, lay in her further subdivision of the
spectra on the basis of the appearance of the spectral lines. She described three different
types of line, the bulk of the stars belonging to class a in which the lines were clearly defined
and of ‘average’ width. In class b, the lines were much broader and hazy, while in class c
the lines were unusually narrow and sharp. Of the 681 stars in her sample, 355 were class a,
only 18 were of class c and 17 were classed as ac intermediate between classes a and c. The
class b stars were mostly rapid rotators or double-lined spectroscopic binaries, although this
was not understood at the time. The division of the stars into classes a and c was to lead to
the discovery of the giant stars by Hertzsprung in 1905.
There was one further wrinkle in the story before the definitive Harvard system of
classification was established. During the solar eclipse of 18 August 1868, spectroscopic
observations were made of the emission from solar prominences, and Rayet discovered an
intense emission line which he identified with the sodium D lines. The line was reobserved by
Norman Lockyer (1836–1920) in October 1868 by placing the spectroscopic slit tangential
to the limb of the Sun. He established that its wavelength did not correspond to either of the
strong sodium D lines, but had a wavelength of 587.6 nm. There was no corresponding line
in the Fraunhofer spectrum of the Sun, nor was there any corresponding feature in the spectra
of the known elements. It was inferred that the line was due to some new element which had
not been isolated in the laboratory and he named it ‘helium’, after Helios, the Greek god
of the Sun. Helium was only discovered in the mineral cleveite by William Ramsay (1852–
1916) in 1895. When the spectrum of the gas was observed in a discharge tube, the line at
587.6 nm was observed along with five other lines. Lockyer showed that some of these lines
were also present in chromospheric spectra and, in particular, that they were present in some
of the Orion stars which had been placed in spectral class B. He recognised that stars which
exhibited helium absorption lines had to be hotter than stars such as Vega on the basis of his
laboratory observations of arc and the hotter spark spectra. Lockyer’s theoretical ideas were
controversial, to say the least, but they foreshadowed future developments. He concluded
that as the temperature increases, the elements are dissociated into ‘proto-metals’. In his
words (Lockyer, 1900),

We have then to face the fact that on the dissociation hypothesis, as the metals which exist at the
temperature of the arc are broken up into finer forms, which I have termed proto-metals, at the fourth
stage of heat (that of the high tension spark) which gives us the enhanced spectrum; so the proto-metals
are themselves broken up at some temperature which we cannot reach in our laboratories into other
simpler gaseous forms, the cleveite gases, oxygen, nitrogen and carbon being among them.
Does the story end here? No there is still a higher stage; as the cleveite gases have disappeared
as the arc lines and enhanced lines did at the lower stages; the raw form of hydrogen to which I have
before drawn attention and which we may think of as ‘proto-hydrogen’, makes its appearance.

What might have been ‘proto-hydrogen’ was discovered by Pickering in 1896 in the star ζ
Puppis. He discovered a sequence of absorption lines resembling the Balmer series, which
became known as the Pickering series (Pickering, 1896). He showed that the lines could
be described by Balmer’s formula provided half-integral values of the principal quantum
 2.3 The Harvard classification of stellar spectra 25

number m were used (Pickering, 1897).8 As a result, Pickering considered that the lines were
associated with hydrogen under conditions of density and temperature not accessible in the
laboratory. In 1912, however, Alfred Fowler (1868–1940) showed that the Pickering series
could be observed in laboratory experiments in which the spectra of mixtures of hydrogen
and helium were measured. In 1913, in his first great paper on the quantum theory of the
hydrogen atom, Niels Bohr (1885–1962) showed that the lines of the Pickering series were
not associated with half-quantum numbers, but rather were due to singly ionised helium
atoms which have twice the nuclear electric charge9 (Bohr, 1913a) – the Pickering series
resulted from transitions from energy levels with principal quantum numbers n > 4 into
the n = 4 level.
The most famous work which led to the standard Harvard classification was carried out
by Annie Cannon. She attended Wellesley College in Norfolk County, Massachusetts, in
1884, one of the first girls of her native state of Delaware to go away to college, and joined the
staff of the Harvard College Observatory in 1896. The Boyden Fund donation of $230 000
was used by Pickering to establish a southern station of the Harvard College Observatory at
Arequipa in Peru in 1887. In addition to the 8-inch telescope, which was no longer needed
for observations in the northern hemisphere, a 13-inch refractor was also purchased from
the Boyden Fund. The objective was to complete the survey of the little-known southern
sky. The first plates of the survey were taken in 1891 and, by 1899, 5961 plates had been
taken of the spectra of 1122 stars. Cannon was assigned the task of classifying 813 of these
stars.
In classifying the stellar spectra, Cannon did not adopt Maury’s scheme, but reverted to
Fleming’s original Harvard scheme with some important amendments. She adopted Maury’s
proposal that the O stars are the hottest classes of star and that the B stars should precede the
A stars in the stellar sequence. She dropped a number of the classes introduced in the original
Harvard sequence so that the basic sequence now became O, B, A, F, G, K, M. In addition,
there was a class P for planetary nebulae and Q for peculiar stars. Another innovation
was the introduction of a decimal notation to represent the spectra of stars intermediate
between the main classes. Thus, B stars were renamed B0 stars and A stars as A0. Stars
intermediate between them could be assigned classes B1, B2, B3, etc. Stars were placed
along this linear sequence on the basis of the presence or absence of different spectral lines,
the intention being that the progression of spectral features should be continuous along the
sequence. Unlike in Maury’s classification system, there was no distinction between lines
of different widths. Although there were subsequent enhancements to Cannon’s system
of spectral types, this is the origin of the basic Harvard classification system which was
published in the Harvard Annals in 1901 (Cannon and Pickering, 1901).
It was already apparent from the investigations of Lockyer that the sequence was basically
a temperature sequence, but it would take a great deal more work before the precise relation
between spectral type and temperature was established through the work of Saha, Fowler
and Milne.† A simplified version of the modern classification system, which is known as
the MK classification, is given in Table 2.1.10

† See Section 3.3.

26 2 The classification of stellar spectra

Table 2.1. The principal features of the modern system of stellar spectra
The MK system does not include the classes R, N, S (Morgan, Keenan and Kellman, 1943;
Johnson and Morgan 1953)

Effective temperature
Class Class characteristics Type (Teff /K)

O hot stars with He II absorption lines; strong ultraviolet O5 40 000

continuum
B He I lines attain maximum strength; no He II lines; H B0 28 000
developing later B5 15 000
A H lines attain maximum strength at A0, decreasing later; A0 9900
Ca II increasing A5 8500
F Ca II stronger; Fe and other metal lines appear F0 6030
F5 6500
G Ca II very strong; Fe and other metals strong; H weaker; G0 6030
solar-type spectrum G5 5520
K neutral metallic lines dominate; CH and CN bands K0 4900
developing; continuum weak in blue K5 4130
M very red; TiO2 bands developing strongly M0 3480
M5 2800
M8 2400
R strong CN bands and C2 bands increasing
N C2 bands; CN bands decreasing
S heavy metal stars; ZrO bands

Cannon went on to classify a further 1477 northern stars according to the 1901 system
and then began the classification of a further 1688 southern stars to fainter magnitudes.
These were published in the Harvard Annals in 1912. The result of the efforts of Pickering,
Maury and Cannon was that, by 1912, almost 5000 stars had been classified over the whole
sky with classifications far superior to those previously available. The quality of the data
was such that in 1912 Pickering was able to begin the analysis of the distribution of stars of
different spectral type throughout the Galaxy. He concluded (Pickering, 1912):
These figures show very clearly that the spectra of Classes A and B are more numerous in the Milky
Way than outside it, and that the maximum point is a little south of the Galactic Equator.
This analysis marks the beginning of the delineation of Galactic structure using stars of
different spectral type.
In addition to these heroic efforts by Pickering, Cannon and Maury, Williamina Fleming
carried out detailed studies of all the stellar spectra which did not fit naturally into what
was to become the standard Harvard classification scheme. From 1899 to 1911, when she
died, she was curator of astronomical photographs for the Observatory and devoted her
astronomical energies to the stars with peculiar spectra, including meticulously describing
the spectra, cataloguing the discoverers of the different types of star and the literature
Notes to Chapter 2 27

references. These stars included novae, gaseous nebulae, peculiar O stars, emission line
A and B stars, spectroscopic binaries, variable stars, N and R stars, Oe5 stars and other
anomalous stars. This list gives some impression of the enormous advances in spectroscopic
knowledge of the stars since the inception of the Henry Draper Memorial project.
At the time of Fleming’s death, Pickering was planning an even more ambitious pro-
gramme, which was to become the Henry Draper (or HD) Catalogue. On 11 October 1911,
Cannon began the classification of 225 300 stars and completed the task just under four
years later. In an extraordinary feat of concentrated effort, she was able to classify spectra
at a rate of about of three per minute, and her classifications were repeatable over the years
of the survey. The huge sample of stars comprised all those brighter than 8th magnitude
in the northern hemisphere and about 9th magnitude in the south. The HD Catalogue was
published between 1918 and 1924.11 Pickering died in 1919, having lived to see the first
three sections of the catalogue published – he had been director of the Harvard College
Observatory for 42 years. He showed no special interest in interpreting his results but
was content to be, in his words, ‘a collector of astronomical facts’. Under Pickering, the
Harvard College Observatory became the worldwide distribution centre for astronomical
information.
Cannon supervised the publication of the other volumes and continued to classify spec-
tra under Pickering’s successor, Harlow Shapley. An extension of the HD Catalogue was
prepared to extend the northern survey to the same magnitude limit as the southern survey.
By the time of her death in 1941, Cannon had classified almost 400 000 spectra. According
to Cecilia Payne-Gaposchkin12
Miss Cannon was not given to theorising; it is probable that she never published a controversial
word or a speculative thought. That was the strength of her scientific work – her classification was
dispassionate and unbiassed.

Cannon was almost completely deaf throughout her career. She received many international
honours. In 1938, she was appointed William Cranch Bond Astronomer, one of the first
women to receive an appointment from the Harvard Corporation, and she was the first
woman to be awarded an honorary degree by Oxford University.

Notes to Chapter 2

1 Many more details on the history of astronomical spectroscopy are contained in John Hearnshaw’s
outstanding book The Analysis of Starlight: One Hundred and Fifty Years of Astronomical Spec-
troscopy (Cambridge: Cambridge University Press, 1986, 1990). In addition to the achievements
of the pioneers of astronomical spectroscopy, Hearnshaw provides many details of the instrumental
innovations which made the astrophysical advances possible.
2 This is a quotation from Huggins’ autobiographical eassy which appeared in the Nineteenth
Century Review, June 1897.
3 This remark appears as a footnote to the 1909 reprint of the 1864 paper by Huggins and Miller in
Huggins’ collected scientific papers in Huggins, Sir W. and Huggins, Lady M., eds, Publications
of Sir William Huggins Observatory, Vol. II (London: W. Wesley and Son).
4 The Elements of Physical Manipulations was published in two volumes by Macmillan and Co.
in the period 1873 to 1876. The second volume contains a major section on practical astronomy.
Pickering wrote in the preface to this volume, ‘One of the most important features of this volume
28 2 The classification of stellar spectra

is the introduction of a chapter on Astronomy. . . A careful examination of the subject seems to

show that the laboratory method may be used to teach Astronomy as successfully as Chemistry
and Physics.’
5 The history of the founding of the Astrophysical Journal was recounted in 1995 by Donald
E. Osterbrock in the centenary edition of the journal (see Astrophysical Journal 438, 1995, 1–8).
6 The many surveys undertaken by Pickering and his team and the details of the different classifi-
cation schemes are described in detail by Hearnshaw, The Analysis of Starlight, Chapter 5.
7 This catalogue was based on the work of Williamina Fleming and was entitled The Draper
Memorial Catalogue of Stellar Spectra.
8 In his paper of 1896, Pickering suggested a modified version of Balmer’s formula, which had only
been discovered 11 years earlier. In this paper, his proposed formula was given by
 
m2
λ = 4650 2 − 1032 Å.
m −4

Note the different ‘Rydberg’ constant as compared with the value appearing in equation (1.1) and
the subtraction of the wavelength 1032 Å. In his paper of 1897, however, he found that the formula

m2
λ = 3646.1 Å (1)
m2 − 16

gave an excellent fit to the data and had the advantage of using the same Rydberg constant as that
of the Balmer series of hydrogen (see equation (1.1)). Pickering suggested that both the Balmer
lines and the lines of the Pickering series could be explained by the original Balmer formula,
provided half-integral values of m were allowed. Thus, replacing m in equation (1.1) by m/2
enables equation (1) to be recovered. Note that, once the Bohr model of the atom was applied
to singly ionised helium, the Pickering series results from transitions into the n = 4 level from
higher principal quantum numbers. The result is that the lines resulting from transitions from even
principal quantum numbers are coincident with those of the Balmer series of hydrogen. The odd
transitions were members of the series described by Pickering in his 1896 paper and which led
to the idea that the lines originated in an even more primitive form of hydrogen, what Lockyer
termed ‘proto-hydrogen’.
9 There is a delightful story concerning Bohr’s identification of the Pickering series with the Balmer
series of singly ionised helium. Bohr argued that singly ionised helium atoms would have exactly
the same spectrum as hydrogen, but the wavelengths of the corresponding lines would be four times
shorter, as observed in the Pickering series. Fowler objected, however, that the ratio of the Rydberg
constants for singly ionised helium and hydrogen was not 4, but 4.00163. Bohr realised that the
problem arose from neglecting the contribution of the mass of the nucleus to the computation of
the moments of inertia of the hydrogen atom and the helium ion. If the angular velocity of the
electron and the nucleus about their centre of mass is ω, the condition for the quantisation of
angular momentum is given by
nh
= μω R 2 ,
2π

where μ = m e m N /(m e + m N ) is the reduced mass of the atom, or ion, which takes account of the
contributions of both the electron and the nucleus to the angular momentum; R is their separation.
Therefore, the ratio of Rydberg constants for ionised helium and hydrogen should be
⎛ me ⎞
1 +
RHe+ ⎜ M ⎟
= 4⎝ m e ⎠ = 4.00160,
RH 1+
4M
Notes to Chapter 2 29

where M is the mass of the hydrogen atom. Thus, precise agreement was found between the
theoretical and laboratory estimates of the ratio of Rydberg constants for hydrogen and ionised
helium.
In his biography of Bohr, Niels Bohr’s Times, in Physics, Philosophy, and Polity (Oxford:
Oxford University Press, 1991), Pais tells the story of the encounter of George Hevesy (1885–
1966) with Einstein in September 1913. When Einstein heard of Bohr’s analysis of the Balmer
series of hydrogen, he remarked cautiously that Bohr’s work was very interesting, and important if
right. When Hevesy told him about the helium results, Einstein responded, ‘This is an enormous
achievement. The theory of Bohr must then be right.’
10 The Yerkes system of spectral classification was published by W. W. Morgan, P. C. Keenan, and
Edith Kellman (b. 1911) in 1943 (Morgan, Keenan and Kellman, 1943), and was known as the
MKK system. Some further refinements were made by Johnson and Morgan in 1953 (Johnson
and Morgan, 1953) and this revised Yerkes system is known as the MK system.
11 The Henry Draper (HD) Catalogue was published in the Harvard Annals, vols 51 and 55–62
between 1918 and 1924.
12 This quotation is included in Cecilia Payne-Gaposchkin’s obituary of Annie Cannon published in
The Telescope, 8, 1941, 62–63.
 3 Stellar structure and evolution

3.1 Early theories of stellar structure and evolution

The origin of the theory of stellar structure and evolution can be traced to the understanding
of the first law of thermodynamics. As a result of the experimental ingenuity of Julius
Mayer (1814–1878) and, particularly, of James Prescott Joule (1818–1889), and the deep
theoretical insights of Rudolph Clausius (1822–1888) and William Thomson, later Lord
Kelvin (1824–1907), the two laws of thermodynamics were established in the early 1850s.1
In popular terms, they can be stated as follows.

(1) Energy is conserved when heat is taken into account.

(2) The entropy of any isolated system can only increase.

Applying the first law to the stars, the source of energy could be attributed to the heat liberated
when matter is accreted onto their surfaces. The kinetic energy of infall from infinity, which
is equal to the gravitational binding energy of the material at the surface, is converted into
heat when the matter hits the surface. A popular version of the theory involved meteoritic
bombardment of stars as the means of providing the necessary energy release. This proposal
contained, however, the serious flaw that the necessary flux of meteoroids would perturb
the orbits of the inner planets and would also have resulted in a quite unacceptably high rate
of meteoroid bombardment of the Earth. Hermann von Helmholtz (1821–1894) and Kelvin
proposed that, rather than the gravitational potential energy of meteoroids, the contraction
of the Sun itself provided an enormous reservoir of energy (Helmholtz, 1854; Thomson,
1854a,b).2 The Sun was conceived of as a liquid sphere which gradually contracted and
cooled. Energy transport through the Sun was assumed to be by convection.
Kelvin and Helmholtz realised that they could estimate the age of the Sun since its
present luminosity is known and its gravitational potential energy can be estimated. This
timescale, nowadays referred to as the Kelvin–Helmholtz timescale, tKH , for cooling of the
Sun, or any star, can be estimated by dividing its gravitational potential energy, ∼G M 2 /R,
by its present luminosity, L, as follows:
E G M2
tKH ≈ ≈ . (3.1)
L LR
For the Sun, this timescale is only ∼107 years. Kelvin used a similar argument to estimate
the age of the Earth. He knew the temperature gradient in the outer layers of the Earth and
so he could estimate how long it would take the Earth to cool by thermal conduction through

30
3.1 Early theories of stellar structure and evolution 31

its surface. This age turned out to be about 20–40 million years, not so different from the
Kelvin–Helmholtz timescale for the Sun.
These estimates were considerably shorter than those favoured by the geologists, who
suggested that the age of the Earth was greater than 100 million years from stratigraphic
analyses; but these estimates were subject to some uncertainty. Kelvin argued forcibly and
successfully against such long geological timescales, much to the chagrin of the geologists.
The first reliable estimates of the age of the Earth were made in 1904 by Ernest Rutherford
(1871–1937) using the relative abundances of radioactive and stable isotopes of very heavy
elements such as uranium. His age for the Earth was at least 700 million years. Rutherford’s
announcement of this result in Kelvin’s presence is delightfully told in his own words.3

I came into the room which was half dark and presently spotted Lord Kelvin in the audience and
realised that I was in for trouble with the last part of my speech dealing with the age of the Earth
where my views conflicted with his. To my relief, Kelvin fell asleep, but as I came to the important
point, I saw the old bird sit up, open an eye and cock a baleful glance at me!
Then a sudden inspiration came and I said Lord Kelvin has limited the age of the Earth, provided
no new source was discovered. The prophetic utterance refers to what we are now considering tonight,
radium! Behold! the old boy beamed upon me.

Rutherford elaborated and refined these techniques over the following years (Ruther-
ford, 1907). The Kelvin–Helmholtz theory of the energy source of the Sun had remained
unchallenged for almost 50 years, but now there was a major problem concerning the origin
of its luminosity and that of the stars. The origin of the Sun’s energy remained a thorny
issue for the pioneers of stellar structure over the next two decades. As late as 1932, when
Chandrasekhar visited Copenhagen to work with Niels Bohr and his associates, Kameshwar
C. Wali records Bohr’s attitude to the problems of understanding the physics of the stars:4

I cannot be really sympathetic to work in astrophysics because the first question I want to ask when
I think of the Sun is where does the energy come from. You cannot tell me where the energy comes
from, so how can I believe all the other things?

By then, however, the feasibility of nuclear energy as the source of the Sun’s luminosity
was well on the way to becoming established, as we will see. Let us first return to the early
pioneers of stellar structure and evolution.
The first person to investigate the internal structure of the Sun as a gaseous, rather than a
liquid, body was the American J. Homer Lane (1819–1880), who was employed by the US
Patent Office in Washington as an expert examiner from 1847 to 1857, and then by the Office
of Weights and Measures from 1869. He was highly regarded by Joseph Henry (1797–1878)
as a mathematical physicist. During the period of the American Civil War, he carried out the
first calculations to determine whether or not the Sun’s surface properties were consistent
with it being considered to be a sphere of a perfect gas. In 1869, he reversed the calculations
and, assuming the material of the Sun to be a perfect gas, attempted to reproduce its surface
properties and to determine the variation of its density, temperature and pressure with radius
(Lane, 1870). This programme was not particularly successful. Nonetheless, he was the first
person to adopt the correct equations of hydrostatic equilibrium and mass conservation for
the stars. Although not included in his 1870 paper, he was the first person to derive the
important and somewhat non-intuitive result that, if a star loses energy by radiation and
32 3 Stellar structure and evolution

contracts, the temperature increases rather than decreases.5 This occurs because, as the star
contracts through a series of quasi-equilibrium states, the negative gravitational potential
energy becomes more negative and since, according to the virial theorem the internal thermal
energy must be minus one half of the gravitational potential energy, the internal energy of
the gas must increase, raising its temperature.†
In the late 1870s, similar calculations were carried out independently by Augustus Ritter
(1826–1908), Professor of Mechanics at the Polytechnical School at Aachen, who identified
the initial phase of evolution of the star as the contraction of a perfect gas sphere, which then
cooled according to the Kelvin prescription (Ritter, 1883a,b, 1898). The culmination of these
early physical models for stars was the treatise by Robert Emden (1862–1940), Gaskugeln,
published in 1907 (Emden, 1907). Emden was at that time an assistant professor of physics
and meteorology at the Technische Hochshule in Munich, and, delightfully, the intention of
the treatise was to attract students into theoretical physics by giving as ‘practical examples’
the internal structure of the stars.
Like Lane and Ritter, Emden assumed that his stellar models were in convective equilib-
rium, but he went further than Lane and Ritter by introducing polytropic solutions, which
allowed a much wider range of stellar models to be constructed. In particular, they allowed
different variations of the density and pressure of the gas with radius consistent with the
laws of physics. The set of equations used by Lane, Ritter and Emden are very powerful
indeed. The equations of hydrostatic equilibrium and mass conservation are given by
dp G Mρ dM
=− 2 and = 4πr 2 ρ, (3.2)
dr r dr
where p and ρ are the pressure and density of the stellar material, respectively, and M is
the mass of the star within radius r , M = M(≤r ). The equation of state of the gas was
assumed to be of power-law form, p = κρ γ . To simplify equations (3.2), it is convenient
to introduce a dimensionless measure of density, w, through the relation ρ = ρc w n , where
n = (γ − 1)−1 is the polytropic index and ρc is the central density of the star. Writing the
distance r from the centre in terms of the dimensionless radius z, r = az, the Lane–Emden
equation can be written as a second-order differential equation:
  
1 d 2 dw
z + w n = 0, (3.3)
z 2 dz dz
with
⎡ ⎤1/2
1
−1
(n + 1)κρcn
a=⎣ ⎦ .
4π G

This equation determines the dependence of w upon radius and so provides solutions for
the structure of the star for different values of n.
Emden showed that the solutions of equation (3.3) result in stars which have a boundary at
a finite radius. As discussed elegantly by Kippenhahn and Weigert in their classic text Stellar
Structure and Evolution, many important aspects of the astrophysics of stars of different

† A demonstration of this result is presented in Section A3.1.

3.1 Early theories of stellar structure and evolution 33

Figure 3.1: An example of Lockyer’s temperature curve for stellar spectral evolution (Lockyer, 1914).
His assignments of the spectral types to different parts of the arch are indicated.

types can be understood on the basis of this equation alone.6 These theories did not gain
the recognition they deserved because astronomers argued that the physical conditions
inside the stars were unknown and there was no understanding of the process of energy
generation.
While these physical models were being developed, the commonly held view was that
the various schemes of spectral classification must have some evolutionary significance.
Among the most vocal advocates of this picture, Norman Lockyer played an important
role as a populariser of astronomy and of science in general. He was an amateur enthu-
siast of enormous energy, who, following his employment by the British War Office,
became Director of the Solar Physics Observatory, set up in South Kensington, London,
in 1879. In 1869, he founded the weekly science magazine Nature, which he edited for
the next 50 years, and which provided a ready-made forum for promoting his opinions.
In the 1880s, Lockyer attempted to identify the various spectral classes of star with an
evolutionary scheme based upon the meteoritic hypothesis. Figure 3.1 shows a sketch of
his theory of stellar evolution.7 The evolutionary ‘temperature arch’ begins at the bot-
tom left and shows a cloud of meteoroids colliding and vaporising, thus giving rise to
gaseous nebulae and comets. The nebulae condense and contract to form stars, at which
point the star attains its maximum temperature at the peak of the arch. The star was then
assumed to cool to become a compact red star. Lockyer assigned spectral classes to dif-
ferent parts of this evolutionary arch, although the reasons for the assignments to the
ascending and descending branches were often rhetorical rather than physical. Nonethe-
less, it was implicit in his scheme that there should exist large- and small-diameter stars
at the same temperature. There are similarities to Ritter’s more physical picture of stellar
evolution.
34 3 Stellar structure and evolution

3.2 The origin of the Hertzsprung–Russell diagram

With the publication of the Draper Memorial Catalogue of Stellar Spectra in 1890 (Picker-
ing, 1890), it became possible to make the first tests of the hypothesis that the stars evolved
down the spectral sequence from hot A and B stars to cool K stars.8 By 1893, William
Monck (1839–1915) and Jacobus Kapteyn had independently come to the conclusion that
something must be wrong with this simple picture (Kapteyn, 1892; Monck, 1895). Although
parallaxes are the best distance measures for stars, cruder information can be obtained from
their proper motions, meaning their apparent drift motions on the sky relative to very distant
background stars. Generally, nearby stars can have appreciable proper motions while distant
stars have none. Monck and Kapteyn discovered that, among the bright stars, those with the
greatest proper motions, and hence nearby, lowest-luminosity objects, were not the K and
M but the F and G stars. This was scarcely consistent with a scheme in which stars cooled
and grew fainter along the spectral sequence from O to M.
The breakthrough was made by the Danish astronomer Ejnar Hertzsprung (1873–1967).
Hertzsprung had trained as a chemical engineer in Copenhagen and St Petersburg and
then studied electrochemistry with Wilhelm Ostwald (1853–1932) in Leipzig. He returned
to Denmark in 1901 and, as an amateur, began his serious study of astronomy. He was
strongly influenced by the discovery by Max Planck (1858–1947) in 1900 of the formula
for black-body radiation and realised that, if it were assumed that the stars radiated like
black bodies and their distances were known, it is a straightforward calculation to work out
their physical sizes. In 1906 he showed that the diameter of Arcturus is roughly the same as
the diameter of the orbit of Mars – he immediately inferred that some very large stars must
exist (Hertzsprung, 1906).
Hertzsprung had already deduced that there must be a wide range of luminosity among
the stars. From the proper-motion data of Monck and Kapteyn, he inferred that statistically
the A and Orion-type B stars must be of high luminosity. In addition, he related these
data to the information on the distinction between the a and c class spectra, which was
part of Maury’s classification scheme (Maury and Pickering, 1897). It was immediately
apparent that the red c stars were distant luminous stars, with luminosities similar to that of
the A and B stars, while the non-c stars were low-luminosity nearby objects. As he stated
(Hertzsprung, 1905):

The result confirms the assumption of Antonia C. Maury that the c-stars show some intrinsic charac-
teristic.

The distinction between what became known as the dwarf and the giant stars was confirmed
by parallax studies (Hertzsprung, 1905, 1907). It turned out that, among the brightest stars
in the sky, there are more giants than dwarfs, and this accounted for Monck and Kapteyn’s
strange result that the red stars are more luminous than the yellow stars.
To Hertzsprung’s distress, when he received the Revised Harvard Photometry, published
in 1908, Cannon’s new spectral classifications were included, but Maury’s classifications,
which were based on the appearance of the spectral lines, were not mentioned. He wrote to
Pickering on 22 July 1908:
3.2 The origin of the Hertzsprung–Russell diagram 35

In my opinion the separation by Antonia C. Maury of the c- and ac- stars is the most important
advancement in stellar classification since the trials by Vogel and Secchi . . . To neglect the c-properties
in classifying stellar spectra, I think, is nearly the same thing as if the zoologist, who detected the
deciding differences between a whale and a fish, would continue in classifying them together.

Pickering was not convinced. He attached little significance to Miss Maury’s classification,
believing that the characteristics she had identified were too subtle to be real.
In 1909 Hertzsprung was invited by Karl Schwarzschild (1873–1916) to visit Göttingen,
where he was appointed associate professor. In the same year, Schwarzschild was appointed
to the directorship of the Potsdam Astrophysical Observatory and Hertzsprung joined him
there. In 1907, he had turned his attention to star clusters for which it can be safely assumed
that the stars are all at the same distance. As early as 1900, Schwarzschild had recognised
the importance of stellar colours in estimating their spectral types and defined the con-
cept of colour index as the difference between a star’s photographic and visual brightness
(Schwarzschild, 1900b).
A closely related concept was that of the effective wavelength, which described the mean
wavelength of the spectral energy distribution of the star and so was also a measure of its
colour. It had been shown by George Comstock (1855–1934), Director of the Washburn
Observatory in Wisconsin, that the effective wavelength was strongly correlated with spec-
tral type (Comstock, 1897), and this was the approach adopted by Hertzsprung to study
the colours of the stars. In 1911, he published the first luminosity–colour diagrams for the
Pleiades and Hyades star clusters (Hertzsprung, 1911). In these diagrams there was a promi-
nent continuous sequence of stars which he named the main sequence, but there was also
a very wide range of luminosity among the red stars (Figure 3.2). If attention is restricted
only to the main sequence, which is composed of dwarf stars, the red stars are indeed less
luminous than the yellow F and G stars and so the systematic trend of stars becoming
intrinsically fainter and redder along the spectral sequence was correct. These were the first
published colour–magnitude diagrams, which have dominated studies of stellar evolution
ever since. Notice that Hertzsprung’s research was based upon photometric studies of the
stars.
Independently, Henry Norris Russell (1877–1957) arrived at the same diagram by a rather
different route. Russell had graduated from Princeton in 1897 and completed his doctorate
in 1900. From 1902 to 1905, he worked at the Observatories at Cambridge, England, and
began one of the first photographic parallax programmes for stars. With Arthur Hinks (1873–
1945), the chief assistant at the University Observatory, Russell perfected the procedures
for measuring stellar parallaxes photographically. He returned to Princeton in 1905, and
the reduction of the parallax data was completed by 1910. In 1908, Russell made contact
with Pickering, who agreed to provide magnitudes and spectra for the 300 stars in the
parallax programme. The data were supplied by September 1909, and it was immediately
apparent that high- and low-luminosity red stars were present in the sample, a result similar
to Hertzsprung’s. Russell’s famous luminosity–spectral class diagram (Figure 3.3(a)) was
first published simultaneously in Nature and Popular Astronomy in 1914 (Russell, 1914a–e).
The scale on the ordinate is in absolute magnitudes, M (M = constant − 2.5 log L, where L
is the intrinsic luminosity of the star), and the spectral class is plotted along the abscissa. The
36 3 Stellar structure and evolution

Figure 3.2: The effective wavelength–apparent magnitude diagram for the Hyades star cluster pub-
lished by Hertzsprung in 1911. The vertical axis is the effective wavelength, which is proportional
to the colour, or colour index, of the star. The horizontal axis is the magnitude of the star. This was
the first published colour–magnitude diagram. The main sequence runs along the lower part of the
diagram and then moves to longer effective wavelengths at faint magnitudes (Hertzsprung, 1911).
The giant stars lie in the area above the main sequence in this form of colour–magnitude diagram.

correlation between spectral type and luminosity indicated by the bounding diagonal lines in
Figure 3.3(a) is apparent and corresponds to the main sequence described by Hertzsprung.
In addition, there are red stars above the main sequence, in a region which became known
as the giant branch. It can be seen that the luminosities of the K and M stars span about
10 magnitudes, corresponding to a factor of 10 000 in luminosity. It is intriguing that the
diagram includes one low-luminosity A star well below the main sequence. This is the star
40 Eridani B, which was the first white dwarf to be identified (see Section 4.2). Russell’s
famous papers also include the luminosity–spectral class diagram for four star clusters in
which the giant branch is more clearly defined (Figure 3.3(b)).
The luminosity–spectral class (or colour–magnitude) diagram was known as the ‘Russell
diagram’ until 1933 when Strömgren introduced the term Hertzsprung–Russell diagram,
recognising Hertzsprung’s key contributions. It is intriguing to compare Figures 3.2 and 3.3
with the Hertzsprung–Russell diagram, or H–R diagram, derived from observations made
by the Hipparcos astrometric satellite which includes 2927 stars for which parallaxes have
been measured to better than 5% accuracy (Figure 3.4).
Hertzsprung’s pioneering studies had shown how features in stellar spectra could be used
to determine whether stars are dwarfs or giants. Independently, Walter Adams (1876–1956)
and Arnold Kohlschütter (1883–1969) discovered other spectral features which could be
used as luminosity indicators (Adams and Kohlschütter, 1914b). By combining data on the
parallaxes, proper motions and spectral types, they discovered that, within a given spectral
class, certain spectral features were sensitive luminosity indicators. Specifically, for stars
3.2 The origin of the Hertzsprung–Russell diagram 37

Figure 3.3: The first published ‘Russell diagrams’ showing the relation between absolute magnitude
and spectral type. (a) The relation for all nearby stars. (b) The relation derived from studies of four
star clusters.

of the same spectral type, the low-luminosity stars had weaker ultraviolet continua than
the luminous stars, the hydrogen absorption lines were much stronger in the luminous stars
and certain metallic line ratios were shown to be sensitive luminosity indicators. In their
paper of 1914, they showed that, using these criteria, the absolute magnitudes of stars could
be estimated with an accuracy of about 1.5 magnitudes, corresponding to a factor of 4
in intrinsic luminosity. In consequence, the distances of stars could be roughly estimated
from the characteristics of their spectra alone. Although the distance estimates would be
uncertain by a factor of 2, this represented an enormous advance because, without this
additional information, a star’s absolute magnitude could be uncertain by more than 10
magnitudes, a factor of 10 000 in luminosity. The procedure was entirely empirical and
relied upon calibration of the relations using stars of known parallax. Distances estimated
in this way were referred to as spectroscopic parallaxes.
The luminosity indicators were built into the system of spectral classification which
eventually superseded the Harvard classification system. The Morgan, Keenan and Kellman
(MKK), or Yerkes, system was published in 1943 in their Atlas of Stellar Spectra, with an
Outline of Spectral Classification. A two-dimensional classification system was introduced
in which, in addition to the basic spectral types listed in Table 2.1, stars were assigned to five
luminosity classes from type I, the supergiant stars, to type V, main-sequence stars (Morgan,
38 3 Stellar structure and evolution

−5

0
Mv(mag)

10

1 10

15
−0.5 0.0 0.5 1.0 1.5 2.0
B–V(mag)

Figure 3.4: The Hertzsprung–Russell diagram for 2927 nearby stars for which parallaxes have been
determined with an accuracy of better than 5% by the Hipparcos astrometric satellite of the European
Space Agency. (M. A. C. Perryman, The Hipparcos and Tycho Catalogues, vol. 1 (Noorwijk, The
Netherlands: ESA Publication Division, ESTEC, ESA SP-1200, 1997).)

Keenan and Kellman, 1943). The modern MK system was described in 1953 and involves
only minor changes as compared with the MKK system (Johnson and Morgan, 1953).
Another of the great problems of stellar astronomy was the determination of stellar
masses, which could only be found for stars which are members of binary star systems. Pio-
neering analyses for the determination of stellar sizes and masses were carried out by Russell
using the light curves of eclipsing binary star systems. In 1912, these techniques were elab-
orated in collaboration with Russell’s first graduate student, Harlow Shapley (1885–1972)
(Russell, 1912a,b; Russell and Shapley, 1912a,b). Shapley went on to make a detailed study
of the light curves of 90 eclipsing binary stars, and from these he was able to demonstrate
that there is a wide range of diameters among these stars. The classical techniques of dynam-
ical astronomy for binary star systems could be used in conjunction with accurate radial
velocities to make estimates of the masses of the stars. Shapley found that some of them had
mean densities similar to the Sun, while the very brightest yellow and red stars were giant
stars of much lower mean density (Shapley, 1915). It soon became apparent that the range
of luminosities was enormous compared with the range of masses. Russell found at best
weak evidence for a correlation between luminosity and mass for the stars in his samples.
 3.3 The impact of the new physics 39

This evidence was consistent with the prevailing view, favoured by Russell, that the red
giant stars represented the earliest phases of stars, which then contracted and heated up to
join the upper end of the main sequence. The main sequence then represented a cooling
sequence for stars as they grew older. Any weak correlation of mass with luminosity could
be attributed to mass loss. A relic of these early, and quite incorrect, theories remains in the
use of the term early-type stars to mean stars on the upper part of the main sequence and
late-type stars for those on the lower main sequence.

3.3 The impact of the new physics

Within ten years, the picture would change dramatically, many pieces of evidence con-
tributing to these profound changes. Continuing the theme of the mass–luminosity relation
for dwarf stars, contrary to Russell’s assertion that the dwarf stars on the main sequence
had similar masses, first Jacob Halm (1866–1944) in 1911 and then Hertzsprung in 1915
showed that there is a correlation between mass and luminosity along the main sequence
(Halm, 1911). By 1919, Hertzsprung had derived an empirical mass–luminosity relation
for main-sequence stars, L ∝ M x with x ≈ 7 (Hertzsprung, 1919), somewhat greater than
present best values, which are closer to x = 4 for stars with mass roughly that of the Sun.
This finding ran contrary to the expectations of the Russell–Lockyer theory, according to
which stars on the main sequence have the same mass. To rescue the standard theory, the
stars would have to lose mass.
Another important development was the idea that energy could be transported through
the gaseous envelope of the star by radiation rather than by convection. This concept had first
been discussed by Ralph Sampson (1866–1939) in 1894 (Sampson, 1895) and studied in
more detail by Arthur Schuster (1851–1934) and Schwarzschild in the early 1900s (Schus-
ter, 1902, 1905; Schwarzschild, 1906). Specifically, Schwarzschild showed that convection
would only occur if the temperature gradient exceeded the adiabatic gradient for the gas
in the envelope. The radiative transport of energy did not, however, find favour with the
majority of astronomers, but in 1916 Arthur Eddington revived the idea and applied it to
radiative transfer in the envelopes of giant stars (Eddington, 1916b).†
Bohr’s theory of atomic structure, published in 1913, had an immediate impact upon
astrophysics (Bohr, 1913a–c). It enabled the energy levels of elements in different states
of ionisation to be determined, and this had important implications for the measurement
of the temperatures of stellar atmospheres. The earliest attempts to measure the surface
temperatures of the stars had assumed that they emitted like black bodies, the technique
which had been used by Hertzsprung to measure the diameter of Arcturus (Hertzsprung,
1907). This technique had been used to measure temperatures by a number of workers, but it
suffered from the problem of taking proper account of the presence of absorption lines in the
stellar spectra. To estimate the continuum intensity, observations had to be made between
the prominent absorption lines. There remained the problem, however, of the unknown

† The radiative transport of energy in stars and a derivation of the fourth equation of stellar structure are presented
in Section A3.2.2.
40 3 Stellar structure and evolution

extent to which weak absorption lines depressed the continuum, a phenomenon known as
line blanketing, as well as the problem of the Balmer jump, the discontinuity which takes
place at the limit of the Balmer series, for late B and A type stars.
The first astronomer to apply the idea of using the state of ionisation of atoms in stellar
atmospheres as a means of measuring temperatures was the Indian astrophysicist Megh
Nad Saha (1893–1956). In 1919, Saha visited the German physical chemist Walther Nernst
(1864–1941), who was studying the thermodynamic theory of the equilibrium state of
chemical reactions. Saha acknowledged that this work was the inspiration for his formulation
of equilibrium ionisation states. In his own words (Saha, 1920), he described ionisation as

a sort of chemical reaction, in which we have to substitute ionisation for chemical decomposition.

John Eggert (1891–1973), a pupil of Nernst, had already calculated the equilibrium
state for eight-times ionised iron in stellar interiors (Eggert, 1919), and Saha applied the
same formalism to studies of the solar atmosphere. These considerations led to the Saha
equation, which describes the state of ionisation of a gas in thermal equilibrium at a given
temperature (Saha, 1920). Saha combined Boltzmann’s equation with the equations of
ionisation equilibrium and so determined how the state of ionisation depends upon both
the density and the temperature of the gas. To estimate temperatures, he used the method
of ‘marginal appearances’ of lines based upon the first appearance or disappearance of the
different spectral lines employed in the Harvard sequence of stellar types listed in Table 2.1.
He concluded his important paper of 1921 with the remark that (Saha, 1921)

It will be admitted from what has gone before that the temperature plays the leading role in determining
the nature of the stellar spectrum. Too much importance must not be attached to the figures given, for
the theory is only a first attempt for quantitatively estimating the physical processes taking place at
high temperature. We have practically no laboratory data to guide us, but the stellar spectra may be
regarded as unfolding to us, in an unbroken sequence, the physical processes succeeding each other
as the temperature is continually varied from 3000 K to 40 000 K .

These concepts were developed by Ralph Fowler (1889–1944) and Edward A. Milne
(1896–1950), who provided a much more complete description of the equilibrium ionisa-
tion states, including the effects of excited states of atoms and ions (Fowler and Milne,
1923, 1924). Rather than simply use the first appearance or disappearance of different ions
and atoms, they worked out the conditions under which the absorption lines would have
maximum strength. The way was opened up for determining the abundances of the elements
in detail, and this task was undertaken by Cecilia Payne (1900–1979), a pupil of Milne’s,
who carried out these studies at Harvard under the supervision of Shapley. Her doctoral
degree was the first awarded by Harvard in astronomy, and it was published in 1925 as
a monograph with the title Stellar Atmospheres (Payne, 1925). According to Otto Struve
(1897–1963),9

It is undoubtedly the most brilliant Ph.D. thesis ever written in astronomy.

Payne’s dissertation concerned the application of the Saha–Fowler–Milne theory to stellar

atmospheres and summarised everything that was known about laboratory and stellar spectra
at that time. The most famous aspect of her work was the demonstration that, although the
3.3 The impact of the new physics 41

spectra of stars can vary widely, they all have remarkably similar chemical compositions,
the principal cause of the observed differences being the surface temperature of the star. In
her monograph, she stated that

the uniformity of composition of stellar atmospheres appears to be an established fact.

She further showed that these abundances were similar to the terrestrial abundances, with
the exception of the elements hydrogen and helium, which she found to be vastly more
abundant in the stars than on Earth. Although she had obtained the correct answer, she did
not believe it. She wrote:

Although hydrogen and helium are manifestly very abundant in stellar atmospheres, the actual values
derived from the estimates of their marginal appearances are regarded as spurious.

This conclusion simply reflected the prevailing prejudice. Three years later, in 1928,
Albrecht Unsöld (1905–1995) showed that the abundance of hydrogen was indeed very
much greater than all the other elements (Unsöld, 1928). This was confirmed by William
McCrea (1904–1999), who used the relative intensities of flash spectra to show that the num-
ber density of hydrogen atoms at the base of the chromosphere was the same as Unsöld’s
value (McCrea, 1929).
Payne’s brilliant analyses indicated the power of spectroscopy in determining the chem-
ical abundances and physical conditions in stellar atmospheres. The story was taken up
by Russell, who had been one of the pioneers in testing Saha’s theory by comparing the
relative intensities of the spectral lines of potassium and rubidium in the solar atmosphere
and in sunspots – he found excellent agreement with the theory (Russell, 1922). In 1925, he
investigated the problem of understanding the anomalous triplet terms of the alkaline earth
metals, calcium, scandium and barium. In collaboration with Frederick Saunders (1875–
1963), he developed the vector model of the atom of Alfred Landé (1888–1976) to account
for what became known as Russell–Saunders or L–S coupling (Russell and Saunders, 1925).
With this new understanding of atomic spectra, Russell, Walter Adams and Charlotte Moore
(1898–1990) began a detailed study of the chemical abundances in the solar atmosphere
(Russell, Adams and Moore, 1928). They used 1288 absorption lines in 228 different mul-
tiplets to find the relation between the strengths of the absorption features and the number
of absorbing atoms or ions. In agreement with the work of Unsöld and McCrea, they found
that hydrogen was by far the most abundant of the elements. In Russell’s analysis of 1929,
the solar abundances for 56 different elements and 6 diatomic molecules were determined
(Russell, 1929). These abundances were all within a factor of 2 of present estimates.
These procedures were adapted for the determination of the abundances of elements
in the stars by Marcel Minnaert (1893–1970) and Gerard Mulders (1908–1993) in 1930
(Minnaert and Mulders, 1930). They introduced the concept of the equivalent width of
the spectral line, meaning the waveband of continuum radiation of the star which would
correspond to the same amount of radiation removed from the continuum by integrating over
the observed line profile. They developed the procedures for relating the equivalent widths
of lines in the spectrum to the number of absorbing atoms, taking account of the different
processes which broaden the absorption lines – radiation damping, natural damping and
thermal broadening. This led to what Minnaert called the curve of growth technique for
42 3 Stellar structure and evolution

Figure 3.5: The curve of growth due to Minnaert and Mulders showing the dependence of the abun-
dance of an element upon its equivalent width (Minnaert and Mulders, 1930). The continuous curve
shows a mean curve of growth for a wide range of equivalent widths.

relating the equivalent width of the line to the number of absorbing atoms (Figure 3.5). The
same type of procedure was developed independently by Donald Menzel (1901–1976) in
1930 for emission lines (Menzel, 1931). These became the standard techniques for analysing
the abundances of the elements in stars, in particular for investigating chemical differences
between stars as a means of determining their evolutionary status.

3.4 Eddington and the theory of stellar structure and evolution

Arthur Stanley Eddington (1882–1944) was the central figure in the development of the
theory of the internal structure and evolution of the stars. After a very distinguished under-
graduate career in mathematics at Cambridge University, where he became senior wrangler
in only his second year, in 1906 he became a senior assistant at the Royal Greenwich Obser-
vatory, where he obtained valuable experience of practical astronomy. In 1913, he returned
to Cambridge as the Plumian Professor of Astronomy, where he was to remain for the rest
of his career.10
Between 1916 and 1924, Eddington published over a dozen papers, which were collected
and extended in his great book The Internal Constitution of the Stars (Eddington, 1926b).
According to Henry Norris Russell (Russell, 1925), whose theory of stellar evolution was
comprehensively demolished by Eddington,

Several investigators – Jeans, Kramers, Eggert – have contributed to this field, but much the largest
share is Eddington’s.
 3.4 Eddington and the theory of stellar structure and evolution 43

In a perceptive letter to me in 1993, William McCrea wrote

[People] don’t realise that before, say, 1916 astronomers simply had no idea what the inside of a
star was like, and had no idea how to find out anything about this. The speed at which Eddington
transformed the sitution was incredible.

There is no simpler way of describing Eddington’s achievement than to quote Chan-

drasekhar’s assessment:11
In the domain of the internal constitution of the stars, Eddington recognised and established the
following basic elements of our present understanding.
(i) Radiation pressure must play an increasingly important role in maintaining the equilibrium of
stars of increasing mass.
(ii) In parts of the star in which radiative equilibrium, as distinct from convective equilibrium, obtains,
the temperature gradient is determined jointly by the distribution of the energy sources and of
the opacity of the matter to the prevailing radiation field. Precisely,
d pr L(r ) 1 4
= −κ ρ, pr = aT (3.4)
dr 4πcr 2 3
and
r
L(r ) = 4π ρr 2 dr, (3.5)
0

where pr , κ,  and ρ denote, respectively, the radiation pressure, the coefficient of stellar opacity,
the rate of energy generation per gram of stellar material, and the density.† Also a is Stefan’s
constant and c is the velocity of light.
(iii) The principal physical processes contributing to the opacity, κ, is determined by the photo-electric
absorption coefficient in the soft X-ray region, i.e., by the ionisation of the innermost K - and
L-shells of highly ionised atoms.
(iv) With electron scattering as the ultimate source of stellar opacity, there is an upper limit to the
luminosity, L, that can support a given mass M. The maximum luminosity, set by the inequality
4π cG M
L< , (3.6)
σe
where σe denotes the Thomson scattering-coefficient, is now generally referred to as the Edding-
ton limit. This Eddington limit plays an important role in current investigations relating to X-ray
sources and the luminosity of accretion discs around black holes.‡
(v) In a first approximation, in normal stars (that is in stars along the main sequence), the (mass,
luminosity, effective temperature)-relation is not very sensitive to the distribution of the energy
sources through the star. Therefore, a relation is available for comparison with observations even
in the absence of a detailed knowledge of the energy sources of the star.
(vi) The burning of hydrogen into helium is the most likely source of stellar energy.
(vii) The phenomenon of Cepheid variability is due to the adiabatic radial pulsations of these stars.

These great insights were not gained without a considerable struggle and, in particular,
there were continuing disputes with James Jeans (1877–1946) and others about many of the
fundamental issues concerning the internal structure of the stars.12 Many of these heated
discussions are faithfully recorded in the Reports of the meetings of the Royal Astronomical

† The origin of the third and fourth equations of stellar structure are described in Section A3.2.
‡ This result is derived in the context of the physics of active galactic nuclei in Section A11.1.1.
44 3 Stellar structure and evolution

Society published in The Observatory through this remarkable era. Some appreciation of
the heat which was generated may be gained from this extract from a letter which Jeans
published in The Observatory in November 1926 (Jeans, 1926):

May I conclude by assuring Prof. Eddington it would give me great pleasure if he could remove a
long-standing source of friction between us by abstaining in future from making wild attacks on my
work which he cannot substantiate, and by making the usual acknowledgements whenever he finds that
my previous work is of use to him? I attach all the more importance to the second part of the request
because I find that some of the most fruitful ideas which I have introduced into astronomical physics –
e.g., the annihilation of matter as the source of stellar energy, and highly dissociated atoms and free
electrons as the substance of the stars – are by now fairly generally attributed to Prof. Eddington.

The problem was that, to determine theoretically the internal structure of a star, the
four equations of stellar structure, equations (A3.1), (A3.2), (A3.13) and (A3.19) had to
be supplemented by knowledge of the equation of state of the stellar material, as well as
the density and temperature dependence of the energy generation rate, ε(ρ, T ), and the
opacity of the stellar material, κ(ρ, T ). This was far beyond what was feasible in 1916.
Some sweeping approximations were needed to make progress, and Eddington had exactly
the right qualities of fearless imagination and technical skill to make the problem tractable.
For example, in order to simplify the mathematics of his ‘standard model’, Eddington made
what Leon Mestel (b. 1927) refers to as the ‘hair-raising approximation’ that the radiation
pressure is a constant fraction of the total pressure throughout the star. It is no surprise that
the subject of stellar structure provoked heated debate.
In his first paper on the internal structure of the stars, Eddington assumed that the mean
atomic mass of the particles was 54, meaning that the star was predominantly composed of
iron atoms (Eddington, 1916b). As soon as his paper was read at the Meeting of the Royal
Astronomical Society on 8 December 1916, Jeans pointed out in the discussion that (Jeans,
1917)
For these temperatures and energy, we have very hard Roentgen radiation, and so the atoms in the gas
will be smashed up.

In the opening paragraph of his next paper, Eddington acknowledged this important
contribution (Eddington, 1917), stating that

Jeans has convinced me that a rather extreme state of disintegration is possible, and indeed seems
more plausible.

Eddington adopted a mean atomic weight of 2, corresponding to the complete ionisation

of the atoms, assuming that there is a negligible fraction of hydrogen present. As noted in
Section 3.3, it was only in the early 1930s that the high cosmic abundance of hydrogen was
established. The adoption of fully ionised gas as the material of the star was an important
change of perspective in that, if all the atoms of the stellar material are fully ionised, the
perfect gas law could be applied at very much higher densities and temperatures than are
found in terrestrial environments. As a bonus, this change of the mean atomic weight enabled
Eddington to obtain better agreement between his theory and the observed masses of the
stars.
Eddington still adhered to the prevailing Russell–Lockyer picture and so the theory
could not be applied to main-sequence stars. The analyses of the properties of red giants by
3.4 Eddington and the theory of stellar structure and evolution 45

Hertzsprung and Russell had indicated, however, that they were of enormous size and so
their very-low-density envelopes were likely to be gaseous. Eddington was inspired to apply
his theory of radiative transfer to the envelopes of the red giant stars. In his paper of 1917,
he showed that the luminosity is predominantly determined by the mass of the star and that
the meagre observational data on their mass–luminosity relation was in good agreement
with the theory. Furthermore, if the release of gravitational energy was the source of the
luminosity of the giant stars, they could not radiate for more than 100 000 years, which is
very much less than the age of the Earth (Eddington, 1917). To avoid the short timescales,
another source of energy had to be found. At this time, his view was that

Probably the simplest hypothesis . . . is that there may be a slow process of annihilation of matter
(through positive and negative electrons occasionally annulling one another).

Eddington had little doubt about the correctness of his theory of red giants, and in 1919
the opportunity arose of testing directly that they have the large diameters inferred from
theory. Albert A. Michelson (1852–1931) had been developing the techniques of optical
interferometry for almost 30 years (Michelson, 1890) and George Ellery Hale, director of
the Mount Wilson Observatories, decided that the 100-inch Hooker telescope should be
equipped with a Michelson interferometer to determine the separations of close binary stars
and, potentially, the diameters of stars. Michelson originally thought that impractically long
interferometer baselines would have to be used to measure the diameters of red giants but,
being aware that the instrument was in the process of construction, Eddington used his theory
of the structure of red giant envelopes to predict the angular size of Betelgeuse. In the light
of this prediction, Michelson built a 6-metre interferometer, which was mounted on the top
ring of the 100-inch telescope by Francis Pease (1881–1938) and John A. Anderson (1876–
1959) (Figure 3.6). On the night of 13 December 1920, they measured the angular diameter
of Betelgeuse to be 0.047 arcsec, just slightly less than Eddington’s prediction (Michelson
and Pease, 1921). This observation confirmed beyond any doubt the large diameters of the
red giants, that of Betelgeuse being greater than the diameter of the Earth’s orbit about the
Sun. Michelson and his colleagues went on to measure the diameters of four other red giants
(Pease, 1921).
In 1919, Eddington discovered to his surprise that his theory of the structure of red
giants could also account for the observed mass–luminosity relation for main-sequence stars
(Figure 3.7) (Eddington, 1924). The implications were profound – the main-sequence stars
were not slowly contracting incompressible liquid spheres, but rather gaseous spheres. This
cut the foundation from under the standard Russell picture. The conclusion was vigorously
opposed by Jeans, who believed the result was spurious since Eddington’s standard model
avoided addressing the problem of the processes of energy generation inside the stars. Jeans
proposed that the source of energy in the Sun was radioactive decay.
In his early papers in this series, Eddington advocated the annihilation of matter as an
inexhaustible source of energy for the stars. In 1920, he realised that, although there was
no known mechanism by which nuclear energy could be released, at least energetically
this provided a very attractive means of powering the stars. In a remarkably prescient
paragraph of his Presidential Address to the Mathematical and Physics Section of the
British Association for the Advancement of Science at its Annual Meeting, held in Cardiff,
he stated (Eddington, 1920)
46 3 Stellar structure and evolution

Figure 3.6: Michelson’s interferometer mounted on the top ring of the Hooker 100-inch telescope
at Mount Wilson. This instrument was used to measure the angular diameter of Betelgeuse in 1919
(Michelson and Pease, 1921).

−4

−2

0
absolute magnitude

6
first-class
second-class
8 Cepheids
eclipsing variable

10

12

−0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
log mass

Figure 3.7: The observed mass–luminosity relation for stars compared with Eddington’s theoretical
mass–luminosity relation (Eddington, 1924). The different symbols refer to different classes of star
for which Eddington had made estimates of luminosities and masses. More than half the stars are
main-sequence stars.
 3.4 Eddington and the theory of stellar structure and evolution 47

Certain physical investigations in the past year . . . make it probable to my mind that some portion of
this sub-atomic energy is actually being set free in the stars. F. W. Aston’s experiments seem to leave
no room for doubt that all the elements are constituted out of hydrogen atoms bound together with
negative electrons. The nucleus of the helium atom, for example, consists of 4 hydrogen atoms bound
with two electrons. But Aston has further shown conclusively that the mass of the helium atom is less
than the sum of the masses of the 4 hydrogen atoms which enter into it; and in this at any rate the
chemists agree with him. There is a loss of mass in the synthesis amounting to about 1 part in 120, the
atomic weight of hydrogen being 1.008 and that of helium 4. . . . Now mass cannot be annihilated, and
the deficit can only represent the mass of the electrical energy set free in the transmutation. We can
therefore at once calculate the quantity of energy liberated when helium is made out of hydrogen. If
5 per cent of the star’s mass consists initially of hydrogen atoms, which are gradually being combined
to form more complex elements, the total heat liberated will more than suffice for our demands, and
we need look no further for the source of a star’s energy.

Eddington had the good fortune to be working at the Observatories at Cambridge Uni-
versity, only a 20 minute walk from the Cavendish Laboratory, where Francis Aston (1877–
1945) was carrying out his precise measurements of atomic and isotopic masses. At that
time, this could be no more than a hypothesis, but Eddington had indeed hit upon the correct
solution for the energy source of the Sun. Note that Eddington’s insight was based upon the
prevailing view at the time that the nucleus consisted of protons and electrons. Although
Rutherford had postulated as early as 1920 the existence of neutrons as the particles which,
along with the protons, made up the total mass of the nucleus, there was little enthusiasm,
and even less evidence, for their existence (Rutherford, 1920). The beauty of Eddington’s
argument was that it did not depend upon the precise nature of the nucleus, but only upon
the conservation of energy and the mass–energy relation E = mc2 .
It is hardly surprising that the unravelling of the internal structure of the stars resulted in
heated debate. In Leon Mestel’s critique13 of Eddington’s The Internal Constitution of the
Stars, he shows that Eddington made a number of rough approximations and used quite a
bit of sleight of hand in order to find tractable solutions of the equations. In fact, Eddington
was lucky in that, as noted by Chandrasekhar and Mestel, the form of the mass–luminosity
relation is remarkably independent of the precise process of energy production and the
opacity law.† As Mestel remarks, even in an extreme model in which the energy is assumed
to originate in a point source at the centre of the star, Thomas Cowling (1906–1990) found
that the resulting variation of density with radius was not so different from Eddington’s
standard model (Cowling, 1935).
The scene was set for probing much more deeply into the physical processes that deter-
mine the opacity of stellar material and its energy generation rate. In 1923, Hendrik Kramers
(1894–1945) used classical arguments to work out the opacity of a fully ionised plasma for
bremsstrahlung, what become known as free–free radiation in the post-quantum era, and
which is a good approximation for the central regions of stars like the Sun (Kramers, 1923).
To obtain the average energy flux transmitted through the star, the absorption coefficient has
to be averaged over all frequencies; the procedure for doing this was worked out by Svein
Rosseland (1894–1985) in 1924, the resulting opacities being known as Rosseland mean

† These weak dependences of the surface properties of main-sequence stars upon the energy generation rate and the
opacity law can be appreciated from the analyses given in Section A3.3.
48 3 Stellar structure and evolution

opacities (Rosseland, 1924) (see Section A3.3.3). The strong dependence of nuclear energy
generation rates upon temperature was established in the 1930s once the great discover-
ies of quantum mechanics had been assimilated into the toolkit of the astrophysicist (see
Section 3.5).

3.5 The impact of quantum mechanics and the discovery of

new particles

The solution of the problem of energy generation in the Sun was one of the fruits of the
remarkable theoretical developments which led to the discoveries of quantum mechanics
and Dirac’s theory of the electron. Specifically, the key developments were the following.
r The discovery of Fermi–Dirac statistics, which found immediate application in the
equation of state of dense matter in stars (Fermi, 1926).
r The phenomenon of quantum mechanical tunnelling, discovered by George Gamow
(1904–1968), and its application to the inelastic scattering of α-particles by nuclei
(Gamow, 1928).
r The formulation of the theory of β-decay by Enrico Fermi (1901–1954), which included
the proposal of the existence of the neutrino (Fermi, 1934a,b).
Equally important were the experimental discoveries of this golden age of physics.
r Nuclear transmutations were discovered by Ernest Rutherford in collisions between fast
α-particles and nitrogen nuclei (Rutherford, 1919; Rutherford and Chadwick, 1921),
and these were photographed in the remarkable automatic cloud chamber experiments
of Patrick Blackett (1897–1974) (Blackett, 1925).
r The discovery of the positron in cosmic-ray cloud-chamber experiments was announced
by Carl Anderson (1905–1991) in 1931 (Anderson, 1932).
r In the same year, Harold Urey (1893–1981) and his colleagues discovered deuterium
in spectroscopic studies of ‘distilled’ liquid hydrogen (Urey, Brickwedde and Murphy,
1932).
r Also, in 1932, the neutron was discovered by James Chadwick (1891–1974) (Chadwick,
1932).
r Finally, in 1932, the experiments of John Cockcroft (1897–1967) and Ernest Walton
(1903–1995) not only provided the first destruction of lithium nuclei by artificially
accelerated fast protons, but also provided direct experimental confirmation of the
correctness of the mass–energy relation E = mc2 (Cockcroft and Watson, 1932).
These discoveries were quickly assimilated into astrophysics.
From the perspective of astrophysics, a key development was the much improved under-
standing of nuclear structure. Eddington’s proposal that nuclear energy could provide the
luminosity of the Sun could now be placed on a proper physical basis.
The problem was that, even at the high temperatures of stellar interiors, the Coulomb
repulsion between protons and nuclei is so great that, according to classical physics, protons
could not pentrate the nucleus and so this energy source could not be tapped. The solution
3.5 Quantum mechanics and the discovery of new particles 49

of this problem had to await Gamow’s theory of quantum mechanical tunnelling of 1928
(Gamow, 1928). One year later, Robert Atkinson (1893–1981) and Fritz Houtermans (1903–
1966) applied Gamow’s theory to the physics of nuclear reactions in the hot central regions
of stars (Atkinson and Houtermans, 1929). By considering the process of barrier penetration
by a Maxwellian distribution of protons, they established two key features of the process
of nuclear energy generation in stars. Firstly, the most effective energy sources involve
interactions with nuclei of small electric charge since the Coulomb barriers are smaller
than for nuclei with large charges. Secondly, the particles which can penetrate the Coulomb
barriers are those few particles in the high-energy tail of the Maxwellian distribution. As
a result, nuclear reactions can take place at temperatures which are considerably smaller
than might have been expected. These ideas also suggested why the luminosity of the stars
should be a sensitive function of temperature. As the temperature increases, the rate of
barrier penetration increases exponentially and so hotter stars should be more luminous
than less massive stars.
By 1931, the evidence was accumulating that hydrogen is by far the most abundant
element in the stars. In addition to the studies of stellar atmospheres discussed above,
Bengt Strömgren had shown that the mass–luminosity relation is sensitive to the hydro-
gen abundance and, using opacities in which hydrogen constituted about 35% by mass,
much improved agreement with the observations was obtained (Strömgren, 1932, 1933).
Reluctantly, Eddington agreed with Strömgren’s conclusion, but the adopted abundances of
the heavy elements was still large. This was the abundance used by Chandrasekhar in his
influential book An Introduction to the Study of Stellar Structure (Chandrasekhar, 1939).
In the 1930s, it was realised that even lower abundances of the heavy elements would also
be consistent with the data, but it was only in the 1940s that the heavy-element abundance
was further revised downwards, close to the values used today, as a result of rather general
astrophysical arguments by Fred Hoyle (Hoyle, 1946).
Atkinson’s objective was to account for the origin of the chemical elements by the
successive addition of protons to nuclei. He argued that the process of forming helium by
the combination of four protons was very unlikely and proposed instead that helium could
be formed by the successive addition of protons to heavier nuclei which, when they became
too massive for nuclear stability, would eject α-particles and so create helium (Atkinson,
1931a,b). This proposal was the precursor of the carbon–nitrogen–oxygen (CNO) cycle,
which was discovered independently by Carl von Weizsäcker (b. 1912) and Hans Bethe
(1906–2005) in 1938 (Weizsäcker, 1937, 1938; Bethe, 1939). In this cycle, carbon acts as a
catalyst for the formation of helium through the successive addition of protons accompanied
by two β + decays as follows:

12
C+p→ 13
N+γ ; 13
N→ 13
C + e + + νe ; 13
C+p→ 14
N + γ;
+
14
N+p→ 15
O+γ ; 15
O→ 15
N + e + νe ; 15
N + p → He +
4 12
C.

In the meantime, it had become possible to make estimates of the reaction rates for the
simplest nuclear reaction, the combination of pairs of protons to form deuterium nuclei
which can then combine with other deuterons to form 3 He and 4 He. The first calculations
were carried out by Atkinson in 1936 (Atkinson, 1936) and were much refined in 1938
50 3 Stellar structure and evolution

by Bethe and Charles L. Critchfield (1910–1994), who combined Fermi’s theory of weak
interactions with Gamow’s theory of barrier penetration (Bethe and Critchfield, 1938). The
principal series of reactions in the proton–proton (or p–p) chain are as follows:†

p + p → 2 H + e+ + νe ; 2
H + p → 3 He + γ ;
3
He + 3 He → 4 He + 2p.

The crucial first reaction in the chain involves a weak interaction, in which a positron and
neutrino are released in what may be thought of as the transformation of one of the protons
into a neutron. This reaction accounts for most of the energy release in the p–p chain but
it has never been measured experimentally at the energies of interest for nucleosynthesis
in the Sun. Bethe and Critchfield showed that this series of reactions could account for
the luminosity of the Sun. In addition, they found that the rate of energy production, ε, of
the p–p chain depends upon the central temperature of the star as ε ∝ T 4 . In 1939, Bethe
worked out the corresponding energy production rate for the CNO cycle and found a very
much stronger dependence, ε ∝ T 17 (Bethe, 1939). He concluded that the CNO cycle was
dominant in massive stars, whereas the p–p chain was the principal energy source for stars
with mass M ≤ M . These conclusions were confirmed by the much more detailed models
of stellar structure which became available after the Second World War and, in particular,
with the development of computer codes, which have converted the study of stellar structure
into one of the most precise of the astrophysical sciences.
The laws of energy generation, ε(ρ, T ), and the dependence of the opacity of stellar
material upon density and temperature, κ(ρ, T ), were the keys to understanding the theory of
stellar structure and evolution in much more detail. Using simple power-law approximations
for the ε and κ, Fred Hoyle (1915–2001) and Raymond Lyttleton (1911–1995) were able to
derive homology relations, which illustrated clearly how the properties of stars depended
upon the different processes of energy generation and upon the opacity law (Hoyle and
Lyttleton, 1942). A simple illustration of the power of these methods is given in Section
A3.3.
Thus, by the time of the outbreak of the Second World War, much of the basic physics
of main-sequence stars was beginning to be understood.

Notes to Chapter 3

1 I have given an account of the origins of the two laws of thermodynamics in Case Study IV of
Malcolm Longair, Theoretical Concepts in Physics (Cambridge: Cambridge University Press,
2003).
2 The same suggestion had been made earlier by Mayer and John James Waterston (1811–1883) in
two independent unpublished papers (Virginia Trimble, personal communication).
3 This quotation is contained in the biography of Lord Kelvin by C. W. Smith and M. N. Wise,
Energy and Empire: A Biographical Study of Lord Kelvin (Cambridge: Cambridge University
Press, 1989).

† More details of these reactions are given in Section 8.2 in the context of the generation of the flux of solar neutrinos.
 Explanatory supplement to Chapter 3 51

4 Chandrasekhar, S., as quoted by Wali, K. C. Chandra: A Biography of S. Chandrasekhar (Chicago:

University of Chicago Press, 1991).
5 These results are contained in Lane’s unpublished notes in the US National Archives.
6 My favorite recommendations as introductions to the theory of the structure and evolution of the
stars are R. J. Tayler, The Stars: Their Structure and Evolution (Cambridge: Cambridge University
Press, 1994) at the elementary level and R. Kippenhahn and A. Weigert, Stellar Structure and
Evolution (Berlin: Springer-Verlag, 1990) at a more advanced level.
7 Lockyer’s first version of his temperature arch was published in 1887 in his paper ‘Researches on
the spectra of meteorites. A report to the Solar Physics Committee’, published in Proceedings of
the Royal Society of London, 43, 1887, 117–156. He continued to publish various versions of the
temperature arch; that shown in Figure 3.1 dates from 1914.
8 An excellent account of the early history of the theory of stellar structure and evolution is given
by David DeVorkin, Stellar evolution and the origin of the Hertzsprung–Russell Diagram, in The
General History of Astronomy, Vol. 4. Astrophysics and Twentieth-Century Astronomy to 1950:
Part A, ed. O. Gingerich (Cambridge: Cambridge University Press, 1984), pp. 90–108.
9 This remark appears on p. 220 of the book by O. Struve and V. Zebergs, Astronomy of the 20th
Century (New York: Macmillan and Company, 1962).
10 Eddington has been the subject of a number of biographical studies, including A. Vibert Douglas,
Arthur Stanley Eddington (London: Thomas Nelson and Sons Ltd, 1956), S. Chandrasekhar,
Eddington: The Most Distinguished Astrophysicist of His Time (Cambridge: Cambridge University
Press, 1983) and D. S. Evans, The Eddington Enigma (Princeton, New Jersey: Xlibris Corporation,
1998).
11 Chandrasekhar’s short book Eddington: The Most Distinguished Astrophysicist of His Time is
essential reading for anyone wishing to understand Eddington’s thinking on astrophysical and
cosmological problems. The list of six insights into the physics of the stars in that book was
expanded to seven in the Preface, which Chandrasekhar wrote as an introduction to the 1988
reprint of Eddington’s The Internal Constitution of the Stars (Cambridge: Cambridge University
Press). I have included the list of seven in the present text.
12 It is interesting that Eddington acknowledged that Jeans had adopted different positions on many
of the key issues in his bibliographical references at the end of The Internal Constitution of the
Stars, p. 402, but left it to the reader to judge their importance.
13 Mestel’s penetrating critique was presented at a meeting of the Royal Astronomical Society on 12
March 2004 to celebrate the 60th anniversary of Eddington’s death. It is published as L. Mestel,
Stellar Structure and Stellar Atmospheres (Cambridge: Cambridge University Press, 2005).

A3 Explanatory supplement to Chapter 3

A3.1 The virial theorem and Homer Lane’s insights
We can demonstrate Lane’s important result quantitatively from the first two equations of
stellar structure. The first is the equation of hydrostatic equilibrium:
dp G Mρ
=− 2 , (A3.1)
dr r
which must apply at all points in a quasi-static star. By quasi-static, we mean that, although
the star is losing energy and so changing slowly, the star can be considered to be in hydrostatic
equilibrium to a very high degree of accuracy so far as its structure is concerned. This
equation describes the balance between the gravitational attraction of the mass M = M(≤r )
within radius r and the pressure gradient of the hot gas pushing outwards.
52 3 Stellar structure and evolution

The second is the equation of conservation of mass, namely that the mass in the spherical
shell of thickness dr is dM = 4πr 2 ρ dr and so
dM
= 4πr 2 ρ. (A3.2)
dr
Note that in this analysis M is a variable, the mass within radius r inside the star.
We first use these equations to derive a form of the virial theorem as applied to stars in
hydrostatic equilibrium. Dividing equation (A3.1) by equation (A3.2), we find
dp GM
=− . (A3.3)
dM 4πr 4
Reorganising equation (A3.3) and integrating from the centre to the surface of the star, we
find
 
GM
4πr d p = 3V d p = −
3
dM, (A3.4)
r

ps Ms  
GM
3V d p = − dM. (A3.5)
pc 0 r
We recognise that the quantity on the right-hand side of equation (A3.5) is the total
gravitational energy of the star, which we write as , noting that is a negative quantity.
Assuming that the surface pressure, ps , is negligible compared with the central pressure pc ,
we integrate the left-hand side by parts to find
Vs
−3 p dV = . (A3.6)
0

Finally, we write dV in terms of the corresponding mass element dM, dM = ρ dV , and so

Ms
p
3 dM + = 0. (A3.7)
0 ρ
This is the virial theorem for stars. Many important general results can be derived from the
virial theorem.
For our present purposes, let us relate the integral on the left-hand side of equation (A3.7)
to the internal thermal energy of the star. The energy density per unit mass is given by the
general relation,
p
u= , (A3.8)
(γ − 1)ρ
where γ is the ratio of specific heats of the material of the star. Therefore, the integral on
the left-hand side of equation (A3.7) becomes
Ms Ms
p
3 dM = 3 (γ − 1)u dM = 3(γ − 1)U, (A3.9)
0 ρ 0

where U is the total internal thermal energy of the star. For a monatomic gas, which is an
excellent approximation for a fully ionised gas, γ = 5/3, and so
2U + = 0. (A3.10)
 Explanatory supplement to Chapter 3 53

Thus, the magnitude of the gravitational potential energy is twice the internal thermal energy
of the star. This explains why the Kelvin–Helmholtz timescale is often referred to as the
thermal timescale of the star. The thermal energy is half the gravitational energy, and so the
time it takes the star to radiate away its internal thermal energy is the Kelvin–Helmholtz
timescale.
We can now return to the apparent paradox that, as stars radiate away their energy, they
heat up. The total energy of the star is the sum of its thermal and gravitational potential
energies, E = U + . But the virial theorem tells us that U = − /2, and so the total
energy is given by

E= = −U ; (A3.11)
2
in other words, a negative quantity. Thus, as the star loses energy, the total energy must
become even more negative and so U must increase; in other words, the star becomes
hotter. This was Homer Lane’s great insight, and it is entirely associated with the fact that
the gravitational potential energy is a negative quantity. This insight also influenced early
theories of the evolution of the stars.

A3.2 The third and fourth equations of stellar structure

A3.2.1 Energy generation
The third equation of stellar structure describes the energy generation rate within the star.
The energy generated within the star diffuses outwards and so the contribution to the outflow
of energy from the shell of radius r and thickness dr is given by

dL = 4πr 2 ρε dr, (A3.12)

where ε is the energy generation rate per unit mass and is a function of the local temperature
and density conditions. Note that L is the rate of flow of energy, or the power, passing through
the spherical surface at radius r . Hence, the differential equation for L is given by
dL
= 4πr 2 ρε. (A3.13)
dr

A3.2.2 Radiative transport of energy through a star

The fourth equation describes how radiation diffuses through the star. There are two prin-
cipal mechanisms for the transport of energy through stars: by radiation and convection.
Schwarzschild, in his important paper of 1906 (Schwarzschild, 1906), derived the condi-
tions under which energy would be transported by convection. If the temperature gradient
in the star exceeds the adiabatic gradient, that is, it is superadiabatic, convective motions
stabilise the energy transport so that the variation of temperature with pressure, or density,
is limited to the adiabatic gradient. Specifically, the condition is given by
d ln T γ −1
≥ ,
d ln p γ
54 3 Stellar structure and evolution

where γ is the ratio of specific heats of the material of the star. In practice, what is done is
to work out the structure of the star and then test whether or not there are regions which are
superadiabatic and in which convective transport of energy should be adopted.
Radiative transport of energy is much more important than thermal conduction because
the mean free path for photons, although small, is still very much greater than the mean
free path for electrons. The standard form of the heat diffusion equation is given by
dT
F = −λ , (A3.14)
dr
where F is the power per unit area parallel to the direction of the temperature gradient and
λ is the heat diffusion coefficient. Therefore, the total rate of flow of energy through the
spherical surface at radius r is L = 4πr 2 F.
In the radiative transport of energy within stars, the radiation is scattered many times,
because of the very high density of the material and the large cross-section for scattering.
Because of the very large numbers of scatterings, it is safe to assume that the radiation at
any point inside the star is almost precisely isotropic and has a black-body spectrum at the
local temperature of the material of the star. The diffusion of energy takes place through
the very gradual decrease in temperature with increasing radius.
Rather than work with a heat diffusion coefficient, astrophysicists work with the quantity
κ, which is known as the opacity of the stellar material. It is defined in terms of the fraction
of the flux density of radiation which is absorbed or scattered per unit mass per unit path-
length. Thus, if the increment of flux density, dF, is scattered by the material of the star on
traversing a distance dr , κ is defined by

dF = −κρ F dr. (A3.15)

Astrophysicists make use of the fact that the spectrum of the radiation inside the stars is
very close indeed to a black-body spectrum at the local temperature to rewrite the equation
of radiative transfer in an alternative form that is more directly related to local physical
conditions in the star. The flux density decrease corresponds to a decrease in radiation
pressure with radius through the star. The energy loss per second from the increment of
path length dr , is −κρ F dr , and hence the corresponding change in momentum per unit
area per unit time, that is, the change of radiation pressure, is given by
κρ F
dp = − dr. (A3.16)
c
The radiation is locally black-body radiation at temperature T , however, and so, according
to the Stefan–Boltzmann law, p = 13 aT 4 . Therefore,
dp 4 d p dr
= aT 3 = . (A3.17)
dT 3 dr dT
We have derived an expression for d p/dr from equation (A3.16), however, which involves
the flux density of radiation, F. Therefore, finally we obtain
4 acT 3 dT
F =− , (A3.18)
3 κρ dr
 Explanatory supplement to Chapter 3 55

or, in terms of the luminosity passing through the sphere at radius r ,

16πacr 2 T 3 dT
L=− . (A3.19)
3κρ dr
This is the standard form of the fourth equation of stellar structure. The opacity, κ, is
determined by the most important processes which impede the escape of radiation from the
star and involves a great deal of detailed atomic physics.

A3.3 The origin of the main sequence

At first sight, finding solutions of the equations of stellar structure appears to be a formidable
task, but Hoyle and Lyttleton realised that tractable power-law solutions could be found
which provide important insights into the relevant physics which determine the observed
properties of stars (Hoyle and Lyttleton, 1942). Let us recall the four equations of stellar
structure:
dp G Mρ
=− , hydrostatic equilibrium, (A3.20)
dr r2
dM
= 4πr 2 ρ, conservation of mass, (A3.21)
dr
dL
= 4πr 2 ρε, energy generation, (A3.22)
dr
dT 3κρ
=− L, energy transport. (A3.23)
dr 16πacr 2 T 3
These equations and the notation were discussed in Sections A3.1 and A3.2. It is possible to
make estimates of the central pressure of the star and a lower limit to its central temperature
from these equations, without knowledge of the detailed physics.

A3.3.1 The central pressure and temperature of the Sun

Let us write d p/dr ∼ p/R. Then, equation (A3.20) becomes

G M2
p∼ , (A3.24)
4π R 4
where we have used the approximation M ∼ ρ R 3 . Inserting the values for the Sun, M =
2 × 1030 kg and r = 7 × 108 m, we find p ∼ 2 × 1013 N m−2 ≈ 2 × 108 atmospheres. This
is the typical pressure inside the Sun. Integrating equation (A3.3) from the centre to the
surface of the star shows that the central pressure is twice this value. Thus, the central
pressure in the Sun is enormous.
A lower limit to the central temperature of the Sun is found from the virial theorem,
equation (A3.7), derived in Section A3.1:
Ms
p
3 dM + = 0. (A3.25)
0 ρ
56 3 Stellar structure and evolution

The gravitational potential energy is given by

Ms
G M dM
− = . (A3.26)
0 r
We really need to know the variation of mass with radius, but we can obtain a firm lower
limit to if we replace r by rs , the radius of the star. Hence,
Ms
G M dM G Ms2
− > = . (A3.27)
0 rs 2rs
For a perfect gas, p = nkT , and so the internal thermal energy of the star can be written as
Ms Ms
p 3k 3k
U =3 dM = T dM = T Ms , (A3.28)
0 ρ m 0 m
where m is the average mass of the particles which contribute to the pressure and T is the
mass-weighted average of the temperature inside the Sun. Thus, since − = 2U ,
G Ms m
T > . (A3.29)
6krs
Anticipating the answer we are about to obtain, the temperature is greater than the ionisation
potential of hydrogen, and so it is a good approximation to assume that the gas is fully ionised.
Since hydrogen is the most abundant element, the mean mass of the particles contributing
to the pressure is the average of the proton and electron mass, m = m p /2. Hence,
G Ms m p
T > = 2 × 106 K, (A3.30)
12krs
for the Sun, confirming that it is a good approximation to assume that the material of the
star is fully ionised. This argument also tells us that we need to understand radiative transfer
processes for radiation in the far-ultraviolet and X-ray wavebands since kT > 200 eV. In
fact, we now know that the central temperature of the Sun is about 1.5 × 107 K, and so we
need to know the opacity at energies kT ∼ 2 keV. Eddington had the important insight that
the opacity of stellar material at X-ray wavelengths was the key to understanding stellar
structure.

A3.3.2 Homologous stars

One of the most useful approximations for understanding the physics of the stars is to
assume that the material of the star has the same composition at all radii and that the
same properties of energy generation and transport apply throughout the star. These stellar
models are known as homologous stars. Then, the equations of stellar structure can be
rewritten in such a way that they depend only upon the mass of the star. The variation of
quantities such as the pressure, temperature and luminosity with radius within the star all
follow the same relations which scale as different powers of the total mass of the star.1
If, in addition, we assume that we can represent the opacity, κ, and the energy generation
rate, ε, by power laws over appropriate ranges of temperature and density, we can find
analytic expressions for relations between the mass of the star and its luminosity, radius and
 Explanatory supplement to Chapter 3 57

effective temperature. The essence of these arguments can be appreciated from dimensional
and order-of-magnitude calculations.
Let us replace all the derivatives in expressions (A3.20) to (A3.23) by their order-of-
magnitude values, that is, d p/dr ∼ p/R, dM/dr ∼ M/R and so on. Then, we obtain the
following four relations:
M2
p∝ , hydrostatic equilibrium, (A3.31)
R4
M ∝ R 3 ρ, conservation of mass, (A3.32)

L ∝ ρε R 3 , energy generation, (A3.33)

RT 4
L∝ , energy transport, (A3.34)
κρ
where we have used the approximation M ∼ ρ R 3 . These equations need to be supplemented
by the equation of state of the material of the star, which we take to be that of a perfect gas:
MT
p = nkT, that is p∝ . (A3.35)
R3
Combining equations (A3.31) and (A3.35),
M
T ∝ . (A3.36)
R
We can see immediately that, if the radius of the star is only a weak function of its mass,
T ∝ M and so L ∝ T 4 ∝ M 4 . It will turn out that this is a surprisingly good approximation
for stars of mass M ∼ M .

A3.3.3 Energy generation rates and stellar opacities

Finally, we need to know the density and temperature dependence of the opacity, κ, and the
energy generation rate, ε, upon temperature and density. In the case of the energy generation
rate, the p–p chain and the CNO cycle for the conversion of hydrogen into helium can be
described by ε ∝ ρT α , where α takes the values 4 and 17, respectively.
The opacity, κ, is a more complex function of temperature. It is most convenient to
describe κ by power-law relations of the form κ ∝ ρ β T γ , where the values of β and γ take
different values in different temperature ranges. The values quoted by Tayler are shown in
Table A3.1.
It is interesting that the most important of these processes can be understood in terms
of the classical processes of the emission and absorption of radiation. At the very highest
temperatures, the plasma is fully ionised and the dominant scattering process is Thomson
scattering for which the Thomson cross-section, σT = e4 /6π 02 m 2e c4 = 6.653 × 10−29 m−2 ,
is independent of frequency.
In the intermediate temperature range, one of the dominant processes is free–free or
bremsstrahlung absorption. The appropriate values of β and γ can be found from semi-
classical arguments (Kramers, 1923) and results in the formula known as the Kramers
opacity. In outline, the calculation proceeds as follows.
58 3 Stellar structure and evolution

Table A3.1. The approximate variations of the opacity, κ, with temperature

and density in different temperature ranges for typical densities and
temperatures found in main-sequence stars

Temperature
Temperature range (K) Physical processes β γ

Low 104 –104.5 atomic and molecular absorption 0.5 4

Medium 104.5 –107 bound–free and free–free absorption 1 −3.5

High >107 electron scattering 0 0

The radiation spectrum of a free electron moving through a fully ionised plasma of ions
with charge Z and number density Ni can be worked out by semi-classical methods and is
found to have the following form:
jν ∝ Z 2 Ni T −1/2 g(ν, T ) e−hν/kT ,
where g(ν, T ) is a slowly varying function of ν and T known as the Gaunt factor.2 The
corresponding absorption coefficient can be found from Kirchhoff ’s law, κν = jν /B(ν),
where B(ν) is the black-body spectrum. Finally, we need to sum over all the contributions
of the different frequencies to the average opacity κ. As shown by Kippenhahn and Weigert,3
the correct weighting is given by
∞
1 π 1 ∂B
= dν.
κ acT 3 0 κν ∂ T
This expression is known as the Rosseland mean opacity, first derived by Svein Rosseland in
1924 (Rosseland, 1924). It is then straightforward to work out the dependence of κ upon the
temperature and density of the plasma, κ ∝ ρT −7/2 , where ρ is the density of the plasma.

A3.3.4 The solutions

Given the power-law dependences discussed in Section 3.3.3, equations (A3.33) and (A3.34)
become
M3 α
L∝ T , energy generation, (A3.37)
R3
T 4−β R 4+3γ
L∝ , energy transport. (A3.38)
M 1+γ
We now have sufficient relations to determine the dependence of L, T and R upon the mass
of the star, M. Firstly, we can use equations (A3.37) and (A3.38) to eliminate L:
T 4−β−α R 7+3γ ∝ M 3+γ . (A3.39)
Finally, combining equations (A3.36) and (A3.39), we find
1−α−β −γ
R ∝ M −(∗) where ∗= . (A3.40)
3 + α + β + 3γ
Explanatory supplement to Chapter 3 59

Table A3.2. The approximate variations of the opacity, κ, with temperature

and density in different temperature ranges for typical densities and
temperatures found in main-sequence stars

R ∝ Ma L ∝ Mb L ∝ Teff
c

α β γ a b c

4 0 0 3/7 3 28/5 = 5.6

17 0 0 4/5 3 60/7 ≈ 8.6
4 1 3.5 1/13 71/13 ≈ 5.5 284/69 ≈ 4.1
17 1 3.5 9/13 67/13 ≈ 5.2 268/49 ≈ 4.5

All the other relations between observables and the mass of the star follow immediately.
Let us first work out the relation between R and M for stars like the Sun, a medium-
temperature star, from the above discussion. Inserting α = 4 for the p–p chain, β = −3.5
and γ = 1, we find

R ∝ M 1/13 . (A3.41)

This is a very weak dependence upon the mass of the star. We can immediately use equations
(A3.36) and (A3.39) to find the temperature– and luminosity–mass relations:
M M3 α
T ∝ ∝ M 12/13 , L∝ T ∝ M 71/13 . (A3.42)
R R3
Finally, we need to determine the dependence of the effective temperature, Teff , of the star
upon mass. It might seem as though equations (A3.42) provide the answer, but that is not
correct since the above temperature relation refers to the central temperature of the star, not
to the emission through its surface. The complication is that, for the homologous stars, the
temperature changes from Tc to T = 0 at the surface and so the homology relations cannot
be used. Instead, we know that the luminosity and radius attain finite values at the surface
of the star and so we can use the Stefan–Boltzmann law to define an effective temperature,
Teff , through the relation

L = 4π R 2 aTeff
4
. (A3.43)

Hence,
 1/4
L 1/4
Teff ∝ ∝ M 71/13 M −2/13 , (A3.44)
R2

Teff ∝ M 69/52 . (A3.45)

Hence, from equations (A3.42),

284/69
L ∝ M 71/13 ∝ Teff . (A3.46)

This is the main sequence for stars of mass roughly that of the Sun in terms of their
luminosities and effective temperatures and is roughly L ∝ Teff
4.1
.
60 3 Stellar structure and evolution

Similar calculations can be carried out for the different expressions for the opacity and
energy generation rates; some of those tabulated by Tayler are given in Table A3.2. Thus,
for a wide range of different assumptions about the opacity of the stellar material and the
energy generation rate, there is a power-law relation of the form L ∝ M b , where b lies in
the range 3 to 5.5. For the physical conditions appropriate for main-sequence stars with
mass roughly that of the Sun, the model in the third row of the table with b ≈ 5.5 is a good
approximation to what is observed among the stars. For high-mass main-sequence stars, the
second row is a good approximation with b ≈ 3.
These models are illustrative, but demonstrate how many features of the properties of
stars can be explained in physical terms. In particular, they show that the properties of the
stars are relatively insensitive to the details of the energy generation mechanism, a point fully
appreciated by Eddington. It can be seen from Table A3.2 that changing the dependence
of the energy-generation rate upon temperature by a huge amount makes remarkably little
difference to the mass–luminosity relation.

Notes to Section A3

1 This approach is beautifully explained by Roger J. Tayler in his book The Stars: Their Structure
and Evolution (Cambridge: Cambridge University Press, 1994).
2 See, for example, my version of the calculation in Malcolm Longair, High Energy Astrophysics,
vol. 1 (Cambridge: Cambridge University Press, 1997), Chapter 3.
3 See R. Kippenhahn and A. Weigert Stellar Structure and Evolution (Berlin: Springer-Verlag, 1990),
Section 5.1.
4 The end points of stellar
evolution

4.1 The red giant problem

While the understanding of main-sequence stars proceeded apace through the 1920s and
1930s, there remained the problem of accounting for the red giant stars, which are very
much more luminous than main-sequence stars at the same effective temperatures. Russell
adopted the position that matter existed in different states in the dwarf and giant stars, what
he termed ‘giant stuff’ and ‘dwarf stuff’. Atkinson assumed that different nuclear processes
were responsible for the luminosities of the giant stars.
The stellar models of Eddington are homogeneous, and it was assumed that homogene-
ity was maintained, probably by large-scale meridional circulation driven by the internal
rotation of the star. It was only in the early 1950s, that a number of astrophysicists, Peter
Sweet (1921–2005), Martin Schwarzschild, Ernst Öpik and Leon Mestel, showed that the
mixing assumption was highly implausible.1
The solution to the red giant problem was discovered in 1938 by the Estonian astrophysi-
cist Ernst Öpik (1893–1985), then working at the University of Tartu (Öpik, 1938). Öpik
realised that if the stars are not well mixed, it is inevitable that they become inhomogeneous.
Within the central core of the star, nuclear burning of hydrogen into helium leads to the
depletion of the nuclear fuel in the core. In Öpik’s model it was assumed that the central
core of the star was maintained in convective equilibrium, resulting in a uniform depletion
of hydrogen in this region. Once core burning had exhausted all the available hydrogen,
nuclear burning would continue in a shell about an inert, isothermal core, while the core
itself would begin to collapse, releasing gravitational potential energy. Öpik argued that the
rapid release of energy during these phases would lead to the expansion and cooling of the
envelope, resulting in the red giant phase of a star’s evolution. His picture of the structure
of a red giant is shown in Figure 4.1.
As a result of Öpik’s work, there was no need to seek separate physical processes to
account for the giant stars – they form naturally at the end of the phase of hydrogen burning
on the main sequence. The giant phase could only last a short time compared with the
age of stars on the main sequence because they must burn the available nuclear fuel at
thousands of times the rate at which it is consumed on the main sequence to account
for the huge luminosities of the red giants. Thus, the giant phase is a brief final fling before
the star settles down to some form of dead star. As Öpik pointed out, this picture is entirely

61
62 4 The end points of stellar evolution

Figure 4.1: Öpik’s model for the structure of a giant star. Hydrogen has been exhausted in the contract-
ing convective core, C, which has radius r . Hydrogen is converted into helium in the hydrogen-burning
shell, A, which has outer radius R1 . The extensive envelope of the giant star, B, was assumed to be in
radiative equilibrium (Öpik, 1938).

consistent with the observation that the red giant stars are very much rarer per unit volume of
space than the dwarf stars. The quantitative theory of red giant stars needed a more complete
theory of the nuclear reactions involved in post-main-sequence evolution. Specifically, as
the core continued to collapse and heat up, helium burning in the core would result in the
formation of carbon, and this was first discussed by Öpik and Edwin E. Salpeter (b. 1924)
after the Second World War (Öpik, 1951; Salpeter, 1952) (see Section 8.2).
An important link in the chain was provided in 1942 by the Brazilian astrophysicist Mario
Schönberg (1914–1990) and Subrahmanyan Chandrasekhar (1910–1995). Their studies
concerned the stability of stellar models with inert isothermal cores, which are expected
even if the central regions are in radiative, rather than convective, equilibrium. In the radiative
case, the hydrogen fuel is depleted first in the hottest central regions and the size of the
inert core grows with time. They found the important result that there do not exist stable
stellar models in which the inert stellar core contains more than about 10% of the mass
of the star (Schönberg and Chandrasekhar, 1942). Physically, the pressure at the base of
the hydrogen-burning shell becomes too great and causes the inner regions to collapse. As
shown by Kippenhahn and Weigert, the key quantity is the ratio of the mean molecular
weights in the core and the envelope.2
This result, known as the Schönberg–Chandrasekhar limit, explained the formation of
red giant stars during the course of stellar evolution. Stars spend most of their lifetimes on
the main sequence, the energy source being the conversion of hydrogen into helium, by the
p–p chain for stars with masses less than about 1.5M and by the CNO cycle for stars with
greater masses. Hydrogen is depleted in the central regions, resulting in the formation of
an inert core. When the core grows to about 10% of the mass of the star, core collapse and
 4.2 White dwarfs 63

the formation of the red giant envelope ensues.3 This result also enables good estimates for
the lifetimes of the Sun and main-sequence stars to be made (see Section A4.1).

4.2 White dwarfs

The discovery of white dwarfs is charmingly told in a reminiscence of Henry Norris Russell
delivered at a Princeton colloquium in 1954. In 1910, Russell suggested to Pickering that
it would be useful to obtain the spectra of stars for which parallaxes had been measured.
Russell’s reminiscence continues:4
Pickering said ‘Well, name one of these stars.’ Well, said I, for example, the faint component of
Omicron Eridani. So Pickering said, ‘Well, we make rather a specialty of being able to answer
questions like that’. And so we telephoned down to the office of Mrs. Fleming and Mrs. Fleming said,
yes, she’d look it up. In half an hour she came up and said, I’ve got it here, unquestionably spectral
type A. I knew enough, even then, to know what that meant. I was flabbergasted. I was really baffled
trying to make out what it meant. Then Pickering thought for a moment and then said with a kindly
smile, ‘I wouldn’t worry. It’s just these things which we can’t explain that lead to advances in our
knowledge.’ Well, at that moment, Pickering, Mrs. Fleming and I were the only people in the world
who knew of the existence of white dwarfs.

The remarkable feature of the faint companion of o-Eridani was that it was a very-
low-luminosity star and yet it had the type of spectrum associated with hot stars on the
upper part of the main sequence. Russell included it without comment in his first ‘Russell’
diagram (Figure 3.3(a)), the single A star lying roughly 10 magnitudes (a factor of 10 000 in
luminosity) below typical main-sequence A stars. Adams drew attention to its remarkable
properties in 1914 (Adams, 1914) and discovered another example in the following year,
Sirius B, the faint companion of Sirius A (Adams, 1915).
Eddington realised that these observations immediately implied that white dwarf stars
had to be very dense indeed. Their masses could be determined from the fact that they
were members of binary star systems, and their radii could be estimated using Planck’s
radiation formula and their observed luminosities. Their mean densities had to be about
108 kg m−3 , but Eddington argued that there was nothing inherently implausible about such
large densities (Eddington, 1924). Matter at such high temperatures would be completely
ionised and so there was no reason at that time why matter could not be compressed
to much higher densities than typical terrestrial densities. In fact, he argued that even
nuclear densities were quite conceivable. In his paper of 1924, he also worked out the
magnitude of the gravitational redshift which would be expected from such a compact star
according to general relativity and found that it corresponded to a Doppler shift of the
spectral lines to longer wavelengths of about 20 km s−1 . Adams made careful spectroscopic
observations of Sirius B with the 100-inch telescope in 1925 and, once account was taken of
the orbital motions of the binary stars, a shift of 19 km s−1 was measured (Adams, 1925a,b).5
Eddington was jubilant (Eddington, 1926a):
Prof. Adams has killed two birds with one stone; he has carried out a new test of Einstein’s theory of
general relativity and he has confirmed our suspicion that matter 2000 times denser than platinum is
not only possible, but is actually present in the universe.
64 4 The end points of stellar evolution

The theory of white dwarfs was one of the first triumphs of the new quantum theory of
statistical mechanics as applied to astrophysics. Wolfgang Pauli (1900–1958) enunciated
the exclusion principle in 1922 (Pauli, 1925), and this led to Fermi–Dirac statistics and the
concept of degeneracy pressure. In 1926, Fowler used these concepts to derive the equation
of state of a cold degenerate electron gas (Fowler, 1926) and found the important result
2  
(3π 2 )2/3 −
h ρ 5/3
p= , (4.1)
5 m e μe m u
where μe is the mean molecular weight of the material of the star per electron and m u is
the unified atomic mass constant.† The important aspect of this equation of state is that it is
independent of temperature and so the structure of white dwarfs can be derived directly from
the Lane–Emden equation (3.3).6 Unlike main-sequence stars, in which pressure support
is provided by the thermal pressure of hot gas, the white dwarfs are supported by electron
degeneracy pressure. The source of their luminosity is the internal thermal energy with
which they were endowed on formation. According to Fowler’s picture, the white dwarfs
simply radiate away their internal thermal energies and end up as inert cold stars with all
the nuclei and electrons in their lowest ground states.
In 1929, Wilhelm Anderson (1880–1940) showed that the degenerate electrons in the
centres of white dwarfs with mass roughly that of the Sun become relativistic (Anderson,
1929). In the extreme relativistic limit, the equation of state of the degenerate electron gas
becomes
 
(3π 2 )1/3−
hc ρ 4/3
p= . (4.2)
4 μe m u
Once again, the result is independent of temperature, but the change in the dependence
of pressure upon density from p ∝ ρ 5/3 to p ∝ ρ 4/3 has profound implications. Anderson
and Edmund Stoner (1899–1968) realised that the consequence was that there do not exist
equilibrium configurations for degenerate stars with mass greater than about the mass of the
Sun (Anderson, 1929; Stoner, 1929). The most famous analysis of this result was carried
out by Chandrasekhar, who had begun working on this problem before he arrived to take
up a fellowship at Trinity College, Cambridge, in 1930. According to Wali’s biography,7
he derived the key result while on board the ship Lloyd Triestino which was taking him
as a 19-year-old from Bombay to London. He found the crucial result that, in the extreme
relativistic limit, there is an upper limit to the mass of stable white dwarfs,
 
(3π)3/2 −hc 3/2 2.01824 5.836
MCh = × = M . (4.3)
2 G (μe m u )2 μ2e
This mass is known as the Chandrasekhar mass‡ (Chandrasekhar, 1931). The critical mass
depends upon the chemical composition of the material of the star through the value of
μe , the mean molecular weight of the stellar material per electron. Other than that, the

† Order-of-magnitude calculations demonstrating the origins of the forms of the non-relativistic and relativistic
equations of state for degenerate matter are given in Section A4.2.
‡ Order-of-magnitude calculations illustrating the origins of the Chandrasekhar mass are given in Section A4.3.
4.2 White dwarfs 65

Chandrasekhar mass only depends upon fundamental constants. Since μe ≈ 2 for the mate-
rial of compact stars, the Chandrasekhar mass is usually quoted as MCh = 1.46M .
The cause of the instability is that, in the extreme relativistic limit, both the internal
thermal energy, Uth , and the gravitational potential energy, Ugrav , of the star depend upon
radius in the same way, Uth = (1/2)Ugrav ∝ R −1 . Now, the gravitational potential energy is
proportional to M 2 , whereas the thermal energy is proportional to the mass of the star and
so, for massive enough stars, the gravitational energy term dominates, causing collapse,
which cannot be stabilised by the pressure of the degenerate gas since the two energies
always depend upon radius in the same way. The inference is that there is nothing to prevent
degenerate stars more massive than MCh from collapsing to very high densities indeed and
possibly to a state of complete gravitational collapse.
This conclusion was vigorously challenged by Eddington and led to the famous dispute
with Chandrasekhar. Eddington found the idea of complete gravitational collapse unaccept-
able, believing that there must be some new unspecified physical process which prevented its
occurrence. Chandrasekhar’s work was publicly repudiated by Eddington, employing argu-
ments which were more polemical than physical, at the meeting of the Royal Astronomical
Society on January 11 1935 (Eddington, 1935):
Dr. Chandrasekhar has got this result before, but he has rubbed it in in his last paper; and, when
discussing it with him, I felt driven to the conclusion that this was almost a reductio ad absurdum
of the relativistic degeneracy formula. Various accidents may intervene to save the star, but I want
more protection than that. I think there should be a law of Nature to prevent a star behaving in this
absurd way! . . . The formula is based upon a combination of relativity mechanics and non-relativistic
quantum theory, and I do not regard the offspring of such a union as born in lawful wedlock.

Eddington fully realised that what we would now call a black hole was the natural outcome
of gravitational collapse. In his words (Eddington, 1941)8
If the star is symmetrical and not in rotation, it would contract to a diameter of a few kilometres, until
according to the theory of relativity, gravitation becomes too great for the radiation to escape.

Eddington objected instinctively, however, to what he called elsewhere ‘this stellar buf-
foonery’. His physical concerns centred upon the use of the Anderson–Stoner relativistic
equation of state for a degenerate gas, which is the ultimate cause of the gravitational
collapse of the star. While the vast majority of physicists and astrophysicists agreed with
Chandrasekhar’s analysis, what Leon Mestel calls a ‘ding-dong dispute’ continued in the
literature for a number of years.9
Chandrasekhar, then a brilliant young mathematical physicist recently arrived from India
to an alien environment, took the rebuff badly and it rankled for many years, despite his
lasting respect for, and friendship with, the older man. In his address to the General Assembly
of the International Astronomical Union in Montreal in 1979, he stated (Chandrasekhar,
1980):
It is difficult to understand why Eddington, who was one of the early enthusiasts and staunchest
advocates of general relativity, should have found the conclusion that black holes may be formed
during the course of the evolution of stars so unacceptable. But the fact is that Eddington’s supreme
authority in those years effectively delayed the development of fruitful ideas along these lines for
some thirty years.
66 4 The end points of stellar evolution

Quite independently, Lev Landau (1908–1968) had come to the conclusion in 1932 that
gravitational collapse to a singularity should be taken seriously (Landau, 1932), and in 1938
Robert Oppenheimer (1904–1967) and Hartland Snyder (1913–1962) gave the first general
relativistic analysis of what would be observed in the final stages of gravitational collapse
of a pressureless sphere (Oppenheimer and Snyder, 1939). In their paper they described the
key observed features of what are now termed black holes.

4.3 Supernovae and neutron stars

The neutron was discovered in 1932 by Chadwick (Chadwick, 1932), and the model of
the nucleus consisting of neutrons and protons was quickly adopted, although the problem
of how the nucleus could be held together remained to be resolved. The first mention of
the possibility of neutron stars appears as the famous ‘Additional remark’ to a paper by
Walter Baade (1893–1960) and Fritz Zwicky (1898–1974) of 1934 (Baade and Zwicky,
1934b). In that year, they published two papers on the energetics of what they termed
‘super-novae’.
The extragalactic nature of the spiral nebulae had been established beyond doubt by
Hubble in 1926. Among the objects which played a part in that debate were the novae, or
‘new stars’, which increase rapidly in brightness and fade away again. As will be discussed
in Section 5.3, the nova of 1885 which exploded in the Andromeda Nebula appeared to be
exceptional, being about 100 times more luminous than the more common nova. Lundmark
had suggested that there were two classes of novae, that of 1885 belonging to the upper
class. In their first paper, Baade and Zwicky proposed that the population of novae consists
of two types: the ordinary novae, which are relatively common phenomena and which had
been used by Lundmark as distance indicators for spiral nebulae (see Section 5.3), and the
supernovae, which are very rare but very energetic indeed (Baade and Zwicky, 1934a).
They identified the bright nova observed in the Andromeda Nebula in 1885 as the archetype
of this class of extremely violent explosion – they suggested that Tycho Brahe’s nova of
1572 was another example of this class. The frequency of occurrence of these events was
estimated to be only about once per 1000 years per galaxy but, when they occurred, an
enormous amount of energy was released, corresponding to a significant fraction of the rest
mass energy of the precursor star. In their second paper, they suggested that such events
might be the sources of the cosmic rays which had been discovered by Victor Hess in 1912
(Hess, 1912). Both proposals are remarkably close to the truth. As an addendum to the
second paper (Baade and Zwicky, 1934b), they wrote

With all reserve we advance the view that a super-nova represents the transition of an ordinary star
into a neutron star, consisting mainly of neutrons. Such a star may possess a very small radius
and an extremely high density. As neutrons can be packed much more closely than ordinary nuclei
and electrons, the ‘gravitational packing’ energy in a cold neutron star may become very large, and
under certain circumstances, may far exceed the ordinary nuclear packing fractions. A neutron star
would therefore represent the most stable configuration of matter as such. The consequences of this
hypothesis will be developed in another place, where also will be mentioned some observations that
tend to support the idea of stellar bodies made up mainly of neutrons.
4.3 Supernovae and neutron stars 67

Figure 4.2: The cartoon which appeared in the Los Angeles Times of 19 January 1934 in the comic
strip entitled ‘Be Scientific with Ol’ Doc Dabble’.

It is best to allow Zwicky to describe how these ideas were received in a quotation from the
extraordinary preface to his Catalogue of Selected Compact Galaxies and of Post-eruptive
Galaxies of 1968 (Zwicky, 1968),
In the Los Angeles Times of January 19, 1934, there appeared an insert in one of the comic strips,
entitled ‘Be Scientific with Ol’ Doc Dabble’ quoting me as having stated ‘Cosmic rays are caused
by exploding stars which burn with a fire equal to 100 million suns and then shrivel from 12 million
miles diameter to little spheres 14 miles thick’, says Prof. Fritz Zwicky, Swiss Physicist.’ This, in all
modesty, I claim to be one of the most concise triple predictions ever made in science. More than 30
years were to pass before this statement was proved to be true in every respect.

(See Figure 4.2.) Baade and Zwicky’s idea that a neutron star might be the remnant left
behind after a supernova explosion was proved correct 33 years later with the discovery of
pulsars by Antony Hewish, Jocelyn Bell and their colleagues in 1987 (Hewish et al., 1968).
68 4 The end points of stellar evolution

In the meantime, Gamow showed in 1937 that a gas of neutrons could be compressed to a
much higher density than a gas of nuclei and electrons and estimated the probable densities
of such stars to be about 1017 kg m−3 (Gamow, 1937, 1939). The issue of the maximum mass
of neutron stars was discussed by Landau in 1938 (Landau, 1938) and in much greater detail
by Oppenheimer and George Volkoff (1914–2000) in the following year (Oppenheimer and
Volkoff, 1939). The result they found is not so different from expression (4.3) if we set
μe = 1 and m u = m n . The physics is the same as in the case of the white dwarfs, but now
neutron degeneracy pressure holds up the star. Complications arise because it is necessary
to take into account the details of the equation of state of neutron matter at nuclear densities,
and the effects of general relativity can no longer be neglected. They found an upper mass
limit of about 0.7M . This result is not so different from the best modern estimates, which
correspond to about 2–3M .
This is a very much more serious situation than the case of the white dwarfs. The
neutron stars are so compact that general relativity is no longer a small correction but
is central to the stability of the star. Typically, for neutron stars, the general relativistic
parameter 2G M/Rc2 = Rg ∼ 0.3, and so they have radii which are only about three times
the Schwarzschild radius, Rg , of a spherically symmetric black hole of the same mass.
This work created some theoretical interest but little enthusiasm from the observers. The
radii of typical neutron stars were expected to be about 10 km and so there was no prospect of
detecting significant fluxes of thermal radiation from such tiny stars. Nonetheless, many of
the objects which were to play a leading role in the development of high-energy astrophysics
in the years following the Second World War were already in place in the literature, even if
there was not a great deal that the astronomers could do about them at that time.

Notes to Chapter 4

1 I am most grateful to Leon Mestel for his deep insights into the history of the astrophysics of stellar
structure and evolution.
2 The result quoted by Kippenhahn and Weigert is that the fraction of the mass of the star in the core
should not exceed (μcore /μenv )2 , where the μs are the mean molecular weight of the material per
electron. For the case of a helium core surrounded by an envelope with normal cosmic abundance,
the limit corresponds to about 10% of the mass of the star being in the form of an inert core.
3 One of the more controversial issues concerned the precise cause of the initiation of the red giant
phase. Every numerical calculation of the evolution of stars shows that, as hydrogen burning in
the core is exhausted, the core collapses and the envelope expands, but the dramatic changes
in the star’s structure are associated with a number of different processes taking place almost
simultaneously – the collapse of the central regions, changes in chemical composition of the stellar
material with radius, in particular the discontinuity in atomic weight at the core–envelope boundary,
and consequently changes in its opacity, the development of extensive convective zones in the stellar
envelope and so on. The most detailed discussion of the likely cause of the formation of red giants
has been presented by John Faulkner (see, for example, J. Faulkner, Fred Hoyle, Red giants and
beyond, Astrophysics and Space Science, 285 (2003), 339, and in the memorial volume to celebrate
the life and work of Fred Hoyle, Red giants, then and now, in D. O. Gough, ed., New Frontiers of
Astronomy (Cambridge: Cambridge University Press, 2005)).
4 See A. G. Davis Philip and D. H. DeVorkin, eds, In Memory of Henry Norris Russell (Dudley
Observatory Report no. 13, 1977), pp. 90–107.
 Explanatory supplement to Chapter 4 69

5 See also A. V. Douglas, The Life of Arthur Stanley Eddington (London: Thomas Nelson and Sons
Ltd, 1956), pp. 75–78.
6 I have demonstrated how this analysis can be carried out in Section 15.3 of Malcolm Longair,
High Energy Astrophysics, vol. 2 (Cambridge: Cambridge University Press, 1997). The section
also includes a derivation of the formula for the Chandrasekhar mass. An order of magnitude
analysis is given in Section A4.3.
7 See endnote 4, Chapter 3.
8 This quotation is taken from the discussion of Eddington’s paper, The theory of white dwarf stars
(See A. S. Eddington, The theory of white dwarf stars, in Novae and White Dwarf Stars, ed. A. J.
Shaler (Paris: Herrmann et Cie), pp. 249–262), which immediately preceded the discussion. A more
extended quote from the discussion recorded by the editors is of interest. ‘Sir Arthur Eddington
replies that in stars of mass greater than the critical masses mentioned by Dr. Chandrasekhar there
is no limit to the contraction, so that if the star is symmetrical and not in rotation, it would contract
to a diameter of a few kilometers, until, according to the theory of relativity, gravitation becomes
too great for the radiation to escape. This is not a fatal difficulty, but it is nevertheless surprising;
and, being somewhat shocked by this conclusion, Sir Arthur was led to reexamine the physical
theory and so finally to reject it.’
9 See, for example, S. Chandrasekhar, Eddington: The Most Distinguished Astrophysicist of his Time
(Cambridge: Cambridge University Press, 1983), p. 47 et seq. Leon Mestel has written: ‘Chandra
was anxious to get backing from leading physicists if only because astronomers were overawed
by Eddington’s reputation. Bohr, Rosenfeld, Dirac, Peierls, Pauli, Fowler all supported Chandra
against Eddington, at least in private, though there seems to have been some reluctance to stand up
and be counted.’ This account agrees with the views expressed to me by William McCrea and Lyman
Spitzer, who were present during many of these debates. Both agreed that everyone thought that
Chandrasekhar was right. McCrea wrote to me in 1993 as follows: ‘No one who knew Eddington . . .
would remotely imagine his being unpleasant. The last few years have seen most unfortunate,
completely misleading accounts of this matter. I was present at the core episode. Unfortunately
misunderstandings arose, mainly on Chandra’s part. Unhappily he has allowed them to prey upon
him, which I deeply regret. Eddington did use some sloppy arguments. But, if Chandra was so
certain that he was right, why did he not pursue the consequences at the time?’ Mestel provides
detailed references to the published articles involved in the dispute (see note 13 to Chapter 3).

A4 Explanatory supplement to Chapter 4

A4.1 The lifetimes of the Sun and the stars
The Schönberg–Chandrasekhar limit enables simple estimates of the lifetimes of main-
sequence stars to be made. Most of the lifetime of a main-sequence star is spent burning
hydrogen to helium in its core. This nuclear fusion reaction releases by far the largest fraction
of the nuclear energy available to the star, about 0.7% of the rest mass of the hydrogen nuclei
being liberated in forming helium. Stellar evolution models show that, once the star has
settled onto the main sequence, its luminosity changes very little until it begins to move off
the main sequence when the core contracts and the red-giant phase begins. The subsequent
phases of stellar evolution are all short compared with the main-sequence lifetime.
These considerations enable a simple estimate of the main-sequence lifetime of the star
to be made. The star moves off the main sequence when the central 10% of its mass has been
converted into helium. The energy released in this process is given by E = 0.007 (0.1 ×
M)c2 . Therefore, since the luminosity of the star is L, its main-sequence lifetime is given
70 4 The end points of stellar evolution

by
E 0.007(0.1 × M)c2
TMS = = .
L L
Inserting the values for the Sun, L = 3.9 × 1026 W and M = M = 2 × 1030 kg, we find
T = 1010 years.
We can use this result to find the main-sequence lifetimes of main-sequence stars of
different masses. If the mass–luminosity relation has the form L ∝ M x , where x ∼ 4 for
stars with M ∼ M , then, by exactly the same argument, the lifetime of the star is given by
 
M −(x−1)
T (M) = 10 10
years. (A4.1)
M

A4.2 The equation of state for degenerate matter

In the cases of both white dwarfs and neutron stars, there is no internal heat source – the
stars are held up by degeneracy pressure. The significance of degeneracy pressure comes
about naturally because, in the centres of stars at an advanced stage of evolution, the
central densities become high and the use of the pressure formulae for a classical gas is
inappropriate. The Heisenberg uncertainty principle ensures that, at very high densities,
when the inter-particle spacing becomes small, the particles of the gas must possess large
momenta according to the relation p x ≈ − h. These large quantum mechanical momenta
provide the pressure of the degenerate gas.
First of all, we work out the physical conditions under which degeneracy pressure is
important. If the electron–proton plasma is in thermal equilibrium at temperature T , the
root-mean-square velocity of the particles, v 2 , is given by 12 mv 2  = 32 kT and hence
the typical momenta of the particles are p = mv ≈ (3mkT )1/2 . Because the electrons are
much lighter than the protons and neutrons, they become degenerate at much larger inter-
particle spacings than the protons and neutrons. According to the Heisenberg uncertainty
principle, the inter-electron spacing at which quantum mechanical effects become important
is x ≈ − h/p and hence, setting p = p, the density of a hydrogen plasma, which is
mostly contributed by the protons, is given by
 
3m e kT 3/2
ρ ≈ m p /(x) ≈ m p
3
−
, (A4.2)
h
where m p is the mass of a proton. Thus, the density at which degeneracy sets in for electrons
in the non-relativistic limit is proportional to T 3/2 . A better estimate of the critical density
can be found by equating the degeneracy pressure of a non-relativistic gas, (equation (A4.9)
below) to the pressure of a classical gas. Performing this sum, we find that the critical density
is given by
ρ ≈ 3.3 × 10−4 T 3/2 kg m−3 , (A4.3)
where T is the temperature in kelvin. Hence, for stars like the Sun with central temperature
about 1.6 × 107 K and central density ρ ≈ 1.5 × 105 kg m−3 , the equation of state of the
gas can always be taken to be that of a classical gas. When the star moves off the main
Explanatory supplement to Chapter 4 71

sequence, however, the central regions contract and, although there is a modest increase
in temperature, the matter in the core becomes degenerate, and this plays a crucial role in
the evolution of stars on the giant branch. Ultimately, in the white dwarfs, the densities are
typically about 109 kg m−3 and so they are degenerate stars.
The next consideration is whether or not the electrons are relativistic. To order of magni-
tude, we can find the condition for the electrons to become relativistic by setting p ≈ m e c
in Heisenberg’s uncertainty principle; then, by the same arguments as above, we find that
the density is given by
 3
mp m ec
ρ∼ ∼ mp −
∼ 3 × 1010 kg m−3 . (A4.4)
(x)3 h

A better calculation, with exactly the same physics but expressed in a slightly different way,1
is to require the Fermi momentum of a degenerate Fermi gas in the zero-temperature limit
to be m e c. In this case, the density at which the electrons become relativistic is given by
 3
mu mec
ρ= − μe = 9.74 × 108 μe kg m−3 . (A4.5)
3π 2 h

In this expression, m u is the atomic mass unit and μe = m B /(m u Ye ) is the mean molecular
weight of the material per electron, m B being the mean baryon rest mass of the material,
and Ye is the mean number of electrons per baryon.2 In the centres of the most massive
white dwarfs, the densities attain these values, and so the equation of state for a relativistic
degenerate electron gas has to be used. It is this feature which determines the upper mass
limit for white dwarfs and neutron stars.
Next, we can work out by these rough methods the equations of state for degenerate matter
in the non-relativistic and relativistic regimes. In general, the relation between pressure and
energy density can be written as p = (γ − 1)u, where p is the pressure, u is the energy
density of the matter or radiation which provides the pressure and γ is the ratio of specific
heat capacities.
In the non-relativistic regime, the energy of an electron in the degenerate limit is given
by
−2
p2 h
E = 12 m e v 2 = = , (A4.6)
2m e 2m e a 2
where a = x is the inter-electron spacing. Therefore, to order of magnitude, the energy
density of the material is given by u ≈ E/a 3 = − 2
h /2m e a 5 . Since the density of matter is
ρ ∼ m p /a 3 , it follows that p ∝ ρ 5/3 , and hence the ratio of specific heat capacities is given
by γ = 5/3. The pressure of the gas is therefore roughly given by
−2 − 2  5/3
h h ρ
p≈ 5
≈ . (A4.7)
3m e a 3m e mp

We can repeat this calculation for a relativistic electron gas, in which case E ≈
pc ≈ −
hc/a and hence u ≈ E/a 3 ≈ −hc/a 4 . Since ρ ∼ m p /a 3 , p ∝ ρ 4/3 and γ = 4/3. The
72 4 The end points of stellar evolution

pressure of the gas is roughly given by

− −  4/3
hc hc ρ
p≈ 4
≈ . (A4.8)
3a 3 mp
The exact results found from the application of statistical mechanics to a Fermi–Dirac
distribution in its ground state are as follows:
2  
(3π 2 )2/3 −
h ρ 5/3
non-relativistic p= ; (A4.9)
5 m e m u μe
 
(3π 2 )2/3−
hc ρ 4/3
relativistic p= . (A4.10)
4 m u μe
We obtain the corresponding results for degenerate neutrons if we substitute neutrons for
electrons in the above expressions and set μe = 1. Then, the expressions for the pressure
of the neutron gas in the two limits are as follows:
 5/3
(3π 2 )2/3 −
h2 ρ
non-relativistic p= ; (A4.11)
5 mn mn
 
hc ρ 4/3
(3π 2 )1/3−
relativistic p= . (A4.12)
4 mn

A4.3 The Chandrasekhar mass

We can use these techniques to illustrate the physical origin of the Chandrasekhar mass.
The total internal energy of the star can be found from our order-of-magnitude derivation
of the equation of state of a relativistic degenerate gas, equation (A4.8). Since p = 13 u,

U = V u = 3V p ≈ V −
hc(ρ/m p )4/3 . (A4.13)

In Section A3.1, we derived the virial theorem for stars, according to which the total internal
energy, U , is related to the total gravitational potential energy, ||, by 3(γ − 1)U = ||.
Setting γ = 4/3 for a relativistic gas, U = ||, and so
 4/3
− ρ G M2
V hc ≈ . (A4.14)
mp R
Now, V ≈ R 3 and ρV = M. Therefore, the left-hand side of equation (A4.14) becomes
−  
hc M 4/3
. (A4.15)
R mp
Note the key point that, because we have used the relativistic equation of state, the left-
hand side of the equation depends upon radius as R −1 , exactly the same dependence as the
gravitational potential energy. Thus, the mass of the star does not depend upon its radius.
From equations (A4.13) and (A4.14), we find
 3/2
1 − hc
M≈ 2 ≈ 2M , (A4.16)
mp G
Notes to Section A4 73

dropping constants of order 1. This is an order-of-magnitude derivation of the Chan-

drasekhar mass.3
Note the physical meaning of the result given in equation (A4.16). For lower-mass
stars, the non-relativistic equation of state, p ∝ ρ 5/3 , should be used, and then equating
the gravitational potential energy and the internal thermal energy results in stars with a
definite radius. The origin of the collapse of relativistic degenerate stars can be understood
as follows. For masses less than the Chandrasekhar mass, inspection of equations (A4.14)
and (A4.15) shows that the internal energy exceeds the gravitational energy and so collapse
does not occur. However, for masses greater than the Chandrasekhar mass, the gravitational
potential energy exceeds the internal thermal energy and so the star collapses. Since both
energies depend upon radius R in the same way, if the gravitational energy once dominates,
it will always dominate. Hence, there is nothing to prevent collapse to a black hole. This
conclusion was the source of contention between Chandrasekhar and Eddington.
Note that the Chandrasekhar mass depends only upon fundamental constants. One of the
more intriguing ways of rewriting the expression (A4.16) is in terms of a ‘gravitational fine-
structure constant’, αG . The standard fine-structure constant is given by α = e2 /4π 0 − hc.
The equivalent formula for gravitational forces can be found by replacing e2 /4π 0 in
the inverse square law of electrostatics by G M 2 from Newton’s law of gravity, where m p
is the mass of the proton. Thus, αG = Gm 2p / − hc. Putting in the values of the constants,
we find α −1 = 137.04 and αG = 5.6 × 10−39 , the ratio of these constants being αG /α =
2.32 × 1040 , reflecting the enormous difference in the strengths of the electrostatic and
gravitational forces. Therefore, the Chandrasekhar mass is roughly given by
−3/2
M ≈ m p αG .
In other words, in terms of the basic constants of physics, stars are typically objects with
1060 protons. Note also that the calculation applies equally to white dwarfs and neutron
stars, the only difference being that the neutron stars are very much denser than the white
dwarfs.

Notes to Section A4

1 R. Kippenhahn and A. Weigert’s book, Stellar Structure and Evolution (Berlin: Springer-Verlag,
1990), deals with these and many other aspects of stellar evolution, bringing out the physical
principles very clearly.
2 S. I. Shapiro and S. A. Teukolsky’s book, Black Holes, Neutrons Stars and White Dwarfs: The
Physics of Compact Objects (New York: Wiley Interscience, 1983), can be thoroughly recom-
mended for a detailed treatment of degenerate stars.
3 I have given a more detailed derivation of the Chandrasekhar mass in Malcolm Longair, High
Energy Astrophysics, vol. 2 (Cambridge: Cambridge University Press, 1997).
Part II
The large-scale structure of the
Universe, 1900–1939
5 The Galaxy and the nature of
the spiral nebulae

The second part of our history concerns the understanding of the large-scale distribution
of matter in the Universe. At the beginning of the period 1900 to 1939, little was known
even about the structure of our own Galaxy; by the end of it, the Universe of galaxies was
established, the system was known to be expanding and general relativity provided a theory
capable of describing the distribution of matter in the Universe on the very largest scales.

5.1 ‘Island universes’ and the cataloguing of the nebulae

The earliest cosmologies of the modern era were speculative conjectures. The ‘island uni-
verse’ model of René Descartes (1596–1650), published in The World of 1636, involved an
interlocking jigsaw puzzle of solar systems. In 1750, Thomas Wright of Durham (1711–
1786) published An Original Theory or New Hypothesis of the Universe, in which the Sun
was one of many stars which orbit about the ‘Divine Centre’ of the star system. Immanuel
Kant (1724–1804) in 1755 and Johann Lambert (1728–1777) in 1761 took these ideas fur-
ther and developed the first hierarchical, or fractal, models of the Universe.1 Kant made the
prescient suggestion that the flattening of these ‘island universes’ was due to their rotation.
The problem with these early cosmologies was that they lacked observational validation, in
particular because of the lack of information on the distances of astronomical objects.
Towards the end of the eighteenth century, William Herschel (1738–1822) was one of
the first astronomers to attempt to define the distribution of stars in the Universe in some
detail on the basis of careful observation. To determine the structure of the Milky Way,
he counted the number of stars in different directions, assuming they all have the same
intrinsic luminosities. In this way, he derived his famous picture for the structure of our
Galaxy, consisting of a flattened disc of stars with diameter about five times its thickness,
the Sun being located close to its centre (Figure 5.1) (Herschel, 1785).
Herschel inferred that the nebulae were island universes similar to our Galaxy. A test of
this picture was to show that the nebulae could be resolved into stars, and he believed this
had been achieved in a number of cases. In others, he assumed that the nebulae were too
distant to be resolved into individual stars. This model came into question, however, when
Herschel discovered that among the nebulae were the planetary nebulae, which consist of
a central star surrounded by a shell of gas. Herschel recognised that these nebulae were

77
78 5 The Galaxy and the nature of the spiral nebulae

Figure 5.1: William Herschel’s model of the Galaxy based upon star counts in different directions.
The Sun is located close to the centre of the disc of stars (Herschel, 1785).

unlikely to be resolved into stars but rather consisted of ‘luminous fluid’ surrounding the
central star.
John Michell (1734–1793) had already warned Herschel that the assumption that the
stars have a fixed luminosity was a poor approximation. This is the same John Michell who
was Woodwardian Professor of Geology at Queens’ College, Cambridge, before becoming
the rector of Thornhill in Yorkshire in 1767. He designed and built what we now know as
the Cavendish experiment to measure the mean density of the Earth. Nowadays, he is also
remembered as the first person to realise that light could not escape from the surface of a
massive enough body, what would now be called a black hole (see Section 11.3) (Michell,
1784). In his remarkable pioneering paper of 1767, he introduced statistical methods into
astronomy in order to show that binary stars and star clusters must be physical associations
and not random associations of stars on the sky (Michell, 1767). Consequently, there must
be a dispersion in the absolute luminosities of the stars from the observed range of apparent
magnitudes in bright star clusters, such as the Pleiades. Despite this warning, Herschel
proceeded to produce a number of different versions of his model for the structure of our
Galaxy, adding appendages to account for various features of the star counts in different
directions.
In 1802, Herschel measured the magnitudes of visual binary stars and was forced to
agreed with Michell’s conclusion about the wide dispersion in the luminosities of the stars
(Herschel, 1802). Equally troubling was the fact that observations with his magnificent
40-foot telescope showed that as he studied fainter samples of stars, the more he continued
to find. Evidently, the stellar system was unbounded – there was no edge to the Galaxy.
Eventually, Herschel lost faith in his model of the Galaxy. On top of all these problems,
the importance of extinction by interstellar dust was not appreciated – it was only in the
 5.2 The structure of our Galaxy 79

1930s that its central importance for studies of our Galaxy was finally established (see
Section 5.6).
Meanwhile, the cataloguing of the nebulae was progressing steadily. Among the first lists
of bright nebulae to be published was that of Charles Messier (1730–1817), whose catalogue
of 103 objects was compiled during the years 1771–1784 (Messier, 1784).2 The list was
not intended primarily as a catalogue of interesting nebulae, but rather as a list of objects
to be avoided by comet-hunters, of whom Messier was a leading exponent. Many of the
bright nebulae are still referred to by their Messier numbers, for example, the Andromeda
Nebula, being the 31st entry in the catalogue, is M31; the Orion Nebula is M42 and the
Crab Nebula is M1.
The systematic cataloguing of the nebulae was begun by William Herschel, assisted by
his sister Caroline (1750–1848), using his 20-foot reflector at Slough. The first catalogue
was published in 1786 and consisted of 1000 nebulae. This was followed by a further 1000
entries in 1789 and 500 more in 1802. This work was continued by William’s son John
Herschel (1792–1871), who took the 20-foot telescope to the recently completed Royal
Observatory at the Cape of Good Hope in South Africa where he surveyed the Southern Sky
for nebulae. In 1864, John Herschel published the General Catalogue of Nebulae containing
5079 objects, of which all but 449 were discovered by the Herschels (Herschel, 1864). The
catalogue was compiled entirely from visual observations, using the 20-foot telescope as
a transit instrument, before the use of photographic methods became practicable for these
studies. This catalogue provided a large fraction of the entries in the New General Catalogue
of Nebulae and Clusters of Stars published by John Dreyer (1852–1926) in 1888 (Dreyer,
1888). This catalogue, the NGC Catalogue, is still the fundamental catalogue of bright
nebulae and contains positions and descriptions of the catalogued nebulae. Dreyer produced
two supplements to the NGC catalogues known as the Index Catalogues and the objects
therein are referred to by their IC numbers (Dreyer, 1895, 1908). In all, these catalogues
contained some 15 000 nebulous objects. The process of cataloguing bright nebulae was
completed by 1908.3
Among the nebulae there were undoubtedly numerous star clusters, and a common view
was that many of the diffuse nebulae were simply too distant to be resolved into stars.
Some of them were certainly ‘enormous masses of luminous gas or vapour’, as had been
convincingly demonstrated by Huggins and Miller’s important spectroscopic observations
of the 1860s (see Section 2.1). There remained, however, the issue of the nature of the spiral
nebulae – the definitive solution of this problem had to await the 1920s (see Section 5.3).

5.2 The structure of our Galaxy

The determination of the large-scale distribution of stars in our Galaxy and the understanding
of its internal dynamics became feasible once measurements of stellar motions and distances
became available. One of the objectives of these studies was the measurement of the motion
of the Sun relative to the nearby stars. This was accomplished by measuring the proper
motions of the stars, that is, their apparent angular motions on the sky relative to very distant
stars. In 1718, Edmund Halley (1656–1742) had noted that the positions of Aldebaran,
80 5 The Galaxy and the nature of the spiral nebulae

Betelgeuse and Sirius differed from the positions listed in the Almagest of Claudius Ptolemy
(second century AD), but it was not clear whether these differences were due to the motion
of the Sun, to the motions of the stars or to some combination of both (Halley, 1718). By
1783, William Herschel had measured the mean motion of the Sun relative to 13 bright stars
(Herschel, 1783), and in 1837 Friedrich Argelander (1799–1875) used the proper motions of
330 stars to find a mean motion of the Sun similar to Herschel’s estimate (Argelander, 1838).
When Hermann Kobold (1858–1942) extended the analysis to a fainter sample of over 1000
stars in 1895, however, a different answer was found – these observations showed that the
motions of the stars were not random but contained a systematic component (Kobold, 1895).
This was the situation when Jacobus Kapteyn (1851–1922) began his studies of the
proper motions of 2400 stars. By 1904, he had confirmed Kobold’s result that the stars did
not move randomly but rather that, when averages were taken within 28 separate areas,
there were systematic motions (Kapteyn, 1905). He found that the stars tended to move in
two preferred opposite directions. This result was confirmed by Eddington in 1908, who
suggested that there were two interpenetrating streams of stars (Eddington, 1908).
Just as the Harvard Observatory had become pre-eminent in the classification of stellar
spectra, so the Lick Observatory, first under James E. Keeler and then William Campbell
(1862–1938), became the leading observatory for the measurement of the radial velocities
of stars through spectroscopic measurements of their Doppler shifts. Although Huggins
had claimed to have measured the Doppler shift of Sirius A as early as 1868, the errors
were large, so large in fact that he claimed a positive radial velocity when it should have
been negative. Keeler, at the Lick Observatory, and Herman C. Vogel and Julius Scheiner
(1858–1913) at Potsdam understood the necessity of taking great care to eliminate the many
systematic errors which can enter into radial velocity measurements and reduced the typical
error to a few kilometres per second.
In 1896, the Lick Observatory received a benefaction from Mr D. O. Mills (1825–
1910) for the construction of a spectrograph, which Campbell optimised for radial velocity
measurements. Campbell already had ambitious plans for a very large survey of the radial
velocities of stars in the northern and southern hemispheres. When he became Director
of the Lick Observatory following Keeler’s death in 1900, he obtained a further major
grant from Mills for the construction and maintenance of a southern observing station at
Cerro San Cristobal, near Santiago in Chile, to undertake the southern part of the survey.
The initial part of the southern survey, consisting of 899 spectrograms, was completed by
1906 and the results for 150 stars were published in 1911 (Wright et al., 1911). The radial
velocity programme continued for many years, and it was 1928 before the catalogue of 2771
radial velocities for stars with magnitudes brighter than 5.51 was completed (Campbell and
Moore, 1928).
One of Campbell’s main interests was the determination of the solar motion, and as early
as 1901 he had derived a value of 19.89 ± 1.52 km s−1 , as well as determining the direction
of the apex of the solar motion. His estimates improved steadily over the years of the survey,
and by 1910 he had shown that the relative velocity of the two streams discovered by Kapteyn
was 40 km s−1 (Campbell, 1910). In 1907, Karl Schwarzschild suggested that it was not
necessary to think in terms of two star streams but rather that the velocity distribution of
the stars could be described in terms of a velocity ellipsoid, meaning that the local velocity
5.2 The structure of our Galaxy 81

Figure 5.2: Kapteyn’s model for the distribution of stars in the Galaxy (Kapteyn, 1922). The diagram
shows the distribution of stars in a plane perpendicular to the Galactic plane. The curves are lines of
constant number density of stars and are in equal logarithmic steps. The Sun, S, is slightly displaced
from the centre of the system.

dispersion of the stars had different values along three orthogonal directions,
  
1 vx2 v 2y vz2
p(vx , v y , vz ) ∝ exp − + 2+ 2 ,
2 σx2 σy σz

in which σx = σ y = σz (Schwarzschild, 1907). The longest axis of the ellipsoid lay along
the direction of Kapteyn’s two star streams.
To determine the scale and structure of the system of stars in the Galaxy, Kapteyn drew
up a plan of 206 Selected Areas in which deep star counts and proper motions would be
measured (Kapteyn, 1906). By this time, it was well known that there is a very wide range
of intrinsic luminosities among random samples of stars and so, to interpret the star counts,
it was necessary to work out the distribution of their luminosities in a typical volume of
space – this distribution is known as the luminosity function of the stars. By 1920, Kapteyn
and Pieter van Rhijn (1886–1960) had determined the luminosity function of stars near the
Sun and found that it could be approximated by a Gaussian distribution with mean absolute
magnitude M = 7.7 and half-width a few magnitudes (Kapteyn and van Rhijn, 1920).
Assuming this luminosity function applied throughout the Galaxy, they were able to work
out the space distribution of stars from star counts in different directions (Kapteyn, 1922).
They found that the Galaxy was highly flattened, with dimensions 1500 pc perpendicular
to the plane and about eight times that size in the Galactic plane (Figure 5.2).
Kapteyn used this model of the Galaxy to work out the gravitational acceleration which
binds the stars to the plane of the Milky Way (Kapteyn, 1922). The distribution of stars
perpendicular to the Galactic plane can be taken to be a plane parallel atmosphere in which
the number density of stars is described by a Boltzmann distribution perpendicular to
the plane, n = n 0 exp −(|z|/z 0 ), where z 0 is the scale height of the distribution. In a simple
picture, the scale height, z 0 , can be related to the gravitational acceleration, g, perpendicular
to the Galactic plane, since z 0 ≈ 12 mvz2 /g, and hence to the mass distribution once the
velocity dispersion of stars, vz2 , perpendicular to the plane is known. The mass density in
the Galactic plane was found to be 10−20 kg m−3 or 0.15M pc−3 , very close to modern
values. This density is often referred to as the Oort limit and provides an upper limit to the
total mass of stars, interstellar matter and dark objects of all types in the plane of the Galaxy
(Oort, 1932).
82 5 The Galaxy and the nature of the spiral nebulae

Meanwhile, Harlow Shapley had adopted a different approach to the determination of

Galactic structure. One of the most important methods for measuring astronomical distances
was discovered as a result of the systematic studies carried out at the Harvard Observatory’s
Arequipa observing station in Peru. From 1893 to 1906, the nearby companions of our
own Galaxy, the Large and Small Magellanic Clouds, were systematically surveyed pho-
tographically by the 24-inch telescope. At Harvard, Henrietta Leavitt (1868–1921), who,
like Annie Cannon, was extremely deaf, was assigned the task of finding variable stars in
the Magellanic Clouds. She had graduated from Radcliffe College in 1892 and after 1902
became the head of the photographic photometry department of the observatory. Whilst she
is best remembered for her work on the Cepheid variables, her main work was the estab-
lishment of the North Polar Sequence, the accurate determination of the magnitude scale
for stars in a region of sky which would always be accessible to observers in the northern
hemisphere. By the time of her death, in 1921, she had extended the North Polar Sequence
from 2.7 to 21st magnitude with errors less than 0.1 magnitudes. To achieve this, she used
observations from 13 telescopes ranging from 0.5 to 60 inches in diameter and compared
her scale using 5 different photographic photometric techniques. From 1915 to 1940, these
pioneering efforts were refined and developed almost single-handedly by Frederick Seares
(1873–1964) at the Mount Wilson Observatory.
The advantage of studying systems such as the Magellanic Clouds is that, although
their absolute distances may not be known, it is safe to assume that all the stars are at
the same distances, and hence that the relative luminosities of the stars can be found.
Leavitt’s technique for measuring small variations in the brightnesses of stars was to make
a positive plate of the star field; by overlaying this plate over negative plates taken at later
epochs, small changes in brightness could be measured rather precisely. Among the 1777
variable stars which she discovered in the Clouds were a number of Cepheid variable stars.
These stars were named after the variable star δ Cephei in which the periodic light curve
has a distinctive temporal behaviour, the brightness of the star increasing rapidly and then
decaying slowly to minimum light. By 1908, she had identified eight Cepheid variables in the
Small Magellanic Cloud and noted that the long-period variables had greater luminosities
than the short-period variables. In her famous paper of 1912 (Leavitt, 1912), the periods and
apparent magnitudes of 25 Cepheid variables were reported and their remarkable period–
luminosity relation, which has played such a prominent role in twentieth-century astronomy,
was displayed for the first time (Figure 5.3).
This discovery provided a powerful means for measuring astronomical distances because
the Cepheid variables are intrinsically luminous stars and their distinctive light curves can
be recognised in stars in distant systems. Once the absolute luminosities of the Cepheid
variables had been determined from studies of nearby examples, the period–luminosity
relation could be calibrated. Therefore, by measuring the period of a Cepheid variable, its
absolute luminosity could be found, from which the distance can be estimated from its
apparent magnitude using the inverse square law. This procedure was first carried out in
1913 by Hertzsprung, who derived a distance of 10 kpc for the Small Magellanic Cloud, the
greatest distance for any astronomical object measured at that time (Hertzsprung, 1913).4
In fact, this value is five times smaller than the present best estimate for the distance of the
Cloud.
5.2 The structure of our Galaxy 83

Figure 5.3: A plot of the period–luminosity relation for the 25 Cepheid variables discovered by Leavitt
in the Small Magellanic Cloud (Leavitt, 1912). The upper locus is found for the maximum light of
the Cepheid variables and the lower line for their minimum brightnesses.

The Cepheid variables were the tools used by Harlow Shapley to determine the structure
of the Galaxy through his studies of globular clusters. Globular clusters were among the
objects classed among the ‘nebulae’, but they have a spherically symmetric appearance
and can be clearly resolved into individual stars. In contrast to most of the components
of the Milky Way, they extend to high Galactic latitudes. By 1918, Shapley had realised
that the system of globular clusters provided a means of determining the scale of our
Galaxy. Among the stars which could be distinguished in the globular clusters were the
Cepheid variables, which he used to establish their distances. The distances found in this
way were entirely consistent with estimates based on observations of giant stars and other
characteristic stars found in these clusters. The scale of the system of globular clusters
was found to be enormous, the most distant globular cluster having a distance of 67 kpc.
Furthermore, the globular cluster system was not centred upon the Solar System, but rather
most of the globular clusters were found in a direction centred upon the constellation of
Sagittarius. Shapley plotted a map of the globular clusters and found that the Solar System
is located towards one edge of the globular cluster system (Figure 5.4) (Shapley, 1918).
He estimated the distance to the Galactic Centre to be about 20 kpc. Shapley’s picture of
the Galaxy differed radically from Kapteyn’s Sun-centred Universe, Shapley arguing that
Kapteyn’s studies referred only to the nearby part of the Galactic system.
84 5 The Galaxy and the nature of the spiral nebulae

Figure 5.4: The distribution of globular clusters in the Galaxy according to Shapley’s distance mea-
surements (Shapley, 1918). The scales on the abscissa and ordinate are in units of 100 pc and corre-
spond to distances in and perpendicular to the Galactic plane, respectively. The Sun, located at zero
coordinates on the abscissa and ordinate, lies towards one edge of the globular cluster system.

5.3 The Great Debate

These different approaches to the determination of the size and structure of our Galaxy and
the nature of the spiral nebulae led to what came to be known as the Great Debate.5 There
were two separate questions to be resolved, the first concerning the scale and structure of
our Galaxy and the second concerning the nature of the spiral nebulae. The first question
concerned the contrast between Shapley’s model of the Galaxy, which had dimensions of
at least 60 kpc and in which the Sun was located towards the outer boundary of the system,
and the Sun-centred model of Kapteyn, in which the Galaxy has dimensions of 10 kpc. The
second question concerned the issue of whether the spiral nebulae were ‘island universes’,
or whether they were constituent members of our own Galaxy.
In 1917, George W. Ritchey (1864–1945), the brilliant optician of the 60-inch and
100-inch telescopes at Mount Wilson, discovered by chance a nova6 in the spiral nebula
NGC 6946 (Ritchey, 1917). This led to searches through the plate archives of the major
observatories for further examples of novae in spiral nebulae. Heber D. Curtis (1872–1942)
and Shapley announced the discovery of several other novae, so that, by the end of 1917, 11
novae were known to have taken place in 7 spiral nebulae, 4 of them having been observed
in the Andromeda Nebula (Curtis, 1917; Shapley, 1917). Curtis noted that, at maximum,
the novae in our Galaxy typically have apparent magnitudes about 5.5, whereas those in the
spiral nebulae were about 10 magnitudes fainter. In consequence, if they were the same types
of object, the spiral nebulae would have to be 100 times more distant than their Galactic
5.3 The Great Debate 85

counterparts. Shapley drew the same conclusion, estimating the distance to the Andromeda
Nebula to be 50 times the distance of the nearby novae, that is, about 300 kpc.
There were, however, two big problems. The first was that an extraordinarily bright nova
had been observed in 1885 in the Andromeda Nebula, M31. It was 6 magnitudes brighter
than the novae which had been used to measure the distance to the Andromeda Nebula and
so, if it really were at a distance of 300 kpc, the nova of 1885 would have been more than
100 times brighter than the typical nearby novae. If the nova of 1885 were regarded as a
typical nova, the distance of the nebula would have been 10 times smaller.
The second problem was that Adriaan van Maanen (1884–1946) claimed to have mea-
sured proper motions in the arms of the bright spiral nebula M101 (van Maanen, 1916).
Similar results were reported for the galaxies M33, M51 and M81 in 1921 (van Maanen,
1921).7 The motions seemed to correspond to both rotational and radial motions, van
Maanen favouring motions along the spiral arms. His measurements of the rotational com-
ponents of these motions corresponded to rotation periods about the centres of the nebulae
of between 45 000 and 160 000 years. If these spiral nebulae had sizes similar to that esti-
mated by Shapley for our own Galaxy, the speed of rotation would exceed the speed of light.
As Shapley remarked,

Measurable internal proper motions, therefore, can not well be harmonised with ‘island universes’ of
whatever size, if they are composed of normal stars.

This was the background to the Great Debate between Shapley and Curtis, which took
place at the National Academy of Sciences in Washington on 20 April 1920 (Curtis, 1921;
Shapley, 1921).The course of the debate was complex, and the two separate issues identified
at the beginning of this section became interwoven. Shapley took the ‘scale of the Universe’
to mean the size of the globular cluster system, which he found to have dimensions of about
30 kpc. He accepted van Maanen’s observations of proper motions in the arms of spiral
nebulae, which he assumed must form part of an extensive halo about the Galaxy. He also
pointed out that the surface brightnesses of the spiral nebulae are very much greater than
the surface brightness of the plane of our Galaxy in the vicinity of the Sun and so it was not
evident that the spiral nebulae were the same class of object as our own Galaxy. There was
also the question of whether or not individual stars had been resolved in the spiral nebulae.
Furthermore, if Shapley’s large dimensions for our Galaxy were adopted, then, even if a
distance as large as 300 kpc were adopted for the spiral nebulae, our own Galaxy would
have been very much larger than the typical spiral nebula and so retain a unique position in
the Universe.
Curtis defended the smaller distances inferred from Kapteyn’s statistical studies and the
‘island universe’ picture. He made what turned out to be the correct inference that van
Maanen’s reported proper motions of the spiral arms of nebulae were spurious8 and placed
considerable weight upon the use of the novae as distance indicators, regarding the nova of
1885 in the Andromeda Nebula as an abnormality. He remarked further:

With one, and only one, exception, all known genera of celestial objects show such a distribution with
respect to the plane of our Milky Way, that there can be no reasonable doubt that all classes, save this
one, are integral members of our galaxy. We see that all the stars, whether typical, binary, variable, or
temporary, even the rarer types, show this unmistakable concentration towards the galactic plane. So
86 5 The Galaxy and the nature of the spiral nebulae

also for the diffuse and the planetary nebulae and, though somewhat less definitely, for the globular
star clusters.
The one exception is formed by the spirals; grouped about the poles of our galaxy, they appear
to abhor the regions of greatest star density. They seem clearly a class apart. Never found in the
Milky Way, there is no other class of celestial objects with their distinctive characteristics of form,
distribution, and velocity in space.

This was the origin of the term zone of avoidence, coined by Hubble. As Virginia Trimble
has pointed out, the arguments used by Shapley and Curtis were sound, so long as they were
discussing the areas in which they were experts (see endnote 5).
There were two problems in reconciling these different pictures. Most serious was the
neglect of interstellar extinction, that is, absorption and scattering of light by interstellar
dust, which affected Shapley’s and Kapteyn’s analyses in different ways. Curtis was well
aware of the importance of obscuring matter in the discs of spiral nebulae, as revealed by
his images of ‘a band of absorbing or occulting matter’ observed in those spiral nebulae
observed edge-on (Curtis, 1918b). Interstellar dust absorption in the plane of the Galaxy
was indeed responsible for the observation that the spiral nebulae avoid the Milky Way.
The central regions of our Galaxy in fact have very similar surface brightness to those of
the spiral nebulae, but interstellar extinction prevents us observing these regions directly
in the optical waveband. The second problem was that it cannot be assumed that the local
luminosity function of stars necessarily applies throughout the Galaxy.
Gradually, the discrepancies between the two pictures were resolved. Between 1917 and
1919, the Swedish astronomer Knut Lundmark (1889–1958) discovered 22 novae in the
Andromeda Nebulae and, if these were assumed to be similar to Galactic novae, a distance
of 200 kpc was found (Lundmark, 1920). Lundmark made the distinction between two
classes of novae, those used to make the distance measurements belonging to the ‘lower
class’, while novae such as that of 1885 were assigned to the ‘upper class’ – these were to
be identified with ‘super-novae’ by Baade and Zwicky in 1934 (see Section 4.3).
In 1899, Julius Scheiner had obtained a spectrum of the central regions of M31 in a 71/2
hour exposure and found that its spectrum was similar to that of the Sun (Scheiner, 1899).
In 1921, Lundmark extended this observation, making a detailed spectroscopic study of the
spiral arms of M33, as well as some of its brightest stars, and found them to be typical of
luminous stars in our Galaxy. If he assumed the brightest stars in the spiral nebulae had
absolute magnitude M = −6, the distance of M33 would be about 300 kpc. In 1921, he
wrote (Lundmark, 1921):

Some objects [in the arms] have a nebular spectrum but most of the objects belonging to the spiral
show a strong continuous spectrum without bright lines. It is of course hard to give an accurate spectral
type but a solar or somewhat earlier type seems to be predominant. From the spectral evidence, it
seems probable that the spiral nebula consists of ordinary stars, clusters of stars, and some nebular
[i.e. gaseous] material.

Further evidence for the extragalactic nature of M31 was provided by Ernst Öpik, who
used measurements of its rotational velocity by Francis Pease to show that if its mass-to-
luminosity ratio were similar to that of stars in our Galaxy, its distance would have to be
 5.4 Hubble and the Universe of galaxies 87

about 480 kpc, in fact a more accurate estimate than that found subsequently by Hubble
(Öpik, 1922).
The conclusive proof of the extragalactic nature of the spiral nebulae was provided by
Edwin Hubble (1889–1953) in 1925 (Hubble, 1925). Using the Hooker 100-inch telescope,
he discovered 22 Cepheid variables in M33 and 12 in M31. These displayed exactly the
same form of period–luminosity relation found for Cepheids in the Magellanic Clouds.
He was therefore able to make good distance estimates for the spiral nebulae, which he
found to be 285 kpc, much greater than Shapley’s largest estimate for the size of our
Galaxy.

5.4 Hubble and the Universe of galaxies

The extragalactic nature of the spiral galaxies was established and Hubble immediately
began to use the galaxies as tools for studying the large-scale structure of the Universe. He
realised that the galaxies provided the means by which fundamental cosmological problems
could be addressed by astronomical observation. In the next year, 1926, he published a
major study of galaxies which begins with his famous classification scheme, distinguishing
between the main classes of galaxies – the ellipticals, normal spirals, barred spirals and
irregulars (Hubble, 1926). Elliptical galaxies were ordered according to the ellipticity of
their images, and the spirals and barred spirals were divided into subclasses labelled a,
b and c according to the tightness of the winding of the spiral structure and the relative
importance of the disc and bulge in the distribution of stars in the galaxy. This classification
scheme was eventually presented in the form of a ‘tuning-fork’ diagram published in 1936
(Figure 5.5) (Hubble, 1936). Hubble interpreted the diagram as an evolutionary sequence in
which the galaxies were supposed to evolve from spherical elliptical galaxies at the left of
the diagram through the sequence of spiral galaxies. This speculation proved to be wholly
incorrect, but the terms ‘early-type’ galaxy and ‘late-type’ galaxy are still used, reflecting
Hubble’s original prejudice.
Of particular significance for cosmology was his realisation that the number counts
of galaxies brighter than a given apparent magnitude provide a test of the homogeneity
of the distribution of galaxies in the Universe. It is a simple calculation to show that, if
the galaxies are distributed uniformly in local Euclidean space, the number brighter than
limiting apparent magnitude, m, is expected to be log N = 0.6m + (constant), independent
of the luminosity function of the galaxies (see Section A5.1). In 1926, Hubble’s galaxy
counts extended to 16.7 magnitude, and he found that the number of galaxies increased with
increasing apparent magnitude exactly as expected for a uniform distribution. This result
was to have profound implications for the construction of cosmological models because it
meant that, as a first approximation, the Universe could be taken to be homogeneous on the
large scale.
Next, Hubble worked out the typical masses of galaxies, and from this he estimated the
mean mass density in the Universe. The value he found was ρ = 1.5 × 10−28 kg m−3 .
Already in this paper of 1926, Hubble recognised that this figure had cosmological
88 5 The Galaxy and the nature of the spiral nebulae

Figure 5.5: Hubble’s ‘tuning-fork’ diagram illustrating the sequence of nebular types. As Hubble
noted in the caption to this diagram, which appears in The Realm of the Nebulae (Hubble, 1936),
‘The diagram is a schematic representation of the sequences of classification. A few nebulae of mixed
types are found between the two sequences of spirals. The transitional stage, S0, is more or less
hypothetical. The transition between E7 and SBa is smooth and continuous. Between E7 and Sa, no
nebulae are definitely recognised.’ The S0 galaxies were later recognised in photographic surveys
of nearby galaxies and may be thought of as disc galaxies with central bulges but without spiral
arms.

significance. Adopting Einstein’s static model for the Universe (Einstein, 1917) (see Sec-
tion 6.2.2), he found that the radius of curvature of the spherical geometry was 27 000 Mpc
and that the number of galaxies in this closed Universe was 3.5 × 1015 . In the last para-
graph of this 1926 paper, he noted that the 100-inch telescope could observe typical
galaxies to about 1/600 of the radius of the Einstein Universe and bright galaxies such as
M31 to several times this distance. He concluded this remarkable paper by remarking that
(Hubble, 1926)

. . . with reasonable increases in the speed of plates and sizes of telescopes it may become possible
to observe an appreciable fraction of the Einstein Universe.

Thus, by 1926, the first application of the ideas of relativistic cosmology to the Universe
of galaxies had been made. It comes as no surprise that in 1928 George Ellery Hale, Director
of the Mount Wilson Observatory, began his campaign to raise funds for the construction of
the Palomar 200-inch telescope – the study of the Universe of distant galaxies needed the
largest telescopes that could be built (Hale, 1928). In the great American tradition of private
sponsorship of observational astrophysics, in which the USA had taken a decisive lead, Hale
was successful in obtaining a grant of $6 000 000 from the Rockefeller Foundation for the
telescope before the year was out.
Before tackling the remarkable story of the discovery of the expanding Universe and the
development of theoretical cosmology, let us complete the story of the understanding of
Galactic structure and the key role of interstellar extinction.
 5.5 The discovery of Galactic rotation 89

5.5 The discovery of Galactic rotation

The first clues which were to lead to the discovery of Galactic rotation came from the radial
velocity programmes established at the Lick and Mount Wilson Observatories. Most of the
radial velocities of the stars were less than about 50 km s−1 , but, by 1901, Campbell at the
Lick Observatory had noted seven stars with velocities greater than 76 km s−1 (Campbell,
1901). In 1914, Adams and Kohlschütter found that among the high-velocity stars with
velocities greater than 50 km s−1 , those approaching the Solar System outnumbered those
moving away from it by a ratio of three to one (Adams and Kohlschütter, 1914a). Their
sample included two stars with radial velocities of −325 and −242 km s−1 . In 1918,
Benjamin Boss (1880–1970) showed that all the stars with negative velocities greater than
75 km s−1 lay in the range of Galactic longitude 140◦ to 340◦ (Boss, 1918). These results
were soon confirmed by Adams and Alfred Joy (1882–1973) and by Gustaf Strömberg
(1882–1962), who by 1924 had discovered about 100 high-velocity stars, all within a range of
Galactic longitude 143◦ to 334◦ , indicating that this system of stars had a highly asymmetric
velocity distribution relative to the Sun and the bulk of the nearby stars (Adams and Joy,
1919; Strömberg, 1924). On average, the high-velocity stars moved at about 300 km s−1
relative to the Sun. Strömberg pointed out that a possible interpretation of these data was
that the system of high-velocity stars was in fact at rest and that the Sun and the nearby
stars were moving through it at high velocities.
In 1925, Bertil Lindblad (1895–1965) introduced the idea that the local system of stars
is rotating about the Galactic centre (Lindblad, 1925). His model of the Galaxy was similar
to Shapley’s, but he divided the system of stars into a number of separate subsystems which
rotated at different velocities about the Galactic centre, which he identified with the centre
of the distribution of globular clusters. In this picture, the high-velocity stars define the
local standard of rest for the Galaxy as a whole and the local stellar distribution drifts
through it. The nearby nebulae partake in the same motion as the local stars. He deduced
a local rotational velocity of 350 km s−1 , a Sun-centre distance of 12 kpc and a Galactic
mass of 1.8 × 1011 M . In a paper of 1927, he went on to show that, if the orbits of stars
moving in purely circular orbits about the centre are perturbed, the result is epicyclic motion
about the circular orbit (Lindblad, 1927). The epicycles are elongated along the circular
orbits, but the major axis of the velocity ellipsoid associated with a system of stars in orbit
about the Galactic centre is perpendicular to this direction. The magnitude of the perturbing
velocities associated with the epicyclic motion was of the same order as the amplitude of
the observed velocity ellipsoid. The beautiful result of this analysis was that it predicted that
the major axis of the velocity ellipsoid should be perpendicular to the streaming velocity
of the high-velocity stars, which is what was observed.
Following Lindblad’s work, Jan Oort (1900–1992) realised that a direct way of testing
the hypothesis of Galactic rotation was to look for the effects of differential rotation in the
distribution of local stellar velocities. In a model such as Lindblad’s, it is expected that
the stars should not rotate as a solid body but rather that the effects of differential rotation
as a function of distance from the Galactic centre should be detectable. Oort first showed
90 5 The Galaxy and the nature of the spiral nebulae

Table 5.1. The residual radial velocities of c-stars from data by Schilt used by Oort
(1927) to demonstrate the local differential rotation of stars in our Galaxy
To match the observed sinusoidal variation of the radial velocity residuals, it had to be
assumed that the direction of the centre of rotation was in the direction l = 325◦ . If the
numbers in the final column are multiplied by 10, good agreement within the limits of the
experimental errors is obtained.

Average Average peculiar Mean error

longitude velocity (km s−1 ) (km s−1 ) sin 2(l − 325◦ )

30 +8 ±3.5 +0.77
90 −8 ±2.7 −0.94
150 0 ±3.6 +0.17
210 +10 ±3.9 +0.77
270 −7 ±4.3 −0.94
330 0 ±3.5 +0.17

by a simple geometrical argument that the variation of radial and tangential velocities,
vr and vt , respectively, observed from the Sun at distance r and Galactic longitude l are
given by

vr = Ar sin 2l, vt = Ar cos 2l + Br, (5.1)

where
       
1 V0 dV 1 V0 dV
A= − and B=− + (5.2)
2 R0 dR R=R0 2 R0 dR R=R0

(Oort, 1927).† Note that V (R) is the circular velocity at radial distance R from the Centre
and that the subscript 0 refers to the orbit of the local standard of rest at the Solar System.
Oort’s A and B constants contain information about the velocity of the local standard of
rest about the Galactic centre, the distance of the Sun from the centre and the local variation
of the rotational velocity with distance from the centre.
In his paper of 1927, Oort found clear evidence for this sinusoidal variation of the radial
velocities with Galactic longitude in all the sets of data in which it might be expected
to be found. The sinusoidal effect was particularly noticeable for the c-stars (Table 5.1).
Taking the rotational velocity of the Sun about the Galactic centre to be 272 km s−1 from
the radial velocities of globular clusters, a new distance estimate for the Galactic centre
of 5.9 kpc was found, which was subsequently revised to 5.1 kpc. Note that this analysis
gave an independent estimate of the direction of the centre of the Galaxy, which agreed
with Shapley’s estimate for the centre of the distribution of the globular clusters. Another
attractive feature of this picture is that the epicyclic angular frequency, κ, is related to
Oort’s B constant by κ 2 = −4Bω0 , where ω0 = V0 /R0 is the angular velocity of the Sun
about the Galactic centre. Oort recognised that this model of the Galaxy differed from that

† A simple proof of Oort’s equations is given in Section A5.2.

 5.6 Interstellar matter and extinction by dust 91

deduced by Kapteyn, but he proposed that the discrepancy was likely to be due to interstellar
extinction.

5.6 Interstellar matter and extinction by dust

In 1904, Johannes Hartmann (1865–1936) reported the observation that the narrow H and K
lines of ionised calcium do not share the periodic displacement of the lines seen in the double
star δ Orionis (Hartmann, 1904). He concluded that the absorption must occur along the
line of sight to the star. In 1909, Edwin Frost (1866–1935) detected similar stationary lines
in several bright, young stars, and Vesto Slipher, in the same year, confirmed Hartmann’s
findings for binary stars (Frost, 1909; Slipher, 1909). These observations provided the first
evidence for diffuse absorbing gas along the line of sight to the stars.
In 1923, John Plaskett (1865–1941) of the Dominion Astrophysical Observatory in
Victoria, British Columbia, showed that the velocities of the calcium absorption lines differ
from those of their background stars by up to 50 km s−1 (Plaskett, 1923). In 1933, he and
Joseph Pearce (1893–1988) completed an exhaustive study of the radial velocities of O and
B stars (Plaskett and Pearce, 1933). The stars exhibited the expected sinusoidal distribution
of radial velocities with Galactic longitude due to differential rotation (equations (5.1)),
but in addition the interstellar absorption lines also showed a sinusoidal behaviour with
amplitude half that of the stellar absorption lines. Plaskett and Pearce inferred correctly that
the interstellar material also partakes in the Galactic differential rotation and is, on average,
smoothly distributed along the line of sight to the stars.
It had been suspected that interstellar absorption by dust was an important influence upon
the magnitudes and spectra of distant stars, but most observers up till about 1930 preferred
to assume that interstellar space was transparent. Kapteyn, for example, could find no
definite evidence of extinction in his data. The first conclusive evidence that interstellar
extinction could not be ignored came from the analysis of Robert Trumpler (1886–1956)
of the properties of open clusters. In a paper published in 1930, he determined the absolute
magnitudes of the stars in the clusters from their spectral properties and so was able to
measure their distances, knowing their apparent magnitudes (Trumpler, 1930). Employing
this procedure, he found that the clusters were systematically larger in physical size with
increasing distance as compared with the nearby clusters. Assuming the clusters all had the
same physical size, he estimated that the extinction along the line of sight to the clusters
amounted to 0.67 mag kpc−1 .
An even more compelling estimate of the importance of interstellar extinction was made
by Alfred Joy, who published the results of his study of the radial velocities of Cepheid
variables in 1939 (Joy, 1939). For these stars absolute magnitudes could be estimated from
the period–luminosity relation. He found that the more distant Cepheids were fainter than
expected and demonstrated that a result consistent with Galactic rotation could be found,
provided the light from the distant stars is attenuated by interstellar dust extinction amounting
to 0.85 mag kpc−1 (Figure 5.6).
By 1940, Joel Stebbins (1878–1966), Charles Hufford (1894–1981) and Albert
Whitford (1905–2002) had used photoelectric photometric techniques to show that
92 5 The Galaxy and the nature of the spiral nebulae

Figure 5.6: Illustrating the evidence for differential rotation within the disc of the Galaxy from Joy’s
data of 1939 for 156 Cepheid variables (Joy, 1939). The radial velocities, V , are plotted against
Galactic longitude, l, for four groups of stars at progressively greater distances from the Sun. The
solid lines show the predicted sinusoidal variation of the radial velocities with Galactic longitude
assuming Oort’s formula, V = Ar sin 2l . A mean value of A = 21 km s−1 kpc−1 was found. Joy used
these data to show that the interstellar extinction amounts to 0.85 mag kpc−1 .

interstellar extinction was sufficient to reconcile the dimensions of the Galaxy derived
from the globular clusters and from the distribution of nearby stars (Stebbins, Hufford and
Whitford, 1940). They also showed, however, that the distribution of absorbing dust was
patchy and that the standard extinction quoted in the literature of 1 mag kpc−1 is only a
global average. For precise work, it is necessary to evaluate the extinction separately along
each line of sight.

5.7 The Galaxy as a spiral galaxy

The result of these studies was to demonstrate that our Galaxy, although among the more
luminous and massive of the spiral galaxies, was by no means exceptional. The obvious
question to ask was whether or not our Galaxy possesses spiral arms similar to those
observed in the spiral galaxies. The means for studying this question was provided by
Walter Baade, who undertook a magnificent set of observations of the Andromeda Nebula
and its companion galaxies M32 and NGC 205 in 1943. At that time, Los Angeles and
Hollywood were subject to a wartime black-out, which enabled plates of quite exceptional
quality to be taken. The acquisition of these plates was a virtuoso performance of observing
5.7 The Galaxy as a spiral galaxy 93

Figure 5.7: Evidence for the spiral arms of our Galaxy from the spatial distribution of associations of
bright O and B stars. The position of the Sun is indicated by the letter S and the centre of the Galaxy
is shown by the double circle at the bottom of the diagram. This diagram is taken from the paper of
Morgan, Whitford and Arthur Code (b.1923) (Morgan, Whitford and Code, 1953), which included
better distance measurements for the OB associations than the earlier abstract by Morgan, Sharpless
and Osterbrock (1951).

technique, since the focus of the telescope changed during the exposure and Baade worked
out how to guide and correct the focus from the comatic image of an off-axis star observed
at a magnification of 2800. The importance of these observations was that he was able to
resolve not only the brightest blue stars but also the fainter red stars.
In his famous paper of 1944, Baade reported the important discovery that different
spectral classes of stars formed different populations within spiral galaxies (Baade, 1944).
He divided the stars into two classes. Population I consisted of open star clusters and highly
luminous O and B stars which were found exclusively in the regions of spiral arms. In a
subsequent paper, he noted that gas clouds and dense regions of interstellar dust were also
characteristic of Population I objects and made the correct inference that the young O and B
stars had recently been formed in these dusty clouds (Baade, 1951). In contrast, Population
II objects comprised the bulk of the old red population of stars which define the central
bulge of the Galaxy and the old disc population. The globular clusters and the high-velocity
stars are typical members of Population II. The differences between the populations are
summarised in Table 5.2.9 One of the important consequences of these studies was that he
also discovered a difference in absolute magnitude between the Cepheid variables belonging
to the two populations (Baade, 1952). This was to have important consequences for the
cosmological distance scale.
Table 5.2. These data, taken from Allen (1973), give some impression of the differences in properties of stars belonging to different stellar
populations in our Galaxy

Population I Population II

Extreme Older Older disc Intermediate Extreme or halo

General description very hot blue stars; often halo stars, old red stars; globular
associated with HII clusters belong to this
regions and dust population
spiral arm population disc population old disc population
Properties of stars ←−metal-rich stars−→ ←−metal-poor stars−→ extremely metal-poor
Heavy-element abundance 0.04 0.02 0.01 0.004 0.001
Ages of stars in units of 109 years ≤ 0.1 0.1–1.5 1.5–5 5–6 >6
Spatial distribution
Extent perpendicular to Galactic 120 160 400 700 2000
plane (kpc)
Mean speed perpendicular to 8 10 16 25 75
Galactic plane (kpc)
Axial ratio of stellar distribution 100 50 20 5 2
Smoothness of distribution very patchy patchy smooth smooth smooth
Concentration to Galactic centre little little strong strong strong
Notes to Chapter 5 95

Figure 5.8: The COBE composite infrared image of the Galaxy in an Aitoff projection made from
observations at 1.25, 2.2, 3.5 and 4.9μ, showing clearly its disc–bulge structure. Analysis of the
photometry of the central bulge has suggested that it is elongated, as expected in a barred spiral
galaxy (Dwek et al., 1995).

In 1951, William Morgan (1906–1994), Stewart Sharpless (b. 1926) and Donald Oster-
brock (b. 1924) published their study of the distribution of O and B stars within 2 kpc of
the Sun and found clear evidence for ‘spiral arms’, the Sun lying on the inner edge of the
local spiral arm (Figure 5.7) (Morgan, Sharpless and Osterbrock, 1951).
The final question in this saga is whether our Galaxy is a normal or a barred spiral.
This is not a trivial question because of the effects of interstellar absorption and the fact
that our Solar System is located within the disc of the Galaxy. The most recent infrared
images of the Galaxy obtained by the COBE satellite, which are free from the effects of
interstellar extinction, show clearly the disc and bulge structure of the old stars in our Galaxy
(Figure 5.8). The same observations have suggested that the central bulge of old stars may
be somewhat ellipsoidal, and so it may well be that we actually live in a barred spiral galaxy
(Dwek et al., 1995).

Notes to Chapter 5

1 Edward Harrison gives a delightful brief survey of these ideas in his text Cosmology (Cambridge:
Cambridge University Press, 2001). Many of these early ideas are also reviewed by John North in his
important book The Measure of the Universe (Oxford: Clarendon Press, 1965). Detailed studies of
the contributions of Wright, Kant and Lambert are included in Section B, A century of speculative
cosmologies, of Michael Hoskin’s Stellar Astronomy (Chalfont St Giles, Buckinghamshire: Science
History Publications, 1982).
2 The compilation of Messier’s catalogue and how amateur astronomers can observe the objects listed
are included in S. J. O’Meara’s book, Deep Sky Companions: The Messier Objects (Cambridge:
Cambridge University Press, 1998). Note that further objects were added to Messier’s original
catalogue, the list finally totalling 110 objects.
96 5 The Galaxy and the nature of the spiral nebulae

3 The most up-to-date catalogue of the brightest galaxies in the sky was published by the de Vau-
couleurs and their colleagues: G. de Vaucouleurs, A. Vaucouleurs, H. G. Corwin, Jr., R. J. Buta,
G. Paturel and P. Fouque, Third Reference Catalogue of Bright Galaxies: Containing Informa-
tion on 23,024 Galaxies With Reference to Papers Published Between 1913 and 1988 (Berlin:
Springer-Verlag, 1991).
4 There was a numerical error in Hertzsprung’s paper in which a distance of only 3000 light years,
about 1 kpc, for the Magellanic Clouds was quoted. This was still a large enough distance to place
the Clouds well above the plane of the Galaxy.
5 This story is told in R. Berendzen, R. Hart and D. Seeley, Man Discovers the Galaxies (New York:
Science History Publications, 1976) and R.W. Smith, The Expanding Universe: Astronomy’s ‘Great
Debate’ 1900–1931 (Cambridge: Cambridge University Press, 1982). As a number of authors have
noted, the Great Debate is a somewhat exaggerated title for the two half-hour talks presented at
the National Academy of Sciences (see M. A. Hoskin, The ‘Great Debate’: what really Happened,
Journal of the History of Astronomy, 7, 1976, 169–182 and V. Trimble, The 1920 Shapley–Curtis
discussion: background, issues, and aftermath, Publications of the Astronomical Society of the
Pacific, 107, 1995, 1133–1144).
6 Novae, or ‘new stars’, are stellar explosions in which the luminosity of the star increases suddenly
by a factor of hundreds to a million. It remains at about this luminosity for about three days to several
months, after which it returns to its pre-nova luminosity. Novae in our Galaxy occur about four
times per year. The explosions are associated with mass transfer in binary systems which contain a
white dwarf. As the mass transfer onto the white dwarf continues, the surface temperature increases
until hydrogen burning takes place in a thermonuclear runaway explosion.
7 Van Maanen continued to publish a series of papers describing these motions in a number of
galaxies during the 1920s.
8 In fact, it took quite some time and effort to demonstrate that van Maanen’s results were incorrect.
The problem was that he was attempting to measure the proper motions of diffuse structures, which
are notoriously difficult observations and very sensitive to the observing conditions. Van Maanen’s
proper motions were only finally laid to rest by Hubble’s observations of 1935 (Hubble, 1935). Van
Maanen retracted his claims as overestimates of the magnitudes of the proper motions in the same
year (van Maanen, 1935).
9 Allen’s compendium of astronomical data was a valuable resource and represented the state of
knowledge in the 1960s and 1970s: C. W. Allen, Astrophysical Quantities, 3rd edn (London:
Athlone Press, 1973).

A5 Explanatory supplement to Chapter 5

A5.1 Euclidean number counts of galaxies
In his famous monograph The Realm of the Nebulae, Edwin Hubble used counts of galax-
ies to the limit of the Mount Wilson 100-inch telescope to demonstrate that, overall, the
distribution of galaxies is homogeneous on the large scale (Hubble, 1936). The argument
goes as follows.
Suppose the galaxies have a luminosity function n(L) dL and that they are uniformly
distributed in Euclidean space. The numbers of galaxies with flux densities greater than
different limiting values, S, in a particular solid angle  on the sky is denoted N (≥ S).
Consider first galaxies with luminosities in the range L to L + dL. In a survey to a limiting
flux density S, these galaxies can be observed out to some limiting distance r , given by the
inverse square law, r = (L/4π S)1/2 . The number of galaxies brighter than S is therefore
 Explanatory supplement to Chapter 5 97

the number of galaxies within distance r in the solid angle :

 3
N (≥ S, L) dL = r n(L) dL .
3
Therefore, substituting for r , the number of galaxies brighter than S is given by
 
 L 3/2
N (≥ S, L) dL = n(L) dL .
3 4π S
Integrating over the luminosity function of the galaxies, we obtain

 −3/2
N (≥ S) = S L 3/2 n(L) dL ,
3(4π)3/2
that is, N (≥ S) ∝ S −3/2 , independent of the luminosity function n(L). The result
N (≥ S) ∝ S −3/2 is known as the Euclidean number counts for any class of extragalactic
object. In terms of apparent magnitudes, m = constant − 2.5 log10 S, the Euclidean number
counts become
N (≤ m) ∝ 100.6m , that is, log N (≤ m) = 0.6m + constant.
This was the homogeneity test carried out by Hubble with the results shown in Figure 6.2.

A5.2 Oort’s A and B constants

Oort’s famous paper of 1927 includes an elegant demonstration of how the rotation curve for
our Galaxy in the vicinity of the Sun can be determined (Oort, 1927). By rotation curve, we
mean the variation of the circular velocity, V (R), of stars about the centre of the Galaxy as a
function of radial distance, R, from the centre. Oort’s argument demonstrated conclusively
that the disc of our Galaxy is in a state of differential rotation, provided a revised estimate
of the direction of the Galactic centre and enabled an estimate of the distance of the Sun
from the Galactic centre to be made.
The geometry of a differentially rotating disc is shown in Figure A5.1. Oort’s analysis
involved determining the radial and azimuthal components of the velocities of nearby stars
as observed from the frame of reference of the Sun at O, which itself is rotating about the
Galactic centre. Consider first the radial component, vr , of stars at S, as observed from O. The
projected velocities of the Sun and the stars at S along the direction OS are V0 cos (π/2 − l)
and V (R) cos α, and so, subtracting, yields
π
vr = V (R) cos α − V0 cos −l (A5.1)
2
= Rω(R) cos α − R0 ω0 sin l, (A5.2)
where ω and ω0 are the angular velocities of the rotating disc about the centre at R and
R0 , respectively. Similarly, the azimuthal velocity, vθ , in the direction of increasing l is the
difference of the projected velocities in the direction perpendicular to OS as shown:
vθ = V (R) cos(π/2 − α) − V0 cos l (A5.3)
= Rω(R) sin α − R0 ω0 cos l. (A5.4)
98 5 The Galaxy and the nature of the spiral nebulae

V0

V(R)

Figure A5.1: The geometry of a differentially rotating disc. The Sun is located at O and the Galactic
latitude, l, of a star at distance r from O is indicated. Differential rotation means that the disc of stars
does not rotate as a solid body, for which the angular velocity, ω, would be a constant for all radii r .
The dashed arrows show the radial and azimuthal directions as observed from O.

Now apply the sine rule to the sides OC and SC of the triangle OCS:
sin(π/2 + α) sin l
= ; (A5.5)
R0 R
R cos α = R0 sin l. (A5.6)
Therefore, equation (A5.3) becomes
vr = R0 (ω − ω0 ) sin l. (A5.7)
Now consider the triangle OCP, where CP is the perpendicular to the radial direction from
O, OS. The length OP is given by R0 cos l and also r + R cos(π/2 − α). Therefore,
R0 cos l = r + R sin α. (A5.8)
Therefore, substituting for R sin α in equation (A5.4), the azimuthal component of the
velocity of the stars at S as observed from O is given by
vθ = R0 (ω − ω0 ) cos l − ωr. (A5.9)
So far, the analysis is exact. Oort noted that, for small distances, such that r R0 , a
simpler relation is found for the expected motions of the stars. Performing a Taylor expansion
to first order in r/R0 , we obtain
 
dω
ω = ω0 + (R − R0 ). (A5.10)
dR R0
Explanatory supplement to Chapter 5 99

Therefore, writing ω = V /R, equation (A5.10) becomes

   
1 dV (R)
ω − ω0 = 2 R0 − V0 (R − R0 ). (A5.11)
R0 dR R0
For small values of r , (R0 − R) ≈ r sin l, and so we find
   
V0 dV (R)
vr = − r sin l cos l, (A5.12)
R0 dR R0
or
vr = Ar sin 2l, (A5.13)
where Oort’s A constant is given by
   
1 V0 dV (R)
A= − . (A5.14)
2 R0 dR R0

Similarly, for the azimuthal component of apparent motion,

   
V0 dV (R)
vθ = − r cos2 l − ω0 r, (A5.15)
R0 dR R0
or
vθ = Ar cos 2l + Br, (A5.16)
where Oort’s B constant is given by
   
1 V0 dV (R)
B= + . (A5.17)
2 R0 dR R0

Thus, the combination of the radial velocity residuals, vr , and the proper motion residuals,
vθ , enables A and B to be found and hence V0 /R0 and (dV /dR) R0 to be determined. Once
V0 has been determined, the distance of the Galactic centre, R0 , can be found.
 6 The origins of astrophysical
cosmology

6.1 Physical cosmology up to the time of Einstein

Gravity is the one long-range force which acts upon all matter. Soon after Isaac Newton had
completed the unification of the laws of gravity and celestial physics through his discovery
of the inverse square law of gravity, he appreciated that the unique form of this law has
important consequences for the large-scale distribution of matter in the Universe.1 In 1692–
1693, the cosmological problem was addressed in a remarkable exchange of letters between
Newton and the young clergyman Richard Bentley (1662–1742), later to become master
of Trinity College, Cambridge. The correspondence concerned the stability of a Universe
uniformly filled with stars under Newton’s law of gravity.2 The attractive nature of the
force of gravity meant that matter tends to fall together, and Newton was well aware of this
problem. His first solution was to suppose that the distribution of stars extends to infinity in
all directions so that the net gravitational attraction on any star in the uniform distribution
is zero. As he wrote,3

The fixt Stars, everywhere promiscuously dispers’d in the heavens, by their contrary attractions destroy
their mutual actions.

Newton made star counts to test the hypothesis that the stars are uniformly distributed
in space and found that the numbers increased more or less as expected with increasing
apparent magnitude. The problem, which was fully understood by Newton and Bentley, was
that a uniform distribution of stars is dynamically unstable. If any star is slightly perturbed
from its equilibrium position, the attractive force of gravity causes the star to continue to
fall in that direction. Newton had to adopt the unsatisfactory assumption that the Universe
had been set up and remained in a perfectly balanced state.
During the late eighteenth century, non-Euclidean geometries began to be taken seriously
by mathematicians who realised that Euclid’s fifth postulate, that parallel lines meet only
at infinity, might not be essential for the construction of a self-consistent geometry.4 The
first proposals that the global geometry of space might not be Euclidean were discussed
by Girolamo Saccheri (1667–1733) and Johann Lambert. In 1766, Lambert noted that, if
space were hyperbolic rather than flat, the radius of curvature of the space could be used
as an absolute measure of distance.5 In 1816, Carl Friedrich Gauss (1777–1855) repeated
this proposal in a letter to Christian Gerling (1788–1864) and was aware of the fact that a

100
6.1 Physical cosmology up to the time of Einstein 101

test of the local geometry of space could be carried out by measuring the sum of the angles
of a triangle between three high peaks in the Harz mountains: the Brocken, Hohenhagen
and Inselberg. In 1818, Gauss was asked to carry out a geodetic survey of the state of
Hanover and he devoted a large effort to carrying out and reducing the data himself. He
was certainly aware of the fact that the sum of the angles of the triangle formed by the three
Harz Mountains was 180◦ within the limits of geodetic measurements.6
The fathers of non-Euclidean geometry were Nicolai Ivanovich Lobachevsky (1792–
1856), who became rector of Kazan University in Russia in 1827, and János Bolyai (1802–
1860) in Transylvania, then part of Hungary. In the 1820s, they independently solved the
problem of the existence of non-Euclidean geometries and showed that Euclid’s fifth pos-
tulate could not be deduced from the other postulates (Bolyai, 1832; Lobachevsky, 1829,
1830). In his papers entitled ‘On the principles of geometry’, Lobachevsky also proposed an
astronomical test of the geometry of space. If the geometry were hyperbolic, the minimum
parallax of any object would be given by
a
θ = arctan , (6.1)
R

where a is the radius of the Earth’s orbit and R is the radius of curvature of the geometry.7
He found a minimum value of R ≥ 1.66 × 105 AU = 2.6 light-years. It is intriguing that this
estimate was made eight years before Bessel’s announcement of the first successful parallax
measurement of 61 Cygni. In making this estimate, Lobachevsky used the observational
upper limit of 1 arcsec for the parallax of bright stars. In a statement which will warm the
heart of observational astronomers, he remarked

There is no means other than astronomical observations for judging the exactness which attaches to
the calculations of ordinary geometry.

Non-Euclidean geometry was placed on a firm theoretical basis by Bernhard Riemann

(1826–1866) (Riemann, 1854), and the English-speaking world was introduced to these
ideas through the works of William Kingdon Clifford (1845–1879) and Arthur Cayley
(1821–1895). In 1900, Karl Schwarzschild returned to the problem of the geometry of space
and was able to set more stringent limits to its radius of curvature. Repeating Lobachevsky’s
argument, he found R ≥ 60 light-years if space were hyperbolic (Schwarzschild, 1900a).
If space were closed, he could set limits to the radius of curvature of the closed geometry
because the total volume of the closed space is V = 2π 2 R3 . Since there were only 100 stars
with measurable parallaxes and at least 108 for which no parallax could be measured, he
concluded that R ≥ 2500 light-years. He also noted that, if space were spherical, it should
be possible to observe an image of the Sun in the direction precisely 180◦ away from its
direction on the sky at any time.
Until Einstein’s discovery of the general theory of relativity, considerations of the geom-
etry of space and the role of gravity in defining the large-scale structure of the Universe
were separate questions. After 1915, they were inextricably linked.
102 6 The origins of astrophysical cosmology

6.2 General relativity and Einstein’s Universe

The history of the discovery of general relativity is admirably told by Abraham Pais in his
scientific biography of Albert Einstein (1878–1955), Subtle is the Lord . . . The Science and
Life of Albert Einstein,8 where many of the technical details of the papers published in the
period 1907 to 1915 are discussed. Equally recommendable is the survey by John Stachel of
the history of the discovery of both theories of relativity.9 In seeking a fully self-consistent
relativistic theory of gravity, Einstein was entering uncharted territory, and for many years
he ploughed a lone furrow, making the ultimate spectacular success of the theory in 1915
all the more remarkable.

6.2.1 Einstein’s route to general relativity

It is simplest to quote Einstein’s words from his Kyoto address of December 1922 (Einstein,
1922b).

In 1907, while I was writing a review of the consequences of special relativity, . . . I realised that all the
natural phenomena could be discussed in terms of special relativity except for the law of gravitation.
I felt a deep desire to understand the reason behind this . . . It was most unsatisfactory to me that,
although the relation between inertia and energy is so beautifully derived [in special relativity], there
is no relation between inertia and weight. I suspected that this relationship was inexplicable by means
of special relativity.

In the same lecture, he remarked

I was sitting in a chair in the patent office in Bern when all of a sudden a thought occurred to me: ‘If
a person falls freely he will not feel his own weight.’ I was startled. This simple thought made a deep
impression upon me. It impelled me towards a theory of gravitation.

In his comprehensive review of the special theory of relativity published in 1907, Ein-
stein devoted the whole of the last section, Section V, to ‘The principle of relativity and
gravitation’ (Einstein, 1907). In the very first paragraph, he raised the following question:

Is it conceivable that the principle of relativity also applies to systems that are accelerated relative to
one another?

He had no doubt about the answer and stated the principle of equivalence explicitly for the
first time:

. . . in the discussion that follows, we shall therefore assume the complete physical equivalence of a
gravitational field and a corresponding acceleration of the reference system.

From this postulate, he derived the time-dilation formula in a gravitational field:

 

dt = dτ 1 + 2 , (6.2)
c
where  is the gravitational potential, recalling that  is always negative, τ is proper
time and t is the time measured at zero potential. Then, applying Maxwell’s equations to
6.2 General relativity and Einstein’s Universe 103

the propagation of light in a gravitational potential, he found that the equations are form-
invariant, provided the speed of light varies as
 
(r )
c(r ) = c 1 + 2 , (6.3)
c
according to an observer at zero potential. Einstein realised that, as a result of Huyghens’
principle, or equivalently Fermat’s principle of least time, light rays are bent in a non-uniform
gravitational field. He was disappointed to find that the effect was too small to be detected
in any terrestrial experiment.
Einstein published nothing on gravity and relativity until 1911, although he was undoubt-
edly wrestling with these problems through the intervening period. In his paper of that year,
he reviewed his earlier ideas, but noted that the gravitational dependence of the speed of
light would result in the deflection of the light of background stars by the Sun (Einstein,
1911). Applying Huyghens’ principle to the propagation of light rays with a variable speed
of light, he found the standard ‘Newtonian’ result that the angular deflection of light by a
mass M would amount to
2G M
θ = , (6.4)
pc2
where p is the collision parameter of the light ray, which in this case is just the radius of
the Sun. The deflection amounts to 0.87 arcsec, although Einstein estimated 0.83 arcsec.
Einstein urged astronomers to attempt to measure this deflection. Intriguingly, equation
(6.4) had been derived by Johann von Soldner (1776–1833) in 1801 on the basis of the
Newtonian corpuscular theory of light (Soldner, 1804).10
Following the famous Solvay conference of 1911, Einstein returned to the problem of
incorporating gravity into the theory of relativity, and, from 1912 to 1915, his efforts were
principally devoted to formulating the relativistic theory of gravity. It was to prove to be a
titanic struggle. In summary, his thinking was guided by four ideas.
r The influence of gravity on light.
r The principle of equivalence.
r Riemannian space-time.
r The principle of covariance.
During 1912, Einstein realised that he needed more general space-time transformations
than those of special relatively. Two quotations illustrate the evolution of his thought. The
first is from Einstein (1912):
The simple physical interpretation of the space-time coordinates will have to be forfeited, and it cannot
yet be grasped what form the general space-time transformations could have.

The second is from Einstein (1922b):

If all accelerated systems are equivalent, then Euclidean geometry cannot hold in all of them.

Towards the end of 1912, Einstein realised that what was needed was non-Euclidean
geometry. From his student days, he vaguely remembered Gauss’s theory of surfaces, which
had been taught to him by Karl Friedrich Geiser (1843–1934). Einstein consulted his old
104 6 The origins of astrophysical cosmology

school friend, the mathematician Marcel Grossmann (1878–1936), about the most general
forms of transformation between frames of reference for metrics of the form

ds 2 = gμν dx μ dx ν . (6.5)

Although outside Grossmann’s field of expertise, he soon came back with the answer that
the most general transformation formulae were the Riemannian geometries, but that they
had the ‘bad feature’ that they are non-linear. Einstein instantly recognised that, on the
contrary, this was a great advantage, since any satisfactory theory of relativistic gravity
must be non-linear.
The collaboration between Einstein and Grossmann was crucial in elucidating the fea-
tures of Riemannian geometry, which were essential for the development of the general
theory of relativity, Einstein fully acknowledging the central role which Grossmann had
played. At the end of the introduction to his first monograph on general relativity, Einstein
wrote (Einstein, 1916a)

Finally, grateful thoughts go at this place to my friend the mathematician Grossmann, who by his
help not only saved me the study of the relevant mathematical literature but also supported me in the
search for the field equations of gravitation.

The Einstein–Grossmann paper of 1913 was the first exposition of the role of Riemannian
geometry in the search for a relativistic theory of gravity (Einstein and Grossmann, 1913a,b).
The details of Einstein’s struggles over the next three years are fully recounted by Pais. It was
a huge and exhausting intellectual endeavour which culminated in the presentation of the
theory in its full glory in November 1915 (Einstein, 1915, 1916b). In that month, Einstein
discovered that he could account precisely for the perihelion shift of the planet Mercury.
In 1859, Urbain Le Verrier (1811–1877) had discovered that, once account was taken of
the influence of the planets, there remained an unexplained component of the advance of
the perihelion of Mercury’s elliptical orbit about the Sun, amounting to about 40 arcsec per
century (Le Verrier, 1859). In a feat of extraordinary technical virtuosity, Einstein showed
in November 1915 that the advance of the perihelion of Mercury expected according to
the general theory of relativity amounted to 43 arcsec per century, a value in excellent
agreement with the present best estimates. He knew he must be right.
The theory also predicted the deflection of light by massive bodies because of the cur-
vature of space-time in their vicinity. For the Sun, the predicted deflection of light rays
from stars just grazing the limb of the Sun amounted to 1.75 arcsec. This deflection is a
factor of two greater than that expected according to a Newtonian calculation (see above
and endnote 10). This prediction resulted in the famous eclipse expeditions of 1919 led by
Arthur Eddington and Andrew Crommelin (1865–1939) (Dyson, Eddington and Davidson,
1920). The Astronomer Royal, Frank Dyson (1868–1939), had long realised that the eclipse
of 1919 would take place under the most advantageous conditions and had begun planning
accordingly. Not only was the totality of the eclipse unusually long, about six minutes, but
also the Sun would then be observed against the background of the Hyades star cluster,
providing many bright target stars for the deflection experiments.
The eclipse of 29 May 1919 passed over Northern Brazil, across the Atlantic Ocean
through the island of Principe and then across Africa. The British Government awarded
 6.2 General relativity and Einstein’s Universe 105

a grant of £1100 to enable two expeditions to be made to photograph the eclipse, one to
Sobral in Northern Brazil, led by Crommelin, and one to Principe, led by Eddington. The
results were in agreement with Einstein’s prediction, the Sobral result being 1.98 ± 0.16
arcsec and the Principe result 1.61 ± 0.4 arcsec.11 These results were widely publicised,
and Einstein’s reputation was established in the public mind as the epitome of scientific
genius.
The theory also predicted the gravitational redshift of light originating close to massive
compact objects. As already described in Section 4.2, Adams’ careful observations of the
spectrum of the white dwarf Sirius B in 1925 showed a gravitational redshift amounting to a
Doppler shift of 19 km s−1 , in precise agreement with the expectations of general relativity
(Adams, 1925a). Thus, by the mid 1920s, the theory had triumphantly passed the three tests
proposed by Einstein.12

6.2.2 Einstein’s Universe

In 1916, the year after the discovery of the general theory of relativity was announced,
Willem de Sitter (1872–1934) and Paul Ehrenfest (1880–1933) suggested in correspon-
dence that a spherical four-dimensional space-time would eliminate the problem of the
boundary conditions at infinity, which pose insuperable problems for Newtonian cosmo-
logical models.13 In 1917, Einstein published his famous paper in which he derived a static
closed model for the Universe which seemed to resolve the problems inherent in Newtonian
cosmological models (Einstein, 1917).
Einstein’s standard field equations can be written in the form
Rmn − 12 gmn R = −κ Tmn , (6.6)
where Rmn is the Ricci tensor, gmn is the metric tensor, Tmn is the energy-momentum tensor,
R is the contracted Ricci tensor and κ = 8π G/c2 . A strict relativist would adopt the point of
view that the concept of gravitational force is unnecessary and the full content of the theory
can only be appreciated in terms of the bending of the Riemannian geometry of space-
time throughout the Universe. For illustrative purposes, we can work with the Newtonian
analogue of equation (6.6) which becomes Poisson’s equation,
∇ f = −4π Gρ, (6.7)
where f is the body force per unit mass due to gravity. The equivalence with the general
relativistic formulation can be seen by writing f = −∇, where  is the (negative) gravi-
tational potential, so that ∇ 2  = 4π Gρ. The metric coefficients, gmn , thus play the role of
‘gravitational potentials’ in general relativity.
Einstein realised that, in general relativity, he had a theory which could be used to
construct models of the Universe as a whole. His motivation for taking this problem seriously
was his objective of incorporating what he designated Mach’s principle into the structure of
general relativity. By Mach’s principle, he meant that the local inertial frame of reference
should be determined by the frame of reference of the distant stars. There were two obstacles
to constructing self-consistent physical models. The first was that the static Newtonian model
was unstable, in the sense that, even in the case of an infinite distribution of stars, local regions
106 6 The origins of astrophysical cosmology

would collapse under gravity. The second problem concerned the boundary conditions at
infinity. Einstein proposed to solve all these problems at one fell swoop by introducing an
additional term into the field equations (6.6), the famous cosmological constant,14 λ. The
equations become
Rmn − 12 gmn R − λgmn = −κ Tmn . (6.8)
The corresponding modification to Poisson’s equation would be given by
∇ f = −4π Gρ + λ. (6.9)
Note the key point that the gravitational force depends upon the density of the medium
but the cosmological term is independent of density and is proportional to distance, f =
1
3
λr . Inspecting equation (6.9), it can be seen that a static solution exists with constant
gravitational potential , f = −∇ = 0, and
λ = 4π Gρ0 , (6.10)
where ρ0 is the density of the static Universe. Since λ is positive, the geometry of the Universe
is closed and the radius of curvature of the geometrical sections is R = c/(4π Gρ0 )1/2 . This
geometry eliminated the problem of the boundary conditions at infinity since this model
Universe is finite and closed. The volume of the spherical geometry is given by V = 2π 2 R3
and there is a finite number of galaxies in the Universe. Furthermore, Einstein believed he
had incorporated Mach’s principle into general relativity. The essence of the argument was
that static solutions of the field equations did not exist in the absence of matter. In other
words, according to equation (6.10), if λ = 0, ρ0 = 0 and so, in the absence of the stabilising
term λ, the only solution is a completely empty Universe. The cosmological constant was
essential in creating a static closed model of the Universe with finite density.
This was the first fully self-consistent cosmological model, but it had been achieved at
the cost of introducing the cosmological constant. This was to remain a thorn in the flesh
of cosmologists from the time of its introduction in 1917, until the last few years of the
twentieth century, when it came into its own. Einstein was somewhat uncomfortable about
its introduction, acknowledging that the term was ‘not justified by our actual knowledge of
gravitation’ but was merely ‘logically consistent’.
In 1919, Einstein realised that a term involving the cosmological constant would appear
in the field equations of general relativity, quite independent of its cosmological significance
(Einstein, 1919). In the derivation of the field equations, the λ term appears as a constant
of integration, which is normally set equal to zero in the development of standard general
relativity. The significance of the cosmological constant can be appreciated by inspection of
equation (6.9), which shows that, even if there is no matter present in the Universe, ρ = 0,
there is still a repulsive force acting on a test particle. As Yakov Borisovich Zeldovich (1914–
1987) remarked, the term corresponds to the ‘repulsive effect of a vacuum’ (Zeldovich,
1968). This type of force has no physical meaning according to classical physics. With the
development of quantum field theory, the concept of the vacuum changed dramatically, but
this is running far ahead of our story.
For most of the twentieth century, cosmologists adopted ambivalent views about the λ
term. In 1919, Einstein was not enthusiastic about the term, remarking that it ‘detracts from
 6.3 De Sitter, Friedman and Lemaı̂tre 107

the formal beauty of the theory’. Willem de Sitter had similar views (de Sitter, 1917b) and
wrote in 1919 that the term
detracts from the symmetry and elegance of Einstein’s original theory, one of whose chief attractions
was that it explained so much without introducing any new hypotheses or empirical constant.

Others regarded it as a constant which appears in the development of general relativity

and its value should be determined by astronomical observation. Throughout the twentieth
century, the cosmological constant made regular reappearances in the literature in response
to various cosmological problems, and these will be recounted in the course of this chapter
and in Part V. Suffice to say that none of these arguments withstood detailed scrutiny until,
in the last decade of the twentieth century, compelling evidence for a non-zero value of the
cosmological constant was found (see Chapters 13 and 15).

6.3 De Sitter, Friedman and Lemaı̂tre

In the same year that Einstein’s first paper on cosmology was published, de Sitter showed
that one of Einstein’s objectives had not been achieved (de Sitter, 1917a). He found solutions
of Einstein’s field equations in the absence of matter, ρ = p = 0, and derived the following
metric for isotropic world models with constant space curvature κ = R−2 :
r  r 
ds 2 = −dr 2 − R 2 sin2 (dψ 2 + sin2 ψ dθ 2 ) + cos2 c2 dt 2 . (6.11)
R R
Thus, although there is no matter present in the Universe, a test particle still has a perfectly
well defined geodesic along which it can travel. As de Sitter asked, ‘If no matter exists
apart from the test body, has this inertia?’ At that time, the principal issues at stake were
the origin of inertia and Mach’s principle, rather than any thought that these considerations
might be of relevance to astronomical observation.
It was soon discovered that this solution could be written in terms of an expanding
metric. In 1922, Cornelius Lanczos (1893–1974) showed that the de Sitter solution could
be written alternatively in the form of a metric in which the test particles move apart at an
exponentially increasing rate (Lanczos, 1922). To achieve this he separated the spatial and
time components of the metric (6.11), so that it became
ds 2 = −dt 2 + cosh2 t[dr 2 + cos2 r (dθ 2 + cos2 θ dφ 2 )]. (6.12)
Lanczos added the remark which has since become a platitude:
It is interesting to observe how one and the same geometry can appear with quite different physical
interpretations . . . according to the interpretation placed upon the particular coordinates.

At almost exactly the same time, Alexander Alexandrovich Friedman (1888–1925) pub-
lished the first of two classic papers for both static and expanding world models (Friedman,
1922, 1924). Friedman15 was a brilliant mathematician whose principal interests were the
application of fluid and gas dynamics to meteorology. In 1922, he was employed by the
Main Geophysical Observatory in Petrograd and led the theoretical division in studies of
the physics of the atmosphere. As a result of the Soviet Revolution of 1917, the Civil War and
108 6 The origins of astrophysical cosmology

the subsequent blockade of the Soviet Union, there was considerable delay before Soviet
scientists became aware of Einstein’s general relativity. Friedman was one of the first to
appreciate fully the significance of the theory and sent his book The World as Space and
Time to the publishers in 1922 (Friedman, 1923).
In the first of his classic papers published in 1922, Friedman wrote down explicitly the
general equations for the dynamics of homogeneous isotropic world models. These can be
written in an exactly equivalent form in terms of the scale factor R which, in the notation
I will use, is the function which describes how the distance between any two points in the
expanding Universe changes with time, R being normalised to the value unity at the present
epoch. For the case of a uniform isotropic model, the field equations reduce to the following
two equations:
 
4π G R 3p 1
R̈ = − ρ + 2 + λR; (6.13)
3 c 3
8π Gρ 2 c2 1
Ṙ 2 = R − 2 + λR 2 . (6.14)
3 R 3
In these equations, ρ is the mean density of the Universe, p is the pressure and R is the
radius of curvature of the geometry of space at the present epoch. Note that the pressure
term appears as a relativistic correction to the inertial mass density of the Universe.16 To
recover Einstein’s Universe, we set p = 0, R̈ = Ṙ = 0 and R = 1. We can find Lanczos’s
solution by setting p = ρ = 0 and then, in the limit of large times, the scale factor changes as
R(t) = R0 exp[(λ/3)1/2 t], the exact analogue of the cosh t time dependence of the coefficient
in front of the spatial part of the metric, equation (6.12).
Friedman explored the solutions of equations (6.13) and (6.14) for a variety of special
cases, including the closed-world model in which the Universe eventually collapses back
to a singular state. In 1925 Friedman died of typhoid in Leningrad before the fundamental
significance of his work was appreciated. The neglect of Friedman’s work in these early days
is somewhat surprising since his papers appeared in the authoritative journal Zeitschrift für
Physik. The problem may have been associated with a brief note by Einstein criticising a step
in Friedman’s paper of 1922 (Einstein, 1922a). Friedman wrote to Einstein pointing out that
the criticism was incorrect and Einstein immediately published a brief note accepting that
Friedman had not made an error (Einstein, 1923). It was not until Abbé Georges Lemaı̂tre
(1894–1966) independently discovered the same solutions in 1927, and then became aware
of Friedman’s contributions, that the pioneering nature of these papers was appreciated. The
standard world models of general relativity, with or without the cosmological constant, are
nowadays usually referred to as the Friedman world models.
The significance of Lemaı̂tre’s work was that he was seeking solutions of the field equa-
tions which avoided the problems which afflicted Einstein’s Universe, which was of finite
density, closed and static, and de Sitter’s Universe, which was open, empty and expanding
(Lemaı̂tre, 1927). In his independent discovery of the expanding solutions in his important
paper of 1927, he ended by remarking, ‘We still have to explain the cause of the expansion
of the Universe.’
One of the problems facing the pioneers of relativistic cosmology was the interpretation
of the space and time coordinates used in their calculations. De Sitter’s solution could be
 6.4 The recession of the nebulae 109

written in apparent stationary form or as an exponentially expanding solution. From the

metric, de Sitter had shown that a distance–redshift relation must exist for his empty-world
model, but it was not clear whether or not this was relevant to the observable Universe.
The answer came resoundingly in the affirmative with Hubble’s discovery of the velocity–
distance relation for galaxies in 1929, which ushered in a new epoch in astrophysical
cosmology.

6.4 The recession of the nebulae

In 1917, Vesto Slipher (1874–1969) published a paper in which he reported heroic spec-
troscopic observations of 25 spiral galaxies made with the Lowell Observatory’s 24-inch
telescope (Slipher, 1917). He realised that, for the spectroscopy of low-surface-brightness
objects such as the spiral nebulae, the crucial factor was the f -ratio, or speed, of the spec-
trograph camera, not the size of the telescope. Exposures of 20, 40 and even 80 hours were
made to secure these spectra. He found that the velocities of the galaxies inferred from the
Doppler shifts of their absorption lines were typically about 570 km s−1 , far in excess of the
velocity of any known Galactic object. Furthermore, most of the velocities corresponded to
the galaxies moving away from the Solar System, that is, the absorption lines were shifted
to longer (red) wavelengths. This phenomenon became known as the redshift, z, of the
galaxies, and it is defined by the relation
λo − λe
z= , (6.15)
λe
where λe is the emitted wavelength of some spectral feature and λo is its observed wave-
length.17 Slipher noted that
This might suggest that the spiral nebulae are scattering but their distribution on the sky is not in
accord with this since they are inclined to cluster.

In 1921, Carl Wilhelm Wirtz (1876–1939) searched for correlations between the veloci-
ties of the spiral galaxies and other observable properties (Wirtz, 1922) and concluded that,
when the data were averaged in a suitable way,
an approximate linear dependence of velocity upon apparent magnitude is visible. This dependence
is in the sense that the nearby nebulae tend to approach our galaxy whereas the distant ones move
away . . . The dependence of the magnitudes indicates that the spiral nebulae nearest to us have a
lower outward velocity than the distant ones.

By 1925, Hubble had established the extragalactic nature of the spiral nebulae and, as
discussed in Section 5.3, in his remarkable paper of 1926, he set out the basic types of
galaxies, their masses and the contribution they made to the mass density of the Universe.
He was fully aware of the importance of his estimate of the mean mass density of the
Universe since, according to Einstein’s Universe, once this was prescribed, all the other
properties of the Universe followed immediately.
By 1929, he had assembled data on the distances of galaxies for which velocities had
been measured (Hubble, 1929). It is interesting to note the methods used to estimate the
110 6 The origins of astrophysical cosmology

distances of the 24 galaxies. The distances of the nearest seven objects, all within 500 kpc,
were the best determined and used the Cepheid variable technique; the distances for the
next 13 objects were found by assuming that the most luminous stars in the galaxies had an
upper limit of absolute magnitude M = −6.3; the last four objects, believed to be members
of the Virgo cluster, had distances assigned on the basis of the mean luminosities of the
nebulae in the cluster. As Hubble acknowledged in The Realm of the Nebulae, most of the
velocities used in his 1929 paper were due to Slipher. In fact, of the 44 galaxy redshifts
known in 1925, 39 of them had been measured by Slipher. From these meagre data, Hubble’s
famous velocity–distance relation was derived (Figure 6.1(a)). It is intriguing that the main
objective of Hubble’s paper was not to derive the velocity–distance relation, but rather to
use the velocities of the galaxies to derive the velocity of the local standard of rest of the
Solar System relative to the extragalactic nebulae.
With hindsight, it is remarkable that Hubble found the redshift–distance relation from
such a nearby sample of galaxies, but there was other evidence even at that time. He noted in
his brief paper that Milton Humason (1891–1972) had measured a velocity of 3910 km s−1
for the galaxy NGC 7619, the brightest galaxy in a cluster. If the velocity–distance relation
were correct, the absolute magnitude of this galaxy would be of the same order as those of
the brightest galaxies in nearby clusters.
Although Hubble did not write down what is often referred to as the Hubble relation
v = H0r in this paper, he noted that ‘the velocity–distance relation may represent the de
Sitter effect’. De Sitter had shown that, according to his static metric, equation (6.11), there
would be a redshift of spectral lines which increases with distance. It had already been
appreciated, however, that the velocity–distance relation is a natural outcome of uniformly
expanding world models. Both Lemaı̂tre and Howard Robertson (1903–1961) were aware of
the fact that the Friedman solutions result locally in a velocity–distance relation (Lemaı̂tre,
1927; Robertson, 1928). Lemaı̂tre derived what he termed the ‘apparent Doppler effect’, in
which ‘the receding velocities of extragalactic nebulae are a cosmical effect of the expan-
sion of the Universe’ with v ∝ r . Robertson found a similar result stating that ‘we should
expect . . . a correlation v ≈ (cl/R)’, where l is distance and v the recession velocity. From
nearby galaxies, he found a value for what we now know as Hubble’s constant, H0 , of
500 km s−1 Mpc−1 .
The subsequent story is told in Hubble’s Silliman Lectures given at Yale University in
1935 and published as the famous and influential monograph The Realm of the Nebulae
in the following year (Hubble, 1936). The task of extending the measurement of the radial
velocities of galaxies to much greater distances was undertaken by Humason using the 100-
inch Hooker Telescope at Mount Wilson. By 1935, he had measured the velocities of almost
150 further galaxies out to distances inferred to be 35 times greater than the distance of the
Virgo cluster and to radial velocities of 42 000 km s−1 , roughly one-seventh of the speed of
light. Although distances could not be measured directly, Hubble and Humason found that
the luminosity functions of the galaxies in clusters are remarkably similar, and so they used
the fifth brightest member of each cluster as a measure of its relative distance (Hubble and
Humason, 1934). The resulting redshift–apparent magnitude relation is expected to follow
the relation log v = 0.2m + constant if the galaxies follow a velocity–distance relation,
6.4 The recession of the nebulae 111

(a)

(b)

Figure 6.1: (a) Hubble’s first version of the velocity–distance relation for nearby galaxies (Hubble,
1929). The filled circles and the full line represent a solution for the solar motion using the nebulae
individually; the open circles and the dashed line represent a solution combining the nebulae into
groups. The cross is an estimate of the mean distance of the other 20 galaxies for which radial
velocities were available. (b) The velocity–apparent magnitude relation for the fifth brightest member
of clusters of galaxies, corrected for galactic obscuration (Hubble and Humason, 1934). Each cluster
velocity is the mean of the various individual velocities observed in the cluster, the number being
indicated by the figure in brackets.

v ∝ r . The results of these arduous programmes of observation are shown in Figure 6.1(b)
and are in excellent agreement with a linear velocity–distance relation. Even today, the
apparent magnitude–redshift relation for the brightest galaxies in clusters remains among
the most convicting evidence for the extension of Hubble’s law to significant cosmological
distances.
112 6 The origins of astrophysical cosmology

n
tio

t io
bu

la
re
tr i
3.0

dis

ed
log N (square degree)

rv
rm

se
ifo

ob
un
2.5 1.00

m
tures
2.0 depar 0.50

0.20
18 19 20 21
m (limiting magnitude)
Figure 6.2: Hubble’s counts of faint galaxies published in 1936 (Hubble, 1936). The line labelled
‘Uniform Distribution’ corresponds to the relation log N (≤ m) = 0.6m + constant. The points repre-
sent the observed counts with a best-fitting line shown. At the bottom of the diagram is the difference
in magnitude between the uniform counts and the observations, which Hubble interpreted as the effect
of redshift.

In Hubble’s monograph, he described the number counts of faint galaxies made with
the 100-inch telescope, which, by 1935, extended to an apparent magnitude limit of 21,
the faintest counts feasible with the 100-inch telescope and the available photographic
plates. The counts followed the expected relation N (≤ m) = 0.6m + constant for a uni-
form distribution of galaxies down to 18th magnitude (see Section A5.1), but at fainter
magnitudes fewer galaxies were observed than were expected for a uniform distribution
(Figure 6.2). Hubble correctly concluded that the counts extended to such faint magni-
tudes, and consequently large distances, that the effects of redshift upon the number counts
had to be taken into account. He also correctly concluded that the counts were evidence
for the overall homogeneity of the Universe as far the 100-inch telescope could observe
distant galaxies. In the last pages of the book, he speculated that the convergence of the
number counts may be associated with the curvature of space. His conclusion that the
Universe must have positive curvature was incorrect, but this can be attributed to the fact
that it took some time before the proper relativistic formulation of the relations between
observables and the intrinsic properties of galaxies were worked out.18 His concluding
remarks strike a resonance with anyone who has attempted these difficult cosmological
observations:

Eventually, we reach the dim boundary – the utmost limits of our telescopes. There, we measure
shadows, and we search among the ghostly errors of measurement for landmarks that are scarcely
more substantial. The search will continue. Not until the empirical resources are exhausted, need we
pass on to the dreamy realms of speculation.
 6.5 The Robertson–Walker metric 113

6.5 The Robertson–Walker metric

The discovery of the velocity–distance relation for galaxies acted as a major stimulus to
the study of the Friedman models. Of prime importance was the need to place the world
models of de Sitter, Friedman and Lemaı̂tre on a firm theoretical foundation. There remained
confusion about the notions of time and distance to be used in the world models of general
relativity because the basis of that theory was that the field equations could be set up in
any frame of reference one pleased. The principle of special relativity meant that observers
located on galaxies moving relative to one another could not agree on the synchronisation
of their clocks. By 1935, the problem was solved independently by Robertson and Arthur
Walker (1909–2001) (Robertson, 1935; Walker, 1936).
A key concept was enunciated by Hermann Weyl (1885–1955) in 1923 and is known
as Weyl’s postulate (Weyl, 1923). To eliminate the arbitrariness present in the choice of
coordinate frame, Weyl introduced the idea that, to quote Hermann Bondi (1919–2005),

The particles of the substratum (representing the nebulae) lie in space-time on a bundle of geodesics
diverging from a point in the (finite or infinite) past.19

The most important aspect of this statement is the postulate that the geodesics which
represent the world-lines of galaxies do not intersect, except at a singular point in the
past. Note that this postulate predates Hubble’s discovery of the velocity–distance relation,
but follows the pioneering works of Lanczos and Friedman. By the term ‘substratum’
Bondi meant an imaginary medium which can be thought of as a fluid which defines the
kinematics of the system of galaxies. The consequence of Weyl’s postulate is that there is
only one geodesic passing through each point of space-time, except at the origin. Once this
postulate is adopted, it is possible to assign a notional observer to each world-line, known
as a fundamental observer, and to define a time coordinate, known as cosmic time. Each
fundamental observer can be provided with a standard clock which measures proper time
along the geodesic, or world-line, of that observer. The clocks can be synchronised at the
time when the geodesics were all together at the singular point at the origin, so that cosmic
time is defined to be the proper time measured by such a fundamental observer.
One further assumption is essential before the framework of the standard models can be
derived. This is known as the cosmological principle and is the statement that our Galaxy is
not located in a privileged or special position in the Universe. In other words, a fundamental
observer located on any other galaxy at the same epoch would observe the same large-scale
features of the Universe that we observe. The implication is that our Galaxy is located at a
typical point in the Universe. We therefore have to decide what large-scale features should
be common to all fundamental observers.
One requirement is that all observers should observe the same velocity–distance rela-
tion at the same cosmic epoch. A second requirement is that, on large enough scales, the
Universe should present the same appearance in all directions and that, on average, matter
and radiation should be homogeneously distributed. Hubble’s galaxy counts were important
empirical evidence that the distribution of galaxies is isotropic and homogeneous, since
they followed closely the relation log N = 0.6m + (constant) expected of a uniform distri-
bution (see Section A5.1). Although there is clustering in the distribution of galaxies on a
114 6 The origins of astrophysical cosmology

local scale, it appeared that, on a large enough scale, these irregularities average out and
the galaxies are isotropically distributed. It was a simple calculation to show that, in any
uniformly expanding homogeneous fluid, an observer located on any particle of the fluid
observes the same velocity–distance relation for the relative motion of other particles of the
fluid.
Putting these concepts together, Robertson and Walker independently showed that the
metric for any isotropically expanding substratum had to have the form
 
R 2 (t) dr 2
ds 2 = dt 2 − + r 2
(dθ 2
+ sin2
θ dφ 2
) (6.16)
c2 (1 + κr 2 )

(Robertson, 1935; Walker, 1936). There are a number of important features of this metric,
which is appropriately known as the Robertson–Walker metric.
r The first is that it can be derived solely from Weyl’s postulate, the cosmological principle,
the special theory of relativity and the assumptions of isotropy and homogeneity and that
the Universe expands isotropically. It contains all the permissible isotropic geometries
consistent with the assumptions of isotropy and homogeneity and these are described
by the curvature κ = R−2 , where R is the radius of curvature of the spatial sections
of the isotropic curved space. If κ is positive, the geometry is spherical; if κ is zero,
the geometry is flat; if κ is negative the geometry is hyperbolic – no other geometries
are allowed. The point of importance is that this metric is of very general form and is
correct whatever forces act upon the substratum. In particular, it does not depend upon
the dynamics of the Universe being described by general relativity.
r The physics of the expansion has been absorbed into the scale factor, R(t), which,
in the form of equation (6.16), takes the value unity at the present epoch. An impor-
tant result, which can be derived directly from this metric, is the relation between the
scale factor, R, and redshift, z: R = (1 + z)−1 . This formula elucidates the signifi-
cance of redshift in cosmology – it provides directly the value of the scale factor of
the Universe when the radiation was emitted and is independent of the cosmological
model.
r Care has to be taken over the definition of the radial coordinate, r . In the above
form, r is defined to be the radial comoving distance coordinate. Once this metric
of isotropic expanding space-time was derived, it was a straightforward task to use
it to derive relations between the intrinsic properties of objects and their observed
properties.20
It was some time before all the subtleties of the analysis were fully appreciated.

6.6 Milne–McCrea and Einstein–de Sitter

The most important solutions for the variation of the scale factor, R(t), with cosmic time are
those derived from general relativity, including those in which the cosmological constant is
non-zero. The solutions for R(t) can be written down formally as the integral of Friedman’s
6.6 Milne–McCrea and Einstein–de Sitter 115

second equation (6.14):

 R dR
t= . (6.17)
8π Gρ 2 c2
R − 2 + 3 λR
1 2
0 3 R
Note that it is assumed that the pressure, p, can be neglected. There is no simple general
closed solution for this equation, and it has been the subject of a great deal of study.
One of the most important contributions to understanding the physical content of the
solutions was provided by Milne and McCrea in 1934 (Milne and McCrea, 1934a,b). They
showed that, despite the fact that Newtonian mechanics cannot provide a fully self-consistent
cosmological model, simple ideas from Newtonian physics can provide insight into the
solutions derived from equation (6.14). They realised that the requirements of isotropy and
homogeneity are very powerful constraints upon the properties of the models. In the simplest
form of their argument, it is supposed that our Galaxy is located at the centre of a uniformly
expanding sphere. This is precisely what we, and any other fundamental observer anywhere
in the Universe, must observe. Suppose we work out the deceleration of a galaxy of mass
m at distance x from our Galaxy. Applying Gauss’s theorem for gravity without bothering
about the boundary conditions at infinity, we can find the deceleration of the galaxy as
follows:
d2 x 4π x 3 Gmρ
m 2
=− . (6.18)
dt 3 x2
The mass of the galaxy, m, cancels out on either side of the equation – the dynamics refer to
the sphere as a whole. Now, we replace the x and ρ by their values at some reference epoch,
t0 , by writing x = R(t)r , where r is the comoving radial distance coordinate, which is a
label attached to a galaxy for all time, and ρ = ρ0 R −3 (t). It is convenient to set R(t) = 1
at the present epoch t0 , and then
4π G R 4π Gρ0
R̈ = − ρ=− . (6.19)
3 3R 2
The first integral of this equation is given by
8π Gρ 2 8π Gρ0
Ṙ 2 = R −C = − C, (6.20)
3 3R
where C is a constant of integration. We have derived formulae of exactly the same forms as
are found in the full theory. Not unexpectedly, the Newtonian argument cannot cope with the
curvature of space, and a relativistic expression for the inertial mass density is needed which
includes the pressure of the gas. It turns out that the constant in equation (6.20) involves
the curvature of space at the present epoch, C = c2 κ = c2 /R2 . It is straightforward to add
terms representing the cosmological constant, λ, to equations (6.19) and (6.20) to recover
equations (6.13) and (6.14).
The reason this argument works is that, because of the postulates of isotropy and homo-
geneity, local physics is also cosmic physics and is applicable on the large scale. Every
fundamental observer would perform exactly the same calculation and obtain the same
answer. The analysis of Milne and McCrea was of considerable importance because it
showed that, despite the problems with the boundary conditions at infinity, the Newtonian
116 6 The origins of astrophysical cosmology

model can be used successfully on large scales in the Universe, and, in particular, on scales
less than the horizon scale, r ≈ ct, where t is cosmic time, it is perfectly adequate to use
Newtonian arguments.
With the discovery of the expansion of the system of galaxies and Hubble’s law, Einstein
regretted the inclusion of the cosmological constant into the field equations. According to
Gamow, Einstein stated that the introduction of the cosmological constant ‘was the greatest
blunder of my life’ (Gamow, 1970). In 1932, Einstein and de Sitter demonstrated that one
particularly simple solution of the field equations for an expanding Universe seemed to
be in good accord with observations (Einstein and de Sitter, 1932). They noted that, if the
cosmological constant is set equal to zero, there is a special solution of the equations in which
the spatial curvature is zero, κ = 0 and R → ∞, corresponding to Euclidean space sections.
This model is often referred to as the Einstein–de Sitter model, and it has particularly simple
dynamics, R(t) = (t/t0 )2/3 , where t0 = 2/3H0 and H0 is Hubble’s constant, the constant
of proportionality in the velocity–distance relation. The model has average density at the
present epoch, ρ0 = 3H02 /8π G. This density is often referred to as the critical density, and
the Einstein–de Sitter model is often referred to as the critical model, because it separates the
ever-expanding models with open, hyperbolic geometries from the models which eventually
collapse to a singularity and which have closed, spherical geometry (Figure 6.3). When
Einstein and de Sitter inserted the value of Hubble’s constant from Hubble’s observations,
H0 = 500 km s−1 Mpc−1 , into the expression for the critical density, they found a mean
value for the density of the Universe of 4 × 10−25 kg m−3 . Although recognising that this
value was somewhat greater than the value derived by Hubble, they argued that the density
was of the correct order of magnitude and, in any case, there might well be considerable
amounts of what would now be called ‘dark matter’ present in galaxies and in the Universe
at large.
Evidence of dark matter in the Universe was not long in coming. Astronomers of the
1930s had two ways of measuring the masses of individual galaxies. The best procedure
was to measure the rotation curves of the discs of spiral galaxies, meaning the variation
of the rotational speed of the disc as a function of distance from its centre. Since the
rotational velocities can be assumed to be in centripetal equilibrium about the centre of the
galaxy, the distribution of mass and total mass can be found by equating the centripetal
acceleration to the gravitational force needed to maintain that circular velocity. The second,
less satisfactory, method was to assume that there is a constant ratio between the masses of
galaxies and their luminosities so that, once the mass-to-luminosity ratio of one galaxy, or
one class of galaxies, has been determined, the masses of others can be found from their
luminosities.
In 1933, Zwicky, working at the Mount Wilson Observatory, made the first studies of
rich clusters of galaxies, in particular of the Coma cluster, which is one of the largest regular
clusters in the northern sky (Zwicky, 1933, 1937). The method Zwicky used to estimate the
total mass of the cluster had been derived by Eddington in 1916 to estimate the masses of
star clusters. Using methods familiar in the theory of gases, Eddington derived the virial
theorem,21 which relates the total internal kinetic energy, T , of the stars or galaxies in a
cluster to the total gravitational potential energy, |U |, assuming the system is in a state of
6.6 Milne–McCrea and Einstein–de Sitter 117

Ω =0
4 critical model
Ω =1
3

2 Ω =2

1
Big Crunch
Big
Bang 0 2 4 6 8
cosmic time in units of H 0t
Figure 6.3: Examples of the dynamics of the standard Friedman world models with λ = 0. The scale
factor, R(t), has been normalised to unity at the present epoch. The critical model, with density
parameter = 1, separates the re-collapsing models with > 1 from those which expand forever
and which have ≥ 1. In this presentation, the trajectories of the world models have been scaled to
the same value of Hubble’s constant at the present epoch, R = 1.

statistical equilibrium under gravity (Eddington, 1916a). The kinetic energy can be written
T = 12 M v 2 , where v 2 is the mean-square velocity of the stars, or galaxies, and |U | =
G M 2 /2Rcl , where Rcl is some suitably defined radius that depends upon the distribution of
mass in the cluster. For a cluster of stars or galaxies in statistical equilibrium, Eddington
showed that T = 12 |U |. Therefore, if the cluster is known to be in statistical equilibrium,
the total mass of the cluster can be found from the virial theorem, M ≈ 2Rcl v 2 /G.
In rich regular clusters of galaxies, such as the Coma cluster, there is convincing evidence
from the radial distribution of galaxies within the cluster that they have reached statistical
equilibrium and so good estimates can be made of its total mass. Zwicky measured the
velocity dispersion of the galaxies in the Coma cluster and found that there was much more
mass in the cluster than could be attributed to the visible masses of galaxies. In solar units
of M /L , the ratio of mass-to-optical luminosity of a galaxy such as our own is about
3, whereas for the Coma cluster the ratio was found to be about 500. In other words, there
must be about 100 times more dark, or hidden, matter as compared with visible matter in
the cluster.
It was some time before Zwicky’s results were accepted by the astronomical community,22
but they have been confirmed by all subsequent studies of rich clusters of galaxies. The
nature of the dark matter remains an open and crucial question for physics and cosmology.
It is now generally agreed that much of the mass in the Universe is in some form of dark
matter, but its nature is unknown.
118 6 The origins of astrophysical cosmology

6.7 Eddington–Lemaı̂tre

Despite Einstein’s renunciation of the cosmological constant, this was very far from the end
of the story, because there remained one grave problem for world models in which the cos-
mological constant is set equal to zero. It is a simple calculation to show that, if λ = 0, the age
of the Universe must be less than H0−1 . Using Hubble’s estimate of H0 = 500 km s−1 Mpc−1 ,
the age of the Universe had to be less than 2 × 109 years old, a figure in conflict with the
age of the Earth derived from the ratios of abundances of long-lived radioactive species.
The present best estimate for the age of the Earth is about 4.6 × 109 years.
Eddington and Lemaı̂tre (Eddington, 1930; Lemaı̂tre, 1931b) immediately recognised
that this problem could be eliminated if the cosmological constant is positive. The effect
of a positive cosmological constant is to counteract the attractive force of gravity when the
Universe grows to a large enough size. There are special solutions of the integral in equation
(6.17) which correspond to the Einstein stationary Universe, but not necessarily at the present
epoch. It was possible to find models which had remained in the static Einstein state for an
arbitrarily long period in the past and which then began to expand away from that state under
the influence of the cosmological term. In this type of Eddington–Lemaı̂tre model, the age
of the Universe could be arbitrarily long. As Eddington expressed it, the Universe would
have a ‘logarithmic eternity’ to fall back on,23 and so resolve the conflict between estimates
of Hubble’s constant and the age of the Earth. The dynamics of the Eddington–Lemaı̂tre
model, and a closely related Lemaı̂tre model, are illustrated in Figure 6.4. In the Lemaı̂tre
model, the Universe does not quite attain a static state, but undergoes a long ‘coasting phase’
when its velocity of expansion is small and the total age of the model can be much greater
than H0−1 .

6.8 The cosmological problem in 1939

Thus, by the end of the 1930s, the basic problems of what I call classical cosmology had
been clearly identified. The solution of the cosmological problem lay in the determination
of the parameters which defined the Friedman world models. This was the goal of the great
programmes of observation to be carried out by the 200-inch telescope and the subse-
quent generation of 4-metre class telescopes. The challenge was to measure the following
parameters which characterise the Universe:
r Hubble’s constant, H0 = Ṙ/R at the present epoch, is a measure of the present rate of
expansion of the Universe;
r the deceleration parameter, q0 = − R̈/ Ṙ 2 at the present epoch, describes the present
deceleration of the Universe, noting that if q0 is negative, the Universe is accelerating;
r the curvature of space κ = R−2 ;
0
r the mean density of matter in the Universe, ρ, in particular the question of whether or
not it attains the critical density, ρcrit ;
r the age of the Universe, T0 , as given by the integral in equation (6.17);
r the cosmological constant, λ.
6.8 The cosmological problem in 1939 119

λ > λc C

Lemaître’s model

R (t)

λ = λc C

(a) Einstein model

(c) Eddington–Lemaître model

R (t)

(c)
(a)

(b)
t

Figure 6.4: Examples of the dynamics of world models with λ = 0. (After H. Bondi, Cosmology
(Cambridge: Cambridge University Press, 1960), p. 84.) The Einstein static model is illustrated by the
model for which R(t) is constant for all time. The Eddington–Lemaı̂tre model, in which the Universe
expanded from the Einstein static universe in the infinite past, is illustrated, as is a Lemaı̂tre model,
in which the value of λ is slightly different from that of the static model. In model (b) the Universe
approaches the stationary Einstein model from a singular origin at t = 0.

These are not independent if the Friedman models are a correct description of the large-
scale dynamics of the Universe. For example, for the models which include the cosmological
constant,

( − 1) + 13 (λ/H02 ) 1 λ
κ = R−2 = , q0 = − , (6.21)
(c/H0 )2 2 3 H02

where = ρ/ρcrit is known as the density parameter. Note that, if λ = 0, there is a simple
one-to-one relation between the geometry of the world models, their densities and dynamics,
q0 = /2 and κ = R−2 = ( − 1)/(c/H0 )2 .
The determination of these parameters has turned out to be among the most difficult
observational challenges in the whole of astronomy, and progress was much slower than the
optimists of the 1930s must have hoped. In compensation, completely new vistas were to
120 6 The origins of astrophysical cosmology

open up after the Second World War as the whole of the electromagnetic spectrum became
available for astronomical observation. The precise determination of these parameters only
became possible in the very last years of the twentieth century by quite different approaches
from those envisaged by the pioneers of the 1930s.

Notes to Chapter 6

1 The history of cosmology and the development of relativistic cosmologies is described in the
monographs by J. D. North, The Measure of the Universe (Oxford: Clarendon Press, 1965), H.
Bondi, Cosmology, 2nd edn (Cambridge: Cambridge University Press, 1960) and E. Harrison,
Cosmology: The Science of the Universe (Cambridge: Cambridge University Press, 2001).
2 A delightful discussion of this correspondence is given by E. R. Harrison, Darkness at Night: A
Riddle of the Universe (Cambridge: Cambridge University Press, 1987), Chapter 6.
3 This remark appears in the second edition of Newton’s Principia Mathematica as a result of the
exchange of letters with Richard Bentley.
4 An excellent history of the discovery of non-Euclidean geometry is provided by R. Bonola in
his book Non-Euclidean Geometry, and the Theory of Parallels by Nikolas Lobachevsky, with a
Supplement Containing the Science of Absolute Space by John Bolyai, trans. H. S. Carslaw (New
York: Dover, 1955).
5 See North, The Measure of the Universe, pp. 74–75.
6 It is often stated that Gauss carried out the measurements himself, but this is probably a myth.
In his biography of Gauss, Bühler writes as follows (W. K. Bühler, Gauss: A Biographical Study
(Berlin: Springer-Verlag, 1981), p. 100): ‘The often-told story according to which Gauss wanted to
decide the question [of whether space is non-Euclidean] by measuring a particularly large triangle
is, as far as we know, a myth. The great triangle Hohenhagen–Inselberg–Brocken was a useful
control for the smaller triangles which it contains. Gauss was certainly aware of the fact that the
error of measurement was well within the possible deviation from 180 degrees from which, under
strict conditions, one could have derived the non-Eucludean nature of space.’
7 I have given a simple introduction to isotropic curved spaces in Chapter 17 of Malcolm Longair,
Theoretical Concepts in Physics (Cambridge: Cambridge University Press, 2003).
8 A. Pais, Subtle is the Lord . . . The Science and Life of Albert Einstein (Oxford: Clarendon Press,
1982).
9 J. Stachel, History of relativity, in Twentieth Century Physics, vol. 1, eds L. M. Brown, A. Pais
and A. B. Pippard (Bristol: Institute of Physics Publishing and New York: American Institute of
Physics Press, 1995), pp. 249–356.
10 In fact, Henry Cavendish (1731–1810) appears to have come to essentially the same result as
von Soldner in an unpublished manuscript of about 1784, inspired by John Michell’s paper of the
previous year on the escape of light from a massive body (Michell, 1784). Clifford Will provides
an intriguing comparison of how Cavendish may have derived his result with that of von Soldner
in C. Will, Henry Cavendish, Johann von Soldner, and the deflection of light, American Journal
of Physics, 56, 1988, 413–415.
11 An account of the trials and tribulations associated with these famous expeditions and the analysis
of the results is presented by P. Coles, Einstein, Eddington and the 1919 eclipse, in Historical
Development of Modern Cosmology, ASP Conference Series 252, eds V. J. Martı́nez, V. Trimble
and M. J. Pons-Bordeı́a (San Francisco: ASP, 2001), pp. 21–41.
12 A fourth test of general relativity was proposed by Irwin Shapiro (b. 1929) in 1964 (Shapiro,
1964). There is a small time delay for electromagnetic waves which pass by the limb of the Sun
at grazing incidence; this is because, according to an observer at infinity, the speed of light is
variable in a changing gravitational field. The time delay amounts to only 200 μs for the round trip
Notes to Chapter 6 121

to Mars and back, but very precise time delays within the gravitational field of the Sun have now
been measured and are in agreement with the predictions of standard general relativity within the
accuracy of the measurements, which is within 0.2% of the expected delay.
13 See the discussion in North, The Measure of the Universe, p. 80.
14 In Einstein’s original paper of 1917, the cosmological constant is written as a lower-case λ. It is
common practice to write the term as upper-case Greek  nowadays.
15 See E. A. Tropp, V. Ya. Frenkel and A. D. Chernin, Alexander A. Friedman: The Man Who Made
the Universe Expand (Cambridge: Cambridge University Press, 1993). Note that I prefer to spell
Friedman’s name with one ‘n’ since he was Russian, not German.
16 I have given a detailed discussion of these equations and their solutions in Malcolm Longair,
Galaxy Formation (Berlin: Springer Verlag, 1998).
17 An unfortunate convention adopted by the early cosmologists was to convert the redshift, which
is a splendid dimensionless number, into a velocity by multiplying by the speed of light. As
discussed in Section 6.5, in all isotropic, homogeneous world models, the cosmological redshift
is directly related to the scale factor, R, at which the radiation was emitted through the relation
R = 1/(1 + z).
18 The correct results for any cosmological model are given in Longair, Galaxy Formation, Section
17.2.
19 See H. Bondi, Cosmology, p. 100.
20 See, for example, Longair, Galaxy Formation, Section 5.5 or Longair, Theoretical Concepts in
Physics, Section 19.4.
21 I have given a simple proof of this result for clusters of stars and galaxies in Section 3.4 of Longair,
Galaxy Formation.
22 Zwicky was a remarkable and original astronomer who ploughed his own furrow and, in the
process, rubbed up many distinguished astronomers the wrong way. A classic example of his
vitriolic style can be found in the Preface to F. Zwicky, Catalogue of Selected Compact Galaxies
and Post-eruptive Galaxies (Guemlingen, Switzerland: F. Zwicky, 1968).
23 See North, The Measure of the Universe, p. 125.
Part III
The opening up of the
electromagnetic spectrum
7 The opening up of the
electromagnetic spectrum and
the new astronomies

7.1 Introduction

Until 1945, astronomy meant optical astronomy. The commissioning of the Palomar 200-
inch telescope in 1949 highlighted the dominance of the USA in observational astrophysics
in the period immediately after the Second World War. The need for greater light-gathering
power to detect faint galaxies for cosmological studies led to George Ellery Hale’s concept
of the 200-inch telescope (Hale, 1928). Hale symbolised the entrepreneurial approach of
US astronomers to the sponsorship of private US observatories, such as the Lick, Harvard,
Yerkes and Mount Wilson Observatories, which began in the late nineteenth century. James
Lick (1796–1876), for example, was a successful maker and seller of pianos and an enthusiast
for astronomy who, on his death in 1876, left a bequest of $700 000 to build ‘a powerful
telescope, superior to and more powerful than any telescope ever yet made . . . and also
a suitable observatory connected therewith’. The observatory was constructed on Mount
Hamilton and officially opened in 1888 with the completion of the 36-inch telescope, under
which James Lick was buried, according to the terms of his bequest.
Hale’s record of observatory and telescope construction is remarkable by any measure.1
He persuaded Charles T. Yerkes (1837–1905), the entrepreneur who built and electrified the
Chicago street-train system and who was regularly on the verge of legal embarrassment,
to provide the funds to build and equip the Yerkes Observatory as part of the University
of Chicago. At Mount Wilson, Hale first constructed the 60-inch telescope and then the
Hooker 100-inch telescope, completed in 1917, with funds provided by John D. Hooker and
the Carnegie Institution of Washington. At the time, each of these was the largest telescope
of its type in the world. Hale’s greatest achievement was, however, the construction of
the 200-inch telescope on a good, dark site on Mount Palomar in Southern California.
Technologically, the 200-inch telescope was a masterpiece of engineering which stretched
mirror and telescope technology to the limit.2
The programme of construction was delayed by the Second World War, but the 200-inch
telescope was finally completed in 1948, ten years after Hale’s death. Like most of the
other major observatories in the USA, the 200-inch telescope was a private telescope which
was used more or less exclusively by the astronomers employed at the host institutions,
the Astrophysics Department of the California Insitute of Technology, the Mount Wilson

125
126 7 The new astronomies

Observatory and the Carnegie Institute of Washington. Thus, in the early 1950s, the most
important telescope in the world for all types of astrophysical and cosmological research was
in the hands of a relatively small group of privileged astronomers. There is no question that
the 200-inch telescope dominated observational astrophysics and cosmology from the time
it was commissioned until the 1980s, when a new generation of 4-metre class telescopes on
better sites provided astronomers with superior observing facilities.
The astronomical scene was, however, about to change with the development of new
ways of tackling astrophysical and cosmological problems. There were several reasons for
this major change in outlook of observational and theoretical astrophysicists since 1945.

(i) The expansion of the wavebands for astronomical observation

The most important reason has been the expansion of the wavebands that have become
available for astronomical observation. A plot of the temperature of a black body against the
frequency (or wavelength) at which most of the radiation is emitted is shown in Figure 7.1(a).
In the lower panel, Figure 7.1(b), the transparency of the atmosphere to radiation as a function
of frequency is presented, showing how high a telescope must be placed above the surface
of the Earth for the atmosphere to become transparent to radiation of different wavelengths.
With the development of radio astronomy and the capability of placing telescopes for
different wavebands in space, the expansion of the accessible electromagnetic spectrum has
led to a vast increase in the range of temperatures which are accessible for astronomical
study. In turn, this had led to a much more complete description of our physical Universe and
to the discovery of new physical phenomena which are important for fundamental physics
as well as astronomy.
Figure 7.1(a) shows that observations in the optical waveband correspond to studying
the Universe in a rather narrow wavelength interval, 300–800 nm, and hence to black-
body temperatures in the range 3000–10 000 K. Of course, a somewhat wider range of
temperatures can be studied since bodies at temperatures outside this range emit some
radiation in the optical waveband, but this is a fair representation of the temperatures of
most of the objects observed at optical wavelengths, for example stars, hot gas clouds and
their associations into galaxies, clusters and so on. The capability of making observations
from above the Earth’s atmosphere opened up the far-infrared, ultraviolet, X-ray and γ -
ray wavebands, so that very much hotter and cooler objects could be studied. It comes
as no surprise that, as observational capabilities in these new wavebands developed, new
and unexpected phenomena were discovered which added important new dimensions to
astrophysical and cosmological research.

(ii) Non-electromagnetic astronomy

Equally important has been the development of non-electromagnetic means of tackling
astrophysical and cosmological problems. The oldest of these is the study of cosmic rays, the
high-energy electrons, protons and nuclei accelerated in a variety of astrophysical environ-
ments, including the Sun, supernovae and active galaxies. In addition, different approaches
to observational astronomy have been developed. Neutrino astronomy has already made
 log (wavelength/m)
2 0 −2 −4 −6 −8 −10 −12 −14
12

109 K
9

millimetre

optical
radio infrared ultraviolet X-ray γ -ray
log (temperature/K)

104 K

3K
0

−3
0 8 10 12 14 16 18 20 22 24
log (frequency/Hz) (a)
log (photon energy/eV)
−6 −4 −2 0 2 4 6 8 10
−9
150
140
−8 130
120
110
log (fraction of atmosphere)

−7
100
−6 90

−5 80

altitude, km
70
−4
60
50
−3
40
−2 30
20
−1
10
0 0
6 8 10 12 14 16 18 20 22 24
log (frequency/Hz)
(b)
Figure 7.1: (a) The relation between the temperature of a black body and the frequency (or wavelength)
at which most of the energy is emitted. The frequency (or wavelength) plotted is that corresponding
to the maximum of a black body at temperature T . Convenient expressions for this relation are
νmax = 1011 (T /K) Hz or λmax T = 3 × 106 nm K. The ranges of wavelength corresponding to the
different wavebands – radio, millimetre, infrared, optical, ultraviolet, X-rays and γ -rays – are shown.
(b) The transparency of the atmosphere for radiation of different wavelengths. The solid line shows the
height above sea-level at which the atmosphere becomes transparent for different wavelengths. After
R. Giacconi, H. Gursky and L. P. van Speybroeck, Annual Review of Astronomy and Astrophysics,
6, 1968, 373. Both diagrams are from M. S. Longair, The new astrophysics, in The New Physics, ed.
P. C. W. Davies (Cambridge: Cambridge University Press, 1988), p. 94.
128 7 The new astronomies

spectacular contributions to astrophysics and fundamental physics. In addition, there are

a number of emerging astronomies which are bound to have fundamental significance for
astrophysics. Gravitational-wave astronomy is fully expected to become a major tool for
the high-energy astrophysicist, while astroparticle physics, the search for stable massive
particles predicted by theories of elementary particles, has a key role to play in understanding
the dark matter problem as well as in fundamental physics.

(iii) Technological advance and computation

None of these developments would have been possible without remarkable technologi-
cal developments in the design and construction of telescopes, instruments and detectors
for all wavebands. In optical astronomy, the photographic plate has been largely replaced
by highly efficient digital detectors. In many of the new astronomies, technologies were
imported from non-astronomical disciplines and modified for the special needs of astro-
nomical observation.
As in all the sciences, the semiconductor and computer revolutions have been cru-
cial to the advance of observation, data collection and analysis, interpretation and theory.
Astronomers were among the very first to capitalise upon the possibilities opened up by
fast digital computers. For example, the rapid growth of the radio technique of aperture
synthesis was wholly dependent upon the availability of the first computers to be made
generally available to the scientific community. The role of computation in astrophysics
and cosmology has completely changed many aspects of research in these disciplines. An
important consequence has been that theory and observation can now be compared with
a precision which would have been quite inconceivable to the pioneers of the pre-War
years.

(iv) The growth of the astronomical community

There has been a huge increase in the volume of activity in the astronomical sciences.3 At
least part of the growth has been associated with an influx of physicists whose research
interests and expertise led them to consider astrophysical problems. By the same pro-
cess of symbiosis between the astrophysical and laboratory sciences, which is a recur-
ring theme throughout this history, astronomy has assimilated new tools from physics
and theoretical physics, most obviously in general relativity and particle physics, but also
from fields such as chemistry, solid state physics, plasma physics, superconductivity and
biophysics.
Some measure of this increase in activity is provided by the membership of the Interna-
tional Astronomical Union, which is open to all professional astronomers and which was
founded in 1919. At the first General Assembly held in Rome in 1922, there were just over
200 members from 19 adhering countries. By 1938, the numbers had risen to 550 from 26
countries. The number was roughly the same immediately after the Second World War. By
the time of the 2003 General Assembly held in Sydney, Australia, the membership had risen
to 9100 from 67 adhering countries.
7.1 Introduction 129

(v) Astronomy as ‘big science’

Astronomy, astrophysics and cosmology have become one of the ‘big sciences’. The case
has already been made that research in astrophysics has always been ‘big science’. At the
beginning of the modern era, huge resources were needed by Tycho Brahe to advance the
study of the motions of the Sun, Moon and planets and, in the nineteenth century, Pickering’s
vast projects, albeit with small telescopes, needed very considerable resources to place the
science of astrophysics on a secure foundation. The 200-inch telescope broke all records
for the cost of an individual instrument in the 1930s and 1940s.
After the Second World War, there was a spectacular increase in investment in basic
research in the USA, largely stimulated by the huge contributions which the very best
research scientists had made during the period of hostilities and the realisation of the
enormous potential for economic growth, as well as strategic defence requirements, which
the fruits of basic research could bring. In Europe, it took somewhat longer to recover from
the ravages of the War, but, in due course, these countries began to invest heavily in pure
and applied research. The attitude of many of the best research workers had been changed
by their wartime experiences. To quote Bernard Lovell (b. 1913), they adopted an approach
to research that was (Lovell, 1987)

. . . utterly different from that deriving from the pre-war environment. The involvement with massive
operations had conditioned them to think and behave in ways which would have shocked the pre-war
university administrators. All these facts were critical in the large-scale development of astronomy.

In due course, the astronomers rode the post-war wave of investment in the fundamental
sciences, but these initiatives had to be seen in a national or international context rather than
as sponsorship by private institutions, as had occurred in the USA. Whilst the increase in
the numbers of astronomers alone made the construction of more large telescopes a priority,
the major discoveries of radio, X-ray and γ -ray astronomy, as well as the rise of high-energy
astrophysics in the 1960s and 1970s, had a considerable impact upon the case for increasing
investment in large astronomical facilities. Examples of the culmination of this historical
progression were the NASA–ESA Hubble Space Telescope, which in the end cost in excess
of $2 billion, and the European Southern Observatory’s Very Large Telescope, consisting
of four 8-metre optical–infrared telescopes on Cerro Paranel in Chile, which cost roughly
$500 million. Thus, the telescopes needed to carry out frontier research have become very
complex and costly, international collaboration often being essential to construct and operate
them.

In Part III, we trace the development of astronomy from the point of view of the open-
ing up of new regions of the electromagnetic spectrum and the changing ways in which
astronomical research is carried out. At the start of this period, astronomical research was
primarily carried out by small groups of astronomers working with their own dedicated
telescopes. By the end of the period, most of the large telescopes for all wavebands were
national or international facilities, operated by specialist teams for the benefit of the commu-
nities of astronomers. Most of the world-leading telescopes are now very high-technology
instruments of great sophistication, and astronomy is more than ever one of the ‘big
130 7 The new astronomies

sciences’. The science, however, is still small science in the sense that the big facilities
provide the data needed by a very wide range of different astronomical disciplines and by a
large population of astronomers. Typically, each large facility will provide data for hundreds
of different astronomical projects, ranging from our own Solar System to the earliest phases
of the Big Bang.

7.2 The discovery of subatomic particles and cosmic rays

The first hints that there was more to the Universe than just gas, dust and stars came from
the discovery of cosmic rays, the history of which is intimately related to the understanding
of subatomic particles. Let us first review some of the key discoveries which were to lead
to the new astronomies in the post Second World War period.

7.2.1 The discovery of α-, β- and γ -rays and the neutron

In 1895, Wilhelm Röntgen (1845–1923) discovered, by accident, that wrapped unexposed
photographic plates left close to Crookes discharge tubes were darkened. In addition, flu-
orescent materials left close to them glowed in the dark. Röntgen came to the correct
conclusion that both phenomena were associated with some new form of radiation emitted
by the Crookes tube, and he named these X-rays (Röntgen, 1895).4 His discovery caused
an immediate sensation when the first X-ray photographs showing the bones of the body
were published. Overnight, X-rays became a matter of the greatest public interest and were
very rapidly incorporated into the armoury of the doctor’s surgery. The X-rays were more
penetrating than cathode rays, since they could blacken photographic plates at a consider-
able distance from the hot spot on the Crookes tube, which was known to be their source.
Their identification with ‘ultra-ultraviolet’ radiation was only convincingly demonstrated in
1906, when Charles Barkla (1877–1944) found that the X-radiation was polarised (Barkla,
1906) and, even more convincingly, when Max von Laue (1879–1960) had the inspiration
of looking for their diffraction by crystals in 1912 (Friedrich, Knipping and Laue, 1912;
Laue, 1912), in the process opening up the new field of X-ray crystallography.
The association of X-rays with fluorescent materials led to the search for other sources
of X-radiation. In 1896, Henri Becquerel (1852–1908) tested several known fluorescent
substances before investigating some samples of potassium uranyl disulphate. The photo-
graphic plates were wrapped in several sheets of black paper, the phosphorescent material
was exposed to sunlight and then the plate was developed to find if it had been darkened
by X-rays. Becquerel’s remarkable discovery was that the plates became darkened even
when the phosphorescent material was not exposed to light. This was the discovery of
natural radioactivity (Becquerel, 1896). In further experiments carried out in the same
year, Becquerel showed that the amount of radioactivity was proportional to the amount of
uranium in the substance and that the radioactive flux of radiation was constant in time.
Another important discovery was that the radiation from the uranium compounds discharged
electroscopes.
Other radioactive substances were soon identified. Thorium was discovered in 1898
(Schmidt, 1898) and then followed the isolation of polonium and radium, both much
7.2 Discovery of subatomic particles and cosmic rays 131

stronger sources of radioactivity than uranium, by Pierre Curie (1859–1906) and Marie
Sklodowska-Curie (1867–1934) (Curie and Sklodowska-Curie, 1898; Curie, Sklodowska-
Curie and Bémont, 1898). In his first publication on radioactivity, Rutherford established
that there are at least two separate types of radiation emitted by radioactive substances
(Rutherford, 1899). He called the component which is most easily absorbed α-radiation (or
α-rays) and the much more penetrating component β-radiation (or β-rays). It took another
ten years before Rutherford conclusively demonstrated that the α-radiation consisted of
what we now know as the nuclei of helium atoms (Rutherford and Royds, 1909). In con-
trast, β-rays were convincingly shown by Walter Kaufmann (1871–1947) to have the same
mass-to-charge ratio as the recently discovered electron (Kaufmann, 1902). Subsequently,
γ -radiation was discovered in 1900 by Paul Villard (1860–1934) as an extremely penetrating
form of radiation emitted in radioactive decays (Villard, 1900a,b). The γ -rays were con-
clusively identified as electromagnetic waves 14 years later when Rutherford and Edward
Andrade (1887–1971) observed the reflection of γ -rays from crystal surfaces (Rutherford
and Andrade, 1913).
The α-, β- and γ -rays were the only known radiations which could cause the ionisation
of air. The characteristic properties which distinguished them were their penetrating power.
In quantitative terms, they had the following properties.
r The α-particles ejected in radioactive decays produce a dense stream of ions and are
stopped in air within about 0.05 m. This is called the range of the particles.
r The β-particles have greater ranges, but there is not a well-defined value for any
particular radioactive decay. We now understand that the spread in range is due to the
fact that the electrons are emitted as part of a three-body process involving the emission
of a neutrino as well as an electron.
r The γ -rays were found to have by far the longest ranges, a few centimetres of lead
being necessary to reduce their intensity by a factor of 10.
The unravelling of the nature of the atomic nucleus continued throughout the period
1911–1930. It was soon established that typical nuclei have mass about two or more times
that which can be attributed to the protons alone. The commonly held explanation for this
difference was that the nucleus was composed of electrons and protons, the ‘inner’ electrons
neutralising the extra protons. The fact that certain nuclei ejected electrons in radioactive
β-decays supported this point of view. Rutherford had speculated in 1920 that the neutral
mass in the nucleus might be in the form of some new type of particle, similar to the
proton but with no electric charge (Rutherford, 1920). During the 1920s Rutherford and his
colleagues, particularly James Chadwick, made a number of unsuccessful attempts to find
evidence for these particles, which became known as neutrons.
In 1930, Walther Bothe (1891–1957) and Herbert Becker (b. 1906) in Germany and,
in 1932, Irene Joliot-Curie (1897–1956) and her husband Frederic Joliot (1900–1958) in
France discovered that neutral penetrating radiation was emitted when light elements were
bombarded by α-particles. Both groups believed that the radiation was some form of γ -
radiation. Chadwick guessed that the penetrating radiation was a flux of the elusive neutrons.
He rapidly performed a classic series of experiments in which the neutral radiation collided
with different substances, including hydrogen and nitrogen, and then, from the recoil effects
of the collisions between the unseen particles and the ambient gas, he could estimate the mass
132 7 The new astronomies

of the particles. This measurement of the mass of the neutral radiation showed conclusively
that it could not be γ -radiation but rather neutral particles ejected from the nucleus with
mass roughly the same as that of the proton (Chadwick, 1932). This was the discovery of
the neutron.

7.2.2 The discovery of cosmic rays

The cosmic-ray story begins in about 1900, when it was discovered that electroscopes dis-
charged even if they were kept in the dark well away from sources of natural radioactivity.5
The electroscope was a key instrument in many of the early experiments in radioactiv-
ity because the rate at which the leaves of the electroscope came together provided a
measure of the amount of ionisation. The origin of this behaviour was a major puzzle,
and various ingenious experiments were carried out to discover the origin of the ionising
radiation. A good example is this quotation from C. T. R. (Charles) Wilson (1869–1959)
(Wilson, 1901):

The experiments with this apparatus were carried out at Peebles. The mean rate of leak when the
apparatus was in an ordinary room amounted to 6.6 divisions of the micrometer scale per hour. An
experiment made in the Caledonian Railway tunnel near Peebles (at night after the traffic had ceased)
gave a leakage of 7.0 divisions per hour . . . There is thus no evidence of any falling off of the rate of
production of ions in the vessel, although there were many feet of solid rock overhead.

Later, Rutherford showed that most of the ionisation was due to natural radioactivity,
either in rocks or from radioactive contamination of the equipment. The big breakthrough
came in 1912 and 1913 when first Victor Hess (1883–1964) and then Werner Kolhörster
(1887–1946) made manned balloon ascents in which they measured the ionisation of the
atmosphere with increasing altitude (Hess, 1912; Kolhörster, 1913). By late 1912, Hess had
flown to 5 km and then, by 1913, Kolhörster had made ascents to 9 km, all these dangerous
experiments being carried out in open balloons. It was Hess who discovered the first definite
evidence that the source of the ionising radiation was extraterrestrial.
Hess and Kolhörster found the startling result that the average ionisation increased with
respect to the ionisation at sea-level above about 1.5 km (see Table 7.1). This was clear evi-
dence that the source of the ionising radiation must be located above the Earth’s atmosphere.
From the data in Table 7.1, the attenuation constant, α, defined by n(l) = n 0 exp(−αl), was
found to have values of 10−3 m−1 or less. The ionising radiation was much more more pen-
etrating than the γ -rays found in radioactive decays. Hess made the immediate inference:

The results of the present observations seem to be most readily explained by the assumption that a
radiation of very high penetrating power enters our atmosphere from above, and still produces in the
lower layers a part of the ionisation observed in closed vessels.

Even at sea-level, there is residual ionisation due to the extraterrestrial ionising radiation,
amounting to about 1.4 × 106 ion pairs m−3 .
It was not too much of an extrapolation to assume that the cosmic radiation, or cosmic
rays as they were named in 1925 by Robert Millikan (1868–1953), were γ -rays with greater
penetrating power than those observed in natural radioactivity. In 1929, Dmitri Skobeltsyn
7.2 Discovery of subatomic particles and cosmic rays 133

Table 7.1. The variation of ionisation with altitude from the

observations of Kolhörster (1913)

Altitude Difference between observed ionisation and

(km) that at sea-level (×106 ions m−3 )

0 0
1 −1.5
2 +1.2
3 +4.2
4 +8.8
5 +16.9
6 +28.7
7 +44.2
8 +61.3
9 +80.4

(1892–1992), working in his father’s laboratory in Leningrad, constructed a cloud chamber

to study the properties of the β-rays emitted in radioactive decays. The experiment involved
placing the chamber within the jaws of a strong magnet so that the curvature of their tracks
could be measured. Among the tracks, he noted some which were hardly deflected at all and
which looked like electrons with energies greater than 15 MeV (Figure 7.2). He identified
them with secondary electrons produced by the ‘Hess ultra γ -radiation’. These were the
first pictures of the tracks of cosmic rays (Skobeltsyn, 1929).
The year 1928 saw the invention of the Geiger–Müller detector by Hans Geiger
(1882–1945) and Walther Müller (1905–1979) which enabled individual cosmic rays to
be detected and their arrival times determined very precisely (Geiger and Müller, 1928,
1929). In 1929, Bothe and Kolhörster carried out one of the key experiments in cosmic
ray physics and introduced the important concept of coincidence counting to eliminate
spurious background events (Bothe and Kohlhörster, 1929). This coincidence technique
is now standard practice in many different types of cosmic ray, X-ray and γ -ray experi-
ment. By using two counters, one placed above the other, they found that simultaneous
discharges of the two detectors occurred very frequently, even when a strong absorber
was placed between the detectors, indicating that charged particles of sufficient pene-
trating power to pass through both of them were very common. In the crucial experi-
ment, they placed slabs of lead and then gold up to 4 cm thick between the counters
and measured the decrease in the number of coincidences when the absorber was intro-
duced. The mass absorption coefficient agreed very closely with that of the atmospheric
attenuation of the cosmic radiation. The experiment strongly suggested that the cosmic
radiation consists of charged particles. As they wrote in their classic paper (Bothe and
Kohlhörster, 1929):
One can perhaps summarise the whole discussion in a single argument: the mean free path of a
γ -ray between two electron ejecting processes would be 1/μ = 10 m in water 1/μ = 0.9 m in lead
and 1/μ = 0.52 m in gold for the high latitude radiation. Hence one can see that a quite exceptional
accident must be supposed to happen if two electrons produced by the same γ -ray should display the
necessary penetrating power and the correct direction to strike both counters directly.
134 7 The new astronomies

Figure 7.2: An image of the first photographic record of the arrival of a cosmic-ray particle by
Skobeltsyn in 1929. The track of the particle in the cloud chamber is indicated by the two white
arrows and one black arrow. From Y. Sekido and H. Elliot, eds, Early History of Cosmic Ray Studies
(Dordrecht: D. Reidel Publishing Company, 1985), p. 47.

They also showed that the flux of these particles could account for the observed intensity
of cosmic rays at sea-level and, because of their long ranges in matter, the energies of the
particles had to be about 109 −1010 eV.
The experiments carried out using cloud chambers showed that showers of cosmic ray
particles are often observed. Most of the cosmic-ray particles observed at the surface of the
Earth are, in fact, secondary, tertiary or higher products of very-high-energy cosmic rays
entering the top of the atmosphere. The full extent of some of these extensive air showers
was established by Pierre Auger (1899–1993) and his colleagues from observations with
a number of separated detectors (Auger et al., 1939). To their surprise, they found that
the air showers could extend over areas greater than 100 metres on the ground and consist
of the arrival of millions of ionising particles. The particles responsible for initiating the
showers must have had energies exceeding 1015 eV at the top of the atmosphere. This
was direct evidence for the acceleration of charged particles to extremely high energies in
extraterrestrial sources.

7.2.3 Cosmic rays and the discovery of elementary particles

From the 1930s to the early 1950s, the cosmic radiation provided a natural source of very-
high-energy particles which were energetic enough to penetrate into the nucleus. This
procedure was the principal technique by which new particles were discovered until the
7.2 Discovery of subatomic particles and cosmic rays 135

early 1950s. In 1930, Millikan and Anderson used an electromagnet ten times stronger
than that used by Skobeltsyn to study the tracks of particles passing through the cloud
chamber. Anderson observed curved tracks identical to those of electrons but corresponding
to particles with positive electric charge (Anderson, 1932). This discovery was confirmed by
Patrick Blackett and Giuseppe Occhialini (1907–1993) in 1933 using an improved technique
in which the cloud chamber was only triggered after it was certain that a cosmic ray had
passed through (Blackett and Occhialini, 1933). They obtained many excellent photographs
of the positive electrons, on many occasions showers containing equal numbers of positive
and negative electrons created by cosmic-ray interactions with the body of the apparatus
being observed.
The discovery of the positive electron, or positron, coincided almost exactly with Paul
Dirac’s theory of the electron (Dirac, 1928a,b). In one of the great theoretical extensions
of quantum mechanics, Dirac (1902–1984) succeeded in deriving the relativistic wave
equation for the electron which not only predicted its spin and magnetic moment, but also
the existence of what we would now call the antiparticle to the electron, the positron.6
There were more surprises in store, however. Anderson noted that there were often much
more penetrating positive and negative particle tracks in the cloud chamber pictures. These
particles displayed little evidence of interaction with the gas in the chamber. By 1936,
Anderson and Seth Neddermeyer (1907–1988) were sufficiently confident of their results
to announce the discovery of particles with mass intermediate between that of the electron
and the proton (Anderson and Neddermeyer, 1936). These mesotrons had mass between
about 50 and 400 times the mass of the electron. This discovery coincided rather nicely with
a theoretical prediction by Hideki Yukawa (1907–1981) concerning the strong force which
binds neutrons and protons together in the nucleus. According to Yukawa’s theory, the strong
short-range force could be understood in terms of the exchange of particles about 250 times
as massive as the electron (Yukawa, 1935). In fact, the particles discovered by Anderson and
Neddermeyer, nowadays known as muons, are not the particles which bind nuclei together.
The identification was somewhat unsatisfactory because the mesotrons showed so little
interaction with the nuclei in the chamber, whereas the exchange particle is expected to
show a strong interaction with nuclei.
The same procedures were used immediately after the Second World War by George
Rochester (1908–2001) and Clifford Butler (1922–1999), who constructed a new cloud
chamber to use with a large electromagnet obtained by Blackett before the War. In 1947
they reported the discovery of two cases of particle tracks in the form of ‘V’s with apparently
no incoming particle (Rochester and Butler, 1947). They correctly suggested that the Vs
resulted from the spontaneous decay of an unknown particle, the mass of which could be
estimated from the decay products. Both had mass about half that of the proton. To obtain
higher fluxes of cosmic radiation, the experiments were repeated at much higher altitudes.
Two years later, the experiments were carried out by Blackett’s group working at the Pic du
Midi Observatory in the Pyrenees and by Anderson and his colleagues on White Mountain
in California. Many more examples of Vs were found, and this class of particle became
known as strange particles. Both neutral and charged strange particles were discovered.
Most of them had mass about half that of the proton and are what are now referred to as
charged and neutral kaons (K+ , K− , K0 ). There were a few examples, however, of neutral
particles with mass greater than the mass of the protons – these are now known as lambda
136 7 The new astronomies

particles (). What puzzled physicists was their long lifetimes, 10−8 and 10−10 s, many
orders of magnitude greater than the timescale associated with the strong interactions.
Meanwhile, another powerful tool for the study of particle collisions and interactions
had been developed by Cecil Powell (1903–1969) at Bristol University. Photographic plates
had played a key role in the discovery of X-rays and radioactivity in the 1890s. Powell, in
collaboration with the Ilford photographic company, developed special ‘nuclear’ emulsions
which were sufficiently sensitive to register the tracks of protons, electrons and all the other
types of charged particle which had been discovered. Powell and his colleagues mastered the
techniques of producing thick layers of emulsion by stacking layer upon layer of emulsion,
resulting in a three-dimensional picture of the interactions taking place in the emulsion.
Among the first discoveries using this high-precision technique was that of the pion (π ) in
1947, which was the particle predicted by Yukawa in 1936 (Lattes, Occhialini and Powell,
1947).
By 1953, accelerator technology had developed to the point where energies comparable
with those available in cosmic rays could be produced in the laboratory with known energies
and directed precisely onto the chosen target. After about 1953, the future of high-energy
physics lay in the accelerator laboratory rather than in the use of cosmic rays. The interest
in cosmic rays shifted to the problems of their origin and their propagation in astrophysical
environments from their sources to the Earth.

7.3 Radio astronomy

The expansion of the observable electromagnetic spectrum began with Karl Jansky’s
announcement of the discovery of the radio emission from the Galaxy in May 1933.7
Working at the Bell Telephone Laboratories at Holmdel, New Jersey, Jansky (1905–1950)
was assigned the task of identifying naturally occurring sources of radio noise which would
interfere with radio transmissions. In what turned out to be a classic series of observations
made at the long wavelength of 14.6 metres (20.5 MHz), he discovered the radio emission
from the Galaxy (Figure 7.3) (Jansky, 1933).
This discovery was confirmed by Grote Reber (1911–2002), a radio engineer and enthu-
siastic amateur astronomer. With his home-built radio antenna and receiving system oper-
ating at a wavelength of 1.87 metres (160 MHz), he made a radio scan along the plane
of the Galaxy which was published in the Astrophysical Journal in 1940 (Reber, 1940).
Comparison of Jansky’s and Reber’s observations showed that the emission could not be
black-body radiation, and Reber proposed that it was bremsstrahlung, or free–free emission.
In the immediately following paper in the Astrophysical Journal, Louis Henyey (1910–1970)
and Philip Keenan (1908–2000) showed that, whilst the radiation at 1.87 m might be the
bremsstrahlung of gas at 10 000 K, the intensity observed by Jansky at the longer wavelength
was far too great for this to be the emission process (Henyey and Keenan, 1940). Other than
this negative conclusion, these observations attracted little attention from professional
astronomers. The culmination of Reber’s work was the publication of the first map of the
radio emission from the Galaxy (Figure 7.4) in the Astrophysical Journal in 1944 (Reber,
1944).
 7.3 Radio astronomy 137

Figure 7.3: Karl Jansky’s radio antenna with which he discovered the radio emission of the Galaxy
in 1933. (Courtesy of the US National Radio Astronomy Observatory.)

Figure 7.4: Reber’s map of the radio emission of the Galaxy, made at a radio frequency of 160 MHz
(1.87 m). The contours of radio emission are plotted in celestial coordinates and are more or less
coincident with the Milky Way (Reber, 1944).

7.3.1 The first discrete radio sources

The development of radar during the Second World War had two immediate consequences
for radio astronomy. Firstly, sources of radio interference which might confuse radar location
had to be identified. In 1942, James Hey (1909–2000) and his colleagues at the Army
Operational Research Group in the UK discovered intense radio emission from the Sun
which coincided with a period of unusually high sun-spot activity (Hey, 1946).8 Towards
138 7 The new astronomies

the end of the War, Hey and his colleagues continued to improve the sensitivities of the
receivers in order to detect incoming V2 rockets. To their consternation, they discovered
that the noise performance of the telescope system did not improve. They soon realised
that the background radio emission from the Galaxy itself was the factor which limited the
sensitivity of the telescope system, and not the receivers. At the end of hostilities, Hey and
his colleagues began mapping the sky at 5 m wavelength, and in 1946 they discovered the
first discrete source of radio emission, which lay in the constellation of Cygnus – the source
became known as Cygnus A (Hey, Parsons and Phillips, 1946). The second consequence was
that the extraordinary research efforts to design powerful radio transmitters and sensitive
receivers for radar resulted in new technologies which were to be exploited by the pioneers
of the new science of radio astronomy, all of whom came from a background in radar.
Immediately after the War, a number of the radar scientists began the systematic study of
the astronomical phenomena discovered, more or less by chance, as a result of the War effort.
The three main groups were headed by Martin Ryle (1918–1984) at Cambridge University,
by Bernard Lovell at Manchester University in the UK and by Joseph Pawsey (1908–1962) at
Sydney. Further discrete sources of radio emission were discovered, and radio interferometry
provided the best means of measuring their positions with improved precision. In 1948, Ryle
and Francis Graham Smith (b. 1923) discovered the most powerful source in the northern
hemisphere, Cassiopeia A (Ryle and Graham Smith, 1948), and in 1949 the Australian
radio astronomers John Bolton (1922–1993), Gordon Stanley (1921–2001) and Bruce Slee
(b. 1924) succeeded in associating three of the discrete radio sources with remarkable nearby
astronomical objects. One was associated with the supernova remnant known as the Crab
Nebula and the others, Centaurus A and Virgo A, were associated with the strange galaxies
NGC 5128 and M87, respectively (Bolton, Stanley and Slee, 1949). In addition to the
diffuse radio emission of our own Galaxy, these early surveys established the existence of a
population of discrete radio sources, some concentrated towards the plane of the Galaxy, but
many lying outside it. There was some uncertainty as to whether the isotropic component of
the source population was primarily associated with nearby radio stars in our own Galaxy
or with distant extragalactic objects.9
The radio astronomers could not answer this question from the radio data alone, since
the radio spectra were found to be continuous, without any spectral features from which a
redshift could be estimated. Distances could only be determined by finding an associated
optical object and measuring its distance. In 1951, Graham Smith measured interferometri-
cally the positions of the two brightest sources in the northern sky, Cygnus A and Cassiopeia
A, with an accuracy of about 1 arcmin (Graham Smith, 1951). This led to their optical iden-
tification by Walter Baade and Rudolph Minkowski (1895–1976) from observations with
the Palomar 200-inch telescope (Baade and Minkowski, 1954). Cassiopeia A was associ-
ated with a young supernova remnant in our own Galaxy, while Cygnus A was associated
with a faint and distant galaxy. The latter observation immediately showed that the radio
sources could be used for cosmological studies. By 1960, another of the brightest radio
sources in the sky, 3C 295, had been associated with the brightest galaxy in a cluster of
galaxies at the largest redshift, z = λ/λ = 0.461, measured for any galaxy at that time
(Minkowski, 1960b). This remained the largest redshift for any galaxy until the mid 1970s.
The cosmological importance of radio astronomical observations of discrete sources was
 7.3 Radio astronomy 139

thus apparent by the mid 1950s – fainter radio sources would lie at greater cosmological
distances and hence probe the Universe at epochs much earlier than the present.

7.3.2 Synchrotron radiation

The nature of the Galactic radio emission and, by analogy, of the discrete radio sources, was
solved in the late 1940s. During the 1930s and 1940s, particle accelerators of increasing
size, such as cyclotrons and betatrons, were constructed in which protons or electrons
moved in circular paths in a uniform magnetic field. It was realised that radiation losses
associated with the centripetal acceleration of electrons in their circular orbits would become
important as their energies increased. Dmitri Ivanenko (1904–1994) and Isaak Pomeranchuk
(1913–1966) published their calculations in 1944 showing that these losses would limit the
maximum energy of a betatron to about 500 MeV (Ivanenko and Pomeranchuk, 1944). The
energy loss rate of an accelerated relativistic electron had been worked out by George Schott
(1868–1937) in 1912 (Schott, 1912), a modern form of the loss rate formula being
 
dE e4 γ 2 B 2 v 2 sin2 θ
− = , (7.1)
dt 6π 0 c3 m 2e
where γ = E/mc2 = (1 − v 2 /c2 )−1/2 is the Lorentz factor of the electron, B is the magnetic
flux density and θ is the pitch angle of the electron, the angle between its direction of
motion and the magnetic field direction. This energy loss was first observed in the 100 MeV
betatron at the General Electric Research Laboratory in Schenectady, New York, by John
Blewett (1910–2000) in 1946, but the radiation of the electrons themselves was not observed
(Blewett, 1946). It was thought at that time that most of the radiation would be emitted at
low harmonics of the electrons’ orbital frequency and so radio receivers sensitive in the 50
to 1000 MHz waveband were used. In the meantime, Julian Schwinger (1918–1994) had
worked out in great detail the expected radiation spectrum of highly relativistic electrons
orbiting in a uniform magnetic field and had shown that, because of the extreme effects
of aberration, the radiation is most intense at a very much higher frequency, ν ≈ γ 2 νg ,
where νg = eB/2πm e is the non-relativistic gyrofrequency of the electron (Schwinger,
1946, 1949). The next accelerator built at the General Electric Laboratory was a 70 MeV
synchrotron accelerator with a transparent glass vacuum tube. Intense optical radiation
from the synchrotron accelerator was first seen in April 1947 and was named synchrotron
radiation (Elder et al., 1947). The characteristic properties of the radiation are that its
spectrum is a broad-band continuum and that it is highly polarised and directional.
The first application of synchrotron radiation in an astronomical context was proposed
by Hannes Alfvèn (1908–1995) and Nicolai Herlofson (1916–2004), who, in 1950, sug-
gested that the emission of the ‘radio stars’, which had just been discovered, might be
the synchrotron radiation of high-energy electrons gyrating in magnetic fields with flux
density 10−10 –10−9 T within a ‘trapping volume’ of about 0.1 light-year radius about
the star (Alfvén and Herlofson, 1950). Then, Karl-Otto Kiepenheuer (1910–1975) and
Vitali Ginzburg (b. 1916) made the much better suggestion that the Galactic radio emis-
sion observed by Jansky and Reber is the synchrotron radiation of ultrarelativistic electrons
gyrating in the interstellar magnetic field (Kiepenheuer, 1950; Ginzburg, 1951). By the
140 7 The new astronomies

mid 1950s, the power-law spectrum of the Galactic radio emission and its high degree of
polarisation convinced everyone of the correctness of the synchrotron hypothesis, a story
which is taken up in Section 11.1. Radio emission is observed throughout the disc of the
Galaxy and so provides direct evidence for an interstellar flux of very-high-energy electrons
being present throughout the disc of the Galaxy. At that time, it was likely that the cosmic
ray protons and nuclei, which had been detected in balloon flights at high altitude, were
of interstellar origin, but the flux of high-energy electrons was difficult to distangle from
secondary electrons created in the upper atmosphere.
The power-law spectra of the discrete radio sources and the polarisation of their radio
emission were naturally interpreted as evidence for synchrotron radiation of ultra-relativistic
electrons, but the energy requirements of some of the most luminous radio emitters, such as
Cygnus A, were enormous. Furthermore, the radio emission did not originate from the body
of the galaxies but from vast radio lobes, very often of much greater dimension than the
galaxies themselves. Somehow, the radio galaxies were capable not only of accelerating huge
fluxes of electrons to ultra-relativistic energies but also of ejecting them into intergalactic
space. These observations and their interpretation in terms of the properties of relativistic
plasmas and magnetic fields provided a powerful stimulus for the new discipline of high-
energy astrophysics.

7.3.3 Radio telescopes, aperture synthesis and VLBI

From the early 1950s onwards, radio astronomy developed as a discipline in its own right.
Large steerable reflectors were constructed, the radio analogues of the large optical reflec-
tors. The culmination of these efforts were the construction of the Jodrell Bank 76-metre
telescope in the UK (1957), now named the Lovell Telescope, the Parkes 64-metre telescope
in Australia (1961), the NRAO 300-foot telescope at Greenbank (1962)10 and the Effels-
berg 100-metre telescope in Germany (1971). These telescopes had excellent sensitivity,
but the angular resolution, θ, was limited by the diameter, D, of the telescope, according to
the Rayleigh criterion θ ≈ λ/D.
Higher angular resolution could be obtained by radio interferometry, which had been
developed as part of the radar development programmes during the Second World War and
which led to the concept of aperture synthesis. As Peter Scheuer has remarked,11

By the beginning of 1954 the principles of aperture synthesis were fully understood all over the world,
but the world of radio astronomy was then very small, and the world I mean, in which radio astronomy
was controlled by radio engineers who were learning astronomy, was smaller still. In the Netherlands
and in the United States radio astronomy was in the hands of real astronomers, to whom a telescope
meant a paraboloidal mirror and nothing else; their contribution was of a different kind. So the little
world that understood aperture synthesis consisted of CSIRO Radiophysics Division in Sydney, the
English radio astronomers at Cambridge and Manchester, and the French Group at Nançay.

Although the underlying principles of aperture synthesis at radio wavelengths had been
clearly set out by Ronald Bracewell (b. 1921) and James Roberts (b. 1927) in 1954 (Bracewell
and Roberts, 1954), there were many technical problems to be overcome before these
concepts could be converted into a reality. The important realisation was that, by measuring
7.3 Radio astronomy 141

58° 37

40° 38'
58° 25

58° 33
40° 36'

58° 31

40° 34'

19h 57m 50s 19h 57m 45s 19h 57m 40s (1950-0) 23h 21m 30s 23h 21m 10s 23h 20m 50s (1950-0)

(a) (b)
Figure 7.5: The first radio maps of (a) Cygnus A and (b) the supernova remnant Cassiopeia A, as
observed by the Cambridge One-Mile Telescope (Ryle et al., 1965). The observations were made at
a frequency of 1.4 GHz, at which the angular resolution of the telescope was 23 arcsec.

both the amplitudes and phases of the incoming signals, the distribution of radio brightness
across celestial sources could be completely reconstructed. The technical problems were
solved by Martin Ryle and his colleagues at Cambridge. There were two major hurdles to be
overcome. The first arose from the need to add together coherently the signals from separated
telescopes. As the separate telescopes follow the same patch of sky, varying delays had to
be switched into the cables from each antenna to compensate for the varying electrical
path-length from the radio source to the correlator. The second was the need for high-
speed computation. The essence of aperture synthesis is that the correlated signals between
different pairs of telescopes sample the Fourier transform of the radio brightness distribution
on the sky and so, to reconstruct the image, a two-dimensional inverse Fourier transformation
of the correlated signals had to be made. The availability of the first high-speed digital
computers by the end of the 1950s made this computational challenge a feasible undertaking.
The principles of Earth-rotation aperture synthesis were demonstrated by the remarkable
image of the region about the north celestial pole created by Ryle and Ann Neville (Ryle
and Neville, 1962). The success of this programme led to the construction of the Cambridge
One-Mile Telescope, the first Earth-rotation aperture synthesis telescope system with fully
steerable telescopes. The first images of radio sources taken with the telescope were a
startling achievement (Ryle, Elsmore and Neville, 1965), the angular resolution of 23 arcsec
corresponding to a fully filled aperture of diameter one mile (Figure 7.5).
The success of this programme led to the construction of a number of large-aperture
synthesis radio telescopes, including the Westerbork Synthesis Telescope in the Netherlands
(1970) and the next generation 5-km telescope, subsequently named the Ryle Telescope, in
Cambridge (1971). The culmination of these efforts was the construction of the Very Large
Array in New Mexico in the USA (1981), the Australia Telescope (1988) and the Giant
Metrewave Radio Telescope in India (1999). During the 1960s, interferometric techniques
142 7 The new astronomies

were extended to intercontinental baselines, the technique known as very long baseline
interferometry (VLBI), resulting in angular resolutions of the order of a milliarcsec. In this
technique, the receivers at the different observing stations were equipped with very precise
stable clocks, which enabled the radio signals to be recorded separately on magnetic tape
and correlated at a later time. The first successful VLBI observations were carried out in
1967 (Broten et al., 1967; Moran et al., 1967).
The new discipline of radio astronomy resulted in many key discoveries for contempo-
rary astrophysics, those of quasars, the cosmic microwave background radiation, neutron
stars as the parent bodies of radio pulsars, interstellar molecules through their millimetre
line emission and superluminal motions being of special importance. The history of these
developments is described in Parts IV and V.

7.4 X-ray astronomy

Immediately after the Second World War, those physicists and astronomers interested in
ultraviolet and X-ray astronomy made the first observations from above the Earth’s atmo-
sphere.12 The atmosphere is opaque to all radiation with wavelengths shorter than about
300–310 nm and so ultraviolet, X-ray and γ -ray astronomy have to be conducted from above
the Earth’s atmosphere. Above about 150 km, absorption by the Earth’s atmosphere is no
longer important (see Figure 7.1). The German V2 rocket programme had made enormous
strides in rocket technology during the War, and the German scientists who had built them,
led by Werner von Braun (1912–1977), as well as 300 box cars full of V2 parts, were taken
to the USA where they formed the core of the US Army’s rocket programme. The US Army
announced that these rockets would be available for scientific research.13
One of the prime targets of the early rocket experiments was the ultraviolet and X-ray
emission of the Sun.14 It was known that the Sun possessed a very hot corona, and it was
surmised that the Earth’s ionosphere might be ionised by its ultraviolet and X-ray emission.
The first successful rocket ultraviolet observations of the Sun were made in October 1946
by the group led by Richard Tousey (1908–1997) at the Naval Research Laboratory (Baum
et al., 1946). Then, in September 1949, Herbert Friedman (1916–2000) and his colleagues
made the first successful X-ray observations of the Sun, confirming the expectation that the
Sun’s corona is very hot (Friedman, Lichtman and Byram, 1951). These rocket experiments
continued throughout the 1950s and elucidated many of the X-ray properties of the Sun.
The flights of Sputniks 1 and 2 in late 1957 and the orbital flight by Yuri Gagarin (1934–
1968) in 1961 came as a profound shock to the US administration, which realised that the
USA had fallen behind the USSR in space technology and therefore was strategically vul-
nerable. The US response was to set up the National Aeronautics and Space Administration
(NASA) in July 1958 as a civilian organisation to begin the process of catching up with the
USSR. As part of that endeavour, the American Science and Engineering group (AS&E)
was set up in association with the Massachusetts Institute of Technology to work on military
and civilian contracts.
The AS&E group, led by Riccardo Giacconi (b. 1931), developed plans for making
astronomical observations in the X-ray waveband, but their theoretical calculations did
7.4 X-ray astronomy 143

Figure 7.6: The payload of the rocket containing the X-ray detectors which made the first observation
of the discrete X-ray source Sco X-1 and the X-ray background radiation in June 1962. The payload
was constructed by the AS&E group (American Science and Engineering). From W. Tucker and
R. Giacconi, The X-ray Universe (Cambridge, Massachusetts: Harvard University Press, 1985).

450
counter 3 Moon magnetic field vector
2
350 7.0 mg cm Mica
number of counts

counter 2
1.4
250
mg cm 2 Mica

150

50

0° 60° 120° 180° 240° 300° 360°

N E S W N
Figure 7.7: The discovery record of the X-ray source Sco X-1 and the X-ray background emission by
Giacconi and his colleagues in a rocket flight of June 1962. The prominent source was observed by
both detectors, as was the diffuse background emission (Giacconi et al., 1962).

not promise much success with the sensitivities available at that time. The best target
seemed to be the search for fluorescent X-rays from the Moon, which would result from
the impact of streams of energetic solar particles hitting its surface.15 The first successful
flight took place in June 1962 (Figure 7.6). In the five minutes of observing time during
which the rocket payload was above the Earth’s atmosphere, Giacconi and his colleagues
failed to detect any X-rays from the Moon, but discovered an intense discrete source of
emission in the constellation of Scorpius, which became known as Sco X-1 (Figure 7.7)
(Giacconi et al., 1962). In addition, an intense background of X-rays was observed which
was remarkably uniformly distributed over the sky (Gursky et al., 1963). These observations
144 7 The new astronomies

were soon confirmed by other rocket flights by the AS&E group as well as by Friedman’s
group at NRL, which also discovered X-rays from the supernova remnant, the Crab Nebula
(Bowyer et al., 1964). As in the case of radio astronomy, these were entirely unexpected
discoveries.
The next decade saw a flurry of activity in which a dozen or more X-ray astronomy
groups made numerous rocket flights of increasing sophistication to understand the nature
of the X-ray sky. These experiments were of considerable ingenuity and involved making
optimum use of the five minutes observing which was possible during each rocket flight.
The Lockheed group introduced the technique of spin-stabilisation, which enabled small
regions of the sky to be scanned slowly with much greater sensitivity. By 1967, more than 30
X-ray sources were known, including the detection of X-rays from a number of supernovae,
the quasar 3C 273 and the radio galaxy M87. The angular resolution of the X-ray telescopes
was improved by the development of sophisticated X-ray collimators which were placed
in front of the X-ray detectors, and this enabled the optical counterpart of the bright X-ray
source Sco X-1 to be detected – it was a faint blue, variable nova-like star (Sandage et al.,
1966). These pioneering observations provided tantalising glimpses of the richness of the
X-ray sky, but the picture was confused. Some of the sources were highly variable, since
they would be present in one rocket flight and then disappear on the next.
These problems were resolved with the launch, in December 1970, of the UHURU X-ray
observatory, which was the first satellite dedicated to X-ray astronomy and which initiated
the successful series of Explorer satellites sponsored by NASA. The satellite was built by
Giacconi and his colleagues at AS&E and was designed as a simple, robust survey telescope
with two gas-filled proportional counter detectors with angular resolutions of 0.5◦ × 5◦ and
5◦ × 5◦ . The UHURU observatory conducted the first survey of the X-ray sky and revealed
the true nature of the X-ray population (Giacconi et al., 1971b). The X-ray sources turned
out to include a wide variety of very hot objects – X-ray binaries with neutron stars and black
holes as the ‘invisible’ companion, supernova remnants, young radio pulsars, active galactic
nuclei and the intergalactic gas in clusters of galaxies. Some impression of the variety of
sources present in the X-ray sky can be gained from the plot of the sources listed in the
fourth UHURU catalogue (Figure 7.8). The history of the astrophysics of these sources is
described in Part IV.
Over the next seven years following the launch of UHURU seven satellites with X-ray
detectors were flown, including the Netherlands ANS satellite and the UK Ariel V satellite.
The latter satellite included an X-ray spectrometer, which made the first detection of the
X-ray emission line of 26 times ionised iron, Fe+26 , from the hot gas in the Perseus cluster
of galaxies (Mitchell et al., 1976). The next major survey instrument, launched in August
1977, was the NASA High Energy Astrophysical Observatory-A, HEAO-A, which can be
considered to be a super-UHURU with about seven times greater sensitivity.
The next step was the development of X-ray telescopes with imaging capabilities, and
this was achieved by the NASA HEAO-B satellite, which was named the Einstein X-ray
Observatory. The problem to be overcome was the fact that ordinary mirrors do not reflect
X-rays but absorb them. The only means of focussing X-rays with energies of about 1 keV
is to make use of the phenomenon of grazing incidence reflection, in which the X-rays
are deflected through angles of less than about 5◦ . As a result, imaging X-ray telescopes
7.4 X-ray astronomy 145

Coma Virgo
super-cluster ? 3C 273

Her X-1 Sco X-1

NGC 3783

Cen X-3
Cyg A

Cyg X-3

Crab

Perseus

M31 LMC
Cyg X-1 SMC
NGC 6624

Figure 7.8: A map of the X-ray sky showing the objects contained in the fourth UHURU catalogue
(Forman et al., 1978). The names of X-ray sources associated with well known astronomical objects
are indicated.

are very long, often consisting of a paraboloid–hyperboloid configuration to focus the

X-rays onto the distant detector. These concepts were tested in rocket flights in the mid
1960s and were used for solar X-ray studies in the Skylab mission in the early 1970s. In
addition, microchannel plate array detectors were developed to register the X-ray images.
The Einstein X-ray Observatory was a complete X-ray observatory consisting of a suite of
telescopes and instruments for high- and low-angular-resolution imaging as well as X-ray
spectroscopy. The telescope was launched in November 1978 and opened up the detailed
study of the astrophysics of the classes of source so far detected. Among the most intriguing
finding was that essentially all classes of star can be X-ray emitters. The imaging quality
of a few arcsec was achieved by the high-resolution camera, among the most impressive
images being those of well known supernova remnants (Figure 7.9).
There was a natural progression from facilities such as the Einstein Observatory to
large dedicated observatories for use by the astronomical community at large. In the NASA
astronomy programme, the concept was developed of a series of ‘Great Observatories’ which
were designed to be long-lived observatories in space. The four observatories involved in this
programme were the Hubble Space Telescope (HST), the Gamma-Ray Observatory (GRO),
the Advanced X-ray Astronomy Facility (AXAF) and the Space Infrared Telescope Facility
(SIRTF). They were all planned to take advantage of the launch and service capabilities
provided by the NASA Space Transportation System, more commonly known as the Space
Shuttle. These very large and expensive missions were very major undertakings and were
all subject to long development and construction phases.
Returning to the development of X-ray capabilities in space, the AXAF observatory
was launched by the Columbia Space Shuttle in July 1999 and was then boosted into
an elliptical high-Earth orbit, allowing long-duration uninterrupted exposures of X-ray
sources. The observatory was named the Chandra X-ray Observatory. In parallel, the
146 7 The new astronomies

Figure 7.9: X-ray images of four well known supernova remnants observed by the Einstein
X-ray Observatory. Notice the intense central source in the Crab Nebula which is associated with
a rapidly rotating neutron star. From W. Tucker and R. Giacconi, The X-ray Universe (Cambridge,
Massachusetts: Harvard University Press, 1985).

European Space Agency developed the concept of the XMM mission, the acronym meaning
X-ray Multi-Mirror Telescope. The primary objective of this mission was high sensi-
tivity and high spectral resolution, and this was achieved by a telescope system involv-
ing 58 nested paraboloid–hyperboloid mirrors and high-sensitivity CCD X-ray detectors.
In addition, the telescope package included an optical monitor, allowing simultaneous
optical–X-ray observations. These capabilities complemented those of the Chandra Obser-
vatory. The XMM mission was launched in December 1999 and renamed the XMM–
Newton X-ray Observatory. Both missions have been very successful and have provided
astronomers with data of extraordinarily high quality for the study of X-ray sources of all
types.
 7.5 Gamma-ray astronomy 147

At the same time, it was understood that, while these powerful facilities for X-ray astron-
omy enabled individual objects to be studied in exquisite detail, there remained the need to
carry out surveys of the whole sky to understand in detail the nature of the population of
X-ray sources. This objective was achieved by the German-led ROSAT Observatory, ROSAT
standing for Roentgen Satellit, which was launched in June 1990. After nine years in space,
an all-sky catalogue of 150 000 objects was compiled, many of these sources providing
important targets for the Chandra and XMM–Newton Observatories.

7.5 Gamma-ray astronomy

In the early 1960s there was already military interest in placing γ -ray detectors in space in
order to monitor the atmospheric nuclear test-ban treaties concluded between the USA and
the USSR.16 The Vela series of satellites was launched for this purpose in the 1960s, but
there was no intention that they should have any astronomical role. Cosmic γ -rays were first
detected in observations made by the Explorer II satellite in 1965 (Kraushaar et al., 1965),
but this experiment did little more than show that there existed γ -rays which originated from
beyond the Earth’s atmosphere. The first important astronomical observations were made
by the third Orbiting Solar Observatory (OSO-III) launched in March 1967. The prime
discovery of this mission was the detection of γ -rays with energies E γ > 100 MeV from
the general direction of the Galactic centre (Clark, Garmire and Kraushaar, 1968). This
γ -ray flux was convincingly interpreted as the γ -ray emission associated with the decay of
neutral pions created in collisions between relativistic protons and the cold plasma of the
interstellar gas.
These pioneering observations were followed up by balloon observations, but these suf-
fered from severe contamination problems because of the production of secondary γ -rays
as a result of interactions of primary cosmic rays with the nuclei of atoms in the atmosphere.
These experiments provided important experience with compact spark chambers, which had
originally been developed as detectors in high-energy physics experiments. The Small Astro-
nomical Satellite, SAS-2, was launched in November 1972 and included an array of spark
chambers to detect the electron–positron pairs created when an incoming γ -ray is converted
into a pair within the instrument. Although it operated for only eight months and detected
about 8000 γ -rays of cosmic origin, these were sufficient to make a number of key astronomi-
cal discoveries (Fichtel, Simpson and Thompson, 1978). Firstly, it was confirmed that there is
a general concentration of γ -rays towards the plane of the Galaxy. Secondly, discrete sources
of γ -rays were present, in particular two of the sources were associated with the pulsars in
the Crab and Vela supernova remnants. Thirdly, evidence was found for diffuse extragalactic
γ -ray background radiation.
The SAS-2 mission was followed in 1975 by the equally successful COS-B satellite
lauched by a European consortium. It also consisted of an array of spark chambers sensitive
to γ -rays with energies greater than about 70 MeV. It continued to take data continuously for
six and a half years and resulted in a detailed map of the Galactic plane as well as evidence
for 24 discrete γ -ray sources (Mayer-Hasselwander et al., 1982).
148 7 The new astronomies

The first evidence of γ -ray line emission came from balloon-borne telescopes in the
early 1970s by the Rice University Group (Johnson and Haymes, 1973). In 1977, balloon
observations confirmed that the line was the electron–positron annihilation line at 511 keV
originating in the direction of the Galactic centre (Leventhal, MacCallum and Stang, 1978).
Then, in 1984, definitive observations were made of the 1.809 MeV line of radioactive 26 Al
by the HEAO-C satellite (Mahoney et al., 1984), this line also being detected from the
direction of the Galactic centre.
To everyone’s surprise, the Vela satellites, which were designed as γ -ray monitors for
atmospheric nuclear tests, led to a key astronomical discovery in its own right. During the
course of the monitoring, γ -ray bursts of astronomical origin were discovered, each γ -ray
burst lasting typically less than one minute (Klebesadel, Strong and Olson, 1973). During
that time, each burst was the brightest γ -ray source in the sky. The first of these was detected
by the Vela satellite in 1967, but the bursts were not reported in the scientific literature until
1973. Their nature remained a mystery since the angular resolution of the γ -ray telescopes
was very low and the short duration of the bursts made follow-up observations a very serious
challenge.
The second of NASA’s Great Observatories, the Gamma-Ray Observatory, was success-
fully launched by the Space Shuttle in April 1991 and was renamed the Compton Gamma-ray
Observatory (CGRO), providing γ -ray astronomers with four different types of detector to
explore different aspects of the γ -ray sky. These included a scintillation spectrometer for
the detection of γ -ray emission lines (OSSE), a Compton telescope for exploring the dif-
ficult energy range 1–30 MeV (COMPTEL), an energetic γ -ray experiment consisting of
the spark chamber detector (EGRET) sensitive to γ -rays with energies Eγ ≥ 100 MeV and
eight detectors placed at the corners of the satellite designed to monitor γ -ray transients and
γ -ray bursts (BATSE). For all types of γ -ray study, these instruments were about an order of
magnitude more sensitive than the previous generation of γ -ray telescopes. The result was
a definitive map of the whole sky in γ -rays (Figure 7.10), showing clearly the plane of our
Galaxy and a variety of discrete Galactic and extragalactic sources – in the third EGRET
catalogue over 250 sources with photon energies greater than 100 MeV were listed. It was
discovered that the distribution of the γ -ray bursts was isotropic over the sky, an important
clue to their nature. In addition, the plane of the Galaxy was mapped in the radioactive decay
line of 26 Al, and the most extreme active galactic nuclei were established as the sources of
the intense, variable extragalactic γ -ray sources. The 13-ton space observatory remained
nine years in orbit before being destroyed in a controlled atmospheric burn-up in June 2000.
As discussed in Section 7.2.2, extensive air-showers are initiated by very-high-energy
cosmic rays entering the atmosphere, and among the products of the collisions between
these particles and the nuclei of nitrogen and oxygen atoms are neutral pions, which sub-
sequently decay into electron–positron pairs. In turn, the electrons and positrons produce
high-energy γ -rays by bremssstrahlung, which are then converted into electron–positron
pairs in interactions with nuclei, and so on. In 1948, Blackett realised that the speeds of the
ultrarelativistic electrons and positrons were so close to the speed of light that they exceeded
the local speed of light in the atmosphere, v = c/n, where n is the refractive index of the
atmosphere. As a result, the ultrarelativistic electrons and positrons should emit optical
Cherenkov radiation (Blackett, 1948). This prediction was confirmed several years later
7.5 Gamma-ray astronomy 149

Figure 7.10: This map of the γ -ray sky at energies E ≥ 100 MeV was made by the EGRET telescopes
of the Compton Gamma-ray Observatory. The emission from the Galactic plane consists mostly of
γ -rays produced in the decay of neutral pions, π 0 , generated in collisions between cosmic ray protons
and nuclei and the interstellar gas. In addition, various Galactic and extragalactic discrete sources
of γ -rays have been detected, including the pulsars in the Crab and Vela supernova remnants, the
strange object Geminga and the luminous active galactic nuclei such as the quasar 3C 273. (Courtesy
of NASA and the CGRO Science Team.)

by William Galbraith (b.1925) and John Jelley (1918–1997) (Galbraith and Jelley, 1953).
Exactly the same technique can be used to detect ultra-high-energy γ -rays incident upon
the top of the atmosphere.17
This technique can be used to detect γ -rays with energies typically in the range 300 GeV
to 30 TeV, but it is a particularly demanding discipline since the fluxes of γ -rays are very
low and have to be distinguished from very much more common optical pulses associated
with extensive air-showers excited by cosmic rays. The Cherenkov light is distributed over
an area similar to that of an extensive air-shower, and so a large light collector or array
of smaller light collectors is used to detect the weak signal. For many years instruments
such as the 10-metre Fred Lawrence Whipple Telescope at Mount Hopkins in the USA, the
18-mirror array at the Tata Institute of Fundamental Research at Ootacamund in India and
the four-element γ -ray telescope of the University of Durham located at Dugway, Utah,
produced results at the few sigma significance level, but it was not wholly convincing that
significant detections had been made.
The breakthrough came with the development of imaging cameras for the light collectors,
which enabled weak γ -ray sources to be discriminated with high efficiency against the
background of cosmic-ray-induced extensive air-showers. In 1989, the group at the Fred
Lawrence Whipple Observatory used a 37-pixel camera to detect γ -rays from the Crab
Nebula at the 9σ level (Weekes et al., 1989). The higher the angular resolution of the
camera, the better the discrimination against background events. Two years later, with an
150 7 The new astronomies

10−8
n Fn (erg cm−2 s−1) May 1996
April 1995
May 1994
10−9
1977 to 1995

10−10

10−11

10−12

10−13

y
ra

ra
γ–

γ–
V
IR

ke
UV al

eV

V
y
R
o

tic

ra
V
ar

Te
rI
m
di

G
XU
X–
ne

op

10
ra

fa

1
10−14
10 12 14 16 18 20 22 24 26 28
log n (Hz)

Figure 7.11: The energy spectrum, νFν , of the extreme BL Lac object Markarian 421 observed from
the radio to the ultra-high-energy γ -ray waveband. The γ -ray source is highly variable. The lower
energy γ -ray observations were made by the Compton Gamma-ray Observatory and the high-energy
γ -ray points from observations by Atmospheric Cherenkov Imaging Telescopes. From M. Catanese
and T. C. Weekes, Very high energy gamma ray astronomy, Publications of the Astronomical Society
of the Pacific, 111, 1999, 1193–1222.

upgraded camera with 109 pixels, the signals were detected at the 20σ level (Vacanti et al.,
1991). The detected photon rate from the Crab Nebula amounted to about two photons
per minute. These detectors, known as Atmospheric Cherenkov Imaging Telescopes, were
successfully developed by a number of groups and all detected the signal from the Crab
Nebula. The importance of the Crab Nebula is that the flux of ultra-high-energy γ -rays is
constant, being associated with the synchro-Compton radiation of high-energy electrons
within the diffuse nebula as a whole, and so the source can be used as a calibrator. In
subsequent years, ultra-high-energy γ -rays have been detected from the extreme BL Lac
objects Markarian 421 and 501. Unlike the Crab Nebula, these sources are highly variable,
but a large fraction of their emitted energy takes place in the E ≥ 300 GeV energy band
(Figure 7.11). The success of these observations has led to the development of a number of
next-generation Atmospheric Cherenkov Imaging Telescopes.18

7.6 Ultraviolet astronomy and the Hubble Space Telescope

Among the earliest beneficiaries of the opening up of space for astronomy were the ultraviolet
astronomers. The central figure in this story is Lyman Spitzer (1914–1997), who, in 1946,
7.6 Ultraviolet astronomy and the HST 151

Figure 7.12: The discovery of the Lyman-γ absorption line of deuterium in the interstellar medium in
the far-ultraviolet spectrum of β Cen observed by the Copernicus satellite (Rogerson and York, 1973).
The broad absorption feature is the interstellar Lyman-γ line of hydrogen centred at 97.2537 nm. The
deuterium line is the weaker feature at wavelength 97.2272 nm.

was invited by the RAND project of the US Air Force to write a report on the utilisation of
space for astronomical purposes.19 Although it was many years before any of these ideas
could become a reality, the seeds were sown early. The development of space ultraviolet
astronomy followed the same pattern as that of X-ray astronomy. Firstly, there were rocket
experiments, which gave a flavour of the scientific potential of the new waveband. Then,
in 1957, as soon as the USSR space achievements galvanised the US space programme
into action, Spitzer and his colleagues planned a series of three space observatories, to be
known as the Orbiting Astrophysical Observatories (OAO), which were to be dedicated to
spectroscopy in the ultraviolet waveband between 90 and 330 nm.
Unlike the other new astronomical wavebands, the astrophysical objectives of ultravio-
let astronomy were well defined. The resonance transitions of essentially all the common
elements lie in the ultraviolet rather than the optical waveband, and so studies of the dif-
ferent phases and chemical composition of interstellar matter are very effectively carried
out in this waveband. OAO-3 was named the Copernicus satellite and was the great suc-
cess of the series. The spectrographs were of very high spectral resolution and had the
capability of exploring the wavebands to the short-wavelength side of the Lyman-α line
at 121.6 nm (Rogerson et al., 1973a). This capability was of special significance because,
among the many resonance lines, those of deuterium are of great cosmological impor-
tance – one of the key discoveries of the mission was the detection of interstellar Lyman-β
to Lyman- absorption lines of deuterium in the spectra of luminous blue stars such as
β Cen (Figure 7.12) (Rogerson and York, 1973). In addition, observations of these deu-
terium transitions towards different stars showed that its interstellar abundance is remark-
ably constant wherever one looks in the local interstellar medium, corresponding to an
152 7 The new astronomies

abundance by mass relative to hydrogen of about 1.5 × 10−5 , a key result for astrophysical
cosmology.
In addition, abundances of the common elements were measured in the interstellar
medium for the first time. The mission also found evidence for a hot component of the
interstellar gas through observations of the absorption lines of highly ionised oxygen, O5+ .
The OAO observatories led in turn to the launch of the International Ultraviolet Explorer
(IUE) in 1978, a joint UK–European Space Agency–NASA project. This was a spectacularly
successful space astronomy mission and has had an impact upon essentially all branches of
astronomy.20
In many ways the IUE was the precursor of the Hubble Space Telescope (HST).21 Optical
astronomers had long been aware of the fact that large ground-based telescopes never
achieve their theoretical angular resolution because of refractive index fluctuations in the
atmosphere which blur the images of stars to typically about 1 arcsec, the phenomenon
known as astronomical ‘seeing’. This figure should be compared with the theoretical angular
resolution of a 4-metre telescope of about 0.03 arcsec. In addition to providing very much
sharper astronomical images, the increase in angular resolution brings with it an increased
sensitivity to point sources since, even on the darkest sites, a diffuse sky background is
present due to atmospheric light scattering and emission. These problems are eliminated by
placing the telescope above the Earth’s atmosphere.
In the 1960s, plans were formulated for the construction of a Large Space Telescope,
which would be an all-purpose optical–ultraviolet space observatory and which would build
on the achievements of space ultraviolet astronomy. The plans were initially for a 3-metre
space telescope, but this was reduced to 2.4 metres when the decision was taken to launch
the telescope using the Space Shuttle. Not only would the telescope be launched by the
Shuttle, but it would also be regularly serviced by it, enabling malfunctioning components
to be replaced and new scientific instruments to replace the old. Since the telescope would
be placed in low Earth orbit, it had to be reboosted into a higher orbit every few years during
the servicing missions.
The approval process for the Hubble Space Telescope was not straightforward. In many
ways, the leap from the small OAO class specialist missions of the 1960s to a fully equipped
multi-purpose 2.4-metre telescope, which would operate at the diffraction limit, was a huge
one and brought with it many technical problems. The biggest problem, however, was the
fact that the project was very expensive, the cost estimates being greater than for any pure
science programme ever undertaken. In the end, international collaboration was secured
with the European Space Agency, which negotiated a 15% involvement in the telescope
for European astronomers. The Ford administration approved the project in 1977, but the
budget was very tight indeed.
The programme encountered major technical and financial difficulties within a couple
of years of approval. The initial launch date was scheduled for the last quarter of 1983,
but this soon proved to be over-optimistic. The crisis came in 1981 when the programme
almost ran out of money. Managerial changes were made and a new, more realistic, budget
was set for the programme. The programme was further delayed by the tragic loss of the
Challenger Space Shuttle in 1986, which resulted in a cessation of Shuttle launches for
a few years. Eventually, the telescope was launched in April 1990. Within weeks, it was
 7.7 Infrared astronomy 153

found that the primary mirror had been figured to the wrong shape – the telescope system
had an unacceptable amount of spherical aberration that blurred the images so that, without
computer enhancement of the data, they were little better than what could be obtained
from the ground. By the process of deconvolution, it proved possible to recover the full
angular resolution capabilities of the telescope for imaging, but with very much reduced
sensitivity.
NASA and the Space Telescope Science Institute immediately set about seeking solutions
to the spherical aberration problem and came up with the concept of introducing correction
optics into the optical train for the scientific instruments, which would restore the full
capability of the HST for astronomical imaging and spectroscopy. On the first servicing
mission of 1993, the correction optics were successfully installed, along with a new Wide
Field Camera with correction optics built into the optical design. The full capability of the
HST was restored and the results have been quite spectacular. The astronomical results have
far exceeded the most optimistic expectations of the astronomers, and there is no field of
astronomy which has not been impacted by its discoveries. NASA made the wise decision to
make a significant investment in the public dissemination of the science and images obtained
by the HST, with the result that the international public was immediately exposed to some
of the most important and spectacular images ever taken by an astronomical telescope.
There is no more remarkable picture than that of the Hubble Ultra-Deep Field, the results
of a three-month exposure taken with the most recent camera, the Advanced Camera for
Surveys (ACS), and the Near Infrared Camera and Multi-object Spectrometer (NICMOS)
(Figure 7.13).
The remarkable success of the Hubble Space Telescope has encouraged NASA to plan
the next leap forward in astronomical imaging and spectroscopy from space with the devel-
opment of a 6–8-metre space telescope, optimised for observations in the near-infrared
waveband, 1–5 μm, with capabilities to wavelengths as short as 0.5 nm. It is planned that
this telescope, named the James Webb Space Telescope, will be launched in 2011 and will
be placed at the second Lagrange point (L2), approximately 1.5 million kilometres from
Earth, outside the orbit of the Moon. The region about L2 is a gravitational saddle point,
where the telescope will remain at a roughly constant distance from the Earth throughout
the year through a series of small spacecraft manoeuvres.

7.7 Infrared astronomy

The infrared radiation of the Sun was first detected in 1800 by William Herschel in his
famous experiments in which he placed mercury-in-glass thermometers with blackened
bulbs beyond the red end of an optical spectrum of the Sun. He found a greater temperature
increase in what he termed the ‘ultra-red’ as compared with the red region of the spectrum
and noted that these rays were refracted less than optical light (Herschel, 1800a–d). In his
words,
there are rays coming from the Sun, which are less refrangible than any of those which affect the
sight. They are invested with a high power of heating bodies, but with none of illuminating objects;
and this explains why they have hitherto escaped unnoticed.
154 7 The new astronomies

Figure 7.13: This image of the Hubble Ultra-Deep Field (HUDF) was obtained in a three-month
set of observations using the Advanced Camera for Surveys (ACS) and the Near Infrared Camera
and Multi-object Spectrometer (NICMOS) of the Hubble Space Telescope. Within a square area of
angular size 3 arcmin, about 10 000 very distant galaxies have been detected. (Courtesy of NASA,
ESA, Dr Steve Beckwith, the HUDF Team and the Space Telescope Science Institute.)

Note that this work was carried out two years before Thomas Youngs’ pioneering papers
on the ‘Theory of light’. Because of the ‘invisibility’ of infrared radiation, this type of
astronomy depended wholly upon the development of either ‘thermal’ detectors, in which
the incident radiation gave rise to a temperature rise in the detector, or of ‘non-thermal’
devices, in which chemical or electronic transitions were excited by the infrared radiation.22
Highlights of nineteenth-century developments in infrared astronomy include measure-
ments by Claude-Servais Pouillet (1790–1868) of the total heat flux of the Sun using his
pyrheliometer, a small waterbath in a blackened enclosure with a thermometer to measure
the temperature rise (Pouillet, 1838). Including a correction for absorption in the Earth’s
atmosphere, he found a value for the solar constant, the incident solar energy flux at the
top of the atmosphere, of 1.44 kW m−2 , in remarkable agreement with the present value
7.7 Infrared astronomy 155

of 1.37 kW m−2 . Soon after, John Herschel discovered the broad absorption bands in the
spectrum of the Sun, due to molecular absorption in the Earth’s atmosphere, now called
the telluric bands (Herschel, 1840). Herschel’s experiments were carried out using a sheet
of paper blackened by soot, which had been soaked in alcohol. The dispersed spectrum
of the Sun dried out the alcohol in regions where the solar spectrum was strong and left
damp the regions where four deep absorption troughs were found. In 1847, this remarkable
observation was confirmed by Fizeau (1819–1896) and Foucault, who used sensitive alcohol
thermometers with miniature bulbs (Fizeau and Foucault, 1847). In this same paper, they
also showed that the infrared rays displayed the same properties as light, namely interference,
polarisation and diffraction.
Thermoelectricity was discovered by Thomas Seebeck (1770–1831) at Jena in 1822 and
resulted in the invention of the thermocouple, which consisted of pairs of dissimilar metallic
strips, such as bismuth and copper, and which provided a better means of measuring tiny
temperature differences than Herschel’s thermometers. The thermocouples could be built
into arrays which were known as thermopiles. The first observations of stars in the infrared
waveband were made by William Huggins in 1868–1869, who used a thermopile with a
small number of elements (Huggins, 1869). Huggins used the technique which would now
be called ‘nodding’ to make observations of the star. This involved observing a blank field
close to the star, allowing the galvanometer needle to settle, then moving the telescope onto
the star, making the observation and then reobserving the same blank field. In his words,

The needle was then watched during five minutes or longer; almost always the needle began to move
as soon as the image of the star fell upon it. The telescope was then moved, so as to direct it again
to the sky near the star. Generally, in one or two minutes, the needle began to return to its original
position. In a similar manner twelve to twenty observations of the same star were made.

By this means, Huggins made successful observations of Regulus, Arcturus, Sirius and
Pollux, later confirmed by Edward Stone (1831–1897).
Perhaps the most important nineteenth-century figure in developing the techniques of
infrared astronomy was Samuel Pierpoint Langley (1834–1906), who perfected the use of the
bolometer for astronomical spectroscopy in the infrared region of the spectrum. Langley’s
bolometer used the fact that the resistance of platinum is highly temperature dependent.
Platinum strips a few microns wide could be used and the tiny resistance changes were
measured by a precision Wheatstone bridge so that it was possible to measure temperature
changes as small as 10−4 K. Langley devoted most of his efforts to bolometric observations
of the Sun and, from observations made from the summit of Mount Whitney, extended
measurements of the solar spectrum to 5.3 μm (Langley, 1886). In subsequent observations,
he mapped the absorption lines in the solar spectrum out to 5.3 μm, observing about 700
lines and measuring accurate wavelengths for 222 of them (Langley, 1900).
Towards the end of the nineteenth century, yet another approach to infrared photometry
was developed based upon the radiometer invented by William Crookes (1832–1919). The
device consisted of vanes which were blackened on one side and mounted on a support, such
that when radiation was incident upon the vanes, they rotated. In the torsion radiometer,
the vanes were blackened on the same side and, if there was a difference in the intensity
of radiation on one compared with the other, there would be a net deflection of the vanes.
156 7 The new astronomies

This device was used by Charles Abbot (1872–1973) to carry out the first stellar spectrora-
diometry of stars. Using the device at the Coudé focus of the 100-inch Hooker telescope at
Mount Wilson, Abbot measured the spectra of 9 stars in 15 wavebands in the wavelength
region 0.437 to 2.224 μm, including four wavebands longer than 1 μm. This was the first
time that spectrophotometric energy distributions were available for stars over a wide wave-
length range, and they enabled black-body curves to be fitted to their spectra (Abbot, 1924).
He found that the temperatures ranged from about 2500 K for α Herculis to 16 000 K for
Rigel. Assuming that the stars behave like black bodies, their angular sizes could be worked
out using the Stefan–Boltzmann law, and then, knowing the distances of the stars, their
physical sizes could be estimated. Whereas stars such as Procyon had diameters similar to
the Sun, Rigel and α Herculis had radii several hundred times that of the Sun. The results
were generally in good agreement with the sizes found by Michelson and Pease’s stellar
interferometric observations (Michelson and Pease, 1921).
The final part of the pre Second World War story developed from the invention, by
Peyotr Lebedev (1866–1912), of the vacuum thermocouple, which had markedly superior
performance to the standard thermocouple because of the large decrease in thermal losses by
conduction and convection. These detectors were optimised for use in astronomy by Herman
Pfund (1878–1949) at the Allegheny Observatory and by William Coblentz (1873–1962) at
the US Bureau of Standards. The most important of these observations were carried out by
Edison Pettit (1889–1962) and Seth Nicholson (1891–1963), who overcame the inherent low
efficiency of these detectors by using the Hooker 100-inch telescope at Mount Wilson (Pettit
and Nicholson, 1928). During the 1920s, they carried out a programme of observations of
124 bright stars and came to similar conclusions to those of Abbot, but with much better
statistics. The diameters of stars determined by the photometric technique were compared
with those derived by Pease by optical interferometry, and reasonable agreement was found,
the discrepancies being attributed to the fact that the photometric technique assumed that
the spectra were black bodies.
Like many astronomical disciplines, infrared astronomy benefited from technological
developments stimulated by the needs of the military, in particular the need to develop heat-
seeking missiles. The first of a new generation of detectors for infrared astronomy was the
lead sulphide cell (PbS), which was pioneered by Charles Oxley and Robert Cashman (1906–
1988) and which could operate to wavelengths as long as 3.6 μm. The great advantage of
this new class of semiconductor detector was that they were about 1000 times more sensitive
than a thermopile. The first infrared spectroscopy of bright stars and planets was carried out
by Gerard Kuiper (1905–1973) and his colleagues in 1947 (Kuiper, Wilson and Cashman,
1947), who incorporated the technique of rapid chopping of the beam to overcome the
problem of detecting faint sources against the bright infrared background, which was also
incident on the detector.
One of the major astronomical advantages of observing in the infrared waveband is that
interstellar dust, which obscures many of the most interesting regions in gas clouds and
galaxies, becomes transparent. The dependence of interstellar extinction upon wavelength
had been determined in the optical waveband by Trumpler in the early 1930s (Trumpler,
1930) and, in more detail, by Stebbins, Hufford and Whitford using photoelectric techniques
in 1940 (see Section 5.6). Interstellar extinction can be described by an extinction coefficient,
7.7 Infrared astronomy 157

−6

−5

−4

−3

−2
Δm

−1

+1

+2

3 2 1 1/ λ 0

U B V R I J K L MN

Figure 7.14: An example of the form of interstellar extinction curve derived by Johnson in 1965
(Johnson, 1965). Frequency is plotted on the abscissa in units of λ−1 , where the wavelength, λ, is
measured in micrometres. The extinction is measured in magnitudes relative to magnitude 0 in the V
waveband. At infinite wavelength, the extinction is zero, corresponding to m = −3.5.

α, such that I = I0 e−αr , where r is the distance. Joel Stebbins and his colleagues found that
α was inversely proportional to wavelength in the optical waveband (Stebbins, Hufford and
Whitford, 1940). This work was extended by Whitford, who observed distant supergiant
stars and concluded that this power-law extinction law was a reasonable approximation out
to 2 μm (Whitford, 1948) (see Figure 7.14).
The background radiation in the detector and its enclosure could be significantly reduced
by cooling, and Harold Johnson (1921–1980) carried out the first systematic surveys of the
stars of a wide range of spectral types, first using cooled PbS detectors and then, after 1961,
a cooled indium antimonide detector, which had the advantage of operating at wavelengths
as long as 5 μm. In 1962, he defined what have become the standard broad-band J (1.2 μm),
K (2.2 μm), L (3.6 μm) and M (5.0 μm) infrared wavebands by means of interference filters
(Johnson, 1962).
Also in 1961, Frank Low (b. 1933) pioneered observations at yet longer wavelengths
through his development of the gallium-doped germanium bolometer, which operated at
liquid helium temperatures (Low, 1961). These developments led to the definitions of the N
(10 μm) and Q (22.2 μm) wavebands, at which the atmosphere has sufficient transparency
to allow observations to be made from good high-altitude, ground-based sites (Low and
Johnson, 1964; Low, 1966). In the late 1950s and early 1960s, Johnson and his colleagues
measured the magnitudes of several thousand stars in a programme of UBVRIJKLMN
photometry (Johnson et al., 1966). These careful studies confirmed the strong dependence
of interstellar extinction upon wavelengths out to 2.2 μm (Figure 7.14). In addition, they
enabled effective temperatures and bolometric corrections to be determined for stars of a
wide range of spectral types.
158 7 The new astronomies

At about the same time, Gerry Neugebauer (b. 1932) and Robert Leighton (1919–1997) of
the California Institute of Technology began a major survey of the sky north of declination
δ = −33◦ at 2.2 μm with a 62-inch telescope, or rather light-collector, which they built
themselves (Neugebauer and Leighton, 1969). The images were about 4 arcmin in size
but, using an array of eight lead sulphide detectors matched to the beam of the telescope,
they found 5612 infrared sources brighter than K = 3. The Two-micron Sky Survey was of
particular importance for astronomy because of its reliability and completeness. In many
cases, the infrared emission represented the extension of the optical spectra of stars into
the infrared waveband but, in addition, many more strong infrared emitters were discovered
than were expected. About 50 stars were found to have (I–K) colour temperatures of only
about 1000 K, while others turned out to be very intense emitters at far-infrared wavelengths,
including objects such as the late M-supergiant NML Cygni and the heavily reddened carbon
star IRC+10216.
At longer wavelengths, 4, 10 and 20 μm, the infrared sky was surveyed using helium-
cooled germanium bolometers in a series of rocket flights by Russell Walker (b. 1931)
and Stephan Price (b. 1941) of the Air Force Cambridge Research Laboratories (AFCRL)
(Walker and Price, 1975). An important result of the AFCRL survey was that many of
the brightest 20 μm sources were extended regions of ionised hydrogen, such as the Orion
Nebula, with temperatures of only about 250 K, associated with hot dust grains within the
ionised clouds.
In 1966, a key discovery was made by Eric Becklin (b. 1940) and Neugebauer, who
made the first painstaking maps of the Orion Nebula at 1.65, 2.2, 3.5 and 10 μm with
the Palomar 200-inch telescope (Becklin and Neugebauer, 1967). To their surprise, a very
intense infrared ‘star’ was detected, not from the prominent optical nebula but from an
obscured area to the north of the four trapezium stars which are responsible for illuminating
the visible nebula (Figure 7.15(a)). The source was as luminous as the prominent trapezium
stars, but there was no detectable optical emission from it. Their preferred interpretation was
that the object was a massive protostar which was still enshrouded in a dusty envelope that
absorbed the energy emitted by the protostar and reradiated it at far-infrared wavelengths. In
another heroic paper of 1968, Becklin and Neugebauer made the first infrared maps of the
Galactic centre in the H, K and L wavebands (Becklin and Neugebauer, 1968). The optical
extinction to the Galactic centre was found to be about 25 magnitudes but, because of the
rapid decrease in extinction with increasing wavelength, the Galactic centre region itself
was observable for the first time (Figure 7.15(b)). They found evidence for an increase in the
stellar density towards the centre as well as a compact region, coincident with the compact
radio source Sagittarius A, which is associated with the dynamical centre of our Galaxy.
The great potential of the infrared waveband for these types of study led to the con-
struction of telescopes optimised for observations in the infrared region of the spectrum.
The UK Infrared Telescope (UKIRT) and the NASA Infrared Telescope Facility (IRTF),
both located on the summit of Mauna Kea, began operation in the late 1970s and played a
major role in making infrared observations an integral part of observational astrophysics.
In the mid 1970s, lead sulphide detectors were replaced by the more sensitive indium anti-
monide detectors, and, by the late 1970s, detector technology and observing techniques had
advanced to such an extent that galaxies as faint as K = 18 could be detected.
7.7 Infrared astronomy 159

(a)

–28°56'
APERTURE

57' 2

4
6
58' 2
DECLINATION (1950)

59' 6

6 2
70
–29°00'
2
4 2

01' 2
4 4
2 2

02'

03'

17h42m45s 40s 35s 30s 25s 20s

RIGHT ASCENSION (1950)

Figure 7.15: (a) The spectrum of the bright infrared star in the Orion Nebula, which became known as
the Becklin–Neugebauer object. The dashed line shows the spectrum of a black body at a temperature
of 700 K (Becklin and Neugebauer, 1967). (b) The first infrared map of the Galactic centre region
mapped with an angular resolution of 0.25 arcmin at a wavelength of 2.2 μm. The central source
coincides with the bright radio source Sagittarius A, which is located at the dynamical centre of the
Galaxy (Becklin and Neugebauer, 1968). The dashed lines show the directions of the scans made to
construct the infrared map.
160 7 The new astronomies

In the mid 1980s, infrared array technology was declassified by the US military agencies,
who had invested heavily in this technology for use as guidance devices for cruise missiles.
These arrays had to be specially modified for astronomical use, particularly with regard to
their uniformity and the reduction of the dark current in the detectors to very low values.
With these developments, it at last became possible to build infrared array cameras and
spectrographs with which to take images and spectra in the infrared waveband.23
The far-infrared regions of the spectrum cannot be observed from the surface of the
Earth because of atmospheric absorption (see Figure 7.1). Pioneering experiments were
carried out from high-flying aircraft and from balloon-borne platforms in the 1970s. Low
carried out a number of exploratory programmes from a modified executive Learjet (Low
and Aumann, 1970; Low, Aumann and Gillespie, 1970), which led in due course to the
development by NASA of the Kuiper Airborne Observatory. This facility consisted of a
Lockheed C-141 transport aircraft with a hole cut in the side to enable observations to be
made with a 91-cm diameter telescope. Typically, the aircraft flew at an altitude of about
13 km, and observations could be carried out for about eight hours at high altitude.
The next natural step was to construct a dedicated satellite to undertake a systematic
survey of the far-infrared sky. The Infrared Astronomy Satellite (IRAS), an international
venture involving the Netherlands, the USA and the UK, was launched in January 1983, and
the mission lasted ten months in space until the cryogens were exhausted. The whole sky
was mapped in those infrared wavebands which are inaccessible from the ground, namely
bands centred on 12, 25, 60 and 100 μm. About 250 000 infrared sources were discovered,
and broad-band colours were measured for many of these. These observations have had a
major impact upon essentially all branches of astronomy, but the outstanding contributions
were made in the study of regions of star formation and the realisation that many galaxies
emit as much radiation in the far-infrared waveband as they do at optical wavelengths.24
The success of the IRAS mission led to the development of the Infrared Space Obser-
vatory (ISO) by the European Space Agency. This satellite observatory operated in space
from November 1995 to May 1998 and was a cryogenically cooled telescope which could
operate in conditions of very low thermal background into the far-infrared region of the
spectrum. The duration of the mission was determined by the lifetime of the cryogens which
maintained the mirror and structure of the telescope at 4 K. The observatory was 1000 times
more sensitive than IRAS with 100 times better angular resolution at 12 μm. In following
up the IRAS survey, this mission began the exploitation of the science of the infrared wave-
bands which are inaccessible from the ground with array detectors and spectrographs.25
This mission was followed by the NASA Space Infrared Telescope Facility (SIRTF), sub-
sequently named the Spitzer Space Telescope, which was successfully launched in August
2003. The European Space Agency will launch the Far-Infrared Space Telescope (FIRST),
named the Herschel Space Observatory, in 2008.
While these developments were taking place in the near-, mid- and far-infrared wave-
bands, millimetre astronomers were pushing their techniques to higher and higher fre-
quencies. The story of the discovery of molecular lines and their role in understanding
the physics of the interstellar medium is taken up in Chapter 9. Radio telescopes operat-
ing at centimetre and millimetre wavelengths were constructed, for example the Kitt Peak
12-metre telescope, but to make observations in the submillimetre waveband considerably
7.7 Infrared astronomy 161

Figure 7.16: Illustrating the transparency of the atmosphere in the millimetre and submillimetre wave-
bands for 0.5 mm precipitable water vapour, a very good figure for a high, dry observing site such as the
summit of Mauna Kea or the Cajnantor plateau in Chile. At frequencies greater than 300 GHz (1 mm
wavelength), the main broad-band windows for which the transmission is adequate for astronomical
observations are centred at about 350 GHz (850 μm), 450 GHz (650 μm), 650 GHz (450 μm) and
850 GHz (345 μm). For optimum detection of the astronomical signals, the submillimetre filters are
matched to the transmission function of the atmosphere. (Data taken from the Caltech Submillimetre
Observatory’s atmospheric transmission estimator.)

higher surface accuracy was required. As a result, many early observations in submillime-
tre astronomy were carried out on optical or infrared telescopes. New types of heterodyne
receivers had to be developed to enable spectral line observations at wavelengths less than
1 mm to be carried out. Schottky barrier diodes were superseded by SIS detectors, which
made enormous strides in sensitivity throughout the period 1980 to 2000. These detectors
enabled spectral line observations to be made in all the submillimetre ‘windows’ which are
accessible to ground-based telescopes from high, dry observing sites (Figure 7.16). The big
advantage of observing in these wavebands is that the spectral lines are relatively stronger
than at millimetre wavelengths and probe deeper into regions which are optically thick
at the longer wavelengths. In addition, the higher rotational transitions of molecules can
be observed, for example the j = 3 → 2 rotational transition of CO at 346 GHz and the
j = 6 → 5 transition at 691 GHz. There is a myriad of molecular lines in the submillime-
tre waveband, and, in high-resolution, high-sensitivity spectral scans, the noise signal is
associated with the multitude of weak spectral lines in this region of the spectrum.
The opening up of the submillimetre waveband for continuum astronomical observa-
tions required the development of sensitive bolometric detectors which operated at liquid
helium temperatures (4 K) or less in order to minimise the effects of thermal noise in the
detectors. Germanium bolometer detectors were successfully developed by Low in the early
1960s and were used by Low and Hartmut Aumann (b. 1940) to make observations in the
162 7 The new astronomies

50–300 μm waveband from the Learjet observatory (Low and Aumann, 1970). The impor-
tant breakthrough came with the construction of composite bolometers, in which a tiny piece
of germanium crystal acts as a very sensitive thermometer to detect the small temperature
rise of an absorbing film consisting of a conducting metal film deposited on a dielectric
substrate. These single-element detectors were developed during the 1970s and were used
successfully on the UKIRT and IRTF telescopes in Hawaii, both of which had the great
advantage of having chopping secondary mirrors. The first common-user bolometer for the
submillimetre wavebands, UKT14, was built for UKIRT in the early 1980s. Heroic observa-
tions with this instrument established the fact that dust emission from star-forming regions
and star-forming galaxies can be successfully observed in the submillimetre waveband.
The potential of these line and continuum observations led to the construction of dedicated
submillimetre telescopes, such as the 15-metre James Clerk Maxwell Telescope and the
10.4-metre CalTech Submillimeter Observatory (CSO) on Mauna Kea in Hawaii and the
15-metre Swedish–European Southern Observatory Telescope (SEST) in Chile.
The discipline of submillimetre astronomy really came of age in the mid 1990s with
the development of the first common-user detector arrays, which were built into the mas-
sive SCUBA submillimetre camera on the James Clerk Maxwell Telescope (JCMT). The
SCUBA array consisted of two back-to-back arrays of composite bolometers, principally
for use in the 850 and 450 μm wavebands, and enabled observations in these wavebands
to be made 1000 times more rapidly than had been previously possible. Among the most
important discoveries was a population of very distant star-forming galaxies, which provide
important information about the star-formation history of the Universe (see Chapter 14).
The success and importance of these observations led to the Atacama Large Millimetre
Array (ALMA) project, which involves constructing an array of 64 12-metre submillimetre
telescopes operating as a synthesis array on the Cajnantor plateau at 5000 m altitude in
the Atacama Desert in Chile. This international venture involving the USA, the European
Southern Observatory and the Japanese National Astronomy Observatory is planned to
begin observations in about 2011 and is expected to do for submillimetre astronomy what
the HST has done for optical astronomy.

7.8 Optical astronomy in the age of the new astronomies

In parallel with the technical advances which expanded enormously the wavebands available
for astronomical observation, optical astronomy developed out of all recognition over the
same period. In the 1960s, the construction of a number of 4-metre class optical telescopes
was begun with the intention of providing improved access for many more astronomers to
world-class observing facilities.
In the USA, the need for national telescopes in addition to the private observatories was
recognised with the founding in May 1960 of the Association of Universities for Research in
Astronomy (AURA). This federally funded organisation had the responsibility of building
and operating telescopes for the US astronomical community, the principal facilities being
the 4-metre Mayall Telescope at the Kitt Peak National Observatory (KPNO) in Arizona,
the Sacramento Peak Solar Observatory in New Mexico and the 4-metre Blanco telescope
 7.8 Optical astronomy 163

at the Cerro Tololo Inter-American Observatory (CTIO) at La Serena in Chile, building

upon the American astronomers’ long association with observatories in Chile.26
Within Europe, it was realised that, while the USA had capitalised upon its advantage
in the northern hemisphere, the many riches of the southern hemisphere, including the
Galactic centre and the Magellanic Clouds, had yet to be explored with large telescopes.
The European Southern Observatory (ESO) was set up in 1962 ‘to establish and operate an
astronomical observatory in the southern hemisphere, equipped with powerful instruments
with the aim of furthering and organising collaboration in astronomy’. Operation of the
3.6-metre telescope at La Silla, a 2400 m mountain bordering the southern extremity of the
Atacama desert in Chile, began in 1977.
The UK had long-standing interests in the southern skies, dating from the time of the
Herschels and the foundation of the Cape and Pretoria Observatories in South Africa. The
UK did not become a founding member of ESO but joined with Australia to construct the
3.9-metre Anglo-Australian Telescope on Siding Spring Mountain in New South Wales,
which was completed in 1975.27 France, Canada and Hawaii agreed to collaborate in the
construction of the 3.6-metre telescope on the summit of Mauna Kea in Hawaii, one of
the best and highest all-round sites for astronomy in the world, and the telescope began
operations in 1979. The UK, Spain and the Netherlands constructed an observatory for the
northern skies on the island of La Palma in the Canary Islands, including the 4.2-metre
William Herschel Telescope, which was completed in 1987.

7.8.1 Electronic detectors for astronomical telescopes

Just as important as the availability of large telescopes for the community as a whole was the
development of electronic detectors, which, by the end of the twentieth century, had largely
replaced the photographic plate as the preferred means of recording astronomical images and
spectra.28 The photoelectric effect was discovered by Heinrich Hertz (1857–1894) while he
was carrying out his brilliant experiments which demonstrated that electromagnetic waves
have all the properties of optical light in 1885–1887. It was not until the 1920s, however, that
photoelectric photometry began to make an impact upon astronomy with the development
of electronic vacuum tubes. These devices had the advantage of having a linear response
over a wide dynamic range and so enabled the calibration of the magnitudes of stars and
galaxies to be carried out much more effectively. The first photomultiplier tubes which had
a major impact upon astronomy were constructed by Vladimir Zworykin (1889–1953) at
the RCA laboratories. The principle of these devices is that the incoming photon causes a
secondary electron cascade to take place though a series of dynodes so that each detected
photon results in a very short burst of electrons at the anode. The efficiency of detection of
the photons is only limited by the quantum efficiency of the first stage of photon detection.
In fact, the primary use of these photomultiplier tubes was for the soundtracks of movies in
the motion picture industry. These devices were first used by Albert Whitford and Gerald
Kron (b. 1913) as an autoguider for the 60-inch telescope at Mount Wilson, but became the
preferred means of calibrating magnitude scales after the Second World War.
The next step in the application of advanced electronics to optical astronomy came with
the development of image intensifiers. These were off-shoots of the television industry
164 7 The new astronomies

Figure 7.17: Willard S. Boyle (left) and George E. Smith, the inventors of the charge-coupled device
(CCD), demonstrating the imaging capabilities of their patented CCD camera in 1974. (Courtesy of
the AT&T Laboratories.)

and, in particular, their development for low-light level applications for military purposes
during the 1960s and 1970s. The principle of these devices is that each photon detected
by the photocathode results in an electron cascade, as in the photomultiplier, but now the
electron beam is focussed onto a light-emitting screen which is scanned by a television
camera. The arrival of each detected photon is registered and the image reconstructed
by photon counting. These types of systems, including the Vidicon System developed
at the Westinghouse Corporation and the Image Photon Counting Systems developed by
Alexander Boksenberg (b. 1936), completely transformed the spectroscopy of faint objects
during the 1970s. They are ideal for faint objects since the counting rate is limited to about
one photon per pixel during the time it takes the television system to register the arrival of
the photon. The Faint Object Camera of the Hubble Space Telescope used this technology
for the imaging of faint objects in the ultraviolet region of the spectrum.
In 1969, the charge-coupled device (CCD) was invented by Willard Boyle (b. 1924) and
George Smith (b. 1936), who were working at the Bell Telephone Laboratories at Murray
Hill, New Jersey (Boyle and Smith, 1970). Their objective was to develop the technology for
a ‘Picturephone’, which would enable telephone callers to see each other (Figure 7.17). The
semiconducting materials which detect the photons can have very high quantum efficiencies
and then the ejected electrons are stored in potential wells within the semiconductor material.
The problem is how to extract the signals without undue losses. This is where the process of
charge-coupling plays a key role. Once the signal is accumulated on the chip, the electrons
are shuffled along the rows of the detector array and are read out by a single amplifier at
the end of the row. The first 100 × 100 arrays were introduced in 1973 and the patent for
the device was received in 1974. The astronomers realised the potential of these devices
 7.8 Optical astronomy 165

for astronomy and they were developed under contract from the Jet Propulsion Laboratory
by Texas Instruments, who built the first devices specifically for astronomy in 1976. The
development of these devices for astronomy received an enormous boost by their selection
as the preferred detectors for the Wide Field Camera of the Hubble Space Telescope in 1977.
Since then, CCDs have dominated optical astronomy in providing directly digital images
with very high quantum efficiency. Rather than the 1% achievable by photographic plates,
the CCDs can have quantum efficiencies of up to about 80%, the equivalent of increasing
the collecting area of the telescope by a factor of 80. In parallel with their development for
astronomy, CCD detectors have come to dominate the domestic camera market, and digital
cameras now contain CCD chips with many millions of pixels. In the same way, the sizes of
astronomical quality CCD chips have become very large, so that, except for very-wide-field
imaging, the days of the photographic plate have passed into history.

7.8.2 New technologies for large telescopes

The Palomar 200-inch telescope was the ultimate in traditional telescope building. As one
writer put it, ‘the Hale telescope stands among all telescopes as the climax of dreadnought
design’ and represented the pre-Second World War ‘brute force’ approach to telescope
construction. There have been many changes in the concepts for large-telescope design
since then. During the 1970s, for example, telescopes were built short and stubby, which
had the advantage of making the structure more rigid for a given amount of steel and also
meant that the building which enclosed the telescope could be much smaller. These changes
both brought with them large cost advantages.
For many years, 4 metres was regarded as the largest feasible aperture for an optical
telescope because, if the traditional approach is taken to maintaining the figure of the
mirror as it points at different elevations and to preventing the telescope structure bending
under gravity, the cost of the telescope increases as a high power of its diameter, D, roughly
proportional to D 4 . The most important revolution in telescope design, which began in the
1980s, was the realisation that the cost–diameter relation can be profoundly changed if the
telescope and mirror are allowed to deform under gravity, but computer-controlled actuators
change the figure of the mirror and the pointing of the telescope so that the telescope always
remains in correct focus as it points to different parts of the sky. To express this in another
way, the cost of high-speed computers decreased so dramatically that it proved much more
cost effective to put the money into compensating for the floppiness of the mirror and
telescope by computer control, rather than by making them rigid structures. These concepts
were developed by ESO in the design and construction of the 4-metre New Technology
Telescope (NTT). The first 8–10-metre-class telescope to exploit similar concepts was the
Keck 10-metre telescope, which has a segmented mirror consisting of 36 hexagonal off-axis
mirrors which are computer-controlled and which provide excellent subarcsecond imaging
at the Mauna Kea site in Hawaii.
In 1987, the European Southern Observatory obtained approval for the programme to
construct the Very Large Telescope (VLT), to consist of four 8.2-metre telescopes located at
Cerro Paranal in the Atacama Desert in northern Chile. Their combined collecting aperture
is equivalent to a 16-metre optical–infrared telescope. The thickness of each 8-metre mirror
166 7 The new astronomies

was less than 20 cm and their figures were maintained by computer-controlled actuators in
the mirror cell. A similar approach was adopted by the Japanese 8-metre Subaru Telescope
located on Mauna Kea and the two 8-metre telescopes of the International Gemini Obser-
vatory, one located on Mauna Kea and the other at Cerro Pachon close to Cerro Tololo in
Chile. The latter project involves the USA, the UK, Canada, Brazil, Australia, Argentina and
Chile. Several other 8-metre telescope projects are currently approaching completion. These
telescopes are very complex systems involving a great deal of computer control. Likewise,
the astronomical instruments are very large and complex so that it is most effective if the
observations are made by the experts at the telescopes rather than in the traditional mode
in which the astronomer travels to the telescope to make the observations. The high cost of
observing time makes it much more cost effective to carry out the observing programme in
‘queue mode’, meaning that observations are planned to maximise the time the telescope
is taking astronomical data for a large number of approved programmes, rather than each
programme being allocated a specific observing period.
Another great advance has been a much deeper understanding of the phenomenon of
astronomical seeing and ways in which it can be minimised for observations with ground-
based telescopes. In the design of the new generation of 8–10-metre telescopes, many
precautions are taken to eliminate the effects of local seeing caused by the fact that the
telescope is located within a telescope dome. For example, the temperature of the mirror is
carefully controlled so that it is not a source of thermal convection cells and the telescope
domes have huge thermal vents so that, when observing, the telescope is essentially in the
open air (Figure 7.18). The net result of all these precautions is that the intrinsic seeing of
the new generation of large telescopes is about 0.4 arcsec, this residual figure being caused
by refractive index fluctuations in the upper layers of the atmosphere.
The next challenge facing the astronomical technologists has been to eliminate the effects
of astronomical seeing. Once local effects associated with the enclosure have been elim-
inated, the residual seeing at optical–infrared wavelengths is due to refractive index fluc-
tuations in the upper atmosphere which cause distortions of the wavefronts of the incom-
ing signals. The scale of these distortions is measured by Fried’s parameter, r0 , which
is the diameter over which the root-mean-square wavefront fluctuation is 1 radian (Fried,
1965). For a good astronomical site, r0 ∼ 15–20 cm at 500 nm, giving a seeing-limited
resolution of θ ≈ λ/r0 ∼ 0.6 arcsec. The Fried parameter increases with wavelength as
r0 ∝ λ6/5 so that the intrinsic seeing improves at longer wavelengths as λ−1/5 . The prin-
ciple of adaptive optics is to compensate for these wavefront distortions by measuring
them in real time and introducing compensating wavefront corrections in order to ‘flat-
ten’ the wavefront, thus obtaining diffraction-limited images. Although these principles
have been understood for a number of years, their effective implementation for astron-
omy only became a priority with the construction of the new generation of 8-metre tele-
scopes and the need to understand how to construct the next generation of 20–50-metre
telescopes.
Another important development in astronomical technology has been the development
of aperture synthesis techniques for optical and infrared wavelengths. The principles of
optical interferometry had been laid down by Michelson and implemented in his pioneering
measurements of the angular diameters of red giant stars (Michelson and Pease, 1921). In
7.8 Optical astronomy 167

Figure 7.18: A view of the interior of the Gemini North telescope enclosure at sunset showing the
fully open thermal vents and the fully open observing slit. (Courtesy of the International Gemini
Observatory.)

these observations, the visibility of the target star at different baselines was used to estimate
the angular size of the star. Michelson built a larger interferometer with a baseline of 50 feet
(15.2 m), but this was at the limit of mechanical stability (Pease, 1931).
The techniques of aperture synthesis at radio wavelengths were well understood by the
time Antoine Labeyrie (b. 1943) published his seminal paper in which he observed inter-
ference fringes from the bright star Vega using a pair of small telescopes separated by
12 metres (Labeyrie, 1975). These observations indicated how optical interferometry could
produce images of very high angular resolution, but there were many technical problems to
be overcome. The sky fluctuations, which cause the stars to twinkle, needed to be recorded
at kilohertz frequencies and required the development of sensitive photon-counting detec-
tors. In order to combine coherently the light from separated telescopes, micrometre-level
metrology of variable optical delay lines had to be constructed, which only became possible
when stabilised lasers became available. Finally, in order to produce images, the phases
as well as the amplitudes of the correlated signals had to be measured. This problem had
been solved by the radio astronomers involved in VLBI observations using the technique of
closure phases, which enables phases to be determined when many separate baselines are
available (Rogers et al., 1974). This procedure is now standard in radio interferometry and
is referred to as self-calibration (Pearson and Readhead, 1984). These technical challenges
required advanced control systems engineering, state-of-the-art detectors and high-speed
computation.
168 7 The new astronomies

100 100

relative declination (milliarcsec)

relative declination (milliarcsec)

50 50

0 0

−50 −50

−100 −100

100 50 0 −50 −100 100 50 0 −50 −100

relative R. A. (milliarcsec) (a) relative R. A. (milliarcsec) (b)

Figure 7.19: The first images from an optical aperture synthesis array made with three telescopes of
the COAST interferometer at a wavelength of 830 nm (Baldwin et al., 1996). These images of the
binary star system Capella were taken on (a) 13 September 1995 and (b) 28 September 1995. The
20 arcsec restoring beam is shown in the bottom left of each image.

From the mid 1980s onwards, John Baldwin (b. 1931) and his colleagues at Cambridge
used their experience of aperture synthesis at radio wavelengths to construct an optical
aperture synthesis array, COAST, which enabled these technical problems to be overcome.
Their first maps at an angular resolution of about 20 milliarcsec were published in 1996,
showing milliarcsecond motion of the stars of the binary system Capella over an interval
of 15 days (Figure 7.19) (Baldwin et al., 1996). Since that time, a number of imaging
optical–infrared aperture synthesis arrays have been developed and have produced important
scientific results. These include precise measurements of stellar diameters, orbital studies
of close binary stars enabling precise masses to be determined, brightness profiles across
stellar discs, providing tests of the theory of stellar atmospheres, and angular diameter
changes in pulsating stars, testing the theory of stellar pulsation. The next generation of
optical interferometers will be observatory-class facilities with the sensitivity to measure
features in active galactic nuclei on the scale of a milliarcsecond.29

7.8.3 Survey astronomy

Astronomical surveys of the whole sky are at the heart of many important studies because
they provide statistical information about the relative importance of different classes of stars
and galaxies and are also the means for discovering rare classes of astronomical objects
such as quasars. In all the new wavebands described in the previous sections, sky surveys
were among the most important priorities in opening them up for scientific exploration.
In the optical waveband, large-scale sky surveys were first carried out using very-wide-
field telescopes, which enable large regions of sky to be observed in a single exposure.
The widest field telescopes for all sky surveys were the Schmidt telescopes, which used an
innovative optical design invented by Bernhard Schmidt (1879–1935) in 1929 (Schmidt,
1931). Observations with this type of telescope were pioneered by Zwicky in the 1930s
7.8 Optical astronomy 169

(see Section 8.10). Immediately after the Second World War, a large Schmidt telescope
of effective aperture 1.2 metres (48 inches) was constructed to support observations made
with the 200-inch telescope at Mount Palomar. The size of each plate was 14 inches, corre-
sponding to about 6◦ on the sky and, over a period of eight years, this telescope completed a
photographic survey of the whole sky north of declination −20◦ in blue and red wavebands.
Photographic copies of the survey plates were made available to the worldwide community
of astronomers, and the resulting Palomar Sky Atlas proved to be an important research tool
for astronomers from all wavebands.
The northern hemisphere had a monopoly of large telescopes, and it was only in the
1960s that this imbalance began to be rectified. The European Southern Observatory and
the UK constructed Schmidt telescopes similar to the Palomar telescope in the southern
hemisphere to carry out the same type of survey as had been completed in the northern sky.
With the use of new emulsions, particularly the IIIaJ emulsions, these Schmidt telescopes
were able to reach significantly fainter magnitudes than the northern surveys. The surveys
also took about seven years to complete and have provided crucial databases for the whole of
astronomy, including an astrometric database for observations to be made with the Hubble
Space Telescope. These large sky surveys contain an enormous amount of statistical data
of importance for astronomy, but quantitative data could only be extracted if suitable high-
speed measuring machines were built for this purpose. UK astronomers took the lead in these
developments with the construction of the COSMOS High-Speed Measuring Machine at
the Royal Observatory, Edinburgh, and the Automatic Plate Measuring Machine (APM) at
Cambridge. These studies have provided many of the most important targets to be observed
by the 4-metre-class telescopes, for example in the discovery of large complete samples of
radio-quiet quasars.
While these surveys provided targets for large telescopes, they yielded little spectral
information. In particular, the redshifts of the galaxies and quasars found in these surveys
had to be determined individually. To overcome this problem, multi-object spectrographs
were developed to enable large numbers of spectra of faint objects to be obtained in a
single exposure. An excellent example of this approach was the 2◦ field (2dF) multi-object
spectrograph designed for the Anglo-Australian Telescope. The top end of the telescope was
redesigned to provide a 2◦ field of view and within that area 400 spectra of faint galaxies
and quasars could be measured simultaneously. The resulting 2dF Galaxy Redshift Survey
(2dFGRS) and the 2dF quasar survey were major spectroscopic surveys which took full
advantage of these unique capabilities. The 2dF survey obtained spectra for almost 250 000
objects, mainly galaxies, including about 25 000 quasars. These data have enabled a very
wide range of astronomical and cosmological questions to be addressed.
Even more ambitious is the Sloan Digital Sky Survey, which has been carried out by a
dedicated 2.5-metre telescope with a 3◦ field of view at the Apache Point Observatory, in
Sunspot, New Mexico. Unlike previous sky surveys in the optical waveband, the survey used
entirely digital CCD detectors to map about one-quarter of the whole sky. The survey uses
the technique of shifting the image on the detector electronically at the sidereal rate as the sky
moves over the telescope, a technique known as the Time-Delay and Integrate (TDI) mode.
This technique was first used to undertake quasar searches using the 200-inch telescope in
transit mode by Donald Schneider (b. 1955), Maarten Schmidt and James Gunn (Schmidt,
170 7 The new astronomies

Schneider and Gunn, 1986; Schneider, Schmidt and Gunn, 1994). The Sloan Survey detector
array consists of 30 2048 × 2048 CCD detectors which provide simultaneous observations
in five different wavebands. The telescope can also be operated in spectroscopic mode,
enabling 600 spectra to be taken in a single exposure. It is planned that the positions and
magnitudes of more than 100 million celestial objects will be determined. The telescope
will also measure the redshifts of up to a million galaxies, providing a three-dimensional
map of the universe, as well as measuring the redshifts of about 100 000 quasars. The results
of the Sky Survey are made available to the scientific community electronically, both as
images and in the form of catalogues.

7.9 Other types of astronomy

In addition to exploiting the unique properties of electromagnetic radiation for astrophysical

studies, other disciplines have already contributed to astrophysics and cosmology in impor-
tant ways. The development of cosmic ray physics was discussed in Section 7.2, and these
observations provide direct evidence for high-energy particles accelerated in cosmic envi-
ronments. In addition, neutrino studies of the Sun have provided information of importance
for particle physics and astrophysics and have become an important growth area. The sig-
nificance of neutrino astrophysics was reinforced by the remarkable discovery of neutrinos
from the supernova 1987A. Astroparticle physics has become a major growth area through
the attempt by laboratory physicists to detect the particles which may constitute the dark
matter in our Galaxy. Gravitational waves have been inferred to be emitted by the binary
pulsar PSR 1913+16, but gravitation waves themselves have not yet been detected directly.
The new generation of large gravitational-wave detectors is now attaining the sensitivities
at which a positive result can be reasonably expected in the near future.
The history of these developments will be told in their astrophysical context in Parts IV
and V.

Notes to Chapter 7

1 For more details of Hale’s remarkable contributions, see H. Wright, J. N. Warnow and C. Weiner,
eds, The Legacy of George Ellery Hale (Cambridge, Massachusetts: MIT Press, 1972).
2 From the historical perspective, the 200-inch telescope included a number of major advances
which were built into succeeding generations of large telescopes.
r The primary mirror was constructed of Pyrex r by the Corning Glass Works company because
of its low coefficient of expansion.
r The mirror was mass-reduced by creating a hexagonal cellular structure within the Pyrex r.
r The f -ratio of the telescope was reduced to f /3.3 to reduce the length of the telescope tube
and so reduce the size of the enclosure.
r The weight of the telescope was supported on oil-pads rather than floated in mercury.
r Serrurier trusses were used to maintain the separation of the primary and secondary mirrors.
The result was that, even when the telescope tube bent under gravity, the primary and secondary
mirrors remained parallel and aligned.
r The mirrors were coated with aluminium rather than silver, with the result that the mirror
needed to be recoated less frequently.
Notes to Chapter 7 171

3 Information on the development of astronomy internationally can be obtained from the

Proceedings of the General Assemblies of the International Astronomical Union, which have
been held at regular three-yearly intervals since 1922, except during the years of the Second
World War.
4 Röntgen’s paper was published in December 1895. It was also published, in English, in 1896 in
Nature, 53, 274–276.
5 An excellent documentary history of studies of cosmic rays is provided by A. Michael Hillas,
Cosmic Rays (Oxford: Pergamon Press, 1972).
6 In Dirac’s two great papers of 1928, he used the relativistic wave equation, known in his honour as
the Dirac equation, to show that the magnitude of the intrinsic angular momentum, or spin, of the
electron is given by h̄s(s + 1), where s = 1/2, that its magnetic moment is eh̄/m and that there
exist negative-energy solutions as well. There was considerable debate before the physical nature
of the negative solutions was understood. At first, they were thought to correspond to protons, but
they had to have the same mass as the electron. Only in 1931 did Dirac come down decisively in
favour of the interpretation that the negative solutions correspond to positively charged electrons,
the positrons (Dirac, 1931). Anderson discovered these particles in the following year.
7 The early history of radio astronomy has been surveyed in W. T. Sullivan III, ed., The Early Years
of Radio Astronomy (Cambridge: Cambridge University Press, 1984). Sullivan has also edited a
compilation of important early papers on radio astronomy: W. T. Sullivan III, Classics in Radio
Astronomy (Dordrecht: D. Reidel Publishing Company, 1982). An amusing account of some of
the personalities involved in the history of radio astronomy is contained in the conference volume:
K. Kellermann and B. Sheets, eds, Serendipitous Discoveries in Radio Astronomy (Green Bank,
Virginia: National Radio Astronomy Observatory Publications, 1983).
8 This paper reported the observation of intense radio emission associated with a large solar flare
which occurred on 27 and 28 February 1942. The information was declassified after the War.
Hey wrote his own account of the early history of radio astronomy in James Hey, The Evolution
of Radio Astronomy (New York: Science History Publications, 1973). He also wrote a touching
autobiographical account of these discoveries in his short book, James Hey, The Secret Man
(Eastbourne: Care Press, 1992).
9 See the discussion ‘The origin of cosmic radio noise’ at the Conference on Dynamics of Ionised
Media held in 1951 at University College, London.
10 The original 300-foot telescope, which operated in transit mode, collapsed in 1988 and has been
replaced by a new, fully steerable, 100-metre telescope which can operate at frequencies as high
as 100 GHz (3 mm wavelength).
11 The history of Martin Ryle’s discovery of Earth-rotation aperture synthesis is delightfully told by
Peter Scheuer in his article ‘The development of aperture synthesis at Cambridge’ in W. T. Sullivan
III, ed., The Early Years of Radio Astronomy (Cambridge: Cambridge University Press, 1984),
pp. 249–265.
12 The history of X-ray astronomy is told by W. Tucker and G. Giacconi, The X-ray Universe
(Cambridge, Massachusetts: Harvard University Press, 1985).
13 The history of space exploration is described in M. Rycroft, ed., The Cambridge Encyclopaedia
of Space (Cambridge: Cambridge University Press, 1990).
14 For the early history of ultraviolet observations of the Sun, see H. Friedman, Sun and Earth (New
York: Scientific American Library, 1986).
15 In fact, this fluorescent emission was observed by the ROSAT soon after its launch in 1991 (see
Figure 14.5(a)).
16 The development of γ -ray astronomy is described in P. Ramana Murthy and A. A. Wolfendale,
Gamma-ray Astronomy (Cambridge: Cambridge University Press, 1993).
17 Reviews of the recent history of ultra-high-energy γ -ray astronomy are given by F. A. Aharonian
and C. W. Akerlof, Gamma-ray astronomy with imaging atmospheric Cherenkov telescopes,
Annual Reviews of Nuclear Science, 47, 1997, 273–314, and M. Catanese and T. C. Weekes,
172 7 The new astronomies

Very high energy gamma-ray astronomy, Publications of the Astronomical Society of the Pacific,
111, 1999, 1193–1222.
18 A survey of planned Atmospheric Cherenkov Imaging Telescopes, many of which have now come
into operation, is given by B. L. Dingus, 26th International Cosmic-ray Conference, Conference
Proceedings 516 (Melville, New York: American Institute of Physics, 2000), pp. 351–364.
19 See R. W. Smith, The Space Telescope: A Study of NASA, Science, Technology and Politics (Cam-
bridge: Cambridge University Press, 1989), p. 30.
20 A comprehensive survey of the many areas of astronomy in which the IUE made major contribu-
tions is contained in Y. Kondo, ed., Exploring the Universe With the IUE Satellite (Dordrecht: D.
Reidel Publishing Company, 1987).
21 The history of the Hubble Space Telescope project is splendidly told by Smith, The Space Tele-
scope. The second edition (1994) includes a discussion of the problem of the spherical aberration
of the primary mirror and its solution using correction optics.
22 A splendid history of the trials and tribulations of the pioneers of infrared astronomy is given
in Chapter 6 of John Hearnshaw, The Measurement of Starlight: Two Centuries of Astronomical
Photometry (Cambridge: Cambridge University Press, 1996). An introduction to the post-War
history of infrared astronomy is given by D. A. Allen, Infrared: The New Astronomy (Shaldon,
Devon: Keith Reid Ltd, 1975).
23 Some of the very first astronomical images taken with infrared array cameras in the infrared
waveband are included in C. G. Wynn-Williams and E. E. Becklin, eds, Infrared Astronomy with
Arrays (Honolulu: Institute for Astronomy, University of Hawaii, 1987). The papers in this volume
indicate the technological challenges that had to be overcome to enable these array technologies
to be adapted for astronomical purposes.
24 The proceedings of the first symposium dedicated to the results of the IRAS mission were reported
in F. P. Israel, ed., Light on Dark Matter (Dordrecht: D. Reidel Publishing Company, 1985).
25 Good examples of the quality of the science achieved by the ISO mission are given in the review
by R. Genzel and C. Cesarsky, Extragalactic results from the Infrared Space Observatory, Annual
Reviews of Astronomy and Astrophysics, 38, 2000, 761–814.
26 The history of AURA’s involvement in these telescope projects is recounted by F. K. Edmondson
in AURA and its US National Observatories (Cambridge: Cambridge University Press, 1997).
27 The history of the Anglo-Australian Telescope is described in detail in the excellent book by
S. C. B. Gascoigne, K. M. Proust and M. O. Robins, The Creation of the Anglo-Australian Obser-
vatory (Cambridge: Cambridge University Press, 1990).
28 An excellent introduction to all types of electronic imaging detectors is given by I. S. McLean
in Electronic Imaging in Astronomy: Detectors and Instrumentation (Chichester: Wiley-Praxis
Series in Astronomy and Astrophysics, 1997).
29 An excellent review of the principles and status of optical aperture synthesis is given by
J. E. Baldwin and C. A. Haniff, ‘The application of interferometry to optical astronomical imag-
ing’, Philosophical Transactions of the Royal Society of London, A360, 2002, 969–986.
Part IV
The astrophysics of stars and
galaxies since 1945

Many astrophysicists and cosmologists refer to the years since 1945 as the ‘golden age’ of
astrophysics and cosmology. The areas pioneered before the Second World War began to
flourish vigorously, and completely new vistas were opened up as a result of the expansion
of the wavebands accessible for astronomical observation and of discoveries in fundamental
physics. The background to these developments was outlined in Part III, and, in Part IV, the
astrophysics of stars, the interstellar gas, galaxies, clusters of galaxies and high-energy astro-
physical phenomena are discussed. Part V is devoted to the achievements of astrophysical
cosmology.
8 Stars and stellar evolution

8.1 Introduction

By 1945, many of the physical processes involved in the evolution of stars on the main
sequence were beginning to be understood, but there remained an enormous amount of
detailed work to be undertaken before a precise comparison between theory and observation
could be made. To build detailed models of the stars, three types of data are required. The first
is the equation of state of the material of the star; the second are accurate nuclear reaction
rates; and the third is the opacity of stellar material for the transfer of radiation. These
quantities need to be known for the wide ranges of temperature and density encountered
inside the stars. Then, the problems of radiation transfer through the body of the star and
its surface layers have to be solved so that meaningful comparisons can be made between
the theory and observations. As a result, the astrophysicists had to have access to a very
wide range of data from nuclear, atomic and molecular physics, which began to become
available with the great expansion in the funding for the physical sciences after the Second
World War.
Then, there was the need to develop models for the evolution of stars from one region
of the Hertzsprung–Russell diagram to another. It was a daunting task, but there was light
at the end of the tunnel with the development of high-speed digital computers in the 1950s
and 1960s, which was to convert the study of the structure and evolution of the stars into
a precise astrophysical science. The new wavebands brought important new insights into
many of the key phases of stellar evolution using techniques which could not have been
imagined by the pioneers of the first half of the twentieth century.

8.2 Nucleosynthesis and the origin of the chemical elements

Two of the problems in the development of stellar astrophysics discussed above were closely
related. The first concerned the processes responsible for the synthesis of the chemical
elements, and the second related to the nuclear processes responsible for energy generation
once stars had moved off the main sequence. Although the CNO cycle could convincingly
account for the synthesis of helium in stars with masses greater than about 1M , how were
the carbon, nitrogen and oxygen created in the first place?
The big problem was that there are no stable isotopes with mass numbers 5 and 8. As
a result, there is no straightforward way in which protons, neutrons and α-particles can be

175
176 8 Stars and stellar evolution

added successively to helium nuclei as the first of a sequence of reactions which lead to
the formation of carbon. The solution was first proposed by Öpik in 1951 and was worked
out independently in more detail by Salpeter in 1952. They pointed out that, when the
central temperature of the star reaches about 4 × 108 K, the triple-α reaction, in which three
α-particles come together to form carbon, can take place (Öpik, 1951; Salpeter, 1952).
The process may be thought of as consisting of the formation of 8 Be, which, being highly
unstable, exists for a very short time before disintegrating into two α-particles; during that
time, however, there is a small probability that a third α-particle reacts with the 8 Be to form
12
C. There remained the problem that the cross-section for the reaction 8 Be + α →12 C was
too small to create a significant abundance of carbon.
The problem was solved by Fred Hoyle in 1953, who realised that the cross-section for
the interaction would be increased if there is a resonance associated with formation of 12 C in
an excited state (Hoyle, 1954). Hoyle estimated that the excited state of 12 C should occur at
about 7.7 MeV. This was a remarkable prediction in that, at that time, models of nuclei were
not sufficiently well developed that any resonance state of any nucleus could be predicted.
Ward Whaling (b. 1923) and his colleagues were persuaded to search for the resonance
and found it at exactly the energy predicted by Hoyle (Dunbar et al., 1953). Subsequent
experiments by William Fowler (1911–1995) and his colleagues in 1957 established the
details of the sequence of reactions. The energetics of the formation of 12 C are as follows:

2α + 94 keV → 8 Be, (8.1)

∗
8
Be + α + 278 keV → 12
C , (8.2)
∗
12
C → 12
C + 2γ + 7.654 MeV. (8.3)

The inclusion of the carbon resonance increased the cross-section for the formation of carbon
by the triple-α process by a factor of 107 . Hoyle went on to show that helium burning can
take place at a temperature of 108 K, the temperature deduced by Allan Sandage (b. 1926)
and Martin Schwarzschild (1912–1997) for the cores of red giant stars at the tip of the giant
branch (Sandage and Schwarzschild, 1952).
Öpik and Salpeter realised that once the carbon had been created, heavier elements,
such as oxygen and neon, could be created by the successive addition of α-particles. In his
paper of 1954, Hoyle went on to argue that, once the star had exhausted the helium in its
core, massive enough stars would continue to contract, increasing the central temperature
in the star so that the nuclear burning of 12 C into 24 Mg would take place and, at a slightly
higher temperature, 16 O would be converted into 32 S. The process of nuclear burning would
continue in massive enough stars until all the nuclear energy resources were used up, that
is when the core of the star consists of 56 Fe, the element with the greatest nuclear binding
energy of the chemical elements.
In 1956, Hans Suess (1909–1993) and Harold Urey published their detailed analysis of
the cosmic abundances of the elements (Suess and Urey, 1956). The primary sources for their
abundance determinations were the chondritic meteorites, which have abundances similar
to those of the photosphere of the Sun – it is commonly assumed that these meteorites have
preserved the primordial chemical composition out of which the Sun and the Solar System
were formed. These abundances were in reasonable agreement with the solar abundances
8.2 The origin of the chemical elements 177

Figure 8.1: A schematic curve of the abundances of the chemical elements as a function of atomic
weight based on the data of Suess and Urey (1956), who used relative isotopic abundances to determine
the slope and general trend of the curve. Burbidge and her colleagues drew special attention to the
overabundances associated with α-particle nuclei with atomic weights A = 16, 20 and 40, the peak at
the iron group, and the twin peaks at A = 80 and 90, at 130 and 138 and at 194 and 208 (Burbidge
et al., 1957).

and those of Population I stars. Suess and Urey not only worked out elemental abundances,
but also the isotopic abundances, which proved to be important discriminators of different
processes of nucleosynthesis. The abundances of the chemical elements fall off rapidly with
increasing atomic weight, but there are important features of the abundance curves which
provide clues to the processes of nucleosynthesis (Figure 8.1).
The nuclear processes involved in the synthesis of the elements were described in two
famous papers published in 1957, one by Margaret Burbidge (b. 1919), Geoffrey Burbidge
(b. 1925), Fowler and Hoyle, commonly known by the acronym B2 FH, and the other by
Alastair Cameron (1925–2005) (Burbidge et al., 1957; Cameron, 1957). The B2 FH paper
drew attention to the overabundance of the ‘α-particle’ nuclei such as those with 16, 20 and
32 nucleons, as well as to the iron-group elements and the peaks of stability at N = 50, 82
and 126. These peaks of stability corresponded to the ‘magic numbers’ of nuclear physics,
which had already been noted by Walter Elsasser (1904–1991) in 1933 (Elsasser, 1933). In
the B2 FH paper, eight nuclear processes by which the elements could be synthesised were
described. In addition to hydrogen burning, helium burning and the α-process, they drew
178 8 Stars and stellar evolution

Figure 8.2: Examples of the products of explosive nucleosynthesis from calculations by Arnett and
Clayton (1970). In these computer simulations, shells of carbon, oxygen and silicon were heated
rapidly to a very high temperature, as in a supernova explosion, and the nucleosynthesis takes place in
an expanding cooling shell. The peak temperatures reached were (a) 2 × 109 K in the case of carbon
burning, (b) 3.6 × 109 K in the case of oxygen burning and (c) 4.7–5.5 ×109 K in the case of silicon
burning. The circles represent the observed solar abundances and the crosses show the products of
explosive nucleosynthesis.

special attention to processes involving the addition of neutrons to pre-existing nuclei, the
slow (s) and rapid (r) processes. These reactions provide the means by which nuclei with
mass numbers greater than the iron group can be synthesised. If iron nuclei are found in
neutron-rich environments, they can absorb neutrons, and the process is termed ‘rapid’ or
‘slow’ depending upon whether or not the product nucleus decays before another neutron is
added to the nucleus. In the r-process, several neutrons are added before decay occurs. At
high enough temperatures and densities, the inverse β-decay process results in the formation
of large numbers of neutrons. Supernova explosions were identified as the sites in which
such reactions could take place. The s-process was believed to occur at an earlier stage in
the evolution of stars on the giant branch.
The general picture that emerged from these nucleosynthesis studies was that, the more
massive the star, the further it would proceed through the sequence of nuclear burning
before it collapsed to some form of dead star, such as a white dwarf or neutron star. Thus,
many of the abundant elements, such as carbon, oxygen and silicon, are synthesised through
steady nuclear burning, which occurs at an advanced stage of evolution on the giant branch.
Cameron drew particular attention to the importance of nucleosynthesis in supernova explo-
sions, a process that is now referred to as explosive nucleosynthesis. He realised that dif-
ferent chemical abundances are created if the process of nucleosynthesis takes place in a
non-stationary manner, as in the case of supernova explosions. With the development of
high-speed computers, it became possible to quantify these predictions. Although it was
not possible to simulate the explosion of complete stars, it was possible to carry out explo-
sive nucleosynthesis calculations for shells of particular elements. In 1970, David Arnett
(b. 1940) and Donald Clayton (b. 1935) showed how many of the element abundances could
be naturally attributed to explosive nucleosynthesis for shells of carbon, oxygen and silicon1
(Figure 8.2) (Arnett and Clayton, 1970).
 8.3 Solar neutrinos 179

8.3 Solar neutrinos

While the astrophysicists were refining their models of the Sun on the basis of the study of
its surface properties, Raymond Davis (b. 1914) suggested in 1955 that it might be possible
to search for the electron neutrinos liberated in the nuclear reactions which take place in the
CNO cycle (see Section 3.5) (Davis, 1955). This proposal was made before the neutrino,
predicted as long ago as 1934 by Fermi, was first measured experimentally in the laboratory
by Clyde Cowan (1919–1974) and Frederick Reines (1918–1998) in 1956 (Cowan et al.,
1956; Reines and Cowan, 1956). Because of their very small cross-sections for interaction
with matter, neutrinos escape essentially unimpeded from their point of origin within the
central 10% of the Sun by radius, and thus the detection of the flux of solar neutrinos
provides a direct test of the processes of nucleosynthesis. Davis proposed detecting the
solar neutrinos by the nuclear transformations which they would produce in a fluid which
contained a large number of chlorine atoms. Specifically, the nuclear reaction
37
Cl + νe → 37
Ar + e− (8.4)

has a threshold energy of 0.814 MeV. The argon created in this reaction is radioactive, and
the amount produced can be measured from the number of radioactive decays of the 37 Ar
nuclei. Unfortunately, as pointed out by Isadore Epstein (1919–1996) and Beverley Oke
(1928–2004), the p–p chain rather than the CNO cycle is the principal source of energy in
the Sun (Epstein, 1950; Oke, 1950). Neutrinos are, however, emitted in the p–p chain:

p + p → 2 H + e+ + νe : 2
H + p → 3 He + γ . (8.5)

The first reaction, in which deuterium is formed, is the principal source of neutrinos from the
Sun; these neutrinos are of low energy, however, the maximum energy being 0.420 MeV, and
so could not be detected by a chlorine detector. In 1958, Cameron and Fowler independently
pointed out that more energetic neutrinos are emitted in a side-chain of the main p–p chain
(Cameron, 1958; Fowler, 1958). There are three routes for the formation of helium, the
most straightforward and likely being the pp1 branch:

pp1 : 3 He + 3 He → 4 He + 2p. (8.6)

The other routes involve the formation of 7 Be as a first step:

3
He + 4 He → 7 Be + γ . (8.7)

Then, 7 Be can either interact with an electron (the pp2 branch) or, very rarely, a proton (the
pp3 branch) to form two 4 He nuclei:

pp2 : 7 Be + e− → 7 Li + ν : 7 Li + p → 4 He + 4 He, (8.8)

pp3 : 7 Be + p → 8 B + γ : 8 B → 8 Be∗ + e− + νe , (8.9)

8
Be∗ → 2 4 He. (8.10)

The electron neutrinos emitted in the decay of 8 B nuclei have maximum energy 14.06 MeV
and so could be detected in the type of experiment proposed by Davis.
180 8 Stars and stellar evolution

Figure 8.3: The observed flux of solar neutrinos from the 37 Cl experiment carried out by Davis and
his colleagues for the period 1970–1988. The solid line at 8 SNU is the expectation of the standard
solar model of Bahcall and Ulrich (Bahcall, 1989).

The first detailed predictions of the solar neutrino flux were made by John Bahcall (1934–
2005) in 1964 (Bahcall, 1964), and, at about the same time, the famous solar neutrino
experiment was begun by Davis and his colleagues using a 100 000 gallon tank of per-
chloroethylene C2 Cl4 located at the bottom of the Homestake gold-mine in South Dakota.
Davis found the neutrino flux to be significantly less than the value predicted by the solar
models. Indeed, by the time of their 1976 review, Bahcall and Davis found a positive signal
at only the one-sigma level (Bahcall and Davis, 1976). The predictions of the neutrino flux
improved over the years as the solar models were refined and the nuclear cross-sections
determined with greater accuracy.2 As the statistics improved over succeeding years, a sig-
nificant flux of neutrinos was detected but, over the 18 years illustrated in Figure 8.3, it
corresponded to only about one-quarter of the flux predicted by the standard solar models.
This discrepancy is the famous solar neutrino problem. The results quoted by Bahcall in
1989 were as follows:

observed flux of neutrinos : 2.1 ± 0.9 SNU, (8.11)

predicted flux of neutrinos : 7.9 ± 2.6 SNU, (8.12)

where 1 SNU = 1 Solar Neutrino Unit = 10−36 absorptions per second per 37 Cl nucleus
(Bahcall, 1989). The errors quoted are formal 3σ errors for both the observations and the
predictions.
8.3 Solar neutrinos 181

250
Ee ≥ 7.5 MeV
events/bin/590 days
200
590 days (June 1988–Apr.1990)
150

100

0
50

0
−1.0 −0.5 0.0 0.5 1.0
cos( Sun)
Figure 8.4: The distribution in cos(θSun ) for the 590-day sample for E e ≥ 7.5 MeV, where θSun is the
angle between the momentum vector of an electron observed at a given time and the direction of the
Sun. The isotropic background, which is roughly 0.1 events day−1 bin−1 , is due to spallation products
induced by cosmic ray muons, γ -rays from outside the detector and radioactivity in the detector water.
The angular resolution of the detector system has been taken into account in calculating the expected
distribution of arrival directions of the neutrinos from the Sun, which is indicated by the histogram
(Hirata et al., 1990).

The origin of this discrepancy became one of the most controversial topics in astrophysics
after Davis’s results were first reported in 1968. They suggested that there must be something
wrong either with the nuclear physics, or with the astrophysics of the Sun, or with both.3
Confirmation that the flux of high-energy neutrinos indeed originated within the Sun
was provided in 1990 by the Japanese Kamiokande II experiment, which had the great
advantage that the arrival directions of the incoming neutrinos are measured (Hirata et al.,
1990). The high-energy neutrinos scatter electrons which recoil with relativistic velocities.
The Cherenkov detectors which line the walls of the Kamiokande II experiment measure the
direction of travel of the scattered electrons, and thus the arrival directions of the neutrinos
can be found. The results of 590 days of observation are shown in Figure 8.4. There was
a small, but significant, excess flux of neutrinos coming from the direction of the Sun, but
it was less than that expected from the standard solar model of Bahcall and Roger Ulrich
(b. 1942) (Bahcall and Ulrich, 1988). As the Kamiokande II team stated:
These provide unequivocal evidence for the production of 8 B by fusion in the Sun.

The final results quoted by the Kamiokande II team from 1036 days of observations from
January 1987 to February 1995 were:
measured flux of neutrinos = 2.56 ± 0.16 (stat) ± 0.16 (syst), (8.13)
where (stat) refers to the statistical errors and (syst) refers to the systematic errors (Fukuda
et al., 1996).
182 8 Stars and stellar evolution

The Kamiokande II experiment was upgraded with an active volume of 32 000 tons of
pure water and renamed super-Kamiokande. The rate of detection of high-energy neutrinos
was greatly enhanced and, from 1258 days of observation, their flux was found to be:

measured flux of neutrinos = 2.32 ± 0.03 (stat) +0.008

−0.007 (syst), (8.14)

again in agreement with the earlier results and those of Davis (Fukuda et al., 2001).
A key test of the solar models is the detection of the low-energy neutrinos from the first
interaction, (8.5), of the p–p chain, since this is the essential first step in the synthesis of
helium and is much more directly related to the luminosity of the Sun than the high-energy
neutrinos. The best approach for measuring the much more plentiful low-energy neutrinos
is to use gallium as the detector material and to measure the neutrino flux from the number
of radioactive germanium nuclei created by the neutrino interaction:

νe + 71
Ga → e− + 71
Ge. (8.15)

During the early 1990s, two international collaborations, GALLEX and SAGE, reported the
results of these demanding experiments, which typically require the use of 30 tons of gallium
to produce a significant result. Over the period 1992 to 1997, the GALLEX collaboration
provided successively improved estimates for the flux of low-energy neutrinos from the
detector located in the Laboratori Nazionali del Gran Sasso in the Abruzzi region of central
Italy. The final result of the experiment was:

measured flux of neutrinos = 77.5 ± 6.2 SNU (8.16)

(Hampel et al., 1999), significantly less than the flux of 129 +8

−6
SNU expected from the
improved standard solar models of Bahcall and his colleagues (Bahcall et al., 1997). The
result reported by the SAGE experiment, located at the Baksan Neutrino Observatory in the
northern Caucasus mountains of Russia, was similar:

measured flux of neutrinos = 70.9 +5.3

−5.2
(stat) +3.7
−3.2
(syst) (8.17)

(Abdurashitov et al., 2002, 2003).

There had been a great deal of speculation about the solution of the solar neutrino prob-
lem, but, by the early 1990s, helioseismological experiments were beginning to demonstrate
that the standard models of the Sun were a remarkably precise description of its internal
structure right into the central nuclear-burning regions (see Section 8.4), and so the focus
of theorists began to centre upon the physics of the neutrino. In 1990, Bahcall and Bethe
proposed that the phenomenon of neutrino oscillations could account for the observed dis-
crepancy (Bahcall and Bethe, 1990). The phenomenon known as the MSW effect had been
discussed by Lincoln Wolfenstein (b. 1923) and by Stanislav Mikheyev (b. 1940) and Alexei
Smirnov (b. 1951) and involves physics beyond the standard model of particle physics,
in which neutrinos can change their type in the presence of matter (Wolfenstein, 1978;
Mikheyev and Smirnov, 1985). Thus, although the neutrinos created in the nuclear reac-
tions in the Sun are electron neutrinos, and this is the number predicted by the standard
solar models, as they propagate through the matter of the Sun, they can change into muon
and tau neutrinos. A consequence of this process is that the neutrinos have finite rest mass.
 8.4 Helioseismology 183

The test of this picture is to determine what fraction of the detected high-energy neutrinos
are electron neutrinos; this experiment has been carried out by combining the results of the
super-Kamiokande experiment with those of the Sudbury Neutrino Observatory (SNO)
located at Sudbury in Ontario, Canada. The SNO detector consists of 1000 tons of heavy
water, D2 O, and can be operated in a mode which is sensitive to the electron neutrinos alone,
whereas super-Kamiokande is sensitive to all types of neutrino. The SNO consortium found
that the electron neutrino flux was indeed significantly less than the flux quoted in equation
(8.14), that is

measured flux of νe = 1.75 ± 0.07 (stat) +0.12

−0.11
(syst) ± 0.05 (theory) (8.18)

(Ahmad et al., 2001). Using the ratio of these fluxes, they estimated that the total flux
of electron neutrinos emitted by the Sun is 5.44 ± 0.99 SNU, in good agreement with
the most recent estimates of the predicted neutrino flux by Bahcall and his colleagues.
This is undoubtedly one of the most remarkable discoveries of modern astrophysics and
demonstrates again the role of astrophysics in making discoveries which strike right to the
heart of fundamental physics. More recently, the same phenomenon of neutrino oscillations
has been measured in laboratory experiments (Eguchi et al., 2003).

8.4 Helioseismology

The study of the internal structure of the Sun remained the province of theoretical astro-
physics with little prospect of testing the theory directly until the discovery of solar oscil-
lations in the 1960s. These oscillations were first observed by Robert Leighton and his col-
leagues (Leighton, 1960; Leighton, Noyes and Simon, 1962) who discovered ‘five-minute’
oscillations in their studies of the velocity field of the solar atmosphere. The nature of these
oscillations was an unsolved problem throughout the 1960s. In a prescient paper of 1968,
Edward Frazier (b. 1939) suggested that the oscillations were trapped acoustic waves in the
outer layers of the Sun (Frazier, 1968). The first detailed analyses of the normal modes of
oscillation of the Sun were carried out by Roger Ulrich in 1970 and by John Leibacher
(b. 1941) and Robert Stein (b. 1935) in 1971 (Ulrich, 1970; Leibacher and Stein, 1971). The
‘five-minute’ oscillations were identified with standing acoustic waves confined to the outer
layers of the Sun. Frazier had made the first plot of the modes of oscillation of the Sun on a
frequency–wave number plot, but it showed no structure. The first analysis to show clearly
the ‘ridges’ in the dispersion relations for the different modes of oscillation of the Sun,
the (k, ω) diagram, was carried out by Franz-Ludwig Deubner (b. 1934) in 1975 (Deubner,
1975). He compared these results with predictions of improved models of the spectrum of
solar oscillations by Hiroyasu Ando (b. 1946) and Yoji Osaki (b. 1938) (Ando and Osaki,
1975) and found a disagreement between the observations and the theory. The discrepancy
was resolved by Douglas Gough (b. 1941) in 1978, who pointed out that the depth of the
convection zone in the standard models of the Sun had been underestimated by about 50%
(Gough, 1977). These studies revealed the great potential of this approach for investigating
the internal structure of the Sun.
184 8 Stars and stellar evolution

In 1976, Andrei Severny (1913–1987) at the Crimea Astrophysical Observatory and the
Birmingham group led by George Isaak (1933–2005) reported the discovery of oscillations
with much longer periods of 20–60 minutes and 160 minutes which had quite different
properties from the ‘five-minute’ oscillations (Brooks, Isaak and van der Raay, 1976; Sev-
erny, Kotov and Tsap, 1976). Oscillations of the diameter of the Sun were also reported by
Henry Hill (b. 1933) in 1976 as a by-product of his high-precision experiments to measure
the oblateness of the Sun (Brown, Stebbins and Hill, 1976).
It is no exaggeration to say that studies of the astrophysics of the Sun were revolutionised
by these discoveries. In terrestrial seismology, the resonance modes of the Earth can be found
by tracing the paths of sound waves inside the Earth. Exactly the same procedure can be
employed to study the resonant modes of the Sun, and this new astrophysical discipline was
named helioseismology. The theory of the modes of oscillation of the Sun is a beautiful
example of the power of classical mathematical physics applied to an astrophysical problem.
It is remarkable that many of the basic techniques of analysis, involving the description of
the normal modes of oscillation of the Sun in terms of associated Legendre polynomials, is
contained in the sixth edition of 1932 of the classic text Hydrodynamics by Horace Lamb
(1849–1934) (Lamb, 1932). The modes of oscillation of the Sun can be thought of as
standing waves resulting from the interference of oppositely directed propagating waves.
The modes of oscillation are of two types: acoustic or p modes, in which the restoring force
is provided by pressure gradients; and gravity or g modes, for which the restoring force is
buoyancy. The modes of greatest interest for the study of the internal structure of the Sun
are the acoustic modes of small degree, l, since these probe into the central nuclear burning
regions. The existence of these low-degree modes was deduced by Jörgen Christensen-
Dalsgaard (b. 1950) and Gough from the data published by the Birmingham group in 1979
(Christensen-Dalsgaard and Gough, 1980).
During the 1980s and 1990s, there were major observational campaigns to determine the
dispersion relations for the modes of oscillation of the Sun. Examples of the quality of the
data available at that time are shown in Figures 8.5(a) and (b). The first diagram shows
the (k, ω) diagram determined from four months of observation in 1986 and 1988 with the
helioseismology telescope at the Big Bear Observatory, showing the ‘ridges’ in the (k, ω)
relation associated with different modes of oscillation of the Sun (Woodard and Libbrecht,
1988). The error bars shown are 1000σ errors, giving some impression of the extraordinary
power of these techniques. The second is the power spectrum of the total luminosity of
the Sun obtained from the IPHIR experiment on board the PHOBOS space probe while
in transit between the Earth and Mars in 1988 and 1989 (Toutain and Frölich, 1992). The
power spectrum displays both large and small splittings of the spectral lines.
Ground-based campaigns continued through the 1980s and 1990s with projects such as
the GONG and Birmingham global networks of solar telescopes providing 24-hour coverage
of the solar oscillations. These studies were greatly advanced by the launch of the European
Space Agency’s Solar and Heliospheric Observatory (SOHO) in 1995, which provided an
ideal observatory for continuous observation of the Sun in space. The instrumental payload
included a Michelson Doppler Imager (MDI), which enabled the two-dimensional velocity
structure of the solar oscillations on the surface of the Sun to be imaged. These observations
enabled the three-dimensional internal structure of the Sun to be determined with remarkable
precision.
8.4 Helioseismology 185

4
3
frequency (mHz)
2
1

0 20 40 60 80 100 120 140

spherical harmonic degree, l (a)
14

12

10
ppm 2 / ( μHz)

8 l=1

6
l=2
4 l=0

0
2000 2500 3000 3500 4000
frequency (μHz) (b)

Figure 8.5: (a) The (k, ω) relation for the normal modes of oscillation of the Sun determined from
four months of observations in 1986 and 1988 with the helioseismology telescope at the Big Bear
Solar Observatory. The modes are p modes, and each ‘ridge’ contains modes with a fixed number of
radial nodes, n, the lowest frequency ridge having n = 1. The error bars shown are 1000σ (Woodard
and Libbrecht, 1988). (b) An example of the frequency spectrum of solar oscillations showing some
of the normal modes of oscillation of the Sun. These data were derived from 160 days of observation
by the IPHIR experiment on board the PHOBOS spacecraft (Toutain and Frölich, 1992). This power
spectrum of the low-degree p modes shows an alternating pattern of double and single peaks; the
double peaks are the l = 0, 2 modes and the single peaks the l = 1 modes.
186 8 Stars and stellar evolution

0.004

0.002
c 2/c 2

−0.002

0 0.2 0.4 0.6 0.8 1

r/R
Figure 8.6: A comparison of the theory and observations in terms of the fractional deviations of
the sound speed inferred from observation relative to the best-fitting standard model of the interior
structure of the Sun. The agreement is better than 0.2% throughout most of the Sun, except at the
boundary between the radiation and convective zones at 0.7R . (Courtesy of ESA and the SOHO
Science Team.)

A measure of the quality of data obtained from the SOHO measurements is provided
by the precision with which the speed of sound can be determined as a function of radius
within the Sun. As in the case of terrestrial seismology, the frequencies of the standing
waves depend upon the variation of the speed of sound as a function of depth within the
Sun. Modes of different order, l, sample different volumes of the solar interior, and so the
solar oscillation data can be inverted to determine how the speed of sound varies with radius
in the Sun. In Figure 8.6, the difference between the square of the speed of sound and the
predictions of a standard solar model are shown as a function of radius within the Sun. The
observations are in agreement with the predictions of the standard solar models, within a
few tenths of one percent.
One important feature is the prominent deviation from the predicted relation at about
70% of the solar radius. This region corresponds to the boundary between the inner regions
of the Sun, in which the transfer of energy is by radiation, and the outer regions, in which
energy is transferred by convection. This sharp feature has been associated with a turbulent
layer that results from the transition from radiative to convective energy transport. The
remarkable success of the standard solar models has enabled non-standard models, for
example those involving large amounts of core mixing, rapid rotation of the core of the Sun,
or the presence of weakly interacting massive particles in the core of the Sun, to be ruled
out.
8.4 Helioseismology 187

Figure 8.7: A cross-section through the Sun showing the internal relative rotation speeds of different
regions. Dark grey regions are slightly slower and the light grey regions are slightly faster relative to
the mean velocity distribution given by equation (8.19). (Courtesy of ESA and the SOHO Project.)

These results were of crucial importance in understanding the origin of the solar neutrino
problem, discussed in Section 8.3. The helioseismological data are now of such precision
that the astrophysics of the solar interior is very well defined and the origin of the deficit
of solar neutrinos must lie in some new type of neutrino physics, rather than in uncertain-
ties about the astrophysics of the temperature distribution inside the Sun, as discussed in
Section 8.3.4
The SOHO observatory also mapped the two-dimensional structure of oscillations on
the surface of the Sun, and this has enabled remarkable three-dimensional mapping of the
kinematics of its interior. An example of the results of this mapping is shown in Figure 8.7.
The dark grey and light grey regions indicate deviations from a mean angular velocity
distribution described by the expression

= 0 (1 − α2 cos2 θ − α4 cos4 θ ), (8.19)

where θ is the angle measured from the pole of the Sun; the constants α2 and α4 are both
about 0.14. It can be seen that, both on its surface and in its interior, the speed of rotation
is less at the poles than it is at the equator, relative to the above expression. In addition, the
pattern of rotation is different within the interior radiation-dominated zone and the exterior
188 8 Stars and stellar evolution

convection-dominated zone. Within the convective zone, there is a broad rapidly rotating
band, as well as polar convection currents, which are indicated by the black flow-lines.
These phenomena provide clues to the nature of the dynamo processes responsible for
maintaining the Sun’s magnetic field, one of the most challenging areas of contemporary
cosmic magnetohydrodynamics.
One of the most exciting prospects for the future study of the stars is the extension of
these techniques to the internal structures of nearby stars, the discipline known as astero-
seismology. Our understanding of other types of star will only become secure when similar
types of observation and analysis can be carried out on nearby stars, and this is already
within the grasp of present state-of-the-art technology.

8.5 Evolving the stars

Eddington’s pioneering studies of the internal structure of the stars (Eddington, 1926b)
and Hoyle and Lyttleton’s construction of homologous stellar models (Hoyle and Lyttleton,
1942) put the structure of main-sequence stars on a secure physical basis, but these were
no more than the first steps in what was to grow into a major industry in the second half of
the twentieth century. To do better than scaling relations, it was necessary to integrate the
equations of stellar structure through the star using the best available data on the opacities,
equations of state and nuclear energy generation rates for stellar material. In addition,
although the homologous models had finite radii, the boundary conditions at the surface of
the star needed careful attention. On top of these challenges, Öpik had come to the correct
conclusion that the solution of the red giant problem was to consider inhomogeneous stars,
which he argued must form when the nuclear energy sources were exhausted in the central
regions of the star (Öpik, 1938). The inevitability of this process was demonstrated by the
calculations of Schönberg and Chandrasekhar, who constructed models in which energy was
generated in a thin shell about a central inert core and found that there were no solutions when
the core constituted more than 10% of the mass of the star (Schönberg and Chandrasekhar,
1942).
The integration of the equations of stellar structure by numerical calculation was a
laborious procedure before the development of digital computers. As Martin Schwarzschild
remarked in 1958 (Schwarzschild, 1958),

A person can perform more than twenty integration steps per day . . . so that for a typical single
integration of, say, forty steps, less than two days is needed.

In 1952, Allan Sandage and Schwarzschild developed theoretical evolutionary tracks

for stars in globular clusters on the Hertzsprung–Russell diagram, stimulated by Sandage’s
important observations of this diagram for old globular clusters in which a continuous
distribution of stars from the main sequence onto the giant branch was observed (Arp, Baum
and Sandage, 1952). These computations demonstrated the rapid evolution of the star from
the main sequence to the giant branch, and they used the main-sequence termination point
to estimate the ages of the clusters, which turned out to be about 3 × 109 years (Sandage
and Schwarzschild, 1952). This has remained one of the most powerful tools for estimating
8.5 Evolving the stars 189

the ages of star clusters. The theoretical astrophysics of these studies was taken very much
further by Hoyle and Schwarzschild, who followed the detailed physics of the evolution of
stars from the main sequence to the tip of the giant branch, showing that, at this point, the
temperature of the core would be sufficiently high to initiate helium burning (Hoyle and
Schwarzschild, 1955).
The necessary atomic, molecular and nuclear data that were required to perform more
exact computations gradually became available as a result of the work of the national
laboratories and because of the military interest in understanding the physics of nuclear
explosions. For example, Arthur Cox (b. 1927) and his collaborators working at Los Alamos
National Laboratory carried out a major programme to determine the opacities of stellar
material using all the best available data, and these became the standard opacities used by
stellar modellers for many years (Cox, 1965).
By the late 1950s and early 1960s, electronic computers had become available to theo-
retical astrophysicists, and numerical procedures were developed to integrate the equations
of stellar structure accurately and efficiently. The pioneers of these numerical procedures
included Louis Henyey, Rudolph Kippenhahn (b. 1926), Icko Iben (b. 1931) and Robert
Christy (b. 1916). Thanks to their efforts, the study of the stars became one of the most pre-
cise of the astrophysical sciences. The issues which had to be addressed are clearly described
by Kippenhahn and Weigert in their authoritative book Stellar Structure and Evolution.5
The full complexity of stellar evolution once hydrogen burning in the centre of the star is
completed became apparent. All numerical studies showed the formation of red giant stars,
but the exact physical reason for this behaviour has been controversial since a number of
different processes take place almost simultaneously. The complexity of post main-sequence
evolution is most simply illustrated by the diagrams showing the evolution of a 5M star
by Kippenhahn and Alfred Weigert (b. 1927) (Figure 8.8). The key episodes in the star’s
history are as follows.

r From A to C, the energy source is hydrogen burning in the core, all the hydrogen in
this region being exhausted by the point C.
r Immediately following C, the core begins to collapse and hydrogen burning takes place
in a shell about the collapsing core. This phase is very rapid and the core collapse is
accompanied by a huge expansion in the radius of the envelope, a factor of 25 in this
simulation, resulting in the formation of a red giant star. The rapid evolution of the star
across this region of the Hertzsprung–Russell diagram accounts for the fact that very
few stars are found in this region, a phenomenon known as the Hertzsprung gap. A deep
outer convective zone is formed, corresponding to the star approaching the Hayashi
limit, which is the locus of fully convective stars (see Section 9.3). This process enables
mixing of the products of nucleosynthesis to take place in the outer envelope of the
star.
r In this massive star, the core is non-degenerate and so heats up until, at T ≈ 108 K,
the onset of helium core-burning begins at E. The process of nuclear helium burning,
creating carbon as well as oxygen and neon, continues around the loop E → F → G.
r Helium-core burning ends at G and then helium-shell burning takes over, with the result
that energy generation takes place in two nuclear burning shells.
190 8 Stars and stellar evolution

Figure 8.8: (a) The evolution of the internal structure of a 5M star, illustrating the nuclear processes
taking place in its interior. The abscissa shows the age of the model star after the ignition of hydrogen
in its core in units of 107 years. The ordinate shows the radial coordinate in terms of the mass, m, within
a given radius relative to the total mass, M, of the star. The cloudy regions indicate convective zones.
Heavily hatched areas indicate high rates of nuclear energy generation. (b) The corresponding positions
of the star on a Hertzsprung–Russell diagram at each stage in its evolution. From R. Kippenhahn and
A. Weigert, Stellar Structure and Evolution (Berlin: Springer-Verlag, 1990).

r Expansion of the outer envelope cools the hydrogen burning shell, which ceases nuclear
energy generation and a deep outer convective zone forms, enabling further mixing of
the processed material to take place in the envelope of the star. This process is referred
to as the second dredge up.
In the case of low-mass stars, the evolution is quite different. The core is in radiative,
rather than convective, equilibrium and so nuclear burning results in the growth of the
8.5 Evolving the stars 191

helium core. Once hydrogen core burning ceases, the core contracts to form an isothermal
degenerate core and nuclear energy generation takes place in a hydrogen-burning shell
about this inert core. The slow growth of the core and the gradual increase in the luminosity
of the star does not result in the rapid movement across the Hertzsprung–Russell diagram
seen in Figure 8.8(b). Furthermore, the low-mass stars are closer to the Hayashi limit as
they gradually move away from the main sequence. Ultimately, as the luminosity of the
hydrogen-burning shell increases, the temperature of the degenerate core reaches ≈108 K,
at which temperature helium burning takes place in a thermal runaway, known as the helium
flash, following which the degeneracy is relieved and a deep outer convection zone is formed
in the star’s envelope.
These features of the evolution of stars of different masses could only be understood in
detail once the numerical calculations had pointed the way to the physical understanding of
post-main-sequence evolution.6 One of the important developments was the understanding
of pulsating stars, such are the RR-Lyrae and Cepheid variables. Both classes of pulsating
star lie within a narrow band in the Hertzsprung–Russell diagram, which is known as the
instability strip, when stars of different masses evolve onto the horizontal branch. The first
serious analysis of the stability of stars and the problem of maintaining their pulsations was
carried out by Eddington in The Internal Constitution of the Stars (Eddington, 1926b). He
demonstrated that the pulsations would be damped out with a characteristic timescale of
8000 years, a figure which was subsequently shown to be a very significant overestimate.7
He proposed two mechanisms which would provide a driving mechanism for the pulsations,
the more important being a ‘valve’ mechanism associated with pulsations in the envelope
of the star. In its simplest form, the mechanism involves reducing the heat leakage during
the compression phase of the pulsation and increasing it during the expansion phase. It
was first proposed by Sergei Zhevakin (1916–2001) that the appropriate opacity changes
could be associated with the ionisation zones of ionised helium in the outer envelope of the
star in which He+ He++ (Zhevakin, 1953). Many studies, including those by John Cox
(1926–1984), Norman Baker (b. 1931) and Kippenhahn, demonstrated the correctness of
this picture, which can account for the pulsational properties of most types of pulsating star,
including the RR-Lyrae and Cepheid variable stars. The appropriate ionisation conditions
are found in the envelopes of stars of different masses as they cross the instability strip
when they have joined the horizontal branch. The process of pulsation is inherently non-
linear, and powerful computer codes were needed to determine the light-curves of the
stellar pulsations. Robert Christy successfully developed numerical codes to account for
the observed properties of RR-Lyrae variable stars (Christy, 1964), and he and Robert Stobie
(1941–2002) developed the corresponding models for Cepheid variable stars (Christy, 1968;
Stobie, 1969).
These very considerable achievements made stellar evolution computations a key tool
for the understanding of the evolution of the stellar populations of galaxies. The ages of star
clusters, for example, could be estimated from the forms of their Hertzsprung–Russell dia-
grams. There were, however, inconsistencies which a number of authors suspected might be
associated with uncertainties in the stellar opacities and the equations of state. In the 1980s,
two large programmes, the OPAL project and the Opacity Project (OP), were undertaken to
improve knowledge of the opacities used in stellar structure and evolution calculations. The
192 8 Stars and stellar evolution

OPAL code was developed at the Lawrence Livermore National Laboratory to compute
opacities of elements with low to mid atomic numbers and included a very large num-
ber of improvements as compared with the earlier opacities (Rogers and Iglesias, 1994).8
The approach adopted by the Opacity Project, an international consortium led by Michael
Seaton (b. 1923), was to calculate opacities based on a new formalism for the equation of
state and on the computation of accurate atomic properties such as energy levels, f -values
and photoionisation cross-sections (Hummer and Mihalas, 1988; Seaton et al., 1994).
These massive programmes resulted in much improved opacities for all types of astro-
physical application. Interestingly, the two projects took quite different approaches to the
determination of the opacities, but they ended up being in very good agreement. As Forrest
Rogers (b. 1938) and Carlos Iglesias (b. 1951) described in their review paper of 1994, the
new opacities were generally somewhat larger than the previous values, much of the increase
being associated with transitions of iron, and this had the effect of resolving a number of
astrophysical puzzles.
r Some Cepheid variables pulsate in different modes at different frequencies, and the
ratios of these frequencies are sensitive to the mass of the star. Using the old Los
Alamos opacities, the deduced masses were up to about 50% smaller than expected.
Using the new opacities, this discrepancy has been removed and the masses agree with
those expected from stellar evolution from the giant branch without mass loss. Similar
results are found for other classes of Cepheid variable.
r A similar discrepancy was found for the RR-Lyrae stars. It was thought that the new
opacities would have only a very small effect because these stars are metal-poor. Despite
this, the new opacities result in masses which are now consistent with the masses
expected from the theory of stellar evolution in globular clusters.
r The new opacities provide better agreement with the p-mode frequencies observed in
helioseismic experiments. The resulting changes in the sound speed deduced from the
solar models now place the inner boundary of the outer convective zone at 0.713 ± 0.003
of the solar radius.
r The new opacities have also contributed to understanding the lithium depletion problem
observed in the Sun and hotter stars. The new opacities increase the depth of the outer
convective zone so that the 7 Li is exposed to temperatures T ≈ 2.4 × 106 K at which
the reaction 7 Li + p → 2 4 He can take place.

8.6 The discovery of neutron stars

The existence of neutron stars had been predicted as long ago as 1934 by Baade and Zwicky
(1934a), soon after the discovery of the neutron, but models of neutron stars suggested
that the only detectable emission would be the thermal X-ray radiation from their sur-
faces. Neutron stars were expected to be very hot on formation but, after an initial period
of rapid cooling by neutrino emission, they remain very hot and cool slowly by thermal
emission from their surfaces. Because they are very compact stars, with radii only about
10–20 km, it was believed that the only chance of observing them would be as weak X-ray
emitters. In a prescient paper of 1967, Franco Pacini (b. 1939) predicted that they might
8.6 The discovery of neutron stars 193

be observable at long radio wavelengths if they were magnetised and were oblique rotators
(Pacini, 1967).
The neutron stars were discovered, more or less by chance, as the parent bodies of radio
pulsars by Antony Hewish (b. 1924) and Jocelyn Bell (b. 1943) in 1967 at Cambridge, UK
(Hewish et al., 1968).9 Hewish had pioneered the technique of observing the scintilla-
tion or ‘twinkling’ of radio sources at long radio wavelengths due to irregularities in the
plasma density along the lines of sight to the radio sources. His work in this area had been
fundamental in understanding the influence of irregularities in the electron density in the
ionosphere upon radio astronomical observations. By the early 1960s, it became apparent
that the same technique could be used to study the properties of the interplanetary plasma,
the phenomenon of interplanetery scintillation. It turned out that this technique could be
used, not only as a means of studying density fluctuations in the interplanetary plasma and
their motions, but also as a means of discovering radio quasars, many of which are compact
radio sources and hence expected to display large radio scintillations at low frequencies.
Hewish designed a large array to undertake these studies and was awarded a grant of
£17 286 by the Department of Scientific and Industrial Research to construct it, as well
as outstations for measuring the velocity of the solar wind. To obtain adequate sensitivity
at the low observing frequency of 81.5 MHz (3.7 m wavelength), the array had to be very
large, 4.5 acres (1.8 hectares) in area, in order to record the rapidly fluctuating intensities of
bright radio sources on a timescale of one-tenth of a second. This was the key technological
development which led to the discovery of radio pulsars since normally radio astronomical
observations require long integrations to detect faint sources.10
The first sky surveys began in July 1967, and Jocelyn Bell, Hewish’s research student,
discovered a strange source which seemed to consist entirely of scintillating radio signals
(Figure 8.9(a)). The source was not always present, and its nature remained a mystery. In
November 1967, the source reappeared and was observed using a receiver with a much
shorter time-constant. It was found to consist entirely of a series of pulses with a pulse
period of about 1.33 s (Figure 8.9(b)). This pulsating radio source, PSR 1919+21, was the
first to be identified, and over the next few months three further sources were discovered
with pulse periods ranging from 0.25 to almost 3 seconds. The name pulsar was coined
soon after the announcement of the discovery.11
Within a year, more than 20 more pulsars were discovered and a flood of theoretical papers
appeared. The favoured picture for the nature of the pulsar phenomenon was described by
Thomas Gold (1920–2004) in 1968, similar in many respects to Pacini’s proposal of 1967,
and consisted of an isolated, rotating, magnetised neutron star in which the magnetic axis
of the star and its rotation axis are misaligned (Pacini, 1967; Gold, 1968). The radio pulses
were assumed to originate from beams of radio emission emitted along the magnetic axis.
The key observations that supported this picture were the very short stable periods of the
pulses and the observation of polarised radio emission within the pulses. Two of the pulsars
discovered in 1968 were of special importance. A pulsar of period 0.089 s was discovered in
the young supernova remnant in the constellation of Vela (Large, Vaughan and Mills, 1968),
and soon afterwards a pulsar with the very short period of 0.033 s was discovered in the
centre of the Crab Nebula, the supernova which exploded in 1054 and which was extensively
studied by the Chinese astronomers at that time (Staelin and Reifenstein, 1968). Optical
194 8 Stars and stellar evolution

(a)

(b)
Figure 8.9: The discovery records of the first pulsar to be discovered, PSR 1919+21 (Hewish et al.,
1968; see also note 10). (a) The first record of the strange scintillating source labelled CP 1919. Note
the subtle differences between the signal from the source and the neighbouring signal due to terrestrial
interference. (b) The signals from PSR 1919+21 (top trace) observed with a shorter time-constant
than the discovery record, showing that the signal consists entirely of regularly spaced pulses with
period 1.33 s. The lower trace shows one-second time markers.

pulses from the Crab Nebula pulsar with precisely the same pulse period were discovered in
1969 (Cocke, Disney and Taylor, 1969), and within three months rocket flights by the teams
from the Naval Research Laboratory (Fritz et al., 1969) and the Massachesetts Institute of
Technology (Rossi, 1970) had discovered X-ray pulses as well.12
The short periods of these pulsars proved beyond any shadow of doubt that the parent
bodies of the pulsars had to be neutron stars. By pushing all the parameters to their limits,
it had just been possible to find models of white dwarfs which could rotate with periods
of about 1 s before they would break up, but periods as short as 0.1 s were excluded. Fur-
thermore, the formation of neutron stars in supernova explosions was conclusively demon-
strated by the coincidence of these short-period pulsars with young supernova remnants.
8.6 The discovery of neutron stars 195

One further prediction made by Gold was soon confirmed observationally. He had noted
that a spinning magnetic dipole emits magnetic dipole radiation and that this energy is
extracted from the rotational energy of the neutron star. Consequently, the periods of pul-
sars should increase as they radiate away magnetic dipole radiation. In 1969, the period of
the pulsar in the Crab Nebula was found to be increasing steadily with time (Richards and
Comella, 1969).
This result solved another long-standing problem in understanding the physics of the
Crab Nebula. The polarisation and non-thermal spectrum of the optical radiation from the
Crab Nebula had been interpreted as the synchrotron radiation of high-energy electrons
(Shklovsky, 1953). The problem with this interpretation was that the high-energy electrons
would lose all their energy in a time short compared with the age of the nebula. The problem
was exacerbated by the fact that the X-ray emission from the nebula was attributed to the
same mechanism and the lifetimes of these electrons were even shorter. In consequence,
some means had to be found of providing a continuous supply of energy to the nebula.
In 1969, Gold pointed out that energy was being supplied to the nebula by the slowing
down of the pulsar. The rate of loss of rotational energy by the pulsar turned out to match
almost precisely the rate at which energy had to be supplied to the nebula to account for the
non-thermal radio, optical and X-ray emission (Gold, 1969).
In the same year, the first papers exploring the electrodynamics of pulsars were published
by Peter Goldreich (b. 1939) and William Julian (b. 1939) and by James Gunn (b. 1938) and
Jeremiah Ostriker. In 1968, Pacini had shown that the magnetic field strengths at the surfaces
of neutron stars had to be enormous, B ∼ 106 −108 T (Pacini, 1968). These magnetic fields
were so strong that the Lorentz (v × B) force extracted electrons from the surface layers of
the neutron star so that electric currents must be present in its magnetosphere (Goldreich and
Julian, 1969). Ostriker and Gunn showed how electrons could be accelerated to very high
energies in the strong electromagnetic waves emitted by the rotating magnetised neutron
star (Ostriker and Gunn, 1969).
Studies of the internal structure of neutron stars were pursued with renewed vigour as
theorists realised that they were no longer theoretical ornaments but an integral part of
observational astrophysics. The equation of state of neutron matter had been the subject
of numerous studies prior to the discovery of pulsars. An equation of state had been given
by Kent Harrison (b. 1934) and John Wheeler (b. 1911) in 1958 (Harrison, Wakano and
Wheeler, 1958) and in convenient analytic forms by Harrison and his colleagues in 1965
(Harrison et al., 1965). Much improved calculations were carried out by Gordon Baym
(b. 1935), Hans Bethe and David Pines (b. 1924) in 1971, and these were used to construct
the standard models for neutron stars (Baym, Bethe and Pethick, 1971a; Baym, Pethick and
Sutherland, 1971b). The equation of state was well understood up to about nuclear densities,
ρ ∼ 3 × 1017 kg m−3 , but there remained uncertainties at higher densities which may be
found in the centres of the most massive neutron stars.
In addition, the interior is threaded by an intense magnetic field. The magnetic field does
not have a strong influence upon the structure of the neutron star, but it does influence its
internal dynamical properties. Long before the pulsars were discovered, Arkadii Migdal
(1911–1993) in 1959, as well as Vitali Ginzburg and David Kirzhnits (1926–1998) in 1964
and Vittorio Canuto (b. 1937) and Györgi Marx13 in 1965, proposed that the interiors of
196 8 Stars and stellar evolution

neutron stars should be superfluid (Migdal, 1959; Ginzburg and Kirzhnits, 1964). Studies of
the properties of the neutron–proton–electron fluid in different regimes inside the neutron
stars showed that the inner crust and the neutron liquid phases are superfluid and that
the protons are superconducting. In 1969, Baym, Christopher Pethick (b. 1942) and Pines
showed that the magnetic field is therefore quantised into vortices which are pinned to the
crust of the neutron star (Baym, Pethick and Pines, 1969a).
The link between the interior structure of the star and observable phenomena was provided
by the phenomenon of glitches, which were discovered in the Vela pulsar in 1969 (Radhakr-
ishnan and Manchester, 1969; Reichley and Downs, 1969). All radio pulsars were known to
slow down, but, occasionally, discontinuous changes in the slow-down rate were observed
in which the period decreased abruptly. A natural interpretation of this phenomenon was
to associate it with internal changes in the structure of the neutron star. One possibility
proposed by Malvin Ruderman (b. 1927) and by Baym, Pethick and Pines in 1969 was that,
as the neutron star slows up, its crust takes up a new shape in a ‘star-quake’ in which the
structure changes discontinuously to take up a new equilibrium configuration (Baym et al.,
1969b; Ruderman, 1969). This process cannot, however, be the whole story since, in pulsars
such as the Vela pulsar, the glitches occur too frequently. The preferred interpretation is
that the glitches may be associated with the unpinning of the superconducting vortices with
the crust of the neutron star. One important aspect of the glitches is that the recovery of the
rotation speed to a steady value is remarkably slow. This is direct observational evidence
that the bulk of the moment of inertia of the neutron star must be in some superfluid form,
which is only weakly coupled to the crust. Thus, the significance of these observations goes
far beyond astronomy since this is the only environment we know of in which matter in bulk
can be studied at nuclear densities.
Intriguingly, one of the most difficult problems of pulsar studies was the understanding
of the mechanism of radio emission. The brightness temperature of the radio pulses often
exceeded 1028 K, indicating that the emission must involve some form of coherent radiation
mechanism. The most promising model was proposed by Venkataraman Radhakrishnan
(b. 1929) and his colleagues in 1969 and involved the radiation of high-energy electrons
as they streamed from the poles of the magnetised neutron star along the curved field
lines (Radhakrishnan and Cooke, 1969; Radhakrishnan et al., 1969). The radiation process,
known as curvature radiation, is similar in many ways to synchrotron radiation, but the
radius of curvature of the particles’ trajectories is defined by the curvature of the field lines
emerging from the poles of the neutron star and must involve coherent bunches of electrons to
produce the extraordinarily high brightness temperatures. This model can naturally account
for the observed variation of the direction of the linear polarisation of the radiation as the
beamed radiation sweeps past the observer.
At the 1970 IAU General Assembly, the distinguished Soviet astrophysicist Iosef
Shklovsky asked me to introduce him to Jocelyn Bell. He told her, ‘Miss Bell, you have
made the greatest astronomical discovery of the twentieth century.’ In many ways, it is
difficult to disagree with his assessment. The existence of these stars shows that neutron
stars form as a result of stellar collapse. Furthermore, these are the last stable stars. The
radius of a solar mass neutron star is only about three times the radius of a black hole of the
same mass, and so the neutron stars may be thought of as objects which have just failed to
 8.7 X-ray binaries and the search for black holes 197

become black holes. Even in neutron stars, general relativity is no longer a small correction
factor but is crucial in determining their stability.

8.7 X-ray binaries and the search for black holes

The next event, which was to have a profound influence upon thinking in high-energy
astrophysics, was the discovery of binary X-ray sources by the UHURU satellite in 1971.14
From previous sounding rocket experiments, it had been suspected that some of the X-ray
sources were variable in intensity because sometimes they were present and at other times
they seemed to have faded or disappeared. To resolve these problems, it was necessary to
carry out long-term systematic studies of these sources. The UHURU X-ray observatory
was the first satellite dedicated exclusively to X-ray astronomy, and it carried out the first
complete survey of the whole sky. The nature of the variability of the sources was one of
the key objectives of the mission. In late 1970, the variable source Cygnus X-1 (Cyg X-1)
was observed and, although for some time it was thought that it had an X-ray periodicity
of 0.073 s, this result was spurious and the source displayed somewhat random variability
on timescales as short as 100 ms, indicating that the source region must be very compact
(Oda et al., 1971; Rappaport, Doxsey and Zaumen, 1971).
Much more remarkable were the observations of the source Centaurus X-3 (Cen X-3)
which were first made in January 1971. Unlike the case of Cyg X-1, Cen X-3 showed a clear
periodicity with a pulse period of about 5 s, longer than that of any known radio pulsar.
Furthermore, the pulsation period was not stable but seemed to vary with time (Giacconi
et al., 1971a). The source was reobserved in May 1971, and it was found that the period
of the X-ray pulsations varied sinusoidally with a period of 2.1 days. This suggested that
the X-ray source was a member of a binary system, the change in period of the pulses
being due to the Doppler shift of the X-ray pulses in the binary orbit. Then, on 6 May, the
source disappeared, only to reappear half a day later. This pattern repeated roughly every
two days. Clearly, the X-ray source was being occulted by the primary star in the binary
system (Schreier et al., 1972). With these clues, Wojciech Krzeminski (b. 1933) was able to
identify the primary star, which turned out to be a massive blue star with the same binary
period of 2.1 days as the X-ray source (Krzeminski, 1973, 1974). Soon after this discovery,
another similar source was discovered (Tananbaum et al., 1972), the source Hercules X-1
(Her X-1), which had a pulse period of 1.24 s and an orbital period of 1.7 days (Figure 8.10).
The short period of the X-ray source in Her X-1 was strong evidence that the parent
body must be a neutron star, similar to those of the radio pulsars. Furthermore, the source
of energy for the system was immediately identified as accretion. The idea of accretion as
a source of energy for the X-ray sources had already been suggested by Satio Hayakawa
(1923–1992) and Masaru Matsuoka (b. 1939) in 1964 (Hayakawa and Matsuoka, 1964),
who considered normal close binary systems, and by Shklovsky in 1966 (Shklovsky, 1967),
who proposed accretion from a binary companion onto a neutron star as the energy source
for the brightest X-ray source in the sky, Sco X-1. In 1968, Kevin Prendergast (b. 1929)
and Geoffrey Burbidge made the important point that, in the accretion of matter from
the primary star onto a compact secondary in a binary system, the accreted matter would
198 8 Stars and stellar evolution

Figure 8.10: The discovery records of the pulsating X-ray source Her X-1. The histogram shows the
number of counts observed in successive 0.096 s bins. The continuous line shows the best-fitting
harmonic curve to the observations, taking account of the varying sensitivity of the telescope as it
swept over the source (Tananbaum et al., 1972).

necessarily have a considerable amount of angular momentum and so an accretion disc

would form about the compact star (Prendergast and Burbidge, 1968). Accretion of matter
from the primary star onto a compact neutron star is a very powerful energy source. A simple
Newtonian calculation shows that the luminosity due to accretion onto an object of mass M
and radius r is roughly 0.5 ṁc2 (rg /r ), where rg = 2G M/c2 is the Schwarzschild radius of
an object of mass M and ṁ is the mass accretion rate (see Section A11.1.2). According to
this simple estimate, the accretion of matter onto a 1M neutron star with radius 10 km can
liberate about 5–10% of the rest mass energy of the infalling matter. When the effects of
general relativity are taken into account, the upper limit to the energy release is 5.72% for
accretion onto a solar mass black hole, roughly an order of magnitude greater than can be
liberated by nuclear fusion reactions.15
The study of pulsating X-ray binaries was of particular importance astrophysically
because the masses of the neutron stars could be estimated using the standard procedures
of celestial mechanics. The gratifying result was found that the masses of the seven binary
X-ray sources for which this analysis was possible lay in the range 1.2 to 1.4 M , entirely
consistent with the upper limit to the masses of neutron stars, which is similar to the
Chandrasekhar limit for white dwarfs (Rappaport and Joss, 1983).
Distances could be estimated for many of the X-ray sources and so their X-ray lumi-
nosities could be found. In 1973, Bruce Margon (b. 1948) and Ostriker showed that the
luminosities of the binary sources extended up to about L = 1031 W, which is very close
to the Eddington limiting luminosity for spherical accretion onto objects with mass 1M ,
the precise limit being L ≤ 1.3 × 1031 (M/M ) W (Margon and Ostriker, 1973). This was
8.7 X-ray binaries and the search for black holes 199

a key result because it demonstrated that sources existed which could radiate X-rays at
luminosities close to the maximum permissible luminosity (see Section A11.1.3). Further-
more, it was natural that these sources should emit most of their radiation in the X-ray
waveband. Assuming that the X-ray emission originates close to the surface of the neutron
star, application of the Stefan–Boltzmann law showed that the temperature of the emitting
region had to be greater than about 107 K to produce such large luminosities, and so it was
natural that the energy of accretion should be emitted in the X-ray waveband.
From these considerations, a standard picture of the nature of the pulsating binary X-
ray sources developed.16 In the case of a massive companion, such as a blue supergiant
star, the neutron star is located in a strong stellar wind, and the matter within a certain
radius of the neutron star is accreted onto it. In the low-mass binary systems, mass transfer
takes place through the process of Roche lobe overflow, in which the primary star fills its
Roche lobe, the equipotential surface joining the two stars, and so matter attains a lower
gravitational potential by collapsing to form an accretion disc about the neutron star.17 The
X-ray pulsations are attributed to accretion onto the poles of the rotating neutron star, the
strong non-aligned magnetic field channelling the matter into the polar regions. Evidence
for the presence of strong magnetic fields in the source Her X-1 was found by Joachim
Trümper (b. 1933) and his colleagues in 1978; they identifed a cyclotron radiation feature
in its X-ray spectrum at about 58 keV (Trümper et al., 1978). This feature was subsequently
observed by the Japanese GINGA satellite and was interpreted as an absorption feature at
34 keV, corresponding to a magnetic field strength of 3 × 108 T (Mihara et al., 1990).
The next obvious step was to ask whether or not there was any evidence for black holes
among the binary X-ray sources. Isolated black holes are very difficult to detect, and it is
only when they are close to sources of fuel that their presence can be readily detected. In
1965, Yakov Zeldovich and Oktay Guseynov (b. 1938) had proposed that the observation of
X-rays or γ -rays from single-line spectroscopic binaries might be the signature of either a
neutron star or a black hole (Zeldovich and Guseynov, 1966). In these binary systems, only
the bright component is observable, and the unseen secondary must be some form of dark
star, either a very-low-mass star, a neutron star or a black hole. In 1969, Virginia Trimble
(b. 1943) and Kip Thorne (b. 1940) investigated whether or not any of the dark companions
of known single-line spectroscopic binaries could be massive enough to be black holes, but
no likely candidate was found and none coincided with known X-ray sources (Trimble and
Thorne, 1969).
The first strong candidate for a black hole companion was found in the bright X-ray
source Cyg X-1. The positions of sources determined by the UHURU observatory were not
normally accurate enough to make an identification of the optical counterpart unless there
was some identifiable feature, such as a binary orbital period. In 1971, with an improved
position provided by an MIT rocket flight, radio astronomers at the National Radio Astron-
omy Observatory in the USA and at the Westerbork Observatory in the Netherlands searched
the field for a variable radio source which might be associated with Cyg X-1. These searches
were successful, and resulted in an accurate radio position which coincided with the 9th-
magnitude blue supergiant star (Braes and Miley, 1971; Hjellming and Wade, 1971). In
the next year, Louise Webster (1941–1990) and Paul Murdin (b. 1942) and, independently,
Thomas Bolton (b. 1943) showed that the star was the primary star of a binary system with
200 8 Stars and stellar evolution

Figure 8.11: The mass distribution of neutron stars and black holes. Those systems which are known
to possess neutron stars all have masses about 1.4M . The masses of the black holes in X-ray binary
systems for which good mass estimates are available all exceed the upper limit for neutron star masses
of M ≈ 3M (Charles, 1998).

period 5.6 days (Bolton, 1972; Webster and Murdin, 1972). These observations enabled
the ratio of the masses of the two components to be estimated. Assuming the mass of the
supergiant B star was greater than 10M , the mass of the invisible companion had to be
greater than 3M , the most likely masses being 20M for the blue supergiant and 10M for
the unseen companion. The latter mass exceeded the upper limit for stability as a neutron
star, and it was concluded that it must therefore be a black hole.
Over the succeeding years, three other good examples of X-ray binaries with massive
invisible companions were discovered: the X-ray binary sources LMC X-1, LMC X-3 and
A0620-00 (Cowley, 1992; McClintock, 1992). In each of these, the X-ray intensity exhibits
short-period variability but no signature of pulsed X-ray emission. In a recent compilation
of neutron star and black hole mass estimates, the numbers of black hole candidates has con-
tinued to increase, with the black hole candidates having quite different X-ray spectral and
variability characteristics as compared with those cases in which pulsating X-rays are found
and which are confidently associated with accreting neutron stars (Figure 8.11) (Charles,
1998). It is now generally accepted that the simplest interpretation of the properties of these
systems is that they contain black holes. Their great advantage is that they are relatively
nearby systems and so the behaviour of matter in the vicinity of the strong gravitational
fields associated with black holes can be studied in some detail. There will be much more
to say about the astrophysics of black holes in Chapter 11 in the context of the physics of
active galactic nuclei.
 8.8 Radio pulsars and tests of general relativity 201

8.8 Radio pulsars and tests of general relativity

By a strange twist of astronomical fortune, it turned out that the radio pulsars provide some
of the very best tests of general relativity. Observations by Joseph Taylor (b. 1941) and his
colleagues made with the Arecibo radio telescope demonstrated that the arrival times of the
radio pulses from pulsars are among the most stable clocks available to us (Taylor, 1992).
The most important systems from the point of view of testing general relativity are the binary
pulsars. As the techniques of pulsar discovery and precise timing became more and more
refined, many binary pulsars were discovered, the Australia Telescope National Facility
Pulsar Catalogue of 2003 showing that, of almost 1500 pulsars listed, 90 are members of
binary systems.
The most important binary pulsars are those in which both stars are neutron stars in a
close binary orbit. The first of these, the pulsar PSR 1913+16, was discovered by Russell
Hulse (b. 1950) and Taylor in 1974 (Hulse and Taylor, 1975),18 and it has been observed
with very precise timing since then. The system has a binary period of only 7.75 hours
and the orbital eccentricity is large, e = 0.617. This system is pure gold for the relativist –
to test general relativity a perfect clock in a rotating frame of reference is needed, and
systems such as PSR 1913+16 are ideal for this purpose. The neutron stars are so inert and
compact that the binary system is very ‘clean’, in the sense that the neutron stars behave
like point masses in their mutual gravitational fields. Various parameters of the binary orbit
can be measured very precisely, and these provide estimates of different quantities which
involve the masses of the two neutron stars, M1 and M2 , in different ways. The fact that
the loci of these quantities intersect precisely at one point in the (M1 , M2 ) plane indicates
that there is no discrepancy with the expectations of general relativity (Figure 8.12) (Taylor,
1992; Will, 2001). Assuming general relativity is the correct theory of gravity, the mass
estimates of the two neutron stars are the most accurate values for any stars, besides the
Sun: M1 = 1.4411 ± 0.0007M and M2 = 1.3873 ± 0.0007M .
A second remarkable measurement has been the rate of loss of rotational energy of
the binary system because of the emission of gravitational radiation. The rate of loss of
rotational energy can be precisely predicted once the masses of the two neutron stars are
known and the binary orbit determined. The rate of change of the angular frequency, , of
the orbit due to gravitational radiation energy loss is precisely known, −d /dt ∝ 5 (see
Section A8). The change of orbital phase of the system PSR 1913+16 has been observed
since its discovery in 1974, and the observed changes agree precisely with the predictions
of general relativity (Figure 8.13) (Taylor, 1992; Will, 2001). The data up to 1992 were
presented by Taylor in his 1993 Nobel prize lecture. The gap in the data between 1992
and 1998 was caused by the refurbishment of the Arecibo radio telescope with which
these observations were made. From the data presented by Clifford Will (b. 1946) in his
review of the confrontation between general relativity and experiment, the change of phase
has continued to follow the expected relation after 1998, as can be seen in Figure 8.13
(Will, 2001). Thus, although the gravitational waves themselves have not been detected,
exactly the correct energy loss rate has been observed. This is an important result since
it enables a wide range of alternative theories of gravity to be excluded. For example, the
gravitational waves derived from standard general relativity are quadrupolar in nature, and
202 8 Stars and stellar evolution

3
dP
1.41 b /d

dω
2
mass of companion (solar masses)

t( a

/ dt
0.
4%

(
1

0.
00
)

04
1.40

%
0 1 2 3

)
1.39

.0 7%)
γ (0
1.38

1.37

1.42 1.43 1.44 1.45 1.46

mass of pulsar (solar masses)

Figure 8.12: Constraints on masses of the pulsar PSR 1913+16 and its companion from precise
timing data, assuming general relativity to be valid. The width of each strip in the (M1 , M2 ) plane
reflects the observational uncertainty, shown as a percentage. The inset shows the three constraints
on the complete (M1 , M2 ) plane; the intersection region (a) has been magnified 400 times in the main
figure (Will, 2001).

any theory which, say, predicted the emission of dipole or scalar gravitational radiation can be
excluded.
The same techniques of accurate pulsar timing can also be used to determine whether
or not there is any evidence for the gravitational constant varying with time (Taylor, 1992).
These tests are slightly dependent upon the equation of state used to describe the interior
of the neutron star, but, for the range of plausible equations of state, the limits to Ġ/G are
less than about 10−11 year−1 . Thus, there can have been little change in the value of the
gravitational constant over cosmological timescales. Einstein’s standard theory of general
relativity has therefore passed the most precise tests devised so far and can be used with
some confidence in the study of black holes and cosmology.

8.9 The search for gravitational waves

It is generally assumed that Figure 8.13 is convincing evidence for the existence of gravita-
tional waves, and this has acted as a spur to their direct detection by the gravitational wave
observatories currently coming into operation. The detection of gravitational waves is one
of the most demanding challenges facing astronomical technologists.
8.9 The search for gravitational waves 203

Figure 8.13: The change of orbital phase as a function of time for the binary neutron star system PSR
1913+16 compared with the expected change due to gravitational radiation energy loss by the binary
system (solid line) (Taylor, 1992; Will, 2001).

The search for these waves was begun by Joseph Weber (1919–2000) in a pioneering
set of experiments carried out in the 1960s. Weber was an electrical engineer by training
and, having become fascinated by general relativity in the late 1950s, he was inspired
to devise the means of relating the theory to experiments which could be carried out in
the laboratory. He published these ideas in his book General Relativity and Gravitational
Radiation (Weber, 1961) and set about constructing an aluminium bar detector to measure
gravitational radiation from cosmic sources (Weber, 1966). He estimated that with this
detector he could measure strains as small as 10−16 , this figure referring to the fractional
change in the dimensions of the detector caused by the passage of gravitational waves. His
first published results caused a sensation when he claimed to have found a positive detection
of gravitational waves by correlating the signals from two gravitational wave detectors
separated by a distance of 1000 km at the University of Maryland and the Argonne National
Laboratory (Weber, 1969). In a subsequent paper, he reported that the signal originated
from the general direction of the Galactic centre (Weber, 1970). These results were received
with considerable scepticism by the astronomical community since the reported fluxes far
exceeded what even the most optimistic relativists would have predicted for the flux of
gravitational waves originating anywhere in the Galaxy. The positive effect of Weber’s
experiments was that a major effort was made by experimentalists to disprove his results,
and, in the end, his results could not be reproduced.
The challenge to the experimental community was important in stimulating interest in
how the extremely tiny strains expected from strong sources of gravitational waves could
204 8 Stars and stellar evolution

be detected. The outcome was the approval of a number of major national and international
projects in order to detect the elusive gravitational waves. Many new technologies had to be
developed, and the programmes initiated in the 1980s can best be regarded as development
programmes which were to be fully approved in the 1990s. For example, the LIGO project,
an acronym for Laser Interferometer Gravitational-Wave Observatory, was initiated by
Kip Thorne and Rainer Weiss (b. 1932) in 1984, but final approval by the US National
Science Foundation was only given in 1994 at a revised cost for development, construction
and early operations of $365 million.19 The project consists of two essentially identical
interferometers, each with 4 km baselines located at Livingston, Louisiana and Hanford
near Richland, Washington. Similarly, the VIRGO project is a French–Italian collaboration
to construct an interferometer with a 3 km baseline at a site near Pisa, Italy. The GEO600
experiment is a German–British interferometer project with a 600 m baseline, while the
Japanese TAMA project is a 300 m baseline interferometer located at Mitaka, near Tokyo.
For all these projects, there was a long development programme to reach the sensitivities
at which there is a good chance of detecting gravitational waves from celestial sources.
In the best commissioning runs from experiments carried out in 2003, strain sensitivities
in the range 10−20 to 10−21 have been achieved. The importance of having a number of
gravitational wave detectors operating simultaneously is that they form a global network and
any significant events should be detectable by all of them. By precise timing, the direction
of arrival of the gravitational waves can be estimated.
At the time of writing, all the gravitational wave observatories are entering their oper-
ational phases with more or less their design sensitivities. None of them has yet detected
gravitational waves, but it will be no surprise if they are discovered in the next few years.
The potential sources of detectable radiation include the collapse of stellar cores in super-
nova explosions, collisions and coalescences of neutron stars or black holes, rotations of
neutron stars with deformed crusts, the continuous emission of very close binary neutron
stars and black holes (see Section A8) and primoridal gravitational radiation created during
the very earliest phases of our Universe. In addition, like all other new astronomies, the
observational study of gravitational waves is likely to produce some real surprises.

8.10 Supernovae

When Walter Baade and Fritz Zwicky wrote their famous papers of 1934 about the existence
of ‘super-novae’ (Baade and Zwicky, 1934a,b), their conclusions were based upon the
properties of 12 supernovae which had occurred between 1900 and 1930, as well as a few
historical supernovae, including Tycho Brahe’s ‘new star’ of 1572.20 Zwicky began the
systematic search for supernovae in 1934, first with a 3 14 -inch Wollensack lens camera
mounted on the roof of the Robinson Astrophysics Laboratory of the California Institute
of Technology. Then, in 1936, he supervised the construction of a wide-angle 18-inch
Schmidt telescope, which was built in the workshops of the Laboratory and sited at the new
observatory at Palomar. The first supernova was discovered in NGC 4157 in March 1937
and the second in the dwarf spiral galaxy IC 4182 on August 26 1937. As chance would
have it, this second supernova reached apparent magnitude 8.4, six magnitudes brighter
 8.10 Supernovae 205

than the luminosity of the dwarf galaxy, and it remained the brightest supernova discovered
in the twentieth century until the appearance in 1987 of the supernova SN 1987A. Zwicky
continued to discover about four supernovae per year in relatively nearby galaxies with
the 18-inch Schmidt telescope. With the construction of the 48-inch Schmidt telescope at
Palomar in 1949, searches could be made for fainter supernovae and typically about 20 were
discovered each year.21

8.10.1 Type I and Type II supernovae

In 1938, Baade and Zwicky gave the first description of the typical light-curves of supernovae
consisting of an initial outburst lasting for a few weeks, following which the brightness
decreases exponentially with a half-life of about 60 days (Baade and Zwicky, 1938). The
light-curves of the first dozen supernovae discovered by Zwicky were remarkably similar,
but in 1941 Minkowski discovered that there are two quite distinct types of supernovae
(Minkowski, 1941). The primary distinction between the two types concerns differences in
their spectral evolution.
r The spectra of Type I supernovae consist of broad emission bands, the nature of which
were not understood until almost 30 years after Minkowski’s paper when Kirshner and
Oke showed that they could be interpreted as the superposition of hundreds of lines of
Fe+ and Fe++ (Kirshner and Oke, 1975). A key feature of their spectra is the absence
of hydrogen lines. The spectral and luminosity evolution of the Type I supernovae are
essentially identical, so that it is possible to tell how old the supernova is from its
luminosity and spectrum. They are found in all types of galaxy.
r In contrast, the Type II supernovae show the Balmer series of hydrogen soon after max-
imum light. They display a much wider range of properties than the Type II supernovae
and are only found in spiral galaxies, generally within the spiral arms.
In 1960, Minkowski showed that there are also differences in the masses ejected by
the two types of supernovae from fragmentary evidence on historical supernova remnants.
Typically, the expanding shells of the Type I supernovae have masses about 0.1M and
expansion velocities of about 1000 km s−1 , resulting in a kinetic energy of about 1041 J,
whereas for the Type II supernovae the ejected masses amount to several solar masses and
typical velocities are about 6000 km s−1 , corresponding to energies of the order of 1044 J
(Minkowski, 1960a). Although some of his classifications would probably be different
nowadays, Minkowski had come to the correct conclusion.
In both cases, the energetics of the explosions were so great that they must involve the
collapse of the star to some form of compact remnant, either a neutron star or a black hole.
The Type II supernovae were identified with the collapse of the central regions of very
massive stars, M ≥ 8M , which have relatively short lifetimes and cannot have travelled
far from the spiral arm regions within which they formed. To account for the presence of
the hydrogen absorption lines, it was assumed that the explosion of the core takes place
within a red giant, and models by Sydney Falk (b. 1947) and David Arnett showed that their
light curves could be accounted for by the outward passage of a strong shock wave through
a red giant atmosphere (Falk and Arnett, 1973).
206 8 Stars and stellar evolution

In contrast, for the Type I supernovae, the most attractive picture was that they are formed
by the accretion of mass onto a white dwarf in a binary system. The process by which the
explosion takes place is not established, but it is thought to be associated with the heating of
the surface of the white dwarf by the infalling matter. If nothing else happened, the process
of accretion would eventually take the total mass of the star over the Chandrasekhar limit,
and collapse to a neutron star would then ensue. It is believed, however, that before this
can happen the temperature of the surface layers resulting from the process of accretion
becomes high enough for nuclear burning to begin in the star’s surface layers. There then
ensues a nuclear deflagration which propagates through the star, causing a violent explosion
in which the whole star is disrupted. It is plausible that this critical point occurs when the
total mass of the white dwarf has reached a certain critical mass and this can account for the
uniformity of their properties. This scenario can also account for the facts that no hydrogen
absorption lines are observed, that their spectra are iron rich and that they are observed in
all types of galaxy. The formation of neutron stars in binary systems can also account for
the high space velocities of radio pulsars if the binary is disrupted in the explosion.
Just how uniform the population of Type IA supernovae is was demonstrated by Mark
Phillips (b. 1951), who analysed the light-curves of nine examples and showed that there is
indeed a small dispersion in their absolute luminosities at maximum, amounting to 0.6 and
0.5 magnitudes in the V and I wavebands, respectively. In addition, he found that the peak
luminosity is correlated with the rate at which the supernova subsequently decayed (Phillips,
1993), a relation first suggested by Yuri Pskovskii (1926–2004) (Pskovskii, 1977, 1984).
Another version of this relation using the light-curve shapes of the supernova outbursts was
developed by Adam Riess (b. 1969), William Press and Robert Kirshner (b. 1949) (Riess,
Press and Kirshner, 1995). This procedure took account of the correlation between the
luminosity of the supernova at maximum and the increasing timescale of the outburst, and
resulted in a ‘corrected’ dispersion in the maximum luminosity of only 0.21 magnitudes in
the V waveband, indicating how uniform the population of these supernovae must be. These
supernovae have turned out to be important ‘standard candles’ for measuring distances out to
large redshifts, and an enormous effort has been devoted to discovering many more of them,
particularly at large redshifts, and in understanding the astrophysics of these explosions in
detail (see Chapter 13).
The long exponential decay of the luminosities of supernovae following their initial
outbursts was a puzzle since the characteristic decay time was the same for essentially all
supernovae. In 1962, Titus Pankey suggested in his Ph.D. dissertation that the decay might
be associated with the decay of radioactive nuclides created in the explosion (Pankey, 1962).
This proposal was put on a firm astrophysical basis by Stirling Colgate (b. 1925) and Chester
McKee (b. 1942) in 1969 (Colgate and McKee, 1969). The basic idea was that, in the process
of collapse to form a neutron star, explosive nucleosynthesis takes place, and among the
products is the radioactive isotope of nickel, 56 Ni. This isotope then decays as follows:

β+ β+
56
Ni −→ 56
Co −→ 56
Fe. (8.20)

The first β-decay of 56 Ni has a half-life of only 6.1 days, while the second β-decay, which
has a half-life of 77.1 days, is presumed to be the source of energy for the exponential decay
 8.10 Supernovae 207

of the luminosity of the supernova, 3.5 MeV being liberated in the form of γ -rays in each
decay of a 56 Co nucleus. Thus, the exponential decay of the luminosity of the supernova
is attributed to the creation of radioactive nickel in the explosion, which is ejected into the
expanding envelope of the supernova and so contributes to the enrichment of the abundance
of heavy elements in the interstellar medium.

8.10.2 Supernova 1987A

Unquestionably, the most important event in supernova studies in the twentieth century was
the explosion of a supernova in one of the dwarf companion galaxies of our own Galaxy, the
Large Magellanic Cloud. This supernova, known as SN 1987A, was first observed optically
on 24 February 1987 and reached about third visual magnitude by mid May 1987.22 It was
the brightest supernova since Kepler’s supernova of 1604 and the first bright supernova to
be studied with all the power of modern instrumentation. Ironically, it appears to have been
a peculiar Type II supernova because the light-curve showed a much more gradual increase
to maximum light than is typical of Type II supernovae. It took 80 days to reach maximum
light, and its bolometric luminosity then remained roughly constant for about two months
after that time, despite the fact that there was a rapid decline in its surface temperature. It
was also subluminous as compared with the typical Type II supernova.
The supernova coincided precisely with the position of the bright blue supergiant star
Sanduleak − 69 202, which disappeared following the supernova explosion. This observa-
tion indicated that the progenitor of the supernova was a massive early-type B3 star. The
fact that the progenitor was a highly luminous blue star was a surprise because it might have
been expected to be a red supergiant. The early phases of development of the light-curve
suggested that the progenitor star must have been massive, M ≈ 20M , consistent with
the mass of the B star Sanduleak −69 202. Stellar evolution models, which begin with this
mass, were developed in which the progenitor first becomes a red giant and then, because of
strong mass loss, moves to the blue region of the Hertzsprung–Russell diagram for 104 years
before exploding as a supernova. According to this picture, the star must have had a smaller
envelope than is usual for the B star and a lower abundance of heavy elements than the stan-
dard cosmic abundances, roughly one-third the solar value. This abundance is consistent
with the general trend of heavy element abundances in the Large Magellanic Cloud.
One piece of great good fortune was that, at the time of the explosion, neutrino detectors
were operational at the Kamiokande experiment in Japan and at the Irvine–Michigan–
Brookhaven (IMB) experiment located in an Ohio salt-mine in the USA. Both experiments
were designed for an entirely different purpose, which was to search for evidence of proton
decay, but the signature of the arrival of a burst of neutrinos was convincingly demonstrated
in both experiments.23 Only 20 neutrinos with energies in the range 6 to 39 MeV were
detected, 12 at Kamiokande and 8 at IMB, but they arrived almost simultaneously at the
two detectors, the duration of this pulse being about 12 seconds. The timescale of 12 seconds
is consistent with what would be expected when allowance is made for neutrino trapping in
the core of the collapsing star. What made this identification of the neutrino pulse with the
supernova convincing was that the supernova was only observed optically some hours after
the neutrino pulse. The neutrinos escape more or less directly from the centre of the collapse
208 8 Stars and stellar evolution

log10(L(erg s−1)) 42

40 56Co

38 5x57Co

44Ti

36 22Na

0 500 1000 1500

days since outburst

Figure 8.14: The light-curve of SN 1987A (Chevalier, 1992). The bolometric luminosity of the
supernova in the ultraviolet, optical and infrared wavebands during the first five years. The energy
deposited by radioactive nuclides (dashed lines) is based upon the following initial masses: 0.075M
of 56 Ni (and subsequently 56 Co), 10−4 M of 44 Ti, 2 × 10−6 M of 22 Na and 0.009M of 57 Co, the last
being five times the value expected from the solar ratio of 56 Fe/57 Fe.

of the progenitor star, whereas the optical light has to diffuse out through the supernova
envelope. This observation, coupled with the measured energies of the neutrinos, enabled
limits to be set to the rest mass of the neutrino. If the electron neutrino had a finite rest
mass, the more energetic neutrinos would be expected to arrive before the less energetic
ones since they would have velocities closer to that of light. The limits derived from these
considerations corresponded to m νe ≤ 20 eV.
The observation of the neutrino flux from the supernova is uniquely important for the
theory of stellar evolution. Adopting standard cross-sections for the neutrino interactions
and the size of the detectors, it turns out that the neutrino luminosity of the supernova was of
the same order as that expected from the formation of a neutron star (E ≈ 1046 J). Note that
these observations provide strong support for the essential correctness of our understanding
of the late stages of stellar evolution.
Figure 8.14 shows how the bolometric light curve of SN 1987A evolved over the five years
since the initial explosion. After the initial outburst, the luminosity decayed exponentially
with characteristic half-life of 77 days until roughly 800 days after the explosion when the
rate of decline decreased (Chevalier, 1992). One of the most intriguing observations has
been the search for the products of the radioactive decay chains in the γ -ray and optical–
infrared spectra of the supernova. To account for the luminosity of the supernova, about
0.07M of 56 Ni must have been deposited in the supernova envelope, and this figure agrees
8.10 Supernovae 209

very well with the theoretical expectations of explosive nucleosynthesis. It is therefore

expected that the ratio of abundances of 56 Ni and 56 Co to iron should decrease as these
radioactive elements decay. Evidence was found for the γ -ray line of 56 Co at the 1238 keV
line within six months of the explosion from observations with the γ -ray spectrometer on
board the Solar Maximum Mission (Matz et al., 1988). In addition, the fine-structure lines
of singly ionised cobalt at 10.52 μm and singly ionised nickel at 10.62 μm appeared in the
infrared spectrum of the supernova and increased in strength once the exponential decrease
in luminosity began (Aitken et al., 1988).
There are other important features of the light-curve shown in Figure 8.14. The optical
light-curve showed a break at about 500–600 days but, at that same time, the far-infrared
flux increased so that the total luminosity continued to decrease exponentially. In addition,
observations of the near-infrared lines of iron showed that less than 0.075M of iron was
present. At about the same time, the emission lines showed absorption of the redshifted gas,
indicating that absorption was taking place within the supernova. All these observations are
consistent with the formation of dust within the supernova ejecta after about 500 days.
After about 900 days, the rate of decline of the total luminosity of the supernova
decreased. The natural interpretation of this phenomenon is that another, longer lived,
radioactive nuclide had taken over from 56 Co as the energy source for the remnant, and the
expected candidate is 57 Co. In Figure 8.14, the expected light-curve is shown, assuming
that 57 Co is five times more abundant than its local abundance; also shown are the expected
contributions of the longer lived radionuclides 44 Ti and 22 Na. The totality of these observa-
tions provides direct confirmation for the radioactive origin of supernova light-curves and
the formation of iron peak elements in supernova explosions.
Observations of SN 1987A have also provided a great deal of information about the
pre-history of the explosion and the surrounding interstellar medium. The burst of radiation
associated with the explosion illuminated the material ejected in previous mass loss events,
particularly from the period of strong mass loss during the red giant phase. Evidence has
been found for bipolar emission structures, as well as for ‘light echoes’ from clouds or dust
‘sheets’ along the line of sight to the supernova.
Perhaps the most remarkable feature has been the discovery by the Hubble Space Tele-
scope of a ring of emission in the line of doubly ionised oxygen [OIII] about the supernova
(Figure 8.15). This ring was excited by the initial outburst of ultraviolet radiation from the
supernova and is in the form of a perfect ellipse. The ultraviolet spectrum of the supernova
had been monitored regularly by the International Ultraviolet Explorer, and, at a certain
time after the explosion, forbidden ultraviolet emission lines appeared which increased in
strength over the subsequent weeks. It was natural to associate this event with the illumina-
tion of the ring. These observations enabled a rather precise estimate of the distance of the
supernova to be made, a value which was entirely consistent with independent estimates of
the distance of the Large Magellanic Cloud (Panagia et al., 1991).
It was predicted by Richard McCray (b. 1937) that eventually the ring would be hit by
the blast-wave from the supernova, which travels at a speed much less than the speed of
light, and then it would be expected that the ring would be illuminated at optical and X-ray
wavelengths (McCray, 1993). As he remarked in that review,
210 8 Stars and stellar evolution

Figure 8.15: Images of SN 1987A taken by the Hubble Space Telescope, the Chandra X-ray Observa-
tory and the Australia Telescope National Facility, showing the evolution of the structure of the ring
over the period 1996 to 2003 at optical, X-ray and radio wavelengths. (Courtesy of NASA, the Space
Telescope Science Institute, the Chandra Science Team and the ATNF.)

. . . the impact will occur c.2004 ± 3 AD, allowing for the uncertainty in the density, n b = 8 cm−3 , of
the shocked wind from the blue giant progenitor of SN 1987A.

Observations by the Hubble Space Telescope, the Chandra X-ray Observatory and the
Australia Telescope in its compact configuration have observed this remarkable event
(Figure 8.15). In all three wavebands, enhanced emission has been observed, the opti-
cal image showing a ‘ring of pearls’ where the shock-wave has encountered higher density
regions within the ring. The collision of the shock-wave with the ring has heated the station-
ary material to high temperatures and accounts for the increase in X-ray intensity observed
from the ring. Equally impressive is the fact that the radio emission, which is associated with
the synchrotron radiation of high-energy electrons accelerated in the shock-wave, shows a
clear increase in intensity, providing a real-time example of the acceleration of high-energy
electrons in a supernova shock-wave (see Section 11.4 for a discussion of the acceleration
process).
Notes to Chapter 8 211

Notes to Chapter 8

1 This work was carried out during a remarkable period in the late 1960s when Fred Hoyle founded
the Institute of Theoretical Astronomy in Cambridge, UK. During the summer months, Fowler,
Arnett, Clayton, Wagoner, the Burbidges and their students were regular visitors. Many of the key
problems of nuclear astrophysics could be addressed by the new generation of high-speed digital
computers.
2 A historical review of the solar models and how the predicted solar neutrino flux changed from
the 1960s to 2003 is given by J. N. Bahcall, Solar models: an historical review, Nuclear Physics B
(Proceedings Supplement), 118, 2003, 77–86.
3 An excellent history of the solar neutrino problem up to 1989 is given by Bahcall in his book
Neutrino Astrophysics (Bahcall, 1989).
4 Bahcall has provided an interesting commentary on the subject of why it was so long before the
particle physicists took the solar neutrino problem seriously. He argues that the particle physicists
did not realise just how robust the helioseismological determinations of the internal structure of
the Sun were; see Bahcall, Solar models.
5 An excellent survey of the role of computation in developing models of stellar structure is given by
R. Kippenhahn and A. Weigert, Stellar Structure and Evolution (Berlin: Springer-Verlag, 1990).
6 The physics of the evolution of stars from the main sequence to the red giant branch has remained
one of the classic ‘unsolved’ problems of stellar astrophysics. John Faulkner has provided deep
insights into the physical cause of the phenomenon in his paper ‘Low-mass red giants as binary
stars without angular momentum’ (Faulkner, 2001). He identifies the key feature as the ratio of
the densities in the core and in the hydrogen-burning shell and the need to be able to store a
considerable amount of mass in the red giant envelope. The consequence of these is an anti-
homology theorem, according to which, under these conditions, the density of the envelope must
decrease as the core density increases.
7 An excellent survey of the history of the understanding of the theory of stellar pulsation is included
in John P. Cox, The Theory of Stellar Pulsation (Princeton: Princeton University Press, 1980).
8 On the OPAL web-site, the following message indicates what was included in these massive
computations. ‘Briefly, the calculations are based on a physical picture approach that carries out
a many-body expansion of the grand canonical partition function. The method includes electron
degeneracy and the leading quantum diffraction term as well as systematic corrections necessary
for strongly-coupled plasma regimes. The atomic data are obtained from a parametric potential
method that is fast enough for in-line calculations while achieving an accuracy comparable to single
configuration Dirac–Fock results. The calculations use detailed term accounting; for example, the
bound–bound transitions are treated in full intermediate or pure LS coupling depending on the
element. Degeneracy and plasma collective effects are included in inverse bremsstrahlung and
Thomson scattering. Most line broadening is treated with a Voigt profile that accounts for Doppler,
natural width, electron impacts, and for neutral and singly ionized metals broadening by H and He
atoms. The exceptions are one-, two-, and three-electron systems where linear Stark broadening
by the ions is included.’
9 Hewish was awarded the Nobel prize for physics for the discovery of neutron stars. The prize was
awarded jointed with Martin Ryle, the citation describing his work in developing the principles of
aperture synthesis.
10 The history of the discovery of pulsars is described by A. Hewish, The pulsar era, Quarterly
Journal of the Royal Astronomical Society, 27, 1986, 548–558. A delightful description of the
discovery of pulsars is given by Jocelyn Bell-Burnell, The discovery of pulsars, in Serendipitous
Discoveries in Radio Astronomy, eds. K. Kellermann and S. Sheets) (Green Bank, West Virginia:
National Radio Astronomy Observatory Publications), pp. 160–170.
11 According to William McCrea, the term ‘pulsar’ was invented by Anthony Michaelis, then science
correspondent for the Daily Telegraph.
212 8 Stars and stellar evolution

12 A survey of observations and the physics of pulsars, including references to many original papers, is
given by A. G. Lyne and F. Graham Smith, Pulsar Astronomy (Cambridge: Cambridge University
Press, 1998).
13 The preprint on this subject is referred to by A. G. W. Cameron in his review Neutron stars Annual
Review of Astronomy and Astrophysics, 8, 1970, 179–208.
14 The history of the UHURU satellite and its discoveries is described by W. Tucker and R. Giacconi,
The X-ray Universe (Cambridge, Massachusetts: Harvard University Press, 1985).
15 I have given a proof of this result in Section 17.10 of Malcolm Longair, Theoretical Concepts in
Physics (Cambridge: Cambridge University Press, 2003).
16 For details of the physics of accreting binary systems, see S. I. Shapiro and S. A. Teukolsky, Black
Holes, White Dwarfs and Neutron Stars: The Physics of Compact Objects (New York: Wiley
Interscience, 1983).
17 Many more details of these processes are contained in J. Frank, A. King and D. Raine, Accretion
Power in Astrophysics, 3rd edn (Cambridge: Cambridge University Press, 2002).
18 Hulse and Taylor were awarded the Nobel prize for physics in 1993 for this discovery.
19 A very readable account of black holes, gravitational waves and their significance for physics and
astronomy is contained in Kip Thorne’s book Black Holes and Time Warps: Einstein’s Outrageous
Legacy (New York: W. W. Norton and Company, 1994).
20 For a comprehenive review of the properties of supernovae, see V. Trimble, Supernovae. Part 1:
the events, Reviews of Modern Physics, 54, 1982, 1183–1224 and V. Trimble, Supernovae. Part II:
the aftermath, Reviews of Modern Physics, 55, 1983, 511–563.
21 For an entertaining description of Zwicky’s researches on supernovae, see F. Zwicky, Review of
the research on supernovae, in Supernovae and Supernova Remnants, ed. C. Battali Cosmovici
(Dordrecht: D. Reidel Publishing Company, 1974), pp. 1–16.
22 There is a vast literature on SN 1987A. Symposia devoted to the supernova include: Supernova
1987A in the Large Magellanic Cloud, eds M. Kafatos and A. G. Michalitsianos (Cambridge:
Cambridge University Press, 1988) and Supernova 1987A and Other Supernovae, eds I. J. Danziger
and K. Kjär (Garching bei München: European Southern Observatory, 1991). Excellent summaries
of the observations up to 1992 are given by R. Chevalier, Nature, 355, 1992, 691–696, and by R.
McCray, Annual Reviews of Astronomy and Astrophysics, 31, 1993, 175–216.
23 An excellent discussion of these observations is given by Bahcall (1989).

A8 Explanatory supplement to Chapter 8

A8.1 Gravitational radiation loss rate by the binary pulsar
The observation that the rate of change of the angular velocity, , of the binary pulsar PSR
1913+16 is proportional to 5 is considered to be very strong evidence that the system is
losing energy by the radiation of gravitational waves. The following elementary arguments
give an impression of how the result comes about. The excellent essay by Bernard Schutz
(b. 1946) can be recommended as a gentle introduction to the physics of gravitational
radiation.1 The theory of gravitational radiation is highly non-trivial and has not been without
controversy.2 These notes are intended to provide some insight into how this dependence
comes about and the energies involved.
We use the analogy between the inverse square laws of electrostatics and gravity to work
out the rate of radiation of quadrupole radiation from a system of charges or masses. In
1906, J. J. Thomson (1856–1940) presented a simple argument to derive the formula for
the instantaneous rate of loss of energy by radiation of an accelerated charged particle.3 I
Explanatory supplement to Chapter 8 213

have used this approach in my book Theoretical Concepts in Physics, the result being, in
SI notation,
 
dE | p̈|2
− = , (A8.1)
dt 6π 0 c3
where p is the dipole moment of the accelerated charge relative to some origin. In the case
of an oscillating dipole, we can write p = p 0 e−i t = qr 0 e−i t , where is the angular
frequency of oscillation and p = qr . Therefore, the expression for the time-averaged rate
of radiation loss is given by
 
dE q 2 4 |r 0 |2 4
| p 0 |2
− = = . (A8.2)
dt 12π 0 c 3 12π 0 c3
The equivalent expression for gravity results in no energy loss because a gravitational dipole
has zero dipole moment, both ‘gravitational charges’ having the same sign. Thus, there is
no dipole emission of gravitational waves.
A system of gravitating masses can, however, have a finite quadrupole moment, and, if this
is time varying, there is an energy loss by gravitational radiation. The corresponding equation
to equation (A8.2) for the time-averaged rate of loss of energy by electric quadrupole
radiation is given by
  6 
dE
− = |Q jk |2 , (A8.3)
dt 1440π 0 c j,k
5

where the elements of the electric quadrupole moment tensor, Q jk , are given by

Q jk = (3x j xk − r 2 δ jk )ρ(x) d3 x, (A8.4)

and ρ(x) is the electric charge density.4

We obtain an approximate expression for quadrupole gravitational radiation in this
‘Coulomb approximation’ if we use the equivalence e2 /4π 0 → Gm 2 ; then,
 
dE G 6 
− = |Q jk |2 , (A8.5)
dt gr 360c5 j,k

where the Q jk are now the components of the gravitational quadrupole moment tensor. The
exact expression quoted by Schutz is as follows:
   
dE G  ... ... ...
1 2
− = 5 Q jk Q jk − 3 Q , (A8.6)
dt gr 5c j,k

where the Q jk are the components of spatial quadrupole tensor, or matrix, the second
moment of the mass, or charge, distribution,5

Q jk = ρx j xk d3 x. (A8.7)

Following Schutz’s pleasant order-of-magnitude estimates, let us estimate the rate of

radiation of a binary star system, each star having mass M moving in a circular orbit of
214 8 Stars and stellar evolution

radius r at speed v about their common centre of mass. To order of magnitude, the third
derivative of the quadrupole moment is then given by

d3 Q Q Mr 2 Mv 3
∼ ∼ ∼ , (A8.8)
dt 3 t3 t3 r
where r and t are the typical spatial and time scales over which the quadrupole moment
varies. Inserting this value into equation (A8.6) for the term in large brackets on the right-
hand side, we find the gravitational radiation loss rate:
   2
dE G Mv 3
− ∼ . (A8.9)
dt gr 5c5 r

Since v = r , we obtain the result

 
dE G
− ∼ 5 M 2r 4 6
. (A8.10)
dt gr 5c

Note that equations (A8.5) and (A8.10) have exactly the same dependence upon the
quantities which appear in the loss rate for gravitational radiation since Q jk ∼ Mr 2 . The
angular velocity of the binary star is given by 2 = G M/4r 3 , and so equation (A8.10) can
be rewritten as follows:
   
dE c5 GM 5
− ∼ , (A8.11)
dt gr 80G r c2

where, following Schutz, it is assumed that four comparable terms contribute to the total
gravitational loss rate. As Schutz points out, the term c5 /G corresponds to a luminosity of
3.6 × 1052 W, an enormous value, which is modified by the ‘relativistic factor’, (G M/r c2 )5 .
Thus, really close binary neutron stars are expected to be extraordinarily powerful sources
of gravitational waves. In particular, the very last phases of inspiralling of a pair of neutron
stars in a very close binary system, before they coalesce into a black hole, is one of the most
important targets for observation by the present generation of gravitational wave detectors.
The key relation from our present perspective is that equations (A8.5) and (A8.10) show
that the rate of energy loss by gravitational waves is proportional to 6 . This energy is
extracted from the binding energy of the binary system, E = − 12 I 2 , and so we can write
  
dE d( 12 I 2
)
− = ∝ 6
; (A8.12)
dt gr dt
˙ ∝ 5
. (A8.13)

This relation was found by Hulse and Taylor for the binary pulsar PSR 1913+16 (see Figure
8.13). They did much more than this because the orbits of the two neutron stars and their
masses were very well determined by their precise timing observations, and so the exact
radiation loss rate for their highly elliptical orbits, e = 0.615, could be estimated and was
found to be in excellent agreement with the observed spin-up rate of the binary system.
Notes to Section A8 215

Notes to Section A8

1 B. Schutz, Gravitational radiation, Encyclopaedia of Astronomy and Astrophysics, vol. 2, ed.

P. Murdin (Bristol and Philadelphia: Institute of Physics Publishing and London: Nature Publishing
Group, 2001), pp. 1030–1042.
2 The care with which the theory has to be formulated is carefully described by Ray D’Inverno in
Introducing Einstein’s Gravity (Oxford: Clarendon Press, 1995). A brief history of some of the
problems involved in understanding the existence of gravitational waves in general relativity is
given by John Stachel in his review History of relativity, in Twentieth Century Physics, vol. 1, eds
L. M. Brown, A. Pais and A. B. Pippard (Bristol and Philadelphia: Institute of Physics Publishing
and New York: American Institute of Physics Press, 1995), pp. 249–356.
3 The argument was first presented by Thomson in his book Electricity and Matter (London:
Archibald Constable and Company, 1906), the published version of his Silliman Memorial Lec-
tures delivered at Harvard in 1903. This same formula was subsequently used by Thomson in his
book Conduction of Electricity through Gases (Cambridge: Cambridge University Press, 1907)
to derive the expression for the cross-section for the scattering of X-rays by free electrons, the
Thomson cross-section.
4 Equation (A8.3) is derived by J. D. Jackson in his excellent book Classical Electrodynamics, 3rd
edn (New York: John Wiley and Sons, Inc., 1999), Section 9.3.
5 Note the difference of a factor of 3 between the definitions of the quadrupole moment tensor for
electromagnetism, equation (A8.4), and gravity, equation (A8.6). There is a compensatory factor of
3 in the term in brackets on the right-hand side of equation (8.6). The difference in the gravitational
case is because of technicalities about the choice of gauge to be used in the full development of
the theory of gravitational radiation.
 9 The physics of the interstellar
medium

9.1 The photoionisation of the interstellar gas

By 1939, the existence of various forms of interstellar matter had been established. From
the study of interstellar absorption lines and the variation of interstellar extinction with
distance, it was known that diffuse gas and dust are present in the interstellar medium
(Plaskett and Pearce, 1933; Joy, 1939). Gaseous nebulae had been known to be constituents
of the Galaxy since the time of Huggins’ pioneering observations in the 1860s. During the
first two decades of the twentieth century, Edward Barnard (1857–1923) made extensive
studies of the forms of the dark clouds apparent in photographs of the Milky Way (Barnard,
1919). The nature of these clouds was studied by Max Wolf (1863–1932), who determined
the amount of extinction they cause by making star counts in their vicinity (Wolf, 1923).
He correctly attributed the extinction to dust grains rather than gas because in the latter
case the strong dependence of Rayleigh scattering upon wavelength would have resulted in
much greater reddening of background stars than was observed.
On the theoretical side, it was recognised in the early 1920s that both the central stars
of planetary nebulae and the O stars are very hot and so radiate a great deal of energy in
the ultraviolet waveband. Russell suggested that the excitation of the emission lines seen
in gaseous nebulae and planetary nebulae were due to photoexcitation (Russell, 1921), and
Eddington showed that, as a result, the gas would attain a temperature of about 10 000 K
(Eddington, 1926b). In 1926, Donald Menzel applied this model to planetary nebulae and
suggested that the Balmer emission lines were associated with the photoionisation of hydro-
gen by the ultraviolet radiation of the central star followed by recombination of the protons
and electrons (Menzel, 1926). He believed, incorrectly, that the temperatures of the central
stars would have to be unreasonably high for this process to be effective. In 1927, Herman
Zanstra (1894–1972), however, had no hesitation in postulating that the central stars of plan-
etary nebulae had temperatures of about 30 000 K (Zanstra, 1926). In a subsequent paper,
he described the processes of photoionisation and recombination of hydrogen, in which
each photon with energy greater than 13.6 eV ionises one hydrogen atom (Zanstra, 1927).
In recombining, the electron does not necessarily return to the ground state but cascades
through a number of energy levels, emitting the Balmer series of hydrogen in the process.
In 1928, Ira Bowen (1898–1973) at last identified the ‘nebulium’ lines which had baffled
spectroscopists since their discovery in 1868 (Bowen, 1927, 1928). In a brilliantly argued
paper, he showed that the nebulium lines can all be associated with what are now termed

216
 9.2 Neutral hydrogen and molecular line astronomy 217

forbidden transitions between low-lying metastable states of the ions of common elements
and the ground state. For example, the pair of lines at 500.7 and 495.9 nm are associated with
forbidden transitions of doubly ionised oxygen [OIII]. These lines had not been observed
in laboratory experiments because the densities are so high that the metastable states are
de-excited by collisions. In constrast, in the diffuse nebulae, the rate of collisions can be
very low and so, although the transition probabilities are low, radiative transitions are the
means by which the ions are de-excited and reach the ground state.
These discoveries set the scene for the development of the theory of photoionisation and
of recombination and forbidden-line spectra under interstellar conditions during the 1930s
and 1940s, studies particularly associated with the names of Menzel and his colleagues
Lawrence Aller (1913–2003) and James Baker (b. 1914).1 The result of these studies was to
establish powerful means of estimating the densities and temperatures of regions of ionised
gas. The importance of the process of photoionisation for clouds of ionised gas in the
interstellar medium was described by Bengt Strömgren (1908–1987) in 1939 (Strömgren,
1939). He solved the problem of the dependence of the ionisation of the interstellar gas
upon its density and the temperature of the exciting star. The radius of the ionisation zone
about a star is a very strong function of its surface temperature because it depends upon the
flux of ionising radiation with wavelengths shorter than the Lyman limit at λ ≤ 91.6 nm.
For example, if the interstellar density were 106 m−3 and the radius of the star the same as
that of the Sun, an O5 star would ionise a region of radius 54 pc, whereas, for a B5 star, the
radius would be only 1.6 pc. Strömgren also showed that the regions of ionised gas, often
referred to as Strömgren spheres, have very sharp edges. As he expressed it (Strömgren,
1939):

Once the proportion of neutral atoms begins to increase, the absorption of the ionising radiation
increases, leading to an accelerated increase of neutral atoms.

These conclusions were in very good agreement with the observed properties of the regions
of ionised hydrogen known as HII regions. The hottest O stars are found embedded in
regions of ionised hydrogen, the most extensive regions being excited by a number of O
stars.
This picture of the interstellar gas, as it was understood in 1939, was entirely derived
from optical observations. Little did these pioneers realise that they had detected only a
tiny fraction of the total amount of material present in the interstellar medium. After the
Second World War, observations became possible in the radio, infrared, ultraviolet and
X-ray wavebands and revealed the full complexity of the interstellar medium.

9.2 Neutral hydrogen and molecular line astronomy

The prediction of the 21 cm line of neutral hydrogen is one of the more remarkable stories
attending the birth of radio astronomy. Jan Oort was the director of the Leiden Observatory
throughout the period of the German occupation of Holland during the Second World War,
and he kept astronomical activity alive by encouraging theoretical studies among those
staff and students who had escaped detention. Conditions were very difficult, and Oort
218 9 The physics of the interstellar medium

was personally in considerable danger because of his support for Jewish professors in
the University. Astronomical seminars were held in secret in the basement of the Leiden
Observatory. Copies of the Astrophysical Journal somehow continued to reach Leiden, and
among the papers was Reber’s continuum map of the Galaxy. Oort asked Hendrik van de
Hulst (1918–2000) the following question:2

Is there a spectral line at radio frequencies we should in principle be able to detect? If so, because at
radio wavelengths absorption should be negligible, we should be able to derive the structure of the
Galaxy. We might even be able to detect spiral arms, if they exist.

Van de Hulst took up the challenge and studied the many ways in which atoms, ions
and molecules could radiate line emission in the radio waveband (Van de Hulst, 1945). His
results were presented at the last of the underground seminars held in April 1944. The most
significant prediction was that neutral hydrogen should emit line radiation at a wavelength
of 21.106 cm because of the minute change of energy when the relative spins of the proton
and electron in a hydrogen atom change, that is, when transitions take place between its
hyperfine states. Although this is a highly forbidden transition with a spontaneous transition
probability of only once every 12 million years, so much neutral hydrogen was expected
to be present in the Galaxy that there was a good chance that the line would be detectable.
Oort and Lex Muller (1923–2004) would probably have been the first to detect the 21 cm
line, but the first receiver was destroyed in a fire. The first detection of the line was made
by Harold Ewen (b. 1922) and Edward Purcell (1912–1997) at Harvard University in 1951
(Ewen and Purcell, 1951); six weeks later, Oort and Muller measured the same line (Muller
and Oort, 1951). The neutral hydrogen line proved to be one of the most important tools for
diagnosing velocity fields in our own and other galaxies. The first maps of the distribution
of neutral hydrogen in the Galaxy and the determination of the Galactic rotation curve
appeared in 1952. By 1953, neutral hydrogen had been detected in the Magellanic Clouds,
and in 1954 the high-velocity features in the Galactic centre and 21 cm absorption spectra
were first measured.3
In 1958, Oort, Frank Kerr (1918–2000) and Gart Westerhout (b. 1927) published their
famous map of the distribution of neutral hydrogen in the plane of the Galaxy (Figure 9.1)
(Oort, Kerr and Westerhout, 1958). This map was made by combining observations made
in the Netherlands with those made at Sydney, Australia. The derivation of the density
distribution along the line of sight through the Galaxy was not a trivial exercise because the
neutral hydrogen observations do not provide any distance measure. Nonetheless, subse-
quent analyses have generally agreed with the large-scale features seen in Figure 9.1. Neutral
hydrogen is omnipresent throughout the plane of the Galaxy and extends well beyond the
Sun’s radius. Various ‘spiral features’ can be distinguished on the map, although there is
only a general correlation with the arms outlined by the O and B stars and other spiral arm
indicators. Observations of the 21 cm line have also enabled the Galactic rotation curve to
be determined. Of particular interest is the rotation curve beyond the Sun’s orbit about the
Galactic centre because it is flat, v(r ) = constant, suggesting the presence of dark matter in
the halo of the Galaxy (Fich and Tremaine, 1991).
Long before the advent of radio astronomy, it was known that there exist significant
adundances of molecules in interstellar space. The molecules CH, CH+ and CN possess
9.2 Neutral hydrogen and molecular line astronomy 219

160° 147°S 130°

170° 120°
180° 110°
R = 16 kpc
190° 100°
200°
90°
210° 14

80°

220° 12
70°

230° 10

60°

240° 8

6 50°

250°
4
40°

260°

30°
C+

270°

4 20°

6
280°

8.2
1.6 10°
1.0
0.6 10
0.2
0.05
atoms cm−2
12
290°

300° 310° 127°S 340° 350° 0°

Figure 9.1: The distribution of atomic hydrogen in the Galactic plane (Oort et al., 1958). The Sun–
Galactic centre distance is assumed to be 8.2 kpc. The numbers around the outside of the figure denote
Galactic longitudes. In deriving this map, it is assumed that the neutral hydrogen rotates in circular
orbits about the Galactic centre, with velocities given by a standard rotation curve.

electronic transitions in the optical waveband, and absorption features associated with these
were well known in the spectra of bright stars. As radio astronomy technology advanced, it
became possible to observe at higher and higher frequencies. The first interstellar molecule
to be detected at radio wavelengths was the hydroxyl radical OH, which was observed in
absorption against the bright radio source Cassiopeia A by Sander Weinreb (b. 1936), Alan
Barrett (1927–1991) and their colleagues at a wavelength of 18 cm in 1963 (Weinreb et al.,
1963). Soon afterwards, in 1965, the hydroxyl lines were observed in emission by Harold
Weaver (b. 1917) and his group at Berkeley (Weaver et al., 1965; Weinreb et al., 1965). The
surprise was that the sources were very compact and variable in intensity. The corresponding
brightness temperatures were very great indeed, Tb ≥ 109 K, implying that the emission
process must involve some form of maser action. This was the beginning of the intensive
search for other interstellar molecules. In 1968, ammonia, NH3 , was detected (Cheung
et al., 1968), and in the following year water vapour, H2 O, and formaldehyde, H2 CO, were
220 9 The physics of the interstellar medium

discovered (Cheung et al., 1969; Snyder et al., 1969). The processes of line formation in
these molecules involve doubling processes, and the observed intensities involved various
forms of maser action.
A key discovery was the great intensity of the carbon monoxide molecule, CO, which was
first observed by Robert Wilson (b. 1936), Keith Jefferts (b. 1931) and Arno Penzias (b. 1933)
in 1970 (Wilson, Jefferts and Penzias, 1970). CO is one of the simplest molecules which
emits line radiation by electric dipole transitions between neighbouring rotational states.
Rotational transitions are only observed from molecules which have a finite dipole moment,
and so the molecule which is expected to be by far the most abundant in the interstellar gas,
molecular hydrogen, H2 , does not emit rotational molecular lines. Therefore, CO is expected
to be the most abundant species which can be detected by its millimetre and submillimetre
line emission and it acts as a tracer for the distribution of molecular hydrogen. Molecular
abundances and the temperatures of the clouds can be determined since the molecules can be
assumed to be in collisional equilibrium. Since that time, the number of detected molecular
species has multiplied rapidly. Among these has been the detection of ethyl alcohol in the
Galactic centre (Zuckerman et al., 1975), the discovery paper remarking that this
truly astronomical source of ethyl alcohol . . . would yield approximately 1028 fifths at 200 proof.

By August 2004, 125 different molecules had been found in the interstellar medium, a
number of the species being unstable in the laboratory but able to survive in the low-density
conditions of interstellar space. The discovery of interstellar molecules came as a surprise
because it had been assumed that the ultraviolet radiation in the diffuse interstellar medium
would dissociate any but the most tightly bound molecules. This argument had neglected
the shielding role of dust, which protects the molecules from the ultraviolet radiation.
Molecular gas is present throughout the plane of the Galaxy, and, as soon as the first maps
of its distribution were made, it was found that a large fraction of the molecules belong to
giant molecular clouds, which typically have masses about 106 M . By the late 1970s, it
was apparent that these clouds contained a great deal of fine-scale structure and that these
are the sites of star formation. The giant molecular clouds are transparent to millimetre and
submillimetre radiation and so the narrow molecular lines are excellent probes of the internal
dynamics of the clouds. The presence of so many different molecular species and the large
densities of interstellar dust gave rise to the new discipline of interstellar chemistry.4 To
understand the existence and abundances of the many molecules now observed in molecular
clouds requires an understanding, not only of gas-phase reactions, but also of molecular
processes occurring on the surfaces of dust grains.
Among the more remarkable aspects of these discoveries was the interaction between
chemists and astronomers. Harold Kroto (b. 1939) and his colleagues at the University
of Sussex in the UK had successfully synthesised the linear acetylenic chain molecule
cyanoacetylene, HC5 N, in the laboratory and measured its rotational spectrum. Despite
expectations that the molecule would only be observed in very low abundances in the inter-
stellar medium, Kroto and his Canadian colleagues from the National Research Council
discovered the molecule in the giant molecular cloud Sagittarius B2 towards the Galac-
tic centre (Avery et al., 1976). Next, they performed the same exercise for the molecule
HC7 N, expected to be even rarer, but again they found the molecule with very much greater
9.2 Neutral hydrogen and molecular line astronomy 221

abundance than expected from simple thermodynamic arguments (Kroto et al., 1978). Simi-
lar success was achieved in the same year in detecting HC9 N, the longest chain molecule yet
detected (Broten et al., 1978). The importance of these observations was that they demon-
strated conclusively that the long-chain molecules could not be synthesised by equilibrium
gas-phase reactions. The clue was provided by sources such as IRC+10216, a cool red giant
carbon star, which was a plentiful source of the acetylenic chain molecules.
The sequel to this story is just as remarkable. In 1985, Kroto carried out experiments
to find out if the chain molecules could be created in the laboratory using a cluster beam
machine constructed by Richard Smalley (1943–2005) and his colleagues at Rice University,
Houston, modified to use a carbon target. This was successfully achieved in 1985 (Kroto
et al., 1987), confirming the idea that these molecules could be created in cool red giant
envelopes. In the same experiments, which involved creating a hot nucleating carbon plasma,
they also found evidence for a very large abundance of molecules consisting of 60 carbon
atoms, the discovery of the C60 molecule, buckminsterfullerene (Kroto et al., 1985).5
While much of the study of interstellar dust grains had focussed upon the properties
of particles roughly 1 μm in size, evidence was found by Kristen Sellgren (b. 1955) for a
population of very much smaller grains from studies of the infrared continuum spectra of
reflection nebulae (Sellgren, 1984). She found that the colour temperature of the emission
was high, about 1000 K, and could not be attributed to the standard continuum emission
processes. Her proposal was that the emission was associated with transient heating of very
small dust grains. In the case of grains with dimension 1 μm, the energy of the absorbed
photons is thermalised and reradiated at the temperature to which the grains are heated. In
the case of grains only about 1 nm in size, this is no longer the case. An incident ultraviolet
photon can raise the temperature of the grain to a temperature of about 1000 K and then
cool very rapidly, resulting in a quite different continuum spectrum. Sellgren showed that
the necessary number of very small dust grains could be explained as an extrapolation of
the grain size distribution from larger sizes. These tiny grains can be thought of as large
molecules, rather than as solid materials.
This idea was taken further by Alain Leger (b. 1943) and Jean-Loup Puget (b. 1947), who
sought to explain the strong unidentified emission features observed in the infrared region of
the spectrum (Leger and Puget, 1984). These prominent lines are observed at wavelengths of
3.28, 6.2, 7.7, 8.6 and 11.3 μm in the spectra of a wide variety of Galactic and extragalactic
sources (Figure 9.2). Leger and Puget proposed that these lines are associated with various
bending and stretching modes of the small aromatic molecules known as polycyclic aromatic
hydrocarbons, or PAHs. These molecules typically consist of about 50 carbon atoms in the
form of planes of benzene rings. Taking as an example the PAH coronene, they computed
that, at a temperature of 600 K, spectral features should be observed at 3.3, 6.2, 7.6, 8.8
and 11.9 μm. These features were identified as follows: the feature at 3.3 μm with the C−H
stretching mode, those at 6.2 and 7.7 μm with the C−C stretching modes, that at 8.6 μm
with the in-plane bending mode and that at 11.3 μm with the C−H out-of-plane bending
mode. In the last case, other features are expected depending upon the number of nearby
hydrogen atoms. The excitation of these modes is again associated with the absorption
of a single UV photon, which transiently raises the temperature of the molecule to about
1000 K.
222 9 The physics of the interstellar medium

Figure 9.2: The PAH emission features in the 5–15 μm spectrum of the reflection nebula NGC 7023
obtained with the ISO Observatory by Diego Cesarsky and his colleagues (Draine, 2003).

Other features in the absorption spectrum of dust, such as the diffuse interstellar bands
seen in the optical region of the spectrum and the prominent feature at 217.5 nm, have so
far eluded agreed identification. The net result of these studies is that interstellar dust must
be composed of a number of different components. An excellent discussion of the range of
different types of dust particles necessary to account for the observations is given by Bruce
Draine (b. 1947) (Draine, 2003).

9.3 The multi-phase interstellar medium

In the years immediately after the Second World War, significant progress was made in
understanding the properties of the diffuse interstellar medium. The realisation that inter-
stellar extinction was due to dust particles had led Carl Schalén (1902–1993) to apply the
theory of Mie scattering to the problem of interstellar extinction in the 1930s, and he found
that metallic particles with size roughly 10−6 m could reproduce the observed dependence
of extinction upon wavelength (Schalén, 1936). The problem was addressed in much more
detail during and after the War by Hendrik van de Hulst, who considered a very wide range of
possible properties for interstellar dust, ranging from perfectly scattering dielectric spheres
to perfectly absorbing metallic spheres (Van de Hulst, 1949b). He showed how the observed
extinction as a function of wavelength provides information about both the size distribution
of the particles and their chemical composition. It is certain that a range of particle sizes
must be present in the interstellar medium, but the exact chemical composition of the grains
has remained the subject of controversy.
 9.3 The multi-phase interstellar medium 223

60

40

20
latitude

−20

−40

−60
180 150 120 90 60 30 0 330 300 270 240 210 180

longitude

Figure 9.3: The polarisation of stars as a function of galactic coordinates. The magnitudes of the
vectors are proportional to the percentage polarisations, and the directions of the vectors indicate the
plane of polarisation of the light (Matthewson and Ford, 1970).

These studies were relevant to the discovery of interstellar polarisation by William

Hiltner (1914–1991) and John Hall (1908–1991) in 1949 (Hall, 1949; Hiltner, 1949). Their
objective was to observe the polarisation of light emitted from the limbs of stars due to elec-
tron scattering in their atmospheres, an observation suggested by Chandrasekhar in 1946
(Chandrasekhar, 1946). The idea was to search for polarisation effects in binary star systems
as one of the stars is occulted by the other. Hiltner indeed discovered large polarisations,
but they were independent of the phase of the binary star, and so he attributed the result to
interstellar polarisation. Hall, who was collaborating with Hiltner in this project, showed
that the percentage polarisation was correlated with the extinction by dust.
In the same year, Van de Hulst proposed that the polarisation could be attributed to dust
grains which are elongated and aligned so that there is preferential absorption of one of the
states of linear polarisation of the starlight (Van de Hulst, 1949a). This led to a search for
the alignment mechanism for the grains, and in 1951 Leverett Davis Jr (1914–2003) and
Jesse Greenstein (1909–2002) proposed that the alignment is produced by the presence of
a large-scale interstellar magnetic field (Davis and Greenstein, 1951). The orientation of
the dust grains depends upon the dielectric properties of the grains but, in their preferred
model, the grains are aligned with their long axes perpendicular to the magnetic field and
so the magnetic field would be expected to lie parallel to the polarisation vectors of the
starlight. Davis and Greenstein also estimated the strength of the magnetic field necessary
to align the grains and found values of about 0.2–0.3 nT. Magnetic field strengths of the
same order were found by interpreting the diffuse non-thermal radio emission from the
Galaxy as synchrotron radiation, which is also strongly polarised. The most convincing
evidence for the presence of a large-scale magnetic field in the vicinity of the Sun came
from polarisation observations of about 7000 stars by Donald Matthewson (b. 1929) and
Vincent Ford (b. 1943) in 1970 (Matthewson and Ford, 1970) (Figure 9.3). This picture
is broadly consistent with the polarisation properties of the Galactic radio synchrotron
emission.
The theory of the diffuse interstellar medium was the subject of a series of papers by
Lyman Spitzer beginning in 1948. In 1950, Spitzer and Malcolm Savedoff (b. 1928) studied
224 9 The physics of the interstellar medium

in detail the processes of heating and cooling of the diffuse interstellar gas (Spitzer and
Savedoff, 1950). While they confirmed the results of Eddington and Strömgren that, in the
vicinity of O and B stars, the temperature of the ionised gas is about 10 000 K, they also
showed that the temperature of the diffuse medium between these phases was determined
by the balance between heating by the ionisation losses of cosmic rays and by energy losses
associated with the low-lying excited states of CI, CII and SiII. The typical temperature of
the diffuse interstellar medium was found to be about 60 K. These low temperatures were
confirmed some years later by observations of the 21 cm line of neutral hydrogen.
There remained the problem of confining the cool neutral hydrogen clouds, and this
was elegantly solved by George Field (b. 1929), Donald Goldsmith (b. 1943) and Harm
Habing (b. 1937) in 1969, who considered the thermal stability of a medium heated by
interstellar high-energy particles (Field, Goldsmith and Habing, 1969). In 1965, Field had
made a detailed study of thermal instabilities in different astrophysical contexts (Field,
1965) and he had shown that, in the intermediate range of temperatures between about 100
and 3000 K, cosmic ray heating in the presence of cooling by electron excitation of the
low-lying states of neutral and singly ionised species results in an unstable situation, the
stable phases consisting of a low-density, high-temperature phase at T ≈ 8000 K in pressure
balance with a low-temperature phase at T ≈ 20 K. This became known as the two-phase
model for the interstellar medium. The temperature of the diffuse neutral hydrogen clouds
was too low, and one solution was that the interstellar carbon, which is one of the principal
coolants of the gas, was depleted relative to its cosmic abundance because of condensation
into interstellar grains.
This picture of the diffuse interstellar medium was soon confronted by observations from
the Copernicus ultraviolet spectroscopic satellite, which was launched in 1972. Spitzer and
his colleagues had designed the Copernicus satellite specifically to make very high spectral
resolution observations of the interstellar medium and, in particular, to study the resonance
absorption lines of common elements which lie in the wavelength range 91.2 < λ < 120 nm.
Many important discoveries resulted from these studies. The depletion of the chemical
abundances of the heavy elements in the interstellar medium as compared with their cosmic
abundances was confirmed (Field, 1974). A very hot component of the medium was detected
through observations of the lines of O5+ in absorption in the direction of hot stars in the
Magellanic Clouds (Rogerson et al., 1973b). There was evidence that this hot gas formed a
hot flattened halo about the Galaxy. Molecular hydrogen was also detected in the direction
of reddened stars. It was found that the greater the extinction, the greater the column density
of molecular hydrogen, suggesting that its formation is catalysed by dust grains (Rogerson
and York, 1973).
The discovery of the soft X-ray background from the interstellar medium further compli-
cated the picture. The first observations of the soft X-ray background were made by Stuart
Bowyer (b. 1934), Field and John Mack (b. 1942) in 1968 (Bowyer, Field and Mack, 1968),
and soft X-ray maps of the sky were made by William Kraushaar (b. 1920) and his col-
leagues at Wisconsin from a series of rocket flights in 1972 and 1973 (Sanders et al., 1977).
These observations showed a strong anti-correlation between the intensity of the X-ray
emission and the column density of neutral hydrogen, the inference being that the soft X-
rays suffer photoelectric absorption by interstellar neutral hydrogen. Despite the absorption,
 9.4 The formation of stars 225

this was clear evidence that there exists diffuse soft X-ray background emission from the
interstellar medium.
In 1974, Donald Cox (b. 1943) and Barham Smith (b. 1947) provided a convincing
explanation for this component (Cox and Smith, 1974). The UHURU satellite had shown
that young supernova remnants are strong X-ray sources in the 1–10 keV waveband, and this
was naturally interpreted as the bremsstrahlung of gas heated to a very high temperature
by the shock-wave resulting from the supernova explosion. Cox and Smith worked out
the subsequent evolution of these supernova remnants and found that supernovae occur
sufficiently often in our Galaxy that old, hot supernova remnants overlap and, by percolation,
result in a series of tunnels of hot gas through the interstellar medium. This picture can
naturally account for the soft X-ray emission of the Galaxy. This model can also account
for the presence of a halo of hot gas about the Galaxy, since bubbles of the hot gas rise up
the potential gradient out of the Galactic plane. Detailed studies of the local distribution of
neutral and hot gas have suggested that the Sun is located in a local hole of hot gas of radius
about 50 pc which was evacuated by a supernova explosion more than 106 years ago.6
These ideas were synthesised into a picture of the violent interstellar medium, an ambi-
tious attempt being described by Christopher McKee (b. 1942) and Ostriker in 1977 (McKee
and Ostriker, 1977). The very hot component at 106 K, the hot neutral gas at about 104 K
and the cool diffuse medium at about 100 K are roughly in pressure balance but, in addition,
there are the giant molecular clouds. The medium is constantly being buffeted by supernova
explosions, which may give rise to the formation and cooling of the giant molecular clouds.
The gas is also strongly perturbed by the gravitational influence of spiral arms, which results
in large density enhancements on the trailing edges of the arms.

9.4 The formation of stars

While impressive progress was being made in understanding the evolution of stars once
nuclear burning in their cores begins, the process of star formation proved a tougher propo-
sition. In 1945, Alfred Joy had drawn attention to the T-Tauri variable stars, which are
embedded in nebulous clouds of dust and gas (Joy, 1945). Although there are prominent
emission lines, these are superimposed upon an absorption-line spectrum consistent with
those of stars with mass roughly equal to the mass of the Sun. In 1947, Viktor Ambartsumian
(1908–1996) argued that the T-Tauri stars were low-mass main-sequence stars in the process
of formation (Ambartsumian, 1947), and in 1952 George Herbig (b. 1920) suggested that
they are low-mass stars which lay above the main sequence (Herbig, 1952). This idea was
proved correct by Merle Walker (b. 1926), who studied the extremely young star cluster
NGC 2264. In addition to luminous O and B stars, Walker studied the fainter T-Tauri stars
in the cluster and showed that they lay above the standard main sequence (Walker, 1956).
Walker interpreted these results as indicating that the T-Tauri stars were in the process of
contracting towards the main sequence.
The models which Walker had used to study the evolution of pre-main-sequence stars
were due to Salpeter, Henyey and their colleagues, and they assumed that, as the star
contracted quasi-statically towards the main sequence, energy transport through the star
226 9 The physics of the interstellar medium

was by radiation. This conclusion was shown to be incorrect by Chushiro Hayashi (b. 1920)
in 1961. Hayashi’s analysis concerned the stability of quasi-static models of stars and, in
particular, those in which energy transport is entirely by convection rather than by radiation
(Hayashi, 1961). He had already studied the stability of stars on the giant branch and
had shown that there is a limiting locus for stellar models when the energy transport by
convection extends throughout the whole of the star. This Hayashi limit is what eventually
stops the expansion of the envelopes of red giant stars. The corresponding Hayashi tracks
for stars on the Hertzsprung–Russell diagram are almost vertical and occur well to the
right of the diagram, in the region of low surface temperatures, T ∼ 3000−5000 K. As
Kippenhahn and Weigert remark,7

One may even say that the importance of the Hayashi track is only surpassed by that of the main
sequence.

Once a star evolves off the main sequence and moves across the Hertzsprung–Russell
diagram, its passage to the right is halted at the point at which it reaches the Hayashi limit,
when it is observed as a red giant star. The star then reorganises its internal structure and
moves up the Hayashi track (see Section A9.1).
The opposite process occurs during star formation. There are no quasi-static solutions
for stars to the right of the Hayashi track (see Section A9). Rather, the stars take up internal
structures in which the outward energy transport is by convection and they evolve down the
Hayashi track until energy transport by radiation becomes more important than convection
as the central temperature rises. Examples of the evolutionary tracks for stars of different
masses worked out by Hayashi in his famous paper are shown in Figure 9.4. Hayashi showed
that the T-Tauri stars lie precisely in the regions of the Hertzsprung–Russell diagram spanned
by his evolutionary tracks.
It was already apparent that star formation was associated with gas clouds in which there
are large amounts of dust, but the central role of dust in the star-formation process was only
appreciated when Eric Becklin and Gerry Neugebauer discovered what they identified as an
infrared protostar in the Orion Nebula (Becklin and Neugebauer, 1967). Neugebauer was
one of the pioneers of infrared astronomy, and, during the winter of 1965, a map was made
of the Orion Nebula by scanning across it with a single-element detector operating at 2.2 μm
with angular resolution 13 arcsec. Seven point sources were found which could be identified
with known stars, but one was not associated with any optical object. During the winter of
1966, photometric observations were made of this source at infrared wavelengths of 1.65,
2.2, 3.4 and 10 μm, and its temperature turned out to be 700 K, well below the surface
temperature of the coolest stars (Figure 7.15(a)). Observations were attempted of the same
object by Douglas Kleinmann (b. 1942) and Frank Low at a wavelength of 20 μm in the same
year, but they failed to find it. Instead, just to the south of the Becklin–Neugebauer object,
they discovered a source with an enormous far-infrared luminosity at the low temperature
of only 70 K (Kleinmann and Low, 1967). The story developed very rapidly from this point
onwards. Although the mapping of regions of star formation with single-element detectors
was time consuming, intense compact far-infrared sources were found in the vicinity of
regions of star formation. The far-infrared luminosities of some of these sources turned
out to be enormous, 103 –105 L  . Molecular-line surveys had revealed dense condensations
9.4 The formation of stars 227

logt 1=4.7
5.6
−1 3.6M
logt 2=5.8
3.1M
5.9 4.9
0 2.7M
5.1
6.2 2.3M
1 5.4
M bol 6.4 1.9M
5.7
2
6.5 1.5M 6.0
3
1.2M 6.3
6.7
4
6.9 M 6.6
5 main
sequence 0.6 M

6 0.4 M
7.3

8
4.2 4.1 4.0 3.9 3.8 3.7 3.6 7.7 3.5 3.4 3.3 3.2 3.1
log Te

Figure 9.4: Hayashi’s evolutionary tracks and ages for stars with different masses under gravitational
contraction; the times t1 and t2 denote the ages, in years, at the turning point and at the point where
the star joins the main sequence (Hayashi, 1961).

within giant molecular clouds and within dark clouds, and extremely luminous infrared
sources were often associated with these.
In their discovery paper, Becklin and Neugebauer (1967) remarked

It is well known that the Orion Nebula is a very young association and that the probability of finding
a star in the process of forming should be relatively high . . . Thus an attractive interpretation of the
observation is that the infrared object is a protostar.

Although this picture was attractive, it proved to be very difficult to demonstrate that the
Becklin–Neugebauer object is indeed a protostar. Part of the problem arose from the use
of words – what exactly was meant by a ‘protostar’? An unambiguous definition would
be that it is a star in which the energy is derived from the release of gravitational energy
of the matter accreted by the star and not from nuclear energy. In cases like the Becklin–
Neugebauer source, it was quite possible that an O or B star has already formed but that it
is still embedded deep within the dense dusty molecular cloud from which it formed.
The most important contribution to many aspects of these studies came from observations
with the Infrared Astronomy Satellite (IRAS), which was launched in 1983. IRAS was
designed to carry out the first deep surveys of the whole sky in the far-infrared wavebands
at 12, 25, 60 and 100 μm, which are inaccessible from the ground and showed that regions
of star formation are among the most intense far-infrared sources. These observations
confirmed beyond any doubt the significance of dust in the formation of stars. The dust acts as
228 9 The physics of the interstellar medium

a transformer, absorbing the optical and ultraviolet radiation of protostars and young stellar
objects and reradiating the absorbed energy at the temperature to which the dust is heated.
This proves to be an extremely efficient means of getting rid of the gravitational binding
energy of matter accreting onto a protostar since the envelope of the star is transparent to
radiation in the far-infrared wavebands.
The significance of the IRAS observations was that not only were luminous far-infrared
sources found in regions of star formation, but lower luminosity sources were discovered
as well with luminosities in the range 1–100L  and these are probably the progenitors of
solar-mass stars. The infrared spectra of the IRAS sources have provided a number of clues
to their evolutionary state. One of the most important early discoveries of the mission was
the detection of dust discs about young stars, which are interpreted as the material out of
which planets are formed. These discs could be identified by their spectral signatures in
T-Tauri stars.
The infrared spectra can be divided into three classes (Figure 9.5), the most extreme
class, Type I, consisting of exclusively far-infrared sources in which no optical radiation
is observed. These are interpreted as protostars that are still in their accretion phases.
The Type II spectra show evidence for dust emission from an accretion disc as well as
the photospheric emission from a reddened pre main-sequence star like the T-Tauri stars.
The Type III spectra appear to be unshrouded, pre main-sequence stars.
The theory of star formation has been one of the more contentious areas of modern
astrophysics because it can no longer be assumed that the star is in a quasi-static state
whilst evolving from a density enhancement in a giant molecular cloud to a fully fledged
main-sequence star. In 1902, James Jeans had worked out the condition for gravitational
collapse, namely that the force of gravity should exceed the force associated with the
pressure gradients within the cloud which resist the collapse, the famous Jeans criterion
for collapse (Jeans, 1902). On large enough scales, gravity will always be the dominant
force.
In the modern version of the theory, the collapsing cloud heats up as it collapses but,
so long as the cloud is optically thin to radiation, it can cool by radiation. Once the cloud
becomes optically thick to radiation, the cloud begins to heat up and the details of the collapse
have to be followed by computer simulations. Among the first of these were calculations by
Richard Larson (b. 1941) in 1969, who showed that a compact core forms within the collaps-
ing cloud and that the star forms by accretion of matter onto this core (Larson, 1969a,b). He
showed that the dust envelope would absorb the optical and ultraviolet radiation, leading to
the types of source which had just been discovered by Becklin and Neugebauer. These ideas
led in due course to a standard picture of the process of formation of spherically symmetric
stars due to Frank Shu (b. 1943) and his colleagues in 1980; this has been widely used
to interpret observations of the star-forming regions (Figure 9.6) (Stahler, Shu and Taam,
1980). The core of the protostar comes into hydrostatic equilibrium and matter is accreted
onto the core through an accretion shock. The binding energy of the accreted matter is
released as radiation, which is absorbed in the dust envelope, which cannot have a tempera-
ture greater than about 1000 K or else the dust grains are evaporated. The dust shell reradiates
away the absorbed energy in the far-infrared waveband. Shu and his colleagues carried out
radiative transfer calculations of the predicted spectra of these objects, and they can account
9.4 The formation of stars 229

36

35 04016+2610

34
log [vLv(r3)]

33

32

31

(a)
30 11 12 13 14 15
10

35

VSSG 23

34
log [vLv ]

33

32

(b)
31 15 16
11 12 13 14

35

34.5 SU Aur

34
log [vLv ]

33.5

33

32.5

(c)
32 14 16
11 12 13 15
log [v ]

Figure 9.5: The energy spectra of young stellar objects in the Taurus and Ophiuchus molecular clouds.
The three sources have mass of the order of 1M . These spectra are interpreted as follows. (a) The
protostar during the infall phase, in which the star and protoplanetary disc are embedded in an infalling
dust envelope. (b) Intense winds blow away the dust along the rotation axis, revealing the new-born
star surrounded by a nebular disc. (c) A T-Tauri star with only a small infrared excess. (Adams, Lada
and Shu, 1987.)

naturally for the details of the observed spectra of the far-infrared sources discovered by
IRAS.
There must, however, be much more to the story than this. In the late 1970s, maps
of young stellar objects were made at millimetre wavelengths, and in 1980 Ronald Snell
(b. 1951) and his colleagues discovered that, rather than showing evidence of infall, the
molecular gas in the source L1551 seemed to be expelled from the source in the form
of what became known as a bipolar outflow (Snell, Loren and Plambeck, 1980). These
molecular outflows have velocities up to about 150 km s−1 and extract large momentum and
230 9 The physics of the interstellar medium

Figure 9.6: Illustrating the structure of an accreting protostar according to Shu and his colleagues
(Stahler et al., 1980).

energy fluxes from the star, the total energies corresponding to about 1036 –1040 J. These
outflows are believed to be responsible for a variety of energetic phenomena seen in the
vicinity of young stellar objects, including the Herbig–Haro objects, high-velocity water-
maser sources, shock-excited molecular hydrogen and optically visible jets. The origin of
these outflows is not established, but they seem to be found wherever protostellar or young
stellar objects are found. It is suspected that they are associated in some way with two
other problems of star formation which have yet to be satisfactorily resolved, namely the
way in which protostellar clouds get rid of angular momentum and magnetic fields. In both
cases, the process of collapse amplifies the energies associated with both of them. It seems
reasonable to suppose that the bipolar outflows are in some way associated with the means
by which the protostellar cloud gets rid of its angular momentum and magnetic field, but
there is no generally agreed mechanism by which this occurs.
In 1987, Shu and his colleagues put together what has become a standard picture for the
formation of stars (Figure 9.7), which seems to account for many aspects of the observations
(Shu, Adams and Lizano, 1987).
One of the challenges of the theories of star formation was to account for the initial
mass function, which describes the rate of formation of stars of different masses, ξ (M).
The first analysis of this problem was carried out by Edwin Salpeter in 1955 and involved
corrections to the observed luminosity function of main-sequence stars for their different
main-sequence lifetimes (Salpeter, 1955). For stars with mass roughly equal to the mass of
9.4 The formation of stars 231

Figure 9.7: A plausible scenario for the formation of stars (Shu et al., 1987). (a) Density inhomo-
geneities collapse under their own gravity. (b) Main accretion phase, in which an accreting core has
formed and infall of matter onto the core takes place. The binding energy of the matter is removed by
radiation, which is absorbed by dust and reradiated in the far-infrared waveband. (c) Jets of material
burst out of the accreting star along its rotation axis, producing the characteristic bipolar outflows.
(d) The accretion of material ceases and the system is left with a young hydrogen-burning star and a
rotating dust disc.

the Sun, Salpeter found that the initial mass function could be approximated by a power-law
distribution in their masses:

dN = ξ (M) d(log M) ∝ M −1.35 d(log M). (9.1)

Salpter’s analysis needed estimates of the main-sequence lifetimes of stars of different

masses, and new determinations of the initial mass function were made as better stellar
models became available and the stellar statistics improved. In 1979, Glenn Miller (b. 1953)
and John Scalo (b. 1948) showed that the initial mass function was better defined by a log-
normal distribution over a wider range of masses (Figure 9.8) (Miller and Scalo, 1979). In
fact, over the mass range 1 ≤ M/M ≤ 10, the Salpeter initial mass function is a good
approximation to the function proposed by Miller and Scalo. One of the goals of the theory
of star formation is to account for this initial mass function, and the log-normal distribution
has the intriguing property of describing random multiplicative processes. These initial
mass functions are useful global averages for the rate of star formation in galaxies, but it
is not certain how accurately they describe star formation in individual regions. They are
essential for studying the evolution of the ultraviolet, optical and infrared spectra of galaxies
and also for studies of the chemical evolution of galaxies (see Section A13.1).
232 9 The physics of the interstellar medium

log ξ

−1 0 1 2
log (M/M )

Figure 9.8: An estimate of the initial mass function of stars derived by Miller and Scalo (1979)
showing their best-fitting log-normal distribution, ξ (log M) dM ∝ exp[−C1 (log M − C2 )2 ] dM (solid
line). Also shown (by the dashed line) is the initial mass spectrum of power-law form proposed by
Salpeter.

9.5 Extrasolar planets and brown dwarfs

The search for planets about nearby stars has been a cherished ambition of astronomers
for centuries, but it has proved to be a difficult challenge. In the 1950s, it was claimed that
the wobbles reported in the position of the nearby star, Barnard’s star, were evidence for
a planetary companion, but this turned out to be spurious. The first definite detection of
extrasolar planets, or exoplanets, came from a quite unexpected direction, the observation
of systematic variations in the radial velocity of the pulsar PSR 1257+12 by Alex Wolszczan
(b. 1943) and Dale Frail (b. 1961) in 1992 (Wolszczan and Frail, 1992). The reason for their
success was the very high precision with which radial velocities of pulsars can be determined
by very precise timing of the arrival times of the pulses. In fact, they found evidence for
three planets, two with masses roughly that of the Earth and one about 50 times less. This
was a wholly unexpected discovery since it was assumed that, when a neutron star forms,
any planets orbiting the pre-supernova star would not remain in bound orbits. A favoured
view is that these planets formed from an accretion disc about the neutron star after the
supernova exploded. In any case, the mere existence of this system shows that planets can
be formed in a wide range of different astronomical environments.
The first detection of a Jupiter-mass planet orbiting a normal star was made by Michel
Mayor (b. 1959) and Didier Queloz (b. 1966) of the University of Geneva in 1995 (Mayor
9.5 Extrasolar planets and brown dwarfs 233

Figure 9.9: The variation of the radial velocity of the star 51 Peg as a function of orbital phase. The
period of the planet’s orbit about the barycentre of the system is 4.231 days (Mayor and Queloz, 1995).

and Queloz, 1995). Their success can be attributed to the development of very stable
spectrographs with very high spectral resolution. For comparison, the Sun orbits the barycen-
tre of the Solar System at a typical speed of about 13 m s−1 . Thus, to have a hope of being
able to detect the presence of Jupiter-mass planets in systems such as our own, the spectro-
graphs have to be able to resolve radial velocities of a few metres per second. The discovery
record of the sinusoidal variation of the radial velocity of the nearby star 51 Peg is shown
in Figure 9.9, from which it can be seen that the amplitude of the motion of this solar-type
star is very much greater than would be expected of a planetary system such as our own.
Furthermore, the period of the planet about the star is only 4.231 days. Analysis of the
orbital data has shown that the mass of the planet is at least 0.46 Jupiter masses and its
semi-major axis is only 0.052 AU.
This discovery stimulated a huge effort to discover further examples of planets about
nearby stars, and it has been extraordinarily successful. By mid 2004, 122 extrasolar planets
were known in 107 planetary systems, including 13 multiple-planet systems.8 If there were
any doubts about the correctness of the interpretation of the radial velocity data, they were
dispelled by the observation in 1999 of a very small dip in the intensity of HD 209458 due
to the transit of the companion across the face of the star, these dips occurring with the
same period that was derived from the radial velocity data (Charbonneau et al., 2000). The
star HD 209458 is a G0 V dwarf star, similar to the Sun. Assuming the stellar radius and
mass are 1.1R and 1.1M , respectively, the ‘eclipse’ data shown in Figure 9.10 have been
interpreted as being due to the transit of a gaseous giant planet with radius 1.27 times the
radius of Jupiter in an orbit with an inclination of 87◦ .
234 9 The physics of the interstellar medium

Figure 9.10: The discovery record of the photometric time series for the star HD 209458 for 9 and
16 September 1999 plotted as a function of time. The data have been averaged in five-minute bins.
(Charbonneau et al., 2000).

These discoveries have resulted in two major surprises, which have forced the theory
of the formation of planetary systems to be considerably revised. The first was the large
fraction of Jupiter-like companions at orbital radii about 100 times closer to the parent
star than in our Solar System. For example, more than half of these gaseous giant planets
orbit within 1 AU of the host star and a significant fraction orbit within 0.1 AU. From the
statistics of the detected extrasolar planets, it seems that our Solar System is the odd man
out. A favoured solution is that, since such gaseous giants could not have formed so close
to the primary star, the Jupiter-sized planets must have been formed much further away and
then undergone orbital migration under the influence of tidal forces.
The second great surprise was the fact that the orbits of many of the Jupiter-sized planets
are highly elliptical. This poses problems for the standard picture of planet formation in
which the planets are formed by accretion in a protoplanetary disc. In this picture, dissipative
processes rapidly circularise the planetary orbit. Therefore, the elliptical orbits must have
come about through some other process. Suggestions have included that they formed directly
by gravitational condensation, rather than by accretion within a protoplanetary disc, or their
orbits may have been strongly perturbed by a companion star, which may have been the
case in a system such as 16 Cyg A and B, or maybe a sling-shot mechanism took place
which ejected the gaseous giant into an elliptical orbit through a gravitational encounter
with another planet.
The study of extrasolar planets is one of the major growth areas of modern astrophysics
and has opened the way to the study of bio-astrophysics and the possibility of determining
by observation whether conditions exist in other planetary systems in which life could
have formed. The challenge is to measure the spectrum of a very faint companion, about
100 000 times fainter than the star itself and very close to it. Remarkably, evidence has
already been found for the presence of an atmosphere in the extrasolar planet associated
with HD 209458. Observations with the very stable STIS spectrograph of the Hubble Space
Telescope showed that, when the planet transited across the face of the star, the depth of the
9.5 Extrasolar planets and brown dwarfs 235

(a) (b)
Figure 9.11: (a) The discovery image of the faint brown dwarf companion to the solar-type star
Gliese 229 obtained at the Palomar Observatory on 27 October 1994 (Nakajima et al., 1995).
(b) A confirmatory image taken by the Hubble Space Telescope on 17 November 1995. (Courtesy
of T. Nakajima, S. Kulkarni, S. Durrance and D. Golimowski, NASA, ESA and the Space Telescope
Science Insitute.)

absorption lines of sodium, the D-lines, increased significantly. This was direct evidence
for the presence of sodium in the atmosphere of this extrasolar planet (Charbonneau et al.,
2002).
A closely related study has been the search for the class of ‘star’ known as brown dwarfs.
These stars have masses less than about 0.08M and so their central temperatures are too
low for the nuclear burning of hydrogen into helium to take place in their cores. There is
thus a range of masses between those of Jupiter-like gaseous giant planets and objects about
1000 times more massive in which the central temperatures are not great enough to tap their
nuclear energy resource. Conventionally, brown dwarfs are taken to be gravitationally bound
objects with masses in the range 0.01 to 0.08M , that is about 10 to 80 Jupiter masses. The
lower limit corresponds to the mass at which even deuterium burning ceases to be possible
in the core of the ‘star’.
Searches for brown dwarfs have been pursued for many years, but the first really con-
vincing case was discovered in 1995 by direct imaging of the companion of the nearby star
Gleise 229 at Palomar using the 60-inch and 200-inch telescopes (Figure 9.11) (Nakajima
et al., 1995). The spectrum of this faint companion, known as Gliese 229B, showed strong
methane and water vapour absorption, similar to the spectrum of Jupiter, the inferred surface
temperature being less than 1000 K (Oppenheimer et al., 1995). This surface temperature
is too low for nuclear burning to take place in its core and so it must be a bone fide brown
dwarf. Many further candidates have since be found through systematic sky surveys in the
near-infrared wavebands, including the 2 Micron All-Sky Survey (2MASS) and the Sloan
Digital Sky Survey. Numerous candidates have also been found in deep-infrared surveys of
nearby star-forming regions, such as the Pleiades, Orion and ρ Ophiuchus clusters.
236 9 The physics of the interstellar medium

9.6 Cosmic-ray astrophysics and the interstellar medium

The history of cosmic-ray astrophysics and its role in the discovery of elementary particles
up to the period immediately after the Second World War was summarised in Section 7.2. As
described by Michael Hillas9 (b. 1932), after the War, cosmic-ray studies were vigorously
pursued from high-altitude balloons and by ground-based studies of extensive air-showers.
The problem for the cosmic-ray physicists involved in the balloon studies was to discrimi-
nate between the primary cosmic rays and the vast numbers of secondary particles created
by interactions of very-high-energy cosmic rays with the nuclei of molecules of the atmo-
sphere. Balloons were flown to progressively greater altitudes, until, at about 36 km, the
residual atmosphere10 amounted to only about 50 kg m−2 , which was much less than the
interaction mean free path for the incoming cosmic rays, which was about 800 kg m−2 ; for
reference, the total depth of the atmosphere is about 10 000 kg m−2 . The detectors used in
these balloon experiments consisted of stacks of nuclear emulsions, and these were reg-
ularly flown at high altitude. Provided the stacks were safely returned to the ground, the
highly developed techniques for the analysis of nuclear emulsion stacks provided informa-
tion about the energies and charges of the cosmic rays. From the 1960s onwards, cosmic-ray
detectors were successfully flown in satellites and space probes, eliminating the problems
of contamination by secondary particles.11 The technologies employed to build the tele-
scopes and detectors were much more akin to those used in high-energy particle physics
experiments. The instruments were no longer retrievable and so they had to be miniaturised
and space-qualified to survive the harsh environment of space.
The majority of the incoming cosmic rays were protons, but it was immediately apparent
that the cosmic radiation also included fluxes of relativistic helium and heavier atomic nuclei
(Freier et al., 1948). To complicate the problem, the spectrum of the cosmic rays with kinetic
energies less than about 1 GeV per nucleon were strongly influenced by their passage through
the interplanetary medium from interstellar space, and these changes varied with the phase
of the solar cycle, the phenomenon known as solar modulation. The distortions are caused
by the scattering of the cosmic rays by irregularities in the interplanetary magnetic field.
The variation with the solar cycle is caused by variations in the amplitude of the spectrum
of irregularities, more scattering occurring at solar maximum than at solar minimum.
The key results for studies of the astrophysics of our Galaxy and the interstellar medium
were the energy spectrum and chemical composition of the primary cosmic radiation. The
energy spectra of the cosmic rays at energies greater than 1 GeV per nucleon were found to
follow a power-law energy distribution:
N (E) dE ∝ E −x dE, (9.2)
with x ∼ 2.5–2.7. This relation was found for protons, helium and heavier nuclei with
energies in the range 109 –1014 eV (Figure 9.12) (Simpson, 1983). Such a relativistic gas
was inferred to be present throughout the interstellar medium from two quite independent
types of observation. Firstly, the Galactic γ -ray emission at energies E > 100 MeV can
be attributed to the decay of neutral pions, π 0 , created in collisions between cosmic-ray
protons and nuclei with the nuclei of atoms, ions and molecules in the interstellar gas
(see Section 7.5). The agreement between these estimates of the local number density of
9.6 Cosmic-ray astrophysics and the interstellar medium 237

Figure 9.12: The differential energy spectrum of cosmic rays as measured from above the Earth’s
atmosphere as summarised by John Simpson (1916–2000) (Simpson, 1983). The solid line shows an
estimate of the proton spectrum once allowance is made for the effects of solar modulation.

cosmic-ray particles and those of the cosmic-ray particles observed at the top of the atmo-
sphere showed that the latter are part of the population of high-energy particles pervading
the whole Galaxy.
Secondly, the synchrotron radiation of the interstellar population of ultra-relativistic
electrons gyrating in the Galactic magnetic field is detected in the radio waveband as
the Galactic radio emission (see Section 7.3.2). It proved a much more challenging task to
measure the primary cosmic-ray electron spectrum since large fluxes of relativistic electrons
are produced as secondary particles, even at very high altitudes. For a number of years there
was uncertainty about the exact form of the primary electron spectrum, but the observations
of William Webber (b. 1929) and his colleagues finally showed that the energy spectrum
of very-high-energy electrons with E ≥ 10 GeV is of power-law form similar to equation
(9.1), with a slightly greater spectral index, x = 3.3 (Webber, 1983). This spectrum can
be smoothly joined onto the spectrum of high-energy electrons inferred to be present in
interstellar space for reasonable values of the interstellar magnetic field strength.12 These
238 9 The physics of the interstellar medium

Figure 9.13: The cosmic abundances of the elements in the cosmic rays (solid line) compared with the
Solar System abundances (solid histogram). The data have been normalised to a relative abundance
of hydrogen of 1012 (Lund, 1984).

data are compelling evidence that the primary cosmic rays detected in the vicinity of the
Earth are samples of the relativistic gas present in the interstellar medium of our Galaxy.
Following the discovery of heavy nuclei in the cosmic rays, their chemical composition
was measured in balloon observations by the Minnesota–Rochester group, and the key
results were reported by Helmut Bradt (d. 1950) and Bernard Peters (1910–1993) in 1950
(Bradt and Peters, 1950). The important discovery was the large abundance of the light
elements lithium, beryllium and boron relative to their Solar System abundances. The study
of the detailed chemical abundances in the cosmic rays continued through the following
decades from balloons and satellites, culminating in the measurements by the HEAO-C
space observatory, which was launched in 1979. The abundances of the elements in the
cosmic rays relative to the Solar System abundances are illustrated in Figure 9.13, which
is taken from the summary by Niels Lund (b. 1938) (Lund, 1984). It can be seen that the
abundance peaks at the carbon, nitrogen and oxygen group and the iron group are present
in both the Solar System and cosmic-ray abundances. The excess of lithium, beryllium
and boron in the cosmic rays is very great indeed relative to their cosmic abundances,
and there is also an excess of elements with atomic and mass numbers just less than iron.
These differences are naturally attributed to the process of spallation in the interstellar
medium between the sources of the cosmic rays and the arrival of these particles at the
top of the Earth’s atmosphere. In this process, energetic cosmic-ray nuclei collide with the
protons and nuclei in the interstellar medium and nucleons are chipped off, resulting in
the formation of lighter nuclei with smaller mass numbers. To carry out a detailed analysis
of the products of these nuclear interactions, spallation cross-sections are needed which
describe the probability of the formation of different secondary nuclei in each collision.
Thus, the primary source of the lithium, beryllium and boron is the spallation of the abundant
9.6 Cosmic-ray astrophysics and the interstellar medium 239

at sources

surviving
180 36 primordial

160 32 arriving
secondaries

140 28

120 24

100 20

80 16

60 12

40 8

20 4

6 7 7 9 10 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 36 40 52 54 56
Li Li Be Be Be B B C C N N O O O F Ne Ne Ne Na Mg Mg Mg Al Si Si Si P S S S Ar Ca Cr Fe Fe

Figure 9.14: The relative abundances of the cosmic rays as observed near the Earth and as inferred to
have been present in their sources, once account has been taken of the effects of spallation between
their sources and the Solar System. The grey histogram shows the inferred source abundances; the
black histogram shows the spallation products; and the white histogram shows the surviving primary
elements (Shapiro, 1991).

carbon, nitrogen and oxygen group, while the elements just lighter than iron are naturally
formed by the spallation of iron group elements.13
In their pioneering paper, Bradt and Peters showed quantitatively that the lithium,
beryllium and boron in the cosmic rays could be attributed to the spallation of carbon
group elements (Bradt and Peters, 1950). Since that time, these studies have been the
subject of increasingly refined calculations which take account of the range of diffusion
path-lengths between the sources of the cosmic rays and the Earth, as well as adopt-
ing much improved spallation cross-sections. The computations start from the transfer
equation for the evolution of the number densities of all types of primary and secondary
nuclei as they travel through the interstellar medium. The typical result of this type of
calculation is shown in Figure 9.14, which is taken from the survey by Maurice Shapiro
(b. 1915) (Shapiro, 1991). On average, the cosmic rays must have traversed about 50 kg m−2
between their sources and the Solar System, but it is also necessary to assume that there
is a distribution of path-lengths in order to obtain the correct observed abundances of
the products of spallation of both the carbon and iron group elements. These calcula-
tions show how the observed abundances of lithium, beryllium and boron and elements
such as 15 N, 17 O, 18 O, 19 F and 21 Ne can all be accounted for by spallation interactions
between the sources of the cosmic rays and the Solar System. It can also be seen that
quite large fractions of the common elements, carbon, oxygen, neon, magnesium and iron,
survive intact between their sources and the Solar System. These computations show that
the cosmic rays were accelerated with chemical abundances similar to the Solar System
abundances.
240 9 The physics of the interstellar medium

Figure 9.15: (a) The isotopic mass distribution of the isotopes of beryllium as observed by the cosmic-
ray telescopes onboard the IMP-7 and IMP-8 space probes. (b) Calibration of the expected distribution
of beryllium isotopes in laboratory experiments showing the resolution of the isotopes of 9 Be and
10
Be. (Garcia-Munoz et al., 1977.)

One of the important issues for the propagation of cosmic rays in the interstellar medium
is the typical time they take to diffuse from their sources to the Solar System. Fortunately,
some of the species created in the spallation interactions are radioactive and so can act
as cosmic-ray clocks. The most important of these is the 10 Be isotope, which has half-life
3.9 × 106 years. Its abundance can be compared with those of the other stable isotopes
of beryllium. This experiment requires excellent resolution of the relative abundances of
the isotopes of beryllium, and this was successfully achieved first by Simpson and his
colleagues in their experiments onboard the IMP-7 and IMP-8 spacecraft in 1977 (Garcia-
Munoz, Mason and Simpson, 1977). Figure 9.15 shows that there is very little 10 Be in the
cosmic rays and consequently there has been time for most, but not all, of these isotopes
to undergo radioactive decay. These data indicate that the average time it takes the cosmic
rays to reach the Earth is about 107 years. Thus, the cosmic rays cannot have travelled
directly from their sources to the Solar System, but rather they must have been scattered
Notes to Chapter 9 241

many times by irregularities in the interstellar magnetic field. It is simplest to think of the
cosmic rays performing a random walk from their sources to the Solar System. Then, if
the interstellar density is taken to have a typical value of 3 × 105 m−3 and the cosmic rays
traverse 50 kg m−2 , the distance travelled at the speed of light would be 107 light-years, in
pleasant agreement with the results of the 10 Be experiment.
These observations indicate that relativistic matter and magnetic fields are important
components of the interstellar medium. They have roughly equal energy densities, a conve-
nient figure being that both correspond to about 1 eV m−3 .

Notes to Chapter 9

1 These works are summarised in Aller’s classic book Atoms, Stars and Nebulae, 3rd edn (Cam-
bridge: Cambridge University Press, 1991).
2 Some reminiscences of Oort and the pursuit of Dutch astronomy under the occupation are contained
in Jan Oort Astronomer, eds J. Katgert-Merkelijn and J. Damen (Leiden: Kleine publicaties van
de Leidse Universiteitsbibliotheek no. 35, 2000). The minutes of the underground seminar in
which van de Hulst first reported the prediction of the 21 cm line are reproduced in facsimile
in D. Hartmann and W. B. Burton, Atlas of Galactic Neutral Hydrogen (Cambridge: Cambridge
University Press), p. 84. On p. 85, the following translation is given: ‘Then Mr Van de Hulst
speaks about the spectrum of the Milky Way at wavelengths of several metres . . . Furthermore,
one may perhaps expect an observable intensity at several discrete wavelengths as a consequence
of transitions between hyperfine structure levels of the ground state of hydrogen . . . After this the
chairman [Oort] thanks Mr Van de Hulst for his fine presentation.’
3 Many of these early discoveries using 21 cm hydrogen line observations were summarised at the
Paris Symposium on Radio Astronomy (Bracewell, 1959) and in the survey of HI observations by
Wim Rougoor (1931–1967) and Oort (Rougoor and Oort, 1960).
4 An introduction is given by W. W. Duley and D. A. Williams, Interstellar Chemistry (London:
Academic Press, 1984).
5 Kroto gives a splendid blow-by-blow account of the discovery of the C60 molecule and the role
of interstellar molecules in its discovery in his Nobel prize lecture of December 1996, Symmetry,
Space, Stars and C60 . Kroto, Robert Curl (b. 1933) and Richard Smalley were jointly awarded the
Nobel prize for chemistry in 1996 ‘for their discovery of fullerenes’.
6 See, for example, F. C. Bruhweiler and A. Vidal-Madjar, in Exploring the Universe with the IUE
Satellite, ed. Y. Kondo (Dordrecht: D. Reidel Publishing Company, 1987), pp. 467–484.
7 See R. Kippenhahn and A. Weigert, Stellar Structure and Evolution (Berlin: Springer-Verlag,
1990).
8 An excellent online encyclopaedia of all the known extrasolar planets is maintained by Jean
Schneider, entitled The Extrasolar Planet Encyclopaedia at http://www.obspm.fr/planets. This
web-site also includes many excellent references to different detection methods for discovering
extrasolar planets and to theories of the origin of the types of planetary system discovered.
9 Hillas provides an excellent documentary study of the development of cosmic-ray astrophysics
from its beginnings up to about 1970. His book describes clearly the many technical difficulties
which had to be overcome before reliable estimates could be made of the spectrum, isotropy and
chemical composition of the cosmic rays. See A. M. Hillas, Cosmic Rays (Oxford: Pergamon
Press, 1972).
10 The units used to describe the amount of material traversed by the cosmic ray is the path length,
ρ dx, where ρ is the density of the material and dx is the increment of distance. The traditional
units used are g cm−2 , but I have translated these into SI units. For reference, the total depth of the
atmosphere is 10 000 kg m−2 .
242 9 The physics of the interstellar medium

11 I have given details of the different types of cosmic ray detectors and telescopes in Malcolm Lon-
gair, High Energy Astrophysics, vol. 1 (Cambridge: Cambridge University Press, 1994). Several
examples are given of the types of data obtained in these experiments.
12 I have discussed the problems of relating the spectrum and emissivity of the Galactic radio emission
to the local spectrum of high-energy electrons in Section 18.2 of Malcolm Longair, High Energy
Astrophysics, Volume 2 (Cambridge: Cambridge University Press, 1994).
13 I have discussed the problems of accounting for the observed abundances of spallation nuclei in
Section 20.2 of Malcolm Longair, High Energy Astrophysics, Vol. 2.

A9 Explanatory supplement to Chapter 9

A9.1 Notes on the Hayashi track
In Hayashi’s pioneering paper, his analysis concerns the stability of fully convective stars.
The condition that a region of a star is in convective, rather than radiative, equilibrium
is that the temperature gradient exceeds the adiabatic gradient of the stellar material. In
this context, the term ‘gradient’ refers to the derivative of the temperature with respect to
pressure, which is a monotonically increasing function of decreasing radius within the star.
Conventionally, the temperature gradient is written, in the case of the radiative transport of
energy, as
 
d ln T
∇rad = (A9.1)
d ln p rad
and is related to the opacity of the stellar material as discussed in Section A3.2.2. If the stellar
material has ratio of specific heat capacities γ , the adiabatic relation is p ∝ T γ /(γ −1) and so
∇ad = (γ − 1)/γ . If the structure of the star is such that the temperature gradient exceeds
this value, the material of the star becomes unstable and convection ensues. The simplest
picture of what happens physically is to consider a ‘bubble’ of material that is slightly
compressed and which then rises up the temperature gradient because of the buoyancy of
the perturbed region. Convection transports energy more rapidly through the star than does
radiation, and the structure of the star reorganises itself under these convective motions
until the temperature and pressure stratification satisfy the relation ∇ad = (γ − 1)/γ . Thus,
for stars in which convection is maintained throughout the whole star, the temperature and
pressure stratification is given almost exactly by the adiabatic gradient, since even a tiny
departure to greater values of ∇rad results in convective motions.
In Hayashi’s brief paper, he considers the stability of the polytropic models of stars in
which the adiabatic relation holds throughout the star (Hayashi, 1961). If the perfect gas ratio
of specific heats is used, γ = 5/3, these models have polytropic index n = (γ − 1)−1 = 3/2.
He then showed that there is an upper limit to a dimensionless parameter involving the
mass, radius, temperature and pressure of the gas, beyond which there exist no quasi-static
solutions. This condition translates into the steep regions of the loci shown in Figure 9.4 for
stars of different mass. There are no quasi-static solutions to the right of these loci.
A more detailed physical discussion of the structure of fully convective stars is given by
Kippenhahn and Weigert,1 who show that, by adopting the adiabatic relation for the structure
of the star, the pressure and temperature structure are decoupled from its luminosity. An
Note to Section A9 243

log L/L

5
M/M =10

4 5

3 1
ma

0.5
in
se

2
qu
en
ce

−1

4.5 3.5
log Teff

Figure A9.1: Theoretical Hayashi tracks for fully convective stars of different masses presented by
Kippenhahn and Weigert, after computations by Ezer and Cameron. From R. Kippenhahn and A.
Weigert, Stellar Structure and Evolution (Berlin: Springer-Verlag, 1990), Chapter 24.

atmosphere in which energy transport is by radiation is needed to determine the luminosity–

temperature relation, and this is found by joining together solutions for the body of the star
to its atmosphere. These calculations demonstrate the steepness of the Hayashi track for
stars of different mass (Figure A9.1).

Note to Section A9

1 See R. Kippenhahn and A. Weigert, Stellar Structure and Evolution (Berlin: Springer-Verlag,
1990).
10 The physics of galaxies and
clusters of galaxies

10.1 The galaxies

The Hubble sequence of galaxy types shown in Figure 5.5 gives some impression of the
diversity of forms found among the galaxies. Hubble planned to publish an atlas of galaxies
illustrating the different galaxy types but, although all the plates for this project were taken
with the 60-inch and 100-inch telescopes by 1948, he died in 1953, before what became the
Hubble Atlas of Galaxies was published. The project was completed by Allan Sandage, who
was Hubble’s last research assistant, and it was published in 1961 (Sandage, 1961b). The
basic Hubble sequence was preserved, including the S0 galaxies, and the irregular galaxies
were placed at the end of the sequence.1
The morphological classification of large samples of galaxies was pursued by Antoinette
(1921–1987) and Gérard de Vaucouleurs (1918–1995), who published a series of Reference
Catalogues of Bright Galaxies, in which the Hubble classification was refined, the basic
linear sequence being preserved (de Vaucouleurs et al., 1991). The distinction between
the normal and barred spirals was maintained, but they showed that all intermediate types
between pure barred spirals and normal spirals are also observed. What gave this morpho-
logical scheme physical significance was the fact that certain physical properties of galaxies
are correlated with their position along the sequence. In de Vaucouleurs’ survey of 1974, he
showed that the mass fraction in the form of gas is a function of position along the sequence,
the elliptical galaxies having less than 0.01% of their mass in the form of interstellar gas,
whereas in the irregular galaxies as much as 30% of the mass can be in gaseous form (de
Vaucouleurs, 1974). The colours of the galaxies also showed a systematic trend along the
sequence, the ellipticals being the reddest and the irregulars the bluest galaxies.2
The availability of the Palomar 48-inch sky survey plates led Halton Arp (b. 1927)
to publish his Atlas of Peculiar Galaxies in 1966 (Arp, 1966). As Arp remarked in his
introduction to the Atlas,
The greatest deviations from the normal are emphasised in this atlas.

The corresponding catalogue for the southern hemisphere was published in 1987 (Arp,
Madore and Roberton, 1987). Similar catalogues were prepared by Boris Vorontsov-
Velyaminov (1904–1994) entitled Atlas and Catalogue of Interacting Galaxies, published in
two parts in 1959 and 1977 (Vorontsov-Velyaminov, 1959, 1977). In both cases, the authors
drew attention to the fact that there are many pathological types of galaxies which do not fall

244
10.1 The galaxies 245

−4.5

−1.5

1.5

4.5

7.5

Figure 10.1: Illustating a prograde close encounter between two spiral galaxies modelled by Juri and
Alar Toomre (Toomre and Toomre, 1972).

naturally into the Hubble sequence of types. In many cases these strange structures could
be interpreted as strong gravitational interactions, or collisions, between galaxies, and in
1972 Juri (b. 1940) and Alar Toomre (b. 1937) showed how even some of the strangest
images could be accounted for by such close encounters (Toomre and Toomre, 1972). As
an example, Figure 10.1 shows how it is possible to produce long tails and bridges between
galaxies as a result of gravitational interactions between them. In this example, the tails are
associated with a prograde collision between two spiral galaxies, the outer rings of stars in
246 10 The physics of galaxies and clusters of galaxies

the spirals being ripped off because, in the prograde collision, the stars in the outer regions
feel the same accelerating force for an extended period of time. It is generally the case that
most of the objects in the catalogues of peculiar galaxies are the results of collisions or
close encounters between normal galaxies and not new types of galaxy. As de Vaucouleurs
(1974) remarked,
After a collision, a car is a wreck, not a new type of car.

The study of these galaxies emphasised the importance of interactions between galaxies
in their evolution. The interactions were often associated with large amounts of dust and hot
gas as the interstellar material in the galaxies came together. A great deal of star formation
was to be expected, and this was confirmed by the surveys conducted by the IRAS satellite
in the mid 1980s, which showed that many of the most luminous IRAS galaxies in the
far-infrared wavebands are those which involve strongly interacting galaxies.3
The masses of galaxies range from systems which have mass only about 107 M to super-
giant elliptical galaxies which, in the most extreme cases, have masses as great as 1013 M .
The dwarf galaxies were recognised by Zwicky in the late 1930s through observations made
with the Palomar 18-inch Schmidt telescope (Zwicky, 1942). Although there are some vari-
ations, the distribution of luminosities among galaxies of all types per unit volume can be
remarkably well described by the form of luminosity function introduced by Paul Schechter
in 1976 (Schechter, 1976):
(L) dL = AL −α exp(−L/L ∗ ) dL , (10.1)
where L is the luminosity of the galaxy and L ∗ is a characteristic ‘break’ luminosity – for
greater values, the numbers of galaxies diminish exponentially.
In 1977, James Felten (b. 1934) surveyed a large number of different determinations of
the luminosity function for galaxies and showed that, once account was taken of the different
corrections and assumptions made by different authors, they were all consistent with the
form of luminosity function proposed by Schechter (Figure. 10.2) (Felten, 1977). The value
of L ∗ , corresponding to the break in the luminosity function, has value L ∗ ≈ 1010 L  . The
index α was found to have a value of about 0.25, indicating that the luminosity function has
a long tail which extends to the dwarf galaxies. Our own Galaxy has luminosity about 0.5L ∗ ,
and so it is typical of bright spiral galaxies found in statistical samples, but is by no means
among the most luminous galaxies known. Integration over the luminosity function, equation
(10.1), shows that most of the background light in the Universe is produced by galaxies
with L ∼ L ∗ . Both spiral and elliptical galaxies span the whole range of luminosities, but
the irregular galaxies are mostly found with L < L ∗ . While the Schechter function is, in
general, a good description of the probability distribution of galaxy luminosities, it does not
give a good description of the most luminous galaxies in clusters, which are anomalously
luminous (see Section 10.5.1).
Extensive studies have been made of correlations between various properties of elliptical
galaxies, specifically their luminosities, sizes, central velocity dispersions, their abundances
of heavy elements and so on. Of these, two studies are of particular importance. The first is
the analysis of Sandra Faber (b. 1944) and Robert Jackson (b. 1949), who in 1976 discovered
a strong correlation between luminosity, L, and central velocity dispersion, σ , of the form
10.1 The galaxies 247

−1

−2

−3
log (φ (MB(0))

−4

−5

−6

−7

−24 −23 −22 −21 −20 −19 −18 −17 −16 −15 −14
Absolute B(0) magnitude
Figure 10.2: The luminosity function for galaxies derived by Felten in 1977 using data from a large
number of different surveys of galaxies (Felten, 1977). The solid line shows the best-fitting Schechter
function to the data.

L ∝ σ x , where x ≈ 4 (Faber and Jackson, 1976). This correlation has been studied by other
authors, who have found values of x ranging from about 3 to 5. The significance of this
relation is that, if the velocity dispersion, σ , is measured for an elliptical galaxy, its intrinsic
luminosity can be inferred from the Faber–Jackson relation and hence its distance found.
This procedure was refined in 1987 by Alan Dressler (b. 1948) and his colleagues and
by George Djorgovski (b. 1956) and Marc Davis (b. 1947), who introduced the concept of
the fundamental plane for elliptical galaxies (Djorgovski and Davis, 1987; Dressler et al.,
1987). The fundamental plane lies in a three-dimensional space in which luminosity, L,
is plotted against central velocity dispersion, σ , and the surface brightness, e , within the
half-light radius. An even stronger correlation than the Faber–Jackson relation was found
when the surface brightness was included:

L ∝ σ 8/3 e−3/5 . (10.2)

This empirical formula enables the distances of elliptical galaxies to be determined indepen-
dent of their redshifts. Dressler and his colleagues estimated that, using these correlations,
the distances of individual galaxies can be determined to about 25% and for clusters of
galaxies to about 10%.
In 1975, Brent Tully (b. 1943) and Richard Fisher (b. 1943) discovered that, for spiral
galaxies, the widths of the profiles of the 21 cm line of neutral hydrogen, once corrected for
the effects of inclination, are strongly correlated with their intrinsic luminosities (Tully and
Fisher, 1977). In their studies, they correlated the total B luminosities with the corrected
velocity width, V , of the 21 cm line and found the relation

L B ∝ V α , (10.3)
248 10 The physics of galaxies and clusters of galaxies

where α = 2.5. A much larger survey carried out by Marc Aaronson (1950–1987) and
Jeremy Mould (b. 1949) in 1983 found a somewhat steeper slope, α = 3.5, for luminosities
measured in the optical B waveband, and an even steeper slope, α = 4.3, in the near-infrared
H waveband at 1.65 μm (Aaronson and Mould, 1983). This correlation, the infrared Tully–
Fisher relation, is very much tighter in the infrared than in the blue waveband because
the luminosities of spiral galaxies in the latter waveband are significantly influenced by
interstellar extinction within the galaxies themselves – in the infrared waveband the dust
becomes transparent. As a result, measurement of the 21 cm velocity width of a spiral galaxy
can be used to infer its absolute H magnitude, and hence, by measuring its flux density,
its distance can be estimated. This procedure has resulted in some of the best distance
estimates for spiral galaxies and has been used in programmes to measure the value of
Hubble’s constant.

10.2 Dark matter in galaxies

In the simplest picture, the distribution of light in spiral galaxies can be decomposed into two
components, the bulge and the disc (Figure 5.8). The discs are in centrifugal equilibrium,
and so measurements of the projected rotational velocity as a function of distance from
the centre provides dynamical information about the mass distribution in the galaxy. In the
optical waveband, these are very demanding spectroscopic observations because the stellar
absorption features in the spectra of galaxies are weak and their surface brightnesses are
low. To circumvent this problem, Margaret and Geoffrey Burbidge and Kevin Prendergast
used the narrow emission lines of ionised gas clouds in the discs of spiral galaxies as tracers
of the velocity distribution. The masses and mass-to-light ratios for the galaxies NGC 5866
and NGC 681 are good examples of these pioneering studies (Burbidge, Burbidge and
Prendergast, 1960, 1965). It is conventional to describe the masses of galaxies in terms of
the ratio of mass to luminosity, normalised to this ratio for the Sun, M /L  . Typically, mass-
to-light ratios in the range 2 to 5 were found, which could be explained by assuming that,
within the radius to which the rotation curve had been determined, the light was dominated
by old stars somewhat less massive than the Sun.
These observations referred to the central regions of galaxies, but it proved very much
more difficult to study the faint outer regions using photographic techniques. From the late
1970s onwards, image tubes and CCD detectors became available which greatly increased
the capability for two-dimensional spectroscopy with long spectrographic slits. Vera Rubin
(b. 1928) and her colleagues pioneered systematic studies of the rotation curves of galaxies
using very long slits, which enabled rotation curves to be determined throughout the bodies
of the galaxies, the narrow emission lines again being the preferred tracer of the velocity
fields (Rubin, Thonnard and Ford, 1980).4 This work was complemented by observations
of the 21 cm line of neutral hydrogen, which enabled the velocity curves of spiral galaxies to
be determined to much greater radial distances than the optical observations (Figure 10.3).
Both types of observation showed that, in the outer regions of galaxies, the velocity
curves are generally remarkably flat, vrot ≈ constant, as far as the rotation curves could
be measured. The significance of this result can be appreciated from a simple Newtonian
10.2 Dark matter in galaxies 249

Figure 10.3: Examples of the rotation curves of spiral galaxies from optical and 21 cm neutral hydro-
gen observations (Bosma, 1981).

calculation. If the galaxy is taken to be spherical and the mass within radius r is M(<r ),
the circular rotational velocity at distance r is found by equating the inward gravitational
acceleration, G M(<r )/r 2 , to the centripetal acceleration, vrot
2
/r , and so
 1/2
G M(<r )
vrot = . (10.4)
r

Thus, if vrot = constant, it follows that M(<r ) ∝ r so that the total mass within radius
r increases linearly with distance from the centre. This result contrasts strongly with the
variation of the surface brightness distributions of spiral galaxies, which decrease much
more rapidly with distance from the centre than as r −2 . Kenneth Freeman (b. 1940) found
that models in which the luminosity per unit surface area decreases exponentially with radius
provided an excellent fit to these data (Freeman, 1970). To rephrase this important result,
the mass-to-light ratio must increase dramatically in the outer regions of spiral galaxies.
The conventional way of expressing this result is to state that there must be a large amount
250 10 The physics of galaxies and clusters of galaxies

of dark matter in the haloes of galaxies. A typical figure for giant spiral galaxies is that they
must contain about ten times as much dark as visible matter.5
An astrophysical argument for the existence of dark matter haloes in spiral galaxies was
presented in 1973 by Jeremiah Ostriker (b. 1937) and James Peebles (b. 1935), who pointed
out that rotating discs are subject to a bar instability, unless a stabilising halo is present
which contains a significant fraction of the mass of the system (Ostriker and Peebles, 1973).
Their criterion for stability was that the ratio of the ordered kinetic energy of the disc, Torb ,
to the total potential energy, |U |, should be less than 0.14, a result confirmed by subsequent
analytic and numerical analyses. A halo of dark matter about spiral galaxies would provide
stabilisation of the galactic disc.
The presence of about ten times more dark than visible matter was also found in the
nearby giant elliptical galaxy M87 by John Huchra (b. 1948) and Jean Brodie (b. 1953) in
1987 (Huchra and Brodie, 1987) from the velocity dispersions of globular clusters in the
halo of the galaxy. A quite different diagnostic tool for studying the mass distribution in
galaxies was first used in 1980 by Daniel Fabricant (b. 1952) and his colleagues, who used
the X-ray surface brightness distribution of elliptical galaxies as a means of determining the
gravitational potential of the system (Fabricant, Lecar and Gorenstein, 1980) (see Section
A10.1). Once again, it was inferred that there must be more mass present than would be
deduced from the optical surface brightness distribution.6 We will find the same result for
the mass-to-luminosity ratios of rich clusters of galaxies (see Section 10.4).
The nature of the dark matter in galaxies remains unknown. Many ideas have been
proposed, including various types of baryonic and non-baryonic dark matter. Examples
of low-luminosity discrete objects include planets, brown dwarfs, low-mass stars, neutron
stars and small black holes. Non-baryonic examples include the types of particles predicted
by particle theorists, for example the lightest supersymmetric particle, massive neutrino-
like particles or the gravitino – these types of particles are known collectively as weakly
interacting massive particles, or WIMPs; these have not yet been detected in particle physics
experiments. The possibility that some form of WIMP exists gains credence from studies
of the origin of the large-scale structure of the Universe (see Chapter 15).
Like the spiral galaxies shown in Figure 10.3, the rotation curve of our own Galaxy is
remarkably flat at radii r > 3 kpc (Fich and Tremaine, 1991), and so it is expected that there
should be a dark matter halo about our Galaxy.7 One of the most impressive approaches
to setting limits to the contribution which discrete low-mass objects, collectively known
as massive compact halo objects, or MACHOs, could make to the dark matter in our own
Galaxy has been the search for the gravitational microlensing signatures of such objects
as they pass in front of background stars. These are very rare events and so very large
numbers of background stars have to be monitored. The beauty of this technique is that it
is sensitive to MACHOs with a very wide range of masses, from 10−7 M to 100M , and
the contributions of a very wide range of candidates for the dark matter can be constrained.
Two very-large-scale projects, the MACHO and the EROS projects, have made systematic
surveys over a number of years to search for these events. The MACHO project, which
ran from 1992 to 1999, used stars in the Magellanic Clouds and in the Galactic bulge as
background stars, and millions of stars were monitored regularly (Alcock et al., 1993b). The
first example of a microlensing event was discovered in October 1993 (Figure 10.4), the mass
10.2 Dark matter in galaxies 251

Figure 10.4: The gravitational microlensing event recorded by the MACHO project in February and
March 1993. The horizontal axis shows the number of days measured from day zero on 2 January 1992.
The vertical axis shows the amplification of the brightness of the lensed star relative to the unlensed
intensity in blue and red wavebands. The solid lines show the expected variations of brightness of a
lensed star with time. The same characteristic light-curve is observed in both wavebands, as expected
for a gravitational microlensing event (Alcock et al., 1993a).

of the invisible lensing object being estimated to lie in the range 0.03M < M < 0.5M
(Alcock et al., 1993a).
By the end of the MACHO project, many lensing events had been observed, including
over 100 in the direction towards the Galactic bulge, about three times more than expected.
In addition, 13 definite and 4 possible events were observed in the direction of the Large
Magellanic Cloud (Alcock et al., 2000). The numbers are significantly greater than the
two to four detections expected from known types of star. The technique does not provide
distances and masses for individual objects, but, interpreted as a Galactic halo population,
the best statistical estimates suggest that the mean mass of these MACHOs is between
0.15M and 0.9M . The statistics are consistent with MACHOs making up about 20%
of the necessary halo mass, the 95% confidence limits being 8–50%. Somewhat fewer
microlensing events were detected in the EROS project, which found that less than 25% of
the mass of the standard dark matter halo could consist of dark objects with masses in the
252 10 The physics of galaxies and clusters of galaxies

range 2 × 10−7 M to 1M at the 95% confidence level (Afonso et al., 2003). The most
likely candidates for the MACHOs observed by the MACHO project would appear to be
white dwarfs, which would have to be produced in large numbers in the early evolution of
the Galaxy, but other more exotic possibilities cannot be excluded. The consensus is that
MACHOs alone cannot account for all the dark matter in our Galaxy, and so some form of
non-baryonic matter must make up the difference.
Useful astrophysical limits can be set to the number densities of different types of
neutrino-like particles in the outer regions of giant galaxies and in clusters of galaxies. The
WIMPs and massive neutrinos are collisionless fermions, and therefore there are constraints
on the phase-space density of these particles, which translate into a lower limit to their masses
since, for a given momentum, only a finite number of particles within a given volume is
allowed. This calculation was first presented by Scott Tremaine (b. 1950) and James Gunn
in 1979, who showed that the masses of neutrino-like particles necessary to bind the halo of
our Galaxy would exceed 30 eV, well in excess of the upper limit to the mass of the electron
neutrino at that time (Tremaine and Gunn, 1979).
The search for evidence for different types of dark matter particles has developed into
one of the major areas of the discipline known as astroparticle physics. An important class
of experiments involves the search for weakly interacting particles with masses m ≥ 1 GeV,
which could make up the dark halo of our Galaxy. In order to form a bound dark halo about
our Galaxy, the particles would have to have velocity dispersion v 2 1/2 ∼ 230 km s−1 and
their total mass is known. Therefore, the number of WIMPs passing through a terrestrial
laboratory each day is a straightforward calculation. When these massive particles interact
with the sensitive volume of the detector, the collision results in the transfer of momentum
to the nuclei of the atoms of the material of the detector, and this recoil can be measured
in three different ways. (i) There is a small temperature increase, which can be measured
in a cryogenically cooled detector, (ii) or the ionisation caused by the recoiling nucleus can
be measured in an ionisation chamber, (iii) or the light emitted by the passage of the recoil
nucleus through the material can be detected by a scintillation detector. The challenge is to
detect the very small number of events expected because of the very small cross-section for
the interaction of WIMPs with the nuclei of atoms. A typical estimate is that less than one
WIMP per day would be detectable by 1 kg of detector material. These are very demanding
experiments, and they have to be located deep underground to avoid contamination by
cosmic rays and to be heavily shielded against natural radioactivity in the surrounding
rocks. Such experiments have been carried out in deep underground laboratories such as
those at Gran Sasso in Italy, the Soudan Underground Laboratory in Minnesota, USA, and
the Boulby Underground Laboratory in Yorkshire, UK.
The importance of these measurements for particle physics is that one of the strong
candidates for the WIMPs is the lightest particle predicted by supersymmetry theories of
elementary particles and which are expected to be stable. Many experiments have been
developed to search for these elusive particles, but as yet no convincing positive detection
has been made. A good example of the quality of the data now available is provided by the
results of the Cryogenic Dark Matter Search at the Soudan Underground Laboratory, which
provided 19.4 kg days of observation with a germanium detector cooled to temperatures
less than 50 mK. At the most sensitive mass, 60 GeV/c2 , the scalar cross-section for the
WIMP–nucleon interaction must be less than σw = 4 × 10−47 m2 (Akerib et al., 2004). It
 10.4 The large-scale distribution of galaxies 253

is intriguing to compare this value with the weak-interaction cross-section for neutrino–
electron scattering, σ = 3 × 10−49 (E/m e c2 )2 m 2 , where E is the energy of the neutrino.
These limits to σw are already sufficiently low to rule out some supersymmetry models.

10.3 The dynamics of elliptical galaxies

It might be thought that the internal dynamics of elliptical galaxies would be a relatively
straightforward problem. The axial ratios of these galaxies range from 1:1 to about 3:1, and
the velocity dispersions of the stars can be measured spectroscopically within the galaxies.
These measurements have been compared with the amounts of rotation and internal velocity
dispersion which would be expected if the flattening of the elliptical galaxies were wholly
attributed to the rotation of an axisymmetric distribution of stars. Francesco Bertola (b. 1937)
and Massimo Capaccioli (b. 1944) in 1975 and Garth Illingworth (b. 1947) in 1977 showed,
however, that luminous elliptical galaxies rotate only very slowly (Bertola and Capaccioli,
1975; Illingworth, 1977). An analysis of a sample of elliptical galaxies, as well as the bulges
of spiral galaxies, by Roger Davies (b. 1954) and his colleagues in 1983 confirmed that, in
general, the most luminous elliptical galaxies do not possess enough rotation to account for
their ellipticity (Davies et al., 1983) (Figure 10.5). The implication of this result is that the
assumptions of an axisymmetric distribution of stars and an isotropic velocity distribution
at all points in the galaxies must be wrong. Consequently, these massive elliptical galaxies
may well be triaxial systems, that is systems with three unequal axes and with anisotropic
velocity distributions. There is no problem in assuming that the velocity distribution is
anisotropic because normally the timescale for the exchange of energy between stars through
gravitational encounters is greater than the age of the Galaxy. Therefore, if the velocity
distribution began by being anisotropic, it would not have been isotropised by now.
Further evidence for the triaxial nature of the stellar distribution in elliptical galaxies has
been discovered by Bertola and Giuseppe Galletta (b. 1954), who found that the major axis
of elliptical galaxies varies with distance from their centres (Bertola and Galletta, 1979).
Furthermore, in some of these galaxies, rotation has been observed along the minor as well
as along the major axis (Bertola et al., 1991).
The theoretical position was clarified by Martin Schwarzschild in 1979, who used the pro-
cedures of linear programming to determine the orbits of particles in general self-gravitating
systems (Schwarzschild, 1979). His analysis showed that there exist stable triaxial systems.
In a result similar to that found in classical dynamics, he showed that there exist stable orbits
about the largest and smallest axes of a triaxial system, but not about the intermediate axis.
Thus, elliptical galaxies can be classified as oblate–axisymmetric, prolate–axisymmetric,
oblate–triaxial, prolate–triaxial and so on.

10.4 The large-scale distribution of galaxies

Groups and clusters of galaxies come in a wide variety of different types, ranging from rich
regular clusters, which have smooth galaxy density profiles and roughly circular appear-
ance, to irregular systems, which have a ragged appearance without any prominent central
254 10 The physics of galaxies and clusters of galaxies

1.2

1.0

0.8
vm
σ
0.6

0.4

0.2

0.1 0.2 0.3 0.4 0.5

ε
Figure 10.5: The ellipticities, ε, of elliptical galaxies as a function of their rotational velocities,
vm , normalised to the velocity dispersion, σ , of the bulge. The open circles are luminous elliptical
galaxies; the filled circles are lower luminosity ellipticals; and the crosses are the bulges of spiral
galaxies. If the ellipticity were entirely due to rotation, with an isotropic stellar velocity distribu-
tion at each point, the galaxies would be expected to lie along the solid line. The diagram shows
that, at least for the massive ellipticals, this simple picture of rotational flattening cannot be correct
(Davies et al., 1983).

concentration of galaxies. While the Palomar 48-inch Schmidt Sky Survey was being car-
ried out, the principal observers undertook various research projects on the wide field plates
which were taken each night, and, among the most important of these, George Abell (1927–
1983) classified and catalogued the rich clusters of galaxies. The catalogue was compiled by
visual inspection of the plates according to strict selection criteria such that only the most
prominent clusters were included. The apparent magnitudes of the galaxies were estimated
by visual inspection. The Abell Catalogue of Clusters of Galaxies was published in 1958
and contains about 2400 of the richest clusters of galaxies north of declination −20◦ away
from the Galactic plane (Abell, 1958). The survey was extended to include the southern
hemisphere when the UK Schmidt Telescope survey plates were completed, and a cata-
logue of over 4000 rich clusters over the whole sky was prepared by Abell, Harold Corwin
(b. 1943) and Ronald Olowin (b. 1945) in 1989, sadly six years after Abell’s untimely death
in 1983 (Abell, Corwin and Olowin, 1989).
In Abell’s catalogue, there is a bias towards the richest, symmetrical systems, reflecting
the strict criteria which he adopted in selecting the clusters. In fact, the clustering of galaxies
10.4 The large-scale distribution of galaxies 255

occurs on a very wide range of physical scales, from small groups containing only a few
galaxies to giant clusters and superclusters of galaxies. Following Hubble’s pioneering
studies of the counts of galaxies in the 1930s, a major effort was made after the Second
World War to define the large-scale structure of the Universe of galaxies using the large-scale
plates of the Lick Northern Proper Motion Surveys. These large plates had been taken with
the 51-cm Carnegie double astrograph for the purpose of measuring the proper motions of
stars, but they also included a great deal of information about the number counts of galaxies.
Counts of galaxies on the Lick plates were made by Donald Shane (1895–1983), Carl
Wirtanen (1910–1990) and their colleagues during the 1950s and were published as counts
of galaxies in 1◦ × 1◦ boxes for the sky north of δ = −23◦ (Shane and Wirtanen, 1957).
Jerzy Neyman (1894–1981) and Elizabeth Scott (1917–1988) used correlation functions to
analyse the variance of the numbers of galaxies in these cells (Neyman, Scott and Shane,
1954). In turn, these studies led to the use of two-point correlation functions to describe the
large-scale clustering properties of galaxies, and these have become the preferred tool for
quantifying the large-scale distribution of galaxies statistically. The two-point correlation
function, ξ (r ), for galaxies can be written as
N (r ) dV = N0 [1 + ξ (r )] dV, (10.5)
where N (r ) is the number density of galaxies at radial distance r from any given galaxy
and ξ (r ) describes the excess probability of finding a galaxy at distance r over a uniform
distribution N0 .
One of the important issues was whether or not there are preferred scales of clustering of
galaxies in the Universe. Abell and de Vaucouleurs had shown that there exists non-random
clustering of clusters of galaxies, but it was not clear whether or not there is a continuous
range of clustering (Abell, 1962; de Vaucouleurs, 1971). Tao Kiang (b. 1928) and William
Saslaw (b. 1944) proposed that there were no preferred scales, but that clustering could
occur on all scales (Kiang and Saslaw, 1969). In 1969, Hiroo Totsuji (b. 1943) and Taro
Kihara (b. 1917–2001) were the first to show that the galaxy correlation function, ξ (r ),
can be approximated by a power-law over a wide range of scales, and this approach was
developed extensively by James Peebles and his colleagues in an important series of papers
in the 1970s.8 The function ξ (r ) can be described by a power-law function,
 −γ
r
ξ (r ) = , (10.6)
r0
where γ = 1.77 and r0 = 5h −1 Mpc, where h = H0 /100 km s−1 Mpc−1 is Hubble’s constant
measured in units of 100 km s−1 Mpc−1 . This function gives a good representation of the
clustering of galaxies on scales from about 10h −1 kpc to 10h −1 Mpc, but on scales r ≥
20h −1 Mpc the function decreases more rapidly with increasing physical size (Figure 10.6)9
(Maddox et al., 1990; Peebles, 1993).
It should be emphasised that this form of correlation function is spherically symmetric
about any point and washes out a great deal of information about the structure of the
clustering. Nonetheless, this form of function makes the important point that clustering
occurs on a very wide range of physical scales, from small groups of galaxies to systems
much greater than even the richest clusters of galaxies. The rich clusters are no more than
256 10 The physics of galaxies and clusters of galaxies

Figure 10.6: The angular two-point correlation function for galaxies over a wide range of angular
scales. (a) The scaling test for the homogeneity of the distribution of galaxies derived from the APM
surveys at increasing limiting apparent magnitudes in the range 17.5 < m < 20.5. The correlation
functions are displayed at intervals of 0.5 magnitudes. (b) The two-point correlation function scaled
to the correlation function derived from the Lick counts of galaxies (Maddox et al., 1990).

the most prominent features of a continuous spectrum of clustering. The origin of this
form of correlation function is one of the goals of the theory of origin of structure in the
Universe.
Figure 10.6 shows that the angular two-point correlation function scales with increasing
apparent magnitude, as expected, if the distribution of galaxies exhibits the same degree
of clustering with increasing distance from our own Galaxy. As Peebles expresses this
important result,

the correlation function analyses have yielded a new and positive test of the assumption that the galaxy
space distribution is a stationary (statistically homogeneous) random process.

In fact, the distribution of galaxies is much more complicated than this. In the 1970s,
Peebles and his colleagues reanalysed the Lick counts of galaxies using the original 10 ×
10 cells used by Shane and Wirtanen and demonstrated that clustering exists on a very
wide range of scales, and, in particular, that, on scales greater than those of clusters of
galaxies, the distribution of galaxies has a stringy, cellular appearance (Seldner et al., 1977)
(Figure 10.7).
In parallel with these studies, increasing numbers of redshifts for nearby galaxies were
becoming available, and these enabled the three-dimensional distribution of galaxies to be
defined directly. Jaan Einasto (b. 1929) and his colleagues were the first to demonstrate
the reality of structures on scales very much greater than those clusters of galaxies and to
appreciate the significance of these for theories of the origin of structure in the Universe10
(Jöeveer and Einasto, 1978). The mapping of the local distribution of galaxies proceeded
though the 1980s, culminating in the map of the local Universe created by Margaret Geller
(b. 1947) and John Huchra, derived from a complete survey of the redshifts of over 14 000
10.4 The large-scale distribution of galaxies 257

Figure 10.7: A map of the galaxy counts in the northern galactic hemisphere derived by Peebles,
Seldner and their colleagues from a reanalysis of the Lick counts of galaxies carried out by Shane
and Wirtanen. The northern galactic pole is at the centre; the galactic equator is the white bounding
circle. The galactic latitude is a linear function of radius from the pole. Galactic longitude increases
clockwise, with l = 0◦ at the bottom of the map. The prominent ‘cluster’ in the centre of the image
is the Coma cluster. (Seldner et al., 1977.)

bright galaxies (Geller and Huchra, 1989) (Figure 10.8). If the galaxies were uniformly
distributed in the local Universe, the points would be uniformly distributed over the diagram.
It can be seen that there are large ‘holes’ in which the local number density of galaxies is
significantly lower than the mean and also long ‘filaments’ or ‘walls’ of galaxies. The scale
of the large holes seen in Figure 10.8 is about 30 to 50 times the scale of a cluster of
galaxies.
Richard Gott (b. 1947) and his colleagues (Gott, Melott and Dickinson, 1986) showed
that the topology of the distribution of the galaxies on the large scale is ‘sponge-like’. The
material of the sponge represents the location of the galaxies and the holes represent the
large voids. Both the holes and the distribution of galaxies are continuously connected
throughout the local Universe. These are the largest known structures in the Universe, and
258 10 The physics of galaxies and clusters of galaxies

Figure 10.8: The distribution of galaxies in the nearby Universe as derived from the Harvard–
Smithsonian Center for Astrophysics survey of galaxies. The map contains over 14 000 galaxies,
which form a complete statistical sample around the sky between declinations δ = 8.5◦ and 44.5◦ .
Our Galaxy is located at the centre of the map, and the radius of the bounding circle corresponds to
a redshift of 0.05, or a distance of 150h −1 Mpc. The galaxies within this slice around the sky have
been projected onto a plane to show the large-scale features in the distribution of galaxies. Rich clus-
ters of galaxies, which are gravitationally bound systems with internal velocity dispersions of about
103 km s−1 , appear as ‘fingers’ pointing radially towards our Galaxy at the centre of the diagram.
(Courtesy of Margaret Geller and John Huchra.)

one of the great cosmological problems is to reconcile the gross irregularity in the large-
scale distribution of galaxies with the remarkable smoothness of the cosmic microwave
background radiation.
A key question for cosmology was whether or not this ‘cellular’ structure persists out to
much greater distances, and this was resoundingly answered in the affirmative by a number
of large-scale surveys of the distribution of galaxies in various sectors of the sky. Firstly,
the Las Campanas Redshift Survey used measurements of the redshifts of 26 418 galaxies
 10.5 The physics of clusters of galaxies 259

Figure 10.9: A 3◦ slice through the Anglo-Australian Telescope 2dF galaxy survey showing the
cellular structure of the distribution of galaxies extending out to redshift z ≈ 0.25. This map extends
to about five times the distance of the image shown in Figure 10.8. (Courtesy of Dr M. Colless and
the 2dF Galaxy Survey Team.)

to extend the depth of the survey to about four times that of the Geller–Huchra survey, and
it showed the ‘cellular’ structure extending out to the limit of their survey (Lin et al., 1996).
This was followed by the 2dF survey carried out using the 2◦ field multi-object spectro-
graph at the prime-focus of the Anglo-Australian Telescope, which carried out a similar
survey of almost 180 000 galaxies, extending the survey to five times that of the Huchra–
Geller survey with large statistics (Figure 10.9) (Colless et al., 2001). Most recently, the
first redshift maps produced by the Sloan Digital Sky Survey have defined the large-scale
distribution of galaxies to a similar depth but with even larger statistics, about 200 000
galaxies being plotted in their first images (Stoughton et al., 2002). Statistical analyses of
these maps have shown that the cellular structure is present throughout the distribution of
galaxies at the present epoch. Thus, although the distribution of galaxies is highly non-
uniform, the same degree of non-uniformity is present throughout the distribution of galax-
ies, consistent with the overall isotropy and homogeneity of the Universe on large enough
scales.

10.5 The physics of clusters of galaxies

The richest clusters, such as the Coma cluster of galaxies, have crossing times, tc = R/v,
which are much less than the age of the Universe. In this expression, v is the mean speed
of a galaxy in the cluster and R is a characteristic size of the cluster. The radial distribution
260 10 The physics of galaxies and clusters of galaxies

of galaxies is found to be similar to that of an isothermal gas sphere, for which N (r ) ∝ r −2

out to some outer radius. Thus, it is certain that such clusters have come to a state of
statistical equilibrium under gravity. It is for this reason that the virial theorem can be
applied with confidence to these clusters in order to determine their masses (Eddington,
1916a). The masses of the clusters can range up to 3 × 1015 M  , very much greater than
would be inferred from the light of the galaxies. A convenient way of expressing this
result is in terms of the mass-to-luminosity ratio of rich clusters of galaxies, which is of
the order 200–300M /L  , very much greater than the values found in the visible regions
of galaxies, which are at most about 10–20M /L  . It is also interesting to compare this
figure with the mass-to-luminosity ratio necessary to attain the critical cosmological density,
which amounts to 1600h M /L  . These observations provide compelling evidence for the
presence of dark matter in the clusters, but not in sufficient quantities for the Universe to
attain the critical cosmological density.

10.5.1 Hot gas in clusters of galaxies

One of the most important discoveries of the UHURU X-ray observatory was that some rich
clusters of galaxies are intense X-ray sources. In clusters such as the Coma cluster, the X-
ray emission is diffuse and fills the core of the cluster (Gursky et al., 1971). This radiation
has been convincingly identified with the bremmstrahlung, or free–free emission, of hot
intracluster gas, the clinching piece of evidence being the discovery of the emission lines of
very highly ionised iron, Fe XXV and Fe XXVI, at 8 keV by the Ariel-V satellite (Mitchell
et al., 1976). The hot gas forms an extended atmosphere within the gravitational potential
of the cluster, and the same technique already discussed in the context of measuring the
masses of galaxies can be used to estimate the mass and mass distribution within the cluster
of galaxies (see Sections 10.2 and A10.1). Observations by the ROSAT X-ray observatory
enabled maps of the X-ray surface brightness distribution within nearby clusters to be
measured and the distributions of the mass in galaxies, in hot gas and in the dark matter,
to be determined. Hans Böhringer (b. 1952) applied these procedures to the Perseus cluster
of galaxies, in which the X-ray emission could be traced out to a radius 1.5h −1 Mpc. From
the X-ray observations, it was possible to determine both the total gravitating mass within
radius r , M(<r ), the mass of gas within radius r , Mgas (<r ), and then to compare these
with the mass in the visible parts of the cluster galaxies (Böhringer, 1994). In Figure 10.10,
it can be seen that the mass of hot intracluster gas is about five times greater than the mass
in galaxies, but that it is insufficient to account for all the gravitating mass that must be
present. Some form of dark matter must be present to bind the cluster gravitationally. Note
also that the observation of iron emission from the intracluster gas indicates that the iron
created in the stars in galaxies must have been circulated through the intracluster medium.
If the density of the hot intracluster gas is large enough, its cooling rate can be sufficiently
large for it to cool over cosmological timescales. At high enough temperatures, the principal
energy loss mechanism for the gas is the same thermal bremsstrahlung process that is
responsible for the X-ray emission of the cluster. The characteristic cooling time for the gas
10.5 The physics of clusters of galaxies 261

Figure 10.10: Integrated radial profiles for the mass in the visible parts of galaxies, hot gas and total
gravitating mass for the Perseus cluster of galaxies, as determined by observations with ROSAT.
The upper band indicates the range of possible total masses and the central band shows the range of
gaseous masses (Böhringer, 1994).

is given by
T 1/2
tcool = 1010 years, (10.7)
N
where the temperature is measured in kelvin and the number density of ions or electrons is
measured in particles per cubic metre. Thus, if the typical temperature of the gas is 107 to
108 K, the cooling time is less than 1010 years if the electron density is greater than about
3 × 103 to 104 m−3 . These conditions are found in many of the clusters of galaxies which
are intense X-ray emitters. As a result, the central regions of these hot gas clouds can cool
and, to preserve pressure balance, the gas density increases, resulting in the formation of
a cooling flow. The evidence for these cooling flows was discovered by Andrew Fabian
(b. 1948) and his colleagues when it became possible to image clusters of galaxies and
determine the variation of the X-ray surface brightness and the temperature of the gas as a
function of radius from the centre (Fabian, 1994). A good example of these flows is found in
the cluster Abell 478, in which the temperature of the gas in the central regions is less than
in the outer regions and a mass inflow rate of about 600–800M year−1 is inferred. Thus,
over a period of 1010 years, such cooling flows can contribute significantly to the mass of
262 10 The physics of galaxies and clusters of galaxies

Figure 10.11: The radio maps of the radio trail source 3C 83.1B associated with the galaxy NGC
1265 in the Perseus cluster of galaxies. (a) The radio jets are swept back by the ram pressure of the
intergalactic gas in the cluster, the galaxy moving at a speed of at least 2200 km s−1 relative to the local
standard of rest of the cluster. (Courtesy of Dr Alan Bridle and the NRAO/AUI.) (b) A high-resolution
map of the core of the radio source showing the radio jets ejected from the core of the galaxy, which
is indicated by the black dot.

the central galaxy. According to Fabian, about half of the clusters detected by the Einstein
X-ray observatory have high central X-ray surface brightnesses and cooling times less than
1010 years. Abell 478 is a particularly massive flow. Typically, the inferred mass flow rates
are about 100 to 300M year−1 .
The presence of hot gas in the cluster has a number of important consequences for
the physics of the galaxies in clusters. The galaxies are in motion with respect to the
intergalactic gas and so the ram pressure of the intergalactic gas can sweep the inter-
stellar gas out of spiral galaxies. This process of ram pressure stripping of galaxies is
a plausible means of forming S0 galaxies, which are found in much greater numbers
in rich clusters of galaxies and regions of high galaxy density than in the general field
(Gunn, 1978; Dressler, 1980).
Another consequence of the presence of hot gas in the cluster is that the structures of
double radio sources, which are powered by beams of relativistic particles, are distorted
by the motion of radio galaxies as they pass through the intergalactic gas. In powerful
extragalactic double radio sources, the jets penetrate the intergalactic gas to large distances
from the nucleus of the active galaxy and create characteristic double structures (see Chapter
11 and Figure 11.1(b)). The jets in the double radio sources associated with nearby cluster
galaxies are much less luminous than the most powerful double radio sources, and the relative
motion of the radio galaxy through the intergalactic gas causes the beams of energetic
electrons to be swept back by the intergalactic gas, resulting in the formation of what are
known as radio trail sources (Miley et al., 1972) (Figure 10.11).11
 10.5 The physics of clusters of galaxies 263

A further effect of the X-ray-emitting gas in the cluster is that the hot electrons can
scatter photons of the cosmic microwave background radiation which pass through the
cluster. This process was first described by Rashid Sunyaev (b. 1943) and Yakov Zeldovich
in 1970; in this process the photons of the background radiation are scattered to higher
energies by Compton scattering (Sunyaev and Zeldovich, 1970a). The result is that the
Planck spectrum of the background radiation is shifted to slightly higher energies, resulting
in a decrease in the intensity of the background radiation in the direction of the cluster of
galaxies in the Rayleigh–Jeans region of the black-body spectrum and an increase in the
Wien region of the spectrum. The Sunyaev–Zeldovich effect amounts to only about one part
in 3000 of the intensity of the background for a rich cluster of galaxies, and it was only
observed with certainty by Mark Birkinshaw (b. 1954) and his colleagues in 1990 after many
years of difficult observation (Birkinshaw, 1990). Maps of these holes in the microwave
background have since been made in the direction of rich clusters, which were known to be
X-ray sources, and the shapes of the holes in the background have been determined directly
(Jones et al., 1998). The importance of the detection of the Sunyaev–Zeldovich effect is that
it enables the pressure of the gas in the cluster to be determined. Combining this observation
with the known temperature and emissivity of the gas enables the size of the gas cloud to
be determined independent of its distance. This is one of the most promising methods of
determining Hubble’s constant.
One intriguing aspect of the imaging of these holes in the cosmic microwave background
radiation is that the size of the decrement is independent of the redshifts of the clusters,
if their properties are independent of redshift. This phenomenon has been observed by
John Carlstrom (b. 1957) and his colleagues, who made maps of the Sunyaev–Zeldovich
effect in clusters of galaxies spanning the redshift range 0.14 to 0.89 (Carlstrom et al.,
2000) (Figure 10.12). Visual inspection of Figure 10.10 shows that the amplitude of the
temperature decrement observed in these clusters is independent of redshift. This is therefore
a powerful method for detecting clusters of galaxies at very large redshifts, and a number
of projects are currently underway to search for these clusters by detecting the holes in the
cosmic microwave background radiation.
Most recently, the excess emission expected at wavelengths shorter than the maximum
of the cosmic background radiation has been detected from a number of clusters (Benson
et al., 2003). These precise measurements enable an upper limit to be set to the peculiar
motions of clusters of galaxies with respect to the local standard of rest through observations
of the kinetic Sunyaev–Zeldovich effect.

10.5.2 The dynamical evolution of clusters of galaxies

The simplest picture of the dynamical evolution of a cluster of galaxies begins by assuming
that the galaxies can be considered to be point masses. When the collapse of a protocluster
gets underway, large gravitational potential gradients are set up since the collapse is unlikely
to be spherically symmetric and the system of galaxies relaxes under the influence of these
large-scale perturbations. This process was first described by Donald Lynden-Bell (b. 1935)
in 1967 and is known as violent relaxation (Lynden-Bell, 1967). He showed that, under
264 10 The physics of galaxies and clusters of galaxies

03'
47'
35' 04'
48'
36'
05'
37' 49'
06'
38' 50'

39' 07'
51'
40' −03°08'
−04°52'
−03° 41'
10 57 10 5 57 0 55 50 20 53 52 48 44 40 36 4 51 48 44 40 36 32

44'
18'
30' 45'
17'

28' 46'
16'
47'
26' 15'
48'
14'
24'
49'
13'
16°22' −01°50'
58° 12'

0 18 54 48 42 36 30 24 18 2 37 30 25 20 15 10 14 52 0 50 40 30 20

36'
02'
35' 56'
03'
34'
54'
04'
33'
52' 05'
32'
06'
31' 50'
07'
36°30'
02° 48'
−01° 08'
8 40 0 54 48 42 36 30 14 1 18 12 6 10 54 48 42 13 9 5 90 55 50 45

54'
54' 28'

52'
52' 26'

50'
50' 24'

48'
48' 22'

37°46'
65° 46' 23° 20'

8 31 45 30 15 31 0 45 30 15 14 26 30 20 10 26 0 50 40 11 55 36 30 24 18 12 6 55 0

Figure 10.12: Images of the Sunyaev–Zeldovich decrement in 12 distant clusters with redshifts in the
range 0.14 to 0.89 (Carlstrom et al., 2000). Each of the images is plotted on the same intensity scale.
The data were taken with the OVRO and BIMA millimetre arrays. The filled ellipse at the bottom left
of each image shows the full-width half-maximum of the effective resolution used in reconstructing
the images.

the influence of these large potential gradients, the galaxies in the cluster rapidly attain an
equilibrium configuration in which galaxies of all masses have the same velocity distribution,
consistent with observations of the velocities of galaxies of different masses in clusters. In
the process of violent relaxation, the system has to get rid of half of its kinetic energy so
that the cluster ends up being bound and satisfying the virial theorem.
Just as in the case of a Maxwellian gas of particles, the galaxies can exchange kinetic
energy, but now by gravitational encounters. Unlike the case of particles in a gas, the
10.5 The physics of clusters of galaxies 265

encounters are rather infrequent, but the statistical result is the same in that the galaxies
tend towards equipartition of energy so that the more massive galaxies slow up and tend
to drift towards the centre of the cluster. This process of deceleration of the galaxies is
known as dynamical friction and was first discussed by Chandrasekhar in the context of the
dynamical evolution of star clusters in 1943 (Chandrasekhar, 1943a–c). This result suggests
a reason why, in the regular relaxed clusters, the most massive galaxies are found towards
the centre.
The absolute magnitudes of the brightest galaxies in clusters display a remarkably narrow
dispersion, resulting in a very well defined velocity distance relation (see Figure 13.1).
There was some controversy over the issue of whether this could be explained by randomly
sampling the high-luminosity end of the luminosity function, or whether there is some
special property of the first-ranked cluster member which is independent of the richness of
the cluster. In 1977, Tremaine and Douglas Richstone (b. 1949) compared the dispersion in
absolute magnitudes of the first-ranked members with the mean value of the difference in
magnitude between the first- and second-ranked members (Tremaine and Richstone, 1977).
They showed that, for any statistical luminosity function, there is much less dispersion in
the absolute magnitudes of first-ranking cluster galaxies than would be expected if they
were simply randomly sampled from the luminosity function, although they recognised
that their results were not conclusive. John Kormendy (b. 1948) emphasised that the central
cD galaxies in clusters are quite distinct from normal giant elliptical galaxies (Kormendy,
1982).
There are other effects which are important because of the finite sizes of galaxies. Just as in
the cases of the peculiar and interacting galaxies, strong tidal effects can cause the disruption
of galaxies, and, in particular, large galaxies tend to tear apart and consume smaller galaxies.
This process, described by Ostriker and Tremaine in 1975, is often referred to as galactic
cannibalism and is likely to be particularly important in rich clusters of galaxies (Ostriker
and Tremaine, 1975). Computer simulations have shown how the coalescence of galaxies
can occur in close encounters between galaxies.
This process, as applied to clusters of galaxies, may explain the effect discovered by
Allan Sandage and Eduardo Hardy (b. 1941) in 1973. Laura Bautz (b. 1940) and William
Morgan classified rich clusters of galaxies according to the magnitude difference between the
brightest member and the next-brightest members (Bautz and Morgan, 1970). Sandage and
Hardy found what they termed the Bautz–Morgan effect in which the brighter the brightest
galaxy in a cluster, the fainter are the second- and third-brightest members, suggesting
that the brightest galaxy has grown at the expense of the next-brightest members of the
cluster (Sandage and Hardy, 1973). The process of galactic cannibalism seems to provide
an explanation of the origin of the huge giant elliptical galaxies observed at the centre
of many of the richest clusters of galaxies, and it has been confirmed by supercomputer
simulations of the formation of clusters of galaxies (see Chapter 15). It may also account
for the fact that the absolute magnitudes of the brightest galaxies in clusters seem to have
remarkably constant absolute luminosity. Marc Hausman and Jeremiah Ostriker showed in
1977 that the galaxies become physically larger and more bloated as they consume galaxies
and that therefore they become larger in physical size as well as becoming more luminous
(Hausman and Ostriker, 1977). When photometric observations are made of these galaxies
266 10 The physics of galaxies and clusters of galaxies

within an aperture of fixed physical size at the galaxy, the increase in luminosity due to
cannibalism is offset by the fact that the galaxy has grown in size, and these two effects
more or less compensate for each other.

Notes to Chapter 10

1 Much more detailed discussions of the properties of galaxies and the physics involved are contained
in the books by J. Binney and M. Merrifield, Galactic Astronomy (Princeton: Princeton University
Press, 1998), L. S. Sparke and J. S. Gallagher, Galaxies in the Universe (Cambridge: Cambridge
University Press, 2000) and J. Binney and S. Tremaine, Galactic Dynamics (Princeton: Princeton
University Press, 1987).
2 A review of the global parameters of galaxies as a function of stage along the Hubble sequence
was presented by M. S. Roberts and M. P. Haynes, Physical parameters along the Hubble sequence,
Annual Reviews of Astronomy and Astrophysics, 32, 1994, 115–152.
3 The proceedings of the first symposium dedicated to the results of the IRAS mission were reported
in: Light on Dark Matter, ed F.P. Israel (Dordrecht: D. Reidel Publishing Company, 1985).
4 See also V. C. Rubin, Field and cluster galaxies: do they differ dynamically?, in Large-scale
Motions in the Universe, eds V. C. Rubin and G. V. Coyne (Vatican City: Pontificia Academia
Scientiarum, 1988), pp. 541–558.
5 Virginia Trimble provides a comprehensive historical survey of the evidence for dark matter in
galaxies up to 1987 in her review Existence and nature of dark matter in the Universe, Annual
Reviews of Astronomy and Astrophysics, 25, 1987, 425–472.
6 I have given details of the various approaches to determining the distribution and amount of dark
matter in galaxies in Section 4.3 of Malcolm Longair, Galaxy Formation (Berlin: Springer-Verlag,
1998).
7 Fich and Tremaine also provide an interesting historical introduction to estimates of the mass of
our Galaxy (Fich and Tremaine, 1991).
8 References to the numerous papers of Peebles and his colleagues on correlation functions for
galaxies can be found in Peebles’ excellent monograph: P. J. E. Peebles, Principles of Physical
Cosmology (Princeton: Princeton University Press, 1993).
9 The diagrams in Figure 10.6 are shown in terms of the angular two-point correlation function on
the sky, defined by
N (θ) d = n g [1 + w(θ)] d ,

where w(θ) describes the excess probability of finding a galaxy at an angular distance, θ, in the
solid angle d from any given galaxy; n g is a suitable average surface density of galaxies. With a
number of reasonable assumptions, w(θ ) can be related to ξ (θ ). For example, if w(θ ) is described
by a power law, w(θ) ∝ θ −x , the spatial two-point correlation function is given by ξ (r ) ∝ r −(x+1) .
The homogeneity test shown in the diagram follows from the fact that if the same degree of
clumpiness is present throughout the region of the Universe surveyed, the angular two-point
correlation function scales as
 
D0 D
w(θ, D) = wL θ ,
D D0

where the function w L (θ) has been determined to the distance D0 . This is the scaling tested in
Figure 10.6(b).
10 Einasto and his colleagues carried out their studies at the Tartu Observatory in Estonia at a time
when relations with the West were strained and it was difficult for scientists from the Soviet Union
to communicate their researches and ideas. Einasto (2001) gives some impression of the problems
facing astronomers from the former Soviet Union during these years.
 Explanatory supplement to Chapter 10 267

11 Figure 10.11(b) was created from the observations made by C. P. O’Dea and F. N. Owen,
Astrophysical Journal, 301, 1986, 845 (see M. S. Longair in The New Physics ed. P. C. W.
Davies (Cambridge: Cambridge University Press 1988), p. 169).

A10 Explanatory supplement to Chapter 10

A10.1 The mass distribution in galaxies and clusters from X-ray observations
The X-ray emission of the hot gas in galaxies and clusters of galaxies provides a very
powerful probe of their gravitational potentials (Fabricant et al., 1980). It is assumed that the
cluster is spherically symmetric so that the total gravitating mass within radius r is M(<r ).
The gas is assumed to be in hydrostatic equilibrium within the gravitational potential defined
by the mass distribution in the cluster, that is by the sum of the visible and dark matter, as
well as the gaseous mass. If p is the pressure of the gas and is its density, both of which
vary with position within the cluster, the requirement of hydrostatic equilibrium is given by
dp G M(<r )
=− . (A10.1)
dr r2
The pressure is related to the local gas density, , and temperature, T , by the perfect gas
law:
kT
p= , (A10.2)
μm H
where m H is the mass of the hydrogen atom and μ is the mean molecular weight of the
gas. For a fully ionised gas with the standard cosmic abundance of the elements, a suitable
value is μ = 0.6. Differentiating equation (A10.2) with respect to r and substituting into
equation (A10.1), we find
 
kT 1 d 1 dT G M(<r )
+ =− . (A10.3)
μm H dr T dr r2
Reorganising equation (A10.3), we obtain
 
kT r 2 d(log ) d(log T )
M(<r ) = − + . (A10.4)
Gμm H dr dr
Thus, the mass distribution within the cluster can be determined if the variation of the gas
density and temperature with radius are known. Assuming the cluster is spherically sym-
metric, these can be derived from high-sensitivity X-ray intensity and spectral observations.
A suitable form for the bremsstrahlung spectral emissivity of a plasma is given by
 
1 Z 2 e6  m e 1/2 hν
κν = g(ν, T )N Ne exp − , (A10.5)
3π 2 ε03 c3 m 2e kT kT
where Ne and N are the number densities of electrons and nuclei, respectively, Z is the
charge of the nuclei and g(ν, T ) is the Gaunt factor, which can be approximated by
√  
3 kT
g(ν, T ) = ln . (A10.6)
π hν
268 10 The physics of galaxies and clusters of galaxies

The spectrum of thermal bremsstrahlung is roughly flat up to X-ray energies ε = hν ∼ kT ,

above which it cuts off exponentially.1 Thus, by making precise spectral measurements,
it is possible to determine the temperature of the gas from the location of the spectral
cut-off and the particle density along the line of sight from the emissivity of the gas. In
practice, the spectral emissivity has to be integrated along the line of sight through the
cluster. Performing this integration and converting it into an intensity, the observed surface
brightness at projected radius a from the cluster centre is given by
 ∞
1 κν (r )r
Iν (a) = dr. (A10.7)
2π a (r − a 2 )1/2
2

Alfonso Cavaliere2 (b. 1933) noted that this is an Abel integral which can be inverted to
find the emissivity of the gas as a function of radius as follows:
 ∞
4 d Iν (a)a
κν (r ) = da. (A10.8)
r dr r (a − r 2 )1/2
2

Notes to Section A10

1 I have given a derivation of this expression for bremsstrahlung in Malcolm Longair, High Energy
Astrophysics, vol. 1 (Cambridge: Cambridge University Press, 1992).
2 Cavaliere’s paper, Models of X-ray emission from clusters of galaxies, was published in X-
ray Astronomy, eds R. Giacconi and G. Setti (Dordrecht: D. Reidel Publishing Company, 1980),
pp. 217–237.
11 High-energy astrophysics

11.1 Radio astronomy and high-energy astrophysics

The early history of radio astronomy was recounted in Section 7.3; that story ended in
the mid 1950s, by which time the Galactic and extragalactic nature of the discrete radio
sources was established. From the point of view of astrophysics, the key realisation was
that, in most cases, the radio emission was the synchrotron radiation of ultra-high-energy
electrons gyrating in magnetic fields within the source regions. The synchrotron radiation
process began to be applied to other astronomical objects in which there was evidence for
high-energy astrophysical activity.
In 1942, Rudolph Minkowski showed that the emission of the supernova remnant known
as the Crab Nebula consists of two components, the filaments, which form a network
defining the outer boundary of the remnant, and diffuse continuum emission originating
within the nebula, which contributes most of its optical luminosity (Minkowski, 1942).
The continuum emission had a featureless spectrum and could not be accounted for by
any form of thermal spectrum. In 1949, John Bolton and Gordon Stanley found that the
flux density of the Crab Nebula at radio wavelengths was about 1000 times greater than in
the optical waveband (Bolton and Stanley, 1949). To account for the continuum emission,
Iosif Shklovsky (1916–1985) proposed in 1952 that both the radio and optical continuum
was synchrotron radiation, the energies of the electrons radiating in the optical waveband
being very much greater than those radiating in the radio waveband (Shklovsky, 1953).
One consequence of this hypothesis was that the optical continuum of the nebula should be
linearly polarised, and this was discovered by Viktor Dombrovskii (1914–1972) and Mikhail
Vashakidze (1909–1956) in 1954 (Dombrovski, 1954; Vashakidze, 1954). This observation
was confirmed by Jan Oort and Théodore Walraven (b.1916) in 1956, using superb plates
taken with the 200-inch Palomar telescope by Walter Baade in 1955 (Oort and Walraven,
1956). The famous optical jet in the nearby giant elliptical galaxy M87 had been discovered
by Heber Curtis in 1918 (Curtis, 1918a), and this strange feature was shown to be linearly
polarised by Baade in 1956 (Baade, 1956). The emission of the jet was interpreted as another
example of optical synchrotron radiation.
Observations of the diffuse Galactic radio emission were important because they provided
a direct quantitative test of the synchrotron hypothesis. The early observations of Jansky
and Reber indicated that the radiation was not the thermal emission of hot gas, but the
precise determination of its radio spectrum was technically difficult since the emission
is diffuse and the observed intensity is sensitive to the exact beam pattern of the radio

269
270 11 High-energy astrophysics

telescope. The spectrum of the Galactic radio emission was eventually determined using
geometrically scaled aerials and was shown to be of power-law form with spectral index
α ≈ 0.4 at low radio frequencies, α being defined by I (ν) ∝ ν −α (Turtle et al., 1962;
Turtle, 1963). According to synchrotron radiation theory, this spectral index is related to
the differential energy spectral index of the radiating electrons x by x = 2α + 1, where
the electron spectrum is described by N (E) dE ∝ E −x dE. Thus, the spectral index of the
radiating electrons would have to be about 1.8, not too different from that of cosmic-ray
protons and nuclei. In fact, the radio spectra of supernovae and extragalactic radio sources
were found to be somewhat steeper than this, α ≈ 0.75, corresponding to x = 2.5, precisely
the spectral index of the cosmic rays.
A key test of the synchrotron hypothesis was the search for polarised radio emission
from the diffuse interstellar medium, and this was observed by Vladimir Razin (b. 1930) in
1958 and by Gart Westerhout and his colleagues at Dwingloo in the Netherlands in 1962
(Razin, 1958; Westerhout et al., 1962). It was to be many years before the spectrum of the
cosmic-ray electrons would be measured reliably, but the radio astronomical observations
were convincing evidence that very-high-energy electrons and magnetic fields are present
throughout the plane of our Galaxy (see Section 9.6). It was inferred that the cosmic rays
observed at the top of the atmosphere are a sample of a general Galactic distribution of
high-energy particles. Estimates of the strength of the interstellar magnetic field could be
obtained from observations of the optical polarisation of starlight and from the Faraday
rotation of the polarised emission of discrete radio sources. The inferred energy density of
the electrons turned out to be roughly one-hundredth of that in the cosmic-ray protons.
The associations of the radio sources Cassiopeia A and Cygnus A with optical objects by
Baade and Minkowski in 1954 were crucial (Baade and Minkowski, 1954). Cassiopeia A was
identified with a young supernova remnant which must have exploded about 250 years ago, as
deduced from the measured expansion velocities of its optical filaments, though no historical
record of this event has been found. Assuming the radio emission of the remnant is syn-
chrotron radiation, the great radio luminosity of Cassiopeia A provided direct evidence for
the acceleration of huge fluxes of very-high-energy electrons in supernova remnants, an idea
foreshadowed by Baade and Zwicky’s remarkable paper of 1934 (Baade and Zwicky, 1934a).
Even more remarkably, the radio source Cygnus A was identified with a galaxy at a
redshift of 0.057, implying that its radio luminosity was enormous, more than a million
times greater than that of our Galaxy. It must therefore be the source of vast quantities
of relativistic material. Just as unexpected was the fact that the radio emission did not
originate from the galaxy itself. In 1953, Roger Jennison (b. 1922) and Mrinal Kumar Das
Gupta (b. 1923) at Jodrell Bank used interferometric techniques to show that the radio
emission originated from two huge lobes (Jennison and Das Gupta, 1953), and it turned
out that these were located on either side of the radio galaxy, once the identification was
made in the following year (Figure 11.1). Thus, not only must the radio galaxy accelerate an
enormous amount of material to relativistic energies, but this material also has to be ejected
into intergalactic space in opposite directions.
The intensity of synchrotron radiation depends upon both the magnetic field strength
and the flux of high-energy electrons in the source region.1 In 1959, Geoffrey Burbidge
worked out the minimum amount of energy in high-energy particles and magnetic fields
 11.2 Quasars and their close relatives 271

(b)

Figure 11.1: (a) A reconstruction of the radio structure of the radio source Cygnus A from radio
interferometric observations by Jennison and Das Gupta at a frequency of 125 MHz (Jennison and
Das Gupta, 1953). (b) A map of Cygnus A made by the Very Large Array in the USA at a frequency
of 5 GHz (Perley, Dreher and Cowan, 1984).

which had to be present in the source regions to account for the radio emission (Burbidge,
1959). The energies proved to be enormous, in some sources corresponding to a rest-mass
energy of about 106 M . In many ways, this result marked the beginning of high-energy
astrophysics in its modern guise. Some astrophysical means had to be found for converting
a significant fraction of the rest-mass energy of a galaxy into high-energy particles and
magnetic fields. The problem was not, however, confined to galaxies – supernovae remnants
such as Cassiopeia A were also able to convert a significant fraction of their mass into high-
energy particles and magnetic fields.

11.2 The discovery of quasars and their close relatives

The radio astronomical discoveries of the 1950s stimulated a great deal of astrophysical
interest and led to major investments being made in the construction of radio telescope
272 11 High-energy astrophysics

Figure 11.2: Hubble Space Telescope image of the quasar 3C 273, showing the optical jet ejected
from the quasar nucleus. The faint smudges to the south of the quasar are now known to be galaxies
at the same distance as the quasar. (Courtesy of Dr John Bahcall, NASA and the Space Telescope
Science Institute.)

systems. The radio observatories began systematic surveys of the sky in order to understand
in more detail both the astrophysics of these intense sources of radio waves and their use as
cosmological probes. The discrete sources of intense radio emission were often associated
with faint galaxies, and so accurate radio positions were needed to find the associated galaxy
among the large numbers of unrelated stars and galaxies. Among the extragalactic radio
sources which could be securely identified, most were found to be associated with some
of the most massive galaxies known, which are very luminous and so could be observed
to large redshifts. By 1960, the largest redshift known for any galaxy was that of the radio
galaxy2 3C 295 at z = 0.46 (Minkowski, 1960b).
By 1962, Thomas Matthews (b. 1927) and Sandage had identified three of the brightest
radio sources, 3C 48, 3C 196 and 3C 286, with ‘stars’ of an unknown type with strange
optical spectra (Matthews and Sandage, 1963). The breakthrough came in 1962 when Cyril
Hazard (b. 1928) measured very precisely the position of the radio source 3C 273 by the
method of lunar occulations using the recently completed Parkes 210-foot (64-metre) radio
telescope in New South Wales, Australia (Hazard, Mackey and Shimmins, 1963). These
remarakable observations enabled 3C 273 to be identified with what appeared to be a 13th
magnitude star (Figure 11.2). The new identification was rapidly relayed to Maarten Schmidt
(b. 1929), who used the Palomar 200-inch telescope to obtain its optical spectrum. The
spectrum contained prominent emission lines, but at unexpected wavelengths. The clue lay
11.2 Quasars and their close relatives 273

in the familiar pattern of lines, which Schmidt realised was the Balmer series of hydrogen,
but shifted to longer wavelengths with redshift z = λ/λ0 = 0.158 – this was certainly no
ordinary star (Schmidt, 1963); 3C 273 was the first, and brightest, of this class of hyperactive
galactic nuclei to be discovered. Its optical luminosity is about 1000 times greater than the
luminosity of a galaxy such as our own. To make matters even more intriguing, searches
through the Harvard plate archives revealed that the enormous luminosity of 3C 273 varied
on a timescale of years (Smith and Hoffleit, 1963). Nothing like this had been observed
in astronomy before. This discovery opened the floodgates for the identification of many
more examples of radio quasars. The radio source 3C 48 was found to have a redshift of
0.3675 (Greenstein and Matthews, 1963), while 3C 47 and 3C 147 have redshifts z = 0.425
and z = 0.545, respectively (Schmidt and Matthews, 1964). These sources were termed
quasi-stellar radio sources, and within a year this term had been contracted to the word
quasar.
One of the first discussions of the implications of these remarkable discoveries was held
at the First Texas Symposium on Relativistic Astrophysics held in Dallas in 1963.3 For the
first time, the optical and radio astronomers got together with the theoretical astrophysicists
and, in particular, the general relativists to thrash out what was known about these objects
and the role which general relativity might play in these studies. Coincidentally, William
Fowler and Fred Hoyle had been investigating the properties of supermassive stars with
masses as great as 106M , but, although these might radiate enormous luminosities, they
proved to be notoriously unstable objects (Hoyle and Fowler, 1963a,b). Perhaps the most
important conclusion of the meeting was the realisation that the quasars must involve strong
gravitational fields and so general relativity must play a central role in understanding their
properties. It was to be a number of years before general relativity became one of the
standard tools of the high-energy astrophysicist because many technical questions still had
to be addressed, but the writing was already on the wall. At the closing dinner of the 1963
Texas Symposium, Thomas Gold made the following remark:4

Everyone is pleased: the relativists who feel they are being appreciated, who are suddenly experts
in a field which they hardly knew existed; the astrophysicists for having enlarged their domain, their
empire by the annexation of another subject – general relativity.

The remarkable upsurge of interest in general relativity was undoubtedly stimulated by the
discovery of radio galaxies and quasars.
For some time, there was concern about the issue of whether the redshifts of the quasars
really were of cosmological origin or whether they might be due to some other physical
process. The short timescale of variability of their enormous luminosities seemed so extreme
that some astronomers, the most prominent of whom were Fred Hoyle, Geoffrey Burbidge
and Halton Arp, suggested that the quasars were actually relatively nearby objects and not
at the large distances implied by their redshifts. If this were the case, their luminosities
would be far less extreme and their variability would be much less of a problem because
they would be local objects, and perhaps some form of new variable star. Although widely
discussed, this hypothesis never attracted wide support, largely because no other satisfactory
explanation for the origin of the redshifts of the quasars was forthcoming.5 By 1965, the
redshifts of the radio quasars had reached z = 2.012 for the quasar 3C 9 (Schmidt, 1965).
274 11 High-energy astrophysics

Gravitational redshifts could not account for the observations, as discussed by Greenstein
and Schmidt in 1964 (Greenstein and Schmidt, 1964), and Doppler redshifts were ruled
out because of the absence of quasars with large blueshifts. The extreme luminosities and
short timescales of variability of quasars pointed inevitably to the presence of very compact
luminous sources in the nuclei of galaxies.
In many ways, the quasars were discovered too early. It eventually became apparent
that they are among the most extreme examples of what are now termed generically active
galactic nuclei. Quite by chance, the radio astronomers had stumbled upon a very effective
means of discovering the most luminous of them. In fact, the first examples of active
galactic nuclei had been discovered in the early 1940s by Carl Seyfert (1911–1960), who
studied a number of spiral galaxies, such as NGC 1068 and NGC 4151, which possessed
star-like nuclei (Seyfert, 1943). On taking the spectra of these nuclei, he found that they
possessed very intense emission lines of the Balmer series and forbidden lines such as
[OII], [OIII], [NII], [NeIII], [SII] and [SIII]. Furthermore, the lines were very broad, the
Doppler velocities required to broaden the lines being up to 8500 km s−1 . The presence and
breadth of these lines were quite unlike those observed in regions of ionised hydrogen in
galaxies. Furthermore, the spectrum of the continuum radiation was smooth, quite unlike
the spectrum of starlight. Seyfert’s pioneering work was largely neglected until the 1960s.
Among the galaxies associated with the radio sources was a class of galaxy which
William Morgan referred to as N-galaxies, by which he meant elliptical galaxies with star-
like nuclei (Matthews, Morgan and Schmidt, 1964). In the N-galaxies, the host galaxy is
often barely distinguishable because of the intensity of the star-like nucleus, the optical
spectrum of which displays intense broad emission lines, similar to those observed in the
Seyfert galaxies.
In 1963, just before the large redshifts of the quasars were discovered, Geoffrey Burbidge,
Margaret Burbidge and Allan Sandage surveyed a wide range of evidence concerning
activity in the nuclei of galaxies in an influential paper entitled ‘Evidence for the occurrence
of violent events in the nuclei of galaxies’ (Burbidge, Burbidge and Sandage, 1963). Their
survey included a very wide range of evidence, including the Seyfert galaxies, the radio
galaxies and other galaxies in which explosions seemed to have occurred, such as the
irregular galaxy M82 and the giant elliptical galaxy M87, which possesses the prominent
optical jet described by Heber Curtis (Curtis, 1918a).
In 1965, radio-quiet counterparts of the radio quasars were discovered by Allan Sandage,
and these radio-quiet quasars turned out to be about 100 times more common than the radio-
loud variety (Sandage, 1965). The similarity between the properties of quasars and the
nuclei of Seyfert galaxies was reinforced by the discovery in 1967 that both the continuum
emission and the strong emission lines in Seyfert galaxies are variable (Fitch, Pacholczyk
and Weymann,1967). It gradually became apparent that there is a continuous sequence of
high-energy activity in the nuclei of galaxies, from weak nuclei, such as the centre of our
own Galaxy, to the most extreme examples of quasars in which the starlight of the galaxy
is completely overwhelmed by the intense non-thermal radiation from the nucleus.
Among the most extreme examples of active galactic nuclei were the BL Lacertae or
BL Lac objects, which were discovered in 1968 as extremely compact and highly variable
radio sources by John McLeod (b. 1937) and Brian Andrew (b. 1939) (McLeod and Andrew,
 11.3 General relativity and active galactic nuclei 275

1968). Their optical spectra are smooth and featureless and display very rapid variability
at optical and radio wavelengths from timescales of less than a day to weeks. Indeed,
BL Lac itself was so named because it appeared in lists of variable stars. Eventually, when
spectrographs with very large dynamic ranges became available, redshifts for the BL Lac
objects were measured by detecting faint features in the spectra of the very faint underlying
galaxies. These redshifts showed that the BL Lac objects are indeed extragalactic objects,
but that their luminosities are not as great as those of the most luminous quasars.
Once the characteristic features of active galactic nuclei were established, many surveys
were undertaken to find more examples of them. The surveys for galaxies with ultraviolet
excess carried out by Beniamin Markarian (1913–1985) and his colleagues at the Byu-
rakan Observatory in Armenia proved to be a rich source of Seyfert galaxies, Markarian’s
pioneering surveys beginning in 1967 and culminating in 1981 with paper 15 of the series
(Markarian, 1967; Markarian, Lipovetsky and Stepanian, 1981). About 10% of what became
known as Markarian galaxies are Seyfert galaxies. Special surveys were also undertaken
to discover quasars, both the radio-loud and the radio-quiet varieties.

11.3 General relativity and models of active galactic nuclei

In parallel with these great discoveries of observational astronomy, enormous progress

was made in understanding the role of general relativity, not only in cosmology, but
also in the physics of matter in strong gravitational fields.6 The exact solution of the
Einstein’s field equations for a point mass in general relativity was discovered by Karl
Schwarzschild7 in 1916, the year after the final version of Einstein’s general theory was pub-
lished (Schwarzschild, 1916). As Einstein commented on receiving Schwarzschild’s paper,

I had not expected that one could formulate the exact solution of the problem in such a simple way.

Schwarzschild had derived the solution in order to provide exact results for tests of
general relativity, all of which involve taking the weak field limit at large distances from a
point object of mass M. The Schwarzschild metric can be written as follows:
⎡ ⎤
 
2G M 1 ⎢ dr 2 ⎥
ds 2 = 1 − 2
dt 2 − 2 ⎢
⎣   + r 2 (dθ 2 + sin2 θ dφ 2 )⎥
⎦, (11.1)
rc c 2G M
1−
r c2
where θ and φ are polar coordinates and r is a coordinate distance. Schwarzschild did not
remark upon the fact that this metric contains two singularities, one at r = 0 and the other
at coordinate distance r = 2G M/c2 . In fact, the second singularity is not a real physical
singularity, but is associated with the particular choice of coordinate system in which the
Schwarzschild metric was written, as was first demonstrated by Martin Kruskal (b. 1925)
in the mid 1950s, but only published in 1960 (Kruskal, 1960). The singularity at r = 0 is,
however, a real physical singularity in space-time.
If the discovery of quasars in 1963 was symbolic of a turning point in modern astro-
physics, the same could be said of general relativity, with the discovery by Roy Kerr (b. 1934)
276 11 High-energy astrophysics

in the same year of what is now known as the Kerr metric, one of the most important exact
solutions of Einstein’s field equations (Kerr, 1963). It turned out that the Kerr solution
describes the metric of space-time about a rotating black hole and is a generalisation of the
Schwarzschild metric.8 In 1965, a further generalisation of the Kerr metric was discovered
by Ezra (Ted) Newman (b. 1929) and his colleagues for the case of a system with finite elec-
tric charge by solving the combined Einstein and Maxwell field equations (Newman et al.,
1965). Only later was it realised that the metric describes a rotating black hole with finite
electric charge. Whereas the Schwarzschild solution is determined entirely by the mass of
the black hole, the Kerr metric depends upon both its mass and its angular momentum. It
took some years and a great deal of analysis before it was realised just how powerful these
solutions are. In 1971, Brandon Carter (b. 1942) showed that the only possible solutions for
uncharged axisymmetric black holes were the Kerr solutions (Carter, 1971), and in 1972
Stephen Hawking (b. 1942) showed that all stationary black holes must be either static or
axisymmetric, so that the Kerr solutions indeed included all possible forms of black hole
(Hawking, 1972).
These theorems led to important conclusions about the fate of collapsing bodies in
general relativity. No matter how complex the object and its properties are before collapse
to a black hole, all other properties, except its mass, angular momentum and electric charge,
are radiated away during the collapse. Another way of expressing these results is that all
multipole moments are radiated away in the process of formation of the black hole, leaving
it only with mass, angular momentum and electric charge. This result is often referred to as
the no-hair theorem for black holes.9 Thus, as the end points of stellar evolution of massive
remnants, black holes are quite remarkably simple objects. Notice that these theorems apply
to isolated black holes. Although the black hole cannot possess a magnetic dipole moment,
magnetic fields can penetrate into the black hole provided they are firmly attached to the
external medium. As shown by Kip Thorne and his colleagues, electrodynamics in the
vicinity of black holes can be precisely described by considering the surface of infinite
redshift, or event horizon, to consist of a membrane with the resistivity of free space,
Z = (μ0 /0 )1/2 (Thorne, Price and Macdonald, 1986).
One of the key questions addressed by the relativists was whether or not there must
be a physical singularity as a result of gravitational collapse. It had been argued by some
relativists that the singularity in the Schwarzschild solution was a special case and that, in
general, the presence of a singularity might depend upon the initial conditions from which
the collapse proceeded. The problem was solved by Roger Penrose (b. 1931) in 1965, who
showed quite generally that, once a surface from which light cannot escape outwards has
formed, what is known as a closed trapped surface, there is inevitably a singularity inside
that surface (Penrose, 1965). The nature of the singularity is described by the Kerr metric.
These same techniques were applied to the Universe as a whole by Penrose and Hawking,
and they were able to show in 1969 that, according to classical general relativity, it is also
inevitable that there exists a singularity at the origin of the Hot Big Bang models of the
Universe, subject to some rather general physical conditions (Hawking and Penrose, 1969).
They later extended their results to a much larger class of theories of gravity, the results
of these endeavours being summarised in the book The Large Scale Structure of Space-
Time by Stephen Hawking and George Ellis (b. 1939) (Hawking and Ellis, 1973). There are
11.3 General relativity and active galactic nuclei 277

numerous versions of the singularity theorems, an example of the Hawking and Penrose
theorem stating that:

A space-time which

(i) contains no time-like curves;

(ii) satisfies Einstein’s equations (without the cosmological term) and the energy condition ( +
pi ; + pi ≥ 0);
(iii) is sufficiently general; and
(iv) contains a closed spacelike hypersurface,

cannot be geodesically complete in all timelike and null directions.

In less technical language, applying the singularity theorems to our Universe, particu-
larly to the cosmic microwave background radiation, and subject to some rather general
geometrical conditions, there is inevitably a singularity at the origin of the world model,
provided gravity is attractive and the equation of state of the matter satisfies the energy
condition + pi ≥ 0.
From the point of view of astrophysics, the most important results of the study of
black holes, as they were named by John Wheeler in 1968 (Wheeler, 1968), concerned
the behaviour of matter in the vicinity of the event horizon and the maximum amount of
gravitational binding energy which could be released when matter falls into the black hole
from infinity. The Schwarzschild radius, rg = 2GM/c2 , of a spherically symmetric black
hole is the surface of infinite redshift, meaning that radiation emitted from this surface is
observed with infinite wavelength at infinity. This is the general relativistic version of the
insight of John Michell that, for a massive enough star, the escape velocity from the object
exceeds the speed of light (Michell, 1767). The interpretation in general relativity is, how-
ever, quite different because the geometry changes on crossing the Schwarzschild radius,
as can be appreciated by inspection of the metric (11.1). There is also a last stable circular
orbit about the black hole at radius 3r g – within this radius there are no stable circular orbits
and the matter spirals inevitably through the surface of infinite redshift, adding to the mass
of the black hole. The maximum gravitational binding energy which can be released by an
element of mass falling into the black hole from infinity is 5.72% of its rest-mass energy.10
Corresponding calculations were carried out for Kerr black holes. There is a maximum
amount of angular momentum which the hole can possess11 or else it cannot form, Jmax =
GM 2 /c. For a maximally rotating black hole, the surface of infinite redshift shrinks to
rg = GM/c2 and, in the case of corotating orbits, up to 42% of the rest-mass energy of the
infalling matter can be released. The rotational energy of the black hole can also be tapped,
as demonstrated by Roger Penrose in 1969 (Penrose, 1969). The part of the rest-mass energy
of the black hole associated with its rotation is, in principle, accessible to external observers
and can amount to a maximum of 29% of the rest-mass energy of the black hole. These are
very important results astrophysically and showed that the accretion of matter onto black
holes is potentially an extremely powerful source of energy – the synthesis of 4 He from four
protons, for example, can only release 0.7% of the rest-mass energy of the matter.12
Immediately following the discovery of quasars, a plethora of models appeared in the
literature, all of them attempting to account for their huge variations in luminosity on short
278 11 High-energy astrophysics

timescales. In 1964, Yakov Zeldovich in Moscow and Edwin Salpeter at Cornell indepen-
dently pointed out that accretion of matter onto black holes is potentially a very powerful
energy source (Salpeter, 1964; Zeldovich, 1964a) (see Section A11.1.2). An effective way
of releasing the binding energy of the infalling material is through the formation of an accre-
tion disc about the black hole. These discs form naturally because matter is most unlikely
to fall directly into the black hole from infinity since it must acquire some small amount of
angular momentum by random gravitational perturbations, which is then greatly amplified
by conservation of angular momentum as the material collapses towards the black hole.
Therefore, the infalling matter collapses along its rotation axis and forms an accretion disc
in centrifugal equilibrium about the black hole. If the disc were frictionless, there would
be no energy release. The presence of viscous forces in the disc performs two important
functions. The first is that the viscous forces enable angular momentum to be transferred
outwards and so the matter of the accretion disc can gradually drift inwards. The second
inevitable consequence of this process is that the frictional forces heat up the material disc,
and this is the means by which the matter releases its gravitational binding energy.13
Surprisingly, it was some time before this model was analysed in detail. In 1969, the first
serious analysis of thin accretion discs about black holes was carried out by Donald Lynden-
Bell, who showed how, in principle, they could account for the most extreme active galactic
nuclei known at that time (Lynden-Bell, 1969). Lynden-Bell had assumed that the black hole
was of the static spherically symmetric variety, but James Bardeen (b. 1939) pointed out in
1970 that the black hole is likely to possess angular momentum as the infalling matter brings
angular momentum with it (Bardeen, 1970). The energy release could be correspondingly
greater, up to 42% of the rest-mass energy of the infalling matter for a maximal corotating
black hole, in which the accretion disc could extend in to rg = GM/c2 . The luminosities of
the binary X-ray sources, discovered in 1972, could be naturally accounted for by accretion
onto neutron stars, and, in a number of cases, these approached the Eddington limiting
luminosity, indicating just how effective this energy source could be (Margon and Ostriker,
1973). Detailed models of accretion discs about black holes were published by Nikolai
Shakura (b. 1945) and Rashid Sunyaev in 1973 (Shakura and Sunyaev, 1973), and they
showed that, although the nature of the viscosity, which is responsible for the outward
transport of angular momentum and energy release, was not well understood, many of the
properties of thin accretion discs are independent of the specific form of the viscosity. Since
that time, accretion models for active galactic nuclei fuelled by accretion discs have been
adopted by many authors as their preferred model for active galactic nuclei. Shakura and
Sunyaev had pointed out that the viscous stresses might be associated with magnetic stresses
within the accretion, and in 1991 Steven Balbus (b. 1953) and John Hawley (b. 1958) revived
the instability first discussed by Evgenii Velikhov (b. 1935) and Chandrasekhar and showed
that magnetic instabilities in the differentially rotating disc could provide the necessary
transfer of angular momentum to provide a physical realisation for the viscous parameter
α introduced by Shakura and Sunyaev (Velikhov, 1959; Chandrasekhar, 1981; Balbus and
Hawley, 1991).
The thin accretion discs were successfully applied to cataclysmic variable stars and
relatively modest active galactic nuclei, but it proved very much more difficult to find self-
consistent solutions for the most luminous sources. As the luminosity of the disc increases,
 11.4 The spectroscopy of active galactic nuclei 279

most of the thin-disc approximations break down. In particular, the central regions of the disc
become very hot and radiation pressure can no longer be neglected. The innermost regions
of the disc are inflated and are expected to take up a toroidal configuration rather than a thin
disc. An attempt to construct a self-consistent thick-disc model applicable to active galactic
nuclei was proposed by Marek Abramowicz (b. 1945) and his colleagues in 1978, in which
the torus was inflated to such a degree that there were funnels along the axis of the torus
which were thought to be relevant to the formation of the jets observed to be ejected from
many active nuclei (Abramowicz, Jaroszyński and Sikora, 1978). Many of these ideas were
synthesised in the review of 1982 by Martin Rees and his colleagues, who pointed out the
many complications which arose in the attempt to construct self-consistent models of thick
discs (Rees et al., 1982). The problem with thick tori, just as in the case of the supermassive
stars studied by Hoyle and Fowler in 1964, is that they are grossly unstable. In 1984, John
Papaloizou (b. 1947) and James Pringle (b. 1949) showed that the most unstable modes of
thick discs are non-axisymmetric and global in character (Papaloizou and Pringle, 1984).

11.4 The spectroscopy of active galactic nuclei

Quasars and the nuclei Seyfert galaxies were ideal objects for spectroscopic study because
their spectra are rich in strong emission lines. The Seyfert galaxies were of particular
importance because their nuclei are generally much brighter than those of quasars and
excellent spectra could be acquired. The objective prism surveys for galaxies with strong
blue continua carried out by Markarian and his colleagues at the Byurakan Astrophysical
Observatory in Armenia were of special importance because, although about 90% of the
objects turned out to be what would now be called star-burst galaxies, the remaining 10%
were Seyfert galaxies. In studying these Seyfert galaxies, Eduard Khachikian (b. 1928) and
Daniel Weedman (b. 1942) discovered in 1974 that there are basically two types of emission-
line spectra (Khachikian and Weedman, 1971, 1974). Those with relatively narrow emission
lines, with line widths corresponding to velocities of about 500 km s−1 , were classed as Type
2 Seyfert galaxies. In contrast, the Type 1 Seyfert galaxies were found to have very broad
permitted emission lines, the line-widths corresponding to about 5000 km s−1 . Narrow
forbidden lines were also observed in the Type 1 Seyferts, but no forbidden counterparts
of the broad permitted lines were detected. These observations suggested that the broad
permitted emission lines originate close to the source of excitation in high-density gas
clouds in which the forbidden lines are suppressed; the narrow-line regions are of lower
density and are located much further away from the nucleus.
The optical continua of quasars were found to be of roughly power-law form and
polarised, and so, by the mid 1960s, it was commonly accepted that the smooth contin-
uum radiation observed in the spectra of active galactic nuclei is synchrotron radiation.
The intense power-law continuum spectrum suggested that the process of excitation of the
emission-line regions was the photoionisation of cool clouds in the vicinity of the nucleus.
The power-law nature of the continuum spectrum had the advantage that a wide range of
different states of ionisation would be expected to be present in the clouds, in agreement
with the observations.14 This working hypothesis was used throughout the 1970s and was
280 11 High-energy astrophysics

Figure 11.3: The optical–ultraviolet spectrum of the quasar 3C 273. The continuum spectrum has been
decomposed into a ‘power-law’ component, a component associated with recombination radiation,
and a ‘blue-bump’ component, which has been represented by a black-body curve. The prominent
Balmer series in the optical waveband which led to the discovery of the large redshifts of quasars can
be seen (Malkan and Sargent, 1982).

confirmed by observations of quasars and active galaxies by the International Ultraviolet

Observatory (IUE) in the late 1970s and early 1980s. A particularly striking observation
made by the IUE in 1982 was of the line and continuum spectrum of 3C 273 which shows
a ‘blue-bump’ continuum component extending into the ultraviolet region of the spectrum
(Figure 11.3) (Malkan and Sargent, 1982).
Convincing evidence for this picture was provided by the correlated variability of the
continuum ultraviolet radiation and the strength of the broad-line spectrum which was found
in a prolonged observing campaign with the IUE (Figure 11.4). More remarkable still was
the observation that there is a time delay between an outburst occurring in the nucleus and
the surrounding clouds responding to this burst of ionising radiation. In the Seyfert galaxy
NGC 4151, the delay is about 10 days, and so the size of the broad emission line regions
must be about a few light-days (Ulrich et al., 1984). Combining this dimension with the
velocities of the clouds inferred from their Doppler widths, the mass within the central
regions was estimated to be about 109M .
The natural extension of this technique is to study the time variations of a variety of
different emission lines in a single object since these are expected to originate at different
distances from the source of ionising radiation.15 The technique depends upon the fact
that, when clouds are excited by incident ultraviolet radiation, they respond essentially
instantaneously, and also upon the inference that the line-emitting clouds occupy only a
11.4 The spectroscopy of active galactic nuclei 281

Figure 11.4: Typical emission line profiles of the lines of CIV, CIII] and MgII in the Type 1 Seyfert
galaxy NGC 4151. The scales on the abscissa are velocities (km s−1 ). (a) The spectra in the vicinity of
these lines taken on 21 April 1980 when the source was in its low state. The continuum emission from
the nucleus is weak and the narrow-line spectra observed. (b) The spectra observed on 19 October
1978. The nucleus was bright, the continuum emission strong and broad wings are observed in the
CIV line, as in a Type 1 Seyfert galaxy (Ulrich et al., 1984).

very small fraction of the volume of the nucleus, that is they have a small filling factor, so
that the ultraviolet ionising photons propagate unhindered from the nucleus to the clouds.
Consequently, the time delay between variations in the continuum intensity and those in
the line-emitting clouds depend only upon the light-travel time from the nucleus and the
geometry of the clouds.
Obtaining high-quality data sets suitable for this type of analysis is a challenge because
of the need to coordinate observations from space- and ground-based observatories for
observing campaigns extending over periods of months or years. Examples of the application
of cross-correlation procedures to a beautiful set of optical and ultraviolet observations of
the Seyfert 1 galaxy NGC 5548 are shown in Figure 11.5 (Clavel et al., 1991; Peterson et al.,
1991). It is apparent from these results that the average time delays are different for atoms
282 11 High-energy astrophysics

Figure 11.5: Light-curves showing the time variation of the ultraviolet (135 nm) and optical (510 nm)
continuum intensity, as well as the fluxes of three prominent emission lines, Lyman-α (121.6 nm), CIV
(154.9 nm) and Hβ (510 nm) in the Seyfert 1 galaxy NGC 5538. The ultraviolet observations were
made with the IUE observatory and the optical observations are from ground-based telescopes (Clavel
et al., 1991; Peterson et al., 1991). In the right-hand column, the cross-correlation functions of the time
variation of the ultraviolet continuum intensity with the optical continuum and the different emission
lines are shown, the first panel being the autocorrelation function of the ultraviolet continuum. The
emission lines show the same pattern of variation as the ultraviolet and optical continuum, but with a
time delay due to the light-travel time from the nucleus to the broad-line regions.
 11.5 Masses of black holes in active galactic nuclei 283

and ions in different states of ionisation. As expressed by Bradley Peterson (b. 1951), lines
which are prominent in highly ionised gases, HeII, NV and CIV, have shorter lag times
than those prominent at lower ionisation levels, for example the Balmer lines. There must
therefore be stratification in the ionisation structure of the clouds about the active nucleus.
For example, the CIV lines originate from a quite different region from those of CIII].

11.5 The masses of black holes in active galactic nuclei

Important aspects of the study of active galactic nuclei were the search for unambigious
evidence for the presence of supermassive black holes and the accurate determination of
their masses. As early as 1964, Zeldovich and Igor Novikov (b. 1935) pointed out that
the masses of quasars had to be very large because their luminosities could not exceed the
Eddington limiting luminosity, L = 1.3 × 1031 (M/M ) W (Zeldovich and Novikov, 1964).
If this value were exceeded, radiation pressure of the source itself would blow it apart (see
Section A11.1.3). Since the luminosity of 3C 273 is at least 1040 W, it follows that the source
of energy must have mass greater than 109 M .
The general arguments in favour of the presence of supermassive black holes in active
galactic nuclei were persuasive, but finding unambiguous evidence proved to be far from
trivial.16 There are two separate parts to the argument. First of all, reliable estimates are
needed of the mass contained within the nuclear regions of galaxies. Useful quantities to
be estimated are the mass-to-luminosity ratio, M/L, of the central regions, which can be
compared with the values typically found in galaxies, and the mass density of the nuclear
regions. Large values of M/L imply the presence of a dark massive object, while the mass
densities can be compared with those of dense star clusters. The second step is to find
convincing evidence that the dark massive object is a black hole, rather than, say, a dense
cluster of stars.
In 1978, pioneering studies of the surface brightness distribution and velocity disper-
sion of stars in the nuclear regions of the nearby active galaxy M87 were carried out by
Peter Young (1954–1981), Wallace Sargent (b. 1935) and their colleagues, who found that
the mass-to-luminosity ratio increased towards the nucleus (Sargent et al., 1978; Young
et al., 1978). Assuming the velocity dispersion, or more precisely the velocity ellipsoid,
is isotropic, the presence of a dark central object with mass 3 × 109M was inferred. This
proved to be a controversial result, however, because, in the case of M87, the velocity field
of the stars showed negligible rotation and so the mass estimate depended strongly upon the
assumption that the velocity distribution of the stars was isotropic throughout the central
regions of the galaxy. Studies by Garth Illingworth, James Binney (b. 1950) and others had
already shown that the flattening of elliptical galaxies could not be wholly attributed to
rotation and hence the velocity ellipsoid must be anisotropic (Illingworth, 1977; Binney,
1978). The possibility could not be ruled out that, if the velocity ellipsoid in M87 were
anisotropic, the radial velocity dispersion might become larger closer to the nucleus, thus
removing the necessity of a massive central black hole.
A different approach was taken by Holland Ford (b. 1940) and his colleagues, who
discovered a disc of ionised gas about the nucleus of M87 from imaging observations made
284 11 High-energy astrophysics

Figure 11.6: The central regions of the giant elliptical galaxy M87 (NGC 4486) in the nearby Virgo
cluster of galaxies, as observed by the Hubble Space Telescope, showing the famous optical jet and the
disc of ionised gas surrounding the luminous point-like source in the nucleus. The kinematics of the
gaseous disc enabled the mass of the nuclear regions to be determined (Ford et al., 1994). (Courtesy
of Dr Holland Ford, NASA and the Space Telescope Science Institute.)

with the Hubble Space Telescope (Ford et al., 1994). As can be seen from Figure 11.6, the
famous optical jet lies along the axis of the nuclear disc. Spectroscopic observations by
Richard Harms (b. 1946) and his colleagues established that the disc is in Keplerian rotation
about the nucleus, the inferred mass of the nuclear regions being 3 × 109M (Harms et al.,
1994). The agreement between the mass inferred from the kinematics of the ionised disc and
the estimates of Sargent and Young is reassuring. It would be surprising if it were fortuitous.
As more data on the physical properties of active galactic nuclei and the timescales of their
variability became available in the 1980s, Amri Wandel (b. 1954) and Richard Mushotzky
(b. 1947) analysed the spectra of quasars and Seyfert galaxies which were known to be
intense variable X-ray sources (Wandel and Mushotzky, 1986). They estimated the masses
of the central objects in two ways. Firstly, they used optical spectroscopic data to estimate
the masses of the nuclei using the width of the emission lines as a measure of the typical
velocities of the clouds and photoionisation models to find the distances of these clouds
from the nuclei. The second approach was to assume that the X-ray variability originates
from roughly the last stable orbit about the black hole, r ≈ 3rg ≈ 10(M/M ) km, and so,
assuming the timescale of variability is of the order of r/c, estimates of the mass of the black
hole could be found (see Section A11.1.1). These two estimates were in good agreement, and
the luminosities of the sources were all significantly less than the corresponding Eddington
11.5 Masses of black holes in active galactic nuclei 285

Figure 11.7: (a) Comparison of mass estimates for active galactic nuclei from the variability of their
X-ray emission and from dynamical estimates. (b) Comparison of the inferred masses and luminosities
with the Eddington limiting luminosity, L E = 1.3 × 1031 (M/M ). All the objects lie well below the
Eddington limit (Wandel and Mushotzky, 1986).

luminosities (Figure 11.7). Thus, for these active galaxies and quasars, there was no problem,
in principle, in accounting for their extreme luminosities and the short timescales of their
variability, if it is assumed that their ultimate sources of energy were supermassive black
holes in their nuclei.
While these observations were strongly suggestive of the presence of supermassive black
holes in the nuclei of these active galaxies, conclusive evidence was provided by a beautiful
set of radio observations of the nearby galaxy NGC 4258, or M106, in the water vapour
(H2 O) maser line at a wavelength of 1.3 cm. Makoto Miyoshi (b. 1962) and his colleagues
mapped the location and velocities of the H2 O masers with an angular resolution of better
than one milliarcsec using the technique of very long baseline interferometry with the US
Very Long Baseline Array (VLBA) (Miyoshi et al., 1995). The water vapour maser lines are
very narrow, with typical velocity widths of about 1 km s−1 . Maser emission is only observed
if the radiation passes through a sufficiently long path-length of the masing medium without
the Doppler shift of the moving clouds shifting the wavelength of the radiation from the
masing wavelength; that is, the maser emission must originate from regions in which the
velocity gradient is zero. This occurs in regions in which the molecular ring is observed
tangentially, and also along the line of sight directly towards the nucleus. The rotation curve
derived from these observations is shown in Figure 11.8. It is important that, outside the
central ±3 milliarcsec, the rotational velocities follow a Keplerian law and so there must be
a compact massive object in the nucleus.
The mass of the central dark object was found to be about 3.7 × 107M , and the mass
density within the nuclear region was greater than 4 × 109M pc−3 . The latter figure is
about 40 000 times greater than the density of the densest globular clusters known in our
Galaxy, and it is so high that it is unlikely that the mass could be a cluster of any known
286 11 High-energy astrophysics

1500

1000
LSR velocity (km s−1)

500

550

0
500

450

400
−500 0.4 0.0 −0.4

9 8 7 6 5 4 3 2 1 0 −1 −2 −3 −4 −5 −6 −7 −8 −9

distance along major (milliarcsec)

Figure 11.8: The rotation curve of the nuclear regions of the galaxy NGC 4258 from VLBI observa-
tions of the water vapour line at 1.3 cm. Note that outside the inner ±3 milliarcsec, the radial velocities
follow a Keplerian rotation law, indicating the presence of a massive compact object in the nucleus
(Miyoshi et al., 1995).

type of stellar object. If the dark mass consisted of a cluster of normal stars, their number
density would be so great that the cluster would evaporate by dynamical friction. Only if the
objects were of low mass, M ≤ 0.03M , could the cluster survive for the age of the Galaxy.
Such objects would have to be planets or brown dwarfs, in which case their collision times
would be very short, roughly 105 years, and so their debris would rapidly fall towards the
centre of the system, resulting in the formation of a massive black hole. It seems inevitable
that there is a black hole with mass 3.7 × 107M in the nucleus of NGC 4258.
Equally spectacular have been the observations of the motions of stars about the centre
of our own Galaxy in the infrared waveband at 2 μm with the Keck 10-metre telescope.
Near-infrared observations were made by Andrea Ghez (b. 1965) and her colleagues with
a very-high-speed infrared camera designed for speckle observations. Those frames which
were diffraction-limited were selected and co-added. From observations over a period of
four years, the orbits of stars within half an arcsec of compact radio source Sgr A∗ , which
is assumed to define the Galactic nucleus, were measured (Figure 11.9(a)). These observa-
tions enabled the acceleration vectors of three of the stars to be determined, and, within the
uncertainties with which these are determined, the centre of gravitational attraction coin-
cides with the position of Sgr A∗ (Figure 11.9(b)) (Ghez et al., 2000). These data implied
that there is a dark object of mass (2.61 ± 0.35) × 106 M within the central 0.015 pc of
our Galaxy. The lower limit to the mass density within this region was 2.2 × 1012 M pc−3 .
11.5 Masses of black holes in active galactic nuclei 287

0.3

0.2
S0−2

offset (arcsec)
0.1

0 S0−1

−0.1
S0−4
−0.2

0.4 0.3 0.2 0.1 0 −0.1

offset (arcsec) (b)

Figure 11.9: (a) The orbits of stars about the Galactic centre determined by measurements of their
proper motions over a period of four years using near-infrared speckle techniques at 2 μm with the
Keck 10-metre optical–infrared telescope. (b) Estimates of the directions of the acceleration vectors of
the infrared stars S0-1, S0-2 and S0-4, the opening angles indicating the uncertainty in the measured
directions. The most likely location of the central attractor includes the non-thermal radio source Sgr
A∗ . (Ghez et al., 2000.)

This value far exceeds the mass density inferred for the nucleus of NGC 4258, and the
arguments given in the preceding paragraph apply with added force. For comparison, the
Schwarzschild radius of a black hole of mass 2.6 × 106 M is 2.6 × 10−7 pc – the size of
the radio source Sgr A∗ measured by VLBI is only about 20 times the Schwarzschild radius
of the central black hole.
A remarkable series of similar observations has been made using the facilities of the
European Southern Observatory. Ten years of high-resolution astrometric imaging enabled
Reinhard Genzel (b. 1952) and his colleagues to trace two-thirds of the orbit of the star
labelled S0-2 in Figure 11.9(a) about Sgr A* (Figure 11.10) (Schödel et al., 2002). These
observations show that the star is moving in a bound, highly elliptical orbit with an orbital
period of 15.2 years and that Sgr A∗ lies in one focus, as expected from Kepler’s laws of
planetary motion. The inferred mass of the black hole is (3.7 ± 1.5) × 106 M . Equally
impressive is the fact that the distance of closest approach to the nucleus is only 17 light-
hours, or 5 × 10−4 pc. To complete the story, the same set of infrared observations have
revealed the very faint infrared counterpart of Sgr A∗ . In addition, these observations have
discovered ‘flares’, probably associated with the process of accretion of mass onto the
central black hole from the accretion disc (Genzel et al., 2003).
Another, quite different, approach to demonstrating that supermassive black holes are
present in the nuclei of active galaxies has been provided by X-ray spectral observations of
asymmetric iron fluorescence lines in the spectra of Seyfert 1 galaxies. The most convincing
case involves the X-ray properties of the Seyfert galaxy MGC-6-30-15, which was known
to be a highly variable X-ray source, significant variations in its X-ray flux density being
observed on the timescale of hours, indicating that the emission originates from very close
to the nucleus itself.
288 11 High-energy astrophysics

Figure 11.10: (a) An image of the 2 × 2 arcsec2 field centred on the compact radio source Sgr A∗ ,
which is assumed to be located in the nucleus of our Galaxy. (b) The orbit of the infrared star S2 (the
same as S0-2 in Fig. 11.9(a), as determined by infrared observations taken with the 3.5-metre New
Technology Telescope at the ESO La Silla Observatory and the 8.2-metre YEPUN telescope of the
ESO VLT at Cerro Paranal. (Schödel et al., 2002.)

The most compelling model for the origin of the fluorescence lines involves a cool accre-
tion disc embedded in an X-ray-emitting halo which is the source of the X-ray continuum
radiation from the active galactic nucleus. Since the material of the disc is cooler than the
surrounding X-ray emission, elements such as iron are not fully ionised, and, although they
may be partially ionised, they possess filled K- and L-shells. The fluorescence process first
involves the absorption of an X-ray photon by an iron ion, the greatest cross-section being
for removal of one of the two K-shell electrons. The energy threshold for this process is
7.1 keV for neutral iron and it increases as the iron becomes more and more stripped of
electrons. There is a 34% probability that the vacancy in the K-shell is filled by the emission
of an X-ray photon of energy 6.4 keV in a permitted electromagnetic transition from the
L-shell. Computations by Andrew Fabian and his colleagues have shown that the 6.4 keV
fluorescent line of iron is by far the strongest of the fluorescent lines because of the large
cross-section for photoelectric absorption by the K-shell of iron and iron’s large cosmic
abundance (Matt, Fabian and Reynolds, 1997). The 6.4 keV fluorescent iron line therefore
acts as a tracer of the velocity field in the accretion disc, which extends into regions in which
special and general relativistic effects are large and can strongly influence the shape of the
profile of the 6.4 keV line.
The best example of this to date has been the observation of asymmetric broadening
of the 6.4 keV line in the spectrum of the Seyfert 1 galaxy MCG-6-30-15. Figure 11.11
11.5 Masses of black holes in active galactic nuclei 289

Figure 11.11: The broad iron line seen in a long ASCA observation of MCG-6-30-15 (Tanaka et al.,
1995). The dashed line shows a best fit of the data to the profile expected of an accretion disc about a
Schwarzschild black hole. The inclination of the disc is i = 30◦ .

shows the spectrum of the 6.4 keV fluorescent line as observed in a long integration by the
Japanese X-ray satellite ASCA (Tanaka et al., 1995). The profile of this emission line has
been determined by subtracting from the overall spectrum a smooth continuum spectrum.
The key features of this remarkable observation are that the spectral line has an abrupt cut-off
at energies greater than 6.7 keV, but extends to energies as low as about 4 keV to the low-
energy side of the line. This type of asymmetry occurs very naturally if the thin accretion
disc extends inwards towards the last stable orbit about the black hole. Two relativistic
effects lead to asymmetries in the line. The first is the gravitational redshift, which shifts the
spectrum to lower X-ray energies, and the second is the transverse Doppler effect, which
appears in the expression for the Doppler shift of the observed energy of photons emitted
from a source moving at an angle θ to the line of sight:

0
obs = v . (11.2)
γ 1 − cos θ
c

Thus, if the plane of the accretion disc lies at a large angle to the line of sight, θ → π/2, the
transverse Doppler shift associated with the Lorentz factor, γ , in the denominator dominates
and shifts the line to lower energies. Therefore, the asymmetric profile of the 6.4 keV line
has a natural interpretation if the largest redshifted parts of the line originate from close to
the last stable orbit of an accretion disc about a massive black hole and the disc is observed
more or less face-on.
In 1995, John Kormendy and Douglas Richstone plotted their best estimates of the black-
hole masses in the nuclei of nearby galaxies against the absolute magnitude of their spheroids
290 11 High-energy astrophysics

11
10 M87
N3115
9
N3377 N4594
8
N4258
°
log M•/M

7 M31
M32
6 Galaxy
5 M33
4
3 M15

2
−5 −10 −15 −20 −25
MB, bulge
Figure 11.12: A plot of the masses of supermassive black holes in the nuclei of nearby galaxies against
the absolute magnitude of their bulges, or spheroids (Kormendy and Richstone, 1995). The authors
warn that the correlation may only be the upper envelope of a distribution which extends to lower
values of M• .

for the cases in which they believed the masses were reasonably secure (Figure 11.12). It is a
challenge to work out all the selection effects that went into the construction of Figure 11.12
(Kormendy and Richstone, 1995). In particular, good estimates are found in those elliptical
galaxies which possess rapidly rotating nuclear discs. The black holes plotted may well
form only the upper envelope of the distribution of black-hole masses at a given absolute
magnitude, on the reasonable assumption that only those in which the evidence is compelling
are included.
It is intriguing that, interpreted literally, Figure 11.12 suggests that the black-hole mass
is proportional to the luminosity of the bulge of the galaxy. Adopting a standard mass-to-
luminosity ratio for the bulge, Kormendy and Richstone suggest that the black-hole mass
fraction is M• /Mbulge ∼ 0.002–0.003. This conclusion is the subject of debate, in view of
the many selection effects which go into the construction of Figure 11.12, particularly the
absence of low-mass black holes in luminous elliptical galaxies.

11.6 Non-thermal phenomena in active galactic nuclei

From the very beginning, the non-thermal properties of active galactic nuclei posed intrigu-
ing astrophysical puzzles, many of them traceable to the very short timescales of their
variability. In 1965, for example, Fred Hoyle, William Fowler and Wallace Sargent showed
 11.6 Non-thermal phenomena in active galactic nuclei 291

that the quasars are susceptible to what became known as the inverse Compton catastrophe
(Hoyle, Burbidge and Sargent, 1966). Using the causality relation r ≤ cτ , an upper limit, r ,
to the size of the emitting region could be found from the timescale, τ , of variability of the
source and hence a lower limit to the energy density of radiation could be determined. If the
continuum optical emission were synchrotron radiation, the energy densities of radiation
would be so great that the radiating relativistic electrons would lose much more energy by
the inverse Compton scattering of the optical photons to higher energies than by the emission
of the optical radiation itself, in the process creating intense fluxes of X- and γ -rays. The
same argument can be applied to the variable radio emission of the cores of radio quasars
and radio galaxies. It is found that there is a critical brightness temperature of about 1012 K,
above which radiation is lost preferentially by inverse Compton scattering rather than by
synchrotron radiation (see Section A11.2). In both cases, the electrons in the source regions
would have very short lifetimes and would have to be replenished to maintain the sources’
luminosities.
In a remarkably prescient paper of 1966, Martin Rees (b. 1942) showed how these prob-
lems could be overcome if the source components moved out of the nuclear regions at
relativistic speeds (Rees, 1966, 1967). Several relativistic effects contribute to the allevia-
tion of the problem of the short variability timescales, particularly aberration effects, which
make the sources appear to be more luminous and larger than they are in their rest frames
(see Section A11.4).

11.6.1 Double radio sources and the acceleration of high-energy electrons

During the late 1960s and early 1970s, the first high-resolution radio maps were made of
extragalactic radio sources with the new generation of Earth-rotation aperture synthesis
radio telescopes, and these began to reveal the details of the structures of these sources.
The culmination of these studies were maps of superb quality, such as that of Cygnus A
(Figure 11.1(b)) made with the US Very Large Array (Perley et al., 1984). In 1974, Bernard
Fanaroff (b. 1947) and Julia Riley (b. 1947) noticed that the morphologies of these radio
structures depended very strongly upon their radio luminosities. In powerful double radio
sources such as Cygnus A, the maximum radio surface brightness of the lobe structures
is observed in hot-spots towards the outer ends of the diffuse radio lobes, and these are
referred to as Fanaroff–Riley Class 2, or FR2, radio sources (Fanaroff and Riley, 1974). In
contrast, those sources in which the maximum surface brightness occurs less than halfway
from the active galactic nucleus to the edge of the diffuse radio lobe structure, the Fanaroff–
Riley Class 1, or FR1, sources, have radio luminosities which are much lower than those of
Class 2. Radio maps of the FR1 sources 3C 31 and 3C 66B are shown in Figure 11.13. The
distinction between the two classes occurs rather abruptly at a well defined radio luminosity
(Figure 11.14).
Many of the astrophysical problems associated with extragalactic radio sources were
derived from the pioneering radio maps made by the Cambridge one-mile and 5-kilometre
radio telescopes and the Westerbork Synthesis Radio Telescope. The features of particular
interest in these maps were the ‘hot-spots’ observed towards the leading edges of the double
292 11 High-energy astrophysics

(a) (b)

Figure 11.13: Examples of Fanaroff–Riley Class 1, or FR1, radio sources. (a) 3C 31 (courtesy of the
US National Radio Astronomy Observatory); (b) 3C 66B (Hardcastle et al., 1996), both observed by
the US Very Large Array at 6 cm wavelength.

Figure 11.14: A plot of radio luminosity at 1.4 GHz against optical absolute magnitude, illustrating
the distinction between FR1 and FR2 double radio sources (Owen and Ledlow, 1994).
11.6 Non-thermal phenomena in active galactic nuclei 293

radio structures. These components were inferred to have high energy densities in relativistic
particles and magnetic fields, and there were two reasons why these regions had to be con-
tinuously replenished with relativistic material (Longair, Ryle and Scheuer, 1973). Firstly,
the synchrotron lifetimes of the ultra-relativistic electrons in the most compact hot-spots
were less than the age of the source and so they would have to be continuously accelerated or
replenished. Secondly, the energy densities of the high-energy particles and magnetic fields
were too great to be confined within the hot-spots of the radio sources. Peter Scheuer, Martin
Ryle and I (b. 1941) inferred that the hot-spots are temporary phenomena which are contin-
uously supplied with energy from the active galactic nucleus by beams or jets of particles.
Direct evidence for these jets was subsequently found in the radio maps of many radio
galaxies and quasars, including Cygnus A. Variants of this model were proposed by Martin
Rees in 1971 (Rees, 1971) and by Peter Scheuer (1930–2001) in 1974 (Scheuer, 1974),
and the generic picture of continuous-flow models has remained the preferred picture to
account for the radio emission of the extended radio sources. Numerical simulations during
the 1980s demonstrated that many of the detailed features of the radio source structures
could be accounted for by gas dynamical processes involving the interaction of gaseous jets
with the ambient interstellar and intergalactic gas. The most elegant analytic model which
describes many of the key features involved in these processes was presented by Christian
Kaiser (b. 1970) and Paul Alexander (b. 1960), in which the gas dynamics of the double
sources is described by a self-similar solution of the equations of gas dynamics (Kaiser and
Alexander, 1997).
One of the difficulties associated with the presence of high-energy electrons in many
different astronomical environments is the fact that, as a gas of relativistic electrons and
its frozen-in magnetic field expand, energy is lost adiabatically. Thus, if the radio-emitting
electrons had been accelerated in the initial explosion of a supernova, such as that which
gave rise to the remnant Cassiopeia A, they would have lost all their energy in the adiabatic
expansion of the sphere of hot gas as it does work against the ambient interstellar gas. Simi-
lar problems were found in accounting for the high energy densities of relativistic electrons
observed in extended radio sources and the fluxes of energetic electrons in the interplane-
tary medium. Up till the mid 1970s, various acceleration mechanisms had been proposed,
but none of them gained any general acceptance. One popular mechanism proposed by
Enrico Fermi in 1949 involved stochastic acceleration of particles in elastic collisions with
interstellar clouds (Fermi, 1949). The problem was that the Fermi acceleration process was
second-order in the ratio of the velocity of the clouds to the speed of light and so was very
slow. Electromagnetic acceleration in the magnetospheres of neutron stars was proposed
by Gunn and Ostriker in 1969 (Ostriker and Gunn, 1969), but this did not overcome the
problem of accelerating the particles in extended source regions.
In 1977 and 1978, a number of independent authors, including Ian Axford (b. 1933),
Egil Leer (b. 1942), George Skadron (b. 1936), Germogen Krymsky (b. 1937), Anthony
Bell (b. 1952), Roger Blandford (b. 1949) and Jeremiah Ostriker showed that first-order
Fermi acceleration in strong shock-waves is a remarkably effective means of accelerating
particles and creating a power-law energy spectrum (Axford, Leer and Skadron, 1977;
Krymsky, 1977; Bell, 1978; Blandford and Ostriker, 1978). The idea is as follows: if
high-energy particles are present in the vicinity of a shock-wave, they can be scattered
294 11 High-energy astrophysics

back and forth across the shock-wave, and, in each crossing in either direction, the particles
obtain a fractional increase in energy of order v/c, where v is the velocity of the shock-wave.
Particles are lost from the accelerating region by the bulk motion of the material behind
the shock downstream, and it is straightforward to show that these two competing effects
lead to the formation of a power-law energy spectrum of the form N (E) dE ∝ E −2 dE.
The beauty of this process is that it depends only upon the particles encountering a strong
shock-wave and being scattered in the interstellar media on either side of the shock. The
energy acquired by the particles is derived from the kinetic energy of the matter behind the
shock-wave.17
This mechanism had great attractions because it meant that the particles are accelerated
wherever there are strong shocks. Thus, in supernova shells and at the interface between
jets and the interstellar or intergalactic media, where there are strong shock-waves, this
mechanism provides a means of accelerating the high-energy particles where they are
needed. There remained a number of problems with the mechanism, including the fact
that it has proved difficult to obtain spectra steeper than N (E) dE ∝ E −2 dE. In addition,
as applied to supernova remnants, particles could not be accelerated to energies much greater
than 1015 eV, at which energy the gyroradii of the electrons become comparable to the size
of the supernova remnant itself (Lagage and Cesarsky, 1983). Thus, some other mechanism
is needed to account for the very highest energy cosmic rays. Nonetheless, the discovery of
this mechanism has gone a long way to resolving the problem of accelerating high-energy
particles in a wide range of astronomical environments.
The origin of the magnetic fields in supernova remnants was solved by Stephen Gull
(b. 1949) in 1975 when he showed that an expanding supernova shell becomes unstable
to Rayleigh–Taylor instabilities as it decelerates (Gull, 1975). The instability results in
turbulent motions at the contact discontinuity between the expanding sphere of hot gas and
the shocked interstellar material. The turbulent eddies can stretch and wind up the magnetic
field until its energy density is roughly the same as that in the turbulent motions. Through
numerical simulations, Gull showed that the magnetic field in supernova remnants could
attain values up to about B ∼ 10 nT, more or less the values observed. Similar instabilities
are expected to take place at the interface between the jets and the intergalactic gas in
extragalactic radio sources and can account for the presence of magnetic fields in the source
components.

11.6.2 Superluminal radio sources

The continuous-flow model of double extragalactic radio sources laid responsibility for their
extended structures on jets of material ejected from the quasar or radio galaxy nucleus. The
radio jets observed in these sources were clear evidence for this process and were very close
relatives of the optical jets seen in objects such as M87 and 3C 273. Even more remarkable
was the evidence from very long baseline interferometric observations of structural changes
in the compact radio cores of radio quasars and radio galaxies by Marshall Cohen (b. 1926)
and his colleagues (Cohen et al., 1971; Whitney et al., 1971). By 1980, there was clear
evidence that the structures observed in these sources on the scale of milliarcseconds were
not stationary; the source components appeared to be moving apart at speeds exceeding the
 11.6 Non-thermal phenomena in active galactic nuclei 295

Figure 11.15: VLBI images of the nuclear regions of the quasar 3C 273 from 1977 to 1980. The
radio component is observed to move a distance of 25 light-years in 3 years, implying a superluminal
velocity of about ten times the speed of light (Pearson et al., 1981).

speed of light. In the case of 3C 273 (Figure 11.15), the fainter component appeared to move
a distance of about 25 light-years between 1977 and 1980, implying an apparent separation
velocity of about ten times the speed of light (Pearson et al., 1981, 1982). Observations
of these sources since that time have shown that the phenomenon is remarkably common
among radio sources with compact radio cores.18
There immediately followed a flurry of theoretical speculation about the origin of this
phenomenon. The simplest, and still the most plausible, explanation had already been
described by Martin Rees in his papers from the 1960s (Rees, 1966, 1967). If a source
component is ejected at a relativistic speed from the nucleus at a small angle to the line
of sight, the component can have an apparent velocity perpendicular to the line of sight
of up to γ v, where γ = (1 − v 2 /c2 )−1/2 is the Lorentz factor. This maximum value is
found when the angle of ejection of the component to the line of sight is sin θ = γ −1 ,
that is, θ ≈ γ −1 . In addition, the luminosity of the approaching component is enhanced by
the effects of aberration, and this may account for the one-sidedness of many of the jets
observed in active galactic nuclei.† It is generally assumed that these superluminal motions
are associated with the origin of the beams or jets responsible for fuelling the hot-spots and
extended radio structures.

† Some of the more important results concerning the effects of aberration and Doppler shifting upon the observed
intensities of sources of radiation are included in the Section A11.4.
296 11 High-energy astrophysics

Figure 11.16: (a) The evolution of the radio structure of the binary X-ray source GRS 1915+105 over
a three-week period in 1994 (Mirabel and Rodrigues, 1994). (b) A comparison between the properties
of the radio jets in quasars and microquasars (Mirabel and Rodrigues, 1998).

The problems of understanding the physics of the double radio structures observed
in extragalactic radio sources was, for a long time, the province of extragalactic radio
astronomers. In 1994, however, evidence was found for relativistic jets in Galactic binary
X-ray sources in which there is evidence for the presence of a black hole as the invisible
companion. Felix Mirabel (b. 1944) and Luis Rodrigues (b. 1948) observed the intense hard
X-ray source GRS 1915 + 105 during a radio outburst; they discovered that its radio struc-
ture was double and that the components were separating from the radio core at 1.25 times
the speed of light (Mirabel and Rodrigues, 1994) (Figure 11.16(a)). The X-ray properties
of the source strongly suggested that the energy source in the binary system is a black
hole. Since this discovery in 1994, several other examples have been discovered. Mirabel
and Rodrigues pointed out the remarkable similarities between the radio properties of these
sources, which they term microquasars, and the radio galaxies and radio quasars (Mirabel
and Rodrigues, 1998) (Figure 11.16(b)). The big advantage of studying the microquasars is
that the timescales of the phenomena associated with their variability and energy produc-
tion scale as the mass of the black hole, and so phenomena which would take thousands or
millions of years for a 109M black hole would be expected to take place on the timescale of
minutes or days in a 10M Galactic black hole. Such rapid changes in the radio and X-ray
emission of these sources have been observed.
 11.6 Non-thermal phenomena in active galactic nuclei 297

Relativistic beaming has also been invoked to solve the problems associated with the rapid
variability of the intense γ -ray sources discovered by the Compton Gamma-ray Observatory.
The Observatory carried out a complete survey of the γ -ray sky and among the most intense
sources discovered were variable γ -ray sources associated with those extreme radio quasars
which exhibited superluminal motions (Kniffen, Chipman and Gehrels, 1994). The γ -
ray luminosities were so great and the timescales of variability so short that the energy
density of γ -rays was very large, so large in fact that the γ -rays would be degraded into
electron–positron pairs through the pair-production process. The importance of this process
is determined by the compactness factor, C, defined to be the quantity
L γ σT
C= , (11.3)
4πm e c4 t
where L γ is the γ -ray luminosity at 1 MeV, σT is the Thomson cross-section and t is the
timescale of variability. Note that C is a measure of how far a 1 MeV γ -ray can propagate
through the source before it is destroyed by electron–positron production (see Section
A11.3). If the compactness factor is very much greater than unity, the γ -rays cannot escape
from the source. In addition, the γ -ray luminosities are extreme, particularly when account
is taken of the ultra-high-energy γ -radiation associated with some of the sources detected
by the Compton Gamma-ray Observatory (see Figure 7.11). As demonstrated in Section
A11.4, if it is assumed that the source components are relativistically beamed, the γ -ray
luminosity is enhanced by a factor of roughly κ 5 , where κ = γ [1 + (v/c) cos θ ] and θ is the
angle between the axis of ejection and the line of sight, and so the γ -rays can survive without
suffering degradation by electron–positron pair production, consistent with the observation
of superluminal motion of the radio components.

11.6.3 Jets in active galactic nuclei

These advances enabled an empirical picture of the physics of active galaxies and extra-
galactic radio sources to be developed, but there are many pieces of the story which are
uncertain. For example, the origin of the jets in active nuclei is unclear. One possibility is
that they are associated with the funnels which are formed along the rotation axes of thick
accretion discs, but these are thought to be unstable configurations. Other models ascribe
the jets to electromagnetic processes occurring close to the black hole itself and, in the case
of rotating black holes, there is a preferred axis along its rotation axis. In the type of model
considered by Martin Rees and his colleagues in 1982, electromagnetic torques are used to
extract the rotational energy of a rotating black hole, resulting in an outflow parallel to its
rotation axis (Rees, 1976; Blandford and Znajek, 1977).
The most promising models associate the initial collimation of the jets with the winding
up of magnetic fields lines frozen into the ionised gas in the accretion disc about a rotating
black hole. The most detailed studies of these complex relativistic magnetohydrodynamic
problems have been carried out by Richard Lovelace (b. 1941) and his colleagues, who have
performed analytic and numerical simulations of the force-free magnetic field structures
which arise as an initial dipolar magnetic field threading the accretion disc is wound up.
They find solutions in which ‘magnetic bubbles’ are expelled at relativistic speeds along the
298 11 High-energy astrophysics

3C266

3C368

3C324

3C280

3C65

50 kpc

Figure 11.17: A series of optical and infrared images of the structures associated with 3CR radio
galaxies at a redshift z ∼ 1 on the same physical scale. The images are in order of increasing physical
size. The images in the left-hand column were taken with the Hubble Space Telescope; those in the
right-hand column were taken at 2.2 μm using the UK Infrared Telescope. The characteristic double
radio structures of the radio sources as observed with the VLA are shown as white contour lines (Best
et al., 1996).

rotation axis of the disc (Lovelace and Romanova, 2003). Even the most extreme relativistic
jets can be accounted for by these models.
The jets which power the outer radio lobes in the most luminous radio sources can have
a dramatic influence upon any interstellar or intergalactic clouds surrounding the active
galaxy as they are engulfed by the expanding cocoon of high-energy particles and magnetic
field. This is most dramatically illustrated by the alignment effect, which was discovered
by Patrick McCarthy (b. 1961), Kenneth Chambers (b. 1956) and their colleagues in 1987
(Chambers, Miley and van Breugel, 1987; McCarthy et al.,1987). They found that the
optical images of large-redshift 3CR radio galaxies were aligned with the radio axes of the
double radio sources. The nature of the alignment effect was revealed by high-resolution
images of these radio galaxies with the Hubble Space Telescope obtained by Philip Best
(b. 1972), Huub Röttgering (b. 1963) and me (Best, Longair and Röttgering, 1996). In the
montage shown in Figure 11.17, the optical, infrared and radio structures of five luminous
3CR radio galaxies are shown on the same linear scale. The right-hand panels show the
infrared images of the galaxies at 2.2 μm and the associated radio sources, revealing the
underlying old populations of the radio galaxies. The HST images in the left-hand panels
show that the optical emission is aligned along the axis of the radio source, this emission
vastly outshining the optical emission from the galaxies themselves. There is little evidence
for the presence of the galaxy in the HST images, but rather the structures are dominated
by emission regions aligned with the direction of passage of the radio jets which power the
 11.6 Non-thermal phenomena in active galactic nuclei 299

outer radio lobes. The spectroscopic evidence suggests that the optical structures observed
in the radio sources with physical size less than about 120 kpc is associated with shock
excitation of pre-existing cool clouds in the vicinity of the radio galaxy (Best, Longair
and Röttgering, 2000). The natural interpretation of these observations is that the optical
emission is excited with shock-waves associated with the highly supersonic passage of the
radio jet through the interstellar and intergalactic medium in the vicinity of the host galaxy.

11.6.4 Unified models for active galaxies

One of the more remarkable industries of the last 20 years has been the attempt to ‘unify’
different types of active galactic nuclei into a coherent, unified picture. The concept of
unification resulted from the realisation that projection effects must be important in deter-
mining the observed properties of active galactic nuclei, and the question is how much of
their diversity can be accounted for in terms of a model in which a single class of active
galaxy is viewed at different angles to the line of sight.19
The importance of projection effects in distinguishing between Seyfert 1 and Seyfert
2 galaxies was first convincingly demonstrated by Robert Antonucci (b. 1954) and Joseph
Miller (b. 1941) in 1985 in their study of the nearby Seyfert 2 galaxy NGC 1068 (Antonucci
and Miller, 1985). When observed spectropolarimetrically, the polarised line emission was
found to be as broad as the broad permitted lines seen in Seyfert 1 galaxies. Antonucci and
Miller interpreted these observations in terms of a picture in which there is an ‘obscuring
torus’ about the nucleus so that, when the galaxy is observed at a small angle to the axis of the
torus, the nuclear regions, which contain the active nucleus itself and the broad-line emitting
regions, are observed and the galaxy is classified as a Seyfert 1 galaxy. The characteristic
Seyfert 1 broad-line regions and strong blue and UV continuum are only observed along
lines of sight close to the axis of the torus. If the axis of the torus is observed at a large
angle to the line of sight, the nuclear regions are obscured and only the narrow-line regions,
which are located much further from the nucleus than the broad-line regions, are observed
in direct light. The obscured central regions can, however, be observed in the light reflected
from clouds outside the region of the torus, perhaps by gas and dust in the narrow-line
regions. The continuum and broad-line emission from the nuclear regions are polarised in
the process of reflection by Thomson or Rayleigh scattering. In Antonucci’s review of 1993,
he provided a number of other compelling arguments for this unification picture (Antonucci,
1993).
Another unification scenario concerns those active galaxies which are strong radio
sources: the radio quasars and the radio galaxies. The importance of projection effects
in determining the observed properties of extragalactic radio sources was advocated by
Peter Barthel (b. 1952). Barthel was principally concerned with the unification of radio
galaxies with the radio quasars, and his analysis centred upon the properties of the classical
double radio sources in the 3CR catalogue (Barthel, 1989, 1994). The principal arguments
concerned the observation of superluminal motions in the radio quasars, suggesting that
they are observed at small angles to the line of sight, the one-sided jet emission observed in
the radio quasars, but not in the radio galaxies, the relative sizes of the radio structures of the
quasars and radio galaxies, and so on. In the unification picture, radio quasars are observed
300 11 High-energy astrophysics

Figure 11.18: Illustrating the unified model for the radio galaxies and radio quasars observed in
samples of bright radio sources. Radio quasars are observed when the axis of the radio source lies
within about 45◦ of the line of sight. When observations are made almost along the axis of the radio
jet, superluminal radio sources and blazars are observed.

when the nucleus is observed within a cone of half-angle roughly 45◦ and a radio galaxy is
observed when the nuclear regions are hidden by the obscuring torus (Figure 11.18). Highly
collimated radio jets are assumed to be emitted by the nucleus along the axis of the torus,
and these jets are responsible for powering the outer radio hot-spots and extended radio
lobes. If the radio jet is observed at an angle close to the line of sight, the emission of the
relativistic jet is strongly enhanced by aberration and Doppler effects and so superluminal
motions can be observed. It is natural to attribute the extreme BL Lac or blazar phenomena
to radio sources, in which the relativistic jets point almost precisely along the line of sight
to the observer. In a few extreme blazars, high-dynamic-range observations have enabled
the underlying double radio source to be observed. Another unification scenario concerns
the relation between the BL Lac objects and radio galaxies. A good case can be made for
BL Lac objects being the relativistically beamed jets originating in the nuclei of FR1 radio
galaxies (Urry and Padovani, 1994).

11.7 The γ-ray bursts

The discovery of γ -ray bursts was briefly recounted in Section 7.5. The paper by Ray
Klebesadel (b. 1932), Ian Strong (b. 1930) and Roy Olson of the Los Alamos National
Laboratory was published in 1973, six years after the first of these events had been discovered
by the Vela surveillance satellites (Klebesadel et al., 1973). The delay in publication was not
caused by security considerations, but rather because the authors wished to be certain that
the γ -ray bursts were genuine astronomical phenomena. The bursts lasted from a fraction
of a second to several minutes (Figure 11.21), and, during that time, the γ -ray bursts were
the most luminous objects in the γ -ray sky.
11.7 The γ -ray bursts 301

Figure 11.19: The locations in galactic coordinates of the 2704 γ -ray bursts recorded by the Burst
and Transient Source Experiment of NASA’s Compton Gamma-ray Observatory during its nine-year
mission. (Courtesy of NASA, G. J. Fishman and the CGRO Science Team.)

The Compton Gamma-ray Observatory (CGRO), launched in 1991, was equipped with
a detector system specifically designed to detect and locate γ -ray bursts, the Burst and
Transient Source Experiment (BATSE). Over the following nine years, this instrument
detected 2704 bursts and established beyond any question that they are uniformly distributed
over the sky (Figure 11.19). Even early in the mission, it was apparent that the faint and
bright bursts were uniformly distributed over the sky. Furthermore, the number counts of
the bursts showed departures from those expected of a uniform Euclidean distribution of
sources. Initially, it was thought that the uniformity of the distribution of the γ -ray bursts
could be most naturally interpreted in terms of a distribution of nearby sources in the
vicinity of the Solar System. The theorists had a field day, coming up with a very wide
variety of exotic explanations,20 the most plausible of which involved a nearby distribution
of neutron stars. In this picture, it was difficult, however, to account for both the isotropy of
their distribution and the uniform convergence of the numbers of events at low fluxes over
the whole sky.
The big problem was that the γ -ray bursts were of very short duration and the angular
resolution of the BATSE instrument was only a few degrees, so that there was not time or
adequate positional accuracy to make secure identifications of the sources. The solution
to the problem was stimulated by a theoretical paper by Peter Mészáros (b. 1943) and
Martin Rees in 1993 (Mészáros and Rees, 1993). They argued that, if the γ -ray bursts
were extragalactic phenomena, the energy densities during the event would be so extreme
that a relativistic shock-wave would be created in which electrons would be accelerated
to very high energies. These electrons would emit synchrotron radiation with a power-
law spectrum, consistent with the observed spectra of the bursts. Although the intense
γ -ray emission would last only a short time at γ -ray energies, the emission would remain
302 11 High-energy astrophysics

observable at lower energies. In other words, there would be an afterglow which would
appear progressively in the X-ray, then optical, infrared and radio wavebands.
The search for afterglows of γ -ray bursts became possible with the launch of the Italian–
Dutch BeppoSAX satellite. In 1997, the X-ray telescope was pointed at the position of
the γ -ray burst GRB 970228, within 8 hours of the event having taken place, and its X-
ray afterglow was discovered (Costa et al., 1997). Subsequently, afterglows were observed
throughout the electromagnetic spectrum. These observations enabled a precise position
for the burst to be determined and the association of this burst with a very faint galaxy
established by observations with the Hubble Space Telescope (Sahu et al., 1997). Within
four years, over 40 γ -ray-burst afterglows had been detected, and 30 of these could be
associated with distant host galaxies. Most of the galaxies associated with the γ -ray bursts
show evidence for active star formation, implying the presence of young massive stars. These
observations demonstrated that the γ -ray bursts are extreme events occurring in galaxies at
cosmological distances – the most distant γ -ray burst detected to date has redshift z = 4.5
(Andersen et al., 2000).
The γ -ray bursts are the most extreme high-energy astrophysical events yet discovered.
The very short timescales of variability of their energy release indicate that the sources
must originate from regions less than about 100 km in size on timescales much less than one
second. Hence, the bursts must involve stellar-mass objects, and the huge energy releases,
typically 1047 J if the emission is isotropic, would require the conversion of about one
solar mass to be converted into high-energy particles. Furthermore, the energy density
in γ -rays in the source region would be so great that the same problems afflicting the
extreme γ -ray luminosities of blasars and quasars, the degradation of the γ -rays by photon–
photon interactions, would apply equally severely. Simply from the energy and scale of
the emitting region, it followed that the source region must expand relativistically, and
therefore, just as in the extreme γ -ray sources, the γ -ray bursts must involve relativistic bulk
motions.21
Many of the results derived by Martin Rees in his papers of 1966 and 1967 can be
applied directly to the observed emission (Rees, 1966, 1967). In the simplest picture, the
source region can be taken to be a relativistic fireball, meaning a relativistically expanding
sphere which heats the surrounding gas and drives a relativistic shock-wave into it. If
the sphere is optically thick, it is expected that the radiation would be thermalised and
a thermal spectrum observed. However, the observed spectra of the γ -ray bursts are of
power-law form with N (ε) ∝ ε−α , where α ∼ 2–3 at energies greater than about 0.1–
1 MeV, and therefore the radiation must originate from optically thin regions later in the
expansion. The most attractive model involves the acceleration of electrons in the relativistic
shock-front by the first-order Fermi acceleration mechanism discussed in Section 11.6.1.
Thus, the general picture involves a sphere of ultra-relativistic electrons and magnetic fields
expanding relativistically. The characteristic evolution of the synchrotron spectrum of such
a sphere has proved to be a good match to the time evolution of the afterglows throughout
the electromagnetic spectrum (Figure 11.20).
Relativistic beaming alleviates the problems of γ γ annihilation, but does not solve the
energy problem. The fireball need not, however, be isotropic. In the case in which the
energy of the γ -ray burst is emitted in a narrow collimated beam, the energy requirements
11.7 The γ -ray bursts 303

100 101
Vi GRB 990510
−1 t =0.72 day
10

−1
10
−2 Vc
10
Fν(mJy)

Fν(mJy)
−3
10
10−3

−4 8.7 GHz
10 I band
V band χ2 = 42/69d.f.
5.0 keV
10
−5
10−5
101 100 101 102 109 1011 1013 1015 1017 1019
time (days) (a) ν (Hz) (b)

Figure 11.20: (a) Model light-curves (dotted lines) at X-ray, optical and radio wavelengths compared
with the observed evolution for the γ -ray burst GRB 990510. (b) The spectrum of the γ -ray burst
GRB 990510 at a wide range of wavelengths after 0.72 day. The dotted line shows the expectation of
the standard synchrotron radiation model (Mészáros, 2002).

would be very significantly reduced. The same general picture described above would apply,
provided the opening angle, θ, of the beam is greater than γ −1 , where γ is the Lorentz factor
associated with the bulk motion of the emitting region. As the jet sweeps up the surrounding
gas and decelerates, the Lorentz factor of the beam decreases, and the beaming associated
with the collimation of the jet itself is more important than the relativistic beaming. This
change is expected to be associated with a more rapid decline in the intensity of the γ -ray
burst, and this has been observed in the γ -ray burst GRB 990510 Figure 11.20(a).
If this jet model is adopted, the total energy requirements of the γ -ray bursts can be
evaluated, and it has been found that, rather than a dispersion in intrinsic luminosities from
about 1044 to 1047 J, if the radiation is isotropic, there is only about an order of magnitude
spread about the value 8 × 1043 J, a value typical of the energy released in the core collapse
of a supernova explosion. In fact, in a number of cases, the bursts are roughly coincident with
supernova explosions occurring in distant galaxies. The supernova –γ -ray burst association
was established with certainty for the γ -ray burst GRB 030329 observed on 29 March 2003,
for which the characteristic broad lines of an extremely energetic supernova were observed
within days of the event (Hjorth et al., 2003).
The distribution of the durations of the γ -ray bursts is bimodal, as illustrated in Fig-
ure 11.21. The results discussed above have been derived from studies of the longer duration
bursts, and it is not so clear that the same considerations apply to the bursts with duration
less than one second. While the relativistic fireball model involving relativistic jets can
account for many of the features of the bursts, there is less agreement about the ultimate
energy source for the γ -ray bursts. An appealing picture is to relate the phenomena to the
collapse of a very massive stellar remnant to form a Kerr black hole and then to associate
the relativistic jet with the presence of electric and magnetic fields in the vicinity of the
hole and the axisymmetric collapse of the outer envelope onto the collapsed core. This
picture is often referred to as a collapsar model for γ -ray bursts. The example of the γ -ray
burst GRB 030329, which is definitely associated with a supernova explosion, may be an
304 11 High-energy astrophysics

80

60
number of bursts

40

20

0
0.001 0.01 0.1 1 10 100 1000
T90 (s)
Figure 11.21: The distribution of γ -ray-burst durations from the bursts observed by the BATSE
experiment of the CGRO. (Courtesy of NASA, G. J. Fishman and the CGRO Science Team.)

example of such an event. There are, however, many other important possibilities, including
the merger of binary neutron stars, or of a neutron star and a black hole, or of two black holes,
which might be associated with the very short γ -ray bursts. Many of these possibilities are
discussed in the excellent review by Peter Mészáros (Mészáros, 2002).

Notes to Chapter 11

1 The luminosity, L(ν), of a source of synchrotron radiation depends upon the magnetic flux density,
B, and the number density of ultra-relativistic electrons, defined by N (E) dE = κVE x dE, as
L ν ∝ κVB (x−1)/2 , where V is the volume of the source region. Thus, a fixed luminosity can be
produced by a small number of electrons in a strong magnetic field, or a large number of electrons
gyrating in a weak magnetic field. Between these extremes, there is a minimum energy requirement
which corresponds closely to the equipartition of energy between the magnetic field and the high-
energy electrons. This minimum energy requirement is often referred to as the equipartition values
of the magnetic field strength and total electron energy. For details, see Malcolm Longair, High
Energy Astrophysics, vol. 2 (Cambridge: Cambridge University Press, 1997), Chapter 19.
2 The designation 3C refers to radio sources listed in the Third Cambridge Catalogue of Radio
Sources (Shakeshaft et al., 1955). An improved version of the catalogue was compiled by Andrew
Bennett entitled the Revised 3C Catalogue, or 3CR (Bennett, 1962).
3 The proceedings of the First Texas Symposium contains many of the most important papers
on radio sources, active galaxies and quasars published up to 1963: Quasi-stellar Sources and
Gravitational Collapse, eds, I. Robinson, A. Schild and E. L. Schucking (Chicago: University of
Chicago Press, 1965). A summary of the early observations of quasars is given by G. R. Burbidge
and E. M. Burbidge, Quasi-stellar Objects (San Francisco: W. H. Freeman and Company, 1967).
4 See Gold’s after-dinner speech in Robinson et al., eds, Quasar-stellar Sources and Gravitational
Collapse.
11.7 The γ -ray bursts 305

5 The arguments involved in the controversy concerning the origin of quasar redshifts are contained
in the volume by G. B. Field, H. Arp and J. N. Bahcall, The Redshift Controversy (Reading,
Massachusetts: W. A. Benjamin, 1973).
6 An excellent review of the history of black holes is presented by Werner Israel (b. 1931), Dark stars:
the evolution of an idea, in 300 Years of Gravitation, eds S. W. Hawking and W. Israel (Cambridge:
Cambridge University Press, 1987), pp. 199–276.
7 Schwarzschild volunteered for military service at the outbreak of the First World War in 1914 and
served in Belgium, France and Russia. He wrote the paper on his exact solution of Einstein’s field
equations while on service in Russia. Tragically, Schwarzschild died in May 1916 from the illness
pemphigus contracted while on military service in Russia.
8 The Kerr metric is somewhat more complicated than the Schwarzschild metric and can be written
in Boyer–Lundquist coordinates as follows:
 
2G Mr 1 4G Mra sin2 θ ρ
ds 2 = 1− dt 2 − 2 dt dφ + dr 2
ρc 2 c ρc 
  
2G Mra 2 sin2 θ
+ ρ dθ 2 + r 2 + a 2 + sin2
θ dφ 2
) ,
ρc2

where the black hole rotates in the positive φ direction; a = (J/Mc) is the angular momentum
of the black hole per unit mass,  = r 2 − (2GMr/c2 ) + a 2 and ρ = r 2 + a 2 cos2 θ. If the black
hole is non-rotating, J = a = 0 and the Kerr metric reduces to the Schwarzschild metric, equation
(11.1).
9 John Archibald Wheeler had conjectured that all multipoles of a black hole except its mass, angular
momentum and charge would be radiated away. It was only with the work of Carter and Hawking
that this remarkable result was proved for black holes.
10 These results can be derived from simple analyses of the form of the Schwarzschild metric,
as I demonstrate in Malcolm Longair, Theoretical Concepts in Physics (Cambridge: Cambridge
University Press, 2003), Chapter 17.
11 This result can be deduced from the Kerr metric given in endnote 8, noting that, for a maximally
corotating black hole, the event horizon occurs at rg = GM/c2 .
12 The physics of black holes is described in C. W. Misner, K. S. Thorne and J. A. Wheeler,
Gravitation (New York: W. H. Freeman and Company, 1973) and S. I. Shapiro and S. A. Teukolsky,
Black Holes, White Dwarfs and Neutron Stars: The Physics of Compact Objects (New York: Wiley
Interscience, 1983).
13 Many of the key features of accretion discs in astrophysics are described in the review by J. Pringle,
Accretion discs in astrophysics, Annual Review of Astronomy and Astrophysics, 19, 1981, 137–
162, and in the excellent monograph by J. Frank, A. King and D. Raine, Accretion Power in
Astrophysics, 3rd edn (Cambridge: Cambridge University Press, 2003).
14 Many astronomers contributed to the understanding of the optical spectra of active galactic nuclei.
Many of the basic ideas are included in the excellent monograph by Donald E. Osterbrock,
Astrophysics of Gaseous Nebulae and Active Galaxies (Mill Valley, California: University Science
Books, 1989).
15 Excellent examples of the results of this type of study and the problems of interpretation are
given by Bradley Peterson in An Introduction to Active Galactic Nuclei (Cambridge: Cambridge
University Press, 1997).
16 The problems of finding definitive evidence for black holes in the nuclei of galaxies have been
surveyed by Kormendy and Richstone (1995) and S. M. Faber, Black holes in galaxy centers,
Formation of Structure in the Universe, eds A. Dekel and J. P. Ostriker (Cambridge: Cambridge
University Press, 1999), pp. 337–359.
17 I have given an elementary derivation of these results in Chapter 21 of Longair, High Energy
Astrophysics, Vol. 2.
306 11 High-energy astrophysics

18 Many important references to early studies of superluminal motions can be found in J. A. Zensus
and T. J. Pearson, eds, Superluminal Radio Sources (Cambridge: Cambridge University Press,
1987).
19 Many different aspects of unification scenarios are discussed in G. V. Bicknell, M. A. Dopita and
P. J. Quinn, eds, First Stromlo Symposium: Physics of Active Galactic Nuclei, ASP Conference
Series, vol. 34 (San Francisco: ASP, 1994), and in the book by E. I. Robson, Active Galactic Nuclei
(Chichester: John Wiley & Sons, in association with Praxis Publishing, 1996).
20 In his brief review of 1994, Robert Nemiroff (b. 1960) lists 100 models for γ -ray bursts. These
include the merger of binary neutron stars or of a neutron star and a black hole, asteroids or comets
falling onto neutron stars and black holes, and so on (Nemiroff, 1994).
21 An excellent survey of the physics of γ -ray bursts is given by Peter Mészáros in his review article
(Mészáros, 2002).

A11 Explanatory supplement to Chapter 11

A11.1 Black holes as energy sources in high-energy astrophysics
The preferred model for the energy source in compact non-thermal sources involves accre-
tion of matter onto black holes. The reasoning involves the following physical arguments.

A11.1.1 Time variability

One of the characteristic features of many luminous compact sources is their rapid time
variability. For any mass M, the smallest physically meaningful dimension is the radius of
the surface of infinite redshift, or event horizon, about a black hole of this mass. Radiation
emitted from this radius is detected with zero frequency, or infinite wavelength, by an
observer at infinity. For spherically symmetric, Schwarzschild black holes, this radius is
at rg = 2GM/c2 = 3(M/M ) km, and for maximally rotating Kerr black holes it occurs at
rg /2 = GM/c2 . Therefore, to order of magnitude, causality implies that the shortest possible
timescale associated with objects of mass M is given by

tmin ∼ rg /c = 10−5 (M/M ) s. (A11.1)

There are two points to be made about this relation. The first is that there is a last stable orbit
about any black hole. For Schwarzschild black holes, this lies at 3rg , and for maximally
rotating Kerr black holes it is at rg /2. Nonetheless, to order of magnitude, equation (A11.1)
is a useful lower limit to the timescale of variation associated with any mass M. The second
point is that this calculation neglects the possibility that the sources of radiation might be
moving relativistically. We discuss these effects in Section A11.4.

A11.1.2 The efficiency of accretion of matter onto black holes as an energy source
Accretion of mass onto compact objects is a very powerful source of energy for high-
energy astrophysical objects. Suppose a mass element, m, is dropped from infinity onto
a star with mass M and radius r . The kinetic energy acquired by the mass m increases
as it falls towards the surface, and this is dissipated as heat when the material hits the
surface. Thus, by conservation of energy, the energy released is GMm/r , and, if the
 Explanatory supplement to Chapter 11 307

material is continuously accreted at a rate ṁ, the luminosity due to accretion is given
by GM ṁ/r . Expressing this result in terms of the Schwarzschild radius, rg = 2GM/c2 , we
obtain

GM ṁ ṁc2 rg
L= = . (A11.2)
r 2 r
Thus, for matter accreting onto the surface of 1M neutron stars with radii 10 km, the
luminosity is expected to be 0.15ṁc2 . In fact, this is an overestimate since this Newtonian
calculation has neglected the effects of general relativity, which become important when
rg /r approaches unity. In the case of a spherically symmetric black hole, the maximum
luminosity is L = 0.057ṁc2 . Nonetheless, this efficiency of energy release is still about an
order of magnitude greater than that associated with nuclear processes. For comparison,
the conversion of hydrogen into helium only releases 0.7% of the rest-mass energy of the
hydrogen nuclei. The corresponding energy release for a maximally rotating Kerr black
hole is 0.42mc 2 . Thus, in principle, the process of accretion onto black holes can release a
substantial fraction of the rest-mass energy of the infalling material.

A11.1.3 The Eddington luminosity

From the considerations of Section A11.1.2, it might seem that, by increasing the mass
accretion rate, ṁ, an arbitrarily large luminosity could be obtained, but this is not the case,
because the radiation emitted by the compact object provides a radiation pressure which can
halt the accretion process. The force of radiation pressure acts upon the electrons, which
are strongly coupled electrostatically to the protons and nuclei in the plasma. Consequently,
the inward force acting on 1 m3 of the infalling matter at radius r is G(m e + m p )Np M/r 2 ≈
Gm p Np M/r 2 , where Np = Ne is the number density of electrons or protons. The radiation
pressure acting on 1 m2 of material at radius r is p = σT Ne Urad , where σT is the Thomson
cross-section and the energy density of radiation at radius r from a point source of lumi-
nosity L is Urad = L/4πr 2 c. Equating the radiation pressure and gravitational forces, we
find
σT N e L Gm p Np M
= . (A11.3)
4πr 2 c r2

Note that both forces depend upon radius as 1/r 2 and so the limiting luminosity, known as
the Eddington luminosity, is independent of radius:
 
4π Mm p c M
L Edd = = 1.4 × 1031 W. (A11.4)
σT M

Note that, in this calculation, we have adopted the minimum possible cross-section for
the scattering of radiation by electrons. Thus, for isotropic radiation, the Eddington limit
provides a rather firm upper limit to the luminosity of any source powered by accretion.
Eddington’s name is associated with this formula since the same result is found in consid-
ering the upper limit for the stability of massive stars which are radiation-dominated.
308 11 High-energy astrophysics

A11.2 Synchro-Compton radiation and the inverse Compton catastrophe

Wherever there are large number densities of soft photons, the presence of ultra-relativistic
electrons in the same region must result in the production of high-energy photons, X-rays
and γ -rays, by the process of inverse Compton scattering. The case of special interest in
this chapter is that in which the same relativistic electrons which are the source of the soft
photons are also responsible for scattering these photons to X-ray and γ -ray energies – this
process is known as synchro-Compton radiation. One case of special importance is that in
which the number density of low-energy photons is so great that most of the energy of the
electrons is lost by synchro-Compton radiation rather than by synchotron radiation. This
line of reasoning leads to what is known as the inverse Compton catastrophe.
We can derive the essential results from the formulae for synchrotron and inverse Comp-
ton radiation.1 The mean energy loss rates for an electron of Lorentz factor γ in a magnetic
field of flux density B and radiation field of energy density Urad are, respectively, given by
   2
dE B
− = 3 σT c
4
γ 2, (A11.5)
dt sync 2μ0
 
dE
− = 43 σT cUrad γ 2 . (A11.6)
dt IC

Thus, the ratio, η, of the rates of loss of energy of an ultra-relativistic electron by inverse
Compton and synchrotron radiation is given by

(dE/dt)IC Uphoton
η= = 2 . (A11.7)
(dE/dt)sync B /2μ0

The synchro-Compton catastrophe occurs if this ratio is greater than unity. In that case,
low-energy photons, say, radio photons produced by synchrotron radiation, are scattered to
X-ray energies by the same flux of relativistic electrons. Since η is greater than unity, the
energy density of the X-rays is greater than that of the radio photons and so the electrons
suffer an even greater rate of loss of energy by scattering these X-rays to γ -ray energies.
In turn, these γ -rays have a greater energy density than the X-rays . . . and so on. It can be
seen that, as soon as the ratio given in equation (A11.7) becomes greater than unity, all the
energy of the electrons is lost at the very highest energies and so the radio source should
instead be a very powerful source of X-rays and γ -rays. Before considering the higher-order
scatterings, let us study the first stage of the process for the case of a compact synchrotron
self-absorbed radio source.
We need to determine the energy density of radiation within a synchrotron self-absorbed
radio source. The flux density of such a source is given by

2kTe r2
Sν = , where  ≈ θ2 = ; (A11.8)
λ2 D2
 is the solid angle subtended by the source, r is the size of the source and D is its distance.
For a synchrotron self-absorbed source, the electron temperature of the relativistic electrons
is the same as the brightness temperature of the source, Te = Tb . The radio luminosity of
Explanatory supplement to Chapter 11 309

the source is given by

8π kTe 2
L ν = 4π D 2 Sν = r . (A11.9)
λ2
Therefore, the energy density of the radio emission, Urad , is given by
Lν ν 2kTe ν
Urad ∼ = 2 . (A11.10)
2
4πr c λc
Note that we have used the fact that L ν is the luminosity per unit bandwidth, and so the
bolometric luminosity is roughly ν L ν . Therefore,
 
2kTe ν
λ2 c 4kTe νμ0
η=  2  = 2 2 . (A11.11)
B λ cB
2μ0
We can now use the theory of self-absorbed radio sources to express the magnetic flux
density, B, in terms of observables. The frequency of emission, ν, is related to the non-
relativistic gyrofrequency, νg = eB/2πm e , by
νg = ν/γ 2 , (A11.12)
and the relation between temperature and Lorentz factor in the relativistic limit is given by
3kTb = 3kTe = γ m e c2 , (A11.13)
where Tb is the brightness temperature of the source. Reorganising these relations, we find
 2
2πm e m e c2
B= ν. (A11.14)
e 3kTe
Therefore, the ratio of the loss rates, η, is given by
 
(dE/dt)IC 81e2 μ0 k 5
η= = νTe5 . (A11.15)
(dE/dt)sync π 2 m 6e c11
The important result is that the ratio of the loss rates depends very strongly upon the
brightness temperature of the radio source. Putting in the values of the constants, the critical
brightness temperature is given by
−1/5
Tb = Te = 1012 ν9 K, (A11.16)
where ν9 is the frequency at which the brightness temperature is measured in units of
109 Hz, that is, in gigahertz. Thus, according to this calculation, no compact radio source
should have brightness temperature greater than TB ≈ 1012 K if the emission is incoherent
synchrotron radiation.
The most compact sources, which have been studied by VLBI at centimetre wavelengths,
have typical brightness temperatures TB ≈ 1011 K which are less than the synchro-Compton
limit. Note that this is direct evidence that the radiation is the emission of relativistic
electrons, since the temperature of the emitting electrons must be at least 1011 K. This is
not, however, the whole story. If the timescales of variability, τ , of the compact sources
310 11 High-energy astrophysics

are used to estimate their physical sizes, l ∼ cτ , the source regions must be considerably
smaller than those inferred from VLBI, and then values of TB exceeding 1011 K are found.
It is likely that relativistic beaming is the cause of this discrepancy.
It might appear that higher-order scatterings would result in a divergent situation in which
the X-rays would be scattered to γ -ray energies. In fact, this does not occur because, at rel-
ativistic energies hν ≥ 0.5 MeV, the Klein–Nishina cross-section rather than the Thomson
cross-section should be used for photon–electron scattering. In the ultra-relativistic limit,
the cross-section is given by

π 2 re2 1
σKN = ln(2hν) + , (A11.17)
hν 2
and so the cross-section decreases as (hν)−1 at high energies. Consequently, higher-order
scatterings result in much reduced luminosities as compared with the non-relativistic cal-
culation.

A11.3 The compactness parameter

The compactness parameter arises in considerations of whether or not a γ -ray source is
opaque for γ γ collisions because of pair production. Let us carry out a simple calculation
which indicates how the compactness parameter arises. We carry out the calculation for
the flux of γ -rays at the threshold for electron–positron pair production, ε ∼ m e c2 , for
simplicity. The mean free path of the γ -ray for γ γ collisions is λ = (Nγ σ )−1 , where Nγ is
the number density of photons with energies ε = hν ∼ m e c2 . If the source has luminosity
L γ and radius r , the number density of photons within the source region is given by
Lγ
Nγ = . (A11.18)
4πr 2 cε
The condition for the source to be opaque is r ≈ λ, that is
4πr 2 cm e c2 Lνσ
r∼ , that is ∼ 1. (A11.19)
Lνσ 4π m e c3 r
The compactness factor, C, is defined to be the quantity
Lνσ
C= . (A11.20)
4π m e c3r
Note that sometimes the compactness parameter is defined without the factor of 4π in the
denominator. If the compactness parameter is very much greater than unity, the γ -rays are
all destroyed by electron–positron pair production, resulting in a huge flux of electrons
and positrons within the source region. Consequently, the source would no longer be a
hard γ -ray source. The significance of the compactness parameter can be appreciated from
observations of some of the intense γ -ray sources observed by the Compton Gamma-ray
Observatory. These have enormous luminosities, L γ ∼ 1041 W, and have been observed to
vary significantly in intensity over timescales of the order of days. Inserting these values
into equation (A11.20), it is found that C 1, and so there is a problem in understanding
why these sources exist. Fortunately, an answer is at hand since all the ultra-luminous
 Explanatory supplement to Chapter 11 311

Figure A11.1: The relativistic ballistic model of superluminal sources.

γ -ray sources are associated with compact radio sources, which exhibit synchrotron self-
absorption, and many of which display superluminal motions. The inference is that the
luminosities of the γ -ray sources and the timescales of variation have been significantly
changed by the relativistic motion of the source region; we turn to this issue next.

A11.4 Superluminal motion and relativistic beaming

The most popular model for superluminal sources is known as the relativistic ballistic model.
The simplest part of the calculation is the determination of the kinematics of relativistically
moving source components. The aim is to determine the observed transverse speed of a
component ejected at some angle, θ, to the line of sight at a high velocity v (Figure A11.1).
The observer is located at a distance D from the source. The source component is
ejected from the origin, O, at some time t0 , and the signal from that event sets off towards
the observer, where it arrives at time t = D/c later. After time t1 , the component is located at
a distance vt1 from the origin and so is observed at a projected distance vt1 sin θ according
to the distant observer. The light signal bearing this information arrives at the observer at
time
D − vt1 cos θ
t2 = t1 + , (A11.21)
c
since the signals have to travel a slightly shorter distance D − vt1 cos θ to reach the observer.
Therefore, according to the distant observer, the transverse speed of the component is given
312 11 High-energy astrophysics

Figure A11.2: Illustrating the geometry of the propagation of light from a circular annulus on the
surface of the Sun to an observer in a spaceship moving radially towards the Sun at speed v.

by

vt1 sin θ vt1 sin θ v sin θ

v⊥ = = = . (A11.22)
t2 − t vt1 cos θ v cos θ
t1 − 1−
c c
It is a simple sum to show that the maximum observed transverse speed occurs at an angle
cos θ = v/c and is given by v⊥ = γ v, where γ = (1 − v 2 /c2 )−1/2 is the Lorentz factor.
Thus, provided the source component moves at a speed close enough to the speed of light,
apparent motions on the sky, v⊥ > c, can be observed without violating causality and the
postulates of special relativity. For example, if the source component were ejected at a speed
0.98c, transverse velocities up to γ c = 5c are perfectly feasible.
This is the easy bit of the story. The trickier bit is to understand the effects of what
is loosely referred to as ‘relativistic beaming’ upon the observed intensities of the source
components. Let us consider first a classical undergraduate problem in relativity:

A rocket travels towards the Sun at speed v = 0.8c. Work out the luminosity, colour, angular size
and brightness of the Sun as observed from the spaceship when it crosses the orbit of the Earth. It
may be assumed that the Sun radiates like a uniform disc with a black-body spectrum at temperature
T0 .

This problem includes many of the effects found in relativistic beaming problems. Let
us work out the separate effects involved in evaluating the intensity of radiation observed
in the moving frame of reference. Consider the radiation from an annulus of angular width
θ at angle θ with respect to the centre of the Sun (Figure A11.2).
Explanatory supplement to Chapter 11 313

The frequency shift of the radiation

It is simplest to use four-vectors to work out the frequency shifts and aberrations. The
frequency four-vector in the frame of the Solar System, S, in Rindler’s notation2 is given by
ω 
0
K = , −k0 cos θ, −k0 sin θ, 0 , (A11.23)
c
where the light rays are assumed to propagate towards the observer at the orbit of the Earth,
as illustrated in Figure A11.2. The frequency four-vector in the frame of reference of the
spaceship, S , is given by

ω
K = , −k cos θ , −k sin θ , 0 . (A11.24)
c
We use the time transform to relate the ‘time’ components of the four-vectors:
 
Vx
ct = γ ct − , (A11.25)
c
and so
 
ω ω0 V k0 cos θ
=γ + . (A11.26)
c c c
Since k0 = ω0 /c,
 
V
ν = γ ν0 1+ cos θ = κν0 . (A11.27)
c
This is the expression for the ‘blueshift’ of the frequency of the radiation due to the motion
of the spacecraft.

The waveband ν
This waveband, in which the radiation is observed, is blueshifted by the same factor,
ν = κν0 . (A11.28)

Time intervals
These are also different in the stationary and moving frames. This can be appreciated by
comparing the periods of the waves as observed in S and S :
1 1
ν = ; ν0 = , (A11.29)
T T0
and so
T ν0
= . (A11.30)
T ν
Since the periods T and T can be considered to be the times measured on clocks, the
radiation emitted in the time interval t is observed in the time interval t by the observer
in S such that
t = t/κ. (A11.31)
314 11 High-energy astrophysics

Solid angles
Finally, we need to work out how the solid angle subtended by the annulus shown in
Figure A11.2 changes between the two frames of reference. It is simplest to begin with the
cosine transform, which is derived from the ‘x’ Lorentz transformation of the frequency
four-vector:
V
cos θ +
cos θ = c . (A11.32)
V
1+ cos θ
c
Now, differentiating with respect to θ and θ on both sides of this relation, we find
sin θ dθ sin θ dθ
sin θ dθ =  2 = . (A11.33)
V κ2
γ2 1+ cos θ
c
This result has been derived for an annular solid angle with respect to the x-axis, but we
can readily generalise to any solid angle since dφ = dφ and so
sin θ dθ dφ d
sin θ dθ dφ = ; d = . (A11.34)
κ2 κ2
Thus, the solid angle in S is smaller by a factor κ 2 as compared with that observed in S.
This is a key aspect of the derivation of the aberration formulae. Exactly the same form of
beaming occurs in the derivation of the formulae for synchrotron radiation.

We can now put these results together to work out how the intensity of radiation from the
region of the Sun within solid angle d changes between the two frames of reference. First
of all, the intensity I (ν) is defined to be the power arriving at the observer per unit frequency
interval per unit solid angle from the direction θ . The observer in the spacecraft observes the
radiation arriving in the solid angle d about the angle θ , and we need to transform its other
properties to those observed in S . Let us enumerate how the factors change the observed
intensity. The energy, hν N (ν), received in S in the time interval t, in the frequency interval
ν and in solid angle  is observed in S as an energy hν N (ν ) in the time interval t ,
in the frequency interval ν and in solid angle  , where N (ν) = N (ν ) is the invariant
number of photons. Therefore, the intensity observed in S is given by
κ × κ × κ2
I (ν ) = I (ν) × = I (ν)κ 3 . (A11.35)
κ
Now, let us apply this result to the spectrum of black-body radiation, for which
2hν 3  hν/kT −1
I (ν) = 2
e −1 . (A11.36)
c
Then,
2hν 3 κ 3  hν/kT −1 2hν 3 hν /kT −1
I (ν ) = e − 1 = e −1 , (A11.37)
c2 c2
Explanatory supplement to Chapter 11 315

where T = κ T . In other words, the observer in S observes a black-body radiation spectrum

with temperature T = κ T . A number of useful results follow from this analysis. For exam-
ple, equation (A11.37) describes the temperature distribution of the cosmic microwave
background radiation over the sky as observed from the Solar System, which is moving
through the frame of reference in which the sky would be perfectly isotropic on the large
scale at a velocity of about 600 km s−1 . Since V /c ≈ 2 × 10−3 and γ ≈ 1, the temperature
distribution is rather precisely a dipole distribution, T = T0 [1 + (V /c) cos θ ] with respect
to the direction of motion of the Solar System through the cosmic microwave background
radiation (see Figure 15.9(b)).
In the example of the spacecraft travelling at v = 0.8c towards the Sun, we can illustrate
a number of the features of relativistic beaming. In this case, γ = 5/3 and the angle at which
there is no change of temperature, corresponding to γ [1 + (V /c) cos θ ] = 1, is θ = 60◦ .
Let us now turn to the case of relativistically moving source components. Evidently, all
we need do is determine the value of κ for the source component moving at velocity V at
an angle θ with respect to the line of sight from the observer to the distant radio source, as
illustrated in Figure A11.1. In this case, a straightforward calculation shows that the value
of κ is given by
1
κ=  , (A11.38)
V cos θ
γ 1−
c
where the source is moving towards the observer as illustrated in Figure A11.1. Just as in
the above example, the observed flux density of the source is therefore
L(ν0 )
S(νobs ) = × κ 3, (A11.39)
4π D 2
where νobs = κν0 . In the case of superluminal sources, the spectra can often be described
by a power law, L(ν0 ) ∝ ν0−α , and so

L(ν0 )
S(ν0 ) = × κ 3+α . (A11.40)
4π D 2
Thus, if the superluminal sources consisted of identical components ejected from the
radio source at the same angle to the line of sight in opposite directions, the relative intensities
of the two components would be in the ratio
⎛ v ⎞3+α
1 + cos θ
S1 ⎜ c ⎟
=⎝ v ⎠ . (A11.41)
S2 1 − cos θ
c
It is therefore expected that there should be large differences in the observed intensities of
the jets. For example, if we adopt the largest observed velocities for a given value of γ ,
cos θ = v/c, then, in the limit v ≈ c,
S1
= (2γ 2 )3+α . (A11.42)
S2
316 11 High-energy astrophysics

Thus, since values of γ ∼ 10 are quite plausible and α ∼ 0-1, it follows that the advancing
component would be very much more luminous than the receding component. It is, therefore,
not at all unexpected that the sources should be one-sided.
Another complication is the fact that the emission is often assumed to be associated with
jets. Care has to be taken because, if the jet as a whole is moving at velocity v, then the time
dilation formula, equation (A11.31), shows that the advancing component is observed in
a different proper time interval as compared with the receding component, the time which
has passed in the frame of the source being t1 = κt0 , where t0 is the time measured in
the observer’s frame of reference. If the jet consisted of a stream of components ejected at
a constant rate from the active galactic nucleus, the observed intensity of the jet would be
enhanced by a factor of only κ 2+α . Thus, the precise form of the relativistic beaming factor
is model-dependent, and care needs to be taken about the assumptions made.
Let us first apply these considerations to the cases of sources exceeding the limiting
surface brightness, Tb = 1012 K, discussed in Section A11.2 and the compactness param-
eter discussed in Section A11.3. In the case of the inverse Compton catastrophe, equation
(A11.15) shows that the ratio of the loss rates for inverse Compton scattering and synchrotron
radiation depends upon the product νTb5 . Since the brightness temperature Tobs = κT0 and
νobs = κν0 , it follows that η ∝ κ 6 , and so the observed value of Tb can exceed 1012 K if the
source is moving at such a high velocity that κ 1.
In the case in which the compactness parameter,
L ν σT
C= , (A11.43)
4π m e c3 × ct
far exceeds unity, the relativistic beaming factors enable us to understand why these sources
should exist. In equation (A11.36), it is assumed that the dimensions of the source are l ≈ ct
from its rapid time variability. The observed luminosity is enhanced by a factor κ 3+α , and,
in addition, because the timescale of variability appears on the denominator of equation
(A11.36), the observed value is shorter by a factor κ, and so the compactness parameter is
increased by relativistic beaming by a factor of roughly κ 4+α . Since α ≈ 1, it can be seen
that C ∝ κ 5 , and so, in the frame of the source components themselves, the value of the
compactness parameter can be reduced below the critical value.

Notes to Section A11

1 I have given detailed derivations of the formulae for synchrotron and inverse Compton radiation
in Malcolm Longair, High Energy Astrophysics, vols I & II (Cambridge: Cambridge University
Press, 1997, revised editions).
2 In Rindler’s notation, the components of the four-vectors transform exactly as [ct, x, y, z] according
to the standard Lorentz transformation ct = γ (ct − V x/c), x = γ (x − V t), y = y, z = z. The
invariant norm of the four-vector is |R|2 = c2 t 2 − x 2 − y 2 − z 2 .
Part V
Astrophysical cosmology since
1945

The final five chapters of this book concern different aspects of cosmology. From a subject
which barely existed before the Second World War, it developed into one of the central
pillars of modern physics and astrophysics.
12 Astrophysical cosmology

This chapter concerns the development of astrophysical cosmology from 1945 to the early
1970s, by which time the success of the standard Big Bang models convinced the com-
munity at large that these provided the most satisfactory framework for the investigation
of cosmological models. Then, in Chapter 13, we describe the endeavours to determine
the values of the cosmological parameters and the problems which faced the observational
cosmologists. It turned out that many of these endeavours encountered the problems of the
evolution of the properties of the objects studied with cosmological epoch, and this is the
subject of Chapter 14. In Chapter 15, we trace the development of ideas about the formation
and evolution of galaxies and the large-scale structure of the Universe. These studies have
provided many of the tools necessary to ask physical questions about the very early stages
of the Universe, which is the subject of Chapter 16.
Many of the issues covered in this chapter on astrophysical cosmology up to the early
1970s are described in the book Cosmology and Controversy by Helge Kragh.1

12.1 Gamow and the Big Bang

During the 1930s, there were two reasons why the synthesis of the chemical elements in the
early stages of evolutionary world models was taken seriously. Firstly, the studies of Cecilia
Payne and Henry Norris Russell had shown that the abundances of the elements in stars
were remarkably uniform, suggesting a common origin for the elements (see Section 3.3).
The second consideration was that the interiors of stars seemed not to be hot enough for
the nucleosynthesis of the chemical elements to take place. The starting point for studies
of primordial nucleosynthesis was therefore to work out the equilibrium abundances of the
elements at some high temperature and assume that, if the density and temperature decreased
sufficiently rapidly, these abundances would remain ‘frozen’ as the Universe expanded and
cooled.
Detailed calculations were carried out in 1942 by Chandrasekhar and Louis Henrich, who
confirmed the expectation of equilibrium theory that, if the elements were in equilibrium at a
high temperature, their abundances would be inversely correlated with their binding energies
(Chandrasekhar and Henrich, 1942). The typical physical conditions under which this result
was found involved densities of ρ ≈ 109 kg m−3 and temperatures T ≈ 1010 K. There were,
however, several gross discrepancies between their predictions and the observed abundances
of the elements. The light elements, lithium, beryllium and boron, were predicted to be

319
320 12 Astrophysical cosmology

vastly overproduced relative to their cosmic abundances, and iron was predicted to be
underproduced, as were all the heavier elements with mass numbers greater than about 70.
This result was referred to as the ‘heavy-element catastrophe’. It was concluded that all
the chemical elements could not have been synthesised at a single density and temperature.
Chandrasekhar and Henrich suggested that some non-equilibrium process was required.
In contrast to this equilibrium picture, Georges Lemaı̂tre proposed in 1931 that the
Friedman models had evolved from an initial state which he termed a ‘primaeval atom’,
consisting of vast numbers of protons, electrons and α-particles packed together at nuclear
densities (Lemaı̂tre, 1931a). Such a huge ‘atom’ is necessarily unstable, and Lemaı̂tre
proposed that the process of disintegration would give rise to the formation of the chemical
elements. He also suggested that the energy released in the nuclear fission processes could
account for the high energies of cosmic rays.
Lemaı̂tre’s ideas provided the starting point for George Gamow’s attack on the problem of
the origin of the chemical elements.2 In 1946, he accepted the conclusion that the synthesis
of the chemical elements had to take place through non-equilibrium processes and he
postulated that the early phases of the Friedman models were the most likely location where
this might occur (Gamow, 1946). He extrapolated the Friedman models back to very early
cosmological epochs, at which the densities were high enough for the nucleosynthesis to
take place, and he found that the timescale of the Universe was then too short to establish an
equilibrium distribution of the elements. In his original proposal, the initial state consisted
of a sea of neutrons and subsequent β-decays and neutron capture processes would move
nuclei towards the locus of nuclear stability.
Ralph Alpher (b. 1921) joined Gamow as a graduate student in 1946 and was given the
task of working out the products of nucleosynthesis according to Gamow’s prescription.
Neutron capture cross-sections were available as a by-product of the nuclear physics pro-
grammes carried out during the Second World War, and these showed the encouraging result
that there is an inverse correlation between the relative abundances of the chemical elements
and their neutron capture cross-sections. In Alpher’s first calculations, a smooth curve was
fitted to the available data, and it was assumed that the initial conditions consisted of a sea of
free neutrons. As protons became available as a result of the β-decay of the neutrons, heav-
ier elements were synthesised by neutron capture. The nuclear reactions were assumed to
begin only after the temperature had fallen below that corresponding to the binding energy
of deuterium, kT = 0.1 MeV, and the Universe was assumed to be static. This theory was
published in 1948 by Alpher, Bethe and Gamow, Bethe’s name being added to complete
the αβγ pun, and they found reasonable agreement with the observed abundances of the
elements (Alpher, Bethe and Gamow, 1948). The importance of the paper was that it drew
attention to the necessity of a hot, dense phase in the early Universe if the chemical elements
were to be synthesised cosmologically.
In the same year, Alpher and Robert Herman (1914–1997) began improved calculations
of primordial nucleosynthesis, but now including the dynamics of the expansion of the
early Universe (Alpher and Herman, 1948). They realised that, at the necessary very high
temperatures at early epochs, the Universe was then radiation- rather than matter-dominated,
and they could then work out the subsequent thermodynamic history of the Universe. They
found that the temperature history of the thermal background radiation corresponded closely
12.1 Gamow and the Big Bang 321

Figure 12.1: The thermal history of the Universe containing many of the key features described by
Alpher and Herman (1948). This diagram was published by Wagoner, Fowler and Hoyle in 1967,
following an earlier version by Robert Dicke and his colleagues (Dicke et al., 1965).

to the adiabatic expansion of a photon gas, T ∝ R −1 , where R is the scale factor of the
Universe (Figure 12.1). From these results, they came to the far-reaching conclusion that the
cooled remnant of these hot early phases should be present in the Universe today, and they
estimated that the temperature of the thermal background should be about 5 K. This was the
first prediction that there should exist diffuse background radiation in the centimetre and
millimetre wavebands associated with what became known as the Big Bang theory3 of the
evolving Universe. Penzias and Wilson announced their discovery of the cosmic microwave
background radiation in 1965 (Penzias and Wilson, 1965).
There was, however, a major problem with this picture, which Gamow and his colleagues
were well aware of – there are no stable nuclei with mass numbers 5 and 8, and hence it was
difficult to understand how elements such as carbon, nitrogen and oxygen could have been
created by the addition of further protons, neutrons or α-particles to helium nuclei. Enrico
Fermi and Anthony Turkevich (1916–2002) carried out calculations of the evolution of the
322 12 Astrophysical cosmology

Figure 12.2: The evolution of the fraction (by number) of the light nuclei in a radiation-dominated
Universe, according to calculations by Fermi and Turkevich, published by Alpher and Herman in 1950
(Alpher and Herman, 1950). The models began with 100% of the material in the form of neutrons.

abundances of the light elements, including 28 nuclear reactions for elements up to mass
number 7 in a radiation-dominated expanding Universe, and their results were published
by Alpher and Herman in 1950 (Alpher and Herman, 1950). These calculations showed
that only about one part in 107 of the initial mass was converted into elements heavier than
helium, far less than the cosmic abundances of the heavy elements (Figure 12.2).
In 1950, another key link in the chain was provided by Chushiro Hayashi, who pointed
out that, in the early phases of the Universe at temperatures only ten times greater than
those at which nucleosynthesis takes place, the neutrons and protons were maintained in
thermodynamic equilibrium through the weak interactions

e+ + n ↔ p + ν̄e , ν e + n ↔ p + e− (12.1)

(Hayashi, 1950). Furthermore, at about the same temperature, electron–positron pair pro-
duction ensures a plentiful supply of positrons and electrons. The result was that, rather
 12.2 Steady state cosmology 323

than assume arbitrarily that the initial conditions consisted of a sea of neutrons, the equilib-
rium abundances of protons, neutrons, electrons and all the other constituents of the early
Universe could be calculated exactly. In 1953, Alpher, James Follin (b. 1919) and Herman
worked out the evolution of the proton–neutron ratio as the Universe expanded, and they
obtained answers remarkably similar to modern calculations (Alpher, Follin and Herman,
1953). They left, however,

for future study to re-examine the formation of the elements by thermonuclear reactions as a subsequent
part of the picture developed here.

They had come very close indeed to the modern picture of the thermal and nuclear evolution
of the early Universe, including the important result that, in the standard Big Bang picture,
about 25% of the primordial material by mass is converted into helium. Before this result
became an established feature of astrophysical cosmology, however, steady state cosmology
and the nucleosynthesis of the chemical elements in stars occupied centre stage.

12.2 Steady state cosmology

As Hermann Bondi4 has remarked, the period immediately after the Second World War
was one of considerable uncertainty concerning the physical basis of cosmology. Observa-
tionally, there was a timescale problem because Hubble’s estimate of the rate of expansion
of the Universe corresponded to a value of H0−1 of only 2 × 109 years. This is the maximum
age which any of the Friedman models can have, if the cosmological constant is set equal
to zero, and it was known that the age of the Earth was greater than this value. The only
solution within the standard picture was to introduce the cosmological constant so that the
timescale of the Universe could be stretched out as illustrated in Figure 6.4. This model did
not have much appeal for Bondi who felt that it was a contrived solution to suppose that the
cosmological constant had precisely the value that would result in a Universe which almost
reaches the Einstein stationary state at the present epoch, but not quite.
There were many new ideas in the air. Milne had developed his theory of kinematic
relativity in which he supposed that there are two different times, one associated with
dynamical phenomena and another with electromagnetic phenomena (Milne, 1948). Dirac
had been profoundly impressed by coincidences between the very large numbers in physics
and the properties of the Universe – for example, the square of the ratio of the strengths of
electromagnetic and gravitational forces is roughly equal to the numbers of protons in the
Universe. A consequence of his identification of these large numbers was his inference that
the gravitational constant might change with time (Dirac, 1937). Eddington was completing
his Fundamental Theory, in which he attempted to account for the values of the fundamental
constants of physics and in which the cosmological constant appeared as a fundamental
constant of nature (Eddington, 1946).5 What was remarkable about these new ideas was
that they were based upon concepts about what the underlying physics might be without any
strong physical motivation. Perhaps the most extreme example was the remark of Eddington,
in which he asserted
324 12 Astrophysical cosmology

Generalisations that can be reached epistemologically have a security which is denied to those that
can only be reached empirically.

These views can scarcely have appealed to a generation of physicists who were coming to
terms with the completely new concepts of relativistic quantum mechanics, which accounted
spectacularly for the experimentally determined properties of matter at the atomic level.
Herbert Dingle (1890–1978) was particularly outspoken in condemning Eddington, Milne,
Dirac and others for what he termed ‘Modern Aristotelianism’ (Dingle, 1937).
It was in this atmosphere that the idea of steady state cosmology was born. Fred Hoyle
has recorded the delightful story of the flash of inspiration which gave rise to the concept
of steady state cosmology. In his reminiscences of late 1946 or early 1947, he wrote:6

In a sense, the steady-state theory may be said to have begun on the night that Bondi, Gold and
I patronised one of the cinemas in Cambridge. The picture, if I remember rightly, was called The
Dead of Night. It was a sequence of four ghost stories, seemingly disconnected as told by the several
characters in the film, but with the interesting property that the end of the fourth story connected
unexpectedly with the beginning of the first, thereby setting-up the potential for a never-ending cycle.
When the three of us returned that evening to Bondi’s rooms in Trinity College, Gold suddenly said:
‘What if the Universe is like that?’

In an earlier version of his reminiscences,7 Hoyle remarked:

One tends to think of unchanging situations as being necessarily static. What the ghost-story film did
sharply for all three of us was to remove this wrong notion. One can have unchanging situations that
are dynamic, as for example a smoothly flowing river. The universe had to be dynamic, since Hubble’s
red-shift law proved it to be so . . . From this position, it did not take us long to see that there would
need to be a continuous creation of matter.

There the matter rested until 1948 when the three of them returned to the cosmological
problem. Bondi and Gold took a quite different approach from Hoyle to the exposition
of the theory. Bondi and Gold derived the theory by very general, almost philosophical,
arguments (Bondi and Gold, 1948), whereas Hoyle built his theory up from a field-theoretical
description of the process of continuous creation of matter, which he described by what he
called the C-field (Hoyle, 1948). The creation rate of matter amounted to only one particle
per cubic metre every 300 000 years. Both theories resulted in the same unique form for the
metric of space-time.
In the approach of Bondi and Gold, the cosmological principle was extended to what they
termed the perfect cosmological principle, according to which the Universe presents the
same large-scale picture to all observers at all times. Hence, Hubble’s constant becomes a
fundamental constant of nature, on a par with the charge on the electron or the gravitational
constant. This hypothesis also disposed of a problem, which concerned Bondi and Gold,
namely the general question of whether or not it was safe to assume that the laws of physics
are unchanging with time – by definition, all the laws were to be unchanging with time. They
showed that the perfect cosmological principle leads to a unique metric for the dynamics
of the Universe with zero spatial curvature8 and a scale factor which changes with time
as R(t) = R0 exp(t/t0 ), where t0 = H0−1 . Consequently, the Universe was infinite in age,
but the age of typical objects observed in the local Universe is only 13 H0−1 . Thus, our own
Galaxy had to be rather older than the typical object in the Universe, but that was not too
12.2 Steady state cosmology 325

unreasonable because, if Hubble’s constant were 500 km s−1 Mpc−1 , our own Galaxy would
be exceptionally large, much larger than other spiral galaxies. Bondi and Gold went on to
evaluate the counts of galaxies expected in the steady state model and found that they agreed
remarkably well with Hubble’s counts of galaxies.
The papers by Bondi and Gold and by Hoyle were published in 1948, and they immedi-
ately attracted considerable attention, both within the astronomical community and, through
Hoyle’s radio broadcasts, among the public at large. One consequence of the theory of the
greatest importance for astrophysics was that Hoyle set about attempting to find an alter-
native means of understanding the formation of the chemical elements, and this was one of
the motivations for his remarkable prediction of the carbon resonance (Hoyle, 1954) and
the subsequent fundamental paper on the processes of nucleosynthesis in stars by Burbidge
et al. (1957) (see Section 8.2). Bondi asked what evidence there was for any relics of the hot
early phases of the Universe and, with these new insights, the abundances of the chemical
elements disappeared as evidence.
The idea of the continuous creation of matter was a major stumbling block for many
physicists and astronomers, but as early as 1951, William McCrea had a deep insight into
the physics of Hoyle’s proposal (McCrea, 1951). McCrea realised that there was a quite
different interpretation of the metric of steady state cosmology, which evokes a resonance
with contemporary cosmology. To quote McCrea,

The single admission that the zero of absolute stress may be set elsewhere than is currently assumed on
somewhat arbitrary grounds permits all of Hoyle’s results to be derived within the system of General
Relativity theory. Also, this derivation gives the results an intellectual physical coherence.

McCrea wrote the physics of the steady state picture in terms of a negative energy
equation of state, p = −ρc2 , and recovered the three features of the theory, namely,
r the density of the Universe is a constant;
r the spatial geometry is flat;
r the scale factor varies as exp[H0 (t − t0 )].
It is intriguing that McCrea had realised that there is nothing intrinsically implausible about
a negative energy equation of state. In fact, these ideas had already been foreshadowed
by Lemaı̂tre in 1933. There is a close relation between the mathematics of a Universe
expanding exponentially under the influence of the cosmological constant  and the steady
state picture.9 Lemaı̂tre had suggested that the -term could be interpreted in terms of a
negative energy equation of state (Lemaı̂tre, 1933). In his words,

Everything happens as though the energy in vacuo would be different from zero.

In the 1950s, two important results were reported of central importance for cosmology.
The first concerned the value of Hubble’s constant. At the meeting of the International
Astronomical Union in Rome in 1952, Walter Baade announced that the distance of the
Andromeda Nebula (M31) had been underestimated by a factor of 2 (Baade, 1952). The
reason for this change was his discovery of the different types of stellar populations in galax-
ies (Section 5.7). The principal indicators used to determine the distance of M31 were the
Cepheid variables, and Baade discovered that there is a difference in the period–luminosity
326 12 Astrophysical cosmology

relations for Cepheids of Populations I and II. By using the same type of Cepheid variable
in our own Galaxy, in the Magellanic Clouds and in M31, the distance to M31 increased
by a factor of 2. This result also eliminated the problems that the globular clusters in M31
appeared to be intrinsically fainter than those in our Galaxy and that our Galaxy appeared
to be exceptionally large. From the point of view of cosmology, Hubble’s constant was
reduced to 250 km s−1 Mpc−1 and hence H0−1 increased to 4 × 109 years. In 1956, Huma-
son, Nicholas Mayall (1903–1993) and Sandage published their redshift–magnitude relation
for 474 galaxies and Hubble’s constant was revised downwards again to 180 km s−1 Mpc−1
and hence H0−1 increased to 5.6 × 109 years (Humason, Mayall and Sandage, 1956). These
revisions reduced, and quite possibly eliminated, the discrepancy between the age of the
Earth and the age of the Universe according to the standard Friedman models if the cosmo-
logical constant  is set equal to zero.

12.3 The counts of radio sources

The second piece of evidence resulted from the surveys of extragalactic radio sources
which began in the early 1950s. The central figure in this story was Martin Ryle, who
was leading the initiatives in radio astronomy at the Cavendish Laboratory in Cambridge.
Initially, he had been strongly wedded to the idea that most of the radio sources were radio
stars belonging to our own Galaxy, but, by early 1954, he had been converted to the idea
that most of the radio sources observed at high galactic latitudes are extragalactic. Ryle and
Anthony Hewish designed and constructed a large four-element interferometer to carry out
a new survey of the sky at 81.5 MHz, which, being an interferometer, would be sensitive to
small angular diameter sources. The second Cambridge (2C) survey of radio sources was
completed in 1954, and the first results were published in the following year (Shakeshaft
et al., 1955). Ryle and his colleagues found that the small-diameter radio sources were
uniformly distributed over the sky and that the numbers of sources increased enormously as
the survey extended to fainter and fainter flux densities. In any uniform Euclidean model,
the numbers of sources brighter than a given limiting flux density, S, are expected to follow
the relation N (≥ S) ∝ S −3/2 (see Section A5.1). In contrast, Ryle found a huge excess of
faint radio sources, the slope of the source counts between 20 and 60 Jy being described by
N (≥ S) ∝ S −3 (Figure 12.3). He concluded that the only reasonable interpretation of these
data was that the sources were extragalactic, that they were objects similar in luminosity to
the radio galaxy Cygnus A and that there was a much greater comoving number density of
radio sources at large distances than there are nearby. As Ryle expressed it in his Halley
Lecture in Oxford in 1955 (Ryle, 1955),

This is a most remarkable and important result, but if we accept the conclusion that most of the radio
stars are external to the Galaxy, and this conclusion seems hard to avoid, then there seems no way in
which the observations can be explained in terms of a Steady-State theory.

These remarkable conclusions came as a surprise to the astronomical community. There

was enthusiasm, and also some scepticism, that such profound conclusions could be drawn
12.3 The counts of radio sources 327

Figure 12.3: The integral number counts of radio sources from the 2C survey of radio sources; N is
the number of radio sources brighter than flux density I , the units of I being 10−25 W m−2 sr−1 . The
dashed line shows the ‘Euclidean’ number counts of radio sources. The observations show a very
large excess of faint radio sources relative to the expectations of the Euclidean world model (Ryle,
1955).

from the counts of radio sources, particularly when their physical nature was not understood
and only the brightest 20 or so objects had been associated with relatively nearby galaxies.
The Sydney group, led by Bernard Mills (b. 1920), were carrying out similar radio surveys
of the southern sky at about the same time with the large cross-shaped radio telescope known
as the Mills Cross, and they found that the source counts could be represented by the relation
N (≥ S) ∝ S −1.65 , which they argued was not significantly different from the expectation of
uniform world models. In 1957 Mills and Bruce Slee stated (Mills and Slee, 1957):

We therefore conclude that discrepancies, in the main, reflect errors in the Cambridge catalogue, and
accordingly deductions of cosmological interest derived from its analysis are without foundation. An
analysis of our results shows that there is no clear evidence for any effect of cosmological importance
in the source counts.
328 12 Astrophysical cosmology

The problem with the Cambridge number counts was that they extended to surface den-
sities of radio sources such that the flux densities of the faintest sources were overestimated
because of the presence of faint sources in the beam of the telescope, a phenomenon known
as confusion. Peter Scheuer, who was Martin Ryle’s research student from 1951 to 1954,
devised a statistical procedure for deriving the number counts of sources from the survey
records themselves without the need to identify individual sources (Scheuer, 1957). The
technique which he discovered, which he referred to as the P(D) technique and which has
since been adopted in many other astronomical contexts, showed that the slope of the source
counts was actually −1.8. Ironically, this result, which is exactly the correct answer, was
not trusted, partly because the mathematical techniques used by Scheuer were somewhat
forbidding and also because his result differed from the prejudices of both Ryle and Mills.
The dispute reached its climax at the Paris Symposium on Radio Astronomy in 1958, but
the conflicting positions were not resolved (Bracewell, 1959). The strong feelings aroused
by the contrasting views of the proponents of the evolutionary and steady state models are
vividly described in Kragh’s book Cosmology and Controversy (see endnote 1).
The resolution of the controversy only came with the construction of the next generation
of radio telescopes, which had higher angular resolution and hence were less sensitive to
the effects of source confusion. In the next Cambridge catalogues, the 3C catalogue (Edge
et al., 1959) and, particularly, the revised 3C catalogue (Bennett, 1962), much more care was
taken to eliminate the effects of confusion, and the accuracy of the radio source positions
improved so that identifications could be made with fainter galaxies. These showed that
Ryle’s conclusions of 1955 were basically correct, but that the magnitude of the excess had
been considerably overestimated. More radio sources were identified with distant galaxies,
and the optical identification programmes led to the discovery of quasars in the early 1960s
(see Section 11.2). The lessons of the 1950s had been taken to heart, and detailed studies were
made of the effects of confusion and the absence of sources of large angular size from the
catalogues. The radio source counts derived from the 4C catalogues (Pilkington and Scott,
1965; Gower, Scott and Wills, 1967) showed an excess over the expectations of Euclidean
world models (Gower, 1966). In fact, the discrepancies with the uniform Friedman and
steady state models were much greater than this simple comparison suggested because the
predicted radio source counts converge rapidly as soon as the source populations extended
to significant redshifts (Longair, 1971; Scheuer, 1975). By the mid 1960s, the evidence was
compelling that there was indeed an excess of sources at large redshifts and this was at
variance with the expectations of the steady state theory.10

12.4 The helium problem

In 1964, while I was completing my first year of research at Cambridge, Fred Hoyle gave a
course of lectures on the problems of extragalactic research. He would arrive with, at best,
a scrap of paper with some notes and expound an area of current research. One week, the
topic was the problem of the cosmic helium abundance. Helium is one of the more difficult
elements to observe astronomically because of its high excitation potential, and so it can
only be observed in very hot stars. Donald Osterbrock and John Rogerson had shown in
 12.5 Cosmic microwave background radiation 329

1961 that the abundance of helium seemed to be remarkably uniform wherever it could
be observed and that it corresponded to about 25% by mass (Osterbrock and Rogerson,
1961). A further important observation was reported by O’Dell (b. 1937) in 1963 of the
helium abundance in a planetary nebula in the old globular cluster M15 (O’Dell, Peimbert
and Kinman, 1964). Despite the fact that the heavy elements were deficient relative to their
cosmic abudances, the helium abundance was still about 25%.
The evidence on the cosmic helium abundance was reviewed by Hoyle, and he then
described the work of Gamow, Alpher, Herman and Follin concerning the problems of
synthesising the heavy elements in the early phases of the Big Bang (see Section 12.1).
Although helium is synthesised in the central regions of stars during their long phases of
evolution on the main sequence, it is most unlikely that this process could have created as
much helium as 25% by mass of the baryonic matter in the Universe. Most of the luminosity
of galaxies is associated with the burning of hydrogen into helium in main-sequence stars
and so, if the luminosity of our Galaxy had remained more of less the same throughout its
lifetime, an upper limit of about 1% of the mass of the Galaxy could have been converted
into helium. Furthermore, the stars move off the main sequence when only about 10%
of their mass has been converted into helium, and then the helium is burned into heavier
elements. It was difficult to understand why there should be a universal abundance of about
25% by mass if the helium was created in stars.
By 1964, when Hoyle was delivering his lecture, it was possible to carry out primordial
nucleosynthesis calculations more accurately. At that time, Roger Tayler (1929–1997) had
just returned to Cambridge and was present in the audience. Hoyle and Tayler realised that
they could undertake much more precise calculations and, over the following week, they and
Tayler’s research student, John Faulkner (b. 1937), worked out the details of the formation of
helium in the early phases of the Big Bang. The audience had the privilege of being present
as a key piece of modern astrophysics was created in real time in a graduate lecture course.
Hoyle and Tayler obtained the result that about 25% helium by mass is synthesised in the
Big Bang, in remarkable agreement with observation and essentially independent of the
overall baryonic matter density in the Universe. The reason for the constancy of the cosmic
helium abundance is that it is primarily determined by the thermodynamics of the early
Universe, rather than by the microphysics involved in the nuclear reactions (see Section
A12.1). Their paper was published in Nature in 1964 (Hoyle and Tayler, 1964).

12.5 The discovery of the cosmic microwave background radiation

One consequence of the Big Bang model which Hoyle and Tayler did not mention explicitly
in their paper was that the cooled remnant of the thermal radiation present during the very hot
early phases should be detectable at centimetre and millimetre wavelengths.11 Alpher and
Herman’s prediction had been more or less forgotten when Gamow’s theory of primordial
nucleosynthesis had failed to account for the creation of the chemical elements. The idea
of searching for thermal radiation from the Big Bang was revived in the early 1960s by
Yakov Zeldovich and his colleagues in Moscow and by Robert Dicke (1916–1997) and his
colleagues in Princeton.12
330 12 Astrophysical cosmology

In 1964, Andrei Doroshkevich (b. 1937) and Igor Novikov reanalysed the physics of
the Big Bang model and showed that the thermal background radiation with a Planck
spectrum at radiation temperature between about 1 and 10 K should be present in the
Universe at the present day (Doroshkevich and Novikov, 1964). They pointed out that
this prediction provided a key test of the Big Bang scenario. They also noted that useful
limits to the background radiation temperature could be obtained from the measurements of
Edward Ohm (b. 1926) in 1961 of the radio background emission at centimetre wavelengths
published in the reports of the Bell Telephone Laboratories (Ohm, 1961). In fact, Ohm had
discovered an excess noise temperature of 3.3 K in his experiments, but believed that this
figure was within the measurement errors of the total signal detected by his antenna and
receiver system, which was 22.3 K. There were, however, earlier indications of a diffuse
extragalactic component of the background radiation. In 1955, Émile Le Roux (b. c1930)
detected a uniform background of 3 ± 2 K at a wavelength of 33 cm at the Nançay Radio
Observatory, somewhat greater than expected from the population of discrete radio sources
(Dennise, Le Roux and Steinberg, 1957). In 1957, Tigran Shmaonov (b. 1930) published a
measurement of the radiation temperature of the background radiation at 3.2 cm of 4 ± 3 K,
but it had been forgotten, and he only brought the result to the attention of Igor Novikov13
much later, in 1983 (Shmoanov, 1957).
The very next year, in 1965, the microwave background radiation was discovered by
Arnold Penzias and Robert Wilson, more or less by accident. They had joined the Bell
Telephone Laboratories in the early 1960s with the intention of using the same 20-foot
horn reflector used by Ohm, which had been built to test telecommunication with the Echo
satellite, for radio astronomical observations. Penzias and Wilson had the responsibility of
calibrating the antenna for use at these frequencies, for which they had built a 7.35 cm cooled
maser receiver. The understanding was that the telescope could be used for astronomical
observations for some fraction of the observing time. Wherever they pointed the telescope
on the sky, they found an excess antenna temperature, which could not be accounted for by
noise sources in the telescope or receiver system. A list of contributions to the total detected
signal is given in Table 12.1. Having carefully calibrated all parts of the telescope and
receiver system, they found that there remained about 3.5 ± 1 K excess noise contribution
(Penzias and Wilson, 1965).
At almost exactly the same time, Robert Dicke’s group in Princeton were preparing
exactly the same type of experiment to detect the cooled remnant of the Big Bang. Discus-
sions with the Princeton group ensued, and it became apparent that Penzias and Wilson had
discovered the diffuse cosmic microwave background radiation, exactly what the Princeton
physicists were searching for. Within a few months, the Princeton group had measured a
background temperature of 3.0 ± 0.5 K at a wavelength of 3.2 cm, confirming the black-
body nature of the background in the Raleigh–Jeans region of the spectrum (Roll and
Wilkinson, 1966).
Remarkably, there was earlier evidence for a diffuse component of millimetre radiation
with this radiation temperature from the study of several faint interstellar absorption lines
associated with the molecules CH, CH+ and CN. In the case of CN, for example, absorption
was observed from the first rotationally excited state of the molecule as well as the ground
state. In 1941, Andrew McKellar (1910–1960) had shown that the necessary excitation
 12.6 The helium problem revisited 331

Table 12.1. Contributions to the total measured

radio signal in Penzias and Wilson’s experiments
at 4.08 GHz (7.35 cm)

Noise signal
Signal T /K

Total zenith noise temperature 6.7 ± 0.3

Atmospheric emission 2.3 ± 0.3
Ohmic losses 0.8 ± 0.4
Backlobe response ≤ 0.1
Cosmic background radiation 3.5 ± 1.0

temperature to populate the first excited state was 2.3 K, although the origin of the excitation
was then unknown (McKellar, 1941).
Many measurements of the background radiation were made in the following years.
At millimetre wavelengths, the observations were very difficult because of atmospheric
absorption, and a number of balloon observations were made which were broadly consistent
with a black-body radiation spectrum at a temperature of about 2.7 K. The best way of
avoiding the problems of atmospheric absorption was to carry out the observations from
space, and this was achieved with the launch of the Cosmic Background Explorer (COBE) in
November 1990. This first results of this experiment showed that the spectrum of the cosmic
background radiation is of black-body form, the radiation temperature being 2.725 ± 0.01 K
(Mather et al., 1990). The final results, reported in 1996, show that deviations from a
perfect black-body spectrum amount to less than 0.03% of the maximum intensity over the
waveband 2.5 to 0.5 mm, the most perfect naturally occurring black-body spectrum (Fixsen
et al., 1996) (Figure 12.4). The radiation temperature was 2.728 ± 0.004 K (95% confidence
level).

12.6 The helium problem revisited

The appearance of the paper by Hoyle and Tayler (1964) and the discovery of the cosmic
background radiation stimulated a number of detailed studies of the synthesis of the light
elements during the period of primordial nucleosynthesis when the Universe was a few
minutes old. After Dicke’s group published their results confirming the discovery of the
background radiation, James Peebles published two papers exploring the constraints on
cosmological models which could be derived from the observed abundances of deuterium
and helium in the Universe at the present time (Peebles, 1966a,b). The standard radiation-
dominated Big Bang picture makes quite definite predictions about the abundance of the
light elements, but it depends upon the dynamics of the Universe through the epochs when
the neutrinos decouple from matter and the synthesis of the light elements begins. Peebles
showed that if the early expansion were speeded up by a factor of 10 to 100 as compared with
the standard picture, the neutrinos would decouple earlier and greater amounts of helium
332 12 Astrophysical cosmology

MJy sr -1

Frequency (cm−1)

Figure 12.4: The final spectrum of the cosmic microwave background radiation as measured by the
COBE satellite (Fixsen et al., 1996). The units of the abscissa are inverse centimetres, so that ten units
corresponds to 1 mm and five correspond to 2 mm. The experimental uncertainties are less than the
thickness of the line.

would be produced. This might happen if the gravitational constant had changed with time,
as in the Brans–Dicke cosmology (Brans and Dicke, 1961), or if the early expansion of the
Universe had been anisotropic. Similar calculations were carried out independently by the
Moscow group (Doroshkevich et al., 1971).
In 1967, Robert Wagoner (b. 1938), William Fowler and Fred Hoyle repeated the anal-
ysis carried out by Hoyle and Tayler, but now using all the available cross-sections for
many more nuclear interactions between light nuclei, and with the knowledge that the
cosmic microwave background radiation had a temperature of about 2.7 K (Wagoner,
Fowler and Hoyle, 1967).14 Fowler’s deep understanding of nuclear physics contributed
greatly to all aspects of these computations15 and enabled a very detailed network of the
many nuclear interactions involved in the synthesis of the light elements to be created
(Figure 12.5). These calculations confirmed that about 25% of helium by mass is cre-
ated by primordial nucleosynthesis and that this figure is remarkably independent of the
present density of baryonic matter in the Universe. The reason for this is that the amount
of helium produced depends primarily upon the neutron–proton ratio when the neutri-
nos decouple from the nuclear reactions which maintain the abundances in equilibrium
(see Section A12.1). Of particular importance was their demonstration that the abundances
of other products of nucleosynthesis, deuterium, 3 He and 7 Li are sensitive to the mean
baryon density in the Universe (Figure 12.6). The importance of observations of these
elements is that they are very difficult to synthesise in stars because they have relatively
small nuclear binding energies – deuterium and 3 He are destroyed rather than created in
stars.
12.6 The helium problem revisited 333

Figure 12.5: The network of reactions used by Wagoner in his determination of the primordial abun-
dances of the elements (Wagoner, 1973). This network is an enhanced version of that used by Wagoner,
Fowler and Hoyle in their paper of 1967 (Wagoner et al., 1967).

Interstellar absorption lines of deuterium were discovered in the ultraviolet region of

the spectrum by John Rogerson (b. 1922) and Donald York (b. 1944) in 1973 from obser-
vations made by the Copernicus ultraviolet satellite (Figure 7.12) (Rogerson and York,
1973). An interstellar deuterium abundance of 1.5 × 10−5 by mass relative to hydrogen
was found. Subsequent observations showed that the same deuterium abundance is found
along the line of sight to other stars which could be observed by the Copernicus satel-
lite (Vidal-Madjar et al., 1977). These observations enabled an upper limit to be placed
upon the mean baryonic density of the Universe of 1.5 × 10−28 kg m−3 , corresponding
to B h 2 ≤ 10−2 . If the mean baryonic density of the Universe were any greater, deu-
terium would be underproduced by primordial nucleosynthesis and no other way of creat-
ing deuterium astrophysically was known. This important upper limit to the baryon density
in the Universe was at least an order of magnitude less than the critical cosmological
density.
The story of the 7 Li abundance was more complicated. Observations of this isotope in
Population I stars had shown that there is an upper limit to its abundance of about 1 part in
109 relative to hydrogen by mass. Stars like the Sun have much lower 7 Li abundances than
this value because lithium is destroyed when it is convected or diffuses into regions with
temperatures greater than about 2 × 106 K. It seemed natural to suppose that the primordial
abundance of 7 Li was about 10−9 . In 1982, François Spite (b. 1930) and Monique Spite
334 12 Astrophysical cosmology

h0(g cm −3 )
10−6 10−5 10−4 10−3 10−2 10−1

4
10−1 He

10−2
2H

10−3

10−4

10−5
mass fraction

3He

10−6

10−7

10−8
7Li A > 12
10−9

10−10
6Li
11B
10−11

10−32 10−31 10−30 10−29 10−28

ρb (T = 2.7 K)(g cm−3)

Figure 12.6: The synthesis of the light elements in Big Bang according to the calculations of Wagoner,
Fowler and Hoyle, as revised by Wagoner in 1973 (Wagoner et al., 1967; Wagoner, 1973). These
computations demonstrated the sensitivity of the abundances of deuterium, 3 He and 7 Li to the present
baryon density of the Universe.

(b. 1939) observed a remarkably constant value for the 7 Li abundance of 10−10 in old
Population II stars (Spite and Spite, 1982). The abundance did not vary with the surface
temperatures of the stars, suggesting that the 7 Li was primordial. It was, however, intriguing
that an abundance of 7 Li in the range 10−9 to 10−10 would be entirely consistent with a
cosmological origin (Figure 12.6).
It was appreciated by Hoyle, Tayler, Fowler, Wagoner and Peebles, and also by the
Moscow group working with Zeldovich, that the synthesis of the heavy elements provided an
important diagnostic tool for the dynamics of the Universe at the epoch of nucleosynthesis.
If the Universe expanded too rapidly, the neutron–proton ratio would freeze out at a higher
temperature and helium would be overproduced. This result enabled important constraints
to be placed upon any variations of the gravitational constant with time as well as restricting
the number of permissible neutrino species to three, a result subsequently confirmed by
experiments with the Large Electron–Positron Collider at CERN from studies of the energy
widths of the decay products of the Z0 bosons.16
Thus, by the end of the 1960s, the overwhelming balance of opinion was that there
was convincing evidence for two relics of the hot early phases of the Universe, the cosmic
microwave background radiation and the cosmic abundances of the light elements. Since that
Notes to Chapter 12 335

time the Big Bang picture has become the preferred framework for astrophysical cosmology.
But, equally important, rather than being the preserve of speculation, geometrical and
astrophysical cosmology acquired a quite new status as the province of genuine astrophysical
enquiry. It is not coincidental that from this period onwards particle physicists, and physicists
in general, had to embrace the early Universe as an integral part of physics. The astronomers
had provided the particle physicists with the ultimate particle physics laboratory.

Notes to Chapter 12

1 H. Kragh, Cosmology and Controversy: The Historical Development of Two Theories of the Uni-
verse (Princeton: Princeton University Press, 1996).
2 The history of Gamow’s work on the Big Bang theory is described by R. A. Alpher and R. C.
Herman in Early work on ‘Big Bang’ cosmology and the cosmic black body radiation, Modern
Cosmology in Retrospect, eds B. Bertotti, R. Balbinot, S. Bergia and A. Messina (Cambridge:
Cambridge University Press, 1990), pp. 129–157, and by R. V. Wagoner, Deciphering the nuclear
ashes of the early Universe: a personal perspective, in Modern Cosmology in Retrospect, pp. 159–
185.
3 There had been references to the expansion of the standard Friedman models as an explosion
or ‘bang’, but the term Big Bang entered the cosmological literature with some force following
a series of BBC radio broadcasts by Fred Hoyle in the spring of 1949. His use of the term has
been interpreted as a pejorative remark, in contrast to his enthusiasm for the steady state theory
that he, Bondi and Gold had developed over the preceding years (see Section 12.2). According to
Simon Mitton, Hoyle did not intend the term to be derogatory (S. Mitton, Conflict in the Cosmos:
Fred Hoyle’s Life in Science (Washington D.C.: Joseph Henry Press, 2005)). The lectures were
subsequently published, essentially unmodified, by Hoyle in his book The Nature of the Universe
(Oxford: Basil Blackwell, 1950). The term is not a particularly happy one, since it conjures up a
somewhat misleading impression of how the standard isotropic world models are constructed, but
it is now firmly embedded in the literature.
4 Bondi’s revealing thoughts about observational and theoretical cosmology during the pre- and post-
War years are contained in H. Bondi, The cosmological scene 1945–1952, in Modern Cosmology
in Retrospect, pp. 189–196.
5 Eddington (1946) was published posthumously under the editorship of E. Whittaker.
6 This reminiscence is recounted by Hoyle in An assessment of the evidence against the steady-state
theory, in Modern Cosmology in Retrospect, pp. 221–231.
7 This reminiscence is contained in F. Hoyle, Steady-State Cosmology Re-visited (Cardiff: Cardiff
University Press, 1980).
8 I have shown how these results may be simply derived from the Robertson–Walker metric in Section
19.5 of Malcolm Longair, Theoretical Concepts in Physics (Cambridge: Cambridge University
Press, 2003).
9 There is a delightful sequel to this story. On the occasion of his 80th birthday in 1995, I invited Hoyle
to lecture to the Cavendish Physical Society. He was delighted to accept this invitation because he
had given his first lecture on steady state cosmology to the Cavendish Physical Society in 1948.
Hoyle remarked wryly that his only mistake had been to call his creation field C rather than ψ. The
exponential expansion of the early Universe according to the inflation picture involves a scalar
field ψ which performs exactly the same function as Hoyle’s C-field or Lemaı̂tre’s cosmological
constant .
10 It was a great sadness that relations between Hoyle and Ryle were so soured by the controversy
over the radio source counts. In 1965, Peter Scheuer and I attempted a reconciliation between them
when the four of us got together in Hoyle’s newly founded Institute of Theoretical Astronomy in
336 12 Astrophysical cosmology

Cambridge to try to understand their different positions. Sadly, there was no longer any common
ground, and each simply repeated their entrenched views. It was one of the saddest events of my
career.
11 According to Roger Tayler (personal communication), he had included this result in his draft of
their paper, but it did not appear in the published version.
12 Some appreciation of the scope of Zeldovich’s contributions to astrophysics and cosmology can be
found in Selected Works of Yakob Borisovich Zeldovich, vol. 2, eds J. P. Ostriker, G. I. Barenblatt
and R. A. Sunyaev (Princeton: Princeton University Press, 1993). An account of Dicke’s role in
the revival of the observational study of Gamow’s picture of the Big Bang can be found in P. J. E.
Peebles, Principles of Physical Cosmology (Princeton: Princeton University Press, 1993).
13 Novikov describes the remarkable early history of attempts to detect the cosmic microwave back-
ground radiation in his paper Discovery of the CMB, Sakharov oscillations and polarization,
in Historical Development of Modern Cosmology, ASP Conference Series, vol. 252, eds V. J.
Martínez, V. Trimble, and M. J. Pons-Bordeía (San Francisco: ASP, 2001), pp. 43–53. This also
includes the story of Shmaonov’s measurements. See also, Kragh, Cosmology and Controversy,
p. 343.
14 When Hoyle first presented these results in Cambridge in 1967, many of us were surprised that it
seemed as though he had been converted to the Big Bang picture. As in the paper itself, however,
equal weight was given to the idea that these computations could also be applied at very much higher
baryonic densities to very massive stars which collapsed and ‘bounced’. The nucleosynthesis of
the expansion phase was exactly the same as a Universe of very high baryonic mass density. The
densities were so high that heavy elements could be synthesised in these stars. The subsequent
paper by Wagoner concentrated upon the primordial synthesis of the light elements (Wagoner,
1973).
15 Wagoner’s review of the history of these computations provides a vivid picture of how Hoyle,
Fowler and he interacted during these exciting years. See Deciphering the nuclear ashes of the
early Universe: a personal perspective, Bertotti et al., eds, Modern Cosmology in Retrospect, pp.
159–185.
16 See, for example, the Opal Collaboration, A combined analysis of the hadronic and leptonic decays
of the Z 0 , Physics Letters, B240, 1990, 497–512.

A12 Explanatory supplement to Chapter 12

A12.1 The primordial abundances of the light elements
The reason for the remarkable stability of the prediction of 25% helium by mass in the
standard Big Bang picture can be understood from the following physical arguments.
Consider a particle of mass m at very high temperatures such that its total energy is
much greater than its rest-mass energy, kT  mc2 . If the timescales of the interactions
which maintain this species in thermal equilibrium with all the other species present at
temperature T are shorter than the age of the Universe at that epoch, the equilibrium
number densities of the particle and its antiparticle are given by the standard expression
from statistical mechanics,

4πg ∞ p 2 d p
N=N= 3 , (A12.1)
h 0 e E/kT ± 1
where g is the statistical weight of the particle, p is its momentum and the ± sign depends
upon whether the particles are fermions (+) or bosons (−). Now, photons are massless
Explanatory supplement to Chapter 12 337

bosons for which g = 2, nucleons, antinucleons, electrons and positrons are fermions with
g = 2 and the electron, muon and tau neutrinos are fermions with helicity for which g = 1.
The equilibrium number densities, N , and energy densities, , can be found from this
expression. For (i) photons, (ii) nucleons and electrons and (iii) neutrinos, respectively,
these are given by:
 
2π kT 3 −3
(i) g = 2, N = 0.244 m , = aT 4 ; (A12.2)
hc
 
2π kT 3 −3
(ii) g = 2, N = 0.183 m , = 78 aT 4 ; (A12.3)
hc
 
2πkT 3 −3
(iii) g = 1, N = 0.091 m , = 7
16
aT 4 . (A12.4)
hc

To find the total energy density, we add all the equilibrium energy densities together, to
obtain

total energy density = = χ (T ) aT 4 . (A12.5)

When the particles become non-relativistic, kT mc2 , and the abundances of the different
species are still maintained by interactions between the particles, the non-relativistic limit
of the integral gives an equilibrium number density,
 3/2  
mkT mc2
N =g exp − . (A12.6)
h2 kT

Thus, once the particles become non-relativistic, they no longer contribute to the inertial
mass density which determines the rate of expansion of the Universe.
Now consider the abundances of protons and neutrons in the early Universe. At redshifts
less than 1012 , the neutrons and protons are non-relativistic, kT mc2 , and their abun-
dances are maintained at their thermal equilibrium values by the electron–neutrino weak
interactions

e+ + n → p + ν̄e ; νe + n → p + e− . (A12.7)

For the neutrons and protons, the values of g are the same, and so the relative abundances
of neutrons to protons is given by
     
n (m n − m p )c2 mc2
= exp − = exp − , (A12.8)
p kT kT

where mc2 is the mass difference between the neutron and the proton.
This abundance ratio freezes out when the neutrino interactions can no longer maintain
the equilibrium abundances of neutrons and protons. The condition for ‘freezing out’ is
that the timescale of the weak interactions becomes greater than the age of the Universe.
338 12 Astrophysical cosmology

The variation of the energy density of radiation and temperature during the early radiation-
dominated phases of the Universe are given by
3c2 −2
= χaT 4 = t , (A12.9a)
32π G
 
1/4
3c2
T = t −1/2 = 1010 t −1/2 K, (A12.9b)
32π Gχa
where it is assumed that the number of neutrino species Nν = 3 and so χ = 43/8. The time,
t, is measured in seconds. Note that equations (A12.9 a,b) illustrate how the early expansion
rate depends upon the gravitational constant G and the number of neutrino species through
χ.
The processes which prevent the neutrinos escaping freely are:

e− + e+ → νe + ν̄e ; e± + νe → e± + νe ; e± + ν̄e → e± + ν̄e . (A12.10)

Straightforward calculations1 show that the timescales for the expansion of the Universe
and the decoupling of the neutrinos are the same when the Universe was almost precisely
one second old at a temperature of 1010 K At that time, t = 1 s, the neutron fraction, as
determined by equation (A12.8), was
 
n
= 0.21. (A12.11)
n+p
The neutron fraction decreases very slowly after this time. Detailed calculations show
that after 300 s the neutron fraction has fallen to 0.123. It is at this epoch that the bulk of
the formation of the light elements takes place. Almost all the neutrons are combined with
protons to form 4 He nuclei, so that, for every pair of neutrons, a helium nucleus is formed.
The reactions involved are

p + n → 3 He + γ ; n + D → 3H + γ ; p + 3 H → 4 He + γ ; (A12.12)
n + He → He + γ ;
3 4
d + d → He + γ ;
4 3
He + He → He + 2p.
3 4
(A12.13)

Most of the nucleosynthesis takes place at a temperature less than about 1.2 × 109 K
since, at greater temperatures, the deuterons are destroyed by the γ -rays of the background
radiation. The binding energy of deuterium is E B = 2.23 MeV and so this energy is equal
to kT at T = 2.6 × 1010 K. However, the photons far outnumber the nucleons, and it is only
when the temperature of the expanding gas has decreased to about 26 times less than this
temperature that the number of dissociating photons is less than the number of nucleons.
Although the neutrons begin to decay spontaneously by this time, the bulk of them survive
and so, according to the above calculation, the predicted helium to hydrogen mass ratio is
just twice the neutron fraction:
4 
He
≈ 0.25. (A12.14)
H
The predicted abundance of deuterium is a strong function of the present density of
the Universe, in contrast to the constant abundance of helium. The reasons for this can
Note to Section A12 339

be understood as follows. The helium abundance results from the equilibrium ratio of
protons to neutrons as the Universe cools down; that is, it is primarily determined by the
thermodynamics of the expanding radiation-dominated Universe. On the other hand, the
abundance of deuterium depends upon the number density of nucleons. If the Universe has
a high baryon number density, then essentially all the deuterons are converted into helium,
whereas if the Universe is of low density, not all the deuterium is converted into 4 He. The
same argument applies to 3 He. Thus, the deuterium and 3 He abundances set an upper limit
to the present baryon density of the Universe.

Note to Section A12

1 I have given a simple version of this calculation in Malcolm Longair, Galaxy Formation (Berlin:
Springer-Verlag, 1998), Sections 10.2 and 10.3.
13 The determination of
cosmological parameters

13.1 Sandage and the values of H0 and q0

In 1952, Walter Baade announced that the value of Hubble’s constant, H0 , had been over-
estimated because the distance to the Andromeda Nebula, M31, adopted by Hubble was
about a factor of 2 too small (Baade, 1952). The cause of the discrepancy was that there
is a difference in the period–luminosity relations for Cepheid variables of Populations I
and II (see Section 12.2). By using the same type of Cepheid variable in our own Galaxy,
in the Magellanic Clouds and in M31, the distance to M31 increased by a factor of 2.
Consequently, Hubble’s constant was reduced to 250 km s−1 Mpc−1 and H0−1 increased to
4 × 109 years.
In 1956, Humason, Mayall and Sandage showed that the expected redshift–magnitude
relation, m = 5 log10 z + constant, is observed for galaxies selected at random, but there
is a large scatter about the mean relation because of the breadth of the luminosity function
of galaxies (Humason et al., 1956). It had been known since Hubble’s pioneering studies of
the 1930s, however, that the brightest galaxies in clusters of galaxies follow a very much
tighter relation1 which follows precisely Hubble’s law v = H0 r (Figure 13.1). Thus, in
order to estimate the value of H0 , it was only necessary to calibrate the observed relation by
measuring the distance of the nearest rich cluster of galaxies, the Virgo cluster of galaxies,
by techniques independent of its redshift. Humason, Mayall and Sandage estimated the
distance of the giant spiral galaxy NGC 4321, one of the brightest galaxies in the Virgo
cluster, assuming that the brightest stars and nebulae in that galaxy were the same as
those in M31. Hubble’s constant was revised downwards again to 180 km s−1 Mpc−1 . In
1958, Sandage’s best estimate of H0 was reduced yet again from 180 to 75 km s−1 Mpc−1
(Sandage, 1958). The principal reason for this further downward revision was that what had
been thought to be the brightest stars in some of the most distant galaxies studied turned
out to be regions of ionised hydrogen and star clusters.
Immediately after the Second World War, the prime instrument for cosmological research
was the Palomar 200-inch telescope, which was commissioned in 1948. In 1961, Allan
Sandage published an influential paper entitled ‘The ability of the 200-inch telescope to
discriminate between selected world models’, in which different approaches to the determi-
nation of cosmological parameters with this telescope were discussed critically (Sandage,
1961b). The observed properties of galaxies at large redshifts depend upon the geometry

340
13.1 Sandage and the values of H0 and q0 341

Figure 13.1: The redshift–magnitude relation in the visual (V) waveband for the brightest galaxies in
clusters presented by Sandage in 1968 (Sandage, 1968). The straight line shows the expected relation
if the galaxies all have the same intrinsic luminosity, m = 5 log10 z + constant. The sparsity of points
at redshifts greater than 0.3 illustrates the difficulty of finding clusters of galaxies at large redshifts
that would be suitable for cosmological tests.

of the world model and upon its kinematics between the epochs of emission and reception
of the radiation. To repeat the list of parameters, these are:
r Hubble’s constant, H0 = Ṙ/R, the present rate of expansion of the Universe;
r the deceleration parameter, q0 = − R/ ¨ Ṙ 2 , the present deceleration of the Universe;
r the curvature of space, κ = R0 ; −2
r the mean density of matter in the Universe at the present epoch, ρ0 , and its value relative
to the critical density, ρcrit = 3H02 /8π G;
r the present age of the Universe, T0 ;
r the cosmological constant, λ.

As pointed out in Section 6.8, these are not independent, provided the Friedman models are
a correct description of the large-scale dynamics of the Universe. Thus, for the Friedman
342 13 The determination of cosmological parameters

models,
 
−2 (0 − 1) + 13 λ/H02 0 1 λ
κ=R = ; q0 = − , (13.1)
(c/H0 )2 2 3 H02

where 0 = ρ0 /ρcrit is the density parameter. If λ = 0, there is a simple one-to-one relation

between the geometry of the world models, their densities and dynamics, q0 = 0 /2 and
κ = R−2 = (0 − 1)/(c/H0 )2 . Thus, if the Friedman models with non-zero cosmological
constant are adopted, three independent parameters need to be determined, for example
H0 , q0 and 0 . If the cosmological constant is zero, only two parameters need be deter-
mined, say H0 and q0 . The steady state theory was uniquely defined by the single parameter
H0 . Ideally, the three parameters should be determined independently, and then equations
(13.1) provide a test of the general theory of relatively on the largest scales accessible
to us.
Sandage fully recognised the magnitude of the task involved. He discussed the use of
the redshift–magnitude relation for giant elliptical galaxies, the angular diameter–redshift
test, the number counts of galaxies and the ages of the oldest stars as means of constraining,
if not estimating, the values of the cosmological parameters. The differences between the
world models only become significant at redshifts greater than about 0.3. For example, the
difference in apparent magnitude of a galaxy at redshift 0.5 between the steady state model
with q0 = −1 and a Friedman model with q0 = 1 and λ = 0 was only 0.9 magnitudes.2
Sandage discussed in detail the problems of using these different techniques to determine
cosmological parameters, and he concluded that the most promising route was the use of
the redshift–apparent magnitude relation for the brightest galaxies in clusters for which the
dispersion in absolute magnitude about the mean relation was only about m ≈ 0.3. His
best estimate for q0 was 1 ± 0.5, but it could have ranged between 0 and 3. He warned
that the analysis involves a number of important selection effects which needed to be taken
into account before a convincing estimate could be made. In particular, he emphasised
the importance of the Malmquist bias, according to which intrinsically brighter objects
are selected in studies which extend to the limit of observational capability (Malmquist,
1920).
Sandage also noted that there was a discrepancy between the ages of the oldest
globular clusters, which were estimated to be about 15 × 109 years, and the age of the
Universe, which, for the q0 = 1, λ = 0 model with H0 = 75 km s−1 Mpc−1 , was T0 =
7.42 × 109 years. A solution to this problem would be to assume that the cosmological
constant was positive, which would result in a negative value for q0 . Sandage took the view
that there were probably too large uncertainties in the estimates of H0 , q0 and T0 for this
result to be taken too seriously. He devoted an enormous effort to the determination of the
basic cosmological parameters, Hubble’s constant, H0 , the deceleration parameter, q0 , and
the age of the Universe, T0 using the telescopes at the Palomar Observatory, and, until the
1970s, his work dominated the field.3 Indicative of his approach to observational cosmol-
ogy during these years was the title of his paper, ‘Cosmology – the search for two numbers’
(Sandage, 1970).
 13.2 Hubble’s constant 343

Figure 13.2: Illustrating the ‘cosmological distance ladder’; after M. Rowan-Robinson, The Cosmo-
logical Distance Ladder (New York: W. H. Freeman and Company, 1985). The diagram shows roughly
the range of distances over which different classes of object can be used to estimate astronomical
distances.

13.2 Hubble’s constant

Hubble’s constant, H0 , appears ubiquitously in cosmological formulae, and its value was
the subject of considerable controversy for many years. The use of the redshift–magnitude
relation for the brightest cluster galaxies had the advantage that Hubble’s law is defined well
beyond distances at which there might have been deviations associated with the peculiar
motions of clusters and superclusters of galaxies. Therefore, Hubble’s constant could be
found if the distances to the nearest rich clusters of galaxies could be estimated accurately.
The traditional approach to this calibration involved a hierarchy of distance indicators
to extend the local distance scale from the vicinity of the Solar System to the nearest giant
cluster of galaxies, the Virgo cluster. The only direct methods of distance measurement
involve stellar parallaxes, and these can only be used for stars in the neighbourhood of the
Sun. To extend the distance scale further, it is assumed that objects of the same intrinsic types
can be identified at greater distances. Then, their relative brightnesses provide estimates of
their distances. Examples of the different techniques used are summarised in Figure 13.2.
344 13 The determination of cosmological parameters

The period–luminosity relation for Cepheid variables, discovered by Henrietta Leavitt in

1912 (Leavitt, 1912), provides one of the best means of extending the distance scale from
our own Galaxy to nearby galaxies, but, even using the 200-inch telescope, it was only
possible to use this procedure to distances of about 1–2 Mpc. Other techniques were used
to extend the distance scale from the neighbourhood of our Galaxy to the Virgo cluster,
including the luminosity functions of globular clusters, the brightest stars in galaxies and
the luminosities of Type I supernovae at maximum light. In 1977, Brent Tully and Richard
Fisher discovered the relation between the absolute magnitudes of spiral galaxies and the
velocity widths of their 21 cm line emission (Tully and Fisher, 1977). This relation could
be determined for a number of spiral galaxies in a nearby group or cluster and then relative
distances found by assuming that the same correlation between their intrinsic properties is
found in more distant groups and clusters (see Section 10.1).
From the 1970s until the 1990s, there was an ongoing controversy concerning the value of
Hubble’s constant.4 In a long series of papers, Sandage and Gustav Tammann (b. 1932) found
values of Hubble’s constant about 50 km s−1 Mpc−1 , whereas de Vaucouleurs, Aaronson,
Mould and their collaborators found values of about 80 km s−1 Mpc−1 . The nature of the
discrepancy can be appreciated from their estimates of the distance to the Virgo cluster.
If its distance is 15 Mpc, the higher estimate of H0 is found, whereas if the distance is
22 Mpc, values close to 50 km s−1 Mpc−1 are obtained. Sandage and Tammann repeatedly
emphasised how sensitive the distance estimates are to observational selection effects, such
as the Malmquist effect, and systematic errors.
During the 1990s, a major effort was made to resolve these differences, much of it
stimulated by the capability of the Hubble Space Telescope (HST) to measure Cepheid
variable stars in the Virgo cluster of galaxies. When the HST project was approved in 1977,
one of its major scientific objectives was to use its superb sensitivity for faint star-like
objects to enable the light curves of Cepheid variables in the Virgo cluster to be determined
precisely and so estimate the value of Hubble’s constant to 10% accuracy. This programme
was raised to the status of an HST Key Project in the 1990s, with a guaranteed share of
observing time to enable a reliable result to be obtained.
The Key Project team, led by Wendy Freedman (b. 1957), carried out an outstanding
programme of observations and analysis of these data. Equally important was the fact
that the team used, not only the HST data, but also all the other distance measurement
techniques, to ensure internal self-consistency of the distance estimates. For example, the
improved determination of the local distance scale in our own Galaxy from the parallax
programmes of the Hipparcos astrometric satellite improved significantly the reliability of
the calibration of the local Cepheid distance scale. The great advance of the 1990s was that
the distances of many nearby galaxies became known very much more precisely than they
were previously. As a result, by 2000, there was relatively little disagreement among the
experts about the distances of those galaxies which had been studied out to the distance of the
Virgo cluster. If there were differences, they arose from how the data were to be analysed
once the distances were known, in particular in the elimination of systematic errors and
biases in the observed samples of galaxies. The final result of the project, published in
2001, was 70 ± 7 km s−1 Mpc−1 , where the errors are one-sigma errors (Freedman et al.,
2001).
13.2 Hubble’s constant 345

In addition to the traditional approach, new physical methods of measuring H0 became

available, which have the advantage of eliminating many of the steps involved in the cos-
mological distance ladder. They are based upon measuring a physical dimension l of a
distant object, independent of its redshift, and its angular size θ , so that an angular diameter
distance, DA , can be found from DA = l/θ at a known redshift z. A beautiful example of
the use of this technique was described by Nino Panagia (b. 1943) and his colleagues, who
combined IUE observations of the time-variability of the emission lines from the supernova
SN 1987A in the Large Magellanic Cloud with Hubble Space Telescope observations of
the emission-line ring observed about the site of the explosion to measure the physical size
of the ring (Panagia et al., 1991) (see Section 8.10.2). The distance found for the Large
Magellanic Cloud was as accurate as that found by the traditional procedures.
Another promising method, suggested originally by Walter Baade in 1926 and modi-
fied by Adriaan Wesselink (1909–1995) in 1947, involves measuring the properties of an
expanding stellar photosphere (Baade, 1926; Wesselink, 1947). If the velocity of expansion
can be measured from the Doppler shifts of the spectral lines, and the increase in size
estimated from the change in luminosity and temperature of the photosphere, the distance
of the star can be found. The Baade–Wesselink method was first applied to supernovae
by David Branch (b. 1942) and Bruce Patchett (1948–1996) in 1973 (Branch and Patch-
ett, 1973) and by Robert Kirshner and John Kwan (b. 1947) in 1974 (Kirshner and Kwan,
1974). It was successfully applied to the supernovae SN 1987A in the Large Magellanic
Cloud by Ronald Eastman (b. 1958) and Kirshner, resulting in a distance consistent with
other precise distance measurement techniques (Eastman and Kirshner, 1989). Extending
the Baade–Wesselink technique to ten Type II supernovae with distances ranging from
50 kpc to 120 Mpc, Brian Schmidt (b. 1967) and his colleagues found a value of H0 of
60 ± 10 km s−1 Mpc−1 (Schmidt, Kirshner and Eastman, 1992).
Another approach which has produced promising results involves the use of the hot
gaseous atmospheres in clusters of galaxies, the properties of which can be measured from
their X-ray emission and from the Sunyaev–Zeldovich decrement in the cosmic microwave
background radiation due to inverse Compton scattering (Figure 10.12) (Gunn, 1978). As
discussed in Section 10.5.1, clusters of galaxies contain vast quantities of hot gas which is
detected by its X-ray bremsstrahlung. The X-ray surface brightness depends upon the elec-
−1/2
tron density, Ne , and the electron temperature, Te , through the relation Iν ∝ Ne2 Te dl.
The electron temperature, Te , can be found from the shape of the bremsstrahlung spectrum.
Furthermore, the decrement in the background  due to the2 Sunyaev–Zeldovich
 effect is pro-
portional to the Compton optical depth, y = (kTe /m e c ) σT Ne dl ∝ Ne Te dl. Thus, the
physical properties of the hot gas are over-determined and the physical dimensions of the
X-ray emitting volume can be found. Steven Myers (b. 1962) and his colleagues estimated
a value of H0 = 54 ± 14 km s−1 Mpc−1 from detailed studies of the Abell clusters A478,
A2142 and A2256 (Myers et al., 1997).
Another example of a physical method of measuring H0 is to use gravitational lensing
of distant objects by intervening galaxies or clusters. The first gravitationally lensed quasar,
0957+561 (Figure 13.3), was discovered by Dennis Walsh (1933–2005), Robert Carswell
(b. 1940) and Ray Weymann (b. 1934) in 1979 (Walsh, Carswell and Weymann 1979). The
gravitational deflection of the light from the quasar by the intervening galaxy splits its
346 13 The determination of cosmological parameters

Figure 13.3: An optical image of the double quasar 0957+561 discovered by Walsh, Carswell and
Weymann in 1979 (Walsh et al., 1979). The spectra of the two quasars were identical. The small
extension to the right of image B is the foreground galaxy responsible for the gravitational lensing.
North is to the right of the image. (Image courtesy of Richard Ellis and Jean-Paul Kneib.)

image into a number of separate components. If the background quasar is variable, a time
delay is observed between the variability of the different images because of the different
path-lengths from the quasar to the observer on Earth. For example, a time delay of 418 days
has been measured for the two components of the double quasar 0957+561 (Kundic et al.,
1997). This observation enables physical scales at the lensing galaxy to be determined,
the main uncertainty resulting from the modelling of the mass distribution in the lensing
galaxy. In the case of the double quasar 0957+561, Tomislav Kundic (b. 1968) and his
colleagues claim that the mass distribution in the galaxy is sufficiently well constrained
for the model-dependent uncertainties to be small, and they derived a value of Hubble’s
constant of H0 = 64 ± 13 km s−1 Mpc−1 at the 95% confidence level.
The estimates of Hubble’s constant found by these physical methods are consistent with
the value found by Freedman and her colleagues.

13.3 The age of the Universe, T0

With Baade’s revision of the value of Hubble’s constant in 1952 and the further revision by
Sandage in 1958, the discrepancy between the age of the Earth and H0−1 was eliminated, but
it was known that the ages of the oldest stars in globular clusters were considerably greater
than that of the Solar System. Globular cluster ages were estimated by the method pioneered
13.3 The age of the Universe, T0 347

by Sandage and Schwarzschild in 1952 and involved the comparison of the Hertzsprung–
Russell diagrams of the oldest, metal-poor, globular clusters with the expectations of the
theory of stellar evolution from the main sequence onto the giant branch (Sandage and
Schwarzschild, 1952).
The feature of these diagrams which is particularly sensitive to the age of the cluster
is the main-sequence termination point. In the oldest globular clusters, the main-sequence
termination point has reached a mass of about 0.9M , and in the most metal-poor, and
presumably oldest, clusters the abundances of the elements with Z ≥ 3 are about 150 times
lower than their Solar System values. These facts made the determination of stellar ages
much simpler than might be imagined. As Michael Bolte (b. 1955) pointed out, low-mass,
metal-poor stars have radiative cores and so are unaffected by the convective mixing of
unprocessed material from their envelopes into their cores (Bolte, 1997). Furthermore, the
corrections to the perfect gas law equation of state are relatively small throughout most solar-
mass stars. Finally, the surface temperatures of these stars are high enough for molecules to
be rare in their atmospheres, simplifying the conversion of their effective temperatures into
predicted colours. Taking account of the various sources of uncertainty, Brian Chaboyer
(b. 1965) demonstrated that the absolute magnitude of the main-sequence termination point
is the best indicator of the age of the cluster (Chaboyer, 1998).
As understanding of the theory of stellar evolution advanced, improved estimates of the
ages of the oldest globular clusters became available. A good example of what could be
achieved is illustrated in Figure 13.4, which shows a comparison of the Hertzsprung–Russell
diagram for the old globular cluster 47 Tucanae with the predicted isochrones for various
assumed ages for the cluster. In this case the abundance of the heavy elements is only 20%
of the solar abundance, and the age of the cluster is estimated to be between 12 × 109 and
14 × 109 years (Hesser et al., 1987).
In 1994, André Maeder (b. 1942) reported evidence that the ages of the oldest globular
clusters are about 16 × 109 years (Maeder, 1994), and similar results were reported by
Sandage in 1995 (Sandage, 1995). In 1997, Bolte argued that the ages of the oldest globular
clusters were given by

T0 = 15 ± 2.4 (stat) +4
−1 (syst) Gyears. (13.2)

The first results of the Hipparcos astrometric survey relating to determination of the
local distance scale were announced in 1997, with the result that it increased by about
10% (Feast and Catchpole, 1997). This result meant that the stars in globular clusters
were more luminous than previously thought and so their main-sequence lifetimes were
reduced. In Chaboyer’s review of 1998, the ages of globular clusters were estimated to be
T0 = (11.5 ± 1.3) Gyears (Chaboyer, 1998).
Constraints on the age of the Galaxy can also be obtained from estimates of the cooling
times for white dwarfs. According to Chaboyer, these provide a firm lower limit of 8 Gyears.
The numbers  of white dwarfs observed in the vicinity of the Solar System enable an estimate
+1.1
of 9.5 −0.8 Gyears to be made for the age of the disc of our Galaxy (Oswalt et al., 1996).
Just as Rutherford had used the relative abundances of the radioactive species to set a
lower limit to the age of the Earth in 1904 (see Section 3.1), so lower limits to the age of
348 13 The determination of cosmological parameters

Figure 13.4: The Hertzsprung–Russell diagram for the globular cluster 47 Tucanae (Hesser et al.,
1987). The solid lines show fits to the data using theoretical models of the evolution of stars of different
masses from the main sequence to the giant branch due to VandenBerg (b. 1947). The isochrones shown
have ages of 10, 12, 14 and 16 ×109 years, the best fitting values lying in the range (12–14) ×109 years.
The cluster is metal-rich relative to other globular clusters, the metal abundance corresponding to
about 20% of the solar value.

the Universe can be derived from the discipline of nucleocosmochronology. A secure lower
limit to the age of the Universe can be derived from the abundances of long-lived radioactive
species. In 1963, Edward Anders (b. 1926) used these to determine an accurate age for the
Earth of 4.6 × 109 years (Anders, 1963). Some pairs of long-lived radioactive species, such
as 232 Th–238 U, 235 U–238 U and 187 Re–187 Os, can provide information about nucleosynthetic
timescales before the formation of the Solar System (Schramm and Wasserburg, 1970).
These pairs of elements are all produced by the r-process, in which the timescale for neutron
capture is less than the β-decay lifetime. The production abundances of these elements can
be predicted and compared with their present observed ratios (Cowan, Thielemann and
Truran, 1991).
The best astronomical application of this technique has been carried out by Christo-
pher Sneden (b. 1947), John Cowan (b. 1948) and their colleagues for the ultra-metal-
poor K giant star CS 22892-052, in which the iron abundance is 1000 times less than the
solar value (Sneden et al., 1992). A number of species never previously observed in such
metal-poor stars were detected, for example Tb (terbium, Z = 65), Ho (holmium, Z = 67),
 13.4 The deceleration parameter, q0 349

Tm (thulium, Z = 69), Hf (hafnium, Z = 72) and Os (osmium, Z = 76), as well a single

line of Th (thorium, Z = 90). The thorium abundance is significantly smaller than its scaled
Solar System abundance, and so the star must have been formed much earlier than the Solar
System. A lower limit to the age of CS 22892-052 of (15.2 ± 3.7) × 109 years was found.
A conservative lower bound to the cosmological timescale can be found by assuming that
all the elements were formed promptly at the beginning of the Universe. From this line of
reasoning, David Schramm (1945–1997) found a lower limit to the age of the Galaxy of 9.6 ×
109 years (Schramm, 1997). The best estimates of the age of the Galaxy are somewhat model-
dependent, but typically ages of about (12−14) × 109 years are found (Cowan et al., 1991).

13.4 The deceleration parameter, q0

The hope of the pioneers of observational cosmology was that the value of q0 could be
found from studies of distant galaxies through the apparent magnitude–redshift relation,
the angular diameter–redshift relation or the number counts of galaxies. This programme
proved to be very much more difficult than the pioneers had expected for a number of reasons
that are dealt with in this section. By the end of the twentieth century, real progress was
made by two somewhat different routes, one involving the use of supernovae of Type Ia and
the other involving observations of the spectrum of fluctuations in the cosmic microwave
background radiation (see Section 15.4). The major problem encountered by many of the
traditional approaches concerned the evolution with cosmic time of the properties of the
objects studied. These issues are of key importance in understanding the processes involved
in the formation of galaxies and larger scale structures and are dealt with in more detail
in Chapter 15. The emphasis in this section is upon endeavours to estimate the value of q0
from observations of distant objects.

13.4.1 Redshift–magnitude relation for the brightest galaxies in clusters

The redshift–magnitude relation for the brightest galaxies in clusters showed an impressive
linear relation (Figure 13.1) (Sandage, 1968), but it only extended to redshifts z ∼ 0.5 at
which the differences between the world models are relatively small. Sandage was well
aware of the many effects which needed to be considered before a convincing estimate of
q0 could be found. Some of these were straightforward, such as the need to determine the
luminosities of galaxies within a given metric diameter, but others were more complex. For
example, as discussed in Section 10.5, Sandage and Hardy discovered that the brightest
galaxy in a cluster is more luminous, the greater the difference in magnitude between the
brightest and next brightest galaxies in the cluster (Sandage and Hardy, 1973). In what they
termed the Bautz–Morgan effect, the second- and third-ranked members of the cluster were
intrinsically fainter than the corresponding galaxies in other clusters with less dominant first-
ranked galaxies. It seemed as though the brightest galaxy became brighter at the expense
of the next brightest members, a phenomenon which could plausibly be attributed to the
effects of galactic cannibalism (Hausman and Ostriker, 1977). Sandage adopted an empirical
correction to reduce the clusters to a standard Bautz–Morgan type.
350 13 The determination of cosmological parameters

Sandage was well aware of the need to take account of the evolution of the stellar popula-
tions of the galaxies with cosmic time. These corrections followed naturally from his work
on the Hertzsprung–Russell diagrams of globular clusters of different ages which mimic
the cosmic evolution of the old stellar populations of galaxies. He included evolutionary
corrections in the K -correction to the absolute magnitudes of the galaxies. There were,
however, other worrying pieces of evidence which did not fit easily into a picture of pas-
sive evolution of the galaxies in clusters. Dramatic evidence for the evolution of galaxies
in rich, regular clusters at relatively small redshifts was first described in the pioneering
analyses of Harvey Butcher (b. 1947) and Augustus Oemler (b. 1945) (Butcher and Oemler,
1978, 1984). They found that the fraction of blue galaxies in such clusters increased from
less than 5% in a nearby sample to percentages as large as 50% at redshift z ∼ 0.4. The
Butcher–Oemler effect was the subject of a great deal of study and debate, the major obser-
vational problems concerning the contamination of the cluster populations by foreground
and background galaxies, as well as bias in the selection criteria for the clusters selected
for observation (Dressler, 1984).
The determination of q0 might seem to be easier if the samples of galaxies extended to
larger redshifts, but it proved far from trivial to find suitable clusters at redshifts greater
than 0.5. Those in which the brightest galaxies were observed often turned out to be bluer
than expected. This finding reflects a basic problem with this approach to measuring the
deceleration parameter – the differences between the expectations of the world models only
become appreciable at large redshifts when the Universe was significantly younger than it
is now. Consequently, careful account has to be taken of the evolutionary changes of the
objects which are assumed to have ‘standard’ properties.
By the time of Sandage’s review of the problem in 1993, the uncertainties in the value
of q0 had not decreased, his estimate being q0 = 1 ± 1 (Sandage, 1995). In fact, by that
time, Alfonso Aragòn-Salamanca (b. 1962), Richard Ellis and their colleagues had extended
the infrared apparent magnitude–redshift relation for the brightest galaxies in clusters to
redshift z = 0.9 (Aragòn-Salamanca et al., 1993). They found evidence that the galaxies
were bluer at the larger redshifts, but, perhaps surprisingly, that their apparent magnitude–
redshift relation followed closely a model with q0 = 1, with no corrections for the evo-
lution of the stellar populations of the galaxies, for cluster richness or for Bautz–Morgan
type.

13.4.2 Redshift–magnitude relation for radio galaxies

Another approach to extending the redshift–magnitude relation to large redshifts became
possible in the early 1980s when the use of the first generation of CCD cameras enabled
complete samples of bright 3CR radio sources to be identified with very faint galaxies. These
galaxies turned out to have very strong, narrow emission-line spectra, and spectroscopy by
Hyron Spinrad (b. 1934) and his colleagues showed that many of these radio galaxies had
very large redshifts.5 These observations showed that the 3CR radio galaxies are among the
most luminous galaxies known.
At about the same time, infrared photometry of these galaxies in the 1–2.2 μm waveband
became feasible with the development of sensitive indium antimonide detectors. There
13.4 The deceleration parameter, q0 351

Figure 13.5: The K magnitude–redshift relation for a complete sample of radio galaxies from the 3CR
catalogue. The infrared apparent magitudes were measured at a wavelength of 2.2 μm. The dashed
lines show the expectations of world models with q0 = 0 and 0.5. The solid line is a best fitting line
for standard world models which include the effects of stellar evolution on the old stellar population
of the galaxies (Lilly and Longair, 1984).

were several advantages in defining the redshift–apparent magnitude relation in the infrared
waveband, one of them being that dust becomes transparent in the near-infrared waveband
and so extinction corrections to the luminosities of the galaxies are very small. A second
advantage is that the stars that contribute most of the luminosity at these wavelengths belong
to the old red giant population of the galaxy. As a result, the magnitudes are not affected by
the bursts of star formation, which can profoundly influence the optical magnitudes of the
galaxies and which are largely responsible for the fact that the galaxies at redshifts greater
than 0.5 were found to be significantly bluer than those observed at lower redshifts (see
Section A13.1).
In 1984, Simon Lilly (b. 1959) and I determined the redshift–apparent magnitude rela-
tion for a complete sample of 3CR radio galaxies at an infrared wavelength of 2.2 μm
(Figure 13.5) (Lilly and Longair, 1984). We found that there is a remarkably well defined
K magnitude–redshift relation which extended to redshifts of 1.5. It was also clear that the
galaxies at large redshifts were more luminous than expected for world models with q0 ∼
0–0.5. When simple evolutionary corrections were made for the increased rate at which
stars evolved onto the giant branch at earlier epochs (see Section A13.1), values of q0 in
the range 0 to 1 were found. This appeared to be evidence for the evolution of the stellar
populations of these galaxies over cosmological timescales.
There were, however, problems even with this simple picture. In the late 1980s, as
discussed in Section 11.6.3, Chambers, Miley, McCarthy and their collaborators discovered
the alignment of the radio structures with the optical images of the 3CR galaxies, and
this complicated the interpretation of these data (Chambers et al., 1987; McCarthy et al.,
352 13 The determination of cosmological parameters

1987). Using a combination of surface photometry of these galaxies in the optical and
infrared wavebands, we were able to show that the alignment effect does not have a strong
influence upon the K magnitude–redshift relationships (Best, Longair and Röttgering, 1998).
More serious was the fact that surveys of fainter samples of 6C radio galaxies by Stephen
Eales (b. 1960), Steven Rawlings (b. 1961) and their colleagues found that, although the
K magnitude redshift relation agreed with our relation at redshifts less than 0.6, their
sample of radio galaxies at redshifts z ∼ 1 were significantly less luminous than the 3CR
galaxies by about 0.6 magnitudes (Eales et al., 1997). Our most recent analysis of these
data for a cosmological model with 0 = 0.3 and  = 0.7, including corrections for the
evolution of their stellar populations, have demonstrated clearly that the large-redshift 3CR
radio galaxies are significantly more luminous than their nearby counterparts (Inskip et al.,
2002), and so the apparent success in accounting for the K magnitude-redshift relation for
3CR radio galaxies was an unfortunate cosmic conspiracy.
The lesson of this story is that the selection of galaxies as standard objects at large
redshifts is a hazardous business – we generally learn more about the astrophysics of the
most massive galaxies than about geometrical cosmology.

13.4.3 Redshift–magnitude relation for Type Ia supernovae

The discussion of Section 13.4.2 makes it clear that what is required is a set of standard
objects which are not susceptible to poorly understood evolutionary changes with cosmic
epoch. The use of supernovae of Type Ia to extend the apparent magnitude–redshift relation
to redshifts z > 0.5 has a number of attractive features. First of all, it is found empirically
that these supernovae have a very small dispersion in absolute luminosity at maximum light
(Branch and Tammann, 1992). This dispersion can be further reduced if account is taken of
the correlation between the maximum luminosity of Type Ia supernovae and the duration of
the initial outburst (see Section 8.10.1). This correlation, referred to as the luminosity–width
relation, is in the sense that the supernovae with the slower decline rates from maximum
light are more luminous than those which decline more rapidly. Secondly, there are good
astrophysical reasons to suppose that these objects are likely to be good standard candles,
despite the fact that they are observed at earlier cosmological epochs. The preferred picture
is that these supernovae result from the explosion of white dwarfs, which are members
of binary systems that accrete mass from the other member of the binary. Although the
precise mechanism that initiates the explosion has not been established, the favoured picture
is that mass accreted onto the surface of the white dwarf raises the temperature of the
surface layers to such a high temperature that nuclear burning is initiated and a deflagration
front propagates into the interior of the star, causing the explosion which results in its
destruction.
In 1995 Ariel Goobar (b. 1962) and Saul Perlmetter (b. 1959) discussed the feasibility
of observing Type Ia supernova out to redshift z ≈ 1 in order to estimate the values of the
density parameter, 0 , and the cosmological constant (Goobar and Perlmutter, 1995). In
1996, they and their colleagues described the first results of systematic searches for Type
Ia supernovae at redshifts z ∼ 0.5 using an ingenious approach to detect them before they
reach maximum light (Perlmutter et al., 1996). Deep images of selected fields, including
13.4 The deceleration parameter, q0 353

a number which contain distant clusters of galaxies, are taken during one period of the
new moon and the fields are imaged in precisely the same way during the next new
moon. Using rapid image analysis techniques, any supernovae which appeared between
the first-and second-epoch observations were quickly identified and reobserved photomet-
rically and spectroscopically over the succeeding weeks to determine their types and light
curves.
Using this search technique, Perlmutter and his colleagues discovered 27 supernovae of
Type Ia between redshifts 0.4 and 0.6 in three campaigns in 1995 and 1996 (Perlmutter et al.,
1996, 1997). The team used these data to demonstrate directly the effects of cosmological
time dilation by comparing the light curves of Type Ia supernovae at redshifts z ∼ 0.4–0.6
with those of the same type at the present epoch, thus testing directly the cosmological
time dilation–redshift relation (Goldhaber et al., 1996). The same peak luminosity–width
relation was found as that observed at small redshifts. When account was taken of this
relation, the intrinsic spread in the luminosities of the Type Ia supernovae was only 0.21
magnitudes.
This same technique has been used to discover Type Ia supernovae at redshifts greater than
z = 0.8 as a result of observations with the Hubble Space Telescope. In two independent
programmes, Peter Garnavich (b. 1958), Perlmutter and their colleagues discovered the
Type Ia supernovae SN1997ck at redshift z = 0.97 and SN1997ap at redshift z = 0.83,
respectively (Garnavich et al., 1998; Perlmutter et al., 1998). The great advantage of the
HST observations is that their high angular resolution enables very accurate photometry to
be carried out on stellar objects in distant galaxies.
The redshift–apparent magnitude relation presented by Perlmutter is shown in Figure
13.6 and is similar to that found by Garnavich and his colleagues. The major result of
these observations, which was found by both groups independently, is that the data favour
cosmological models in which the cosmological constant, λ, is non-zero. This was the
first time in the history of observational cosmology that compelling evidence for a finite
value of the cosmological constant had been found. Both groups have continued to extend
this technique to large redshifts through the discovery of Type Ia supernovae at very large
redshifts (Knop et al., 2003; Tonry et al., 2003). The best presentation of these results is
in terms of a diagram in which the density parameter of the matter content of the Uni-
verse, 0 , is plotted against the value of the cosmological constant, written in terms
of the energy density in the vacuum fields through the normalisation  = λ/3H02 –
the reason for this form of presentation is that, if the global geometry of the Universe
were flat, 0 +  = 1. The results of the Supernova Cosmology Project are shown in
Figure 13.7.
There are various ways of interpreting Figure 13.7, particularly when taken in conjunction
with independent evidence on the mean mass density of the Universe and the evidence from
the spectrum of fluctuations in the cosmic microwave background radiation. Perhaps the
most conservative approach is to note that the matter density in the Universe must be greater
than zero, and, as discussed in Section 13.5, all the data are consistent with values of 0 ≈
0.25–0.3. Consequently, the cosmological constant must be non-zero. The data would be
consistent with 0 +  = 1 if 0 ≈ 0.25–0.3. We will come back to these results in
Chapter 15.
354 13 The determination of cosmological parameters

(ΩΜ,ΩΛ) = (0, 1)
(0,5,0,5) (0,0)
(1, 0) (1,0)
24 (1, 5,− 0,5) (2,0)

Λ=0
Flat
22
Supernova
Cosmology
20 Projeet
effective mB

18

Calan/Tololo Survey
16

(a)
14
1.5
(ΩΜ, ΩΛ) =
1.0
mag residual

(0, 1)
0.5
(0.28, 0.72)
0.0
(0, 0)
−0.5 (0,75, 0,25)
−1.0 (b) (1, 0)
−1.5
6
standard deviation

4
2
0
−2
−4
−6 (c)
0.0 0.2 0.4 0.6 0.8 1.0
redshift, z

Figure 13.6: Results of the Supernova Cosmology Project (Perlmutter et al., 1999). (a) The
Hubble diagram for 42 high-redshift Type Ia supernovae and 19 low-redshift Type Ia super-
novae from the Calan/Tololo Supernova survey plotted on a linear redshift scale. The theoreti-
cal lines are for (0 ,  ) = (0, 0); (1, 0); (2, 0) from top to bottom (solid lines) and (0 ,  ) =
(0, 1); (0.5, 0.5); (1, 0); (1.5, −0.5) from top to bottom (dotted lines). (b) The magnitude residuals
from the best-fit flat cosmological model with M = 0.28 and  = 0.72. (c) The uncertainty-
normalised residuals from the best-fit cosmological model.

13.4.4 Number counts of galaxies

In his assessment of approaches to the determination of cosmological parameters using the
number counts of galaxies, Sandage was not optimistic (Sandage, 1961a):

Galaxy counts are insensitive to the model . . . There seems to be no hope of finding q0 from the N (m)
counts because the predicted differences between the models are too small compared with the known
fluctuations of the distribution.

These concerns have been fully justified by subsequent studies. The determination of
precise counts of galaxies has proved to be one of the more difficult areas of observational
cosmology.6 The reasons for these complications are multifold. First of all, galaxies are
13.4 The deceleration parameter, q0 355

3
no Big Bang SCP 2003

68%, 90%, 95%, 99%

ng
rati
ele
ac
c ing
le rat
ce
ΩΛ 1 de

ver
expands fore
0 en tually
recollapses ev
cl
os
fla
op

ed
t
en

−1
0 1 2 3
ΩΜ
Figure 13.7: The 68%, 90%, 95% and 99% confidence limits for the values of 0 and  determined
by the Supernova Cosmology Project. The line labelled ‘flat’ on the diagram shows the condition
0 +  = 1, which corresponds to flat geometry (Knop et al., 2003).

extended objects, often with complex brightness distributions, and great care must be taken
to ensure that the same types of object are compared at different magnitude limits and red-
shifts. Furthermore, the distribution of galaxies is far from uniform on scales less than about
50h −1 Mpc, as illustrated by the large voids and walls in the local distribution of galaxies
(see Figures 10.7–10.9). Even at the faintest magnitudes, this ‘cellular’ structure in the dis-
tribution of galaxies results in fluctuations in the number counts of galaxies which exceed
the statistical fluctuations expected in a random distribution. In addition, the probability of
finding galaxies of different morphological types depends upon the galaxy environment.
Finally, the luminosity function of galaxies is quite broad (Figure 10.2), and so the geomet-
rical differences between models are masked by the convolution of the predictions of the
world models with this function.
Up till about 1980, the deepest counts extended to apparent magnitudes of about 22
to 23 and, although there were disagreements between the results of different observers,
there was no strong evidence that the counts of galaxies departed from the expectations of
uniform world models. In the 1980s, much deeper counts became feasible with the use of
CCD cameras on 4-metre class telescopes, the deepest surveys being carried out by Anthony
Tyson (b. 1940) (Tyson, 1990). It was found that there is an excess of faint galaxies at blue
magnitudes greater than about 22 (Figure 13.8). In contrast, the counts of galaxies at red and
356 13 The determination of cosmological parameters

b j number counts

10 6

no evolution Ω0 = 0
no evolution Ω 0 = 1
10 5 10 4
number of galaxies deg−2 mag−1
100 1000 10
1

15 20 25

b j magnitude

Figure 13.8: The counts of faint galaxies in the blue (B) waveband compared with the expectations
of uniform world models with 0 = 0 and 1. For the references to the data points, see Ellis (1987).

infrared wavelengths showed little evidence of such strong evolution. While there is little
prospect of using these observations to determine cosmological parameters, they are of the
greatest interest in studying the astrophysical evolution of galaxies with cosmic epoch, and
this topic will be taken up in Chapter 14.

13.4.5 Angular diameter–redshift test

The angular diameter–redshift relation provides an attractive route for the determination of
cosmological parameters if accurate proper distances, l, of astronomical objects can be mea-
sured at large redshifts and their corresponding angular sizes, θ, measured. Then, the angular
diameter distance DA = l/θ can be determined as a function of redshift and compared with
the predictions of the standard world models. The physical methods of measuring proper
distances at large redshifts described in Section 13.2, involving the Sunyaev–Zeldovich
effect in conjunction with X-ray observations of the hot gas in clusters, gravitational lenses
and the various versions of the Baade–Wesselink method, all provide means of undertaking
 13.5 The density parameter, Ω0 357

this test, but as yet the techniques are not sufficiently precise to provide useful results. A
possible problem with this programme is the extent to which the predicted angular diameter–
redshift relations are modified by inhomogeneities in the distribution of mass along the line
of sight, which can significantly change the predicted relations (Dashevsky and Zeldovich,
1964; Zeldovich, 1964b; Dyer and Roeder, 1972).7
The alternative approach is to use objects which may be considered to be ‘rigid rods’,
but the problem is to find suitable standard objects that can be used in the test. A distinctive
feature of this test is that there is expected to be a minimum angular diameter, as the objects
are observed at large redshifts. A good example is the use of the separation of the radio
components of double radio sources, such as those illustrated in Figure 11.1(b). Large
samples of these objects can be found spanning a wide range of redshifts. This version of
the angular diameter–redshift test was first carried out by George Miley (b. 1942), who used
the largest angular size of the radio structures of radio galaxies and quasars as a ‘rigid rod’
(Miley, 1968, 1971), but no minimum was found in the observed relation. Vijay Kapahi
(1944–1999) confirmed this result, using instead the median angular separation, θm , of the
radio source components as a function of redshift (Kapahi, 1987), but again no minimum was
found (Figure 13.9(a)). The median angular separation of the source components is observed
to be roughly inversely proportional to redshift, and this was interpreted as evidence that
the median physical separation of the source components, l m , was smaller at large redshifts.
Examples of fits to the observational data using evolution functions of the form lm ∝
(1 + z)−n are shown in Figure 13.9(a) for world models with q0 = 0 and 0.5 – values of
n ≈ 1.5–2.0 can provide good fits to the data. There are, of course, many reasons why
the separation of the radio source components might be smaller in the past; for example
the ambient interstellar and intergalactic gas may well have been greater in the past and so the
source components could not penetrate so far through the surrounding gas. Again, we learn
more about astrophysical changes with cosmic epoch than about geometrical cosmology.
Another version of the same test was described by Kenneth Kellermann (b. 1937) and
involved using only compact double radio structures studied by very long baseline interfer-
ometry (Kellermann, 1993). He argued that these sources are likely to be less influenced
by changes in the properties of the intergalactic and interstellar gas, since the components
are deeply embedded within the central regions of the host galaxy. In his angular diameter–
redshift relation, there is evidence for a minimum in the relation, which would be consistent
with a value of q0 ∼ 0.5 (Figure 13.9(b)). The problem with this type of approach is that
we cannot be certain that precisely the same types of double radio source are being selected
at large and small redshifts.

13.5 The density parameter, Ω0

The critical cosmological density, ρcrit = 3H02 /8π G, depends upon the value of Hubble’s
constant, and so it is convenient to write the density parameter as

ρ0 ρ0 H0
0 = = , where h= (13.3)
ρcrit 2 × 10 h 2 kg m−3
−26 100 km s−1 Mpc−1
358 13 The determination of cosmological parameters

q0= 0 q0= 0.5

median angular size (arcseconds)

median angular size (arcseconds)

100 100

n=1
n=1.5

10 10

n=2
n= 2 n=1.5 n = 2.5

0.1 1 0.1 1
redshift redshift (a)

100
mean angular size (milliarcseconds)

q0= 1

10

q0= 1/ 2

q0= 0

1/z
ss
1
0.01 0.1 1 10
redshift (b)

Figure 13.9: (a) The angular diameter–redshift relation for double radio sources, in which the median
angular separation of the double radio source components, θm , is plotted against redshift (Kapahi,
1987). The observed relation follows closely the relation θm ∝ z −1 . The left-hand panel shows fits
to the observations for a world model with q0 = 0 and the right-hand panel is for a model with
q0 = 0.5; in both cases, the median separation of the components is assumed to change with redshift
as lm ∝ (1 + z)−n . (b) The mean angular diameter–redshift relation for 82 compact radio sources
observed by VLBI (Kellermann, 1993). In addition to the standard Friedman models, the relation for
steady state cosmology (SS) as well as the relation θ ∝ z −1 (dashed line) are shown.

and ρ0 is the present average mass density in the Universe. Estimates of ρ0 for galaxies were
included in Hubble’s first paper on the extragalactic nature of the diffuse nebulae using his
estimates of their average mass-to-light ratios. He found the value ρ0 = 1.5 × 10−28 kg m−3
(Hubble, 1926).
A convenient way of evaluating the average mass density of the Universe was first to
work out the average luminosity density due to galaxies by integrating over the luminosity
function of the galaxies, and then to convert this into a mass density by adopting a suitable
average value for the mass-to-light ratio of the matter in galaxies. An analysis of this
13.5 The density parameter, Ω0 359

nature was carried out by Oort in 1958, who found that the average mass density was
3.1 × 10−28 kg m−3 , assuming that Hubble’s constant was 180 km s−1 Mpc−1 (Oort, 1958).
In 1978, James Gunn expressed the same result in terms of the mass-to-light ratio which
would be needed if the Universe were to attain the critical density (Gunn, 1978). He found
(M/L)crit = 2600h, very much greater than the values found in our vicinity in the plane
of the Galaxy and in the visible parts of galaxies. As described in Sections 10.2 and 10.5,
however, the mass of dark matter in galaxies and clusters of galaxies far exceeds that in
the visible parts of galaxies. If account is taken of the dark matter, the overall mass-to-
luminosity ratio attains values of M/L ∼ 100–150. In well studied rich clusters, such as
the Coma cluster, the value of M/L is of the order of 250, but this value is biassed towards
elliptical and S0 galaxies, which have three times larger values of M/L than the spiral
galaxies, the latter contributing most of the light per unit volume in the Universe at large.
These values of M/L are significantly less than the value needed to close the Universe.
Gunn’s best estimate of the density parameter for bound systems such as galaxies, groups
and clusters of galaxies was about 0.1 and was independent of the value of h.
Neta Bahcall (b. 1942) described many different approaches that can be taken to derive
values of M/L for clusters of galaxies – cluster mass-to-light ratios, the baryon fraction
in clusters and studies of cluster evolution. These have all found the same consistent result
that the mass density of the Universe corresponds to 0 ≈ 0.25, and furthermore that
the mass approximately traces light on large scales (Bahcall, 2000). These results reflect
the generally accepted view that, if mass densities are determined for bound systems, the
total mass density in the Universe is about a factor of 4 less than that needed to close the
Universe.
On scales greater than those of clusters of galaxies, estimates of the mass density in the
general field can be found from the cosmic virial theorem (Peebles, 1976). In this procedure,
the random velocities of galaxies with respect to the mean Hubble flow are compared with
the varying component of the gravitational acceleration due to large-scale inhomogeneities
in the distribution of galaxies. As in the other methods of mass determination, the mass
density is found by comparing the random kinetic energy of galaxies with their gravita-
tional potential energy, this comparison being carried out in terms of two-point correlation
functions for both the velocities and positions of galaxies selected from the general field.
Application to the random velocities of field galaxies suggested that 0 might be larger
than 0.2 (Davis, Geller and Huchra, 1978; Davis and Peebles, 1983).
A similar argument involves studies of the infall of galaxies into superclusters of galaxies.
Galaxies in the vicinity of a supercluster are accelerated towards it, thus providing a measure
of the mean density of gravitating matter within the system. The velocities induced by large-
scale density perturbations depend upon the density contrast, ρ/ρ, between the system
studied and the mean background density. A typical formula for the infall velocity, u, of
test particles into a density perturbation is given by
 
0.6 ρ
u ∝ H0 r 0 (13.4)
ρ 0
(Gunn, 1978). In the case of small spherical perturbations, a result correct to second order
in the density perturbation was presented by Alan Lightman (b. 1948) and Paul Schechter
360 13 The determination of cosmological parameters

(a) (b)

Figure 13.10: Surface density plots of the density field in the local supergalactic plane. (a) The mass
distribution reconstructed from the peculiar velocity and distance information for the galaxies in this
region using the POTENT numerical procedure. (b) The density field of optical galaxies. Both images
have been smoothed with a Gaussian filter of radius 1200 km s−1 . The density contrast is proportional
to the height of the surface above (or below) the plane of the plot. (Hudson et al., 1995.)

(Lightman and Schechter, 1990):

   2
v 4/7 ρ 13/21 ρ
= − 13 0 + 4
 . (13.5)
v ρ 0
63 0
ρ 0

In Gunn’s analysis, this method resulted in values of 0 of about 0.2 to 0.3. In the 1990s,
complete samples of IRAS galaxies became available, and they were used to define the
local density and velocity fields. A mean density close to the critical density was found
(Lynden-Bell et al., 1988).
In an ambitious programme, Avishai Dekel (b. 1951) and his colleagues devised numer-
ical procedures for deriving the distribution of mass in the local Universe entirely from the
measured velocities and distances of complete samples of nearby galaxies, the objective
being to determine a three-dimensional map of velocity deviations from the mean Hubble
flow. Then, applying Poisson’s equation, the mass distribution responsible for the observed
peculiar velocity distribution can be reconstructed numerically. Figure 13.10 shows an
example of a reconstruction of the local density distribution using this procedure (Hudson
et al., 1995). Despite using only the velocities and distances, and not their number densities,
many of the familiar features of our local Universe are recovered – the Virgo supercluster
and the ‘Great Attractor’ can be seen, as well as voids in the mean mass distribution. These
procedures tended to produce somewhat larger values of 0 , Dekel stating that the density
parameter is greater than 0.3 at the 95% confidence level.
The issue of the total amount of dark matter present in the Universe was the subject of
heated debate throughout the 1990s. Some flavour for the points of contention among the
experts in the field can be appreciated from the discussion in 1996 at the Princeton meeting
Critical Dialogues in Cosmology between David Burstein (b. 1947) and Avishai Dekel, the
discussion being moderated by Simon White (b. 1951) (Dekel, Burstein and White, 1997).
The upshot of these considerations was that there was agreement that the value of 0 is
greater than 0.1, and that a value of 0.2 to 0.3 would probably be consistent with most of
 Notes to Chapter 13 361

the data, the only concern being the somewhat larger values favoured by Dekel and his
colleagues.
The infall test became feasible on very large scales following the completion of the two-
degree-field (2dF) survey of galaxy redshifts carried out at the Anglo-Australian Telescope.
The survey involved measuring redshifts for over 200 000 galaxies randomly selected from
the Cambridge APM galaxy survey. A cut through that survey was shown in Figure 10.9.
The concept behind the test was that superclusters of galaxies generate a systematic infall
of other galaxies in their vicinities and this would be evident in the pattern of recessional
velocities, resulting in anisotropy in the inferred spatial clustering of galaxies. Using the
redshifts of more than 141 000 galaxies from the 2dF galaxy redshift survey, John Peacock
and his colleagues discovered convincing statistical evidence for infall and estimated the
overall density parameter to be 0.6 0 = 0.43b ± 0.07, where b is the bias parameter, the
factor by which visible matter is more clustered than the dominant dark matter. When this
result was combined with data on the anisotropy of the cosmic microwave background, their
result favoured a low-density Universe with 0 ≈ 0.3 (Peacock et al., 2001).
When taken in conjunction with the results derived from the power spectrum of fluc-
tuations in the cosmic microwave background radiation discussed in Section 15.4, the
consensus view is that the best estimate of the overall density parameter for the Universe is
given by 0 ≈ 0.25–0.3. An immediate consequence of this result is that most of the mass
cannot be in the form of baryonic matter, which is constrained by the production of the light
elements in the early stages of the Big Bang. To rephrase the results discussed in Section
12.6, a conservative upper limit to the density parameter in baryons is bar ≤ 0.0375h −2 ;
otherwise, less than the observed abundance of deuterium is expected to be synthesised
during the epoch of nucleosynthesis. Thus, adopting h = 0.7, there cannot be sufficient
baryons to account for the observed mass density. Most of the mass in the Universe must
be in some non-baryonic form.

13.6 Summary

From a subject dogged by controversy and strong feeling for most of the twentieth cen-
tury, classical cosmology saw a dramatic change in perspective during the last decade of
that century. New methods were developed which eliminated many of the problems of the
pioneering efforts of previous decades. Whilst the emphasis in this chapter has been upon
the traditional route to the determination of cosmological parameters, the consensus pic-
ture received a remarkable boost from analyses of the fluctuation spectrum of the cosmic
microwave background radiation, a story which is told in the context of the understand-
ing of the formation of large-scale structures in the Universe in Chapter 15. Indeed, many
cosmologists would now look first to these observations as providing the key to unlocking
many of the problems of classical cosmology.

Notes to Chapter 13

1 In Sandage’s historical review of efforts to determine the redshift–apparent magnitude relation for
galaxies, he reproduces many different versions of this relation, from Hubble’s first example of
1929 to his preferred relation of 1993 (Sandage, 1995).
362 13 The determination of cosmological parameters

2 The reader may wonder why the pioneers of observational cosmology worked in terms of the
deceleration parameter, q0 , rather than in terms of the density parameter, 0 , and the cosmological
constant, λ. The reason is that the expression for the comoving radial distance coordinate, r ,
and what I call the ‘effective distance’, D, used to relate intrinsic properties to observables are
independent of the density parameter, 0 , and the curvature of space, κ, to second order in the
redshift, z. Explicitly, we find

c z2
D=r = z + (1 + q0 ) + . . . .
H0 2

I have given details of this result in Section 8.2 of Malcom Longair, Galaxy Formation (Berlin:
Springer Verlag, 1998). A briefer version is given by James Gunn in his review of 1978 (Gunn,
1978). Thus, provided the observations do not extend to redshifts greater than, say, 0.4–0.5, the var-
ious cosmological tests provide direct information about the present deceleration, or acceleration,
of the Universe.
3 In his survey of his contributions to observational cosmology of 1995, Sandage provided an excel-
lent overview of the history of the determination of cosmological parameters by the techniques
listed above using optical observations (Sandage, 1995). The text includes many intriguing com-
ments and footnotes about the problems of theory and observation which faced the pioneers of
the subject from the 1920s to the 1990s. His delightful and touching remark about the comparison
between his and my expositions of the foundations of cosmology is a measure of the differences in
approach between the 1950s and the 1990s to the study of astrophysical cosmology (see Sandage
(1995), footnote 8, p. 25).
4 For a detailed discussion of the different approaches to the determination of Hubble’s constant
during the 1970s and 1980s, see M. Rowan-Robinson, The Cosmological Distance Ladder (New
York: W. H. Freeman and Company, 1985); Rowan-Robinson’s conclusions were updated in M.
Rowan-Robinson, The extragalactic distance scale, Space Science Reviews, 48, 1988, 1–71.
5 An account of Spinrad’s many contributions to the study of large-redshift radio galaxies are
described in the volume celebrating his 65th birthday, The Hy-redshift Universe, APS Confer-
ence Series, vol. 193, eds. A. J. Bunker and W. J. K van Breugel (San Francisco: APS, 1999).
6 An excellent account of the problems of determining and interpreting the counts of galaxies has
been given by Richard Ellis, Faint blue galaxies, Annual Review of Astronomy and Astrophysics,
35, 1997, 389–443.
7 I have given simple derivations of the angular diameter–redshift relations for inhomogeneous world
models in Section 7.4 of Longair, Galaxy Formation.

A13 Explanatory supplement to Chapter 13

A13.1 Observing distant galaxies in the infrared waveband
There are advantages in carrying out studies of the redshift–apparent magnitude relation in
the 1–2.2 μm wavebands. To demonstrate these, a model has been developed for the spectral
energy distribution of a giant elliptical galaxy at the present epoch, involving a wide range
of stellar masses and detailed information about their spectra throughout the ultraviolet,
optical and infrared wavebands. Then, models for the evolution of such a spectrum with
cosmic epoch can be determined and these depend upon the assumed initial mass function
of the stars in the galaxy and the star formation rate as a function of cosmic epoch.
The example shown in Figure A13.1 illustrates a number of important aspects of such
computations (Inskip et al., 2002). It is assumed that all the stars in the model giant elliptical
Explanatory supplement to Chapter 13 363

Figure A13.1: Illustrating the spectral evolution of a giant elliptical galaxy in which the stellar popula-
tion is formed in an initial star-burst of duration 108 years. The computations used the stellar synthesis
codes for the spectral energy distribution of galaxies developed by Gustavo Bruzual and Stefane Char-
lot (b. 1964) (Bruzual and Charlot, 2003). The different spectra show the predicted spectrum at later
times, the intensity at infrared wavelengths decreasing monotonically with increasing age, the ages of
the galaxies being shown by the key to the top right of the diagram. The computations were designed
so that the spectrum of a gaint elliptical galaxy would be reproduced at the present epoch (Inskip
et al., 2002). The vertical line labelled K is the standard K waveband at 2.2 μm. The vertical lines
labelled z = 1, 2 and 3 show the regions of the galaxy spectrum observed at the present epoch if the
galaxy were observed at these redshifts.

galaxy were formed in the first 108 years of its history, during which time the star-formation
rate was taken to be constant. The diagram shows the evolution of the spectral energy
distribution of the galaxy at subsequent times in the galaxy’s rest frame, resulting in the
observed spectrum of a giant elliptical galaxy at the present epoch. There is a one-to-one
relation between the mass of a star, its luminosity and the waveband in which it emits most of
this luminosity, so that stars radiating in the ultraviolet waveband are massive and have short
lifetimes. This is reflected in the rapid decrease with cosmic epoch of the spectral energy
distribution in the ultraviolet regions of the spectrum. In contrast, in the infrared waveband,
the stars are long-lived, and the gradual decrease in luminosity reflects the decreasing rate
at which stars evolve from the main sequence onto the red giant branch.
The origin of the slow decline with cosmic epoch of the infrared luminosity of the old
stellar populations of galaxies can be appreciated from the following simple calculation
due to Gunn (1978). Most of the luminosity of the galaxy in the infrared waveband 1–3 μm
is associated with red giant stars. Since the time stars spend on the giant branch, tg , and
364 13 The determination of cosmological parameters

their luminosities are relatively independent of their main-sequence masses, we need only
determine the rate at which stars evolve from the main sequence onto the giant branch as a
function of main-sequence mass in order to find the change in luminosity of a galaxy with
redshift. In the passive evolution model of galactic evolution, it is assumed that all the stars
were formed in an initial brief star-burst and that the subsequent luminosity evolution of
the galaxy is due to the stellar evolution of this population. For illustration, let us assume
that the initial mass function of the stars is of Salpeter form, dN = N (M) dM ∝ M −y dM,
where y = 2.35. It is a straightforward calculation to show that the number of stars on the
giant branch, Ng , is given by
  
dN dN dM
Ng = tg = tg . (A13.1)
dt dM dt
Thus, using the relation between mass and main-sequence lifetime derived in Section A4.1,
t = t (M/M )−(x−1) , we find
L(t) = L(t0 ) t −(x−y)/(x−1) . (A13.2)
Inserting the values x = 5 and y = 2.35, we find L ∝ t −0.66 . For the case of the critical
world model, t/t0 = (1 + z)−3/2 , and so, to a good approximation, L ∝ (1 + z). Thus, at a
redshift of unity, the old stellar populations of galaxies should be about twice as luminous
as they are at the present epoch, and at redshift z = 3, four times as luminous. This rough
calculation explains the slow decrease in the luminosity of the model galaxy in the infrared
(K) waveband seen in Figure A13.1.
The vertical line towards the right of Figure A13.1 shows the observing wavelength
of 2.2 μm at zero redshift; the vertical lines to the left show the region of the spectrum
observed at 2.2 μm at redshifts of 1, 2 and 3. It can be seen that, even at a redshift of 3, the
evolution of the stellar energy distribution is quite smooth, and relatively stable corrections
can be made for the effects of evolution of the stellar populations. This behaviour contrasts
strongly with what would be observed if the same exercise had been carried out in the optical
waveband. Then, the spectrum at large redshifts would originate from the strongly varying
ultraviolet regions of the spectrum. Furthermore, the predicted spectrum at large redshifts
is very sensitive to assumptions about the star-formation rate as a function of epoch. If there
were bursts of star formation, for example, the ultraviolet region of the spectrum would
be very strongly enhanced and the corrections for such effects would become difficult to
predict.
Thus, from the point of view of studying the properties of the integrated stellar popu-
lations of galaxies, there are real advantages in carrying out these studies in the 1–2.2 μm
waveband.
14 The evolution of galaxies
and active galaxies with
cosmic epoch

Evidence for strong evolutionary changes in the properties of extragalactic objects with
cosmic epoch was first found in the 1950s and 1960s as a result of surveys of radio sources
and quasars. An excess of faint sources was found in radio source and quasar surveys, as
compared with the expectations of uniform world models. The inference was that there
were many more of these classes of object at early cosmic epochs as compared with their
number at the present epoch. During the 1980s, as the first deep counts of galaxies became
available, a large excess of blue galaxies at faint apparent magnitudes was discovered. These
studies culminated in the remarkable observations of the Hubble Deep Field in 1998 and
the Hubble Ultra-Deep Field in 2004 by the Hubble Space Telescope.
In the 1990s, the first deep surveys of the X-ray sky were carried out by the ROSAT X-ray
observatory, and evidence for an excess of faint X-ray sources was found, similar in many
ways to the evolution inferred from studies of extragalactic radio sources and quasars. In the
thermal infrared wavebands, the IRAS survey, although not extending to as large redshifts as
the surveys mentioned above, also provided evidence for an excess of faint sources, which
appear to be evolving in a manner similar to the active galaxies. Then, in the last few years
of the century, evidence was found for a large population of submillimetre or far-infrared
galaxies at large redshifts.
The primary evidence for these evolutionary changes came from counting the numbers
of objects in well defined complete samples. In addition, since the 1970s, vast amounts of
new data accumulated on many different aspects of galaxy formation and evolution – neutral
hydrogen absorption-line systems, abundances of the elements in large-redshift absorption
systems, star-formation rates as a function of cosmic epoch and so on.1

14.1 The cosmological evolution of active galaxies

14.1.1 Source counts for the standard world models
In his earliest studies of galaxies as extragalactic systems, Hubble realised that the number
counts of galaxies potentially contain information about the large-scale structure of the
Universe. In his monograph The Realm of the Nebulae, he used counts of galaxies to the
limit of the Mount Wilson 100-inch telescope to demonstrate that, overall, the distribution

365
366 14 Cosmological evolution of galaxies and active galaxies

Figure 14.1: The predicted differential source counts, N , normalised to the Euclidean prediction,
N0 ∝ S −5/2 S, for a single luminosity class of source having spectral index α = 0.75 for different
cosmological models with λ = 0. The numbers opposite each point are the redshifts at which the
sources are observed (Longair, 1978).

of galaxies is homogeneous on the large scale (Hubble, 1936). As illustrated in Figure 6.2,
he compared the number counts with the expectations of a uniform Euclidean world model,
which were derived in Section A5.1:

N (≥S) ∝ S −3/2 or log N (≤ m) = 0.6m + constant, (14.1)

both results being independent of the luminosity function, N0 (L).

As the number counts extended to significant redshifts, however, departures from this
relation are expected. These are most simply demonstrated by a plot of the differential
number counts, normalised to the differential Euclidean number counts, N0 ∝ S −2.5 S, as
a function of flux density, S, (Figure 14.1). The Euclidean prediction, N /N0 = constant,
is represented by the abscissa, log10 (N /N0 ) = 0. The numbers opposite each point on
these relations are the redshifts at which the sources are observed. It can be seen that the
predicted differential counts depart rapidly from the Euclidean prediction, even at relatively
small redshifts. For example, for the case 0 = 1, the source counts at redshift z = 0.5 have
differential slope −2.08 rather than −2.5, corresponding to a slope of the integral source
counts of −1.08 rather than −1.5. In fact, the source populations cannot be represented by
a single luminosity, but, rather, the counts shown in Figure 14.1 must be convolved with the
luminosity function, N0 (L), of the sources.

14.1.2 Extragalactic radio sources and radio quasars

Historically, the considerations of the Section 14.1.1 were important because the counts of
radio sources at high flux densities had integral slope β = 1.8, much steeper than expected in
14.1 The cosmological evolution of active galaxies 367

the uniform world models. In the 1960s, this was strong evidence that there must have been
many more radio sources at large redshifts than were predicted by the uniform, isotropic
models, indicating that the source population must have evolved with cosmic epoch.
Pioneering studies of how strong the evolution had to be to account for the observations
were carried out by William Davidson (b. 1924), who showed how extreme the divergence
was between the expectations of the steady state theory and the observed number counts
(Davidson, 1962). He also constructed the first empirical expressions for the evolution of
the population of radio sources (Davidson and Davies, 1964).
As increasing numbers of optical identifications became available, the local luminosity
function of extragalactic radio sources became better defined, and so the nature of the
cosmological evolutionary changes in the radio source population could be quantified in
more detail. By 1966, I was able to impose important constraints on the form of the evolution,
the key features of the successful models being strong evolution of either the luminosities
or comoving number densities of radio sources out to redshifts z ∼ 2–3, beyond which the
strong evolution could no longer continue (Longair, 1966). Otherwise, the convergence of
the counts indicated by the observations of Martin Ryle and Ann Neville (b. 1938) could not
be reproduced (Ryle and Neville, 1962) and the isotropic radio background radiation would
be exceeded. Typically, the luminosities of the most luminous radio sources had to change
with cosmic epoch as L(z) ∝ (1 + z)3 out to redshifts z ∼ 2–3. Because of the steepness
of the radio luminosity function, the corresponding increase in the comoving space density
of sources of a given luminosity was even more extreme, n(z) ∝ (1 + z)5.5 .
The discovery of the radio quasars, and the large numbers of redshifts which rapidly
became available for them, confirmed that there was an excess of these objects at large
redshifts. In 1968, Michael Rowan-Robinson (b. 1942) and Maarten Schmidt independently
developed a procedure, known as the luminosity-volume or V /Vmax test, to show that the
radio quasars were located towards the limits of their observable volumes, implying the
same type of strong evolutionary effect inferred somewhat less directly from the counts
themselves (Rowan-Robinson, 1968; Schmidt, 1968).
From the 1960s onwards, number counts of radio sources were derived at frequencies
throughout the radio waveband, all of them displaying the excess of radio sources observed
at high flux densities, followed by convergence at low flux densities. Examples of the differ-
ential number counts at frequencies from 150 MHz to 8.44 GHz compiled by Jasper Wall
(b. 1942) are shown in Figure 14.2 (Wall, 1996). While the number counts were well estab-
lished by the efforts of many radio astronomers, the determination of how the luminosity
function of the sources changed with cosmic epoch required the redshifts of the sources.
This meant identifying very faint galaxies and quasars reliably, which required very accurate
radio positions, and preferably radio structures, and then measuring the redshifts of these
extremely faint objects.
An example of an ambitious study using large complete samples of radio sources with
redshifts was completed by James Dunlop (b. 1962) and John Peacock (b. 1956) in 1990
(Dunlop and Peacock, 1990). The inferred changes of the radio luminosity function with
cosmic epoch are very strong (Figure 14.3) and can be well represented by a model in
which the luminosity function was shifted to greater radio luminosities at large redshifts.
This increase cannot continue, however, beyond redshifts z ∼ 2–3. There needs to be some
368 14 Cosmological evolution of galaxies and active galaxies

Figure 14.2: The differential, normalised counts of extragalactic radio sources at a wide range of
frequencies throughout the radio waveband (Wall, 1996). The points show the number counts derived
from surveys of complete samples of radio sources. The boxes indicate extrapolations of the source
counts to very low flux densities using the P(D) technique described in Section 12.3. (Diagram
courtesy of Dr Jasper Wall.)

form of cut-off, and there has been considerable debate about how the luminosity function
of the radio sources changes at redshifts greater than z ∼ 2–3. Dunlop pointed out that,
although the precise form of the evolution of the radio luminosity function at large redshifts
is still uncertain, the integrated radio emissivity
 of all the radio sources as a function of
redshift is rather well defined. The quantity L N (L , z) dL attains a maximum at redshifts
z ≈ 2–3 and decreases rapidly at larger redshifts (Dunlop, 1994). Thus, the epoch corre-
sponding to z ∼ 2–3 was the era when the radio quasars and radio galaxies were most
populous.

14.1.3 The radio-quiet quasars

Intensive searches for radio-quiet quasars2 began in 1965 as soon as Sandage announced
their discovery in that year (Sandage, 1965). In these early days, the clue to finding
14.1 The cosmological evolution of active galaxies 369

Figure 14.3: Illustrating the evolution of the luminosity function of extragalactic radio sources with
(a) steep and (b) flat radio spectra as a function of redshift, z, or cosmic epoch. The luminosity
functions shown describe the number of radio sources with different radio luminosities per unit
comoving volume, that is in a coordinate system which expands with the Universe, so that the figure
shows changes over and above the changing density due to the expansion of the Universe (Dunlop
and Peacock, 1990).

radio-quiet quasars among the vastly more populous foreground stars was the fact that
their optical spectra were non-thermal, their continua being of roughly power-law form. As
a result, they displayed ultraviolet and infrared excesses in comparison with stellar spec-
tra. Quasars were sought in catalogues of blue stellar objects, and surveys were made to
search specifically for objects with ultraviolet excesses. The problem with these surveys
was that the objects with ultraviolet excesses might be white dwarfs, and so spectroscopic
confirmation that the objects had large redshifts with the characteristic quasar spectra was
required. In 1970, Alessandro Braccesi (b. 1937) and his colleagues compiled a list of 175
quasar candidates which resulted in a complete sample of 19 quasars to apparent magnitude
B = 18 with confirmed redshifts in an area of 36 square degrees (Braccesi, Formiggini and
370 14 Cosmological evolution of galaxies and active galaxies

Gandolfi, 1970). These were the first observations to show convincingly that the number
counts of optically selected quasars had slope steeper than β = 1.5.
In 1972, Richard Green (b. 1949) and Maarten Schmidt began a major survey using the
Palomar 18-inch Schmidt telescope to find all the bright quasars in a region of 10 700 square
degrees of the northern sky away from the Galactic plane (Schmidt and Green, 1983). The
survey was made in two colours, U and B, and several thousands of stellar objects with
ultraviolet excesses were discovered. Of these, 108 turned out to be quasars which formed
a complete sample with apparent magnitudes less than B = 16.2. Comparison of these
numbers with those of fainter radio-quiet quasars again showed that the counts of optically
selected quasars were significantly steeper than the Euclidean expectation.
The ultraviolet excess technique was only useful out to redshifts z ≤ 2.2 because at
that redshift the Lyman-α emission line is redshifted into the B filter and so the quasars
no longer exhibit ultraviolet excesses. The extension of this technique involved the use of
multicolour photometry to discriminate between stars and objects with the typical spectra
of large redshift quasars. David Koo (b. 1951) and Richard Kron (b. 1951) used (U, J, F, N)
photometry to find radio-quiet quasars with ultraviolet excesses to B = 23 (Koo and Kron,
1982). This survey provided the first evidence for the convergence of the number counts of
radio-quiet quasars at redshifts z ≥ 2.
The technology for undertaking quantitative surveys of large samples of stars and galaxies
improved greatly with the availability of high-speed measuring machines, such as the APM
facility in Cambridge and the COSMOS machine at the Royal Observatory, Edinburgh.
These enabled scans to be made of the very large numbers of objects present on the sky
survey plates taken with the UK 48-inch Schmidt telescope at the Siding Spring Observatory
in Australia. In 1987, the multicolour technique was extended by Stephen Warren (b. 1957)
and his colleagues to four-colour photometry using observations in the U, J, V, R and I
wavebands, providing four colours (U–J, J–V, V–R, R–I) (Warren et al., 1987). Stars lie along
a rather narrow locus in this four-dimensional colour space. By searching for objects which
lay well away from that locus, they found the first quasar with redshift z > 4. In 1991, this
technique was further refined by Michael Irwin (b. 1952) and his colleagues, who realised
that quasars with redshifts greater than 4 could be found by means of two-colour photometry
in the (BJ , R, I) wavebands because the colours of these very large redshift quasars are very
different from those of stars (Irwin, McMahon and Hazard, 1991). The reason for the
success of this approach was that, at these very large redshifts, the redshifted Lyman-α
forest enters the BJ waveband and so strongly depresses their continuum intensities. The
largest redshift found in their survey was z = 4.8. At the same time, Brian Boyle (b. 1960)
and his colleagues used the ultraviolet-excess technique to derive a complete sample of over
400 quasars down to B = 21, all confirmed by spectroscopy (Boyle et al., 1991).
Other techniques were developed to search for quasars. One approach made use of the fact
that the Lyman-α and CIV emission lines are always very strong in the spectra of quasars and
are superimposed upon a roughly power-law continuum energy distribution. If a dispersion
prism, or grating, is used in conjunction with a wide-field telescope, low-resolution spectra
are obtained for each object in the field, and this proved to be a powerful means of discovering
quasars with redshifts z > 2, since these strong lines are then redshifted into the optical
waveband. Pioneering surveys were carried out at the Cerro Tololo Observatory in Chile
14.1 The cosmological evolution of active galaxies 371

Figure 14.4: The evolution of the optical luminosity function for 6000 optically selected quasars in
the redshift range 0.35 ≤ z ≤ 2.3 observed in the 2dF survey carried out at the Anglo-Australian
Telescope (Boyle et al., 2000).

using the Curtis Schmidt telescope and the 4-metre telescope by Malcolm Smith (b. 1942),
Art Hoag (1921–1999) and Patrick Osmer (b. 1943) in the late 1970s (Hoag and Smith,
1977). By the early 1980s, it was known that the optically selected quasars exhibited strong
cosmological evolution, similar in character to the radio quasars. Osmer found that there
seemed to be a lack of large-redshift objects as compared with the expectations of the
evolutionary models, which could explain the redshift distributions and number counts
of objects to a redshift of about 2 (Osmer, 1982). The problem with these surveys was
the question of the completeness of the samples and the many complex selection effects
involved in selecting the candidate objects.
One of the most successful applications of this technique was the survey of Maarten
Schmidt and his colleagues, who used the Palomar 200-inch telescope as a fixed transit
instrument in conjunction with a grism and a large-area CCD camera, which was clocked at
the sidereal rate. In this way, six narrow bands across the sky were scanned both photometri-
cally in the v and i wavebands, as well as spectroscopically, resulting in a total scanned area of
62 square degrees. Of 1660 candidate emission-line objects, 141 were found to be quasars
in the redshift interval 2.0 < z < 4.7 (Schmidt, Schneider and Gunn, 1986, Schneider
et al., 1994).
Perhaps the most ambitious programme to date to determine the optical luminosity
function of radio-quiet quasars and its evolution with cosmic epoch has resulted from
372 14 Cosmological evolution of galaxies and active galaxies

the 2dF quasar survey carried out at the Anglo-Australian telescope in conjunction with
the 2dF Galaxy Redshift Survey. The quasar candidates were selected by the multicolour
technique previously employed by Boyle and his colleagues and resulted in a sample of
over 6000 quasars in the redshift interval 0.35 ≤ z ≤ 2.3 (Boyle et al., 2000). From these
data, they reconstructed the evolving comoving luminosity function of quasars which can
be conveniently described by luminosity evolution of the quasar luminosity function as
L(z) = L 0 exp(7τ ) out to redshift z = 2.3, where τ is the fractional look-back time, τ =
(t0 − t)/t0 , t0 being the present age of the world model and t being the time when the light
was emitted by the quasar (Figure 14.4).
Systematic surveys were made to determine the large-redshift evolution of the radio-quiet
quasars by a number of authors (Warren, Hewett and Osmer, 1994; Kennefick, Djorgovski
and de Carvalo, 1995; Schmidt et al., 1995). Good agreement was found between the results
of these surveys. There are fewer large-redshift optically selected quasars than expected if
the comoving optical luminosity function had remained constant at all redshifts at z ≥ 2 :
the comoving number densities of luminous quasars decreased by a factor of about 5 to 7
over the redshift interval 2 ≤ z ≤ 4. The general consensus was that both the radio sources
and the optically selected quasars show a maximum in their comoving space densities at
redshifts z ∼ 2–3 and decline steeply at both lower and higher redshifts.

14.1.4 X-ray sky surveys

Following the bright X-ray source survey carried out by the UHURU X-ray Observatory in
the early 1970s, the first deep surveys of the X-ray sky were carried out by the Einstein X-ray
Observatory in 1979 (Giacconi et al., 1979). One of the prime motivations of these studies
was to understand the origin of the X-ray background radiation, which was discovered in
Giacconi’s pioneering rocket flight in 1962 and which is remarkably bright (Figure 14.5(a)).
Giacconi and his colleagues found that about 26% of the background intensity could be
attributed to discrete X-ray sources, most of which were associated with active galaxies and
quasars (Giacconi et al., 1979).
Definitive evidence for cosmological evolution of the population of X-ray sources was
derived from the deep surveys carried out by the German–American–British ROSAT satel-
lite, which was launched in 1991. The principal objective of the mission was to carry out a
complete survey of the sky in the X-ray energy band 0.1 to 2.4 keV, and this was success-
fully achieved, resulting in a catalogue of about 60 000 sources and information about their
X-ray spectra in four X-ray ‘colours’. In addition to the sky survey, very deep observations
were carried out in a small region of sky to define the X-ray source counts to the faintest
achievable flux densities. The deep survey was made in the ‘Lockman Hole’, a region of sky
in which the neutral hydrogen column density has a very low value, NH = 5.7 × 1019 cm−2 ,
so that there is minimum photoelectric X-ray absorption by the interstellar gas.
The X-ray source counts were derived by Günther Hasinger (b. 1954) and his colleagues
in two ways from these observations (Hasinger et al., 1993). For the medium deep survey,
particular care was taken to understand the effects of source confusion at low X-ray flux
densities. The very deep survey was analysed using a P(D) analysis similar to that described
in Section 12.3. The resulting differential, normalised X-ray source counts are shown in
14.1 The cosmological evolution of active galaxies 373

(a)

Figure 14.5: (a) An image of the Moon taken by the ROSAT Observatory showing the fluorescent
X-ray emission from the sunlit side of the Moon. On the dark side, the Moon is seen occulting
the diffuse X-ray background emission. (Courtesy of Professor J. Trümper, Max Planck Institute for
Extraterrestrial Physics, Garching.) (b) The differential, normalised counts of faint X-ray sources
observed by the ROSAT X-ray observatory. The dot–dash line is the best-fit source count from the
Einstein Observatory surveys. The dotted area at faint flux densities shows the 90% confidence limits
from the fluctuation analysis of the deepest ROSAT survey in the Lockman Hole (Hasinger et al.,
1993).

Figure 14.5, and these bear a strong resemblance to the differential counts of radio sources
(Figure 14.2). At high X-ray flux densities, S > 3 × 10−14 erg s−1 cm−1 , the differential
counts have slope (β + 1) = 2.72 ± 0.27, and below this flux density the counts converge
with source count slope (β + 1) = 1.94 ± 0.19. The identifications of sources in the medium
deep survey were consistent with a picture in which the X-ray sources follow the same type
374 14 Cosmological evolution of galaxies and active galaxies

of cosmological evolutionary behaviour as the radio galaxies, radio quasars and optically
selected quasars. The background intensity which could be attributed to discrete sources with
flux densities greater than 2 × 10−15 erg cm−2 s−1 amounted to 59% of the total background
intensity in this energy range. Extending the counts to the limit of the deep survey, discrete
X-ray sources can account for about 75% of the background intensity.

14.1.5 Infrared and submillimetre wavebands

The IRAS satellite carried out the first essentially complete sky survey in those infrared
wavebands between 12.5 and 100 μm which can only be observed from space. Among the
many important discoveries of the mission was the realisation that many galaxies are intense
far-infrared emitters, the most intense being the star-burst galaxies. The catalogues of IRAS
galaxies proved to be important cosmologically because they provided complete samples
of galaxies unaffected by obscuration by interstellar dust. As a result of a major effort by
many astronomers, the redshifts of complete samples of IRAS galaxies were measured and
the local luminosity function at 60 μm determined.
Counts of IRAS galaxies were made at 60 μm from the IRAS Point Source Catalogue, the
IRAS Faint Source Survey and a survey in the region of the ecliptic poles (Oliver, Rowan-
Robinson and Saunders, 1992). The normalised differential counts showed that there are
more faint IRAS sources than expected, in the same sense as the counts of radio sources,
X-ray sources and quasars. The counts did not, however, extend deep enough to constrain
the large-redshift behaviour of the source population.
An interesting consequence of the evolution of the IRAS galaxies resulted from the
strong correlation between the radio emission of normal and star-burst galaxies and their far-
infrared emission (Helou, Soifer and Rowan-Robinson, 1985). The proportionality extends
over many orders of magnitude and can be written S(60 μm) = 90S(1.4 GHz), where both
flux densities are measured in janskies (Jy). As a result, it is straightforward to predict the
counts of star-burst and normal galaxies in the radio waveband. It turns out that it is possible
to account for the flattening of the radio source counts at radio flux densities S ≤ 10−3 Jy
seen in Figure 14.2 in terms of the evolution of the population of IRAS galaxies (Rowan-
Robinson et al., 1993). This conclusion is supported by the identification content of the
millijansky radio sources, many of which are blue and have spectra similar to those of
star-burst galaxies (Windhorst, Dressler and Koo, 1987; Windhorst, 1995).
Galaxies undergoing bursts of star formation are not only sources of intense ultraviolet
continuum radiation, but are also strong emitters in the far-infrared waveband because of
the presence of dust in the star-forming regions. The ultraviolet and optical radiation are
absorbed by the dust which reradiates the absorbed energy at the temperature to which
the dust is heated, which is typically about 30–60 K, and so corresponds to emission in
the far-infrared and submillimetre wavebands. In a study of star-forming galaxies in the
Markarian catalogues of ultraviolet-excess galaxies, Joseph Mazzarella (b. 1961) and Vicki
Balzano (b. 1961) found that star-forming galaxies are, on average, stronger emitters in the
far-infrared than in the ultraviolet waveband (Mazzarella and Balzano, 1986). Similarly, in
a sample of star-forming galaxies studied by the International Ultraviolet Explorer, Daniel
Weedman found that most of the galaxies emit much more strongly in the far-infrared rather
14.1 The cosmological evolution of active galaxies 375

(a) (b)

Figure 14.6: (a) The flux density–redshift relations expected for a source with a power-law spectrum
of the form S ∝ ν −1 and for a galaxy with an inverted dust spectrum at a temperature of 45 K, as
observed at a submillimetre wavelength of 850 μm, for world models with 0 = 1 and 0. (b) The
predicted differential normalised counts of sources at 450 and 1100 μm assuming the galaxies have
typical dust spectra and that they have a far-infrared luminosity function given by that of IRAS galaxies
at 60 μm (Saunders et al., 1990). No evolution of the source population is assumed. The predictions
for dust temperatures of 30 and 60 K are shown (Blain and Longair, 1993).

than in the ultraviolet region of the spectrum (Weedman, 1994). These observations had
two consequences. First of all, some of the star-forming galaxies may well be obscured by
dust and so not be present in optical–ultraviolet multicolour surveys. Secondly, it is quite
possible that a significant fraction of the radiation associated with the formation of stars
and the heavy elements was not radiated in the ultraviolet–optical region of the spectrum,
but at far-infrared wavelengths.
One way of estimating the significance of absorption by dust is to carry out observa-
tions in the millimetre and submillimetre wavebands. The dust emission in the millimetre/
submillimetre waveband provides a complementary measure of the star-formation rate to
that inferred from optical observations. What makes this approach feasible observationally
is the fact that the spectrum of dust in the submillimetre waveband is strongly ‘inverted’
with spectra of the form Sν ∝ ν x , where x ∼ 3–4. As a consequence, the ‘K -corrections’
are so large and negative that a typical star-forming galaxy is expected to have essentially
the same flux density, whatever its redshift in the range 1 < z < 10 (Figure 14.6(a)) (Blain
and Longair, 1993). A further consequence is that the number counts are expected to be
inverted, even if there if no evolution of the source population (Figure 14.6(b)).
Deep surveys in the submillimetre waveband became feasible with the construction of
bolometer array receivers, in particular the SCUBA instrument operating on the James
Clerk Maxwell Telescope. Ian Smail (b. 1967), Robert Ivison (b. 1966) and Andrew Blain
(b. 1970) made the first deep submillimetre surveys in the fields of two clusters of galaxies
with SCUBA and discovered a large population of faint background submillimetre sources
(Smail, Ivison and Blain, 1997). What was remarkable was that the excess of faint sub-
millimetre sources was even greater than the largest excess predicted by our most extreme
evolutionary models of 1993. All subsequent observations have confirmed the existence of
this large population of faint submillimetre sources, which are almost all star-burst galaxies
at redshifts z ≥ 1. One of the more remarkable SCUBA observations was the submillimetre
376 14 Cosmological evolution of galaxies and active galaxies

Figure 14.7: The submillimetre image of the Hubble Deep Field observed at a wavelength of 850 μm
by the SCUBA submillimetre camera on the James Clerk Maxwell Telescope. Although only a few
sources are detected, they contribute significantly to the star-formation rate at large redshifts (Hughes
et al., 1998; Serjeant et al., 2003).

image of the Hubble Deep Field (Hughes et al., 1998; Serjeant et al., 2003), the central bright
source being associated with a very faint extremely red object with an estimated redshift
z ≈ 4 (Dunlop et al., 2004) (Figure 14.7). The galaxies associated with the SCUBA sources
are very faint optically, but they make a major contribution to the global star-formation rate
at large redshifts (see Section 14.6).

14.2 The counts of galaxies

The determination of precise counts of galaxies proved to be one of the more difficult areas
of observational cosmology.3 The reasons for these complications are multifold. First of all,
galaxies are extended objects, often with complex brightness distributions, and great care
must be taken to ensure that the same types of object are compared at different magnitude
limits and redshifts. Secondly, the distribution of galaxies is far from uniform, as illustrated
by the huge voids and walls in the local distribution of galaxies (Figure 10.8). Even at the
faintest magnitudes, this ‘cellular’ structure in the distribution of galaxies results in fluctu-
ations in the number counts of galaxies which exceed the statistical fluctuations expected of
14.2 The counts of galaxies 377

Figure 14.8: The counts of faint galaxies observed in the B, I and K wavebands compared with
the expectations of various uniform world models, as well as other models in which various forms
of the evolution of the luminosity function of galaxies with redshift are assumed (Metcalfe et al.,
1996). The galaxy counts follow closely the expectations of uniform world models at magni-
tudes less than about 21, but there is a excess of galaxies in the B and I wavebands at fainter
magnitudes.

a random distribution (Figure 10.9). Finally, the probability of finding galaxies of different
morphological types depends upon the galaxy environment.
A major complication concerns the K -corrections which should be used for galaxies
of different types. There have been remarkably few systematic surveys of the ultraviolet
spectra of normal galaxies, and only in a few cases have images of galaxies in the ultraviolet
waveband been obtained (Giavalisco et al., 1996). The problem is exacerbated by the fact
that the ultraviolet spectra of galaxies can be dominated by bursts of star formation, and
this fact alone makes the comparison of the optical images of galaxies at the present epoch
with those at redshifts of one and greater problematic. On the other hand, as discussed in
Section A13.1, counts of galaxies in the infrared K waveband at 2.2 μm have a number of
advantages and enable the old, stable populations of galaxies to be studied out to redshifts
z ∼ 3.
Number counts in the B (440 nm), I (800 nm) and K (2.2 μm) wavebands from a number
of independent determinations by ground-based optical and infrared telescopes, as well as
deep number counts in the Hubble Deep Field, are shown in Figure 14.8 (Metcalfe et al.,
378 14 Cosmological evolution of galaxies and active galaxies

Ngal

Figure 14.9: The redshift distribution of galaxies in the deep survey of Cowie and his colleagues
in the magnitude interval 22.5 < B < 24 (Cowie et al., 1996; Metcalfe et al., 1996). The solid line
and the long-dashed line show the expected redshift distributions if there were no evolution of the
local luminosity function of galaxies. It can be seen that there is a long high-redshift tail of blue
galaxies, which, from the presence of strong [OII] lines, are inferred to be undergoing rapid star
formation.

1996). The lines labelled ‘No evolution’ show the expectations of uniform world models and
include appropriate K -corrections for the types of galaxy observed in bright galaxy samples.
In the infrared K waveband (2.2 μm), the counts follow reasonably closely the expectations
of uniform world models with q0 ∼ 0–0.5. In contrast, in the B and I wavebands, there is
a large excess of faint galaxies, particularly in the B waveband. The departure from the
expectations of the uniform models sets in at about B = 23, and, at fainter magnitudes, there
is a large excess of faint blue galaxies. The lines on the diagram illustrate the results of
various modelling exercises to account for the observed counts on the basis of models for
the evolution of the stellar populations of spiral and elliptical galaxies.
The nature of the excess of faint blue galaxies is a key cosmological problem. Redshift
distributions for complete samples of galaxies at which the excess of blue galaxies is
observed are required, but this was beyond the capabilities of the generation of 4-metre
class telescopes. Richard Ellis (b. 1950) noted that the complete redshift surveys with such
telescopes are effectively limited to B ≤ 24, I ≤ 22 and K ≤ 18 (Ellis, 1997). An indication
of what is possible with the 8–10-metre optical–infrared telescopes was provided by the
first surveys of faint galaxies carried out with the Keck 10-metre telescope (Cowie, Hu
and Songaila, 1995; Cowie et al. 1996). Figure 14.9 shows the redshift distribution for an
almost complete sample of galaxies in the magnitude interval 22.5 < B < 24, compared
 14.3 The Lyman-α clouds 379

with the expectations of the ‘no evolution’ models. There is a significant excess of blue
galaxies extending to redshifts z = 1.7. According to Lennox Cowie (b. 1950) and his
colleagues, the distribution is composed of a mixture of normal galaxies at small redshifts
plus galaxies undergoing rapid star formation from z = 0.2 to beyond z = 1.7 (Cowie et al.,
1996). There is unquestionably an increase in the numbers of star-forming galaxies with
increasing redshift. At the same time, they found little change in the K-band luminosity
function out to redshifts z ≈ 1, suggesting that most of their stellar populations were already
in place by a redshift of 1.
The nature of the excess of blue galaxies has been at least partially elucidated by stud-
ies with the Hubble Space Telescope. By combining observations of the HST Medium
Deep Survey with those of the Hubble Deep Field, number counts have been determined
for galaxies of different morphological types (Figure 14.10). The high-resolution images
enabled the morphologies of galaxies to be classified into spheroidal/compact, spiral and
irregular/peculiar/merger categories. The counts for the different morphological classes
show that the spheroidal and spiral galaxies more or less follow the expectations of the uni-
form world models, while the objects classified as irregular/peculiar/merger systems show a
distinct excess relative to their populations in bright galaxy samples (Abraham et al., 1996).
These results are consistent with visual impressions of the Hubble Deep Field, which sug-
gests that about 25% of the galaxies seem to be irregular/interacting/merging systems. They
are also consistent with the imaging results of Cowie, David Schade (b. 1953) and their col-
leagues, which indicate that about the same fraction of the blue galaxies in their surveys
have peculiar morphologies (Cowie et al., 1995; Schade et al., 1995).

14.3 The Lyman-α clouds

As soon as the first quasar with redshift z > 2 was discovered in 1965, the radio quasar
3C 9, James Gunn and Bruce Peterson (b. 1941), and independently Peter Scheuer, realised
that the continuum radiation to the short-wavelength side of the redshifted Lyman-α line
was observable in the optical waveband and provided a sensitive test for the presence
of intergalactic neutral hydrogen (Gunn and Peterson, 1965; Scheuer, 1965). The Gunn–
Peterson test makes use of the fact that the cross-section for the Lyman-α transition at
121.6 nm is very large, and so, when the ultraviolet continuum of distant quasars is shifted
to the redshift at which it has wavelength 121.6 nm, the radiation is scattered many times, so
that, if sufficient neutral hydrogen is present at these redshifts, an absorption trough would
be expected to the short-wavelength side of the redshifted Lyman-α line.
The predicted absorption trough was searched for in those quasars with redshifts 1 + z ≥
(330 nm)/λLy , z ≥ 2. No evidence for such a depression to the short-wavelength side of the
Lyman-α line was observed in 3C 9 or in any of the other large-redshift quasars which
were observed over the succeeding years. As larger redshift quasars were discovered, it was
possible to search for a feature to the short-wavelength side of the corresponding line of
neutral helium, HeI, which has rest wavelength 58.4 nm, but no evidence of an absorption
trough was observed. A typical upper limit to the number density of neutral hydrogen atoms
at a redshift z = 3 was NH ≤ 10−5 m−3 , which is very small indeed compared with typical
380 14 Cosmological evolution of galaxies and active galaxies

Figure 14.10: The number–magnitude relation for morphologically segregated samples of galaxies
from the Medium Deep Survey (MDS) and the Hubble Deep Field (Abraham et al., 1996). The
morphological classifications were carried out independently by Richard Ellis, Sidney van den Bergh
(b. 1929) and by an automated machine-based classification algorithm. The dotted line in (a) shows
the total counts of galaxies from a field observed by the Keck Telescope. The solid lines show the
expected counts of the different morphological classes, assuming their properties do not change with
cosmic epoch (Glazebrook et al., 1995).

cosmological baryonic densities. Therefore, if there were significant amounts of hydrogen

in the intergalactic medium, it would have to be very highly ionised.
Although there was little evidence for continuum absorption due to diffuse neutral
gas, as more large-redshift quasars were discovered, various types of absorption lines
and absorption-line systems were observed.4 The first quasar to be discovered with a
14.3 The Lyman-α clouds 381

rich absorption-line spectrum was 3C 191 with redshift z = 1.953, which was observed
by the Burbidges, Roger Lynds (b. 1928) and Alan Stockton (b. 1942) (Burbidge, Lynds and
Burbidge, 1966; Stockton and Lynds, 1966). The Burbidges found that the absorption-
line systems were mostly observed in quasars with redshifts greater than about z = 1.9
(Burbidge and Burbidge, 1967). Some of the lines could be associated with strong reso-
nance absorption lines, such as Lyman-α and CIV, which enabled an absorption redshift to
be measured, but others remained unidentified. Soon, examples of multiple-absorption-line
systems of the common elements were found in large-redshift quasars (Bahcall, Greenstein
and Sargent, 1968; Burbidge, Lynds and Stockton, 1968).
These studies benefitted enormously from the electronic revolution in astronomical spec-
troscopy during the 1970s. For example, the image photon counting system developed by
Alexander Boksenberg revolutionised these studies and enabled superb absorption-line
spectra to be obtained.5 Similar spectrographs were eventually built for most of the 4-metre
class telescopes and ultimately for the generation of 8- to 10-metre telescopes. These studies
culminated in absorption-line spectra of the extraordinary quality seen in Figure 14.11. The
study of such spectra became a major field in its own right, and these developments impacted
a wide range of different astrophysical and cosmological topics, including the physics of the
intergalactic gas, the evolution of galaxies and the history of metal enrichment in galaxies.6
While some of the absorption-line systems are associated with matter ejected from the
quasar, there are two broad classes of absorption-line system which are of direct importance
for cosmological studies. By far the most common are those belonging to the Lyman-α
forest, which dominates the spectra of large-redshift quasars, such as that illustrated in
Figure 14.11. These systems have neutral hydrogen column densities in the range 1016 ≤
NHI ≤ 1021 m−2 . In those systems with NHI ≥ 1019 m−2 , evidence for low abundances of the
heavy elements, corresponding to about 1% of the solar value, is usually found (Boksenberg,
1997). The absorbers responsible for the Lyman-α forest are interpreted as intergalactic
clouds, containing largely unprocessed primordial material. There must, however, be some
mild enrichment of these primordial clouds, the chemical abundances of which are similar
to those of halo stars in our Galaxy.
In contrast, the rarer Lyman limit and damped Lyman-α systems have much larger col-
umn densities, 1021 ≤ NHI ≤ 1026 m−2 , and have correspondingly larger optical depths for
Lyman-α absorption. The Lyman-limit systems are those with column densities in the range
1021 ≤ NHI ≤ 2 × 1024 m−2 . A continuum break is observed at the redshifted wavelength
of the Lyman limit at 91.2 nm and the spectra display the corresponding absorption lines
of the common elements. Jacqueline Bergeron (b.1942) showed that these systems can be
associated with the extended gaseous haloes of galaxies, similar to those observed about
nearby galaxies (Bergeron, 1988). The abundances of the heavy elements are less than 10%
of their cosmic abundances, consistent with the inference that the haloes are very extensive,
∼50–100 kpc, and so the gas in these regions is not expected to be as enriched as the gas
in the disc of a galaxy. These systems probably contain most of the mass density of neutral
gas in the Universe (Lanzetta, Wolfe and Turnshek, 1995).
The damped Lyman-α systems are those with the greatest column densities, 2 × 1024 ≤
NHI ≤ 1026 m−2 . In 1988, Arthur Wolfe (b. 1939) showed that these can be convincingly
associated with galactic discs, and, in his pioneering paper, he identified them with the
382 14 Cosmological evolution of galaxies and active galaxies

(a)

(b)
Figure 14.11: (a) The spectrum of the quasar Q1422+2309 at an emission redshift z = 3.62 showing
the remarkable ‘Lyman-α forest’ to the short-wavelength side of the strong redshifted Lyman-α emis-
sion line, which has observed wavelength 560 nm. (Courtesy of Dr W. L. W. Sargent, from Boksenberg
(1997).) (b) To the long-wavelength side of the Lyman-α line, the spectrum is very much smoother,
only weak metal absorption lines being observed, most of them associated with the CIV absorption
line.

progenitors of the stellar discs of present-day spiral galaxies (Wolfe, 1988). This is a natural
assumption, since stars form in the coldest regions of the interstellar gas. It was reinforced
by the important analysis of Robert Kennicutt (b. 1951), who made detailed observations of
a number of nearby spiral galaxies and showed that active star formation only takes place in
their discs if the column density of neutral gas exceeds 2 × 1024 m−2 (Kennicutt, 1989). It
is striking that this criterion is identical to the lower limit at which absorbers are identified
as damped Lyman-α systems. Kennicutt argued that this criterion is consistent with the
stability criterion for rotating thin gaseous discs. Thus, the damped Lyman-α systems can
provide information about the chemical evolution of disc galaxies.
An important aspect of the distribution of both classes of absorption system is their
variation with redshift. It is convenient to parameterise the variation of the number density
of absorbers with redshift by a power-law distribution of the form

N (z) dz = A(1 + z)γ dz. (14.2)

Typically, it is found that, for the Lyman-α forest systems, A ≈ 10 and γ = 2–3 whereas,
for the Lyman-limit systems, A ≈ 1 and γ ∼ 1. These variations with redshift can be
compared with the expected distribution if the properties of the absorbers were unchanging
 14.4 Abundances of elements in Lyman-α absorbers 383

with cosmic epoch, that is if the absorbers had the same proper cross-sections and constant
comoving number density.7 If 0 = 1, N (z) ∝ (1 + z)1/2 and, if 0 = 0, N (z) ∝ (1 + z).
Thus, the observed number density of Lyman-α forest absorbers changes more rapidly with
increasing redshift than expected according to the uniform absorber model. The sense of
the evolution is that there were more Lyman-α forest systems at large redshifts as compared
with low redshifts. On the other hand, the Lyman-limit systems seem to show little variation
with redshift other than that expected if their cross-sections and comoving number densities
remained unchanged with cosmic epoch.

14.4 The abundances of elements in Lyman-α absorbers

Michael Fall (b. 1951) has emphasised that one of the great attractions of using absorption
lines in Lyman-α absorbers to estimate the relative abundances of any species along the line
of sight to distant quasars is that, provided the absorbers are randomly oriented, average
relative abundances of different species, such as atoms, ions and molecules, can be found
independent of the structures or clumpiness of the clouds (Fall, 1997). There will generally
be a distribution of column densities, N (NX ) dNX , for any species X, and this is expected to
change with cosmic epoch. Gas is condensed into stars, and the interstellar gas in galaxies
is enriched as a result of nucleosynthesis in stars and the subsequent recycling of processed
material to the gas. The beauty of this result is that, if we average over many lines of sight,
we take averages over all systems in all orientations which contribute to the mean density
parameter of that species, X (z), at that redshift. Furthermore, we can determine how the
global metallicity, Z = m /g , and other relative abundances change with redshift by
taking appropriate ratios.
The damped Lyman-α systems are of particular importance in understanding the evo-
lution of neutral gas and the build-up of the heavy elements with cosmic epoch (Lanzetta
et al., 1995; Storrie-Lombardi, McMahon and Irwin, 1996; Pettini et al., 1997). The pic-
ture which emerges is that, at redshifts z ∼ 3, the comoving density parameter in neutral
hydrogen is HI ≈ (1 − 2) × 10−3 h −1 and this decreases with decreasing redshift until, at
z = 0, its value is only about 2 × 10−4 , a value consistent with independent measures of
the amount of neutral hydrogen present in galaxies and their environs at the present epoch
(Figure 14.12). It is striking that the density parameter in neutral gas at a redshift z = 2.5 is
of the same order of magnitude as the density parameter corresponding to the visible mass
of galaxies at the present epoch, which is indicated by the hatched area in Figure 14.12.
The big advantage of studying the damped Lyman-α systems is that they have such large
column densities that relatively rare species can be used to probe the chemical abundances
of the elements. Max Pettini (b. 1949) and his colleagues, for example, used observations of
singly ionised zinc, Zn+ or ZnII, which has a number of advantages as a tracer of the overall
abundance of the heavy elements (Pettini et al., 1997). Zinc shows little affinity for dust
and is predominantly in the form of Zn+ in HI regions. Although zinc is a relatively rare
species, the Solar System value corresponding to [Zn/H] = 3.8 × 10−8 , this is an advan-
tage since the absorption lines are optically thin and so accurate column densities can be
determined.
384 14 Cosmological evolution of galaxies and active galaxies

Figure 14.12: The evolution of the mass density of neutral gas as a function of redshift as determined
by the mass density of damped Lyman-α absorbers in the spectra of distant quasars (Storrie-Lombardi
et al., 1996). The circles show the estimates of HI corrected for the effects of dust extinction and the
boxes show the corrected values which take account of neutral hydrogen not associated with damped
Lyman-α systems. The hatched area indicates estimates of the density parameter in stars, that is the
mass associated with the visible light of galaxies, at the present epoch.

In Figure 14.13, the abundance of zinc relative to hydrogen is shown for a number of
large-redshift damped Lyman-α clouds. Recalling that we should take averages over all
systems at a given redshift, it can be seen that the heavy element abundances have built
up from values of only about 10% of the present solar abundances at a redshift z ≈ 2
to Z ≈ Z  at the present epoch. Over the same redshift interval, the mean dust-to-gas
ratio has increased by roughly the same factor, while the mean dust-to-metals ratio has
remained roughly constant (Fall, 1997). These results strongly suggest that a large fraction
of the build-up of the heavy elements in galaxies took place over the redshift interval
1 < z < 3.

14.5 The Lyman-break galaxies

In an important paper of 1987, Simon Lilly and Lennox Cowie first showed how the rate of
formation of heavy elements could be inferred from the flat blue continuum spectra of star-
forming galaxies and that these estimates are independent of the choice of cosmological
model (Lilly and Cowie, 1987). A prolonged burst of star formation has a flat intensity
14.5 The Lyman-break galaxies 385

Figure 14.13: The [Zn/H] abundance in 24 damped Lyman-α systems at large redshifts relative to the
solar value, which is indicated by the dashed line (Pettini et al., 1997). The usual convention is used
of plotting the logarithm of the abundance ratio [Zn/H] on the ordinate.

spectrum at wavelengths longer than the Lyman limit at 91.2 nm, as illustrated by the
model star-bursts synthesised by Gustavo Bruzual (b. 1949) using his spectral synthesis
codes to predict the spectra of galaxies at different phases of their evolution. Figure 14.14
shows the spectrum of a star-burst galaxy which lasts 12 Gyears, as observed at different
ages, assuming that the star-formation rate is constant with the same Salpeter initial mass
function. The flatness of the spectrum is associated with the fact that, although the most
luminous blue stars have short lifetimes, they are continuously being replaced by new stars.
To a good approximation, it can be assumed that the spectrum of the star-forming galaxy
can be described by a power law, I (ν) ∝ ν −α , with α = 0 at wavelengths λ > 91.2 nm,
and zero intensity at shorter wavelengths. The intensity of the flat part of the spectrum
is directly proportional to the rate of formation of heavy elements in the star-burst since
the conversion of hydrogen into helium is the essential first stage in the synthesis of the
heavy elements in the central regions of the massive stars which are primarily responsible
for the ultraviolet continuum. The important result found by Cowie and Lilly was that
the background intensity associated with flat-spectrum star-forming galaxies is directly
related to the rate at which elements are formed, and this rate, as a function of redshift, is
independent of the cosmological model.8
Cowie, Lilly and their colleagues undertook deep multicolour surveys to discover flat-
spectrum star-forming galaxies at large redshifts. Their survey was successful in finding such
galaxies, which have roughly equal intensities in the U, B and V wavebands, the background
due to such objects amounting to about 10−24 W m−2 Hz−1 sr−1 . They interpreted this result
as meaning that a significant fraction of the heavy elements, about 1.5 × 10−31 kg m−3 , must
have been synthesised at redshifts of about unity (Cowie et al., 1988).
In a remarkable pioneering set of observations, Charles Steidel (b. 1962) and his col-
leagues extended the multicolour technique to find star-forming galaxies at redshifts z > 3.
The objective was to search for star-forming galaxies in which the Lyman limit, clearly seen
in the models of star-burst galaxies in Figure 14.14, is redshifted into the optical waveband.
The technique is illustrated in Figure 14.15(a), in which the spectrum of a star-forming
galaxy at redshift z = 3.15 is observed through carefully chosen filters in the ultraviolet,
386 14 Cosmological evolution of galaxies and active galaxies

Figure 14.14: Synthetic spectra of a 12 Gyear star-burst with constant star-formation rate as observed
at the ages indicated. A Salpeter initial stellar mass function, N (M) dM ∝ M −2.35 dM, is assumed with
cut-offs at 75M and 0.08M . The spectra were generated by Gustavo Bruzual using his evolutionary
synthesis programmes (White, 1989).

blue and red spectral regions. The signature of such a star-burst galaxy is that its image
should be roughly equally bright in the two longer wavebands, but should not be present in
the ultraviolet waveband. Steidel’s original intention was to use this technique to identify
the galaxies responsible for the Lyman-limit absorption systems in the spectra of distant
quasars, and this programme turned out to be remarkably successful (Steidel and Hamil-
ton, 1992). It was soon found, however, that the technique was also a remarkably effective
means of discovering star-forming galaxies in the general field at z > 3 (Steidel, 1998).
Observations of the Hubble Deep Field have proved to be ideal for exploiting this approach
because, in addition to very precise photometry in four wavebands spanning the wavelength
range 300 < λ < 900 nm, high-resolution optical images have enabled the morphologies
of these galaxies to be studied. HST images in four wavebands of one of the Lyman-break
galaxies in the Hubble Deep Field are shown in Figure 14.15(b).
This technique has been successful in identifying star-forming galaxies at redshifts z ≥ 3.
Spectroscopic confirmation of their large redshifts has been obtained for almost 200 of these
galaxies as a result of observations with the Keck telescope. An example of the redshift
distribution found for the galaxies in Steidel’s surveys is shown in Figure 14.16, in which
the observed redshift distribution more or less follows that expected according to the colour
selection criteria. There is a large ‘spike’ at z = 3.09, which is probably associated with
the large-scale clustering of galaxies at that redshift. Thus, not only the properties of these
star-forming galaxies, but also their three-dimensional spatial distribution, can be studied.
 14.6 The global star-formation rate 387

Figure 14.15: (a) Illustrating how multicolour photometry can be used to discover star-forming galax-
ies at redshifts z > 3, at which the Lyman limit is redshifted into the optical region of the spectrum
(Steidel, 1998). (b) Images of a distant star-forming galaxy present in the Hubble Deep Field. From
left to right, the images were taken at red (I), green (V), blue (B) and ultraviolet (U) wavelengths.
Because the Lyman limit has been redshifted beyond the U waveband, no image of the galaxy appears
on the U image (Macchetto and Dickinson, 1997).

These data can be analysed to determine the average star-formation rates within these
galaxies, and they have turned out to be not so different from the typical overall star-
formation rate in galaxies at the present epoch (see Section 14.6 and Figure 14.17). In
addition, it has been possible to learn a great deal about the astrophysical nature of these
objects from spectroscopic observations with the Keck telescope, for example their internal
kinematics, their stellar populations and so on.

14.6 The global star-formation rate

In an important paper of 1996, Piero Madau (b. 1958) and his colleagues attempted to
synthesise all the information discussed in Sections 14.2 to 14.5 into a unified picture for
388 14 Cosmological evolution of galaxies and active galaxies

Figure 14.16: The redshift distribution of Lyman-break galaxies in a single 9 × 18 area of sky, all of
which have spectroscopically confirmed redshifts. The differently shaded histograms reflect slightly
different selection criteria. The dotted curve shows the expected redshift distribution defined by the
colour selection criteria for the complete sample of objects. The ‘spike’ at z = 3.09 is significant at
the 99.9% confidence level (Steidel, 1998).

the evolution of the global star-formation rate with cosmic epoch (Madau et al., 1996). By
global star-formation rate, we mean the rate of star formation integrated over all types of
galaxies as a function of cosmic epoch, or redshift. In Figure 14.17, the results are presented
as the number of solar masses per year per cubic megaparasec, derived from an analysis of a
large body of data. The sources of these include the global star-formation rate at the present
epoch (Gallego et al., 1995), the 280 nm continuum emission of star-forming galaxies at
z < 1 (Lilly et al., 1995), the Hα emission of star-forming galaxies in the redshift interval
1 < z < 2 (Connolly et al., 1997) and the Lyman-limit galaxies, including those observed
in the Hubble Deep Field, which set limits to the star-formation rate at redshifts z > 3
(Madau et al., 1996; Madau, Pozzetti and Dickinson, 1998).
A literal interpretation of Figure 14.17 suggests that the cosmic rate of star formation
peaked in the redshift interval 1 ≤ z ≤ 2, at which time it was about an order of magnitude
greater than it is at the present epoch. At larger redshifts, z ≥ 3, the star-formation rate
was less than this maximum, and not so different from that occurring at the present epoch,
consistent with the observations of Steidel and his colleagues.
14.6 The global star-formation rate 389

Figure 14.17: Comparison of the observed variation of the rate of star formation with cosmic epoch
with that predicted on the basis of the absorption history of the interstellar gas in galaxies (Madau
et al., 1998). Both the observed rates and the model predictions have been corrected in a self-consistent
manner for dust extinction.The model predictions are due to Pei and Fall (1995).

In 1995, Michael Fall and Yichuan Pei (b. 1962) developed a simple formalism for
interpreting the emission and absorption histories of star formation and the build-up of
the chemical elements with cosmic epoch (Pei and Fall, 1995). This approach was based
upon the equations of cosmic chemical evolution, which were originally derived by Beatrice
Tinsley (1941–1981) for the chemical evolution of galaxies (Tinsley, 1980). Many more
details of these types of calculation and of the physics of the chemical evolution of stars
and galaxies are given by Bernard Pagel (b. 1930) (Pagel, 1997).
In the approach adopted by Pei and Fall, all mention of galaxies disappears, all that
remains being the global averages of the density parameters for the different constituents
of the Universe. The density parameter in stars, s , is zero when star formation begins
and builds up to the mean value s (t0 ) ≈ (4–8) ×10−3 by the present epoch. At the same
time, the density parameter in gas, g , initially comprised 100% of the baryonic matter in
galaxies and has decreased to only about 5% of its initial value by the present epoch. The
density parameter in heavy elements, m , was initially zero and has built up to about 1% of
the baryonic mass by the present epoch. Finally, dust was also formed as the abundance of
the heavy elements built up. A typical figure would be that about 50% of the heavy elements
in the interstellar media of galaxies is in the form of dust.
390 14 Cosmological evolution of galaxies and active galaxies

The objective of the approach is to relate the absorption history of the Universe, as defined
by the build-up of the heavy elements and the decrease in the abundance of neutral hydrogen,
to the emission history, which provides direct information about the global star-formation
rate. To give a flavour for this approach, the first equation describes the conservation of
mass:
d  
g + s = 
˙ f, (14.3)
dt
where  ˙ f is the rate of infall into, or of the expulsion of baryonic matter from, galaxies.
If this term is zero, equation (14.3) states that  ˙ g = − ˙ s , that is the rate at which gas is
depleted is equal to the rate at which mass is condensed into stars.
The second equation describes the rate at which the mass of heavy elements in diffuse
gas changes with time. The metallicity of the gas at any epoch is defined to be Z = m /g ,
and so the rate of change of m is given by

dm d   ds ds

= Z g = y −Z ˙ f.
+ Z f (14.4)
dt dt dt dt
The first term on the right-hand side, y ds /dt, describes the rate of increase of the mass
of heavy elements associated with the rate of star formation. The quantity y is called the
yield. This term means that, in the time dt, a mass of stars ds is formed per unit comoving
volume and these eventually return a mass of heavy elements, y ds , to the interstellar
medium. In the spirit of this analysis, the yield is assumed to be independent of cosmic
epoch. The second term on the right-hand side of equation (14.4) describes the loss of
heavy elements because of the formation of stars from gas which has already attained a
metallicity Z . In the time dt, the loss of heavy elements per unit comoving volume is given
by −dm = −Z ds . The third term on the right-hand side represents enhancement of
the heavy element abundance by the infall of baryonic material from intergalactic space,
assumed to have metallicity Z f . This term takes account of the fact that the primordial gas
might have been enriched by early generations of star formation.
Pei and Fall use these equations to predict the emission history from the absorption
history, taking account of the effect of dust in a self-consistent manner. The results of their
calculations are shown in Figure 14.17 by the pair of continuous lines which bound the
observed star-formation rate determined by Madau and his colleagues. It is notable that
their analysis is sensitive to dust obscuration of the quasars, which are the prime source of
information on the absorption-line history of metal enhancement, as Pei and Fall (1995)
demonstrate.
Another concern is the completeness of the information on the global star-formation rate,
which, in Figure 14.17, was entirely determined by optical observations. Fortunately, the
advances in submillimetre astronomy described in Section 14.1.5 have enabled estimates
to be made of the incompleteness of the optical surveys. The observations of the Hubble
Deep Field and other areas by the SCUBA array have shown that a population of obscured
star-forming galaxies exists at large redshifts and that these sources increase the global star-
formation rate significantly above the rates shown in Figure 14.17 (Hughes et al., 1998).
Subsequent analyses of the large-redshift global star-formation rate, including improved
 Notes to Chapter 14 391

corrections for obscured star formation, have suggested that the rate may well be roughly
constant at redshifts beyond the peak at z ∼ 1–2.

14.7 Conclusion

The developments described in this chapter have been among the most important in astro-
physical cosmology in the second half of the twentieth century. Most of the developments
described above have concerned the evolution of discrete objects once they have formed.
The major challenge is to relate these to theories of the origin of galaxies and larger-scale
structures, which are the subject of Chapter 15.

Notes to Chapter 14

1 I have given many more details of the observations and data involved in the studies described in
this chapter in Part IV of Malcolm Longair, Galaxy Formation (Berlin: Springer-Verlag, 1998).
2 Lodewijk Woltjer (b. 1930) has presented a concise summary of the relative merits of the different
selection procedures for finding radio-quiet quasars and other types of active galaxy in Woltjer,
Phenomenology of active galactic nuclei, in Active Galactic Nuclei by Blandford, R. D. Netzer,
H. and Woltjer, L., eds T. Courvoisier and M. Mayor (Berlin: Springer-Verlag, 1990).
3 The review of counts of faint galaxies by Richard Ellis (Ellis, 1997) describes vividly the compli-
cations in determining precisely and in interpreting the counts of galaxies.
4 A contemporary account of the early years of quasar studies is given in the book by Geoffrey and
Margaret Burbidge (Burbidge and Burbidge, 1967). The book includes much of the early history
of quasar studies, both the observations and interpretations of the data. The puzzles presented by
the absorption-line spectra are surveyed as they were understood in 1967.
5 Boksenberg’s image photon counting system was a superb instrument which was particularly
effective in astronomical spectroscopy. During the 1970s, it was so superior to all other instruments
that he was awarded very large amounts of time on the premier telescopes available at that time. In
consequence, he and his support team, affectionately known as Boksenberg’s Flying Circus, were
involved in many of the most important programmes in the spectroscopy of faint objects.
6 The nature and properties of these absorption-line systems are vast subjects; for many more details,
reference should be made to the following volumes: J. C. Blades, D. Turnshek and C. A. Norman,
eds, QSO Absorption Lines: Probing the Universe (Cambridge: Cambridge University Press,
1988); G. Meylan, ed., QSO Absorption Lines (Berlin: Springer-Verlag, 1995); and N. R. Tanvir,
A. Aragón-Salamanca and J. V. Wall, eds, The Hubble Space Telescope and the High Redshift
Universe (Singapore: World Scientific Publishing Company, 1997).
7 I have derived these relations in Section 18.3.2 of Longair, Galaxy Formation.
8 I have derived this result in more detail in Section 18.4.1 of Longair, Galaxy Formation.
15 The origin of galaxies and
the large-scale structure
of the Universe

Galaxies and clusters of galaxies are complex systems, but the aim of the cosmologist is not
to explain all their detailed features. Rather, it is to explain how large-scale structures formed
in the expanding Universe in the sense that, if δρ is the enhancement in density of some
region over the average background density ρ, the density contrast δρ/ρ reached amplitude
1 from initial conditions which must have been remarkably isotropic and homogeneous.
Once the initial perturbations have grown in amplitude to δρ/ρ ∼ 1, their growth becomes
non-linear and they rapidly evolve towards bound structures in which star formation and
other astrophysical process lead to the formation of galaxies and clusters of galaxies as
we know them. The cosmologist’s objectives are therefore twofold – to understand how
density perturbations evolve in the expanding Universe and to derive the initial conditions
necessary for the formation of structure in the Universe.1
Galaxies, clusters of galaxies and other large-scale structures of our local Universe must
have formed relatively late in the history of the Universe. The average density of matter in
the Universe today corresponds to a density parameter 0 ∼ 0.3. The average densities of
gravitationally bound systems, such as galaxies and clusters of galaxies, are much greater
than this value, typically their densities being about 106 and 1000 times greater than the
mean background density, respectively. Superclusters have mean densities a few times the
background density. Therefore, the density contrasts δρ/ρ for galaxies, clusters of galaxies
and superclusters at the present day are about ∼ 106 , 1000 and a few, respectively. Since the
average density of matter in the Universe, ρ, changes as R −3 = (1 + z)3 , where R is the scale
factor and z is redshift, it follows that typical galaxies must have had δρ/ρ ∼ 1 at a redshift
z ≈ 100. They could not have separated out as discrete objects at larger redshifts, or else
their mean densities would be greater than those observed at the present epoch. The same
argument applied to clusters and superclusters indicates that they could not have separated
out from the expanding background at redshifts greater than z ∼ 10 and 1, respectively.
Therefore galaxies and larger-scale structures must have separated out from the expand-
ing Universe at redshifts significantly less than 100. This epoch occurred long after the epoch
of recombination at z ≈ 1000 when the primordial plasma recombined (see Section 15.2.1)
and well into the matter-dominated phase of the standard Big Bang. Thus, these structures
were not formed in the inaccessibly remote past, but at redshifts which are accessible to
observation.

392
 15.1 Gravitational collapse structure formation 393

15.1 Gravitational collapse and the formation of structure in the

expanding Universe

The standard Friedman world models are isotropic and homogeneous, and so the diversity
of structure we observe in the Universe about us is absent. The next step is to include
density perturbations of small amplitude, δρ/ρ, into the Friedman models and study their
development under gravity. This problem was solved by James Jeans in 1902 for the case of
a stationary medium (Jeans, 1902). He derived the dispersion relation, the relation between
wavenumber, k, and angular frequency, ω, for perturbations in a medium of density ρ0 and
pressure p0 :
ω2 = cs2 k 2 − 4π Gρ0 , (15.1)
where cs is the speed of sound in the medium and k = 2π/λ, λ being the wavelength
of the perturbation.2 Thus, for small wavelengths and large wavenumbers, the right-hand
side is positive and the perturbations behave as propagating sound waves. If the right-
hand side is negative, gravitational collapse occurs and the density perturbations grow
exponentially with characteristic timescale τ ∼ (Gρ0 )−1/2 in the limit of long wavelengths.
The criterion for collapse is thus that the size of the perturbation should exceed the Jeans’
length, λJ = cs /(Gρ0 /π)1/2 , meaning that the gravitational force of attraction by the matter
of the perturbation exceeds the pressure gradient which resists collapse. This criterion is
important in studies of star formation.
The analysis was repeated for the case of an expanding medium in the 1930s by Georges
Lemaı̂tre and by Richard Tolman (1881–1948) for the case of spherically symmetric pertur-
bations (Lemaı̂tre, 1933; Tolman, 1934), and the solution for the general case was found by
Evgenii Lifshitz (1915–1985) in 1946 (Lifshitz, 1946). Lifshitz found that the condition for
gravitational collapse is exactly the same as the Jeans’ criterion at any epoch, but, crucially,
the growth of the density contrast is no longer exponential but only algebraic (see Section
A15.1). In the case of a matter-dominated Universe with the critical density 0 = 1, the
density contrast grows linearly with the scale factor R, that is
δρ/ρ ∝ R = (1 + z)−1 ∝ t 2/3 . (15.2)
This is one of the most important equations in astrophysical cosmology. A similar result
is found for radiation-dominated universes, δρ/ρ ∝ R 2 = (1 + z)−2 . For other Friedman
world models, the growth rate given by equation (15.2) is a good approximation for redshifts
z > −10 , but, at smaller redshifts, the perturbations are stabilised. The implication of these
results is that the fluctuations from which the large-scale structure of the Universe formed
cannot have grown from infinitesimal statistical perturbations in the number density of
particles – they must have developed from perturbations of finite amplitude. For this reason,
Lemaı̂tre, Tolman and Lifshitz inferred that galaxies were not formed by gravitational
collapse.
Other authors took the view that finite perturbations should be included in the initial con-
ditions from which the Universe evolved and then the evolution of the perturbation spectrum
with cosmic time should be studied in detail. The Moscow school, led by Yakov Zeldovich,
Igor Novikov and their colleagues, and James Peebles at Princeton pioneered the study of
394 15 The large-scale structure of the Universe

the development of structure in the Universe in the 1960s. If perturbations on a particular

physical scale were tracked backwards into the past, it was found that, at some large redshift,
the scale of the perturbation was equal to the horizon scale of the Universe at that time, that is
r = ct, where t is the age of the Universe. In 1964, Novikov showed that, to form structures
on the scales of galaxies and clusters of galaxies, density perturbations on the scale of the
horizon had to have amplitude δρ/ρ ∼ 10−4 in order to guarantee the formation of galaxies
by the present epoch (Novikov, 1964). These perturbations are certainly not infinitesimal,
and their origin had to be ascribed to processes occurring in the very early Universe.

15.2 The thermal history of the Universe

The discovery of the cosmic microwave background radiation by Penzias and Wilson in
1965 had an immediate impact upon these studies, since the thermal history of the matter
and radiation content of the Universe could be determined in detail. In turn, this enabled the
variation of the speed of sound with cosmic epoch to be determined; this is needed to apply
the Jeans’ stability criterion on different physical scales at different cosmic epochs. The
thermal history of the Big Bang established by Alpher and Herman (Alpher and Herman,
1948) could now be placed on a firm observational foundation. In the simplest picture,
the temperature of the background radiation changes with scale factor as T = T0 /R =
T0 (1 + z), exactly the same as the adiabatic expansion of a photon gas, but there are some
important elaborations (Figure 15.1).

15.2.1 The epoch of recombination and the last scattering surface

At a redshift z = 1500, the background radiation attained a temperature T = T0 (1 + z) ≈
4000 K, at which there were sufficient photons in the Wien region of the Planck distribution to
ionise all the intergalactic hydrogen.3 This epoch is referred to as the epoch of recombination
since the hydrogen was fully ionised at earlier cosmic epochs.
The details of the process of recombination of the primordial plasma as the Universe
expands and cools are important in understanding the origin of the temperature fluctuations
in the cosmic microwave background radiation. These recombination calculations were first
carried out by James Peebles and by Yakov Zeldovich, Vladimir Kurt (b. 1933) and Rashid
Sunyaev independently in the late 1960s (Peebles, 1968; Zeldovich, Kurt and Sunyaev,
1968). Recombinations to the ground state of hydrogen release Lyman continuum photons
which can immediately reionise any neutral hydrogen atoms which have recombined. If
the recombination takes place to an excited state, the liberated photon excites a neutral
hydrogen atom to an excited state from which it can be readily ionised. Therefore, the
recombination rate is determined by the rate at which Lyman-α photons are destroyed by
the rare two-photon process. The result is that the process of recombination takes place over
a finite redshift range. Detailed calculations show that the pre-galactic gas was 50% ionised
at a redshift z r ≈ 1500. At earlier epochs, z ≈ 6000, helium was 50% ionised and rapidly
became fully ionised before that time.
15.2 The thermal history of the Universe 395

Figure 15.1: A summary of the thermal history of the cosmic microwave background radiation,
showing some of the important events at different cosmic epochs. This diagram is a simplified version
of Figure 12.1, which was first derived by Alpher and Herman in 1948 (Alpher and Herman, 1948;
Wagoner et al., 1967).

The most important consequence is that, at redshifts greater than about 1000, the Uni-
verse became opaque to Thomson scattering. This is the simplest scattering process, which
impedes the propagation of photons from their sources to the Earth through an ionised
plasma, the photons being scattered without loss of energy by free electrons. The inter-
galactic gas was essentially fully ionised at z > 1000, and so the optical depth at larger
redshifts is given by

B
τT = 0.035 1/2
hz 3/2 . (15.3)
0

For reasonable values of B , 0 and h, τT  1. Therefore, the Universe beyond a redshift

of 1000 is unobservable. Any photons originating from larger redshifts were scattered many
times before they propagated to the Earth, and consequently all the information they carry
about their origin is lost. There is therefore a photon barrier or last scattering surface at a
redshift of 1000, beyond which we cannot obtain information directly using photons. This
396 15 The large-scale structure of the Universe

is the surface on which the ripples in the cosmic microwave background radiation were
imprinted by matter perturbations.

15.2.2 The radiation-dominated era

If the matter and radiation were not thermally coupled, they would cool independently,
the hot gas having a ratio of specific heat capacities γ = 5/3 and the radiation γ = 4/3,
corresponding to variations of the temperature of the matter and radiation with scale factor
R as Tm ∝ R −2 and Tr ∝ R −1 , respectively. The cosmic microwave background radiation
provides by far the greatest contribution to the energy density of radiation in intergalactic
space. Therefore, comparing the inertial mass density in the radiation and the matter as a
function of redshift, we obtain

ρr aT 4 (z) 2.48 × 10−5 (1 + z)

= = . (15.4)
ρm 0 ρc (1 + z)3 c2 0 h 2
Hence, the Universe became radiation- rather than matter-dominated at redshifts z  4 ×
104 0 h 2 . It would be expected that the matter would cool more rapidly than the radiation,
and this is indeed what is expected to take place during the post-recombination era. This
is not the case, however, during the pre-recombination and immediate post-recombination
eras because the matter and radiation are strongly coupled by Compton scattering. The
optical depth of the pre-recombination plasma for Thomson scattering is so large that the
small energy transfers which take place between the photons and the electrons in Compton
collisions cannot be ignored.
The details of this process were worked out in pioneering papers by Raymond Weymann
(Weymann, 1966) and in much more detail by Zeldovich and Sunyaev (Zeldovich and
Sunyaev, 1969).4 The analyses of Zeldovich and Sunyaev were based upon the theory of
induced Compton scattering developed by Alexander Kompaneets (1914–1974), which had
been published in 1956, long after this remarkable classified work had been completed
in 1950 (Kompaneets, 1956).5 What these papers showed was that, during the radiation-
dominated epochs, the matter and radiation were maintained in very close thermal contact
by Compton scattering so long as the intergalactic gas remained ionised. Therefore, since
the radiation had much greater heat capacity than the matter, the matter cooled at the same
rate as the radiation during the radiation-dominated epochs, Tm ∝ R −1 .
Zeldovich and Sunyaev showed how significant distortions of the spectrum of the
microwave background radiation could take place if the electrons were heated to a tem-
perature greater than the radiation temperature by some process. These might involve, for
example, the dissipation of primordial sound waves or turbulence, matter–antimatter annihi-
lation, the evaporation of primordial black holes by the Hawking mechanism or the decay of
heavy unstable leptons. If no photons were created, the spectrum of the radiation would be
distorted from its Planckian form to a Bose–Einstein spectrum with a finite dimensionless
chemical potential μ:
   −1
2hν 3 hν
Iν = 2 exp +μ −1 . (15.5)
c kTr
 15.2 The thermal history of the Universe 397

This is the form of equilibrium spectrum expected when there is a mismatch between the
number of photons and the energy to be distributed among them in statistical equilibrium.
Very strong upper limits to the value of μ were derived from the COBE spectral observations
of the cosmic microwave background radiation. As illustrated in Figure 12.4, the microwave
background radiation has a perfect black-body spectrum, the upper limit to μ being |μ| ≤
10−4 (Page, 1997). In general terms, this means that there cannot have been major injections
of energy into the pre-galactic gas in the redshift interval 107 ≥ z ≥ 2 × 104 .
As noted above, during the post-recombination era, z ≤ 1000, matter and radiation were
decoupled and so the matter cooled more rapidly with redshift than the radiation. Since
Tm ∝ R −2 and Tr ∝ R −1 , it would be expected that the matter would now be roughly a factor
of 1000 colder than the background radiation. In fact, as indicated by the Gunn–Peterson
test (Section 14.3), the intergalactic gas must be highly ionised at redshifts z ≤ 6 because of
the absence of absorption troughs to the short-wavelength sides of the redshifted Lyman-α
emission lines observed in the spectra of large-redshift quasars. Therefore, at some time
between the epoch of recombination and a redshift certainly greater than 5, the intergalactic
gas must have been ionised and reheated, presumably by some form of activity associated
with the early formation of stars and galaxies. We will return to this topic in Section 15.13.

15.2.3 Earlier epochs

Extrapolating back to redshifts z ≈ 3 × 108 , T = 109 K, the radiation temperature was
sufficiently high for the background photons to attain γ -ray energies, ε = kT = 100 keV. At
this high temperature, the high-energy photons in the Wien region of the Planck distribution
were energetic enough to dissociate light nuclei such as helium and deuterium. At earlier
epochs, all nuclei were dissociated into protons and neutrons. When the clocks were run
forward, this was the epoch when primordial nucleosynthesis of the light elements took
place, a topic discussed in Sections 12.4 to 12.6.
At redshift z ≈ 109 , electron–positron pair production from the thermal background
radiation took place and the Universe was flooded with electron–positron pairs, one pair for
every pair of photons present in the Universe now. When the clocks were run forward from
an earlier epoch, the electrons and positrons annihilated at about this epoch and their energy
transferred to the photon field – this accounts for the little discontinuity in the temperature
history when the electrons and positrons were annihilated (Figure 15.1). At a slightly earlier
epoch, the opacity of the Universe for weak interactions became unity, resulting in a neutrino
barrier, similar to the photon barrier at z ≈ 1000.
We can extrapolate even further back in time to z ≈ 1012 when the temperature of the
background radiation was sufficiently high for baryon–antibaryon pair production to take
place from the thermal background. Just as in the case of the epoch of electron–positron
pair production, the Universe was flooded with baryons and antibaryons, one pair for every
pair of photons present in the Universe now. Again, there is a little discontinuity in the
temperature history at this epoch.
These considerations lead to one of the great cosmological problems, the baryon asym-
metry problem. In order to produce a matter-dominated Universe at the present epoch, there
must have been a tiny asymmetry between matter and antimatter in the very early Universe.
398 15 The large-scale structure of the Universe

Table 15.1. Planck units

Unit Defining expression SI value

Time tP = (Gh/c5 )1/2 10−43 s

Length lP = (Gh/c3 )1/2 4 × 10−35 m
Mass–energy m P = (hc/G)1/2 5.4 × 18−8 kg ≡ 3 × 1019 GeV

Roughly, for every 109 antibaryons, there must have been 109 + 1 baryons. When the clocks
were run forward, the 109 baryons annihilated with the 109 antibaryons, leaving one baryon,
which became the Universe as we know it with the correct photon-to-baryon ratio.
In 1965, Zeldovich showed that, if the Universe were completely symmetric with respect
to matter and antimatter, the present-day photon-to-baryon/antibaryon ratio would be about
1018 (Zeldovich, 1965), very much greater than the observed value of about 109 . Baryon-
symmetric models of the Universe were proposed by Hannes Alfvén (1908–1995) and
Oskar Klein (1894–1977) in 1962 (Alfvén and Klein, 1962) and by Roland Omnes in 1969
(Omnes, 1969), but none of these convincingly demonstrated how the matter and antimatter
could be separated in the early Universe. The baryon asymmetry must have originated in
the very early Universe. Fortunately, it is known that there is a slight asymmetry between
matter and antimatter because of CP violation observed in the decays of K 0 mesons. We
take this subject up in Section 16.4.
The process of extrapolation can be carried further and further back into the mists of the
early Universe, as far as we believe we understand high-energy particle physics. Probably
most particle physicists would agree that the standard model of elementary particles has
been tried and tested to energies of at least 100 GeV, and so we can probably trust laboratory
physics back to epochs as early as 10−6 s, although more conservative cosmologists might
be happier to accept 10−3 s. The most ambitious theorists have no hesitation in extrapolating
back to the very earliest Planck eras, tP ∼ (Gh/c5 )1/2 = 10−43 s, when the relevant physics
was certainly very different from the physics of the Universe from redshifts of about 1012
to the present day (see Table 15.1).6 Some aspects of these ideas are taken up in Chapter
16.

15.3 The development of small perturbations with cosmic epoch

15.3.1 The speed of sound as a function of cosmic epoch
An important quantity for understanding the physics of the formation of structure in the
Universe is the variation with cosmic epoch of the speed of sound, cs ,
 
∂p
cs =
2
, (15.6)
∂ρ S
where the subscript S means ‘at constant entropy’, that is adiabatic sound waves. From the
epoch when the energy densities of matter and radiation were equal, to beyond the epoch of
recombination, the dominant contributors to p and ρ changed dramatically as the Universe
 15.3 The development of small perturbations with cosmic epoch 399

changed from being radiation- to matter-dominated. Since the matter and radiation were
closely coupled throughout the pre-recombination era, the square of the sound speed can
be written as
c2 4ρr
cs2 = , (15.7)
3 4ρr + 3ρm
where ρr and ρm are the inertial mass densities in radiation and matter, respectively. Thus,
in the radiation-dominated era, z  4 × √10 0 h , ρr  ρm , and the speed of sound tended
4 2

to the relativistic sound speed, cs = c/ 3. At smaller redshifts, the sound speed decreased
as the contribution of the inertial mass density of the matter became more important.
Specifically, between the epoch of equality of the matter and radiation energy densities and
the epoch of recombination, the pressure of the sound waves was provided by the radiation,
but the inertia was due to the matter.
After the decoupling of matter and radiation, the sound speed became the thermal sound
speed of the matter which, because of the close coupling between the matter and the radiation,
1/5
had temperature Tr = Tm at redshifts z ≥ 550 h 2/5 0 . Thus, at a redshift of 500, the
temperature of the gas was about 1300 K.
In 1968, Joseph Silk (b. 1942) realised that, during the pre-recombination epochs, sound
waves in the radiation-dominated plasma were damped by the diffusion of radiation out of
the perturbation by repeated electron scatterings (Silk, 1968). The effect of damping was to
dissipate fluctuations with masses less than M = MD = 1012 (B h 2 )−5/4 M by the epoch
of recombination. Thus, for adiabatic perturbations, all fine-scale structure was wiped out
and only objects with masses greater than those of large galaxies or clusters of galaxies
survived to the epoch of recombination.

15.3.2 The evolution of perturbations on different physical scales

During the 1960s and 1970s, it was generally assumed that the principal sources of inertial
mass in the Universe were baryonic matter and the cosmic microwave background radiation.
The dark matter problem was fully appreciated, but within the limits of observational
uncertainty at that time, the dark matter could well have been in some dark baryonic form.
Consequently, the development of the spectrum of initial perturbations could be worked out
assuming that the principal constituents of the Universe were baryonic matter and radiation
and so, once the the variation of the speed of sound with cosmic epoch was established,
the evolution of the primordial perturbation spectrum could be worked out. Of particular
importance was the variation with cosmic epoch of the Jeans’ mass, the mass which is just
stable against collapse under gravity, in other words the mass contained within a region of
dimensions the Jeans’ length, λJ = cs /(Gρ0 /π)1/2 .
In the adiabatic picture developed by Zeldovich and his colleagues, it was assumed
that a spectrum of small adiabatic perturbations was set up in the early Universe and their
evolution followed according to the physical rules developed above. Figure 15.2(a) shows
how perturbations on different mass scales evolve with cosmic epoch in the standard Big
Bang (Sunyaev and Zeldovich, 1970b). Since the speed of sound in the radiation-dominated
phases was close to the speed of light, the Jeans’ mass was roughly equal to the mass
400 15 The large-scale structure of the Universe

(a) (b)

Figure 15.2: The ‘stability diagram’ of Sunyaev and Zeldovich published in 1970. (a) The region
of stability is to the left of the solid line. The two superimposed graphs illustrate the evolution of
adiabatic perturbations with different masses from early times, through the times when they enter
the horizon up to the epoch of recombination. (b) Perturbations corresponding to different masses
arrive at the epoch of recombination with different phases, resulting in a periodic dependence of the
amplitude of the perturbations upon mass (Sunyaev and Zeldovich, 1970b).

contained within the horizon scale, rH = ct, during these epochs. As soon as masses on
these scales came through the horizon, they were stabilised by the internal pressure of
the photon gas and became sound waves. Specifically, during the pre-recombination era,
after the perturbations came through the horizon, those perturbations with masses less than
MJ = 3.75 × 1015 /(B h 2 )2 M were sound waves.
The sound speed decreased after the epoch of equality of the energy densities in the
matter and radiation until the epoch of recombination was approached. Then, after the
decoupling of matter and radiation, the speed of sound dropped dramatically to the thermal
sound speed in the baryonic matter, with the result that all masses greater than about
MJ = 1.6 × 105 (0 h 2 )−1/2 M ∼ 106 M became unstable and began to grow in amplitude
as δρ/ρ ∝ (1 + z)−1 . It is therefore apparent why the adiabatic perturbations had to be
of finite amplitude when they came through the horizon since they could only grow after
the epoch of recombination and then only as (1 + z)−1 . Figure 15.2(a), due to Sunyaev
and Zeldovich, shows diagrammatically the oscillations of perturbations on different mass
scales in the pre-recombination era.
Remarkably, Andrei Sakharov (1921–1989) studied the evolution of density perturbations
in a cold universe in the years before the discovery of the cosmic microwave background
radiation and showed that there would be preferred mass scales in the mass spectrum
of large-scale structures in the Universe (Sakharov, 1965). Zeldovich and his colleagues
applied these ideas to the evolution of adiabatic perturbations in the standard hot Big Bang
model and determined the amplitudes and scales of the preferred scales which survived
to the epoch of recombination, as illustrated in Figure 15.2(b) (Sunyaev and Zeldovich,
1970b). The oscillations seen in Figure 15.2(b) result from the fact that the fluctuations which
develop into bound structures at late epochs are those with large amplitudes when they came
through the horizon. Figure 15.2(a) shows examples of two perturbations coming through
15.3 The development of small perturbations with cosmic epoch 401

their particle horizons and oscillating as sound waves until the epoch of recombination. The
amplitude of the oscillations at the epoch of recombination depended upon the phase of
oscillation of the sound waves at that time. Those oscillations which completed an integral
number of oscillations would have maximum amplitude as they began to collapse under
gravity after the decoupling of matter and radiation. In contrast, those oscillations which
had phases such that they had zero amplitude at the decoupling epoch did not form objects at
all. The mass spectrum of perturbations at the decoupling epoch is shown in Figure 15.2(b).
This spectrum of oscillations as a function of mass is sometimes called the spectrum of
acoustic or, more appropriately, Sakharov oscillations, and is a general prediction of theories
of structure formation which involve primordial adiabatic sound waves.
In the early 1970s, Zeldovich and Edward Harrison (b. 1919) independently put together
information about the spectrum of the initial fluctuations on different physical scales and
showed that observed structures in the Universe could be accounted for if the mass fluctua-
tion spectrum had the form δ M/M ∝ M −2/3 in the very early Universe, corresponding to a
power spectrum of initial fluctuations of the form P(k) ∝ k n , with n = 1 (Harrison, 1970;
Zeldovich, 1972).7 The amplitude of the power spectrum had to be ∼ 10−4 . Such a spectrum
has the attractive feature that fluctuations on different mass scales had the same amplitude
when they came through the horizon; in other words, it results in a fractal universe. This
spectrum is known as the Harrison–Zeldovich spectrum of initial perturbations.
A key test of these models was provided by the fact that the presence of density fluctua-
tions at the epoch of recombination should leave some imprint upon the cosmic microwave
background radiation. In the simplest picture, if the process of recombination were instan-
taneous, the adiabatic perturbations would be expected to result in temperature fluctuations
of T /T = 13 ρ/ρ on the last scattering surface (Silk, 1968). In fact, the problem is much
more complicated than this because the process of recombination was not instantaneous.
The fluctuations which were imprinted upon the background radiation depended upon their
sizes and optical depths relative to the thickness of this last scattering surface. The principal
sources of temperature fluctuations on small scales were expected to be associated with
first-order Doppler scattering due to the collapse of the perturbations (Sunyaev and Zel-
dovich, 1970b). These predictions provided a challenge for the observers since the predicted
amplitudes of the fluctuations in these early theories were of the order T /T ≥ 10−3 –10−4 .
At the opposite extreme from the adiabatic pertubations were the isothermal perturba-
tions. In the radiation-dominated phase of the standard Big Bang, these fluctuations in the
baryon density took place against the uniform cosmic background radiation. In the case of
perfect gases, any pressure and density distribution in the radiation-dominated phases can
be represented as the superposition of a distribution of adiabatic and isothermal perturba-
tions. The perturbations were isothermal in the sense that they caused no fluctuations in
the background radiation temperature during the radiation-dominated phases. Their internal
temperature was the same as that of the uniform radiation background, and they were frozen
into the radiation-dominated plasma. Throughout the radiation-dominated era, the timescale
for the expansion of the radiation-dominated Universe was much shorter than the collapse
timescale and so the isothermal perturbations scarcely grew at all. A simple calculation
shows that the amplitude of these perturbations grew by only a factor of about 2.5 from
the time they entered the horizon to the epoch of equality of matter and radiation energy
402 15 The large-scale structure of the Universe

densities. Subsequently, the perturbations grew according to the usual result δρ/ρ ∝ R.
This important effect will reappear in a slightly different guise in considerations of the
evolution of isocurvature perturbations involving collisionless cold dark-matter particles.
This phenomenon was first described by Peter Mészáros, and is known as the Mészáros
effect (Mészáros, 1974).

15.4 The adiabatic and isothermal scenarios for galaxy formation

In the 1970s, the concepts described in Section 15.3 gave rise to two different scenarios for
the origin of structure in the Universe. In the adiabatic scenario, the initial perturbations
were adiabatic sound waves before recombination and structure in the Universe formed
by the fragmentation of the large-scale structures which reached amplitude δρ/ρ ∼ 1 at
relatively late epochs. A realisation of this scenario was described by Doroshkevich and his
colleagues in 1974 (Doroshkevich, Sunyaev and Zeldovich, 1974).
The alternative picture, favoured by James Peebles and his colleagues at Princeton, was
one in which the perturbations were not sound waves but simply isothermal perturbations
in the pre-recombination plasma which were in pressure balance with the background
radiation. Low-mass perturbations were not damped out and so masses on all scales survived
to the recombination epoch. Galaxies and clusters of galaxies then formed by the process of
hierarchical clustering of low-mass objects under the gravitational influence of perturbations
on larger scales.
Both models predicted similar amplitudes for the temperature fluctuations imprinted on
the cosmic microwave background radiation at the epoch of recombination as the perturba-
tions began to collapse to form bound objects. Their subsequent behaviours were, however,
entirely different.

15.4.1 The adiabatic ‘pancake’ model

In the adiabatic picture developed by Zeldovich and his colleagues, only large-scale pertur-
bations with masses M ≥ 1014 M survived to the epoch of recombination, all fluctuations
on smaller mass scales being damped out by photon diffusion. During the pre-recombination
era, after the perturbations came through their particle horizons, those with masses less than
MJ = 1016 –1017M were sound waves, which oscillated until the epoch of recombination,
when their internal pressure support vanished and the Jeans’ mass dropped to MJ ∼ 106 M .
Following recombination, all the surviving perturbations grew in amplitude as δρ/ρ ∝
(1 + z)−1 until the epoch at which 0 z ∼ 1 (see Section A15.1). In the early 1970s, the
density parameter in baryons, B , was known to be less than about 0.05h −2 from the
constraints provided by primordial nucleosynthesis (Section 10.4), and so, even if h = 0.5,
the perturbations would grow slowly at redshifts z ≤ 5. In order to ensure the formation of
galaxies and larger-scale structures, the amplitudes of the perturbations must have attained
δρ/ρ = 1 by z ∼ 5. This was a satisfactory result, since quasars were known to exist at
redshifts greater than 2, and the number counts of quasars and radio sources indicated that
these objects had flourished at these early epochs (see Section 14.1). Zeldovich and his
 15.4 Adiabatic and isothermal scenarios 403

colleagues inferred that galaxies and the large-scale structure of the Universe began to form
at relatively late epochs, z ∼ 3–5. Since the fluctuations had attained amplitude δρ/ρ ∼ 1
at z ∼ 5 and δρ/ρ ∝ (1 + z)−1 , the amplitude of the density perturbations at the epoch of
recombination must have been at least δρ/ρ ≥ 3 × 10−3 .
The structures which survived on the scale of clusters and superclusters of galaxies
were unlikely to be perfectly spherical and, in a simple approximation, could be described
by ellipsoids with three unequal axes. In 1970, Zeldovich derived an analytic solution for
the non-linear collapse of these structures and showed that such ellipsoids collapsed most
rapidly along their shortest axis, with the result that flattened structures, which Zeldovich
called ‘pancakes’, were formed (Zeldovich, 1970). The density became large in the plane
of the pancake, and the infalling matter was heated to a high temperature as the matter col-
lapsed into the pancake, a process sometimes called the ‘burning of the pancakes’. Galaxies
were assumed to form by fragmentation or thermal instabilities within the pancakes. In
this picture, all galaxies formed late in the Universe, once the large-scale structures had
collapsed. This baryonic pancake theory was developed in some detail by Zeldovich and
his colleagues in the 1970s and can be thought of as a ‘top-down’ scenario for galaxy for-
mation (Doroshkevich et al., 1974). Among the successes of the theory was the fact that
it accounted naturally for the large-scale structure in the distribution of galaxies. In three
dimensions, the pancakes formed interconnected, flattened, stringy structures, not unlike
the great holes and sheets of galaxies observed in the local Universe.

15.4.2 The isothermal model and hierarchical clustering

In contrast, Peebles and his Princeton colleagues favoured the isothermal picture, in which
masses began to collapse on all scales greater than M = MJ ∼ 106 M immediately after the
epoch of recombination. This scenario had the attractive feature that the first objects to form
would have masses similar to those of globular clusters, which are the oldest known objects in
our Galaxy. The process of galaxy and structure formation was ascribed to the hierarchical
clustering of these small-scale structures under the influence of the power spectrum of
perturbations, which extended up to the largest scales. One of the attractive features of this
picture was that there would be early enrichment of the chemical abundances of the elements
as a result of nucleosynthesis in the first generations of massive stars. This process could
account for the fact that, even in the largest redshift quasars, the abundances of the elements
were not so different from those observed locally. Many of these ideas were developed by
Peebles in his important monograph, The Large-Scale Structure of the Universe (Peebles,
1980).8
The process of structure formation by hierarchical clustering was put on a formal basis by
William Press (b. 1948) and Paul Schechter (b. 1948) in a remarkable paper of 1974 (Press
and Schechter, 1974). Their objective was to provide an analytic formalism for the process of
structure formation once the density perturbations had reached amplitude δρ/ρ ∼ 1. Their
analysis started from the assumptions that the power spectrum of the primordial density
perturbations was of power-law form, P(k) ∝ k n , and that the phases of the waves were
random, what are known as Gaussian fluctuations. When the amplitude of the perturbations
reached a critical value δc , it was assumed that they formed bound systems with mass M.
404 15 The large-scale structure of the Universe

Figure 15.3: The variation of the form of the Press–Schechter mass function as a function of cosmic
time in the Einstein–de Sitter world model, 0 = 1, according to equation (15.8). (Courtesy of
Dr Andrew Blain.)

With these assumptions, they showed that the evolution of the spectrum of bound objects
with cosmic time could be written in the following remarkably simple form:
     
¯ γ M γ /2 M γ
N (M) = √ exp − , (15.8)
π M2 M∗ M∗
where γ = 1 + (n/3) and M ∗ = M ∗ (t0 )(t/t0 )4/3γ . The variation of this mass function with
time is shown in Figure 15.3. Press and Schechter were well aware of the limitations of their
approach, but it turned out that their mass function and its evolution with cosmic epoch
were in good agreement with more detailed analyses and with the results of subsequent
supercomputer simulations.9 The Press–Schechter formalism has proved to be a very use-
ful tool for studying the development of galaxies and clusters of galaxies in hierarchical
scenarios for galaxy formation. In contrast to the adiabatic picture, the isothermal scenario
is a ‘bottom-up’ picture, in which galaxies and larger-scale structures were assembled out
of smaller objects by clustering and coalescence.

15.4.3 Confrontation with the observations

Despite these advances, there were major problems with both scenarios. First of all, it
gradually became apparent that the dominant form of matter in the Universe was unlikely to
be baryonic. The constraints from primordial nucleosynthesis of the light elements strongly
suggested that the mean baryonic mass density of the Universe was about an order of
 15.5 Hot dark matter 405

magnitude less than the mean total mass density, 0 ≈ 0.2–0.3. In addition, in 1980, the
concept of the inflationary expansion of the early Universe, pioneered by Alan Guth and
his colleagues, caught the imagination of theorists (Guth, 1981). One of the consequences
of that picture, which could resolve a number of the fundamental cosmological problems,
was that the Universe should have flat spatial geometry, and so, if  = 0, it followed that
0 = 1. In this case, there was no question but that most of the mass in the Universe had
to be in some non-baryonic form.
In addition, there was observational conflict with the expected amplitude of the temper-
ature fluctuations in the cosmic microwave background radiation. As discussed above, after
the epoch of recombination, both adiabatic and isothermal perturbations began to collapse
and, for masses on the scales of clusters of galaxies and greater, their behaviour is similar.
In purely baryonic theories, these fluctuations were expected to have large amplitudes on
the last scattering surface, and these would cause observable fluctuations in the radiation
temperature of the cosmic microwave background radiation.
The theory of these processes for both adiabatic and isothermal baryonic perturbations
was worked out by Sunyaev and Zeldovich (Sunyaev and Zeldovich, 1970b), who found
the important result that, for both types of perturbation, the root-mean-square temperature
fluctuations were predicted to be
  1/2  1/2 1/2
δT 2 −5 M0
= 2 × 10 (1 + z 0 ), (15.9)
T 1015 M
−1/2
for masses M ≥ 1015 0 M , where z 0 is the redshift at which δ/ = 1.
Throughout the 1970s, increasingly sensitive searches were made for fluctuations in the
cosmic microwave background radiation, these observations being analysed critically by
Bruce Partridge (b. 1940) in his review of 1980 (Partridge, 1980a). His own observations
had reached sensitivities of T /T ≈ 10−4 or slightly better by that time (Partridge, 1980b).
Thus, by the early 1980s, the upper limits to the intensity fluctuations in the cosmic back-
ground radiation were beginning to constrain quite severely purely baryonic theories of
structure formation.

15.5 Hot dark matter – neutrinos with finite rest mass

A potential solution to these problems appeared in 1980, when Valentin Lyubimov (b. 1929)
and his collaborators reported that the electron neutrino had a finite rest-mass of about 30 eV
(Lyubimov et al., 1980). As early as 1966, Semion Gershtein (b. 1929) and Zeldovich had
noted that relic neutrinos of finite rest mass could make an appreciable contribution to
the mass density of the Universe (Gershtein and Zeldovich, 1966). In the 1970s, Györgi
(George) Marx (1927–2002) and Alexander Szalay (b. 1949) had considered the role of
neutrinos of finite rest mass as candidates for the dark matter, as well as studying their
role in galaxy formation (Marx and Szalay, 1972; Szalay and Marx, 1976). The intriguing
aspect of Lyubimov’s result was that, if the relic neutrinos had this rest mass, this Universe
would just be closed, 0 = 1. Zeldovich and his colleagues developed a new version of the
406 15 The large-scale structure of the Universe

adiabatic model in which the Universe was dominated by neutrinos with finite rest mass
(Doroshkevich et al., 1980a,b; Zeldovich and Sunyaev, 1980).
In the new picture, most of the inertial mass of the Universe was in the form of neutrinos
of rest-mass energy 30 eV. The neutrinos were therefore highly relativistic during the epoch
of nucleosynthesis and so none of the predictions of the standard Big Bang were changed.
The differences began to appear at later epochs when the neutrinos became non-relativistic.
This occurred at about the same redshift that the Universe changed from being radiation-
to matter-dominated. This was not a coincidence. Prior to this epoch the energy densities
in the photons and neutrinos were roughly the same, and so, when the neutrinos became
non-relativistic, their inertial masses no longer decreased as the Universe expanded, unlike
the photons, and so the Universe became matter-dominated. The neutrino fluctuations began
to grow under gravity as soon as they became non-relativistic, but, since the neutrinos are
weakly interacting, they streamed freely out of the perturbations and so the small-scale
perturbations were damped out. Because of the decoupling of the matter and radiation from
the neutrinos, except through their gravitational influence, the amplitudes of the sound
waves in the matter and radiation remained at the same level they had when they came
through the horizon. After recombination, the baryonic matter collapsed into the surviving,
larger-amplitude neutrino fluctuations.
Only neutrino perturbations on the very largest scale with masses ≥ 1016 M survived
to the epoch of recombination, and so, just as in the old adiabatic model, the largest-
scale structures formed first and then the smaller-scale structures formed by a process of
fragmentation. This model had the advantage of reducing very significantly the expected
amplitude of the fluctuations in the microwave background radiation since the fluctuations in
the baryonic matter were of low amplitude during the critical phases when the background
photons were last scattered. The subsequent evolution of the perturbations was not so
different from the adiabatic model. This scenario for galaxy formation became known as
the hot dark matter picture of galaxy formation since the neutrinos were highly relativistic
when they decoupled from thermal equilibrium.
There were, however, concerns about this picture. First of all, there were grave reser-
vations about the experiments that claimed to have measured the rest mass of the electron
neutrino, and it is now believed that the result was erroneous – the upper limit to the rest
mass of the electron neutrino is now found to be m ν ≤ 3 eV (Weinheimer, 2001).10 Sec-
ondly, constraints could be set to the mass of the neutrinos if they were to constitute the dark
matter in galaxies, groups and clusters of galaxies, as discussed in Section 10.2. Gunn and
Tremaine showed that, while 30 eV neutrinos could bind clusters and the haloes of giant
galaxies, those needed to bind dwarf galaxies would have to have masses much greater than
30 eV (Tremaine and Gunn, 1979).

15.6 Cold dark matter and structure formation

Once it was appreciated that non-baryonic dark matter had to be taken really seriously,
many possibilities were proposed by the particle physicists. Examples included the axions,
supersymmetric particles such as the gravitino or photino and ultraweakly interacting
15.6 Cold dark matter and structure formation 407

neutrino-like particles, all of which might be relics of the very early Universe. From roughly
1980 onwards, the particle physicists began to take the early Universe seriously as a labora-
tory for particle physics at energies which could not be achieved in terrestrial laboratories.
According to James Peebles, Richard Bond (b. 1950) introduced the term cold dark mat-
ter in 1982 to encompass many of these exotic types of particle suggested by the particle
physicists (Peebles, 1993). The matter was ‘cold’ in the sense that these particles decou-
pled from the thermal background after they had become non-relativistic. In the same year,
Peebles demonstrated how the presence of such particles could reduce the amplitude of the
predicted fluctuations in the cosmic microwave background radiation to levels consistent
with the observational upper limits (Peebles, 1982).
A change of terminology was also introduced about this time. In the purely baryonic
picture, the perturbations in the early Universe could be decomposed into isothermal and
adiabatic modes. Now, another independent component, the cold dark matter, was added to
the picture. In the three-component case, the decomposition can be made into similar modes,
but the names ‘isothermal’ and ‘adiabatic’ were scarcely appropriate for fluids containing
collisionless dark-matter particles. The corresponding modes were referred to as curvature
and isocurvature modes.
r The curvature modes were the equivalent of the adiabatic modes in that, when these
perturbations entered the horizon during the radiation-dominated era, the amplitudes of
the perturbations in the radiation, the baryonic matter and the dark matter were all more
or less the same. As result, there were variations in the local mass–energy density from
point to point in the Universe, resulting in local perturbations to the curvature of space.
r In the isocurvature modes, the mass–energy density is constant throughout space and
so there were no perturbations to the spatial curvature of the background world model,
despite the fact that there might be fluctuations in the mass–energy density of each of
the three components from point to point in the Universe.
The cold dark matter scenario was similar in many ways to the isothermal model (Davis
et al., 1992a). Since the matter was very cold, the perturbations were not damped by free
streaming. Fluctuations on all scales survived, and so, when the pre-recombination Universe
became matter-dominated, these perturbations begin to grow, decoupled from the matter
and radiation. As in the hot dark matter scenario, after the epoch of recombination, the
baryonic matter collapsed into the growing potential wells in the dark matter. Galaxies,
groups and clusters formed by a process of hierarchical clustering, which can be modelled
by the Press–Schechter formalism (Press and Schechter, 1974).
An important aspect of these models for the formation of large-scale structures was
the fact that the initial power spectrum of the perturbations, taken to be of Harrison–
Zeldovich form when they entered the horizon, was modified by various physical processes,
and this modified spectrum became the input spectrum of perturbations for simulations
of the subsequent post-recombination evolution. Thus, in the hot dark matter model, the
free streaming of neutrinos damped out all perturbations on all scales up to about 1016 M .
In the adiabatic cold dark matter model, the evolution of the perturbations was driven by
the perturbations in the radiation until the epoch at which the Universe became matter-
dominated. In the isocurvature cold dark matter model, the growth of the perturbations
408 15 The large-scale structure of the Universe

Figure 15.4: The ‘processed’ spectrum of primordial density fluctuations, (λ) = δρ/ρ(λ), observed
at some time after the epoch of equality of matter and radiation energy densities (Kolb and Turner,
1990). Rather than mass, the abscissa is plotted in terms of the wavelength associated with the
perturbations. The three models show the processed fluctuation spectrum for hot and cold dark matter
models, assuming the fluctuations are adiabatic, and for an isocurvature cold dark matter model. In
all three cases, it is assumed that the input power spectrum is of Harrison–Zeldovich form, P(k) ∝ k,
which is only preserved on the very largest physical scales.

under gravity was retarded by the Mészáros effect until the matter-dominated era. The
net result is that each model involves a different transfer function which modifies the
primordial Harrison–Zeldovich form of the spectrum (Figure 15.4). In an important paper,
James Bardeen and his colleagues worked out the transfer functions for a number of different
scenarios for structure formation and provided a convenient set of analytic formulae for
astrophysical applications (Bardeen et al., 1986).
By the early 1980s, the power of digital computers had developed to such an extent that
simulations of the non-linear evolution of the various scenarios for galaxy and large-scale
structure formation could be carried out successfully. Figure 15.5 shows a sample of the
results of computer simulations of the hot and cold dark matter models carried out by Marc
Davis, Carlos Frenk (b. 1951) and their colleagues (Frenk, 1986). These simulations repre-
sented the state of the art in numerical simulations of non-linear gravitational clustering in
the mid 1980s, their N -body codes involving periodic boundary conditions and 32 768 par-
ticles. The models were run for different cosmological models and showed the evolution of
the spectrum of perturbations during the post-recombination eras over a factor of 16 in length
scale. In the hot dark matter picture, flattened structures like pancakes were produced very
15.6 Cold dark matter and structure formation 409

Figure 15.5: Simulations of the expected distribution of galaxies on the sky according to (a) the cold
dark matter scenario and (b) the hot dark matter scenario of galaxy formation including biassing
compared with (c) the observed distribution of galaxies in the Harvard Center for Astrophysics
northern sky survey (Frenk, 1986). The outer circle represents Galactic latitude +40◦ and the empty
regions lie at declinations less than 0◦ .

effectively. The model was, in fact, far too effective in producing flattened, stringy struc-
tures. Essentially everything collapsed into thin pancakes, resulting in a much more highly
structured Universe than is actually observed. Furthermore, it was difficult to understand
how stars and galaxies could be formed before the structures on scales M ≥ 4 × 1015 M .
In the cold dark matter picture, masses on all scales began to collapse soon after recom-
bination, and star clusters and the first generations of stars could be old in this picture. In the
post-recombination epoch, large-scale systems, such as galaxies and clusters of galaxies,
were assembled from their component parts by non-linear clustering. Figure 15.5 shows
that structure indeed developed, but in the high-density models with 0 = 1, it was not
as pronounced on the large scale as is observed in the local Universe. The physical reason
for this was that it was difficult to produce elongated structures by gravitational clustering
alone, which tended to make more symmetrical structures than the sheets and filaments of
galaxies found in the Universe on the largest scales.
One of the important successes of the cold dark matter picture was that it could account for
the observed two-point correlation function of galaxies.11 A simulation in which the initial
spectrum was of standard Harrison–Zeldovich form in a 0 = 1 universe and the phases of
the waves were random, showed the two-point correlation function, ξ (r ), evolving towards
a power law (Figure 15.6). The process of non-linear gravitational clustering converts the
modified input power spectrum into one much more closely resembling the observed two-
point correlation function, ξ (r ), with the observed slope of γ = 1.8 over a wide range of
scales (Davis et al., 1985). As the model evolved, the correlation function became steeper.
410 15 The large-scale structure of the Universe

3.5

2.4
2 Ω = 1
1.8

1.4

γ =1.8

1
log ξ(r )

0
1.0

−1

−2 −1
log(r/L)
Figure 15.6: The two-point correlation functions, ξ (r ), at different scale factors. The error bars show
the uncertainties in the correlation functions from five independent runs of the computational model.
The dashed line shows the slope, γ = 1.8, of the observed two-point correlation function for galaxies
(Davis et al., 1985; Efstathiou, 1990).

As explained by George Efstathiou (b. 1955), this was very far from the end of the story
(Efstathiou, 1990). In particular, in this realisation of the model, the velocity dispersion of
galaxies chosen at random from the field was too large, but the match to observation could
be improved if it was assumed that galaxy formation occurred preferentially in overdense
regions, as proposed by Kaiser (1986). This led to the issue of the extent to which the visible
matter traces the distribution of the dominant dark matter.

15.7 Biassing

So far, it has been assumed that the visible parts of galaxies trace the distribution of the
dark matter, but one can imagine many reasons why this might not be so. The generic term
for this phenomenon is biassing, meaning the preferential formation of galaxies in certain
 15.8 Reconstructing the initial power spectrum 411

regions of space rather than in others. Part of the motivation behind the introduction of
biassing was to improve the agreement between the predictions of the preferred cold dark
matter scenario and the observed distribution of galaxies. In the hot dark matter picture,
anti-biassing would be needed so that the formation of galaxies is not so highly concentrated
into sheets and filaments.
Was there any observational evidence for biassing in the Universe? In the Coma Cluster
and in other large-scale systems, for example, the mass of the dark matter amounts to about a
factor of 10 greater than that in the visible matter, but this factor was only about one-quarter
or one-third of the value necessary to attain a density parameter 0 = 1. If the Universe
really had the critical density 0 = 1, there must be biassing by a factor of 3 towards the
formation of galaxies on the scale of clusters and superclusters as opposed to the general
field. Many possible biassing and anti-biassing mechanisms were described by Dekel and
Rees (Dekel, 1986; Dekel and Rees, 1987).
One of the most important process for biassing was described by Nicholas Kaiser
(b. 1954), who realised that inherent in the notion of the power spectrum of the pertur-
bations is the fact that the perturbations have a Gaussian distribution of amplitudes about
the root-mean-square value. If we write  = δρ/ρ, then, on any scale,  has mean value,
, with variance 2 , so that the probability of encountering a density contrast, , at some
point in space is proportional to exp (−2 /2 ). Kaiser argued that galaxies are most likely
to form in the highest peaks of the density distribution. Thus, if the density perturbations
had to exceed some value, crit , in order that structures form, galaxy formation would be
biassed towards the highest density perturbations over the mean background density (Kaiser,
1986). This picture can account for the fact that the clusters of galaxies are more strongly
clustered than galaxies in general. If structure only forms when the density contrast exceeds
a certain value crit , then galaxy formation within a large-scale density perturbation, which
will eventually form a cluster of galaxies, is strongly favoured. This model was worked
out in detail by John Peacock and Alan Heavens (b. 1959) and by James Bardeen and his
colleagues (Peacock and Heavens, 1985; Bardeen et al., 1986). The numerical simulations
described by Efstathiou illustrated clearly how the density peaks of a Gaussian random
field result in a much more highly structured distribution of galaxies as compared with
the underlying mass distribution (Efstathiou, 1990). He showed that a standard cold dark
matter model with 0 = 1 and b = 2.5 could be reconciled with a number of independent
aspects of the large-scale distribution of galaxies, including the amplitude and slope of the
two-point correlation function and the mean velocity dispersion of galaxies in the general
field.

15.8 Reconstructing the initial power spectrum

Granted that biassing plays a role in determining the amplitudes of the correlation functions
on different scales, was it possible to produce self-consistent models for the formation of
structure from a single initial power spectrum? In a pioneering analysis, Andrew Hamilton
(b. 1951) and his colleagues showed how it was possible to relate analytically the observed
spectrum of perturbations well into the non-linear regime, ξ (r )  1, to the initial spectrum
412 15 The large-scale structure of the Universe

Figure 15.7: Comparison of the inferred power spectrum of large-scale structures at the present epoch
with different variants of the cold dark matter model. The models have scale-invariant input spectra
which have been modified by the cold dark matter transfer functions. Different values of the fitting
parameter 0 h are shown, 0 h = 0.5, 0.45, . . . , 0.25, 0.2, in decreasing order of power at short
wavelengths (Peacock and Dodds, 1994).

in the linear regime (Hamilton et al., 1991). John Peacock and Stephen Dodds (b. 1970)
extended this analysis to eight separate determinations of the present power spectrum of
perturbations in the non-linear regime and then determined biassing factors for different
samples of galaxies and clusters in order to generate a smooth power spectrum (Peacock
and Dodds, 1994). This power spectrum could then be compared with the expectations of
different versions of the cold dark matter picture for different values of the density parameter,
0 .
Peacock and Dodds found that there must be significant bias present such that 0.6 0 /b =
1.0 ± 0.2, where the bias factor, b, is defined by
   
δρ δρ
ξgal (r ) = b ξD (r ) or
2
=b , (15.10)
ρ gal ρ D
the subscripts referring to galaxies (gal) and dark matter (D). The shapes of the curves were
derived assuming that the initial power spectrum was of scale-invariant form, P(k) ∝ k n ,
and was modified by the transfer function for different values of the parameter 0 h. The
 15.9 Variations on a theme of cold dark matter 413

best-fitting value of 0 h was

0 h = 0.255 ± 0.017 + 0.32(n −1 − 1). (15.11)
Thus, for the standard Harrison–Zeldovich spectrum, n = 1, this analysis suggested that
the ‘standard’ cold dark matter model, with 0 = 1, was a poor fit to the data. Even if
h = 0.5, the best-fitting models would have 0 significantly less than unity. This result was
found by a number of workers, namely that the simplest cold dark matter models predict
too much power on scales corresponding to 0.1 < k/h < 1 Mpc−1 if the large-scale power
spectrum in the linear region is to be accommodated. The other key constraint, which has
not yet been built into this picture, is the observation of intensity fluctuations in the cosmic
microwave background radiation. The predicted power spectra shown in Figure 15.7 could
be extrapolated to the scale of the COBE observations and reasonable consistency found
with the observed fluctuation spectrum.
Although the ‘standard’ cold dark matter picture with 0 = 1,  = 0 and n = 1 had
some success in reproducing a number of features of the observed structures in the Universe
on large scales, the consensus of opinion among cosmologists was that it is probably not
good enough. These concerns were sufficiently worrying for Marc Davis and his colleagues
to entitle their Nature review paper of 1992 ‘The end of cold dark matter?’ (Davis et al.,
1992a). The outcome of these studies was the development of a number of variants of the
standard picture.

15.9 Variations on a theme of cold dark matter

Many variants of the standard cold dark matter picture were proposed, all involving the
introduction of additional parameters. In parallel with these developments, the ability to
carry out large-scale simulations of the origin of large-scale structure increased dramati-
cally, mirroring the exponential increase in computing power. These enabled many of these
possibilities to be tested by supercomputer simulations. Some examples of the outcome of
these simulations are shown in Figure 15.8, which shows some of the outputs of the pro-
grammes of the Virgo consortium (Kauffmann et al., 1999). In these examples, the models
were evolved from similar initial conditions with a power-law input spectrum of Harrison–
Zeldovich form and were constrained to reproduce the same large-scale structure at the
present epoch.
These models were debated in some detail in 1996 at the Princeton meeting Critical
Dialogues in Cosmology, the proceedings of which convey the atmosphere of excitement
aroused by these debates (Turok, 1997). Some of the more promising alternatives were the
following.

Open cold dark matter (OCDM)

As the second row of Figure 15.8 shows, OCDM models could account satisfactorily for the
observations and produce the type structures observed in large-scale redshift surveys. The
differences as compared with the standard cold dark matter picture were, firstly, that the
414 15 The large-scale structure of the Universe

z =3 z =2 z =1 z =0

ΛCDM

OCDM

τ CDM

SCDM

Figure 15.8: Some examples of the predicted large-scale structure in the distribution of galaxies
from supercomputer simulations by the Virgo consortium. Each panel has side 240h −1 Mpc and
the gravitational interactions of 2563 = 1.7 × 107 particles were followed. The four models shown
involve the standard cold dark matter (SCDM), open cold dark matter (OCDM), cold dark matter
with a finite cosmological constant (CDM) and cold dark matter with decaying neutrinos (τ CDM).
The parameters of the models have been chosen to reproduce the observed large-scale structure in the
distribution of galaxies at the present epoch (Kauffmann et al., 1999).

epoch of equality of matter and radiation occurs rather later and, secondly, that the growth of
structure proceeds over a somewhat smaller range of redshifts, only until the epoch at which
0 z ≈ 1. Consequently, the break in the power spectrum takes place at greater masses,
resulting in less power at short wavelengths. This would be consistent with the dynamical
evidence that the overall density parameter was about 0 = 0.3.

Cold dark matter with a finite cosmological constant (ΛCDM)

A way of preserving the flat geometry of space is to include the cosmological constant into
the model so that 0 +  = 1. There is not a great deal of difference in the dynamics of
 15.10 Cosmic microwave background radiation fluctuations 415

the underlying model as compared with the open cold dark matter model (see the top row
of Figure 15.8). This is because the dynamics only differ from the  = 0 case at redshifts
1/3
(1 + z) ≤ 0 . Thus, the differences occur in the late stages of evolution when the effect
of the cosmological constant is to stretch out the timescale of the model, allowing some
further development of the perturbations.

Cold dark matter with decaying neutrinos (τ CDM)

In this scenario, the ratio of radiation to matter energy densities is enhanced so that the
epoch of equality of matter and radiation energy densities is shifted to lower redshifts, as
in the case of the open cold dark matter picture. It is essential to ensure that the predictions
of primordial nucleosynthesis are not violated in that, if there were additional relativistic
components present in the Universe during the epoch of nucleosynthesis, the expansion rate
would be increased and excessive amounts of helium would be produced. The trick was to
suppose that there existed particles which decay after the epoch of nucleosynthesis, thus
enhancing the radiation relative to the matter energy densities and so delaying the epoch of
equality of matter and radiation energy densities.

By the end of the conference, the four scenarios illustrated in Figure 15.8 seemed equally
plausible, but the whole picture was about to change with the new, highly suggestive,
evidence that the cosmological constant was non-zero (see Section 13.4.3).

15.10 Fluctuations in the cosmic microwave background radiation

The study of the power spectrum of spatial fluctuations in the cosmic microwave background
radiation has provided one of the most important means of confronting theories of the
origin of the large-scale structure of the Universe with observation. Limits to the amplitude
of the fluctuations continued to improve until, in 1992, a positive detection was made
in the whole sky survey carried out by the Cosmic Background Explorer (COBE) led
by George Smoot (b. 1945) (Smoot et al., 1992). The objectives of the project were to
measure precisely the spectrum and spatial distribution of the radiation over the whole
sky, and in both cases these were achieved with outstanding success. The maps of the
whole sky demonstrated how uniform the background radiation is on large angular scales
(Figure 15.9(a)). The dipole distribution of intensity expected due to the motion of the
Earth through the frame of reference in which the radiation would have been isotropic
on a large scale was clearly defined (Figure 15.9(b)). Finally, at the limiting sensitivity
of the survey, fluctuations of cosmological origin were discovered on an angular scale
of 10◦ with an amplitude of I /I ≈ 10−5 in directions away from the Galactic plane
(Figure 15.9(c)).
These observations marked the beginning of a new era in astrophysical cosmology since
they provided a direct link to the processes of formation of large-scale structures in the
Universe when they were still in their linear stage of development. Many new ground-based
experiments were developed, and approval was given for further space missions, specifically
the Wilkinson Microwave Anisotropy Probe (WMAP) of NASA, which was launched in
416 15 The large-scale structure of the Universe

Figure 15.9: Maps of the whole sky in galactic coordinates as observed at a wavelength of 5.7 mm
by the COBE satellite (Smoot et al., 1992). (a) The distribution of total intensity over the sky.
(b) Once the uniform component was removed, the dipole component associated with the motion
of the Earth through the background radiation was observed, as was the emission from the Galactic
plane. (c) Once the dipole component was removed, radiation from the plane of the Galaxy was seen
as a bright band across the centre of the picture. The fluctuations seen at high galactic latitudes were
largely noise from the telescope and the instruments, the root-mean-square value at each point being
36 μK. When averaged statistically over the whole sky at high latitudes, an excess sky noise signal of
cosmological origin amounting to 30 ± 5 μK was observed.

2001 and which was named after David Wilkinson (1935–2002), and the Planck project of
the European Space Agency, scheduled for launch in 2007.
From the early pioneering efforts of Peebles, Zeldovich and their colleagues, studies of
the cosmic microwave background radiation have developed into one of the most important
fields of cosmology. The subject is very rich astrophysically, and many different aspects of
physics contribute to the interpretation of the observations.
 15.10 Cosmic microwave background radiation fluctuations 417

15.10.1 The ionisation of the intergalactic gas through the epoch of

recombination
As discussed in Section 15.2.1, primordial intensity fluctuations in the cosmic microwave
background radiation originate in a rather narrow-redshift range at z ≈ 1000. Following the
pioneering studies of Zeldovich, Kurt and Sunyaev and Peebles (see Section 15.2.1), detailed
calculations of the degree of ionisation through the critical redshift range were carried out
by Bernard Jones (b. 1946) and Rosemary Wyse (b. 1957). These studies enabled the range
of redshifts from which the photons of the background radiation observed today were last
scattered to be determined. This probability distribution could be closely approximated by
a Gaussian distribution with mean redshift 1070 and standard deviation σ = 80 in redshift.
Thus, half the photons observed today were last scattered between redshifts of 1010 and
1130.12
The thickness of this last scattering layer corresponds to a physical scale of
10(0 h 2 )−1/2 Mpc at the present epoch, where 0 is the density parameter in matter
for the Universe as a whole. The mass contained within this scale is given by M ≈
3 × 1014 (0 h 2 )1/2 M , corresponding roughly to the mass of a cluster of galaxies. The
1/2
corresponding angular scale is θ = 60 arcmin. For scales smaller than this value, a num-
ber of independent fluctuations are expected to be present along the line of sight through
the last scattering layer. The random superposition of these perturbations leads to a statis-
tical reduction in their amplitude by a factor of roughly N −1/2 , where N is the number of
fluctuations along the line of sight.
This angular scale can be compared with the horizon scale, r = 3ct, at a redshift of 1000
1/2
which is 1.80 degrees. Perturbations on this scale are the largest which can be in causal
contact at that epoch and correspond to comoving scales of about 200(0 h 2 )−1/2 Mpc at
the present day. Thus, the smallest angular scales observed by the COBE satellite, θ = 10◦ ,
corresponded to structures on scales about five to ten times the horizon scale on the last
scattering surface. The next task is to relate the density perturbations and their associated
velocities to temperature, or intensity, fluctuations in the last scattering layer.

15.10.2 Large angular scales

On the very largest scales, the dominant source of intensity fluctuations results from the
gravitational redshift associated with density perturbations at the last scattering layer. The
fluctuations on these scales far exceeded the horizon scale at the epoch of recombination, and
so they are more appropriately referred to as metric perturbations. Although the discussion
of this subsection is concerned with perturbations on the very largest scales, super-horizon
perturbations present a general problem for the formation of structure in the expanding Uni-
verse. Perturbations on all scales of astrophysical interest exceeded their particle horizons13
early enough in the Universe.
Super-horizon perturbations need careful treatment because of the problems of defin-
ing the appropriate hypersurfaces corresponding to the same cosmic time. For sub-horizon
scales, this is not a problem because the background metric can be taken to be of Robertson–
Walker form and a synchronous time coordinate used to defined hypersurfaces of constant
 418 15 The large-scale structure of the Universe

cosmic time. In principle, radar methods could be used to ensure the synchronisation of
clocks on scales less than the particle horizon. This is not possible on super-horizon scales
since the scale of the perturbation exceeds the particle horizon. These problems were elu-
cidated in 1980 in an important paper by James Bardeen, in which he showed how these
difficulties could be solved by introducing a gauge-invariant formalism (Bardeen, 1980).
These issues are also discussed by Efstathiou and Hu (Efstathiou, 1990; Hu, 1996). As Efs-
tathiou emphasises, there are no ambiguities so long as a well defined background model
on super-horizon scales in the very early Universe is adopted.
To evaluate the expected temperature fluctuations in the last scattering layer, a general
relativistic treatment was needed, and this was first performed by Raymond Sachs (b. 1932)
and Arthur Wolfe (Sachs and Wolfe, 1967). The result was that the temperature fluctuation
T /T = (1/3)φ/c2 was expected, recalling that the Newtonian gravitational potential,
φ, is a negative quantity – this source of temperature fluctuations is known as the Sachs–
Wolfe effect.14
Although the photons pass through gravitational potential fluctuations during their sub-
sequent propagation to the Earth, what they gain by falling into them is compensated by the
gravitational redshift coming out, so long as the perturbations continue to grow linearly with
redshift.15 The amplitudes of the temperature fluctuations as a function of angular scale
due to the Sachs–Wolfe effect depended only upon the spectral index, n, of the initial power
spectrum of the fluctuations; specifically, in a simple analysis, T /T ∝ θ (1−n)/2 . Thus, if
n = 1, the amplitudes were expected to be independent of angular scale. According to the
COBE team and Partridge, the power spectrum at small multipoles, l < 30, determined from
the four-year data set from the COBE experiment, provided an estimate of n = 1.1 ± 0.3
(Bennett et al., 1996; Partridge, 1999). This value was consistent with the expectations of
the scale-free Harrison–Zeldovich spectrum, n = 1.
Whilst this agreement was encouraging, another source of fluctuations on these very
large angular scales was associated with primordial gravitational waves. According to some
inflationary theories of the early Universe, the quantum fluctuations which are responsible
for the density perturbations from which the large-scale structure of the Universe developed
might well be accompanied by a background of gravitational waves (Starobinsky, 1985;
Davis et al., 1992b; Crittenden et al., 1993). This is claimed to be a general feature of a
wide class of inflationary models for the formation of structure. According to these ideas,
to a good approximation the spectral indices of the tensor (gravitational wave) and scalar
(density perturbation) modes, n t and n s , respectively, are related by n t ≈ 1 − n s , and the
ratio of the amplitudes of their quadrupole power spectra, r = C2t /C2s , depends upon the
spectral index of the scalar perturbations as r = 7(1 − n s ). Although there is no direct
evidence for such modes being present in the COBE observations, the importance of these
ideas is that they provide a possible probe of the very early Universe indeed.

15.10.3 Intermediate angular scales – the acoustic or Sakharov oscillations

The COBE results acted as a spur to search for the acoustic, or Sakharov, oscillations in
the power spectrum of the temperature fluctuations expected on angular scales θ ≈ 1◦ . The
different variants of the cold dark matter model discussed in Section 15.9 made different
15.10 Cosmic microwave background radiation fluctuations 419

predictions about the amplitude of the power spectrum of the fluctuations as a function
of angular scale. The temperature fluctuations are distributed over the surface of a sphere,
and so the spherical polar equivalents of Fourier transforms are needed to define their
power spectra.16 The appropriate complete sets of orthonormal functions are the spherical
harmonics Ylm (θ, φ), each characterised by the amplitude, alm , the square of which, when
averaged over the azimuthal harmonic, m, is a measure of the power, Cl , on angular scale
θ ≈ π/l. In the simplest picture, the density perturbations are assumed to be Gaussian,
meaning that the phases of the waves which make up the spherical harmonic decomposition
over the sky are random, and the coefficients alm follow a Gaussian probability distribution
with phases uniformly distributed between 0 and 2π (Kogut et al., 1996). However, the seeds
for the formation of structure need not necessarily be Gaussian. The temperature fluctuations
on the sky might display non-Gaussian features, such as abrupt temperature discontinuities,
intense hot spots, linear structures and so on. These types of feature are predicted by theories
in which large-scale structures were seeded by topological defects, cosmic strings, or by
cosmic textures (see Section 16.3) (Shellard, 2003). The non-Gaussian features would result
in strongly correlated values of the coefficients alm .
To predict the power spectrum of temperature fluctuations, the evolution of the power
spectrum of density perturbations on all scales is followed from the time they entered the
horizon through the epoch of recombination. This involved using the collisional Boltzmann
equation to follow the evolution of the independent Fourier modes of the perturbations in the
dark matter, the baryonic matter and the radiation field. These computations were described
by James Peebles and Jer Tsang Yu (b. 1942) for baryonic perturbations (Peebles and Yu,
1970) and by George Efstathiou and Wayne Hu (b. 1968) and Naoshi Sugiyama (b. 1961) for
dark-matter cosmologies (Efstathiou, 1990; Hu and Sugiyama, 1995).17 A good example
of the predicted power spectrum of temperature fluctuations for such models is shown in
Figure 15.10 (Bersanelli et al., 1995). Let us disentangle some of the physical processes
contributing to the temperature fluctuations expected on scales 0.1◦ ≤ θ ≤ 1◦ .
The abundances of the light elements created by primordial nucleosynthesis indicated that
the density parameter in baryonic matter is low, B h 2 ≈ 10−2 (see Section 12.6). In this case,
according
√ to equation (15.7), the sound speed throughout most of the pre-recombination era
was c/ 3 and so the Jeans’ length was of the same order as the horizon scale. Therefore,
as soon as they came through the horizon, the coupled photon–baryon perturbations were
stabilised and oscillated as sound waves. At the same time, the perturbations in the decoupled
dark matter continued to grow in amplitude from the time the Universe became matter-
dominated. A good analogy is that the photon–baryon perturbations can be considered to
be forced oscillations within the growing potential wells in the dark-matter perturbations.
Thus, essentially all baryonic perturbations with wavelengths less than the horizon scale
were acoustic waves. As a result, they gave rise to Sakharov oscillations on the last scattering
surface.
The longest-wavelength Sakharov oscillations had wavelengths roughly equal to the
sound horizon, λs = cs t, on the last scattering layer, where cs is the speed of sound and t is
the age of the Universe. This is the maximum distance over which coherent oscillations could
have existed at the epoch of recombination and sets an upper limit to the wavelengths which √
acoustic waves could have at that epoch. Adopting the relativistic sound speed cs = c/ 3
420 15 The large-scale structure of the Universe

Figure 15.10: The predicted power spectrum of temperature fluctuations in the cosmic microwave
background radiation plotted as a function of the multipole, l, for an inflationary cold dark matter
cosmology (Bersanelli et al., 1995). The quantity plotted on the ordinate is l(l + 1)Cl , where Cl is the
power spectrum defined in endnote 15. The multipole, l, corresponds to an angular scale θ = π/l.
The curve labelled scalar shows the contribution of small density perturbations and that labelled
tensor shows the contribution of gravitational waves. The relative amplitudes of these contributions
depend upon the specific inflationary model. In the case of the scale-invariant spectrum, n = 1, there
is expected to be no contribution of gravitational waves and the spectrum of density perturbations
would be flat, corresponding to n = 1, at multipoles l ≤ 30. The estimates shown in the diagram are
for n s = 0.86. The bars along the top of the diagram show the ranges of multipoles probed by COBE
and by the Planck mission of the European Space Agency.

as an upper limit to cs , the sound horizon corresponds to a comoving distance scale of

32(0 h 2 )−1 Mpc at the present epoch and to an angular scale of
λs (comoving) 1/2
θ≈ = 0.3 0 degrees. (15.12)
D
Detailed solutions of the coupled Boltzmann equations for baryonic matter, dark matter
and radiation showed that the first acoustic peak should occur at l = 200 if 0 = 1 and at
l = 430 if 0 = 0.2. Note that the wavelength of the first Sakharov oscillation acts as a
‘rigid rod’ and so can be used to determine cosmological parameters. This is the beginning
of a long story which involves using the details of the observed angular power spectrum of
15.10 Cosmic microwave background radiation fluctuations 421

the fluctuations in the cosmic microwave background radiation to estimate many different
cosmological parameters.
As discussed in Section 15.3.2, the amplitudes of the Sakharov oscillations depend upon
the phase difference from the time they came through the horizon to last scattering surface,
that is they depend upon

dφ = cs dt. (15.13)

The first peak in the temperature spectrum corresponds to waves with wavelength equal to the
sound horizon at the last scattering layer. Oscillations which are nπ out of phase with the first
acoustic peak also correspond to maxima in the temperature power spectrum at the epoch
of recombination. Note that the odd harmonics correspond to maximum compression of
the waves and so to increases in the temperature, whereas the even harmonics correspond
to rarefactions of the acoustic waves and so to temperature minima. Those with phase
differences π(n + 12 ) relative to that of the first acoustic peak have zero amplitude at the
last scattering layer and correspond to minima in the power spectra. Thus, the maxima
correspond to frequencies

ωtrec = nπ. (15.14)

Adopting the short-wavelength dispersion relation for the oscillations, the condition
becomes
nπ
cs kn trec = nπ ; kn = = nk1 . (15.15)
λs
Thus, the acoustic peaks are expected to be evenly spaced in wavenumber.
If the contribution of baryons to the speed of sound at the last scattering surface is
neglected, that is R = 3ρB /4ρrad 1, the amplitudes of the maxima in the power spectrum
shown in Figure 15.10 would change smoothly with increasing wavenumber. When the
inertia of the baryons can no longer be neglected, however, the amplitude of the temperature
fluctuations at maximum compression is (1 + 6R) times that of the Sachs–Wolfe effect.
Furthermore, the amplitudes of the oscillations are asymmetric, the temperature excursions
varying between −(φ/c2 )(1 + 6R) for kλs = (2n + 1)π and (φ/c2 ) for kλs = 2nπ .
These results account for some of the features of the temperature fluctuation spectrum
shown in Figure 15.10. The temperature perturbations associated with the acoustic peaks
are expected to be much larger than the Sachs–Wolfe fluctuations. The asymmetry between
the even and odd peaks in the fluctuation spectrum is associated with the extra compression
at the bottom of the gravitational potential wells when account is taken of the inertia of the
baryonic matter.
The damping of the oscillations seen in Figure 15.10 at multipoles l ∼ 1000 are princi-
pally associated with the effects of photon diffusion, or Silk damping, discussed in Section
15.3.1. The suppression factor for waves of wavenumber k can be written

dτ −τ (z) −k/kD (z)

e e dz, (15.16)
dz
422 15 The large-scale structure of the Universe

where kD (z) = 2π/λD (z) and λD (z) is the scale on which fluctuations were damped out at
redshift z. The effects of damping and the random superposition of the perturbations lead
to a strong suppression of all primaeval temperature fluctuations with wavenumbers greater
than about 2000, that is on the scale of a few arcminutes.
The upshot of this discussion is that measurements of the details of the power spectra of
fluctuations in the cosmic microwave background radiation contain a wealth of information
about many key cosmological parameters, in addition to being a key probe of the very
early Universe. As a result, throughout the 1990s and the early years of the twenty-first
century, many ground-based and balloon-borne experiments were carried out to search for
the Sakharov oscillations, these efforts culminating in the observations of the NASA WMAP
space observatory.

15.11 The discovery of Sakharov oscillations

The race was on to discover whether or not Sakharov oscillations were present in the power
spectrum of the fluctuations in the cosmic microwave background radiation. Strong hints
that they were present were suggested by experiments during the 1990s, but the definitive
results were published from 2000 onwards. The first results of the Boomerang experiment,
which involved a balloon flight around the Antarctic continent, were published in April 2000
(de Bernardis et al., 2000); the first results of the Maxima experiment, another balloon-borne
project launched from Palestine in Texas, appeared in November 2000 (Hanany et al., 2000);
the power spectrum of the DASI experiment, a ground-based interferometer based at the
South Pole, was published in March 2002 (Halverson et al., 2002); the Archeops team, who
developed a balloon-borne detector which was flown across northern Europe and Russia
containing prototype detectors for the Planck project (Benôit et al., 2003a), published their
power spectrum in March 2003, while the Very Small Array (VSA), located on a high
site on Tenerife, resulted in a power spectrum which was published in June 2003 (Scott
et al., 2003); the CBI experiment, another interferometer located in the Atacama Plateau at
5000 m in Chile, resulted in a power spectrum for wavenumbers up to about 2000, which
was published in July 2003 (Pearson et al., 2003). The results of these experiments are
summarised in Figure 15.11.
The Wilkinson Microwave Anisotropy Probe (WMAP), the successor to COBE, was
launched in August 2001, and the results of the first year’s observations were pub-
lished in September 2003. Both the intensity and polarisation power spectra are shown in
Figure 15.12, which shows the extraordinary quality of the WMAP data (Bennett et al.,
2003). These data have defined the angular power spectrum and the temperature–polarisation
cross-spectrum with very high precision to multipole moments, l, beyond the second peak
of the power spectrum, clearly showing the Sakharov oscillations. Both the scalar power
spectrum and the linear polarisation of the perturbations are in excellent agreement with the
expectations of an adiabatic CDM model, as illustrated by the solid lines in Figure 15.12.
Particularly striking are the peak and trough in the cross-polarisation spectrum at l = 300
and l = 150 (Figure 15.12(b)), respectively. The polarisation signal is associated with Thom-
son scattering of the background radiation in the last scattering layer. The anti-correlation of
 15.12 The determination of cosmological parameters 423

8000
COBE
Boomerang
DASI
Maxima
6000 VSA
CBI
l (l +1)Cl /2π (μK2)

Archeops

4000

2000

0
10 100 1000
l

Figure 15.11: The power spectrum of fluctuations in the cosmic microwave background radiation
published by the Archeops team, comparing their results with those of the other experiments listed
on the diagram (Benôit et al., 2003b).

these features with the first maximum of the power spectrum is consistent with the adiabatic
CDM model.
A surprise was the large polarisation signal observed at small multipoles. This signal
must arise long after the epoch of recombination because it occurs on very large angular
scales. It has been interpreted as polarisation induced during the re-ionisation epochs when
the intergalactic gas was reheated and ionised.

15.12 The determination of cosmological parameters

The discussion of Section 15.10 indicated how various features of the power spectrum of
fluctuations in the cosmic microwave background radiation enable cosmological parameters
to be determined. On their own, the WMAP observations provide remarkable limits to
many cosmological parameters, but they become even more impressive when taken in
combination with independent data on the large-scale structure of the Universe. Some
appreciation of what became possible by combining the WMAP observations with the two-
point correlation function for galaxies determined by the Sloan Digital Sky Survey (SDSS)
was provided by the analysis of Max Tegmark (b. 1967) and his colleagues (Tegmark et al.,
2004) (Figure 15.13).
424 15 The large-scale structure of the Universe

angular scale
90° 2° 0.5° 0.2°
6000

TT cross-power
5000 spectrum
ΛCDM all data
WMAP
4000 CBI
ACBAR
l (l+1)Cl /2π (μK2)

3000

2000

1000

(a)
0

TE cross-power
3 reionisation spectrum

2
(l+1)Cl /2π (μK2)

(b)
−1
0 10 40 100 200 400 800 1400
multipole moment (l )

Figure 15.12: The WMAP angular power spectrum (Bennett et al., 2003). (a) The points show the
measured power spectrum from the first year of observations. The grey band shows the cosmic variance
associated with the observations. The solid line shows the fit of a CDM model to the observations.
The grey points at l > 800 are derived from the CBI and ACBAR experiments. (b) The temperature–
polarisation cross-spectrum. The peak in the spectrum near l = 300 is out of phase with the total
power spectrum, as predicted for adiabatic initial conditions.

The procedure involves first defining the range of parameters to be included in the sim-
ulations in order to reproduce the observed spatial correlation function for galaxies and the
power spectrum of fluctuations in the cosmic background radiation. A list of 13 parameters
which span a very wide range of possible models is shown in Table 15.2. Many of these
parameters have been discussed in previous sections, for example B , D , h,  , As , n s , b,
and the definitions are summarised in Table 15.2. The new parameters include the density
15.12 The determination of cosmological parameters 425

Figure 15.13: The determination of cosmological constants using both the WMAP power spectrum
and the three-dimensional power spectrum for galaxies from the Sloan Digital Sky Survey (SDSS)
(Tegmark et al., 2004). The light shaded areas use the information from WMAP alone, while the
darker shaded areas show the result of including the data from the SDSS. Each distribution has
been marginalised over all other quantities in the six-parameter family (τ,  , ωd , ωb , As , n s ) of the
‘vanilla’ CDM model, as well as the bias factor, b, when the SDSS data are included. The horizontal
dashed lines show the 1σ and 2σ limits for each parameter.

parameter associated with space curvature, k , which is zero in the simplest inflationary
model of the early Universe, the curvature of the initial power spectrum as parameterised by
α = d(ln n s )/d(ln k), the ratio of gravitational wave to scalar perturbations, r , the spectral
index of the gravitational waves, n t , and the ratio of mass densities in neutrinos to dark
matter, f ν .
Figure 15.13 shows how cosmological information can be extracted from these modelling
procedures. Tegmark and his colleagues used Monte Carlo Markov Chain methods to deter-
mine the probability distributions with which the parameters are determined. The panels in
Figure 15.13 include a large number of parameters, 13 of which are variables; the others are
426 15 The large-scale structure of the Universe

Table 15.2. The parameters involved in the construction of cosmological models for the
origin of structure in the Universe (Tegmark et al., 2004)

Parameter Definition Status WMAP alone WMAP + SDSS

ωB = B h 2 baryon density parameter not optional 0.0245 +0.0050

−0.0019
0.0232 +0.0013
−0.0010
ωD = D h 2
dark-matter density parameter not optional 0.115 +0.020
−0.021
0.1222 +0.0090
−0.0082
 dark-energy density parameter not optional 0.75 +0.10
−0.10
0.699 +0.042
−0.045
w dark-energy equation of state
τ reionisation optical depth not optional 0.21 +0.24
−0.11
0.124 +0.083
−0.057
k a space curvature
As amplitude of scalar power spectrum not optional 0.98 +0.56
−0.21
0.81 +0.15
−0.09
ns scalar spectral indexb
1.02 +0.16
−0.06
0.977 +0.039
−0.025
α running of scalar spectral index
r tensor–scalar ratio
nt tensor spectral index
b bias factor not optional no constraint 1.009 +0.073
−0.083
f ν = ρν /ρD neutrino fraction

a
B + D +  + k = 1.
b
n s = 1 is the preferred value according to the standard inflation picture.

derived from them. The light grey regions of the diagrams show the inferred values of the
parameters using only the WMAP data. It can be seen that these strongly constrain many of
the most important cosmological parameters. For example, there is unambiguous evidence
for the presence of a non-zero cosmological constant, as parameterised by  = /3H02 .
Likewise, there are strong constraints on ωD and ωB .
When the evidence of the two-point correlation function for galaxies from the Sloan
Digital Sky Survey is included, as shown by the dark grey areas of Figure 15.13, the param-
eters are determined with roughly a factor of 2 greater precision. This arises because the
two-point correlation function for galaxies provides a direct measure of the matter power
spectrum with very much smaller error bars than could be inferred from the WMAP data
alone.
There are many ways of interpreting the results of this remarkable analysis, as explained
by Tegmark et al. (2004). One conservative approach is to derive the simplest set of param-
eters needed to account satisfactorily for all the WMAP and SDSS data. For illustration,
such sets of parameters are shown in Table 15.2. Of the 13 parameters listed, six are abso-
lutely necessary in order to account for the observations, and these have been labelled ‘not
optional’. In this ‘vanilla’ model, it is assumed that there are no primordial gravitational
waves or relic neutrinos, the equation of state of the dark energy is p = −ρc2 , the initial
power spectrum is defined by a constant spectral index, n s , and the spatial geometry of the
Universe is flat, k = 0. It turns out that satisfactory and self-consistent estimates of the six
parameters can be found for this model. The estimates of the parameters for WMAP alone
are listed in column 4 and for WMAP plus SDSS in column 5 of Table 15.2. Particularly
striking results are the following.
 15.13 The post-recombination Universe 427

r The spectral index of the primordial power spectrum is very close to n s = 1, as expected
for inflation models of the early Universe.
r The dark-energy density parameter  ≈ 0.7, while the dark matter has density param-
eter D h 2 ≈ 0.122.
r The baryonic matter has density parameter B h 2 ≈ 0.023.
r Hubble’s constant can be inferred from combining the parameters in appropriate ways;
it is found to be h = 0.695 +0.039
−0.031
.
r When the SDSS data are included, the bias parameter, b, is very close to unity, meaning
that the baryonic matter in galaxies traces the distribution of the cold dark matter.
What is spectacular about these results is that they are entirely consistent with the inde-
pendent estimates of these parameters discussed in some detail in Chapter 13. For example,
the estimates of Hubble’s constant from the HST key project (Section 13.2) are in excellent
agreement with that given in the above list, the estimates of  from the Type Ia supernovae
projects agree very well with those listed above, as do the estimates of the overall density
parameter in matter, 0 = B + D , and the density parameter in baryonic matter, B .
Including the data derived from the Type Ia supernovae with the WMAP and SDSS data
leads to an estimate of the age of the Universe of t0 = 14.1 +1.0
−0.9
Gyears.
As expected, if additional parameters are included in the model-fits, the uncertainty in
some of these parameters increases, but many of the results are robust. It is remarkable,
however, how well the simplest models can account for many diverse data sets. As Tegmark
et al. (2004) remark
Readers wishing to choose a concordance model for calculational purposes using Ockham’s razor can
adopt the best fit ‘vanilla lite’ model
(τ,  , ωd , ωb , A s ) = (0.17, 0.72, 0.12, 0.024, 0.89). (15.17)
Note that this is even simpler than 6-parameter vanilla models, since it has n s = 1 and only 5 free
parameters.

It is significant that these results are in excellent agreement with the independent con-
clusions of the WMAP team, who used the data on the two-point correlation function for
galaxies from the 2dF Galaxy Redshift Survey (Spergel et al., 2003). The data in equation
(15.17) correspond to the set of concordance parameters listed in Table 15.3.

15.13 The post-recombination Universe

15.13.1 The epoch of re-ionisation
Among the important results of the above analysis is the estimate of the optical depth of the
intergalactic gas to Thomson scattering, τ ≈ 0.17, which is associated with the reheating
and reionisation of the intergalactic gas, presumably by the earliest generations of stars or
active galactic nuclei in galaxies. If the intergalactic gas were instantly 100% reionised at
some redshift z ion , the reionisation redshift was found to be z ion = 14.4 +5.2
−4.7
according to
the analysis of Tegmark and his colleagues. Note that this result is strongly dependent upon
the polarisation information shown in Figure 15.12.
428 15 The large-scale structure of the Universe

Table 15.3. A concordance set of cosmological parameters for

calculational purposes using Ockham’s razor (Tegmark et al., 2004)

Parameter Definition Value

H0 Hubble’s constant 72 km s−1 Mpc−1

k space curvature 0
 dark-energy density parameter 0.72
0 = B + D total matter density parameter 0.28
B baryon density parameter 0.047
D dark-matter density parameter 0.233
ns scalar spectral index 1
As amplitude of scalar power spectrum 0.89
τ reionisation optical depth 0.17

This result can be contrasted with the information on the absence of a Gunn–Peterson
trough in the spectra of large-redshift quasars (see Section 14.3). The quasars with redshifts
up to z ∼ 5 showed no evidence for neutral hydrogen absorption by diffuse intergalactic
gas, but a Gunn–Peterson trough has been found in the spectrum of the largest-redshift
quasar discovered as part of the Sloan Digital Sky Survey. Four quasars with redshifts 5.80,
5.82, 5.99 and 6.28 were discovered by the technique of searching for i-band drop-outs,
similar to the technique illustrated in Figure 14.15, but now optimised for very-large-redshift
quasars. The spectra of these four quasars observed with the Keck-2 telescope are shown
in Figure 15.14 (Becker et al., 2001).
As the redshift of the quasar increases, the Lyman-α forest, shown in Figure 14.11,
depresses the continuum to the short-wavelength side of Lyman-α. As shown by Palle Møller
(b. 1954) and Peter Jakobsen (b. 1953), the effect increases dramatically with increasing
redshifts, particularly when account is taken of the evolution of the number of absorption
systems per unit redshift, as discussed in Section 14.3 (Møller and Jakobsen, 1990). The
first three quasars show increasing absorption to the short-wavelength side of the redshifted
Lyman-α line, as expected if the absorption were due to an increasing number of discrete
absorbing clouds. At the very largest redshifts, 5.95 ≤ z ≤ 6.15, however, the continuum
flux drops dramatically to zero to the short-wavelength side of the Lyman-α line in the
largest-redshift quasar at z = 6.28. Robert Becker (b. 1946) and his colleagues interpreted
this result as showing that the Gunn–Peterson trough has at last been observed, implying
that the fractional abundance of diffuse neutral hydrogen begins to increase with increasing
redshift beyond z ∼ 6. This result was reinforced by observations of the trough to the
short-wavelength side of the Lyman-β line. In the words of Becker et al. (2001),

the Universe is approaching the reionisation epoch at z ∼ 6.

The study of the Universe at redshifts z ≥ 6, often referred to as the dark ages, is one of
the great challenges for twenty-first-century astrophysical cosmology. The first generations
of stars in galaxies must have formed between redshifts 30 ≥ z ≥ 6. The ultraviolet radi-
ation of these newly formed stars, and any black holes which had formed in the nuclei of
galaxies, must have resulted in heating and reionisation of the intergalactic gas. It is a great
 15.13 The post-recombination Universe 429

Figure 15.14: The optical spectra of four very-large-redshift quasars (z ≥ 5.8) observed with the
Keck-2 telescope. In each spectrum, the wavelength of prominent emission lines, as well as the
Lyman limit, are indicated by dotted vertical lines (Becker et al., 2001). The key observation is
the zero continuum flux to the short-wavelength side of the Lyman-α line in the spectrum of the
largest-redshift quasar, which can be contrasted with the residual Lyman-α forest in the lower-redshift
quasars.

observational challenge to determine observationally the history of structure formation and

the evolution of the intergalactic gas during these critical epochs.

15.13.2 The effects of dissipation

Most of the discussion of the formation of galaxies and large-scale structures has focussed
upon the dynamical evolution of the potential wells defined by the dark matter, in other
words the linear and non-linear evolution of perturbations under the influence of gravity
430 15 The large-scale structure of the Universe

Figure 15.15: The cooling rate per unit volume, (T ), of an astrophysical plasma of number
density 1 nucleus cm−3 by radiation for different cosmic abundances of the heavy elements, ranging
from zero metals to the present abundance of heavy elements as a function of temperature, T (Silk
and Wyse, 1993). In the zero-metal case, the two maxima in the cooling curve are associated with the
recombination of hydrogen ions and doubly ionised helium.

alone. In addition, the role of dissipation, meaning energy loss by radiation and which
results in the loss of thermal energy from the system, needs to be taken into account. If the
radiation process is effective in removing pressure support from the system, this can result
in a runaway situation, known as a thermal instability (Field, 1965).
Dissipative processes play a dominant role in the formation of stars, and this leads to
the question of whether or not similar processes are important in the formation of larger-
scale structures. In 1977, the role of dissipative processes in galaxy formation was elegantly
described by Martin Rees and Jeremiah Ostriker (Rees and Ostriker, 1977), who considered
the cooling of the primordial plasma consisting of the primaeval abundances of hydrogen
and helium. In 1993, Joseph Silk and Rosemary Wyse included cooling by heavy elements
at different levels of enrichment from the primordial values into their cooling curves. The
key relation is the energy loss rate of the plasma by radiation as a function of temperature
(Figure 15.15) (Silk and Wyse, 1993). The cooling rate is presented in the form dE/dt =
−N 2 (T ), where N is the number density of hydrogen ions. In the absence of heavy metals,
the dominant loss mechanism at high temperatures, T > 106 K, is thermal bremsstrahlung,
the energy loss rate being proportional to N 2 T 1/2 . At lower temperatures, the main loss
mechanisms are free–bound and bound–bound transitions of hydrogen and ionised helium,
corresponding to the two maxima in the cooling curve. As the abundance of the heavy
15.13 The post-recombination Universe 431

Figure 15.16: A number density–temperature diagram showing the locus defined by the condition
that the collapse time of a region, tdyn , should be equal to the cooling time of the plasma by radiation,
tcool , for different abundances of the heavy elements (Silk and Wyse, 1993). Also shown are lines of
constant mass, a cooling time of 1010 years (dotted lines) and the density at which the perturbations
are of such low density that they do not collapse in the age of the Universe.

elements increases, the overall energy loss rate can be more than an order of magnitude
greater than that of the primordial plasma at temperatures T ≤ 106 K.
For the case of a fully ionised plasma, the cooling time is defined to be the time it takes
the plasma to radiate away its thermal energy, that is
E 3N kT
tcool = = 2 . (15.18)
|dE/dt| N (T )
This timescale can be compared with the timescale for gravitational collapse, tdyn ≈
(Gρ)−1/2 ∝ N −1/2 . The significance of these timescales is best appreciated by inspecting the
locus of the equality, tcool = tdyn , in a temperature–number density diagram (Figure 15.16).
The locus tcool = tdyn is a mapping of the cooling curve of the hydrogen–helium plasma loss
rate into the (T, N ) plane. Inside this locus, the cooling time is shorter than the collapse
time, and so it is expected that dissipative processes are more important than dynamical
processes in the collapse of the baryonic matter. Also shown in Figure 15.16 are lines of
constant mass, as well as loci corresponding to the radiation loss time being equal to the age
of the Universe, and to the perturbations having such low density that they do not collapse
gravitationally in 1010 years. It can be seen that the range of masses which lie within the
critical locus, and which can cool in 1010 years, corresponds to 1010 ≤ M/M ≤ 1013 – this
is the key conclusion of this analysis. The fact that the masses lie naturally in the range of
observed galaxy masses suggests that the typical masses of galaxies are not only determined
432 15 The large-scale structure of the Universe

by the initial fluctuation spectrum, but by astrophysical processes as well. As can be seen
in Figure 15.16, the greater the abundance of the heavy elements, the shorter the timescale
for cooling of a region of a given temperature and density.
Figure 15.16 can be used astrophysically in the following way. For any theory of the
origin of the large-scale structure, the density and temperature of the gas can be worked
out at each epoch. The diagram can then be used to determine whether or not cooling by
radiative losses is important. A good example was found in the various versions of the
adiabatic pancake theory. When the gas cloud collapses to form a pancake, the matter falls
into a singular plane and, as a result, a shock-wave passes out through the infalling matter,
heating it to a high temperature. In this picture, galaxies can form by thermal instabilities
in the heated gas. Inspection of Figure 15.16 shows that, if the gas is heated above 104 K,
there is no stable region for masses in the range 1010 to 1013 M .
A second exercise, carried out by George Blumenthal (b. 1945) and his colleagues,
was to plot the observed location of galaxies on the temperature–number density diagram
(Blumenthal et al., 1984). The effective temperature associated with the velocity dispersion
of the stars in a galaxy or the galaxies in a cluster, 12 kTeff ≈ 12 mv 2 , was plotted, rather than
the thermal temperature of the gas. The irregular galaxies fell well within the cooling locus,
and the spirals, S0, and elliptical galaxies all lay close to the critical line. On the other hand,
the clusters of galaxies lay outside the cooling locus. Thus, cooling is expected to be an
important factor in certain scenarios for the formation of galaxies.
The success of the cold dark matter models in accounting for the large-scale distribu-
tion of galaxies encouraged the numerical cosmologists to tackle the next step of includ-
ing diffuse baryonic matter into the computations and following the evolution of the gas
using the procedures of single-particle hydrodynamics and similar computational algorithms
(Hernquist et al., 1996; Katz et al., 1996; Miranda-Escudé et al., 1996). These programmes
involved incorporating the role of dissipative processes in enabling the baryonic compo-
nents of galaxies to form the types of structures observed in the real Universe. Once star
formation got underway, the intense ultraviolet radiation of young massive stars contributed
to the ionisation of the diffuse intergalactic gas. In addition, quasars were formed, and their
ionising ultraviolet radiation had an important influence upon the physical state of the gas.
Figure 15.17 shows an example of the resulting structure of neutral hydrogen clouds at
redshift z = 2. The simulations result in a network of filaments, with dense knots of neutral
hydrogen forming in the vicinity of galaxies. The column densities found in these simu-
lations span the range from about 1018 to 1026 m−2 , and their number density distribution
follows closely the observed power-law relation N (NH ) ∝ NH−1.5 over this range. According
to the simulations, the high column density knots, shown as white blobs in Figure 15.17, arise
from radiatively cooling gas associated with galaxies which form in high-density regions.
In contrast, the low-density systems are associated with a wide range of different types of
structure. To quote Lars Hernquist (b. 1954) and his colleagues (Hernquist et al., 1996),

the low column density absorbers are physically diverse: they include filaments of warm gas; caustics in
frequency space produced by converging velocity flows; high-density halos of hot collisionally ionised
gas; layers of cool gas sandwiched between shocks; and modest local undulations in undistinguished
regions of the intergalactic medium. Temperatures of the absorbing gas range from below 104 K to
above 106 K.
Notes to Chapter 15 433

Figure 15.17: A supercomputer simulation of the expected structure of neutral hydrogen in the inter-
galactic medium at a redshift z = 2 in a standard cold dark matter cosmology with 0 = 1. The size of
the box corresponds to a comoving scale of 22.22 Mpc. The simulation includes self-shielding of the
neutral hydrogen from the background ultraviolet ionising radiation. The grey-scale is such that the
white blobs correspond to column depths N (HI) ≥ 1020.5 m−2 . The faint filamentary structures corre-
spond to column densities 1019.5 ≥ N (HI) ≥ 1018.5 m−2 and the black ‘voids’ correspond to regions
with N (HI) ≤ 1018.5 m−2 (Katz et al., 1996).

Notes to Chapter 15

1 Part III of Malcolm Longair, Galaxy Formation (Berlin: Springer-Verlag, 1998) is devoted to an
exposition of the detailed physics of the processes involved in structure formation in the Universe,
and covers much of the content of this chapter.
2 The corresponding equation for the electrostatic case was only derived after the discovery of
plasma oscillations by Irving Langmuir (1881–1957) and Lewi Tonks (1897–1971) in the 1920s,
and describes the dispersion relation for longitudinal plasma oscillations, or Langmuir waves:
Ne e2
ω2 = c2s k 2 + ,
m e ε0

where Ne is the electron density and m e is the mass of the electron (Tonks and Langmuir, 1929).
The formal similarity of the physics may be appreciated by comparing the attractive gravitational
434 15 The large-scale structure of the Universe

acceleration of a region of mass density ρ0 and the repulsive electrostatic acceleration of a region
of electron charge density Ne e. The equivalence of −Gρ0 and Ne e2 /4π 0 m e is apparent.
3 It may appear strange at first that the temperature at which the intergalactic gas was fully ionised
is not closer to 150 000 K, the temperature at which hν = kT = 13.6 eV, the ionisation potential
of neutral hydrogen. The important point is that the photons far outnumber the baryons in the
intergalactic medium by a factor of 3.6 × 107 /B h 2 , and there is a broad range of photon energies
present in the Planck distribution. Roughly speaking, the intergalactic gas will be ionised, provided
there are as many ionising photons with hν ≥ 13.6 eV as there are hydrogen atoms, and this occurs
when the temperature is only about 1/25 of the ionisation potential of hydrogen. This type of
calculation appears in a number of different guises in astrophysics: the nuclear reactions which
power the Sun take place at a much lower temperature than expected; the temperature at which
regions of ionised hydrogen become fully ionised is only about 10 000 K; light nuclei are destroyed
in the early Universe at much lower temperatures than would be expected. In all these cases, the
tails of the Planck and Maxwell distributions contain large numbers of photons and particles,
respectively, with energies very much greater than the mean.
4 The exchange of energy between photons and electrons is an enormous subject and has been
treated in more detail by Sunyaev, Zeldovich and their colleagues (Sunyaev and Zeldovich, 1980;
Pozdnyakov, Sobol and Sunyaev, 1983).
5 The topic of induced Compton scattering and the physics of radiation transfer had been the subject
of detailed studies by Zeldovich, who led the team which developed the Soviet hydrogen bomb
in the 1950s. Some of the very best Soviet physicists, including Lev Landau, contributed to the
development of what is now called the Kompaneets equation. Kompaneets was the first physicist
to receive security clearance so that the work could be published in the open literature.
6 It is intriguing that, as soon as Planck had introduced what he called the ‘unit of action’, h, into
the expression for black-body radiation, he appreciated that it became possible for the first time to
define a set of ‘natural units’ by combining the gravitational constant, G, and the speed of light,
c, with what we now call Planck’s constant, h, as listed in Table 15.1 (Planck, 1900).
7 The power spectrum P(k), of the perturbations is defined in the following way. First of all, we
define the Fourier transform pair for the density perturbations on scale r , δρ/ρ(r ) = (r ):

V
(r ) = k e− ik·r d3 k;
(2π)3
1
k = (r ) e ik·r d3 x.
V

We now use Parseval’s theorem to relate the integrals of the squares of (r ) and its Fourier
transform k :

1 V
2 (r ) d3 x = |k |2 d3 k.
V (2π )3

The quantity on the left-hand side of this relation is the mean square amplitude of the fluctuation
per unit volume, and |k |2 is the power spectrum of the fluctuations, which is often written as
P(k). Therefore, we can write

V V
2  = |k |2 d3 k = P(k) d3 k.
(2π)3 (2π )3

8 Peebles recounts the history of the development of many of these ideas in his major book, The
Principles of Physical Cosmology (Peebles, 1993).
9 I have given a simple derivation of the Press–Schechter formula in Section 16.3 of Longair, Galaxy
Formation.
Notes to Chapter 15 435

10 The mass of the electron neutrino/antineutrino was measured in tritium β-decay experiments. The
decay results in a 3 He nucleus, an electron and an electron antineutrino. If the antineutrinos had
non-zero mass, the spectrum of the electrons would be deformed at high energies, that is the
antineutrino mass determines the maximum energy of emitted electrons (Weinheimer, 2001). The
results of the Mainz experiments provide an upper limit to the mass of the electron antineutrino
of 2 eV. The particle data book suggests a conservative upper limit of 3 eV (see http://www-
pdg.lbl.gov/pdg.html).
11 The review by George Efstathiou can be thoroughly recommended for those who wish to enter
much more deeply into the details of these remarkable simulations (Efstathiou, 1990).
12 Further detailed calculations for the probability distribution of last scattering of photons through
the epoch of recombination were presented by Hu and Sugiyama (1995).
13 Various types of horizon appear in cosmological discussions. The term particle horizon refers to
the maximum separation which points could have had and could still be causally connected at some
epoch t. In a simple approximation, the particle horizon is given by r H ≈ ct. When account is taken
of the fact that the Universe was expanding more rapidly in the past, the scale becomes rH = 3ct
in a matter-dominated Universe and rH = 2ct in a radiation-dominated Universe. Another term
found in the literature is the event horizon, which was introduced by Wolfgang Rindler (b. 1924)
in 1956 (Rindler, 1956). This is defined to be the greatest comoving radial distance coordinate
which an object can have at a particular epoch if it is ever to be observable, however long the
observer waits.
14 Thanu Padmanabhan (b. 1957) shows how this result can be derived by perturbing the Friedman
metric and relating the temperature fluctuation to the perturbation in the Newtonian gravitational
potential φ (see also his even better solution of 1996) (Padmanabhan, 1993, 1996). Peter Coles
(b. 1963) and Francesco Lucchin (1944–2002) rationalised how the Sachs–Wolfe answer can be
derived (Coles and Lucchin, 1995). In addition to the Newtonian gravitational redshift, because
of the perturbation of the metric, the cosmic time, and hence the scale factor R, at which the
fluctuations are observed, are shifted to slightly earlier cosmic times. Temperature and scale factor
change as T /T = −R/R. For all the standard models in the matter-dominated phase R ∝ t 2/3 ,
and so the increment of cosmic time changes as R/R = (2/3)t/t. But ν/ν = −t/t is just
the Newtonian gravitational redshift, with the net result that there is a positive contribution to
T /T of −(2/3)φ/c2 . The net temperature fluctuation is given by T /T = 13 φ/c2 .
15 There are variants on the theme of the Sachs–Wolfe effect. If the time evolution of the perturbations
does not grow linearly with time, the integrated effect of the gravitational perturbation along the
line of sight needs to be evaluated. This is known as the integrated Sachs–Wolfe effect. Another
effect, noted by Rees and Sciama (1926–1999), is that, if the background radiation passes through
large-scale density perturbations and the depth of the potential well increases during propagation
through the perturbation, temperature fluctuations are induced in the cosmic background radiation
(Rees and Sciama, 1968).
16 The first step is to make a spherical harmonic expansion of the temperature distribution over the
whole sky:

∞
T T (θ, φ ) − T0 m=l
(θ, φ ) = = alm Ylm (θ, φ ), (1)
T T0 l=0 m=−l

where the normalised functions, Ylm , are given by the expression

 
2l + 1 (l − |m|)! 1/2
Ylm (θ, φ ) = Plm (cos θ ) e imφ (2)
4π (l + |m|)!
(−1)m for m ≥ 0
× , (3)
1 for m < 0
436 15 The large-scale structure of the Universe

and Plm (cos θ) are the associated Legendre polynomials of order l. The values of alm are found by
∗
multiplying the temperature distribution over the sphere by Ylm and integrating over the sphere:
T ∗
alm = (θ, φ )Ylm d . (4)
4π T
If the fluctuations can be represented by a superposition of waves of random phase, each of the
(2l + 1) coefficients of alm associated with the multipole, l, provides an independent estimate of the
amplitude of the temperature fluctuations associated with that multipole. If the power spectrum is
∗
assumed to be circularly symmetric about each point in the sky, the mean value of alm alm , averaged
over the whole sky, provides an estimate of the power associated with the multipole, l:
1 ∗
Cl = alm alm = |alm |2  . (5)
2l + 1 m

Because of the assumption that the fluctuations are Gaussian, the power spectrum, Cl , provides a
complete statistical description of the temperature fluctuations.
17 Excellent surveys of the physics of temperature fluctuations in the cosmic microwave background
radiation have been given by Hu and Sugiyama (1995), Hu (1996) and by Hu, Sugiyama and Silk
(1997).

A15 Explanatory supplement to Chapter 15

A15.1 The development of small perturbations in the expanding universe
The results concerning the development of small perturbations in the expanding Universe
are so important that it is worthwhile giving a simple demonstration of the origin of the
fundamental results.1
The development of a spherical perturbation in the expanding Universe can be modelled
by embedding a spherical region of density ρ + δρ in an otherwise uniform Universe of
density ρ (Figure A15.1). According to Gauss’s law for gravity, the spherical region behaves
dynamically like a Universe of slightly higher density than the background model. It is
simplest to begin with the parametric solutions for the dynamics of the Friedman world
models, which can be written

R = a(1 − cos θ), t = b(θ − sin θ ), (A15.1)

0 0
a= , b= . (A15.2)
2(0 − 1) 2H0 (0 − 1)3/2

Firstly, we find the solutions for small values of θ , corresponding to early stages of the
matter-dominated era. Expanding to third-order in θ, cos θ = 1 − 12 θ 2 , sin θ = θ − 16 θ 3 ,
we find the solution
 2/3
1/3 3H0 t
R = 0 . (A15.3)
2
This solution shows that, in the early stages, the dynamics of all world models tend towards
those of the Einstein–de Sitter model, 0 = 1, λ = 0, that is R = (3H0 t/2)2/3 , but with a
different constant of proportionality.
Explanatory supplement to Chapter 15 437

ρ + δρ

Figure A15.1: Illustrating a spherical perturbation with slightly greater density than the average in a
uniformly expanding Universe. The region with slightly greater density behaves dynamically, exactly
like a model Universe with density ρ + δρ.

Now consider the region of slightly greater density embedded within the background
model. We expand the expressions for R and t to fifth-order in θ : cos θ = 1 − 12 θ 2 + 24 θ −
1 4

. . . , sin θ = θ − 6 θ + 120 θ − · · · . The solution is then given by

1 3 1 5

 2/3   
1/3 3H0 t 1 6t 2/3
R = 0 1− . (A15.4)
2 20 b

We can now write down an expression for the change of density of the spherical perturbation
with cosmic epoch:
 
−3 3 (0 − 1)
ρ(R) = ρ0 R 1+ R . (A15.5)
5 0
Note that, if 0 = 1, there is no growth of the perturbation. The density perturbation may
be considered to be a mini-Universe of slightly higher density than 0 = 1 embedded in an
0 = 1 model. Therefore, the density contrast changes with scale factor as
δρ (R) − 0 (R) 3 (0 − 1)
= = = R. (A15.6)
ρ 0 (R) 5 0
This result indicates why density perturbations grow only linearly with cosmic epoch.
The instability corresponds to the slow divergence between the variation of the scale factors
with cosmic epoch of the model with 0 = 1 and one with slightly greater density. This
behaviour is illustrated in Figure A15.2. This is the essence of the argument developed by
Tolman and Lemaı̂tre in the 1930s and, more generally, by Lifshitz in 1946 to the effect that,
because the instability develops only algebraically, galaxies could not form by gravitational
collapse.
438 15 The large-scale structure of the Universe

Figure A15.2: Illustrating the growth of a spherical perturbation in the expanding Universe as the
divergence between two Friedman models with slightly different densities.

It can also be appreciated why the perturbations only grow at redshifts z ≥ 1/0 . This is
because at redshifts z ≥ 1/0 the dynamics of the world model tend to those of the critical
Einstein–de Sitter model, and so the arguments given above apply to the divergence of the
slightly denser model, which will then follow the behaviour shown in Figure A15.2. At
redshifts z ≤ 1/0 , the background model follows the ‘open’ trajectory, and it would need
a very large density perturbation to move the perturbed region onto the collapsing branch
in Figure A15.2.

Note to Section A15

1 I have given details of the full derivations in Chapter 11 of Malcolm Longair, Galaxy Formation
(Berlin: Springer-Verlag, 1998).
16 The very early Universe

16.1 The big problems

The history recounted in the preceding four chapters represents quite extraordinary progress
in understanding the astrophysical origins and evolution of our Universe. The contrast
between the apparently insuperable problems of determining precise values of cosmological
parameters up till the 1990s and the era of precision cosmology of the early years of the
twenty-first century is startling.
Yet, despite the undoubted success of the concordance model, it raises as many problems
as it solves. The picture is incomplete in the sense that, within the context of the standard
world models, the initial conditions listed in Tables 15.2 and 15.3 have to be put in by hand in
order to create the Universe as we observe it today. How did these initial conditions arise? As
the quality of the observations improved, a number of fundamental issues for astrophysical
cosmology became apparent. The resolution of these problems will undoubtedly provide
insight into the laws of physics under physical conditions which at the moment can only be
studied by cosmological observations.1

16.1.1 The horizon problem

This problem, clearly recognised by Robert Dicke in 1961, can be restated, ‘Why is the
Universe so isotropic?’ (Dicke, 1961). At earlier cosmological epochs, the particle horizon
r ∼ ct encompassed less and less mass and so the scale over which particles could be
causally connected became smaller and smaller. A vivid example of this problem is to
work out how far light could have travelled along the last scattering layer at z = 1000 since
the Big Bang. In matter-dominated models, this distance is r = 3ct, corresponding to an
angle of θH = 1.80 ◦ on the sky. Thus, regions of the sky separated by greater angular
1/2

distances could not have been in causal communication. Why then is the cosmic microwave
background radiation so isotropic? How did causally separated regions ‘know’ that they
had to have the same temperature to better than one part in 105 ?

16.1.2 The flatness problem

Why is the Universe so close to the critical density, 0 = 1? The flatness problem was
also recognised by Dicke in his paper of 1961 and was reiterated by Dicke and Peebles
in 1979 (Dicke, 1961; Dicke and Peebles, 1979). The problem arises from the fact that,

439
440 16 The very early Universe

according to the standard world models, if the Universe were set up with a value of the
density parameter differing even slightly from the critical value 0 = 1, it would diverge
very rapidly from 0 = 1 at later epochs. It is straightforward to show that, if the Universe
has density parameter 0 today, at redshift z, (z) would have been given by
   
1 1
1− = f (z) 1 − , (16.1)
(z) 0

where f (z) = (1 + z)−1 for the matter-dominated era and f (z) ∝ (1 + z)−2 during the
radiation-dominated era. Thus, since 0 ∼ 1 at the present epoch, it must have been
extremely close to the critical value in the remote past. Alternatively, if (z) had departed
from (z) = 1 at a very large redshift, 0 would be very far from 0 = 1 today. Thus,
the only ‘stable’ value of 0 is 0 = 1. There is nothing in the standard world models that
would lead us to prefer any particular value of 0 . This is sometimes referred to as the
fine-tuning problem.

16.1.3 The baryon-asymmetry problem

The baryon-asymmetry problem arises from the fact that the photon-to-baryon ratio today
is given by

Nγ 4 × 107
= = 1.6 × 109 , (16.2)
NB B h 2
where B is the density parameter in baryons and the values of B and h have been taken
from Table 15.3. If photons are neither created or destroyed, this ratio is conserved as the
Universe expands. At temperature T ≈ 1010 K, electron–positron pair production takes
place from the photon field. At a correspondingly higher temperature, baryon–antibaryon
pair production takes place, with the result that there must have been a very small asymmetry
in the baryon–antibaryon ratio in the very early Universe if we are to end up with the correct
photon-to-baryon ratio at the present day. As explained in Section 15.2.3, at these very early
epochs there must have been roughly 109 + 1 baryons for every 109 antibaryons to guarantee
the observed ratio at the present epoch. If the Universe had been symmetric with respect
to matter and antimatter, the photon-to-baryon ratio would now be about 1018 , in gross
contradiction with the observed value (Zeldovich, 1965). Therefore, there must be some
mechanism in the early Universe which results in a slight asymmetry between matter and
antimatter.

16.1.4 The primordial fluctuation problem

What was the origin of the density fluctuations from which galaxies and large-scale struc-
tures formed? According to the analyses of Chapter 15, the amplitudes of the density per-
turbations when they came through the horizon had to be of finite amplitude, δρ/ρ ∼ 10−5 ,
over a very wide range of mass scales. These cannot have originated as statistical fluctua-
tions in the numbers of particles on, say, the scales of superclusters of galaxies. There must
 16.1 The big problems 441

have been some physical mechanism which generated finite-amplitude perturbations with
a power spectrum close to P(k) ∝ k in the early Universe.

16.1.5 The values of the cosmological parameters

The horizon and flatness problems were recognised before compelling evidence was found
for the finite value of the cosmological constant, or, in modern parlance, the density param-
eter in the vacuum fields  , but these problems remain unchanged. The concordance
values for the cosmological parameters create their own problems. The Universe seems to
be geometrically flat, k = 0, and so the sum of the density parameters in the matter and
the dark energy must sum to unity,  + m = 0.72 + 0.28 = 1. Even if the sum of these
two parameters were not precisely unity, it is a surprise that the two parameters are of the
same order of magnitude at the present epoch. The matter density evolves with redshift as
(1 + z)3 , while the dark-energy density parameter is unchanging with cosmic epoch. Why
then do we live at an epoch when they have more or less the same values?
Inspection of the Friedman equation (6.13) shows that if  is positive, the term may
be thought of as representing, in the words of Yakov Zeldovich, the ‘repulsive force of
a vacuum’, the repulsion being relative to an absolute geometrical frame of reference
(Zeldovich, 1968). There was no obvious interpretation of this term in the context of classical
physics. There is, however, a natural interpretation according to quantum field theory.
The key insight was the introduction of the Higgs fields into the theory of weak interac-
tions (Higgs, 1964). These and other ideas of quantum field theory were clearly described
by Zeldovich (1968). The Higgs fields were introduced into the electro-weak theory of
elementary particles in order to eliminate singularities in that theory and to endow the
W ± and Z 0 bosons with mass. Precise measurements of the masses of these particles at
CERN confirmed that theory very precisely, although the Higgs particles themselves have
not yet been detected – this is one of the major goals of the Large Hadron Collider (LHC)
experiment at CERN, due to begin taking data in 2007. The Higgs fields are scalar fields,
which have negative energy equations of state, p = −ρc2 . Fields of this nature, associ-
ated with phase transitions when the strong force decoupled from the electro-weak force in
the early Universe, is a possible candidate for a cosmological negative-energy equation of
state.2
In their review of the problem of the cosmological constant, Sean Carroll (b. 1966),
William Press and Edwin Turner (b. 1949) described how a theoretical value of  could
be estimated using simple concepts from quantum field theory. They found the mass density
of the repulsive field to be ρv = 1095 kg m−3 , about 10120 times greater than permissable
values at the present epoch which correspond to ρv ≤ 10−27 kg m−3 (Carroll, Press and
Turner, 1992).3 This is quite a problem, but it should not be passed over lightly. If the
inflationary picture of the very early Universe is taken seriously (see Section 16.4), this is
exactly the type of force which drove the inflationary expansion. Then, we have to explain
why ρv decreased by a factor of about 10120 at the end of the inflationary era. In this
context, 10−120 looks remarkably close to zero, which would correspond to the standard
Friedman picture with  = 0, but there is now clear evidence that  is finite with
value 0.72.
442 16 The very early Universe

As if these problems were not serious enough, they are compounded by the fact that the
natures of the dark matter and the dark energy are unknown. One of the consequences of
precision cosmology is the troubling result that we do not understand the nature of about
95% of the material which drives the large-scale dynamics of the Universe. The concordance
values for the cosmological parameters listed in Table 15.3 really are extraordinary – many
of my colleagues regard them as crazy. Rather than being causes for despair, however, these
problems should be seen as the great challenges for the astrophysicists and cosmologists
of the twenty-first century. It is not too far-fetched to see an analogy with Bohr’s theory of
the hydrogen atom, which was an uneasy mix of classical and primitive quantum ideas, but
which was ultimately to lead to completely new insights with the development of quantum
mechanics.

16.1.6 The way ahead

In the standard Big Bang model, the problems are solved by assuming that the Universe was
endowed with appropriate initial conditions in its very early phases. It is postulated that our
Universe evolved from an initial state which was isotropic with flat geometry, was slightly
matter–antimatter asymmetric, contained fluctuations with a Harrison–Zeldovich spectrum
and had initial values of m and  such that they ended up being roughly equal at the
present day. To put it crudely, we get out at the end what we put in at the beginning.
I have suggested five possible approaches to solving these problems (Longair, 1997).
r That is just how the Universe is – the initial conditions were set up that way.
r There are only certain classes of Universe in which intelligent life could have evolved.
The Universe has to have the appropriate initial conditions, and the fundamental con-
stants of nature should not be too different from their measured values or else there
would be no chance of life forming as we know it. This approach involves the anthropic
cosmological principle, according to which it is asserted that the Universe is as it is
because we are here to observe it.
r The inflationary scenario for the early Universe can be adopted and its consequences
studied.
r We should seek clues from particle physics and extrapolate that understanding beyond
what has been confirmed by experiment to the earliest phases of the Universe.
r The solution may well turn out to be something else we have not yet thought of. This
would certainly involve new physical concepts.
Let us consider each of these approaches.

16.2 The limits of observation

Even the first, somewhat defeatist, approach might be the only way forward if it turned out
to be just too difficult to disentangle convincingly the physics responsible for setting up the
initial conditions from which our Universe evolved. In 1970, William McCrea considered
the fundamental limitations involved in asking questions about the very early Universe, his
 16.3 The anthropic cosmological principle 443

conclusion being that we can obtain less and less information the further back in time one
asks questions about the early Universe (McCrea, 1970). A modern version of this argument
would be framed in terms of the limitations imposed by the existence of a last scattering
surface for electromagnetic radiation at z ≈ 1000 and those imposed on the accuracy of
observations of the cosmic microwave background radiation and the large-scale structure
of the Universe because of their cosmic variances.
In the case of the cosmic microwave background radiation, the observations are already
cosmic variance limited for values of spherical harmonic l ≤ 354 – we will never be able to
do any better than what we already know on these scales. Observations by the Planck Satellite
will extend the cosmic variance limit to l ≈ 1500. In these studies, the search for new physics
will depend upon discovering discrepancies between the standard concordance model and
future observations. The optimists, of whom the present author is one, would argue that the
advances will come through extending our technological capabilities so that these classes of
observation become cosmic variance limited and new approaches are adopted. For example,
the detection of primordial gravitational waves, dark-matter particles and even the nature
of the vacuum energy are likely to become the cutting edge of astrophysical cosmology
during the twenty-first century. These approaches will be accompanied by discoveries in
particle physics with the coming generations of ultra-high-energy particle experiments.
There will also be surprises which open up completely new ways of tackling these apparently
insuperable problems – for example, what will be discovered in ultra-high-energy cosmic-
ray experiments, such as those to be carried out with the Auger array?
It is folly to attempt to predict what will be discovered over the coming years, but we might
run out of luck. How would we then be able to check that the theoretical ideas proposed
to account for the properties of the very early Universe are correct? Can we do better
than boot-strapped self-consistency? The great achievement of modern observational and
theoretical cosmology has been that we have made enormous strides in defining a convincing
framework for astrophysical cosmology which has been observationally validated, and the
big problems can now be addressed as areas of genuine physical enquiry.

16.3 The anthropic cosmological principle

There is certainly some truth in the fact that our ability to ask questions about the origin
of the Universe says something about the sort of Universe we live in. The ‘cosmological
principle’ asserts that we do not live at any special location in the Universe, and yet we are
certainly privileged in being able to make this statement at all. In this line of reasoning,
there are only certain types of Universe in which life as we know it could have formed.
For example, the stars must live long enough for there to be time for biological life to
form and evolve into sentient beings. This line of reasoning is embodied in the anthropic
cosmological principle, first expounded by Brandon Carter in 1974 (Carter, 1974) and
dealt with in extenso in the books by John Barrow (b. 1952) and Frank Tipler (b. 1947) and
by John Gribbin (b. 1946) and Martin Rees (Barrow and Tipler, 1986; Gribben and Rees,
1989). Part of the problem stems from the fact that we have only one Universe to study – we
cannot go out and investigate other Universes to see if they have evolved in the same way
444 16 The very early Universe

as ours. There are a number of versions of the principle, some of them stronger than others.
In extreme interpretations, it leads to statements such as the strong form of the principle
enunciated by John Wheeler (Wheeler, 1977),
Observers are necessary to bring the Universe into being.

It is a matter of taste how seriously one wishes to take this line of reasoning. To many
cosmologists, it is not particularly appealing because it suggests that it will never be possible
to find physical reasons for the initial conditions from which the Universe evolved, or for the
values of the fundamental constants of nature. On the other hand, Steven Weinberg (b. 1933)
found it such a puzzle that the vacuum energy density,  , is so very much smaller than
the values expected according to current theories of elementary particles, that he invoked
anthropic reasoning to account for its smallness (Weinberg, 1989, 1997). I prefer to regard
the anthropic cosmological principle as the very last resort if all other physical approaches
fail.

16.4 The inflationary Universe and clues from particle physics

The most important conceptual development for studies of the very early Universe can
be dated to 1980 and the proposal by Alan Guth (b. 1947) of the inflationary model for
the very early Universe (Guth, 1981).4 There had been earlier suggestions foreshadowing
his proposal. Zeldovich had noted in 1968 that there is a physical interpretation of the
cosmological constant, , associated with zero-point fluctuations in a vacuum (Zeldovich,
1968). Andrei Linde (b. 1948) in 1974 and Sidney Bludman (b. 1927) and Malvin Ruderman
in 1977 showed that the scalar Higgs fields have similar properties to those which would
result in a positive cosmological constant (Linde, 1974; Bludman and Ruderman, 1977).
Guth realised that, if there were an early exponential expansion of the Universe, this could
solve the horizon problem and drive the Universe towards a flat spatial geometry, solving the
flatness problem at the same time. Suppose the scale factor, R, increased exponentially with
time as R ∝ et/T . Such exponentially expanding models were found in some of the earliest
solutions of the Friedman equations, in the guise of empty de Sitter models, driven by what
is now termed the vacuum energy density,  (see Section 6.3 (Lanczos, 1922)). Consider a
tiny region of the early Universe expanding under the influence of the exponential expansion.
Particles within the region were initially very close together and in causal communication
with each other. Before the inflationary expansion began, the region had physical scale less
than the particle horizon, and so there was time for it to attain a uniform, homogeneous
state. The region then expanded exponentially so that neighbouring points were driven to
such large distances that they could no longer communicate by light signals – the causally
connected regions were swept beyond their particle horizons by the inflationary expansion.
At the end of the inflationary epoch, the Universe transformed into the standard radiation-
dominated Universe and the inflated region continued to expand as R ∝ t 1/2 . In Guth’s
original inflationary scenario, the exponential expansion was associated with the symmetry
breaking of ‘grand unified theories’ of elementary particles at very high energies through a
first-order phase transition. At high enough energies, the strong and electroweak forces were
16.4 The inflationary Universe particle physics 445

unified, and only at lower energies did they appear as distinct forces. The grand unification
phase transition was expected to take place at a characteristic energy E ∼ 1014 GeV, known
as the GUT scale, only about 10−34 s after the Big Bang.
To order of magnitude, the argument ran as follows. The timescale 10−34 s was also
the characteristic e-folding time for the exponential expansion. Over the interval from
10−34 s to 10−32 s, the radius of curvature of the Universe increased exponentially by a
factor of about e100 ≈ 1043 . The horizon scale at the beginning of this period was only
r ≈ ct ≈ 3 × 10−26 m, and this was inflated to a dimension of 3 × 1017 m by the end of
the period of inflation. This dimension then scaled as t 1/2 , as in the standard radiation-
dominated Universe, so that the region would have expanded to a size of 3 × 1042 m by
the present day – this dimension far exceeds the present particle horizon, r ≈ cT0 , of the
Universe, which is about 1026 m. Thus, our present Universe would have arisen from a
tiny region in the very early Universe which was much smaller than the horizon scale at
that time. This guaranteed that our present Universe would be isotropic on the large scale,
resolving the horizon problem. At the end of the inflationary era, there was an enormous
release of energy associated with the ‘latent heat’ of the phase transition, and this reheated
the Universe to a very high temperature indeed (Figure 16.1).
The exponential expansion also had the effect of straightening out the geometry of the
early Universe, however complicated it may have been to begin with. Suppose the tiny
region of the early Universe had some complex geometry. The radius of curvature of the
geometry Rc (t) scales as Rc (t) = Rc (t0 )R(t), where Rc (t0 ) is the radius of curvature of the
geometry at the present epoch t0 , and so radius of curvature of the geometry is inflated to
dimensions vastly greater than the present size of the Universe, driving the geometry of the
inflated region towards flat Euclidean geometry, k = 0, and consequently the Universe
must have 0 +  = 1. It is important that these two aspects of the case for the inflation-
ary picture can be made independently of a detailed understanding of the physics of the
inflation.
Guth demonstrated how concepts from theories of elementary particles could lead to a
physical realisation of this picture. Although the particles associated with any scalar fields
have not yet been detected, scalar fields are now common in theories of elementary particles.
The most important of these are the Higgs bosons associated with electroweak unification
and which are expected to have mass about 100–200 GeV and which should be discovered
in the Large Hadron Collider experiments at CERN. In Guth’s picture, the masses of the
scalar particles were of the order of 1014 GeV and the symmetry breaking took place at a
temperature of about T ∼ 1027 K. An important feature of this picture was that, although
occurring at energies far exceeding those accessible by laboratory experiments, the inflation
took place long after the Planck era, tP ∼ 10−43 s, and so the physics of inflation does not
require a fully developed theory of quantum gravity before progress can be made.
In Guth’s original proposal, the Universe was in a symmetric state, referred to as a false
vacuum state, at a very high temperature before the inflationary phase took place. As the
temperature fell, spontaneous symmetry breaking took place through the process of barrier
penetration from the false vacuum state and the Universe attained a lower-energy state,
the true vacuum. At the end of this period of exponential expansion, the phase transition
took place, releasing a huge amount of energy. The problem with this realisation was that
446 16 The very early Universe

Figure 16.1: Comparison of the evolution of the scale factor and temperature in the standard Big
Bang and inflationary cosmologies.

it predicted ‘bubbles’ of true vacuum embedded in the false vacuum, with the result that
huge inhomogeneities were predicted.
Another concern about Guth’s original proposal was that an excessive number of
monopoles were created during the GUT phase transition. Thomas Kibble (b. 1923) showed
that, when this phase transition took place, a variety of topological defects are expected to be
created, including point defects (or monopoles), line defects (or cosmic strings) and sheet
defects (or domain walls) (Kibble, 1976). Kibble showed that one monopole is created for
each correlation scale at that epoch. Since that scale cannot be greater than the particle
horizon at the GUT phase transition, it is expected that huge numbers of monopoles were
created. According to the simplest picture of the GUT phase transition, the mass density
in these monopoles in the standard Big Bang picture would vastly exceed 0 = 1 at the
present epoch (Kolb and Turner, 1990).
The model was revised in 1982 by Linde and by Andreas Albrecht (b. 1957), and Paul
Steinhardt (b. 1952), who proposed instead that, rather than through the process of barrier
penetration, the transition took place through a second-order phase transition, which did
 16.5 The origin of the spectrum of primordial perturbations 447

not result in the formation of ‘bubbles’ and so excessive inhomogeneities (Albrecht and
Steinhardt, 1982; Linde, 1982, 1983). This picture, often referred to as new inflation, also
eliminated the monopole problem, since the likelihood of even one being present in the
observable Universe was very small.
The original hope that a physical realisation for the inflationary expansion could be found
within the context of particle physics beyond the standard model has not been achieved,
but the underlying concepts of the inflationary picture have been used to define the nec-
essary properties of the inflaton potential needed to create the Universe as we know it.
The successful realisations are similar to those involved in the new inflationary picture.
Once the inflationary expansion began at some stage in the early Universe, the change from
the false to true vacuums states took place through a process of slow roll-over, meaning
that the inflationary expansion took place over many e-folding times before the huge energy
release takes place. An excellent introduction to these concepts and the changing perspec-
tive on the inflationary picture of the early Universe is contained in the book Cosmological
Inflation and Large-Scale Structure by Andrew Liddle (b. 1965) and David Lyth (b. 1940)
(Liddle and Lyth, 2000). Many different versions of the inflationary picture of the early
Universe have emerged, an amusing table of over 100 possibilities being presented by Paul
Shellard (b. 1959) (Shellard, 2003).

16.5 The origin of the spectrum of primordial perturbations

As discussed in the preceding section, there were hopes that an origin for the perturbations
from which the large-scale structure of the Universe formed would be found in the phase
transitions associated with the end of the inflationary era when the Universe was reheated.
Whenever a first-order phase change occurs in nature, such as when water freezes or boils,
large fluctuations are found as nucleation takes place. It was soon established that the
amplitudes of the perturbations expected according to the old inflation model were huge,
far exceeding the amplitude δρ/ρ ∼ 10−5 necessary for the formation of cosmic structure.
The situation changed dramatically with the development of the new inflationary sce-
nario, in which the phase transition is second-order. The history of the development of
the idea that quantum fluctuations in the early Universe could create the observed scale-
free spectrum of Harrison–Zeldovich type has been recounted by Guth (Guth, 1997, 2003).
Stephen Hawking and Gary Gibbons (b. 1946) worked out the important result that quantum
fluctuations in expanding de Sitter space produce thermal radiation with a well defined tem-
perature (Gibbons and Hawking, 1977). This acted as a stimulus to apply similar ideas to the
new inflationary picture with a view to estimating the perturbation spectrum. These ideas
were thrashed out at the 1982 Nuffield Workshop held in Cambridge (Gibbons, Hawking and
Siklos, 1983). The key result of these discussions was the agreement that the spectrum of
quantum fluctuations of the vacuum Higgs fields were scale-free and result naturally in adi-
abatic curvature perturbations with a spectrum strikingly similar to the Harrison–Zeldovich
spectrum with n ≈ 1. For any particular model of the inflaton potential, the vacuum fluctu-
ation spectrum can be determined exactly. This field has become one of the major growth
areas of theoretical cosmology.5 The success of these theories in suggesting a physical
448 16 The very early Universe

origin for the spectrum of initial perturbations is impressive, although the predicted ampli-
tude of the perturbations is model-dependent. According to Liddle and Lyth (2000),

Although introduced to resolve problems associated with the initial conditions needed for the Big Bang
cosmology, inflation’s lasting prominence is owed to a property discovered soon after its introduction.
It provides a possible explanation for the initial inhomogeneities in the Universe that are believed to
have led to all the structures we see, from the earliest objects formed to the clustering of galaxies to
the observed irregularities in the microwave background.

At the same time, we should not neglect the possibility that there are other sources of
perturbations resulting from various types of topological defect, such as cosmic strings,
domain walls, textures and so on (Shellard, 2003). However, the startling success of
inflationary ideas in accounting for the observed spectrum of fluctuations in the cosmic
microwave background radiation has made it the model of choice for studies of the early
Universe.

16.6 Baryogenesis

A key contribution of particle physics to the physics of the early Universe concerns the
baryon-asymmetry problem, a subject referred to as baryogenesis. In a prescient paper of
1967, Andrei Sakharov enunciated the three conditions necessary to account for the baryon–
antibaryon asymmetry of the Universe (Sakharov, 1967). Sakharov’s rules for the creation
of non-zero baryon number from an initially baryon-symmetric state are:
r baryon number must be violated;
r C (charge conjugation) and CP (charge conjugation combined with parity) must be
violated;
r the asymmetry must be created under non-equilibrium conditions.
The reasons for these rules can be readily appreciated from simple arguments (Kolb
and Turner, 1990). Concerning the first rule, it is evident that, if the baryon asymmetry
developed from a symmetric high-temperature state, baryon number must have been violated
at some stage – otherwise, the baryon asymmetry would have to be built into the model
from the very beginning. The second rule is necessary in order to ensure that a net baryon
number is created, even in the presence of interactions which violate baryon conservation.
The third rule is necessary because baryons and antibaryons have the same mass and so,
thermodynamically, they would have the same abundances in thermodynamic equilibrium,
despite the violation of baryon number and C and CP invariance.
There is evidence that all three rules can be satisfied in the early Universe from a combi-
nation of theoretical ideas and experimental evidence from particle physics. Thus, baryon
number violation is a generic feature of grand unified theories, which unify the strong and
electroweak interactions – the same process is responsible for the predicted instability of
the proton. C and CP violation have been observed in the decay of the neutral K 0 and
K¯ 0 mesons. The K 0 meson should decay symmetrically into equal numbers of particles
 16.7 The Planck era 449

and antiparticles, but, in fact, there is a slight preference for matter over antimatter, at the
level 10−3 , very much greater than the degree of asymmetry necessary for baryogenesis,
∼10−8 . The need for departure from thermal equilibrium follows from the same type of
reasoning that led to the primordial synthesis of the light elements. As in that case, so long
as the timescales of the interactions which maintained the various constituents in thermal
equilibrium were less than the expansion timescale, the number densities of particles and
antiparticles of the same mass would be the same. In thermodynamic equilibrium, the num-
ber densities of different species did not depend upon the cross-sections for the interactions
which maintain the equilibrium. It is only after decoupling, when non-equilibrium abun-
dances were established, that the number densities depended upon the specific values of the
cross-sections for the production of different species.
In a typical baryogenesis scenario, the asymmetry is associated with some very massive
boson and its antiparticle, X, X , which are involved in the unification of the strong and elec-
troweak forces and which can decay into final states which have different baryon numbers.
Edward (Rocky) Kolb (b. 1951) and Michael Turner (b. 1949) provide a clear description of
the principles by which the observed baryon asymmetry can be generated at about the epoch
of grand unification, or soon afterwards, when the very massive bosons could no longer
be maintained in equilibrium. Although the principles of the calculations are well defined,
the details are not understood, partly because the energies at which they are likely to be
important are not attainable in laboratory experiments, and partly because predicted effects,
such as the decay of the proton, have not been observed. Thus, although there is no definitive
evidence that this line of reasoning is secure, well understood physical processes of the type
necessary for the creation of the baryon–antibaryon asymmetry exist. The importance of
these studies goes well beyond their immediate significance for astrophysical cosmology.
As Kolb and Turner (1990) remark,

in the absence of direct evidence for proton decay, baryogenesis may provide the strongest, albeit
indirect, evidence for some kind of unification of the quarks and the leptons.

16.7 The Planck era

It should be borne in mind that there is no evidence for the inflationary picture beyond the
need to solve the big problems listed in Section 16.1. Enormous progress has been made in
understanding the types of physical process necessary to resolve the four great problems, but
it is not clear how independent evidence for them can be found. The methodological problem
with these ideas is that they are based upon extrapolations to energies vastly exceeding those
which can possibly be tested in terrestrial laboratories. Cosmology and particle physics come
together in the early Universe and they boot-strap their way to a self-consistent solution.
This may be the best that we can hope for, but it would be preferable to have independent
constraints upon the theories.
A representation of the evolution of the Universe from the Planck era to the present day
is shown in Figure 16.2. The Planck era is that time in the very remote past when the energy
450 16 The very early Universe

Figure 16.2: A schematic diagram illustrating the evolution of the Universe from the Planck era to
the present time. The shaded area to the right of the diagram indicates the regions of known physics.

densities were so great that a quantum theory of gravity is needed. On dimensional grounds,
this era must have occurred when the Universe was only about tP ∼ (hG/c5 )1/2 ∼ 10−43 s
old. Despite enormous efforts on the part of theorists, there is no quantum theory of gravity,
and so we can only speculate about the physics of these extraordinary eras.6
Being drawn on a logarithmic scale, Figure 16.2 encompasses the evolution of the whole
of the Universe, from the Planck area at t ∼ 10−43 s to the present age of the Universe,
which is about 4 × 1017 s, or 13 × 109 years, old. Halfway up the diagram, from the time
when the Universe was only about one millisecond old, to the present epoch, we can be
reasonably confident that we have the correct picture for the Big Bang, despite the basic
problems discussed above.
At times earlier than about one millisecond, we quickly run out of known physics. This
has not discouraged theorists from making bold extrapolations across the huge gap from
10−3 s to 10−43 s using the current understanding of particle physics and concepts from
string theories. Some impression of the types of thinking involved in these studies can be
found in the ideas expounded in Gibbons et al. (2003). Maybe many of these ideas will turn
Notes to Chapter 16 451

out to be correct, but there must be some concern that some fundamentally new physics may
emerge at higher and higher energies before we reach the GUT era at t ∼ 10−36 s and the
Planck era at t ∼ 10−43 s. This is why the particle physics experiments to be carried with the
Large Hadron Collider at CERN are of such importance for astrophysics and cosmology, as
well as for particle physics. It is fully expected that definite evidence will be found for the
Higgs boson. In addition, there is the possibility of discovering new types of particles, such
as the lightest supersymmetric particle or new massive ultra-weakly interacting particles,
as the accessible range of particle energies increases from about 100 GeV to 1 TeV. These
experiments should provide clues to the nature of physics beyond the standard model of
particle physics and will undoubtedly feed into the understanding of the physics of the early
Universe.
It is certain that at some stage a quantum theory of gravity is needed which may help
resolve the problems of singularities in the early Universe. The singularity theorems of
Roger Penrose and Stephen Hawking show that, according to classical theories of gravity
under very general conditions, there is inevitably a physical singularity at the origin of the
Big Bang, that is as t → 0, the energy density of the Universe tends to infinity. However, it
is not clear that the actual Universe satisfies the various energy conditions required by the
singularity theorems. All these considerations show that new physics is needed if we are to
develop a convincing physical picture of the very early Universe.

Notes to Chapter 16

1 The physics of the early Universe is splendidly described in Kolb and Turner (1990).
2 In the modern picture of the vacuum, there are zero-point fluctuations associated with the zero-
point energies of all quantum fields. The stress–energy tensor of a vacuum has a negative-energy
equation of state, p = −ρc2 . This pressure may be thought of as a ‘tension’ rather than a pressure.
When such a vacuum expands, the work done, p dV , in expanding from V to V + dV is given
by p dV = −ρ c2 dV , so that, during the expansion, the mass density of the negative-energy field
remains constant. The same result can be found from application of the first law of thermodynamics
for a relativistic fluid in a cosmological setting,
 p
dρ ρ +
+3 c2 = 0.
dR R

(See Malcolm Longair, Galaxy Formation (Berlin: Springer-Verlag, 1998).) If the vacuum energy
density is to remain constant, ρvac = constant, it follows from this equation that p = −ρc2 .
3 John Peacock sets this as example 7.5 in his book Cosmological Physics (Cambridge: Cambridge
University Press, 1999). Heisenberg’s uncertainty principle states that a virtual pair of particles
of mass m can exist for a time t ∼ h− /mc2 , corresponding to a maximum separation x ∼ h/mc. −

Hence, the typical density of the vacuum fields is given by ρ ∼ m/x 3 ≈ c3 m 4 /h− 3 . The mass density
in the vacuum fields is unchanging with cosmic epoch and so, adopting the Planck mass for m (see
Table 15.1), the mass density corresponds to about 1097 kg m−3 .
4 A popular account of the history of the development of ideas about the inflationary picture of the
early Universe is contained in Alan Guth’s book The Inflationary Universe: The Quest for a New
Theory of Cosmic Origins (Guth, 1997). The pedagogical review published by Charles Lineweaver
(b. 1954) is very readable and can also be strongly recommended. He adopts a healthily sceptical
452 16 The very early Universe

attitude to the concept of inflation and our ability to test inflationary models through confrontation
with observations (Lineweaver, 2005).
5 An excellent survey of these important ideas is contained in the book Cosmological Inflation and
Large-Scale Structure by Liddle and Lyth (2000).
6 An excellent survey of many of these ideas is contained within the volume celebrating Stephen
Hawking’s 60th birthday, The Future of Theoretical Physics and Cosmology (Gibbons, Shellard
and Rankin, 2003).
References

Aaronson, M. and Mould, J. (1983). A distance scale from the infrared magnitude/H I
velocity–width relation. IV – The morphological type dependence and scatter in the
relation; the distances to nearby groups, Astrophysical Journal, 265, 1–17.
Abbot, C. G. (1924). Radiometer observations of stellar energy spectra, Astrophysical Jour-
nal, 60, 87–107.
Abdurashitov, J. N., Bowles, T. J., Cleveland, B. T. et al. (2002). Results of the SAGE
experiment, Journal of Experimental and Theoretical Physics, 95, 181–193.
Abdurashitov, J. N., Bowles, T. J., Cleveland, B. T. et al. (2003). Measurement of the solar
neutrino capture rate, Nuclear Physics B Proceedings Supplements, 118, 39–46.
Abell, G. O. (1958). The distribution of rich clusters of galaxies, Astrophysical Journal
Supplement, 3, 221–288.
(1962). Membership of clusters of galaxies, in Problems of Extragalactic Research, ed.
McVittie, G. C. (New York: Macmillan), pp. 213–238.
Abell, G. O., Corwin Jr, H. G. and Olowin, R. P. (1989). A catalogue of rich clusters of
galaxies, Astrophysical Journal Supplement, 70, 1–138.
Abraham, R. G., Tanvir, N. R., Santiago, B., Ellis, R. S., Glazebrook, K. and van den Bergh,
S. (1996). Galaxy morphology to I = 25 mag in the Hubble Deep Field, Monthly
Notices of the Royal Astronomical Society, 279, L47–L52.
Abramowicz, M. A., Jaroszyński, M. and Sikora, M. (1978). Relativistic, accreting disks,
Astronomy and Astrophysics, 63, 221–224.
Adams, F. C., Lada, C. J. and Shu, F. H. (1987). Spectral evolution of young stellar objects,
Astrophysical Journal, 312, 788–806.
Adams, W. S. (1914). An A-type star of very low luminosity, Publications of the Astronom-
ical Society of the Pacific, 26, 198.
(1915). The spectrum of the companion of Sirius, Publications of the Astronomical
Society of the Pacific, 27, 236–237.
(1925a). The relativity displacement of the spectral lines in the companion of Sirius,
Proceedings of the National Academy of Sciences, 11, 382–387.
(1925b). The relativity displacement of the spectral lines in the companion of Sirius,
Observatory, 48, 336–342.
Adams, W. S. and Joy, A. H. (1919). The motions in space of some stars of high radial
velocity, Astrophysical Journal, 49, 179–185.
Adams, W. S. and Kohlschütter, A. (1914a). The radial velocities of one hundred stars with
measured parallaxes, Astrophysical Journal, 39, 341–349.

453
454 References

Adams, W. S. and Kohlschütter, A. (1914b). Some spectral criteria for the determination of
absolute stellar magnitudes, Astrophysical Journal, 40, 385–398.
Afonso, C., Albert, J. N., Andersen, J. et al. (2003). Limits on Galactic dark matter with
5 years of EROS SMC data, Astronomy and Astrophysics, 400, 951–956.
Ahmad, Q. R., Allen, R. C., Andersen, T. C. et al. (2001). Measurement of the rate of
νe + d → p + p + e− interactions produced by 8 B solar neutrinos at the Sudbury
Neutrino Observatory, Physical Review Letters, 87, 071301(1–6).
Aitken, D. K., Smith, C. H., James, S. D., Roche, P. F., Hyland, A. R. and McGregor, P. J.
(1988). 10 micron spectral observations of SN 1987A – the first year, Monthly Notices
of the Royal Astronomical Society, 235, 19P–31P.
Akerib, D. S., Alvaro-Dean, J., Armel-Funkhouser et al. (2004). First results from the
cryogenic dark matter search in the Soudan Underground Lab, Physical Review Letters,
93, 211301(1–5).
Albrecht, A. and Steinhardt, P. J. (1982). Cosmology for grand unified theories with radia-
tively induced symmetry breaking, Physical Review Letters, 48, 1220–1223.
Alcock, C., Akerlof, C. W., Allsman, R. A. et al. (1993a). Possible gravitational microlensing
of a star in the Large Magellanic Cloud, Nature, 365, 621–623.
Alcock, C., Allsman, R. A., Axelrod, T. S. et al. (1993b). The MACHO project – a search
for the dark matter in the Milky-Way, in Sky Surveys: Protostars to Protogalaxies,
ed. Soifer, T. (San Francisco: Astronomical Society of the Pacific Conference Series),
pp. 291–296.
Alcock, C., Allsman, R. A., Alves, D. R. et al. (2000). The MACHO project: microlensing
results from 5.7 years of Large Magellanic Cloud observations, Astrophysical Journal,
542, 281–307.
Alfvén, H. and Herlofson, N. (1950). Cosmic radiation and radio stars, Physical Review,
78, 616.
Alfvén, H. and Klein, O. (1962). Matter–antimatter annihilation and cosmology, Arkiv für
Fyzik, 23, 187–194.
Alpher, R. A. and Herman, R. C. (1948). Evolution of the Universe, Nature, 162, 774–775.
(1950). Theory of the origin and relative distribution of the elements, Reviews of Modern
Physics, 22, 153–212.
Alpher, R. A., Bethe, H. and Gamow, G. (1948). The origin of the chemical elements,
Physical Review, 73, 803–804.
Alpher, R. A., Follin, J. W. and Herman, R. C. (1953). Physical conditions in the initial
stages of the expanding Universe, Physical Review, 92, 1347–1361.
Ambartsumian, V. A. (1947). Stellar Evolution and Astrophysics (Yerevan: Armenian
Academy of Sciences).
Anders, E. (1963). Meteorite ages, in The Moon, Meteorites and Comets – The Solar System
IV, eds Middlehurst, B. M. and Kuiper, G. P. (Chicago: University of Chicago Press),
pp. 402–495.
Andersen, M. I., Hjorth, J., Pedersen, H. et al. (2000). VLT identification of the optical after-
glow of the gamma-ray burst GRB 000131 at z = 4.50, Astronomy and Astrophysics,
364, L54–L61.
References 455

Anderson, C. D. (1932). The apparent existence of easily deflected positives, Science, 76,
238–239.
Anderson, C. D. and Neddermeyer, S. H. (1936). Cloud chamber observations of cosmic
rays at 4300 metres elevation and near sea-level, Physical Review, 50, 263–271.
Anderson, W. (1929). Gewöhnliche Materie und Strahlende Energie als Verschiedene
‘Phasen’ eines und Desselben Grundstoffes (Ordinary matter and radiation energy
as different phases of the same underlying matter), Zeitschrift für Physik, 54,
433–444.
Ando, H. and Osaki, Y. (1975). Nonadiabatic nonradial oscillations: an application to the
five-minute oscillation of the Sun, Publications of the Astronomical Society of Japan,
27, 581–603.
Antonucci, R. R. (1993). Unified models for active galactic nuclei and quasars, Annual
Review of Astronomy and Astrophysics, 31, 473–521.
Antonucci, R. R. and Miller, J. S. (1985). Spectropolarimetry and the nature of NGC 1068,
Astrophysical Journal, 297, 621–632.
Aragòn-Salamanca, A., Ellis, R. S., Couch, W. J. and Carter, D. (1993). Evidence for
systematic evolution in the properties of galaxies in distant clusters, Monthly Notices
of the Royal Astronomical Society, 262, 764–794.
Archer, F. S. (1851). On the use of collodion in photography, The Chemist, 2 (March)
257–258.
Argelander, H. (1838). Ueber die eigene Bewegung des Sonnensystems (On the proper
motion of the Solar System), Astronomische Nachrichten, 16, 45–48.
Arnett, W. D. and Clayton, D. D. (1970). Explosive nucleosynthesis in stars, Nature, 227,
780–784.
Arp, H. C. (1966). Atlas of Peculiar Galaxies (Pasadena: California Institute of Technology).
Arp, H. C., Baum, W. A. and Sandage, A. R. (1952). The HR diagrams for the globular
clusters M92 and M3, Astronomical Journal, 57, 4–5.
Arp, H. C., Madore, B. F. and Roberton, W. E. (1987). A Catalogue of Southern Peculiar
Galaxies and Associations (Cambridge: Cambridge University Press).
Atkinson, R. d’E. (1931a). Atomic synthesis and stellar energy I, Astrophysical Journal,
73, 250–295.
(1931b). Atomic synthesis and stellar energy II, Astrophysical Journal, 73, 308–347.
(1936). Atomic synthesis and stellar energy III, Astrophysical Journal, 84, 73–84.
Atkinson, R. d’E. and Houtermans, F. G. (1929). Zur Frage der Aufbaumöglichkeit der
Elemente in Sternen (On the possible synthesis of the elements in stars), Zeitschrift
für Physik, 54, 656–665.
Auger, P., Ehrenfest Jr, P., Maze, R., Daudin, J., Robley, X. and Fréon, A. (1939). Extensive
air showers, Reviews of Modern Physics, 11, 288–291.
Avery, L. W., Broten, N. W., Macleod, J. M., Oka, T. and Kroto, H. W. (1976). Detection of
the heavy interstellar molecule cyanodiacetylene, Astrophysical Journal, 205, L173–
L175.
Axford, W. I., Leer, E. and Skadron, G. (1977). The acceleration of cosmic rays by shock
waves, Proceedings of the 15th International Cosmic Ray Conference, 11, 132–135.
456 References

Baade, W. A. (1926). Über eine Möglichkeit, die Pulsationstheorie der δ-Cephei-

Veränderlichen zu Prüfen (On a possible method of testing the pulsation theory of
the variations of δ-Cephei), Astronomische Nachrichten, 228, 359–362.
(1944). The resolution of Messier 32, NGC 205, and the central region of the Andromeda
Nebula, Astrophysical Journal, 100, 137–146.
(1951). Galaxies – present day problems, Publications of the Observatory of the Univer-
sity of Michigan, 10, 7–17.
(1952). A revision of the extra-galactic distance scale, Transactions of the International
Astronomical Union, 8, 397–398.
(1956). Polarization in the jet of Messier 87, Astrophysical Journal, 123, 550–551.
Baade, W. A. and Minkowski, R. (1954). Identification of the radio sources in Cassiopeia,
Cygnus A, and Puppis A, Astrophysical Journal, 119, 206–214.
Baade, W. A. and Zwicky, F. (1934a). On super-novae, Proceedings of the National Academy
of Sciences, 20, 254–259.
(1934b). Cosmic rays from super-novae, Proceedings of the National Academy of Sci-
ences, 20, 259–263.
(1938). Photographic light-curves of the two supernovae in IC 4182 and NGC 1003,
Astrophysical Journal, 88, 411–421.
Bahcall, J. N. (1964). Solar neutrinos. I. Theoretical, Physical Review Letters, 12, 300–302.
(1989). Neutrino Astrophysics (Cambridge: Cambridge University Press).
Bahcall, J. N. and Bethe, H. (1990). A solution of the solar neutrino problem, Physical
Review Letters, 65, 2233–2235.
Bahcall, J. N. and Davis Jr, R. (1976). Solar neutrinos: a scientific puzzle, Science, 191,
264–267.
Bahcall, J. N. and Ulrich, R. (1988). Solar models, neutrino experiments and helioseismol-
ogy, Reviews of Modern Physics, 60, 297–372.
Bahcall, J. N., Greenstein, J. L. and Sargent, W. L. W. (1968). The absorption-line spec-
trum of the quasi-stellar radio source PKS 0237-23, Astrophysical Journal, 153,
689–698.
Bahcall, J. N., Pinsonneault, M. H., Basu, S. and Christensen-Dalsgaard, J. (1997). Are
standard solar models reliable?, Physical Review Letters, 78, 171–174.
Bahcall, N. A. (2000). Clusters and cosmology, Physics Reports, 333, 233–244.
Balbus, S. A. and Hawley, J. F. (1991). A powerful local shear instability in weakly magne-
tized Disks. I – Linear analysis. II – Nonlinear evolution, Astrophysical Journal, 376,
214–233.
Baldwin, J. E., Beckett, M. G., Boysen, R. C. et al. (1996). The first images from an optical
aperture synthesis array: mapping of Capella with COAST at two epochs, Astronomy
and Astrophysics, 306, L13–L16.
Balmer, J. J. (1885). Note on the spectral lines of hydrogen, Annalen der Physik und Chemie,
25, 80–87.
Bardeen, J. M. (1970). Kerr metric black holes, Nature, 226, 64–65.
(1980). Gauge-invariant cosmological perturbations, Physical Review D, 22, 1882–
1905.
References 457

Bardeen, J. M., Bond, J. R., Kaiser, N. and Szalay, A. S. (1986). The statistics of peaks of
gaussian random fields, Astrophysical Journal, 304, 15–61.
Barkla, C. G. (1906). Polarisation of secondary Röntgen radiation, Proceedings of the Royal
Society of London, A77, 247–255.
Barnard, E. E. (1919). On the dark markings of the sky. With a catalogue of 182 such objects,
Astrophysical Journal, 49, 1–24.
Barrow, J. D. and Tipler, F. J. (1986). The Anthropic Cosmological Principle (Oxford:
Oxford University Press).
Barthel, P. D. (1989). Is every quasar beamed?, Astrophysical Journal, 336, 606–611.
(1994). Unified schemes of FR2 radio galaxies and quasars, in First Stromlo Symposium:
Physics of Active Galactic Nuclei, ASP Conference Series, vol. 54, eds Bicknell, G. V.,
Dopita, M. A. and Quinn, P. J. (San Francisco: ASP), pp. 175–186.
Baum, W. A., Johnson, F. S., Oberly, J. J., Rockwood, C. C., Strain, C. V. and Tousey, R.
(1946). Solar ultraviolet spectrum to 88 kilometers, Physical Review, 70, 781–782.
Bautz, L. and Morgan, W. W. (1970). On the classification of the forms of clusters of
galaxies, Astrophysical Journal Letters, 162, L149–L153.
Baym, G., Pethick, C. and Pines, D. (1969a). Superfluidity in neutron stars, Nature, 224,
673–674.
Baym, G., Pethick, C., Pines, D. and Ruderman, M. (1969b). Spin up in neutron stars: the
future of the Vela pulsar, Nature, 224, 872–874.
Baym, G., Bethe, H. A. and Pethick, C. J. (1971a). Neutron star matter, Nuclear Physics A,
175, 225–271.
Baym, G., Pethick, C. J. and Sutherland, P. (1971b). The ground state of matter at high
densities: equation of state and stellar models, Astrophysical Journal, 170, 299–317.
Becker, R. H., Fan, X., White, R. L. et al. (2001). Evidence for reionisation at z ∼ 6:
detection of a Gunn–Peterson trough in a z = 6.28 quasar, Astronomical Journal, 122,
2850–2857.
Becklin, E. E. and Neugebauer, G. (1967). Observations of an infrared star in the Orion
Nebula, Astrophysical Journal, 147, 799–802.
(1968). Infrared observations of the Galactic center, Astrophysical Journal, 151, 145–161.
Becquerel, A. E. (1842). Mémoire sur la constitution du spectre solaire (Memoir concerning
the constitution of the solar spectrum), Bibliothéque Universelle de Gèneve, 40, 341–
367.
Becquerel, H. (1896). Sur les radiations invisibles émises par les corps phosphorescents
(On the invisible radiation emitted by phosphorescent bodies), Comptes Rendus de
l’Academie des Sciences, 122, 501–503.
Bell, A. R. (1978). The acceleration of cosmic rays in shock fronts. I, Monthly Notices of
the Royal Astronomical Society, 182, 147–156.
Bennett, A. S. (1962). The Revised 3C Catalogue of radio sources, Memoirs of the Royal
Astronomical Society, 67, 163–172.
Bennett, C. L., Banday, A. J., Górski, K. M. et al. (1996). Four-year COBE DMR cosmic
microwave background observations: maps and basic results, Astrophysical Journal,
464, L1–L4.
458 References

Bennett, C. L., Halpern, M., Hinshaw, G. et al. (2003). First-year Wilkinson Microwave
Anisotropy Probe (WMAP) observations: preliminary maps and basic results, Astro-
physical Journal Supplement Series, 148, 1–27.
Benôit, A., Ade, P., Amblard, A. et al. (2003a). The cosmic microwave background
anisotropy power spectrum measured by Archeops, Astronomy and Astrophysics, 399,
L19–L23.
(2003b). Cosmological constraints from Archeops, Astronomy and Astrophysics, 399,
L25–L30.
Benson, B. A., Church, S. E., Ade, P. A. R. et al. (2003). Peculiar velocity limits from
measurements of the spectrum of the Sunyaev–Zel’dovich effect in six clusters of
galaxies, Astrophysical Journal, 592, 674–691.
Bergeron, J. (1988). Properties of the heavy-element absorption systems, in QSO Absorption
Lines: Probing the Universe, eds Blades, J. C., Turnshek, D. and Norman, C. A.
(Cambridge: Cambridge University Press), pp. 127–143.
Bersanelli, M., Bouchet, F. R., Efstathiou, G. et al. (1995). Phase A Study for the
Cobras/Samba Mission (Paris: European Space Agency), D/SCI(96)3.
Bertola, F. and Capaccioli, M. (1975). Dynamics of early type galaxies. I – The rotation
curve of the elliptical galaxy NGC 4697, Astrophysical Journal, 200, 439–445.
Bertola, F. and Galletta, G. (1979). Ellipticity and twisting of isophotes in elliptical galaxies,
Astronomy and Astrophysics, 77, 363–365.
Bertola, F., Bettoni, D., Danziger, J., Sadler, E., Spark, L. and de Zeeuw, T. (1991). Testing
the gravitational field in elliptical galaxies: NGC 5077, Astrophysical Journal, 373,
369–390.
Bessel, F. W. (1839). Bestimmung der Entfernung des 61sten Sterne des Schwans (Determi-
nation of the distance of 61-Cygni), Astronomische Nachrichten, 16, 64–96.
Best, P. N., Longair, M. S. and Röttgering, H. J. A. (1996). Evolution of the aligned structures
in z ∼ 1 radio galaxies, Monthly Notices of the Royal Astronomical Society, 280, L9–
L12.
(1998). HST, radio and infrared observations of 28 3CR radio galaxies at redshift z ∼ 1.
II – Old stellar populations in central cluster galaxies, Monthly Notices of the Royal
Astronomical Society, 295, 549–567.
(2000). Ionization, shocks and evolution of the emission-line gas of distant 3CR radio
galaxies, Monthly Notices of the Royal Astronomical Society, 311, 23–36.
Bethe, H. A. (1939). Energy production in stars, Physical Review, 55, 434–456.
Bethe, H. A. and Critchfield, C. L. (1938). The formation of deuterons by proton combina-
tion, Physical Review, 54, 248–254.
Binney, J. (1978). On the rotation of elliptical galaxies, Monthly Notices of the Royal
Astronomical Society, 183, 501–514.
Birkinshaw, M. (1990). Observations of the Sunyaev–Zeldovich effect, in The Cosmic
Microwave Background: 25 Years Later, eds Mandolesi, N. and Vittorio, N. (Dordrecht:
Kluwer Academic Publishers), pp. 77–94.
Blackett, P. M. S. (1925). The ejection of protons from nitrogen nuclei, photographed by
the Wilson method, Proceedings of the Royal Society of London, A107, 349–360.
References 459

Blackett, P. M. S. (1948). A possible contribution to the light of the night sky from the
Cherenkov radiation emitted by cosmic rays, in The Emission Spectra of the Night
Sky and Aurorae, Gassiot Committee Report (London: Physical Society of London),
pp. 34–35.
Blackett, P. M. S. and Occhialini, G. P. S. (1933). Some photographs of the tracks of
penetrating radiation, Proceedings of the Royal Society of London, A139, 699–722.
Blain, A. W. and Longair, M. S. (1993). Sub-millimetre cosmology, Monthly Notices of the
Royal Astronomical Society, 264, 509–521.
Blandford, R. D. and Ostriker, J. P. (1978). Particle acceleration by astrophysical shocks,
Astrophysical Journal, 221, L29–L32.
Blandford, R. D. and Znajek, R. L. (1977). Electromagnetic extraction of energy from Kerr
black holes, Monthly Notices of the Royal Astronomical Society, 179, 433–456.
Blewett, J. P. (1946). Radiation losses in the induction electron accelerator, Physical Review,
69, 87–95.
Bludman, S. A. and Ruderman, M. A. (1977). Induced cosmological constant expected
above the phase transition restoring the broken symmetry, Physical Review Letters,
38, 255–257.
Blumenthal, G. R., Faber, S. M., Primack, J. R. and Rees, M. J. (1984). Formation of galaxies
and large-scale structure with cold dark matter, Nature, 311, 517–525.
Bohr, N. (1913a). On the constitution of atoms and molecules: Part 1, On the constitution
of atoms and molecules, Philosophical Magazine, Sixth Series, 23, 1–25.
(1913b). On the constitution of atoms and molecules: Part 2, Systems containing only a
single nucleus, Philosophical Magazine, Sixth Series, 23, 476–502.
(1913c). On the constitution of atoms and molecules: Part 3, Systems containing several
nuclei, Philosophical Magazine, Sixth Series, 23, 857–875.
Böhringer, H. (1994). Clusters of galaxies, in Frontiers of Space and Ground-based Astron-
omy, eds Wamsteker, W., Longair, M. S. and Kondo, Y. (Dordrecht: Kluwer Academic
Publishers), pp. 359–368.
Boksenberg, A. (1997). Quasar absorption lines: reflections and views, in The Hubble Space
Telescope and the High Redshift Universe, eds Tanvir, N. R., Aragón-Salamanca, A.
and Wall, J. V. (Singapore: World Scientific Publishing Company), pp. 283–294.
Bolte, M. (1997). Globular clusters: old, in Critical Dialogues in Cosmology, ed. Turok, N.
(Singapore: World Scientific), pp. 156–168.
Bolton, C. T. (1972). Identifications of Cyg X-1 with HDE 226868, Nature, 235, 271–273.
Bolton, J. and Stanley, G. J. (1949). The position and probable identification of the source of
galactic radio-frequency radiation Taurus A, Australian Journal of Scientific Research,
A2, 139–148.
Bolton, J. G., Stanley, G. J. and Slee, O. B. (1949). Positions of three discrete sources of
galactic radio-frequency radiation, Nature, 164, 101–102.
Bolyai, J. (1832). Appendix: Scientiam spatii absolute veritam exhibens (Appendix explain-
ing the absolutely true science of space). Published as an appendix to the essay by his
father, F. Bolyai, An Attempt to Introduce Studious Youth to the Elements of Pure
Mathematics (Maros Vásárhely, Transylvania).
460 References

Bondi, H. and Gold, T. (1948). The steady-state theory of the expanding Universe, Monthly
Notices of the Royal Astronomical Society, 108, 252–270.
Bosma, A. (1981). 21-cm line studies of spiral galaxies II. The distribution and kinematics
of neutral hydrogen in spiral galaxies of various morphological types, Astronomical
Journal, 86, 1825–1846.
Boss, B. (1918). Real stellar motions (abstract), Popular Astronomy, 26, 686.
Bothe, W. and Kolhörster, W. (1929). The nature of high-altitude radiation, Zeitschrift für
Physik, 56, 751–777.
Bowen, I. S. (1927). The origin of the chief nebular lines, Publications of the Astronomical
Society of the Pacific, 39, 295–297.
(1928). The origin of the nebular lines and the structure of the planetary nebulae, Astro-
physical Journal, 67, 1–15.
Bowyer, C. S., Field, G. B. and Mack, J. E. (1968). Detection of an anisotropic soft X-ray
background, Nature, 217, 32–34.
Bowyer, S., Byram, E. T., Chubb, T. A. and Friedman, H. (1964). Lunar occulation of X-ray
emission from the Crab Nebula, Science, 146, 912–917.
Boyle, B. J., Jones, L. R., Shanks, T., Marano, B., Zitelli, V. and Zamorani, G. (1991).
QSO evolution and clustering at z < 2.9, Proceedings of the Workshop on The Space
Distribution of Quasars: Astronomical Society of the Pacific Conference Series, 21,
191–201.
Boyle, B. J., Shanks, T., Croom, S. M. et al. (2000). The 2dF QSO redshift survey – I.
The optical luminosity function of quasi-stellar objects, Monthly Notices of the Royal
Astronomical Society, 317, 1014–1022.
Boyle, W. S. and Smith, G. E. (1970). Charge coupled semiconductor devices, Bell System
Technical Journal, 49, 587–593.
Braccesi, A., Formiggini, L. and Gandolfi, E. (1970). Magnitudes, colours and coordinates
of 175 ultraviolet excess objects in the field 13h , +36◦ , Astronomy and Astrophysics,
5, 264–279.
Bracewell, R. N., ed. (1959). Paris Symposium on Radio Astronomy (Stanford: Stanford
University Press).
Bracewell, R. N. and Roberts, J. A. (1954). Aerial smoothing in radio astronomy, Australian
Journal of Physics, 7, 615–640.
Bradley, J. (1728). An account of a new discovered motion of the fixed stars, Philosophical
Transactions of the Royal Society, 35, 637–661.
Bradt, H. L. and Peters, B. (1950). Abundance of lithium, beryllium, boron and other light
nuclei in the primary cosmic radiation and the problem of cosmic-ray origin, Physical
Review, 80, 943–953.
Braes, L. L. E. and Miley, G. K. (1971). Radio emission from Scorpius X-1 at 21.2 cm,
Astronomy and Astrophysics, 14, 160–163.
Branch, D. and Patchett, B. (1973). Type I supernovae, Monthly Notices of the Royal Astro-
nomical Society, 161, 71–83.
Branch, D. and Tammann, G. A. (1992). Type I supernovae as standard candles, Annual
Review of Astronomy and Astrophysics, 30, 359–389.
References 461

Brans, C. and Dicke, R. H. (1961). Mach’s principle and a relativistic theory of gravitation,
Physical Review, 124, 925–935.
Brooks, J. R., Isaak, G. R. and van der Raay, H. B. (1976). Observatons of free oscillations
of the Sun, Nature, 259, 92–95.
Broten, N. W., Legg, T. H., Locke, J. L. et al. (1967). Radio interferometry with a baseline
of 3074 km, Astrophysical Journal, 72, 787–800.
Broten, N. W., Oka, T., Avery, L. W., Macleod, J. M. and Kroto, H. W. (1978). The detection
of HC9 N in interstellar space, Astrophysical Journal, 223, L105–L107.
Brown, T. M., Stebbins, R. T. and Hill, H. A. (1976). Observed oscillations of the apparent
solar diameter, in Solar and Stellar Pulsation Conference, eds Cox, A. N. and Dupree,
R. G. (Los Alamos: Los Alamos Scientific Laboratory), pp. 1–6.
Bruzual, G. and Charlot, S. (2003). Stellar population synthesis at the resolution of 2003,
Monthly Notices of the Royal Astronomical Society, 344, 1000–1028.
Burbidge, E. M., Burbidge, G. R., Fowler, W. A. and Hoyle, F. (1957). Synthesis of the
elements in stars, Reviews of Modern Physics, 29, 547–650.
Burbidge, E. M., Burbidge, G. R. and Prendergast, K. H. (1960). The rotation, mass distri-
bution and mass of NGC 5866, Astrophysical Journal, 131, 282–292.
(1965). The rotation and mass of the SA Galaxy NGC 681, Astrophysical Journal, 142,
154–159.
Burbidge, E. M., Burbidge, G. R. and Sandage, A. R. (1963). Evidence for the occurence
of violent events in the nuclei of galaxies, Reviews of Modern Physics, 35, 947–
972.
Burbidge, E. M., Lynds, C. R. and Burbidge, G. R. (1966). On the measurement and
interpretation of absorption features in the spectrum of the quasi-stellar object 3C 191,
Astrophysical Journal, 144, 447–451.
Burbidge, E. M., Lynds, C. R. and Stockton, A. N. (1968). Further observations of quasi-
stellar objects with absorption-line spectra: Ton 1530, PKS 0237-23, and PHL 938,
Astrophysical Journal, 152, 1077–1093.
Burbidge, G. R. (1959). Estimates of the total energy in particles and magnetic field in the
non-thermal radio sources, Astrophysical Journal, 129, 849–851.
Burbidge, G. R. and Burbidge, E. M. (1967). Quasi-stellar Objects (New York: Freeman
and Company).
Butcher, H. and Oemler, Jr., A. (1978). The evolution of galaxies in clusters. I – ISIT
photometry of Cl 0024+1654 and 3C 295, Astrophysical Journal, 219, 18–30.
(1984). The evolution of galaxies in clusters. V – A study of populations since z ∼ 0.5,
Astrophysical Journal, 285, 426–438.
Cameron, A. G. W. (1957). Nuclear reactions in stars and nucleogenesis, Publications of
the Astronomical Society and the Pacific, 69, 201–222.
(1958). Nuclear astrophysics, Annual Review of Nuclear Science, 8, 299–326.
Campbell, W. W. (1901). A preliminary determination of the motion of the Solar System,
Astrophysical Journal, 13, 80–89.
(1910). Some peculiarities in the motions of the stars, Lick Observatory Bulletin 6,
No. 196, 125–133.
462 References

Campbell, W. W. and Moore, J. H. (1928). Radial velocities of stars brighter than visual
magnitude 5.51 as determined at Mount Hamilton and Santiago, Publications of the
Lick Observatory, 16, 1–399.
Cannon, A. J. and Pickering, E. C. (1901). Spectra of bright southern stars photographed
with the 13-inch Boyden telescope as part of the Henry Draper Memorial, Annals of
the Harvard College Observatory (Part II), 28, 131–263.
Carlstrom, J. E., Joy, M. K., Grego, L. et al. (2000). Imaging the Sunyaev–Zel’dovich
effect, in Particle Physics and the Universe: Proceedings of Nobel Symposium 198,
eds Bergström, L., Carlson, P. and Fransson, C. (Stockholm: Physica Scripta), pp. 148–
155.
Carroll, S. M., Press, W. H. and Turner, E. L. (1992). The cosmological constant, Annual
Review of Astronomy and Astrophysics, 30, 499–542.
Carter, B. (1971). Axisymmetric black hole has only two degrees of freedom, Physical
Review Letters, 26, 331–333.
(1974). Large number coincidences and the anthropic principle in cosmology, in Con-
frontation of Cosmological Theories with Observational Data, IAU Symposium no. 63,
ed. Longair, M. S. (Dordrecht: D. Reidel Publishing Company), pp. 291–298.
Chaboyer, B. (1998). The age of the Universe, Physics Reports, 307, 23–30.
Chadwick, J. (1932). Possible existence of a neutron, Nature, 129, 312.
Chambers, K. C., Miley, G. K. and van Breugel, W. J. M. (1987). Alignment of radio and
optical orientations in high-redshift radio galaxies, Nature, 329, 604–606.
Chandrasekhar, S. (1931). The maximum mass of ideal white dwarfs, Astrophysical Journal,
74, 81–82.
(1939). An Introduction to the Study of Stellar Structure (Chicago: Chicago University
Press).
(1943a). Dynamical friction I. General considerations: the coefficient of dynamical fric-
tion, Astrophysical Journal, 97, 255–262.
(1943b). Dynamical friction II. The rate of escape of stars from clusters and the evidence
for the operation of dynamical friction, Astrophysical Journal, 97, 263–273.
(1943c). Dynamical friction III. A more exact theory of the rate of escape of stars from
clusters, Astrophysical Journal, 98, 54–60.
(1946). On the radiative equilibrium of a stellar atmosphere. X., Astrophysical Journal,
103, 351–370.
(1980). The role of general relativity in astronomy: retrospect and prospect, Highlights
of Astronomy, vol. 5, ed. Wayman, P. A. (Dordrecht: D. Reidel Publishing Company),
pp. 45–61.
(1981). Hydrodynamic and Hydromagnetic Stability (New York: Dover Publications).
Chandrasekhar, S. and Henrich, L. R. (1942). An attempt to interpret the relative abundances
of the elements and their isotopes, Astrophysical Journal, 95, 288–298.
Charbonneau, D., Brown, T. M., Latham, D. W. and Mayor, M. (2000). Detection of planetary
transits across a Sun-like star, Astrophysical Journal, 529, L45–L48.
Charbonneau, D., Brown, T. M., Noyes, R. W. and Gilliland, R. L. (2002). Detection of an
extrasolar planet atmosphere, Astrophysical Journal, 568, 377–384.
References 463

Charles, P. (1998). Black holes in our Galaxy: observations, in Theory of Black Hole Accre-
tion Disks, eds Abramowicz, M. A., Björnsson, G. and Pringle, J. E. (Cambridge:
Cambridge University Press), pp. 1–20.
Cheung, A. C., Rank, D. M., Townes, C. H., Thornton, D. D. and Welch, W. J. (1968).
Detection of NH3 molecules in the interstellar medium by their microwave emission,
Physical Review Letters, 221, 1701–1705.
(1969). Detection of water in interstellar regions by its microwave radiation, Nature, 221,
626–628.
Chevalier, R. A. (1992). Supernova 1987A at five years of age, Nature, 355, 691–696.
Christensen-Dalsgaard, J. and Gough, D. O. (1980). Is the Sun helium deficient?, Nature,
288, 544–547.
Christy, R. F. (1964). The calculation of stellar pulsation, Reviews of Modern Physics, 36,
555–571.
(1968). The theory of Cepheid variables, Quarterly Journal of the Royal Astronomical
Society, 9, 13–39.
Clark, G. W., Garmire, G. P. and Kraushaar, W. L. (1968). Observation of high-energy
cosmic gamma rays, Astrophysical Journal Letters, 153, L203–L207.
Clavel, J., Reichert, G. A., Alloin, D. et al. (1991). Steps toward determination of the size and
structure of the broad-line region in active galactic nuclei. I – An 8 month campaign
of monitoring NGC 5548 with IUE, Astrophysical Journal, 366, 64–81.
Cockcroft, J. D. and Walton, E. T. S. (1932). Disintegration of lithium by swift protons,
Nature, 129, 649.
Cocke, W. J., Disney, M. J. and Taylor, D. J. (1969). Discovery of optical signals from pulsar
NP 0532, Nature, 221, 525–527.
Cohen, M. H., Cannon, W., Purcell, G. H. et al. (1971). The small-scale structure of radio
galaxies and quasar-stellar sources at 3.8 centimetres, Astrophysical Journal, 170,
207–218.
Coles, P. and Lucchin, F. (1995). Cosmology – The Origin and Evolution of Cosmic Structure
(Chichester: John Wiley & Sons).
Colgate, S. A. and McKee, C. (1969). Early supernova luminosity, Astrophysical Journal,
157, 623–643.
Colless, M., Dalton, G., Maddox, S. et al. (2001). The 2dF galaxy redshift survey: spec-
tra and redshifts, Monthly Notices of the Royal Astronomical Society, 328, 1039–
1063.
Comstock, G. C. (1897). On the application of interference methods to the determination
of the effective wavelength of starlight, Astrophysical Journal, 5, 26–35.
Connolly, A. J., Szalay, A. S., Dickinson, M., SubbaRao, M. U. and Brunner, R. J. (1997).
The evolution of the global star formation history as measured from the Hubble Deep
Field, Astrophysical Journal, 486, L11–L14.
Costa, E., Frontera, F., Heise, J. et al. (1997). Discovery of an X-ray afterglow associated
with the gamma-ray burst of 28 February 1997, Nature, 387, 783–785.
Cowan, J. J., Thielemann, F.-K. and Truran, J. W. (1991). Radioactive dating of the elements,
Annual Reviews of Astronomy and Astrophysics, 29, 447–497.
464 References

Cowan Jr, C. L., Reines, F., Harrison, F. B., Kruse, H. W. and McGuire, A. D. (1956).
Detection of the free neutrino: a confirmation, Science, 124, 103–104.
Cowie, L. L., Hu, E. M. and Songaila, A. (1995). Detection of massive forming galaxies at
redshifts z > 1, Nature, 377, 603–605.
Cowie, L. L., Lilly, S. J., Gardner, J. and McLean, I. S. (1988). A cosmologically significant
population of galaxies dominated by very young star formation, Astrophysical Journal,
332, L29–L32.
Cowie, L. L., Songaila, A., Hu, E. M. and Cohen, J. D. (1996). New insight on galaxy for-
mation and evolution from Keck spectroscopy of the Hawaii deep fields, Astronomical
Journal, 112, 839–864.
Cowley, A. P. (1992). Evidence for black holes in stellar binary systems, Annual Reviews
of Astronomy and Astrophysics, 30, 287–310.
Cowling, T. G. (1935). The stability of gaseous spheres, Monthly Notices of the Royal
Astronomical Society, 96, 42–60.
Cox, A. N. (1965). Stellar absorption coefficients and opacities, in Stellar Structure – Stars
and Stellar Systems: Compendium of Astronomy and Astrophysics, Vol. VIII, eds Aller,
L. H. and McLaughlin, D. B. (Chicago: University of Chicago Press), pp. 195–268.
Cox, D. P. and Smith, B. W. (1974). Large-scale effects of supernova remnants on the Galaxy:
generation and maintenance of a hot network of tunnels, Astrophysical Journal Letters,
189, L105–L108.
Crittenden, R., Bond, R., Davis, R. L., Efstathiou, G. and Steinhardt, P. J. (1993). Imprint
of gravitational waves on the cosmic microwave background, Physical Review Letters,
71, 324–327.
Curie, M. P. and Sklodowska-Curie, M. (1898). On a new radioactive substance contained
in pitchblende, Comptes Rendus, 127, 175–178.
Curie, M. P., Sklodowska-Curie, M. and Bémont, G. (1898). On a new, strongly radioactive
substance, contained in pitchblende, Comptes Rendus, 127, 1215–1217.
Curtis, H. D. (1917). New stars in spiral nebulae, Publications of the Astronomical Society
of the Pacific, 29, 180–181.
(1918a). Descriptions of 762 nebulae and clusters photographed with the Crossley reflec-
tor, Publications of the Lick Observatory, 13, 11–42. The reference to ‘a curious straight
ray’ in NGC 4486 (M87) appears on page 31.
(1918b). A study of occulting matter in the spiral nebulae, Publications of the Lick
Observatory, 13, 45–54.
(1921). The scale of the Universe, Bulletin of the National Research Council, 2, 194–217.
Dashevsky, V. M. and Zeldovich, Y. B. (1964). Propagation of light in a nonhomogeneous
non-flat Universe II, Astronomicheskii Zhurnal, 41, 1071–1074. Translation in Soviet
Astronomy, 8, 1965, 854–856.
Davidson, W. (1962). The cosmological implications of the recent counts of radio sources,
I. Analysis of the results and their immediate interpretation, Monthly Notices of the
Royal Astronomical Society, 123, 425–435.
Davidson, W. and Davies, M. (1964). Interpretation of the counts of radio sources in terms
of a 4-parameter family of evolutionary universes, Monthly Notices of the Royal Astro-
nomical Society, 127, 241–255.
References 465

Davies, R. L., Efstathiou, G., Fall, S. M., Illingworth, G. and Schechter, P. L. (1983). The
kinematic properties of faint elliptical galaxies, Astrophysical Journal, 266, 41–57.
Davis Jr., L. and Greenstein, J. L. (1951). The polarization of starlight by aligned dust grains,
Astrophysical Journal, 114, 206–240.
Davis, M. and Peebles, P. J. E. (1983). A survey of galaxy redshifts. V – The two-point
position and velocity correlations, Astrophysical Journal, 267, 465–482.
Davis, M., Efstathiou, G., Frenk, C. and White, S. D. M. (1985). The evolution of large-
scale structure in a universe dominated by cold dark matter, Astrophysical Journal,
292, 371–394.
(1992a). The end of cold dark matter?, Nature, 356, 489–494.
Davis, M., Geller, M. J. and Huchra, J. (1978). The local mean mass density of the Uni-
verse – new methods for studying Galaxy clustering, Astrophysical Journal, 221,
1–18.
Davis, R. (1955). Attempt to detect the antineutrinos from a nuclear reactor by the
Cl37 (ν − , e− )Ar37 reaction, Physical Review, 97, 766–769.
Davis, R. L., Hodges, H. M., Smoot, G. F., Steinhardt, P. J. and Turner, M. S. (1992b).
Cosmic microwave background probes models of inflation, Physical Review Letters,
69, 1856–1859.
de Bernardis, P., Ade, P. A. R., Bock, J. J. et al. (2000). A flat Universe from high-resolution
maps of the cosmic microwave background radiation, Nature, 404, 955–959.
de Sitter, W. (1917a). On Einstein’s theory of gravitation and its astronomical consequences,
Monthly Notices of the Royal Astronomical Society, 78, 3–28.
(1917b). On the relativity of inertia. Remarks concerning Einstein’s latest hypothesis,
Proceedings of the Royal Academy of Amsterdam, 19, 1217–1225.
de Vaucouleurs, G. (1971). The large-scale distribution of galaxies and clusters of galaxies,
Publications of the Astronomical Society of the Pacific, 83, 113–143.
(1974). Structure, dynamics and statistical properties of galaxies (invited paper), in
The Formation and Dynamics of Galaxies, IAU Symposium 58, ed. Shakeshaft, J. R.
(Dordrecht: D. Reidel Publishing Company), pp. 1–53.
de Vaucouleurs, G., de Vaucouleurs, A., Corwin Jr, H. G., Buta, R. J., Paturel, G. and Fouque,
P. (1991). Third Reference Catalogue of Bright Galaxies: Containing Information on
23,024 Galaxies With Reference to Papers Published Between 1913 and 1988 (Berlin:
Springer-Verlag).
Dekel, A. (1986). Biased galaxy formation, Comments on Astrophysics, 11, 235–256.
Dekel, A. and Rees, M. J. (1987). Physical mechanisms for biased galaxy formation, Nature,
326, 455–462.
Dekel, A., Burstein, D. and White, S. D. M. (1997). Measuring , in Critical Dialogues in
Cosmology, ed. Turok, N. (Singapore: World Scientific), pp. 175–192.
Dennise, J.-F., Le Roux, E. and Steinberg, J. C. (1957). Novelles observations du rayon-
nement du ciel sur la longeur d’onde 33 cm (New observations of the background
radiation at a wavelength of 33 cm), Comptes Rendus, 244, 3030–3033.
Deubner, F.-L. (1975). Observations of low wavenumber nonradial eigenmodes of the Sun,
Astronomy and Astrophysics, 44, 371–375.
Dicke, R. H. (1961). Dirac’s cosmology and Mach’s principle, Nature, 192, 440–441.
466 References

Dicke, R. H. and Peebles, P. J. E. (1979). Big Bang cosmology – enigmas and nostrums, in
General Relativity: An Einstein Centenary Survey, eds Hawking, S. W. and Israel, W.
(Cambridge: Cambridge University Press), pp. 504–517.
Dicke, R. H., Peebles, P. J. E., Roll, P. G. and Wilkinson, D. T. (1965). Cosmic black-body
radiation, Astrophysical Journal, 142, 414–419.
Dingle, H. (1937). Modern Aristotelianism, Nature, 139, 784–786.
Dirac, P. A. M. (1928a). The quantum theory of the electron, Proceedings of the Royal
Society of London, A117, 610–624.
(1928b). The quantum theory of the electron II, Proceedings of the Royal Society of
London, A118, 351–361.
(1931). Quantum singularities in the electromagnetic field, Proceedings of the Royal
Society of London, A133, 60–72.
(1937). The cosmical constants, Nature, 139, 323.
Djorgovski, S. G. and Davis, M. (1987). Fundamental properties of elliptical galaxies,
Astrophysical Journal, 313, 59–68.
Dombrovski, V. A. (1954). On the nature of the radiation from the Crab Nebula, Dokladi
Akademiya Nauk SSSR, 94, 1021–1024.
Doroshkevich, A. G. and Novikov, I. D. (1964). Mean density of radiation in the metagalaxy
and certain problems in relativistic cosmology, Dokladi Akademiya Nauk SSSR, 154,
809–811. Translation in Soviet Physics Doklady, 9, 1964, 111–113.
Doroshkevich, A. G., Novikov, I. D., Sunyaev, R. A. and Zeldovich, Y. B. (1971). Helium
production in the different cosmological models, in Highlights of Astronomy, vol. 2,
ed. de Jager, C. (Dordrecht: D. Reidel Publishing Company), pp. 313–327.
Doroshkevich, A. G., Sunyaev, R. A. and Zeldovich, Y. B. (1974). The formation of galaxies
in Friedmannian universes, in Confrontation of Cosmological Theories with Observa-
tional Data, IAU Symposium no. 63, ed. Longair, M. S. (Dordrecht: D. Reidel Pub-
lishing Company), pp. 213–225.
Doroshkevich, A. G., Zeldovich, Y. B., Sunyaev, R. A. and Khlopov, M. Y. (1980a).
Astrophysical implications of the neutrino rest mass – Part II. The density-perturbation
spectrum and small-scale fluctuations in the microwave background, Pis’ma v Astro-
nomicheskii Zhurnal, 6, 457–464.
(1980b). Astrophysical implications of the neutrino rest mass – Part III. The non-linear
growth of perturbations and hidden mass, Pis’ma v Astronomicheskii Zhurnal, 6,
465–469.
Draine, B. T. (2003). Interstellar dust grains, Annual Reviews of Astronomy and Astrophysics,
41, 241–289.
Draper, H. (1879). On photographing the spectra of the stars and planets, American Journal
of Science and Arts, 18, 419–425.
Dressler, A. (1980). Galaxy morphology in rich clusters – implications for the formation
and evolution of galaxies, Astrophysical Journal, 236, 351–365.
(1984). The evolution of galaxies in clusters, Annual Review of Astronomy and Astro-
physics, 22, 185–222.
Dressler, A., Lynden-Bell, D., Burstein, D. et al. (1987). Spectroscopy and photometry of
elliptical galaxies. I – A new distance estimator, Astrophysical Journal, 313, 42–58.
References 467

Dreyer, J. L. E. (1888). New general catalogue of nebulae and clusters of stars, Memoirs of
the Royal Astronomical Society, 49, 1–237.
(1895). Index catalogue of nebulae, Memoirs of the Royal Astronomical Society, 51,
185–228.
(1908). Index catalogue of nebulae and cluster stars, Memoirs of the Royal Astronomical
Society, 59, 105–198.
Dunbar, D. N. F., Pixley, R. E., Wenzel, W. A. and Whaling, W. (1953). The 7.68-MeV state
of C12 , Physical Review, 92, 649–650.
Dunlop, J. S. (1994). The cosmological evolution of active galaxies, in Frontiers of Space
and Ground-Based Astronomy: The Astrophysics of the 21st Century, eds Wamsteker,
W., Longair, M. S. and Kondo, Y. (Dordrecht: Kluwer Academic Publishers), pp. 395–
407.
Dunlop, J. S. and Peacock, J. A. (1990). The redshift cut-off in the luminosity function of
radio galaxies and quasars, Monthly Notices of the Royal Astronomical Society, 247,
19–42.
Dunlop, J. S., McLure, R. J., Yamada, T. et al. (2004). Discovery of the galaxy counterpart
of HDF 850.1, the brightest submillimetre source in the Hubble Deep Field, Monthly
Notices of the Royal Astronomical Society, 350, 769–784.
Dwek, E., Arendt, R. G., Hauser, M. G. et al. (1995). Morphology, near-infrared luminos-
ity, and mass of the Galactic bulge from COBE DIRBE observations, Astrophysical
Journal, 445, 716–730.
Dyer, C. C. and Roeder, R. C. (1972). The distance–redshift relation for universes with no
intergalactic medium, Astrophysical Journal, 174, L115–L117.
Dyson, F. W., Eddington, A. S. and Davidson, C. (1920). A determination of the deflection
of light by the Sun’s gravitational field, from observations made at the total eclipse of
May 29, 1919, Philosophical Transactions of the Royal Society, 220, 291–333.
Eales, S., Rawlings, S., Law-Green, D., Cotter, G. and Lacy, M. (1997). A first sample of
faint radio sources with virtually complete redshifts. I – Infrared images, the Hubble
diagram and the alignment effect, Monthly Notices of the Royal Astronomical Society,
291, 593–615.
Eastman, R. G. and Kirshner, R. P. (1989). Model atmospheres for SN 1987A and the
distance to the Large Magellanic Cloud, Astrophysical Journal, 347, 771–793.
Eddington, A. S. (1908). On the mathematical theory of two star drifts, and on the systematic
motions of zodiacal stars, Monthly Notices of the Royal Astronomical Society, 68, 588–
605.
(1916a). The kinetic energy of a star cluster, Monthly Notices of the Royal Astronomical
Society, 76, 525–528.
(1916b). On the radiative equilibrium of the stars, Monthly Notices of the Royal Astro-
nomical Society, 77, 16–35.
(1917). Further notes on the radiative equilibrium of the stars, Monthly Notices of the
Royal Astronomical Society, 77, 596–612.
(1920). The internal constitution of the stars, Observatory, 43, 341–358.
(1924). On the relation between the masses and luminosities of the stars, Monthly Notices
of the Royal Astronomical Society, 84, 308–332.
468 References

Eddington, A. S. (1926a). Diffuse matter in interstellar space, Proceedings of the Royal

Society, A111, 424–456.
(1926b). The Internal Constitution of the Stars (Cambridge: Cambridge University Press).
Reprinted 1988.
(1930). On the stability of Einstein’s spherical world, Monthly Notices of the Royal
Astronomical Society, 90, 668–678.
(1935). Remark in the paper ‘Relativistic degeneracy’ read to the meeting of the Royal
Astronomical Society on January 11 1935, Observatory, 58, 37–39.
(1941). Discussion of Sir Arthur Eddington’s contribution, in Novae and White Dwarfs,
ed. Shaler, A. J. (Paris: Herrmann et Cie), pp. 262–267.
(1946). Fundamental Theory (Cambridge: Cambridge University Press).
Edge, D. O., Shakeshaft, J. R., McAdam, W. B., Baldwin, J. E. and Archer, S. (1959). A
survey of radio sources at a frequency of 159 Mc/s, Memoirs of the Royal Astronomical
Society, 68, 37–60.
Efstathiou, G. (1990). Cosmological perturbations, in Physics of the Early Universe, eds
Peacock, J. A., Heavens, A. F. and Davies, A. T. (Edinburgh: SUSSP Publications),
pp. 361–463.
Eggert, J. (1919). Über den Dissoziationszustand der Fixsterngase (On the dissociation state
of gases in the fixed stars), Physikalische Zeitschrift, 20, 570–574.
Eguchi, K., Enomoto, S., Furuno, K. et al. (2003). First results from KamLAND:
evidence for reactor anti-neutrino disappearance, Physical Review Letters, 90,
id. 021802(1–6).
Einasto, J. (2001). Dark matter and large scale structure, in Historical Development of
Modern Cosmology, ASP Conference Series vol. 252, eds Martínez, V. J., Trimble, V.
and Pons-Bordeía, M. J. (San Francisco: ASP), pp. 85–107.
Einstein, A. (1907). Über das Relativitätsprinzip und die aus demselben gezogenen Fol-
gerungen (On the relativity principle and the conclusions drawn from it), Jahrbuch der
Radioaktivität und Elektronik, 4, 411–462.
(1911). Über den Einfluss der Schwerkraft auf die Ausbreitung des Lichtes (On the
influence of gravitation on the propagation of light), Annalen der Physik, 35, 898–908.
(1912). Relativität und Gravitation. Erwiderung auf eine Bemerkung von M. Abraham
(Relativity and gravitation. Reply to a comment by M. Abraham), Annalen der Physik,
38, 1059–1064.
(1915). Die Feldgleichung der Gravitation (The field equations of gravitation), Sitzungs-
berichte, Königlich Preussische Akademie der Wissenschaften (Berlin), II (1915), 844–
847.
(1916a). Die Grundlage der Allgemeinen Relativitätstheorie (The Foundation of the Gen-
eral Theory of Relativity). (Leipzig: J. A. Barth). Acknowledgement to the contribution
of Marcel Grossmann is printed on p. 6.
(1916b). Die Grundlage der Allgemeinen Relativitätstheorie (The foundation of the gen-
eral theory of relativity), Annalen der Physik, 49, 769–822.
(1917). Kosmologische Betrachtungen zur Allgemeinen Relativitätstheorie (Cosmolog-
ical considerations in the general theory of relativity), Sitzungsberichte, Königlich
Preussische Akademie der Wissenschaften (Berlin), I (1917), 142–152.
References 469

Einstein, A. (1919). Spielen Gravitationsfelder im Aufbau der materiellen Elemen-

tarteilchen eine wesentliche Rolle? (Do gravitational fields play a significant role
for the structure of elementary particles?), Sitzungsberichte, Königlich Preussische
Akademie der Wissenschaften, I (1919), 349–356.
(1922a). Bemerkung zu der Arbeit von A. Friedmann ‘Ueber die Kruemmung des
Raumes’ (Remark on the work of A. Friedmann ‘On the curvature of space’), Zeitschrift
für Physik, 11, 326.
(1922b). Kyoto address of December 1922, in Einstein Köen-Roku, ed. Ishiwara, J.
(Tokyo: Tokyo-Tosho, 1977).
(1923). Notiz zu der Arbeit von A. Friedmann ‘Über die Krümmung des Raumes’ (A
note on the work of A. Friedmann ‘On the curvature of space’), Zeitschrift für Physik,
16, 228.
Einstein, A. and de Sitter, W. (1932). On the relation between the expansion and the mean
density of the Universe, Proceedings of the National Academy of Sciences, 18, 213–214.
Einstein, A. and Grossmann, M. (1913a). Entwurf einer verallgemeinerten Rela-
tivitätstheorie und einer Theorie der Gravitation (Outline of a Generalised Theory
of Relativity and of a Theory of Gravitation) (Leipzig: Teubner).
(1913b). Entwurf einer verallgemeinerten Relativitätstheorie und einer Theorie der Grav-
itation (Outline of a generalised theory of relativity and of a theory of gravitation),
Zeitschrift für Mathematik und Physik, 62, 225–259.
Elder, F. R., Gurewitsch, A. M., Langmuir, R. V. and Pollock, H. C. (1947). Radiation from
electrons in a synchrotron, Physical Review, 71, 829–830.
Ellis, R. (1987). Galaxy surveys at high redshift – past, present and future, in High Redshift
and Primaeval Galaxies, eds Bergeron, J., Kunth, D., Rocca-Volmerange, B. and Tran
Thanh Van, J. (Gif sur Yvette: Edition Frontières), pp. 3–16.
Ellis, R. G. (1997). Faint blue galaxies, Annual Review of Astronomy and Astrophysics, 35,
389–443.
Elsasser, W. M. (1933). Sur le principe de Pauli dans les noyaux (On Pauli’s principle for
nuclei), Journal de Physique et le Radium, Series 7, 4, 549–556.
Emden, R. (1907). Gaskugeln (Leipzig and Berlin: B. G. Teubner).
Epstein, I. (1950). A note on energy generation, Astrophysical Journal, 112, 207–210.
Ewen, H. I. and Purcell, E. M. (1951). Radiation from Galactic hydrogen at 1420 MHz,
Nature, 168, 356.
Faber, S. M. and Jackson, R. E. (1976). Velocity dispersions and mass-to-light ratios for
elliptical galaxies, Astrophysical Journal, 204, 668–683.
Fabian, A. C. (1994). Cooling flows in clusters of galaxies, Annual Review of Astronomy
and Astrophysics, 32, 277–318.
Fabricant, D. G., Lecar, M. and Gorenstein, P. (1980). X-ray measurements of the mass of
M87, Astrophysical Journal, 241, 552–560.
Falk, S. W. and Arnett, W. D. (1973). A theoretical model for Type II supernovae, Astro-
physical Journal, 180, L65–L68.
Fall, S. M. (1997). A global perspective on star formation, in The Hubble Space Telescope
and the High Redshift Universe, eds Tanvir, N. R., Aragón-Salamanca, A. and Wall,
J. V. (Singapore: World Scientific Publishing Company), pp. 303–308.
470 References

Fanaroff, B. L. and Riley, J. M. (1974). The morphology of extragalactic radio sources

of high and low luminosity, Monthly Notices of the Royal Astronomical Society, 167,
31P–36P.
Faulkner, J. (2001). Low-mass red giants as binary stars without angular momentum, in
Evolution of Binary and Multiple Star Systems; A Meeting in Celebration of Peter
Eggleton’s 60th Birthday, ASP Conference Series vol. 229, eds Podsiadlowski, P.,
Rappaport, S., King, A. R., D’Antona, F. and Burder, L. (San Francisco: ASP),
pp. 3–14.
Feast, M. W. and Catchpole, R. M. (1997). The Cepheid period-luminosity zero-point from
Hipparcos trigonometrical parallaxes, Monthly Notices of the Royal Astronomical Soci-
ety, 286, L1–L5.
Felten, J. (1977). Study of the luminosity function for field galaxies, Astronomical Journal,
82, 861–878.
Fermi, E. (1926). Sulla quantizzazione del gas perfecto monoatomica (On the quantisation
of a perfect monatomic gas), Rendiconti della Accademia Nazionale dei Lincei, 3,
145–149.
(1934a). Tentativo di una teoria dei raggi β (An attempt at a theory of β-rays), Nuovo
Cimento, 11, 1–19.
(1934b). Versuch einer Theorie der β(-Strahlen) (An attempt at a theory of β-rays),
Zeitschrift für Physik, 88, 161–177.
(1949). On the origin of the cosmic radiation, Physical Review, 75, 1169–1174.
Fich, M. and Tremaine, S. (1991). The mass of the Galaxy, Annual Review of Astronomy
and Astrophysics, 29, 409–445.
Fichtel, C. E., Simpson, G. A. and Thompson, D. J. (1978). Diffuse gamma radiation,
Astrophysical Journal, 222, 833–849.
Field, G. B. (1965). Thermal instability, Astrophysical Journal, 142, 531–567.
(1974). Interstellar abundances: gas and dust, Astrophysical Journal, 187, 453–469.
Field, G. B., Goldsmith, D. W. and Habing, H. J. (1969). Cosmic-ray heating of the interstellar
gas, Astrophysical Journal Letters, 55, L149–L154.
Fitch, W. S., Pacholczyk, A. G. and Weymann, R. J. (1967). Light variations of the Seyfert
galaxy NGC 4151, Astrophysical Journal, 150, L67–L70.
Fixsen, D. J., Cheng, E. S., Gales, J. M., Mather, J. C., Shafer, R. A. and Wright, E. L.
(1996). The cosmic microwave background spectrum from the full COBE FIRAS data
set, Astrophysical Journal, 473, 576–587.
Fizeau, H. and Foucault, L. (1847). Recherches sur les interférences des rayons calorifiques
(Researches on the interference of heat rays), Comptes Rendus de l’Academie des
Sciences, 25, 447–450.
Ford, H. C., Harms, R. J., Tsvetanov, Z. I. et al. (1994). Narrowband HST images of M87:
evidence for a disk of ionized gas around a massive black hole, Astrophysical Journal
Letters, 435, L27–L30.
Forman, W., Jones, C., Cominsky, L. et al. (1978). The fourth UHURU catalog of X-ray
sources, Astrophysical Journal Supplement Series, 38, 357–412.
Foucault, L. (1849). Lumière électrique (Electric light), L’Institut, Journal Universal des
Sciences, 17, 44–46.
References 471

Fowler, R. H. (1926). On dense matter, Monthly Notices of the Royal Astronomical Society,
87, 114–122.
Fowler, R. H. and Milne, E. A. (1923). The intensities of absorption lines in stellar spectra,
and the temperatures and pressures in the reversing layers of stars, Monthly Notices of
the Royal Astronomical Society, 83, 403–424.
(1924). The maxima of absorption lines in stellar spectra (second paper), Monthly Notices
of the Royal Astronomical Society, 84, 499–515.
Fowler, W. A. (1958). Completion of the proton–proton reaction chain and the possibility
of energetic neutrino emission by hot stars, Astrophysical Journal, 127, 551–556.
Fraunhofer, J. (1817a). Bestimmung des Brechungs- und Farbenzerstreuungs-Vermögens
Verschiedener Glasarten, in Bezug auf die Vervollkommnung Achromatischer
Fernröhre (On the refractive and dispersive power of different species of glass in
reference to the improvement of achromatic telescopes, with an account of the lines or
streaks which cross the spectrum), Denkschriften der königlichen Akademie der Wis-
senschaften zu München, 5, 193–226. Translation: Edinburgh Philosophical Journal,
9, 288–299 (1823); 10, 26-40 (1824).
(1817b). Bestimmung des Brechungs- und Farbenzerstreuungs-Vermögens Ver-
schiedener Glasarten, in Bezug auf die Vervollkommnung Achromatischer Fernröhre
(On the refractive and dispersive power of different species of glass in reference to the
improvement of achromatic telescopes, with an account of the lines or streaks which
cross the spectrum), Gilberts Annalen der Physik, 56, 264–313.
(1821). Neue Modifikation des Lichtes durch Gegenseitige Einwirkung und Beugung
der Strahlen, und Gesetze Derselben (New modifications of light through interactions
and diffraction of rays and its laws), Denkschriften der königlichen Akademie der
Wissenschaften zu München, 8, 1–76.
(1823). Kurzer Bericht von den Resultaten neuerer Versuche über die Gesetze des Lichtes,
und die Theorie Derselben (A short account of the results of recent experiments upon
the laws of light and its theory), Gilberts Annalen der Physik, 74, 337–378. Translation:
Edinburgh Journal of Science, 7, 101–113, 251–262 (1827); 8, 7–10 (1828).
Frazier, E. N. (1968). A spatio-temporal analysis of velocity fields in the solar photosphere,
Zeitschrift für Astrophysik, 68, 345–356.
Freedman, W. L., Madore, B. F., Gibson, B. K. et al. (2001). Final results from the Hubble
Space Telescope Key Project to measure the Hubble constant, Astrophysical Journal,
533, 47–72.
Freeman, K. C. (1970). On the disks of spiral and S0 galaxies, Astrophysical Journal, 160,
811–830.
Freier, P., Lofgren, E. J., Ney, E. P., Oppenheimer, F., Bradt, H. L. and Peters, B. (1948).
Evidence for heavy nuclei in the primary cosmic radiation, Physical Review, 74, 213–
217.
Frenk, C. (1986). Galaxy clustering and the dark-matter problem, Philosophical Transac-
tions of the Royal Astronomical Society, A320, 517–541.
Fried, D. L. (1965). Statistics of a geometric representation of wavefront distortion, Journal
of the Optical Society of America, 55, 1427–1435.
Friedman, A. A. (1922). On the curvature of space, Zeitschrift für Physik, 10, 377–386.
472 References

Friedman, A. A. (1923). The World as Space and Time (Petrograd: Academia).

(1924). On the possibility of a world with constant negative curvature, Zeitschrift für
Physik, 12, 326–332.
Friedman, H., Lichtman, S. W. and Byram, E. T. (1951). Photon counter measurements of
solar X-rays and extreme ultraviolet light, Physical Review, 83, 1025–1030.
Friedrich, W., Knipping, P. and Laue, M. von (1912). Interferenz-Erscheinungen bei
Röntgenstrahlen (Interference effects with Röntgen rays), Sitzberichte der Königlich
Bayerischen Akademie der Wissenschaften, pp. 303–322.
Fritz, G., Henry, R. C., Meekins, J. F., Chubb, T. A. and Friedman, H. (1969). X-ray pulsar
in the Crab Nebula, Science, 164, 709–712.
Frost, E. B. (1909). Spectroscopic notes, Astrophysical Journal, 29, 233–239.
Fukuda, S., Fukuda, Y., Ishitsuka, et al. (2001). Solar 8 B and hep neutrino measurements
from 1258 days of super-Kamiokande data, Physical Review Letters, 86, 5651–5655.
Fukuda, Y., Hayakawa, T., Inoue, K. et al. (1996). Solar neutrino data covering solar cycle
22, Physical Review Letters, 77, 1683–1686.
Galbraith, W. and Jelley, J. V. (1953). Light pulses from the night sky associated with cosmic
rays, Nature, 171, 349–350.
(1955). Light-pulses from the night sky and Cherenkov radiation, Part 1, Journal of
Atmospheric and Terrestrial Physics, 6, 250–262.
Gallego, J., Zamorano, J., Aragón-Salamanca, A. and Rego, M. (1995). The current star
formation rate of the local Universe, Astrophysical Journal, 455, L1–L4.
Gamow, G. (1928). Zur Quantentheorie der Atomzertrümmerung (On the quantum theory
of atomic destruction), Zeitschrift für Physik, 52, 510–515.
(1937). Atomic Nuclei and Nuclear Transformations (Oxford: Oxford University Press).
(1939). Physical possibilities of stellar evolution, Physical Review, 55, 718–725.
(1946). Expanding Universe and the origin of elements, Physical Review, 70, 572–573.
(1970). My World Line, (New York: Viking Press). The reference to Einstein’s admission
of ‘the greatest blunder of my life’ is on p. 44.
Garcia-Munoz, M., Mason, G. M. and Simpson, J. A. (1977). The age of the Galactic cosmic
rays derived from the abundances of 10 Be, Astrophysical Journal, 217, 857–877.
Garnavich, P. M., Kirshner, R. P., Challis, P. et al. (1998). Constraints on cosmological
models from Hubble Space Telescope observations of high-z supernovae, Astrophysical
Journal Letters, 493, L53–L58.
Geiger, H. and Müller, W. (1928). Das Electronenzählrohr (The electron-counting tube),
Physicalische Zeitschrift, 29, 839–841.
(1929). Technische Bemerkungen zum Electronenzählrohr (Technical remarks on the
electron-counting tube), Physicalische Zeitschrift, 30, 489–493.
Geller, M. J. and Huchra, J. P. (1989). Mapping the Universe, Science, 246, 897–903.
Genzel, R., Schödel, R., Ott, T. et al. (2003). Near-infrared flares from accreting gas around
the supermassive black hole at the Galactic centre, Nature, 425, 934–937.
Gershtein, S. S. and Zeldovich, Y. B. (1966). Rest mass of a muonic neutrino and cosmology,
Pisma v Zhurnal Eksperimentalnoi i Teoreticheskoi Fiziki, 4, 174–177.
Ghez, A. M., Morris, M., Becklin, E. E., Tanner, A. and Krememek, T. (2000). The accel-
erations of stars orbiting the Milky Way’s central black hole, Nature, 407, 349–351.
References 473

Giacconi, R., Gursky, H., Paolini, F. R. and Rossi, B. B. (1962). Evidence for X rays from
sources outside the solar system, Physical Review Letters, 9, 439–443.
Giacconi, R., Gursky, H., Kellogg, E., Schreier, E. and Tananbaum, H. (1971a). Discovery
of periodic X-ray pulsations in Centaurus X-3 from UHURU, Astrophysical Journal,
167, L67–L73.
Giacconi, R., Kellogg, E., Gorenstein, P., Gursky, H. and Tananbaum, H. (1971b). An X-ray
scan of the Galactic plane from UHURU, Astrophysical Journal, 165, L27–L35.
Giacconi, R., Bechtold, J., Branduardi, G. et al. (1979). A high-sensitivity X-ray survey
using the Einstein Observatory and the discrete source contribution to the extragalactic
X-ray background, Astrophysical Journal Letters, 234, L1–L7.
Giavalisco, M., Livio, M., Bohlin, R. C., Macchetto, F. D. and Stecher, T. P. (1996). On the
morphology of the HST faint galaxies, Astronomical Journal, 112, 369–377.
Gibbons, G. W. and Hawking, S. W. (1977). Cosmological event horizons, thermodynamics,
and particle creation, Physical Review, D15, 2738–2751.
Gibbons, G. W., Hawking, S. W. and Siklos, S. T. C., eds (1983). The Very Early Uni-
verse: Proceedings of the Nuffield Workshop, Cambridge, UK, June 21–July 9, 1982
(Cambridge: Cambridge University Press).
Gibbons, G. W., Shellard, E. P. S. and Rankin, S. J., eds (2003). The Future of Theoretical
Physics and Cosmology (Cambridge: Cambridge University Press).
Ginzburg, V. L. (1951). Cosmic rays as a source of Galactic radio-radiation, Doklady
Akademiya Nauk SSSR, 76, 377–380.
Ginzburg, V. L. and Kirzhnits, D. A. (1964). On the superfluidity of neutron stars, Zhur-
nal Experimentalnoi i Teoretichseskikh Fizica, 47, 2006–2007. Translation in Soviet
Physics JETP, 20, 1965, 1346–1348.
Glazebrook, K., Ellis, R. S., Colless, M., Broadhurst, T. J., Allington-Smith, J. R. and Tanvir,
N. R. (1995). The morphological identification of the rapidly evolving population of
faint galaxies, Monthly Notices of the Royal Astronomical Society, 275, L19–L22.
Gold, T. (1968). Rotating neutron stars as the origin of pulsating radio sources, Nature, 218,
731–732.
(1969). Rotating neutron stars and the nature of pulsars, Nature, 221, 25–27.
Goldhaber, G., Boyle, B., Bunclark, P. et al. (1996). Cosmological time dilation using Type
Ia supernovae as clocks, Nuclear Physics B Proceedings Supplements, 51, 123–127.
Goldreich, P. and Julian, W. H. (1969). Pulsar electrodynamics, Astrophysical Journal, 157,
869–880.
Goobar, A. and Perlmutter, S. (1995). Feasibility of measuring the cosmological constant
lambda and mass density omega using Type IA supernovae, Astrophysical Journal,
450, 14–18.
Gott, J. R., Melott, A. L. and Dickinson, M. (1986). The sponge-like topology of large-scale
structure in the Universe, Astrophysical Journal, 306, 341–357.
Gough, D. O. (1977). Random remarks on solar hydrodynamics, in The Energy Balance
and Hydrodynamics of the Solar Chromosphere and Corona, eds Bonnet, R. M. and
Delache, P. (Clermont-Ferrand: G. de Bussac), pp. 3–36.
Gower, J. F. R. (1966). The source counts from the 4C survey, Memoirs of the Royal
Astronomical Society, 133, 151–161.
474 References

Gower, J. F. R., Scott, P. F. and Wills, D. (1967). A survey of radio sources in the declination
ranges -07 to 20 and 40 to 80, Monthly Notices of the Royal Astronomical Society, 71,
49–144.
Graham Smith, F. (1951). An accurate determination of the positions of four radio stars,
Nature, 168, 555.
Greenstein, J. L. and Matthews, T. A. (1963). Red-shift of the unusual radio source 3C 48,
Nature, 197, 1041–1042.
Greenstein, J. L. and Schmidt, M. (1964). Red-shifts of the radio sources 3C 48 and 3C
273, Astrophysical Journal, 140, 1–43.
Gregory, J. (1668). Geometriae Pars Universalis (Padua: Published by the heirs of Paolo
Frambotti), p. 148.
Gribben, J. and Rees, M. J. (1989). Dark Matter, Mankind and Anthropic Cosmology (New
York: Bantam Books).
Gull, S. F. (1975). The X-ray, optical and radio properties of young supernova remnants,
Monthly Notices of the Royal Astronomical Society, 171, 263–278.
Gunn, J. E. (1978). The Friedmann models and optical observations in cosmology, in Obser-
vational Cosmology: 8th Advanced Course, Swiss Society of Astronomy and Astro-
physics, Saas-Fee 1978, eds Maeder, A., Martinet, L. and Tammann, G. (Geneva:
Geneva Observatory Publications), pp. 1–121.
Gunn, J. E. and Peterson, B. A. (1965). On the density of neutral hydrogen in intergalactic
space, Astrophysical Journal, 142, 1633–1636.
Gursky, H., Giacconi, R., Paolini, F. R. and Rossi, B. B. (1963). Further evidence for the
existence of galactic X-rays, Physical Review Letters, 11, 530–535.
Gursky, H., Kellogg, E. M., Murray, S., Leong, C., Tananbaum, H. and Giacconi, R. (1971).
A strong X-ray source in the Coma cluster observed by UHURU, Astrophysical Journal
Letters, 167, L81–L84.
Guth, A. (1981). Inflationary Universe: a possible solution to the horizon and flatness
problems, Physical Review, D23, 347–356.
(1997). The Inflationary Universe: The Quest for a New Theory of Cosmic Origins (Read-
ing, Massachusetts: Addison-Wesley).
(2003). Inflation and cosmological perturbations, in The Future of Theoretical Physics
and Cosmology, eds Gibbons, G. W., Shellard, E. P. S. and Rankin, S. J. (Cambridge:
Cambridge University Press), pp. 725–754.
Hale, G. E. (1928). The possibilities of large telescopes, Harper’s Magazine, 156, 639–646.
Hall, J. S. (1949). Observations of the polarized light from stars, Science, 109, 166–167.
Halley, E. (1718). Considerations on the change of the latitudes of some of the principal
fixt stars, Philosophical Transactions of the Royal Society, 30, 736–738.
Halm, J. (1911). Further considerations relating to the systematic motions of the stars,
Monthly Notices of the Royal Astronomical Society, 71, 610–639.
Halverson, N. W., Leitch, E. M., Pryke, C. et al. (2002). Degree angular scale interferom-
eter first results: a measurement of the cosmic microwave background angular power
spectrum, Astrophysical Journal, 568, 38–45.
Hamilton, A. J. S., Kumar, P., Lu, E. and Matthews, A. (1991). Reconstructing the primordial
spectrum of fluctuations of the Universe from the observed nonlinear clustering of
galaxies, Astrophysical Journal, 374, L1–L4.
References 475

Hampel, W., Handt, J., Heusser, G. et al. (1999). GALLEX solar neutrino observations:
results for GALLEX IV, Physics Letters B, 447, 127–133.
Hanany, S., Ade, P., Balbi, A. et al. (2000). MAXIMA-1: a measurement of the cosmic
microwave background anisotropy on angular scales of 10 −5◦ , Astrophysical Journal,
545, L5–L9.
Hardcastle, M. J., Alexander, P., Pooley, G. G. and Riley, J. M. (1996). The Jets in 3C 66B,
Monthly Notices of the Royal Astronomical Society, 278, 273–284.
Harms, R. J., Ford, H. C., Tsvetanov, Z. I. et al. (1994). HST FOS spectroscopy of M87:
evidence for a disk of ionized gas around a massive black hole, Astrophysical Journal
Letters, 435, L35–L38.
Harrison, B. K., Wakano, M. and Wheeler, J. A. (1958). Matter-energy at high den-
sity; end-point of thermonuclear evolution, in Onzième Conseil de Physique Solvay,
La Structure et l’evolution de l’univers (Brussels: Editions Stoops), pp. 124–
146.
Harrison, B. K., Thorne, K. S., Wakano, M. and Wheeler, J. A. (1965). Gravitational Theory
and Gravitational Collapse (Chicago: University of Chicago Press).
Harrison, E. R. (1970). Fluctuations at the threshold of classical cosmology, Physical
Review, D1, 2726–2730.
Hartmann, J. F. (1904). Investigations of the spectrum and orbit of Delta Orionis, Astro-
physical Journal, 19, 268–286.
Hasinger, G., Burg, R., Giacconi, R. et al. (1993). A deep X-ray survey in the Lockman
Hole and the soft X-ray log N –log S, Astronomy and Astrophysics, 275, 1–15.
Hausman, M. A. and Ostriker, J. P. (1977). Cannibalism among galaxies – dynamically
produced evolution of cluster luminosity functions, Astrophysical Journal, 217, L125–
L129.
Hawking, S. W. (1972). Black holes in general relativity, Communications in Mathematical
Physics, 25, 152–166.
Hawking, S. W. and Ellis, G. R. (1973). The Large Scale Structure of Space-Time (Cam-
bridge: Cambridge University Press).
Hawking, S. W. and Penrose, R. (1969). The singularities of gravitational collapse and
cosmology, Proceedings of the Royal Society, A314, 529–548.
Hayakawa, S. and Matsuoka, M. (1964). Part V. Origin of cosmic X-rays, Supplement of
Progress of Theoretical Physics (Japan), 30, 204–228.
Hayashi, C. (1950). Proton-neutron concentration ratio in the expanding Universe at the
stages preceding the formation of the elements, Progress of Theoretical Physics
(Japan), 5, 224–235.
(1961). Stellar evolution in early phases of gravitational contraction, Publications of the
Astronomical Society of Japan, 13, 450–452.
Hazard, C., Mackey, M. B. and Shimmins, A. J. (1963). Investigation of the radio source
3C 273 by the method of lunar occultations, Nature, 197, 1037–1039.
Helmholtz, H. von (1854). On the interaction of natural forces, Philosophical Magazine
(Series 4), 11, 489–518. Lecture delivered at Königsberg, 7 February 1854.
Helou, G., Soifer, B. T. and Rowan-Robinson, M. (1985). Thermal infrared and nonthermal
radio – remarkable correlation in disks of galaxies, Astrophysical Journal, 298, L7–
L11.
476 References

Henderson, T. (1840). On the parallax of α Centauri, Memoirs of the Royal Astronomical

Society, 11, 61–68.
Henyey, L. G. and Keenan, P. C. (1940). Interstellar radiation from free electrons and
hydrogen atoms, Astrophysical Journal, 91, 625–630.
Herbig, G. H. (1952). Emission-line stars in Galactic nebulosities, Journal of the Royal
Astronomical Society of Canada, 46, 222–233.
Hernquist, L., Katz, N., Weinberg, D. H. and Miralda-Escudé, J. (1996). The Lyman-alpha
forest in the cold dark matter model, Astrophysical Journal, 457, L51–L55.
Herschel, J. F. W. (1840). On the chemical action of the rays of the solar spectrum on
preparations of silver and other substances, both metallic and non-metallic, and on some
photographic processes, Philosophical Transactions of the Royal Society of London,
130, 1–59.
(1864). General catalogue of nebulae and clusters of stars, Philosophical Transactions
of the Royal Society, 154, 1–137.
Herschel, W. (1783). On the proper motion of the Sun and Solar System; with an account
of several changes that have happened among the fixed stars since the time of Mr.
Flamstead, Philosophical Transactions of the Royal Society, 73, 247–283.
(1785). On the construction of the heavens, Philosophical Transactions of the Royal
Society, 75, 213–268.
(1800a). Experiments on the refrangibility of the invisible rays of the Sun, Philosophical
Transactions of the Royal Society, 90, 284–292.
(1800b). Experiments on the solar, and on the terrestrial rays that occasion heat; with a
comparative view of the laws to which light and heat, or rather the rays which occasion
them, are subject, in order to determine whether they are the same, or different. Part I,
Philosophical Transactions of the Royal Society, 90, 293–326.
(1800c). Experiments on the solar, and on the terrestrial rays that occasion heat;
with a comparative view of the laws to which light and heat, or rather the rays
which occasion them, are subject, in order to determine whether they are the same,
or different. Part II, Philosophical Transactions of the Royal Society, 90, 437–
538.
(1800d). Investigation of the powers of the prismatic colours to heat and illuminate
objects; with remarks that prove the different refrangibility of radiant heat. To which is
added, an inquiry into the method of viewing the Sun advantageously, with telescopes
of large apertures and high magnifying powers, Philosophical Transactions of the
Royal Society, 90, 255–283.
(1802). Catalogue of 500 new nebulae, nebulous stars, planetary nebulae, and clusters of
stars; with remarks on the construction of the heavens, Philosophical Transactions of
the Royal Society, 92, 477–528.
Hertzsprung, E. (1905). Zur Strahlung der Sterne I (On the radiation of stars I), Zeitschrift
für Wissenschaftliche Photographie, 3, 429–442.
(1906). Ueber die Optische Stärke der Strahlung des Schwartzen Körpers und das Mini-
male Lichtäquivalent (On the optical intensity of black-body radiation and the equiv-
alent minimum light emission), Zeitschrift für Wissenschaftliche Photographie, 4,
43–54.
References 477

Hertzsprung, E. (1907). Zur Strahlung der Sterne II (On the radiation of stars II), Zeitschrift
für Wissenschaftliche Photographie, 5, 86–107.
(1911). Über die Verwendung Photographischer Effectiver Wellenlängen zur Bestim-
mmung von Farbenäquivalenten (On the use of photographic effective wavelengths
for the determination of equivalent colours), Publikationen des Astrophysikalischen
Observatoriums zu Potsdam, 22, 1–40.
(1913). Über die Räumliche Verteilung der Veränderlichen vom δ Cephei-Typus (On
the spatial distribution of the variables of δ-Cephei type), Astronomische Nachrichten,
196, 201–209.
(1919). Bermerkungen zur Statistik der Sternparallaxen (Remarks on the statisics of
stellar parallaxes), Astronomische Nachrichten, 208, 89–96.
Hess, V. F. (1912). Über Beobachtungen der durchdringenden Strahlung bei sieben Freibal-
lonfahrten (Concerning observations of penetrating radiation on seven free balloon
flights), Physikalische Zeitschrift, 13, 1084–1091.
Hesser, J. E., Harris, W. E., VandenBerg, D. A., Allwright, J. W. B., Shott, P. and Stetson, P.
(1987). A CCD color-magnitude study of 47 Tucanae, Publications of the Astronomical
Society of the Pacific, 99, 739–808.
Hewish, A., Bell, S. J., Pilkington, J. D. H., Scott, P. F. and Collins, R. A. (1968). Observations
of a rapidly pulsating radio source, Nature, 217, 709–713.
Hey, J. S. (1946). Solar radiations in the 4–6 metre radio wave-length band, Nature, 157,
47–48.
Hey, J. S., Parsons, S. J. and Phillips, J. W. (1946). Fluctuations in cosmic radiation at
radio-frequencies, Nature, 158, 234.
Higgs, P. W. (1964). Broken symmetries, massless particles and gauge fields, Physics Letters,
12, 132–133.
Hiltner, W. A. (1949). Polarization of light from distant stars by the interstellar medium,
Science, 109, 165.
Hirata, K. S., Inoue, K., Kajita, T., Kifune, T. and Kihara, K. (1990). Results from one
thousand days of real-time, directional solar-neutrino data, Physical Review Letters,
65, 1297–1300.
Hjellming, R. and Wade, C. (1971). Further radio observations of Scorpius X-1, Astrophys-
ical Journal, 170, 523–528.
Hjorth, J., Sollerman, J., Møller, P. et al. (2003). A very energetic supernova associated with
the γ -ray burst of 29 March 2003, Nature, 423, 847–850.
Hoag, A. A. and Smith, M. G. (1977). Faint emission-line quasi-stellar object candidates,
Astrophysical Journal, 217, 362–381.
Hoyle, F. (1946). The chemical composition of the stars, Monthly Notices of the Royal
Astronomical Society, 106, 225–259.
(1948). A new model for the expanding Universe, Monthly Notices of the Royal Astro-
nomical Society, 108, 372–382.
(1954). On nuclear reactions occurring in very hot stars. I. The synthesis of elements
from carbon to nickel, Astrophysical Journal Supplement, 1, 121–146.
Hoyle, F. and Fowler, W. A. (1963a). On the nature of strong radio sources, Monthly Notices
of the Royal Astronomical Society, 125, 169–176.
478 References

Hoyle, F. and Fowler W. A. (1963b). Nature of strong radio sources, Nature, 197, 533–535.
Hoyle, F. and Lyttleton, R. A. (1942). On the internal constitution of the stars, Monthly
Notices of the Royal Astronomical Society, 102, 177–193.
Hoyle, F. and Schwarzschild, M. (1955). On the evolution of type II stars, Astrophysical
Journal Supplement, 2, 1–40.
Hoyle, F. and Tayler, R. J. (1964). The mystery of the cosmic helium abundance, Nature,
203, 1108–1110.
Hoyle, F., Burbidge, G. R. and Sargent, W. L. W. (1966). On the nature of the quasi-stellar
sources, Nature, 209, 751–753.
Hu, W. (1996). Concepts in CMB anisotropy formation, in The Universe at High-z, Large-
Scale Structure and the Cosmic Microwave Background, eds Martinez-Gonzales, E.
and Sanz, J. L. (Berlin: Springer-Verlag), pp. 207–240.
Hu, W. and Sugiyama, N. (1995). Anisotropies in the cosmic microwave background: an
analytic approach, Astrophysical Journal, 444, 489–506.
Hu, W., Sugiyama, N. and Silk, J. (1997). The physics of microwave background
anisotropies, Nature, 386, 37–43.
Hubble, E. P. (1925). Cepheids in spiral nebulae, Publications of the American Astronomical
Society, 5, 261–264.
(1926). Extra-galactic nebulae, Astrophysical Journal, 64, 321–369.
(1929). A relation between distance and radial velocity among extra-galactic nebulae,
Proceedings of the National Academy of Sciences, 15, 168–173.
(1935). Angular rotations of spiral nebulae, Astrophysical Journal, 81, 334–335.
(1936). The Realm of the Nebulae (New Haven: Yale University Press).
Hubble, E. P. and Humason, M. (1934). The velocity–distance relation among extra-galactic
nebulae, Astrophysical Journal, 74, 43–80.
Huchra, J. and Brodie, J. (1987). The M87 Globular Cluster System. I – Dynamics, Astro-
nomical Journal, 93, 779–784.
Hudson, M. J., Dekel, A., Courteau, S., Faber, S. M. and Willick, J. A. (1995).  and
biasing from optical galaxies versus POTENT mass, Monthly Notices of the Royal
Astronomical Society, 274, 305–316.
Huggins, W. (1868). Further observations on the spectra of some of the stars and nebulae,
with an attempt to determine therefrom whether these bodies are moving towards
or from the Earth, also observations on the spectra of the Sun and of the Comet II,
Philosophical Transactions of the Royal Society of London, 158, 529–564.
(1869). Note on the heat of the stars, Proceedings of the Royal Society of London, 17,
309–312.
Huggins, W. and Miller, W. A. (1864a). On the spectra of some of the fixed stars, Philo-
sophical Transactions of the Royal Society of London, 154, 413–435.
(1864b). On the spectra of some of the nebulae; a supplement to the paper ‘On the spectra
of some fixed stars’, Philosophical Transactions of the Royal Society of London, 154,
437–444.
Hughes, D. H., Serjeant, S., Dunlop, J. et al. (1998). High-redshift star formation in the
Hubble Deep Field revealed by a submillimetre-wavelength survey, Nature, 394, 241–
247.
References 479

Hulse, R. A. and Taylor, J. H. (1975). Discovery of a pulsar in a binary system, Astrophysical

Journal Letters, 195, L51–L53.
Hulst, H. C. van de (1945). Radio waves from space: origin of radiowaves, Nederlands
Tijdschrift voor Natuurkunde, 11, 210–221.
(1949a). Interstellar polarization and magneto-hydrodynamic waves, in Problems of
Cosmical Aerodynamics: Proceedings of IUTAM–IAU Symposium on Cosmical Gas
Dynamics, eds Burgers, J. M. and van de Hulst, H. C. (Dayton, Ohio: Central Air
Documents Office), pp. 45–58.
(1949b). The solid particles of interstellar space, Recherches Astronomiques de
l’Observatoire d’Utrecht, no. 11, Part 2, 1–50.
Humason, M. L., Mayall, N. U. and Sandage, A. R. (1956). Redshifts and magnitudes of
extra-galactic nebulae, Astronomical Journal, 61, 97–162.
Hummer, D. G. and Mihalas, D. (1988). The equation of state for stellar envelopes. I –
An occupation probability formalism for the truncation of internal partition functions,
Astrophysical Journal, 331, 794–814.
Illingworth, G. (1977). Rotation (?) in 13 elliptical galaxies, Astrophysical Journal Letters,
218, L43–L47.
Inskip, K. J., Best, P. N., Longair, M. S. and MacKay, D. J. C. (2002). Infrared magnitude-
redshift relations for luminous radio galaxies, Monthly Notices of the Royal Astronom-
ical Society, 329, 277–289.
Irwin, M., McMahon, R. G. and Hazard, C. (1991). APM optical surveys for
high redshift quasars, Proceedings of the Workshop on The Space Distribution
of Quasars: Astronomical Society of the Pacific Conference Series, 21, 117–
126.
Ivanenko, D. and Pomeranchuk, I. (1944). On the maximal energy attainable in a betatron,
Physical Review, 65, 343.
Jansky, K. G. (1933). Electrical disturbances apparently of extraterrestrial origin, Proceed-
ings of the Institution of Radio Engineers, 21, 1387–1398.
Jeans, J. H. (1902). The stability of a spherical nebula, Philosophical Transactions of the
Royal Society of London, 199, 1–53.
(1917). Remark in discussion of the Royal Astronomical Society of 8 December 1916,
Observatory, 40, 43.
(1926). Diffuse matter in interstellar space: Letter to the Editor, Observatory, 49, 333–
335. See the final paragraph.
Jennison, R. C. and Das Gupta, M. K. (1953). Fine structure of the extra-terrestrial radio
source Cygnus 1, Nature, 172, 996–997.
Jöeveer, M. and Einasto, J. (1978). Has the Universe the cell structure?, in The Large Scale
Structure of the Universe, eds Longair, M. S. and Einasto, J. (Dordrecht: D. Reidel
Publishing Company), pp. 241–251.
Johnson, H. L. (1962). Infrared stellar photometry, Astrophysical Journal, 135, 69–77.
(1965). Interstellar extinction in the Galaxy, Astrophysical Journal, 141, 923–942.
Johnson, H. L. and Morgan, W. W. (1953). Fundamental stellar photometry for standards of
spectral type in the revised system of the Yerkes spectral atlas, Astrophysical Journal,
117, 313–352.
480 References

Johnson, H. L., Mitchell, R. I., Iriate, B. and Wisniewski, W. Z. (1966). UBVRIJKL pho-
tometry of the bright stars, Communications of the Lunar Planetary Laboratory, 4,
99–110.
Johnson, W. N. III and Haymes, R. C. (1973). Detection of a gamma-ray spectral line from
the Galactic-center region, Astrophysical Journal, 184, 103–126.
Jones, M., Saunders, R., Alexander, P. et al. (1998). An image of the Sunyaev–Zel’dovich
effect, Nature, 365, 320–323.
Joy, A. H. (1939). Rotational effects, interstellar absorption, and certain dynamical con-
stants of the Galaxy determined from Cepheid variables, Astrophysical Journal, 89,
356–376.
(1945). T Tauri variable stars, Astrophysical Journal, 102, 168–195.
Kaiser, C. R. and Alexander, P. (1997). A self-similar model for extragalactic radio sources,
Monthly Notices of the Royal Astronomical Society, 286, 215–222.
Kaiser, N. (1986). Evolution and clustering of rich clusters, Monthly Notices of the Royal
Astronomical Society, 222, 323–345.
Kapahi, V. K. (1987). The angular size-redshift relation as a cosmological tool, in Observa-
tional Cosmology, eds Hewitt, A., Burbidge, G. and Fang, L.-Z. (Dordrecht: D. Reidel
Publishing Co.), pp. 251–265.
Kapteyn, J. C. (1892). To what stellar system does our Sun belong?, Publications of the
Astronomical Society of the Pacific, 4, 259–260.
(1905). Star streaming, Report of the British Association for the Advancement of Science,
257, pp. 237–265.
(1906). Plan of Selected Areas (Groningen: Astronomical Laboratory).
(1922). First attempt at a theory of the arrangement and motion of the sidereal system,
Astrophysical Journal, 55, 302–328.
Kapteyn, J. C. and Rhijn, P. J. van (1920). On the distribution of the stars in space especially
in the high galactic latitudes, Astrophysical Journal, 52, 23–38.
Katz, N., Weinberg, D. H., Hernquist, L. and Miranda-Escudé, J. (1996). Damped Lyman-
alpha and Lyman-limit absorbers in the cold dark matter model, Astrophysical Journal,
457, L57–L60.
Kauffmann, G., Colberg, J. M., Diaferio, A. and White, S. D. M. (1999). Clustering of
galaxies in a hierarchical Universe: I. Methods and results at z = 0, Monthly Notices
of the Royal Astronomical Society, 303, 188–206.
Kaufmann, W. (1902). Die Elektromagnetische Masse des Elektrons (On the electromag-
netic mass of the electron), Physikalische Zeitschift, 4, 54–56.
Kellermann, K. I. (1993). The cosmological deceleration parameter estimated from
the angular-size/redshift relation for compact radio sources, Nature, 361, 134–
136.
Kennefick, J. D., Djorgovski, S. G. and de Carvalo, R. R. (1995). The luminosity function
of z > 4 quasars from the second Palomar Sky Survey, Astronomical Journal, 110,
2553–2565.
Kennicutt, R. C. (1989). The star formation law in galactic discs, Astrophysical Journal,
344, 685–703.
Kerr, R. P. (1963). Gravitational field of a spinning mass as an example of algebraically
special metrics, Physical Review Letters, 11, 237–238.
References 481

Khachikian, E. Y. and Weedman, D. W. (1971). A spectroscopic study of luminous galactic

nuclei, Astrofizika, 7, 389–406.
(1974). An atlas of Seyfert galaxies, Astrophysical Journal, 192, 581–589.
Kiang, T. and Saslaw, W. C. (1969). The distribution in space of clusters of galaxies, Monthly
Notices of the Royal Astronomical Society, 143, 129–138.
Kibble, T. W. B. (1976). Topology of cosmic domains and strings, Journal of Physics A:
Mathematical and General, 9, 1387–1398.
Kiepenheuer, K. O. (1950). Cosmic rays as the source of general Galactic radio emission,
Physical Review, 79, 738–739.
Kirchhoff, G. (1859). Ueber den Zusammenhang zwischen Emission und Absorption von
Licht und Wärme (On the connection between emission and absorption of light and
heat), Berlin Monatsberichte, pp. 783–787.
(1861). Untersuchungen über das Sonnenspektrum und die Spectren der Chemischen Ele-
mente (Investigations of the solar spectrum and the spectra of the chemical elements),
Part 1, Abhandlungen der königlich Preussischen Akademie der Wissenschaften zu
Berlin, pp. 63–95.
(1862). Untersuchungen über das Sonnenspektrum und die Spectren der Chemischen
Elemente (Investigations of the solar spectrum and the spectra of the chemical ele-
ments), Part 1 (continued), Abhandlungen der königlich Preussischen Akademie der
Wissenschaften zu Berlin, pp. 227–240.
(1863). Untersuchungen über das Sonnenspektrum und die Spectren der Chemischen Ele-
mente (Investigations of the solar spectrum and the spectra of the chemical elements),
Part 2, Abhandlungen der königlich Preussischen Akademie der Wissenschaften zu
Berlin, pp. 225–240.
Kirshner, R. and Kwan, J. (1974). Distances to extragalactic supernovae, Astrophysical
Journal, 193, 27–36.
Kirshner, R. P. and Oke, B. (1975). Supernova 1972e in NGC 5253, Astrophysical Journal,
200, 574–581.
Klebesadel, R. W., Strong, I. B. and Olson, R. A. (1973). Observations of gamma-ray bursts
of cosmic origin, Astrophysical Journal Letters, 182, L85–L88.
Kleinmann, D. E. and Low, F. J. (1967). Discovery of an infrared nebula in Orion, Astro-
physical Journal, 149, L1–L4.
Kniffen, D. A., Chipman, E. and Gehrels, N. (1994). The gamma-ray sky according to
Compton: a new window to the Universe, in Frontiers of Space and Ground-Based
Astronomy, eds Wamsteker, W., Longair, M. S. and Kondo, Y. (Dordrecht: Kluwer
Academic Publishers), pp. 5–16.
Knop, R. A., Aldering, G., Amanullah, R. et al. (2003). New constraints on M ,  , and
w from an independent set of 11 high-redshift supernovae observed with the Hubble
Space Telescope, Astrophysical Journal, 598, 102–137.
Kobold, H. A. (1895). Untersuchungen des Eigenbewegung des Auwers-Bradley Cat-
alogs nach Bessel’schen Methode (Investigation of the proper motions from the
Auwers–Bradley Catalogue using Bessel’s method), Abhandlungen der Kaiserlicher
Leopoldinisch-Carolinschen Deutschen Akademie der Naturforscher, 64, 213–365.
Kogut, A., Banday, A. J., Bennett, C. L. et al. (1996). Tests for non-Gaussian statistics in
the DMR four-year sky maps, Astrophysical Journal, 464, L29–L33.
482 References

Kolb, E. W. and Turner, M. S. (1990). The Early Universe (Redwood City, California:
Addison–Wesley Publishing Co.).
Kolhörster, W. (1913). Messungen der Durchdringenden Strahlung im Freiballon in
Grösseren Höhen (Measurements of penetrating radiation in free balloon flights at
great altitudes), Physikalische Zeitschrift, 14, 1153–1156.
Kompaneets, A. (1956). The establishment of thermal equilibrium between quanta and
electrons, Zhurnal Eksperimentalnoi i Teoreticheskoi Fiziki, 31, 876–885. Translation
in Soviet Physics, 4, 1957, 730–737.
Koo, D. C. and Kron, R. (1982). QSO counts – a complete survey of stellar objects to
B = 23, Astronomy and Astrophysics, 105, 107–119.
Kormendy, J. (1982). Observations of galaxy structure and dynamics, in Morphology and
Dynamics of Galaxies: Twelfth Advanced Course of the Swiss Society of Astronomy
and Astrophysics, eds Martinet, L. and Mayor, M. (Sauverny, Switzerland: Geneva
Observatory), pp. 113–288.
Kormendy, J. and Richstone, D. O. (1995). Inward bound – the search for supermassive black
holes in galactic nuclei, Annual Review of Astronomy and Astrophysics, 33, 581–624.
Kramers, H. A. (1923). On the theory of X-ray absorption and of the continuous X-ray
spectrum, Philosophical Magazine, 46, 836–871.
Kraushaar, W. L., Clark, G. W., Garmire, G. P., Borken, R., Higbie, P. and Agogino, M.
(1965). Explorer XI experiment on cosmic gamma rays, Astrophysical Journal, 141,
845–863.
Kroto, H. W., Heath, J. R., O’Brien, S. C., Curl, R. F. and Smalley, R. E. (1985). C(60):
buckminsterfullerene, Nature, 318, 162–163.
Kroto, H. W., Kirby, C., Walton, D. R. M. et al. (1978). The detection of cyanohexatriyne,
H(C≡C)3 CN, in Heiles’s Cloud 2, Astrophysical Journal, 219, L133–L137.
Kroto, H. W., Heath, J. R., O’Brien, S. C., Curl, R. F. and Smalley, R. E. (1987).
Long carbon chain molecules in circumstellar shells, Astrophysical Journal, 314,
352–355.
Kruskal, M. D. (1960). Maximal extension of Schwarzschild metric, Physical Review, 119,
1743–1745.
Krymsky, G. F. (1977). A regular mechanism for the acceleration of charged particles on
the front of a shock wave, Doklady Akademiya Nauk SSSR, 234, 1306–1308.
Krzeminski, W. (1973). International Astronomical Union Circular no. 2612.
(1974). The identification and UBV photometry of the visible component of the Centaurus
X-3 binary system, Astrophysical Journal Letters, 192, L135–L138.
Kuiper, G. P., Wilson, W. and Cashman, R. J. (1947). An infrared stellar spectrometer,
Astrophysical Journal, 106, 243–250.
Kundic, T., Turner, E. L., Colley, W. N. et al. (1997). A robust determination of the time
delay in 0957+561A, B and a measurement of the global value of Hubble’s constant,
Astrophysical Journal, 482, 75–82.
Labeyrie, A. (1975). Interference fringes obtained on Vega with two optical telescopes,
Astrophysical Journal, 196, L71–L75.
Lagage, P. O. and Cesarsky, C. J. (1983). The maximum energy of cosmic rays accelerated
by supernova shocks, Astronomy and Astrophysics, 125, 249–257.
Lamb, H. (1932). Hydrodynamics, 6th edn (Cambridge: Cambridge University Press).
References 483

Lanczos, K. (1922). Bemerkung zur de Sitterschen Welt (Remarks on de Sitter’s world

model), Physikalische Zeitschrift, 23, 539–543.
Landau, L. D. (1932). On the theory of stars, Physicalische Zeitschrift der Sowjetunion, 1,
285–288.
(1938). Origin of stellar energy, Nature, 141, 333–334.
Lane, J. H. (1870). On the theoretical temperature of the Sun; under the hypothesis of a
gaseous mass maintaining its volume by its internal heat, and depending on the laws
of gases as known by terrestrial experiment, American Journal of Science and Arts,
2nd Series, 50, 57–74.
Langley, S. P. (1886). On hitherto unrecognised wave-lengths, American Journal of Science,
32, 83–106.
(1900). The absorption lines in the infra-red spectrum of the Sun, Annals of the Smith-
sonian Astrophysical Observatory, 1, 5–21.
Lanzetta, K. M., Wolfe, A. M. and Turnshek, D. A. (1995). The IUE Survey for damped
Lyman-α and Lyman-limit absorption systems, Astrophysical Journal, 440, 435–457.
Large, M. I., Vaughan, A. E. and Mills, B. Y. (1968). A pulsar supernova association?,
Nature, 220, 340–341.
Larson, R. B. (1969a). The emitted spectrum of a proto-star, Monthly Notices of the Royal
Astronomical Society, 145, 297–308.
(1969b). Numerical calculations of the dynamics of collapsing proto-star, Monthly
Notices of the Royal Astronomical Society, 145, 271–295.
Lattes, C. M. G., Occhialini, G. P. S. and Powell, C. F. (1947). Observations on the tracks
of slow mesons in photographic emulsions, Nature, 160, 453–456.
Laue, M. von (1912). Eine quantative Prüfung der Theorie für die Interferenzerschein-
ungen bei Röntgenstrahlung (A quantitative test of the theory of X-ray interference
phenomena), Sitzberichte der Königlich Bayerischen Akademie der Wissenschaften,
pp. 363–373.
Le Verrier, U. J. J. (1859). Sur la théorie de Mercure et sur le mouvement du périhélie de
cette planète (On the theory of Mercury and the movement of the perihelion of this
planet), Comptes Rendus, 49, 379–383.
Leavitt, H. S. (1912). Periods of 25 variable stars in the Small Magellanic Cloud, Harvard
College Observatory Circular, No. 173, 1–2.
Leger, A. and Puget, J. L. (1984). Identification of the ‘unidentified’ IR emission features
of interstellar dust?, Astronomy and Astrophysics, 137, L5–L8.
Leibacher, J. W. and Stein, R. F. (1971). A new description of the solar five-minute oscilla-
tion, Astrophysical Letters, 7, 191–192.
Leighton, R. B. (1960). (In Discussion on) Considerations on local velocity fields in stellar
atmospheres: Prototype – the solar atmosphere, in Aerodynamic Phenomena in Stellar
Atmospheres, ed. Thomas, R. N. (Bologna: Nicola Zanichelli), pp. 321–327.
Leighton, R. B., Noyes, R. W. and Simon, G. W. (1962). Velocity fields in the solar atmo-
sphere. I. Preliminary report, Astrophysical Journal, 135, 474–499.
Lemaı̂tre, G. (1927). A homogeneous Universe of constant mass and increasing radius,
accounting for the radial velocity of extra-galactic nebulae, Annales de la Société
Scientifique de Bruxelles, A47, 29–39. Translation in Monthly Notices of the Royal
Astronomical Society, 91, (1931), 483–490.
484 References

Lemaı̂tre, G. (1931a). The beginning of the world from the point of view of quantum theory,
Nature, 127, 706.
(1931b). The expanding Universe, Monthly Notices of the Royal Astronomical Society,
91, 490–501.
(1933). Spherical condensations in the expanding Universe, Comptes Rendus de
L’Academie des Sciences de Paris, 196, 903–904.
Leventhal, M., MacCallum, C. J. and Stang, P. D. (1978). Detection of 511 keV positron
annihilation radiation from the Galactic center direction, Astrophysical Journal Letters,
225, L11–L14.
Liddle, A. R. and Lyth, D. (2000). Cosmological Inflation and Large-Scale Structure (Cam-
bridge: Cambridge University Press).
Lifshitz, E. (1946). On the gravitational stability of the expanding Universe, Journal of
Physics, Academy of Sciences of the USSR, 10, 116–129.
Lightman, A. P. and Schechter, P. L. (1990). The omega dependence of peculiar velocities
induced by spherical density perturbations, Astrophysical Journal Supplement Series,
74, 831–832.
Lilly, S. J. and Cowie, L. L. (1987). Deep infrared surveys, in Infrared Astronomy with
Arrays, eds Wynn-Williams, C. G. and Becklin, E. E. (Honolulu: Institute for Astron-
omy, University of Hawaii Publications), pp. 473–482.
Lilly, S. J. and Longair, M. S. (1984). Stellar populations in distant radio galaxies, Monthly
Notices of the Royal Astronomical Society, 211, 833–855.
Lilly, S. J., Tresse, L., Hammer, F., Crampton, D. and LeFevre, O. (1995). The Canada–
France redshift survey. VI. Evolution of the galaxy luminosity function to z ∼ 1,
Astrophysical Journal, 455, 108–124.
Lin, H., Kirshner, R. P., Shectman, S. A., Landy, S. D., Oemler, A. and Tucker, D. L.
(1996). The power spectrum of galaxy clustering in the Las Campanas Redshift Survey,
Astrophysical Journal, 471, 617–635.
Lindblad, B. (1925). Star-streaming and the structure of the stellar system, Arkiv för Matem-
atik, Astronomi och Fysik, 19A (21), 1–8.
(1927). On the cause of the ellipsoidal distribution of stellar velocities, Arkiv för Matem-
atik, Astronomi och Fysik, 20A (17), 1–7.
Linde, A. D. (1974). Is the Lee constant a cosmological constant?, Zhurnal Experimentalnoi
i Teoretichseskikh Fizica (JETP) Letters, 19, 183–184.
(1982). A new inflationary Universe scenario: a possible solution of the horizon, flatness,
homogeneity, isotropy and primordial monopole problems, Physics Letters, 108B,
389–393.
(1983). Chaotic inflation, Physics Letters, 129B, 177–181.
Lineweaver, C. H. (2005). Inflation and the cosmic microwave background, in The New
Cosmology: Proceedings of the 16th International Physics Summer School, Canberra,
ed. Colless, M. (Singapore: World Scientific).
Lobachevsky, N. I. (1829). On the principles of geometry, Kazanski Vestnik) (Kazan Mes-
senger).
(1830). On the principles of geometry, Kazanski Vestnik (Kazan Messenger).
References 485

Lockyer, J. N. (1900). Inorganic Evolution (London: Macmillan and Co.).

(1914). Notes on stellar classification II, Nature, 94, 618–619.
Longair, M. S. (1966). On the interpretation of radio source counts, Monthly Notices of the
Royal Astronomical Society, 133, 421–436.
(1971). Observational cosmology, Reports of Progress in Physics, 34, 1125–1248.
(1978). Radio astronomy and cosmology, in Observational Cosmology: 8th Advanced
Course, Swiss Society of Astronomy and Astrophysics, Saas-Fee 1978, eds Maeder, A.,
Martinet, L. and Tammann, G. (Geneva: Geneva Observatory Publications), pp. 125–
257.
(1997). The Friedman Robertson–Walker models: on bias, errors and acts of faith, in
Critical Dialogues in Cosmology, ed. Turok, N. (Singapore: World Scientific), pp. 285–
308.
Longair, M. S., Ryle, M. and Scheuer, P. A. G. (1973). Models of extended radio sources,
Monthly Notices of the Royal Astronomical Society, 164, 253–270.
Lovelace, R. V. E. and Romanova, M. M. (2003). Relativistic Poynting jets from accretion
disks, Astrophysical Journal, 596, L159–L162.
Lovell, A. C. B. (1987). The emergence of radio astronomy in the UK after World War II,
Quarterly Journal of the Royal Astronomical Society, 28, 1–9.
Low, F. J. (1961). Low-temperature germanium bolometer, Journal of the Optical Society
of America, 51, 1300–1304.
(1966). The infrared brightness temperature of Uranus, Astrophysical Journal, 146, 326–
328.
Low, F. J. and Aumann, H. H. (1970). Observations of Galactic and extragalactic
sources between 50 and 300 microns, Astrophysical Journal Letters, 162, L79–L85.
Low, F. J. and Johnson, H. L. (1964). Stellar photometry at 10 μm, Astrophysical Journal,
139, 1130–1134.
Low, F. J., Aumann, H. H. and Gillespie, C. M. (1970). Closing astronomy’s last frontier –
far infrared, Astronautics and Aeronautics, 8, 26–30.
Lund, N. (1984). Cosmic ray abundances, elemental and isotopic, in Cosmic Radiation
in Contemporary Astrophysics, ed. Shapiro, M. M. (Dordrecht: D. Reidel Publishing
Company), pp. 1–26.
Lundmark, K. (1920). The relations of the globular clusters and spiral nebulae to the stellar
system, Küngliga Svenska Vetenskaps-Akademiens Handlingar, 60, no. 8.
(1921). The Spiral Nebula Messier 33, Publications of the Astronomical Society of the
Pacific, 33, 324–327.
Lynden-Bell, D. (1967). Statistical mechanics of violent relaxation in stellar systems,
Monthly Notices of the Royal Astronomical Society, 136, 101–121.
(1969). Galactic nuclei as collapsed old quasars, Nature, 223, 690–694.
Lynden-Bell, D., Faber, S. M., Burstein, D. et al. (1988). Spectroscopy and photometry of
elliptical galaxies, Astrophysical Journal, 326, 19–49.
Lyubimov, V. A., Novikov, E. G., Nozik, V. Z., Tretyakov, E. F. and Kozik, V. S. (1980). An
estimate of the νe mass from the β-spectrum of tritium in the valine molecule, Physics
Letters, 138, 30–56.
486 References

Maanen, A. van (1916). Preliminary evidence of internal motion in the spiral nebula Messier
101, Astrophysical Journal, 44, 210–228.
(1921). Internal motion in four spiral nebulae, Publications of the Astronomical Society
of the Pacific, 33, 200–202.
(1935). Internal motion in spiral nebulae, Astrophysical Journal, 81, 336–337.
McCarthy, P. J., van Breugel, W. J. M., Spinrad, H. and Djorgovski, G. (1987). A cor-
relation between the radio and optical morphologies of distant 3CR radio galaxies,
Astrophysical Journal, 321, L29–L33.
Macchetto, F. D. and Dickinson, M. (1997). Galaxies in the young Universe, Scientific
American, 276, 66–73.
McClintock, J. E. (1992). Black holes in the Galaxy, in Proceedings of the Texas ESO/CERN
Symposium on Relativistic Astrophysics, Cosmology and Fundamental Particles, eds
Barrow, J. D., Mestel, L. and Thomas, P. A. (New York: New York Academy of Sci-
ences), pp. 495–502.
McCray, R. (1993). Supernova SN 1987A revisited, Annual Review of Astronomy and
Astrophysics, 31, 175–216.
McCrea, W. H. (1929). The hydrogen chromosphere, Monthly Notices of the Royal Astro-
nomical Society, 89, 483–497.
(1951). Relativity theory and the creation of matter, Proceedings of the Royal Society of
London, 206, 562–575.
(1970). A philosophy for Big Bang cosmology, Nature, 228, 21–24.
McKee, C. F. and Ostriker, J. P. (1977). A theory of the interstellar medium – three compo-
nents regulated by supernova explosions in an inhomogeneous substrate, Astrophysical
Journal, 218, 148–169.
McKellar, A. (1941). Molecular lines from the lowest states of the atomic molecules com-
posed of atoms probably present in interstellar space, Publications of the Dominion
Astrophysical Observatory (Victoria), 7, 251–272.
McLeod, J. M. and Andrew, B. H. (1968). The radio source VRO 42.22.01, Astrophysical
Letters, 1, 243.
Madau, P., Ferguson, H. C., Dickinson, M. E., Giavalisco, M., Steidel, C. C. and Fruchter,
A. (1996). High-redshift galaxies in the Hubble Deep Field: colour selection and star
formation history to z ∼ 4, Monthly Notices of the Royal Astronomical Society, 283,
1388–1404.
Madau, P., Pozzetti, L. and Dickinson, M. (1998). The star formation history of field galaxies,
Astrophysical Journal, 242, 106–116.
Maddox, S. J., Efstathiou, G., Sutherland, W. G. and Loveday, J. (1990). Galaxy correlations
on large scales, Monthly Notices of the Royal Astronomical Society, 242, 43P–47P.
Maeder, A. (1994). A selection of 10 most topical stellar problems, in Frontiers of Space
and Ground-Based Astronomy, eds Wamsteker, W., Longair, M. S. and Kondo, Y.
(Dordrecht: Kluwer Academic Publishers), pp. 177–186.
Mahoney, W. A., Ling, J. C., Wheaton, W. A. and Jacobson, A. S. (1984). HEAO 3 discovery
of Al-26 in the interstellar medium, Astrophysical Journal, 286, 578–585.
Malkan, M. and Sargent, W. L. (1982). The ultraviolet excess of Seyfert 1 galaxies and
quasars, Astrophysical Journal, 254, 22–37.
References 487

Malmquist, K. G. (1920). A study of stars of spectral type A, Meddelanden från Lunds

Astronomiska Observatorium, Series II, no. 22 (Lund: Scientia Publishers), pp. 1–69.
Margon, B. and Ostriker, J. P. (1973). The luminosity function of Galactic X-ray sources: a
cut-off and a ‘standard candle’?, Astrophysical Journal, 186, 91–96.
Markarian, B. E. (1967). Galaxies with an ultraviolet continuum, Astrofizica, 3, 24–38.
Markarian, B. E., Lipovetsky, V. A. and Stepanian, D. A. (1981). Galaxies with ultraviolet
continuum XV, Astrofizica, 17, 619–627. Translation in Astrophysics, 17, 1982, 321–
332.
Marx, G. and Szalay, A. S. (1972). Cosmological limit on neutretto mass, in Neutrino ’72,
vol. 1 (Budapest: Technoinform), pp. 191–195.
Mather, J. C., Cheng, E. S., Eplee Jr, R. E. et al. (1990). A preliminary measurement of the
cosmic microwave background spectrum by the Cosmic Background Explorer (COBE)
Satellite, Astrophysical Journal, 354, L37–L40.
Matt, G., Fabian, A. C. and Reynolds, C. S. (1997). Geometrical and chemical dependence
of K-shell X-ray features, Monthly Notices of the Royal Astronomical Society, 289,
175–184.
Matthews, T. A. and Sandage, A. R. (1963). Optical identification of 3C 48, 3C 196 and 3C
286 with stellar objects, Astrophysical Journal, 138, 30–56.
Matthews, T. A., Morgan, W. W. and Schmidt, M. (1964). A discussion of galaxies identified
with radio sources, Astrophysical Journal, 140, 35–49.
Matthewson, D. S. and Ford, V. L. (1970). Polarization observations of 1800 stars, Memoirs
of the Royal Astronomical Society, 74, 139–182.
Matz, S. M., Share, G. H., Leising, M. D., Chupp, E. L. and Vestrand, W. T. (1988). Gamma-
ray line emission from SN 1987A, Nature, 331, 416–418.
Maury, A. C. and Pickering, E. C. (1897). Spectra of bright stars photographed with the
11-inch Draper telescope as part of the Henry Draper Memorial, Annals of the Harvard
College Observatory (Part I), 28, 1–128.
Mayer-Hasselwander, H. A., Kanbach, G., Bennett, K. et al. (1982). Large-scale distribution
of Galactic gamma radiation observed by COS-B, Astronomy and Astrophysics, 105,
164–175.
Mayor, M. and Queloz, D. (1995). A Jupiter-mass companion to a solar-type star, Nature,
378, 355–359.
Mazzarella, J. M. and Balzano, V. A. (1986). A catalog of Markarian galaxies, Astrophysical
Journal Supplement Series, 62, 751–819.
Menzel, D. H. (1926). The planetary nebulae, Publications of the Astronomical Society of
the Pacific, 38, 295–312.
(1931). The general theory of absorption and emission lines, Publications of the Lick
Observatory, 17, 213–243.
Messier, C. (1784). Catalogue de Nébuleuses et des Amas d’Étoiles (Catalogue of nebulae
and clusters of stars), in Connaissance des Temps, ou Connaisance des Mouvemans
Céleste, Pour l’année bissextile 1784 (Paris: L’Académie Royal des Sciences), pp. 117–
269.
Mészáros, P. (1974). The behaviour of point masses in an expanding cosmological substra-
tum, Astronomy and Astrophysics, 37, 225–228.
488 References

Mészáros, P. (2002). Theories of gamma-ray bursts, Annual Review of Astronomy and

Astrophysics, 40, 137–169.
Mészáros, P. and Rees, M. J. (1993). Gamma-ray bursts: multiwaveband spectral predictions
for blast wave models, Astrophysical Journal, 418, L59–L62.
Metcalfe, N., Shanks, T., Campos, A., Fong, R. and Gardner, J. P. (1996). Galaxy formation
at high redshifts, Nature, 383, 236–237.
Michell, J. (1767). An inquiry into the probable parallax, and magnitude of the fixed stars,
from the quantity of light which they afford us, and the particular circumstances of
their situation, Philosophical Transactions of the Royal Society, 57, 234–264.
(1784). On the means of discovering the distance, magnitude, etc. of the fixed stars, in
consequence of the diminution of the velocity of their light, in case such a diminution
should be found to take place in any of them, and such other data should be procured
from observations, as would be farther necessary for that purpose, Philosophical Trans-
actions of the Royal Society, 74, 35–57.
Michelson, A. A. (1890). On the application of interference methods to astronomical mea-
surements, Philosophical Magazine, 30, 1–21.
Michelson, A. A. and Pease, F. G. (1921). Measurement of the diameter of Alpha Orionis
with the interferometer, Astrophysical Journal, 53, 249–259.
Migdal, A. B. (1959). Superfluidity and the moments of inertia of nuclei, Zhurnal Experi-
mentalnoi i Teoretichseskikh Fizica, 37, 249–263. Translation in Soviet Physics JETP,
10, 1960, 176–185.
Mihara, T., Makashima, K., Ohashi, T., Sakao, T. and Tashiro, M. (1990). New observations
of the cyclotron absorption feature in Hercules X-1, Nature, 346, 250–252.
Mikheyev, S. P. and Smirnov, A. Y. (1985). Resonance enhancement of oscillations in matter
and solar neutrino spectroscopy, Soviet Journal of Nuclear Physics, 42, 913–917.
Miley, G. K. (1968). Variation of the angular sizes of quasars with red-shift, Nature, 218,
933–934.
(1971). The radio structure of quasars – a statistical investigation, Monthly Notices of the
Royal Astronomical Society, 152, 477–490.
Miley, G. K., Perola, G. C., van der Kruit, P. and van der Laan, H. (1972). Active galaxies
with radio trails in clusters, Nature, 237, 269–272.
Miller, G. E. and Scalo, J. M. (1979). The initial mass function and stellar birthrate in the
solar neighborhood, Astrophysical Journal Supplement Series, 41, 513–547.
Mills, B. Y. and Slee, O. B. (1957). A preliminary survey of radio sources in a limited region
of the sky at a wavelength of 3.5 m, Australian Journal of Physics, 10, 162–194.
Milne, E. (1948). Kinematic Relativity (Oxford: Clarendon Press).
Milne, E. A. and McCrea, W. H. (1934a). Newtonian expanding Universe, Quarterly Journal
of Mathematics, 5, 64–72.
(1934b). Newtonian Universes and the curvature of space, Quarterly Journal of Mathe-
matics, 5, 73–80.
Minkowski, R. (1941). Spectra of supernovae, Publications of the Astronomical Society of
the Pacific, 53, 224–225.
(1942). The Crab Nebula, Astrophysical Journal, 96, 199–213.
References 489

Minkowski, R. (1960a). International cooperative efforts directed toward optical identifica-

tion of radio sources, Proceedings of the National Academy of Sciences of the United
States of America, 46, 13–19.
(1960b). A new distant cluster of galaxies, Astrophysical Journal, 132, 908–910.
Minnaert, M. and Mulders, G. (1930). Intensity measurement of the Fraunhofer lines in the
wavelength region 5150 to 5270 Å, Zeitschrift für Astrophysik, 1, 192–199.
Mirabel, I. F. and Rodrigues, L. F. (1994). A superluminal source in the Galaxy, Nature,
371, 46–48.
(1998). Microquasars in our Galaxy, Nature, 392, 673–676.
Miranda-Escudé, J., Cen, R., Ostriker, J. P. and Rauch, M. (1996). The Lyman alpha forest
from gravitational collapse in the CDM + lambda model, Astrophysical Journal, 471,
582–616.
Mitchell, R. J., Culhane, J. L., Davison, P. J. N. and Ives, J. C. (1976). Ariel 5 observations of
the X-ray spectrum of the Perseus Cluster, Monthly Notices of the Royal Astronomical
Society, 175, 29P–34P.
Miyoshi, M., Moran, J., Herrnstein, J. et al. (1995). Evidence for a black-hole from high
rotation velocities in a sub-parsec region of NGC4258, Nature, 373, 127–129.
Møller, P. and Jakobsen, P. (1990). The Lyman continuum opacity at high redshifts – through
the Lyman forest and beyond the Lyman valley, Astronomy and Astrophysics, 228, 299–
309.
Monck, W. H. S. (1895). The spectra and colours of the stars, Journal of the British Astro-
nomical Association, 5, 416–419.
Moran, J. M., Crowther, P. P., Burke, B. F. et al. (1967). Spectral line interferometry with
independent time standards at stations separated by 845 kilometers, Science, 157,
676–677.
Morgan, W. W., Keenan, P. C. and Kellman, E. (1943). Atlas of Stellar Spectra, with an
outline of Spectral Classification. (Chicago: University of Chicago Press).
Morgan, W. W., Sharpless, S. and Osterbrock, D. (1951). Some features of Galactic structure
in the neighbourhood of the Sun (abstract), Astronomical Journal, 57, 3.
Morgan, W. W., Whitford, A. E. and Code, A. D. (1953). Studies in Galactic structure. I. A
preliminary determination of the space distribution of the blue giants, Astrophysical
Journal, 118, 318–322.
Muller, C. A. and Oort, J. H. (1951). The interstellar hydrogen line at 1420 MHz and an
estimate of galactic rotation, Nature, 168, 356–358.
Myers, S. T., Baker, J. E., Readhead, A. C. S., Leitch, E. M. and Herbig, T. (1997). Mea-
surements of the Sunyaev–Zeldovich effect in the nearby clusters A478, A2142, and
A2256, Astrophysical Journal, 485, 1–21.
Nakajima, T., Oppenheimer, B. R., Kulkarni, S. R., Golimowski, D. A., Matthews, K. and
Durrance, S. T. (1995). Discovery of a cool brown dwarf, Nature, 378, 463–465.
Nemiroff, R. J. (1994). A century of gamma ray burst models, Comments on Astrophysics,
17, 189–205.
Neugebauer, G. and Leighton, R. B. (1969). Two-micron Sky Survey: A Preliminary Cata-
logue (Washington: NASA SP-3047).
490 References

Newman, E. T., Couch, K., Chinnapared, K., Exton, A., Prakash, A. and Torrence, R. (1965).
Metric of a rotating charged mass, Journal of Mathematical Physics, 6, 918–919.
Neyman, J., Scott, E. L. and Shane, C. D. (1954). The index of clumpiness of the distribution
of images of galaxies, Astrophysical Journal Supplement, 1, 269–293.
Novikov, I. D. (1964). On the possibility of [the] appearance of large scale inhomogeneities
in the expanding Universe, Journal of Experimental and Theoretical Physics, 46, 686–
689.
Oda, M., Gorenstein, P., Gursky, H., Kellogg, E., Schreier, E., Tananbaum, H. and Giacconi,
R. (1971). X-ray pulsations from Cygnus X-1 observed from UHURU, Astrophysical
Journal, 166, L1–L7.
O’Dell, C. R., Peimbert, M. and Kinman, T. D. (1964). The planetary nebulae in the globular
cluster M15, Astrophysical Journal, 140, 119–129.
Ohm, E. A. (1961). Project Echo: receiving system, Bell System Technical Journal, 40,
1065–1094.
Oke, J. B. (1950). A theoretical Hertzsprung–Russell diagram for red dwarf stars, Journal
of the Royal Astronomical Society of Canada, 44, 135–148.
Oliver, S. J., Rowan-Robinson, M. and Saunders, W. (1992). Infrared background constraints
on the evolution of IRAS galaxies, Monthly Notices of the Royal Astronomical Society,
256, 15P–22P.
Omnes, R. (1969). Possibility of matter–antimatter separation at high temperatures, Physical
Review Letters, 23, 38–40.
Oort, J. H. (1927). Observational evidence confirming Lindblad’s hypothesis of a rotation
of the Galactic system, Bulletin of the Astronomical Institutes of the Netherlands, 3,
275–282.
(1932). The force exerted by the stellar system in the direction perpendicular to the
Galactic plane and some related problems, Bulletin of the Astronomical Institutes of
the Netherlands, 6, 249–287.
(1958). Distribution of galaxies and density in the Universe, in Solvay Conference on The
Structure and Evolution of the Universe (Brussels: Institut International de Physique
Solvay), pp. 163–181.
Oort, J. H. and Walraven, T. (1956). Polarization and composition of the Crab Nebula,
Bulletin of the Astronomical Institutes of the Netherlands, 12, 285–311.
Oort, J. H., Kerr, F. J. and Westerhout, G. (1958). The Galactic system as a spiral nebula,
Monthly Notices of the Royal Astronomical Society, 118, 379–389.
Öpik, E. (1922). An estimate of the distance of the Andromeda Nebula, Astrophysical
Journal, 55, 406–410.
(1938). Stellar structure, source of energy, and evolution, Publications of the Astronomical
Observatory of the University of Tartu, 30 (3), 1–115.
(1951). Stellar models with variable compositions. II Sequences of models with energy
generation proportional to the fifteenth power of temperature, Proceedings of the Royal
Irish Academy, 54, 49–77.
Oppenheimer, B. R., Kulkarni, S. R., Matthews, K. and Nakajima, T. (1995). Infrared
spectrum of the cool brown dwarf GL229B, Science, 270, 1478–1479.
References 491

Oppenheimer, J. R. and Snyder, H. (1939). On continued gravitational contraction, Physical

Review, 56, 455–459.
Oppenheimer, J. R. and Volkoff, G. M. (1939). On massive neutron cores, Physical Review,
55, 374–381.
Osmer, P. S. (1982). Evidence for a decrease in the space density of quasars at z more than
about 3.5, Astrophysical Journal, 253, 28–37.
Osterbrock, D. E. and Rogerson, J. B. (1961). The helium and heavy-element content of
gaseous nebulae and the Sun, Publications of the Astronomical Society of the Pacific,
73, 129–134.
Ostriker, J. P. and Gunn, J. E. (1969). On the nature of pulsars. I. Theory, Astrophysical
Journal, 157, 1395–1417.
Ostriker, J. P. and Peebles, P. J. E. (1973). A numerical study of the stability of flattened
galaxies: or, can cold galaxies survive?, Astrophysical Journal, 186, 467–480.
Ostriker, J. P. and Tremaine, S. D. (1975). Another evolutionary correction to the luminosity
of giant galaxies, Astrophysical Journal, 202, L113–L117.
Oswalt, T. D., Smith, J. A., Wood, M. A. and Hintzen, P. (1996). A lower limit of 9.5 gyr on
the age of the galactic disk from the oldest white dwarf stars, Nature, 382, 692–694.
Owen, F. N. and Ledlow, M. J. (1994). The FR I/II break and the bivariate luminosity
function in Abell clusters of galaxies, in First Stromlo Symposium: Physics of Active
Galactic Nuclei, ASP Conference Series, vol. 34, eds Bicknell, G. V., Dopita, M. A.
and Quinn, P. J. (San Francisco: ASP), pp. 319–323.
Pacini, F. (1967). Energy emission from a neutron star, Nature, 216, 567–568.
(1968). Rotating neutron stars, pulsars and supernova remnants, Nature, 219, 145–146.
Padmanabhan, T. (1993). Structure Formation in the Universe (Cambridge: Cambridge
University Press).
(1996). Cosmology and Astrophysics Through Problems (Cambridge: Cambridge Uni-
versity Press), p. 437–440.
Page, L. (1997). Review of observations of the CMB, in Critical Dialogues in Cosmology,
ed. Turok, N. (Singapore: World Scientific), pp. 343–362.
Pagel, B. E. J. (1997). Nucleosynthesis and Chemical Evolution of Galaxies (Cambridge:
Cambridge University Press).
Panagia, N., Gilmozzi, R., Macchetto, F., Adorf, H.-M. and Kirshner, R. P. (1991). Properties
of the SN 1987A circumstellar ring and the distance to the Large Magellanic Cloud,
Astrophysical Journal, 380, L23–L26.
Pankey Jr., T. (1962). Possible Thermonuclear Activities in Natural Terrestrial Minerals,
Ph.D. thesis, Howard University.
Papaloizou, J. C. B. and Pringle, J. E. (1984). The dynamical stability of differentially
rotating discs with constant specific angular momentum, Monthly Notices of the Royal
Astronomical Society, 208, 721–750.
Partridge, R. B. (1980a). Flucutations in the cosmic microwave background radiation at
small angular scales, Physica Scripta, 21, 624–629.
(1980b). New limits on small-scale angular fluctuations in the cosmic microwave back-
ground, Astrophysical Journal, 235, 681–687.
492 References

Partridge, R. B. (1999). Current status of the cosmic microwave background radiation, in

Cosmological Parameters and the Evolution of the Universe, IAU Symposium no. 183,
ed. Sato, K. (Dordrecht: Kluwer Academic Publishers), pp. 74–87.
Pauli, W. (1925). Über den Zusammenhang des Abschlusses der Elektronengruppen im
Atom mit der Komplexstruktur der Spektrum (On the connection of filled shell phe-
nomena in atoms with complex structure of atomic spectra), Zeitschrift für Physik, 31,
765–783.
Payne, C. H. (1925). Stellar Atmospheres: Harvard College Observatory Monographs, No. 1
(Cambridge, Massachusetts: Harvard University Press).
Peacock, J. A. and Dodds, S. J. (1994). Reconstructing the linear power spectrum of cos-
mological mass fluctuations, Monthly Notices of the Royal Astronomical Society, 267,
1020–1034.
Peacock, J. A. and Heavens, A. F. (1985). The statistics of maxima in primordial density
perturbations, Monthly Notices of the Royal Astronomical Society, 217, 805–820.
Peacock, J. A., Cole, S., Norberg, P. et al. (2001). A measurement of the cosmological
mass density from clustering in the 2df galaxy redshift survey, Nature, 410, 169–
173.
Pearson, T. J. and Readhead, A. C. S. (1984). Image formation by self-calibration in radio
astronomy, Annual Reviews of Astronomy and Astrophysics, 22, 97–130.
Pearson, T. J., Unwin, S. C., Cohen, M. H. et al. (1981). Superluminal expansion of quasar
3C273, Nature, 290, 365–368.
Pearson, T. J., Unwin, S. C., Cohen, M. H. et al. (1982). Superluminal expansion of 3C273,
in Extragalactic Radio Sources, eds Heeschen, D. S. and Wade, C. M. (Dordrecht:
D. Reidel Publishing Company), pp. 355–356.
Pearson, T. J., Mason, B. S., Readhead, A. C. S. et al. (2003). The anisotropy of the microwave
background to l = 3500: mosaic observations with the cosmic background imager,
Astrophysical Journal, 591, 556–574.
Pease, F. G. (1921). The angular diameter of α Bootis by the interferometer, Publications
of the Astronomical Society of the Pacific, 33, 171–173.
(1931). Interferometer methods in astronomy, Ergebnisse der Exakten Naturwis-
senschaften, 10, 84–96.
Peebles, P. J. E. (1966a). Primeval helium abundance and the primeval fireball, Physical
Review Letters, 16, 410–413.
(1966b). Primordial helium abundance and the primordial fireball II, Astrophysical Jour-
nal, 146, 542–552.
(1968). Recombination of the primeval plasma, Astrophysical Journal, 153, 1–11.
(1976). A cosmic virial theorem, Astrophysics and Space Science, 45, 3–19.
(1980). The Large-Scale Structure of the Universe (Princeton: Princeton University
Press).
(1982). Large-scale background temperature fluctuations due to scale-invariant primaeval
perturbations, Astrophysical Journal, 263, L1–L5.
(1993). Principles of Physical Cosmology (Princeton: Princeton University Press).
Peebles, P. J. E. and Yu, J. T. (1970). Primeval adiabatic perturbation in an expanding
Universe, Astrophysical Journal, 162, 815–836.
References 493

Pei, Y. C. and Fall, S. M. (1995). Cosmic chemical evolution, Astrophysical Journal, 454,
69–76.
Penrose, R. (1965). Gravitational collapse and space-time singularities, Physical Review
Letters, 14, 57–59.
(1969). Gravitational collapse: the role of general relativity, Rivista Nuovo Cimento, 1,
252–276.
Penzias, A. A. and Wilson, R. W. (1965). A measurement of excess antenna temperature at
4080 MHz, Astrophysical Journal, 142, 419–421.
Perley, R. A., Dreher, J. W. and Cowan, J. J. (1984). The jet and filaments in Cygnus A,
Astrophysical Journal, 285, L35–L38.
Perlmutter, S., Boyle, B., Bunclark, P. et al. (1996). High-redshift supernova discoveries
on demand: first results from a new tool for cosmology and bounds on q0 , Nuclear
Physics B, 51, 20–29.
Perlmutter, S., Gabi, S., Goldhaber, G. et al. (1997). Measurements of the cosmological
parameters omega and lambda from the first seven supernovae at z > 0.35, Astrophys-
ical Journal, 483, 565–581.
Perlmutter, S., Aldering, G., della Valle, M. et al. (1998). Discovery of a supernova explosion
at half the age of the Universe, Nature, 391, 51–54.
Permutter, S., Aldering, G., Goldhaber, G., et al. (1999). Measurements of  and  from
42 high–redshift supernovae, Astrophysical Journal, 517, (2), 565–586.
Peterson, B. M., Balonek, T. J., Barker, E. S. et al. (1991). Steps toward determination of the
size and structure of the broad-line region in active galactic nuclei. II – An intensive
study of NGC 5548 at optical wavelengths, Astrophysical Journal, 368, 119–137.
Pettini, M., King, D. L., Smith, L. J. and Hunstead, R. W. (1997). The metallicity of high-
redshift galaxies: the abundance of zinc in 34 damped Ly-α systems from z = 0.7 to
3.4, Astrophysical Journal, 486, 665–680.
Pettit, E. and Nicholson, S. B. (1928). Stellar radiation measurements, Astrophysical Jour-
nal, 68, 279–308.
Phillips, M. M. (1993). The absolute magnitudes of Type IA supernovae, Astrophysical
Journal, 413, L105–L108.
Pickering, E. C. (1890). The Draper catalogue of stellar spectra photographed with the
8-inch Bache telescope as part of the Henry Draper Memorial, Annals of the Harvard
College Observatory, 27, 1–388.
(1896). Stars having peculiar spectra. New variable stars in Crux and Cygnus, Astro-
physical Journal, 4, 369–370. This note is also Harvard College Observatory Circular
no. 12.
(1897). Stars having peculiar spectra, Astrophysical Journal, 5, 92–94. This note is also
Harvard College Observatory Circular no. 16.
(1908). Revised Harvard photometry, Annals of the Harvard College Observatory, 50,
1–252.
(1912). Distribution of stellar spectra, Annals of the Harvard College Observatory, 56
(1), 1–26.
Pilkington, J. D. H. and Scott, P. F. (1965). A survey of radio sources between declinations
20 and 40, Monthly Notices of the Royal Astronomical Society, 69, 183–224.
494 References

Planck, M. (1900). On the theory of the laws of the energy distribution in the normal
spectrum, Verhandlungen der Deutschen Physikalische Gesellschaft, 2, 237–245.
Plaskett, J. S. (1923). The H and K lines of calcium in O-type stars, Monthly Notices of the
Royal Astronomical Society, 84, 80–93.
Plaskett, J. S. and Pearce, J. A. (1933). The problems of diffuse matter in the Galaxy,
Publications of the Dominion Astrophysical Observatory, 5, 167–237.
Pouillet, C.-S. (1838). Mémoire sur la chaleur solaire, sur les pouvoirs rayonnants et
absorbants de l’air atmosphérique, et sur la temperature de l’espace (Memoir on the
heat of the Sun, on the radiative and absorptive powers of atmospheric air and on the
temperature of space), Comptes Rendus de l’Academie des Sciences, 7, 24–65.
Pozdnyakov, L. A., Sobol, I. M. and Sunyaev, R. A. (1983). Comptonization and the shaping
of X-ray source spectra: Monte Carlo calculations, Astrophysics and Space Science
Reviews, 2, 189–331.
Prendergast, K. H. and Burbidge, G. R. (1968). On the nature of some Galactic X-ray
sources, Astrophysical Journal, 151, L83–L88.
Press, W. H. and Schechter, P. (1974). Formation of galaxies and clusters of galaxies by
self-similar gravitational condensation, Astrophysical Journal, 187, 425–438.
Pskovskii, Y. P. (1977). Light curves, color curves, and expansion velocity of Type I super-
novae as functions of the rate of brightness decline, Astronomicheskii Zhurnal, 54,
1188–1201. Translation in Soviet Astronomy, 21, 1977, 675–682.
(1984). Photometric classification and basic parameters of Type I supernovae, Astro-
nomicheskii Zhurnal, 61, 1125–1136. Translation in Soviet Astronomy, 28, 1984, 658–
664.
Radhakrishnan, V. and Cooke, D. J. (1969). Magnetic poles and the polarisation structure
of pulsar radiation, Astrophysics Letters, 3, 225–229.
Radhakrishnan, V. and Manchester, R. N. (1969). Detection of a change of state in the pulsar
PSR 0833−45, Nature, 222, 228–229.
Radhakrishnan, V., Cooke, D. J., Komesaroff, M. M. and Morris, D. (1969). Evidence in
support of a rotational model for the pulsar PSR 0833-45, Nature, 221, 443–446.
Rappaport, S., Doxsey, R. and Zaumen, W. (1971). A search for X-ray pulsations from
Cygnus X-1, Astrophysical Journal, 168, L43–L47.
Rappaport, S. A. and Joss, P. C. (1983). X-ray pulsars in massive binary systems, in Accre-
tion Driven Stellar X-ray Sources, eds Lewin, W. H. G. and van den Heuvel, E. P. J.
(Cambridge: Cambridge University Press), pp. 1–39.
Razin, V. A. (1958). The polarization of cosmic radio radiation at wavelengths of 1.45 and
3.3 meters, Astronomicheskii Zhurnal, 35, 241–252. Translation in Soviet Astronomy,
2, 1958, 216–225.
Reber, G. (1940). Cosmic static, Astrophysical Journal, 91, 621–624.
(1944). Cosmic static, Astrophysical Journal, 100, 279–287.
Rees, M. J. (1966). Appearance of relativistically expanding radio sources, Nature, 211,
468–470.
(1967). Studies in radio source structure – I. A relativistically expanding model for
variable quasi-stellar radio sources, Monthly Notices of the Royal Astronomical Society,
135, 345–360.
References 495

Rees, M. J. (1971). New interpretation of extragalactic radio sources, Nature, 229, 312–317.
(1976). Beam models for double sources and the nature of the primary energy source,
in The Physics of Non-thermal Radio Sources, ed. Setti, G. (Dordrecht: D. Reidel
Publishing Company), pp. 107–120.
Rees, M. J. and Ostriker, J. P. (1977). Cooling, dynamics and fragmentation of massive gas
clouds – clues to the masses and radii of galaxies and clusters, Monthly Notices of the
Royal Astronomical Society, 179, 541–559.
Rees, M. J. and Sciama, D. W. (1968). Large-scale density inhomogeneities in the Universe,
Nature, 217, 511–516.
Rees, M. J., Phinney, E. S., Begelman, M. C. and Blandford, R. D. (1982). Ion-supported
tori and the origin of radio jets, Nature, 295, 17–21.
Reichley, P. E. and Downs, G. S. (1969). Observed decrease in the periods of pulsar PSR
0833−45, Nature, 222, 229–230.
Reines, F. and Cowan Jr, C. L. (1956). The neutrino, Nature, 178, 446–449.
Richards, D. W. and Comella, J. M. (1969). The period of pulsar NP 0532, Nature, 222,
551–552.
Riemann, B. (1854). Über die Hypothesen welche der Geometrie zu Grunde liegen (On
the Hypotheses that Lie at the Foundations of Geometry) (Göttingen: University of
Göttingen). Habilitationschrift.
Riess, A. G., Press, W. H. and Kirshner, R. P. (1995). Using Type IA supernova light curve
shapes to measure the Hubble constant, Astrophysical Journal, 438, L17–L20.
Rindler, W. (1956). Visual horizons in world models, Monthly Notices of the Royal Astro-
nomical Society, 116, 662–677.
Ritchey, G. W. (1917). Novae in spiral nebulae, Publications of the Astronomical Society
of the Pacific, 29, 210–212.
Ritter, A. (1883a). Untersuchungen über die Constitution Gasförmiger Weltkörper
(Researches on the Constitution of Gaseous Celestial Bodies), Wiedemanns Annalen,
20, 897–927.
(1883b). Untersuchungen über die Constitution Gasf örmiger Weltkörper (Researches
on the Constitution of Gaseous Celestial Bodies), Wiedemanns Annalen, 20, 137–
160.
(1898). On the constitution of gaseous celestial bodies, Astrophysical Journal, 8, 293–
315.
Robertson, H. P. (1928). On relativistic cosmology, Philosophical Magazine, 5, 835–848.
(1935). Kinematics and world structure, Astrophysical Journal, 82, 284–301.
Rochester, G. D. and Butler, C. C. (1947). Evidence for the existence of new unstable
elementary particles, Nature, 160, 855–857.
Rogers, A. E. E., Hinteregger, H. F., Whitney, A. R. et al. (1974). The structure of radio
sources 3C 273B and 3C 84 deduced from the ‘closure’ phases and visibility ampli-
tudes observed with three-element interferometers, Astrophysical Journal, 193, 293–
301.
Rogers, F. J. and Iglesias, C. A. (1994). Astrophysical opacity, Science, 263, 50–55.
Rogerson, J. B. and York, D. G. (1973). Interstellar deuterium abundance in the direction
of Beta Centauri, Astrophysical Journal, 186, L95–L98.
496 References

Rogerson, J. B., Spitzer, L., Drake, J. F. et al. (1973a). Spectrophotometric results from the
Copernicus satellite. I. Instrumentation and performance, Astrophysical Journal, 181,
L97–L102.
Rogerson, J. B., York, D. G., Drake, J. F., Jenkins, E. B., Morton, D. C. and Spitzer, L.
(1973b). Spectrophotometric results from the Copernicus satellite. III. Ionization and
composition of the intercloud medium, Astrophysical Journal Letters, 181, L110–
L115.
Roll, P. G. and Wilkinson, D. T. (1966). Cosmic background radiation at 3.2 cm – support
for cosmic black-body radiation, Physical Review Letters, 16, 405–407.
Röntgen, W. C. (1895). Über eine neue Art von Strahlen (On a new type of ray. Preliminary
communication), Erste Mittheilung: Sitzungsberichte der Physikalisch-Medizinische
Gesellschaft, Würzburg, 137, 132–141.
Rosseland, S. (1924). Note on the absorption of radiation within a star, Monthly Notices of
the Royal Astronomical Society, 84, 525–528.
Rossi, B. (1970). An X-ray pulsar in the Crab Nebula, in Non-Solar X- and Gamma-Ray
Astronomy, IAU Symposium No. 37, ed. Gratton, L. (Dordrecht: D. Reidel Publishing
Company), pp. 183–184.
Rougoor, G. W. and Oort, J. H. (1960). Distribution and motion of interstellar hydrogen
in the Galactic system with particular reference to the region within 3 kiloparsecs of
the center, Proceedings of the National Academy of Sciences of the United States of
America, 46, 1–13.
Rowan-Robinson, M. (1968). The determination of the evolutionary properties of quasars
by means of the luminosity-volume test, Monthly Notices of the Royal Astronomical
Society, 141, 445–458.
Rowan-Robinson, M., Benn, C. R., Lawrence, A., McMahon, R. G. and Broadhurst, T. J.
(1993). The evolution of faint radio sources, Monthly Notices of the Royal Astronomical
Society, 263, 123–130.
Rubin, V. C., Thonnard, N. and Ford, W. K. (1980). Rotational properties of 21 Sc galaxies
with a large range of luminosities and radii from NGC 4605 (R = 4 kpc) to UGC2885
(R = 122 kpc), Astrophysical Journal, 238, 471–487.
Ruderman, M. (1969). Neutron-Starquakes and Pulsar Periods (New York: Department of
Physics, New York University).
Russell, H. N. (1912a). On the determination of the orbital elements of eclipsing variable
stars I, Astrophysical Journal, 35, 315–340.
(1912b). On the determination of the orbital elements of eclipsing variable stars II,
Astrophysical Journal, 36, 54–74.
(1914a). Relations between the spectra and other characteristics of stars, Popular Astron-
omy, 22, 275–294.
(1914b). Relations between the spectra and other characteristics of stars, Popular Astron-
omy, 22, 331–351.
(1914c). Relations between the spectra and other characteristics of the stars, I. Historical,
Nature, 93, 227–230.
(1914d). Relations between the the spectra and other characteristics of the stars, II.
Brightness and spectral class, Nature, 93, 252–258.
References 497

Russell, H. N. (1914e). Relations between the the spectra and other characteristics of the
stars, III., Nature, 93, 281–286.
(1921). Response at the meeting of the Royal Astronomical Society on Friday 11 February
1921 on receipt of the Gold Medal of the Society, Observatory, 44, 71–2.
(1922). The theory of ionization and the sun-spot spectrum, Astrophysical Journal, 55,
119–144.
(1925). The problem of stellar evolution, Nature, 116, 209–212.
(1929). On the composition of the Sun’s atmosphere, Astrophysical Journal, 70,
11–82.
Russell, H. N. and Saunders, F. A. (1925). New regularities in the spectra of the alkaline
earths, Astrophysical Journal, 61, 38–69.
Russell, H. N. and Shapley, H. (1912a). On darkening at the limb in eclipsing variables I,
Astrophysical Journal, 36, 239–254.
(1912b). On darkening at the limb in eclipsing variables II, Astrophysical Journal, 36,
385–408.
Russell, H. N., Adams, W. S. and Moore, C. E. (1928). A calibration of Rowland’s scale of
intensities for solar lines, Astrophysical Journal, 68, 1–8.
Rutherford, E. (1899). Uranium radiation and the electrical conduction produced by it,
Philosophical Magazine, Series 5, 47, 109–163.
(1907). Some cosmical aspects of radioactivity, Journal of the Royal Astronomical Society
of Canada, 1, 145–165.
(1919). Collisions of α particles with light atoms, IV. An anomalous effect in nitrogen,
Philosophical Magazine, Series 6, 37, 581–587.
(1920). Nuclear constitution of atoms, Proceedings of the Royal Society of London, A97,
374–400.
Rutherford, E. and Andrade, E. N. da C. (1913). The reflection of γ -rays from crystals,
Nature, 92, 267.
Rutherford, E. and Chadwick, J. (1921). The artificial disintegration of light elements,
Philosophical Magazine, Series 6, 42, 809–825.
Rutherford, E. and Royds, T. (1909). The nature of the α particle from radioactive substances,
Philosophical Magazine, Series 6, 15, 281–286.
Ryle, M. (1955). Radio stars and their cosmological significance, The Observatory, 75,
137–147.
Ryle, M. and Graham Smith, F. (1948). A new intense source of radio-frequency radiation
in the constellation of Cassiopeia, Nature, 162, 462–463.
Ryle, M. and Neville, A. C. (1962). A radio survey of the north polar region with a
4.5 minutes of arc pencil-beam system, Monthly Notices of the Royal Astronomical
Society, 125, 39–56.
Ryle, M., Elsmore, B. and Neville, A. C. (1965). High resolution observations of the radio
sources in Cygnus and Cassiopeia, Nature, 205, 1259–1262.
Sachs, R. K. and Wolfe, A. M. (1967). Perturbations of a cosmological model and angular
variations in the microwave background, Astrophysical Journal, 147, 73–90.
Saha, M. N. (1920). Ionization in the solar chromosphere, Philosophical Magazine, 40,
479–488.
498 References

Saha, M. N. (1921). On the physical theory of stellar spectra, Proceedings of the Royal
Society of London, 99A, 135–153.
Sahu, K. C., Livio, M., Petro, L. et al. (1997). The optical counterpart to gamma-ray burst
GRB 970228 observed using the Hubble Space Telescope, Nature, 387, 476–478.
Sakharov, A. D. (1965). The initial stage of an expanding Universe and the appearance of a
nonuniform distribution of matter, Zhurnal Eksperimentalnoi i Teoreticheskoi Fiziki,
49, 345–358. Translation in Soviet Physics JETP, 22, 1966, 241–249.
(1967). Violation of CP invariance, C asymmetry, and baryon asymmetry of the Universe,
Zhurnal Experimentalnoi i Teoretichseskikh Fizica (JETP) Letters, 5, 32–35.
Salpeter, E. E. (1952). Nuclear reactions in stars without hydrogen, Astrophysical Journal,
115, 326–328.
(1955). The luminosity function and stellar evolution, Astrophysical Journal, 121, 161–
167.
(1964). Accretion of interstellar matter by massive objects, Astrophysical Journal, 140,
796–800.
Sampson, R. A. (1895). On the rotation and mechanical state of the Sun, Memoirs of the
Royal Astronomical Society, 51, 123–183.
Sandage, A. R. (1958). Current problems in the extragalactic distance scale, Astrophysical
Journal, 127, 513–526.
(1961a). The ability of the 200-inch telescope to discriminate between selected world
models, Astrophysical Journal, 133, 355–392.
(1961b). The Hubble Atlas of Galaxies (Washington D.C.: Carnegie Institution of Wash-
ington); Publication 618.
(1965). The existence of a major new constituent of the Universe: the quasistellar galaxies,
Astrophysical Journal, 141, 1560–1578.
(1968). Observational cosmology, The Observatory, 88, 91–106.
(1970). Cosmology – the search for two numbers, Physics Today, 23, 34–41.
(1995). Practical cosmology: inventing the past, in The Deep Universe, by Sandage, A.
R., Kron, R. G. and Longair, M. S., eds Binggeli, B. and Buser, R. (Berlin: Springer-
Verlag), pp. 1–232.
Sandage, A. R. and Hardy, E. (1973). The redshift-distance relation. VII. Absolute mag-
nitudes of the first three ranked cluster galaxies as a function of cluster richness and
Bautz–Morgan cluster type: the effect on q0 , Astrophysical Journal, 183, 743–758.
Sandage, A. R. and Schwarzschild, M. (1952). Inhomogeneous stellar models II. Models
with exhausted cores in gravitational contraction, Astrophysical Journal, 116, 463–
476.
Sandage, A. R., Osmer, P., Giacconi, R. et al. (1966). On the optical identification of Sco
X-1, Astrophysical Journal, 146, 316–321.
Sanders, W. T., Kraushaar, W. L., Nousek, J. A. and Fried, P. M. (1977). Soft diffuse
X-rays in the southern Galactic hemisphere, Astrophysical Journal Letters, 217,
L87–L91.
Sargent, W. L. W., Young, P. J., Lynds, C. R., Boksenberg, A., Shortridge, K. and Hartwick,
F. D. A. (1978). Dynamical evidence for a central mass concentration in the galaxy
M87, Astrophysical Journal, 221, 731–744.
References 499

Saunders, W., Rowan-Robinson, M., Lawrence, A. et al. (1990). The 60 μm and far-infrared
luminosity functions of IRAS Galaxies, Monthly Notices of the Royal Astronomical
Society, 242, 318–337.
Schade, D., Lilly, S. J., Crampton, D., Hammer, F., LeFevre, O. and Tresse, L. (1995).
Canada–France redshift survey: Hubble Space Telescope imaging of high-redshift
field galaxies, Astrophysical Journal Letters, 451, L1–L4.
Schalén, C. (1936). Über Probleme der Interstellaren Absoption (On problems of interstellar
absorption), Nova Acta Regiae Societatis Scientiarum Upsaliensis, Series IV, Vol. 10,
no. 1 (also Uppsala Astronomiska Observatoriums Meddelanden, no. 64).
Schechter, P. (1976). An analytic expression for the luminosity function of galaxies, Astro-
physical Journal, 203, 297–306.
Scheiner, J. (1899). On the spectrum of the Great Nebula in Andromeda, Astrophysical
Journal, 9, 149–150.
Scheuer, P. A. G. (1957). A statistical method for analysing observations of faint radio stars,
Proceedings of the Cambridge Philosophical Society, 53, 764–773.
(1965). A sensitive test for the presence of atomic hydrogen in intergalactic space, Nature,
207, 963.
(1974). Models of extragalactic radio sources with a continuous energy supply from a
central object, Monthly Notices of the Royal Astronomical Society, 166, 513–528.
(1975). Radio astronomy and cosmology, in Stars and Stellar Systems, vol. 9, eds Sandage,
A. R., Sandage, M. and Kristian, J. (Chicago: University of Chicago Press), pp. 725–
760.
Schmidt, B. P., Kirshner, R. P. and Eastman, R. G. (1992). Expanding photospheres of
Type II supernovae and the extragalactic distance scale, Astrophysical Journal, 395,
366–386.
Schmidt, B. V. (1931). Ein lichtstarkes komafreies Spiegelsystem (A wide-field coma-free
mirror system), Zentralzeitung für Optik und Mechanik, 52, 25–26.
Schmidt, G. C. (1898). Ueber die von den Thorvebindungen und einigen anderen Substanzen
ausgehende Strahlung (On the emitted radiation from thorium compounds and several
other substances), Annalen der Physik und Chemie (Wiedemanns Annalen), 65, 141–
151.
Schmidt, M. (1963). 3C 273: a star-like object with large red-shift, Nature, 197, 1040.
(1965). Large redshifts of five quasi-stellar sources, Astrophysical Journal, 141, 1295–
1300.
(1968). Space distribution and luminosity functions of quasi-stellar sources, Astrophysi-
cal Journal, 151, 393–409.
Schmidt, M. and Green, R. F. (1983). Quasar evolution derived from the Palomar bright
quasar survey and other complete quasar surveys, Astrophysical Journal, 269, 352–
374.
Schmidt, M. and Matthews, T. A. (1964). Redshift of the quasi-stellar radio sources 3C 47
and 3C 147, Astrophysical Journal, 139, 781–785.
Schmidt, M., Schneider, D. P. and Gunn, J. E. (1986). Spectroscopic CCD surveys for
quasars at large redshift. II – A PFUEI transit survey, Astrophysical Journal, 310,
518–533.
500 References

Schmidt , M., Schneider, D. P. and Gunn, J. E. (1995). Spectrscopic CCD surveys for quasars
at large redshift. IV. Evolution of the luminosity function from quasars detected by their
Lyman-alpha emission, Astronomical Journal, 110, 68–77.
Schneider, D. P., Schmidt, M. and Gunn, J. E. (1994). Spectroscopic CCD surveys for
quasars at large redshift. 3: The Palomar transit GRISM survey catalog, Astronomical
Journal, 107, 1245–1269.
Schödel, R., Ott, T., Genzel, R. et al (2002). A star in a 15.2-year orbit around the super-
massive black hole at the centre of the Milky Way, Nature, 419, 694–696.
Schönberg, M. and Chandrasekhar, S. (1942). On the evolution of the main-sequence stars,
Astrophysical Journal, 96, 161–171.
Schott, G. A. (1912). Electromagnetic Radiation (Cambridge: Cambridge University Press).
Schramm, D. N. (1997). The age of the Universe, in Critical Dialogues in Cosmology, ed.
Turok, N. (Singapore: World Scientific), pp. 81–91.
Schramm, D. N. and Wasserburg, G. J. (1970). Nucleochronologies and the mean age of the
elements, Astrophysical Journal, 162, 57–69.
Schreier, E., Levinson, R., Gursky, H., Kellogg, E., Tananbaum, H. and Giacconi, R. (1972).
Evidence for the binary nature of Centaurus X-3 from UHURU X-ray observations,
Astrophysical Journal, 172, L79–L89.
Schuster, A. (1902). The solar atmosphere, Astrophysical Journal, 16, 320–327.
(1905). Radiation through a foggy atmosphere, Astrophysical Journal, 21, 1–22.
Schwarzschild, K. (1900a). Über das zulässige Krümmungsmass des Raumes (On an upper
limit to the curvature of space), Vierteljahrsschrift der Astronomischen Gesellschaft,
35, 337–347.
(1900b). Über die Photographische Vergleichung der Helligkeit Verschiedenfarbiger
Sterne (On the photographic comparison of the brightness of different coloured
stars), Sitzungsberichte der Kaiserlichen Akademie der Wissenschaften in Wien,
Mathematisch-naturwissenschaft, Klasse 2a, 109, 1127–1134.
(1906). Über das Gleichgewicht der Sonnenatmosphäre (On the equilibrium of the solar
atmosphere), Nachrichten von der Königlichen Gesselschaft der Wissenschaften zu
Göttingen, Mathematisch-physikalische Klasse, 41, pp. 1–24.
(1907). Über die Eigenbewegung der Fixsterne (On the proper motion of the fixed stars),
Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, pp. 614–631.
(1916). Über das Gravitationsfeld einis Massenpunktes nach der Einsteinschen Theorie
(On the gravitational field of a point mass according to Einsteinian theory), Sitzungs-
berichte der Königlich Preussischen Akademie der Wissenschaften zu Berlin, 1, 189–
196.
Schwarzschild, M. (1958). Structure and Evolution of the Stars (Princeton: Princeton Uni-
versity Press).
(1979). A numerical model for a triaxial stellar system in dynamical equilibrium, Astro-
physical Journal, 232, 236–247.
Schwinger, J. (1946). Electron radiation in high energy accelerators, Physical Review, 70,
798.
(1949). On the classical radiation of accelerated electrons, Physical Review, 75, 1912–
1925.
References 501

Scott, P. F., Carreira, P., Cleary, K. et al (2003). First results from the Very Small Array –
III. The cosmic microwave background power spectrum, Monthly Notices of the Royal
Astronomical Society, 341, 1076–1083.
Seaton, M. J., Yan, Y., Mihalas, D. and Pradhan, A. K. (1994). Opacities for stellar envelopes,
Monthly Notices of the Royal Astronomical Society, 266, 805–828.
Secchi, A. (1866). Nouvelles Recherches sur l’Analyse Spectrale de la Lumière des Étoiles
(New researches on the spectral analysis of the light of the stars), Comptes Rendus,
63, 621–628.
(1868). Sur les Spectres Stellaires (On the spectra of the stars), Comptes Rendus, 66,
124–126.
Seldner, M., Siebars, B., Groth, E. J. and Peebles, P. J. E. (1977). New reduction of the Lick
catalog of galaxies, Astronomical Journal, 82, 249–256.
Sellgren, K. (1984). The near-infrared continuum emission of visual reflection nebulae,
Astrophysical Journal, 277, 623–633.
Serjeant, S., Dunlop, J. S., Mann, R. G. et al. (2003). Submillimetre observations of the
Hubble Deep Field and flanking fields, Monthly Notices of the Royal Astronomical
Society, 344, 887–904.
Severny, A. B., Kotov, V. A. and Tsap, T. T. (1976). Observations of solar pulsations, Nature,
259, 87–89.
Seyfert, C. K. (1943). Nuclear emission in spiral nebulae, Astrophysical Journal, 97, 28–40.
Shakeshaft, J. R., Ryle, M., Baldwin, J. E., Elsmore, B. and Thomson, J. H. (1955). A radio
survey of radio sources between declinations −38 and +83, Memoirs of the Royal
Astronomical Society, 67, 106–154.
Shakura, N. and Sunyaev, R. A. (1973). Black holes in binary systems. Observational
appearance, Astronomy and Astrophysics, 24, 337–355.
Shane, C. D. and Wirtanen, C. A. (1957). The distribution of galaxies, Publications of the
Lick Observatory, 22, 1–60.
Shapiro, I. I. (1964). Fourth test of general relativity, Physical Review Letters, 13, 789–791.
Shapiro, M. M. (1991). A brief introduction to the cosmic radiation, in Cosmic Rays,
Supernovae and the Interstellar Medium, eds Shapiro, M. M., Silberberg, R. and Wefel,
J. P. (Dordrecht: Kluwer Academic Publishers), pp. 1–28.
Shapley, H. (1915). Orbits of eighty seven eclipsing binaries – a summary, Astrophysical
Journal, 38, 158–174.
(1917). Note on the magnitude of novae in spiral nebulae, Publications of the Astronom-
ical Society of the Pacific, 29, 213–217.
(1918). Studies based on the colors and magnitudes in stellar clusters. VII. The distances,
distribution in space, and dimensions of 60 globular clusters, Astrophysical Journal,
48, 154–181.
(1921). The scale of the Universe, Bulletin of the National Research Council, 2, 171–193.
Shellard, P. (2003). The future of cosmology: observational and computational prospects,
in The Future of Theoretical Physics and Cosmology, eds Gibbons, G. W., Shellard,
E. P. S. and Rankin, S. J. (Cambridge: Cambridge University Press), pp. 755–780.
Shklovsky, I. S. (1953). On the nature of the radiation from the Crab Nebula, Dokladi
Akademiya Nauk SSSR, 90, 983–986.
502 References

Shklovsky, I. S. (1967). The nature of the X-ray source Sco X-1, Astronomicheskii Zhurnal,
44, 930–938. Translation in Soviet Astronomy, 11, 1967, 749–755.
Shmoanov, T. (1957). A method for measuring the absolute effective radiation temperature
of radio emission at low equivalent temperatures, Pribory i Tekhnika Experimenta
(Instruments and Experimental Methods), 1, 83–86.
Shu, F. H., Adams, F. C. and Lizano, S. (1987). Star formation in molecular clouds –
observation and theory, Annual Reviews of Astronomy and Astrophysics, 25, 23–81.
Silk, J. (1968). Cosmic black-body radiation and galaxy formation, Astrophysical Journal,
151, 459–471.
Silk, J. and Wyse, R. F. G. (1993). Galaxy formation and Hubble sequence, Physics Reports,
231, 293–365.
Simpson, J. A. (1983). Elemental and isotopic composition of Galactic cosmic rays, Annual
Reviews of Nuclear and Particle Science, 33, 323–381.
Skobeltsyn, D. (1929). Über eine neue Art sehr schneller β-strahlen (On a new type of very
fast β-ray), Zeitschrift für Physik, 54, 686–702.
Slipher, V. M. (1909). Peculiar star spectra suggestive of selective absorption of light in
space, Bulletins of the Lowell Observatory, 2, 1–2.
(1917). A spectrographic investigation of spiral nebulae, Proceedings of the American
Philosophical Society, 56, 403–409.
Smail, I., Ivison, R. J. and Blain, A. W. (1997). A deep sub-millimeter survey of lensing
clusters: a new window on galaxy formation and evolution, Astrophysical Journal
Letters, 490, L5–L8.
Smith, H. J. and Hoffleit, D. (1963). Light variations in the superluminous radio galaxy 3C
273, Nature, 198, 650–651.
Smoot, G. F., Bennett, C. L., Kogut, A. et al. (1992). Structure in the COBE differential
microwave radiometer first-year maps, Astrophysical Journal, 396, L1–L5.
Sneden, C., McWilliam, A., Preston, G. W., Cowan, J. J., Burris, D. L. and Armosky, B. J.
(1992). The ultra-metal-poor, neutron-capture-rich giant star CS 22892-052, Astro-
physical Journal, 467, 819–840.
Snell, R. L., Loren, R. B. and Plambeck, R. L. (1980). Observations of CO in L1551
– evidence for stellar driven wind shocks, Astrophysical Journal Letters, 239,
L17–L22.
Snyder, L. E., Buhl, D., Zuckerman, B. and Palmer, P. (1969). Microwave detection of
interstellar formaldehyde, Physical Review Letters, 22, 679–681.
Soldner, J. G. von (1804). On the deflection of a light ray from its straight motion due to the
attraction of a world body which it passes closely, Astronomisches Jahrbuch für das
Jahr 1804 (Berlin: Späthen), pp. 161–172.
Spergel, D. N., Verde, L., Peiris, H. V. et al. (2003). First-year Wilkinson Microwave
Anisotropy Probe (WMAP) observations: determination of cosmological parameters,
Astrophysical Journal Supplement Series, 148, 175–194.
Spite, F. and Spite, M. (1982). Abundance of lithium in unevolved halo stars and old disk
stars – interpretation and consequences, Astronomy and Astrophysics, 115, 357–366.
Spitzer, L. and Savedoff, M. P. (1950). The temperature of interstellar matter, Astrophysical
Journal, 111, 593–608.
References 503

Staelin, D. H. and Reifenstein III, E. C. (1968). Pulsating radio sources near the Crab Nebula,
Science, 162, 1481–1483.
Stahler, S. W., Shu, F. J. and Taam, R. E. (1980). The evolution of protostars I – Global
formation and results, Astrophysical Journal, 241, 637–654.
Starobinsky, A. A. (1985). Cosmic background anisotropy induced by isotropic flat-
spectrum gravitational-wave perturbations, Soviet Astronomy Letters, 11, 133–137.
In Russian: Pis’ma k Astronomicheskii Zhurnal, 11, 1985, 323–330.
Stebbins, J., Hufford, C. M. and Whitford, A. E. (1940). The mean coefficient of selective
absorption in the Galaxy, Astrophysical Journal, 92, 193–199.
Steidel, C. C. (1998). Galaxy evolution: has the ‘epoch of galaxy formation’ been found?, in
Eighteenth Texas Symposium on Relativistic Astrophysics and Cosmology, eds Olinto,
A. V., Frieman, J. A. and Schramm, D. N. (River Edge, N.J.: World Scientific Publishing
Company), pp. 124–135.
Steidel, C. C. and Hamilton, D. (1992). Deep imaging of high redshift QSO fields below
the Lyman limit. I – The field of Q0000−263 and galaxies at z = 3.4, Astronomical
Journal, 104, 941–949.
Stobie, R. S. (1969). Cepheid pulsation-III. Models fitted to a new mass-luminosity relation,
Monthly Notices of the Royal Astronomical Society, 144, 511–535.
Stockton, A. N. and Lynds, C. R. (1966). The remarkable absorption spectrum of 3C 191,
Astrophysical Journal, 144, 451–453.
Stoner, E. C. (1929). The limiting density in white dwarf stars, Philosophical Magazine, 7,
63–70.
Storrie-Lombardi, L. J., McMahon, R. G. and Irwin, M. J. (1996). Evolution of neutral gas
at high redshift: implications for the epoch of galaxy formation, Monthly Notices of
the Royal Astronomical Society, 283, L79–L83.
Stoughton, D., Lupton, R. H., Bernardi, M. et al. (2002). Sloan Digital Sky Survey: early
data release, Astronomical Journal, 123, 485–548.
Strömberg, G. (1924). The asymmetry in stellar motions and the existence of a velocity-
restriction in space, Astrophysical Journal, 59, 228–251.
Strömgren, B. (1932). The opacity of stellar matter and the hydrogen content of the stars,
Zeitschrift für Astrophysik, 4, 118–152.
(1933). On the interpretation of the Hertzsprung–Russell Diagram, Zeitschrift für Astro-
physik, 7, 222–238.
(1939). The physical state of interstellar hydrogen, Astrophysical Journal, 89, 526–
547.
Struve, F. G. W. (1840). Über die Parallaxe des Sterns α Lyrae (On the parallax of the star
α Lyrae), Astronomische Nachrichten, 396, 177–180. The page numbers refer to the
columns of the journal.
Suess, H. E. and Urey, H. C. (1956). Abundances of the elements, Reviews of Modern
Physics, 28, 53–74.
Sunyaev, R. A. and Zeldovich, Y. B. (1970a). Interaction of matter and radiation in the hot
model of the Universe, Astrophysics and Space Science, 7, 21–30.
(1970b). Small-scale fluctuations of relic radiation, Astrophysics and Space Science, 7,
3–19.
504 References

Sunyaev, R. A. and Zeldovich, Y. B. (1980). Microwave background radiation as a probe of

the contemporary structure and history of the Universe, Annual Review of Astronomy
and Astrophysics, 18, 537–560.
Szalay, A. S. and Marx, G. (1976). Neutrino rest mass from cosmology, Astronomy and
Astrophysics, 49, 437–441.
Tanaka, Y., Nandra, K., Fabian, A. C. et al. (1995). Gravitationally redshifted emission
implying an accretion disk and massive black-hole in the active galaxy MCG:-6-30-
15, Nature, 375, 659–661.
Tananbaum, H., Gursky, H., Kellogg, E. M., Levinson, R., Schreier, E. and Giacconi,
R. (1972). Discovery of a periodic binary X-ray source in Hercules from UHURU,
Astrophysical Journal, 174, L144–L149.
Taylor, J. H. (1992). Pulsar timing and relativistic gravity, Philosophical Transactions of
the Royal Society, 341, 117–134.
Tegmark, M., Strauss, M. A., Blanton, M. R. et al. (2004). Cosmological parameters from
SDSS and WMAP, Physical Review D, 69, 103501 (1–28).
Thomson, W. (1854a). On the mechanical energies of the Solar System, British Association
Report, Part II.
(1854b). On the mechanical energies of the Solar System, Philosophical Magazine, 8,
409–430.
Thorne, K. S., Price, R. H. and Macdonald, D. A. (1986). Black Holes: The Membrane
Paradigm (New Haven: Yale University Press).
Tinsley, B. M. (1980). Evolution of the stars and gas in galaxies, Fundamentals of Cosmic
Physics, 5, 287–388.
Tolman, R. C. (1934). Effect of inhomogeneity on cosmological models, Proceedings of
the National Academy of Sciences, 20, 169–176.
Tonks, L. and Langmuir, I. (1929). Oscillations in ionized gases, Physical Review, 33,
195–210.
Tonry, J. L., Schmidt, B. P., Barris, B. et al. (2003). Cosmological results from high-z
supernovae, Astrophysical Journal, 594, 1–24.
Toomre, A. and Toomre, J. (1972). Galactic bridges and tails, Astrophysical Journal, 178,
623–666.
Toutain, T. and Frölich, C. (1992). Characteristics of solar p-modes – results from the IPHIR
experiment, Astronomy and Astrophysics, 257, 287–297.
Tremaine, S. and Gunn, J. E. (1979). Dynamical role of light neutral leptons in cosmology,
Physical Review Letters, 42, 407–410.
Tremaine, S. and Richstone, D. O. (1977). A test of a statistical model for the luminosities
of bright cluster galaxies, Astrophysical Journal, 212, 311–316.
Trimble, V. L. and Thorne, K. S. (1969). Spectroscopic binaries and collapsed stars, Astro-
physical Journal, 156, 1013–1019.
Trümper, J., Pietsch, W., Reppin, C., Voges, W., Steinbert, R. and Kendziorra, E. (1978).
Evidence for strong cyclotron line emission in the hard X-ray spectrum of Hercules
X-1, Astrophysical Journal Letters, 219, L105–L110.
Trumpler, R. J. (1930). Preliminary results on the distances, dimensions, and space distri-
bution of open star clusters, Lick Observatory Bulletin, 14, 154–188.
References 505

Tully, R. B. and Fisher, J. R. (1977). A new method of determining distances to galaxies,

Astronomy and Astrophysics, 54, 661–673.
Turok, N., ed. (1997). Critical Dialogues in Cosmology (Singapore: World Scientific).
Turtle, A. J. (1963). The spectrum of the Galactic radio emission, II, Monthly Notices of the
Royal Astronomical Society, 126, 405–417.
Turtle, A. J., Pugh, J. F., Kenderdine, S. and Pauliny-Toth, I. I. K. (1962). The spectrum of
the Galactic radio emission, I. Observations of low resolving power, Monthly Notices
of the Royal Astronomical Society, 124, 297–312.
Tyson, A. (1990). Spectrum and origin of the extragalactic optical background radiation,
in Galactic and Extragalactic Background Radiation, eds Bowyer, S. and Leinert, C.
(Dordrecht: Kluwer Academic Publishers), pp. 245–255.
Ulrich, M. H., Boksenberg, A., Bromage, G. E. et al. (1984). Detailed observations of NGC
4151 with IUE – III. Variability of the strong emission lines from 1978 February to
1980 May, Monthly Notices of the Royal Astronomical Society, 206, 221–238.
Ulrich, R. K. (1970). The five-minute oscillations of the solar surface, Astrophysical Journal,
162, 993–1002.
Unsöld, A. (1928). Über die Struktur der Fraunhoferschen Linien und die Quantitative
Spektralanalyse der Sonnenatmosphäre (On the structure of the Fraunhofer lines and
a quantitative spectral analysis of the solar atmosphere), Zeitschrift für Physik, 46,
765–781.
Urey, H., Brickwedde, F. G. and Murphy, G. M. (1932). A hydrogen isotope of mass 2,
Physical Review, 39, 164–165.
Urry, C. M. and Padovani, P. (1994). Unification of BL Lac objects and FR1 radio galaxies, in
First Stromlo Symposium: Physics of Active Galactic Nuclei, ASP Conference Series,
vol. 34, eds Bicknell, G. V., Dopita, M. A. and Quinn, P. J. (San Francisco: ASP),
pp. 215–226.
Vacanti, G., Cawley, M. F., Colombo, E. et al. (1991). Gamma-ray observations of the Crab
Nebula at TeV energies, Astrophysical Journal, 377, 469–479.
Vashakidze, M. A. (1954). On the degree of polarization of the light near extragalactic
nebulae and the Crab Nebula, Astronomicheskikh Tsirkular, no. 147, 11–13.
Velikhov, E. P. (1959). Stability of an ideally conducting liquid flowing between cylinders
rotating in a magnetic field, Zhurnal Eksperimentalnoi i Teoreticheskoi Fiziki, 36,
1398–1404. Translation in Soviet Physics – JETP, 9, 1959, 995–998.
Vidal-Madjar, A., Laurent, C., Bonnet, R. M. and York, D. G. (1977). The ratio of deuterium
to hydrogen in interstellar space. III – The lines of sight to Zeta Puppis and Gamma
Cassiopaeia, Astrophysical Journal, 211, 91–107.
Villard, P. (1900a). Sur la réflection et la réfraction des rayons cathodique et les rayons
déviables de radium (On the reflection and refraction of cathode rays and the devi-
able rays of radium), Comptes Rendus de L’Academie des Sciences, 130, 1010–
1012.
(1900b). Sur le rayonnement du radium (On the radiation of radium), Comptes Rendus
de L’Academie des Sciences, 130, 1178–1179.
Vogel, H. C. (1874). Spectralanalytische Mitteilungen (Communications on spectral anal-
ysis), Astronomische Nachrichten, 84, 113–124.
506 References

Vorontsov-Velyaminov, B. A. (1959). Atlas and Catalogue of Interacting Galaxies, Part I

(Moscow: Sternberg Institute, Moscow State University).
(1977). Atlas of interacting galaxies, part II, and the concept of fragmentation of galaxies,
Astronomy and Astrophysics Supplement Series, 28, 1–117.
Wagoner, R. V. (1973). Big-Bang nucleosynthesis revisited, Astrophysical Journal, 179,
343–360.
Wagoner, R. V., Fowler, W. A. and Hoyle, F. (1967). On the synthesis of elements at very
high temperatures, Astrophysical Journal, 148, 3–49.
Walker, A. G. (1936). On Milne’s theory of world structure, Proceedings of the London
Mathematical Society, Series 2, 42, 90–127.
Walker, M. F. (1956). Studies of extremely young clusters, Astrophysical Journal Supple-
ment, 2, 365–387.
Walker, R. G. and Price, S. D. (1975). Air Force Cambridge Research Laboratories Infrared
Sky Survey (Cambridge, Massachusetts: AFCRL TR-0373).
Wall, J. V. (1996). Space distribution of radio source populations, in Extragalactic Radio
Sources, IAU Symposium no. 175, eds Ekers, R., Fanti, C. and Padrielli, L. (Dordrecht:
Kluwer Academic Publishers), pp. 547–552.
Walsh, D., Carswell, R. F. and Weymann, R. J. (1979). 0957+561A, B – twin quasistellar
objects or gravitational lens?, Nature, 279, 381–384.
Wandel, A. and Mushotzky, R. F. (1986). Observational determination of the masses of
active galactic nuclei, Astrophysical Journal, 306, L61–L66.
Warren, S. J., Hewett, P. C., Irwin, M. J., McMahon, R. G. and Bridgeland, M. T. (1987).
First observation of a quasar with a redshift of 4, Nature, 325, 131–133.
Warren, S. J., Hewett, P. C. and Osmer, P. S. (1994). A wide-field multicolor survey for
high-redshift quasars, z ≥ 2.2. III: The luminosity function, Astrophysical Journal,
421, 412–433.
Weaver, H., Williams, D. R. W., Dieter, N. H. and Lum, W. T. (1965). Observations of a
strong unidentified microwave line and of emission from the OH molecule, Nature,
208, 29–31.
Webber, W. R. (1983). Cosmic ray electrons and positrons – a review of current measure-
ments and some implications, in Composition and Origin of Cosmic Rays, ed. Shapiro,
M. M. (Dordrecht: D. Reidel Publishing Company), pp. 83–100.
Weber, J. (1961). General Relativity and Gravitational Waves, Interscience Tracts on Physics
and Astronomy (New York: Interscience).
(1966). Observation of the thermal fluctuations of a gravitational-wave detector, Physical
Review Letters, 17, 1228–1230.
(1969). Evidence for discovery of gravitational radiation, Physical Review Letters, 22,
1320–1324.
(1970). Anisotropy and polarization in the gravitational-radiation experiments, Physical
Review Letters, 25, 180–184.
Webster, B. L. and Murdin, P. (1972). Cygnus X-1: a spectroscopic binary with a heavy
companion?, Nature, 235, 37–38.
Weedman, D. (1994). Starburst galaxies at high redshift, in First Stromlo Symposium:
Physics of Active Galactic Nuclei, ASP Conference Series, vol. 34, eds Bicknell,
G. V., Dopita, M. A. and Quinn, P. J. (San Francisco: ASP), pp. 409–415.
References 507

Weekes, T. C., Cawley, M. F., Fegan, D. J. et al. (1989). Observation of TeV gamma rays from
the Crab Nebula using the atmospheric Cerenkov imaging technique, Astrophysical
Journal, 342, 379–395.
Weinberg, S. (1989). The cosmological constant problem, Reviews of Modern Physics, 61,
1–23.
(1997). Theories of the cosmological constant, in Critical Dialogues in Cosmology, ed.
Turok, N. (Singapore: World Scientific), pp. 195–203.
Weinheimer, C. (2001). Neutrino mass from tritium β-decay, in Dark Matter in Astro-
and Particle Physics, Proceedings of the International Conference DARK 2000, ed.
Klapdor-Kleingrothaus, H. V. (Berlin: Springer-Verlag), pp. 513–519.
Weinreb, S., Barrett, A. H., Meeks, M. L. and Henry, J. C. (1963). Radio observations of
OH in the interstellar medium, Nature, 200, 829–831.
Weinreb, S., Meeks, M. L., Carter, J. C., Barrett, A. H. and Rogers, A. E. E. (1965).
Observations of polarized OH emission, Nature, 208, 440–441.
Weizsäcker, C. F. von (1937). Element transformation inside stars. I, Physikalische
Zeitschrift, 38, 176–191.
(1938). Element transformation inside stars. II, Physikalische Zeitschrift, 39, 633–646.
Wesselink, A. J. (1947). The observations of brightness, colour and radial velocity of δ-
Cephei and the pulsation hypothesis, Bulletin of the Astronomical Institutes of the
Netherlands, 10, 91–99. Errata, 10, 258 and 310.
Westerhout, G., Seeger, C. L., Brouw, W. N. and Tinbergen, J. (1962). Polarization of the
Galactic 75-cm radiation, Bulletin of the Astronomical Institutes of the Netherlands,
16, 187–212.
Weyl, H. (1923). Zur allgemeinen Relativitätstheorie (On the theory of general relativity),
Physikalische Zeitschrift, 29, 230–232.
Weymann, R. J. (1966). The energy spectrum of radiation in the expanding Universe, Astro-
physical Journal, 145, 560–571.
Wheeler, J. A. (1968). Our Universe: the known and the unknown, American Scientist, 56,
1–20.
(1977). Genesis and observership, in Foundational Problems in the Special Science, eds
Butts, R. E. and Hintikka, J. (Dordrecht: D. Reidel Publishing Company), pp. 3–33.
White, S. D. (1989). Observable signatures of young galaxies, in The Epoch of Galaxy
Formation, eds Frenk, C. S., Ellis, R. S., Shanks, T., Heavens, A. F. and Peacock, J. A.
(Dordrecht: Kluwer Academic Publishers), pp. 15–30.
Whitford, A. E. (1948). An extension of the interstellar absorption-curve, Astrophysical
Journal, 107, 102–105.
Whitney, A. R., Shapiro, I. I., Rogers, A. E. E. et al. (1971). Quasars revisited: rapid time
variations observed via very-long-baseline interferometry, Science, 173, 225–230.
Will, C. M. (2001). The confrontation between general relativity and experiment,
Living Review in Relativity, 4. Online article: cited on 15 August 2001
http://www.livingreviews.org/Articles/Volume4/2001-4will/.
Wilson, C. T. R. (1901). On the ionisation of atmospheric air, Proceedings of the Royal
Society of London, 68, 151–161.
Wilson, R. W., Jefferts, K. B. and Penzias, A. A. (1970). Carbon monoxide in the Orion
Nebula, Astrophysical Journal Letters, 161, L43–L44.
508 References

Windhorst, R. A., Dressler, A. and Koo, D. A. (1987). Ultradeep optical identifications

and spectroscopy of faint radio sources, in Observational Cosmology, eds Hewitt, A.,
Burbidge, G. and Fang, L.-Z. (Dordrecht: D. Reidel Publishing Co.), pp. 573–576.
Windhorst, R. A., Fomalont, E. B., Kellermann, K. I. et al. (1995). Identification of faint
radio sources with optically luminous interacting disk galaxies, Nature, 375, 471–474.
Wirtz, C. W. (1922). Einiges zur Statistik der Radialgeschwindigkeiten von Spiralnebeln
und Kugelsternhaufen (Some remarks on the statistics of the radial velocities of spiral
nebulae and star clusters), Astronomische Nachrichten, 215, 349–354.
Wolf, M. (1923). On the dark nebula NGC 6960, Astronomische Nachrichten, 219, 109–116.
Wolfe, A. M. (1988). Damped Ly-α absorption systems, in QSO Absorption Lines: Prob-
ing the Universe, eds Blades, J. C., Turnshek, D. and Norman, C. A. (Cambridge:
Cambridge University Press), pp. 306–317.
Wolfenstein, L. (1978). Neutrino oscillations in matter, Physical Review D, 17, 2369–2374.
Wollaston, W. H. (1802). A method of examining refractive and dispersive powers, by
prismatic reflection, Philosophical Transactions of the Royal Society, 92, 365–380.
Wolszczan, A. and Frail, D. (1992). A planetary system around the millisecond pulsar PSR
1257+12, Nature, 255, 145–147.
Woodard, M. F. and Libbrecht, K. G. (1988). On the measurement of solar rotation using
high-degree p-mode oscillations, in Seismology of the Sun and Sun-Like Stars, ed.
Rolfe, E. J. (Noorwijk: ESA Publications), pp. 67–71.
Wright, W. W., Palmer, H. K., Albrecht, S. and Campbell, W. W. (1911). Radial velocities
of 150 stars south of declination −20◦ determined by the D. O. Mills expedition period
1903–1906, Publications of the Lick Observatory, 9 (Part 4), 71–347.
Young, P. J., Westphal, J. A., Kristian, J., Wilson, C. P. and Landauer, F. P. (1978). Evidence
for a supermassive object in the nucleus of the galaxy M87 from SIT and CCD area
photometry, Astrophysical Journal, 221, 721–730.
Young, T. (1802). On the theory of light and colours, Philosophical Transactions of the
Royal Society, 92, 12–48.
Yukawa, H. (1935). On the interaction of elementary particles. I, Proceedings of the
Physical-Mathematical Society of Japan, 17, 48–57.
Zanstra, H. (1926). An application of the quantum theory to the luminosity of diffuse
nebulae, Physical Review, 27, 644.
(1927). An application of the quantum theory to the luminosity of diffuse nebulae, Astro-
physical Journal, 65, 50–70.
Zeldovich, Y. B. (1964a). The fate of a star and the evolution of gravitational energy upon
accretion, Soviet Physics Doklady, 9, 195–197.
(1964b). Observations in a Universe homogeneous in the mean, Astronomicheskii Zhur-
nal, 41, 19–24. Translation in Soviet Astronomy, 8, 1964, 13–16.
(1965). Survey of modern cosmology, Advances of Astronomy and Astrophysics, 3, 241–
379.
(1968). The cosmological constant and the theory of elementary particles, Uspekhi Fizish-
eskikh Nauk, 95, 209–230.
(1970). Gravitational instability: an approximate theory for large density perturbations,
Astronomy and Astrophysics, 5, 84–89.
References 509

Zeldovich, Y. B. (1972). A hypothesis, unifying the structure and the entropy of the Universe,
Monthly Notices of the Royal Astronomical Society, 160, 1P–3P.
Zeldovich, Y. B. and Guseynov, O. H. (1966). Collapsed stars in binaries, Astrophysical
Journal, 144, 840–841.
Zeldovich, Y. B. and Novikov, I. D. (1964). Mass of quasi-stellar objects, Soviet Physics
Doklady, 9, 834–837.
Zeldovich, Y. B. and Sunyaev, R. A. (1969). The interaction of matter and radiation in a
hot-model Universe, Astrophysics and Space Science, 4, 301–316.
(1980). Astrophysical implications of the neutrino rest mass. I – The Universe, Pis’ma v
Astronomicheskii Zhurnal, 6, 451–456.
Zeldovich, Y. B., Kurt, D. and Sunyaev, R. A. (1968). Recombination of hydrogen in the hot
model of the Universe, Zhurnal Eksperimentalnoi i Teoreticheskoi Fiziki, 55, 278–286.
Translation in Soviet Physics – JETP, 28, 1969, 146–150.
Zhevakin, S. A. (1953). On the theory of Cepheids. I., Astronomisheskii Zhurnal, 30, 161–
179.
Zuckerman, B., Turner, B. E., Johnson, D. R. et al. (1975). Detection of interstellar trans-
ethyl alcohol, Astrophysical Journal, 196, L99–L102.
Zwicky, F. (1933). Rotverschiebung von Extragalaktischen Nebeln (The redshift of extra-
galactic nebulae), Helvetica Physica Acta, 6, 110–118.
(1937). On the masses of nebulae and of clusters of nebulae, Astrophysical Journal, 86,
217–246.
(1942). On the large scale distribution of matter in the Universe, Physical Review, 61,
489–503.
(1968). Catalogue of Selected Compact Galaxies and of Post-eruptive Galaxies (Guem-
lingen, Switzerland: F. Zwicky).
 Name index

Dates are given on pages with numbers in bold type.

Aaronson, Marc, 248, 344 Barrow, John, 443

Abbot, Charles, 156 Barthel, Peter, 299
Abell, George O., 254, 255 Bautz, Laura, 265
Abramowicz, Marek, 279 Baym, Gordon, 195, 196
Adams, Walter, 36, 41, 63, 89, 105 Becker, Herbert, 131
Airy, George, 18 Becker, Robert, 428
Albrecht, Andeas, 446 Becklin, Eric E., 158, 226, 227, 228
Alexander, Paul, 293 Becquerel, Edmond, 9
Alfvén, Hannes, 139, 398 Becquerel, Henri, 130
Aller, Lawrence, 217, 241 Bell, Anthony, 293
Alpher, Ralph, 320, 321, 322, 323, 329, 394, 395 Bell(-Burnell), Jocelyn, 67, 193, 196, 211
Ambartsumian, Viktor Amazaspovich, 225 Bennett, Andrew, 304
Anders, Edward, 348 Bennett, Charles, 11
Anderson, Carl, 48, 135, 171 Bentley, Richard, 100, 120
Anderson, John A., 45 Bergeron, Jacqueline, 381
Anderson, Wilhelm, 64 Bergh, Sidney van den, 380
Ando, Hiroyasu, 183 Bertola, Francesco, 253
Andrade, Edward Neville da Costa, 131 Bessel, Friedrich, 6, 8, 101
Andrew, Brian, 274 Best, Philip N., 298
Antonucci, Robert, 299 Bethe, Hans, 49, 50, 182, 195, 320
Arago, François, 9 Binney, James, 283
Aragòn-Salamanca, Alfonso, 350 Birkinshaw, Mark, 263
Archer, Frederick Scott, 9 Blackett, Patrick, 48, 135, 148
Argelander, Friedrich, 80 Blain, Andrew, 375
Arnett, David D., 178, 205, 211 Blandford, Roger D., 293
Arp, Halton C., 244, 273 Blewett, John, 139
Aston, Francis W., 47 Bludman, Sydney, 444
Atkinson, Robert d’E., 49, 61 Blumenthal, George, 432
Auger, Pierre, 134
Bohr, Niels, 11, 25, 28, 29, 31, 39, 69, 442
Aumann, Hartmut H., 161
Böhringer, Hans, 260
Axford, W. Ian, 293
Boksenberg, Alexander, 164, 381, 391
Baade, Walter, 66, 86, 92, 93, 138, 192, 204, 205, 269, Bolte, Michael, 347
270, 325, 340, 345, 346 Bolton, John, 138, 269
Bahcall, John N., 180, 181, 182, Bolton, Thomas, 199
183, 211 Bolyai, János, 101
Bahcall, Neta A., 359 Bond, Richard, 407
Baker, James, 217 Bondi, Hermann, 113, 323, 324, 325, 335
Baker, Norman, 191 Boss, Benjamin, 89
Balbus, Steven, 278 Bothe, Walther, 131, 133
Baldwin, John E., 168 Bowen, Ira S., 19, 216
Balmer, Johann Jakob, 10, 11, 17, 20 Bowyer, Stuart, 224
Balzano, Vicki, 374 Boyle, Brian, 370, 372
Bardeen, James, 278, 408, 411, 418 Boyle, Willard S., 164
Barkla, Charles, 130 Braccesi, Alessandro, 369
Barnard, Edward, 216 Bracewell, Ronald, 140
Barrett, Alan, 219 Bradley, James, 8

510
Name index 511

Bradt, Helmut, 238, 239 Curl, Robert, 241

Brahe, Tycho, 3, 16, 129, 204 Curtis, Heber D., 84, 85, 86, 269, 274
Branch, David, 345
Braun, Werner von, 142 Daguerre, Louis-Jacques-Mandé, 9
Brodie, Jean, 250 Darwin, Charles, 19
Bruzual, Gustavo, 363, 385, 386 das Gupta, Mrinal Kumar, 270, 271
Bunsen, Robert, 7 Davidson, William, 367
Burbidge, E. Margaret, 177, 248, 274, 381 Davies, Roger L., 253
Burbidge, Geoffrey R., 177, 197, 211, 248, 270, 273, Davis, Leverett, 223
274, 381, 391 Davis, Marc, 247, 408, 413
Burstein, David, 360 Davis, Raymond, 179, 180, 181
Butcher, Harvey, 350 de la Rue, Warren, 10
Butler, Clifford, 135 de Sitter, Willem, 105, 107, 108, 109, 110, 113,
116
Calver, George, 13 de Vaucouleurs, Antoinette, 96, 244
Cameron, Alistair G. W., 177, 178, 179 de Vaucouleurs, Gérard, 244, 246, 255, 344
Cameron, Julia Margaret, 9 Dekel, Avishai, 360, 411
Campbell, William, 80, 89 Descartes, René, 77
Cannon, Annie Jump, 22, 25, 26, 27, 34, 82 Deubner, Franz-Ludwig, 183
Capaccioli, Massimo, 253 Dicke, Robert, 321, 329, 330, 331, 336, 439
Carlstrom, John, 263 Dingle, Herbert, 324
Carroll, Sean, 441 Dirac, Paul A. M., 135, 323, 324
Carswell, Robert, 345, 346 Djorgovski, George, 247
Carter, Brandon, 276, 305, 443 Dodds, Stephen, 412
Cashman, Robert, 156 Dollond, John, 8
Cavaliere, Alfonso, 268 Dombrovskii, Viktor A., 269
Cavendish, Henry, 120 Donati, Giovanni Battista, 18
Cayley, Arthur, 101 Doroshkevich, Andrei, 330, 402
Chaboyer, Brian, 347 Draine, Bruce, 222
Chadwick, James, 48, 66, 131 Draper, Anna Palmer, 21
Chambers, Kenneth, 298, 351 Draper, Henry, 10, 11, 19, 21, 22, 23
Chandrasekhar, Subrahmanyan, 31, 43, 47, 49, Draper, John, 11, 12
51, 62, 64, 65, 69, 73, 188, 223, 265, 278, Dressler, Alan, 247
319, 320 Dreyer, John, 79
Charlot, Stefane, 363 Dunlop, James, 367, 368
Christensen-Dalsgaard, Jörgen, 184 Dyson, Frank, 104
Christy, Robert, 189, 191
Clausius, Rudolf, 30 Eales, Steven, 352
Clayton, Donald D., 178, 211 Eastman, Ronald, 345
Clifford, William Kingdon, 101 Eddington, Arthur Stanley, 39, 42, 43, 44, 45, 46, 47,
Coblentz, William, 156 48, 49, 51, 56, 60, 61, 63, 65, 69, 73, 80, 104,
Cockcroft, John, 48 105, 116, 117, 118, 188, 191, 216, 224, 307, 323,
Code, Arthur, 93 324
Cohen, Marshall, 294 Efstathiou, George, 410, 411, 418, 419, 435
Coles, Peter, 435 Eggert, John, 40, 42
Colgate, Stirling, 206 Einasto, Jaan, 256, 266
Common, Andrew A., 13 Einstein, Albert, 16, 29, 88, 101, 102, 103, 104, 105,
Comstock, George, 35 106, 107, 108, 116, 118, 121, 275
Copernicus, Nicholas, 8 Ellis, George F. R., 276
Corwin, Harold, 254 Ellis, Richard, 350, 362, 378, 380, 391
Cowan, Clyde, 179 Elsasser, Walter, 177
Cowan, John, 348 Emden, Robert, 32, 32
Cowie, Lennox, 378, 379, 384, 385 Epstein, Isadore, 179
Cowling, Thomas, 47 Erhenfest, Paul, 105
Cox, Arthur, 189 Euclid of Alexandria, 100
Cox, Donald, 225 Ewen, Harold, 218
Cox, John P., 191, 211
Critchfield, Charles L., 50 Faber, Sandra, 246, 305
Crommelin, Andrew, 104, 105 Fabian, Andrew C., 261, 262, 288
Crookes, William, 155 Falk, Sydney, 205
Crossley, Edward, 13 Fall, Michael, 383, 389, 390
Curie, Pierre, 131 Fanaroff, Bernard, 291
Curie-Sklodowska, Marie, 131 Faulkner, John, 68, 211, 329
512 Name index

Felten, James E., 246, 247 Habing, Harm, 224

Fermi, Enrico, 48, 179, 293, 321, 322 Hale, George Ellery, 14, 15, 22, 45, 88, 125, 170
Field, George B., 224 Hall, John, 223
Fisher, Richard, 247, 344 Halley, Edmund, 79
Fizeau, Armand, 155 Halm, Joseph, 39
Fleming, Williamina P., 22, 23, 26, 27, 28, 63 Hamilton, Andrew, 411
Follin, James, 323, 329 Hardy, Eduardo, 265, 349
Ford, Holland, 283 Harms, Richard, 284
Ford, Vincent, 223 Harrison, Edward, 401
Foucault, Léon, 6, 13, 155 Harrison, Kent, 195
Fowler, Alfred, 25, 28 Hartmann, Johannes, 91
Fowler, Ralph H., 40, 64, 69 Hasinger, Günther, 372
Fowler, William A., 176, 177, 179, 211, 273, 279, 291, Hausman, Marc, 265
321, 332, 334, 336 Hawking, Stephen W., 276, 277, 305, 447, 450, 451,
Fox Talbot, William Henry, 9 452
Frail, Dale, 232 Hawley, John, 278
Fraunhofer, Joseph, 4, 5, 6, 8, 16, 18 Hayakawa, Satio, 197
Frazier, Edward, 183 Hayashi, Chushiro, 226, 227, 242, 322
Freedman, Wendy, 344, 346 Hazard, Cyril, 272
Freeman, Kenneth C., 249 Heavens, Alan F., 411
Frenk, Carlos, 408 Helmholtz, Hermann von, 30
Friedman, Alexander Alexandrovich, 107, 108, 113, Henderson, Thomas, 8
121 Henrich, Louis, 320
Friedman, Herbert, 142, 144 Henry, Joseph, 31
Frost, Edwin, 91 Henyey, Louis, 136, 189, 225
Herbig, George, 225
Gagarin, Yuri, 142 Herlofson, Nicolai, 139
Galbraith, William, 149 Herman, Robert, 320, 321, 322, 323, 329, 394, 395
Galilei, Galileo, 3, 16 Hernquist, Lars, 432
Galletta, Giuseppe, 253 Herschel, Caroline, 79
Gamow, George, 48, 49, 68, 116, 319, 320, 321, 329, Herschel, John, 9, 10, 16, 79, 155, 163
335, 336 Herschel, William, 77, 78, 79, 80, 153, 155
Garnavich, Peter, 353 Hertz, Heinrich, 163
Gauss, Carl Friedrich, 100, 101, 120 Hertzsprung, Ejnar, 24, 34, 35, 36, 39, 45, 82, 96
Geiger, Hans, 133 Hess, Victor, 66, 132, 133
Geiser, Karl Friedrich, 103 Hevesy, George, 29
Geller, Margaret, 256, 258 Hewish, Antony, 67, 193, 211, 326
Genzel, Reinhard, 287 Hey, James, 137, 138, 171
Gerling, Christian Ludwig, 100 Hill, Henry, 184
Gershtein, Semion, 405 Hillas, Michael, 236, 241
Ghez, Andrea, 286 Hiltner, William, 223
Giacconi, Riccardo, 142, 143, 144, 372 Hinks, Arthur, 35
Gibbons, Gary, 447 Hoag, Arthur, 371
Ginzburg, Vitali Lazarevich, 139, 195 Hooker, John D., 14, 125
Gold, Thomas, 193, 195, 273, 304, 324, 325, Houtermans, Fritz, 49
335 Hoyle, Fred, 49, 50, 55, 176, 177, 188, 189, 211, 273,
Goldreich, Peter, 195 279, 291, 321, 324, 325, 328, 329, 331, 332, 334,
Goldsmith, Donald, 224 335, 336
Goobar, Ariel, 352 Hu, Wayne, 418, 419, 435, 436
Gott, J. Richard, 257 Hubble, Edwin P., 15, 19, 66, 87, 88, 96, 109, 110,
Gough, Douglas O., 183, 184 111, 112, 113, 116, 118, 244, 255, 323, 340, 358,
Graham Smith, Francis, 138 361, 365
Green, Richard, 370 Huchra, John, 250, 256, 258
Greenstein, Jesse, 223, 274 Hufford, Charles, 91, 156
Gregory, James, 7 Huggins, William, 10, 11, 18, 19, 20, 21, 27, 79, 80,
Gribbin, John, 443 155, 216
Grossmann, Marcel, 104 Hulse, Russell, 201, 212, 214
Grubb, Howard, 13 Hulst, Hendrik van de, 218
Gull, Stephen F., 294 Humason, Milton, 110, 326, 340
Gunn, James E., 169, 195, 293, 359, 360, 362, 379, Huyghens, Christian, 4
406
Guseynov, Oktay, 199 Iben, Icko, 189
Guth, Alan, 405, 444, 445, 446, 447, 451 Iglesias, Carlos, 192
Name index 513

Illingworth, Garth, 253, 283 Krymsky, Germogen, 293

Irwin, Michael, 370 Krzeminski, Wojciech, 197
Isaak, George, 184 Kuiper, Gerard, 156
Israel, Werner, 305 Kundic, Tomislav, 346
Ivanenko, Dmitri, 139 Kurt, Vladimir, 394, 417
Ivison, Robert, 375 Kwan, John, 345

Jackson, Robert, 246 Labeyrie, Antoine, 167

Jakobsen, Peter, 428 Lamb, Horace, 184
Jansky, Karl, 136, 137, 139, 269 Lambert, Johann, 77, 95, 100
Jeans, James, 42, 43, 44, 45, 51, 228, 393 Lanczos, Cornelius, 107, 108, 113
Jefferts, Keith, 220 Landau, Lev D., 66, 68, 434
Jelley, John, 149 Landé, Alfred, 41
Jennison, Roger, 270, 271 Lane, J. Homer, 31, 32, 51, 53
Johnson, Harold, 29, 157 Langley, Samuel Pierpoint, 155
Joliot, Frederic, 131 Langmuir, Irving, 433
Joliot-Curie, Irene, 131 Larson, Richard, 228
Jones, Bernard, 417 Laue, Max von, 130
Joule, James Prescott, 30 Le Roux, Émile, 330
Joy, Alfred, 89, 91, 92, 225 Le Verrier, Urbain, 104
Julian, William, 195 Leavitt, Henrietta, 82, 83, 344
Lebedev, Peyotr, 156
Kaiser, Christian, 293 Leer, Egil, 293
Kaiser, Nicholas, 410, 411 Leger, Alain, 221
Kant, Immanuel, 77, 95 Leibacher, John, 183
Kapahi, Vijay, 357 Leighton, Robert, 158, 183
Kapteyn, Jacobus, 34, 80, 81, 83, 84, 86, 91 Lemaı̂tre, Georges, 108, 110, 113, 118, 320, 325, 393,
Kaufmann, Walter, 131 437
Keeler, James E., 13, 14, 17, 22, 80 Lick, James, 125
Keenan, Philip C., 29, 37, 136 Liddle, Andrew, 447, 448, 452
Kellermann, Kenneth, 357 Liebig, Justus von, 13
Kellman, Edith, 29, 37 Lifshitz, Evgenii, 393, 437
Kelvin, Lord, see Thomson, William Lightman, Alan, 359
Kennicutt, Robert, 382 Lilly, Simon, 351, 352, 384, 385
Kepler, Johannes, 3, 16 Lindblad, Bertil, 89
Kerr, Frank, 218 Linde, Andrei, 444, 446
Kerr, Roy, 275 Lineweaver, Charles, 451
Khachikian, Eduard, 279 Lobachevsky, Nicolai Ivanovich, 101
Kiang, Tao, 255 Lockyer, Norman, 24, 25, 28, 33, 51
Kibble, Thomas, 446, 446 Longair, Malcolm S., 293, 298, 335, 351, 352, 367,
Kiepenheuer, Karl-Otto, 139 442
Kihara, Taro, 255 Lovelace, Richard, 297
Kippenhahn, Rudolph, 32, 58, 62, 68, 189, 190, 191, Lovell, Bernard, 129, 138
211, 226, 242, 243 Low, Frank, 157, 160, 161, 226
Kirchhoff, Gustav, 4, 7, 18 Lucchin, Francesco, 435
Kirshner, Robert, 205, 206, 345 Lund, Niels, 238
Kirzhnits, David, 195 Lundmark, Knut, 66, 86
Klebesadel, Ray, 300 Lynden-Bell, Donald, 263, 278
Klein, Oskar, 398 Lynds, Roger, 381
Kleinmann, Douglas, 226 Lyth, David, 447, 448
Kobold, Hermann, 80 Lyttleton, Raymond, 50, 55, 188
Kohlschütter, Arnold, 36, 89 Lyubimov, Valentin, 405
Kolb, Edward (Rocky) W., 449, 451
Kolhörster, Werner, 132, 133 Maanen, Adriaan van, 85
Kompaneets, Alexander S., 396 McCarthy, Patrick, 298, 351
Koo, David, 370 McCray, Richard, 209
Kormendy, John, 265, 290, 305 McCrea, William, 41, 43, 69, 115, 211, 325, 442
Kramers, Hendrik, 42, 47 Mack, John, 224
Kraushaar, William, 224 McKee, Chester, 206
Kron, Gerald, 163 McKee, Christopher, 225
Kron, Richard, 370 McKellar, Andrew, 331
Kroto, Harold W., 220, 221, 241 McLeod, John, 274
Kruskal, Martin, 275 Madau, Piero, 387, 390
514 Name index

Maddox, Richard L., 11 Oemler, Augustus, 350

Maeder, André, 347 Ohm, Edward, 330
Margon, Bruce, 198 Oke, J. Beverley, 179, 205
Markarian, Beniamin, 275, 279 Olowin, Ronald, 254
Marx, Györgi (George), 405 Olson, Roy, 300
Matsuoka, Masaru, 197 Omnes, Roland, 398
Matthews, Thomas A., 272 Oort, Jan, 89, 90, 97, 98, 99, 217, 218, 241, 269, 359
Matthewson, Donald, 223 Öpik, Ernst, 61, 62, 86, 176, 188
Maury, Antonia, 22, 23, 24, 25, 26, 34, 35 Oppenheimer, J. Robert, 66, 68
Mayall, Nicholas, 326, 340 Osaki, Yoji, 183
Mayer, Julius, 30, 50 Osmer, Patrick, 371
Mayor, Michel, 232 Osterbrock, Donald E., 93, 95, 305, 328
Mazzarella, Joseph, 374 Ostriker, Jeremiah P., 195, 198, 225, 250, 265, 293,
Menzel, Donald, 42, 216, 217 430
Messier, Charles, 79, 95 Ostwald, Wilhelm, 34
Mestel, Leon, 44, 47, 51, 61, 65, 68, 69 Oxley, Charles, 156
Mészáros, Peter, 301, 304, 402
Michell, John, 8, 16, 120, 277 Pacini, Franco, 192, 193, 195
Michelson, Albert A., 45, 46, 156, Padmanabhan, Thanu, 435
166, 167 Pagel, Bernard, 389
Migdal, Arkadii, 195 Panagia, Nino, 345
Mikheyev, Stanislav, 182 Pankey, Titus, 206
Miley, George, 351, 357 Papaloizou, John, 279
Miller, Glenn, 231, 232 Parsons, William, 12, 13, 21
Miller, Joseph, 299 Partridge, Bruce, 405, 418
Miller, William, 18, 19, 27, 79 Patchett, Bruce, 345
Millikan, Robert, 132, 135 Pauli, Wolfgang, 64, 69
Mills, Bernard, 327, 328 Pawsey, Joseph, 138
Mills, D. O., 80 Payne, Cecilia, 27, 40, 41, 319
Milne, Edward A., 25, 40, 115, 323, 324 Payne-Gaposchkin, Cecilia, see Payne, Cecilia
Minkowski, Rudolph, 138, 205, 269, 270 Peacock, John A., 361, 367, 411, 412, 451
Minnaert, Marcel, 41, 42 Pearce, Joseph, 91
Mirabel, Felix, 296 Pease, Francis, 45, 86, 156
Mitchell, John, 78 Peebles, P. James E., 250, 255, 256, 257, 331, 334,
Miyoshi, Makoto, 285 393, 394, 402, 403, 407, 416, 417, 419, 439
Møller, Palle, 428 Pei, Yichuan, 389, 390
Monck, William, 34 Penrose, Roger, 276, 277, 451
Moore, Charlotte, 41 Penzias, Arno A., 220, 321, 330, 394
Morgan, William W., 29, 37, 93, 95, Perlmutter, Saul, 352, 353
265, 274 Peters, Bernard, 238, 239
Mould, Jeremy, 248, 344 Peterson, Bradley, 283, 305
Müller, Walther, 133 Peterson, Bruce, 379
Mulders, Gerard, 41, 42 Pethick, Christopher, 196
Muller, Lex, 218 Pettini, Max, 383
Murdin, Paul, 199 Pettit, Edison, 156
Mushotzky, Richard, 284 Pfund, Herman, 156
Myers, Steven, 345 Phillips, Mark, 206
Pickering, Edward C., 21, 22, 23, 24, 25, 26, 27, 28,
Neddermeyer, Seth, 135 35, 63, 129
Nemiroff, Robert, 306 Pines, David, 195, 196
Nernst, Walther, 40 Pius IX, Pope, 20
Neugebauer, Gerry, 158, 226, Plücker, Julius, 7
227, 228 Plaskett, John, 91
Neville, Ann, 141, 367 Planck, Max, 7, 16, 34, 434
Newman, Ezra (Ted), 276 Pomeranchuk, Isaak, 139
Newton, Isaac, 3, 8, 16, 100, 120 Pouillet, Claude-Servais, 154
Neyman, Jerzy, 255 Powell, Cecil, 136
Nicholson, Seth, 156 Prendergast, Kevin, 197, 248
Niels Lund, 238 Press, William, 206, 403, 404, 441
Novikov, Igor Davidovich, 283, 330, 336, 393, 394 Price, Stephan, 158
Pringle, James, 279, 305
O’Dell, C. Robert, 329 Pskovskii, Yuri, 206
Occhialini, Giuseppe, 135 Ptolemy (Claudius Ptolemeus), 80
Name index 515

Puget, Jean-Loup, 221 Schmidt, Bernhard, 168

Purcell, Edward, 218 Schmidt, Brian, 345
Schmidt, Maarten, 169, 272, 273, 274, 367, 370, 371
Queloz, Didier, 232 Schneider, Donald P., 169
Schönberg, Mario, 62, 188
Radhakrishnan, Venkataraman, 196 Schott, George A., 139
Ramsay, William, 24 Schramm, David, 349
Rawlings, Steven, 352 Schuster, Arthur, 39
Rayet, Georges A. P., 23, 24 Schutz, Bernard, 212, 213, 214
Razin, V., 270 Schwarzschild, Karl, 35, 35, 39, 53, 80, 101, 275, 305
Reber, Grote, 136, 137, 139, 218, 269 Schwarzschild, Martin, 61, 176, 188, 189, 253, 347
Rees, Martin J., 279, 291, 293, 295, 297, 301, 302, Schwinger, Julian, 139
411, 430, 435, 443 Sciama, Dennis, 435
Reines, Frederick, 179 Scott, Elizabeth, 255
Rhijn, Pieter van, 81 Seares, Frederick, 82
Richstone, Douglas O., 265, 290, 305 Seaton, Michael, 192
Riemann, Bernhard, 101 Secchi, Angelo, 18, 20, 21, 23, 35
Riess, Adam, 206 Seebeck, Thomas, 155
Riley, Julia M., 291 Sellgren, Kristen 221
Rindler, Wolfgang, 313, 435 Severny, Andrei, 184
Ritchey, George W., 15, 84 Seyfert, Carl, 274
Ritter, Augustus, 32, 33 Shakura, Nikolai, 278
Roberts, James, 140 Shane, Donald, 255, 256, 257
Robertson, Howard, 110, 113, 114 Shapiro, Irwin, 120
Rochester, George, 135 Shapiro, Maurice M., 239
Rodrigues, Luis, 296 Shapley, Harlow, 15, 27, 38, 40, 82, 83, 84, 85, 86, 87,
Röntgen, Wilhelm, 130 89, 90
Röttgering, Huub J. A., 298 Sharpless, Stuart, 93, 95
Rogers, Forrest, 192 Shellard, Paul, 447
Rogerson, John B., 328, 333 Shklovsky, Iosef, 196, 197, 269
Rømer, Ole, 3 Shmaonov, Tigran, 330, 336
Rosse, Lord, see Parsons, William Shu, Frank, 228, 230
Rosseland, Svein, 47, 58 Silk, Joseph, 399, 430
Rougoor, G. Wim, 241 Simpson, John, 237
Rowan-Robinson, Michael, 367 Skadron, George, 293
Rubin, Vera, 248 Skobeltsyn, Dmitri, 132, 134, 135
Ruderman, Malvin, 196, 444 Slee, Bruce, 138, 327
Russell, Henry Norris, 35, 36, 37, 38, 39, 41, 42, 45, Slipher, Vesto, 91, 109, 110
61, 63, 216, 319 Smail, Ian, 375
Rutherford, Ernest, 31, 47, 48, 131, 132, 347 Smalley, Richard E., 221, 241
Rutherfurd, Lewis Morris, 11, 18, 20 Smirnov, Alexei, 182
Ryle, Martin, 138, 141, 171, 211, 293, 326, 328, 335, Smith, Barham, 225
367 Smith, George E., 164
Smith, Malcolm G., 371
Saccheri, Girolamo, 100 Smoot, George, 415
Sachs, Rainer K., 418 Sneden, Christopher, 348
Saha, Megh Nad, 25, 40 Snell, Ronald, 229
Sakharov, Andrei, 400, 448 Snyder, Hartland, 66
Salpeter, Edwin E., 176, 225, 230, 231, 232, 278 Soldner, Johann von, 103, 120
Sampson, Ralph, 39 Spinrad, Hyron, 350, 362
Sandage, Allan R., 176, 188, 244, 265, 272, 274, 326, Spite, François, 333
340, 341, 342, 344, 346, 347, 349, 350, 354, 361, Spite, Monique, 333
362, 368 Spitzer, Lyman, 69, 150, 151, 223, 224
Sargent, Wallace L. W., 283, 284, 291 Stanley, Gordon, 138, 269
Saslaw, William, 255 Stebbins, Joel, 91, 156
Saunders, Frederick, 41 Steidel, Charles, 385, 386, 388
Savedoff, Malcolm, 223 Stein, Robert, 183
Scalo, John, 231, 232 Steinhardt, Paul, 446
Schade, David, 379 Steinheil, Karl, 13
Schalén, Carl, 222 Stobie, Robert S., 191
Schechter, Paul, 246, 359, 403, 404 Stockton, Alan, 381
Scheiner, Julius, 80, 86 Stone, Edward, 155
Scheuer, Peter A. G., 140, 293, 328, 379 Stoner, Edmund, 64
516 Name index

Strömberg, Gustav, 89, 89 Walraven, Théodore, 269

Strömgren, Bengt, 36, 49, 217, 224 Walsh, Dennis, 345, 346
Strong, Ian, 300 Walton, Ernest, 48
Struve, Otto, 40 Wandel, Amri, 284
Struve, Wilhelm, 6, 8 Warren, Stephen, 370
Suess, Hans, 176, 177 Waterston, John James, 50
Sugiyama, Naoshi, 419, 435, 436 Waymann, Ray, 345, 346
Sunyaev, Rashid Alievich, 263, 278, 394, 396, 400, Webber, William, 237
402, 405, 417, 434 Weber, Joseph, 203
Sweet, Peter, 61 Webster, Louise, 199
Szalay, Alexander, 405 Weedman, Daniel, 279, 374
Weigert, Alfred, 32, 58, 62, 68, 189, 190, 211, 226,
Tammann, Gustav, 344 242, 243
Tayler, Roger J., 51, 57, 60, 329, 331, 332, 336 Weinberg, Steven, 444
Taylor, Joseph, 201, 212, 214 Weinreb, Sander, 219
Tegmark, Max, 423, 425, 426, 427 Weiss, Rainer, 204
Thomson, J. J., 212, 215 Weizsäcker, Carl von, 49
Thomson, William, 30, 31, 50 Wesselink, Adriaan, 345
Thorne, Kip, 199, 204, 212, 276 Westerhout, Gart, 218, 270
Tinsley, Beatrice, 389 Weyl, Hermann, 113
Tipler, Frank, 443 Weymann, Raymond, 396
Tolman, Richard, 393, 437 Whaling, Ward, 176
Tonks, Lewi, 433 Wheeler, John A., 195, 277, 305, 444
Toomre, Alar, 245 White, Simon, 360
Toomre, Juri, 245 Whitford, Albert, 91, 93, 156, 157, 163
Totsuji, Hiroo, 255 Wilkinson, David, 416
Tousey, Richard, 142 Will, Clifford, 201
Tremaine, Scott, 252, 265, 406 Wilson, Charles T. R., 132
Trimble, Virginia, 86, 199, 266 Wilson, Robert W., 220, 321, 330, 394
Trümper, Joachim, 199 Wirtanen, Carl, 255, 256, 257
Trumpler, Robert, 91, 156 Wirtz, Carl Wilhelm, 109
Tully, Brent, 247, 344 Wolf, Charles J. F., 23
Turkevich, Anthony, 321, 322 Wolf, Max, 216
Turner, Edwin, 441 Wolfe, Arthur, 381, 418
Turner, Michael S., 449, 451 Wolfenstein, Lincoln, 182
Tycho Brahe, see Brahe, Tycho Wollaston, William, 4, 16
Tyson, Anthony, 355 Wolszczan, Alex, 232
Woltjer, Lodewijk, 391
Ulrich, Roger, 180, 181, 183 Wright, Thomas, 77, 95
Unsöld, Albrecht, 41 Wyse, Rosemary, 417, 430
Urey, Harold, 48, 176, 177
Yerkes, Charles T., 125
VandenBerg, Donald, 348 York, Donald, 333
Vashakidze, Mikhail, 269 Young, Peter, 283, 284
Velikhov, E. P., 278 Young, Thomas, 4, 154
Villard, Paul, 131 Yu, J., 419
Vogel, Hermann C., 20, 35, 80 Yukawa, Hideki, 135, 136
Vogel, Hermann W., 20
Volkoff, George, 68 Zanstra, Herman, 216
Vorontsov-Velyaminov, Boris, 244 Zeldovich, Yakov Borisovich, 106, 199, 263, 278, 283,
329, 334, 336, 393, 394, 396, 398, 399, 400, 401,
Wagoner, Robert V., 211, 321, 332, 334, 336 402, 403, 405, 416, 417, 434, 441, 444
Walker, Arthur, 113, 114 Zhevakin, S., 191
Walker, Merle, 225 Zwicky, Fritz, 66, 67, 86, 116, 117, 121, 168, 192,
Walker, Russell, 158 204, 205, 212, 246, 270
Wall, Jasper, 367 Zworkin, Vladimir, 163
Object index

0957+561 (double quasar), 345, 346 γ Draconis, 8

3C 9, 273, 379 o Eridani, 36, 63

3C 31, 291, 292 40 Eridani B, see o Eridani
3C 47, 273
3C 48, 272, 273 Galactic centre, 83, 90, 97, 99, 147, 148, 158, 159,
3C 66B, 291, 292 163, 203, 218, 220, 286, 287
3C 83.1B, 262 Gleise 229, 235
3C 147, 273 Gleise 229B, 235
3C 191, 381 GRB 030329, 303
3C 196, 272 GRB 970228, 302
3C 273, 144, 149, 272, 273, 280, 283, GRB 990510, 303
294, 295 Great Attractor, 360
3C 286, 272 GRS 1915+105, 296
3C 295, 138, 272
HD 209458, 233, 234
A0620-00, 200 Hercules X-1, 197, 198, 199
Abell 478, 261, 262, 345 α Herculis, 156
Abell 2142, 345 Hyades star cluster, 35, 36, 104
Abell 2256, 345
Aldebaran, 18, 79 IC 4182, 204
Andromeda Nebula, see M31 Io, 3
Arcturus, 34, 39, 155 IRC +10216, 158
IRC +10216, 221
Barnard’s star, 232
Becklin–Neugebauer object, 159, 226, Jupiter, 3, 4, 7, 8, 235
227, 228
Betelgeuse, 18, 20, 45, 46, 80 Kepler’s supernova remnant, 146
Kleinmann–Low object, 226
Capella, 168
Cassiopeia A, 138, 141, 146, 219, 270, 271, 293 L1551, 229
α Centauri (α Cen), 8 BL Lacertae (BL Lac), 275
β Centauri (β Cen), 151 Large Magellanic Cloud, 207, 209, 251, 345
Centaurus A, 138 LMC X-1, 200
Centaurus X-3, 197 LMC X-3, 200
δ Cephei, 82 α Lyrae, 8, 10, 16, 19, 24, 167
Coma cluster of galaxies, 116, 257, 259, 260,
359, 411 M1, 79, 138, 144, 146, 147, 149, 150, 193, 195, 269
CP1919, see PSR 1919+21 M15, 329
Crab Nebula, see M1 M31, 14, 66, 79, 84, 85, 86, 87, 88, 92, 325, 326, 340
Crab Nebula pulsar, 193, 194, 195 M32, 92
CS 22892-052, 348, 349 M33, 85, 86, 87
61 Cygni, 8, 101 M42, 12, 79
NML Cygni, 158 M51, 12, 14, 85
Cygnus A, 138, 140, 141, 270, 271, 291, 293, 326 M81, 85
16 Cygnus A, 234 M82, 274
16 Cygnus B, 234 M87, 12, 138, 144, 250, 269, 274, 283, 284, 294, 464
Cygnus X-1, 197, 199 M101, 85

517
518 Object index

M106, 285 Pollux, 155

Magellanic Clouds, 82, 87, 96, 163, 218, 224, 250, Procyon, 156
326, 340 PSR 1257+12, 232
Markarian 421, 150 PSR 1913+16, 170, 201, 202, 203, 212, 214
Markarian 501, 150 PSR 1919+21, 193, 194
Mars, 34, 121, 184 ζ Puppis, 24
Mercury, 104
MGC-6-30-15, 287, 289 Q1422+2309, 382
Moon, 3, 9, 129, 143, 153, 373
Regulus, 155
NGC 205, 92 Rigel, 156
NGC 681, 248
NGC 1068, 274, 299 Sagittarius A, 158, 159
NGC 1265, 262 Sagittarius A∗ (Sgr A∗ ), 287, 288
NGC 2264, 225 Sagittarius B2, 220
NGC 4151, 274, 280, 281 Sanduleak −69 202, 207
NGC 4157, 204 Saturn, 16
NGC 4258, 285, 286, 287 Sco X-1, 143, 144, 197
NGC 4321, 340 Sirius, 5, 8, 10, 20, 80, 155
NGC 4486, 284 Sirius A, 63, 80
NGC 5128, 138 Sirius B, 63, 105
NGC 5548, 281, 282 Small Magellanic Cloud, 82, 83
NGC 5866, 248 SN 1987A, 170, 205, 207, 208, 209, 210, 212, 345
NGC 6946, 84 SN 1987ap
NGC 7083, 222 SN 1997ap, 353
NGC 7619, 110 SN 1997ck, 353
1885 nova in Andromeda Nebula, 66, SNR 1006, 146
85, 86 Supernova 1987A, see SN 1987A

ρ Ophiuchus cluster, 235 Taurus molecular cloud, 229

Ophiuchus molecular cloud, 229 47 Tucanae, 347, 348
Orion Nebula, 13, 79, 158, 159, 226, 227 Tycho Brahe’s supernova, 204
Orion star cluster, 235
Orion stars, 24 Vega, see α Lyrae
δ Orionis, 91 Vela pulsar, 193, 196
Vela supernova remnant, 147
51 Pegasus, 233 Venus, 5
Perseus cluster of galaxies, 144, 260, 261, 262 Virgo A, 138
Pleiades star cluster, 35, 78, 235 Virgo cluster of galaxies, 110, 117, 284, 340, 343, 344
Polaris, 22 Virgo supercluster of galaxies, 360
Subject index

The principal references to broad topics of major significance are highlighted in bold type. Topics described in
more detail in the explanatory supplements to the chapters are also printed in bold type. References to the books
mentioned in the text, notes and explanatory supplements are shown in italic type with the author’s name in
brackets. For books with three or more authors, et al. is used.

Abel integral, 268 acetylenic chain molecules in the interstellar medium,

Abell Catalogue of Clusters of Galaxies, 254–5 220–1
aberration, relativistic, 291, 295, 300 Active Galactic Nuclei (Blandford et al.), 391
aberration of light of stars, Bradley’s measurements, 8, Active Galactic Nuclei (Robson), 306
16 active galactic nuclei, 200, 271–5, 305
absolute magnitude, 35 discovery of different classes of, 271–5
absorption line systems in the spectra of quasars, general relativity and models for, 275–9
381–3 ionisation structure of gas clouds about, 281–3
identification of, 381 Markarian galaxies, 275
abundance determinations for stellar atmospheres, masses of black holes in, 283–90
40–2, 319 models for, 277–9
hydrogen and helium problem, 41, 44, 49 non-thermal phenomena in, 290–300
abundance of elements in Lyman-α absorbers, reverberation mapping of, 280–3, 305
383–4 Seyfert galaxies as, 274
acceleration of electrons, protons and nuclei spectral properties of accretion disc in, 274–5,
Fermi mechanism, 293–4 279–83, 305
first-order, 293–4 spectroscopy of, 279–83
limits to energies produced in, 294 unified models for, 299–300
and formation of power-law energy spectrum, 294 variability of luminosities of, 284–5, 290–1
in pulsar magnetospheres, 195, 293 X-ray fluorescence line from, 288–9
in strong shock waves, 210, 293–4 active galaxies, evolution with cosmic epoch, 365–74
accretion in high-energy astrophysical systems, adaptive optics, 166
197–9, 212, 277–9, 306–7 adiabatic and isothermal scenarios for galaxy
Eddington limiting luminosity for, 307 formation, 402–5
efficiency of, as an energy source, 306–7 adiabatic expansion of photon gas, 320–1, 394
for Kerr and Schwarzschild black holes, 307 adiabatic picture/scenario of structure formation,
for neutron stars, 307 399–401, 402–3, 407
maximum energy available due to, 198, 277–8 acoustic or Sakharov oscillations on last scattering
accretion discs about neutron stars and black holes, surface, 400–1
43, 197–8, 278–9, 305 confrontation with observations, 404–5
formation of, 278–9 constrained by upper limits to fluctuations in the
role of viscous forces in, 278 cosmic microwave background radiation, 405
thick, instability of, 279 fragmentation of large-scale structures at late
as origin of collimation of relativistic jets, 197 epochs, 402
X-ray fluorescence line from, 288–9 late formation of structures, 402–3
accretion discs and planet formation, 233–4 with neutrinos with finite rest mass, 405–6
accretion phase of pre-main sequence stars, 226–8 ‘pancake’ model, 402–3
accretion shocks during, 228–30 Silk damping in, 402
Accretion Power in Astrophysics (Frank et al.), 212, sound waves during the pre-recombination era,
305 400–1
accretion onto white dwarfs in Type 1a supernovae, stability diagram for, 399–401
206 as a ‘top-down’ process, 403

519
520 Subject index

Advanced X-ray Astronomy Facility of NASA Apache Point Observatory, 169

(AXAF), 145–6; see also Chandra X-ray aperture synthesis, 140–2, 166–8, 171–2
Observatory for optical–infrared astronomy, 166–8, 172
age of disc of Galaxy, 347 science programmes for, 168
age of the Earth, 30–1, 45, 118, 323, 326, arc spectrum, 7,16, 24
346–7 Arecibo Radio Observatory, 201
from radioactive dating, 31, 348–9 Arequipa Observatory, Peru, of the Harvard College
from stratigraphic analyses, 31 Observatory, 25, 82
ages of the stars, 63, 69–70 Argonne National Laboratory, 203
in globular clusters, 188–91, 342, 346–7 Arthur Stanley Eddington (Vibert Douglas), 51
age of Universe, 118, 323–6, 341–2, 346–9, Ariel-5 satellite, 260
425–7, 431 Army Operational Research Group, 137–8
from radioactive dating in ultra-metal-poor stars, ASCA X-ray satellite (Japan), 289
348–9 Association of Universities for Research in
timescale problem, 118, 323, 326 Astronomy (AURA), 162–3
Air Force Cambridge Research Laboratories asteroseismology, 188
(AFCRL) infrared sky survey, 158 astrobiology, 234–5
Alexander A. Friedman: The Man Who Made the astronomical community, development of, 128
Universe Expand (Tropp et al.), 121 astronomical parallaxes, 4, 7–9
alignment effect for extragalactic radio sources, astronomical spectroscopy, development of, 4–7
298–9, 351–2 Astronomy and Astrophysics Encyclopedia, The
impact upon redshift–apparent magnitude relation, (ed. Maran), xii
351–2 astronomy as ‘big science’, 129–30
shock waves associated with the, 298–9 impact of Second World War, 129
Allegheny Observatory, 156 Astronomy Through the Telescope (Learner) xii, 17
Almagest (Ptolemy), 80 Astronomy of the 20th Century (Struve and Zebergs),
‘α-particle’ nuclei, 177–8 51
α-particles, discovery and properties of, 130–1 astroparticle physics, 128, 170, 252–3
Alpher–Bethe–Gamow (αβγ ) theory of primordial astrophysical cosmology, origins of, 100–21,
nucleosynthesis, 320 319–39
aluminium-26 γ -ray line from the direction of the in the period immediately after the Second World
Galactic Centre, 148 War, 323
American Civil War, 31 since 1945, 318–452
American Science and Engineering group (AS&E), Astrophysical Journal, 22, 28, 218
142–4 founding of, 22, 28
ammonia molecule, discovery in interstellar medium, Astrophysical Quantities (Allen), 96
219 Astrophysics of Gaseous Nebulae and Active Galaxies
Analysis of Starlight, The: One Hundred and Fifty (Osterbrock), 305
Years of Astronomical Spectroscopy astrophysics of stars and galaxies since 1945,
(Hearnshaw), xii, 27–8 174–316
Anglo-Australian 3.9-metre Telescope (AAT), 163, Atacama Large Millimetre Array (ALMA), 162
372 Atlas and Catalogue of Interacting Galaxies
angular diameter distance, 345 (Vorontsov-Velyaminov), 244–5
angular diameter–redshift relation, 342, 356–8 Atlas of Galactic Neutral Hydrogen (Hartmann and
and Baade–Wesselink method, 356–7 Burton), 241
for compact double radio sources, 357–8 Atlas of Peculiar Galaxies (Arp), 244–6
for double radio sources, 357–8 Atlas of Stellar Spectra (Morgan et al.), 37–8
effect of inhomogeneities along the line of sight, atmosphere, transparency of, 126–7, 331
357, 362 Atmospheric Cherenkov Imaging Telescopes, 149–50,
evidence for evolutionary changes with cosmic 172
epoch, 357–8 Atoms, Stars and Nebulae (Aller), 241
and gravitational lenses, 356–7 Auger cosmic-ray array, 443
for inhomogeneous world models, 362 AURA and its US National Observatories
minimum in the, 357–8 (Edmondson), 172
and the Sunyaev–Zeldovich effect, 356–7 Australia Telescope National Facility (ATNF), 141
Annals of the Harvard Observatory, 21 Australia Telescope National Facility Pulsar
Annual Review of Astronomy and Astrophysics, xii Catalogue, 201
anthropic cosmological principle, 442, 443–4 Automatic Plate Measuring Machine (APM),
strong form of, 444 Cambridge, 169, 370
anti-biassing in the hot dark matter picture of structure axions as cold dark matter particles, 406–7
formation, 411
anti-homology theorem, 211 Baade–Wesselink method of distance estimation, 345,
antiparticles, 135 356
Subject index 521

applied to supernovae, 345 BL Lac (BL Lacertae) objects, 274–5, 300

using Type II supernovae, 345 properties of, 274–5
Bache 8-inch telescope, 22, 25 relativistic jets in, 300
Baksan Neutrino Observatory, 182 black holes, 65–6, 68, 73, 78, 120, 178, 196–7,
Balbus–Hawley instability, 278 199–200, 214, 275–8, 283–90, 305–7
balloon flights accretion of matter onto, 278–9, 306
for far-infrared astronomy, 160 angular momentum of, 305
for the study of cosmic rays, 132–3 in binary X-ray sources, 199–200
Balmer formula, 10–11, 17, 24–5, 28–9 discovery of, 199–200
Balmer jump, 40 masses of, 200
Balmer series of hydrogen, 10–11, 20, 40, 216, 280 relativistic motions, 296
barred spiral galaxies, 87–8 in centre of the Galaxy, 286–7
baryogenesis, 448–9 infrared flares observed from, 287
associated with very massive boson/antiboson charged, 276
decays, 449 closed trapped surface and, 276
Sakharov’s rules for, 448 electrodynamics of, 297
baryon–antibaryon pair production, 440 and formation of jets, 297
baryon asymmetry problem, 397–8, 440, 442, 448–9 energy release by, 277–8
creation under non-equilibrium conditions, 448–9 as energy sources in high-energy astrophysics,
baryon density parameter from primordial synthesis of 306–7
deuterium, 332–5, 361 event horizon of, 276–7, 305, 306
baryon number, violation of, 448 history of, 305
generic feature of grand unified theories, 448 Kerr, 275–8
baryon-symmetric cosmologies, 398, 440 maximum angular momentum of, 277
baryonic dark matter, 250 rotational energy of, 277
Bautz–Morgan classification of clusters of galaxies, last stable circular orbit about, 277–8, 289, 306
265–6, 349 magnetic fields and, 276
‘Be Scientific with Ol’ Doc Dabble’ comic strip, 67 masses of, in active galactic nuclei, 280, 283–90
Bell Telephone Laboratory, Holmdel, New Jersey, correlation with masses of bulges/spheroids, 290
136–7, 330 estimated from kinematics of ionised gas clouds,
Bell Telephone Laboratory, Murray Hill, New Jersey, 283–5
164 estimated from X-ray continuum time variability,
bending of light rays by massive bodies, 103–5 284–5
Henry Cavendish’s analysis, 120 from orbits of stars in nucleus of the Galaxy,
‘Newtonian’ result for, 103–4, 120 286–7
Benediktbraun glassworks, 4 from water vapour maser lines in M106, 285–6
Bentley–Newton correspondence, 100 maximally rotating, 305
BeppoSAX X-ray telescope (Italian–Dutch project), membrane paradigm, 276
302 naming of, 277
Berkeley–Illinois–Maryland Association (BIMA) ‘no-hair’ theorem for, 276, 305
millimetre array, 264 observed properties of, in binary X-ray sources, 200
β-decay, Fermi theory of, 48, 50 properties of, 276–8, 305
β-rays, discovery and nature of, 131 Schwarzschild, 275–8
betatron, electron energy losses in, 139 searches for, in binary stellar systems, 199
bias parameter or factor (b), 361, 411–13, singularity theorems for, 276–7
425–8 surface of infinite redshift, 277
for different samples of galaxies, 412–13 time variability of, 306
biassing and structure formation, 409–13 Black Holes, Neutron Stars and White Dwarfs
evidence for, 411 (Shapiro and Teukolsky), 73, 212, 305
formation of galaxies in highest density Black Holes and Warped Space-Time: Einstein’s
perturbations, 411 Outrageous Legacy (Thorne), 212
Big Bang model of Universe, 319–23 black-body spectrum, 7, 34, 39, 54, 63, 126–7, 156,
origin of the term, 335 263, 280, 314–15, 394, 434
Big Bear Solar Observatory, 184–5 distortions of, by Compton scattering, 263
binary neutron stars, 201–2, 304 of relativistically moving source, 314–15
energy loss rate due to gravitational radiation, Blanco 4-metre telescope at Cerro Tololo, 162–3, 371
201–2, 212–15 blazers, 300
binary stars, 38, 45 blue supergiant stars, 199–200
bioastrophysics, 234–5 associated with X-ray binaries, 199–200
bipolar outflows in star-forming regions, 229–30 ‘blue-bump’ in spectra of active galactic nuclei, 280
association with dissipation of angular momentum blueshift of radiation from a moving source, 313
and magnetic fields, 230 Bohr’s theory of the hydrogen atom, 25, 28–9, 39, 442
Birmingham solar oscillation experiments, 184 Boksenberg’s Flying Circus, 391
522 Subject index

Boltzmann equation, collisionless, 419 Cavendish experiment, 78

for baryonic perturbations, 419 Cavendish Laboratory, Cambridge, 47, 326
for cold dark matter cosmologies, 419–20 Cavendish Physical Society, 335
Bose–Einstein spectrum, 396–7 central pressure and temperature of the Sun, 55–6
Boulby Underground Laboratory, Yorkshire, UK, centre of the Galaxy, 90
252–3 Cepheid variables, 43, 82–3, 191–2
Boyden Fund, 22, 25 due to adiabatic radial pulsations, 43
Boyer–Lundquist coordinates, 305 discovery of, in M31 and M33 (Hubble), 87
Brans–Dicke cosmology, 332 as distance indicators, 82–3, 110, 325, 344
bremsstrahlung and interstellar extinction (Joy), 91–2
cooling time for hot gas in clusters of galaxies, period–luminosity relation for, 82–3, 344
260–2 of different stellar populations, 325–6, 340
cooling time for thermal instabilities, 430–2 theory of, 191–2
opacity in stars, 47, 57–60 in Virgo cluster observed by Hubble Space
X-ray emission of hot gas in clusters of galaxies, Telescope, 344
260–2, 267–8, 345 Cerro Pachon, 166
X-ray emission of hot interstellar gas, 225 Cerro Paranal, site of VLT, 129, 165–6
British Association for the Advancement of Science, Cerro San Cristobal observing station of Lick
45 Observatory, 80
British War Office, 33 Cerro Tololo Inter-American Observatory (CTIO),
broadening processes for stellar absorption lines, 41 163, 370–1
brown dwarfs, 235 Challenger Space Shuttle disaster, 152
discovered in 2MASS and Sloan surveys, 235 Chandra: A Biography of S. Chandrasekhar (Wali),
properties of, 235 51, 64
Bruce Fund, 22 Chandra X-ray Observatory, 145–7, 210
Buckminsterfullerene (C60 ), 221 Chandrasekhar mass, 64–5, 69, 72–3, 206
Butcher–Oemler effect in clusters of galaxies, 350 origin of, 72–3
Byurakan Observatory, Armenia, 275, 279 charge coupled device (CCD) detectors, 17, 146,
164–5, 169–70, 350, 355, 371
C (charge conjugation) violation, 448 invention of, 164–5
C-field (Hoyle), 324–5, 335 quantum efficiencies of, 165
Cajnantor plateau, Chile, 162 in time-delay and integrate (TDI) mode, 169–70,
Calan/Tololo Supernova survey, 354 371
Caledonian railway tunnel near Peebles, 132 for X-ray astronomy, 146
California Institute of Technology, 204 chemical history
CalTech Submillimetre Observatory (CSO), 162 of galaxies, 389
Calver–Common reflecting telescopes, 13–14 of the intergalactic gas, 381
Cambridge 4.5-acre pulsar array, 193 chemical potential, 396–7
Cambridge 5-km radio telescope (Ryle Telescope), Cherenkov detector imaging array cameras, 149–50
141, 291 Cherenkov detectors to measure neutrino trajectories,
Cambridge APM galaxy survey, 361 181
Cambridge Encyclopaedia of Space (ed. Rycroft), 171 Cherenkov radiation of ultra-relativistic electrons and
Cambridge four-element (2C) interferometer, 326 positrons, 148–9
Cambridge Observatories, 35, 47 detection of, at optical wavelengths, 148–9
Cambridge One-Mile Telescope, 141, 291 as a detector for ultra-high-energy γ -rays, 149–50
Cambridge Optical Aperture Synthesis Telescope chondritic meteorites, 176–7
(COAST), 168 classical cosmology, 118–20
Cambridge University, 163 Classical Electrodynamics (Jackson), 215
cannibalism, galactic, 349 Classics in Radio Astronomy (Sullivan), 171
Cannon’s development of the Harvard classification classification of stellar spectra, 18–29
scheme, 25–6 Harvard, 21–8
carbon burning, 176, 178 history of, 20–9
carbon monoxide molecule, discovery in interstellar Secchi’s, 20–1
medium, 220 Vogel’s, 20–1
carbon–nitrogen–oxygen cycle (CNO cycle), 49–50, cleveite, 24
57–60, 62, 175, 179 closed-world models, 108
carbon stars, 20, 221 closure phase techniques of image reconstruction, 167
Carnegie 51-cm double astrograph, 255 cloud-chamber experiments, 48, 133–5
Carnegie Institution of Washington, 14, 125–6 clusters of galaxies, 117, 246, 253–6, 259–68, 406,
cataclysmic variables, accretion discs in, 278 411, 417
Catalogue of Selected Compact Galaxies and Abell catalogue of rich, 254–5
Post-eruptive Galaxies (Zwicky), 67, 121 average densities of, 392
causality, 291, 312, 417, 435, 439, 444 baryon fraction in, 359
Subject index 523

Bautz–Morgan classification of, 265–6 comoving radial distance coordinate, 362, 435
Butcher–Oemler effect, 350 compactness factor in extragalactic γ-ray sources,
cD galaxies in, 265–6 297, 310–11, 316
crossing time for, 259–60 relativistic beaming and, 316
dark matter in, 259–61 Compton Gamma-ray Observatory of NASA
dissipative processes in formation of, 432 (CGRO), 145, 148, 297, 301, 310–12
dynamical evolution of, 263–6 instrument package, 148, 301
epoch of formation of, 392 Compton optical depth, 345
‘fingers’ defining rich, 258 Compton scattering, 263, 345, 396–7, 434
galactic cannibalism in, 265–6, 349 induced, theory of, 396, 434
mass distribution in, from X-ray observations, computer modelling of large-scale structure
260–1, 267–8 formation, 408–10, 413–15, 432–3, 435
mass-to-luminosity ratios for, 260–1, 359 and the evolution of the intergalactic gas, 432–3
masses of, 117, 260–1 ‘computers’, Pickering’s research assistants, 22–3
most luminous galaxies in, 246, 265–6 concordance values for cosmological parameters,
constancy of absolute magnitudes of, 265–6 427–8, 439, 441–3
observed in submillimetre waveband, 375 discrepancies between observations and, 443
pressure of hot gas in, 263 Conduction of Electricity through Gases (Thomson),
problems of discovering, at large redshifts, 250 215
Sunyaev–Zeldovich effect in, 263–4, 345 Conflict in the Cosmos: Fred Hoyle’s Life in Science
maps of, 263–4 (Mitton), 335
violent relaxation in formation of, 263–4 continuous creation of matter in steady state theory,
X-ray emission of, 260–3, 267–8, 345 324–5
COBE images of our Galaxy convective equilibrium, 32, 43, 242–3
in the infrared waveband, 95 convective transport of energy in stars, 39, 53–4, 189
at microwave frequencies, 416 condition for, 39, 53–4, 242–3
Cockcroft and Walton experiment, 48 convective zones in stars, 189–91, 192
coincidence counting, 133–4 outer, 190–1
cold dark matter, 406–10 cooling flows in clusters of galaxies, 261–2
definition of, 407 cooling rate of astrophysical plasma for different
cold dark matter model for large-scale structure element abundances, 430–2
formation, 406–15 physical processes associated with, 430–1
and amplitude of fluctuations in the cosmic cooling timescale
microwave background radiation, 407 for astrophysical plasmas, 431–2
including biasing, 410–13 compared with gravitational collapse time, 431–2
cold dark matter with decaying neutrinos (τ CDM), for hot gas in clusters of galaxies, 260–2
415 for the Sun and stars, 30–1
cold dark matter model with finite cosmological Copernican system of the world, 3
constant (ΛCDM), 414–15, 422–8 Copernicus satellite (OAO-3) and interstellar
early formation of star clusters and heavy elements, abundances, 224, 333
409 Corning Glass Works, 170
evolution of perturbations in, 407, 409 coronene, 221
hierarchical clustering in, 407, 409 COS-B satellite, 147
and observed large-scale structure, 409–10 cosmic abundances of the elements (Suess and Urry),
open cold dark matter model (OCDM), 413–15 176–7
predicted temperature spectra for, 418–19 cosmic abundances of the light elements (Li, Be,
problems with, 409–10, 413 B), 331–5
similarities with isothermal model, 407 constraints on cosmological models from, 331–4
and two-point correlation function for galaxies, destruction of, in stellar interiors, 332
409–10 synthesis in early stage of Friedman models, 331–4
variants of the, 413–15 Cosmic Background Explorer (COBE), 331–2, 397,
collisionless cold dark matter, 402, 407 420, 422–3
collisions between galaxies, 244–6 cosmic chemical evolution, equations of, 389–91
simulations of, 245–6 conservation of mass, 390
star formation associated with, 246 importance of dust extinction, 390–1
collodion process cosmic microwave background radiation, 263–4,
dry, 11, 19–20 277, 320–1, 329–31, 334–5, 336, 345, 394–6,
wet, 9–10,16 400, 405, 415–27
first stellar spectra taken using, 10, 19–21 black-body spectrum of, 330–1, 415
colour index, 35–6 limits to deviations from, 331
colour–magnitude diagram, 36 Dicke’s experiment at Princeton to detect, 330
colours of the rainbow, 4 dipole component due to the Earth’s motion, 315,
Columbia Space Shuttle, 145 415–16
524 Subject index

cosmic microwave background radiation (cont.) parameter, density parameter, Hubble’s

discovery of, 321, 329–31, 394 constant
contributions to the total measured signal, 331 age of the Universe, 346–9, 425–7
distortions of spectrum of, 396–7 from cooling ages of white dwarfs, 347
early limits to the temperature of, 330, 336 methods of estimating, 346–9
energy density of, 396 concordance values for, 427–8
fluctuations in the, 415–27; see also fluctuations in constraints from synthesis of the light elements,
the cosmic microwave background radiation 331–2, 404–5, 419
isotropy of, 415–16, 439 cosmological parameters, relations between, 118–19,
polarisation of, 422–4, 427 340–1
prediction of Alpher and Herman of, 320–1 Cosmological Physics (Peacock), 451
prediction of Doroshkevich and Novikov of, 330 cosmological principle, 113–14, 443
spectrum of, 331–2 cosmological problems in 1939, 118–20
temperature of, 330–2 cosmological time dilation, 353
measured using interstellar molecular lines, cosmology
330–1 fundamental problems of, 439–52
cosmic rays, 66–7, 126, 170 initial condition for concordance model, 439
beryllium-10 in, 240–1 origins of astrophysical, 100–21
chemical composition of, 236–41 precision, 439
transfer equation for, 239 up to the time of Einstein, 100–1
discovery of, 132–4 Cosmology (Bondi), xii, 120
and the discovery of elementary particles, 134–6 Cosmology: The Science of the Universe (Harrison),
electrons, spectrum of, 270 xii, 95, 120
energy spectrum of, 236–7 Cosmology and Controversy: The Historical
in the Galaxy, 140 Development of Two Theories of the Universe
highest energy, 294 (Kragh), xii, 319, 328, 335–6
history of, 132–4, 236–41 COSMOS high-speed measuring machine, Edinburgh,
ionisation losses of, in heating interstellar medium, 169, 370
224 Coulomb repulsion, 48–9
light elements in, 238–40 counts of radio sources, see number counts of radio
path length distribution for, 239 sources
primordial origin for, 320 covariance, principle of, 103
relativistic electrons in, 237–8 CP (charge conjugation combined with parity)
residence time in the Galaxy, 240–1 violation, 398, 448
solar modulation of, 236–7 Creation of the Anglo-Australian Observatory, The
source chemical composition of Galactic, 239 (Gascoigne et al.), 172
techniques for detection of, 236 creation-field, or C-field (Hoyle), 324–5, 335
Cosmic Rays (Hillas), 171, 241 Crimean Astrophysical Observatory, 184
cosmic-ray astrophysics and the interstellar critical brightness temperature in compact radio
medium, 236–41 sources, 291, 309–10, 316
cosmic-ray clocks, 240–1 and relativistic beaming, 316
cosmic star-formation rate, 162, 362–4, 365, 375–6 critical cosmological density, 116–17, 333, 341–2,
cosmic strings, 419 357–9, 439
cosmic textures, 419 and the flatness problem, 439
cosmic time, 113 critical cosmological model, 116–17
cosmic virial theorem, 359 Critical Dialogues in Cosmology (ed. Turok), 360–1,
cosmological constant, 106–9, 114–16, 118, 120, 323, 413–15
325–6, 335, 341–2, 362, 413–15, 441, 444 Crookes discharge tubes, 130
interpretation according to quantum field theory, Crookes radiometer, 155–6
441, 444 Crossley 91-cm reflecting telescope, 13
Cosmological Constants: Papers in Modern Cryogenic Dark Matter Experiment, 252–3
Cosmology (eds Bernstein and Feinberg), xii Curtis Schmidt telescope at Cerro Tololo, 371
‘cosmological distance ladder’, 343–5 curvature modes in models of structure formation,
Cosmological Distance Ladder, The 407–8
(Rowan-Robinson), 362 curvature of space (k ), 113–14, 118–19, 341–2,
cosmological evolution of galaxies and active 425–8, 441
galaxies with cosmic epoch, 365–91 curvature radiation, 196
active galaxies, 365–74 curve of growth techniques for determining chemical
Cosmological Inflation and Large-Scale Structure abundances, 41–2
(Liddle and Lyth), 447, 452 cyanoacetylene, discovery in interstellar medium, 220
cosmological parameters, determination of, cyclotron radiation feature in spectra of X-ray sources,
340–64, 423–8; see also under deceleration 199
Subject index 525

D lines of sodium, 6 Department of Scientific and Industrial Research

daguerreotype process, 9–10, 16 (UK), 193
dark ages, the Universe at z > 6, 428–9 deuterium
dark clouds, extinction in, 216 binding energy of, 338
dark energy, 353–4, 423–8 cosmological significance of abundance of, 151–2,
density parameter for ( ), 425–8, 441, 444 361
equation of state for, 425–8 detection of, by ultraviolet spectroscopy in
nature of, 441, 443 interstellar medium, 151, 333
dark matter, 116–17, 248–53, 260–1, 266, 359–61, discovery of, 48
399, 409–13, 425–8 formation by primordial nucleosynthesis, 338–9
baryonic, 399 uniformity of abundance in interstellar medium,
and biasing, 410–13 333
controversy over amount of, 360–1 deuterium-burning stars, 235
density parameter for (D ), 425–8 development of small perturbations in the
in galaxies, 248–53, 359–61 expanding Universe, 436–8
nature of, 250–3, 399, 441, 443 diameters of stars
particles, discovery of, 443 estimates of, 34, 38, 45
Darkness at Night (Harrison), 120 measurements of, 45–6, 156, 166
de Sitter solution, 107–9, 113, 444 Dictionary of Scientific Biography (ed. Gillespie), xii,
redshift–distance relation, 109–10 16
written as an expanding metric, 107 differential rotation of stars in the Galaxy, 89–90, 97–9
de Sitter space, 447 diffuse interstellar bands, 222
Dead of Night, The, 324 feature at 217.5 nm, 222
deceleration parameter q0 , 118–19, 341–2, 349–57, Dirac’s theory of the electron, 48, 135, 171
362 Dirac’s theory of the expanding Universe with varying
from angular diameter–redshift relation, 349, 356–8 gravitational constant, 323
history of determination of (Sandage), 361–2 dissipation of primordial turbulence, 396
independent of 0 and curvature of space at small distance indicators in cosmology, 343–4
redshifts, 362 brightest stars in galaxies, 344
from redshift–apparent magnitude relation luminosity function of globular clusters, 344
for brightest galaxies in clusters, 349–50 supernovae of Type Ia, 344
for radio galaxies, 350–2 distances of galaxies, 110
for Type Ia supernovae, 349, 352–5 distances of the stars, 7–9, 37
from number counts of galaxies, 349, 354–6 ‘Divine Centre’ of the star system, 77
difficulty of use to determine, 354–6 Doppler shifts in superluminal motions, 295, 300
from observations of distant objects, 349–57 double-slit experiment, Young’s, 4
from spectrum of fluctuations in cosmic microwave doubling processes in the formation of molecular line
background radiation, 349, 353, 423–8 spectra, 220
Deep Sky Companions: The Messier Objects Draper 11-inch telescope, 23
(O’Meara), 95 Draper Memorial Catalogue of Stellar Spectra, 22–3,
deflection of light rays by the Sun, 103–5 28, 34
degeneracy pressure, 64–5, 70–3 Durham University Cherenkov array, Dugway, Utah,
relativistic, 71–3 149
densities of stars, 38 dust extinction and observability of distant
density contrast, 359, 392–4, 436–8, 440 star-forming galaxies, 375
growth rate with redshift, 392, 436–8 dust grains, 216, 351
density of the Universe, 87–8, 116–19, 341–2, 357–8 emission spectrum of, 375–6
from cosmic virial theorem, 359 K-corrections for dust sources at submillimetre
Hubble’s estimate of, 87–8, 358 wavelengths, 375
from infall of galaxies into superclusters, 359–61 small, discovery of, 221
from mass-to-luminosity ratio for Universe as a transient heating of, 221
whole, 358–9 dust-to-gas ratio as a function of redshift, 384
density parameter in baryons B , 404–5, 425–8 dwarf galaxies, 246
density parameter in vacuum fields ΩΛ , 352–5, dwarf stars, 34
425–8, 441 dynamical astronomy, 38
predicted value according to quantum field theory, dynamical friction, 264–5, 286
441 Dynamics of Ionised Media conference, 171
density parameter of the Universe Ω0 , 119, 341–2, dynamo processes to maintain the Sun’s magnetic
352–5, 357–61, 362, 392, 404–5, 425–8, 441 field, 188
density parameters of stars, gas and heavy elements,
389–91 Early Years of Radio Astronomy (Sullivan), 171
density perturbations, development with cosmic early-type galaxies, 87
epoch, 398–402, 436–8 early-type stars, 39
526 Subject index

Earth-rotation aperture synthesis, 141–2 stability of triaxial distribution of matter in, 253
Echo satellite, 330 as triaxial systems, 253
eclipse expedition of 1919, 104–5, 120 Emmanuel College, Cambridge, 4
eclipsing binary star systems, 38 Encyclopaedia of Astronomy and Astrophysics
light curves of, 38 (ed. Murdin), xii, 215
X-ray binaries as, 197–8 end points of stellar evolution, 61–73
Eddington–Chandrasekhar controversy, 65–6, 69 Energy and Empire: A Biographical Study of Lord
Eddington Enigma, The (Evans), 51 Kelvin (Smith and Wise), 50
Eddington–Lemaı̂tre world models, 118–19 energy density of radiation in the Universe, 396
‘coasting phase’ of, 118–19 energy generation, equation of (third equation of
Eddington limiting luminosity, 43, 198–9, 278, stellar structure), 53, 55
283–5, 307 energy generation rate by nuclear reactions, 44, 47,
Eddington: The Most Distinguished Astrophysicist of 48, 50, 53, 55–8
His Time (Chandrasekhar), 51, 69 energy loss of binary systems by gravitational
Eddington’s theory of red giants, 45 radiation, 201–2
Eddington’s theory of stellar structure and evolution, energy loss rate by Crab Nebula pulsar, 195
42–8, 61 energy-momentum tensor, 105
‘effective distance’, 362 energy source of the Sun, 30–1, 44–5, 48–50
effective temperature of a star, 59–60 annihilation of matter as the, 44–5
effective wavelength, 35 gravitational attraction as the, 30
effective wavelength–apparent magnitude diagram, meteoritic bombardment as the, 30, 33
35–6 nuclear energy as the, 31, 43, 45, 47, 48–50
Effelsberg 100-metre radio telescope, 140 radioactive decay as the, 45, 47
Einstein–de Sitter world model, 114–17, 404, 436–7 energy supply from Crab Nebula pulsar to the nebula,
Einstein X-ray Observatory (HEAO-B), 144–6, 262, 195
372–3 energy transport, equation of (fourth equation of
Einstein’s field equations, 105–6, 275 stellar structure), 53–5
Einstein’s Kyoto address, 102 enrichment of interstellar medium by supernova
Einstein’s route to general relativity, 102–5 explosions, 207
Einstein’s static model of the Universe, 88, 105–7, epicyclic motion of stars in the Galaxy, 89
108, 109, 118, 323 epoch of equality of matter and radiation energy
Electricity and Matter (Thomson), 215 densities, 396, 399–400, 406, 408, 414–15
electrodynamics of pulsars, 195 for finite-mass neutrinos, 406
electromagnetic processes in formation of jets, 297–8 epoch of recombination, 392, 394–6, 417, 434
electromagnetic spectrum, opening up of, 125–72 ionisation of the intergalactic gas through, 417
electron–positron annihilation γ -rays from the physics of recombination at, 434
Galactic Centre, 148 temperature of gas at, 434
electron–positron pair production, 147, 297, 310–11, equation of state of stellar material, 44, 48, 64,
440 70–2, 175, 192
electron scattering in hot stars, 43, 57–60 of degenerate matter, 64, 70–2
electronic detectors for optical astronomy, 163–5 of neutron star matter, 195, 202
electronic vacuum tubes, 163–4 equations of stellar structure, 44, 51–60, 188
image intensifiers, 163–4 numerical integration of, 188–91
photomultiplier tubes, 163–4 equilibrium abundances of particles and antiparticles,
Electronic Imaging in Astronomy: Detectors and 336–7
Instrumentation (McLean), 172 for bosons and fermions, 336–7
electrons, high-energy in the interplanetary medium, non-relativistic case, 337
293 equilibrium total energy density in bosons and
electroscope, 132 fermions, 337
Elements of Physical Manipulation (Pickering), 21, equipartition in energy requirements for synchrotron
27–8 radiation, 270–1, 304
elements present in stellar atmospheres, 18–20 equipotential surfaces in binary systems, 199
elements present in the solar atmosphere, 7 equivalence, principle of, 102–3
elliptical galaxies, 87–8, 246–7, 289–90, 432 equivalent width of absorption line, 41
black hole–bulge mass correlation, 289–90 EROS project, 250–2
dark matter in, from X-ray brightness distribution, ESO Schmidt telescope, 169
250 ESO 3.6-metre telescope at La Silla, 163
dissipation processes and, 432 ethyl alcohol, discovery in interstellar medium, 220
dynamics of, 253 Euclidean number counts of galaxies, 96–7, 112
Faber–Jackson relation for, 246–7 effects of redshift upon, 112
fundamental plane for, 247 Euclid’s fifth postulate, 100–1
rotation of, 253–4, 283 European Southern Observatory (ESO), 129, 162–3,
spectral energy distributions for, 362–4 287
Subject index 527

European Space Agency (ESA), 152 Fanaroff–Riley (FR) Classes 1 and 2, 291–2
evolution of galaxies and active galaxies with Faraday rotation of the polarised emission of radio
cosmic epoch, 365–91 sources, 270
convergence of evolution at large redshifts, 367–72 Far-infrared Space Observatory (FIRST) of ESA, also
strong evolution out to z ∼ 2–3, 367–72 known as the Herschel Space Observatory, 160
Evolution of Radio Astronomy, The (Hey), 171 Fermat’s principle of least time, 103
evolution of stars from main sequence to giant branch, filling factor of emission line regions in active galactic
188–92 nuclei, 280–1
for low-mass stars, 190–1 fine structure constant, 73
exoplanets, see extrasolar planets flame spectrum, 7, 16
Expanding Universe, The: Astronomy’s ‘Great flash spectra, 41
Debate’ 1900–1931 (Smith), 96 Fermi acceleration of charged particles, 293–4, 302
expansion of the Universe, limits to anisotropic early Fermi–Dirac statistics, 48, 64, 71–2
expansion of, 332 Fermi momentum, 71
Explorer satellites of NASA, 144 Fermi’s theory of weak interactions, 48, 50
Explorer 11, 147 fine-tuning problem, 439–40
Exploring the Universe with the IUE Satellite first law of thermodynamics for a relativistic fluid, 451
(Kondo), 172, 241 First Stromlo Symposium: Physics of Active Galactic
explosive nucleosynthesis, 178, 206–7 Nuclei (eds Bicknell et al.), 306
extensive air showers, 134, 149 First Texas Symposium on Relativistic Astrophysics,
extinction by dust, discovery of, 91–2 273
extinction coefficient for dust, 156–7 First World War, 30
extragalactic γ -ray sources, 297 flares from nucleus of the Galaxy, 287
compactness factor for, 297, 310–11 fluctuations in the cosmic microwave background
relativistic aberration effects in, 297, 310–11 radiation, 353, 361, 405, 415–28, 448
superluminal radio motions associated with, 297, ACBAR experiment, 424
310–11 acoustic, or Sakharov, oscillations, 418–23
time variability of, 297, 310–11, 316 discovery of, 422–3
extragalactic radio sources Archeops experiment, 422–3
adiabatic losses in, 293–4 Boomerang experiment, 422–3
CBI experiment, 422–3
alignment effect in, 298–9
COBE observations of, 415–16, 417–18
association with massive galaxies, 272
cosmic variance associated with observations of,
double radio sources, 291–300
443
and the angular diameter–redshift test, 357–8
DASI experiment, 422–3
continuous-flow models for, 293–4
and determination of cosmological parameters,
self-similar models for, 293
423–8
early history of studies of, 269–75 consistency with independent estimates, 427
evolution with cosmic epoch, 365–9 discovery of, of cosmic origin, 415–16
hot spots in, 291–4 evolution of perturbations up to the epoch of
Fanaroff–Riley (FR) Classes 1 and 2, 291–2 recombination, 419–20
luminosity function for, 366–9 due to first-order Doppler scattering, 401–2
magnetic fields in, 293–4 on intermediate angular scales, 418–23
number counts for, 326–8, 366–9 amplitudes of Sakharov oscillations, 421
optical identification of, 270–5, 367 first acoustic peak in, 419–21
properties of, 140 as a ‘rigid rod’, 420
radio jets observed in, 294, 299–300 on large angular scales, 417–18
spectra of, 270 due to gravitational redshifts at last scattering
structures of, 140, 270–1 layer, 417–18
surveys of, 326–8 large polarisation signal associated with
unified schemes for radio galaxies and radio reionisation of intergalactic gas, 422–4
quasars, 299–300 due to primordial gravitational waves, 418
Extrasolar Planet Encyclopaedia, The (Schneider), Maxima experiment, 422–3
241 minimum set of parameters to account for, 426–8
extrasolar planets, or exoplanets, 232–5 multipoles associated with, 419–22
atmospheres of, 234–5 non-Gaussian features in, 419
discovery of, 232–3 origin of, in cold dark matter cosmologies, 419–22
eclipsing, 233–4 phases of acoustic oscillations on last scattering
ellipticity of orbits of, 234 surface, 421
Jupiter-sized, 232–3 and physics of ionisation of the intergalactic gas
methods of detection of, 232–4 through the epoch of recombination, 417, 435
properties of, 233–5 polarisation observations of, 422–4
statistics of, 233–4 predicted power spectrum of, 419–20
528 Subject index

fluctuations in the cosmic microwave (cont.) during the radiation- and matter-dominated eras,
random superposition of perturbations on small 440
angular scales, 417, 422 horizon problem, 439, 441
ratio of mass density in neutrinos to dark matter primordial fluctuation problem, 440–1
( f v ), 425–7 values of cosmological parameters, 441–2
ratio of tensor to scalar perturbations (r), 425–7 ways ahead in understanding, 442
Sachs–Wolfe effect, 417–18, 421 clues from particle physics, 442
dependence upon spectral index of perturbation as initial conditions, 442
power spectrum, 418 new concepts, 442
Sakharov oscillations, 418–23 Fundamental Theory (Eddington), 323
Silk damping of, at large multipoles, 421 funding of astronomical research, 22
spherical harmonic power spectrum analyses of, Future of Theoretical Physics (eds Gibbons et al.),
419, 435–6 452
Legendre polynomials and, 436
suppression factor at large multipoles, 421–2 Galactic Astronomy (Binney and Merrifield), 266
‘vanilla’ CDM model to account for, 426–8 Galactic centre, 163, 203, 274
‘vanilla-lite’ CDM model to account for, 426–8 high-velocity HI clouds in, 218
Very Small Array experiment, 421–2 infrared images of, 158–9
width in redshift and physical scale of last Galactic Dynamics (Binney and Tremaine), 266
scattering layer, 417, 435 Galactic halo, hot, 224–5
WMAP experiment, 415–16, 422–7 Galactic radio emission, 269–70
angular power spectrum derived from, 422–5 background radiation due to, 138, 139–40
cross-polarisation power spectrum from, 422–4 nature of, 136, 139–40
fluorescent materials, 130 polarisation of, 270
forbidden lines, identification of by Bowen, 19, spectrum of, 269–70
216–17 Galactic rotation, 89–91
force-free magnetic fields, 297–8 discovery of, 89–91
formaldehyde molecule, discovery in interstellar Oort’s analysis of differential rotation of stars in the
medium, 219–20 Galaxy, 89–90
galaxies, 87–8, 244–68
formation of stars, 225–32
average densities of, 392
Formation of Structure in the Universe (eds Dekel and
background radiation due to, 246
Ostriker), 305
and bursts of star formation, 377
Foucault, The Life and Science of Léon (Tobin), 17
collisions between, 244–6
Fourier transform pairs, 434
dark matter in, 248–53
fourth Cambridge (4C) catalogue of radio sources, 328
de Vaucouleurs classification scheme for, 244
fourth test of general relativity (Shapiro), 120–1
and determination of large-scale structure of the
fractal models of the Universe, 77 Universe, 87–8
fractional look-back time, 372 evolution of stellar populations of, 350
Fraunhofer lines extragalactic nature of, 87–8
in the solar spectrum, 5–7, 24 Hubble’s classification scheme for, 87–8, 244
in stellar spectra, 23 K-corrections for galaxies of different types, 377–8
Fraunhofer’s spectroscopic observations of the planets large-scale distribution of, in the Universe, see
and stars, 5–6 large-scale distribution of galaxies in the
Fred Lawrence Whipple 10-metre Telescope, Mount Universe
Hopkins, USA, 149–50 masses of, 87–8, 116
free–free absorption by stellar material, 47, 57–60 morphological typing of faint HST galaxies, 379–80
free streaming of neutrinos from neutrino nature of the faint blue galaxies, 379
perturbations, 406–7 number counts of
frequency four-vector, 313 and determination of cosmological parameters,
Fried parameter, 166 354–6
Friedman world models, 107–9, 113, 114–19, 320, as a test of the homogeneity of the Universe, 87,
323, 341–2, 358, 393, 435, 438, 441, 444 96–7
Lemaı̂tre’s independent discovery of, 108 observations of in the infrared waveband, 362–4
non-equilibrium nucleosynthesis in early phases of, origin of, 392–438
320–3 physics of, 244–68
parametric solutions for models with  = 0, 436 properties of, 87–8, 109, 244–8
fullerenes, 241 redshift distributions at faint magnitudes, 378–9
fully convective stars, 242–3 role of dissipative processes in formation of, 431–2
fundamental observer, 113, 115 tails and bridges between, 244–6
fundamental problems of cosmology, 405, 439–42 types of, 87–8, 244
baryon-asymmetry problem, 397–8, 440 ultraviolet spectra of, 377
flatness problem (or fine-tuning problem), 439–41 Galaxies in the Universe (Sparke and Gallagher), 266
Subject index 529

Galaxy (‘Milky Way’ galaxy) Gaskugeln (Emden), 32

age of, 324–5 gauge-invariant formalism for superhorizon density
as a barred spiral galaxy, 92–5 perturbations, 418
black hole in nucleus of, 286–7 Gaunt factor, 58
dark halo of, 250–3 Gauss: A Biographical Study (Bühler), 120
dark objects in bulge and halo of, 250–2 Gauss’s geodetic survey of Hanover, 101, 120
discovery of spiral structure of, 92–5 Gauss’s theorem for gravity, 115, 436
distribution of neutral hydrogen in, 218–19 Gauss’s theory of surfaces, 103
‘spiral features’ in, 218–19 Gaussian perturbations, 403, 409–11, 419, 436
distribution of stars perpendicular to the plane of, 81 Geiger–Müller detector, 133
epoch of formation of, 392 gelatin photographic plates, 11, 19
first radio maps of, 136–7 Geller–Huchra redshift survey of galaxies, 256–9
Reber’s map, 136–7, 218 General Catalogue of Nebulae (Herschel), 79
interpenetrating streams of stars in, 80 General Electric Research Laboratory, Schenectady,
mass density in the plane of, 81 139
mass of, 89 General History of Astronomy, The, Volume 4
and nature of the spiral nebulae, 77–99 (ed. Gingerich), xii, 51
orbits of infrared stars in nuclear regions of, 286–7 general relativity and Einstein’s Universe, 102–7
Shapley’s use of globular clusters to determine, General Relativity and Gravitational Radiation
83–4 (Weber), 203
star formation correlated with spiral arms, 225 general theory of relativity, 63, 68, 69, 101–5, 108,
structure of, 79–91 113, 197, 201–4, 273, 275–9
Galaxy Formation (Longair), 121, 266, 339, 362, 391, and high-energy astrophysics, 273
433–4, 438, 451 and models of active galactic nuclei, 275–9
GALLEX experiment for solar neutrinos, 182 origins of, 102–5
γγ annihilation problem, 297, 302, 310–11 tests of
Gamma-ray Astronomy (Ramana Murthy and on the cosmological scale, 342
Wolfendale), 171 and radio pulsars, 201–2
γ-ray astronomy, history and development of, geometry of space, local, 101, 342, 405
147–50 according to the inflationary picture, 405
discovery of discrete γ -ray sources, 147 limits to radius of curvature of, 101
γ -ray background radiation, 147 GEO600 gravitational wave observatory, 204
γ-ray bursts, 300–4, 306 giant branch, 35–6
afterglows associated with, 301–3 Giant Metre-wave Radio Telescope (GMRT), India,
beaming of radiation in, 303 141
collapsar model for, 303–4 giant molecular clouds, 225–8
convergence of number counts of, 301 star formation in, 226–8
deceleration of beam creating, 303 giant stars, 24, 34, 38, 44–5, 61–3
discovery of, 148, 300 discovery of, 34–5
duration of, 300, 303–4 glitches, in pulsar slow-down rate, 196
energy requirements for, 302–3 global star-formation rate as a function of cosmic
energy sources for, 303–4 epoch, 387–91
extragalactic nature of, 302–3 globular clusters, 83–4, 89–90, 93, 326, 347–8, 403
γ γ annihilation problem, 302 ages of the oldest, 347–8
isotropy of, over the sky, 148, 301 and the determination of the structure of the
relativistic blast wave/fireball model for, 301–3 Galaxy, 83–4, 89–90
spectra of, 302–3 evolutionary tracks for stars in, 188–9
spectral evolution of, 302–3 Hertzsprung–Russell diagram for, 188–9
supernova association with, 303 use of, in M87 to measure mass distribution, 250
theories of, 301–3, 305 gold medal of Royal Astronomical Society, 13
γ -ray line emission, 148 GONG experiment, 184
γ -ray map of the sky, 148–9 grand unified theories (GUT) of elementary particles,
γ -ray Observatory, Compton, of NASA (CGRO), 145, 444–7
148–9 GUT era, 451
γ -ray pulsars, 147 symmetry breaking during, 444–7
γ -ray sources, see extragalactic γ -ray sources Gravitation (Misner et al.), 305
γ -rays gravitational collapse, 65–6, 393–4, 436–8
discovery and properties of, 131–2 dispersion relation for, 393
from the Galactic centre region, 147, 236 in an expanding medium, 393–4, 436–8
Gamow’s theory of barrier penetration, 50 growth rate of, 393–4, 436–8
Gamow’s theory of primordial nucleosynthesis, problems of, 393–4, 437
319–23, 329, 335–6 and formation of structure in expanding
Dicke’s role in reviving interest in, 336 Universe, 393–4
530 Subject index

gravitational collapse (cont.) acoustic or p modes, 184–5

Jeans’criterion for, 393 angular velocity distribution inside the Sun, 187–8
timescale for, 393, 431–2 convection currents with the Sun, 187–8
gravitational constant, limits to variation with time, discovery of, 183–4
202, 332, 334 dispersion relation for, 183–5
‘gravitational’ fine-structure constant, 73 ‘five-minute’ oscillations, 183–4
gravitational interactions between galaxies, 244–6 gravity or g modes, 184
gravitational lensing and estimates of Hubble’s IPHIR experiment on the PHOBOS spacecraft,
constant, 345–6 184–5
gravitational microlensing and searches for MACHOs, limits to non-standard models of the solar interior,
250–2 186
gravitational radiation, 212–15 normal modes of oscillation of the Sun, 183–4
energy loss rate by binary systems, 212–15 SOHO experiment of ESA, 184–8
and inspiralling of binary systems, 214 solar oscillations as trapped acoustic waves, 183
quadrupole emission of, 213–14 speed of sound within the Sun, 186
gravitational redshift, 63, 105, 274 two-dimensional structure of the solar interior from
at the surface of white dwarfs, 63, 105 SOHO measurements, 187–8
gravitational wave astronomy, 128, 170, 202–4 helium, 328–9, 331–5, 336–9
gravitational waves, 201–4 abundance predicted to be insensitive to baryonic
detectors for, 203–4 density of the Universe, 329, 336–9
energy loss by binary neutron star system, 201, discovery of, 24–5
212–15 formation by primordial nucleosynthesis, 329,
history of the search for, 203–4 336–9
primordial, search for, 443 constraints on world models, 331–4
search for, 202–4 helium–hydrogen ratio from, 338–9
sources of, 204 nuclear reactions involved in, 338
Weber’s aluminium bar experiments, 203 high abundance in old globular cluster, 329
gravitationally lensed quasars, 345–6 intergalactic neutral, 379
and angular diameter–redshift test, 356–7 problems of observing, 328–9
discovery of, 345–6 uniformity of its abundance, 328–9
time delay of variability of images of, 345–6 helium burning, 62, 176–7, 189–90, 329
gravitinos as cold dark matter particles, 406–7 in a shell about degenerate core, 189–90
grazing incidence optics for X-ray imaging, 144–5 in stellar cores, 189–90
‘Great Debate, The’, 84–7 helium flash, 191
Great Exhibition of 1851, 13 helium problem, 328–9, 331–5
Great Observatories programme of NASA, 145 low abundance produced by stellar nucleosynthesis,
groups of galaxies, 253–5 329
Gunn–Peterson test for intergalactic neutral hydrogen, revisited, 331–5
379–80, 397, 428–9 Henry Draper Catalogue (HD Catalogue), 21, 27, 29
positive result for largest redshift quasar, 428–9 Henry Draper Fund, 21–2
gyrofrequency, 139, 309 Henry Draper Memorial project, 27
Herbig–Haro objects, 230
Harrison–Zeldovich primordial mass spectrum, 401, Herschel Space Observatory of ESA, also known as
407–8, 409–10, 413, 418, 440–2, 447 the Far-infrared Space Observatory (FIRST),
transfer functions for modifications to, when 160
perturbations come through horizon, 407–8 Herschel’s 20-foot telescope, 79
Harvard Annals, 25–6, 29 Herschel’s 40-foot telescope, 9–11, 78
Harvard classification of stellar spectra, 21–9, 34, 40 Herschel’s model of the Galaxy, 77–9
Harvard College Observatory, 21–2, 25, 27, 125, 218 problems with, 78–9
plate archives, 273 Hertzsprung gap, 189
Harvard–Smithsonian Center for Astrophysics survey Hertzsprung–Russell diagram, 34–9, 175, 190,
of galaxies, 256–8, 409 226–7, 346–8
Harz mountains, 101, 120 ages from shape of, for globular clusters, 347–8
Hayashi limit, 189–91, 226–7, 242–3 evolution of stars on, 189–91
Hayashi track, 226–7, 242–3 evolutionary corrections for galaxies from, for
heat diffusion equation, 54 globular clusters, 350
heavy element abundances, 365 main-sequence termination point, 347–8
determination of rate of formation from ultraviolet origin of, 34–9
continuum of star-forming galaxies, 384–6 Hess ‘ultra γ -radiation’, 132–3
evolution with cosmic epoch, 365 hierarchical models of the Universe, 77
Heisenberg uncertainty principle, 70–1, 451 Higgs fields and Higgs particles, 441, 444–5, 451
heliometer built by Fraunhofer for Bessel, 6, 8 High Energy Astrophysical Observatory-A
helioseismology, 182, 183–8, 211 (HEAO-A) of NASA, 144
Subject index 531

High Energy Astrophysical Observatory-B (HEAO-B) Sunyaev–Zeldovich effect in X-ray clusters of

of NASA, 144 galaxies, 345
High Energy Astrophysical Observatory-C (HEAO-C) problems of cosmic evolution and, 349
of NASA, 148, 238 revisions of, 325–6, 340, 346
high-energy astrophysics, 269–316 Sandage and the values of H0 and q0 , 340–3
High Energy Astrophysics, Volume 1 (Longair), 60, Hubble’s velocity–distance relation for galaxies
242, 268, 316 (Hubble’s law), 109–12, 116, 324
High Energy Astrophysics, Volume 2 (Longair), 69, Huyghen’s principle, 103
73, 242, 304–5, 316 Hydrodynamics (Lamb), 184
high-velocity stars, 89, 93 hydrogen, intergalactic neutral, 379–80
Hipparcos astrometric satellite (ESA), 9, 36–7, 344, hydrogen burning, 176–8
347 in shell about degenerate core, 189
Historical Development of Modern Cosmology in stellar cores, 189
(eds. Martı́nez et al.), xii, xiv, 120, 336 hydrostatic equilibrium, equation of, 31–3, 51, 55,
History of Astronomy from 1890 to the Present, A 267
(Leavington), xii hydroxyl molecule, 219
Homestake gold-mine, South Dakota, 180 discovery in interstellar space, 219
homogeneity of the Universe, from galaxy number maser emission of, 219
counts, 112, 365 hyperbolic space, 100–1
from two-point correlation functions for galaxies, minimum parallax in, 101
255–6 Hy-redshift Universe, The (eds Bunker and
homology relations for the structures of the stars, 50, van Breugel), 362
56–60,188, 242–3
horizon scales, 394, 400, 435, 445 Ilford Photographic company, 136
event, 435 Image Photon Counting System (IPCS), 164, 381, 391
particle, in matter- and radiation-dominated eras, IMP-7 and IMP-8 space missions, 240
435, 439 Index Catalogues of Nebulae and Star clusters (IC)
horizontal branch stars, 191–2 (Dreyer), 79
hot dark matter inelastic scattering of α-particles, 48
neutrinos with finite rest mass, 405–6 infall of gas into galaxies, 390
scenario for structure formation, 405–6, 408–9 infall velocity of galaxies into superclusters, 359–60
amplitude of temperature fluctutions in the inflationary model of the very early Universe, 335,
cosmic microwave background radiation, 406 405, 418, 425–7, 441–2, 444–7, 449, 451–2
damping of low-mass perturbations, 406 creation of monopoles in, 446
formation of excessive large-scale structure, 409 exponential expansion at very early epochs, 444–6
hot-spots in double radio sources, replenishment of horizon and flatness problems, 444
energies in, 293, 295 monopole problem, 446
HII regions, 217 new inflationary model, 446–7
Hubble and the Universe of galaxies, 87–8 physical realisation of, from particle physics, 445–7
Hubble Atlas of Galaxies (Sandage), 244 reheating of the Universe at the end of the inflation,
Hubble Deep Field, 365, 376, 377–80, 386, 388, 390 445
number counts of galaxies in, 377–8 slow roll-over picture of, 447
observed in the submillimetre waveband, 376 spontaneous symmetry breaking in, 445
Hubble sequence of galaxies, 87–8, 244 topological defects formed during GUT phase
correlations along the, 244, 266 transition: monopoles, cosmic strings and
Hubble Space Telescope (HST), 129, 145, 152–3, domain walls, 446
162, 164–5, 169, 209, 234, 272, 284, 298, 302, transition from false to true vacuum, 445–7
344–5, 353, 365, 376, 379, 386–7, 391 transition to standard radiation-dominated model,
history of, 152–3 444–6
key project, 344 typical realisation of, 444–6
Medium Deep Survey, 379–80 variants of, 447
scientific instruments of, 152–3, 164–5 Inflationary Universe, The: The Quest for a New
Hubble Space Telescope and the High-Redshift Theory of Cosmic Origins (Guth), 451
Universe, The (eds Tanvir et al.), 391 inflaton potential, 447
Hubble Ultra-Deep Field, 153–4, 365 infrared astronomy
Hubble’s constant (H0 or h), 110, 118, 248, 324, chopping techniques for, 156
325–6, 340–2, 343–6, 357, 359, 425–8 cryogenic techniques for, 157–8, 160
controversy over value of, 344 gallium-doped detectors for, 157
final HST key project result, 344 Herschel’s pioneering observations of the Sun, 153
and the Palomar 200-inch telescope, 340–1 history and development of, 153–62
physical approaches to determining, 345–6 indium-antimonide detectors for, 158, 350
Baade–Wesselink method, 345 infrared array detectors for, 160
gravitationally lensed quasars, 345–6 lead sulphide detectors for, 156–8
532 Subject index

infrared astronomy (cont.) interstellar dust, 156–60, 222–3

‘nodding’ to remove effect of infrared background, alignment mechanisms for, 223
155 extinction curve for, 156–7
observations of star-forming regions and galaxies and star formation, 226–30
in, 162 theory of scattering by, 222
observing distant galaxies in the infrared interstellar extinction, 78–9, 86, 88, 91–2, 216, 222–3
waveband, 362–4 and the structure of the Galaxy, 91–2
semiconductor materials for detectors for, 156–8 interstellar magnetic field, 223,
spectroradiometry of stars, 156 and alignment of interstellar dust grains, 223
thermal detectors for, 153–7 and Galactic synchrotron radiation, 223, 237–8
vacuum thermocouple, 156 large-scale distribution, from polarisation of nearby
Infrared Astronomy Satellite (IRAS), 160, 227–9, stars, 223
246, 374 interstellar medium, 91–2, 216–43, 270
deep infrared survey of the sky, 227–8 depletion of abundances of common elements in,
spectra of star-forming regions, 228–9 224
and star-forming regions, 227–9 discovery of, 91, 216
Infrared Astronomy with Arrays (eds. Wynn-Williams heating and cooling mechanisms for, 224–5
and Becklin), 172 hot component of, 224–5
infrared radiation, shown to have the same properties molecules in the, see molecules in the interstellar
as light, 155 medium
infrared sources, intense sources in Orion, 158–9 photoionisation of, 216–17
Infrared Space Observatory (ISO) of ESA, 160, 222 physics of, 216–43
Infrared Telescope Facility (IRTF) of NASA, 158, 162 thermal stability of, 223–5
Infrared: The New Astronomy (Allen), 172 two-phase model of, 224–5
inhomogeneous stars, 61–3 violent, 225
initial mass function for stars, 230–2, 362–4 interstellar polarisation, 223
and evolution of spectral energy distribution of correlation with extinction by dust, 223
galaxies, 362–4 theory of, 223
and global star-formation rate, 231 Introducing Einstein’s Gravity (d’Inverno), 215
Miller and Scalo, 231–2 Introduction to Active Galactic Nuclei, An (Peterson),
Salpeter, 230–2 305
initial power spectrum of cosmological Introduction to the Study of Stellar Structure, An
perturbations (Chandrasekhar), 49
analytic reconstruction in non-linear phases, 411–12 inverse Compton catastrophe, 290–1, 308–10, 316
curvature of (α), 425 inverse Compton scattering, 291, 308–10, 316
reconstruction of, 411–13 Investigations of the Solar Spectrum and the Spectra
instability strip, 191 of the Chemical Elements (Kirchhoff), 7, 18
Institute of Theoretical Astronomy, Cambridge ionisation of material inside stars, 44
University, 211, 336 ionisation of the atmosphere, 132–3
integration of equations of stellar structure, 188–92 ionisation zones in pulsating stars, 191
intelligent life in the Universe, 442–4 IRAS faint source survey, 374
intensity of radiation, definition of, 314–15 IRAS galaxies, 360, 374
interference and diffraction, discovery of laws of, 4 correlation between radio and far-infrared
intergalactic medium, physics of, 381 luminosities, 374
Internal Constitution of the Stars, The (Eddington), luminosity function of, 374
42, 47, 51, 191 IRAS point source catalogue, 374
internal structure of the Sun as a gaseous body, 31–3 iron fluorescence line in spectra of Type I Seyfert
International Astronomical Union, 65, 128 galaxies, 287–9
Proceedings of the General Assemblies of, 171 iron-peak elements, formation of, 208–9
Rome General Assembly 1922, 128 iron X-ray emission line, 144
Rome General Assembly 1952, 325 irregular galaxies, 87–8
Sydney General Assembly 2003, 128 Irvine–Michigan–Brookhaven (IMB) neutrino
International Gemini project, 165 experiment, 207–8
International Ultraviolet Explorer (IUE), 152, 172, ‘island universes’, 77, 84–5
209, 280–3, 345, 374 isochrones for stars in globular clusters, 347–8
interplanetary medium, scattering of cosmic rays in, isocurvature modes in models of structure formation,
236–7 402, 407–8
interplanetary scintillation, 193 isothermal gas sphere, 260
discovery of quasars by, 193 isothermal picture of structure formation, 401–5, 407
interstellar chemistry, 220 as a ‘bottom-up’ process, 404
gas-phase reactions, 220 confrontation with observations, 404–5
on surfaces of dust grains, 220 constrained by upper limits to fluctuations in the
Interstellar Chemistry (Duley and Williams), 220 cosmic microwave background radiation, 405
Subject index 533

early formation of stars and heavy elements, 403 and cosmic virial theorem, 359
hierarchical growth of structure, 402, 403–4 epoch of formation of, 392
and Press–Schechter formalism for structure holes and walls in the, 257–9, 355, 360, 377–8
formation, 403–4 isothermal picture of formation of, 401–5
slow growth in pre-recombination era, 401–2 at large redshifts, 386–8
isotopic abundances of the elements, 177 origin of, 392–438
isotropic, homogeneous world models, 107–9, preferred scales in the, 255–6
113–14, 115, 335, 442 ‘sponge-like’ distribution of, 257–9
stringy, cellular structure of, 256–9, 403
James Clerk Maxwell Telescope (JCMT), 162, 375–6 topology of, 257–9
James Webb Space Telescope, 153 Large Scale Structure of Space-Time, The (Hawking
Jan Oort Astronomer (Katgert-Merkelijn and Damen), and Ellis), 276–7
241 Large-Scale Structure of the Universe, The (Peebles),
Japanese National Astronomy Observatory, 162 403
Jeans criterion for gravitational collapse, 228, 393–4 large-scale structure of the Universe 1900–1939,
Jeans’ length, 393 77–121
Jeans’ mass, 399 Large Space Telescope (LST), 152
jet in M87, 269, 274, 284, 294 large telescopes, new technologies for, 165–8
Jet Propulsion Laboratory, 165 computer control of, 165
jets in active galactic nuclei, 279, 293–9 cost–diameter relation for, 165
and the alignment effect, 298–9 Las Campanas Redshift Survey, 258–9
magnetohydrodynamic theory of, 297–8 Laser Interferometry Gravitational-Wave Observatory
relativistic jets in superluminal sources, 316 (LIGO) at Livingston, Louisiana and Hanford,
theories of origin of, 297–8 Washington, 204
Jodrell Bank 76-metre radio telescope (Lovell last scattering surface, 395–6, 435, 443
Telescope), 140 horizon scale on, 417, 439
late-type galaxies, 87
Kamiokande neutrino experiment, 207–8 late-type stars, 39
kaons, charged (K± ) and neutral (K0 ), 135–6 latent image, 9
neutral kaons (K0 ) and CP violation, 398, 448–9 Lawrence Livermore National Laboratory, 192
Kapteyn’s model of the Galaxy, 81, 84–6, 90–1 laws of physics, invariance with time, 324
Kazan University, 101 Learjet for far-infrared astronomy, 160, 162
K-corrections for galaxies, 350 Legacy of George Ellery Hale, The (eds Wright et al.),
Keck I and II 10-metre Telescopes, 165, 286–7, 170
379–80, 386, 428 Leiden Observatory, 217–18
Keeler, James E.: Pioneer American Astrophysicist conditions during Second World War, 217–18
(Osterbrock), 17 Lemaı̂tre world model, 118–19
Kelvin–Helmholtz timescale, 30–1, 53 lenticular galaxies, formation of, 262, 432
Kepler’s laws of planetary motion, 3, 287 ‘Leviathan’ reflecting telescope of Lord Rosse, 12–13
Kerr black holes, 276–7, 303, 306–7 Lick counts of galaxies (Shane Wirtanen counts),
Kerr metric and the Kerr solution, 275–7, 305 255–7
Kew Gardens, 10 statistical analyses of, 255–7
Kirchhoff’s law of emission and absorption of Lick Northern Proper Motion Surveys, 255
radiation, 7, 16, 58 Lick Observatory of University of California at Santa
Kitt Peak National Observatory, 162 Cruz, 13–14, 80, 89, 125, 255–7
Kitt Peak 12-metre millimetre telescope, 160–1 Life of Arthur Stanley Eddington, The (Douglas), 69
Klein–Nishina cross-section, 310 lifetimes of the stars, 63, 69–70, 230–1
Kompaneets equation, 434 light elements (Li, Be, B), 331–4, 336–9
Kramers opacity, 47, 57–60 constraints on cosmological models from the
Kuiper Airborne Observatory of NASA, 160 cosmic abundances of, 331–4
destruction of, in stellar interiors, 332
Laboratori Nazionali di Gran Sasso, 182, 252–3 synthesis in early stage of Friedman models,
Lagrange point, second (L2), 153 331–4, 336–9, 434, 449
lambda particles, 135–6 Light on Dark Matter (ed. Israel), 172, 266
Lane–Emden equation, 32–3, 64 limits of observation, 442–3
Langmuir waves, 433–4 line blanketing, 40
Large Electron–Positron (LEP) collider at CERN, 334, lithium abundance, primordial, 333–4
441 constancy in old population II stars, 334
Large Hadron Collider (LHC) at CERN, 441, 445, 451 lithium depletion problem, 192, 333–4
large number coincidences, 323 Lloyd Triestino, 64
large-scale distribution of matter in the Universe, local distance scale, 347
253–9 Lockheed C-141 transport aircraft, 160
adiabatic picture of formation of, 399–405 Lockheed X-ray astronomy group, 144
534 Subject index

‘Lockman hole’, 372 Main Geophysical Observatory, Petrograd, 107

Lockyer’s theory of stellar evolution, 24–5, 33, 39, 44, main sequence, 35–7, 55–60, 69–70
51 theory of the origin of, 55–60, 188
Lord Rosse’s 72-inch reflecting telescope, 12–13 main-sequence termination point, 188
Lorentz factor, 139, 308–9, 312 as an indicator of the age of the cluster, 188–9
Lorentz force, 195 Malmquist bias, 342, 344
Lorentz transformations, 313–14, 316 Man Discovers the Galaxies (Berendzen et al.), 96
Los Alamos National Laboratory, 189, 300 Manchester University, 138
Los Alamos opacities, 189, 192 marginal appearances, method of, 40
Los Angeles Times, 67 Markarian galaxies, 275, 374
Lowell Observatory’s, 24-inch telescope, 109 ultraviolet–far-infrared correlation for, 374–5
LS coupling, 41 Marseille Observatory, 80-cm reflector at, 13
luminosity classes, 37–8 Maryland, University of, 203
luminosity–colour diagram, 35, 243 maser emission from interstellar molecules, 219–20,
luminosity function for galaxies, 96–7, 110, 246 285–6
in clusters, 110–11 mass absorption coefficient for cosmic rays, 133–4
‘break’ luminosity L∗ , 246 mass conservation, equation of, 31–3, 52, 55
luminosity function for main-sequence stars, 230–1 mass distribution in galaxies, 248–50
luminosity indicators, 36–7 mass loss, 39
luminosity–spectral class diagram, 35–6 mass–energy relation E = mc2 , 47–8
luminosity–volume test for quasars, 367 mass–luminosity ratio for galaxies, 86, 116–17,
lunar occultations at radio wavelengths, 272 248–50, 358–9
Lyman-α clouds, 379–83 used to estimate the distance of M31 (Öpik), 86–7
and the detection of diffuse intergalactic neutral mass–luminosity ratio for clusters of galaxies, 260–1,
hydrogen, 379–81 359
and the relative abundances of the elements, 383–4 mass–luminosity ratio for Universe as a whole, 117,
Lyman-α forest, 370, 381–3, 428–9 260, 358–9
evolution with redshift, 382–3, 428–9 mass–luminosity relation for central regions of active
Lyman-α systems, damped, 381–3 galactic nuclei, 283
associated with progenitors of discs of spiral mass–luminosity relation for main-sequence stars, 39,
galaxies, 381–2 43, 45–7, 70
and the build-up of the chemical elements, 383–4 Massachusetts Institute of Technology (MIT), 142,
and evolution of neutral hydrogen with redshift, 194, 199
383–4 masses of clusters of galaxies, 116–17
and density parameter of neutral hydrogen at large masses of stars, 38–9
redshifts, 383–4 massive compact halo objects (MACHOs), 250–2
zinc abundance in, 383–5 matter–antimatter annihilation, 396–7
Lyman-break galaxies, 384–7 Mauna Kea Observatory, 165–6
cosmic star-formation rate determined from, 386–9 Maury’s classification scheme based on the
properties of, 386–7 appearance of spectral lines, 24–5, 34–5
redshift distribution for, 386–8 Mayall 4-metre telescope at Kitt Peak, 162
technique for detecting at large redshifts, 385–7 Measure of the Universe, The (North), xii, 95, 120–1
Lyman limit absorption line systems, 381–3, 386 Measurement of Starlight, The: Two Centuries of
associated with gaseous haloes of galaxies, 381 Astronomical Photometry (Hearnshaw), xii,
evolution with redshift, 382–3 172
Measuring and Modelling the Universe (ed.
Mach’s principle, 105–7 Freedman), xiv
MACHO project, 250–2 Memory of Henry Norris Russell, In (eds Davis Philip
Magellanic Clouds, 82–3, 163, 224, 250–2 and DeVorkin), 68
Large, 82, 207 mergers of neutron stars, black holes, 304
distance of, 209, 345 mesatrons, 135
supernova 1987A in, 207–10 Messier Catalogue of Nebula, 79, 95
neutral hydrogen in, 218 M87, central rotating ionised disc in, 283–4
Small, 82 Mészáros effect, 402, 407–8
distance of, 82, 96 metal enrichment, 381
variable stars in, 82 metallicity, average as a function of redshift, 383–4,
magic number nuclei, 177–8 390–1
magnetic bubbles, 297–8 metric of isotropic world models, 107
magnetic dipole radiation, 195 metric of steady state cosmology, 324–5
magnetic fields McCrea’s interpretation of, 325
origin of, in supernova remnants, 294 metric perturbations, 417
in pulsars, 195 metric tensor, 105–6
magnetosphere of pulsars, 195 Michelson stellar interferometer, 45–6, 156, 166–7
Subject index 535

microchannel plate detector arrays, 145 nebulae

microquasars, similarity to double extragalactic radio cataloguing of, 77–9
sources, 296 nature of, 19, 77
timescales for evolutionary phenomena in, 296 ‘nebulium’ lines, identification of, 19, 216–17
Mie scattering, 222 negative pressure/energy equation of state, 325, 441,
Mikheyev–Smirnov–Wolfenstein effect (MSW effect), 451
182–3 Netherlands ANS satellite, 144
Milky Way, 77 neutral hydrogen
millimetre and sub-millimetre astronomy absorption line systems, 365
composite bolometers for, 162 column density distribution, 432
development of, 160–2 evolution with redshift, 365, 432–3
germanium bolometers for, 161–2 nature of, according to computer simulations, 432
heterodyne receivers for, 161 clouds, confinement of, 224
Mills Cross radio telescope (Australia), 327–8 in galaxies, 247–8
Milne–McCrea Newtonian models of the Universe, map of Galaxy in, 218
114–17 network of neutral hydrogen clouds in cosmological
Milne’s kinematic cosmology, 323 simulations, 432–3
mirrors, 11–13 rotation curve of the Galaxy determined from HI
grinding and polishing of, 11–12 observations, 218–19
silvering of, 12–13 temperature of interstellar, 224
mixing of the products of nucleosynthesis, 189–90 neutral hydrogen and molecular line astronomy,
MK classification system for stellar spectra, 25–6, 29, 217–22
37–8 neutrino astronomy, 128–30, 170, 207–8
MKK classification system for stellar spectra, 29, 37–8 Neutrino Astrophysics (Bahcall), 211–12
Modern Aristotelianism (Dingle), 324 neutrino decoupling, epoch of, 331–2, 337–8
Modern Cosmology in Retrospect (Bertotti et al.), xii, neutrino fluctuations in the hot dark matter picture,
335–6 406
molecular hydrogen in the interstellar medium, 220, free-streaming of neutrinos from, 406
224 neutrino oscillations, 182–3
correlated with interstellar extinction, 224 neutrino scattering, 181, 338
shocked, 230 neutrino species, limits to numbers of, from
molecular line astronomy, 160–1 primordial nucleosynthesis, 334, 338–9
molecules in the interstellar medium, 218–22 neutrino trapping in supernova explosions, 207–8
bipolar outflows in star-forming regions, 229–30 neutrino-like particles in galaxies and clusters, 252–3
discovery, from radio observations of, 219–21 neutrinos
evidence for, from absorption features in optical as candidates for dark matter, 405–6
waveband, 218–19 density parameter in, if rest mass finite, 405
large, 221–2 discovery of, 179
and measurements of the temperature of the cosmic limits to rest mass of, 208, 406, 435
background radiation, 330–1 from tritium β-decay experiments (Mainz), 435
shielding of, 220 measurement of rest mass of, 405–6
Monte Carlo Markov Chain techniques, 425–7 erroneous, 406
Moon, X-ray fluorescence image of, 373 role in galaxy formation, 405–6
Mount Hamilton (Lick) Observatory, 13–14 from SN1987A, 207–8
Mount Whitney, 155 solar, see solar neutrinos
Mount Wilson Observatory, 14–15, 45, 82, 88, 89, neutron
116, 125–6 discovery of, 48, 66, 131–2, 192
multi-phase interstellar medium, 222–5 finite rest mass of, 182–3
muons, discovery of, 135 postulated existence of, 47, 131
neutron–proton ratio during the epoch of
Nançay Radio Observatory, 140, 330 nucleosynthesis, 337–8
National Academy of Science, Washington, 85 freezing out of, 337–8
National Aeronautics and Space Administration neutron stars, 67, 192–202, 206, 232
(NASA), 142, 152–3 accretion onto poles of, as origin of X-ray emission,
National Radio Astronomy Observatory (US), 199 199
National Research Council of Canada, 220–1 discovery of, as the parent body of pulsars, 67,
National Science Foundation (US), 204 192–5
Nature, 33, 35 equation of state of neutron matter within, 195
founding of, 33 formation of, 208
Nature of the Universe, The (Hoyle), 335 neutrino luminosity, resulting from, 208
Naval Research Laboratory of USA (NRL), 142, 144, high space velocities of, 206
194 internal structures of, 195–6
N-body computer codes, 408–10 as last stable stars, 196–7
536 Subject index

neutron stars (cont.) problems of determining, 376–7, 391

magnetic fields of, 199 redshift distributions at faint magnitudes, 378–9
masses of, 198–200, 201–2, 214 in steady state cosmology, 325
maximum mass of, 68 as a test of the large-scale homogeneity of the
prediction of existence of, 66–8, 192–3 Universe, 87
radius of, 196–7 number counts of IRAS galaxies, 365, 374
thermal X-ray emission of, 192 excess of faint far-infrared sources, 365, 374
in X-ray binary systems, 197–8 number counts of radio sources, 326–8, 373
new astronomies, 125–72 conflict with steady state cosmology, 326–8, 335–6,
New General Catalogue of Nebulae and Clusters of 367
Stars (NGC catalogue) (Dreyer), 79 convergence of, when surveys extend to large
New Technology Telescope (NTT) of ESO, 165, 288 redshifts, 328, 366–7
Newtonian world models, problems with, 105–6 at different frequencies, 367–8
Newton’s law of gravity, 3, 73, 100 different results in northern and southern
Newton’s laws of motion, 3 hemispheres, 327–8
N-galaxies, 274 effects of confusion on, 328
Niels Bohr’s Times, in Physics, Philosophy and Polity evidence for evolution of source population with
(Pais), 29 cosmic epoch, 365–9, 402
nineteenth-century astronomy, legacy of, 3–17 evolution of integrated radio emissivity with
Nineteenth Century Review, 27 redshift, 368
Nobel prize, 201, 211, 212, 241 excess of faint radio sources, 326–8, 366–9
non-baryonic dark matter, 405–7 flattening of, at sub-millijansky flux densities, 374
non-electromagnetic astronomy, 126–8 number counts of stars (Newton), 100
non-Euclidean geometry, 100–1, 103–4 number counts of sub-millimetre sources, 365, 374–6
Non-Euclidean Geometry and the Theory of Parallels large excess of faint sub-millimetre sources, 365,
(Bonola), 120 375–6
non-linear gravitational clustering, 409–10 number counts of X-ray sources, 365, 372–4
non-thermal phenomena in active galactic nuclei, convergence of, 373–4
290–300 excess of faint X-ray sources, 365, 372–4
North Polar Sequence, 82 similarity to radio source and quasar number
novae, 66, 84–6, 96 counts, 374
in the Andromeda Nebula, 86
as distance indicators, 84–6 O stars, photoionisation by, 216–17
two classes of, 66, 86 objective prism studies of stellar spectra, 22–7
Novae and White Dwarf Stars (ed. Shaler), 69 oblateness of the Sun, 184
NRAO 300-foot radio telescope at Greenbank, 140, Observatory, The, 44
171 observing distant galaxies in the infrared
nuclear astrophysics, 211 waveband, 362–4
nuclear deflagration in Type 1a supernovae, 206 Ockham’s razor, 427–8
nuclear emulsions, 136 100-inch Hooker telescope at Mount Wilson
nuclear energy generation in stars, 277, 307, 434 Observatory, 14–15, 45–6, 84, 87, 88, 110,
region of, 189–90 112, 125, 156, 244, 365
nuclear reaction rates, 175 Oort limit, 81
nuclear transmutations, discovery of, 48 Oort’s A and B constants, 90–1, 97–9
nucleocosmochronology, 347–8 opacity of the Earth’s atmosphere, 126–7, 142
nucleosynthesis and the origin of the chemical opacity of stellar matter, 43–4, 47, 49–50, 54–5,
elements, 175–8, 319, 325, 329, 406 56–8, 175, 189, 191–2, 211
primordial, 319–23 impact of new opacities from the OPAL and OP
nucleosynthetic processes in stars, 176–8, 325 projects, 192
rapid (r) process, 178, 348 Opacity Project (OP), 191–2
slow (s) process, 178 OPAL (opacity) project, 191–2, 211
Nuffield workshop, 1982, 447 physics involved in the computations for, 211
number counts of galaxies, 87, 96–7, 112–14, 255, and the theory of stellar pulsation, 191
325, 342, 354–6, 362, 365, 376–9 OPAL collaboration (CERN), 336
as a cosmological test, 354–6 open cold dark matter model, 413–15
of different morphological types, 377, 379–80 open clusters and measurements of stellar extinction,
excess of faint blue galaxies in, 355–6, 365, 377–9 91
excess of irregular/peculiar/merger systems at faint opening up of electromagnetic spectrum, 125–72
magnitudes, 379–80 optical astronomy in the age of the new
fluctuations in, 355 astronomies, 162–70
and galactic evolution, 356 optical identification of radio sources, 138, 328
in the infrared waveband, 377–9 and discovery of quasars, 328
predicted Euclidean, 96–7, 112, 326–8, 366 at millijanksy flux densities, 374
Subject index 537

optical interferometry, 45, 166–8 photoelectric effect, 163

used to measure diameter of Betelgeuse, 45 photoelectric photometry, 163
orbital migration, 234 photographic plates, 9–11
Orbiting Astrophysical Observatories (OAO), 151–2 development of, 9–11
OAO-III (Copernicus), 151, 224 hypersensitising, 11, 17
Orbiting Solar Observatory III (OSO-III), 147 photographic terms, invention of, 9
origin of galaxies and the large-scale structure of photography, invention of, 4, 9–11
the Universe, 392–438 photoionsation models of active galactic nuclei,
Origin of Species, The (Darwin), 19 279–80, 284
Original Theory or New Hypothesis of the Universe cross-correlation analyses for, 281–3
(Wright), 77 evidence from correlated continuum-line
oscillations of the diameter of the Sun, 184 variability, 280–3
Owens Valley Radio Observatory (OVRO) millimetre time delay between continuum and line variability,
array, 264 280–3
Oxford University, honorary degree for Annie photoionisation of interstellar gas, 216–17
Cannon, 27 photon barrier, 395
oxygen burning, 176, 178 photon to baryon ratio of intergalactic gas, 434, 440
conservation in expanding Universe, 440
Paine Fund, 22 physics and astronomy, interaction between, 3, 183
Palomar 18-inch Schmidt telescope, 204, 246 physics beyond the standard model, 182, 451
Palomar 48-inch Schmidt telescope, 169, 205, 254 Pic du Midi Observatory, 135
Palomar 60-inch telescope, 235 Pickering series, 24–5, 28–9
Palomar Sky Atlas, 169, 244, 254 ‘Picturephone’, 164
Palomar 200-inch telescope, 15, 88, 118, 125–6, 129, pions, 136, 147, 236
158, 165, 169–70, 235, 272–3, 340–2, 344, 371 discovery of, 136
technical innovations in construction of, 170 decay of neutral, as source of γ -rays, 147, 236
used as a transit instrument, 371 Planck era, 398, 445, 449–51
‘pancakes’, 402–3, 432 and quantum gravity, 449–50
burning of, 403, 432
Planck project (ESA), 416, 420, 422, 443
paraboloid–hyperboloid reflectors for X-ray
Planck units, 398, 434
astronomy, 145–6
Planck mass, 451
parallaxes, stellar, 7–9, 34, 36–7, 343–4
planetary nebulae, 23, 25, 77, 216
first trigonometric, 8
photoionisation of, 216–17
photographic, 35
planets, extrasolar, see extrasolar planets
photometric, 7–8, 16
theories of formation of, 234
spectroscopic, 37
trigonometric, 8–9, 343–4 plasma oscillations, discovery of dispersion relation
Paris Symposium on Radio Astronomy for, 433–4
(ed. Bracewell), 328 plate holder, adjustable, 13
Parkes 64-metre radio telescope, 140, 272 Plumian Professor of Astronomy, 42
Parseval’s theorem, 434 Poisson’s equation for gravity, 105–6, 360
particle horizons, 417–18, 439, 444 polonium, discovery of, 130
particle physicists, their involvement in studies of the polycyclic aromatic hydrocarbons (PAHs), 221–2
early Universe, 335, 407 and infrared emission features, 221–2
particle physics and the early Universe, 444–7 stretching and bending modes of, 221–2
passive evolution of spectral energy distribution of Polytechnical School, Aachen, 32
galaxies, 363–4 polytropic stellar models, 32–3, 242–3
Pauli’s exclusion principle, 64 Popular Astronomy, 35
P(D) technique for deriving shape of number counts positional astronomy, 3
of radio sources, 328, 368 positron, discovery of, 48, 135
of X-ray sources, 372–3 post-main-sequence evolution of stars, 189–91
peculiar galaxies, 244–6 post-recombination Universe, 427–33
peculiar stars, 25–7 effects of dissipation, 429–33
perchloroethylene, 180 epoch of reionisation, 427–9
perfect cosmological principle, 324 potassium uranyl disulphide, 130
perihelion shift of the orbit of Mercury, 104 Potsdam Astrophysical Observatory, 35
perturbations, primordial, 393–4 power spectrum of density fluctuations, 434
amplitude on the horizon scale, 394, 441–2 preferred mass scales in mass spectrum on last
photinos as cold dark matter particles, 406–7 scattering surface, 400–1
photoelectric absorption pre-main-sequence stars, 225–30
in interstellar medium, 224–5, 372 evolutionary tracks for, 226–7
in X-ray fluorescence process, 288 as fully convective stars, 226–7
in X-ray waveband in stars, 43 Hayashi track for, 226–7, 242–3
538 Subject index

Press–Schechter formalism for structure formation, pulsars, discovery and properties of, 192–7, 211
403–4, 407, 434 binary, and test of general relativity, 201–2
Press–Schechter mass function, 403–4 and the Cambridge 4.5 acre pulsar array, 193
Pretoria Observatory, 163 coherent radio emission of, 196
‘primaeval atom’ (Lemaı̂tre), 320 discovery record of, 194
primordial abundances of the light elements, glitches in slow-down rate, 196
336–9 invention of name, 211
primordial black holes, evaporation of, 396–7 as magnetised rotating neutron stars, 193
primordial gravitational waves, 418–19 optical pulses from Crab Nebula pulsar, 194
spectral index of, 418 origin of radio emission, 193
primordial nucleosynthesis, 319–23, 332–9 planets orbiting, 232
constraints on baryonic density parameter, 336–9, prediction of detectability at long radio
404–5 wavelengths, 192–3
constraints on dynamics of Universe during, radio emission, mechanism of, 196
334–5 speed-up due to radiation of magnetic dipole
electron–positron pair production in, 322–3 radiation, 194–5
equilibrium abundances at high temperatures, as stable clocks, 201–2
319–20, 322 star-quakes in, 196
importance of weak interactions in maintaining, and supernova remnants, 193–4
322–3 and tests of general relativity, 201–3
overproduction of light elements in, 319–20 X-ray pulses from Crab Nebula pulsar, 194
formation of helium and the light elements by, pulsating stars, theory of, 191–2
329, 332–5, 336–9 damping of oscillations according to, 191
‘heavy element catastrophe’, 320 mechanisms for, 191
network of nuclear reactions involved in, 332–3 numerical computations of, 191
neutron–proton ratio from, 323, 334 pyrheliometer, 154–5
problem of lack of stable nuclei with mass numbers
5 and 8, 321–2 QSO Absorption Lines (ed. Meylan), 391
results of the analyses of Wagoner, Fowler and QSO Absorption Lines: Probing the Universe
Hoyle, 332–5 (ed. Blades et al.), 391
primordial perturbations, 399–402 quadrupole moments in gravity and electromagnetism,
evolution of, on different physical scales, 213–15
399–402 quadrupole power spectrum of gravitational waves,
finite amplitude on horizon scale, 400 418
Harrison–Zeldovich spectrum of, 401 quantisation of angular momentum, 28–9
origin of spectrum of, 447–8 quantised vortices in magnetic field inside neutron
during phase transitions in inflationary picture, stars, 196
447 quantum fluctuations in the very early Universe, 418,
from quantum fluctuations in the early Universe, 447
447–8 and adiabatic curvature perturbations, 447
power spectrum of, 401 and primordial gravitational waves, 418
Silk damping of adiabatic, 399 quantum mechanical tunnelling, Gamow’s theory of,
Principe, island of, 104–5 48–9
Principia Mathematica (Newton), 120 quantum mechanics, impact upon stellar structure,
Principles of Physical Cosmology (Peebles), 266, 336, 48–50
434 quantum theories of gravity, 445, 449–51
Problems of Extragalactic Research (Hoyle lecture quasars (originally meaning quasi-stellar radio
course), 328–9 sources), 144, 271–5, 279–83
projection effects in active galactic nuclei, 299–300 convergence of comoving density at large redshifts,
in Seyfert 1 and Seyfert 2 galaxies, 299 370–2
proper motions in arms of spiral galaxies, 85, 96 discovery of, 271–5
proper motions of stars, 34, 36, 79–80 early history of, 391
‘proto-hydrogen’, 24, 28 local theory of, 273–4
‘proto-metals’, 24 Lyman-α absorbers in the spectra of, 379–83
proton decay experiments, 207–8 microquasars and, 296
proton–proton (pp) chain, 49–50, 57–60, 62, 179 models for, 277–9
protostars, 158, 226–30 optical spectra of, 272–3, 279–80
definition of, 227 radio-quiet, 368–72
dust discs about, 228–31 spectroscopy of, 279–83
far-infrared luminosities of, 226–7 strong evolution with cosmic epoch, 365–74
infrared, 226–30 variability of, 273
spectra of, 228–9 Quasi-Stellar Objects (Burbidge and Burbidge), 304,
Pulsar Astronomy (Lyne and Graham Smith), 212 391
Subject index 539

Quasi-Stellar Sources and Gravitational Collapse Öpik’s solution to the, 61–2, 188
(eds. Robinson et al.), 304 problems of, 61–3, 68, 189
Queens’ College, Cambridge, 78 redshift, 109, 112, 121
‘queue mode’ observing, 166 definition of, 109
of quasars, nature of, 273–4
Radcliffe College, 82 redshift–apparent magnitude relation
radial comoving distance coordinate, 114–15 for brightest galaxies in clusters, 340–2, 343,
radial velocity programmes for stars, 89–91 349–50
radiation-dominated universe of Alpher and Herman, Bautz–Morgan effect and, 349
320–3 evolution of stellar populations and, 350–2
radiation loss rate, of accelerated electron, 212–13 problems in determining, 349–50
of relativistic electron, 139–40 for galaxies, 326, 340, 362–4
radiation pressure and the Eddington limit, 283, 307 in the infrared waveband, 362–4
radiative transport of energy in stars, 39, 43, 45, 53–5 history of determinations of (Sandage), 361–2
applied to giant star envelopes by Eddington, 39, 45 infrared (K), for radio galaxies, 350–2
radio astronomy, origin and development of, for Type IA supernovae, 352–5
136–42, 269–71 Redshift Controversy, The (eds. Field et al.), 305
discovery of extraterrestrial radio emission, 136–7 reduced masses, 28
first discrete radio sources, 137–9 Rees–Sciama effect, 435
key discoveries of, 142 Reference Catalogue of Bright Galaxies
radio background emission, 367 (de Vaucouleurs et al.), 244
radio galaxies, 138, 144, 270–5, 350–2, 362 reflecting telescopes, 12–15
redshift–apparent magnitude relation for, 350–2 reflection nebulae, 221
3CR, properties of, 350 refracting telescopes, 12
6C, comparison with 3CR radio galaxies, 352 refractive indices of glasses, Fraunhofer’s
radio interferometry, 138, 140–2 measurements of, 4–6
radio jets, 262 regions of ionised gas
radio-quiet quasars, 274–5, 368–72 determination of properties of, 217
Braccesi survey for, 369–70 temperature of, 434
convergence of number counts of, 370–2 reionisation, epoch of, 427–9, 432–3
cosmological evolution of, 368–72 relativistic beaming, 311–16
described by luminosity evolution, 372 aberration effects, 314–15
decrease of comoving space density at large example of spaceship travelling at, 0.8c, 312–15
redshifts, 370–2 time dilation/interval effects, 313–16
dispersion prism/grating techniques for discovering, waveband effects, 313–15
370–1 relativistic degenerate gas, 64–5
methods of discovering, 368–72, 391, 428 relativistic gravity, theories of, 201–2
ultraviolet and infrared excess, 369 relativistic magnetohydrodynamics, 297–8
multicolour techniques for discovering large relativistic motions in nuclei of active galaxies, 291,
redshift, 370–2 306, 310
Schmidt–Green survey for bright, 370 relativistic plasmas in the Universe, 140
steep number counts for, 370–2 relativistic quantum mechanics, 324
‘radio stars’, 139 relativistic sound speed, 399, 419
radio telescopes, 136–8, 140–2, 271–2, 291 relativity, history of, 102, 120
radio trail radio sources, 262 relics of the early phases of the Big Bang, 334–5
radioactivity, discovery of, 130–1 reverberation mapping of active galactic nuclei, 280–3
radium, discovery of, 130–1 Revised Harvard Photometry, 22, 34
ram pressure, stripping of galaxies in clusters, 262 revision of extragalactic distance scale, 325–6
and formation of radio trail sources, 262 Rice University Chemistry Group, 221
RAND project of the US Air Force, 151 Rice University γ -ray group, 148
ranges of α-, β- and γ -rays, 131 Ricci tensor, 105
Rayleigh criterion, 140, 152 Riemannian geometries and space-time, 103–5
Rayleigh scattering, 216, 299 Rindler notation for four-vectors, 313, 316
Rayleigh–Taylor instability, 294 ‘ring of pearls’ in SN 1987A, 210
RCA Laboratories, 163 Ritchey–Chrétien optical design, 15
Realm of the Nebulae, The (Hubble), 96, 110, 365 Ritter’s theory of stellar evolution, 32–3
recession of the nebulae, 109–12 Robertson–Walker metric, 113–14, 335, 417
recombination, theory of, 217 Roche lobe, overflow, 199
reconstructing the initial power spectrum, 411–13 Rockefeller Foundation, 88
reconstruction of local velocity and density Roman College Observatory, 24-cm refractor at, 20
distribution, 360 ROSAT (Röntgen Satellite) X-ray Observatory, 147,
red giants, 61–3, 68, 176, 205, 211, 221, 351, 363–4 171, 260–1, 365, 372–3
formation of, 189–90, 211 Rosseland mean opacity, 47–8, 58
540 Subject index

rotation curve, of galaxies, 97, 116, 248–50 self-calibration techniques, 167

our Galaxy, 218, 250 self-shielding of neutral hydrogen, 433
of nuclear regions of M106, 285–6 Serendipitous Discoveries in Radio Astronomy
rotational speed of Sun about the centre of the Galaxy, (Kellerman and Shields), 171, 211
89–91 Serrurier trusses, 170
rotational transitions of interstellar modecules, 220 Seyfert galaxies, 274–5
Royal Astronomical Society, meetings of, 43–4, 65 objective prism surveys for, 279
Royal Greenwich Observatory, 42 spectra of their active nuclei, 274, 279–83
Royal Observatory at the Cape of Good Hope, 79, 163 Type 1 and Type 2, 279
RR-Lyrae variable stars, 191–2 variability of continuum and line emission, 274
‘Russell diagram’, 36–7, 63 X-ray fluorescence emission from nuclei of, 287–9
Russell’s picture of stellar evolution, 39, 42, 44–5, 61 Shapley’s model of the Galaxy, 83–6, 89–90
Russell–Saunders coupling (LS coupling), 41 Siding Spring Observatory, Australia, 370
Rydberg constants for hydrogen and singly ionised silicon burning, 178
helium, 28–9 Silk damping, 399
Silliman lectures
Sachs–Wolfe effect, 418, 435 at Harvard, 215
dependence upon spectral index of perturbation at Yale University, 110
power spectrum, 418 single-line spectroscopic binaries, 199
integrated, 435 single-particle hydrodynamics, 432–3
physics of, 435 singularity theorems
temperature fluctuations due to, 418 and solutions of Einstein’s field equations, 275–7
Sacramento Peak Solar Observatory, 162 for the Universe as a whole, 276–7, 451
SAGE experiment for solar neutrinos, 182 SIS detectors, 161
Saha equation, 40–1 60-inch telescope at Mount Wilson Observatory, 14,
Sakharov oscillations on last scattering surface, 84, 125, 244
400–1 Skylab space mission, 145
discovery of, 422–4 sling-shot mechanism, 234
Saltpeter initial mass function for stars, 230–2, 364, Sloan Digital Sky Survey (SDSS), 169–70, 235, 259,
386 423–7, 428
Sandage and the values of H0 and q0 , 340–2 Small Astronomical Satellite-2 (SAS-2), 147
scalar (density) perturbations, 418–19 Sobral, island off coast of Northern Brazil, 104–5
scalar fields, 441, 445 sodium D-lines, 6, 24
scale factor, 108, 114–17, 321, 325, 392–4 in atmosphere of exoplanet of HD 209458, 234–5
relation to redshift, 114, 121 Solar and Heliospheric Observatory (SOHO) of ESA,
scattering of electrons in the vicinity of strong shocks, 184–8
294 solar constant, 154–5
Schechter luminosity function for galaxies, 246–7 Solar Maximum Mission of NASA, 209
Schmidt telescopes, 168–9 solar motion, 79–80, 110–11
IIIaJ emulsions for, 169 solar neutrinos, 179–83, 187, 211
Schönberg–Chandrasekhar limit, 62–3, 68, 69, 188 beryllium-8 neutrinos, 179–82
Schottky barrier diodes, 161 chlorine detector for (Davis experiment), 179–81
Schwarzschild black holes, 306–7 experiments to detect, 180–2
Schwarzschild metric, 275–6, 305 gallium experiments for, 182
Schwarzschild radius, 68, 198, 275–7, 287, 306–7 Kamiokande II experiment for, 181
scintillation of radio sources at long wavelengths, 193 low-energy, 179, 182
interplanetary, 193 nuclear reactions involved in creating, 179
due to irregularities in the electron density in predicted emission rates, 180–2
ionosphere, 193 problem of, 180–2
SCUBA, see submillimetre common-user bolometer super-Kamiokande experiment for, 182–3
array solar oscillations, see helioseismology
SCUBA sources, very large redshifts of, 376 Solar Physics Observatory, South Kensington,
Secchi stellar spectral classification scheme, 20–1, 23 London, 33
second Cambridge catalogue of radio sources (2C), solar spectrum, 4–7
326–8 Becquerel’s first daguerreotype of the, 9
effects of confusion on flux densities of sources in, elements responsible for lines in the, 7
328 Fraunhofer lines in, 5–6
second dredge-up, 190 Fraunhofer’s observations of, 4–6
Secret Man, The (Hey), 171 infrared lines in the, 155–6
seeing, astronomical, 152, 166–7 naming of lines in the, 5
selected areas of Kapteyn, 81 Newton’s near miss of discovering lines in the, 16
Selected Works of Yakov Borisevich Zeldovich, Vol. 2 Rutherfurd’s 3-metre photograph of the, 11
(eds Ostriker et al.), 336 Wollaston’s discovery of five lines in the, 4, 16
Subject index 541

solar wind, velocity of, 193 standard model of elementary particles, 398
Solvay conference of 1911, 103 star-burst galaxies, 279, 374–6
Soudan Underground Laboratory, Minnesota, USA, absence in deep optical surveys because of dust
252–3 extinction, 375
sound horizon for perturbations on the last scattering flux density–redshift relation in submillimetre
surface, 419–20 waveband, 375–6
Source Book in Astronomy and Astrophysics, A, intense far-infrared emission of, 374–5
1900–1975 (eds Lang and Gingerich), xii predicted number counts in submillimetre
source counts of . . . , see number counts of . . . waveband, 375–6
integral and differential, 366 star clusters, 35
for standard world models, 365–6 star formation, 225–32, 430
Soviet Union hydrogen bomb programme, 434 bipolar outflows associated with, 229–30
space distribution of stars of different spectral types conditions for, in nearby spiral galaxies, 382
(Pickering), 26 Jeans’ length, 393
Space Infrared Telescope Facility (SIRTF) of NASA, problems of, 230–1
see Spitzer Space Telescope role of dust in, 226–30
Space Telescope: A Study of NASA, Science, standard picture of (Shu et al.), 228–30
Technology and Politics (Smith), 172 theory of, 228–9
Space Telescope Science Institute (STScI), 153 star-formation history of the Universe, 162, 362–4,
Space Transportation System of NASA (Space 365, 375–6, 387–91
Shuttle), 145, 152–3 absorption and emission histories of, 389–91
spallation and formation of light elements in cosmic relations between, 390–1
rays, 238–40 from observations in the ultraviolet and far-infrared
spark chamber detectors for γ -rays, 147–8 wavebands, 375–6
spark spectrum, 7, 16, 24 star-forming galaxies, 162, 384–7
special theory of relativity, 113–14 at large redshifts, 379, 384–7
speckle observations of the nucleus of the Galaxy, obscured, at large redshifts, 390–1
286–7 at present epoch, 388
spectra of nebulae, Huggins’ observations of, 19 signatures of, at large redshifts, 385–7
spectral classification, 20–9, 35–6 spectra of, 384–7
history of, 20–9 stars
spectral energy distribution of giant elliptical galaxy, first generations of (population III stars), 428–9
362–4 luminosity function of, 81
evolution of, 363–4 radial motions of, 80–1
spectral synthesis codes for galaxies, 363–4, 385–7 stars and stellar evolution from 1945, 175–215
spectropolarimetry of active galactic nuclei, 299 Stars, The: Their Structure and Evolution (Tayler), 51,
spectroscope, invention of, 5–6 58–60
spectroscopy of active galactic nuclei, 279–83 statistical methods in astronomy, 78
speculum metal, 13 steady state cosmology, 323–6, 335, 342, 358
speed of light continuous creation of matter, 324–5
Rømer’s measurement of, 3–4 dynamics of world model, 324–5
Foucault’s experiment to measure the, 13 perfect cosmological principle, 324
speed of sound as a function of cosmic epoch, 398–9, Steady State Cosmology Revisited (Hoyle), 335
419 Stefan–Boltzmann law, 54, 59, 156
for adiabatic sound waves, 398, 400 Stellar Astronomy (Hoskin), 16, 95, 199
in matter and radiation-dominated eras, 398–9 stellar astrophysics, origin of, 18–29
spherical aberration problem of HST, 153 stellar atmospheres, 40–2
spherical harmonic transforms, 435–6 ionisation state of, 40
spin stabilisation of rocket-borne experiments, 143 Stellar Atmospheres (Payne), 40–1
spin-up of binary pulsar, 201–3, 214 stellar evolution and the spectral evolution of elliptical
spiral arms of galaxies galaxies, 362–4
discovery of, 12, 92–3 stellar luminosities, 34–9
spectra of, 86 stellar masses, 38–9
spiral galaxies, 87–8, 109, 432 stellar photospheres, 345
dark matter in, 248–50 stellar populations, 92–4, 325–6
light distribution in, 249–50 discovery of, by Baade, 92–3
Slipher’s measurements of the recession velocities properties of different, 93–4, 177
of, 109–10 stellar spectra, classification of, 18–29
stability of the discs of, 250 stellar structure and evolution, 30–60, 188–92
stability of thin rotating gaseous discs in, 382 early theories of, 30–3, 51
spiral nebulae, nature of, 79, 84–7, 109 Eddington’s theory of, 42–8, 188
Spitzer Space Telescope, 160 end points of, 61–73
Sputniks 1 and 2, 142 equations of, 44
542 Subject index

stellar structure and evolution (cont.) supermassive stars, 273, 336

impact of the new physics, 39–42 nucleosynthesis in ‘bouncing’, 336
impact of quantum mechanics on the theory of, Supernova Cosmology Project, 353–5
48–50 supernova 1987A, 207–10
from main sequence to the giant branch, 188–91 bipolar structures observed in, 209
pre-main-sequence stars, 225–30 blast wave from, exciting ring of emission about,
Stellar Structure and Evolution (Kippenhahn and 209–10
Weigert), 32–3, 51, 58, 60, 62–3, 68, 73, discovery of, 207
189–90, 211, 226, 241–3 dust formation in, 209
Stellar Structure and Stellar Atmospheres (Mestel), 51 evidence for radioactive products of explosion of,
stellar temperatures, 39–40 208–9
problem of measuring, 39–40 γ -ray observations of, 208–9
strange particles, 135–6 infrared observations of, 208–9
string theories, 450 light curve of, 207–8
Strömgren spheres, 217 ‘light-echoes’ observed in, 209
strong electromagnetic waves, 195 pre-history of, 207, 209
Subaru 8-metre telescope, 165 mass loss from progenitor red giant, 209
sub-atomic particles, discovery of, 130–6 properties of, 207–10
submillimetre common-user bolometer array radio emission of, 210
(SCUBA), 162, 375–6, 390–1 ring of emission about, 209–10, 345
submillimetre windows for astronomy, 161 Supernova 1987A and Other Supernovae
submillimetre waveband, molecular transitions in, (eds Danziger and Kjär), 212
161 Supernova 1987A in the Large Magellanic Cloud
substratum, 113 (eds Kafatos and Michalitsionios), 212
Subtle is the Lord . . . the Science and Life of Albert supernova remnants, 138, 144–6, 193–5, 225
Einstein (Pais), 102, 104, 120 acceleration of particles in the shock waves of, 294
Sudbury Neutrino Observatory (SNO), 183 and heating of interstellar medium, 225
Sun old, and Galactic soft X-ray emission, 225
age of, 69–70
optical and radio properties of, 269–71
depth of convection zone of, 183, 186, 192
polarised optical and radio emission, 269–70
discovery of hot corona of, 142
radio spectra of, 270
heat flux of, 154–5
X-ray emission of, 144–6, 225
infrared radiation of, 153–5
supernovae, 66–7, 86, 204–10
motion relative to nearby stars, 79–80
core collapse of, 303
Sun and Earth (Friedman), 171
exponential decline of luminosity due to radioactive
sunspot records from Kew Observatory, 10
decay, 206–7
sunspots, radio emission from, 137–8
Sunyaev–Zeldovich effect in clusters of galaxies, frequency of, 66
263–4, 345, 356–7 light curves of, 205, 206
and distance measurements, 345 radioactive origin of decline of, 208–9
independent of redshift of cluster, 263–4 mass ejected in, 205
kinetic, 264 resulting from collapse of core of massive stars, 205
superclusters of galaxies, 255–9, 359 as sites of heavy element synthesis, 178, 205
average densities of, 392 of Type I and II, 205–7
infall of galaxies into, 359 Supernovae and Supernova Remnants (ed. Battali
superconductivity in neutron stars, 195–6 Cosmovici), 212
superfluidity in neutron stars, 195–6 supernovae, Type Ia
supergiant stars, 37, 207 correlation of luminosity with decay rate, 206,
super-horizon perturbations, 417–18 352–3
superluminal motions in active galactic nuclei, 291, and cosmological time-dilation, 353
294–7, 305 deflagration fronts in, 352
aberration effects associated with, 295 evidence for non-zero cosmological constant, 352–5
maximum speed of, 295, 311–12 as exploding, accreting white dwarfs, 206, 352
one-sidedness of, 295, 315–16 luminosity–width relation, 206, 352–3
relativistic ballistic model for, 311–16 observations at large redshifts, 352–4
jets, 316 redshift–apparent magnitude relation for, 352–4
ratio of component flux densities, 315–16 as standard candles for cosmology, 206, 352, 427
and relativistic beaming, 311–16 techniques for observing, 352–3
theory of, 295–7 uniformity of properties of, 206, 352
and unified schemes for active galactic nuclei, supersymmetric particles, 252–3, 406, 451
299–300 as cold dark matter particles, 406
Superluminal Radio Sources (eds Zensus and survey astronomy, 168–70
Pearson), 306 Sussex University, 220
Subject index 543

Swedish–European Southern Observatory Telescope matter-dominated era, 395–8

(SEST), 162 neutrino barrier in the early Universe, 397
Sydney, CSIRO Radiophysics Division, 138 primordial nucleosynthesis, epoch of, 397
symmetry breaking in Guth’s original inflationary radiation-dominated era, 396–7
model, 445 ratio of radiation to matter energy densities, 396
synchro-Compton radiation of ultra-high-energy recombination through two-photon emission, 394
electrons, 150, 291, 308–10 reheating of the intergalactic gas, 397
synchronisation of clocks in world models, 113, 417 speed of sound, variation with cosmic epoch, 394
synchronous time coordinate, 417 summary of, 395
synchrotron radiation, 195–6, 269–71, 291, 301–2, temperature change with scale factor R, 394–6
304, 308, 314, 316 Thomson scattering, opacity due to, 395–6
discovery of, 139–40 very early cosmological epochs, 397–8
equipartition of energy between particles and thermal instabilities, 403, 430–3
magnetic fields, 304 and galaxy formation, 431–2
Galactic, as evidence of interstellar flux of thermal timescale for stars, 53
high-energy electrons, 237–8 thermocouple, 155
lifetimes of relativistic electrons due to the thermodynamics, laws of, 30, 50
emission of, 293 thermoelectricity, 155
minimum energy requirements for, 270–1, 304 thermopile, 155
optical, in quasar spectra, 279–80 first astronomical observations with, 155
optical and X-ray, from the Crab Nebula, 195 Third Cambridge Catalogue of Radio Sources (3C),
self-absorption, 308–9, 311 304, 328
spectrum of, 139–40, 304 revision of (3CR), 304, 328, 350–2
System of the Stars, The (Clerke), 16 Third Reference Catalogue of Bright Galaxies
(de Vaucouleurs et al.), 96
TAMA Japanese gravitational wave observatory, 204 36-inch Lick telescope, 125
Tartu Observatory, Estonia, 266–7 Thomson scattering, 57–60, 215, 299
Tata Institute 18-mirror Cherenkov array, 149 cross-section, 43, 57–60, 215, 297, 307, 310
Technische Hochschule, Munich, 32 in last scattering layer, 422
technological advance and computation, 128 optical depth for, through the reionisation epochs
telescope design, 10–15 (τ ), 425–9
improvements during the 19th century in, 10, and reionisation of the intergalactic gas, 427–9
11–15 thorium, discovery of, 130
John Draper’s contributions to, 11–12 3CR radio galaxies
telluric bands in the solar spectrum, 155 alignment effect in, 298–9
‘temperature arch’ of Lockyer, 33, 51 unified schemes for, 299–300
temperature–frequency relation for black-body Three Hundred Years of Gravitation (eds Hawking and
radiation, 126–7 Israel), 305
tensor (gravitational wave) modes, 418–19 time-delay test of general relativity (Shapiro), 120–1
Texas Instruments, 165 time dilation in a gravitational field, 102–3
Theoretical Concepts in Physics (Longair), x, 16, 50, topological defects in the early Universe, 419, 448
120–1, 212, 213, 305, 335 torus, accretion about active galactic nuclei, 279
Theory of Light and Colours, On the (Young), 4 and formation of jets, 279
Theory of Stellar Pulsation, The (Cox), 211 obscuring, in active galactic nuclei, 299–300
thermal conduction in stars, 54 transfer functions for the spectrum of initial
thermal history of the Universe, 320–3, 329, 337–8, perturbations, 407–8, 412
394–8 transparency of the Earth’s atmosphere, 126–7, 142,
Alpher and Herman’s prediction of microwave 160–1, 331
background radiation at temperature of 5 K, for submillimetre waves, 161
320–1 tri-chromatic theory of colour vision of Young, 4
baryon–antibaryon pair production, 397–8 Trinity College, Cambridge, 100
baryon asymmetry problem, 397–8 triple-α reaction, 176, 325
Compton scattering during pre-recombination era, Hoyle’s prediction of resonance associated with,
396–7 176, 325
coupling of matter and radiation, 396–7, 399 T-Tauri stars, 225–6
dissociation of light elements in the early Universe, as low-mass stars in the process of formation, 225
397 Tully–Fisher relation for spiral galaxies, 247–8, 344
electron–positron pair-production in early Universe, infrared, 248
397 ‘tuning-fork’ diagram for classification of galaxies,
epoch of equality of matter and radiation energy 87–8
densities, 396, 399–400 turbulent motions in supernova remnants, 294
epoch of recombination, 394–6, 399 Twentieth Century Physics (eds Brown et al.), x, xiv,
last scattering surface, 394–6 120, 215
544 Subject index

24-cm Dorpat Telescope for Tartu Observatory, 6 velocity dispersion of galaxies in general field, 411
21-cm line of neutral hydrogen, 217–19, 241 velocity–distance relation for galaxies, 109–12,
observed in absorption against bright radio sources, 113–14, 116
218 Hubble’s discovery of, 109–12, 113
prediction and discovery of, 217–18, 241 velocity ellipsoid, 80–1, 89, 283
‘spiral features’ in Galactic distribution of, 218–19 very early Universe, 439–52
transition probability of, 218 Very Large Array of NRAO (VLA), 141, 291, 298
two-degree field (2dF) of the Anglo-Australian Very Large Telescope (VLT) of ESO, 129, 165–6, 288
Telescope, 169 Very Long Baseline Array (VLBA), 285
multi-object spectrograph for, 169 very long baseline interferometry (VLBI), 141–2,
2dF Galaxy Redshift Survey, 169, 259, 361, 372, 167, 285, 294–5, 309–10
427 Vidicon System of the Westinghouse Corporation, 164
2dF quasar survey, 371–2 violent events in the nuclei of galaxies, 274
Two-micron All Sky Survey (2MASS), 235 violent relaxation, 263–4
Two-micron Sky Survey (Neugebauer and Leighton), VIRGO consortium supercomputer simulations of
158 large-scale structure formation, 413–15
two-photon process of recombination, 394 VIRGO French–Italian gravitational wave
two-point correlation functions for galaxies, 255–7, observatory, 204
266, 359, 411–12, 423–7 virial theorem for star and galaxy clusters, 116–17,
and determination of cosmological parameters, 260
423–7 virial theorem for stars, 32, 51–3, 55–6, 72
scaling test using, 255–6, 266 viscosity, role in accretion discs, 278
Type Ia supernovae, see supernovae, Type Ia viscous parameter, α, in accretion discs, 278
visual photometry, Pickerings’s observations, 22
UHURU X-ray Observatory, 144–5, 197–9, 212, 225, vortices, quantised, in magnetic field inside neutron
260, 372 stars, 196
UHURU X-ray source catalogue, 144–5 unpinning of, 196
UK Ariel V satellite, 144 V2 rockets, 138, 142
UK 48-inch Schmidt telescope, 169, 254, 370 V/Vmax test for quasars, 367
UK Infrared Telescope (UKIRT), 158, 162, 298
ultra-high-energy gamma-ray astronomy, 149–50, W± bosons, 441
171–2, 297 Washburn Observatory, Wisconsin, 35
ultra-high-energy particle experiments, 443 water vapour molecules, discovery in interstellar
ultraviolet astronomy, history and development of, medium, 219
150–2 high-velocity, associated with bipolar outflows, 230
ultraviolet emission of Sun, 142 in nuclear regions of NGC 4258 (M106), 285–6
ultraviolet spectroscopy, 151–2 weak interactions, theory of, 441
discovery of hot component of interstellar medium weakly interacting massive particles (WIMPs), 250,
by, 152 252–3
of resonance transitions of common elements, laboratory searches for, 252–3
151–2 limits to masses of in the halos of galaxies, 252
ultra-weakly interacting particles as cold dark matter Wellesley College, 25
particles, 406–7, 451 Westerbork Synthesis Radio Telescope (WSRT), 141,
Unfolding our Universe (Nicholson), xii 199, 291
unification of quarks and leptons, 449 Weyl’s hypothesis, 113–14
unification of strong and electroweak forces, 444–5, white dwarfs, 36, 63–6, 68, 70–3, 96, 105, 178, 194,
448 252, 369
characteristic energy for (GUT scale), 445 cooling ages of, 347
and instability of the proton, 448–9 discovery of, 63
unified models/unification scenarios for active galactic maximum mass of (Chandrasekhar mass), 64–5, 68
nuclei, 299–300, 305 theory of, 64–5, 70–3
for BL Lac objects and FRI radio galaxies, 300 and Type 1a supernovae, 206
for radio galaxies and radio quasars, 299–300 White Mountain, California, 135
for Seyfert 1 and Seyfert 2 active galaxies, 299 Wilkinson Microwave Anisotropy Probe (WMAP),
Universe, stability under gravity, 100 415–16, 422–7
US Army’s rocket programme, 142 and the determination of cosmological parameters,
US Bureau of Standards, 156 423–8
US Office of Weights and Measures, 31 William Cranch Bond Astronomer (Annie Cannon),
US Patent Office, Washington, 31 27
William Herschel 4.2-metre Telescope (WHT), 163
vacuum energy, see dark energy William Huggins Observatory, Publications of Sir
vector model of the atom, 41 (eds Huggins and Huggins), 27
Vela satellites, 147–8, 300 Wisconsin soft X-ray rocket flights, 224
Subject index 545

Wolf–Rayet stars, 23 X-ray emission lines of Fe XXV and Fe XXVI, 260

Woolensack 31/4 -inch telescope, CalTech, 204 X-ray fluorescence lines from Seyfert I galaxies,
World, The (Descartes), 77 287–9
World as Space and Time (Friedman), 108 asymmetric broadening of, 288–9
special and general relativistic effects in, 288–9
XMM-Newton X-ray Observatory of ESA, 146–7 X-ray imaging telescopes, 144
X-ray astronomy, history and development of, X-ray sky surveys, 365, 372–4
142–7 deep ROSAT, 365, 372–4
first detections of extra-solar-system sources, 143–4 source confusion in deep, 372–3
X-ray Astronomy (eds Giacconi and Setti), 268 X-ray sources, nature of, 144–5
X-ray background radiation, 372–4 X-ray Universe, The (Tucker and Giacconi), 171,
anticorrelation with neutral hydrogen column 212
density, 224–5 X-rays, discovery and early history of, 130
extragalactic background attributed to discrete
sources, 372–4 Yerkes Observatory, 125
Galactic, at soft X-ray energies, 224–5 Yerkes 1-metre refracting telescope, 12
X-ray binaries, 144, 197–200 Yerkes system of stellar spectral classification, 29,
discovery of, 197–8 37–8
eclipsing, 197–8 yield of heavy elements by stars, 390
identification of stellar companions of, 197–8 Young’s double-slit experiment, 4
pulsed signals from, 197–8 Yukawa’s theory of the strong nuclear force, 135
and the search for black holes, 197–200
superluminal motions in, 296 Zeldovich solution for non-linear collapse, 403
X-ray luminosities of, 198–9 zero-point energy of vacuum fields, 325, 451
X-ray detectors, 144–5 zinc as a tracer of build-up of heavy elements,
X-ray emission 383–5
fluorescent, of Moon, 143, 171 zone of avoidance, 86
of Sun, 142 Z0 bosons, 334, 441