Time: Towards a Consistent Theory

Fundamental Theories of Physics

An International Book Series on The Fundamental Theories of Physics:

Their Clarification, Df!velopment and Application

Editor: ALWYN VANDER MERWE

University of Denver, US.A.

Editorial Advisory Board:

ASIM BARUT, University of Colorado, US.A.
BRIAN D. JOSEPHSON, University of Cambridge, UK.
CLIVE KILMISTER, University of London, UK.
GUNTER LUDWIG, Philipps-Universitiit, Marburg, Germany
NATHAN ROSEN, Israel Institute of Technology, Israel
MENDEL SACHS, State University of New York at Buffalo, US.A.
ABD~JS SALAM, International Centre for Theoretical Physics, Trieste, Italy
HANS-JORGEN TREDER, Zentralinstitut fur Astrophysik der Akademie der
Wissenschajten, Germany

Volume 65
Time:
Towards a
Consistent Theory
by

C. K. Raju
Indian Institute ofAdvanced Study,
Rashtrapati Nivas,
Shimla, India
and

Centre for Development of Advanced Computing,

New Delhi, India

SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN 978-90-481-4462-4 ISBN 978-94-015-8376-3 (eBook)

DOI 10.1007/978-94-015-8376-3

Printed on acid-free paper

All Rights Reserved

1994 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 1994
Softcover reprint of the hardcover 1st edition 1994
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner.
Contents
Synoptic table of contents . vi
Preface . . xi
Introduction 1

Part A: Preliminary paradoxes and puzzles

I Philosophical time 11

Part B: The measurement oftime

II Newton's time ...... . 33

IlIA The Michelson-Morley experiment 49
IIIB Einstein's time 59

Part C: The arrow oftime

IV Thermodynamic time 79
VA The electromagnetic field . 102
VB Electromagnetic time 116

Part D: The topology oftime

VIA Bell and non-locality 139

VIB Quantum-mechanical time 161
VII Cosmological time . . . 190

Part E: Towards a consistent model oftime

VIII Mundane time 215

Notes and References 233
Index. . . . .. 247

v
 Synoptic Table of Contents

INTRODUCTION 1 Part B: The measurement of time

1 The structure oftime 1
II NEWTON'S TIME 33
1.1 Mundane time 1
1.2 Superlinear time 2 1 Introduction 33
1.3 The paradox of mundane time 2
2 Newton's laws of motion 33
1.4 Th~ irrelevance of (quantum) indeter-
mlntSm 3 2.1 Bodies 34
1.5 Temporal assumptions in the argu- 2.2 Forces 35
ment from chaos 4 2.3 Sources of forces 35
2.4 Inertial frames 36
2 The physicist's point of view 4
2.5 External forces 37
3 Objectives of the exposition 5 2.6 Newton's laws as defmitions 38
5 2.7 Clocks 38
4 Organization
3 Falsifiability and physical theories 39
4 Laws of motion and law of gravitation 41
Part A: Preliminary paradoxes and 5 Laplace's demon 43
puzzles 5.1 Time in classical mechanics 43
5.2 The structure of time in classical me-
chanics 43
I PHILOSOPHICAL TIME 11
5.3 The argument from complexity 44
1 The definition of time 12 5.4 Laplace's demon 44
12 5.5 Exorcism of Laplace's demon 46
1.1 The operational defmition
1.2 Is time objectively definable? 14 6 Conclusions 47
1.3 Is a definition necessary? 14 Box items:
1. Newton and action at a distance 36
2 The myth of passage 15 2. Karl RainlUnd Popper 40
2.1 The A-series: the flow of time 15 3. Pierre Simon de Laplace 45
2.2 The B-series: the arrow of time 16
2.3 McTaggart's paradox 18 IlIA THE MICHELSON-MORLEY
2.4 Resolution of McTaggart's paradox 20 EXPERIMENT 49
3 The sea fight tomorrow 20
3.1 Fatalism and determinism 21 1 Introduction 49
3.2 Dumett's paradox 21 2 The background of aether theories 50
4 Logic and 'free will' 22 2.1 The blue sky or why aether was intro-
duced 50
4.1 The Master Argument of Diodorus 2.2 The Dragon's head 51
Cronus 22
2.3 Stellar aberration and the finiteness of
4.2 Lukasiewicz's three-valued logic 23 the speed of light 51
4.3 The quasi truth-functional system 24 2.4 The grove of trees 52
5 Zeno's paradoxes of motion 26 2.5 The aether drag 52
5.1 Achilles, dichotomy and the arrow 27 2.6 A delicate point 53
5.2 Tasks and supertasks 28 2.7 The aether dragged 54

6 Conclusions 29 3 Electrodynamics: theory and experiment 54

3.1 Maxwell 54
Box items:
1. Augustine 13 3.2 The experiment proper 55
2. ~cTaggart 17 3.3 Miller's observations 56
3. Sri Harsa 19 3.4 Aether, relativity and metaphysics 57
4. Zeno of Elea 27 3.5 The moral of the story 57

vi
 57 4.2 Poincare recurrence theorem 87
4 Conclusions 4.2.1 The recurrence theorem simpli-
Appendix: Einstein on the Michelson Mor- fled 88
ley experiment 58 4.2.2 Consequences of the recurrence
theorem 88
IIIB EINSTEIN'S TIME 59 4.2.3 Relationship to mechanics:
Liouville's theorem 89
4.3 The recurrence paradox 89
1 Introduction 59
4.4 Refutation of the paradoxes 90
2 Electrodynamics 60 4.5 Objections to the refutations 92
2.1 Lorentz #1 60 4.5.1 Is the cosmos shuffled? 92
2.2 Lorentz #2 61 4.5.2 The meaning(lessness) of large
recurrence times 92
3 Newton's laws again 62
3.1 Velocity-dependent forces 62 5 Resolution of the paradoxes 96
3.2 Homogeneity 63 6 Conclusions 96
3.3 Isotropy 63 Appendix: Proof of Liouville's theorem 97
3.4 Straight-line motion 63 Box items:
3.5 Relative motion and force 64 1. Pictures of entropy and evolution 84
3.6 Newton's time 64 2. Mixing and sortin! Maxwell's demon 90
3. Elements of er~o .c theory 93
4 Poincare 64 4. Rudolph ClausIus 98
4.1 Poincare and the relativity principle 64 5. Ludwig Boltzmann 99
4.2 Poincare and aether 65
4.3 Poincare and mechanical explanations 66
4.4 Relative velocity and the Michelson- VA THE ELECTROMAGNETIC
Morley experinlent 66 FIELD 102
4.5 Poincare on Lorentz's theory 67
4.6 The new mechanics 68 1 Introduction 102
4.7 Poincare on time 69 2 Is the field necessary? 102
5 Einstein 70 2.1 Force at a distance 103
6 Time 70 2.2 Action by contact 103
2.3 The field 104
7 Conclusions 71 2.4 'Contact' and the structure of matter 105
Appendix: The historical record 72 2.5 The meaning of 'contact' 105
3 The electromagnetic arrow of time 106
3.1 Retarded and advanced solutions 106
Part C: The arrow o/time 3.2 Rl/-diative damping and the arrow of
time 107
IV THERMODYNAMIC TIME 79 3.3 Mixed potentials 107
3.4 The paradoxes of advanced action 108
1 Introduction 79 4 The divergences of field theory 109
1.1 Summary of earlier chapters 79 4.1 The Abraham-Lorentz model 109
1.2 Relativity and existence: loss of time
asymmetry 79 4.2 Action at a distance 111
1.3 Physics and the direction of time 80 4.3 Dirac's approach 111
1.4 Time "ymmetry of physics 80 5 Preacceleration 113
1.5 Is time symmetry reasonable? 80
6 Conclusions 115
2 The entropy law 81
2.1 The meaning of entropy 82
VB ELECTROMAGNETIC TIME 116
3 The Boltzmann H-theorem 83
3.1 The Ehrenfest model 86 1 Introduction 116
4 The reversibility and recurrence 2 The two-body problem of electrodynamics 117
paradoxes 87 2.1 Formulation 117
4.1 Loschmidt's paradox 87 2.2 Some definitions 118

vii
 2.3 The recurrence paradox and the past- 2.6 Me~app.y'sics: locality and the voodoo
value problem 119 prmClple 150
2.4 The reversibility paradox: time asym- 3 Bell's inequalities 152
metry of delay 120
2.5 Preaq;ele~ation: the Taylor-series ap- 3.1 Outline of inequalities 152
proxunatlon 122 3.2 Bell's locality postulate 153
2.6 The pond paradox: advanced equations 123 3.3 Experiments on Bell's inequalities 153
2.7 Indeterminism: mixed deviating argu- 3.4 Spookiness: interactions between sepa-
ments 124 rated systems 154
2.8 The Wheeler-Feynman paradox 125 3.5 Bell locality: interpretation of experi-
ments 154
3 The absorber theory of radiation 126 3.5.1 The early objections 154
3.1 Action at a distance 126 3.5.2 The cugument of Barut and
3.2 The Sommerfeld radiation condition 126 Meystre 155
3.3 The Wheeler-Feynman theory 127 3.5.3 The cugument from efficiency 155
3.4 Hogarth's theory 129 4 Further confusion: other notions of local-
3.5 Other theories 130 ity 157
4 Empirical tests 131 4.1 The Aharonov-Bohm effect 157
4.2 Locality in quantum field theory 157
5 Conclusions 132 4.3 Locality and Localizability 157
Appendix: Derivation of the relativistic two- 4.4 Summary 158
body equations of motion with a tilt in the
5 The debate over non-locality in classical
arrow of time 133
mechanics 158
Part D: The topology o/time 5.1 Francis Bacon and magic 158
5.2 Locality, chains of causes, and action
by contact 159
VIA BELLAND NONWCALITY 139 5.3 Descartes and Newton 159
5.4 The field 159
1 The topology of time: a general introduc- 5.5 Hyperbolicity: fmite propagation
tion to Chapters VIA, VIB, VII 139 speeds 159
1.1 The language of time 140 6 Conclusions 160
1.1.1 The U-calculus 140 Box items:
1.1.2 The 'minimal'tense logic 141 1. Time in q.m.: some recognized new
1.1.3 The language of time: further con- features 143
siderations 142 2. Schrodinger's cat 151
1.2 A new logic for physics 142 3. Schrodinger's cat and SQUIDs 152
1.2.1 Structured time and the tilt in the 4. The Aharonov-Bohm effect 156
arrow of time 142
1.3 Objectives 144
1.4 Non-locality 145 VIB QUANTUM.MECHANICAL
2 Background to Bell's inequalities 146 TIME 161
2.1 Basic experiments 146 1 Introduction 161
2.1.1 The two-slit diffraction experi-
ment 146 2 The orthodox formalism of q.m. 162
2.1.2 Quantization of electron spin 147
2.2 Interpretations 147 3 From quantum logic to the formalism of
2.2.1 Hidden variables 147 q.m. 164
2.2.2 The Copenhagen interpretation 148 3.1 Non-commutativity and non-existence
2.3 Controversy: The EPR Paradox 148 of joint distributions 164
2.4 Classical mechanics as a hidden vari- 3.2 Need for a ch~e of logic: failure of
able theory 149 the distributive aw 165
2.5 History: the no-hidden-variable theo- 3.3 Birkhoff-von Neumann app'roach and
rems 149 the orthodox formalism of q.m. 165
2.5.1 von Neumann's 'theorem' 149 3.3.1 The lattice ofprojections 165
2.5.2 Gleason's theorem and context 3.3.2 Order relation and orthocomple-
dependence 150 ment 166
2.5.3 Non-locality of Bohm's theory 150 3.3.3 Geometrical interpretation 166

viii
 3.3.4 Dynamical variables, random 1.3 The cosmological arrow of time 192
variables and self-adjoint opera-
tors 166 2 Darkness of the night sky: "OIbers' .. par-
3.4 The quantum logic approach 167 adox 193
3.4.1 The minimum requirements 167 2.1 Olbers' paradox 193
3.4.2 Orthomodularity 168 2.2 The absorption solution 193
3.4.3 Compatibility 168 2.3 The expansion solution 194
3.5 The J auch-Piron approach 170 2.4 The finite age solution 194
3.6 Defects in the quantum logic approach 172 2.5 Summary 195
4 The structured-time interpretation of 3 Cosmological expansion and redshifts 196
q.m. 173 3.1 Hubble's law 196
4.1 Motivation 173 3.2 Other interpretations of the redshift 197
4.2 Overview of the argument 173
4.3 Fr,om electrodynamics to structured 4 The Friedmann models 197
time 174 4.1 The cosmological principle 197
4.3.1 The analogy with CSP 174 4.2 The FLRW line element 198
4.3.2 Past and present contingents 175 4.3 Id~ntmcation with observable quanti-
4.3.3 The logic of stroctured time 175 ties 199
4.4 From structured time to quantum logic 175
4.4.1 Statements 175 5 The cosmic microwave background 201
4.4.2 Troth-functional worlds 176 6 The beginning of time 202
4.4.3 Quasi troth-functional worlds 176 6.1 The initial singularity 202
4.4.4 Modalities 176 6.2 Singularities and black holes 202
4.4.5 Imducible contingents and ad-
missible worlds 177 6.3 Cosmological singularities 203
4.4.6 States 177 6.4 Interpretation of singularities: shocks 203
4.4.7 Quantum measurements, access 6.5 Shocks and hyperbolicity 204
relations, and selection func- 7 The end of time: dark matter and rota-
tions 177
4.4.8 Incompatibility, joint measure- tion of galaxies 205
ments and repeated measure- 8 The age of the universe and closed
ments 178 timelike curves
4.4.9 Measurability 178 206
4.4.10 'And', 'or'; 'if, 'not' 178 9 Problems about time: a cosmological per-
4.4.11 Preliminary order relation 179 spective 208
4.4.12 Orthogonality, compatibility and
orderrelation 179 9.1 Is there a proper clock? 208
4.4.13Main result 180 9.2 Related asymmetries? 209
4.4.14Example offai/ure of the distribu- 9.3 Imperfect asymmetries? 210
tive law 181 9.4 Local asymmetries? 210
4.4.15 The Lattice strocture 181 9.5 Is asymmetry adequate? 211
4.4. 16Admissibility 182
4.4.17 Example of a selection function 182
4.5 Relation to other interpretations 182
4.5.1 The many-worlds interpretation 182 Part E: Towards a consistent model of
4.5.2 The transactional interpretation 183 time
4.5.3 The modal interpretation 183
4.5.4 The Copenhagen interpretation 183
5 Conclusions 183
VIII MUNDANE TIME 215
Appendix: Proof of the main theorem 184 1 Introduction 215
Box items:
1. Collected definitions 169 2 Mundane time: linear past and branch-
ing future 216
2.1 Reformulation of the A-series 216
VII COSMOLOGICAL TIME 190 2.2 Ignorance, past and future 216
1 Introduction 190 2.3 Epistemic and ontic contingents 217
1.1 Speculation in cosmology 191 3 Comparison with time in physics 218
1.2 Choice of a theory 191 3.1 The B-series from relativity 218

ix
 3.2 Limitations of the arrow simile 218 5.2.4 Laplace's demon 226
3.3 BranchinR; of time and the failure of 5.3 Epistemically broken time: the argu-
physicaflaws 218 ment from chaos 226
5.3.1 Predictability and contingents 226
4 Need for consistency 219 5.3.2 Reversibility of chaos 226
4.1 The problem of consistency 219 5.3.3 Unpredictability of the past 227
4.2 Building a consistent model of experi-
ence 219 6 Consistency through tilting time 227
4.2.1 The two-time theory 219 6.1 Tilting time and consistency within
4.2.2 Mundane time and assumptions phYSICS 227
about time in physics 219 6.2 Preliminary consistency with mundane
4 3 Mundane experience as hallucination: time 228
. the problem of intentional choice 221 6.3 Remaining problems of consistency 228
4.3.1 The flat earth and parallax 221 6.4 The problem of intentionality 228
4.3.2 The problem of intentional choice221 6.4.1 Temporal assumptions underly-
4.4 The complete problem of consistency 222
ing epistemically broken time 228
4.5 Do physical laws somehow fail for 222 6.4.2 Explanation from final causes 228
human beings?
6.4.3 Intention or 'purpose' as future
5 Consistency through broken time 222 data 229
5.1 Ontic~lly broken time: quantum inde- 223 6.4.4 Intentionality with mundane time 229
termlBlSm 6.4.5 Differences between the two no-
5.1.1 Chance vs. choice 223 tions 229
5.1.2 Occasionalism 223 6.5 The problem of macrophysical choice 230
5.1.3 The transition to macrophysics 224 6.5.1 Advanced interactions as pertur-
5.1.4 Quantum 'choice' and measure- bations 230
ment 224 6.5.2 Deep history dependence 230
5.1.5 Summary of quantum indeter- 6.5.3 Anticipatory phenomena
minism and mundane time 224 (choice) and Maxwell's demon 230
5.2 Epistemically broke~ time: the argu- 6.5.4 Ampliatory self-organization 230
ment from compleXIty 225
5.2.1 The complexity of macrophysical 6.6 The problem of past contingents: struc-
systems 225 tured time vs. past linearity 231
5.2.2 Thennodynamics and stochastic 7 Conclusions 231
evolution 225
5.2.3 Comparison with quantum inde-
tenninism 225 Notes and References 233

x
 Preface

T IME is indubitably a matter of life and death! But it is a subject on which miscon-
ceptions are widely prevalent, even amongst physicists. This prompted me to write
a series of expository papers, On Time, which appeared in the pedagogical journal
Physics Education (India) as follows.
I Philosophical time, 7(3), 204-217,1990.
II Newton's time, 8(1), 15-25, 1991.
IlIA The Michelson-Morley experiment, 8(3), 193-200, 1991.
IIIB Einstein's time, 8(4), 293-305,1992.
IV Thermodynamic time, 9(1), 44-62,1992.
VA The Electromagnetic field, 9(2), 119-128, 1992.
VB Electromagnetic time, 9(3), 251-265,1992.
VIA Bell and non-locality, 10(1),55-73,1993.
VIB Quantum-mechanical time, 10(2),143-161,1993.
VII Cosmological time (to appear).

I am grateful to Professor A.W. Joshi, Assistant Editor of that journal, for inviting
me to write this series and for his constant enthusiasm which has been a source of
support. I am also grateful to the Editors for permission to reproduce the papers in a
slightly modified form in this book.
While writing this series of papers, it was hardly possible to avoid the infusion of
some new ideas and a fresh way of looking at old controversies. I am grateful to the
Indian Institute of Advanced Study, Shimla, for a Fellowship which made it possible to
complete the core of the series and shape it into a book. Apart from the Fellowship
which provided the time, the tranquil and beautiful setting of the Institute was a definite
source of inspiration. I am grateful to the authorities of the Institute, particularly to
Professor S. Gopal, Chairman of the Governing Body, and Professor Mrinal Miri,
Director, for readily agreeing to my proposal to have the book published outside so that
it might reach a wider audience.
In the surcharged atmosphere of an organization like C-DAC, invested with the
mission of developing a supercomputer, from scratch, within three years, it would
hardly have been possible to commence (and later complete) this book (even between
delayed flights), but for the enlightened management of Dr. Vijay P. Bhatkar.
I am grateful to Professor Jagdish Mehra for the general encouragement that he
provided.
In addition, there are a large number of people to whom my debt must remain
unacknowledged. These include the readers of Physics Education, and the anonymous
referees who provided useful feedback.
Unfortunately, there is no one whom I can thank for careful preparation of the
typescript, so there is no one to whom I can indirectly pass on even part of the blame
for any errors that remain!
I dedicate this book to Tomu and Jaya for the sudden disruption in their lives that
the writing of this book entailed.

xi
 INTRODUCTION

1 The structure of time

D OES time have a non-trivial logical or topological structure? This possibility was
implicitly recognized long ago; it found formal expression in temporal logics in the
60's and 70's and has even acquired considerable practical significance since the recent
advent of parallel computing. But the question has, so far, not been considered from
the point of view of physics.
A quick way to bring the problem into focus is to think about the sharp contrast
between the notion of time in physics and the notion of time in everyday life.

Fig. 1: Mundane time

/ On the mundane view, time is be-

lieved to be linear towards the past,
and branching towards the future.
There is only one past, and the
possibilities actually realized are
represented by the thick line. The
arrows represent the several possi-
ble futures, while the thin lines rep-
resent the might-have-been
possibilities, now excluded, pre-
sumably due to human choices.

1. 1 Mundane time

Everyday actions tend to presume the picture of a time which branches towards the
future an<l is linear towards the past (Fig. 1). In mundane life, one believes that the
future is uncertain because it is unknown, whereas one believes that the past is certain
even if it is unknown. From the point of view of logic, these mundane beliefs can be
regarded as beliefs about the truth or falsity of assertions about the past and future. The
truth-value of an assertion about the future, such as 'It will rain tomorrow' is undecided
because it is unknown. The truth-value of assertions about the past is believed to be
decided, even if one does not know what it is. (Treating the future on par with the past
rules out human choices and leads to paradoxes like the sea battle, considered in
Chapter I.) One believes that one's actions at least partly decide the future, though they
leave the past unaffected.
2 INIRODUCTION

IPast I --------------------------------
Fig. 2: Superlinear time

In physics, time is both logically and topologically a line. The evolutionary laws ofphysics are fonnulated as
differential equations, presupposing that time has (locally) the topology of the real line. Given the past, the
evolutionary laws of physics decide the future. Usually, both past and future are decided, given the present.
One believes that it is possible to distinguish between past and future, so that time is represented by an arrow.

1. 2 Superlinear time

On the other hand, physics itself treats time as a real variable. The evolutionary
equations of physics (including the Schrodinger equation) involve derivatives with
respect to time, so that they already assume that time has (locally) the topology of the
real line. (More generally, one assumes the existence of a spacetime manifold which is
locally Euclidean or Minkowskian.) Existence and uniqueness theorems for differential
equations now typically ensure that both future and past are (locally) decided, if the
present is. Thus, time in physics is both logically and topologically a line - to coin a
term, time in physics is superlinear (Fig. 2).

1. 3 The paradox of mundane time

In the history of physics, such conflicts between physics and mundane experience have
often arisen. A round earth or a heliocentric world, for instance, seemed entirely
contrary to common sense. Physics has usually triumphed. To conclude inductively, that
mundane time is another hallucination masquerading as common sense would, how-
ever, undermine the very basis of physics.
Physics, one believes, incorporates distilled human experience. One believes that
physics is grounded in empirical reality, and that the predictions arising from physical
theories are vindicated by experiment. A philosopher of science, like Popper, might
adopt the more modest position that physical theories are (or ought to be) logically
falsifiable. It should be possible, at least in principle, to put physical theories to test.
Now, it is a matter of common experience that one is 'free' to perform any
experiment to test physical theories. It is also a matter of common experience that one
is free to conceive of counterfactual possibilities like a bird that is like a swan in all
respects except that it is bright orange. It is this human freedom to conceive alternatives
which decides whether or not the statement 'All swans are white' is refutable. This
common experience of human freedom is, of course, underpinned by mundane time -
one believes one's choices will affect the future, but cannot affect the past, one can
choose which experiment to perform, but one cannot choose (to alter) the results of an
experiment already performed.
So, if one abandons mundane time, the validity of physical theories is in jeopardy.
One is no longer even sure that physical theories are more meaningful than tautologies.
 INfRODUCfION 3

On the other hand, one can hardly accept the mundane picture of time, or abandon the
superlinear time of physics, without pointing to some flaw in the equations of classical
physics (including relativity) or in the mathematical formalism of quantum mechanics.
Perhaps one needs to give up the entire approach to physics via differential equations,
and perhaps one needs an entirely new mathematics on which to base physics.
How does one escape from this paradoxical situation where, prima facie, one can
neither accept nor reject mundane time?

1.4 The irrelevance of (quantum) indeterminism

It is possible, of course, to argue that there is no real conflict between the two pictures.
The typical argument here confuses the conflict between mundane time and superlinear
time with the conflict between 'free will' and 'determinism' in the scholastic tradition.
For example, it is argued that quantum mechanics is indeterministic, so that time in
quantum physics is not the real line used in writing down the Schrodinger equation, but

Fig. 3: Ontically broken time of

quantum mechanics

The figure shows schematic evolution according to quantum mechanics. Evolution takes place according to
the SchrOdinger equation (U(t)), until it is intemtpted at time t by a measurement process, resulting in the
col/apse of the wavejUnction tp. Quantum mechanics fIXes the probabilities of outcome only after the
measurement process has been specified.

is 'ontically broken' (Fig. 3). Now, it is debatable whether quantum mechanics does not
decide the future, given the past; but it is unnecessary to go into such debates.
The point is simply that mere indeterminism is irrelevant. An occasionalistic world
in which books may turn, quite unpredictably, into horses (which soil the library) is quite
indeterministic. Such a world does not even admit probabilistic connections between
future and past. But just because the occasionalistic world evolves arbitrarily it leaves
no scope for human intervention to bring about a particular future or plan a certain
experiment. Thus, an argument against determinism is not necessarily an argument for
mundane time.
The other problem with the current belief in quantum indeterminism is that it
involves the measurement postulate. This postulate fixes the probabilities of the
outcome of a measurement process only after one has decided what measurement to
make. For example, one may decide to measure the spin of an electron, say, in the
x-direction rather thany. This belief, that one is free to decide the direction in which to
measure the spin, already presupposes mundane time, where human choices (partly)
decide the future. Thus, any argument for mundane time, using quantum indetermin-
ism, begs the question, and the current belief in quantum indeterminism turns out to
be incoherent. The argument from chaos, examined below, does not fare much better.
4 INTRODUCTION

Neither argument truly addresses the conflict between mundane time and superlinear
time.

1.5 Temporal assumptions in the argument from chaos

The argument from chaos also seeks to wish away the paradox of mundane time. Popper,
for instance, seeks to demolish the superlinear time of physics by exorcising Laplace's
demon. The argument is that the dynamics of complex systems, like human beings, is
chaotic. Thus, the future may be decided, but it is impossible, in principle, to predict
the future or the truth value of assertions concerning future events. The obvious fallacy
with this case of 'episternically broken' time is that it is time symmetric - chaotic
systems are equally chaotic towards the past, so that retrodiction is equally impossible.
It is irrelevant that one can keep records - the argument from chaos works just because
it is supposed that memory or records cannot, in principle, be accurate. So, on this
argument, the remote past (about which there are no accurate records) is certainly
indeterminate, but so also is the immediate past and the present. Moreover, if unpre-
dictability, in principle, of the future is an argument for the belief that human choices
may affect the future, then non-retrodictability, in principle, of the past ought to be an
argument for the belief that human choices may affect the past.
The less obvious fallacy with the argument from chaos is that it, too, makes further
temporal assumptions. The world, according to Popper, is like a mystery novel - the
murderer may be decided, but his identity is unknown to the reader on p 15. The tacit
assumption, that the reader cannot peek at the last page, is analogous to the temporal
assumption that the future cannot be directly known - it must be inferred. That is,
physical interactions can propagate only from past to future and not vice versa. Holding
on to such a plethora of temporal assumptions invites incoherence - ironically, it turns
out that one can derive a stronger form of the result sought by Popper, and at least partly
reconcile mundane time with time in physics, by denying the preceding assumption.

2 The physicist's point of view

From the point of view of the physicist (or the philosopher) as a social animal, the
question of reconciling the two pictures of time is a dangerous question, because,
without going beyond human experience, it seems to tempt him to tread on academic
taboos by transgressing the academic boundaries established by convention. The inter-
esting thing is that the problem may be stated, and the case argued out entirely from
within physics, without any reference to mundane time.
Briefly, it is possible to relate the need for a structured time to the need for a change
of logic, required at the microphysical level by classical quantum axiomatics. Further-
more, it is possible to arrive at a structured time by re-examining the question of the
arrow of time, leading to the proposal of a tiny tilt in the arrow of time. To this end, I
thought it preferable to embed the argument for a structured time in an exposition of
the problem of time in its entirety, an exposition which avoids the usual confusion
between theory and entrenched speculatory positions.
 INTRODUCTION 5

3 Objectives of the exposition

This exposition has some new angles. Time is a basic concept of physics, but apart from
a few casual references to non-existent 'principles of causality', most physics texts do
not think it necessary to go into further details. Many physicists tend to internalize this
'no problem' attitude. It is assumed that time, like mass or the colour yellow, is
something we all understand. There is not the slightest hint of the immense complexities
and confusion that underlie the subject. Not for no reason did Augustine wail, 'What,
then, is time?' Neither several millennia of speculation nor the elaborate structures of
physics have, so far, quite served to answer Augustine's question. But, today, we are at
least in a position to state some of the problems clearly.
Thus, one of the objectives of this exposition is to formulate some of the major
problems relating to time. To formulate these problems, it is necessary, first, to
acknowledge their existence. It is necessary to adopt a scientific attitude, and not treat
this or that entrenched speculation, even if it originates from an authority, as the last
word. Thus, this exposition of what we do know concerning time emphasizes what we
do not know. In this respect, the present exposition differs radically from all earlier
expositions, including the excellent one by Paul Davies (1974) on The Physics of Time
Asymmetry. (Davies concludes that nothing qualitatively new is likely to emerge from a
study of the physics of time - only philosophical questions are left!)
A difficulty is that the physics of time sweeps across all of physics. Since time is a
subject of wide interest, and very few can reasonably be expected to be experts in all
areas of physics, to maintain intelligibility, I thought it necessary to strip off every
technical detail that was not absolutely essential. I hope this will make the book more
accessible to a wider audience of physicists, and especially to young physicists, who are
less susceptible to entrenched speculations. I also hope that, by pushing technicalities
into the background, this book will initiate an 'honest' dialogue between physicists and
philosophers of science, who are surely interested in the issues involved.

4 Organization

The exposition has five broad divisions. The first part deals with preliminary non-tech-
nical questions about the nature of time that have puzzled thinking people since the
beginning of recorded human history. These questions pertain to the definition of time,
the reality or objectivity of change, the possibility of human choice, and the discrete or
continuous nature of time. Age-old attempts to answer these questions have generated
a series of paradoxes which I elaborate with an indication of the physical context.
The second part takes up the problem with the measurement of time in Newtonian
mechanics. Nearly a century after the birth of relativity, Newton's laws of motion are
still taught as if they constitute a self-contained physical theory. The stock presentation
of the transition to relativity emphasizes the Michelson-Morley experiment rather than
the theoretical difficulty in defining equal intervals of time. Having accidentally arrived
at the much-neglected line of thought used by Poincare, I found it natural to emphasize
it. In the process, I hope I have placed matters in a truer theoretical and historical
6 INfRODUCTION

perspective. In general relativity, of course, all measurement reduces to time measure-

ment.
The solution of the problem of equal intervals of time, by the postulated constancy
of the speed of light, has another consequence. It results in a static (B-series) view of
time - one can no longer adopt the commonplace view that the future has yet to come
into existence or that the past has ceased to exist. The absence of such a self-evident
distinction between past and future, and the near-absence of any distinction between
past and future, in physics, forces consideration of the celebrated problem of the arrow
of time in physics: how does one distinguish physically between past and future? This
question is taken up in the third part.
Though the entropy law is rather close to the intuitive distinction between past and
future in terms of information ('the future is more unknown than the past'), its
dynamical basis is weak. Using a rather general formulation of the recurrence and
reversibility paradoxes, I argue that any attempt to derive time-asymmetric conclusions
from time-symmetric dynamics is bound to fail. Usually this failure remains hidden
because some intuitively plausible (and symmetric-looking) but time-asymmetric as-
sumption, such as mixing (as opposed to sorting), has been introduced at some inter-
mediate stage in the argument. I argue that thermodynamics, independently of
cosmology or field theory, cannot provide the basis for time asymmetry - the thermo-
dynamic arrow of time cannot be fundamental.
Having enumerated all the ways of escaping the recurrence and reversibility
paradoxes, I argue as follows. If one wishes to keep the thermodynamic distinction
between past and future independent of cosmology, then the infinite degrees of
freedom of the field provide the only escape route.
Introduction of the field brings with it the problem of locality and divergences.
Retarded direct action then appears as the most elegant solution - it provides a
time-asymmetric dynamics, free from inconsistencies, and completely resolves the
recurrence and reversibility paradoxes. The problem, of course, is to justify the choice
of retarded direct action. I argue that the paradoxes of advanced action (similar to the
grandfather paradox) (a) assume mundane time, and (b) fail to take into account the
simplest (though perhaps counter-intuitive) qualitative features of solutions of differ-
ential equations with deviating advanced arguments. Hence, advanced interactions may
not be excluded on purely logical grounds.
On the other hand, there is the as-yet unresolved difficulty that advanced potentials
are necessary for a covariant derivation of radiative damping. One could attempt to
resolve this difficulty within the framework of the absorber theory of radiation. Here I
point out that the Wbeeler-Feynman absorber theory has been shown to be internally
inconsistent, while there is a strong likelihood that the Hoyle-Narlikar theory is, at least,
externally inconsistent. The remaining theory (mine) cannot rule out the possibility of
small amounts of advanced interactions or a small tilt in the arrow of time. This
possibility is also not ruled out by the only experiment that has been performed.
The fourth part of the book deals with the problem of the structure or topology of
time. If one takes seriously the possibility of a small tilt in the arrow of time (howsoever
obtained), then one is immediately confronted with the question of locality. In recent
years, locality has been much debated in the context of Bell's inequalities. Without
denying non-locality (and without accepting hidden variable theories), I argue that the
experiments on Bell's inequalities have not demonstrated non-locality or even a viola-
tion of Bell's inequalities, because of the problem with the efficiency of the detectors.
 INTRODUCTION 7

I further argue that experiments on Bell's inequalities can, at best, serve to attach a tag
of non-locality to those hidden variable theories that are empirically successful. On the
other hand, I argue that there may be difficulties in the mathematical formulation of
the notion of locality used by Bell, and that there is no logical reason to regard locality
as sacrosanct since every formulation of locality, in classical mechanics, has failed over
the last few centuries.
If one does take seriously the possibility of a microphysical tilt in the arrow of time,
one must inquire into the qualitative consequences. I argue that a microphysical tilt in
the arrow of time results in a structure of time at the microphysical level. This structure
forces one to abandon classical logic at the microphysical level, and I show that the
resulting logic is a quantum logic. Thus, one could hope to base quantum axiomatics
on the physical postulate of a tilt in the arrow of time.
The last part attempts a tentative consolidation. The chapter on cosmology places
the problem of time in a cosmological perspective, which, I believe, is the only right
one. Some of the fundamental unresolved problems include the global nature of time,
the relation between global and local structure, the conceptual adequacy of 'time
asymmetry', and the physical existence of a proper clock. I hope that more attention
will be paid to these problems, in the near future, if only because time is proving to be
a major hurdle in the programme of reconciling quantum mechanics and gravitation.
The last chapter reverts to the question of consistency with mundane time posed
at the beginning of the introduction. Many arguments have sought to reconcile the
apparent determinism of physics with macrophysical human choice. I bring out the
fundamental weaknesses in the arguments from complexity, chaos, and (classical)
occasionalistic, quantum indeterminism. Finally, I argue that a tilt in the arrow of time
is the most conservative change to the current understanding of physics which provides
a solution to the consistency problem between mundane time and the superlinear time
of physics.
 ............................. :........................................................................ ......,
~

Preliminary paradoxes and

A.
puzzles
I PHILOSOPHICAL TIME
ABSTRACT. Our aim is to present a detailed exposition of problems concerning time in philosophy,
Newtonian mechanics, relativity, thermodynamics, electrodynamics, quantum mechanics, and cosmology,
emphasizing what we do not know.

This is the first part of the exposition and deals with certain preliminary questions regarding time. Age-old
attempts to fmd satisfactory answers to these questions have generated a series of paradoxes. Those
elaborated here are the paradox ofthe vanishing present, the paradox of unreal time, McTaggart's paradox,
Dumett's paradox, Aristotle's sea battle, the Master Argument of Diodorus Cronus, Zeno's paradoxes of
motion, and the Thomson lamp. The relevance of these paradoxes to frontier areas of physics, and the
subsequent parts of this exposition, is briefly indicated.

'XTE begin with the treatment of time in philosophy, because most philosophers have
'" assumed little beyond the common experience of time, and the rules oflogic. Also,
philosophers are adept at demonstrating the confusion that arises when one tries to
analyse one's experiences! In this respect, the questions asked by philosophersl are of
the type that a thoughtful and well-read student of physics might raise.
Philosophers have addressed themselves to the following kinds of questions.
1. What is the definition of time?
2. Does time flow?
3. Is the future determined? or is it open?
4. Is our perception of time affected by the language we use?
5. Is time continuous or discrete?
Attempts to answer these questions are sketched in the respective sections below.

Fig. 1: The paradox of

INOW/ the ever-shrinking
present
According to Aristotle, if
the present has any dura-
tion, some part of this
must be past, and some
part future. Discarding
these parts and repeating
the argument shrinks the
present to a point.
12 GHAPTERI

1 The definition of time

Dead yesterday, unbom tomorrow

Why fret about it, if today be sweet?
Omar KbayyAm

1. 1 The operational definition

Time is that when the growth OT decay of bodies is seen. The same in association with some
action gets the name of the day and the night. - With which action? - With the movement of
the sun. The same being repeatedly associated with it gets the name of month and year.
PataDjali (c 500 BC)
JYdkarana Ma1u1.bh4fya 11.2.5

Time present is absent; for, when an object falls, there are only two possible times: the time
past during which it has fallen, and the time future during which it will yet fall.
(Anti-) Gautam (c 300 BC)
NyO.ya Siitra 11.1.39

Aristotle, naturally, was the first to comment on almost everything! In his Physics,
Aristotle felt that before venturing a definition of time, it was necessary to ensure that
time existed. He was inclined to suspect that time 'either does not exist at all or barely,
and in an obscure way.' Thus, he argued, if something is to exist at all, 'all or some of
its parts must exist.' But, the part of time we call the past has ceased to exist, and the
part which we call the future has yet to come into existence. So the only part that could
possibly exist is the present.
On the other hand, the present itself cannot have any duration. For, if it did, some
part of it must be past or some part of it must be future as in the following example.
Suppose I ask you 'What is playing on the radio?' 'The National Anthem,' you answer.
But, I say, 'The National Anthem has several stanzas, which stanza is being played now?'
'The second.' you reply. But, I persist, 'Which line of the second stanza?' 'The first.' you
say, somewhat testily. 'Which word?' I continue ....
Therefore, 'now' must reduce to something that is point-like and durationless. (See
Fig. 1: The paradox of the ever-shrinking present.) Or, as some philosophers say, the
present must be punctal (not punctual!) or punctiform. (There is a lower threshold to
perception so that the 'present' that humans actually perceive consists of a finite
time-span as measured by clocks. This is called the 'specious present' to distinguish it
from the true present, which is believed to be punctal.)
But Aristotle was loath to believe that something durationless could exist. Neither
was he inclined to accept that time, or duration, could be built up from durationless
'nows't. Hence his conclusion that if at all time exists, then it barely exists in some
obscure way.

t This argument is fallacious. An interval of finite length may consist of an uncountable number
(super-denumerable infinity) of points, each of which has zero length. The issue here is not how one
 PHILOSOPHICAL TIME 13

Nevertheless, Aristotle proceeded to define time by the way time is measured. In

modern terminology, one might say that he gave an operational definition. Operation-
alism is a school of thought which asserts that science is not concerned with the true
nature of things. It is not for science to decide what exists and what does not. Rather,
science deals only with the measurable results of experiments. So, to specify something,
it is enough to specify a series of steps, or operations, by which that thing is to be
measured. This point of view has its merits, until it is used to justify, and maintain, a
state of ignorance, as sometimes happens in quantum mechanics!
So, in an operational vein, Aristotle defined time as the measure of motion or
change. (We need not, at present, distinguish between movement and change.) Accord-
ing to Aristotle, time is not motion, but only the number or measure of motion.
In other words, time is that which is measured by a clock. But what is a clock? If
the hands of my wrist-watch do not move, one could hardly call it a clock. So, to
distinguish a clock from a non-clock one must be able to ascertain whether the hands
of the purported clock move. But the only way to measure movement is to measure
change (of position) with respect to time. Hence, we cannot measure time without some
prior (perhaps subjective) notion of time.
So, Aristotle's definition is circular. Time is defined by movement, and movement
is defined by time. So far as Aristotle was concerned, each defined the other, and it was
time to move on to other things of interest.

Box 1. Augustine (354-430)

Also known as Aurelius Augustinus. Augustine is an influential figure in the history of
western philosophy. He bridges the transition from classical antiquity to the middle ages.
Born at Thagaste, in the Roman province of Numidia, in North Africa, he became Bishop
of Hippo, N. Africa, in 394, and died there during an invasion by Vandals. It was probably
not as evident then, as it is today, that during Augustine's life the Roman empire was
declining. There was a corresponding transition from Roman paganism to Christianity.
Augustine's father was a pagan, and his mother a Christian. His education followed the
standard pattern oflate Roman antiquity - almost entirely literary, and with great stress
on rhetoric. 'Its aim; according to the Encyclopaedia of Philosophy, 'was to enable its
recipients to imitate the great literary masterpieces of the past.' His spiritual quest began
when, at age 18, he read a now lost dialogue by Cicero - the Hortensius. He became a
teacher of rhetoric, first in N. Africa, then in Milan. During this period he acquired a
mistress, a son, and an interest in the teachings of the Persian mystic Mani. He was
converted to Christianity by 386, and thereafter wrote 118 treatises, including the famous
spiritual autobiography The Confessions.
Augustine's considerable influence on the western philosophy of time (and phys-
ics) relates to the formulation of questions about time, and to the rejection, on
essentially religious grounds, of cyclic time (which he identified with the Stoic variety).

defines 'length', but how one defines the sum of an infinity of numbers: a countable infinity of zeros adds
up to zero, whereas an uncountable infinity of zeros need not add up to zero!
14 CHAPTER I

1. 2 Is time objectively definable?

How can we talk about a single action when all we perceive are parts that are not even
co-existent? Action is like a circle offlame formed by moving a buming wand in a circle.
Bhartrhari (c 450)
Va/cyapadiya 111.B.B

Thus, time being of this nature is not rea~ but a mental construct.
Vyisa (c 300 BC)
Yogab~a 111.52

No wonder, some eight centuries later, Augustine was left lamenting, 'What, then, is
time? If no one asks me, I know; if I wish to explain to him who asks, I know not.'
Gale2 sympathizes:
We might try to analyse his [Augustine's] predicament as follows. He has an immediate ...
awareness of time, and moreover, he knows how to use ordinary temporal expressions
about past, present, and future .... Yet oddly enough when he tries to give a verbal
definition of time he is struck dumb; for any definition he may propose winds up being
circular .... But why should this be cause for alarm? After all there are many words,
such as 'yellow,' which stand for simple unanalyzable (indefinable) properties.
There are, however, good reasons why no one has ever asked in anguish, 'What, then,
is yellow... ,' for at least we can ostensively (demonstratively) define yellow, we
can point to an instance of yellow. Time unfortunately admits no such ostensive
definition. Obviously, there is nothing we can point to and say 'This is the past (or
future)' (We must not confuse a photograph or memory with the past event it depicts.)
Neither is it possible for Augustine to point to the present, since it is, as Aristotle
claimed, a knife-edge without thickness which serves merely to connect the past with
the future.
Augustine conclude~ that time is purely subjective - a psychological matter. There
is no past, present or future, but 'only a present of things past, memory; a present of
things present, sight; and a present of things future, expectation.'
Nor is it possible to measure time objectively or say that two minutes are twice as
long as one. For 'the time past that was long, is it long when it is now past or was it long
when it was yet present?'
For the past and future do not exist objectively, 'nor can we measure things that are
not.' But time present, too, cannot be measured since it is punctal. Measurements of
time relate only to measurements of different expanses in our memory.

1. 3 Is a definition necessary?

And a rule fixing the meaning of the terms 'time' and 'sequence' is equally [unnecessary].
PA\lini (c 600 BC)
~!8dhyllyi L2.57

What is the modem view? According to the modem philosophy oflanguage, as put forth
by Wittgenstein, we are certainly able to use, and understand the use of, the word 'time'
in different contexts (e.g. 'I reached just in time,' 'What is the right time,' and so on).
 PHILOSOPHICAL TIME 15

We do not understand the meaning of the word only in the sense that we are unable to
reduce, to a single formula, the variety of ways in which the word could be used.
Friedrich Waismann3 argues:
Here it might be asked: can the word be defmed? But why should I try to find a defInition?
What, after all, is the use of a defInition? It only enables one to replace the word 'time,' in
any given context, by a phrase, or phrases, defIDing time. For example, suppose time is
defmed as 'a form of becoming'. Then the phrase 'there is plenty of time' may be translated
'there is plenty of form of becoming!'

In the Wittgensteinian view, Augustine has lost his way in language. Nouns are often
used to name things or substances, but this is not always the case. So, while 'time' is a
noun, there may not be any single objective entity corresponding to it. To give another
example, the word 'God' is a noun. Not only does the meaning of this word vary with
the context and the person, but there may be no objective entity corresponding to it.
The problem of definition, then, may be regarded as settled ('no definition is
necessary'), except for the depressing fact that we may have to assert side by side that
time, as a unique objective entity, may not exist. Considering that time is as essential to
physics as God is to religion, this answer is not quite satisfactory.

2 The myth of passage

In the definition of Cause it has been stated that the cause should invariably be earlier than the
effect; and here the word earlier is meant to exclude the present and future times; now we have
to consider what is the earlier time which excludes the other two. As a matter offact, however,
no consideration or examination of this is possible ... for what after all is the ~ of the
notion of 'present' and the rest? for is this division of time something intrinsic or adventitious?
Sri Har~a (c 8(0)
KlrJu}rfrma Kha,!tja KhIldya W,139

2. 1 The A -series: the flow of time

And even the smallest moment does not stick for time is ever moving.
\>allanicArya

An event may, at one time, be wholly in the future. For example, one may say 'My
examinations begin next month.' Tension and apprehension increase as the exams come
nearer. Eventually, the first phase of the event becomes present, as when the invigilator
starts distributing answer sheets. Soon the exams are over, and, as the vacation begins,
the exams recede further and further into the past.
In this way of looking at time, time is regarded as a dynamic entity. An event was
future, became present, and is past. At successive instants some events come into
existence while others cease to exist. The unceasing parade of instants makes the head
swim. We imagine a river of time flowing past us.
Mixed metaphors apart, the language has three tenses - past, present and future. 4
When speaking of time, we tend to use these tenses - we speak of time in a dynamic
or tensed way. Formally, this dynamic way of depicting time is called an 'A-series'. The
16 CHAPTER I

pastness, presentness, or futurity of an event is called its 'A-determination'. Each event

has three successive A-determinations: past, present, and future.

2. 2 The B-series: the arrow of time

Time has as its characteristic signs the relations 'prior', 'posterior' , ...
Ka,ada
Vai!e~ikasutra 11.2.6-11

A proposition such as ' Now there is a pot' takes into consideration the motion of the Sun and
so on. When we say 'now' we refer to the position of the Sun .... In other words, when we say
'Devadatta is prior to Yajnadatta' this indicates that Devadatta is related to a larger number of
solar revolutions.
Siddhilnta Muktavalf 45-46

Now a question arises: how can an object be related at all with solar motion?
Kiranavalli Bhdskar-a

There is another way of looking at time as 'a pattern of timeless moments' in the words
of T.S. Eliot. The very same events whose A-determinations are continually changing
are laid out in a permanent order, corresponding to the relation 'earlier than'.
'Earlier than' is an order relation in that it is transitive and anti-symmetric like the
relation < (less than). If M is earlier than N, and N is earlier than L then M is earlier
than L. Similarly, it is asymmetric since both 'M is earlier than N' and 'N is earlier than
M' cannot hold.s The relation 'later than', like 'greater than', has similar properties.
These are called B-relations.
B-relations are thought to be permanent. 'Plato's death, for example, cannot
through diligence and hard work sneak up on Stalin's death, for if one event is earlier
than some other event by so many time-units, then it is always the case that the one is
so many time units earlier than the other.'6 In this static way of conceiving of time, we
take a God's eye view of all events arranged in a given order. Events arranged in this
manner are said to form a B-series.
The A-determinations present in customary tensed speech can be transformed into
formal tenseless B-statements. Thus, instead of saying 'Nehru's death is past' we say
that 'Nehru's death is (so many time units) earlier than 26 October 1988,' or 'Nehru's
death is simultaneous with 27 May 1965'. In such translations, the use of 'is' is to be
regarded as tenseless, since it only depicts a timeless B-relation. 'Is' is here used in the
same tenseless way as in the statement 'There is a real number whose square is 2.'
Here are some more examples of such translations. 'It is now raining' is translated
as 'Rain is simultaneous with this utterance'. 'It will rain tomorrow' is translated as 'Rain
is one day later than this utterance.' The general rule is that an event is past, present,
or future only with respect to some other chosen event, i.e., it is earlier than, simulta-
neous with, or later than the chosen event. The chosen event may be a date, an utterance,
or a mental act of perception.
To summarize, A-determinations change. An event which is now present was
future, and will be past. But B-relations are permanent. If M is ever earlier than N, it is
always earlier. Any sentence involving A-determinations can be transformed into a
sentence involving only B-relations.
 PHILOSOPHICAL nME
Box 2. McTaggart (18661925)
John McTaggart Ellis McTaggart, the
British metaphysicist, was born in lon-
don, the son of Francis and Caroline
Ellis. The alliteration in his name is not
exactly Scottish custom: his father later
took the name McTaggart to fulfill a con-
dition for inheriting a bequest. J.M.E.
McTaggart was a successful teacher at
Cambridge from 1897 until retirement in
1923. McTaggart's paradox first ap-
peared as 'The Unreality of Time', Mind
17 (1908), and has had an abiding influ-
ence on the 20th century philosophy of
time.
In his autobiography (Ex-Prodigy,
Ch. 14), Norbert Wiener writes, 'It is
impossible to describe Bertrand Russell
except by saying that he looks like the
Mad Hatter ... .' He warms to his theme
with a swipe at precognition, ' ... the car-
icature of Teniel almost argues an antic-
ipation on the part of the artist.' He then
generalizes to complete the picture, with
McTaggart as the Dormouse, and G.E.
Moore as the March Hare. The three
men were collectively known as the Mad
Tea Party of Trinity.
McTaggart belongs to the school of
thought which distinguishes between an
apparent mundane reality and a true ul-
timate reality, and regards the latter as
appropriate for philosophical contem-
plation. Ultimate reality, in his view, ex-
cluded space, time and material objects.
Despite the unreality of time, McTaggart
argued that individuals may have an im-
mortal soul re-incarnated in a succession
of (apparent) bodies.
He wrote on Hegel (Studies in
Hegelian Cosmology, Cambridge 1901)
17
and on religion (Some Dogmas of Reli-
gion, London, 1906, 2nd ed., London
1930). In the latter book, he examined a
number of dogmas especially relevant to
Christianity. Unlike Francis Crick, who
misunderstood the meaning of 'dogma',
by 'dogma' McTaggart meant a proposi-
tion baving metaphysical significance.
God was not a dogma he had in mind,
and he thought that there is no real con-
nection between theism and personal
happiness as is often believed.
Other works include The Nature of
Existence where he examines various en-
tities that exist according to common ex-
perience, and attempts to account for the
apparent existence of entities that fail to
satisfy his requirement for existence.
1. B. Priestley, who happened to at-
tend McTaggart's lectures at Cam-
bridge, has the following to say about
him.18
His presence was delightful: He had a curi-
ous high voice, a large moon-baby face with spec-
tacles on the end of its nose, and he held his head
to one side and stared up at the ceiling while he
illustrated his argument by references to pink
elephants. He was one of the grcat originals of
Cambridge. Some odd disability gave him a crab-
like walk, and one met him coming sideways
around buildings, like a sheriff about to shoot it
out with the bad man in a Western. He had a
passion for reading novels, any kind, thousands
and thousands of them; and it was said he kept a
special edition of Disraeli's novels to read in his
bath. He believed in human immortality, was a
staunch supporter of the Church of England, and
having cheerfully argued God out of the universe,
he was an atheist. Now let us see how this extraor-
dinary man deals with Time.
18 CHAPrnRI

2. 3 McTaggart's paradox

In reality time is not conceived using spatial relations, rather it is manifested through action.
Vatsyiyana
Nyllya SUtra Bhil!Ya 11.1.40

What you say is quite true; but as a matter of fact, 'when the connection of the circumstances
with the Time is actually here, we have the idea of "present"; when that connection is destroyed
we have the idea of "past"; and when that connection is yet to come we have the idea of "future". '
This also is not right; as in your expression 'is there' if the Present tense is meant to be significant,
then it comes to this that you seek to explain the notion of the 'present' by means of the idea of
the present; so that you must land yourself either into self-reference or an infinite regress.
SriHar~a
KhaIJrJana KhalJrJa Khlidya JV.142

This suggests that B-relations are more objective. In the first part of his paradox,
McTaggart7 argues that it would be a mistake to think so. He argues that the A-series
is essential to time, and that A-determinations are more fundamental than B-relations.
Thus, he argues that time involves change, that there would be no time if nothing
changed.
And if anything changes, then all other things change with it. For its change must change
some of their relations to it, and so their relational qualities. The fall of a sand-castle on the
English coast changes the nature of the Great Pyramid.
What, then, is change? We find Mr. Russell's views on this subject in his Principles of
Mathematics, Section 442. 'Change is the difference, in respect oftruth or falsehood, between
an entity and the time T, and a proposition concerning the same entity and the time T',
provided that these propositions differ only by the fact that T occurs in the one where T'
occurs in the other.' That is to say, there is change, on Mr. Russell's view, if the proposition
'at time T my poker is hot' is true, and the proposition 'at time T' my poker is hot' is false.

McTaggart elaborates on Russell's definition of change, and rejects it as follows.

It will be noticed that Mr. Russell looks for change, not in the events in the time-series, but
in the entity to which these events happen .... If my poker, for example, is hot on a
particular Monday, and never before or since, the event of the poker being hot does
not change. But the poker changes, because there is a time when this event is happening
to it, and a time when it is not happening to it.

McTaggart maintains that

this makes no change in the qualities of the poker. It is always a quality of the poker that it
is one which is hot on that particular Monday. And it is always a quality of that poker that it
is one which is not hot at any other time. Both these qualities are true of it at any time - the
time when it is hot and the time when it is cold. And therefore it seems to be erroneous to
say that there is any change in the poker.
Let us consider the case of another sort of series. The meridian of Greenwich passes through
a series of degrees of latitude. And we can fmd two points in this series, S and S' , such that
the proposition 'at S the meridian of Greenwich is within the UK' is true, while the
proposition 'at S' the meridian of Greenwich is within the UK' is false. But no one would
say that this gave us change. Why should we say so in the case of the other series?
 PHILOSOPHICAL TIME 19

Hence the contention that there is no change unless facts change, and so there can
be no change without an A-series. Hence, also, it is a mistake to regard B-relations as
more objective. McTaggart concludes that it is the A-series that is fundamental, since
'if we remove the A-series from the prima facie nature of time we are left with a series
which is not temporal, and which allows change no more than a series of latitudes does.'
The second part of McTaggart's paradox consists of a proof that the A-series is
contradictory and so cannot be a part of reality.
Thus, any event has three A-determinations, which are mutually exclusive. For, if
an event is past it cannot, obviously, be either present or future. But this is a contradic-
tion, since any event has all three A-determinations.
We cannot escape from the paradox by suggesting that an event M has these
A-determinations at successive moments of time, e.g., by saying that M is present at the
present moment of time, was future at past moments of time, and will be past at future
moments of time. For, each moment of time is itself an event, and so has all three
A-determinations.
To get rid of the paradox in this way, it would seem that we must introduce another
time, time-2, so that each moment of time has three A-determinations at successive
moments of time-2. But, if time-2 is to be time, moments of time-2 must also form an
A-series so that we again have a contradiction, unless we introduce time-3. Since the
argument applies at any stage, the first A-series never escapes from contradiction.8
To summarize, McTaggart's paradox consists of two parts. In the first part, McTagg-
art argues that time involves change so that the A-series is essential for the existence
of time. In the second part, McTaggart argues that statements about an event M - that
it is present, will be past, and has been future - mean that M is present at a moment
of present time, past at some moment of future time, and future at some moment of
past time. But moments of time are equally events, and so cannot be permanently
characterized by pastness, presentness, or futurity. Hence, the A-series is contradictory,
since each term has all three mutually exclusive A-determinations. Therefore, McTagg-
art concludes, time is unreal.

Box 3. Sri Har~a (c 800)

KhCllJrJana Kh<U}q,a Khlidya means breaking into bits and devouring (the opponents
arguments). This carried to an extreme the tradition of argumentation in India, where
the opposing position (p(uva p~a) was first stated, and then refuted. In Sri Har~a's
case, the opposing point ofview was that of the ancient tradition ofNytiya- VaiSqikawbich
held that time is real. Sri H~a was a supporter of Srupkara'sAdvaita Vedanta position,
which maintained that the whole world is unreal (in an ultimate sense), mayd being
responsible for the illusion of reality. Sri H~a's idea was to use the tools of Nyflya to
demolish Nyflya. Though the traditional Buddhist challenge to Nyaya physically disap-
peared a little after Sal]lkara, the responses to these arguments led to an identifiably
distinct system of philosophy, known as NavyaNyaya (lit. the new Nyflya).
As usual, almost nothing is known about his personal life. According to one
anecdote, h~ started writing poetry at an early age. When Sri Har~a took his poems
to his uncle for comments, the uncle apparently responded: "You should have
shown me these poems earlier. I was preparing a treatise on poetry, and I was sorely
on the look-out for examples of bad poetry." According to the anecdote, Sri Harsa
threw his poems into the river (and very nearly threw himself in). .
20 CHAPI'ERI

2. 4 Resolution of McTaggart's paradox

One way to resolve McTaggart's paradox, proposed by B. Russell, c.J. Ducasse, A

Griinbaum, AJ. Ayer, J.J.C. Smart, and others, is to accept McTaggart's argument for
the unreality of the A-series. These B-theorists, drawing support from relativity, argue
that the flow of time is illusory, and describe it by appellations such as 'the myth of
passage'.
To be sure, we have an overpowering impression that time moves.9 But this may
be likened to the phenomenon of parallax. Sitting in a moving train, we have the distinct
impression that the distant mountains or stars move along with the train, while nearby
object whiz past. The nearby lamp-post, however, is usually at rest with respect to the
distant mountains. So physics cannot accommodate the relative movement that we
actually perceive.
Rather, one studies visual perception: the human eye judges distances through the
phenomenon of parallax. So, we say that the eye (or the brain) perceives relative
movements between objects that are actually relatively at rest. However, there is no
obvious organ, like the eye, for temporal perception. So studies of temporal perception
are still at a rudimentary stage, and one is unable to account for the illusion of a moving
time.
Physics is not really concerned with human perceptions (e.g. under the influence
of drugs). Rather, it is concerned with objective reality, and the objective reality is that
the B-series correctly describes the structure of time. In the words of Hermann Weyl,
'The objective world simply is, it does not happen.'
The first part of McTaggart's argument is dismissed by B-theorists as a dispute over
terminology. According to McTaggart, in a B-series there is no change, hence B-rela-
tions are not temporal relations. In other words, while B-relations may well exist,
McTaggart would not call them temporal relations since the word 'time', in common
language, has a distinctly different meaning. But just as physical theories are not overly
concerned with sense-perceptions, so physical usage need not be tied down to common
usage dictated perhaps by these sense perceptions. The word 'yellow', for instance,
might have an altogether different meaning in physics; it might be defined in terms of
light of a certain frequency, and a colour blind person may simply fail to perceive yellow.
To be sure, spatial relations are extrinsic, while temporal relations are intrinsic.
(Thus, to the audience in front of the stage Desdemona may be to the left of Othello,
but from behind the stage it would seem that Desdemona is to the right of Othello. As
opposed to this, if M is earlier than N, then this seems to be the case for all observers.)
But the reasons for the intrinsic order in temporal relations are to be sought in physical
asymmetries. The future is physically different from the past, whereas the left of some
point need not be physically different from the right of some point.

3 The sea fight tomorrow

There are A-theorists, like AN. Prior,lowho would not be convinced by this answer to
McTaggart's paradox. Modern A-theorists, deriving their support from quantum me-
 PHILOSOPHICAL TIME 21

chanics, argue that the future is logically (and not just physically) different from the past.
Thus, consider Aristotle's perplexity over the sea fight tomorrow.
Aristotle argued that it certainly is true that either there will be a sea fight tomorrow
or there will not be a sea fight tomorrow. For, a statement of the form 'Y or not-Y'
always is a tautology. But to say that 'Y or Z' is true means that either Y is definitely
true or that Z is definitely true. But, if it is definitely true that there will be a naval battle
tomorrow, then a sea fight is fated to occur tomorrow, irrespective of what the naval
commanders may do in the meanwhile. The same argument applies if it is definitely
true that there will not be a sea fight tomorrow. This is contrary to the common belief
that the past is determined and unchanging, while the future is contingent and depends
upon human choices.

3. 1 Fatalism and determinism

To see the point of Aristotle's argument, suppose I am ill. Then it is true that 'I will die
or I will not die.' If it is true that 'I will die,' there is no point in my going to a doctor. If,
on the other hand, it is true that 'I will not die,' then a doctor is somewhat redundant.
It is even true that 'I will suffer or I will not suffer.' And the above argument extends to
show that there is no point in going to a doctor to help alleviate my suffering even. Thus,
I will go to a doctor only if I am fated to go to a doctor.
There are people who try to draw a careful distinction between this kind of 'fatalism'
and the determinism of science. The argument runs somewhat as follows. 'I will not die'
may be perfectly true, and one may legitimately infer that '(Even) in don't go to a doctor
I will not die.' But this does not make doctors superfluous, for the last assertion is
perfectly compatible with the belief that 'If I don't go to a doctor I will die,' and one
may have excellent empirical grounds to believe that. Indeed, it may be that I will not
die is true simply because I intend to go to a doctor. Therefore, from the mere fact that
'I will not die' is true one may not infer that doctors are superfluous. Likewise, from the
truth of 'I will die' one may not conclude that doctors are inefficacious, and determinism
is not the same as fatalism.

3. 2 Dumett's paradox

An interesting reply to this argument is given by Dumett.ll The example concerns a

tribe which has the quaint custom that young men are sent to hunt lions to prove their
manhood. They travel for two days, hunt for two days, and travel back another two days.
The chief of the tribe dances for six days, not because he believes that dancing has some
magical properties, but because he mistakenly holds the system of causal beliefs that
dancing-in this fashion causes the young men to act bravely.
After four days have elapsed, suppose we approach the chief and try to persuade
him to stop dancing. We argue that the young men either have been brave or they have
not been brave, and nothing now can alter that. The chief replies 'If the young men have
been brave, then it is true, in a sense, that if I do not dance they will have been brave.
But this is not inconsistent with my belief that if I do not dance they will not have been
brave. In fact, two years ago, I fell ill, and had to stop dancing after two days. And when
the young men returned they had not been brave. Indeed, although the young men may
have been brave, they will have been brave just because I intend to go on dancing.'
22 CHAYIERI

The chief's argument seems ridiculous, but at the level of formal logic it is identical
with the argument for a distinction between fatalism and determinism. Both arguments
have the following form.
1. Assume p (p: 'I will not die' or 'The young men have been brave').
2. Hence, q => P (q: 'I don't go to a doctor' or 'I don't dance').
3. But q => P is not inconsistent with q => -po
4. There are empirical grounds to believe q => -p (and 4'. -q => p).
5. Hence -q (from 2 and 4).
6. Hence p (from 4').
So, it would seem that if we accept the hair-splitting between fatalism and deter-
minism, we are also forced to accept that the past, too, depends upon human choices,
a conclusion that is not too palatable.
Of course, one may argue that most of physics is deterministic anyway, so one need
not be overly concerned with a contingent future and quaint notions like 'free will'. One
problem with this argument is that though science may be deterministic, the philosophy
of science is not. As Hawking and Ellis 12 remark, 'all of our philosophy of science is
based on the assumption that one is free to perform any experiment.' If, for example,
it is already fated that a certain experiment will never be performed, the empirical
justification offered for scientific theories falls flat.

4 Logic and 'free will'

Therefore, some philosophers at least (particularly A-theorists) would like to believe

that the future is contingent or open. Aristotle, motivated largely by his belief in human
free will, hinted that ordinary two-valued logic was inconsistent with a contingent future.
The disjunction 'there will be a sea fight tomorrow or there will not be a seafight
tomorrow' he regarded as true, but each disjunct, for instance 'there will be a seafight
tomorrow', he regarded as neither true nor false, but only 'possible' or 'contingent' or
'indeterminate'. Aristotle, however, appears to have overlooked that the notion of
'possible' is not unproblematic as illustrated by the following.

4. 1 The Master Argument of Diodorus Cronus

The Master Argument begins by pointing out the incompatibility among the following
three propositions:
1. Everything that is past and true is necessary;
2. The impossible does not follow (from? after?) the possible;
3. What neither is nor will be is possible.
Having demonstrated the incompatibility,13 Diodorus used the plausibility of (1)
and (2) to conclude that (3) must be false so that nothing is possible which neither is
nor will be true. In other words every possibility is realized in the present or the future!
In the words of Cicero '[Diodorus] says that only what either is true or will be true is a
possibility, and whatever will be, he says must necessarily happen and whatever will not
be, according to him cannot possibly happen. You say that things which will not be are
 PHILOSOPHICAL TIME 23

also possible - for instance it is possible for this jewel to be broken even if it never will
be ... '

4.2 Lukasiewicz's three-valued logic

Though the preceding arguments are all logical, a disturbing feature is that they are
entirely so. It seems a bit perverse that logic alone, without reference to empirical facts,
should determine the nature of physical reality. Perhaps the root of the difficulty lies
in the formal apparatus of the logic used. Since we shall encounter a somewhat similar
point of view, again, in connection with quantum mechanics, we give below one such
theory of a three-valued logic developed by the Polish logician Jan Lukasiewicz.14
Theusualdefinitionofthelogicalconnectives- (not), 1\ (and), V (or), =>(implies)
may be set out in a truth table (Table 1). Here p and q are assertions that are either
true (T) or false (F). The truth-value of any assertion compounded fromp and q can be
read off, e.g. p and q is true iff both p and q are true.
Table 1: Truth tables for
-p P Aq pVq P =:.q p~q
ordinary two-valued logic

For each value of p given in the

first column, the two possible
p \q - T FT FT FT F values of q are given in the sec-
ond row. The truth-values for
any connective are in the bottom
F T F FT FT TF T two rows under that column.

T F T FT TT FT F

As distinct from this standard system of two-valued logic, the assertions in

Lukasiewicz's three-valued logic have the values T (true), F (false), or I (indetermi-
nate). The connectives are now defined by another table (see Table 2).

A tautology is still an assertion, such as p => (q => p), which always takes the value
T, no matter what value is ascribed to p, q. A tautology in the three valued system is
always a tautology in the two-valued system, but the converse is not true. For example,
if p takes the value I, so do -p and also p V -po

Thus, Aristotle's argument fails at the first step: an assertion, such as 'there will be
a sea fight tomorrow or there will not be a sea fight tomorrow,' remains indeterminate,
until tomorrow, and then becomes either true or false.

A three-valued system, in which some propositions remain indeterminate up to a

time t, and thereafter become true or false, does not exactly embody the view imputed
to Aristotle (since p or -p fails to be a tautology). A more serious problem is that if
p V -p fails to be a tautology then p 1\ -p fails to be a contradiction.
 24 CHAPTER I

Table 2: Truth tables for

-p pAq pVq p~q poq Lukasiewicz's 3-valued
logic
This table is read exactly like
p \q - TIF TIF TIF TIF Table 1, except that p and q
now have three values each.
Withthissystem,p V -pdoes
T F TIF TTT TIF TIF not remain a tautology. A
somewhat similar system was
used by Reichenbach in his
interpretation of quantum
I I IIF TII TTl ITI mechanics.

F T FFF TIF TTT FIT

4. 3 The quasi truth-functional system

One can get around this difficulty by using a more sophisticated quasi truth-func-
tional (q.t.f.) system (see Table 3).
Technically speaking, this language is obtained by taking the Cartesian product of
two two-valued logics to obtain a logic with four truth values (T,T), (T, F), (F, T), (F,
F). One then compresses this to a three-valued logic by treating (T, T) = T, (T, F) =
(F, T) = ?, (F, F) = F. The semantic interpretation of such a system is that (at any time)
there are two possible worlds. Indeterminate (?) values are assigned to those proposi-
tions that are true in one world and false in another.
Such a picture clearly evades the conclusion of the Master Argument: something
is declared to be possible which neither is nor will be, e.g., the sea battle tomorrow
neither is nor may be, but today it is declared to be possible. Nevertheless, one does not
have to reject the propositions (1) and (2) because the incompatibility between the
three propositions depends on the law of the excluded middle (strictly, the law of
bivalence), which is not true in the above language.
This language also shows a way out of the problem raised by Aristotle's sea battle.
Thus, in the above language p v -p is always a tautology. But this does not force either
p or -p to be true - both may be indeterminate. Thus, it is possible to have a language
or logic in which the future is treated as logically different from the past, and whether
or not this is the case is something that is best decided empirically.
To recapitulate, the A-theorist's answer to McTaggart's paradox is that the future
is logically different from the past. In support of this they point out the common belief
in free will (on which the philosophy of science is based). They also point out (or should
do so) that Aristotle's sea battle and the Master Argument of Diodorus Cronus lead to
the perverse conclusion that purely logical considerations are sufficient to decide
whether or not the world is deterministic. Therefore, the A-theorists argue, the future
must be treated on a logically different footing from the past.
 PHILOSOPHICAL TIME 25

Table 3: Quasi
-p P Aq pVq truth-functional system

The q.t.f. system cannot

be defined using a truth
p \q - T ? F T ? F table. The ? should not
be treated as a third truth-
value. This table should
T F T ? F T T T be seen only as an anal-
ogy. With this system,
p V -p remains a tautol-
ogy.
? ? ? (1orF) F T (1orT) ?

F T F F F T ? F

Aristotle's argument shows that the truth-value of a proposition must be allowed

to vary with time. The Master Argument shows that the key to the matter is that one
must have a language in which events are possible even though they may never occur.
One such language is the q.t.f. system.

Table 4: The conditional

p~q poq in the q.t.f. system

This table shows a possi-

ble 'definition' of the con-
p \q T ? F T ? F ditional, using p ~ q for
-pVq, andpoq for
(p~q) A(q~p). Thepre-
T T ? F T ? F cise definition of 'if is very
important for axiomatic
quantum mechanics.
? T (1orT) ? ? (1orT) ?

F T T T F ? T
26 CHAPTER I

5 Zeno's paradoxes of motion

Just as the atom is the minimal limit of matter, so the instant is the minimal limit of time. Or
the time taken by a moving atom in order to leave one point and reach the next is an instant.
The instants form a sequence ca/led time. Two instants cannot be simultaneous because it is
impossible that there be a sequence between two things that occur simultaneously. Thus, in the
present there is a single moment, and there are no combinations of earlier or later moments.
Accordingly, the whole world mutates in any single instant ....

Vyasa (c 300 BC)

Yosu BlWfya III.52

Zeno's paradoxes of motion15 are more than 2400 years old. Many current texts in
mathematics tend to dismiss Zeno's paradoxes, giving the impression that they are
adequately refuted by the theory of real numbers or the continuum. But, granting the
internal consistency of the theory of the continuum, there is a deeper point in connec-
tion with these paradoxes: they pertain to motion and time. Naively speaking, quantum
theory shows that energy transfers at the atomic level are discrete, but it is far from
clear, today, whether space and time must be treated as 'discrete' in some sense.

As opposed to this, the idea that moments in time form an ordered continuum, like
points on a line, has been of great importance in physics ever since Galileo, who
represented time by a straight line. The intuitive idea is that of Isaac Barrow, who
asserted that 'time has length alone, is similar in all its parts and can be looked upon as
constituted from a simple addition of successive moments ... .' Time is also required to
be a continuum, since the calculus and modern analysis are both based on the notion
of the continuum, so that neither Newton's laws nor the Schrodinger equation could be
written down without appealing to the concept of the continuum. 16

The mathematical theory of the continuum is, today, straightforward enough except
that it is now being applied to the physical world. To see the kinds of problems that this
might create, consider the following. Any interval, say [0,1], can always be cut into two,
in such a way that the result of cutting it is always two intervals, say [0, liz], and [liz, 1].
Now let us apply this to a lump of chocolate: (a) any lump of chocolate may always be
cut into two; and (b) the result of cutting a lump of chocolate into two always consists
of two lumps of chocolate. Clearly, while the theory of the continuum is logically
consistent, it does not quite apply to lumps of chocolate. Thus, a molecule of chocolate
is either not a lump of chocolate, in which case (b) is false, else (a) is false.

Thus, Zeno's paradoxes apply to physical theories which assume that time is either
continuous or discrete. There are four paradoxes in all: Achilles and the tortoise,
dichotomy, the arrow, and the stadium. Though there was no clear concept of the
continuum in Zeno's time, it is believed that the first two pertain to continuous time,
and the last two to discrete time.
 PHILOSOPHICAL TIME 27

Box 4. Zeno of Elea s regarded as the atom of length and t

Not to be confused with Zeno of Citium,
also known as Zeno the Stoic. In his
Parmenides, Plato represents Zeno as
being "close to forty" when Parmenides
was 65 and when Socrates (born 469
B.c.) was a young man. Zeno, the disci-
ple of Parmenides, has been credited
with the origin of the dialectic, Le., the
method of seeking knowledge by ques-
tion and answer, or dialogue, as in the
Upanishads. This method was later de-
veloped and extensively practised by
Socrates. In Plato's Parmenides, Zeno
subjects Socrates to the same kind of
treatment to which, elsewhere in Plato,
Socrates subjects others.
Today, he is remembered princi-
pally for his paradoxes of motion! One of
the lesser-known paradoxes, the sta-
dium or the moving blocks, concerns
three blocks A, B, C with edges of equal
lengths. A is stationary, while B and C
move past A in opposite directions at
equal speeds. B will traverse the distance
s both in time t and also in time t/2 (since
it crosses the block C in this much time).
It would seem logical to regard this as a
paradox against discrete space-time with
regarded as an atom of discrete time,
though this interpretation is disputed. In
any case, one thing is clear: regardless of
Zeno's intentions, the paradox certainly
can be applied to simple-minded quanti-
zation of space time.
Regarding the arrow paradox,
Bertrand Russell says the following.
Zeno belonged to the Eleatic school whose object
was to prove that there could be no such thing as
change. The natural view of the world is that there
are things which change; for example, there is an
arrow which is now here, now there. By bisection
of this view, philosophers have developed two
paradoxes. The Eleatics said there were things but
DO changes; Heraclitus and Bergson said there
were changes but no things. The Eleatics said
there was an arrow, but DO flight; Heraclitus and
Bergson said there was a !light but DO arrow. Each
party conducted its arguments by refutation of the
other party. How ridiculous to say there is no
arrow! say the "static" party. How ridiculous to
say there is no flight! say the "dynamic" party. The
unfortunate man who stands in the middle and
maintains that there is both the arrow and its flight
is assumed by the disputants to deny both; he is
therefore pierced, like St Sebastian, by the arrow
from one side and by its flight from the other.

5. 1 Achilles, dichotomy and the arrow

Neither in the past nor in the future are you flying, 0 crow! And if you are flying ~ the
Himalaya is moving, this entire world is flying ...
Pataiijali (c 500 BC)
Vy~ Mahdb~a 1/1.2.123

The Achilles paradox is well known. The Greek god Achilles, famed for his swiftness,
patronizingly gives the proverbially slow tortoise a start of ten meters in a hundred
meters race. Achilles can run ten times faster than the tortoise, so this is not much of a
start, and Achilles is sure to win. But how does Achilles catch up? By the time the
tortoise has covered one meter, Achilles has covered ten, and the gap is reduced to just
one meter. Even this meter Achilles covers by the time the tortoise has moved another
ten centimeters. And that ten centimeters he covers by the time the tortoise has moved
only a centimeter; and that centimeter he covers by the time the tortoise has moved a
28 CHAPTER I

millimeter. But by the time he moves that millimeter the tortoise is still ahead by a tenth
of a millimeter, and by the time Achilles covers that distance, the tortoise is still a
hundredth of a millimeter ahead. So it seems that the tortoise is always just a little bit
ahead, and Achilles would take an infinite amount of time to catch up with, leave alone
overtake, the tortoise.
The assumption here is that the order of temporal events is a dense order: between
any two events there is always a third. So to say, the motion (or time) is assumed to be
continuous. To go from point A to point B one must necessarily traverse any point C
in-between. This point is brought out most clearly by the companion paradox of
dichotomy. To go from point 0 to 1, one must first travel from 0 to liz. To go from 0 to
112 one must first go from 0 to 14 and so on. Thus, to go from 0 to lone must perform
an infinite number of journeys, which is impossible - so no motion is possible and
motion, hence time or change, is an illusion, in keeping with Eleatic philosophy. It is,
of course, an observed fact that Achilles overtakes the tortoise, and one does manage
to perform a few journeys now and then. So what exactly is wrong here?
A simple way to get around these two paradoxes is to say that Achilles moves in
steps of discrete size and so overtakes the tortoise. Likewise, in dichotomy, it is
meaningless to talk of journeys smaller than one step. The arrow paradox is aimed
against this simple refutation. Suppose all motion takes place in steps, so that the arrow
moves from one position to the next in discrete steps. So, at every point of time, during
its flight, the arrow is still (or motionless). Hence, the motion of the arrow (and so also
time) is an illusion.
The simple refutation of Achilles not being quite satisfactory, one might try to apply
the more sophisticated theory of the continuum. In Achilles as well as in dichotomy we
are asked to contemplate an infinity of motions of ever shorter duration. The sum of
this infinite series of durations is finite. Therefore, Achilles will overtake the tortoise
in a finite distance, equal to the sum of the series 1 + 0.1 + 0.01 + ... = 1.11111...,
and in a finite period of time obtained by dividing the above distance by his speed.

5. 2 Tasks and supertasks

Does the application of the theory of the continuum really imply that all of an infinite
series of tasks can actually be performed? Surely it is one thing for an infinite series to
have a sum and quite another to sum an infinite number of terms.
Suppose it were possible to perform such a supertask - an infinite series of tasks.
Consider the case of the Thomson lamp. On pressing the switch the lamp goes off if it
is on and on if it is off. Suppose the Thomson lamp is off. An odd number of jabs at the
switch would culminate in the lamp being on, while an even number of jabs would leave
it off. Now, suppose we jab the switch at intervals of 112,14, Vs, 16, .... minutes. At the end
of 1I2+14+Vs+16+ ... =1 minute what is the state of the lamp? Such questions are
intimately connected with the renormalization problem of quantum field theory where
one deals with divergent integrals 17 rather than divergent series, such as
1-1+1-1+ ....
One might argue that after Xo jabs the lamp can be neither off nor on, therefore
the process just described is logically impossible: 1-1 + 1-1 + ... cannot be summed.
The conclusion would be that supertasks are not, in general, logically possible. Or one
might argue that at the end of one minute the lamp would, on an average, be on half
 PHILOSOPIDCAL TIME 29

the time and off half the time, so that it would shine with half its intensity:
1-1+1-1+ ... = \12.
Or again, one might make a detailed physical analysis. If the switch has to move an
equal distance each time, then the amount of energy for each jab would be constant so
that the process would require an infinite amount of energy. So, each time, the switch
must move over shorter distances, in such a way that the series of distances converges.
Hence the nth distance must tend to zero. Therefore, after one minute the switch would
not be moving at all, hence the lamp would be on. The trouble with this last argument
of Griinbaum is that one could perfectly well arrange for the circuit to be broken when
the switch makes contact so that finally the lamp could just as well be off.

6 Conclusions

Alice sighed wearily. '1 think you might do something better with the time,' she said, 'than
wasting it in asking riddles that have no answers. '
'If youknew Time as well as I do, , said the Hatter, you wouldn't speak of wasting it It's him. '

Lewis Carroll

Time, of course, is real: ask any woman who has just seen the first wrinkle on her face ....

R.M.Gale
The Philosophy of Time

With all this, however, there must be some difference among the several points of time; as it
cannot be denied that people have the conception of past, present and future. And thus the
existence ofsome sort of difference, in a general way, being an established fact, if we are unable
to determine what that exact difference is, all that this can justify us to do is to say that the exact
nature of the difference is doubtful. This is not right, we reply; as you cannot explain the exact
nature of Doubt either....

Sri Halla
KluuJf/.ana KluuJr!a KMdya

To conclude, no immediate conclusion may be drawn with regard to any of the questions
raised at the beginning of this chapter. Specifically, it is not clear what the definition of
time must be and whether at all any definition is necessary or possible. It is not clear
whether time should be represented by an A-series or a B-series or both as in common
language. It is not clear whether human freedom is an illusion, nor whether the
philosophy of science is based on the invalid assumption that one is free to perform any
experiment. It is not clear whether the future is physically or logically different from
the past nor what system oflogic should be used, if such is the case. It is not clear whether
the concept of the continuum may be applied to physical time.
B. The measurement of time
II NEWTON'S TIME
ABSTRACf. In the previous chapter, a number of paradoxes concerning time were brought out - deeper
problems may yet exist! Here we take up the study of time in Newtonian physics. In the absence of an
independent definition of force and an independent measure of time, Newton's laws of motion, by
themselves, fail to be falsifiable, and hence lack physical content. However, the combination of the laws
of motion and the law of gravitation leads to refutable consequences and good physics, because both
unknowns may be eliminated. We conclude with a discussion of Laplace's demon, and point out that
standard ways to exorcise the demon lead, in fact, to unpleasant time-symmetric contingents.

1 Introduction

Twith
HOUGH philosophers have written much on time, and have analysed many issues
considerable skill, their analysis has remained preliminary since they have
failed in one important respect. They have tended to rely overmuch on the everyday
experience of time. A subtle contradiction may well exist between the mundane
experience of time and other mundane experiences such as seeing leaves falling lazily
from a tree, or basking in the warmth of the sun on a winter's day, or watching the stars
at night and seeing but failing to notice the dark background.
The possible existence of such a contradiction between seemingly unrelated aspects
of our experience is demonstrated in the subsequent parts of this exposition. It may
happen, as it often does in physics, that to resolve this contradiction, we may need
experiments that take us very far beyond the limits of human perception. What seems
a solid brick wall to a human being would appear to be vast tracts of vacuum to a
neutrino. Similarly, the familiar fabric of everyday time, which clothes so many human
languages, may dissolve at different levels of perception.
To examine the possible existence of such a contradiction, it is necessary to study
the notion of time in physical theories. One also expects that such a study would help
to answer some of the questions that were left unanswered at the end of the preceding
chapter.

2 Newton's laws of motion

Let us bepn our study of time in physics with something that everyone understands:
Newton's laws of motion. On the face of it, Newton's laws of motion fall into the
historical tradition of belief in a 'natural' state of motion together with 'causes' (forces)
required for departure from this state. Aristotle, and some of his contemporaries,
thought that the 'natural' state was motion in a circle with uniform speed. Their reason
was roughly that the circle was the perfect geometrical figure, and the heavenly bodies
were perfect. With this point of view, there was a difficulty with the motion of an arrow
- how did the arrow keep moving with only an initial 'force' or cause? For Descartes,
the straight line was the natural geometrical figure (and state of motion).
34 ClIAPTERII

Galileo had used the notion of rectilinear inertia in his derivation of parabolic
trajectories. But these trajectories gave a good description of only projectile motion,
clearly planetary orbits are closed. Galileo had also used something like Newton's third
law in studying the motion of objects on an inclined plane. But Galileo had rejected the
Aristotelian dichotomy between natural and forced motions, he had sought to introduce
a third category of 'neutral' motions. Ironically, Newton in his Principia credited Galileo
with the two laws of motion.
The first two of Newton's laws of motion state:
1. A body continues in its state of rest or uniform motion in the absence of external forces.
2. When an external force F is impressed upon a body ofmass m its acceleration a is given
by F = ma = mdv/dt, v being the velocity.
In the traditional view, the first law characterizes the inertia of a body or its
tendency to stay on course in the absence of external interference. The mass of a body
is a measure of its inertia or 'resistance to movement or deflection'. The notion hardly
needs any explication for we have an immediate intuitive understanding of it: we can
put an eight pound shot much further than a 16 pound one.
The mass of a body is not, of course, quite the same as its weight. If one were to
jump from an aeroplane with a 1 kg mass on one hand, and a 2 kg mass on the other,
both would seem equally weightless and it would be a bit more difficult to measure
either mass (till such time as one opens the parachute or hits the ground!). But it would
be fair to say that if we are standing on earth with a balance, two weights declared equal
by the balance will correspond to equal masses.

2.1 Bodies

Newton's second law, in the traditional view, tells us about the rate at which the velocity
of a body, of a given mass, changes when a force is applied to it.
There are, of course, a few points that need some clarification. What exactly is a
body? My car seems to start perfectly well without the need of an external push, though
it does need petrol from time to time. To be sure, a car would not move on a road in
the absence of friction, but what about a rocket? One might say that the rocket ejects
something, and the resulting reaction should be regarded as an external force exerted
by the ejecta on the rocket. But what of a time bomb which explodes without any
external help? Worse still, imagine a stiff rod or a wheel which is set rotating in deep
space. Conservation of angular momentum assures us that the rod/wheel will keep
rotating in the absence of external forces. But the wheel is neither in a state of rest nor
in a state of uniform motion.
It might seem that there is a bit of cheating in the last two cases. After all, some
external force was required to set the wheel moving in the first place. But Newton's law
are indifferent to the past history of a body. The first law might be stated: at any instant
a body continues in its state of rest or uniform motion in the absence of external forces
at that instant. A quick look at the second law would confirm this alternative form of
the first law.
One way out is to declare that time bombs, gyroscopes, cars and rockets are not
bodies at all - they consist of many smaller bodies. But that brings us back to the
original question, what exactly is a body? Molecules are held together by chemical
bonds that may snap, sending the constituent atoms flying apart; a photon might come
 NEWTON'S TIME 35

whizzing out of an atom, even nuclei might decay, sending out particles and moving
backwards as a result.
As A. K. Raychaudhuri1 remarks, in his book on classical dynamics, he ventured to
ask his student whether electrons were after all 'bodies'. The student apparently
responded that he had thought his teacher to be intelligent and well-read enough to
know that electrons obeyed quantum mechanics rather than Newtonian mechanics!

2.2 Forces

Similarly, on the face of it there appears to be a straightforward contradiction between

everyday experience and Newton's laws. Imagine a drunkard who decides, after one
drink too many, to test Newton's laws by relinquishing his hold on his glass. The glass
does not stay in a state of rest, hanging in mid-air. Instead, it promptly falls to the ground
and shatters. Evidently, the inebriated gentleman had forgotten all about the force due
to gravity.
But what exactly is a force? A push or a pull one can actually feel, and one can
usually identify the manner in which it operates (except where pulls operate from
behind the scenes!). In the case of gravity, we can point to a very tangible source: the
earth. But, does a force have to have a tangible source? Here is an experiment that can
be performed while sober. Put a record on a turntable, put a small ball-bearing on top
of the record, and start the turntable. Very soon, the ball-bearing will be thrown off. In
this case the culprit is the centrifugal force. Anything that rotates experiences a
centrifugal force. The ball-bearing rotates along with the record, so it experiences a
centrifugal force and is thrown off.

2. 3 Sources offorces

The centrifugal force has no tangible source, so it seems a good idea to know when
exactly something rotates. Suppose, instead of switching on the turntable we whiz round
it at 33\1'3 rounds per minute, all the time keeping our eyes fixed on the record. The
record will certainly appear to rotate, but the ball bearing will stay put.
A classic example of this sort is Newton's bucket. Fill a bucket with water, tie it to
a rope and whirl it round. The water surface will become concave. As the bucket is
rotated faster, the surface will become more concave. On the other hand, it is not
possible to alter the shape of the water surface by running, no matter how fast, round
the bucket.
As yet another example, imagine sitting on a giant turntable which is whirling round
inside a dark box. One would have no direct sensation of rotation. And one would not
be able to assign any tangible source to the invisible hand of the centrifugal force trying
to push us off. Forces of this sort, to which no tangible source can be assigned, are called
fictitious forces.
The preceding example is not particularly fantastic. We live on the earth, and there
is day and night on the earth because the earth rotates. We have no direct sensation of
rotation, though we do observe tides in the oceans. So how do we discriminate between
the following two possibilities: (a) the earth rotates and (b) the sun goes round the
earth?
36 CHAPTER II

In the case of the turntable, the earth was an obvious reference point, and we could,
if we chose, distinguish between the two cases (i) turntable turning, and (ii) experi-
menter running around turntable. But in the case of the earth what shall we choose as
a reference point?

2. 4 Inertial frames

One possibility is to choose the distant (fixed t ) stars. A freely falling rigid body or a
frame of three vectors, at right angles to each other, which remains fixed, or moves with
constant velocity, with respect to the distant stars, is called an inertial reference frame.
'True' rotation is rotation with respect to an inertial frame.
The ball-bearing is thrown off when the turn table rotates with respect to the distant
stars. The water surface is concave when the bucket rotates with respect to the fixed
stars. The earth, by the way, rotates with respect to the distant stars; so the earth is not
an inertial frame.
To summarize, a centrifugal force will act only in case the rotation is with respect
to the fixed stars. This roughly amounts to saying that the distant stars are the 'source'
of the centrifugal force. (A similar statement was actually made by Ernst Mach, and
carried to its logical conclusion by Hoyle and Narlikar.2)

Box 1. Newton and action at a distance
Historically, many objections were raised to
Newton's laws of gravitation. One of the
problems of natural philosophy in Newton's
time was to account for actions transmitted
between bodies not in physical contact with
each other. The existence of such forces was
indicated, for instance, by the behaviour of
magnets or the connection between the po-
sition of the moon and the rise and fall of
tides. Aristotle had maintained that force
cannot be communicated except by pres-
sure or impact, and this was carried forward
by Thomas Aquinas, who set the scholastic
tradition.
Rene Descartes, who rejected the
scholastic tradition, was not prepared to ac-
cept these 'actions at a distance' as occult
influences. He postulated that bodies
can act upon each other only when con-
tiguous. This led him to believe that the
space between the moon and the earth,
and indeed the whole of space must be
filled with material of a kind more subtle
than ordinary material things like air.
Space wa , in Descartes' theory, a ple-
num, being filled with a medium, largely
imperceptible to the senses, but capable
of transmitting force and exerting effects
on material bodies immersed in it. He
called this material the aether, a word
which had originally meant the blue sky.
Descartes postulated that the aether had
mechanical properties, and was the pri-
mary occupant of the universe.

I(continued on p 37)

t The word 'fixed' is not tautological but has a technical meaning. The nearby stars exhibit an annual
parallax due to the motion of the earth around the sun.
 NEWTON'S TIME 37

Newton was faulted for having intro-
duced action at a distance and much
more. As Voltaire 9 wrote in 1730, 'A
Frenchman who arrives in London will find
Philosophy, like everything else, very much
changed there. He had left the world a ple-
num, and now he finds it a vacuum' He
continued,
It is tbe language used, and not tbe thing in itself
tbat irritates tbe buman mind. If Newton bad not
used the word attraction in his admirable philos-
ophy, everyone in our Academy would have
opened his eyes to the light; but unfortunately he
used in London a word to which an idea of
ridicule was attached in Paris, and on that alone
be was judged adversely, with a rashness which
will some day be regarded as doing very little
honour to his opponents.
How was Newton led to abandon
Descartes' idea that the only intelligible
'force' was force by impact? Newton's
correspondence with Hooke provides
some insight.
In 1679, Hooke, then the Secretary
of the Royal Society, wrote10 to Newton
asking for his 'objections against any hypothe-
sis or opinion of mine,' particularly 'that of
compounding the celestiall motions of the
planetts of a direct motion by the tangent and
an attractive motion towards the centrall
body ... .' Newton replied with 'a fun")' of my
own about. .. a spiral path that a freely falling
body would follow as it . ,. fell to Earth,' and
'spiralled to (orvery near to) the Earth's centre
after a few revolutions.' Hooke insisted that
the path would 'resemble an Ellipse'. Newton
now agreed that the body would describe
not 'a spiral to the very centre but circu-
late with an alternative ascent and
descent' but maintained that 'the body
will not describe an Ellipsoid'.
In a letter of 6 Jan 1680, Hooke men-
tioned the inverse square law, 'that theAttrac-
tion always is in a duplicate proportion to the
Distance from the Centre Reciprocall' but
stated that 'the Velocity ... as Kepler Supposes
Reciprocall to the Distance.'
Newton did not reply. But, being a
good mathematician, he knew that the
'approximate' law of speed (v IX 11r )
stated by Hooke was inconsistent with
Kepler's area law (the -speed should be
inversely proportional to the perpendic-
ular from one focus to the tangent at that
point), though the two quantities agree
at the apides.
Newton now found out that 'the
Orbs [of the planets] would be such El-
lipses as Kepler had described [when]
the forces which kept them in their Orbs
about the Sun were as the squares of
their ... distances from the Sun reciprocally.'
That is, Newton back-calculatedtheforr:e, given
the orbit.
The story did not have a happy end-
ing. Hooke claimed priority for having
discovered the inverse square law, and
Newton responded by threatening to
stop the publication of his Principia.

2. 5 External forces

Thus, an 'external' force mayor may not have a tangible means of operation (e.g. gravity)
and it mayor may not have a tangible source (e.g. centrifugal force) . All this is somewhat
puzzling. So a skeptical person, determined to test Newton's laws, is bound to enquire:
when can one say that external forces are absent?
More precisely, imagine a person shut up in a black box. Can he, by means of
experiments performed inside the box, and without reference to the world outside, say
whether or not an external force is absent?
38 CHAPTER II

The only reasonable answer to this question seems to be the following. Our
experimenter-in-a-black-box (EIABB) can determine whether or not a body initially at
rest stays at rest. If it does not, external forces are present; if it does, external forces are
absent.

2. 6 Newton's laws as definitions

So to our EIABB Newton's first law seems no more than a tautology. It appears to assert
1. If a body continues in its state of rest or uniform motion, then external forces are absent.
So also, the empirical content of the first law, as judged by possible experimental
tests of the law, seems rather meagre. Indeed, Newton's first law seems not so much an
empirical 'law' as a definition of the term 'force'. By the same reasoning, Newton's
second law defines the magnitude of the force.
2. When a body of mass m accelerates by an amount a, with respect to an inertial frame,
a force of magnitude F = ma is said to act on the body.

2.7 Clocks

Now there are definitions and definitions, some are good, and some are bad. How well
do Newton's laws define the concept of force?
Once again let us look at the situation empirically, from the point of view of our
EIABB. To determine whether or not a force is acting on a body, the EIABB must
determine whether or not the body continues in its state of rest or uniform motion. To
determine whether a body continues in its state of rest is an easy matter, and a body is
said to be in uniform motion if it covers equal distances in equal intervals of time.
Equal distances are clear enough, but what are equal intervals of time? Clearly, it
is not an easy matter to put two intervals of time side by side and compare them. The
best one can do is to declare that the durations between two ticks of a clock are equal.
But which clock should one use?
Are the times between two heartbeats equal? Evidently not - there are any
number of special circumstances in which one can feel the difference! Are the times
between sunrise and sunset equal? Evidently not, for the ecliptic is inclined to the
celestial equatorial plane, so days keep getting longer and then shorter. Are the times
between sunrise and sunrise equal? Again no, because tidal friction slows down the
rotation of the earth, and the earth must eventually stop rotating. (Even if they were
equal, it would be a problem to function with such a large unit of time or to divide a day
and night into 86,400 equal parts!)
Are the times between two swings of a pendulum equal? Unfortunately no, for the
exact theory of the pendulum shows that the period of the pendulum depends upon the
amplitude, and the amplitude changes because of friction.
It seems best to use atomic clocks. But the problem is not really one of precision.
The problem is, as Poincare pointed out, that there is a 'democracy of clocks' in
Newtonian mechanics. Pragmatically one may standardize, but, in principle, there is no
reason (in Newtonian mechanics) to prefer atomic clocks to heartbeats. Heartbeats may
be irregular with respect to atomic clocks, but atomic clocks are irregular with respect
to heartbeats. Any given clock keeps excellent time with respect to itself, and for any
given clock there exists another with respect to which it is imprecise. In fact, there has
 NEWTON'S TIME 39

been a serious suggestion (by Milne 3) that the spiral structure of galaxies arises from
the fact that galaxies keep different time from atoms.
Newton, expanding upon Isaac Barrow's 'even tenour' hypothesis, wrote that
'Absolute, true and mathematical time' flows uniformly on 'and by another name is
called duration: relative, apparent and common time is some sensible and external
(whether accurate or inequable) measure of duration by means of motion, which is
commonly used instead of true time; such as an hour, a day, a month or a year.' Hence,
Newton's laws of motion fail to be physical laws because they fail to specify both the
terms 'force' and 'clock'. In fact, Newton's second law fails, even as a definition offorce,
because the right hand side also remains undefined.

3 Falsifiability and physical theories

Since Newton's laws are part of every first course in physics, it seems necessary to specify
what is meant by 'physical'. Physics concerns empirical reality; physics concerns exper-
iment; and a law that cannot, in principle, be tested against experiments is not a physical
law.
This is a crude summary of Popper's demarcation of physical theories as those that
are falsifiable or refutable. Theories that may be tested by means of experiments are
empirically falsifiable. But, because there is always some possibility of error in an
experiment, and so some (philosophical) doubt regarding its outcome, Popper" settles
for the more modest criterion of logical falsifiability.
Logical falsifiability means that there is at least one conclusion of the theory that
may, in principle, be contradicted. As a famous example, consider the statement (A):
'All swans are white.' Suppose I find a bird which is like a swan in all respects except
that it is black. If one is prepared to accept that the existence of such a bird would
contradict statement A, then statement A is logically falsifiable. On the other hand, one
might hedge by saying that it is an 'essential' property of a swan that it be white; in that
case a black bird cannot, by definition, be a swan and statement A ceases to be
falsifiable, for it cannot be contradicted even in principle, and so cannot have anything
to do with the empirical world. The definition may possibly be based on a large number
of empirical observations, but that does not alter matters in any way, or change the
definition into a statement concerning empirical facts.
The last point may be somewhat confusing. Perhaps the confusion may be lessened
by reiterating that the simple fact here is that a definition does not cease to be a
definition even if it happens to be based on empirical facts! A definition may be well
formulated, well motivated, and generally agreed upon; for all that it remains a
definition. A definition may be 'good' or 'bad', but as a logical statement it is always
trivially true. Empirical facts, therefore, can never be used to test a definition.
As another example, consider the statement (B): 'All men are selfish.' If a man
gives his life to save that of another, would we accept this as an instance of altruism? or
would we engage in an in-depth Freudian analysis to demonstrate that the motives were
basically selfish? If the latter, then all men are selfish by definition, and the statement
B ceases to be falsifiable.
40 CHAPlERII

As a third example, consider the statement 'God exists'. If I pray to God and my
prayers are not answered, is that an indication that God does not exist? Or can we hedge
by saying that I am too insignificant for God to take notice? If someone really important,
like a Pope, prays to God and his prayers are not answered, what do we conclude? Can
we still hedge by saying that the Pope is equally insignificant in the eyes of God?
As another example, consider Newton'st 'even tenour' hypothesis 'Absolute, true
and mathematical time of itself, and by its own nature, flows uniformly on, without regard
to anything external... .' [Italics mine] This disregard for 'anything external' shows that
there is no way to refute the hypothesis of an 'absolute, true and mathematical time'
which is not, therefore, a physical hypothesis.

Box 2. Karl Raimund Popper Economics.From 1945 until his recent

Austrian philosopher of natural and so-
cial sciences, Popper was born in 1902 in
Vienna, and studied mathematics, phys-
ics and philosophy at the University
there. As a student he had the following
insight into induction: while a large num-
ber of experiments in favour do not quite
establish a theory, a single experiment
against suffices to prove the theory false.
The difficulty is that, in practice, theories
accumulate hypotheses to evade falsifi-
cation.
Although not a member of the fa-
mous Vienna Circle oflogical positivists,
and in sharp disagreement with many of
its views, he shared the group's philo-
sophical concerns and was in close touch
with members like Rudolph Carnap.
His first book Logic der Forschung
(1935), later translated as The Logic
of Scientific Discovery, was published
in the Circle's series.
In 1937 he went to Canterbury
University College in Christchurch,
New Zealand. In 1945 he moved to a
Readership at the London School of
retirement, he remained a professor of
logic and scientific method at the LSE
teaching a subject which he describes as
'non-existent'. He was knighted in 1964.
He has written a series of books and
essays (some quite provocative). Particu-
larly, he wrote on time for the joumalNa-
lUre, and published in 1983 a three volume
'Postscript'to The Logic of Scientific Dis-
covery, in one volume ofwruch he takes up
the question of human freedom
Popper believes that scientific or
empirical theories should be refutable.
Thomas Kuhn, on the other hand, proposed
that progress in scientific thought came
about through revolutionary changes in par-
adigms. Much has been made of the Pop-
per-Kuhn debate, and it should be
remembered that Popper is proposing a
normative view, while Kuhn is proposing a
historical view. In his Postscript, Popper
questions the historical validity of Kuhn's
view as well, suggesting that the growth
of quantum mechanics as well as rela-
tivity took place through the falsifica-
tion of Newtonian mechanics and
gravitation.

t The 'even tenour' hypothesis was actually introduced by Isaac Barrow, Newton's teacher and
predecessor to the Lucasian Chair, who declared that 'whether things move or are still, whether we sleep
or wake, time pursues the even tenour of its way.'
 NEWfON'STlME 41

A$ a final example, consider Newton's laws of motion. If we observe deviations

from these laws, would we be ready to accept that the laws might be false? Or would
we regard this as an indication for the existence of a fifth force?
According to the criterion of falsifiability, a theory is physical if at least one conclusion
derived from the theory is, in principle, refutable. If, on the other hand, every piece of
evidence that goes against the theory is accompanied by an ad hoc modification then the
theory is metaphysical, not subtle. The theory is metaphysical for it refuses to confront
physical reality and tries to evade it. Somewhat the same charge may be laid at the door of
Newton's laws of motion. When everyday objects apparently fail to satisfy Newton's laws
we introduce the force of gravitation. When magnets apparently fail to satisfy the laws, we
introduce the magnetic force, and so on up to a fifth or perhaps sixth force.
To summarize, there is no way in which Newton's laws of motion can fail to be
satisfied, since they are no more than definitions of the two unknowns 'force' and 'equal
intervals oftime'. They are not even good definitions since they attempt to define two
unknowns through one equation. There is no place in the universe where the laws are
supposed to hold exactly, and no object to which they definitely apply.

4 Laws of motion and law of gravitation

The argument from falsifiability, and the corresponding conclusions, seem to apply
equally to Newton's law of gravitation, F = Gmt m21r-. Here the right hand side can
be measured, but the left hand side contains the term 'force' which, as we have just seen,
is not defined. The law of gravitation, therefore, is no more than a definition of the
gravitational force, so no experiment can conceivably prove this law to be false.
But there is now one difference. If we use both, the laws of motion and the law of
gravitation, then the unknown quantity 'force' may be eliminated. And 10 and behold!
we have some physics at last! If we throw an object, such as a ball, into the air then,
according to Galileo, the centre of mass should describe a parabola, for he supposed
that the body would be subject to a constant vertical acceleration, so that its speed would
be characterized by v2 = -'lgh + constant. According to Newton's (both) laws, how-
ever, the body will fall with a varying acceleration, directed towards the centre of the
earth, and trajectory of the centre of mass should be an ellipse, while the velocity will
obey the equation

dv Mm (1)
m-=-G-
dt ,:z. ,

M being the mass of the earth, m that of the ball, and r the distance from the centre of
the earth. Fortuitously (!) we have the same m on both sides of the equation, so we may
happily cancel it and, since dvldt = dvldr . drldt = v dvldr, we obtain

(2)
42 CHAPIERII

which gives upon integration

v2(r) = 2G M + constant. (3)

r

If we write r = (R+h), R being the radius of the earth, and h the elevation from
the surface of the earth, and assume h to be small compared to R, then

(4)

so that

v2 = -2G ~ .h + constant. (5)

R

Thus, Newton's laws, when combined, tell us that the trajectory of the ball is an ellipse
which fits the Galilean parabola rather closely so long as hlR is much smaller than 1.
The agreement between Newtonian physics and Galilean physics is rather close for
tennis balls, but TV programmes on the national hook-up would not have been possible
but for the elliptical orbits of satellites. On the other hand, Newton's laws, combined,
also tell us that the trajectory of the centre of mass may, in some instances, be a
hyperbola (or even a parabola).
Moreover, all these statements are falsifiable. Man could reach the moon because
of the difference between the Newtonian and Galilean theories. According to Galileo's
laws, the energy required would have been some 1014 erg/g, corresponding to an initial
velocity of 100 km/s, which is some three times higher than the peak velocity attained
by modern-day rockets, and an order of magnitude larger than the escape velocity
actually required.
Having eliminated force, by combining both of Newton's laws, it is now also
possible to construct a measure of equal times, using Newton's laws of motion. For
example, a simple pendulum measures out approximately equal intervals of time. It
should be emphasized that this is a practical scheme, not a theoretical definition. Thus,
suppose that, as in Dirac's cosmology,S the Newtonian constant of gravitation varies
with the epoch, over geological periods of time. This would mean that, in some sense,
a second measured by the pendulum in the Pre-Cambrian era would differ from a
second measured currently. But if a pendulum were used to define equal intervals of
time, the two seconds in question would be equal by definition.
 NEWfON'S TIME 43

5 Laplace's demon

5. 1 Time in classical mechanics

Having extracted some physics from Newton's laws, let us turn to the questions posed
by philosophers that were listed at the end of the previous chapter. To obtain some
physics from Newton's laws we eliminated the unknown term 'force'. Not so surpris-
ingly, the unknown 'time' has also been eliminated, and the physical content of
Newton's laws (combined) is independent of both unknowns.
The conclusion is that it is not necessary to define the notion of time in Newtonian
physics, and that Newton's own efforts in this direction are irrelevant. Since time is
redundant in Newtonian physics, so also are controversies concerning the dynamic
(A-series) versus the static (B-series) picture of time. This, now, becomes a matter of
taste. One may suppose, if one likes, that the future comes into existence and that the
present passes out of existence every moment. The future is neither physically nor
logically different from the past.
The same conclusion follows from an analysis of the Newtonian notion of a 'body'.
For Newton himself, these bodies may have been either 'heavenly bodies' or 'bodyes
here amongst us', reflecting the fact that his analysis applied equally to planetary motion
and to projectile motion. But the questions about the ultimate structure of matter,
arising from an analysis of the notion of 'body' are perhaps spurious. Newtonian physics
deals not so much with 'bodies' as with trajectories of bodies. The mathematics used is
such that these trajectories are curves, so that bodies are mass points by default. (This
point of view, where only trajectories are studied, is reinforced in general relativity.)
Time, thus, becomes a parameter along the curve, which is not truly relevant to a
description of the geometry of the curve.
Controversies concerning the continuum or discrete nature of time are also largely
irrelevant. It is true that Newtonian physics assumes a continuum picture of time so as
to permit the application of the calculus. But Newtonian mechanics is known to fail at
the microscopic level so that this assumption is hardly helpful in deciding the nature of
physical time.

5. 2 The structure of time in classical mechanics

The only question that does not become redundant is that concerning determinism -
even (traditional) quantum mechanics supposes that it is classical mechanics which
applies at the macrophysicallevel of the measuring apparatus. This question is import-
ant because the justification for accepting Newtonian physics and rejecting Galilean
physics is that Newtonian mechanics provides a better description of reality, as can
easily be determined by performing experiments. But this justification presupposes that
we are free to perform experiments and free to choose between competing scientific
theories. It presupposes the mundane picture of time which branches towards the future
and is linear towards the past.
On the other hand, Newtonian physics seems fully deterministic. In classical
mechanics the evolution of a system, no matter how complex, is governed by an ordinary
differential equation y' (t) = f (t, yet) ), where yet) denotes the phase vector of the
44 CHAPTER II

system. The positions and momenta of all particles in the universe at one time determine
the initial value yeO) and, hence, the past and future history of the universe down to the
last detail. Not only are the past and future decided, they are decided by the present -
time is superlinear.
How did this picture of time evolve? Newton's teacher, Isaac Barrow, contemptu-
ously called 'quacks' those mathematicians who were content to plod along without a
clear picture of time. Presumably inspired by geometry, he forcefully asserted that 'time
has length alone, is similar in all its parts and can be looked upon as ... either a straight
or a circular line. ' [Italics mine] The rather vast difference between a straight line and
a circle is now nearly lost since time became a straight line after Newton.
The interesting thing is that for Barrow there were only two choices for the structure
of time: either a straight or a circular line. There can be no logical reason why Newton
chose the straight line, because in his formulation of 'instantaneous' laws of mechanics
the global structure of time does not matter. Perhaps he was influenced in this choice
by his religious convictions. (The Fifth Ecumenical Council had declared cyclic time to
be anathema.)
The logical linearity of time probably did not bother Newton. He was unable to
prove the stability of planetary orbits, and thought that evolution according to his laws
was supplemented by divine intervention, from time to time, to keep the planets in their
orbits. Laplace proved the stability of planetary orbits. Thus, even though time was
largely irrelevant to the Newtonian study of trajectories, Newtonian mechanics played
a key role in the evolution of the picture of time from cyclic to linear to superlinear.

5. 3 The argument from complexity

One way to avoid superlinear time, and evade the conclusion of determinism, is to argue
as follows. Newtonian physics may be deterministic in principle, but there are a very
large number of particles in the universe, and it seems a rather tedious task to observe
the initial positions and momenta of all the particles in the universe and then solve a
very large number of differential equations for all the particles in the universe. The
statistical-mechanics argument is that even a gas in a box contains such a very large
number of particles that it is pointless, if not impossible, to obtain through observations
the initial conditions for all particles, and compute the solutions of the equations of
motion for all the particles in the box. Therefore, Newtonian mechanics does not quite
determine the future states of the universe, since it is a herculean task, in practice, to
obtain these future states through Newtonian mechanics. This may be called the
argument from complexity, or an argument from ignorance, because it asserts that
though the future may be determined, our ignorance of complex systems precludes an
actual determination from being carried out in practice.

5. 4 Laplace's demon

On the other hand, one might argue, as Laplace did, that practical limitations will sooner
or later be overcome. Laplace philosophizing on probability stated:
We may regard the present state of the universe as the effect of the past and the cause of its
future. An intelligence which at a given moment of time knew all the forces that animated
nature, and the respective positions of the beings that compose it, and further possessing
 NEWTON'S TIME 45

the scope to analyse these data could condense into a single formula the movement of the
greatest bodies of the universe and that of the least atom; for such an intelligence nothing
could be uncertain, and past and future alike would be before its eyes.
As Laplace further remarked to Napoleon, he had no need for God in his system so that
the Laplacean 'intelligence' goes under the name of Laplace's demon. Laplace's demon
is not an omniscient God, merely a super-scientist, super-observer, and supercomputer
rolled into one. Popper6 describes Laplace's demon as follows.
He is not supposed to be able to do anything which human scientists could not do or at least
approximately do: he merely is supposed to be able to carry out his tasks with superhuman
perfection.

Box 3. Pierre Simon de Laplace (17491827)

French astronomer and mathematician, famous for his books on celestial mechanics and
the theory of probability. Born in Nonnandy, he came to Paris and attracted the attention
of D'Alembert, who found him employment in the Ecole Militaire. Here he taught
mathematics to trainee artillery officers, among whom was Napoleon. After the French
revolution of 1789, when the revolutionary government established the Ecole Poly-
technique, Laplace was one of the founding professors. He helped devise the meter,
standard weights and measures, and developed the theory of statistical sampling which
provides an economical and efficient substitute for a census.
Newton could not establish the stability of the solar system, which he thought
needed divine correction from time to time. Laplace showed that all known secular
variations, such as the changing speeds of Saturn and Jupiter, were cyclic and that
the system was entirely stable, and needed no divine maintenance. This was the basis
of his remark to Napoleon that he had no need for God in his system.
Laplace's work in celestial mechanics was extremely tidy, consisting of the final
polishing of the Newtonian theory. In his huge treatise on celestial mechanics (published
in five volumes from 1799 to 1825) he did not at all acknowledge his predecessors. It has
been conjectured that this was the basis ofNapoleon's remarkto Laplace that he had written
a huge book without once acknowledging God as the author of the universe.
As a founder of the theory of probability, he felt compelled to clarify that
probabilities were not 'real', and that they were a measure of ignorance of determin-
istic pbysicallaws which truly represented reality. Regarding his demon, he re-
marked in his Philosophical Essays on Probability that 'The human mind offers, in
the perfection which it has been able to give to astronomy, a feeble idea of this
intelligence.' The superhuman intelligence ('Laplace's demon') was to have no need
for the theory of probability. Laplace would have regarded as ridiculous the idea that
God, if there was one, pJayed dice. He said, 'The curve described by a simple
molecule of air or vapor is regulated in a manner just as certain as the planetary
orbits; the only difference between them is that which comes from our ignorance.'
Laplace also completed the theory of tides and solved another famous Newtonian
problem - the deduction of the speed of sound from first principles. It is a tribute to his
versatility and intellectual suppleness that the curve described by his career only progressed
upward through several revolutionary changes of governments. On the other hand, he was
generous, and assisted many younger men like Guy Lussac, Poisson. and Cauchy.
 46 CUAPTERII

Thus, Laplace would have been ready to admit that human scientists cannot ascertain the
initial conditions of all the physical bodies in the universe; but he would have pointed out
that they can measure all the initial conditions of a solar system if the number of planets is
small. He would also have been ready to admit that scientists cannot obtain absolutely exact
initial conditions; but he would have pointed out that they can improve the degree of
precision in measuring them, and that there is no absolute limit to these improvements.
.. . Laplace would have pointed out that, though we have not solved the general
many-body problem... we might one day [md its solution, which certainly would make
it legitimate to invest the demon with this knowledge; and he might have added that,
even though the general problem might be strictly insoluble, we can, in every par-
ticular case. .. replace the exact solution by an approximation which is precise to any
degree we may choose to stipulate. It is in this sense that Laplace's demon is only an
idealized human scientist.

5. 5 Exorcism of Laplace's demon

Popper gives an interesting argument to exorcise Laplace's demon. In the first place,
Popper argues that something similar to Heisenberg's uncertainty principle might
operate in classical mechanics, so that it may be impossible, in principle, to know initial
conditions exactly. Thus, any measurement, such as the measurement of a length,
involves some error. The relative error in measurement decreases as the length to be
measured increases. If, however, one is measuring something like acceleration, a
problem arises. One gets a more accurate measurement if one observes the body for a
longer period, but the quantity measured is the average acceleration. The larger the
time interval over which this average is taken, the poorer the approximation this
provides for the instantaneous acceleration, which is the quantity that is actually
required by Newtonian theory. Thus, the error increases both when the time interval is
chosen very small and when it is chosen very large, and so there must be an optimum
time interval which minimizes the error in measurement of instantaneous acceleration.
In any case, there must always be some non-zero error, in principle, in the measurement
of acceleration.
In the second place, Popper points to chaotic motions. The first example of chaotic
motion (or an ill-posed initial value problem) was given by Hadamard in the previous
century. The example corresponds to geodesic flow on a manifold of constant negative
curvature. The geodesics (phase-space trajectories) diverge exponentially, so that it is
impossible to predict future states of the system unless the initial conditions are known
exactly. Today, many examples of chaotic motions are known, such as the flow of water
from a tap, the logistic map, or the circle map. Chaotic motions are characterized by
instability, a positive Liapunov exponent, or exponentially diverging trajectories in
phase space. It is believed that complex systems tend to be chaotic.7
Putting together the two parts of Popper's argument, one concludes that Laplace's
demon is unable to predict, in principle, the future states of the universe. Thus, the
human brain is a complex system and the creation of Mozart's 40th symphony or the
creation of the special theory of relativity are both complex forms of activity and must
be modeled in Newtonian physics, if at all, by chaotic dynamical systems. Using the first
part of Popper's argument it is impossible, in principle, for Laplace's demon to know
exactly the initial states of such dynamical systems. On the other hand, if the initial state
 NEWfON'S TIME 47

is not known exactly, Laplace's demon is unable to predict the outcome of Mozart's or
Einstein's thought processes.
To see the relevance of this discussion to the nature and structure of time, let us
analyse the temporal assumptions underlying Popper's arguments. To give an analogy,
Popper exorcises Laplace's demon by likening the world to a mystery novel. The
murderer may be determined, but after reading the first few pages it may be impossible,
in principle, to deduce the identity of the murderer (even if one is as good at deduction
as Sherlock Holmes). But this tends to suppose that one follows the rules and does.not
peek at the last page. Worse still, just when the novel is getting interesting, one may
recall having read it before.
Thus, Popper's argument tends to assume that direct, non-inferential knowledge
of the future is impossible. In particular, the argument assumes that physical interac-
tions may propagate from past to future but not the other way around. (In Chapter VB,
we explore some of the consequences of denying this assumption.) Popper's argument
also tends to assume that one does not have any memory of the future. This assumption
explicitly excludes the possibility of 'circular time', or a cosmology of the G6del type,
where the remote future might blend into the remote past.
Both of Popper's assumptions seem perfectly justified on the basis of the everyday
experience of time. But it must be remembered that neither assumption is logically
necessary and that both assumptions pertain to the nature of the physical world;
therefore, their validity must be decided on the basis of physical theories. Moreover,
Newtonian physics, by itself, is insufficient to decide the validity of these assumptions.
It is possible to give other arguments for indeterminism. For instance, Newtonian
mechanics gives no clear specification of the notion of a 'body', and we do know that
Newtonian mechanics fails at the microscopic level. Again, not every differential
equation possesses a unique global solution.8 All such arguments suffer from one
deficiency - they are time-symmetric: Laplace's demon is equally unable to retrodict
the past. This suggests that the conclusion, too, must be time-symmetric. Thus, if
inability to predict the future is an argument for indeterminism of the future, then
inability to predict the past must be an argument for indeterminism of the past. But this
is contrary to our belief in a past that is fixed and unchanging. The only scientific attitude
towards this belief is not to believe blindly, but to decide its validity on the basis of
appropriate physical theories and experiments. We shall attempt to do this in the
subsequent chapters.

6 Conclusions

Time as a physical entity is largely irrelevant to Newtonian physics, since equal intervals
of time cannot be defined with a democracy of clocks. Newton's laws of motion and
Newton's law of gravitation are individually meaningless; but the combined laws are
falsifiable, for time may be eliminated in discussing their consequences. The combined
laws may also be used to obtain a pragmatic scheme for an approximate measure of the
notion of equal intervals of time. In Newtonian physics it is not essential to define the
concept of time any further. Since time becomes largely redundant, so also do contro-
versies concerning the static or dynamic nature of time. The future is neither logically
48 CHAYIERII

nor physically different from the past. Newtonian physics isprimafacie deterministic,
contradicting the human freedom required to demonstrate its empirical validity. The
future may still be regarded as indeterminate, in the sense of being unpredictable,
modulo the validity of certain commonsense assumptions concerning time. The diffi-
culty with this method of exorcising Laplace's demon is that the conclusions are actually
time-symmetric so that the past, too, must be regarded as indeterminate.
llIA THE MICHELSONMORLEY EXPERIMENT
ABSTRACT. In the previous two chapters, we covered problems related to the physical notion of time in
(i) philosophy and (ii) Newtonian mechanics. Our contention is that a theoretical analysis of the concept
of time in Newtonian theory is and was of greater relevance than the Michelson-Morley experiment in
understanding the special theory of relativity. Here we argue in support of the latter part of the contention.
The Michelson-Morley experiment, though relevant, was not intended to test the constancy of the speed
of light. Rather, the experiment attempted to distinguish between the theories of Fresnel and Stokes, both
of which used the aether hypothesis. Michelson concluded that the Stokes theory was valid, while Miller
was awarded a prize for his claim to have detected an aether drift, which was widely disbelieved.

1 Introduction

One remark, however, appears indispensable. The notion of 'material point' is fundamental
for mechanics. If now we seek to develop the mechanics of a bodily object which itself can not
be treated as a material point - and strictly speaking every object 'perceptible to our senses' is
of this category - then the question arises: How shall we imagine the object to be built up out
of material points, and what forces must we assume as existing between them? The formulation
of this question is indispensable if mechanics is to pretend to describe the object completely.
It is in line with the natural tendency of mechanics to assume these material points, and the
laws offorces acting between them as invariable, since temporal changes would lie outside the
scope of mechanical explanation. From this we can see that classical mechanics must lead us
to an atomistic construction of matter. We now realize, with special clarity, how much in error
are those theorists who believe that theory comes inductively from experience.
Albert Einstein
Physics and ReaIiJy
The Journal of the Franklin Inst. 221, 1936

There was a young man named Fisk

Whose fencing was exceedingly brisk.
So fast was his action
The Fitzgerald contraction
Reduced his rapier to a disk.
Anon.

M popular
OST textbooks in physics introduce the special theory of relativity, not through
limericks, but by stating that the historic Michelson-Morley experiment
demonstrated the constancy of the speed of light.
Physics deals with empirical reality, so it seems desirable that the postulates of
physics should have direct empirical support. Direct support may not always be forth-
coming, but the special theory of relativity seems to be a classical case in the point,
because of the Michelson-Morley experiment.
It seems part of human nature that if one desires something strongly one pretends
that it is true. If the pretence is carried out long enough, it becomes difficult to
distinguish between pretence and reality. The textbook vision of a simple and clear
relationship between theory and experiment is enough to bring a lump in the throat of
50 CHAPTER IlIA

a modern-day physicist struggling hard to find a way through a maze of inconclusive

experiments, hypotheses and counter-hypotheses. But the student, who takes this
charming piece of folklore seriously enough to pursue matters further, is in for a rude
shock.
What are we to make of the following set, which seems somewhat puzzling, and not
quite consistent?
1. The conclusion of the previous chapter: that because of the democracy of clocks in
Newtonian physics, no absolute significance can be attached to the measurement of
(equal intervals of) time, therefore of speed, and, in particular, the speed of light.
2. The fact that Einstein's original paperl of 1905 made only a passing reference to the
'unsuccessful attempts to discover any motion of the earth relatively to the "light
medium" ... ', and that Einstein decided to 'introduce another postulate ... , namely, that
light is always propagated in empty space with a definite velocity c which is indepen-
dent of the state of motion of the emitting body.' [Emphasis original] Einstein later
repeatedly stated that he was largely unaware of the Michelson-Morley experiment,
that he took the results of the experiment for granted, and that it did not much affect
his thinking on relativity. (See Appendix.)
3. The text-book claim that the Michelson-Morley experiment should correctly be
regarded as the basis of the special theory of relativity, since it establishes the
constancy of the speedt of light, and since it inspired the formulas for the Fitzgerald-
Lorentz contraction.
Perhaps it is worthwhile taking a quick look at the history of the experiment.

2 The background of aether theories

2. 1 The blue sky or why aether was introduced

[The universeI is like a rare clock, such as may be that at Strasbourg, where all things are so
skillfully contrived, that the engine being once set a-moving, all things ... per/onn theirfunctions
upon particular occasions, by virtue of the general and primitive contrivance of the whole
engine.
R. Boyle (16271691)

One of the early problems of natural philosophy was to explain how bodies at a distance
from each other could influence each other. By Descartes' time the existence of such
influences was perfectly clear from, say, the behaviour of magnets or the correlation
between tides and the position of the moon. Descartes felt that 'animistic' and 'occult'
explanations of these influences obfuscated the issue. He sought a mechanistic (clock-
work) model of the universe and he followed the Greek atomists, Aristotle, and Thomas
Aquinas in believing that only contiguous bodies can act on each other, and that the
only perfectly intelligible forces were those due to pressure and impact.

t I prefer to speak of constancy of 'speed' rather than 'velocity' to emphasize the novel element.
 THE MICHELSON-MORLEY EXPERIMENT 51

But this meant that the space between the earth and the moon, and all of space, in
fact, should be filled with a medium capable of transmitting forces. The particles of this
medium were supposed to fill even the subtle interstices between particles of ordinary
material things like air, water or the earth. This medium came to be called the aether,
a word which denoted the blue sky or the rarefied upper atmosphere.
Thus, in Descartes' view, space was a plenum filled with aether particles which he
supposed were perpetually moving. But since there was no empty space for an aether
particle to move into, it could only move by occupying the space vacated by an adjacent
particle of aether. Thus, the motion of a single aether particle involved the motion of a
whole closed chain of particles, and came to be called a vortex.
All these marvelous inferences followed logically in the I-think-therefore-I-am
fashion, and carried conviction for a few centuries. Descartes' objective was to banish
the occult and introduce clarity - it seems hardly fair to blame him for not anticipating
that a few centuries later this very hypothesis would come to be regarded with great
suspicion as an exotic source of obfuscation.

2. 2 The Dragon's head

The story now moves on to the star" in the head of the Dragon (Gamma Draconis), as
observed by the astronomer Bradley in 1725. The star had already been observed in
1669 by Robert Hooke, who claimed to have discovered a parallax of about 30 seconds
of the arc, in confirmation of the Copernican theory. This was in line with other
observations of large parallax in the 17th century. (We now know that even the closest
star Alpha Centauri has a parallax of only 0.76".)
Bradley also observed an apparent displacement of about 20", but was sharp enough
to notice that the maximum apparent 'sideways' displacement of the star occurred with
motion along (and not transverse to) the line of sight, when the earth was 'furthest' and
'closest' to the star, so that the displacement could not have been due to parallax.
Bradley referred to the phenomenon of large apparent stellar motions as stellar
'aberration'. Since stellar parallax and stellar aberration are both measured by observ-
ing the change in the apparent position of the star, this can cause some confusion. It is
useful to bear in mind that apart from the above 'phase relationship' to the line of sight,
the difference between parallax and aberration is that parallax, unlike aberration,
depends upon the distance of the star.

2. 3 Stellar aberration and the finiteness of the speed of light

By Bradley's time it had already been guessed that the speed of light is finite, and this
speed had been measured by observing the eclipses of the satellites of Jupiter. Assuming
the finiteness of the speed of light c, simple vector addition with the orbital velocity of
the earth v was sufficient to account for aberration: the apparent position of a distant
star changes in the same fashion as raindrops, falling vertically, slant across the window
of a moving train. Using this explanation and his measurement of the constant of
aberration vic, (v being the orbital speed of the earth) Bradley used the known solar
parallax to deduce the time for light to travel from the sun to the earth to be 8 minutes
and 12 seconds.
52 CHAPTER IlIA

2. 4 The grove of trees

There is no displacement caused in the aether [ = ~a, lit. sky] by things moving in it or striking
against it, as there is in water by a piece of woodpassing through it... . Secondly, the aetheroffers
no obstruction to things moving in it... - and why is this so? - It is because the aether is not
tangible.
VAtsyAyana
Nydya satra BhdfYa IV.2.22

The controversy really began when Young attempted to explain the phenomenon of
stellar aberration on the wave theory of light. 'I am disposed to believe,' he stated, upon
considering the phenomenon of the aberration of the stars, 'that the luminiferous aether
pervades the substance of all material bodies with little or no resistance, as freely
perhaps as the wind passes through a grove of trees.' If the aether is unaffected by the
motion of the earth, then so are the light waves. So, if a telescope is pointed at the true
position of the star, 'the image of the star will appear to be displaced from the central
spider-line at the focus by a distance equal to that which the earth describes while the
light is moving through the telescope,'2 i.e., the angular displacement will be vic, as
before.
Young's explanation of stellar aberration raised many questions. To avoid refrac-
tion of light from a star, should the normal to a plate of glass point to the true position
of a star or to its apparent position? When observing a star towards which the earth is
moving, due to the change in the relative velocity of light, should the focal length of the
telescope be increased? Arago concluded through experiments that in all cases of
reflection and refraction light behaves exactly as if the star were in its apparent position
and the earth at rest, i.e., the apparent refraction in a moving prism is equal to the
absolute refraction in a fixed prism.

2. 5 The aether drag

Though the accuracy of Arago's experiments of 1810 (published only in 1839) was doubtful,
Fresnel, in 1818, published an account of refraction in moving bodies, using the wave theory,
which explained the result. Fresnel also used another suggestion of Young, namely, that the
amount of refraction in a transparent body depends upon the concentration of aetherwithin
it. This he made precise by supposing that the refractive index of a body was proportional
to the square of its aethereal density. Thus, if c and Cl denote respectively the speed of light
in vacuo and in the body, n = clcl is the refractive index, and a, al are the corresponding
aethereal densities, Fresnel supposed that

(1)

In line with Young's 'wind through a grove of trees' suggestion, Fresnel supposed that
a density of aether a remains at rest in a moving body. On the other hand, the moving
body itself was characterized by supposing that it carried along with it that part of the
aether which constituted the excess of its density -over the density of aether in vacuo.
That is, the body carried along an excess amount of aether with the density
 THE MICHELSON-MORLEY EXPERIMENT 53

a1 - a = (n 2 - 1) a, (2)

while an amount of aether with density a remained at rest. (In this picture the aether
around the body does not participate in the motion of the body. But we could, if we
liked, think of the aether as entering the moving body from the front, getting im-
mediately compressed, and issuing from behind and getting immediately rarefied.) For
a body moving with velocity v, the centre of gravity of the aether therefore moves with
the velocity

[en 2 -2 1)]v. (3)

n

The quantity in square brackets became famous as the Fresnel drag coefficient.
Relative to the body, the speed of light is C1, so that the absolute velocity of light in a
moving body is

n2 - 1 (4)
C1 + [--2-]v.
n

This relationship was confirmed in 1859 by Fizeau, who split a beam of light, passed
it through two columns of water, one of which was moving, and showed that the
displacement of the interference fringes depended upon the speed with which the water
moved. The relationship was also confirmed by Michelson and Morley in 1886.
Poincare, who never tired of taking a crack at the aether, remarked that in Fizeau's
experiment 'one believes one can touch the aether with one's fingers.'

2. 6 A delicate point

Applying the above formula (Fresnel's law) to the problem of refraction in moving
bodies, Fresnel concluded that if observations were made with a telescope filled with
water, the observation of stellar aberration would be unaffected. This conclusion
differed from earlier thinking that the thickness of glass or other transparent material
in the telescope would affect the observed amount of stellar aberration, since the light
would move more slowly through the telescope. Fresnel's conclusion of a null result,
because the water dragged the light sideways, was confirmed experimentally by
Respighi in 1861, Hoek in 1868, and Airy in 1871, all working with telescopes filled with
water! The first two found no effect on reflection and refraction if the orientation of
the apparatus was altered relative to the direction of the earth's motion. Airy remarked
that this 'is a question which involves, theoretically, one of the most delicate points in
the Undulatory Theory of Light.'
54 CHAPTER IlIA

2. 7 The aether dragged

While Fresnel's theory was experimentally successful, there seemed no way it could
account for refraction in dispersive media, with different colours being refracted by
different amounts. (Did the aether move with several different velocities simulta-
neously?) The other difficulty with Fresnel's theory was the idea that the aether passed
through the body like 'wind through a grove of trees'. If the aether surrounding the body
did not in the least participate in the motion of the body, then the aether might just as
well not exist!
Stokes attempted to answer the last objection by attempting to explain stellar
aberration in another way. He supposed that the aether in the vicinity of the earth is
dragged along due to the motion of the earth, so that it is at rest relative to the earth.
This motion of the aether must be superposed on the aether vibrations while transmit-
ting the light. If the motion of the aether is irrotational, then a wavefront that was
originally plane remains plane, and light is still propagated in straight lines. But, because
of the superposition, the orientation of the wavefronts will change, and the normal to
the wavefront will deviate from the original direction of propagation by the small angle
vic. Since the direction in which the star is seen is obtained by extending this normal
backwards, the apparent position of the star will differ, as before, from its true position.
It was later pointed out that the type of motion visualized by Stokes could not exist
for the following reason. The irrotational flow of an incompressible fluid is known to
be a potential flow, i.e., the velocity can be derived from a potential. The potential
equation (Laplace's equation) admits a unique solution if the normal derivative is
specified at the boundary (Neuman problem). Therefore, if the normal component of
the velocity was the same as that of the earth, the tangential component would have to
be different.
The luminaries who tackled this problem included Planck and Lorentz. Planck
proposed that if the aether were compressible according to Boyle's law, and also subject
to gravitation, it would be compressed like the atmosphere around the earth, and
Stokes' conditions would be satisfied provided that the speed of light was independent
of the density of the aether. The last assumption, of course, ran counter to Fresnel's
hypothesis about the refractive index.

3 Electrodynamics: theory and experiment

3.1 Maxwell

The next chapter of the story of the Michelson-Morley experiment concerns J.e.
Maxwell, who had already put the final touches to the grand synthesis of the laws of
electricity and magnetism during the above paragraphs. Maxwell's equations led to the
computation of the speed of light and completed the victory of the wave theory of light.
But Maxwell wanted to do for the electrodynamics of moving bodies what Fresnel had
done for the wave theory of moving bodies.
 THE MICHELSON-MORLEY EXPERIMENT 55

In a way it is strange that Maxwell, who applied his electromagnetic theory so

successfully to the propagation of light in isotropic media, in crystals, in metals, and in
plasmas, did not construct a theory of reflection and refraction. A plausible explanation
is that Maxwell believed too strongly in the analogy between the electric displacement
and the displacement of an elastic solid that the aether was sometimes supposed to be.
Therefore, unlike Green or Heaviside, he could not bring himself to believe in the jump
conditions at an interface which required a discontinuity in the derivatives.
Maxwell's thoughts on the subject were published in a posthumous letter3 in which
he thought of distinguishing between the theories of Fresnel and Stokes. We recall that
according to Fresnel4 the aether remained fixed in space so that it could be regarded
as streaming through moving bodies. Stokess on the other hand, proposed that the
aether is dragged along in the vicinity of moving bodies.
Inspired by this posthumous letter from Maxwell, Albert Michelson,6 who believed
in the aether theory, conceived the idea of using the interferometer to determine the
velocity of the earth relative to the aether in order to decide between the theories of
Fresnel and Stokes. Maxwell had suggested, like Roemer, the use of the satellites of
Jupiter, since terrestrial experiments could only show up second order(v2/c2) effects,
which Maxwell felt were beyond the reach of the current instruments. Michelson, on
the contrary, felt that observations on Jupiter made six years apart would be of doubtful
value, and that he could refine the interferometer to detect second order effects.
Michelson felt that this should be done because he thought that major advances in
physics would be confined to the sixth decimal place. To reiterate, the objective of
Michelson's experiment was not to test the aether theory, but to discriminate between the
two versions of the theory due to Fresnel and Stokes.

3. 2 The experiment proper

The idea behind the design of the experiment was simple enough. The local speed of
sound in a fluid is determined by the local properties of the fluid (pressure, density).
Consequently, sound travels faster in the direction of the wind, and slower against it.
So, if a wind is blowing from the echo-point to the hill, the echo will take longer to return
(since the sound travels at a lower speed for longer than it travels at a higher speed, so
the average speed is reduced). On the other hand, sound velocity transverse to the wind
velocity would remain unaffected.
Replacing sound by light, and wind by aether wind, Michelson concluded that the
aether drag, if it existed, would similarly affect the average speed oflight in the direction
of the earth's motion, leaving it unaffected in the transverse direction. His apparatus
consisted of a pair of arms at right angles to each other. Light from a single source (to
ensure coherence) was split by a half-silvered mirror, and traveled down each arm. After
reflection in mirrors set at the end of the arms, the light signals would recombine to
give rise to interference fringes, observable through a telescope. If one arm of the
apparatus was pointed in the direction of the earth's motion around the sun, the reduced
speed of light in that pirection would mean an increased path length which would show
up in the form of a shift in the interference fringes.
Taking the speed of the earth to be about 1 part in 1()4 of the speed of light, for an
arm of length 1 meter, the change in effective path length would be about 2 parts in
56 CHAP'IER IlIA

lOS or 4'100 of a wavelength for yellow light, resulting in the fringes shifting by about ti
the thickness of the fringe.
Though simple, in principle, the experiment itself must have been extraordinarily difficult
to perfonn. Michelson calculated that a difference of 0.01 0 K, between the temperatures of
the two arms would produce an effect three times what he was looking for. Vibrations due to
people moving about would make the fringes disappear altogether, so that it was impossible
to perfonn the experiment in Berlin7 even in the middle of the night The aether was actually
supposed to be at rest with respect to the galaxy, and there were difficulties in judging the exact
direction of motion of the aether. It was later pointed out that neglect of this last factor, in his
calculations, led Michelson to overestimate the expected result. After each reading, the
apparatus was rotated through 900, and the final null result was based on a rough statistical
search for periodicity in fringe shifts from 160 readings, of which 20 were discarded due to
apparatus effects.
The experiment was repeated several times by Michelson, Morley, and Miller.s Michel-
son concluded that the experiment vindicated the Stokes theory.

3. 3 Miller's observations

It would not be exactly correct to state that all these experiments had a null result, since Miller
did claim in 1925 and till as late as 1933 that he had observed an aether drift, and arrived at
the value of 208 km/s for the motion of the solar system relative to the aether. For this Miller
received a prize of a thousand US dollars from the American Association for the Advance-
ment of Science. It is a different matter that few people believed his later results,9 and those
who did kept offering alternative explanations.1o
For example, Einstein wrote to Besso 'I have not for a moment taken [Miller's results]
seriously,' and went on to remark that the positive effect found by Miller could be explained by
tiny temperature differences in the experimental equipment It was on hearing rumours of
Miller's positive results that Einstein had remarked 'Subtle is the Lord, but malicious He is not'
The minor role played by experiments in inspiring the theory of relativity is also apparent
from the earlier reactions to the experimental results of Kaufrnann,l1 on the electromagnetic
mass, refined and repeated in 1906. Kaufmann concluded, 'The measurements are incompati-
ble with the Lorentz-Einstein postulate. The Abraham equation and the Bucherer equations
represent the observations equally well.'
The reaction of the physics community is described by Abraham Pais12 (p 159).
Planck ... took a wait and see attitude....Lorentz vacillated: The experiments 'are decidedly
unfavorable to the ideas of a contraction, such as I attempted to work out. Yet though it
seems very likely that we shall have to relinquish it altogether, it is, I think, worth while
looking into it more closely .. .' ... Einstein was unmoved 'Herr Kaufmann has determined
the relation between [electric and magnetic deflection] of {3-rays with admirable care....
Using an independent method, Herr Planck obtained results which fully agree with [the
computations of] Kaufmann.... It is further to be noted that the theories of Abraham and
Bucherer yield curves which fit the observed curve considerably better than the curve
obtained from relativity theory. However, in my opinion, these theories should be ascribed
a rather small probability because their basic postulates concerning the mass of the
moving electron are not made plausible by theoretical systems which encompass wider
complexes of phenomena. '
Probably, Synge13 deserves the last word. Suppose we believe Miller's result, and
disbelieve the explanation in terms of thermal fluctuations. Would we then abandon
 THE MICHELSON-MORLEY EXPERIMENf 57

the theory of relativity? Not at all; positive results in the Michelson-Morley type of
experiment do not refute the theory of relativity.
When the Michelson-Morley experiment was performed, no doubts were entertained about
the validity of the concept of a rigid body, and a slab of stone was regarded as rigid in the
Newtonian sense ... In fact the experiment was made to enlarge our knowledge of the
behaviour of light, the behaviour of bodies called 'rigid' not being questioned. But actually
the experiment involved both things.... it appears impossible to make any satisfactory
predictions as to the result of an ideal Michelson-Morley experiment performed with mirrors
attached to a rigid three-dimensional base (the slab of stone) for the reason that such a base
has not been satisfactorily defined in the terms of mathematical physics. If the small effect
observed by Miller were ... validated in a repetition of the experiment, we would not interpret
it as evidence for or against the theory of relativity, but rather as information about the way
in which [the slab of stone] changed under the circumstances of the experiment.

3. 4 Aether, relativity and metaphysics

So much for the history of the Michelson-Morley experiment. Those who would jump
to declare aether as a metaphysical vehicle for speculation may stop to consider the
statement of William Maggie 14 in his 1911 presidential address to the American Physical
Society, 'They are asking us to abandon [the aether hypothesis which] has furnished a
sound basis for the interpretation of phenomena and for constructive work in order to
preserve the universality of a metaphysical postulate'. The 'metaphysical postulate' was
the principle of relativity. On the other hand, it is interesting to learn that in 1916, one
of Einstein's friends, Friedrich Adler, assassinated the Austrian prime minister for
violating the country's freedom. He invoked the principle of relativity in his defence.

3. 5 The moral of the story

Every story, especially such a long one, should have a moral! The moral of this story is
that those who do not learn from history are condemned to repeat it. The idyllic situation
portrayed in textbooks is rather far from the reality of the clash of theories and
inconclusive experiments.
The story shows that it is not enough to have a theory of some sort, even a theory
like Fresnel's which made dramatic predictions, supported by experiments. The
conceptual muddles in a theory cannot be wished away with the optimism that the future
would somehow resolve them, because the future often resolves these muddles by
rejecting the theory.
The aethereal gymnastics of the 19th century physicists amuse us. Will the 21st
century physicist be similarly amused by our temporal gymnastics?

4 Conclusions

The objective of the Michelson-Morley experiment was to decide between the theories
of Fresnel and Stokes, and Michelson concluded that Stokes was correct, while Miller
found the existence of an aether drag.
58 CHAPTER IlIA

Appendix: Einstein on the Michelson-Morley experiment

At different points in his life Einstein responded differently to the question "did he
know of Michelson's work prior to 1905?" As Abraham Pais (ref. 12,p 115) remarks,
'Michelson is mentioned neither in the first nor in any of Einstein's later research
papers on relativity. One also looks in vain for his name in Einstein's autobiographi-
cal sketch' (emphasis mine). Einstein responded to Shankland15 in three different
ways.
(i) Shankland recalls the discussion on 4 Feb. 1950 'When I asked him how he had
learned of the Michelson-Morley experiment, he told me that he had become aware
of it through the writings of H. A. Lorentz, but only after 1905 had it come to his
attention!' (Italics Shankland's). Einstein maintained that the observations on stellar
aberration and Fizeau's experiment were enough.
(ii)In another discussion on 24 October 1952 'I asked Professor Einstein when he
had first heard of Michelson and his experiment. He replied, "This is not so easy, I
am not sure when I first heard of the Michelson experiment. I was not conscious that
it had influenced me directly during the seven years that relativity had been my life.
I guess I just took it for granted that it was true." , Einstein went on to say that he
had thought a great deal about Michelson's results during 1905-1909, and then
realized that he had also been conscious of Michelson's results before 1905 partly
through his reading of the papers of Lorentz and more because he had simply
assumed this result of Michelson to be true.
(iii) In a letter to Shankland in Dec 1952 Einstein wrote 'The influence of the crucial
Michelson Morley experiment upon my own efforts has been rather indirect. I
learned of it through H. A. Lorentz's decisive investigation of the electrodynamics
of moving bodies (1895) with which I was acquainted before developing the special
theory of relativity. Lorentz's basic assumption of an ether at rest seemed to me not
convincing in itself and also for the reason that it was leading to an interpretation of
the Michelson-Morley experiment which seemed to me artificial'.
(iv) In an address given at Kyoto on 14 Dec 1922, Einstein stated (in German,
translated into Japanese and retranslated into English) 'As a student I got acquainted
with the unaccountable result of the Michelson experiment ... this is the first route
which led me to what is now called the special principles [sic] of relativity.... I had
just a chance to read Lorentz's 1895 monograph ... in this connection I took into
consideration Fizeau's experiment.'
I am not interested in speculating over Einstein's motivations for making these
conflicting statements - these motive could surely have been varied and complex.
My aim here was only to bring out the fact that the Michelson Morley experiment
played only a minor role in his thinking about the problem.
mB EINSTEIN'S TIME
ABSTRACT. In the previous chapter we pointed out that the Michelson-Morley experiment was not
designed to test either the constancy of the speed of light or the existence of the aether. Here we argue
that a theoretical analysis of the concept of time in Newtonian theory is fundamental to an understanding
of the special theory of relativity. Historically, Poincare played a key role in carrying out this analysis. He
emphasized the principle of relativity in classical mechanics, rejected the aether and the associated notion
of absolute motion, pointed out that the qualitative problem of simultaneity was related to the quantitative
problem of measuring equal intervals of time, or measuring velocity, and correctly postulated the constancy
of the speed of light. We also point out that Abraham Pais' treatment of the origin of relativity repeatedly
misrepresents Poincare. Relativity leads to a deterministic, B-series, continuum view of time.

By chance a friend of mine in Bem (Michele Besso) helped me out. It was a beautiful day when
I visited him with this problem. I started the conversation with him in the following way:
'Recently I have been working on a difficult problem. Today I came here to battle against that
problem with you.' We discussed every aspect of this problem. Then suddenly I understood
where the key to the problem lay. Next day I came back to him again and said, without even
saying hello 'Thank you. I've completely solved the problem.' An analysis of the concept of
time was my solution. Time cannot be absolutely defined, and there is an inseparable relation
between time and signal velocity. With this new concept, I could resolve all the difficulties
completely for the first time.
Within five weeks the special theory of relativity was completed. I did not doubt that the new
theory was reasonable from a philosophical point of view .. ..
A. Einstein
How I Created the Theory of Relativityt

1 Introduction

I not
N the preceding chapter, we saw that the Michelson-Morley-Miller experiment was
designed to test the constancy of the speed of light. Rather, it was designed to
discriminate between the theories of Fresnel and Stokes, and contradictory results were
obtained by Michelson-Morley and Miller. Conversely, Einstein's original paper on
special relativity made only a passing reference to experiments to detect absolute
velocity, and postulated the constancy of the speed of light. None of Einstein's research
papers mention Michelson, and Einstein does not seem to have taken such experiments
very seriously.
Having seen all this, the next question that naturally arises is this: if the Michelson-
Morley experiment was not responsible for the birth of relativity, what was responsible?
A clue is provided by the fact that Einstein's original paper on special relativity was
called 'On the Electrodynamics of Moving Bodies'. Another clue is provided by the
anecdote that Einstein, during the examination for his admission to the Polytechnic
(ETH Zurich), told his examiners that he understood electrodynamics but not Newton-
ian mechanics. The discovery of Einstein's 'first paper' by Jagdish Mehra, some twenty
t Based on a talk given at Kyoto on 14 Dec 1922, when Einstein was unable to attend the Nobel Prize
ceremony at Stockholm, as he had already proceeded to Japan. Tr. from the Japanese by Y oshimasa A.
Ono. In: Spencer R. Weart and Melba Phillips (eds), History of Physics, Amer. Inst. Phys., 1985, p 244.
60 CHAPTER IIIB

years ago shed further valuable light on the subject. From his earliest days, Einstein was
engrossed in electrodynamics. Therefore, one must first examine the electrodynamics
of that time. The principal characters here are Lorentz, Poincare, and Einstein.
Our primary interest is not in the historical record of the origin of relativity. Rather,
our interest is in trying to understand and learn from the process by which the theory
was developed. The usual thing is to suppose that the way to develop theories cannot
be taught. But, the purely formal presentation of a subject as if it had been immaculately
conceived can be quite bewildering! Moreover, we propose to show that an analysis of
the concept of time was of central importance in the passage from Newtonian mechanics
to special relativity.

2 Electrodynamics

The marvelous experiments ofHertz have most ingeniously illuminated the dynamic nature of
these phenomena - the propagation in space, as well as the qualitative identity ofthese motions
with light and heat.
A. Einstein (1894) 1

2. 1 Lorentz #1

Lorentz's earliest analysis2 of the Michelson-Morley experiment went as follows. He

could not bring himself to accept Michelson's conclusion that Stokes' theory was the
correct one, because 'The difficulties which this [Stokes'] theory encounters in explain-
ing aberration seem too great for me to share this [Michelson's] opinion.' (These
difficulties were explained earlier in Chapter IlIA, p 54.)
After the initial troubles with the Michelson-Morley experiment had been sorted
out, there seemed to be only one way of accepting the conclusions of the experiment,
and rejecting Stokes' theory. The expected path difference in the Michelson-Morley
experiment was Lv2/ c2 Therefore, the only way out was to suppose that one of the arms
of the interferometer suffered a contraction by an amount :v'2. Lv2/c 2 Lorentz thought
that the arm of the interferometer which was pointed in the direction of the earth's
motion was shortened by the above amount.
His explanation for this was that the molecular forces, which held together the arm,
were increased by a factor of (1-v2/c 2)v\ perhaps due to the pressure of the aether.
This was the same formula that Lorentz had proposed earlier for the change of the
electrostatic forces.
Lorentz himself did not doubt the existence of the aether3:
Surprising as this hypothesis may appear at fIrSt sight, yet we shall have to admit that it is by
no means far-fetched, as soon as we assume that molecular forces are also transmitted
through the ether, like the electric and magnetic forces of which we are able at the present
time to make this assertion definitely.
 EINSfEIN'S TIME 61

2. 2 Lorentz #2

In his second attack on the problem, Lorentz refers to the experiments of Rayleigh and
Trouton and Noble, which, like the Michelson-Morley experiment, looked for second
order effects of the earth's motion. Rayleigh, in 1902, sought to find out whether the
earth's motion may cause a body to become doubly refracting. Trouton and Noble, in
1903, sought to detect a turning couple which would seek to align a condenser parallel
to the earth's motion. Lorentz, by now, had developed a picture of the electron as an
extended distribution of charge. The picture of a finite charge is suggested if one asks
the question: what is the density of charge in a moving frame? For a moving charge, the
picture of point charges and retarded distances does not give the correct potentials.
Rather, one obtains the correct Lienard-Wiechert potentials by assuming a spherical
charge distribution, and then passing to the limit.4
Lorentz now proposed that it was the individual electrons themselves, not merely
the forces between them, which were affected. He suggested that the individual
electrons shrank in the direction of motion. 'It will be easily seen that the hypothesis
which was formerly advanced in connexion with Michelson's experiment is implied in
what has now been said.'5
LorentzJormally introduces the transformations

X
,
= f3.x , t
, t
= 7J -
v2
{3 - 2 x ' {3
c
... r--;
= -V 1-~2
c
(1)

(Lorentz actually introduces a quantity I, which he later puts equal to unity.) He then
applies these transformations, for a reference frame moving with any velocity less than
that of light, to transform the system

div D = a, div H = 0, (2)

1 aD 1aH
curl H = -;; (at + av), curl D = c at'

F = D + 1. v x H.
c

He shows how this transforms to an essentially identical system in a moving frame,

except for the first equation, which transforms to

div'D' = (l-vu'x)a', a'=*, (3)

where the derivatives 'are now with respect to the transformed quantities, and

u'x = {32 Ux, U'y = {3Uy, u'z = {3uz. (4)

62 CHAPTERIIIB

Next he considers the change in the ponderomotive force (given by the last of the above
system).
Then we shall obtain the forces acting on the electrons of the moving system 1:, if we fIrst
determine the corresponding forces in [the static system] 1:', and next multiply their
components perpendicular to [the direction of motion] by {2/fJ2 [ = 1/fJ2].6

Lorentz goes on to repeat how all forces are influenced by translation in the same
way as the electrostatic forces. In order to retain Newton's law, F = rna, he is now forced
to introduce two different masses: one for acceleration in the direction ofmotion, and one
for acceleration perpendicular to this direction. He called these the 'longitudinal' and the
'transverse' mass respectively. 'I shall suppose that there is no other, no "true" or
"material" mass.'7 [Emphasis original] Thus, we can see how electrodynamics, at the
turn of the century, directly brought into question the validity of Newton's laws.
Lorentz later put forward the view that all the mass of the electron was electromag-
netic in origin. However, he did have a serious difficulty with the new coordinate t'
which he called 'local time' . Later on he commented8
The chief cause of my failure was my clinging to the idea that only the variable t can be
considered as the true time and that my local time t' must be regarded as no more than an
auxiliary mathematical quantity.

3 Newton's laws again

3. 1 Velocity-dependent forces

Why should electrodynamics contradict Newton's laws? After all, there is a marked
similarity between Newton's law of gravitation and the Coulomb force law. The only
difference is that charges may be positive and negative. The first discordant note was
struck by the fact that the seat of the magnetic inverse square force - the magnetic
monopole - could not be isolated (though efforts are still continuing!). The next
discordant note was struck by the fact that currents interact with magnets and with each
other.
With the development of the Lorentz theory of electrons, it came to be believed
that currents were no more than moving charges. Therefore, the force law between
moving charges was no longer a simple inverse square law, but took the more complicated
form of the Heaviside-Lorentz force law

F = e (E + v x B). (5)

The strange thing about this force was that it was velocity dependent. What is so strange
about a force which is velocity dependent? For this, we must go back to our analysis of
Newton's laws.
What exactly do Newton's laws of motion assert?

dv (6)
F = m dt .
 EINSfEIN'S TIME 63

3. 2 Homogeneity

The first thing we notice is that these laws involve the derivatives of position. An arbitrary
constant can always be added to a function without changing its derivative. Thus, the
laws are indifferent to the origin of the coordinates. In physical terms, this means that
space is homogeneous. (Another name for this is invariance under translations.) The
homogeneity of space seems a trivial matter to us, but it is well to recognize the centuries
of intellectual effort it took to arrive at this point of view. According to our immediate
experience, the earth is the natural centre of affairs. Copernicus moved this centre to
the sun. Homogeneity means the absence of a centre or other privileged point in space.
It was Kepler's discovery that the planetary orbits were elliptic (rather than circular)
which removed the Sun from the centre of the solar system and brought it to a focus of
an ellipse, which did not seem such a special place.

3. 3 Isotropy

We know from our earlier analysis that the above equation defines force. So, it is natural
to choose the simplest definition - linear dependence of force upon acceleration. With
this hypothesis, we would have the equation

(7)

Here, subscripts denote the components of force (F), mass (m) and acceleration (a)
respectively. If we require the force to be zero (if and) only if accelerations are zero,
then the mass matrix (mjk) should be invertible. More particularly, we may require
masses to be positive, i.e., the matrix m to be positive definite (all eigenvalues real and
strictly positive). Now it is easy to see that the assumption of a scalar mass (all
eigenvalues equal) is equivalent to assuming the isotropy of space.
Like homogeneity, the isotropy of space is far from obvious. There is an immediate
contradiction with everyday experience - moving up is more difficult than moving
sideways. Therefore, this formulation tacitly assumes an accompanying theory of grav-
itation, and if we abandon this assumption, we must modify the theory of gravitation.

3. 4 Straight-line motion

The other interesting feature of the above equation is that it is a vector equation: when
force is zero, not only the speed but the direction of motion is constant. This differed
from certain earlier beliefs that in the absence of external forces, a body moved with
constant speed (e.g. uniform motion in a circle) rather than constant velocity (which
implies straight-line motion). Kepler's discovery of elliptic orbits also aroused doubts
that uniform motion in a circle was the natural thing. Certainly a great achievement of
Newton's theory was that he could succeed in deriving elliptic orbits starting from the
assumption that straight-line motion was 'natural'.
64 CHAPTERIIIB

Of course there is much more to the vector formulation. The vector formulation
ensures that the laws are independent of the coordinate system employed. Assuming,
as usual, that the underlying model of displacement is Euclidean, this means that the
law is independent of the orientation of the coordinate axes. That is, the law is invariant
under rotations of the coordinate axes.

3. 5 Relative motion and force

The equation involves the second derivative of position. This is not necessary a priori.
The consequence is that one can add a constant velocity Vo to the v appearing in the
right hand side of the law without changing anything. Thus, the law is independent of
the velocity of the coordinate system, so long as this is constant. Two observers moving
with uniform velocity with respect to each other will agree on the definition of force;
or, as it is usually put, force is a Galilean invariant. A velocity-dependent force clearly
does not fit into this framework, and raises awkward questions: are the velocities
entering into the Heaviside-Lorentz force law relative or absolute velocities?
The Galilean invariance of Newton's laws is also sometimes taken to mean the
combination of all the above: that there is no preferred place for parking the coordinate
system (space is homogeneous), nor is there a preferred orientation of the coordinate
axes (space is isotropic), nor even is there a preferred velocity for the coordinate system.
The last observation has the important consequence that the only physically meaningful
velocities are relative velocities. This leads directly to the principle of relative motion.

3. 6 Newton's time

We should recall, of course, the flaw in the formulation of Newton's laws that was
considered in detail in Chapter II. All the above reasoning assumes that there is a
measure of equal intervals of time which is independently available.

4 Poincare

4. 1 Poincare and the relativity principle

Poincare's analysis of Newton's laws was to an extent similar to the above analysis. The
crucial thing is Galilean invariance, which follows from the fact that Newton's equations
are second order differential equations. Poincare's analysis of classical mechanics and
relative motion, of course, goes much deeper, and he considers and rejects the possi-
bility that Newton's laws could be first order differential equations.
Poincare generalized this to the principle of relativity:9 'the movement of any
system whatever ought to obey the same laws whether it is referred to fixed axes or to
the movable axes that are implied in uniform motion in a straight line't He regards this
as one of the basic principles of physics:
t Poincare, of course, mistrusts both concepts of 'uniform motion' and 'straight line'. See, e.g., Ref. 9, p
77
 EINSTEIN'S TIME 65

the principle of relativity [is] not only confirmed by daily experience, not only is it a necessary
consequence of the hypothesis of central forces, but it is irresistibly imposed upon our good
sense.1

It is this principle that seems to have convinced him that the aether did not exist.

4. 2 Poincare and aether

Poincare attacked aether on many grounds, not the least of which was its lack of
refutability. Though in later times Popper has been the staunchest champion of falsifi-
ability, many of these ideas including the 'simplicity of unity' (Occam's razor) were basic
to Poincare's approach.
For a correct understanding of Poincare's views on aether, it is necessary to bear
in mind the following. Being a mathematician, Poincare presumably did think in terms
of mathematical concepts, which tend to be sharply defined. But, by Poincare's time,
the word 'aether' had come to refer to a diversity of loosely associated notions.
Therefore, in an entirely non-mathematical exposition of the subject (at its then-fron-
tier), Poincare accurately, but confusingly, used the word 'aether' to denote two quite
distinct mathematical concepts.
In the context of relativity, 'aether' refers to the substratum needed for absolute
motion. In this sense, Poincare unambiguously rejects the aether. The velocities in the
Heaviside-Lorentz force law are relative velocities. Nothing else would be consistent
with his principle of relativity or the observation that Newton's laws are second-order
differential equations.
Optical and electrical phenomena ... might reveal to us not only the relative motion of
material bodies, but also what would seem to be their absolute motion.... Will this ever be
accomplished? I do not think So ... l1

But, apart from the order of the differential equations governing the evolution of a
physical system, there is also the question of the nature of these equations. Here the
aether performs another function which Poincare explained in detail in his Science and
Hypothesis (p 169).
Does our ether actually exist? We know the origin of our belief in ether. If light takes several
years to reach us from a distant star, it is no longer on the star, nor is on the earth. It must
be somewhere, and supported, so to speak, by some material agency.

He follows up with a remark that may yet be crucial to the origin of time-asymmetry.
The same idea may be expressed in a more mathematical and more abstract form ... in
ordinary mechanics, the state of the system under consideration depends only on its state at
the moment immediately preceding; the system therefore satisfies certain differential equa-
tions. On the other hand, if we did not believe in ether, the state of the material universe
would depend not only on the state immediately preceding, but also on much older states;
the system would satisfy eQ.Uations of finite differences. The aether was invented to escape
the breaking down of the laws of general mechanics.1 2 [Emphasis mine]

Today, the word 'aether' is not part of the physicist's vocabulary, but rejecting the
aether in this latter sense (together with the associated notion of action by contact) is
something yet to be accomplished, as we shall see in Chapters VA. VB, and VIA. In this
chapter, especially in the next two subsections, we will be concerned with the aether
only in the first sense, in relation to absolute motion.
66 CHAPTER IIIB

4. 3 Poincare and mechanical explanations

Descartes used to commiserate the Ionions. Descartes in his tum makes us smile, and no doubt
some day our children will laugh at us. Is there no way ofgetting at once to the gist ofthe matter,
and thereby escaping the raillery which we foresee?
H. Poincare l3

It is impossible to summarize Poincare (but I will nevertheless attempt to do so). In

Poincare's view, the older physics of central forces has been replaced by the physics of
principles, of which he identifies the following as basic: conservation of energy, the
entropy law, Newton's principle of equality of action and reaction, the principle of
relativity, conservation of mass, and the principle of least action.
In Poincare's view, aether and other such 'unknown fluids with which we might fill
space'l4 are introduced to provide a mechanical ('physical') interpretation of phenom-
ena. Such explanations, using 'one or more hypothetical fluids'l5 he regards as unim-
portant because of the following proposition. If the principles are satisfied he proves
that a mechanical explanation is always possible, and in this case an infinity of mechan-
ical explanations are possible.
As an illustrative example he considers the Fresnel theory of aberration:
The Fresnel theory enables us to do today as well as it did before Maxwell's time. The
differential equations are always true, they may always be integrated by the same methods,
and the results of this integration still preserve their value. It cannot be said that this is
reducing physical theories to simple practical recipes; these equations express relations, and
if the equations remain true, it is because the relations preserve their reality... only the
something that we then called motion, we now call electric current. But these are merely
names of the images.... The true relations between ... real objects are the only reality we
can attain.... Ifthe relations are known to us, what does it matter if we think it convenient
to replace one image by another.16

The aether as a basis for mechanical explanation is, therefore, completely useless
and redundant.

4. 4 Relative velocity and the Michelson-Morley experiment

Poincare, therefore, thinks that no experiment can ever reveal the presence of aether
or any other such hypothetical fluid. Moreover, the observation of aether, or absolute
velocity, would contradict the principle of relativity.
Many experiments have been made on the influence of the motion of the earth. The results
have always been negative.... we might expect to find accurate methods giving positive
results. I think that such a hope is illusory.... I do not believe, in spite of Lorentz, that more
exact observation will ever make evident anything else but the relative displacement of
material bodies. l7

For Poincare, principles were decided by the need for simplicity in a theory - they
came logically before experiments.
All attempts to measure the velocity of the earth in relation to the ether have led to negative
results. This time experimental physics has been more faithful to the principle [of relativity]
than mathematical physics; the theorists, to put in accord their other general views, would
 EINSTEIN'S TIME 67

not have spared it; but experiment has been stubborn in confirming it. The means have been
varied; finally Michelson pushed precision to its last limits; nothing came of it.l8
Nevertheless, he was not averse to reminding people about Michelson's experi-
ment.
Recall what we have said of the principle of relativity and the efforts made to save it. And it
is not merely a principle which it is the question of saving, it is the indubitable results of the
experiments of Michelson.19
His belief in the primacy of principle, of course, persists:
Michelson has shown us, I have told you, that the physical procedures are powerless to put
in evidence absolute motion; I am persuaded that the same will be true of the astronomic
procedures, however far precision be carried.20
To summarize, Poincare did not believe in the aether (or absolute motion) because
he believed in the principle of relativity, from his analysis of classical mechanics. He
regarded the aether as a redundant means of mechanical explanation which should be
discarded since it conflicted with the principle of relativity. The principle was primary,
and he regarded support of the Michelson-Morley experiment as incidental.

4. 5 Poincare on Lorentz's theory

Poincare was well aware that a crisis confronted theoretical physics at the tum of the
century. In The Value of Science he devotes a whole chapter to this crisis, pointing out
how every one of the principles he regarded as basic has been put into danger by new
discoveries. In particular, he points to the fact that the Lorentz theory puts into jeopardy
the principles of relativity, Newton, and Lavoisier (constancy of mass). As for resolving
this crisis, he suggests: 'It is a question before all of endeavouring to obtain a more
satisfactory theory of the electrodynamics of bodies in motion. '21
How should one go about it?
The Lorentz theory is very attractive. It gives a simple explanation of certain phenomena,
which the earlier theories - even Maxwell's in its primitive form - could only deal with in
an unsatisfactory manner; for example, the aberration of light, the partial impulse of
luminous waves, magnetic polarization, and Zeeman's experiment.22
Or again,
Look at the ease with which the new Zeeman phenomenon found its place, and even aided
the classification of Faraday's magnetic rotation, which had defied all Maxwell's efforts. This
facility proves that Lorentz's theory is not a mere artificial combination which must eventu-
ally find its solvent. It will probably have to be modified, but not destroyed.23
Poincare finds that the Lorentz theory is in conflict with the principle of relativity:

Consider two electrified bodies; though they seem to us at rest, they are both carried along
by the motion of the earth; an electric charge in motion... is equivalent to a current; these
two charged bodies are, therefore, equivalent to parallel currents of the same sense and
these two currents should attract each other. In measuring this attraction, we shall measure
the velocity of the earth; not its velocity in relation to the sun or the fixed stars, but its absolute
velocity.24
He also points out that the Lorentz theory is in conflict with the principle of
constancy of mass (Lavoisier's principle). '[Lorentz] says that the masses of all the
68 CHAPTER. I1IB

particles be influenced by a translation to the same degree as the electromagnetic mass of

the electrons. [Emphasis original]
So the mechanical masses must vary in accordance with the same laws as the electrodynamic
masses; they can not, therefore, be constant [emphasis mine].

Need I point out that the fall of Lavoisier's principle involves that of Newton's? This latter
signifies that the centre of gravity of an isolated system moves in a straight line; but if there
is no longer a constant mass, there is no longer a center of gravity, we no longer even know
what this is. This is why I said above that the experiments on the cathode rays appeared to
justify the doubts of Lorentz concerning Newton's principle.2S

Lorentz's doubt, as stated by Poincare, was that the constancy of mass

seems true to us only because in mechanics merely moderate velocities are considered, but
that it would cease to be true for bodies animated by velocities comparable to that of light.
Now these velocities, it is believed ... have been realized; the cathode rays ... may be formed
of very minute particles or of electrons which [have] velocities smaller no doubt than that of
light, but which might be its one tenth or one third.26

4. 6 The new mechanics

Poincare, therefore, proposes a new order.

The most satisfactory theory is that of Lorentz ... it still possesses a serious fault ...
[Newton's] principle according to Lorentz cannot be applicable to matter alone, ... it must
take into account the action of the aether on matter and the reaction ofthe matter on aether.
Now, in the new order, it is very likely that things do not happen in this way. 27 [Emphasis mine]

A few years later, in a famous address in 1904, Poincare announced the following
modifications to the theory of Lorentz:
From all these results, if confirmed, would arise an entirely new mechanics, which would be,
above all, characterized by this fact, that no velocity could SU[Jlass that of lightl , allY more
than any temperature can fall below absolute zero. [Original footnote: lBecause bodies
would oppose an increasing inertia to the causes which would tend to accelerate their
motion; and this inertia would become infinite when one approached the velocity ofli&IJt.]28
[Emphasis mine]

What about a moving observer? Could bodies exceed the speed oflight for such an
observer? Poincare continues:
No more for an observer, carried along himself in a translation he does not suspect, could
any apparent velocity surpass that of light; and this would then be a contradiction, if we did
not recall that this observer would not use the same clocks as a fixed observer, hut indeed,
clocks marking 'local time' 29 [Emphasis mine]

What is local time?

 EINSI'EIN'S TIME 69

4. 7 Poincare on time

It is therefore by cause that we define time; but ... when two facts appear to us bound by a
constant relation, how do we recognize which is the cause and which the effect? We assume that
the anterior/act, the antecedent is the cause ... It is then by time that we define cause. How save
ourselves from this petito principi? We say now post hoc, ergo propter hoc; now propter hOt;
ergo post hoc; how shall we escape from this vicious circle?
H. Poincare30

Poincare had emphasized (in an 1898 article on the measure of time, reproduced in The
Value of Science, Ch. II) the difficulty in the measurement of equal intervals of time, in
Newtonian mechanics.
We have not a direct intuition o/the equality o/two intervals o/time. Those persons who believe
they possess this intuition are dupes of an illusion. When I say, from noon to one the same
time passes as from two to three, what meaning has this affirmation? The least reflection
shows that by itself it has none at all. It will only have that which I choose to give it, by a
definition which will certainly possess a certain degree of arbitrarines. 31 [Emphasis original]
We have considered some of these difficulties in Chapters I and II.
But he went further, and emphasized that 'It is difficult to separate the qualitative
problem of simultaneity from the quantitative problem of the measurement of time.'32
He explains by an example concerning the measurement of the speed of light.
Could not the observed facts be just as well explained if we attributed to the velocity of light
a little different value from that adopted, and supposed Newton's law only approximate?
Only this would lead to replacing Newton's law by another more complicated. So for the
velocity of light a value is adopted, such that the astronomic laws compatible with this value
may be as simple as possible. When navigators and geographers determine a longitude, they
have to solve just the problem we are discussing; they must, without being at Paris, calculate
Paris time. How do they accomplish it? They carry a chronometer set for Paris. The
qualitative problem of simultaneity is made to depend upon the quantitative problem of the
measurement of time.33
He concludes with simplicity as the guiding rule:' ''The simultaneity of two events,
or the order of their succession, the equality of two durations, are to be so defined that
the enunciation of the natural laws may be as simple as possible." '34
Therefore, Poincare had no difficulty in accepting that the 'local time' defined by
Lorentz was indeed physical time. He interprets 'local time' as a way of synchronizing
clocks.
The most ingenious idea was that of local time. Imagine two observers who wish to adjust
their timepieces by optical signals; they exchange signals, but as they know that the trans-
mission of light is not instantaneous, they are careful to cross them. The clocks are adjusted
if the slowness equal to t represents the duration of the transmission, and to verify it, station
B sends in its turn a signal.... The timepieces are then adjusted.
And in fact they mark the same hour at the same physical instant, but on the one condition,
that the two stations are fixed. Otherwise ... the watches adjusted in that way will not mark...
the true time; they will mark what may be called the local time, so that one of them will gain
on the other. It matters little. since we have no means ofperceivin& it.3S [emphasis mine]
About the Fitzgerald-Lorentz length contraction, Poincare is not happy with 'that
extraordinary contraction of all bodies', and asks 'could we not make a more simple and
70 CHAP1ER IllB

natural hypothesis?'36 (Poincare derived the group properties of the Lorentz transfor-
mations shortly thereafter; an excellent introduction to the relationship between
groups, geometry, mechanics, and experiment may be found in the first few chapters of
Science and Hypothesis.)

5 Einstein

The 'simple and more natural hypothesis' provided by Einstein was that the Lorentz
length contraction followed as a consequence if moving observers used clocks which
measured 'local time'. It might seem strange that a length measurement should depend
upon the clocks that one uses. But let us tty to understand the length contraction What exactly
is meant by a measurement of length? Suppose I want to measure the length of a rigid rod, ie.,
a straight line segment,AB. I need onlyuse a ruler to mark off the coordinates ofA andB along
the x-axis. But now supposeAB is moving along the x-axis. Thex-coordinate ofB will be zero
at some time, and that ofA will be zero a little later. From this, should one conclude that the
length of the moving rod is zero? Evidently not, for one must measure the coordinates ofA and
Batthesametime,i.e.,simultaneously.Butwehavealreadyseenthattheproblemofsimultaneity
depends upon the problem of measurement of time, so before making a length measurement, I
must decide which clock to use. This is decided by the rule enunciated above: I must use that
clock which leads to the simplest formulation of the physical laws. The rest of relativity follows.
A curious feature of Einstein's original paper is that Einstein shows the formal
covariance of only the equations of electrodynamics. He derives the transformation law
for density by including 'convection currents' in the dynamical equations. Specifically,
he is silent on Coulomb's law. If two relatively static charges are moved apart, and then
remain relatively static, what happens if a relatively static observer at 'infinity' measures
the Coulomb field at infinity? Of course, the Coulomb field decays rather more rapidly
than the radiation field, but is it possible, in principle, to transfer information to 'infinity'
in this manner, at a speed faster than light?

6 Time

Having seen how relativity arose from a re-examination of the concept of time, let us
re-examine the questions about time raised in Chapter I. In the first place, a definition
of 'equal intervals of time' may be obtained by 'bouncing a photon', i.e., repeatedly
reflecting a 'ray' of light, between two mirrors at rest with respect to each other. Since
the mirrors remain the same distance apart, and since the speed of light is constant by
definition, the corresponding intervals of time are equal by definition. Length measure-
ments (and eventually, in general relativity, mass measurements, hence all measure-
ments) are to be madet in terms of such a 'proper clock'.
t For a systematic approach along these lines, see J. L. Synge, Relativity: the General Theory, North
Holland, Amsterdam, 1965.
 EINSTEIN'S TIME 71

Secondly, in the relativistic view of things, it is not possible to tie up future and past
with existence. Because the notion of simultaneity differs for observers moving with
respect to each other, an event which is 'future' for one may be 'now' for another, and
'past' for a third. To speak Lorentz invariantly, one cannot speak of things coming into
existence or passing out of existence; one must speak of the entire past and future
history, the world-line of a particle. A month before his own death, Einstein consoled
the bereaved wife of his friend, Besso, by pointing out that the departed friend
continued to exist in some sense somewhere: 'For us practising physicists, the division
into past, present, and future is merely an illusion, albeit an obstinate one.' Thus,
relativity commits us to a continuum, deterministic, B-series view of time.
Finally, there are a few temporal relationships which remain Lorentz invariant. The
Lorentz transform may rotate the plane of simultaneity, but it leaves unchanged the
path of light rays or the null cone. The null cone, then, can be divided into two parts,
the absolute future and the absolute past of the vertex. These are absolute in the sense
that any event in the future null cone will be to the future of the vertex with respect to
any observer. However, relativity does not provide any physical distinction between the
absolute past and the absolute future.

7 Conclusions

The Michelson-Morley experiment did inspire the Fitzgerald-Lorentz length contrac-

tion hypothesis. Nevertheless, from an analysis of the laws of classical mechanics, and
particularly the concept of time in Newtonian mechanics, one can derive the principle
of relativity, postulate the constancy of the speed of light, and reinterpret the experi-
ment as indicative of the absence of absolute motion. Relativity leads to a deterministic,
B-series picture of time as a continuum, though it does not establish any physical
asymmetry between the past and the future.
72 CUAPTERIllB

Appendix: The historical record

In the Oct-Dec 1990 issue of Physics Education (India), we find a letter to the effect that
it was natural for Einstein to have got angry on hearing that Whittaker had written a
book crediting Poincare with the discovery of relativity. Our presentation may also have
raised questions concerning priorities and credits for the special theory of relativity.
Would it have angered Einstein?
It is not my purpose to answer these questions here. One reason is that such an
investigation would be out of context. Another is that there are many details which
would bear investigation. I do not, for instance, know the month in which Poincare's
Value of Science was published in 1905. I do not know whether Planck was aware of
Poincare's articles in French, when he picked up Einstein's work.
It is not even clear to me how credits ought to be allocated. In a football team does
all the credit go to the striker who shot the goal? In that case should we credit the general
theory of relativity to Hilbert because he published an independent derivation of the
field equations fully five days ahead of Einstein?37
The objective of the exposition was to bring out the process of theoretical analysis
which played a crucial role in the genesis of relativity. In the process, several miscon-
ceptions have, I hope, been cleared. But, some doubts have been raised, and the student
who pursues these further will find other hurdles in the literature.
To learn more about the origin of special relativity, one would naturally tum to a
biography of Einstein. The current authoritative biography of Einstein is the one by
Abraham Pais 'Subtle is the Lord... '. Without any doubt, the book is, on the whole,
extremely well written and deeply researched. But, in examining the origin of special
relativity, Pais has been unfair to Poincare. (This is one reason for the extensive quotes
from Poincare - I did not want to be accused of reading into Poincare what did not
exist.) My primary intention in the following is to attempt to deallocate some of the
discredit that Pais heaps on Poincare.
In the first place, there is the question: did Einstein at all know about Poincare's
work?
A.l Solovine on Einstein and Poincare
It is certain that Einstein had read Poincare's Science and Hypothesis, which first
appeared in 1902.
Not only had Einstein, together with K. Habicht and M. Solovine, read this book,
but Solovine (who maintained a list of the books read) commented that this book
'profoundly impressed us and kept us breathless for weeks on end.'38 To the last
comment I would like to add that after reading Poincare's piece on classical mechanics
I find my own exposition (Chapter II) shabby by contrast; I may have covered the
essential points, but Poincare's exposition dazzles.
A.2 Whittaker, Born and Pais on the origin of special relativity
How have others looked at the matter? Those who have expressed their doubts can
learn from the following example.
In 1951 Whittaker wrote a companion second volume to his book History of the
Theories of Aether and Electricity. The second volume covers the period from 1900 to
1926, and has a chapter called 'The Relativity Theory of Poincare and Lorentz'.
 EINSfEIN'S TIME 73

Whittaker's view seems to have been that Einstein only amplified the relativity theory
of Poincare and Lorentz. Abraham Pais after presenting the above evidence, comments
that Whittaker's book 'shows how well the author's lack of physical insight matches his
ignorance of the literature.'39 Pais seeks to soften this comment, and make it appear
balanced by remarking about the earlier volume that 'colleagues more knowledgeable
on this period than I, confirm my impression that [the first volume] is a masterpiece.'4O
When Max Born voiced similar doubts, Pais says that Born 'did not acquit himself
well'41 because he asserted 'The reasoning used by Poincare was just the same as that
which Einstein introduced in his first paper of 1905 .... Does this mean that Poincare
knew all this before Einstein? It is possible .. .'42
A.3 Pais and Poincare
What, then, was the truth? This is difficult to ascertain, but it is certain that Pais'
treatment of Poincare is biased. Pais' basic claim is that Poincare required, in addition
to relativity and the constancy of the speed of light, a third hypothesis, viz. the contraction
of rods. What is the evidence?

A.3.1 The third and fourth hypotheses. According to Pais (p 168), Poincare himself
asserts the need for a third hypothesis:
I repeat what Poincar6 said in St Louis in 1904. On that occasion he also introduced in
essence the first two postulates and then added, 'Unfortunately, [this reasoning] is not
sufficient and complementary hypotheses are necessary; one must assume that bodies in
motion suffer a uniform contraction in their direction of motion.'
Pais seems to have overlooked something here. There is, in fact, afourth hypothesis in
the same paragraph, two sentences away:
And then, there is still the hypothesis about forces. Forces, whatever be their origin, gravity
as well as elasticity, would be reduced in a certain proportion in a world animated by uniform
translation; or, rather, this would happen for the components perpendicular to the transla-
tion ... Resume then, our discussion of the two electrified bodies.43
This sounds oddly like the Lorentz theory, and one turns to the previous page for
the context: 'Their task was not easy, and if Lorentz got through it, it is only by
accumulating hypotheses.'44
There can be no ambiguity of interpretation. Poincare is discussing the hypotheses
in the Lorentz theory. More specifically, Poincare is discussing his favourite example:
how two charges on earth contradict the principle of relativity, and how the Lorentz
theory would account for this. Science and Hypothesis ended on this note, and the above
discussion ends 'Thus, the principle of relativity has been valiantly defended in these
latter times, but the very energy of the defense proves how serious was the attack.'
Poincare's own view is quite different. He is well aware that the accumulation of
hypotheses, resulting in loss of refutability, renders a theory sterile and in need of a
replacement: 'Is it not evident that from the principle, so understood, one could no
longer hope to infer anything.'4S 'Now, in the new order it is very likely that things do
not happen thuS.'46 The new mechanics foreseen by Poincare, in 1902 and announced
in 1904 has already been described above, in 4.6.

A.3.2 Complementary hypotheses. The importance of Pais' hypothesis is clear: we find

the above assertion repeated on p 128 of his book. He mentions Poincare proposal to
74 CHAPrnR IIIB

synchronize clocks by light signals: 'Clocks regulated in this way will not mark the true
time, rather they mark what one may call the local time', omitting the next sentence 'It
matters little, since we have no means of perceiving it.' Pais continues,
Poincare is getting close. But then he falters: 'Unfortunately [this reasoning] is not sufficient
and complementary hypotheses are necessary [italics Pais']; one must assume that bodies in
motion suffer a uniform contraction in their direction of motion.' The reference to comple-
mentary hypotheses makes clear that relativity theory had not yet been discovered.

As shown above the reference to the complementary hypotheses is Poincare's criticism

of Lorentz's accumulation of hypotheses. I wonder what this criticism makes clear?

A.3.3 'We are not yet there'. Pais is not prepared to give Poincare credit for the new
mechanics, based on the constancy of the speed of light, and the Lorentz transformation,
which he announced in 1904, and the details of which he submitted for publicatidn in
1905. For example,
Poincare concludes this [St Louis] lecture with another one of his marvelous visions:
'Perhaps we must construct a new mechanics, of which we can only catch a glimpse, ... in
which the velocity of light would become an unpassable limit.' But, he added, 'I hasten to
say that we are not yet there and that nothing yet proves that [the old principles] will not
emerge victoriously and intact from this struggle.' 47

Perhaps a more detailed quote will help to settle the context. Having discussed the
crisis in mathematical physics, where each principle has been challenged, and having
proposed the new mechanics (4.6 above), Poincare is discussing what the theoretical
physicist should be doing. He discusses Michelson's experiment and relative motion.
He also discusses electrons and spectra, and recommends:
A Japanese physicist, M. Nagaoka, has recently proposed an explanation; according to him,
atoms are composed of a large positive electron surrounded by a ring formed by a very great
number of very small negative electrons....This is not yet wholly satisfactory, but the attempt
should be renewed.

He points out that 'if a principle ceases to be fecund, experiment without con-
tradicting it directly will nevertheless have condemned it,' so that 'it would be necessary
to rebuild anew.'
What direction would this renewal take?
As 1 have said, we have already passed through a like crisis. 1 have shown you that in the
second mathematical physics, that of principles, we find traces of the first, that of central
forces; it will be just the same if we must know a third.... We can not foresee in what way
we are about to expand; ... perhaps it is the kinetic theory which [would] serve as a model.. ..
Physical law would then assume an entirely new aspect; it would no longer be solely a
differential equation, it would take the character of a statistical law.
Perhaps, too, we sha1l have to construct an entirely new mechanics that we only succeed in
catching a glimpse of, where, inertia increasing with velocity, the velocity of light would
become an impassable limit. The ordinary mechanics, more simple, would remain a first
approximation, since it would be true for velocities not too great, so that the old dynamics
would still be found under the new. We should not have to regret having believed in
principles.... They are so useful, it would be necessary to keep a place for them. To determine
to exclude them altogether would be to deprive oneself of a precious weapon. 1 hasten to
say in conclusion that we are not yet there .... [Emphasis mine]
 EINSI'EIN'S TIME 75

A.3. 4 Pais' derogatory suggestions. Pais' entire account is strewn with derogatory sug-
gestions concerning Poincare. The central question, according to him48 is: 'Why did
Poincare never understand special relativity?' He doesn't have the same difficulty with
Lorentz.49 'For example, Lorentz still opines that the contraction of rods has a dynamic
origin. There is no doubt that he had read and understood Einstein's papers by then.'
According to Pais,so it is not significant that Einstein never cited Poincare:
It cannot be said that the content of Einstein's June 1905 paper depends in any technical
sense on these important remarks of Poincare. Others in Einstein's position might perhaps
have chosen to mention Poincare at the earliest opportunity. However, it does not seem to
me that Einstein had compelling reasons to do so in 1905.
But PaisSl finds it significant that Poincare never acknowledged Einstein: 'Shall I
write of petulance or professional envy. I shall not .. .' It is hard to understand why
Poincare should have felt envious of Einstein. Poincare was an established figure,
recognized as 'a brilliant creative genius', and 'at the present moment unquestionably
the most powerful investigator in the domain of mathematics and mathematical physics
... .' Moreover, one must recall Pais' investigation into Poincare's lack of understanding
of relativity. Why then should the thought be stated, even if only to. be withdrawn?

A.3.5 Future directions. It can be seen that the issues involved are difficult, and getting
to be more so as more is written about them. It would take an expert historian of science
to sort out these issues. I do not claim to be one. But we have before us the testimony
of two expert historians. Whittaker apparently gives the major credit to Poincare and
Lorentz. On the other hand, the manner in which Pais has denigrated Poincare, and
quoted him out of context, suggests the inference that Pais thinks this is the only way in
which credit can be given to Einstein. So this testimony leaves me wondering.
We must wait for another expert historian to clear up the issues that have become
clouded. To quote Poincare one last time, 'While waiting, our doubts remain.'
c. The arrow of time
;;;;;;; ; ;;;;- ;;;;: ~,-;;;;;;;;;;;;;;;;;-:;;:;;;::; :;;;;." ;::;,:;,;';:,;:;,;;:: : :

=
IV THERMODYNAMIC TIME
ABSTRACf. In the previous chapters we saw how relativity resolves the problem of time measurement,
and reduces all measurement to time measurement. Relativity, though, abolishes the intuitive notion of
'becoming'. In this part we begin consideration of the problem of time asymmetry in physics.
We explain how entropy relates to lack of information, and the entropy law to the commonsense belief
that one is surer of the past than of the future. The reversibility and recurrence paradoxes are discussed in
the context of the general recurrence theorem, which is stated and proved. The usual way of resolving these
paradoxes, using the Ehrenfest model and large recurrence times, is explained together with its weaknesses:
recurrence times can be calculated only on the assumption of ergodicity, and the validity of this assumption
for cosmic evolution is unknown. Moreover, there is no external measure of time with respect to which the
cosmos spends most of its time in thermodynamic equilibrium.
We briefly enumerate other attempts to resolve these paradoxes, including Tipler's 'no-return theorem',
but excluding quantum mechanics. We conclude that, classically, the infinite degrees of freedom provided
by the field offer the best hope of an escape route.

1 Introduction

1. 1 Summary of earlier chapters

I concerned
N Chapter I we stated a number of unresolved problems related to time. One problem
the measurement of time, and the difficulty in defining equal intervals of
time - one can't put two intervals of time side by side to compare them. In Chapter II
we pointed out how the problem of time measurement carried over into the difficulty
in Newtonian mechanics of defining two unknowns - 'force' and 'equal intervals of
time' - by means of one equation. In Chapter II1B we showed how this issue became
important in the context of velocity-dependent forces in electrodynamics, and how
Poincare related the quantitative problem of defining equal intervals of time to the
qualitative problem of simultaneity. We also explained how the problem was ultimately
resolved by postulating the constancy of the speed of light, and using a 'bouncing
photon', or a repeatedly reflected ray of light, to construct a proper clock. (In Chapter
IlIA we had shown that the Michelson-Morley experiment was not meant to test the
constancy of the speed of light, and had only an indirect influence on this postulate.)
Relativity theory, thus, rendered equivalent the measurement of length and time.
We had also pointed out (in Chapter IIIB, p 70) that in general relativity mass
measurements, hence all measurements, reduce to the measurement of time.

1. 2 Relativity and existence: loss of time asymmetry

Had Isaac Barrow been right when he said that 'time has length alone, and is similar in
all its parts .. .', all problems concerning time should have been resolved with the
resolution of the problem of time measurement. But the other problems listed in the
conclusions of Chapter I have not become irrelevant.
80 CHAPTER IV

The problem of time measurement was resolved by adopting a 'static' or B-series

view of time. The naive idea of using 'coming into existence' or 'becoming' to distinguish
between past and future has been abolished. In relativity, only entire world-lines
(complete past and future histories) are meaningful.
The difficulty is that along with the naive view, all distinction between the past and
future seems to have been abolished. With a little stretch of the imagination, it may be
possible to accept the B-series view that time does not involve change. But if one cannot
physically distinguish between past and future, one no longer knows what it is that one
has succeeded in measuring!

1. 3 Physics and the direction of time

Does the rest of physics provide any physical distinction between past and future? Is
there any physical asymmetry between past and future? Does physics help to decide the
direction of time or the arrow of time? This question will be central to the remaining
parts of this exposition.

1. 4 Time symmetry ofphysics

The laws of classical physics treat past and future on an equal footing. We say that they
are time-symmetric, meaning that they are unaffected if past is replaced by future, and
future by past. The laws of classical physics can be expressed using differential equa-
tions, which are unaffected if t is replaced by -to (Quantum field theory exhibits CPT
invariance: along with the transformation, T, which replaces t by -t, one must also
conjugate the charge, C, and reverse the parity, P.) These equations (including the
SchrDrlinger equation) can be solved either forward or backward in time, and both
solutions represent possible physical processes.
To get a feel for what this really means, watch a video sequence while rewinding:
to see the time reverse of the original sequence of events. The time symmetry of the
laws of physics means that the sequence of events projected backwards is perfectly
possible according to the laws of physics.
Thus, if one drops a tennis ball, it will bounce a few times and come to rest. At each
bounce, the ball passes on some of its energy to the molecules of air. In the time-reverse
situation, the molecules of air would 'collectively conspire' to pass on their energy to
the ball, which would then start bouncing so that it would bounce higher after each
bounce!
If one drops a stone into a pond, ripples spread outwards. In the time reverse case,
the ripples would converge on to the stone and throw it out of the pond straight into
one's outstretched hand! The time symmetry of physics means that such things are
physically possible.

1. 5 Is time symmetry reasonable?

From an intuitive point of view, this sort of thing seems very disagreeable. The time
reverse situations seem fantastic and contrary to experience. The disturbing thought is
that physical laws, with their time symmetry, are also grounded in experience. How
should we resolve this conflict in our experience?
 THERMODYNAMIC TIME 81

Let us re-examine the notions of past and future. In our mind there is a clear
distinction between the past and the future. For example, we think that we know the
past but not the future. How well-founded is this belief?
Recently the Indian nation was plunged into turmoil because of disputes concern-
ing the remote past, because the remote past is not easily knowable. If two historians
disagree over the authenticity of a manuscript, the dispute is not easy to settle. If two
geologists disagree about the extinctions of dinosaurs 65 million years ago, it is harder
still to settle the dispute.
What about the immediate past? One is usually quite certain regarding some of the
events that took place, say, five minutes back or yesterday or the day before. But how
reliable is this certainty? Surely that piece of paper that got lost was right there on the
table five minutes back? Couldn't one have sworn that one had met that person before?
And if memory occasionally plays tricks, there may be events, such as Rajiv Gandhi's
assassination, which are hard to reconstruct. So, our knowledge of the immediate past,
while certain, cannot be said to be always certain beyond doubt.
On the other hand, one is fairly certain of some of the things one will do tomorrow.
Brushing one's teeth, or having a cup of tea in the morning, is no~ quite what one has
in mind when speaking of the glorious uncertainty of the future. It is true that an
earthquake might disturb the morning schedule, but is that more likely than the
possibility of memory playing tricks?
To sum up, the remote past is as unknowable as the remote future, and one can
predict the immediate future nearly as well as one can remember the immediate past.
Since there seems no qualitative difference between the past and the future, our intuitive
distinction between past and future must be based on a quantitative difference between
the uncertainty regarding the past and the uncertainty regarding the future. We are surer
of the past than we are of the future. Are we?

2 The entropy law

The entropy law, also known as the second law of thermodynamics, is a precise
statement of the above belief.
The classical statement of the second law of thermodynamics is the following.
Statement 0: Heat does not flow by itself from cooler to hotter bodies.
The qualification 'by itself is essential. A refrigerator removes heat from the cooler
cabinet and throws it out into the hotter room. But the refrigerator consumes energy
to carry out this process, and so generates heat: if one were to keep the door of the
refrigerator open, and put the refrigerator in a closed room, the room would only heat
up. For the same reason, an air-conditioner is fixed so that half of it is outside the room
- otherwise the air-conditioner would always heat up the room.
A more modern statement of the second law of thermodynamics is the following.
Statement 1: The entropy of a closed system never decreases.
The phrase 'by itself has been replaced by the phrase 'closed system'. Imagine that all
the bodies concerned have been put into a sealed room (an adiabatic enclosure) and
there is no external interference, so that everything inside the room functions 'by itself
or 'of its own accord'. Then the bodies inside the room form a closed system. If a body
82 CHAl'TERIV

heats up, its entropy increases. Since heat cannot escape from the sealed room, the
modern version says that the room will not cool down. So the modern version seems to
say more or less the same thing as the classical one.

2. 1 The meaning of entropy

Does the flow of heat 'of its own accord' or otherwise have anything to do with the
quantitative difference between memory and expectation? Intuitively, the flow of heat,
or any other irreversible process, is symptomatic of the growing anarchy in the world;
left to itself, a system seems to move irreversibly from an ordered or low entropy state
to the most disordered state of thermodynamic equilibrium, or maximum entropy.
Intuitively, the entropy is a measure of disorder, or lack of information. In an orderly
room, a blind man can find his bearings by identifying one object. In a disorderly room,
he must first construct a mental map. One may make this concept precise, as follows.
(See also Box 1.)
The entropy S is given by a function of the following sort:

= - 2:
(1)
S Pi In Pi,
i

where the Pi correspond to a probability (of event i), Pi>O, and 2: Pi = 1. For an
i
arbitrary (discrete) probability distribution, the meaning of the function S may be
ascertained as follows. Suppose that the random variable X can take on five values with
equal probability. For example, suppose we know the first letter 'p', and the last letter
't' of a three-letter code: we know that the intermediate letter must be one of the five
vowels a, e, i, 0, u, and we suppose that all five of pat, pet, pit, pot, put are equally likely.
Suppose also that one must determine the value of X based only on yes-no type of
questions, and that each question is very expensive so that one follows the best possible
strategy. What is the average number of questions that one would have to ask?
This average may be calculated as follows. Let us choose the following strategy.
First divide the five values into a group of three values, say {i, 0, u}, and a group of two
values {a, e}. Next, divide the group of three into subgroups of one, say {i}, and two.
Only one question (IOU?) is required to decide in which group the actual value lies,
and one more question (I?) to isolate the subgroup. Thus, the actual value can be
ascertained in a maximum of 3 questions, in 2 cases, and a minimum of 2 questions in
3 cases, all cases being equally possible so that each case has probability lis. If one keeps
repeating this process, the number of questions required will vary, and the average or
expected number of questions would be 3 x s + 2 x:Ys = 2.4, which is in good agree-
ment with log2 5 = 2.32, Le., the entropy assuming that all possibilities are equally likely,
so that eachpi = lis in equation (1), except that the base of the logarithm is 2, corre-
sponding to a choice of the unit. Exact agreement obtains whenever the number of
possibilities is a power of 2, e.g., 2 = log2 4 questions are required if there are 4 equally
likely possibilities.
 THERMODYNAMIC TIME 83

If the probabilities were unequal, say 0.3, 0.2, 0.2, 0.15, and 0.15, the average
number of questions turns out to be 2.3, whereas S = 2.27. In general the average
number of yes-no questions is greater than or equal to the entropy S, and S may be
regarded as the minimum average number of yes-no questions that are needed.
The average number of questions one must ask is a measure of one's lack of
information about a system: the less information one has, the more questions one must
ask (which perhaps explains why some students are afraid to ask questions!). Thus, the
entropy represents the lack of information about the system. For other ways of looking
at entropy, see Ash. 1
Let us further acknowledge that the only truly 'closed system' that we have is the
cosmos. Then the entropy law states the following.
Statement 2: The entropy of the cosmos never decreases.
That is, the entropy of the cosmos at a future instant must be at least as large as the
entropy at a past instant. That is, one has at least as much information about the past as
one has about the future. This is the same thing as saying that memory is not less reliable
than expectation.

3 The Boltzmann H-theorem

Ever since Clausius and Boltzmann, physicists have not been happy with the above form
of the entropy law, because it does not actually say that we are surer of the past than we
are of the future. It is not of much use to say that entropy never decreases. What if it
never increases as well, and remains constant? This would correspond to a situation
where one is equally sure of past and future, so that expectation would have to be treated
on par with memory.
But more mundane problems would arise; for if entropy remains constant, heat
need not flow, by itself, from hotter to cooler bodies. But this is something that is actually
observed, and physicists would like to explain this observed phenomenon, using some
general physical law such as the entropy law.
The second reason why physicists tend to be unhappy with statement 2 of the
entropy law is that the statement is semi-empirical: it does not have the force of theory
behind it. In keeping with Panini's principle of laghava or brevity (or Occam's razor),
one would like to show that this form of the entropy law is a logical consequence of the
basic laws of mechanics - Newton's laws of motion, say.
Thus, one would like to replace statement 2 of the entropy law by a statement to
the following effect.
Statement 3: The entropy of a closed system, or the cosmos, goes on increasing with time
until it reaches its maximum value (in thermodynamic equilibrium).
This is the more ambitious H-theorem, called a theorem because one tries to prove it,
starting from the laws of mechanics, and some plausible assumptions. The 'H' refers to
the negative of entropy, which must have a minimum.
If the H-theorem were true, one could use it to define an earlier-later relationship:
later times are those with higher entropy. Thus, the H-theorem, if valid, provides a
thermodynamic asymmetry between past and future, or a thermodynamic arrow of time.
Many physicists regard this as the primary source of time asymmetry.
84 CHAPfERIV

Box 1. Pictures of entropy and evolution .

1. 1hennodynamic entropy and i"eversible evolution

dQIT is a perfect differential for reversible processes, dS = dQIT. For a perfect gas
in equilibrium,s S = N k 10g(JIT:>l). This entropy represents the amount of heat
'unavailable' for 'useful' work.
For beat, or anyform of energy, to perform work, anon~brium situationis necessary.
The ocean is at an absolute temperature of several hundred degrees. But a ship sailing on the
ocean cannot make use of this energy for it is in equilibrium with the ocean. Irrevemble
processes, sucb as friction, transfer energy from non-equilibrium to an equilibrium situation.
Most processes in nature are irreversible, and the notion of revemble proce&'i is an idealization.
2. Boltzmann entropy and shuffling
Conc;ider 6-dimensional ,u-space. Each molecule (point mass) of the gas has six degrees of
freedom, and its dynamical state is completely specified by a single point in,u-space. Thus, the
state of a gas with N molecules is completely specified by prescnbing N points in,u-space.
Because of the difficulty in describing the motion of this swarm of N points,
,u-space is divided or discretized into celIs which are small compared to macroscopic
dimensions, but still contain a large number of molecules. We can, thus, obtain the density
of gas particles or points in,u-space,/ (q,p, t), or the probability clistribution in,u-space,
which is all that is needed to descnbe the macroscopic properties of tbe gas. One often
assumes spatial uniformity, constant energy, and a constant number N of gas particles, so
that only the distribution function I (p, t) orI (v, t) is of concem Equipartition of energy
leads to the Maxwell-Boltzmann distribution.
The H-function is now defined by

H(t) = f I(v) log/(v)dv. (Blot)

H is minimized by the M-B distribution, and in this case of equilibrium it is related

to the entropy S by S = -kH V. Thus, H provides an extension of the notion of
thermodynamic entropy to the non-equilibrium case.
In this picture of evolution, the following is assumed.
(i) Shuffling: Points in.u-space are shuftled at random.
(ii)Moleadarchaos: Velocities (and positions) of molecules are statistically independentbej'ore
collision.
(ill) Principle ofmiao-reversibiJity. The cross-section for a collision is equal to the cross section
for the reverse collision. I (continued on D 8S)
The H-theorem has an immediate intuitive appeal: the inevitability of the approach
towards thermodynamic equilibrium is rather closely analogous to the apparently
irreversible processes of aging and death. Ever since Boltzmann, more than a century
ago, many attempts have been made to prove this theorem, with Newton's laws taken
as the basic laws of mechanics.
 THERMODYNAMIC TIME 85

3. Fine-grained entropy and Hamiltonian evolution

This is the entropy associated with the standard picture of Hamiltonian evolution in
6N dimensional phase space or r -space. The exact state of the g~ at any instant is desaibed
by a ingIe point in phase space, and its evolution by a single trajectory. Uncertainties in the
microphysical state are described at the macroph)'Sicallevel by means of a number of identical
mental copies of the system, an ensemble, represented by a swarm of points in r -space. The
swarm of points in r-space is imaginary, compared to the swann ofN points in,u-space.
The relative density of this swarm of points, or the probability distribution in
r-space, p (ql, ..., q3N.Pl ..... P3N. t) describes the statistical properties of the system;
p = constant is called the uniform ensemble, andp(E) = 6(E) is called the micro-canon-
ical ensemble. The classical liouville theorem states that the swarm of points in r -space
behaves like an incompressible fluid, dp/dt = D. the derivative being taken in coordinates
movingwith the fluid, so that the density of points is conserved, or probability is preserved.
The fine-grained entropy is defined by

set) = JaCt) log aCt) dq dp, (B1.2)

and remains constant by Liouville's theorem.

4. Coarse-grained entropy and Markovian evolution
The discrete cells in,u-space correspond to finite regions. called 'stars'. in r-space.
Due to the shuffiing of points in,u-space, the representative point wanders from star to star in
a stochastic manner.
Specifically, the assumption is that this evolution is Markovian: given an exact
knowledge of the current macrostate. knowledge of the past history is redundant. If
one is interested in the behaviour at only discrete instants of time, n. one obtains a
Markov chain. Mathematically, the requirement is that the conditional probability

Pr {X(n)=x(n) IX(nl)=X(nl),X(n2)=X(n2), ... ,x(nk)=X(nk) }

= Pr{X(n)=x(n) I X(nk)=X(nk)} (B1.3)

for all 'instants of time' ni, such that nl < m < ... < nk . I (continuedonp 86)
The use ofNewton's laws, rather than relativity. may suggest a basic problem right here.
since we know that in Newton's laws there is a conceptual difficulty with time. So the use of
Newton's laws might amount to adding chaos to confusion! With a finite speed of interaction,
there is a fundamental change in the nature of the many-body equations of motion: they
become history-dependent, hence time asymmetric. Nevertheless, we will review the
attempts that have been made to prove the H-theorem using Newton's laws, and point out
the difficulties involved in these attempts before considering alternative means.
86 CHAPTER IV

If the gas is bounded (e,g., kept in a box) there are only a finite number of stars.
Thus, the evolution of the ~as may be modeled by a finite Markov chain, theX(n) takevalues
in a finite set {el, ez, .. ., en f' The transition probabilities,

pjk (n, m) = Pr {X(n) = ek I X(m) = ed, (B1.4)

arecalledstationaryifpjk(n+m, m+m) = pjk(n, m)forallm,n,m.Thatis,theoriginchosen

for time is immaterial, andpjk(n, m) is a function only of n - m. Stationarity is the analogue of
equilibrium, and itis a mathematicalfact9 that any Markov chain wiIlapproach a stationary state.
Ergodicity now means that each state, ej, is accessIble from any other state, ej, pjj(n) >0, for
some n, and so each state is visited some time. In the stationary case, the theory is particularly
simple since pjk(n) =pjk(n+m, m) is the jlcth element of the matrix [Pjk(l)]" by the
Chapman-Kolmogorov (Fokker-Planck) equation.
If Q is the volume of a star, one defines a coarse-grained distribution function

P = -b f Q
a dqdp (B1.5)

and an associated coarse-grained H function

H = f PlogPdqdp. (B1.6)

This coarse-grained entropy increases, except in equilibrium, but the details of the
approach to equilibrium are missing.
5. Entropy and information: See text.

3. 1 The Ehrenfest model

To illustrate the operation and the proof of the H-theorem, take a full pack of cards.
(The original Ehrenfest model involved two urns and balls numbered 1 to N, which
were moved at random from one urn to another.) Divide it into two packs, one consisting
of all red cards, and the other consisting of all black cards. Move the cards according to
the following rule. Take the top card from pack 1 and put it on top of pack 2. Take the
bottom card from pack 2 and put it at the bottom of pack 1. Shuffle the packs separately,
and again interchange according to the above rule. Continue this process.
The number of, say, red cards in one of the packs gives the (macroscopic) state of
the system. The rule for interchanging cards corresponds to the evolutionary (dynami-
cal) law. The process of shuffling corresponds to the 'ergodic hypothesis', or the
property of 'mixing' or 'metric transitivity'. (There are subtle differences between these
concepts which we need not consider here; see Boxes 2 and 3.) The whole system is a
model for a finite Markov chain (Box 1).
 THERMODYNAMIC TIME 87

The initial state of the system, with all red cards in one pack and black cards in the
other, is an ordered or low entropy state, like that of a gas in a box with all molecules
in one half of the box. If we continue shuffling and interchanging the cards, after a short
time there will be about an equal number (= 13) of red and black cards in each pack:
the gas has expanded into vacuum. This is the most disordered state or the state of
maximum entropy or thermodynamic equilibrium.
The moral of the story is that the system, starting from a low entropy state, will
progress irretrievably towards thermodynamic equilibrium.

4 The reversibility and recurrence paradoxes

4. 1 Loschmidt's paradox

Loschmidt objected to Boltzmann's proof of the H-theorem, pointing out that the laws
of mechanics were reversible, so no proof of the H-theorem could depend on the laws
of mechanics alone. When told about Loschmidt's objection that molecular motions
were reversible, Boltzmann is rumoured to have remarked 'Go ahead, reverse them'.
In the above analogy, the laws of mechanics correspond to the rule for interchang-
ing the cards. If the proof of the H -theorem depended only on this rule, then the entropy
of the system would increase whether or not the cards were shuffled each time.
If the cards are not shuffled after each interchange, then (by interchanging the two
packs) one can get back to the initial state, starting from a given state, by the same rule
for interchanging cards. If the increase in entropy depended only on this rule, then the
entropy must increase in both cases. So the entropy of any state must be the same as
th~ entropy of any other state, and entropy must remain constant.
One can arrive at the same conclusion in another way. A universe evolving accord-
ing to the laws of mechanics is time-symmetrically deterministic in the following sense.
The future of the universe is completely determined by its past and vice versa. In such
a deterministic universe, the future is uncertain only to the extent that the past is
uncertain: the process of evolution does not generate any uncertainty.
To summarize Loschmidt's objection, in a universe evolving time-symmetrically
and deterministically, according to the laws of classical mechanics say, the future cannot
be more uncertain than the past. Some shuffling mechanism is necessary. But who or
what is responsible for shuffling the cosmos? and how does one know whether, in fact,
the cosmos is being shuffled?

4. 2 Poincare recurrence theorem

Zermelo raised another objection which applies to any proof of the H-theorem. This
objection is based on the Poincare recurrence theorem. To illustrate the recurrence
theorem, consider first the case where the packs are not shuffled. If we go on inter-
changing the cards between packs, ultimately we will come back to the state where all
the red cards are back in the first pack, and all the black cards in the second pack. The
initial state has recurred.
88 CHAPTER IV

If we continue interchanging, other states will recur as well - the evolution of the
system is completely cyclic. Perhaps this is only because of the particular rule that we
adopted. The Poincare recurrence theorem states that this kind of quasi-periodic
behaviour must occur regardless ofthe particular rule used to interchange the cards. More
specifically, the theorem asserts.
Theorem. Let (X, p, Tt ) be an abstract dynamical system, i.e., (X, p) is a probability
space, and Tt is a group of automorphisms of (X, p), so that each Tt is one-to-one and
preserves measure. Let A be any subset of X Then for almost every x E A, there exist
arbitrarily large (hence infinitely many) t for which Tt x EA.
Proof. Let B = {x E A, Tt x f/. A, V t ;?: to }. Relabel to as 1 for convenience. The sets
B, TIB, ... , TnB, ... are all disjoint, andp(TnB) = pCB), since each Tn preserves measure.
Hence,

so thatp(B) = O.

4. 2. 1 The recurrence theorem simplified. One could restate the theorem and proof as
follows. Suppose we do not quite know the state of a system, we may liken the evolution
of the system to a flow of points in the manner of a fluid. The trajectory of a point initially
at x(O) is being described by the rule x (t) = Ttx (0). Suppose the flow of this fluid
preserves volume, i.e., the fluid behaves as if it were incompressible. Then, for any finite
volume regionA, the trajectory starting fromA must return toA after arbitrarily large
times, hence infinitely often.
The proof depends upon the assumptions that (a) the flow preserves volume and
(b) the total volume of the available state space is finite.
Thus, letA be any region, and let B be that subregion ofA , starting from which the
trajectory never returns toA after 1 s, say. If the fluid initially in B occupies the region
BI after 1 s, then Band BI cannot overlap by definition of B. Similarly, if the fluid is in
B2 after 2 s, then B, BI, and B2 cannot overlap (else BI and B would overlap). But the
flow preserves volume, so that the volume of B must be the same as that of BI, B2, ....
SinceB, BI, B2, ... do not overlap, their volumes may be added. But there is only a finite
volume available, which will sooner or later be exhausted unless B has zero volume.
So, ifA has non-zero volume, all that volume must be occupied by states that return
toA after 1 s. The same argument goes through for to s in place of 1 s. Since to could be
chosen arbitrarily large, the trajectory starting from almost every point ofA will return
toA after arbitrarily large times, hence infinitely often. Since the volume of A could be
chosen as small as we please (so long as it is non-zero), the trajectory must return
arbitrarily close to the initial point infinitely often.

4. 2. 2 Consequences of the recurrence theorem. The consequences of this theorem are

rather dramatic. Suppose, for example, a gas is kept in one half of a box by means of a
membrane, and the membrane is broken. We would expect the gas to expand and fill
the entire box. The theorem predicts that the gas will spontaneously return to the half
 THERMODYNAMIC TIME 89

of the box that it originally occupied, creating a vacuum in the other half. Moreover,
this sort of thing will keep happening: history repeats itself.

4.2.3 Relationship to mechanics: Liouville's theorem. Before exploring this theorem

further, let us settle one question. What does an abstract dynamical system have to do
with Newton's laws of motion or Einstein's equations? The answer is provided by
Liouville's theorem. The classical form of this theorem is that evolution in classical
phase space, according to classical mechanics, preserves volume. We state this in a more
general form, of which many physicists seem unaware, so that it also applies to, for
example, the relativistic case of geodesic flow on a manifold.
Theorem. Let X be a compact Hausdorff space, and Tt a group of homeomorphisms on
X Then there exists a regular Borel probability measure fl on X which is invariant under
Tt, so that Tt is a group of automorphisms of (X; fl).
We postpone the proof to the appendix, and explain here only the hypotheses of
the theorem in the context of Newton's laws. For a point mass, specification of3 position
coordinates and 3 momenta completely specifies the dynamical state. Thus, the dynam-
ical state of a gas is specified by specifying a point in the 6N-dimensional Euclidean
space R6N. (If the gas molecules cannot be treated as point masses, one would need
more degrees of freedom.)
If the gas is confined to a finite volume, and energy is conserved, then the state must
lie in a closed and bounded region of R6N, which is compact by the Heine-Borel theorem.
This provides the compact X of the theorem. The group Tt specifies the evolutionary
law. To say that each Tt is a homeomorphism means, first of all, that every state has a
unique precursor, t seconds in the past, and a unique successor t seconds in the future.
This requirement is met in classical mechanics, and is usually stated in the following
form: 'Through every point of phase space passes a trajectory, and no two trajectories
may cross each other.' (Thus, this requirement excludes singularities of the Hawking-
Penrose type encountered in relativistic evolution. 2) If the evolutionary law is specified,
like Newton's law or Hamilton's equations, by an ordinary differential equation, this
property follows from the existence and uniqueness theorems for ordinary differential
equations.
The homeomorphism requirement implies continuity, so that two states that are
sufficiently close to each other wilVdid stay close to each other afterlbefore t seconds.
Thus, two rockets fired into space, at a slight angle to each other may stray parsecs apart
after a sufficient lapse of time. But we could, in principle, arrange matters so that after
1000 years the rockets are not more than 100 km apart. In the case of classical mechanics
this last requirement follows from the theorem for ordinary differential equations which
ensures continuous dependence of the solution on initial values.
To summarize, in classical mechanics, Liouville's theorem, and hence the Poincare
recurrence theorem, apply provided only that (i) the system is bounded and (ii) has a
finite number of degrees of freedom.

4. 3 The recurrence paradox

Poincare, and later Zermelo, objected to the H-'theorem'. Since every state of the system
recurs, the entropy, rather than progressing irreversibly towards a maximum, must behave
almost periodically. The observed entropy increase must be a local matter. One cannot
90 CHAPTER IV

Box 2. Mixing and sorling: Maxwell's demon

1. Abstract dynamical system

An abstract dynamical ~m (X, 1"t) consists of a compact topological spacex, and a group of
homeomorphisms 1"t onK Alternatively, by Uouville's theorem, an abstract dynamical !'Stem
(X, It, 1"t) is a group of automorphisms (measure preseIVing, bijective maps) on a probability
space (X, It). In classical mechanics, the homeomorphisms T, are obtained ona subset X of the
phase space R2n, from Hamilton's equations.
2 Shuffling and e'Eodicity
One understands what is meant by shuftling a pack of cards. But what does shuJlling mean for
a gas in a box? what does it mean for an abstract dynamical !'Stem?
If the cards are shuffled properly, one would expect every possible combination to be
realized. Since each combination is a 'state', there is a finite probabilily of going from a
given state to any other state in a finite number of steps: every state is accessible from any
other state, so that each state is visited sometime. This is the definition of ergodicily for a
finite Markov chain, and possibly the notion that Boltzmann, the atomist, had in mind.
For a continuous !'Stem, the classical ergodic hypothesis took the form that the trajectory
visits every point in phase space. A plausible consequence of this hypothesis is that space and
time averages coincide. But as stated, the hypothesis is false, and was soon shown to be so.
Dimension is a topological invariant, the Tc's are homeomorphisms (and hence preseIVe
topological invariants), the trajectory is one-dimensional, and phase spare is multi-dimensional.
3. Quasi-e'Eodicity and metric tnlnSitivity
However, what was actually required was the interchange of space and time averages, and this
could be achieved by the quasi-ergodic hypothesis: the trajectory comes arbitrarily close to any
point in pha<;e space.
A more convenient hypothesis, which implies quasi-ergodicity, is that of metric transitivity or
indecomposability: Tc (A) = A implieslt (A) = 0 or It (A) = 1. That is, invariant sets or
their complements must have probability 1, SO that the trajectory does not remain confined to any
part of phase space. There is, once again, an analogue for a finite Markov chain (with stationary
transition probabilities): the state space may be decomposed into recurrence dasses. and the
Markov chain is ergodic if there is precisely one recurrence class. I (continued on D 91J
get around this argument by questioning the definition of entropy, for the same
argument shows that it is impossible to define any persistently monotonic function of
the state. (One can, however, try to argue that the entropy need not be defined as a
function of the state.)

4. 4 Refutation of the paradoxes

The reversibilily and recurrence paradoxes are refuted by postulating a shuffling process,
and using this to prove the H-theorem, as in the Ehrenfest model. This immediately refutes
Loschrnidt's paradox. The H-theorem is a consequence of shuffling, and not merely the
 THERMODYNAMIC TIME 91

4. Mixing
A stronger hypothesis than metric transitivity is 'mixing'. In Gibbs' example, if ink is added to
Welter and stirred, one gets a uniformly indigo mixture. After stirring for some time, the ink has
quite 'forgotten' its original position. The hypothesis takes the form

lim # (TcA nB) =# (A) # (B). (B21)

(-+00

Whatever the regions A and B, after long enough evolution, A is transformed into a region
statistically independent of B.
5. Sorting out Manvell:S denwn
The opposite sort of example of sorting is provided by Maxwell's demon. Consider a gas in a
partitioned box. Maxwell's demon stands at a small aperture in the partition. He stops slow
molerules, and allows fast molecules to pass. After some time, the gas has been sorted: faster
molecules are in one half and slower molecules are in the other. Now one can connect the two
partitions from outside, and allow heat to flow and extract energy from the process.
The demon violates the second law of thermodynamics: it progressively takes the gas from
equilibrium to a non-equilibrium state by sorting.
The demon has been exorcised by Brillouin10: in the process of distinguishing the fast
molecules from the slow, the demon would create more entropy than it destroys.
Brillouin's exorcism of Maxwell's demon, however, misses the crucial point The demon
need not be taken literally as a being who must make measurements. What if sorting processes
were to occur spontaneously but regularly? and what if the demon were able to amplify such
processes (not necessarily in a predictable Wely)? Such ~bilities will be considered later on.
From this point of view, what one has to try to understand is why 'mixing' rather than 'sorting'
is commonly found in nature.
laws of mechanics. Shuffling preserves local time symmetry: each individual inter-
change is reversible, but probabilistic and intermediate time asymmetry is evident.
The process of shuffling can be used to deny the hypothesis of the recurrence
theorem: by asserting that evolution takes place stochastically, and not deterministically
according to Hamilton's equations. However, the conclusion of that theorem remains
valid for the case of Markovian evolution, provided that there are only a finite number
of distinct (macroscopic) states. If macroscopic states are obtained by lumping together
or discretizing the microscopic states, then for a bounded system one would expect to get
only a finite number of distinct macroscopic states.
Thus, shuffling does not affect the recurrence theorem qualitatively; but it does bring
about a quantitative difference. Suppose it takes one second to shuffle the packs and
interchange the cards. (Instead of shuffling and interchanging manually, we may programme
a computer to do the job, in which case one second is not a bad estimate!) In the absence
of shuffling, a given state would recur in 52 s, or a little under a minute. If shuffling is allowed,
a given state would recur, on an average, after 226 s, or more than 10 years. So, shuffling
increases the recurrence time dramatically. In a similar manner, Boltzmann calculated3 that
the recurrence time for 1 C.c. of gas would be of the order of 101019 s, if each of the 1018
92 CHAPrnRIV

molecules were to return to within 10-7 cm of its original position, and to within 102
cm/s of its original velocity.
There is yet another way of denying the consequences of the recurrence theorem, using
the notion of coarse graining(Box 1). Individual systems may recur, but the ensemble as a
whole displays an asymmetry in time. One can argue that it is only this asymmetry which is
relevant to statistical mechanics.

4. 5 Objections to the refutations

Physics is not a science; it is an art.
A.A. (a well-known physicist, off record)

As Lewis Carroll put it in The Hunting ofthe Snark, 'What I tell you three times is true.' If
this is the criterion for the truth, then we should accept the above refutations of the
paradoxes, for these refutations are found in every text book, and even in most critical
expositions of the subject. For example, the book by Davies4 repeats the assertion that
Poincare recurrence is very rare some n times.
The above refutation is, however, questionable onmathematica1, physical and statistical
grounds.
4. 5. 1 Is the cosmos shuffled? The Poincare recurrence time has been calculated only
under two broad assumptions (i) the assumption of quasi-ergodicity or metric transitivity,
and (ii) the assumption, as in the Ehrenfest mode~ that evolution may be represented by
an ergodic Markov chain. In the first case, the recurrence time, for a trajectory to return to
a setA, turns out to be5 proportional to lI,u(A), ,u being the invariant measure. In the second
case, the recurrence time of a given state is proportional to the inverse of the (stationary)
probability that the system is in that state. I know of no other way to derive the recurrence
time in classical statistical mechanics, avoiding this or very similar assumptions. So, for the
recurrence time to be large, the cosmos must be shuffled.
The other part of the story is that everyone agrees that it is almost impossible to decide
whether or not a given realistic physical system is metrically transitive. A century of effort
in this direction has proved fruitless. So, no one knows whether the cosmos is shuffled.
Summer follows winter every year, but if the ergodic hypothesis were valid for the solar
system it would be a long time indeed before the next summer.
To put together both parts of the story, the claim that the mean recurrence time
for realistic systems is astronomically larger than the astronomical is based on an
assumption whose applicability to realistic systems is, by common consent, impossible
to decide. Therefore, the claim, repeated ad nauseum, is a matter of faith.
4. 5. 2 The meaning(lessness) of large recurrence times. The second argument against
very long recurrence times is physical. The question is whether large recurrence times
are physically meaningful. The picture of a gas which progresses quickly towards
thermodynamic equilibrium and spends most of its time there may be legitimate for
statistical mechanics. But a difficulty arises if this picture is applied to the entire
universe, as it must be applied if the entropy law is taken as the basis of the direction of
time. When the cosmos has reached a state of thermodynamic equilibrium, how will
one distinguish between a long time and a short time? When the entire cosmos is in a
state of equilibrium, no measurement, hence no time measurement, is possible, in
 THERMODYNAMIC TIME 93

Box 3. Elements of ergodic theory

1. Space and time averages
Consider a system with n degrees of freedom, described by n generalized coordi-
nates q, and the canonical momenta p. Let H denote the Hamiltonian of the system,
and let the motion of the system be governed by Hamilton's equations

. aH aH (B3.1)
qi = api = [qj,H], Pi = -a qi = (Pi,H],

where [ ] denotes the Poisson bracket.

Ifxo = (ql(O), ... , qn(O);Pl(O), ... ,Pn(O)) denotes the initial point in 2n dimen-
sional phase space, X ~ Rn, and Xt = (ql(t), ..., q,,(t);Pl(t), ...,p,,(t)) denotes the
representative point at time t, then the transformations Tr. defined by

Ttxo = Xt, (B3.2)

form a one parameter group of homeomorphisms of X, for reasons explained in the

text.
The ergodic hypothesis needs to prove that the time average,

rIfT
un -T I(Ttx) d.x =rex),
(B3.3)
0
( .... 00

exists for everyI in an appropriate class, and that it equals the phase space average,
calculated with respect to the conserved measurep. onX

r (x) = J= f x I(x) d#(x). (B3.4)

It is often enougb to consider, instead of (B3.3), the discrete time average,

r
00
(B3.5)
(x) = Nlim 1. '" I (T k x),
.... oo N L.J
k=O

where T == n. I
(continued on p 94)

principle. For a gas in a box, there is a time outside the box with respect to which the
gas spends a large time in thermodynamic equilibrium. But it is meaningless to say that
the cosmos will spend a long time or a short time in thermodynamic equilibrium.
94 CHAPfERIV

2. Invariant functions
The time average r is an invariant function:

r (Tx) =r (x) a.e., (B3.6)

where a.e. stands for 'almost everywhere with respect to /-t'. Constant functions are
other examples of invariant functions. If the only invariant functions are constants
a.e., then T must be metrically transitive, as can be seen by applying this to the
indicator functions XA of any measurable set A. Conversely, if there is an invariant
function.,[, which is not constant, metric transitivity must fail, for we need only consider
the set A = {x,f(x) > a}, and its complement, where a may be chosen so that
neither /-t(A) nor .u(Ae) equals 1, since[is not constant a.e. In terms of tbe Koopmans
operator, U, defined by

(Uf) (x)=[(Tx) (B3.7)

the requirement is that 1 must be a simple eigenvalue of U.

3. Ergodic theorems

There are a number of such theorems, of which we mention only two. The individual
ergodic theorem says that if [E Lt(;t), thenr (x) exists a.e., r
(x) E Lt(;t), and

(B3.8)

That is, the time average of an integrable function exists almost surely, and is an
unbiased estimator of the space average.
The operator U, defined by (B3.7), is a unitary operator on the Hilbert space
L2(;t). The mean ergodic theorem says that, if [ E L2,

(B3.9)
L
N~l

~ Un [ -+ P[ (strong convergence),
n=O

where P is an orthogonal projection on an invariant subspace of U. If only constants

are left invariant,

P[ = <[,1> = f [dp" (B3.1O)

by the Riesz representation theorem, <',' > denoting inner product in L2. The
conclusion is that, for a metrically transitive system, the variance of the time average
may be treated as insignificant. I ron D 95)
 THERMODYNAMIC TIME 95

4. Mixing

Mixing implies metric transitivity, for if A is an invariant set, and if B =AC, then
A nB = 0,sothat

lim .u(TtA nB) = .u(A nB) 0, (B3.H)

1.... 00

which implies that either.u (A) = 0 or.u (B) =.u CAC) = O.

In terms of the unitary operator U on L2 , the requirement is that 1 be the only
eigenvalue. It may be shown that mixing is equivalent to the requirement that

lim <Ut/,g> = <f, 1> <l,g>. (B3.12)

t-+oo

There are many related notions of 'mixing', and for full confusion we refer to the
literature.l l

Furthermore, as pointed out by Kac,5 the variance of the mean recurrence time is
equal to the mean, so that the entropy increase of the cosmos must be a very irregular
process, and fluctuations must occur. If a fluctuation does occur, will we count time as
running forward or backward?
To summarize, not only is the belief in large recurrence times a matter of faith, but
this faith is in something physically quite meaningless, when applied to the cosmos, for
it assumes some notion of time external to the cosmos. Boltzmann's original approach
to this problem (Box 5, p 101) appears to have been more realistic.
There are I ' f statistical problems. According to the ensemble approach, even if
one were to ob :! a gas moving back into half of the box, creating a vacuum in the
other half, one would maintain that the entropy had increased. This approach makes
no provision for the possibility that fresh observations of the single system (cosmos)
under cORsideration may show that the entropy had in fact decreased. At least in
classical mechanics, the: ' 'e ase in uncertainty pertains only to a certain mental model
of the system that has b(;~ . adopted, and this mental model makes no provision for
phenomena that may occu ,,~d be manifestly observable, in the single cosmos open to
observation.
Finally, it somehow ten, . J be assumed that the cosmos will come to an end long
before recurrence occurs. What if it does not?
To conclude, the standard ways of resolving the paradoxes are not good enough.
96 CHAl'TERIV

5 Resolution of the paradoxes

Is there any other way to resolve these paradoxes? Let us review the hypotheses that
led to the paradoxes, and systematically enumerate all the ways in which these hypoth-
eses can be evaded. The basic assumptions underlying the paradoxes and the above
theorems are the following:
1. time symmetry (the assumption that Tt form a group),
2. deterministic evolution (the assumption that each Tt is a homeomorphism),
3. compact phase space, e.g.,
3a. finite number of degrees of freedom, and
3b. boundedness.
Statistical models, such as the Ehrenfest model, give up deterministic evolution,
and may be used to show how the stochastic evolution of, say, a finite Markov chain may
be regarded as macro-irreversible, while preserving micro-reversibility. Moreover, the
ergodic hypothesis ensures that the recurrence times are very large. The problems with
this approach have been explained above: one does not know whether the cosmos obeys
the ergodic hypothesis, and one does not know what physical meaning to assign to large
recurrence times for the cosmos.
In general relativity, it is possible also to give up deterministic evolution by
postulating the existence of singularities of the Hawking-Penrose type, as in Tipler's
no-return theorem. 2 (I use the word 'postulate' because, contrary to common belief,
theorems of the Hawking-Penrose type do not actually prove the existence of singular-
ities. One could, for example, interpret the theorems to imply a breakdown in the
smoothness of the metric tensor within dense matter. Hawking and Ellis' arguments
against such an interpretation are weak; but most physicists believe them. See, further,
Chapter VII.) This approach, however, only rules out exact or eternal recurrence,
leaving open the possibility of quasi-recurrence for a long time to come.
One may give up boundedness by arguing that the universe is open, though only at
the risk of eventual recurrence. Finally, one may argue that taking into account quantum
mechanics, or even fields, destroys the finiteness of the number of degrees of freedom.
Leaving aside the question of quantum mechanics (which will be considered later), I
regard the introduction of the field as the strongest of the above arguments, since it
introduces nothing ad hoc. It is a point of view certainly worth exploring. One consequence
of accepting this point of view is that the thermodynamic time asymmetry cannot be
regarded as basic: it must be regarded as derived from some other type of time asymmetry.
Granting this, how does the introduction of fields bring about time asymmetry?

6 Conclusions

Entropy increase towards the future coincides with the intuitive belief that recall of the
past is more reliable than expectation of the future. However, thermodynamic time
asymmetry cannot be regarded as a consequence of Newton's laws and some statistical
postulates, due to defects in the current refutation of the reversibility and recurrence
paradoxes: the calculation of recurrence time assumes ergodicity, and no one knows
 THERMODYNAMIC TIME 97

whether the evolution of the cosmos is ergodic. Moreover, large recurrence times are
physically meaningless. One can hope to get rid of these paradoxes by taking into
account the electromagnetic field, though this means that the thermodynamic arrow of
time cannot be regarded as fundamental.

Appendix: Proof of Liouville's theorem

I do not recall any reference which gives the proof, hence the proof is given below.

(a) LetX be a compact Hausdorff space, and Tt : X -+ X a group of homeomorphisms.

Let C(X) denote the space of continuous functions on X. Define Ut : C(X) -+ C(X) by
(Utf)(x) =[(Ttx). Then Ut is a group of isometries on C(X). Define the group of
adjoints ut on the dual space M(X) of all regular, complex, Borel measures on X by
<Ut/,p> = <I, Ut*P>, fortE C(X), for allp EM(X), where

<f,p> = f x [dp. (A. 1)

A fixed point for Ut * is an invariant measure for Tt, since Ut *p = p 0 Tt -1 by the change
of variables formula.

(b) Since Ut* is a one parameter group, it is a commuting set of operators. If P is the set
of regular Borel probability measures onX, then, since each Ut * is a positive isometry,
P is invariant under ut. P is trivially convex, so by the Markov-Kakutani fixed point
theorem (Dunford and Schwartz,6 p 456), the group ut has a common fixed point in
P just as soon as we show that P is compact with respect to a vector topology on M(X)
in which Ut * is continuous.

(c) If B is the unit ball in M (X), then B is weak* compact by the Banach-Alaogulu
theorem. Each Ut * is trivially weak* continuous. Since P is a subset of B, it suffices to
prove that P is weak* closed.

(d) Since X is compact, P is weak* sequentially closed. Thus if pn -+ p weakly, and

[(x) = 1 for all x EX, then <f,Pn> = Pn(X) -+ p(X), which implies that p(X) = 1.
Moreover, since X is compact, C(X) is separable, so that B is metrizable (Rudin,7
theorem 3.16) in the weak* topology. Hence, P is metrizable in the weak* topology, so
that it is weak* closed, since it is weak* sequentially closed.
98 CHAPTER IV

Box 4. Rudolph Clausius

Born in Koszalin, Poland, then known as Koslin, Prussia, on 2 Jan 1822, and died in
Bonn, Germany, 24 Aug 1888.
The earlier caloric theory had assumed that (i) the total heat in the universe is
conserved and (ii) the heat present in any substance is a function of the state.
Clausius denied both premises, based on experiments of Joule and others. He
asserted that whenever work is produced by heat, a quantity of heat equivalent to
that amount of work is consumed, and cannot any longer be considered a function
of the state. He argued that there was only free heat, and understood this free heat
as vis viva or kinetic energy.
This tended to affect Carnot's theorem, which required that all ideal engines
must produce the same amount of work when operating between the same temper-
atures, with the same amount of beat. The proof was based on the principle that work
cannot be produced from nothing. With Clausius' re-interpretation, however, tbe
work produced was accompanied by the disappearance of beat. Therefore, Clausius
used, instead, the principle that heat cannot be transferred by itselJfrom a colder to
a hotter body. He called this new principle the theorem of the equivalence of
transformations - since this established the equivalence of two types of transfor-
mations: (i) a conversion of heat into work at one temperature, and (ii) a transfer of
heat from a higher to a lower temperature.
He assigned positive transformation values to these processes, and equal nega-
tive transformation values to their opposites. Hence the word entropy from the
Greek tropae (lit. 'turning') or transformation, to denote the 'transformation
content' of a system in line with energy as work-contents, in German. Hence, also,
Clausius was led to irreversible processes - a negative transformation value would
correspond to a flow of heat from a lower to a higher temperature, contrary to the
normal behaviour of heat, so that the sum of transformation values in any (engine)
cycle could only be zero (reversible) or positive (irreversible).
He encapsulated the two ideas in a famous couplet for the two laws of thermo-
dynamics: 'The energy of the world is constant; the entropy strives to a maximum.'
Clausius had used the term 'entropy' as a measure of the unavailability of
thermal energy for conversion into mechanical work. Some confusio~ was created
over the meaning of the term entropy by the usage of P.G. Tait and Maxwell, with
Tait asserting in 1868 [Sketch Thermodynamics, p 29], 'We shalL .. use the excellent
term Entropy in the opposite sense to that in which Clausius has employed it - viz.
so that the entropy of the Universe tends to zero.' Maxwell, in his 4th edition of The
Theory ofHeat (1875, p 189, note) stated, 'In former editions of this book the meaning
of the term Entropy as introduced by Clausius was erroneously stated to be tbat part
of the energy which cannot be converted into work. The book then proceeded to use
the term as equivalent to the available energy ... '
I(continuedonp 99)
 THERMODYNAMIC TIME 99

Clausius had a major clash with P.G. Tait of Edinburgh over this principle, and
because he got involved in the dispute over the claims of Joule and Julius Mayer to
the discovery of the equivalence of work and heat, by asserting Mayer's priority.
Subsequently, he was involved in other such controversies with Tait, including one
in which he asserted that Tait was claiming more than his rightful share of the theory
of heat.
On the technical side, he staunchly defended his principle, driving Tait to argue,
in desperation, that Maxwell's demon could contradict Clausius' principle. Clausius
retorted that his principle was only concerned with the behaviour of heat by itself,
and not with the aid of demons.

Box 5. Ludwig Boltzmann

Born in Vienna, Austria, 20 Feb 1844, and died at Duino, near Trieste, 5 Sep 1906.
He studied with Josef Stefan at the University of Vienna. where he later held a
professorship. He married in 1876, and had four children.
His first major work was on the Maxwell-Boltzmann velocity distribution. The
velocity distribution was initially derived by Maxwell assuming (i) spatial uniformity,
and (ii) statistical independence of the three components of velocity. The second
assumption being suspect, Maxwell supposed instead (a) that the velocities of
colliding molecules were statistically independent before collision, (b) conservation
of energy, and (c) micro-reversibility of the collision cross-section. (These assump-
tions are reflected in Boltzmann's subsequent proof of the H-theorem.) Boltzmann
extended Maxwell's analysis to the case where one of the molecules was acted upon
by a force with potential V. He then assumed equipartition (uniform distribution) of
energy to derive the velocity distribution

(B5.1)

He identified the factor a2 with kT, a universal constant times the temperature: the
constant later came to be known as Boltzmann's constant, and is inscribed on his
tombstone.

I(continued on p 1()())
100 CHAYr'ERIV

Boltzmann was an atomist even in his mathematics. He preferred combinatoric

analysis, followed by a passage to the limit, to the infinitesimal calculus, which last he felt
was not quite rigorous unless supported by such a combinatoric analysis. Consequently,
his equations look similar to those of modern quantum theory. This is not mere
coincidence since Planck and others used some of Boltzmann's techniques.
The Maxwell-Boltzmann distribution applies to a gas in equilibrium, and both
Maxwell and Boltzmann felt that the kinetic theory should be able to show that a gas
will actually tend to equilibrium. Boltzmann spent much effort in this direction. With
Maxwell's assumption of molecular chaos he derived the transport equation

~ = foo J.t+XI [f(s,t)/(x+x'-t;,t) I(X,t)/(x"t)] ( , l:)d.x',1f:

at 0 0 '1~ ''1x+x'-~ - -IX..fX <p X,X ,5 ~,

(B5.2)

where I represent the distribution of energy (hence the square root), and the
principle of micro-reversibility applies to the collision probability function !p. He
proved that in equilibrium ~ = O.
He introduced the function

E = I "" f
0
I(x, t)
(x, t) [log -IX - 1] d.x
(B5.3)

and proved that dEldt< 0 if I: const x ~-xlkT, and dEldt=O otherwise. The proof
uses the mathematical fact that (a-b) log bla is always negative, based on the
convexity of the function -x log x. This resul t later came to be known as Boltzmann's
minimum theorem or the H-theorem.
Boltzmann showed that, in equilibrium, H was the same as the entropy, apart
from a constant factor, using

s =k log W, (B5.4)

where W is the number of possible 'microstates' or 'molecular configurations'

compatible with a given 'macrostate'. If Wi is the number of microstates compatible
with the given macrostate i, then Wi may also be regarded as proportional to the
probability of state i, so that the expected value of the entropy is, upto a constant
factor,

s =k 2: Wi log Wi, (B5.5)

which agrees with (1). I (connnuedonplOI)

 THERMODYNAMIC TIME 101

When the recurrence paradox was brought to his notice, Boltzmann replied that
tbe recurrence theorem was in complete harmony with the H-theorem. He first
argued that thermodynamic fluctuations would occur with a small probability and
would, hence, take very large times.
He suggested a model of the universe in thermodynamic equilibrium, with no
distinctions between forward and backward directions in time. In small regions, such
as individual galaxies, there are fluctuations large enough to include the ordered
states necessary for tbe existence of life. Thus tbe increase of entropy with time is to
be regarded as a tautology, but different part of the universe will have different
directions in time.
Against the reversibility paradox he argued by initial conditions.
It can never be proved from the equations of motion alone, that the minimum function
H must always decrease. It can only be deduced from the laws of probability that if the
initial state is not specially arranged for a certain purpose, but haphazard governs freely,
the probability that H decreases is always greater than that it increases. (L. Boltzmann,
Nature 51, 414, 1895; emphasis mine).

This, perhaps, was the basis of his reported comment 'Every point of the H curve is
a maximum.' Boltzmann was not quite sure about the origin of ergodicity, and
attributed it to collisions of gas molecules with the boundary.
He strongly defended atomism, against which the various paradoxes were taken
seriously as defects. (Mach, for instance, argued as late as 1912 that atoms were not
real.)
In my opinion it would be a great tragedy for science if the theory of gases were
temporarily thrown into oblivion because of a momentary hostile attitude toward it, as
was for example the wave theory because of Newton's authority.

I am conscious of being only an individual struggling weakly against the stream of time.
But it still remains in my power to contribute in such a way that, when the theory of gases
is again revived, not too much will have to be rediscovered .... [Foreword to second vol.
(1898) ofLeclUres on Gas Theory.]

In the event, his suicide in 1906, perhaps in despair over the rejection of his
life-work, was a tragic irony. for it took place just before the existence of atoms came
to be accepted due to the quantum theory.
VA THE ELECTROMAGNETIC FIELD

ABSTRACI'. We concluded the previous chapter with the hope that the field may provide the way out of
the paradoxes of thermodynamics. We point out that various intuitive ideas underlying field theory, such
as action by 'contact', the idea of 'locality' or a 'chain of causes', or the mechanical model provided by the
field, are respectively unintelligible, illusory, and redundant. Consequently, there is nothing 'spooky' about
action at a distance.

We are left only with the differential equations of field theory, which admit both retarded and advanced
solutions. In the context, the advanced solutions cannot be rejected on grounds of 'causality', and we state
some of the paradoxes of advanced action: the pond, grandfather, and Wheeler-Feynman paradoxes.

We point out the divergences, due to self-action, in field theory. The Abraham-Lorentz model of the
extended electron failed, while the action-at-a-distance approach is equivalent to time-symmetric fields.
Even with Dirac's approach, using the equivalence of advanced and retarded representations, the electron'
preaccelerates. Thus, consideration of interactions propagating into the past seems forced, and, prima
facie, the introduction of the field seems to create more problems than it solves.

1 Introduction

I past
N the previous chapter, we examined the problem of a physical asymmetry between
and future; we examined whether the entropy law could be used to define a
direction of time. We found that the standard ways of resolving the recurrence and
reversibility paradox were unsatisfactory, and reduced to faith in meaninglessly large
mean recurrence times for the cosmos. We concluded with the hope that taking account
of the infinite number of degrees of freedom of the field may provide a way out.
Before proceeding further, let us anticipate some of the problems that might arise
with this approach. Every physics student has been told, at some time or the other, that
the field is not real. If this is correct, and the field is a mental construct, one can hardly
hope to establish a physical asymmetry between past and future using the field. So what
exactly is the field? Is it real? Or is it an imagined entity, a mental construct like force
or the aether?

2 Is the field necessary?

At one level, the field seems not only real, but desirable. The motion of a given body,
like the earth, is determined by the positions of other bodies, like the sun and the moon.
But how does the earth know where the sun and the moon are? Indeed, how does it
know about the positions of all the bodies in the universe? How does an electron know
the positions and velocities of all other charged particles in the universe?
 THE ELECfROMAGNETIC FIELD 103

2.1 Force at a distance

The Newtonian concept of force provided one sort of solution. Starting from the idea
that motion with uniform velocity needed no explanation, it sought to explain deviations
by force! Forces could be exerted between bodies in contact or at a distance, regardless
of intervening bodies. Newton's position was pragmatic rather than philosophical: he
conceded that action at a distance might be explained in terms of impulses or some
other means, but he thought that phenomena certainly compelled the conclusion that
certain bodies behave as if there are attractions and repulsions between them.
This picture became increasingly complicated with electrostatic and magnetostatic
forces being modified by the medium, and electrodynamics leading to curved lines of
force, and velocity-dependent forces. Ultimately, as we saw in Chapters II and III, the
picture had to be abandoned because it was hazy with regard to time. To resolve the
problem of time-measurement, it became necessary to postulate the constancy of the
speed of light, and adopt a different concept of simultaneity. It came to be believed that
forces certainly could not operate instantaneously at a distance.

2. 2 Action by contact

But even in its heyday, the picture of force at a distance could not answer the question:
how do these forces operate at a distance? To get a feel for what is really involved here,
we recall that Newton's predecessor, Francis Bacon, discussed, in all seriousness, cases
of influence at a distance such as the effect of 'the motion of the moon on the spirits of
lunacy'. As Hessel (p 95) suggests:
The phenomena which he [was] most ready to ascribe to action at a distance without any
material medium [were] those which savour most of witchcraft, magic, astrology, and
telepathy, and since these were the beliefs most discredited by the subsequent advance of
physical science, the fact that action at a distance was discredited with them is not surprising.

It is this kind of thing which seems to account for the terminology of 'spooky action at
a distance' in the current debate on Bell's inequalities, ever since Einstein, Podolsky
and Rosen.
The other point of view that action requires 'contact' has an immediate intuitive
appeal, and an old history, dating back to the Vaisesika Sutra and Aristotle. As Leibniz2
asserted:
A body is never moved naturally, except by another body which touches it and pushes it;
after that it continues until it is prevented by another body which touches it. Any other kind
of operation on bodies is either miraculous or imaginary.
An idea closely related to 'contact' is that of a 'chain of causes'. As Hume put it,3
Though distant objects may sometimes seem productive of each other, they are commonly
found upon examination to be linked by a chain of causes which are contiguous among
themselves, and to the distant object; and when in any particular instance we cannot discover
this connection, we still presume it to exist.
Descartes' aether was based on similar ideas and we saw earlier the fate that it suffered.
Nevertheless, very similar ideas are presumably responsible for the current terminology
of 'spooky action at a distance'.
104 CHAPTER VA

2. 3 The field

The field seems to incorporate the underlying ideas of contact and chains of causes,
without the relative disadvantages of the aether and forces at a distance. Unlike the
aether, the field is relativistically covariant. But like the aether, it operates on the
'principle of contact'. An electron 'knows' which way to move because it is in 'contact'
with the field in its immediate neighbourhood; and this field incorporates information
about the position and velocities of all the particles in the universe. As a pleasant bonus,
the field provides a mechanical model: one may model a wave propagating in a field by
an infinity of coupled harmonic oscillators, each oscillator is set into motion and disturbs
the next. This model also shows how the field incorporates the 'chain of causes' idea.

On the other hand, let us recall Poincare's arguments4 (in Chapter IIIB) that such
mechanical models are redundant.

The differential equations are always true, they may always be integrated by the same
methods .... It cannot be said that this is reducing physical theories to simple practical recipes;
these equations express relations, and if the equations remain true, it is because the relations
preserve their reality ... only that something that we then called motion, we now call electric
current. But these are merely names of images.... The true relations between ... real objects
are the only reality we can attain.... If the relations are known to us, what does it matter if
we think it convenient to replace one image by another?

In the above quote we only need to replace some names of images: 'motion' by 'aether'
and 'electric current' by 'field'! So the field concept has some excess baggage.

There are other disadvantages. For, consider the Coulomb field, which is rigidly
attached to a charged particle. This field falls of as lIR , and so is very small at large
distances, R, from the particle. But, in classical mechanics, this field is still measurable,
in principle. So, the movement of the particle can be ascertained, in principle, by
measuring this field. If one is sufficiently far off, information can be transmitted in this
manner, at a speed exceeding that of light.

Mathematically, this has to do with the fact that the Laplace and Poisson equations
are elliptic; they model a stationary or equilibrium situation. As Max Born5 stated:

in this treatment of electrostatics, we are not dealing with a true theory of contiguous action .
... For the differential equations...contain no member that expresses a change in time. Hence
they entail no transmission of electric forces with finite velocity but, in spite of their
differential form, they represent an instantaneous action at a distance.

So, the 'chain of causes' idea is sometimes incorporated in the field in as illusory a
way as it is incorporated in the notion of an incompressible aether, which transmits
actions at a speed exceeding that of light. Worse still, the notion of 'contact' turns out
upon examination to be largely meaningless.
 THE ELECfROMAGNETIC FIELD 105

2.4 'Contact' and the structure of matter

~toms must have pans since they are capable of contact.' 'You misunderstand the nature of
contact. "Contact" resides in the pair of atoms it qualifies; it is not a material property.'
Udyotkara6

It was following Leibniz that the debate on contact became transformed into a debate
on the ultimate structure of matter. 'Contact' was visualized as a collision between
atoms, and this suggested that atoms had parts which came into contact. Leibniz7 argued
against the view that atoms were rigid, indivisible balls of matter: 'there are no elements
of bodies... analysis proceeds to infinity'. VlS viva (kinetic energy) was not lost in contact,
but was indefinitely transferred to ever smaller parts of bodies.
Boscovich used Leibniz's principles of continuity, sufficient reason, and conserva-
tion of vis viva to adopt exactly the opposite point of view. He constructed a tentative
'unified theory' using point particles and a force which was alternately attractive and
repulsive. 'Contact' meant contact between the atmospheres of force grouped around
the particles; the 'atmosphere' extended to infinity. Perhaps the whole controversy
could have been avoided had it been observed, following Udyotkara, that it centred
round a linguistic fallacy.
But it did come to be realized that action by contact was not more intelligible than
action at a distance. People stopped asking for the causes of attraction at a distance.
HumeS commented:
Here is a kind of attraction which in the mental world will be found to have as extraordinary
effects as in the natural.... Nothing is more requisite for a true philosopher than to restrain
the intemperate desire of searching into causes.
In Kant's second antinomy every substance both is and is not made up of simple parts:
both thesis and anti-thesis lead to contradictions. Kant settled for the regulative
hypothesis of infinite divisibility on grounds of economy: at any stage it is conceivable
that matter can be subdivided further.

2. 5 The meaning of 'contact'

To summarize, so far as physics is concerned, action by contact is not more meaningful

than action at a distance. The field incorporates the idea of 'locality', or a chain of causes,
in a deceptive and unsatisfactory way. Moreover, the mechanical model provided by
the field is redundant. Of what use, then, is the field? We will let Poincare9 have the
last word, in the only argument he gave in support of the aether - an argument which
applies perfectly well to the field.
The same idea may be expressed in a more ... abstract form .... in ordinary mechanics, the
state of the system under consideration depends only on the state at the moment immediately
preceding; the system therefore satisfies certain differential equations. On the other hand,
... the state of the material universe would depend not only on the state immediately
preceding, but also on much older states; the system would satisfy equations of finite
differences.
The moral of the story is that, at least classically,10 the field is quite redundant and
if someone proposes to get rid of it, one should not gasp, 'How will the electron know
106 CHAPTER VA

what to do?' Rather, one should enquire about the equations which will replace the
current differential equations. Let us proceed with this inquiry. Are the differential
equations of field theory satisfactory?

3 The electromagnetic arrow of time

Electricity and magnetism were regarded as separate branches of physics till Maxwell
proposed the theory of electromagnetism, unifying the two, and subsuming all earlier
phenomenological laws. The triumph of this field theory was the prediction of electro-
magnetic waves: accelerating charged particles must emit electromagnetic waves. For
the case of a simple harmonic oscillator, the classical Larmor formula, P oc da/dt, shows
that the power radiated, P, is proportional to the rate of change of acceleration. ll
Electromagnetic waves resemble waves on the surface of water and sound waves.
The difference is that electromagnetic waves travel at the speed of light, usually much
larger than the speed of sound. As we have seen, the speed of light must be declared a
constant, in order to arrive at a measure of time. The other difference is that electro-
magnetic waves may travel in a medium, but also in vacuum. They are oscillations of
the 'field', which may exist in vacuum.

3. 1 Retarded and advanced solutions

The similarity is that electromagnetic waves, too, obey the wave equation,

(1)

whereAIt is the 4-potential, andjlt the 4-current. The wave equation, being time-sym-
metric, admits two broad categories of solutions. One such solution, called the retarded
solution, or the retarded potential, is given by

[ AIt(r, t)] ret =

f [" II1
j Cr' t')
r":'r '
ret
dr ' ,
(2a)

where [ ] denotes an evaluation of the bracketed quantities at retarded times

t' =t- IIr-r' 1l1e, c being the speed of light.
The qualitative features of this solution are the following. When one throws a stone
into a pond, one normally observe ripples spreading outwards. The ripples arrive at the
edge of the pond a little while after the stone lands into the pond. There is a delay or
retardation between the cause (stone dropping into the pond) and effect (ripples
arriving at the edge of the pond). A person who laughs two minutes after a joke is called
retarded.
Another type of solution, known as the advanced solution or the advanced potential,
is given by
 THE ELECfROMAGNEI1C FIELD 107

-f
[ AIl(r, t)] adv -
[jp,(r" t')
II r-r' II 1adv
dr'
'
(2b)

where [ ] denotes an evaluation of the bracketed quantities at advanced times

t' =t+ IIr-r' IlIe . For the advanced solutions, the 'effect' precedes the 'cause', just as
it does in the case of a person who laughs in advance of the punch line. In the pond
analogy, advanced solutions correspond to ripples that converge onto a stone and throw
it out of the pond. This sort of thing is not usually observed, unless one has filmed the
movement of the ripples, and plays the film backwards.
Both types of waves, or radiation, or signals may, of course, occur at any frequency.
Retarded radiation travels into the future, and advanced radiation into the past. A
retarded signal from the past, received in the present, carries information about the
past. Similarly, if advanced signals could be observed, they would carry information
about the future.

3. 2 Radiative damping and the arrow of time

When a source emits retarded radiation, it loses energy; if it were to emit advanced
radiation, it would gain energy. This loss of energy results in the motion of the source
being damped, a phenomenon known as radiation damping. In the advanced case, the
source experiences radiative anti-damping. The stone thrown into the pond acts as a
source of retarded radiation. It loses energy, and sinks slowly to the bottom, while the
ripples carry the lost energy outwards. In the other case, the ripples converging onto
the stone provide it with enough energy to jump out of the pond.
A radio transmitter consumes power while broadcasting, on account of radiation
damping. If the transmitter were to broadcast advanced waves, it would generate power
in the process. If this sort of thing were to happen systematically, one could charge a
car battery by installing a radio transmitter in the car.
Obviously enough, this sort of thing does not happen, and the advanced solutions
are rejected for this reason. Only retarded solutions of the wave equation are deemed
physically acceptable. If one charge is set into motion, the radiation from the charge
affects other charges after some time. One can invert this relationship to obtain the
electromagnetic arrow of time: if an electromagnetic signal reaches from event P to
event Q, then event P relates to an earlier time.

3. 3 Mixed potentials

But the wave equation being linear, any convex combination of retarded and advanced
solutions,

A = aAadv + (l-a)Aret, O<a<l, (3)

is also a solution. What if the solution were to be such a mixture, with a tiny element
(ad) of advanced radiation? Because the advanced component is tiny, it would be very
difficult to observe. A radiating source would stilliose energy, but the power required
108 CHAPTER VA

by the source would not quite equal the amount given by the Larmor formula. How can
one be sure that one does not live in the kind of universe where this is the case?

3. 4 The paradoxes of advanced action

One kind of reason that can be offered is 'causality'. Such reasoning will obviously not
do in our context, for we are investigating the very basis of 'causality', the physical
distinction between past and future.
Other kinds of reasons have been offered. It has been suggested that advanced
effects would be paradoxical, leading to a logically impossible situation. We outline
some of these paradoxes below.
(a) The pond paradox: In a series of papers, Popper12 explicitly rejects advanced
solutions on the grounds that any explanation for such phenomena would involve an
infinite regress. Popper asserts that 'an inexplicable conspiracy of causally unrelated
conspirators' would be required to explain the coherence of the generators required to
produce a converging ripple. Again, Popper asserts that a converging ripple may be
'theoretically possible but [is] causally and therefore physically impossible' without the
aid of 'organization from the centre.' He reasons, 'Causes that are not centrally
correlated are causally unrelated, and can cooperate only by accident (or by a miracle).
The probability of such an accident will be zero....'
(b) The grandfather paradox: The grandfather paradox has recently come into promi-
nence because of the related problem13 of closed timelike curves in quantum gravity.
Perturbations of a 'normal' spacetime unavoidably result in a spacetime with closed
timelike curves. In such a spacetime, one can construct and maintain a 'rocketship' or
a 'time machine' which permits one to travel into one's past. The existence of either
closed timelike curves or small amounts of advanced radiation may be generally
categorized as a 'tilt in the arrow of time', the general feature being that interactions
with the past are permitted. Supposing I take the liberty to interact with the past. What
would happen to me if I were to arrange to kill my grandfather when he was a mere
boy? If he dies, then I could not have been born, so that I couldn't have killed him.
(c) The Wheeler-Feynman paradox: A less gruesome, but logically equivalent paradox
was presented by Wheeler and Feynman long ago.14 The paradox apparently eliminates
human intervention, and all questions of human freedom. It concerns two charges which
interact via advanced and retarded signals, leading to a logically impossible situation.
A mechanism shields particle A from a blow at 6 p.m., if and only if it moves at 8 a.m.
Charged particles A and B are located in otherwise charge free space at a distance of 5 light
hours. A clockwork mechanism is set to accelerate A at 6 p.m. Thereby B will be affected,
not only at 11 p.m. via retarded effects, but also at 1 p.m. via advanced forces. This afternoon
motion will cause A to suffer a premonitory movement at 8 a.m. Our dilemma now is this:
Is A hit in the evening or is it not? If it is, then it suffered a premonitory displacement at 8
a.m. which cut off the blow so A is not struck at 6 p.m.! If it is not bumped at 6 p.m. there is
no morning movement to cut off the blow and so in the evening A is jolted.

Something is clearly wrong here. It is one thing to say that the existence of even
small amounts of advanced radiation would be surprising, and quite another to say that
a mathematically valid solution is logically impossible.
 TuE ELECI'ROMAGNETIC FIElD 109

4 The divergences of field theory

For the moment,let us put aside the above question, and ask another question. How
does an electron move? We believe that it moves according to the neat-looking
Heaviside-Lorentz force law: the total force F acting on an electron is given (in the
non-relativistic case) by

F = e (E + v x B), (4)

where e is the charge of the electron, v its (relative) velocity, and E and B denote the
electric and magnetic fields. These are total fields - the sum of the fields of all the
particles in the universe, including the field of the concerned electron (the self-field).
Now, the field is the derivative ofthe potential:

F llv = All, v - A v,ll. (5)

For a charged particle in arbitrary motion the potential may be calculated explicitly,
using the Lienard-Wiechert formula. The final result is, in the non-relativistic limit,1S

E (r, t) = e [<n-v;C1;V2)]
/C 'R ret
+ e[
/C
~'R x{(n-v)xv}] ret , (6)

adv adv

where /C = l-n v, n = RlR, R = r - r' = instantaneous distance. Inspection of

the above formula shows that the self-field of the charged particle diverges at the
position (R -+ 0) of the particle. The electron exerts an infinite force upon itself!
This infinite force, or divergent self-field, may be thought of (not quite rigorously)
as consisting of two components: (a) the near-field which diverges as lIR2 and (b) the
far-field, or radiation field, which diverges as lIR. The divergence of the near-field has
an analogue in the divergence of the gravitational field, and so we will adopt the facile
device of ignoring it. The divergence of the radiation field is something new. What is
to be done about it? One cannot simply ignore it, because this field affects the motion
of the particle, and it is this field, at the position of the particle, which must be used to
calculate radiation damping.

4. 1 The Abraham-Lorentz model

One simple way out is to suggest that there is in fact no divergence because real particles
are not pointlike, but are extended objects. A detailed model of this sort was proposed
by Abraham and Lorentz. Supposing that real electrons (i.e., 'atoms of electricity')
correspond to a uniform, rigid distribution of charge over a sphere, Lorentz calculated
the self action (quite literally the retarded action of one part of the charge distribution
on other parts), leading to the expression for the self-force:
110 CHAPTER VA

4
F -_ - - Wself a + -2e
2 L (-1t a (n)
OCR-n-l),
(7)
3c2 3c2 n!cn

where all derivatives appearing in the summation are with respect to time, and if is
the assumed radius of the charge distribution. The first term suggests that the energy
of self-action, Wse/f, may be regarded as generating the 'electromagnetic mass' of the
electron,

Wself (8)
mself = 2'
c

which resembles Einstein's famous equation. An attractive feature of (8) was that all
the mass of the electron could be attributed to its self-action. This suggested the picture
of a purely electromagnetic electron.
The second term of (7) corresponded to the force of radiative damping, and
suggested the equation of motion

(9)

with m being the mass of the electron, and F ext the external force, provided the
higher-order (structure-dependent) terms in (7) are discarded.
A number of difficulties arose in connection with this model. This equation of
motion (9) involves a third order in time: it involves the derivative of the acceleration.
In fact the higher order, structure-dependent terms have been suppressed. Specification
of initial position and momentum does not, therefore, ensure a unique solution. If one
attempts to eliminate the structure-dependent terms, by going to the limit if -+ 0, the
self-energy diverges.
The second difficulty is that the various parts of the charge distribution must repel
each other according to Coulomb's law, so that the charge distribution would disinte-
grate.
The third difficulty arises from the fact that in passing from (7) to (8) we suppressed
a factor of 4'3 which should appear in front of the mass. This created a major hurdle in
obtaining the appropriate relativistic generalization, since the factor of 4'3 destroys
Lorentz covariance.
Poincare proposed that ad hoc cohesive forces could be used to make the electron
stable. Regardless of their origin, the requirement of stability ensured that the self
energy due to these cohesive forces exactly changed the factor 4'3 to 1, making the
Abraham-Lorentz electron suitable for relativistic generalizations. (It was later pointed
out16 that the factor 4'3 was incorrect.) However, for relativistic generalizations, the
electron was taken to be an oblate spheroid, and the resulting electromagnetic mass
displayed a complicated dependence on velocity.
The theory thus lost its major attraction, which lay in the idea of a purely electro-
magnetic electron. The objections, however, remained to the structure-dependent
 THE ELECfROMAGNETIC FIELD 111

terms as well as to the idea of the electron as a miniature analogue of a macroscopic

object. Quantum mechanics provided the coup de grace.

4. 2 Action at a distance

Short of modifying electrodynamics at the microscopic level, the other simple way of
getting rid of the divergences of field theory was to revert partly to the logically clear
Newtonian picture of action at a distance. Particles act upon other particles, not uppn
themselves. The field is a fiction meant to represent this action. The divergence arises
because we attempt to extend this fiction too far, to self-action.
This point of view was taken up by Schwarzschild, Tetrode and Fokker, who
formulated an action principle. Gauss had regarded this as the 'keystone', which he
failed to fmd.
The basic difficulty with this approach is to balance action and reaction: the field
is no longer available to balance action and reaction locally. Moreover, because of the
finite speed of propagation, if A acts on B, the back reaction from B to A arrives too
late. Poincare emphasized:
Under these conditions can there be compensation between action and reaction ...? Evi-
dently not ....Even if the compensation should be exact, it could not be simultaneous. The
perturbation...propagated with finite velocity... reaches the second electron only after the
first has long ago entered upon its rest.

The only way to balance action and reaction, in a theory with propagated action, is to
allow B to react back on A through an interactionpropagating into the past, i.e., by the
introduction of advanced potentials.
In effect, such a theory turns out to be mathematically equivalent to a field theory
in which advanced and retarded potentials enter symmetrically (a=:v'2, in (3. It is
important to clarify that there is no violation of the relativity principle here. If advanced
potentials could be accessed, it would be possible to transmit information instantane-
ously, across any distance, by means of a retarded interaction traveling into the future
together with an advanced interaction traveling into the past. There is no violation of
the relativity principle because special relativity is not based on the postulate that
information cannot be transmitted faster than the speed of light. This is a wrong
understanding of the postulate which only asserts that the speed of light must be
constant in all inertial frames. An additional assumption of 'causality', or the blanket
exclusion of advanced potentials, is needed to prevent instantaneous communication
at a distance. Not only may one not assume this in the current context, but, as in Dirac's
approach, the introduction of advanced potentials turns out to be necessary for a
covariant theory of radiative damping.

4. 3 Dirac's approach

Dirac's approach to the divergence difficulties was based on the following observation.
Any solution of the field equations can be written using both advanced and retarded
potentials, as

F = Fret + Fout = Fadv + Fin. (10)

112 CHAPTER VA

Here the field F is being represented in a certain volume V, and Fin and Fout fields refer
to fields coming in and going out of this volume. Fret and Fadv refer, of course, to the
retarded and advanced fields obtained from the corresponding potentials, using (5).
Dirac proposed the elegant and natural definition of a radiation field as the flux through
the volume (taken as a sphere, say, of large radius),

Frad = Fout - Fin. (11)

With the radiation field defined in this manner, advanced fields must necessarily be
taken into account in considering radiative phenomena,

Frad = Fret - Fadv. (12)

It may be possible to define the radiation field in a different way, but is it an accident
that this simple and natural definition of the radiation field leads to the correct
expression for radiative damping (after multiplication by 1I2)? Dirac showed that the
radiation field defined by (12) corresponds to a finite force given at the world line of
the electron by

(13)

where the bracket denotes anti-symmetrization, z == z(t) is the world-line of the elec-
tron, and dots denote differentiation with respect to proper time t.
The important thing is that the radiation field is (a) covariantly defined and (b) free
of singularities everywhere including on the world line of the electron, even though the
electron has been assumed to be a point charge, and we have supposed that Maxwell's
equations hold right up to the point charge. Dirac thought of the above definition as a
subtraction procedure, somewhat similar to what he had used in the theory of the
positron.
Dirac also obtained an alternative expression to (13), involving higher derivatives,
but rejected it on the grounds that one could hardly expect the more complicated
expressions to apply to a simple thing like an electron.
We can finally write down the Lorentz-Dirac equation of motionof an electron:

(14)

The radiative damping, once again, gives rise to the very strange term, called the Schott
term, which is of the third order in time,

(15)
2 2
--e 2 2 zp..
= - -e
3 vp. 3
 THE ELECfROMAGNETIC FIELD 113

Dirac, however, regards this as bringing out 'the most beautiful feature of the theory.'

5 Preacceleration

To see the beauty of the theory let us consider the non-relativistic limit of the equations
of motion,

l~v
- = -E+-v
e e
XH+-
b -dt 2 ,
(16)
dt m me

where ~ =~ ;;2' two thirds the time taken by light to cross the classical electron
radius, is known as the relaxation time of the electron. This is the same as the Lorentz
equation of motion (9), with the external force being due to the electromagnetic field.
In the absence of external fields and forces, the above equation has the solution

(17)

known as the runaway solution. The other solution is the trivial solution:

j, = o. (18)

The first solution (the 'runaway solution') diverges for an arbitrary non-zero initial
acceleration. This is clearly unphysical. Dirac proposed that one should, therefore,
choose the second solution, so that the acceleration of a charged particle is zero after
a force is applied. But this means that the acceleration of a charged particle must be
non-zero before a force is applied.
Formally, it is possible17 to write the general solution of the one-dimensional
equation of motion (9) as

v = i t [v (0) - ! I~ e -bt' f(t') dt'],

(19)

(for purely time-dependent external forcesf(t, so that Dirac's proposal that the accel-
eration be zero after the force is applied amounts to the assertion that the velocity of a
charged particle is decided by a weighted average of all future forces:

v(0) = ~ fJO e -bt' f( t') dt'. (20)

m 0
114 CHAPTER VA

Dirac interpreted this as evidence for a new way of looking at the finite size (non-local-
ity) of charged particles, the interior being a region where signals are transmitted faster
than light. The particle, however, extends, in an exponentially weighted manner, all the
way to infinity. The beauty of the theory is that it starts with the assumption that the
electron is essentially a point particle, being 'too simple a thing to require such
complications'.
One can, of course, question this procedure: does one have the freedom to impose
such conditions? As first pointed out by Haag, and stressed by Rohrlich,18 such subsid-
iary conditions, corresponding to the asymptotic condition that the acceleration be zero
at past and future space-like infinity, must be imposed in order to obtain a meaningful
system of equations of motion.
Eliezer19 objected to Dirac's procedure and argued that the equations of motion
did not admit physical solutions for cases such as an attractive or repulsive Coulomb
field. However, Plass20 suggested that Eliezer had overlooked the physical solution from
amongst the infinity of unphysical solutions. For the one-dimensional relativistic equa-
tions of a particle acted upon by forces independent of velocity,

2 (21)
. 1.. 1 vv
v-- v + - - -
b b c2 +v2

he suggested the use of the transformation

(22)
vCr) = c sinh [~]
fC

leading back to the non-relativistic equation considered earlier. Therefore, for this case,
the solution exists whenever the force admits a Laplace transform, and Plass considers
a number of particular cases. All the physical solutions, of course, violate causality.
The actual 'violations of causality', however, take place over a time period which
is very small, only 10-23 s, and one may choose to ignore them. This attitude originated
at a time when21
Bohr had often emphasized that quantum mechanics could not be expected to be applicable
to phenomena in which distances of the order of the classical electron radius e2/mc 2 playa
role, since quantum theory does not purport to give an account of the structure of the
electron,
and Dirac [1934] had claimed,
A more reasonable assumption to make is that the relativistic wave equation fails for energies
of the order of 137mc2
But we seem to be stuck with it. Davies22 comments:
Somehow a feeling of dissatisfaction remains. The runaway solutions are eliminated at the
expense of introducing unpleasant acausal behaviour, which in turn is disregarded because
of the values that the constants of nature (e and m) happen to have. Unless it can one day
be shown that nature is constrained in some fundamental way to choose the values of the
constants of nature to preserve causality, this explanation will not be totally convincing.
To summarize, the most obvious way of removing the divergences of the field, the
Abraham-Lorentz model, seems to have failed. Other ways of removing the divergences
 THE ELECfROMAGNETIC FIElD 115

result in 'violations of causality'. Two questions now arise: (a) Is there some way of
getting rid of these 'causality violations'? (b) If 'causality' is 'violated', or if there is a
small 'tilt in the arrow of time', how does one deal with the paradoxes of advanced
action? Thus it seems that the field, instead of resolving the paradoxes of thermody-
namics, has resulted in some more paradoxes!

6 Conclusions

The intuitive notion of 'contact' between particle and field is unintelligible, while the
field incorporates the idea of 'locality' or a 'chain of causes' in an illusory way: field
theory consists of its differential equations. Advanced radiation cannot be entirely
eliminated on the basis of mundane observation and the differential equations of field
theory. These equations exhibit divergences, and the known ways of removing them, by
action at a distance or by Dirac's approach, necessarily open the possibility of interac-
tions propagating into the past.
VB ELECTROMAGNETIC TIME
ABSTRACT. In the previous chapter we saw that the introduction of the field creates more problems than
it solves. Here we begin by disregarding the problems due to the field, and formulate the two-body problem
of retarded electrodynamics, without radiation reaction. The resulting equations of motion are time-asym-
metric, and fail to satisfy the 'phase-flow' hypothesis underlying the recurrence and reversibility paradoxes.
We present a counter-example to show that the Lorentz-Dirac equation, resulting in preacceleration, may
be invalid since it is derived by replacing a retarded ordinary differential equation (o.d.e.) by a higher-order
standard o.d.e. obtained by Taylor approximation. The solutions of advanced o.d.e.'s branch into the
future, implying in-principle unpredictability from the past and resolving Popper's pond paradox. The
branching and collapse of solutions of mixed o.d.e.'s suggests a resolution of the Wheeler-Feynman and
grandfather paradoxes.

With a direct -action theory, or with Dirac's definition of radiation damping, the elimination of advanced
interactions is a serious problem. We present an exposition of (i) the Sommerfeld condition, pointing out
its arbitrariness; (ii) the Wheeler-Feynman absorber theory, pointing out its internal inconsistency; and
(iii) the Hogarth-Hoyle-Narlikar theory, pointing out its external inconsistency. The remaining absorber
theory predicts the existence of rare advanced interactions. We compare this with the empirical results of
Partridge, and suggest that experiment proposed by Heron and Pegg may now be revived.

1 Introduction

C HAPTER IV ended with the hope that the introduction of the field may help to
resolve the paradoxes of thermodynamics. But in Chapter VA we saw that the
introduction of the field seems to create more problems than it solves.
Let us, for the moment, forget about the field, or else let us adopt the point of view
that the field is a dispensable intermediary between particles. Let us look, instead, at
the nature of the many-body problem (of electrodynamics), and the novel features
arising from the finite speed of interaction: to measure time, or anything else, one must
postulate that the speed of light is constant.
What are the consequences of the finite speed of interaction? We recall Poincare's
remark: 'The state of the world will depend not only on the moment just preceding, but on
much older states.' In both cases we obtain what Poincare called (p 65) 'equations of
finite differences'. As Poincare further argued, the substance of physics lies in its
mathematical formalism - the 'mechanical explanations' are redundant. So, between
fields and particles, it matters little which mental picture we feel comfortable with.
How does the finiteness of the speed of interaction affect the underlying equations? To
see this, consider a system of n particles. In the field picture, an accelerated charged particle
el gives out retarded radiation which is incident upon other charged particles e2, e3, ..., en.
The outgoing retarded wave accelerates other charged particles at later or retarded times.
In the particle picture, only these accelerations matter: the acceleration of el at time t
depends upon the acceleration of e2, e3, ... , en, at past or retarded times, say
t-i2, ..., t-in. Ignoring, for the moment, the question of self-action and radiation damp-
ing, the difference between the field-picture and the particle-picture does not show up
mathematically.
 ELECfROMAGNETIC TIME 117

2 The two-body problem of electrodynamics

2.1 Formulation

For example, the equations of motion of two charged particles i and j, in one dimension,
interacting solely through retarded radiation (and without radiative damping) take the
form: 1

Zi (Si) = !.. [1 +;hSi)]V2 sgn [Zi(Si)-Zj(Sj)] 2 '

(1)

m ([Zi(Si)-Zj(Sj)]Zj(Sj)- IZi(Si)-Zj(Sj) I cij(Sj)}

where (t, z), with the appropriate subscript, denote the coordinates of the world-lines
of the particles, dots denote differentiation with respect to the proper times Si, Sj of the
two particles, and (ij) = (1,2) or (2,1). Givensi, the retarded proper time Sj in the above
equation, corresponds to the point at which the backward null cone from the point
(ti(S;), Zi(Si meets the world line of particle j. This is obtained from (see Fig. 5):

(2)

The difficulty of having two independent variables can be removed by rewriting (1)
as

V'i(t) c-( -l'jvj(t--rji(t (1')

c+( -l'jvj(t--rji(t'

where k = -el ez/c2, the explicit retardations

-rji (t) = t - tj (Sj) (2')

are obtained from equation (2), primes denote differentiation with respect to t, and
Vi = Z'i is the velocity.
Reforming the notation (to use bars rather than subscripts), and using units with
c = 1, these equations may be rewritten:

v'et) b l+v(t--r) (3a)

= -r2 1-v(t--r) ,
[ 1-v2(t) ] Y2

v'et) Ti 1-v(t-f) (3b)

[ 1-v2(t) ] :Y.z 'f2 1+v(t-f) ,

where b = -kim, Ti = -kim, and

118 CHAPJER VB

-e(t) = Ix(t)-x(t--e) I , (4a)

'f(t) = Ix(t)-x(t-i) I . (4b)

We can now see the difference quite clearly: the character ofthe differential equation
has changed. The system of equations (3) is no longer a system of simple differential
equations, but is a system of 'difference-differential equations' or a system of ordinary
differential equations (o.d.e.'s) with retarded deviating ar[J!:lments. The 'accelerations'
of the particles at time t depend upon their velocities v, v at past or retarded times,
t--e, t-'f.
What is the significance of this change? To study the two-body problem of electro-
dynamics, one must study o.d.e.'s with deviating arguments: even the simplest qualita-
tive features of such o.d.e.'s completely destroy the Newtonian paradigm, and suggest
a resolution of the paradoxes of thermodynamics, and the paradoxes of advanced action.

2. 2 Some definitions

A first-order o.d.e. has the form

x'(t) = f(t,x(t, (5)

where f is some function (generally non-linear). It is well known that the most general
system of o.d.e.'s can be reduced to a system of such equations, i.e., to the form (5),
regarding x as a vector, if necessary. It is also well known that (under some mild
requirement of continuity on f) prescription of the 'initial value' xeD), at an instant
t = 0, determines a unique solutionx(t) of(5) in a neighbourhood [-6,6] of t = D. This
is the Newtonian paradigm.
In contrast, a differential equation with deviating arguments has the form

x' = f (t, x(t), x(t--e) ) . (6)

The 'dependent variable', the function x, appears for more than one value of its
argument, the 'independent variable', t.
An equation with deviating arguments is classified as retarded, or history dependent,
if the highest order derivative of the unknown function appears for exactly one value
of the argument, and this argument is not less than all the arguments bf the unknown
function and its derivatives appearing in the equation. For example,

x'(t) = f (t,x(t),x(t--e(t ) (8)

is called retarded ifT(tD.

Similarly, the equations of motion of char' ed p:,,;ticles interacting solely through
advanced radiation correspond to anticipatory behaviour, or an o.d.e. with advanced
deviating arguments,

x'(t) = f (t,x(t),x(t+-e(t)). (7)

 ELECTROMAGNETIC TIME 119

The definition requires the same proviso as above, except that now the argument of the
highest order derivative must be less than or equal to all the other arguments of the
unknown function, i.e., r(t) >0 in (8), or r(t)<O in (6).
More generally, a system of charged particles interacting through both advanced
and retarded radiation displays partly anticipatory behavior, corresponding to an o.d.e.
with mixed-type deviating arguments:

x'(t) =f (t,x(t),x(t-r1(t)),x(t+r2(t))). (9)

From now on, equations of the type (6), (7), and (9) will be referred to as retarded,
advanced, and mixed-type o.d.e.'s. The mathematical theory of such differential equa-
tions with deviating arguments, also known as functional differential equations, differs
from the mathematical theory of the usual differential equations.

2. 3 The recurrence paradox and the past-value problem

Van Dam and Wigner2 considered equations involving both retarded and advanced
fields. They asserted (without proof) that instantaneous positions and velocities were
sufficient to determine unique trajectories.
Now, with electromagnetic interactions taken into account, the many-body equa-
tions of motion (3) are retarded o.d.e.'s. However, for the simplest model of even a
retarded differential equation, modeling a history-dependent situation, it is inadequate,
in general, to provide initial data at a point. Consider, for instance, the o.d.e with
constant retardation n/2,

x'(t) = x(t-I)' t ~ O. (to)

To obtain a unique solution, it is insufficient to specify only the state at one point of
time, sayx(O). Thus,x = cos t andx = sin t are obvious solutions, and since the equation
is linear x = a . cos t + b sin t is a solution for arbitrary constants a and b, and both a
and b cannot be determined from a knowledge of x(O).
Since the behavior is history-dependent, it is more reasonable to ask for a unique
solution after prescribing the past history, i.e., an initial function x = rp, over the relevant
part of the past: the interval of retardation, [-n/2, 0].
In general, a unique solution of the past-value problem for the retarded system,

x'(t) = f (t,x (t-rl(t), t-r2(t), ... , t-rn(t)) ), (11)

may be obtained under the following sufficient conditions.3 (i) All delays, ri, are
bounded, and (ii) some technical conditions such as a local Lipshitz condition and a
continuity condition are satisfied.
From the point. of view of thermodynamics, the interesting conclusion is the
following. The hypotheses underlying the recurrence paradox have been destroyed:
there is no longer a unique trajectory through each point of phase space. More than one
trajectory may pass through each point of phase space; trajectories may intersect (Fig. 1).
120 CHAPTER VB

Effect. of History Dependence

Innlfficlency oflnitIDJ va/ue.J Fig. I: ElTects of
1.3
history-dependence
12
1.1
"""
0 .9
\ -' ~ Three solutions of a model first-
0.6
~J \ 'l.. / K'" '\, ...-- ~ order delay equation, showing the
;r:
0.3
0
If \ '-q /! '\ I"~ ~ "'- effects of history-dependence. All
-0.3
J \.\ / I)'{ ,,~
/ 1/)' three solutions have the same ini-
1 ....... ~lfI
\,. 'Ii / '- ~ Jf
tia/ value x(O), though the past his-
-0.6
IJ .......
-0.9
~ tories prescribed over the
- 12
-1.0 0.0 3 .0 6.0 9.0 time-interval [-1, OJ are different.
Note also the discontinuity in the
- Solution 1 -0-- Solution 2 - 0 - Solution 3 derivative of solution 1, visible at
~--------------------------------------~ t=~

2. 4 The reversibility paradox: time asymmetry of delay

For retarded o.d.e.'s the intersection of trajectories takes place preferentially towards
the future, in a way that destroys the hypothesis underlying the reversibility paradox.
An ordinary differential equation is time symmetric: it may be solved either forward or
backward in time. From a knowledge of the current values, Newton's laws may be used
to predict the future or retrodict the past. However, a retarded o.d.e. relates past causes
to current effects. Such an equation may be solved forward in time, but not, in general,
backwards in time.
Consider the following ordinary, linear, retarded differential equation with con-
stant coefficients, and constant retardation r:

x'(t) == ax(t) + bx(t-r), (12)

with b different from zero and r> O. To solve the equation backwards, it is only necessary
to solve an algebraic equation,

x'(t)-ax(t) (13)
x(t-r)
b

to obtain the solution on [t-2r, t-r], given x == <p on [-r, 0]. For nonlinear equations, this
already means that backwards solutions will not be unique. For the case under consid-
eration, suppose a:;t;-b and we prescribe <pet) == k, a constant, on [-r, 0], and ask for a
backwards solution for t :5 O. Then (13) implies that x(t) == -aklb so that the unique
solution of the algebraic equation (13) fails to be continuous, and hence differentiable.
Therefore, a (continuous) backwards solution of (12) does not exist in general.
Of course, one could think of choosing a final function in such a way that the
solution exists. But then the solution would, in general, fail to be unique. Consider

x' (t) == bet) x(t-1), (14)

where b is any sufficiently smooth (e.g. continuous) function which vanishes outside
[0, 1], and with
 ELECI'ROMAGNETIC TIME 121

Jbet) cit = -1. (15)

For example,

t :sO,
bet) ~ {-\ + ~ 2m o :S t :S 1,
t~1.
(16)

For t :S 0, (14) reduces tox'(t) = 0 so that, for t :S O,x(t) = k for some constant k.
Now if k is any constant then, for t E [0,1],

x(t) = x(O) + f ox'es) ds

= x(O) + f o b(s)x(s-l)ds
= x(O) + x(O) f o b(s) ds, (17)

since x(s-l) == k = x(O) on [-1, 0]. Hence, using (15), x(l) = 0 no matter what k was.
Butx(l) = 0, and bet) = 0 for t ~ 1, implies, by (14), thatx(t) = 0 for all t ~ 1. Conse-
quently, (14) does not admit a unique backwards solution even if we prescribe future
data for all future times t ~ 1. Thus, if cp differs from 0 on [1,00) there are no backward
solutions. But, if cp == 0 on [1, 00), the solutions branch into the past (Fig. 2), and there
is no way to pick a unique solution from the infinity of continuous solutions that are available.

TIme Asymmetry of Delay Fig 2: Time asymmetry of delay

Collapse ofsohltions ofQ l'eJmrJed equalio"
1.2
Three solutions of the retarded
equation (14) which col/apse to-
"'-
0.8
wards the future. The different past

"
0.4
histories, presecribed over the time-
x 0 interval [ 1, OJ, all result in the same
/" future solution for t~l. Retrodic-
L
-0.4
tion is hence impossible from fu-
-0.8 nP" ture data prescribed over t ~ 1.
-1 .2 -1.0 0.0 1.0 2..0 Teleological 'explanations' are im-
possible with history-dependent
evolution.
-0- Solution 1 -<>- Solution 2 -0- Solution 3
122 CHAPTER VB

2. 5 Preacceleration: the Taylor-series approximation

We saw in Chapter VA that the study of radiative damping and, in particular, the Schott
term, leads to equations that are of the third order in time, resulting in the preaccelera-
tion of the electron. Dirac, in 1938, obtained these equations by means of a Taylor
expansion which seems unavoidable.4 Many other authors5 have attempted similar
approximation procedures, using a Taylor series expansion to get rid of retarded/ad-
vanced expressions, in dealing with the two-body problem in electrodynamics and
gravitation. Physically, this procedure means that we model a history-dependent system
by an instantaneous system with additional degrees of freedom.
This procedure is known to be, in general, invalid. This may be seen from the
following counter-example,

x'(t) = -2x(t) + x(t-,-r), (18)

where r> 0 is a small constant. Every solution of this equation is bounded6 and tends to
zero as t-+ oo But if we choose the Taylor series approximation to the right hand side
and truncate after two terms, we obtain

(19)
x'(t) = -2x(t) + [x(t)-rx'(t) + ~ lxl/(t)] ,

which admits exponentially increasing solutionsx(t) = c exp(at), with a > O. Thus, the
Taylor approximation of (18) by (19) leads to qualitatively incorrect behaviour, no
matter how small r is, so long as r>O.
It may be shown that it is not the order of the approximation which is at fault: with
instantaneous data, even an infinite number of degrees of freedom is inadequate. The
order of the approximation does, however, make a difference from the numerical point
of view, as pointed out by El'sgol'ts,7 'since the transition is equivalent to the rejection
of the term with the highest order derivative in an unstable-type differential equation
with a small coefficient before the highest derivative.'[Emphasis mine]
In the usual treatment of the numerical solution of retarded o.d.e.'s, attention is
focused upon the discontinuities that might arise at the ends of delay intervals (e.g.
Solution 1 of Fig. 1). However, one would expect the general electrodynamic many-
body problem to be 'stiff: there could be oscillations at widely varying frequencies. In
view of the Dahlquist barrier,8 A-stability fails for any rule higher than the trapezoidal
rule, so that the Taylor approximation could be numerically misleading for derivatives
of order greater than two. 9 Thus, Dirac was perhaps right in a way when he rejected the
higher order terms as too complex to apply to 'a simple thing like the electron'.
To summarize, the origin of the Schott term in the Lorentz-Dirac equation of
motion is mathematically dubious, and can result in qualitatively incorrect behaviour,
though it may yet provide a more robust numerical approximation than would be
obtained by the inclusion of higher-order relativistically covariant terms. The alterna-
tives that have been proposed,lO to the Lorentz-Dirac equation, have not proved
satisfactory. 11
 ELECI'ROMAGNETIC TIME 123

2. 6 The pond paradox: advanced equations

Popper's pond paradox, considered in Chapter VA, also admits a simple resolution in
the context. The paradox seeks to exclude advanced interactions on the grounds that
'without the aid of organization from the centre' it is impossible to arrange the
coherence of generators required to produce a convergent ripple (advanced wave). The
resolution is that advanced waves, or anticipatory phenomena, cannot be predicted
from past information, and hence cannot be arranged.
The reasons are simple. In the first place, anticipation is as time-asymmetric as
delay. The past-value problem for advanced o.d.e.'s is, in general, insoluble for exactly
the same reasons that retarded o.d.e.'s do not admit backwards solutions (Le., the
future-value problem is insoluble for retarded o.d.e.'s).
As an explicit example, analogous to (14), consider the equation

x'(t) = b(t)x(t+l), (20)

where b has the same properties as before, except that we now require

f bet) dt = 1 (21)

in place of (15). For example,

t :S; 0, (22)
o :S; t :S; 1,
t~1.

For t ~ 1, x' (t) = 0 implies x(t) == k, for some constant k. But, if x(t) = k for t ~ 1,
where k is any constant, then, arguing in the same way that leads up to (17), we find

x(t) = x(l) - IIt x'es) ds

= x(l) - IIt b(s) x(s+ 1) ds
= x(l) - x(l) IIt b(s) ds, (23)

since x(s+ 1) == x(l) = k for s E [0, 1]. Hence, x(O) = 0, which implies x(t) = 0 for all
t :S; 0, since bet) = 0 for t :S; 0, so that, by (20), x' (t), vanishes for t :S; O.
Thus, the solutions of the advanced equation (20) may collapse towards the past in
exactly the same way as solutions of retarded equations collapse towards the future.
 124 CHAPTER VB

nme Asymmetry of Adv.n ce Fig. 3: Time asymmetry of

FIduN blVtlCJrbrg of.Ol>dklS Ofem advanced -"<>"
1.2 advance
0.8
004
,;- Three solutions 0/ the advanced
.,/ equation (20) which branch towards

'" '"
'" 0
the future. Three different future his-
tories prescribed over the interval
.(l,B
{1,2] result in the same solution/or
-1.2
-1.0 0 ,0 1.0 2.0 t S O. Prediction from past data,
TI.ma over t ~ 0 is, thus, impossible.
--0-- Solution 1 - 0 - SolutIon 2 -0-- SolutIon 3 Causal 'explanations' are impossi-
L..-_ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _~ ble/oranticipatory phenomena.

Another way of looking at this is that solutions of the initial function problem for
(20), with coefficient as in (22), branch into the future as shown in Fig. 3. Given the

entire past history, tp, it is impossible to predict the future for an anticipatory system
which obeys (20). Thus, if tpCt) is different from anywhere on (-00, 0], then (20) does
not admit any solution. But if tpCt) == 0 on (-00,0] then (20) admits infinitely many
continuous solutions, and there is no way to pick a unique solution.
With retarded interactions, retrodiction is impossible; it would be absurd to rule
out history-dependent phenomena on the grounds that all phenomena ought to be
retrodictable. With advanced interactions, prediction is impossible; it is equally absurd
to rule out anticipatory phenomena on the grounds that all phenomena ought to be
predictable. It is the last requirement which leads to Popper's paradox.

2. 7 Indeterminism: mixed deviating arguments

Observational limits on advanced interactions, considered later on, show that the above
kind of behaviour is not the norm. Nevertheless, it is possible to combine the two
examples given above to get an idea of the type of situation that might prevail in a
physically plausible universe in which both types of electromagnetic interactions are
present, but advanced interactions form only a very small component, say one part in a
trillion. In this case, the evolutionary equations would involve mixed deviating argu-
ments.
No general methods are known for obtaining or even proving the existence of a
solution of mixed-type o.d.e. But, the examples (14) and (20) may be combined to show
that mixed-type o.d.e. will exhibit both the 'branching' and 'collapse' described earlier
in Figs 2 and 3.
For example, consider

x'Ct) = a(t)x(t-l) + b(t) x(t+ 1), (24)

where bet) has the properties as in (21) and (22). If we prescribe x(t) = tp(t) == 0 for
t S 0, then, on [0, 1], any solution of (20) is a solution of (24) since we have
x(t-l) = 0 on [0, 1]. For t> 1, bet) = 0, so the equation to be solved is of type (14),
x '(t) = a(t)x(t-l), which shows that each (non-unique) solution of (24), on [0,1], has a
unique continuation for t> 1.
 ELECfROMAGNETIC TIME 125

Fig. 4: The mixed case

1.2

0.8 Three solutions of the mixed-type

0.'1
f \ equation (24), which first branch
x 0
.J \, and then col/apse, showing that both
'\ /' prediction and retrodiction are, in
-0.8
\ I genera~ impossible for such equa-
tions. Solutions of such equations
- 1.2
_LO 0 .0 1.0 2 .0 3 .0 4 .0 ' .. 0 may be regarded as intrinsically non-
unique. The many-body equations
- 0 - SolutIon 1 -<>- Solution 2 - 0 - So1utIon 3 of motion are of mixed-type with a
'---- - - - - - - - - -- - - - - - - - - - ' small tilt in the arrow of time.

In particular, (24) does not provide a unique future solution even if the entire past
history is prescribed. The opposite sort of behaviour, with branching into the past, and
'collapse' towards the future, is also possible. For instance, for (24), define a(t) arbi-
trarily for t E (0, 1), a(t) = 0 for t E [1,2], a(t) continuous for t ~ 2 with

f3 a(t) dt = -1, (25)

2

and a(t) == 0 for t ~ 3. Then the solutions of (24), for the same initial function
cp(t) == 0 for t :5 0, are of the form shown in Fig. 4.This follows from the arguments given
for equation (14) and in the above paragraph.
A couple of observations about mixed o.d.e.'s will illustrate the current state of
knowledge. In the physics literature, precisely one exact solution is known, for the case
of circular orbits,12 and some numerical computations have been done for nearly-cir-
cular orbitsP From a general point of view, I am aware of only the works of Driver14
and Schulman,15 who recommends the use of boundary conditions in such a case and
proposes a formal method of solution to the boundary value problem. The kind of result
desired by Schulman was anticipated by Cooke and Krumme,16 who relate the solution
of an initial-boundary value problem for hyperbolic equations to the solution of a
boundary value problem for a differential equation of mixed (neutral) type, describing
the characteristics of the original hyperbolic system.

2. 8 The Wheeler-Feynman paradox

Wheeler and Feynman themselves proposed a resolution of their paradox by asserting

that discontinuous forces do not exist in nature. I do not find this resolution satisfactory17
for a paradox which is so well formulated.
One way out of this paradox is the well-known way of (hard) determinism. For
example, the grandfather paradox may be deemed to have been resolved by the proof18
that the Cauchy problem (initial value problem) is well posed in a spacetime with closed
time like curves. One might chance upon one's grandfather when he was a mere boy,
but it would be mathematically impossible to kill him! We pointed out in Chapter I, the
difficulty with this way: the empirical refutation of scientific theories tends to presup-
pose some indeterminism, or human freedom.
126 CHAPrERVB

The other way out is to observe that the paradox actually demonstrates the logical
incompatibility of the following three assumptions (stated informally):
(i) one is free to perform any experiment,t9
(ii) interactions with the past are permitted,
(iii) the past is fixed and unchanging.
An additional assumption, which usually remains implicit (but was used in the preceding
paragraph) is that
(iv) there is a univocal relationship between past and future (unlike the situation in
Figs 3 and 4), even if interactions with the past are permitted.
The usual thing is to reject (ii) while ignoring the silent contradiction between (i) and
(iv), and the less obvious contradiction between (ii) and (iv). We have shown that there
is some coherence between (i) and (ii), though, if (ii) is assumed, one must give up (iv)
and possibly (iii). A changing or 'dynamic' past would be very surprising, but it is not
logically impossible; its physical validity must be decided by experiment! Current
observational limits already show that advanced interactions, if they exist, must be very
rare. In that case, it would be only very rarely, or at the microphysical level, that the past
could be affected. We will examine this line of thought in greater detail in the next part,
in the context of quantum mechanics.

3 The absorber theory of radiation

3. 1 Action at a distance

From the counter-examples in 2.3, 2.4, and 2.5 above, it is clear that there are
advantages to dispensing with the field picture, and adopting direct action at a distance.
A theory based on retarded direct action would resolve the divergence difficulties of
field theory as well as the paradoxes of thermodynamics. A theory based on mixed
interactions may resolve even deeper problems.
There would be no great loss to intuition, because, as we saw in Chapter VA, the
intuitive notions of 'contact' or 'locality' or 'chain of causes' are incorporated in the
field concept in a way that is unintelligible or illusory. But the problem with the
direct-action approach of Schwarzschild, Tetrode and Fokker was also pointed out: in
order to balance action and reaction, one must have a time-symmetric interaction.
As we saw in 2.6, 2.7, and 2.8 above, while advanced interactions do not lead to
logically impossible situations, they imply situations that are physically fantastic, unless
advanced interactions are very rare. The real world may not be perfectly time-asym-
metric, but, locally at least, it is not time-symmetric. How should one reconcile the
time-symmetric direct-action theory with observation?

3. 2 The Sommerfeld radiation condition

How is the reconciliation achieved in field theory? We saw, in Chapter VA, that the
wave equation admits both retarded and advanced solutions; and, moreover, any field
may be represented, within a certain volume, using both retarded and advanced fields:
 ELECfROMAGNEI1C TIME 127

F = Fret + Fout = Fadv + Fin. (26)

Both expressions describe the same physical situation.

In a famous paper20 recording their disagreement, Einstein and Ritz observe that
retarded and advanced fields are equivalent 'in some situations'. Ritz proposed that
there should be a law against advanced fields! Einstein, on the other hand, thought that
it was enough to restrict the solutions to be retarded: the exclusion of advanced fields
was not law-like, but a matter of physical acceptability. Ritz regarded the choice of
retarded fields as 'one of the roots of the Second Law [of thermodynamics],' while
Einstein believed that 'irreversibility is exclusively based on reasons of probability.'
The analysis in Chapter IV and 2 above shows that there are certainly merits to
Ritz's point of view: it is easier to base irreversibility on retarded interactions than on
reasons of probability. But the blanket exclusion of advanced fields would go against
Dirac's covariant definition of the radiation field, which necessarily involves both
advanced and retarded fields. Dirac adopted the pragmatic point of view that time-sym-
metric fields be accepted for all theoretical purposes, while, in practice, one could
restore time asymmetry, by imposing the condition

Fin = O. (27)

This is known as the Sommerfeld radiation condition, in analogy with scattering theory,
since it fixes the incoming wave. If (27) is imposed, the second equation in (26) fails.
The whole procedure seems a bit arbitrary: one uses advanced fields when conve-
nient or necessary, and rejects them otherwise.21 But even this arbitrary procedure is
not available to the direct-action theory, for the field does not exist as an independent
entity.

3. 3 The Wheeler-Feynman theory

Wheeler and Feynman22 proposed the following procedure. The direct action theory of
Schwarzschild, Tetrode, and Fokker is equivalent to field theory with time-symmetric
fields. Given a collection of charged particles, the total field acting on a charged particle
kamountsto

(28)

The field that ought to act on particle k according to retarded electrodynamics is

(29)
~ F(j) ret + ~ (F(k) ret - F(k) adv).
]'k

That is, the field acting on particle k should consist of the retarded fields of all other
particles together with only the radiation field of the particle k.
128 CHAPTER VB

The necessary and sufficient condition for (28) and (29) to be equal is

(30)

all particles

Is there any physical circumstance in which (30) can be realized? If all the particles are
shut inside an opaque enclosure, the absorber, we would certainly expect all radiation
fields to vanish outside

(31)
(outside).

But the radiation field has no singularities inside - it is a solution of the homogeneous
wave equation - it must, therefore, actually vanish everywhere.
But complete destructive interference is impossible between converging and di-
verging fields. Therefore, we must also have

L Fret = 0,

L Fadv = 0, (32)

which is equivalent to the desired condition (30).

Thus, in a totally absorbing cosmos, time-symmetric interaction is equivalent to
retarded interaction. A picturesque summary of the procedure is this: one sees the sun,
or a distant star, just because its light is absorbed by the eye. The sun would not shine
unless it knew that its light would eventually be absorbed. In this picture, which has
been compared with Faraday's lines of force,23 the source sends an 'offer' wave into the
future, the absorber returns an advanced wave into the past, resulting in a 'transaction',
or a giant handshake across spacetime.
Despite the appeal of the picture, the Wheeler-Feynman result is puzzling. How
did time asymmetry emerge from time symmetry? Thus, inside a totally absorbing
cosmos satisfying (32), we also have the other equality,

1 (33)
L ~ (F(j)ret+F(j)adv) = LF(j)adv -2 (F(k)ret-F(k)ad'V) = o.
j~k j~k

That is, fully advanced fields with radiative anti-damping are also consistent. Is there
any reason to regard (29) as more valid than (33)1 Wheeler and Feynman argued as
follows. When particle k is disturbed, the retarded radiation from the particle disturbs
all other particles, at future times. The advanced (response) fields acting on the particle
are therefore highly correlated. The retarded fields, due to past motions of absorber
particles, may, however, be assumed to be uncorrelated. Therefore, on grounds of
statistical mechanics, one must regard (29) as more valid than (33).
 ELECfROMAGNETIC TIME 129

This argument rather begs the question. With time-symmetric interactions, one
may no longer assume that particle motions are uncorrelated be/ore interaction.
Wheeler and Feynman gave other explicit calculations, which are also circular,24 as are
other similar arguments that have been advanced.2S

3. 4 Hogarth's theory

Hogarth26 pointed out that the Wheeler-Feynman calculation did not apply to an
evolving universe. In the first place, the mean free path of a photon is of the order of a
Hubble radius. That is, the radiation emitted by an antenna, in outer space say, would
travel a long distance - to the very edge of the universe - before being absorbed. This
would take a very long time, some 10 billion years. So, the advanced radiation would
be absorbed some 10 billion years in the past, whereas retarded radiation would be
absorbed some 10 billion years in the future. In an evolving universe, conditions are not
likely to remain the same for such a long period of time, so the past absorber would be
different from the future absorber.
Specifically, the absorption depends upon the (imaginary part of the) refractive
index, which, in turn, depends upon the frequency of the radiation, and the density of
charged particles. In an expanding cosmos, for instance, the density of charged particles
in the past absorber would be much greater than the density of particles in the future
absorber. Moreover, retarded radiation, traveling into the future would be red-shifted,
while advanced radiation, traveling into the past would be blue-shifted. Finally, one can
hope to compute the 'effective radius' of either absorber, without assuming it to be
infinite, as was done by Wheeler and Feynman.
In view of these factors, the qualitative difference between the past and future
absorbers may itself be the source of the asymmetry of radiation. Hogarth proposed a
linear theory: the response R of the past and future absorbers is a linear multiple of
the incident stimulus fields S, so that

Rp = pSp, Rj = IS!, (34)

"-
where the sUbscripts p and / refer to the past and future, while the coefficients p and /
are constants of proportionality. The stimulus field he took to consist of the radiation
field from the particle, and the response from the other absorber.
The effective field of the particle is, therefore,

(35)

From this follow the conditions for the consistency of retarded radiation: / = 1, p . 1.
On the other hand, ifp = 1,f:~1, advanced radiation is consistent.lfp=/, the radiation
is time symmetric, no matter how close p and / are to 1. However, if p =/= 1 the above
equation becomes indeterminate, and leads to no conclusion.
The case p = 1 corresponds to perfect absorption, or opacity, and holds for big bang
cosmological models.27 The case /= 1 holds for models with a final singularity, or those
130 CHAYrERVB

which expand slower than the Dirac model. It has been argued that this case also holds
for steady-state cosmologies.
Hoyle and Narlikar28 propose a theory similar to Hogarth's, in that it demands an
ideal future absorber, suitable for the steady state cosmology. Their theory differs in
some details, however. In the first place they use an action principle in place of
Hogarth's assumption of elementary time symmetric fields. Secondly, they prefer to
base their theory on an argument very similar to the Wbeeler-Feynman argument given
above. Thirdly, they make a number of ad hoc hypotheses in the further development
of the theory including its quantization; for example, they suppose, somewhat conve-
niently for the steady state theory, that the refractive index for advanced radiation
behaves differently: advanced radiation 'builds up' the further it is from the source.
The Hogarth-Hoyle-Narlikar theory is, in any case, externally inconsistent: accord-
ing to this theory, retarded radiation should be consistent in a steady state cosmos, while
advanced radiation should be consistent in a big-bang cosmological model like the
Einstein-de Sitter, and time-symmetric radiation in the closed Friedmann model. The
microwave background radiation, discovered a quarter century ago, and interpreted as
relic radiation from the big bang, seems to have sounded the death knell of the steady
state theory.

3. 5 Other theories

Since the Wbeeler-Feynman theory is internally inconsistent, this is not a very happy
situation. I have proposed an alternative theory, which goes back to the notion of the
electron as a finite-size object29 to justify the hypothesis of a lower signal velocity for
advanced radiation.
To calculate the absorber response, one now has to consider a countable infinity
of stimulus and response fields, which can be summed to arrive at the following
conclusions in terms of the response factors p and / for the past and future absorbers.
Retarded radiation is consistent in those models which satisfy p = 1 or /= -1 provided
Ip/ I ::F-l, while advanced radiation is consistent in those models which satisfy /= 1 or
p = -1, I p/ I ::F- 1. If both p = 1, / = 1, the results are indeterminate, but it is possible
for retarded radiation to be approximately consistent if we only have p=l, /=1.
Specifically, a mixture of the form (1-6)Fret + 6Fadv can exist, with

6 = ~ 1 (36)
1-/

provided (l-p) (I-f). The last condition is not, of course, the same as saying/ p.
,
For instance ifp= 1-10-20 and/= 1-10-1 then the above condition is true though not
/ p in the usual sense. The Einstein-de Sitter model satisfies the conditionp=I,/::F-l,
while the closed Friedmann model satisfies p= 1,/= 1, leading to the situation depicted
by (36).
 ELECI'ROMAGNEIlC TIME 131

4 Empirical tests

Is there any way to put to test any of the above considerations? One experiment has
been performed, and another announced and later abandoned. Both these tests concern
the possible existence of small amounts of advanced radiation, arising from incomplete
absorption in the relevant part of the cosmos.
Partridge30 in 1973 measured the power required to drive a horn antenna which
radiated alternately into free space, and into a local ('totally') absorbing cover, which
was placed far enough (20A.) away to eliminate any confusion due to power reflected
back. The point of a horn antenna is that it provides a directed beam, which could be
pointed towards different parts of the sky. The antenna radiated in the microwave
region to ensure that the emergent radiation, when directed towards free space, would
not be absorbed by the earth's atmosphere, but could be expected to travel cosmological
distances before absorption. Since any change in the power input would be expected to
be small, the changes were measured using a phase-sensitive detector to measure the
change in the current MIl.
Partridge assumed a relationship of the type

Pf = (1 - 6) Pa, (37)

where Pf denotes the power required while radiating into free space, Pa denotes the
power required while radiating into the local absorber, and 6 denotes a small number.
Conventionally, what happens to the radiation after it leaves the source should not affect
the source in any way, so that we should have 6 = 0; at any rate we expect 6 to be small,
which is the reason for the assumed linear relationship. On the Wheeler-Feynman
theory, it is clear from the above discussion that we should have 6 ~ O. Partridge thought
this was true of the Hoyle-Narlikar theory as well. The objective of the experiment was
to place an upper limit on the value of 6.
For a positive value of 6, as predicted by the Wheeler-Feynman theory, the
observed value of Mil should have been negative for the phase rp = ()O, and positive
for the phase rp = 18()O of the phase sensitive detector, as compared to the actual
observed values +(14.77.7)XlO- S, and -(1O.37.7)X10-S, respectively. I pointed
out that31 the phase sensitive detector is most sensitive for these two phase settings, and
both these values correspond to a significant negative value of 6. Such a negative value
of 6 is predicted by my theory mentioned above, while the theories of Wheeler-Feyn-
man and Hogarth-Hoyle-Narlikar are already ruled out on grounds of internal and
external inconsistency. Partridge actually took a weighted average over various phase
settings, to arrive at a mean value of 6 = (-1.1 1.6) x 10-9, still negative but no longer
significant. Undoubtedly, Partridge's procedure is robust, but this robustness is
achieved at the cost of diminished sensitivity of the experiment, since for values of the
phase setting other than ()o and 18()O the errors are systematically greater_
Partridge concluded that the result was consistent with 6 =0, conforming to ordinary
expectations, and that either the analysis in terms of the Wheeler-Feynman theory was
incorrect or that absorption along the future light cone was better than 1 part in 109.
Pegg subsequently suggested that, in order to obtain a non-null result, it was
necessary to use a dynamic absorber, a 'chopper', which would affect past and future
132 CHAFrERVB

absorbers differently. On this basis, an experiment was proposed by Heron and Pegg.32
The experiment also proposed to exploit the fact that, with a dynamic absorber, the
expected change in power input would be time varying. One could use a square wave
input, so that any modulation of this, by the 'chopper' could be easily detected. It was
estimated that the experiment would, thereby, have an increased sensitivity of 1 part in
1012, compared to the accuracy of 1 part in 1()9 in Partridge's experiment.
The experiment was eventually abandoned due to the absence of a picosecond
switch, which now essentially exists.
Experiments of Partridge's type do not exhaust all empirical possibilities, and we
will consider some other possibilities in the next part, including the possibility that a
mixture might result in a time which is structured rather than linear.

5 Conclusions

The many-body equations of motion, based on retarded interactions, help to resolve

the recurrence and reversibility paradoxes of thermodynamics, by destroying the
hypotheses underlying those paradoxes. The derivation of the Lorentz-Dirac equation
of motion for an electron, and hence the origin of preacceleration, is based on an invalid
approximation. The existence of mixed interactions, or a 'tilt in the arrow of time,' is
not logically impossible, and may even help to resolve deeper paradoxes. Such a 'tilt'
may, in fact, have been detected by Partridge's experiment. Further experiments with
increased sensitivity are now possible.
 ELECI'ROMAGNETIC TIME 133

Appendix: Derivation of the relativistic two-body equations of motion with a tilt in

the arrow of time

We give a summary derivation of the two-body equations of motion. Our aim is to make
a clear statement of the physical assumptions used in writing down, say, equation (1) at
the beginning of the chapter. It is not our intention to simplify the algebra.
The assumptions are the following.
(i) The particles are mathematical points with no size, so that one may speak properly
of the world-line of a particle. .
(ii) The two charged particles interact electromagnetically. In the retarded case, equa-
tion (1), it is assumed that the electromagnetic field (potential) due to a moving charge
is that given by the usual (Lienard-Wiechert) retarded potential (equation 2a of
Chapter VA). In the case of a tilt in the arrow of time, one uses mixed potentials
(equation 3 of Chapter VA, with a, presumed small, determined from further theoret-
ical and empirical considerations).
(iii) A charged particle moves in accordance with the Heaviside-Lorentz force law.
(Equation 4 of Chapter VA gives the non-relativistic form.) .
We partly follow the notation of Synge. Spacetime is taken to be flat, the spacetime
coordinates are xll with X4 = ict. This is not the conventional notation, but has the
advantage that it avoids the distinction between covariant and contravariant compo-
nents: all suffixes are written as subscripts. However, we let Greek suffixes range from
1 to 4 and Latin suffixes range from 1 to 3.
The inner product (dot product) of two vectors VIl , WIl is written

(AI)

For a unit vector, (VV) = 1 according as it is space-like or time-like.

We let:

L, L' = world-lines (A2)

m, m' = proper masses
e, e' = charges (in esu)
All, A'1l = unit tangents
tis, tis' = elements of proper time
u" u', = velocities
y-2 = l-u 2/c 2, y,-2 = l-u2/c2
u2 = u,u,
m' ee'
p=- k=--
m' m'c2

The Lienard-Wiechert retarded potential may be written more elegantly as follows.

e All (A3)
All = -~
134 CHAl'TER VB

This notation, however, suppresses the key fact that the potentialA,u refers to a point
P' on the world line L', while the tangent vector ,t,u refers to a different point P on the
world-line L. The retardation vector ~ is the vector connecting P to P'. This fact is
made evident in Fig. 5.
Fig. 5: The retardation vector
IFuture
Given anypoint P' on the world line
L', to obtain the retarded potential
All at P' due to the world line L, let
P be the point at which the back-
ward null cone from P' meets L. Let
Abe the tangent vector to L at P, and
let ~ be the retardation vector join-
ing P to P'. Now apply equation
(A.3).

The corresponding electromagnetic field is given by

(A4)

(AS)

where

,t [ 1+ (dA.)
p =-~ 1 dA.u
-~ ] +--~
(A6)
,u (,t~)3 ds (~)2 ds'

The required equations of motion are, therefore, given by

..l1' ,
,!:!!:...E.. _ fL- r;' l ' (A7)
m ds' - 2 r,uvlt. v,
C

which reduces to

(AS)

Similarly, we obtain the equation of motion for the other particle:

 ELECfROMAGNEllC TIME 135

(A9)

Equations (AS) and (A9) are the generalized forms of equation (1).
The neglect of radiation damping is not in any way essential to the conclusions that
follow from equation (1). One can add the radiation damping term to (A7) to obtain

(A 10)

It is easy to see that this makes the equation messier: though the highest order derivative
now appears in the radiation damping term, this does not change the basic type of the
equation.
The type of the equation does change if we use advanced or mixed potentials in
place of the retarded potential (A3). Mixed potentials are given by a convex combina-
tion of retarded and advanced potentials (equation 3, Chapter VA). From (2b, Chapter
VA), it is clear that advanced potentials are given by the same expression as (A3) except
that one must now construct the forward null cone at P' to obtain the point Q where it
intersects the world line L; .t must be taken to be the tangent vector at Q, and ~ as the
advance vector connecting Q to P' .
We emphasize that a relativistic treatment of this problem is essential. One can
certainly conceive the possibility of posing the problem non-relativistically, using the
non-relativistic expression for the Heaviside-Lorentz force law (5, Chapter IIIB). But
this would not be a very nice thing to do, because the Heaviside-Lorentz force is
velocity-dependent, whereas force is (and ought to be) a Galilean invariant in the
framework of Newton's laws. (Hence, the Heaviside-Lorentz force law was a major step
towards relativity.) Thus, the key feature of these equations of motion (viz. that they
are equations with deviating arguments) is a direct outcome of the postulated finiteness
of the speed of interaction.
Secondly, we have seen that the qualitative effects are not confined to relativistic
velocities. Consequently, there is a case to use these equations as the basic equations
of motion for statistical mechanics.
Finally, we notice that though the field-picture and the particle-picture view the
origin of radiation damping differently, this distinction did not enter in any essential
way in the equations. This is in line with Poincare's basic philosophy (expounded in
Chapter IIIB), that the equations are the key, and so long as we do not change the
equations it does not matter very much which mental picture we feel comfortable with.
The equations, however, do change with the relativistic postulate, and they change
further if one postulates a tilt in the arrow of time.
D. The topology of time
VIA BELL AND NON LOCALITY
ABSTRACf. Does time have a non-trivial local structure? This question is new to physics, but central to
the rest of this exposition. We explain how the notion of a structured time may be formulated. If time in
reality is structured then physics cannot be based on differential equations alone, but needs to be founded
on a new logic. The relation of such a new logic to quantum logic is postponed till Chapter VIB, while
questions about the global structure oftime are postponed till Chapter VII.
The previous chapter concluded by recommending the hypothesis of a tilt in the arrow of time. This
hypothesis has two consequences: a structured time and non-locality. We outline the first consequence
and indicate how most recognized new features about time in q.m. fit into the perspective of a (microphys-
ically) structured time, whereas hidden variable theories deny the existence of this structure by supposing
that 'reality' cannot but be definite. The rest of this chapter examines the physical acceptability of
non-locality in the context of Bell's inequalities.
We develop the background and state the inequalities. We point out the difficulty with inefficient detectors
and try to correct the popular misconception that the inequalities have been conclusively violated in
experiments. Without rejecting q.m., and without accepting hidden variables, we argue that the intuitive
notion of locality is fuzzy and unreliable, while difficulties in the mathematical formulation of the notion
of locality make 'non-locality' merely a disreputable tag. Finally, we point out that nothing in classical
physics justifies the polemic of spookiness applied to non-locality - for several centuries, locality has
remained a desirable metaphysical requirement, rarely satisfied by physical theory.

1 The topology of time: a general introduction to Chapters VIA, VIB, VII

I Nofthis exposition two problems have so far been studied in some detail: the problem
the measurement of time and the problem of the arrow of time. Both problems
were studied in the context of classical (non-quantum) mechanics. But an implicit
assumption in both cases was that time is like a line - that time may be represented by
a real variable, t, so that time is like the real line. The evolutionary equations of physics
(including the SchrOdinger equation) involve derivatives with respect to time. This
notion of a derivative with respect to time involves a limiting notion, and usually
presupposes that time has the topology of the real line.
More generally, partial differential equations, involving a derivative with respect
to time, could be written down on a spacetime manifold - but this is not critical just
now. A spacetime manifold is locally modeled on Minkowski spacetime. Issues pertain-
ing to the global topology of time will be taken up in Chapter VII.
Could the (local) topology of time in the real world be different from that of a line?
Indeed, the topology of time that one assumes in everyday life is different from that of
a line. In everyday life one supposes that the future is uncertain, hence unknown, and
that the past is certain, even if unknown. (This, incidentally, is different from the
thermodynamic characterization of the future as more uncertain than the past.)
Using relatively recent developments in logic,l these beliefs about past and future
may be recast into a belief about the structure of time, or into a belief about the nature
of the logic best-suited to describe the real world. The truth-value of assertions about
the past ('It was the case that it rained here yesterday') is decided, even if one does not
know what it is. On the other hand, the truth value of assertions about the future ('It
 140 CHAYrnRVIA

will be the case that it will rain here tomorrow') is undecided because it is intrinsically
unknowable. (Supposing that the truth-value of assertions about the future is decided,
but only unknown, leads to paradoxes like the sea battle, discussed in Chapter I.) There
are two possible future worlds: one in which it rains, and one in which it does not. (These
are possible logical worlds, and one can talk about them without committing oneself to
their actual physical existence.) There is only one past world: one in which it either
rained or did not rain. The situation is sketched in Fig. 1.
Fig. 1: Mundane time

/ On the mundane view, time is be-

lieved to be linear towards the past,
and branching towards the future.
There is only one past, and the pos-
sibilities actually realized are repre-
sented by the thick line. The arrows
represent the several possible fu-
tures, while the thin lines represent
the might-have-been possibilities,
now excluded, presumably due to
IFuture I human choices.

This everyday picture of time is, in fact, implicitly used by physicists. Indeed, our
belief in the validity of physics is based on the assumption that time has the structure
of Fig. 1. That is to say, one believes that the entire past has not silently decided whether
or not a certain experiment will be performed: the choice is ours. On the other hand,
one believes that experimental records are records of the past: one has no choice about
the past. One believes that the past has not changed by the time the results of an
experiment are published.

1. 1 The language of time

One can conceive of other possible structures of time. The examples of fission-fusion
time, cyclic time, and parallel time streams are sketched in Fig. 2. Each such structure
corresponds to a different topology. For example, removing a point from the real line
always disconnects the real line into exactly two components, whereas removing a point
from a tree may disconnect the tree into three or more components. A structure which
admits closed loops is not simply connected, for a closed loop cannot be continuously
deformed to a point, while remaining within the structure.
Such structures can typically be expressed in either of two formally equivalent2 ways
described below.

1. 1. 1 The U-calculus. The problem of time asymmetry concerns an earlier-later

relationship. But one must recognize that there may be different types of earlier-later
relations. Such relations are formalized in the V-calculus as follows.
Leta, b, c, ... denote 'instants' of time, and let Uab stand for 'a is earlier thanb'. If
p, q, r, ... are propositions let TaP denote 'it is the case at instant a that p'. Taking
two-valued logic as the base, one writes down the postulates
 BElLAND NON-LoCAllTY 141

0
Fig. 2 Some structures of time

-~
Three possible structures of time: (a) fission-
fusion time, (b) cyclic time, (c) parallel time-
streams. Each ofthese stlUCtures corresponds
to a different logic. The vertical line in (a)
represents more complex structures of an in-
stant if time is thought of as parametrized.

1. Ta -p == - Ta p
2. Ta (p=>q) == (TaP => Taq)
together with the definitions
3. Ta G p == V b (Uab => Tb p )
4. TaHp == V b (Uba => TbP)
In ordinary language, statement 1 says, 'not-p is the case at instant a exactly ifp is
not the case at instant a'. Similarly, statement 3 says, 'Gp is the case at instant a exactly
if p is the case at all instants b later than a'.
The V-calculus has been criticized on the ground that it postulates the existence of
possible 'instants' of time which may not actually be realised, as in Fig. 1.

1. 1. 2 The 'minimal'tense logic. The tense logic attempts to remedy this apparent
defect without affecting the results of the V-calculus. Thus, one wants that the theorems
of the 'minimal' tense logic should be exactly those propositions of the V-calculus which
are true at any instant. To this end, one introduces operators like
F: 'It will sometime be the case that ...',
P: 'It was sometime the case that ...',
G: 'It will always be the case that ... ',
H: 'It was always the case that ...'.
While the tense logic avoids assuming the existence of possible 'instants', the semantic
interpretation of these operators, given in quotes, involves possible (logical) worlds, as
in postulates 3 and 4 of the V-calculus.
More formally, one may take either F and P or G and H as logical primitives, and
adopt the definitions G == - F-, H == - P-. There are simple rules of inference, e.g.,
Rule G: to infer Ga from a. The postulates are:
1. G (p => q). => . (Gp => Gq),
2.PGp => p,
together with symmetric rules and postulates for H and P.
The usual notion of an arrow of time implicitly imposes a variety of conditions on
U, such as transitivity, density and linearity. Various conditions that one may want to
impose on the relation U may also be translated into the tense logic. For instance, the
transitivity of U (Uab => Ube => Uac) is equivalent to idempotence of G, i.e., to the
condition Gp => GGp. Similarly the density of U (Uab => 3c, Uac A Ucb) corresponds
to idempotence of F (Fp => FFp). Linearity, i.e., U a total order
[Va,b, (a=b) v Uab v Uba],isexpressedbytrichotomy:PFp=>p v Fp v pp. Similarly,
one can talk about a beginning of time, or time being non-ending etc.
142 CHAPTER VIA

1. 1. 3 The language of time: further considerations. These two equivalent ways may
not be adequate. The order in rational numbers is transitive, dense, and linear; but
rationals do not form the continuum needed for the existence of limits and derivatives,
i.e., the topology of the rationals is not complete. So, all properties implicit in the
metaphor of an arrow of time cannot be captured by means of an earlier-later relation
alone.
A binary earlier-later relation is inadequate to represent cyclic time: the binary
relation must be symmetric and reflexive (and also transitive), so that it becomes an
equivalence relation, and is unable to distinguish between a permutation of points on
the circle. A four-place relation S(a,b,c,d), read as a and b separate c and d, is required.
Correspondingly, instead of one-place operators like F, P, 3-place operators are
needed: 1J pqr, read 'p, q, r are true at tp, tq ,t" where to, tp separate tq, t,.
More importantly, the measurement of time being essential for physics, one may
want to say that each of two possibilities refers to the same 'instant of time', i.e., the two
are simultaneous. With more complex structures of an instant, one needs to reject the
first two postulates of the U-calculus,3 and use logics other than two-valued logic as a
base. For example, if points on the vertical line in the first diagram in Fig. 2 belong to
the same 'instant', then we need the quasi truth-functional logic described in Chapter
I. If the present and past too have structure, one would also need to grant physical
existence, in some sense, to the possible worlds.

1. 2 A new logic for physics

What would be the consequences, for physics, of a non-trivial structure of time? Clearly,
one needs a new mathematics - the calculus is obviously limited. A non-trivial
structure of time means that the logic used to formulate physics would have to be
different from the usual logic. Are there any reasons to suppose that physics needs a
change of logic?
Clearly, there are a-priori grounds. If the notion of a derivative assumes that time
is (locally) linear, existence and uniqueness theorems for differential equations guar-
antee that future and past are logically dependent on the present: that is, time must also
be logically linear. A time which is both logically and topologically linear may be called
superlinear - the formulation of physical laws as differential equations forces time to
be superlinear. There is unquestionable incoherence in using (past-linear, future-
branching) mundane time to validate a physics which cannot conceive of time as other
than superlinear.
There are other grounds as well. A change of logic does seem to be forced by the
traditional axiomatic formulation of quantum mechanics.

1. 2. 1 Structured time and the tilt in the arrow of time. The question of the change of
logic necessitated by q.m. is examined in more detail in the next chapter. But if this
change of logic is seen as arising from a structure of time, there are other implications
and possibilities that need to be explored. Quantum mechanics applies at the micro-
physical level, where classical mechanics fails. Could it be that the failure of classical
mechanics (including the believed failure of relativity) at the microphysical level is due
 BELL AND NON-LoCALITY 143

Box 1. 1bne in q.m.: some recognized new features

It has been recognized for some time that quantum mechanics introduces some new
considerations into the problem of time. It is interesting to quickly review some of these
considerations against the background of questions about the local and global topology of
time.

The measurement oftime. In order to make the measurement of time meaningful, in relativity,
space and time intervals are allowed to intenningle freely. In q.m., treating space and time on
an equitable footing presents a problem. like position, one would like to treat time as an
'observable', or a dynamical variable: as commonly put, the fourth uncertainty relation,
n,
I'1Ellt ~ cannot be quantized. Even if one allows negative energy states, one wants the
energy operator, the Hamiltonian H, to be bounded from below. In that case, there is no
operator T, with spectrwn a(1) contained in the real line, which satisfies the commutation
relations [H,TJ = in. One could interpret the absence of a time operator as follows: there
cannot be any observable corresponding to the 'linear' notion of time. That is, one need not
necessarily reject an operator which is not self-adjoint, but which satisfies the commutation
relations and keeps the Hamiltonian bounded from below.

The arrow oftinut. As already mentioned in Chapter N, in a quantized theory one expects
the CPT symmetry rather than simple time symmetry or T-symmetry. One consequence of
this is that one may regard the asymmetry between matter and anti..:matter as a manifestation
of time asymmetry. Another feature is that processes which violate T-symrnetry, or CP
symmetry, such as the K-meson decay, are known. Perhaps this allows for the definition of a
microphysical time-asymrnetry.36 Wewill reconsider the relevance ofCP-violation in Chapter
vn.
It has also been suggested that the measurement process, unique to q.rn., could be a
further source of time asymmetry. Evolution according to the SchrOdinger equation is
reversible, but the act of measurement results in an irreversible collapse of the wavefunction
- looking at Schrbdinger's cat might kill it
In this context, it seems worth pointing out that the measurement postulate fixes the
probabilities of the outcome only after the measurement process has been chosen. Quantum
mechanics itself says nothing about how this choice is arrived at That is, the measurement
postulate implicitly assumes the structure (Fig. 1) of mundane time (in supposing that the
observer can decide what to measure). Now, a non-trivial topology of time can itself be a
source of time-asymmetry, as with the mundane distinction between past and future.
Therefore, the belief that the measurement process is a new source of time asymmetry only
exposes the incoherence in assumptions about the structure of time in q.m.
.-------\
I (continued onp 1502
to neglect of the possibility that time in the real world has a non-trivial microphysical
structure?
If the answer is 'yes', this suggests a new way of looking at the relationship between
time in relativity and time in q.m. .The previous chapter concluded by recommending
the hypothesis of a tilt in the arrow of time: interactions mostly propagate from past to
144 CHAPTER VIA

The global topology o/time. The problem of time in q.m. has recently come to the
forefront37 because of the great interest in the quantization of gravity. We have already
seen one instance of the difficulties, in the grandfather paradox: perturbations of a
'normal' spacetime unavoidably result in a spacetime with closed timelike curves, in
which some part of the future blends into the past and a 'time machine' may be
constructed and maintained. This sort of behaviour suggests a cyclic structure of time
(Fig. 2b). This is an aspect of the global topology of time, taken up in Chapter W.

Other peculiar behaviour can occur. For example, overlooking the conceptual chaos
in applying a theory back to its logically primitive notions, the naive and direct application
of q.m. to spacetime results38 in a 'frozen universe' which perpetually stays in the ground
state of the Hamiltonian. In such a universe, nothing happens; the whole concept of
change, and with it the notion of time, seems destroyed.

future, but in a tiny fraction of cases they may propagate from future to past. With this
hypothesis, the many-body equations of electrodynamics are functional differential
equations with mixed deviating arguments. Such equations do not admit a unique future
solution even if the entire past history is prescribed. Instead of looking for ultra-spe-
cialized4 situations under which such equations admit unique solutions from point data,
it seems better to accept the non-uniqueness as inherent. Even though time is treated
as a parameter in writing down these equations, the non-uniqueness of the future, given
the entire past, shows that, with a tilt in the arrow of time, time ought to be assigned a
non-trivial logical structure. (That is to say, with these equations, though one continues
to treat time as a (linear) parameter, time is not any longer logically linear; time is
parametrized, but 'instants' have a non-trivial structure.) If the tilt in the arrow of time
is small, it would be reasonable to conjecture that the resulting structure would usually be
confined to microphysics. This would also be a step towards resolving the incoherence
between a logically asymmetric mundane time and the linear time of physics (in a way which
is not time-symmetric like the argument from chaos, considered in Chapter II).

1. 3 Objectives

This suggestion that time may, in fact, have a non-trivial microphysical structure, that
this structure may be related to the change of logic required by q.m., and that this
possibility may be 'sensed' from within classical mechanics, raises a whole host of issues,
of which we will explore only a few in this exposition. Certainly, the larger issue of
incoherence between mundane time and time in physics may be postponed. Accepting
that the nature of time is fundamentally different in q.m., since the Schrodinger
equation continues to use the notion of time as a parameter, there is a possibility that
the notion of time within q.m. is incoherent. One could think of limiting oneself to
questions of this sort, which lie unquestionably within physics.

Before proceeding to answer these questions we need a detour: the proposed tilt
in the arrow of time, in addition to implying a non-trivial structure of time, rather
obviously implies non-locality of some sort. Is 'non-locality' physically acceptable?
 BElL AND NON-LoCALITY 145

1. 4 Non-locality

It seems natural to consider the question of non-locality in the context of Bell's

inequalities: the aspect of q.m. which has been heavily discussed in recent times is the
apparent non-locality of q.m. which seems to result from Bell's inequalities, and
Aspect-type experiments. Though Bell's inequalities concern q.m., the force of the
'locality' requirement still derives from the apparent conflict between non-locality and
classical macrophysical intuition.
One could object to this procedure of discussing non-locality in the context of two
distinct theories - classical mechanics and q.m. - on the ground that the problem of
time in q.m. is known to involve essentially new considerations. However, it is possible
to view most of these new considerations against the perspective of a possibly structured
time (see Box 1). It is in fact necessary to do so if one is to take seriously the above claim
that the structure of time in q.m. is fundamentally different.
The idea that time might have a non-trivial structure also provides a new perspec-
tive to the debate over Bell's inequalities. The inequalities concern hidden variable
theories which, in a way, reject the reality of quantum-mechanical behaviour, and
suggest that q.m. is a substitute for a 'complete' theory along the lines of classical physics.
The question here is not so much one of a 'deterministic sub-quantum theory' or of
'reality' - the basic idea is that reality is definite. Something either happens or it does
not: at any instant ScbrOdinger's cat is either dead or alive, not in some strange limbo
in between.
That is, the 'realism' underlying hidden-variable theories may be seen as an issue
about the logic and structure of time. According to hidden variable theories, in reality,
something either is or is not the case - there are no in-between cases, as in postulate
1 (and, less obviously 2) of the U calculus.
On the other hand, the opponents of hidden-variable theories have argued in a
variety of ways, the most recent argument being that from locality - if a hidden variable
theory is to reproduce q.m. it must be non-local in some sense, so that the notion of
time in (hidden-variable) q.m. is somehow different from the notion of time in relativity.
Now, the suggestion that time in q.m. has a non-trivial local structure, and that this
structure relates to non-locality of some sort, suggests, further, that q.m. ought to be
non-local. It is easy to confound this suggestion with the link between hidden variables
and non-locality proposed by Bell. Since both - the earlier suggestion and Bell's
inequalities - seem to point in roughly the same direction, it is tempting to suppose
that experiments on Bell's inequalities show that q.m. is non-local.
Below we review the entire argument about hidden variable theories, and try to
correct this impression. We also try to correct the popular misconception that experi-
ments have conclusively demonstrated a violation of Bell's inequalities. The world (and
q.m.) may be non-local, but this does not convincingly follow from experiments on Bell's
inequalities. Apart from seeing the debate in a fresh perspective, the other objectives
are to develop some material needed later on, and to clarify the issues involved, without
either accepting hidden-variable theories or denying non-locality.
Bell's inequalities put the hidden-variable theorists' metaphysical demand for
'completeness' or 'realism' in opposition to the (metaphysical) requirement of 'locality' .
We argued above that the requirement of 'realism' is really a demand for linear time.
146 CHAPTER VIA

It is equally important to recognize the fuzzy and metaphysical nature of the demand
for locality, because any denial of locality has often been equated with 'spookiness' of
some sort, and our basic hypothesis implies a structured time while rejecting locality in
some sense. We try to bring out the fuzzy and metaphysical nature of the locality
requirement by raising the following questions. (a) Is the intuitive notion of 'locality'
reliable? (b) Does the intuitive notion of 'locality' admit a unique and precise formu-
lation? (c) Does the experience of classical physics support the belief in locality?
We argue that the usual macrophysical intuition about locality involves irrelevant
thermodynamic considerations; moreover, the accompanying implicit notions of
'localizability' (and/or 'contact') become problematic at the microphysical level. The
intuitive notion of locality is fuzzy because in some phenomena, such as the Aharonov-
Bohm effect, 'locality' is intuitively 'violated', but the explanation for these phenomena
does not require us to abandon locality. On the other hand, Bell's locality requirement
may be violated in situations where there is no intuitive violation of locality.
We speak of Bell-locality because the notion of locality admits a diversity of
apparently unconnected and possibly conflicting mathematical formulations. In the
absence of a unique and precise formulation, the notion of 'locality' is not falsifiable
and, hence, is metaphysical.
Question (c) is relevant because Bell's locality requirement is motivated and
justified by classical considerations and in classical terms. Putting together the various
comments about non-locality scattered throughout the earlier chapters provides an
answer to (c): it turns out that 'locality', far from being a well-established principle of
classical physics, has always been a fuzzy and metaphysical demand that has often been
raised, with a signal lack of success. Therefore, non-locality, by itself, does not make the
world a 'spooky' place.

2 Background to Bell's inequalities

2. 1 Basic experiments

2. 1. 1 The two-slit diffraction experiment. Let us first clarify some of the terms used in
the debate, using the two-slit diffraction experiment. Consider an 'electron gun' which
fires electrons, if necessary one at a time, at a barrier in which there are two slits. Some
of the electrons pass through the slits, and are received on the screen. The screen may
consist of a scintillation counter, which shows a bright dot at whatever place the electron
strikes it. The resulting pattern is an interference pattern, with alternating dark and
bright regions.
The two-slit diffraction experiment was first performed to establish the wave theory
of light. On the wave theory, each wave coherently splits into two parts, half of it goes
through slit A, and the other half through slit B, and the two interfere to produce the
two-slit diffraction pattern.
But electrons are particles, not waves. So, at best, one may imagine that if a number
of electrons are fired simultaneously, half of them would go through slit A, and the other
half through slit B. The puzzling feature of the experiment is that the diffraction pattern
 BELL AND NON-LoCAUIY 147

is obtained even if the electrons are fired one at a time. How does the electron know
that it should not fallon the part of the screen which remains dark?
The remote possibility that the electron splits into two, with half going through each
slit, can be eliminated by experiment. By installing an electron detector at or just behind
each slit, we can know precisely which slit the electron goes through. And, indeed, each
electron is found to go through either slit A or slit B. No electron goes through both
slits at the same time.
But something has changed. We no longer get the diffraction pattern. Instead, we
get a bullet-shot pattern, a superposition of two Gaussians. There are no alternating
dark and bright regions; rather, there is a fairly smooth reduction of intensity out from
the centre. This is just the sort of thing one would expect to get if electrons were like
bullets going through the slits.
All this suggests that, so long as no one is looking, the electron may (and perhaps
does) split into two, half going through either slit, and the two halves re-assembling on
the other side. The moment we look at it, the electron neatly folds up into itself, and all
of it discreetly passes through one and only one slit.
This is usually regarded as the beginning of the real quantum mechanics. The way things
are seems to depend on the way in which one looks at them: the outcome depends upon the
measurement process. Merely looking at the electron seems to collapse it from wave to
particle. Looking at Schrodinger's cat kills it. This sort of behaviour is called the collapse or
reduction of the wave function. (A more precise formulation is given in Chapter VIB.)

2. 1. 2 Quantization of electron spin. Another phenomenon which illustrates this,

paradigmaticallyS rather than generically, is the electron spin. One may imagine that
the electron spins somewhat like a top, except that the way in which the top spins does
not depend upon the way in which one looks at the top. Like a top, the electron may
have its spin in any direction, provided no one is looking; but, upon measurement, the
spin of an electron always is either 'up' or 'down', regardless of the particular direction
in space we choose to call 'up'.

2.2 Interpretations

2. 2. 1 Hidden variables. Let us try to form a classical explanation for this mysterious
behaviour. Since the electron is electrically charged, and because it spins, it behaves
like a tiny magnet (current loop). Now, if a big magnet is brought near a small magnet,
the small one will jump and align itself either parallel or anti-parallel to the big magnet.
Electrons being the tiniest of known magnets, if a big magnet (or a strongly inhomoge-
neous magnetic field) is applied to a beam of electrons, we find that all the electrons
are lined up with their spins either along the big magnet (spin 'up') or against it (spin
'down'). No matter what direction we choose to call 'up' (i.e., no matter in what direction
we orient the big magnet) electrons will always be found upon measurement to have
spin 'up' or spin 'down', never spin 'sideways'.
The experiment we have just performed is called the Stern-Gerlach experiment6
and the 'explanation' we built in may be called the 'clumsy fingers' explanation. That
is, electrons are so small, and our fingers are so big and clumsy that the moment we
poke the electron to find out where, or in what state, it is, it jumps like a quantum, or
148 CHAPTER VIA

perhaps withdraws into its shell like a startled turtle. This is also the kind of explanation
that is offered in many texts for the Heisenberg uncertainty principle.
This kind of thinking suggests that there is no question about the reality of the
existence of 'sideways' spin; it is merely the case that q.m. provides an incomplete
description of reality. There are hidden variables describing the 'real' state of objects
when no one is looking; these variables also describe the interaction of objects with the
measuring apparatus.

2. 2. 2 The Copenhagen interpretation. The other kind of thinking, generically known

as the 'official' or Copenhagen interpretation of quantum-mechanics, is that there is
nothing particularly strange about the tortoise-like behaviour of the electron. The
physicist's job is to describe the way things are, when measured, and not the way they
might have been when no one was looking. A scientific theory, and quantum mechanics
in particular, is concerned only with calculating and predicting the results of experi-
ments. In this respect quantum mechanics works. The ultimate nature of things
(whether or not the electron chooses, of its own 'free will', between spin 'up' or spin
'down') is a matter best left to philosophers. (The last utterance must be accompanied
by a contemptuous gesture with the hand!)
Now, physical theories are undoubtedly marvelous instruments for predic-
tion. But the instrumentalist doctrine - that physical theories are no more than
instruments for prediction - was first adopted by Copernicus, perhaps in order
to avoid conflict with the Christian Church. He did not assert that the Sun really
was at the centre of the solar system; rather, he asserted that for predictions
connected with planetary orbits it was more efficient to suppose that the Sun was
at the centre. The Copenhagen interpretation goes a step further in asserting that
it is irrelevant for physics to decide what really is the case.

2. 3 Controversy: The EPR Paradox

Many of the founding fathers of quantum theory, de Broglie, Schrodinger, Einstein and
Wigner for example, revolted against this kind of thinking: that the tortoise-theory of
the electron was the ultimate. They formulated their paradoxes, like Schrodinger's cat
(see Box 2), Wigner's friend, etc. The Einstein-Podolsky-Rosen (EPR) paradox,? as
presented in a thought-experiment due to D. Bohm, 8 is the following.
Consider a system of two electrons with oppositely oriented spins. Suppose that the
two electrons now fly apart. By conservation of angular momentum, they will continue
to have oppositely oriented spins. So, by measuring the spin of one electron, we can
know the spin of the other electron, without disturbing it in any way. This is the first
idea behind the EPR paradox: that it is possible to peep at a system discreetly, without
disturbing it in any way. So our fingers may not be so big and clumsy after all.
The paradox proper is obtained by supposing that instead of having measured the
spin of electron 1 in the direction in which we did (x-direction, say) we could well have
measured it in the y-direction. As before, this would have amounted to an implicit
measurement of the spin of electron 2. Since electron 2 is not disturbed in any way, it
has no way of knowing the direction in which we will choose to measure the spin of
electron 1. But after having measured the spin of electron 1, in any direction, one can
predict the result of measuring the spin of electron 2. This suggests that the spin of
 BElL AND NON-LoCALITY 149

electron 2 (hence that of 1) was determined prior to the measurement process, although
the measurement process may have altered this direction.
According to the original statement of the paradox, there must exist 'elements of
reality' corresponding to the spin of electron 2 in the x and y directions, since the spin
of that electron, in these directions, can be predicted with certainty, without disturbing
the state of the electron. Since these are complementary observables, q.m. does not
specify their simultaneous values and is, therefore, incomplete.
In brief, the EPR paradox tries to show that (a) our 'clumsy fingers' may be made
adroit, and (b) that 'God does not play dice', or at least that he does not choose spinning
electrons as his dice. Thus, it would seem that the peculiarities of q.m. arise because it
is an incomplete substitute for a deterministic sub-quantum theory (hidden-variable
theory), which would explain the details of the collapse of the wave-function.

2. 4 Classical mechanics as a hidden variable theory

What is a hidden variable theory? The idea may be understood as follows. The classic
picture of a deterministic (or local realistic) 'hidden variable' theory, underlying a
statistical theory, is provided by classical mechanics seen as the substratum for thermo-
dynamics. The exact state, or pure state of a macrophysical system is given by a single
point in 6N-dimensional phase space (r-space), corresponding to the exact values of
the 3N positions and 3N momenta of N particles. But statistical mechanics describes
this by a 'mixed state' or a probability distribution (or a swarm of imaginary points)
which represents our ignorance about the exact state of the system. From the hidden-
variable point of view, the probabilistic nature of q.m. is to be understood by supposing
that there is a missing hidden-variable theory which is to q.m. as classical mechanics is
to thermodynamics.

2. 5 History: the no-hidden-variable theorems

2.5.1 von Neumann's 'theorem'. Long before the first such hidden-variable theory
was formulated, von Neumann in 1933 sowed the seeds of confusion by pretending to
have proved that, in the case of q.m., any such hidden-variable theory was bound to fail
in the sense that it would be internally inconsistent. Von Neumann argued that quantum
probabilities are fundamentally different, and cannot be derived from classical proba-
bilities (which represent ignorance about the exact state). Hence, such an underlying
theory was impossible for q.m.
The, argument runs as follows. Suppose it were possible to associate a classical
random variable a(l) with the quantum mechanical observable (treated as synonymous
with a self-adjoint operator)A. One would expect this theory to reproduce q.m., on an
average, so that the average value of the two would be the same:
E(A) = E(a(l = a, say, whereE denotes the mathematical expectation, or averag-
ing operator. Thinking of a as a sharp-value, or the 'real' value, it is immediately clear
that we require also that E(A2) = a 2. Moreover, one would expect linearity:
E(A+B+C) = a+b+c, whereb andc are associated in the same way with B and C. In
fact, extending the relation between A and a to polynomials and imagining a passage to
the limiting case of continuous functions, one could even require that
150 CHAPTER VIA

E(g(A = g(a), for any continuous function g. Von Neumann proved that this was
impossible. He concluded that hidden-variable theories did not exist.
It was pointed out later that von Neumann had assumed rather more than was
warranted. The problem lies with the trivial-seeming assumption of linearity. IfA and
B are complementary (non-commuting) observables, an experiment designed to mea-
sure A would, in general, interfere with the measurement of B: a measurement of
position interferes with the measurement of momentum. Therefore, the assumption of
linearity fails.

2. 5. 2 Gleason s theorem and context dependence. Kochen and Specker9 later gave a
similar proof, without assuming linearity for non-commuting observables. In between,
Jauch and PironlO and Mackeyll had proved similar results. Mackey's proof is based on
Gleason's theoreml2 : ifll is a measure on the closed subspaces (projection operators)
of a Hilbert-space, with dim~3, then there exists a positive trace-class operator (density
matrix) p, such that Il(S) = Tr(pPs), for every closed subspace S, where Ps is the
orthogonal projection onto S.
Since every Il, hence every averaging procedure, on a Hilbert space is obtained in
this way, to avoid trouble with Gleason's theorem, a hidden-variable theory must be
context dependent: there must be hidden variables also for the measurement apparatus.
The outcome of an experiment depends upon 'the complete disposition of the measure-
ment apparatus'. In current terminology, Gleason's theorem rules out only the non-
contextual hidden variable theories.13

2. 5. 3 Non-locality of Bohms theory. In a way von Neumann was right: hidden-vari-

able theories, in fact, came into existence much later! Despite the impressive list of
theorems against it, and the complexity of some of the proofs, Bohm's hidden variable
theoryI4 showed no signs of being internally inconsistent. However, Bell,IS while sur-
veying the no-hidden-variable theorems, observed that the theory did have a physically
objectionable feature to it: it was non-local. Given two particles, the evolution of 1 is
closely intertwined with the evolution of 2 however far apart they may be.
This line of thinking suggests a resolution of the EPR paradox 'in the way which
Einstein would have liked least'. The EPR paradox assumed 'locality': since the two
electrons are separated, perhaps by a considerable distance, an experiment performed
on one cannot disturb the other. Could this mean that all hidden variable theories were
non-local? If they are, why should this be cause for concern?

2. 6 Metaphysics: locality and the voodoo principle

Part of the reason for concern is that the 'principle of locality' is the popular opposite
of the 'voodoo principle': to cause serious harm to a person, obtain a lock of hair, or
some personal effect, attach it to his likeness, and smash his likeness to bits. The idea
behind the voodoo principle is that since the lock of hair was, at one time, associated
(correlated) with the person, any damage inflicted on the lock of hair is bound to have
serious repercussions on the person, at the precise instant the lock of hair is damaged,
without the necessary intervention of a chain of causes.
On the other hand, 'locality' means that systems that are spatially separated cannot
influence each other, without the intervention of a chain of causes, and in less time than
 BELLAND NON-LoCALfIY
Box 2. Schrooinger'S cat
Most of the dissatisfaction with quantum
mechanics can be traced to von
Neumann's (in)famous projection postu-
late: that a quantum system evolves ac-
cording to Schrodinger's equation until
one makes a measurement (the measuring
apparatus being classical) upon which the
state vector of the system is reduced (dis-
continuously) to one of its eigenstates.
1his dissatisfaction bas found expression
in paradoxes dealing with the measurement
process. ScbrOdinger's cat is one such paradox,
and not Sch.r6dinger's pet, as might be im-
agined.
Consider a double-slit experiment, wilh
electrons, leading to a diffraction pattern The
diffraction pattern persists even if one fires
electrons so slowly that they go through the slit
one at a time. But if one observes which slit
each electron goes through one gets a bullet-
shot pattern instead of a diffraction pattern
The situation may be described by saying that
the electron mysteriously splits into two, half
going through each slit The mystel)' is that
upon measurement we see only full electrons
going through either slit (but then one no
longer sees the diffraction pattern).
SchrOdinger's cat aggravates this mys-
tery (see figure). The cat is kept in a box
containing a bottle filled with the deadly
chemical MICE (Methyl Iso-CyanatE). A
photon falls on a half-silvered mirror. Half
the photon is reflected, and half is transmit-
ted. (Upon measurement, of course, we
would find a photon which is reflected or a
photon which is transmitted.) Hthe photon is
detected behind the screen, a plunger breaks
the bottle and releases the MICE. We will
suppose the MICE can kill the cat in one
minute. Suppose also that we isolate the
whole system - photon source, half-silvered
mirror, plunger, MICE and cat - so that it
evolves quantum-mechanically. What is the
state of the cat after two minutes?
151
The cat is obviously half-dead + half-
alive. H perchance we look and happen to
s:-
~~
:
,I
:,,
:q~1 ;
., ..
,,
,f'
,.
\
, ,
,,, .
j ;
'.
"\ ScIYOcfngers
;
ir Plngef
\
Da1gerous
cat
MICE
find a dead cat, it is the act of looking (the
measurement) which caused the cat to
go into its eigenstate. Hence the old
adage: curiosity killed the cat.
Note that quantum-mechanical proba-
bilities apply also to single systems. H we
observed ten cats we would, of course, ex-
pect to find 5 dead and 5 alive. But each
individual cat is dead with probability half
and alive with the same probability. This
probability is not merely due to our ignorance
(five mental copies of the cat dead and five
alive) since it can produce objective diffrac-
tion patterns. But the probability does dis-
sapear when we attempt to remove our
ignorance by measurement.
Where does one draw the line between
~c rnea5llring apparatus and a quan-
rum system? Howshouldonedoquantumgrav-
ity? What comtitutes a measuring apparatus?
Should one suppa;e, like Wigner, that con-
sciousne&'i is the ultimate measuring appamrus?
Shoold one reformulate theprojectionpaffillate
(or fonnulate it for otservables with continuous
spectra)? Should one suppose that these myste-
rious probabilities are due to some unknown
hick:len wriab1es? There are many answers and
one undecided <at
P .S. The above material is based on a talk
given at the Schrodinger Centenary Sym-
posium organized by the Centre for Ad-
vanced Technology, Indore.
152 CHAPTER VIA

Box 3. Schrodinger's cat and SQUIDs

There is a belief that quantum-mechanical effects will, like preacceleration, forever

remain confined to microphysics. Schrodinger's cat, seen as a macrophysical mani-
festation of quantum-mechanical effects, strikes at this belief. Recent techno-
logical developments suggest that the cat is more than a metaphor - one may reach
out to the cat through a new variety of 'fish' called SQUIDs.
SQUIDs are Superconducting Quantum Interference Devices. It has been
suggested that Schrodinger's cat may be put to empirical test using SQUIDs. A
SQUID ring is macrophysical, with a diameter around 1 cm. According to q.m., a
SQUID placed in a properly adjusted magnetic field can occupy one of two
degenerate states, corresponding to the current circulating clockwise or anti-clock-
wise. The current is typically of the order of a few micro-amperes, so the two states
are macrophysically distinct. However, according to q.m., the SQUID, like
Schrodinger's cat, generally lies 39 in a superposition of the two states. The validity
of this approach has been questioned. 40

a light signal takes to cover the distance: a 'signai' must propagate outwards from the
damaged lock of hair, and there must be a small time-lag between the damage inflicted
on the lock of hair and damage accruing to the person! (Of course, if no such signal is
found to emanate, nothing can happen to the person.)
Bohm's hidden-variable theory does not satisfy the 'principle of locality': the state
at one point of spacetime is inextricably intertwined with the state at another point.
Thus, locality fails for Bohm's theory, and Bell's idea was that 'locality' would fail also
for any other hidden-variable theory. On the other hand the EPR paradox assumed
locality: the result of a measurement performed on electron 2 cannot be affected by the
kind of measurement we choose to perform on electron 1. This is not the same as saying
that the result of a measurement performed on electron 2 cannot be systematically
related to the results of measurements performed on electron 1. While the two
outcomes may be correlated, the value of the correlation may not depend upon the
choices made by the observer. The outcome of measuring the spin of electron 2 (i.e.,
whether the spin is 'up' or 'down') cannot depend upon the direction in which we choose
to measure the spin of electron 1.

3 Bell's inequalities

3. 1 Outline of inequalities

To establish that any hidden-variable theory must be non-local, Bell used a mathemat-
ical formulation of the locality assumption, which I will call 'Bell locality' . Bell derived
certain inequalities to distinguish between q.m. and hidden variable theories satisfying
the 'Bell locality' assumption. The peculiar nature of quantum probabilities enters into
these inequalities through the form of the quantum-mechanical co"elation function.
 BELL AND NON-LoCALTIY 153

Briefly, consider a system of two electrons which initially had oppositely oriented
spins (singlet spin state). Suppose that the electrons are now separated. Let a be the
direction (unit vector in 3 space) in which the spin of electron 1 is measured, and let
A(a) be the result of the measurement. The result is of binary type:A(a) is a random
variable taking on only two values, 1 say. (One could also assign a third value, 0, in
the non-ideal case, corresponding to a failure to detect the particle.) Similarly, let
B(b) denote the result of measuring the spin of electron 2 in the direction b.
If pea, b) is the correlation between A(a), and B(b), then, according to q.m.,
pea, b) = -a'b . For Bell-local hidden variable theories, Bell showed that:

I pea, b)-pea, c) I s 2 [PCb, b')+P(b', c)], (1)

where a, b is one set of directions along which the spins of electrons 1 and 2 respectively
are measured, and b I, C is another such set of directions.
With the quantum-mechanical correlation functionP(a, b) = -a 'b, however, the
inequality (1) must be violated. For, if we choose a, b, b', c as coplanar vectors, and
b' = b, while a .l.. b, then la'cl s 2 - b 2 - bc = 1 - b'c, sinceb is a unit vector. If 0
is the angle between band c, then ab'c = cos 0, while a'c = cos (0+.:rr/2), so that (1)
would imply Isin 0 I s 1- cos O. This is false for a range of values of 0, and leads to Bell's
conclusions that Bell-local hidden variable theories, which must satisfy (1), cannot
reproduce the quantum-mechanical correlation function.

3.2 Bel/'s locality postulate

Bell's locality assumption is the following. Consider the above system from the point of
view of hidden variables. For the results of measurements, we now haveA(a, b, A) and
B(a, b, A), in place ofA(a) and B(b), where A is the hidden variable.
The objectionable feature is that even when the electrons are well separated, the
outcome A (a, b, A) depends upon the disposition of the remote measuring apparatus,
specified by b. Bell's locality postulate is 'that the setting b of a particular instrument
has no effect on what happens, A, in a remote region, and likewise that a has no effect
on B: A(a, A), B(a,A).'16 With these local forms it is not possible to reproduce the
quantum correlations.
To summarize, Bell's conclusions differed marginally from those of von Neumann:
an explanation of quantum-mechanical behaviour, in terms of the way things are (when
no one is looking) was certainly possible, but it would violate locality.

3.3 Experiments on Bel/'s inequalities

A little later, Clauser pointed out that Bell's inequalities could be tested empirically,
and the enthusiasm to test philosophical ideas empirically was greatly fueled by the
initial disagreement over empirical results, and then over their interpretation. A
number of experiments have been performed since then,17 and the outcome seems to
have been decisively against hidden-variable theories, and in favour of q.m. (in terms
of both theory, and the number of papers published!).
The thinking behind the design of the 'delayed choice' kind of experiments may be
explained as follows. One (possibly fantastic) way to account for the two-slit experiment
154 CHAPTER VIA

is to suppose that the electron sends out a probe, in the manner of a general sending
out scouts, to explore the surrounding terrain. Since the experimental setup is usually
fixed much in advance, the electron has enough time to find out the details of the setup.
Thus, the experimental setup is somehow able to influence the behaviour of the
electron, so that one gets a diffraction pattern in one case and a superposition of two
Gaussians in the other case.
Similarly, the experimental setup may somehow influence the outcome of spin
measurement for a pair of electrons, with oppositely oriented spins: 'the settings of the
instruments are made sufficiently in advance to allow them to reach some mutual
rapport by exchange of signals with velocity less than or equal to that of light.' (We recall
that consistent hidden-variable theories must be context-dependent.) The 'delayed
choice' or 'timing' experiment is meant to eliminate this type of explanation by (ideally)
measuring the spins at space like separation so that the electrons do not know about the
experimental setup beforehand. Thus, the experimentally observed violation of Bell's
inequalities suggests that interactions at spacelike separations are possible.

3. 4 Spookiness: interactions between separated systems

Does this mean that the real world, as described by q.m., is non-local? Many physicists
seem to think so, and it would seem as if there is some justification after all for the
voodoo principle. In particular, if q.m. is non-local, it would seem that one is perfectly
justified in setting up non-local hidden-variable theories!
Indeed, there have been some half-serious attempts,IS to generalize the voodoo
principle to astrology. During or just after the big bang, the molecules constituting our
bodies were quite close to the atoms and molecules of which stars and planets are made.
So, by studying the movements of one, one may deduce the fate of the other!
Such misunderstandings bring us to one factor which must be understood in
connection with non-locality. The usual belief that separated regions of space do not
interact (something called the 'law of conditional independence' by Penrose and
PercivaP9) is essentially of 'thermodynamic' origin. In general, there would be no
correlations between separated electrons because any correlation that might have
existed in the past would have been destroyed by a large number of intervening
interactions with other particles.
This notion of locality has nothing to do with Bell's inequalities: one must distin-
guish between the fake 'non-locality' due to the existence of a correlation between
spatially separated objects (expected under the conditions of the experiment), and the
(also possibly spurious) non-locality due to the peculiar form of the correlation function
in q.m .. The idea is that the particular correlation function, pea, b) = -a 'b, cannot be
reproduced by Bell-local hidden-variable theories.

3. 5 Bell locality: interpretation of experiments

3. 5. 1 The early objections. So while Bell's arguments are plausible, 'the fact that so
much follows from such apparently innocent assumptions leads us to question their
innocence.'2o As in the earlier case of von Neumann, Gleason, etc., the suspicion falls
on the mathematical translation, rather than the physical assumption (in this case
locality).
 BELL AND NON-LoCAUlY 155

Bell's argument has been criticized by a number of authors. Bub,21 for instance,
analysed Wigner's reformulation22 of Bell's argument to conclude that Bell's postu-
late has nothing to do with locality. However, Freedman and Wigner23 refuted this
conclusion. Bell24 has replied to various other arguments, essentially reiterating his
earlier position. We do not intend to discuss here the various formulations of Bell
locality and their criticism.25 Some of the problems are brought out in a simple
example due to Barut and Meystre. 26

3. 5. 2 The argument of Barut and Meystre. Barnt and Meystre27 suppose, for the sake
of argument, that the electron spin has a definite direction as in the classical picture of
the electron as a small magnet (dipole). Upon measurement, the electron spin aligns
itself parallel or anti-parallel to the direction of measurement like a small magnet near
a big magnet (Stern-Gerlach field). They introduce the hypothesis that this decision is
not capricious, but the electron always chooses the nearest direction. This nearest
direction is defined except when the 'real' spin is exactly perpendicular to the direction
of measurement, a case which occurs with probability zero (assuming that the 'real' spin
is uniformly distributed).
The point of this picture is not that it is realistic, but that it provides a hidden-vari-
able theory of a certain kind. Consider for simplicity, the case of anti-parallel directions
of measurement. No matter how the measurement axis is oriented in space, the spins
of the electrons will, upon measurement, come out to be anti-parallel. The separated
electrons do not communicate with each other, and, on the assumptions, this apparently
non-local behaviour comes out as an artefact of the measurement process. But Barnt
and Meystre prove that their hypotheses result in a hidden-variable theory which is Bell
non-local. No one would say that there is anything actually non-local in this picture, so
perhaps Bell did err in his mathematical translation.

3. 5. 3 The argument from efficiency. Another argument, this time one which appears
to have much wider acceptance, relates to the efficiency of the detectors used in the
experiments on Bell's inequalities. On the one hand, only those experiments (like
Aspect et aI's) which used optical techniques to measure the polarization correlation
of photon pairs are relatively acceptable. On the other hand, it has been argued28 that
Bell's inequalities (strictly, local realism) cannot be empirically violated if the efficien-
cies of the photon detectors are less than 82.84%. The actual efficiencies were closer
to 0.1%.
Though the actual argument is more complex, the magic figure of 82.84% may be
obtained as follows. We start with Bell's inequalities (1) in the more symmetric form,

(2)

where P;j = pea;, aj), a;, for i = 1,2 are directions for particle 1, and aj, for j =3,4 are
directions for particle 2. Garg and Mermin show that the necessary and sufficient
conditions for violation of 'local realism'29 may be put in the form:

(3)

where
156 CHAPTER VIA

Box 4. The Aharonoy-Bohm effect

The Aharonov-Bohm41 effect demonstrates that intuition may be an inadequate
guide to questions about locality. In the Aharonov-Bohm effect, the motion of an
electron is affected by the presence of an electromagnetic field in a region from the
which the particle is forever absent. Consider a beam of electrons coherently split
into two and recombined to form an interference pattern. A stationary magnetic field
passes through the region between the two beams. The electrons are rigorously
excluded from the region where the field is: the return magnetic flux also avoids the
region where the electrons are permitted, i.e., one arm of the split beam passes
through a region surrounded by a solenoidal magnetic field.
How is the motion of a particle affected by an absentee field? The theory is
simple. The wave function 1/J depends upon the action S:

(B2.1)

in time-independent form, 1po being the wave function in the absence of the magnetic
field. The action S(x) is the line integral of the vector potential A (x) (in Gaussian
units)

Sex) = --ce J.t A(x') dx'. (B2.2)

The region from which the magnetic field is excluded is not simply connected: a closed
path going through the region surrounded by the solenoid and returning from outside
cannot be shrunk to a point, while remaining in the region from which the field is
excluded. Therefore, the line integral around such a closed path cannot vanish, but
must equal the magnetic flux cI> through the coil. This induces a phase shift

M) = (S2 - SI) _ (.)cI> (B2.3)

h he '

where S2 and S1 are the action integrals along the two arms. The interference pattern,
therefore, depends upon the flux through the coil.
This observable effect depends upon the unobservable vector potential rather
than the field. The effect has been observed repeatedly and, it is believed, decisively,
'with exquisite precision and control of the stray field problem.'42
 BElL AND NON-LoCALITY 157

Pile = _'1_ Pile (4)

2-1/ '

assuming that particle detection depends only on the quantum efficiency 'I of the
detectors. Since the left hand side of (2) has a maximum value 2v2, violation of the
inequality in (3) requires 'I > 2(V2-1) = 0.8284.
The additional assumption, that the observed photon correlations were repre-
sentative, had seemed natural to experimenters who concluded that Bell's inequalities
were violated. Today, the gap seems unbridgeable. While some have attempted to
rework30 the inequalities for the non-ideal conditions expected under experiments,
others31 have gone so far as to argue that the preparation of a system violating Bell's
inequalities is impossible, in principle, on account of the uncertainty principle.

4 Further confusion: other notions oflocality

4. 1 1heAharonov-Bohm effect

Perhaps the diffirulty is that Bell's formulation oflocality does not accurately reflect the intuitive
notion of locality. But the intuitive notion itself is rather fuzzy. There are other situations where
locality intuitively seems to be violated; but one may account for these phenomena without
supposing that locality actually breaks down. In the Aharonov-Bohm effect (see Box 3), the
interference pattern produced by an electron beam variesininteractionwith a rurrent, although
the beam never is in a region where it can (classically) interact with the rurrent Though the
interpretations are varied, it is not necessary to suppose that locality breaks down.

4. 2 Lncality in quantwnfield theory

Another ~ to suspect Bell-locality is the following. If it is true that q.m. is non-local, it would
seem that quantum field theory is non-local, a fortiori Nevertheless, quantum field theory
continues to use explicitly the postulate of 'locality' or micro-causality: field operators at spacelike
separations always comnrute. The postulate is derived from the same belief that an observation
performed at one point cannot alter the factual situation at another point at spacelike separation,
ie., in less time than light takes to travel between the two points. It is ~ble to argue that there is
no real conflict between a non-local q.m. and a local quantum field theory - but this sort of
argument only shows that locality is far from the transparent requirement it is made out to be.

4. 3 Locality and Localizability

Like action by contact, locality tends to assume that the particles are localized. For
example, in speaking of particles at space-like separation, one typically imagines that
the particles are point-like, and that the spacelike separation refers to the separation
between the points at which the particles are localized. In q.m., localizability is prob-
lematic. For example, Hegerfeldt's32 no-go theorem shows that localizability is incon-
sistent with 'locality' or 'Einstein causality'. One may want to accept localizability, for
158 CHAPTER VIA

simplicity for example, so that 'locality' would have to be abandoned. Again, an analysis
of the renormalization problem33 suggests non-Iocalisability and non-locality, at least
at the microphysical level, as the long-term solution.

4.4 Summary

To summarize, the experiments on Bell's inequalities suggest, at the intuitive level, that
q.m. is non-local. But, on the one hand, phenomena such as the Aharonov-Bohm effect
make one doubt the intuitive notion oflocality. On the other hand, from a strictly formal
point of view, these experiments require supplementary assumptions concerning the
observed events, due to low detector efficiency. These supplementary assumptions are
not satisfied by all hidden-variable theories.
The mathematical formulation of locality by Bell is suspect. The argument of Bamt
and Meystre suggests that intuitively local situations may be Bell non-local. On the other
hand, a Bell non-local q.m. seems quite consistent with a local quantum field theory.
The existence of unconnected and possibly conflicting mathematical formulations
makes 'locality' a metaphysical requirement. The clarity which seemed to have emerged
from experiments on Bell's inequalities now appears to have evaporated.
One thing, however, is clear. No amount of experimentation on Bell's inequalities
will serve to falsify hidden-variable theories as such: experiments on Bell's inequalities
say nothing, for instance, about the empirical falsity of Bohm's hidden-variable theory.
The best one can hope for is that with the aid of these experiments one would be able
to attach a tag of non-locality to these empirically successful hidden-variable theories.
Why is this tag a matter of such disgrace?

5 The debate over non-locality in classical mechanics

The Bell-locality requirement concerns hidden-variable theories which share the

world-view of classical physics. But is locality a well-established requirement in
classical physics? Indeed, ever since EPR, non-locality has been equated with spook-
iness. Presumably the background to this polemic was classical mechanics and rela-
tivity. Were Einstein, Podolsky, and Rosen justified in their polemic?
To answer these questions, we review the debate over locality in classical mechanics,
collecting together, in the process, the various strands scattered throughout Chapters I-V.
The possibility that one can learn something from the much older debate over locality in
classical mechanics has been sadly neglected in the whole debate over Bell locality, even
though that notion of locality relates to classical (in fact, macrophysical) intuition.

5. 1 Francis Bacon and magic

We recall that the association of non-locality with witchcraft, voodoo and so on

originated with Francis Bacon, who thought that action at a distance was the appropriate
explanation in such cases (Chapter VA). But there seems no logical reason to continue
with this association. In Francis Bacon's time, there was no clear understanding of the
nature or speed of light. Proponents of witchcraft probably would not mind very much
 BELL AND NON-LoCAUTY 159

if the effects of their magic could travel at the rather large speed of light - after all it
would typically make a difference of only a small fraction of a second. So the presence
or absence of action at a distance makes no difference to magic.

5. 2 Locality, chains of causes, and action by contact

Nevertheless, there has been a widespread and persistent feeling that there is something
magical about the interaction between distant objects without the intervention of a 'chain
of causes'. We have seen that this intuitive idea is linked to the idea of action by contact.
The idea of locality - that an explanation is required for the interaction between
distant objects - is essentially the contrapositive of the idea of action by contact. On
the face of it, 'locality' seems to avoid the trouble faced by 'action by contact' in defining
'contact', thereby leading up to questions of the ultimate structure of matter (Chapter
VA). But this seems a psychological trick. Instead of emphasizing contact and seeking
explanations for interactions between separated objects, 'locality' demands an expla-
nation for interactions between separated objects while conceding that interaction
between objects in 'proximity' needs no explanation. Without going further into the
subtle and intangible metaphysical distinctions between 'contact' and 'locality' let us
look at the situation in physical theory.

5. 3 Descartes and Newton

From the point of view of physics, Descartes aether, which proceeded on the basis of
the philosophical principle of 'locality', was a failure. On the other hand, Newton's
back-calculation of the gravitational force, from Kepler's elliptic orbits, was rather more
successful, even though it sacrificed the principle of locality. The lesson is clear: ever
since Newton, physics has been non-local, though physicists have persistently felt that
a good physical theory ought to be local. That is, in classical mechanics, locality has
persisted as an a priori metaphysical prejudice.

5. 4 The field

The field sought to re-establish the notion of action by contact: a particle 'knew' about
the rest of the world for it was in contact with the field. This proved to be a spurious
notion, for charges at any separation could remain in 'contact' through the Coulomb
field or the gravitational field. Both fields were rigidly attached to their sources, and
extended indefinitely, leading to a complete breakdown in the notion of localised
(point) charges and masses tacitly assumed in Newtonian physics.

5. 5 Hyperbolicity: finite propagation speeds

It is strange that the solution of the problem of time measurement provided by relativity
theory (Chapter llIA, B) led to (a) the rejection of aether, and (b) reinforcement ofthe
chain of causes idea: not only must distant effects be related by an intermediate chain
of causes, but the chain of causes could not propagate faster than the speed of light. The
contradiction between relativity and Descartes' aether is a striking demonstration of
the fuzziness in the intuitive notion of 'locality': though the intuitive notion remained
160 CHAPfERVIA

unchanged, the latest formalization is regarded as the epitome ofgood physics, while the earlier
fonnallsation is rejected as confused, metaphysical, and redundant
More mathematically, the idea (that a chain of causes could not propagate faster than the
speed oflight) could be described as the emergence of the paradigm ofhyperbolicity: physical laws
should be fomru1ated aslo/}1eTbolicpartial differential equations. The equations ofgeneral relativity
were developed on the basis ofthis requirement, which is perhaps whyit seemed so natural to EPR
There are, however, two problems with hyperbolicity. Firstly, the notion is a little fuzzy.
The meaning of hyperbolicity is perfectly clear only for the case of a single second-order
p.d.e. with constant coefficients, such as the wave equation. When the coefficients vary, one
may speak of a local notion of hyperbolicity, in the neighbourhood of a point, by freezing
the coefficients at that point. In the more general case of a non-linear p.d.e. one can only
speak of a local notion of hyperbolicity with respect to a solution.34 In the general case of a
system of non-linear (second-order) p.d.e.'s the formalizations are more tentative. 35
Secondly, the Coulomb-field problem does not disappear - Coulomb-type fields satisfy the
Poisson or Laplace equations which are elliptic rather than hyperbolic. The result is that, even with
the existing formalisations, the equations of classical physics (such as the full Navier-Stokes or
Maxwell's equations) and relativity (the Hilbert-Einstein equations) are not, in fact, hyperbolic, but
are mixed-type p.d.e.'s. (This notion of mixed-type p.de.'s concerns p.d.e.'s which are, say,
hyperbolic in one region, with respect to one solution, and elliptic elsewhere. This notion must be
clearly distinguished from the notion of a p.de. with mixed-type deviating arguments.)
To mix up issues and to go back to the earlier notion of differential equations with mixed-type
deviating arguments, eveninthe simplest case of the wave equation, itisnotclearthathyperbolicity
quite captures the intuitive notion of locality. Thus, a tilt in the arrow of time implies non-locality
and interactions at space like separations, by a combination of retarded and advanced interactions,
within 1o/}1eTbolicity. In admitting a tilt in the arrow of time (howsoever obtained), one gives up
locality. But as we saw in Chapter VA, it is simplistic to imagine that interactions at space-like
separations 'violate' the principle of relativity - such interactions only 'violate' the principle of
relativity + 'causality'. On considerations of relativistic covariance alone, it may be necessary, as in
Dirac's derivation of radiative damping, to use advanced potentials. The requirementof'causality',
inteIpreted as time asymmetIy, of course seems natural in view of the asymmetric structure of
mundane time. But that structure has yet to find justification in physics.

6 Conclusions

The Aspect-type experiments on Bell's inequalities did not conclusively demonstrate

non-locality or a violation of Bell's inequalities, since they require additional assump-
tions on account of the low efficiencies of the photon detectors. In any case, experiments
on Bell's inequalities can, at best, serve to classify the empirically successful hidden-
variable theories as 'non-local'. On the other hand 'locality' cannot be taken for granted,
especially at the microphysical level. At the intuitive level the notion of 'locality' is fuzzy
and the notion has been formalized in unconnected or even apparently conflicting ways.
The notion has no basis in classical physics, since every formulation of 'locality', over
the past four centuries, has failed in classical mechanics!
There is possible incoherence about the topology of time in quantum physics.
VIB QUANTUM-MECHANICAL TIME
ABSTRACf. We present a brief exposition of the orthodox axiomatic approach to q.m., indicating the
relation to the text-book approach. We explain why the usual axioms force a change of logic. We then
explain the attempts to derive the Hilbert space and the probability interpretation from a new type of 'and'
and 'or' or a new type of 'if and 'not'. Included are the Birkhoff-von Neumann, Jauch-Piron, and quantum
logic approaches, together with an account oftheir physical and mathematical obscurities.
Instead of entering the labyrinth of subsequent developments, which seek new algebraic structures while
accepting the old physical motivation, we present an exposition of the structured-time interpretation of
q.m., which seeks a new physical motivation.
We saw in Chapter VB that, with a tilt in the arrow of time, the solutions of the many-body equations of
motion are intrinsically non-unique. In Chapter VIA we had indicated how this non-uniqueness relates to
a change in the logic of time. We now explain how the resulting changes in the logic and structure of time
lead to a new type of 'if and 'not', of the kind required by q.m., while escaping from the criticism which
applies to the earlier 'quantum logic' approaches.
We briefly indicate the analogy between this logic and the temporal logic required for the formal semantics
of parallel-processing languages like OCCAM, and distinguish the structured-time interpretation from
the superficially similar many-worlds interpretation and the transactional interpretation of q.m.

1 Introduction

T HE preceding chapter introduced the problem of a non-trivial structure of time:

the (local) topology of time, in the real world, might be different from that of the
real line. The real-line topology differs from the mundane view of a past-linear future-
branching time, used to demarcate and validate physics. Moreover, there is possible
incoherence about the structure of time, even within physics, as different structures may
be simultaneously implicit.
We explained how the notion of a structure of time could be formalized in terms
of properties of the earlier-later relation (V-calculus) or, more generally, using an
appropriate (temporal) logic.
This chapter deals with two earlier claims (a) that an appropriately structured time
could be related to the change of logic required by the axiomatic formulation of q.m.,
and (b) that the hypothesis of a tilt in the arrow of time implies such an appropriate
structure. The other consequence of the basic hypothesis, viz. non-locality, is hardly a
serious drawback since we saw in Chapter VIA that locality is a fuzzy and metaphysical
requirement which lacks a basis even in classical physics.
To reiterate, the aim of this chapter is to present an exposition of the structured-
time interpretation of q.m. which relates the many-body equations of motion of non-
local classical (relativistic) mechanics, the emergence of a logical structure, or a
non-trivial topology of time,! and the mathematical formalism of q.m.
2 presents an exposition of the orthodox Hilbert-space axiomatics of q.m. and
relates it to the usual textbook approach. 3 explains why the orthodox axiomatic
approach forces a change of logic and goes on to present an exposition of the 'quantum
logic' approach, its relation to the Hilbert-space axiomatics, and its obscurities. The
162 CHAPTERVIB

idea is to distill the body of q.m. to its algebraic and logical 'skeleton'. Finally, 4
presents the structured-time approach, relating a structure of time to the 'quantum
logic' approach. The programme is to reconstruct the algebraic skeleton from the
hypothesis of a tilt in the arrow of time (presented in Chapter VB). It remains to be
seen whether this procedure generates the body of q.m.l
Since the structured-time approach involves inputs from diverse areas at the
frontiers of knowledge (such as the many-body problem, the theory of counterfactuals
and conditionals in general, parallel processing, and temporal logic), an in-depth
understanding may require some mathematical sophistication. But I believe the basic
ideas may be grasped by anyone who has some familiarity with the modern (Lebesgue)
integral and Hilbert spaces. To relate this approach to conventional q.m., we need the
connections established in 2 and 3 below, for which there is unfortunately no
convenient reference.

2 The orthodox formalism of q.m.

The chief features of the orthodox formalism of quantum mechanics are the following.
For those who skip this section: the main point is to get rid of the usual hang-up with
11jJ I 2 as a probability density, and to introduce some form of the projection postulate
to work with.
(i) State space: An abstract separable Hilbert space H to which the state vectors of a
system belong.
(ii)Operator representation: A correspondence between classical dynamical variables
and densely defined (and maximally extended) self-adjoint operators on H, subject to
the restriction that canonically conjugate variables satisfy the commutation relations

~, qJ = iii, (1)

where Ii denotes the Planck constant divided by 2n, the bracket denotes the commuta-
tor, and the operator on the right is a scalar multiple of the identity. In the textbook
approach, in the configuration-space representation, one takes q to be the multiplica-
tion operator, q :1jJ(q) -+ q1jJ(q), and p to be the differentiation operator,
p : 1jJ(q) -+ ili(iJ1jJ/iJq) (with appropriate boundary conditions).
(iii)The probability interpretation: The probability interpretation is closely related to
the operator representation, a fact sometimes obscured by texts. The relation is ob-
tained by means of the spectral theorem: an observable or a self-adjoint operator T
generates a 'resolution of the identity', or a spectral measure or a projection-valued
measure E on H, such that

T = f u(7) AdE(A), (2)

where (J(1), the spectrum of T, is a subset of the real lineR . In general, ifp is a dynamical
variable, and Ep is the corresponding spectral measure, the probability that p lies in the
Borel subset2 A of R is
 QUANTUM-MECHANICAL TIME 163

Pr(pEA) = <Ep(A)1p, tp>, (3)

where < " . > denotes inner product in H, and tp is the state with II tp II = 1. For an
arbitrary tp, <Ep (.)tp, tp> is a regular Borel measure, which is a probability measure
when the state is normalized.
The correspondence with the textbook approach is obtained as follows. In the first
place, the form of the operator (whether q above or in a/ap) is unimportant. A
self-adjoint operator T is really characterized by giving its spectrum u(n, together with
multiplicities. For an observable T, the spectrum u(n k R and corresponds to mea-
surable values.
For, say, the position operator q, if we believe that the spectrum u (q ) = R, with
no multiplicity, the spectral multiplicity theorem3 (Hahn-Hellinger theorem) allows us
to recover the usual configuration-space representation. The theorem provides a
unitary map, between the abstract Hilbert space H and L2 ( u (q == L2(R), which
carries q to the multiplication operator on L2(R), and the spectral measure Eq to the
spectral measure E on L2(R) corresponding to multiplication by characteristic func-
tions. Thus,

Pr (qEA) = <Eq (A) cp, cp> = <XA tp, tp> = fR XA 1/J1P = f A !tp! 2 (4)

recovers the more usual form of the probability interpretation. The uncertainty princi-
ple is an easy consequence of non-commutativity (1) and the Schwartz inequality.
(iv) Schrooinger equation: The SchrOdinger equation describes unitary evolution in
this Hilbert space

tp(t) = U(t) tp(O). (5)

By Stone's theorem, any such (strongly continuous) one-parameter unitary group may
be written as

U (t) = e -iHt, (6)

where H is self-adjoint. Hence, the infinitesimal form of (5) reads:

at
atp = - l'H tp,
(7)

which is the more usual form of the SchrOdinger equation.

Physically, the infinitesimal generator H is identified with the Hamiltonian, or
energy operator. It is an extraordinarily curious fact that, modulo commutativity, the
quantum Hamiltonian.is the same function of the canonical variables as the classical
Hamiltonian (when the latter exists).
(v) The projection postulate: The naive formulation of the projection postulate, for the
paradigmatic observable with discrete spectrum, is the following. 4
' ... any result of a measurement of a real dynamical variable is one of its eigenvalues ... ,
164 CHAPTER VIB

... if the measurement of the observable x for the system in the state corresponding to
Ix > is made a large number of times, the average of all the results obtained will be
< xI ~ Ix > ... ,
... a measurement always causes the system to jump into an eigenstate of the dynamical
variable that is being measured .... '
The result of measuring such an observable always is an eigenvalue. As a
'consequence' of the measurement process, the system is thrown (discontinuously) into
the eigenstate corresponding to the measured eigenvalue. If the same measurement is
repeated immediately, the same eigenvalue results. Von Neumann incorrectlyS sup-
posed that the postulate could be generalized in a straightforward way to observables
with continuous spectra.
The precise formulation, even for an observable with discrete spectrum, is messy.
One must account for incomplete measurements, possible degeneracy (non-zero mul-
tiplicity in the spectrum), and superpositions. The last can be achieved by viewing states
as density matrices (trace-class operators, provided by Gleason's theorem, normalized
to trace unity; this theorem was described in Chapter VIA).
Now, if Tis an observable with discrete spectrum, the integral in (2) reduces to a
sum,

T = L AjPj, (8)

where Aj are the eigenvalues and Pj are the corresponding eigenprojections. Let
A E BR (A is a Borel subset of R). The act of measurement conditioned on the statement
TEA (partial measurement) transforms the state p to the (non-normalized) state p',

p' = L PipPi, (9)

AjEA

the sum being taken over those indices i for which AiEA. In the case where A isa
singleton, A = {lJ. (9) reduces to a complete measurement, and the system is thrown
into an eigenstate. This process does not work if the observable is degenerate.

3 From quantum logic to the formalism of q.m.

3. 1 Non-commutativity and non-existence ofjoint distributions

The relationship of the axiomatic approach to the textbook approach may be clear, but
the relation to phenomena remains a mystery. Understanding this mystery may require
a long process of distillation.
What are the chief new features of this formalism? We have already encountered
the Hilbert space and the picture of unitary evolution in the context of classical
statistical mechanics (Chapter IV, Box 3). One could extend this picture to represent
dynamical variables by self-adjoint operators.
 QUANTUM-MECHANICAL TIME 165

The new feature however is non-commutativity: the approach of classical statistical

mechanics always results in a commuting algebra of observables. Since non-commuting
operators cannot be simultaneously diagonalized, this non-commutativity gives rise to
a peculiar difference between classical and quantum probabilities: a joint probability
distribution does not exist for canonically conjugate (non-commuting) dynamical vari-
ables, as observed6 and later proved by Wigner.7 The chief problem then would seem8
to be the explanation of the origin of these peculiar quantum probabilities.

3. 2 Need for a change of logic: failure of the distributive law

Now probabilities may be defined on a a-algebra M of subsets of a given set X, using

U, n and ~, or on a logic of sentences. In the latter setting, the usual measure-theoretic
approach to probability is recovered by identifying the usual set-theoretic operations
with the logical operations required to define them: 'not' with complement " 'and' with
n, 'or' with U, and ~ with~. The usual calculus of sentences results in a Boolean
algebra isomorphic to the algebra of subsets of a given set.9 In the Birkhoff-von
Neumann (BN) approach10 the peculiarities of quantum probabilities are explained by
asserting that the logic of q.m. differs from classical logic in that the distributive law
between 'and' and 'or' fails. In a double-slit experiment, to say that 'the electron reached
the screen and passed through slit Aor slit B', is not the same as saying that 'the electron
reached the screen and passed through slit A or the electron reached the screen and
passed through slit B'. In one case one gets a diffraction pattern, in the other case a
superposition of two Gaussians.
The failure of the distributive law means that a joint probability distribution cannot
be defined; for example, the marginal distributions would fail to be additive.

Pr{aEA & (bEB or bEC)} Pr{aEA & bEB} + Pr{aEA & bEC}, (10)

even if the 'or' is exclusive, i.e., B and C are disjoint.

The BN approach, therefore, advocates a change of the logic on which the proba-
bilities are defined. Probabilities, such as those on the left hand side of (4), are defined
on sentences, but the 'and' and 'or' used to compound these sentences are such that the
distributive law fails. One therefore obtains a more general algebraic structure, rather
than the usual Boolean algebra (ora-algebra), on which probabilities are to be defined,
as countably additive, positive functionals with total mass 1.

3. 3 Birkhoff-von Neumann approach and the orthodox formalism of q.m.

3. 3. 1 The lattice ofprojections. The BN approach begins by noticing that the sub-
spacesll of a Hilbert space form a lattice, or a 'logic', with the desired properties. One
may identify a subspace of a Hilbert space with the orthogonal projection onto that
subspace. We now define an 'and' (A) and 'or' (v) as follows. If P1 and P2 are two
orthogonal projections on the subspaces R(P1) and R(P2) respectively, then P1 AP2 is
n
the projection on the subspace R(P1) R(P2), while P1 VP2 is the projection on the
smallest subspace which contains both R(P1) and R(P2).
If the dimension of the Hilbert space is ~2, and P1, Pz, P3 are taken, as in Fig. 1, as
 166 CHAPTERVIB

the projections on thex-axis,y-axis, and the line y=x, then it is clear that the distributive
law fails: PlA(P2VP3) = PI, but PlAP2 = 0, PlAP3 = 0, and OvO = 0, so that
PIA (P2 VP3)-:;t:. (PI AP2) V(PlVP3). Conversely, the algebraic structure corresponding
to the usual sentence calculus is a distributive lattice or a Boolean algebra. Thus, the
Hilbert space is related naturally to the failure of the distributive law.

3. 3. 2 Order relation and orthocomplement. The axioms for the usual sentence calcu-
lus may be formulated in terms of 'not', 'and' (A) and 'or' (V), or, more usually, in terms
of 'not' ( - ) and implication (~ ). In algebraic terms, following the algebraization oflogic
initiated by Boole, 'if and 'not' may be described respectively by an order relation l2 s,
and an orthocomplement '. These may be used to describe the properties of the lattice
of projections P: PI SP2 exactly if R(Pl) !: R(P2), i.e., the subspace onto which PI
projects must be a subset of the subspace onto whichP2 projects. The orthocomplement
P' is the projection on the null space of P denoted by N(P).
Fig. 1: Failure of the distributive law

If the projections Pl, P2, P3 are defined

by Pla = a Pza = ai, PJa = a.e,
for any vector a, then the join ofany two
of these is the projection on the plane,
while their meet is zero. Hence the join
of any two meets is zero, and cannot
equal the meet of anyone projection
with the join of the other two.

Given an order relation, the 'and' and 'or' may be re-interpreted: PlAP2 is the
greatest lower bound (g.l.b., infimum), while PI VP2 is the least upper bound (l.u.b.,
supremum) of the two-element set {PI, P2}. One would expect de Morgan's laws to
hold. One may also define the notion of orthogonality, P.lQ if pSQ '.

3. 3. 3 Geometrical interpretation. All these notions have simple geometrical mean-

ings in 3-dimensional Euclidean space. The closed subspaces are: the point at the origin,
lines through the origin (extended to infinity in both directions), planes through the
origin, and the whole space. The orthogonal projections are precisely that: if P is the
projection onto a line or a plane, the result of applying P to a vector is obtained by
dropping a perpendicular to the line or the plane in question. The partial order is
set-theoretic inclusion, and the infimum is the set-theoretic intersection. The l.u.b. of
two lines is the plane they span. The orthocomplement of a line is the plane perpendic-
ular to it. Orthogonality just means perpendicularity of the corresponding subspaces.

3. 3. 4 Dynamical variables, random variables and self-adjoint operators. Apart from

the failure of the distributive law, what is the point of studying the lattice of projections?
An immediate application is that it leads to the operator representation of observables
(hence the probability interpretation) in a natural way.
The first step is that classically an observable or a dynamical variable is a random
 QUANfUM-MECHANICAL TIME 167

variable. One observes the dynamical variable, and the observed values vary or show
some dispersion or scatter.
Now, in the usual measure-theoretic approach to probabilities, a random variable
is a measurable function. Given a set X, and a Boolean a-algebra, M, of subsets of X, a
(real-valued) random variable is a functionj:X-+R, such that f-l(A) EM whenever
A EBR. What one actually requires for this mysterious textbook definition is the inverse
map,j- 1:BR -+M, which is an isomorphism between the two a-algebras:

r 1( U Ai) = U r 1 (Ai ),

r1 ( nAi) = nr1 (Ai),

r1 (Ac) = (r 1(A)" (11)

where i will always denote an index running over finite or countable values.
When the setting is changed from a a-algebra to the lattice P (H) of projections on
a Hilbert space H, it is, therefore, natural to define a random variable in this way as an
isomorphism m: BR -+ P(H) into a Boolean subalgebra of P (H):

m( U Ai) = V m(Ai),

m (nAi) = A m(Ai),

(12)

The definition (12) makes sense for a broader class of algebraic structures. But if
the basic algebraic structure is the lattice of projections on a Hilbert space, then (12)
means that an observable, or a random variable, is automatically a projection-valued
measure or a self-adjoint operator.
Similarly, one can define a state, or a probability measure on the lattice of projec-
tions. The text-book definition of a (positive) measure is the following. Given (X, M)
as above, a measure p. is a function p. : M -+ R+ such that (i) p.(rfJ) = 0, (ii) if Ai'S are
pairwise disjoint,p. (UAi) = L P. (Ai). A probability measure satisfiesp.(X) = 1.
In the lattice of projections, the notion of disjointness is replaced by orthogonality:
if P.l.Q then P I\Q = o. So, one uses the conditions that (i) p.(0) = 0, (ii)
p.( V Pi) = LP.(Pi), if the Pi'S are pairwise orthogonal. A probability measure satisfies
p.(1) = 1. Gleason's theorem now recovers the usual density matrix approach.

3. 4 The quantum logic approach

3. 4. 1 The minimum requirements. To summarize the preceding section, the crucial

distinguishing feature of quantum probabilities is thenon-uzStence ofjoint distributions.
In the BN approach, this is achieved by defining the probabilities on the non-distributive
168 CHAPTER VIB

lattice of projections on a Hilbert space. The operator representation of dynamical

variables (or the 'probability interpretation') emerges as a bonus in this approach.
But the mysterious Hilbert space is still in the background. Can one account for
this too, starting only from a new type of 'and' and 'or' or a new type of 'if and 'not'?
The quantum logic approach tries to whittle down, a little further, the mystery of
quantum axiomatics. What is the minimum structure necessary to speak of probabilities
without a joint distribution?
Suppose we are given a set P, together with an order relation s, and an or-
thocomplement '. Such a triple (P, s, ') is called an orthoposet. For the above
definition of random variable and probability measure to go through, the right hand
sides should make sense. The minimum requirement is that P should be a-or-
thocomplete, i.e., every countable collection of mutually orthogonal elements of P
should admit a supremum in P.

3. 4. 2 Orthomodularity. There is one more technical requirement: P should be or-

thomodular (see box). Orthomodularity is a weak form of the distributive law:
distributivity => modularity => orthomodularity.
The lattice of projections on a Hilbert space His orthomodular, but fails to be modular
if H is infinite dimensional.
The point of orthomodularity is this: if P fails to be orthomodular, measures on P
may fail to be monotone. That is, we could have 13 asb, but fl(a)'<?fl(b), which is
unacceptable.
Thus, the minimum necessities for life seem to be that P should be an orthomodu-
lar, a-orthocomplete orthoposet. Unfortunately, such an object need not admit any
probability measures at all. If one further postulates the existence of 'enough' proba-
bility measures on P, or a 'full' or order-determining set of states on P, one obtains a
quantum-logic proper, in the sense of Gudder. 'Enough' probability measures means
that if a '<? b then there is a probability measurep such thatp(a) '<? p(b).

Fig. 2: Example of a
non-orthomodular lattice

A 'benzene ring' Hasse diagram pro-

vides an example of a non-or-
thomodular ortholattice. A measure
defined on such a lattice could fail to
be monotone. Hence the requirement
of orthomodularity.
For the definitions and the notion of
a Hasse diagram see Box 1: Collected
definitions.

3. 4. 3 Compatibility. In the general setting of a quantum logic, there need be no

operation corresponding to the usual algebraic product of two projections on a Hilbert
space. So what happens to the relationship between non-commutativity and incompat-
ibility or the non-existence of joint distributions?
One may recover this by defining commutativity or compatibility via the relation
 QUANTUM-MECHANICAL TIME 169

Box 1. Collected Definitions

Basics
Aposet is a pair (P, s), where P is a set and s is an order relation. That is, "I a, b, c E P, (i)
aSa (s is reflexive), (ii) asb and bsc ~ aSc (s is transitive), (iii) asb and bSa ~ a =b
(s is anti-symmetric).
In a poset (P, s), avb denotes the SlqJremum of {a, b}. That is, (i)a s aVb, and bs
a vb (a vb is an upper bound), (ii) if c E P is such that aSc, and bsc, then a vb s c (a vb
is the least upper bound). Similarly, a Ab denotes the infimum or the greatest lower bound
These notions may be extended to sets with an arbitrary number of elements. 0 is the g.l.b.
for P, while 1 is the Lu.b. for P. A poset is called bounded if 0 and 1 exist.
bcovmaifab,asb,and"lxEP,a < x s b ~a =b. InaHat\'ediagram,anelement
b which covers a is placed directly above a. An element of P which covers 0 is called an
atom. P is called atomic if, for any xEP, there is an atom aSx.

Lattice

A 1aJtice is a set L with binary operations V, A, which satisfy, for all a, b, c E L,

(i)(associativelaws)aV(bVc) = (aVb)Vc, aA(bAc)=(aAb) Ac,
(li) (commutative laws) a vb =b Va, a Ab = bAa,
(iii)(absorptionlaws) av(aAb) = a, aA(avb) = a,
(iv) (idempotentlaws)aVa=a, aAa=a
[(iv) is a consequence of (i)-(iii)]. One may define an order relation in a lattice by
asb ~a=aAb, or, equivalently,asb ~b =avb.
In a a-lattice, countable sets admit suprema and infima. In a complete lattice, all sets
admit suprema and infima
Orthogonality and orthocomplement

An orlhocomplementation on a bounded poset is a unary operation' which satisfies,

"I a,b,
(i) (antitone) asb :;. b' Sa',
(ti) (period 2) a" = a,
(iii) (orthogonal decomposition) a Va' =1, a Aa' = O.
An orthoposet is a poset with an ortbocomplementation In an orthoposet, a is orthogonal
to b, written a J. b, if asb'. Orthogonality is a symmetric relation
I (continued on p 170)
x Cy ~ Y == xA(yVx'), (13)

which could be interpreted to mean: x is compatible with y if whenever x is 'true', and

yvx' is 'true', theny is'true'.
In the lattice of projections, this notion of commutativity agrees with the usual
notion. The relation to distributivity is this: the triple {a, b, c} is distributive if anyone
is compatible with the other two. This provides one more reason to believe in or-
170 CHAYrERVIB

Distributive and modular laws

Given a lattice I.., for a, b, e E I.., a triple (a, b, e) is called a distributive triple if tbe distributive
laws

(aAb)ve = (ave) A (bVe), (aVb)Ae = (aAe) V (bAc)

hold. A distributive ortholattice is a Boolean algebra. Lis modular exactly if a $ b implies

that (a, b,x) is a distnbutive triple, for every x E L. Lis orlhomoduiar if a$ b implies that
(a, b, a') is a distributive triple. In general, we have the relations

distributive law ~ modular law ~ orthomodular law,

so that a Boolean algebra is a modular ortholattice, which, in turn, is an orthomodular

lattice.
In a lattice I.., {a, b} is called a modularpair, ifV'xE I.,xsb ~ (a, b,x) is a distributive
triple. L is called semi-modular jf whenever {a, b} is a modular pair, so is {b, a}.
Modularity holds for the lattice P(H) of projections on a Hilbert space, when H is
finite-dimensional, but fails otherwise since the sum M + N, of two subspaces, M, N, need
not be closed. The onhomodular law,

a$b ~av(a' Ab) = b,

always holds for P(H).
One can also think of other notions of modularity in terms of a generalized associative
law. One has, for instance, the modular law

a$e ~aV(bAc) = (aVb)Ae.

Similarly, {a, b} could be called a modular pair if" x E l.,x$b => (xva)l\b = xv(al\b).
Orthomodular lattices are characterized among ortholattices by the property that they
have no subalgebra isomorphic to the ortholattice shown in the 'benzene ring' Hasse
diagram (Fig. 2).
On the above ortholattice, one could define a measure which would fail to be
monotone: .u(0) = 0, .u(a) =2, .u(b) =1, .u(x)= 00, for any other x. However, on an or-
thomodular lattice, a non-negative additive map is always monotone.
I. -(continued
--------1
on p 171)

tho modularity - an orthoposet is orthomodular when the relation of compatibility is

symmetric.

3. 5 The laueh-Piron approach

In a quantum logic, the need to postulate a full set of states ('enough' probability
measures) introduces a jarring note into the program of 'deriving' the Hilbert space
 QUANI1JM-MECHANICAL TIME 171

Compatibility
An orthomodular poset may contain a Boolean subalgebra. An example is the propo-
sition range of a classical dynamical variable m, given by {p == mE E lEE BR }. A
block in an orthomodular lattice L is a maximal Boolean subalgebra. Two elements x
andy commute, xC y, ifx andy are contained in a block. Commutativity is a symmetric
relationship in an ortbomodular lattice. Equationally, the relationship of commutativity
in an ortbomodular lattice is defined as follows.
x C y if and only if anyone of the following holds:

x= (xAy) V (xAi),
X= (XVY)A (XVy'),
XA(X'Vy) =xAy.

The commu/a/orc(x,y) ofx,y is defined by

c(x,y) = (xAy)V (xAi)V(x' Ay )V(x' Ai).

= 1. The centre C(L) of an ortbomodular lattice Lis the set of
xC y if and only if c (x,y)
those elements which commute with every other element:

C(L) = {zE~zCx, V'xEL}.

C(L) is a Boolean subalsebra of L An ortbomodular lattice is called irreducible if

the centre is trivial: C(L) = {O, 1}.
Hilbert lattice
The lattice L satisfies the exchange axiom, if whenever a covers aAb, then aV b covers b.
A complete, atomic, ortbomodular, infinite-dimensional lattice, satisfying the exchange
axiom, is called a Hilbert lattice. A Hilbert lattice does not necessarily arise as the lattice
of closed subspaces of a Hilbert space (real, complex, or quaternionic).

from a new type of 'and' and 'or' or 'if and 'not'. Can one give sufficient conditions on
the algebraic structure so that this requirement is automatically satisfied?
An example of such sufficient conditions is provided by Piron's representation
theorem. One needs to assume that the 'if and 'not' give rise to
(i) an orthomodular lattice L, such that
(ii) L is complete(suprema and infima exist for arbitrary sets),
(iii) L is irreducible (0 and 1 are the only elements of L which commute with all other
elements), and
(iv) Lis semimodular (XSy ~ yA(y V x) = x).I

Such an algebraic structure is called an irreducible CROC. Piron's representation

theorem asserts that if L is an atomic, irreducible CROC, then L is isomorphic to the
lattice of closed subspaces of a Hilbert space.
 172 CHAPTERVIB

The significance of atomicity arises from the relation of atoms to J auch-Piron states.
A Jauch-Piron state is a subset S ~ L satisfying
(i) 0 ~ S, (ii) if a E S and a S x, then xES, (iii) if aj E S, then" t aj E S, (iv) Sis
maximal with respect to the above properties.
Every Jauch-Piron state defines an atom, and vice versa.

3. 6 Defects in the quantum logic approach

Though the logico-algebraic approach initially seemed promising, it now seems riddled
with physical and mathematical obscurities.
As one example of a mathematical difficulty, consider the following. The Piron
representation theorem is one of the deeper results of the theory. But the Hilbert space
provided by Piron's theorem is not necessarily the usual Hilbert space over the field of
complex numbers: it is a Hilbert space over a 'division ring with involution', F. If this
Hilbert space admits an 'observable' which is 'maximal' in some sense, one may show
thatF is an extension of the reals. If one further assumes that the degree of this extension
is finite, the Frobenius theorem14 tells us that F must be R or C (complex numbers)
or the Quaternions. The final result is tantalizing, but what is one to make of the
intermediate hypotheses?
As another example, consider the following. A lattice with all the nice properties
of the lattice of projections is called a Hilbert lattice (see box). But a Hilbert lattice
cannot necessarily be represented by the lattice of projections on a Hilbert space.
Clearly there are limits to what can be achieved by a purely algebraic approach.
One could live with the mathematical difficulties, but the physical obscurities are
fatal.
(i) The whole idea behind the approach is to motivate quantum axiomatics from a
physical standpoint. Now we have already seen the emergence of some distinct struc-
tures
- Hilbert lattice (Birkhoff-von Neumann)
- quantum logic (Gudder)
- atomic irreducible CROC (Jauch-Piron)
The actual list is very much longer. How is one to choose between these structures?
For this one must go to the physical interpretation. The physical interpretation relates
to the elements of the lattice: these are 'yes-no' type 'questions' (Jauch-Piron), or
true-false 'outcomes of measurement' (Birkhoff-von Neumann). But one cannot hope
to simultaneously measure incompatible observables, so the 'and' does not make
physical sense. Hence one cannot, in principle, have a lattice, and out goes the Hilbert
space. One can form alternative, better motivated structures, but this does not lead to
quantum physics as used by physicists for calculations.
(ii) While the algebraic approach sheds some light on the algebraic structures underly-
ing classical mechanics and q.m., a problem common to all such approaches is the
following. They push deeper into oblivion the link between classical and quantum
dynamics, or the link between classical Hamiltonian evolution and evolution according
to the Schrodinger equation.
(iii) This approach does not explain why the change of logic should be needed at the
microphysical level, while usual Boolean logic is good enough at the macrophysical
level, especially since the concepts of 'micro' and 'macro' tend to be anthropocentric.
 QUANTUM-MECHANICAL TIME 173

Going back to the 'clumsy fingers' kind of idea would defeat the whole purpose of the
approach, irrespective of experiments on Bell's inequalities.
Thus, the logico-algebraic approach fails to explain the origin of the Hilbert space
and the peculiar nature of quantum probabilities. The most sympathetic view that one
can adopt, and the one that I will adopt in the sequel, is that this approach, while partly
satisfying, needs further development, not so much at the level of formal axiomatic
systems for quantum logics, but in terms of a priori motivation for the change of logic
at the microphysical level. It is better to clarify that this is more or less the exact opposite
of the typical approach in this area.

4 The structured-time interpretation of q.m.

4. 1 Motivation

The structured-time interpretation15 may be motivated as follows. That classical me-

chanics fails at the microphysical level is an established fact. But can one identify the
intrinsic or a priori shortcomings of classical mechanics which led to this failure? Can
one identify those microphysical shortcomings of classical mechanics which need to be
remedied in order to obtain the formalism of q.m.? As an analogy, there was a
theoretical problem with the measurement of time in classical mechanics, the identifi-
cation and resolution of which led to relativity theory (Chapters II, IlIA, IIIB).
According to the structured-time interpretation of q.m., the problem is once again
with the notion of time in classical mechanics. The particular shortcoming in question
is the simplistic concept of linear time used in classical mechanics. At the tum of this
century, Poincare's analysis of the deficiencies in the Newtonian concept of time,
particularly the concept of simultaneity and equal intervals of time, led to the origin of
the special theory of relativity. However, according to the structured-time interpreta-
tion of q.m., this analysis did not go far enough, for it assumed that time was like a line,
by assuming the existence of intervals of time, and dealing only with the problem of the
equality of the lengths of these intervals. According to the structured-time interpreta-
tion, the assumption of time as a line (either a straight line or a timelike curve) needs
to be given up.

4. 2 Overview of the argument

As the argument is a little complex, we present an overview first. We argue below ( 4.3)
that a microphysically structured time follows as a consequence if one accepts small
departures from strict time-asymmetry or a small 'tilt in the arrow of time '. We recall
that the hypothesis of a tilt in the arrow of time itself emerged naturally in the process
of attempting to establish a physical direction of time, going from the entropy law
(Chapter IV) to the electromagnetic field (Chapter VA) to electrodynamics (Chapter
VB).
The link with q.m. is obtained as follows. The difference between classical mechan-
ics and q.m. was related above to a difference oflogic. Now, a microphysically structured
time, too, entails a change of logic at the microphysical level: the logic of a structured
 174 CHAPI'ER VIB

time differs from the logic of linear time in somewhat the same way as the logic
(required for the formal semantics) of a parallel computer program differs from the
logic of a sequential program. What remains to be seen is whether the logic arising from
a microphysically structured time corresponds to a quantum logic.
The above claims may be represented schematically.
time-asymmetry in classical mechanics ~ tilt in the arrow of time ~
structured time ~ change of logic ~ quantum logic ~ q.m.
The slogan formulation of these claims would be: (1) 'Non-locality is unavoidable
in classical mechanics', and (2) 'Non-local classical mechanics is quantum mechanics'.

4. 3 From electrodynamics to structured time

As a specific and empirically viable example of a tilt in the arrow of time let us consider
direct-action electrodynamics with incomplete absorption. That is, we go back to the
situation, considered in Chapter VB, where interactions are both retarded and ad-
vanced, but advanced interactions form only a small part, say 1 part in a trillion. 'Strict
causality' or perfect time asymmetry has broken down, but the departures from the
standard picture are quantitatively rather small.
The particle16 equations of motion are now differential equations with mixed-type
deviating arguments considered in Chapter VB. As pointed out there, such equations
do not, in general, admit a unique solution, in either direction in time, even if the entire
past (or future) history is prescribed. Typical solutions will branch and collapse both
into the future and the past. The picture is somewhat similar to what one would get by
taking paths, in the path integral formalism, literally as particle trajectories.
Accepting the existence of small amounts of advanced radiation as an empirical
hypothesis, what are we to make of this non-uniqueness, and the branching and collapse
of solutions? In the absence of any mathematical or physical criteria which enable one
to select a particular solution, the structured-time interpretation grants equal validity
to all the solutions.
Granting equal validity to all solutions means that the state of the world is not
uniquely specifiable: the world has split into a number of equally real states.17 The
statement, 'the x-coordinate of the electron (at "time t") is xo' cannot any longer be
regarded as either true or false: its truth or falsity depends upon the particular branch
of the solution we choose. Let us refer to this state of affairs, involving intrinsic
non-uniqueness, and the branching of solutions as 'branching time')8
It is impossible to describe branching time, or an 'ambiguous reality', with the usual
logic (of linear time), where a given statement must be either true or false, where the
x-coordinate of an electron (at time t) either definitely is, or is not, Xo, regardless of our
state of knowledge. A new logic is needed which takes into account the generic break-
down of uniqueness of solutions, and the resulting breakdown of determinism, at least
at the microphysical level.

4. 3. 1 The analogy with esp. The salient features of this new logic are not difficult to
grasp. For those familiar with parallel computing, there is an interesting and complete
analogy with the branching of processes in Hoare's language of communicating sequen-
tial processes (CSP), and its practical implementation in OCCAM. Unlike the case of
 QUANTUM-MECHANICAL TIME 175

the transactional interpretation of q.m., the analogy here is incidental rather than
essential, and operates at the deeper level of the temporal logic required for the formal
semantics of Hoare's CSP language. 19 Consider an OCCAM program of the following
type.

PAR - - do indented processes in parallel

PARi=OFORn - - Proc 1: replicate in parallel
SEQ - - do in sequence
A(i) - - some process which generates B(i)
Chan [i] ! B(i) - - output B(i) to Chan (i)
ALTi=OFORn - - Proc 2: do indeterministically
Chan [i]? B - - input from Chan (i) to B

The formal indeterminism of the ALT construct should not be confused with
'randomness': the ALT construct does not assert, 'select one of th~ B(i) at random'. B
is not a classical random variable, since it is impossible, in principle, to specify uniquely20
the value of the number B, or the final state of the program (assuming that the numbers
B(i) are distinct). The PAR construct corresponds to branching: the n processes evolve
in parallel. The ALT corresponds to indeterministic collapse: one of the n values is
selected indeterministically.

4.3.2 Past and present contingents. We note that, with the hypotheses of a tilting
arrow of time, time branches into both the future and the past, as compared to the
everyday picture of time as future-branching but past-linear .21 In the usual picture (built
into the grammatical structure of Indo-European languages), which we take seriously
at the level of everyday life (and in the belief that one is 'free' to perform any
experiment), the future is uncertain or contingent, but the belief is that the past is
decided (even if we do not know all about it). But with a tilt in the arrow of time not
only the future, but the past and the present also become contingent.22

4. 3. 3 The logic of structured time. The simplest logic that describes such a structured
time is a quasi truth-functional (q.tJ.) logic, considered in Chapter I. In this logic there
are precisely two branches, corresponding to a world with two possibilities (at any
'instant'). By considering collections of such 'worlds' one may model a situation with
many possibilities.
We now explain how such a logic naturally gives rise to a quantum logic.

4. 4 From structured time to quantum logic

4. 4. 1 Statements. To model the concept of a world with many possibilities, in a logical

setting, each 'branch' or possibility of the real world, at any 'instant', is modeled by the
Wittgensteinian concept of a world as 'all that is the case'. In other words, a world, or
rather a truth-functional world, is specified by a collection of statements deemed to be
true. Intuitively, the statements declared to be true are tautologies, and the empirical
facts valid at that 'point of time'.
176 CHAl'TER VIB

The primary model of empirical facts being that of classical mechanics, the basic
statements, p, are taken to be of the type

p == mEE, (14)

where m is a classical dynamical variable, and EEBR is a Borel subset of R. Reference

to (3) and (4) shows that the probabilities of interest in q.m. are the probabilities of
precisely such statements. As explained earlier, the 'quantum' nature of these proba-
bilities is decided by the logic used to compound these basic statements.
The set of all propositions23 of type (14) will be denoted by P. This seems simple
enough, except that there is a subtle but crucial difference right here. As Van Fraassen24
points out, 'The description of the set of propositions is usually quite abstract, the
relation to q.m. being left to the imagination of the reader.'
The point is that it is problematic to assign truth-values T, F or 0, 1 or 'yes', 'no' to
sentences of the type (14). This is straightforward in classical mechanics, but in quantum
mechanics only probabilities of statements of type (14) are meaningful. In the BN
approach, the elements ofP are related to true-false experiments, and in the Jauch-Piron
approach to yes-no questions. In all such motivations, the measurement process enters
as a logically primitive notion, even though all the controversy in q.m. centres around
the measurement process.
In the present approach, elements of P correspond to well-formed statements of
classical mechanics, which admit a truth-value if a classical state is prescribed. The
lattice theory approach has been regarded as suspect precisely because, for quantum-
mechanically incompatible statements p, q, experiments for p and q interfere, so that
p A q is not physically meaningful. So the objection to a lattice structure is no longer
valid with the present approach. The set of propositions may form a lattice, even if one
does not expect this for the set of measurable propositions.

4. 4. 2 Truth-functional worlds. A truth-functional world a is, then, a collection of

statements p, such that (i) every tautology is in a, (ii) conjuncts (p A q), disjuncts
(p V 'q), and logical consequences of elements p, q of a are in a, and (iii) for any p,
eitherp is in a or its negation ( -p) is in a. Since the only empirical facts of concern are
those of classical mechanics, a truth-functional world may be thought of as specified by
a point in classical phase space.

4. 4. 3 Quasi truth-functional worlds. In a truth-functional world every statement is

either true or false. In a world admitting contingents, or in a world in which time
branches, there must be at least two possibilities or branches. The simplest logical
model of such a world is, therefore, a quasi truth-functional world ltJ = {a, p}, an
unordered pair of truth functional worlds a, p. In the structured-time interpretation, a
q.t.f. world is regarded as the basic element of (microphysical) reality.

4.4.4 Modalities. In a q.t.f. world, modalities (or the notions of 'possibility' and
'necessity') may be formally defined. A statement p, which is true in both a and ,0, is
necessary (or necessarily true) inltJ = { a,p}, denoted byltJ E lop I. Astatementp
which is true in at least one of a, ,0 is called possible (or possibly true) in ltJ, denoted by
 QUANfUM-MECHANICAL TIME 177

(J) E I pI. A statement which is false in both a, f3 is, of course, necessarily false in
(J).

4.4.5 I"educible contingents and admissible worlds. A q.t.f. world represents a snap-
shot of a world with two possibilities. Since there is only an inconsequential difference
between a q.t.f. world (J) = {a, a} and the truth functional world a, such worlds are
ruled out by deeming them not admissible. The precise criteria for admissibility will be
considered later: they have to do with preserving a minimum or irreducible contingency.
The impossible world 0, in which every statement is true, and the true world 1, in which
only tautologies are true, will be deemed, by convention, to be admissible.

4. 4. 6 States. The collection of all admissible q.t.f. worlds is the (logical) universe U
(and a q.t.f. world will by default be admissible from now on). A collection ofadmissible
q.t.f. worlds is a state.25 A state, thus, represents a snapshot of a world in which there
are many possibilities, but the basic units of which are also irreducibly contingent. Using
the correspondence of truth-functional worlds with points in phase space, a state may
be identified with a swarm of points in classical phase space, with the proviso that the
'real' state of the system is not represented by any single point. Astate represents the
real world at an instant of time, or rather an instant of time may be defined by a state.
It will become clear later on that the connection between states and probability
measures on the set of propositions is the following. Each q.t.f. world (J) induces a
probability measure Pw defined on P as follows.

if (J) E I[}7 I
if (J) E I pl.
otherwise

(15)

4. 4. 7 Quantum measurements, access relations, and selection functions. The idea is

to model the process of quantum measurement as follows. A measurement can be
carried out only if there is some certainty: it is pointless to speak of the length of a moving
earthworm. The situation is worse if there are several possibilities at the same 'instant
of time', as happens with structured time. In a world with many possibilities one can
measure only that which is certain. That is, a dynamical variable can be measured only
in a world in which it necessarily assumes a certain value.
Thus, the measurement process for a given dynamical variable, m, must, by defini-
tion, transform the given world (or state) into one in which a statement ofthe type (14)
is necessary. Therefore, the process of measurement, conditioned on a given statement,
(discontinuously) 'selects' the 'closest' 'accessible' world in which the statement is
necessary. We have already defined 'necessary', and we now formalize the terms in
quotes.
Accessibility is modeled by means of a relation. An access relation R is defined
between two q.t.f. worlds, {J)1 and {J)2, by requiring that {J)1 R {J)2 if and only if {J)1 and {J)2
have a truth-functional world in common, {J)ln {J)2 -:;t:. 0. The impossible world will be
deemed, by convention, to be accessible from any world. We observe that this access
relation is not transitive.
178 CHAPTER VIB

'Selection' is modeled by a selection function. A selection function, conditioned on

statement p, is a map between q.t.f.worlds such that (i) /p(w) is accessible from w, (ii)
p is necessary in/pew), and (iii) ifp is necessary in w, then/pew) = w. More formally, a
map /:px U-+U, such that /(P, .) satisfies the above properties, for each p, is called a
selection function.
The process of measurement is specified by a selection function,26 which is expected
to 'select' the 'closest' accessible q.t.f. world in which p is necessary. In a general logical
setting, defining 'closest' is known to create difficulties27: if Bizet and Verdi are friends,
is the closest world one in which they are both French or both Italian? In the present
case, remembering that a truth-functional world may be specified by a point in classical
phase space, it is possible, though not essential, to define 'closest' as closest in the sense
of distance in classical phase space, subject to admissibility (or the exclusion of a certain
'minimum' phase space volume in order to ensure that contingency is irreducible). The
actual existence of a selection function is demonstrated later on.

4. 4. 8 Incompatibility, joint measurements and repeated measurements. In the above

view of measurement, two statements p, q, are deemed to be incompatible if there is
no q.t.f. world, hence no state, in which they are jointly necessary, but q is possible in
some worlds in whichp is necessary and vice versa. It is impossible, in principle, to jointly
measure two incompatible statements: a measurement of p results in a world in which
there is no definite information about q. It may be easily checked that, in terms of
selection functions, this corresponds to a failure of commutativity:/p 0 /q ~ /q o/p. That
is, in terms of repeated measurements, forincompatible statements p, q, a measurement
of p followed by a measurement of q does not, in general, result in the same world as a
measurement of q followed by a measurement ofp.

4. 4. 9 Measurability. If p, q are incompatible statements, then p A q is a statement

which is possible in some worlds but necessary nowhere. Such a statement is not open
to measurement. We now define the quasi-measurable statements as those p for which,
given a world w in which p is possible, it is always possible to select a (unique)
non-impossible world/pew) ~ 0, in whichp is necessary. A propositionp is regarded as
measurable ifbothp and its negation -p are quasi-measurable. The set of all measur-
able propositions will be denoted by M. Formally,

M = {p EP I/p(w) ~ 0, for all wE I <>p I},

M = {p E pip EM A -p EM} (16)

4. 4. 10 'And~ 'or~' 'if~ 'not'. As observed earlier, the origin of the difference between
classical and quantum probabilities, as also many of the differences between the two
formalisms, can be traced to the fact that 'and' and 'or' between statements must be
interpreted differently in q.m.. As also seen earlier, it is possible and more convenient
to work with a new type of 'if' or a new type of order relation (,implicate order') which
we now introduce, using the concepts developed above.
It is, first of all, clear that in a world admitting contingents, a new type of 'if' is
required. What would one say about the truth ofP'*q in a world in which p is possible,
 QUANTUM-MECHANICAL TIME 179

but q is false? One may adopt different positions. Two simple possibilities are the
following:

Dp ~ Dq, (17)

D(p ~ q). (18)

The first says that 'if p then q' is deemed true exactly if whenever p is necessarily true
so is q. Ifp is only possible, nothing is implied about q. The second (strong implication)
says that p~q must necessarily be true, and is the stronger formulation: this implica-
tion is false if there is a world in which p is possible but q is false. A variety of other
formulations exist. 28

4. 4. 11 Preliminary order relation. The preliminary order is defined on the set P of all
propositions29 p of the form (14), using the selection function introduced above. We
introduce a proposition pq as follows:

wE Ip q I == fqofp(w) = fp(w). (19)

This says that pq is true in a world W exactly if the closest world in which p is necessary
is also a world in which q is necessary. One now gets an order relation p > q by the
proposition Du (pq),

p > q == Du (pq), (20)

i.e., by demanding that pq be true in every admissible world.

It is clear that the 'if' ('counterfactual implication') or order relation defined by
(20) is weaker than that defined by (18). One of the achievements of the theory is the
rather easy proof30 that, for measurable propositions, the order relation, defined above
in terms of repeated measurements, may be characterized in terms of strong implica-
tion:

p > q <> Du (D (p ~ q, p, q E M. (21)

This order relation is not, in general,antisymmetric, for a proposition p which is

possible in some worlds but necessary nowhere satisfies bothp>O and O>p, without
being identically false. But we may adopt the usual algebraic technique of identification
and equivalence classes: we regard p and q as weakly equivalent, p = q, if both p > q
and q > p hold, in order to obtain an order relation on the set of equivalence classes
P / =. This is not algebraic juggling - the measurement process cannot distinguish
between weakly equivalent statements. For p, qEM, of course, weak equivalence is the
same thing as usual logical equivalence.

4. 4. 12 Orthogonality, compatibility and order relation. Before defining the final 'if,
180 CHAPTER VIB

we need one last concept. Two propositions P and q are regarded as orthogonal, written
P .i q, if P > -q holds,

p.i q == p > -q. (22)

Using orthogonality, we can finally define an order relation among measurable propo-
sitions as follows: p > >q if and only if there exists a measurable proposition s orthog-
onal to p such that q =p V s:

p q == 3s EM, s.ip and q = pVs. (23)

The earlier order relation seemed fairly natural, but what is the significance of this
new relation? It may be shown31 that this order relation is equivalent to (i) the earlier
order relation p >q together with (ii) compatibility, or commutativity of the selection
functions. Algebraically, the advantage of defining the order relation in this way is that
orthomodularity is an immediate consequence.

4. 4. 13 Main result. The main result is as follows.

Theorem.32 The set of all measurable propositions, M, ordered by the 'if defined by
(23), > >, and complemented by the usual 'not', -, is a quantum logic which is also
atomic.
(The terminology used in reference 15 and the current terminology may be related
as follows. Given a countable collection of measurable propositions PiEM, which are
mutually orthogonal,Pi.i Pi for i #.j, the measurability of V t Pi follows because not more
than one Pi may be true in a truth-functional world. For the same reason, Pw , defined
by (15), is countably additive, hence a probability measure. Hence, using (23), if
P > >q thenPw(q) = Pw(P) + Pw(s). Ifs#.O, then there exists anw for which Pw(s) #. 0;
for this w, Pw(P) <Pw(q). That is, there exists an order-determining set of states.)
We can now summarize. Starting from a purely classical hypothesis (the existence
of small amounts of advanced radiation), we have arrived at the basic minimum
algebraic structure required for q.m., and particularly for the mysterious behaviour of
quantum probabilities.
There is undoubtedly a gap between the slogan formulation and what has been
achieved above. Some questions about semi-modularity, irreducibility, and the lattice
structure are considered below. The more serious deductive gap pertains to the irre-
ducibility of contingents assumed below, in the criteria of admissibility. The examples
in Chapter VB only showed that contingents are possible; below we make the inductive
generalization that they are necessary. To put matters crudely, the physically plausible
assumption is that the 'degree' of contingency is proportional to the amount of advanced
interactions.
In the absence of a well-developed mathematical theory of functional differential
equations, one may argue for the plausibility of this assumption as follows. Consider
the motion of a macrophysical object such as a rocket. Emission of a single advanced
photon from the rocket (= absorption of a single retarded photon) is closely analogous
to the usual statistical perturbation (even though this perturbation could not, in princi-
ple, have been determined from an exact knowledge of the entire past history of the
rocket and the enveloping universe). In this situation, one would expect possible
 QUANTUM-MECHANICAL TIME 181

deviations from the classical trajectory to be small, just because these would be
proportional to the strength of the perturbation, i.e., the amount of advanced interac-
tions, which has been assumed to be very small.

4.4.14 Example offailure of the distributive law. With the above order relation, the
distributive law, of course, fails. Let us call a q.tJ. world in whichp is necessary ap-world.
Choose measurable propositions p, q, s, such that no p-world or q-world is an s-world.
Then s Ap=O, s Aq=O, but sA (p Vq)~O, if p is true in one branch of an s-world while
q is true in the other branch.
The failure of the distributive law arises through a 'squeezing in' of possibilities:
the necessity of p Vq does not imply the necessity of p or the necessity of q, for p may
be true in one branch while q is true in the other. In the two-slit experiment, in the case
of the diffraction pattern, it is necessary that the electron came through slit A or slit B.
In the case of the Gaussians, it is necessary that the particle came through slit A or it is
necessary that the particle came through slit B. The two statements are not equivalent.
The possibility that has got 'squeezed in' is not that the electron split into two, with
one part going through slit A and the other part going through slit B. Rather, the real
world has two components: in each the electron came through one slit. On this
interpretation, the diffraction pattern is an empirical consequence of a contingent past.

4. 4. 15 The Lattice structure. We had remarked earlier that alternative approaches to

quantum axiomatics tend to reject the lattice structure, but that such a structure is
possible in the present approach, because the propositions are not motivated either in
terms of extant quantum theory or using some logically primitive notion of the mea-
surement process. Thus, there is no question of the physical interpretation of the lattice
structure. The B-N question, 'What is the physical significance of A, V?', is to be seen
as a wrong question dating back to the days when physics was regarded as no more than
an axiomatic theory together with an interpretation.
Rather, the lattice is to be seen as a technical convenience (like Borel sets), and
obtained by enlarging, if necessary, the set of measurable propositions. Thus, the
question is 'What is the most economical enlargement of M which will give a lattice?'.
Using the work ofPooP3 one may proceed as follows. Eachp E M may be identified
with the functionfp. We let

(24)

L containS a copy P(L) of the quantum logic M, which can be extended to a lattice if

(25)

It is possible to interpret (24) as repeated measurements regarded as instantaneous in

a coarse treatment, and (25) in terms of the micro time symmetry of repeated measure-
182 CHAPTERVIB

ments; from the present point of view, however, the more important question is whether
any of the selection functions defined below satisfy (25).

4.4.16 Admissibility. We pointed out earlier that q.tJ. worlds of the type w = {a, a}
cannot be deemed to be admissible. On the other hand, permitting q.t.f. worlds of the
type w = {a, p}, where a,p are arbitrarily close (in the sense that-the corresponding
points in phase space are arbitrarily close) would permit contingency to disappear by
the back door.
Contingency may be maintained only if 13 is excluded from a certain 'finite' (as
opposed to infinitesimal) region around a. To do this invariantly, one must exclude a
'finite' volume around a. In the one-dimensional case, if a is specified by the phase
space coordinates (Po,qo), and 13 = (p,q), then w={a, p} is admissible, provided
Ip - po I Iq - qo I ~ 1 (in appropriate units). The existence of incompatible propo-
sitions, assumed earlier, is an immediate consequence of this criterion.
Such a criterion is open to the charge of having introduced the uncertainty principle
by the back door. However, the structured-time interpretation is based on a dynamical
hypothesis regarding the nature of the equations of motion. So, the above criterion has
the status, not of an absolute principle, but of a conjecture concerning the solutions of
the equations of motion. (It may not be so easy to settle this conjecture, since that would
amount to relating the value of Planck's constant to the amount of advanced interactions
assumed to exist.)

4. 4. 17 Example of a selection function. Given the above criterion of admissibility,

one may construct a selection function as follows. Let m be a classical dynamical
variable, E E BR, andp == mE E the proposition described by (14). Let w = {a,p}
be a q.tJ. world, where a and 13 are described by the same phase space coordinates as
above. If (po, qo) E m- 1(E), i.e., the statement mE E is true in a = (Po, qo), we let
fp(w) = Wi = {a, Pi}, where the subscript p refers to the statement mEE, and 13' is
specified by the unique phase space point (Pl, ql), closest to (p, q), which satisfies the
criterion of admissibility and also (Pl, ql)E m- 1(E). In case such a point fails to exist or
fails to be unique, we letfp(w) = O.
This example of a selection function is not unique. Instead of making the 'minimum'
change in w, we could allow the selection function to minimize contingency by selecting
the world closest to 13 from among the admissible worlds closest to a. The particular
selection function one chooses decides the nature of the measurement process.

4. 5 Relation to other interpretations

4. 5. 1 The many-worlds interpretation. Though it is the world, rather than the elec-
tron, which suffers from schizophrenia, the structured-time interpretation must be
distinguished from the superficially similar Everett-Wheeler 'many-worlds' interpreta-
tion of q.m. The many-worlds interpretation takes quantum-mechanics seriously
enough at the macrophysicallevel to set up a wavefunction for the universe, leading to
questions about the observer being included in the wave-function or consciousness
being included in physics but excluded from the wave-function. To account for the
collapse of the wave function, in the 'many-worlds' interpretation, the world constantly
splits into possibilities (for no earthly reason). These are entirely separate worlds, and
 QUANI1JM-MECHANICAL TIME 183

one observer is aware of only one possibility. Moreover, that interpretation does not
provide any route to the formalism of q.m..

4. 5. 2 The transactional interpretation. The transactional interpretation34 of q.m. is

also superficially similar, in that it uses the Wheeler-Feynman theory and the
'handshake' analogy of the transputer implementation of CSP to account for non-local-
ity: a retarded 'offer wave' generates an advanced 'confirmation' wave, resulting in a
giant handshake across spacetime. The interpretation is primarily concerned with this
analogy, which it regards as analogous to Faraday's lines of force!
However, we have already pointed out (Chapter VB) the internal inconsistency of
the Wbeeler-Feynman theory. The transactional interpretation also shows a curious
disregard for any empirical consequences of non-locality beyond the possibility of
interactions at space-like separations in Aspect-type experiments.

4. 5. 3 The modal interpretation. Though the structured time approach is construc-

tive, rather than axiomatic, it is very similar to the modal interpretation.35 One of the
crucial differences is the use of quasi truth-functional semantics which greatly clarifies
certain obscurities in the modal interpretation.36

4. 5. 4 The Copenhagen interpretation. The Copenhagen interpretation is, in a way,

closest to the structured-time interpretation. The difference lies in the fact that the
structured-time interpretation begins with a dynamical hypothesis rather than abstract
philosophical principles like 'complementarity'. Rather general connections37 between
'equilibrium' (or indifference to the origin of time) and unitary evolution, together with
the analogy to the path integral formalism, suggest that there is at least some hope of
linking classical Hamiltonian evolution with SchrOdinger evolution, within the struc-
tured-time interpretation.

5 Conclusions

Non-local classical mechanics admits qualitative features that have been regarded as
typically quantum mechanical. Probabilities in a world admitting irreducible contin-
gents manifest features characteristic of quantum probabilities: specifically, the micro-
physically structured time, resulting from a microphysical tilt in the arrow of time, forces
a change of logic, resulting in a quantum logic.
184 CHAPIERVIB

Appendix: Proof of the main theorem

We recall that statementsp, q, r... are of the type

p=mEE (1)

whereE E BR is a Borel subset of the real lineR. We recal1that a q.t.f. world w consists
of a pair of truth-functional worlds denoted by a, p, ... . We assume that we are given a
collection of all admissible q.t.f. worlds - the universe U. Truth-functional proposi-
tions p consist of truth-functional worlds, whereas q.t.f. propositions consist of q.t.f.
worlds:

I p I = {a,p E a},
wE IOpl .=. ('Va Ew)p Ea

wEI <>p I .=. (3a E w) pEa.

Thus, a E Ip I means thatp is true ina, while wE lOp I means thatp is necessary in
w. We will use the notion of universal necessity, Du , to indicate a q.t.f. proposition
which is true for all (admissible) q.t.f. worlds.

Remark 1. Here, and elsewhere, we use the 'dot notation' obtained as follows.
Imagine that the requisite parentheses are replaced by an overscore. Now imagine that
the overscore is removed except for its tip near the appropriate connective.

We recall the definition of the access relation R,

1. wRO ('Vw)
2. if W2 ~ 0, Wt R W2 iff Wt n W2 ~ 0,
the definition of the selection function,
1. w Rfp (w)
2. fp (w) E lop I
3. wE lOp I "* fp (w) = w,
and the definition of the order relation,p > q .=.fp < fq,

wEI p > q I .=. fq 0 fp (w) = fp (w) (2)

Theoreml. Ou (O(p"*q ."*. Du (Op"*Oq) ."*. Du(p > q).

Remark 2. The theorem says that if, in every q.t.f. world,p "* q is necessary, then, in
every q.t.f. world the necessity of p implies the necessity of q, and every q.t.f. world
belongs to the proposition p > q. To paraphrase, 'strong implication' implies 'weak
implication' implies the 'angle' implication, under universal necessity.
 QUANI1JM-MECHANICAL TIME 185

Proof. The first implication is trivial, since

wE 10(p *q)1 .*. wE lOp *Oq I.

(Ifp*q is true in both branches of a world w, then ifp is true in both branches, so must
be q.) The second implication follows immediately from the assumed universal neces-
sity, since Op * Oq holds also in the worldfp (w), so that q must be necessary in
fp (w), so thatfq 0 fp (w) = fp (w), from the definition offq.
Unlike strong and weak implication, whether or not p>q is true in a given q.t.f.
world w (i.e., wEI p>q I) is decided by what happens in the worldfp (w). This is
emphasized by the remarks below.

'*
Remark 3. In general, it may happen that w E 10 (p q) I but w f/. I p>q I. Thus,
let p be a proposition, w = W1 = {a1,lh}, and fp (W1) = W2 = {a2,,82}. Suppose
p, qE a1, -p E ,81. Since W1 R w2, we may suppose, by relabeling if necessary, that
a1 = a2. We must have p E,82, since fp (w) E lOp I, but it may not be true that
W2 E 10 (p * q) I, so q may be false in ,82. The simplicity of this reasoning is more
apparent from the following diagram.
Fig. 3: Reasoning with q.t.r. worlds

The figure summarizes the reasoning

used above. Each q.t.f. world is shown
as a pair of branches, and the relevant
statements which are true in each
branch are displayed. The anvw gives
the action oftI'"

Remark 4. 0 P 1\ -0 p .*. p > 0. That is, a statement which is necessary nowhere,

Du
but possible somewhere, satisfies p > O.

Remark 5. The preceding remark shows that it may happen that

wEI p > q I but w f/. 10 (p * q) I. Thus, the reverse implications in Theorem 1 are
non-trivial. The next theorem states a natural condition under which the reverse
implication is valid.

Theorem2. Iffp (w);t:. 0, \lw E I Op I thenw EI p > q I .*.W E 10(p *q)l.

Proof. wE lOp I .*.fp (w) = ro, and fqofp = fp .*.fq (w) = wE 10 q I. That is,
under the hypothesis, a world in which p is necessary must also be a world in which q
is necessary. On the other hand, if wEI Op I andw f/. lOp I,
w = {a,,8}, pEa, p fI.,8, thenw Rfp (w) * wnfp (w) ;t:. 0 (sincefp (w) ;t:. 0), which
implies a E fp (w) = fq 0 fp (w), by hypothesis, which implies q E a, from the definition
offq. Hence, q must be true in anyp-branch of an admissible q.t.f. world. This argument
is summarized in the following two diagrams.
186 CHAPmRVIB

=>

=>

Fig. 4: Proof or Theorem 2

Cor. Iflp (m) 0, 'If mEl p I then

Du (p > q) .=>. Du (0 (p => q
Remark 6. Combining this corollary with Theorem 1, we see that, under the hypoth-
esis, Du (Op=>Oq) .=>. Du(p=> q). That is, under the hypothesis, for any q.tJ.
world, if the necessity ofp implies the necessity of q, then the possibility ofp implies
the possibility of q. This corresponds to Assumption 2 of Van Fraassen (ref. 24, p 87),
and Axiom 1.3 of Pool, which states that if, for any state, Pr ~ p = 1 } => Pr ~ q = 1 }
then Pr ~ p -:;f:. o} => Pr { q -:;f:. o}. The assumption was needed to prove a result of the
kind of'ibeorem 3 below:

Such a result follows from the above corollary using - - =0 .

Remark 7. This is assumed by the above authors presumably because it does not seem
possible to derive this result in any standard modal logic where the result seems
unnatural. (As Van Fraassen says, '[It] could be doubted.') On the other hand we have
found a natural condition for this to be true: Ip (I'Ll) -:;f:. O. With an ambiguous reality, a
dynamical variable has a definite value only in a world in which it is necessary. The
condition therefore refers to those propositions for which starting from any world in
which the proposition is possible one can always find a closest world in which the
 QUANTUM-MECHANICAL TIME 187

proposition is necessary. Had there been no quantum theory of measurement, it would

still have been quite natural to call such a proposition quasi-measurable!

Remark 8. We recall that ap-world is a world in whichp is either true or necessary

(depending upon whether the world in question is a truth-functional world or a q.t.f.
world).

Theorem 3. Iffp (w ) ::F- 0, V wEI 0 p 1 then

Du(P > q) .=>. Du(-p > -q)

Proof It follows from Theorem 2 that there does not exist an w with a p A -q
truth-functional subworld. Hence,

Du(P > q) .=>. Du(D (P => q .=>. Du(D (-q => -p

and the conclusion follows by Theorem 1.

Remark 9. We recall that the quasi-measurable propositions, p EM, are precisely

those for which p satisfies the hypothesis of the previous two theorems. We also recall
that p ..L q == Du (P > -q), so that there is no w with a truth-functional subworld in
which p A q is true.

Theorem 4. Ifp, q EM, andp ..L q thenp V q EM.

Proof
If wEI (p V q) l,Jp Vq (w) ::F- 0 trivially, while if wElD -(p V q) 1
there is nothing to prove. Hence, we need to consider only the case where p V q is true
in exactly one branch: wEI 0 (P Vq) 1 but w fI. 1 D(p V q) I. But since p ..L q, exactly
one ofp, q can be true in this branch. Ifp happens to be true thenfp Vq (w) = fp(w),
since no world with w' E 1 Oq 1 is accessible, the (unique) closest p world (which
exists, since p E Mby hypotheses) must necessarily be the (unique) closestp V q world.
Sirnilarly,Jp Vq (w) = fq (w) for wEI Oq I.

Cor. ~f {Pi };:1 are such that eachpi EM, and Pi'S are mutually orthogonal, then
00

V Pi EM.
l=l

Proof Consider w = {a, P} . If all Pi'S are false in both a and p, there is nothing to
prove. If some Pi is true in a, and some pj is true in p, there is nothing to prove, since
188 CHAPTERVIB

then oi V pi. If some Pi is true in a and all otherpj's are false inp, thenfVpi = fp; as
above. No other case is possible, since at most one Pi is true in any a.

Remark 10. We recall that the measurable propositions,p E M,are precisely those
for which both p and -p are quasi-measurable. We also recall that
p q .=. q = p vs,forsomesEM suchthatpl.s.

Remark 11. Itistrivialthatp q .=>. ou(p>q).Thenextpropositionestablishes

the conditions of compatibility or commutativity under which the converse is valid.
N (fp) = {w E U, fp = 0 }.

Theorem 5. Ifp, q EM and 0 u (p>q) then the following are equivalent.

(i)(a)fpofq = fqofp = Ip
& (b) f -q 0 f -p = f -p 0 f -q = fp
(ii)(a)N (fp ofq) = N(fqofp) = N(fp) = 10 -pi
& (b)N(f-qol-p) = N(f-pof-q) = N(f-q) = 10ql

(iii)(a)fq ( 10 -p I ) = 10 -p I
& (b) I -p ( 10 q I ) = I 0 q I
(iv) -p II q E M

(v)p q

Proof (i) => (ii) is obvious.

(iia) => (iiia). If wE lo-p I, andlq (ll) E I p I thenfp (fq (w '. 0, since p EM.
But this contradicts (ii).
(iii) => (iv). Leaving aside the trivial cases which are mapped to 0 or to themselves, we
must have the situation depicted in Fig. 5, since fq leaves lo-p I invariant. Hence

Fig. 5: Diagram for (iii) ~

(iv)
The diagram summarises
the reasoning used (as in
Fig. 3). Under the hypothesis,
a world in which -p is neces-
sary must be mapped by fq to
a world in which both -p and
q are necessary.
 QUANIUM-MECHANICAL TIME 189

f-p "q = fq on ID-p 1 since the world selected by fq is the (unique) closest q world
and also a -p 1\ q world. The other non-trivial case follows similarly from (iiib) since
f-p "q = f-p on IDq I. Since, for p, q E M, D u (p>q) implies p ..L -q, Theorem 4
completes the proof.
(iv) ~ (v). Take s = -p 1\ q E M. Then q = p v s, and p ..L s .
(v) ~ (i) Letq = p vs, andp ..L s. Now, Du (D(p ~ pvs .~. Du (QJ ~ D(pvs so
that fq 0 fp = fp. But p ..L s implies fpvs = fp on ID-s 1 :2 IQJ 1 and fpvs = fs on
1D-p 1 :2 1D s 1 since p ..L s, and p, s E M, so that Theorem 2 applies. But
fp vs (ill) = ill if ill E 1 <> P 1\ <> s I, since p ..L s. Hence fp 0 fq = fp, and the following
lemma completes the proof.

Lemma. p q ~ -q -p

Proof Since p ..L s and p,s E M, q = P Vs ~ -p = -q V s, sincep, q,s E M

by hypothesis, so that weak equivalence is the same thing as strong equivalence by
Theorem 2. Also, -q > > (-p 1\ -s) > > -s ~ -q..L s.

Cor. 1. > > is transitive.

Cor. 2. M ordered by > > is a complete relatively orthocomplemented orthomodular

poset.

Proof Orthomodularity follows directly from the definition of the order relation.
Completeness follows from the corollary to Theorem 4.
VII COSMOLOGICAL TIME
ABSTRACT. Our objective is to formulate questions about time in their true setting - cosmology. Even
to formulate these questions one needs a theory. But, for cosmology, the available theories are unsatisfac-
tory and the slender empirical evidence admits a diversity of interpretations.

We review the evidence for an expanding, fmite-age cosmos, using general relativity and in relation to
questions about the global structure of time. In interpreting the evidence we emphasize the implicit
assumptions, which often go unstated.

We conclude by formulating a number of questions about time. Is there a proper clock? What is the exact
relationship between the time asymmetries? Is time asymmetry imperfect? Could it be local? Is the notion
of an asymmetry conceptually adequate to capture a sufficient variety of structures of time?

1 Introduction

I the
N the previous parts of this exposition some things have, I hope, emerged clearly. With
solution of the problem of time measurement, in relativity, two observers may
differ on 'presentness' or the simultaneity of events. This forces a (static) B-series view
of time (Chapters I-III); simultaneously, one must abandon the mundane distinction
between a dead past which has ceased to exist and an unborn future yet to come into existence.
More insidiously, the differential equations of physics presuppose that time has
(locally) the topology of the real line. This view is fortified by the use of 'instantaneous'
differential equations, which implies that past and future depend in a determinate way
on the present. This forces us to abandon also the past-linear, future-branching,
tree-like topology of mundane time (Chapter VIA).
Conventionally, therefore, time in physics is the real line, the only problem being
to distinguish its head from its tail. But the conventional way of using thermodynamics
to distinguish between past and future fails. Given the information-theoretic interpre-
tation of entropy, the entropy law is indeed close to the intuitive belief that the future
is more uncertain than the past. But the century-old attempts to derive the entropy law
from Newtonian dynamics cannot possibly succeed: in view of the generalized recur-
rence and reversibility paradoxes, entropy must stay constant if evolution is either
deterministic and time symmetric (e.g. evolutionary laws given by ordinary or partial
differential equations), or else stochastic and history-independent (Markovian). Such
attempted derivations necessarily require either an ad hoc time-asymmetric postulate
(such as ergodicity or mixing as opposed to sorting) or an assumption about the
(empirical) nature of the cosmos. Even with these assumptions, with the 'instantaneous'
formulation of physical laws, the resolution of the paradoxes remains unsatisfactory
(Chapter IV) and the physical origin of time asymmetry remains mysterious.
One can attempt to avoid both types of assumptions by resorting to the infinite
number of degrees of freedom provided by the field; but, in view of the divergence
problems of field theory (Chapter VA), it is more attractive to work with the essentially
equivalent approach via direct-action electrodynamics. It is possible to resolve the
recurrence and reversibility paradoxes using time-asymmetric evolution given by the
 COSMOLOGICAL TIME 191

many-body functional differential equations of motion of retarded electrodynamics.

The 'infinite number of degrees of freedom' of the field are now seen to refer, more
properly, to history-dependence introduced through the finiteness of the speed with
which interactions propagate. But the assumption of retarded interactions can be
justified only in the context of cosmology, as in the absorber theory of radiation. The
Wheeler-Feynman-Davies absorber theory, which sought to base the electromagnetic
arrow of time on the thermodynamic arrow, is internally inconsistent. The Hogarth-
Hoyle-Narlikar absorber theory is externally inconsistent, and the only remaining
theory (mine) predicts the existence of small amounts of advanced interactions, or a
microphysical tilt in the arrow of time (Chapter VB). The non-locality due to such a tilt
is not a serious problem (Chapter VIA)
Thus, neither thermodynamics nor electrodynamics can, by itself, provide a distinc-
tion between past and future. Inevitably, cosmological issues tend to creep in. The
hypothesis of a tiny tilt in the arrow of time, resulting in a microphysical structure of
time, the proposed raison d'etre for the change of logic required by quantum mechanics
(Chapter VIB), is also founded on cosmology. Questions about the global structure of
time ('does time have a beginning?') are rather obviously questions of cosmology. Thus,
any assertion about time in physics relates, implicitly or explicitly, to a cosmological
model. (The more usual, but less satisfactory, statement is: 'the cosmological arrow of
time is fundamental'.)
Hence, cosmology is the only appropriate context for the study of time.

1. 1 Speculation in cosmology

But it is difficult to confront the riddle of time in its proper setting: cosmology. Unlike
the other areas of physics covered so far, cosmology is very speculative. Sciama1
provides an interesting conclusion for papers on cosmology:
The reader may be appalled by the amount of speculation in this article. All I can say in my
defence is that I find it hard to believe that it is all wrong and/or misleading. Even if only a
small part of it is found to be on the right lines, we would still be witnessing the birth of
imaginative new possibilities for our understanding of the universe ....

Should one, therefore, speculate more so that more, in absolute terms, would be
on the right lines? Surely Sciama doesn't intend this: when speculations are piled on
the conclusion tends to have probability zero. Moreover, opening too many new
possibilities can have the unintended effect of prematurely closing basic questions for
the worst of all reasons - that they are old hat.
While it is obviously more interesting to talk of 'imaginative new possibilities', and
while spe<:ulation is unavoidable in cosmology, it seems desirable that speculation
should proceed on the principle that a directed search is more efficient than a random
search. To direct the search one needs to identify and focus upon the important
questions. The objective of this chapter is to identify and sharpen the basic questions
about time in the setting of cosmology.

1. 2 Choice of a theory

Which theory should one use to formulate these questions? All the available choices
192 CHAPrERVII

are unsatisfactory. Newtonian mechanics is hardly a contender, especially if the object

of doing cosmology is to study time.
The problem with relativity is that it addresses only the problem of time measure-
ment in the Newtonian theory. We saw in Chapter IT that Newtonian mechanics is
inseparably tied to gravitation: force and equal intervals of time are both unknowns, so
that Newtonian mechanics is meaningless without an accompanying force law. General
relativity theory (GRT) modifies Newtonian physics, by introducing a proper clock and
dispensing with the notion of force; but it cannot do without the umbilical connection
between mechanics and gravitation which reappears in the form of the principle of
equivalence: the 'inertial' mass must be the same as the 'gravitational' mass. (Perhaps
the connection is a historical accident arising from Newton's back-calculation of force
from Kepler's laws.) Similarly, GRT, like Newtonian mechanics, pays no attention to
questions about the structure of time or matter - time is a line, bodies are points,
trajectories are curves, physics is geometry. These questions are far more serious in
GRT, where there is no satisfactory 'statistical' continuum model of matter, and a
point-mass model of matter seems impossible. (Nevertheless, the geodesic hypothesis
asserts that geodesics are the trajectories of point-like 'test-particles'.) Traditionally,
one expects GRT to fail when these questions become important, i.e., when matter is
at high densities, at epochs such as the 'beginning of the universe'.
But quantum gravity is not satisfactory either, since it has yet to find an internally
consistent formulation. Perhaps the wisest course, therefore, is to follow the
Wittgensteinian dictum, 'Whereof one cannot speak, thereof one must be silent'.
Physics must, however, proceed with the available theories, and I will choose
classical GRT. Time may not be a line, bodies may not be points, trajectories may not
be curves and physics may not be geometry, but at least the internal consistency of GRT
is not in question. At the physical level, even though unsatisfactory, it is still the most
satisfactory theory available. Moreover, as argued earlier, relativistic many-particle
dynamics, with a tilt in the arrow of time, provides at least a qualitative model for
classical axiomatic quantum mechanics.

1. 3 The cosmological arrow of time

Granting the assumption of linearity of time (B-series) implicit in GRT, is there a basis
in cosmology for the earlier-later relation? Was there a beginning of time? Will there
be an end of time? Is the global topology of time such that a binary earlier-later relation
is adequate?
The earlier-later relation, or the cosmological arrow of time, is based on cosmo-
logical expansion. In 2-5, we review the basic lines of observation and theory which
seem to converge towards the belief in cosmological expansion. We also review the
theory and evidence which support the belief in a finite age for the universe, and a
beginning and end of time. The empirical evidence available for cosmology is even more
suspect than the theory: it is slender and admits a diversity of interpretations. The point
of going over this material is to stress the underlying assumptions (which often go
unstated), so that the nature of the speculation becomes clear - there is an evirlent
need for robustness in dealing with so fundamental an issue as time. Finall', we
formulate some problems concerning the nature of time.
 COSMOLOGICAL TIME 193

2 Darkness of the night sky: "OIbers' " paradox

The simplest cosmological observation, one which doesn't require any sophisticated
instrumentation, is to look up at the sky and see the stars. On a clear night, far away
from the lights of any city, the poet in us is likely to marvel at the beauty of the
star-studded sky. The cosmologist, however, is more likely to wonder at the extraordi-
nary fact that the night sky has a dark background.
Why are stars visible at all? Why don't they obscure each other by lighting up the
entire sky, like the sun in the day or the city-lights at night? Why is it, the cosmologist
tends to ask, that every point of the night sky is not as bright as any other?

2. 1 Olbers' paradox

Indeed, Olbers' paradox2 suggested that this is exactly what should happen if an infinity
of stars blazed, as was once supposed, for an eternity of time. (Newton thought that the
universe, since static, must be infinite, else it would collapse; today, this argument would
be regarded as a misunderstanding of infinity.)
If the universe were infinite, and uniformly populated with stars, every line of sight
would ultimately intersect a star. The standard analogy is to a forest packed with trees,
where every line of sight would ultimately intersect a tree. Where are the missing stars?
One could also think of the paradox in another way. While a star would certainly
get fainter, the further away it is, it would also look smaller. Moreover, the apparent
brightness of a star decreases with distance at exactly the same rate as its apparent size.
So the ratio of the two (apparent brightness/apparent size) would remain constant. The
conclusion is that each point of the sky would be as bright as any other point, so that
each point would be as bright as the sun! What happened to the missing starlight?
Henrich Wilhelm Olbers himself thought, in 1823, that such intense starlight would
not inconvenience life on earth, except for astronomers, who would find it difficult to
see the stars! This did not seem so strange some one and a half centuries ago, when
William Herschel had thOUght that life might exist on the surface of the sun.
The more popular version of Olbers' paradox, which develops an argument given
by Edmund Halley in 1721, points out that, with an infinite sphere of uniformly
distributed stars, the intensity of starlight should actually diverge. Dividing space into
thin concentric shells of equal thickness, and assuming the stars to be uniformly
distributed, each shell contributes an equal flux of starlight, regardless of its radius.
Therefore, the total intensity of starlight should actually diverge.

2. 2 The absorption solution

One way to resolve the paradox is to suppose that starlight is absorbed somehow before
it reaches the eye. Cheseaux, in 1744, proposed that the distant stars were hidden from
view because of an intervening obscuring medium. Cheseaux's analysis suggested that
the stars could be seen only out to some 3 X 1015 light years. This prevented the starlight
catastrophe.
The flaw in the absorption argument is that the hypothetical interstellar medium
would absorb and re-emit starlight, and would eventually reach a stage where absorp-
194 CHAPTER VII

tion would equal emission. Long before such an equilibrium is reached, the stellar
background would become appreciably brighter than it actually is. So one must look for
an invisible means of absorption.

2. 3 The expansion solution

Bondi suggested that, in a steady-state cosmology, the expansion of the cosmos provided
the required invisible means of absorption. In an expanding universe, light from the
distant stars would be redshifted and enfeebled. Thus, in the steady-state cosmology,
the darkness of the night sky assumes a profound significance - it indicates the
expansion of the universe.
The redshift solution does not work for a general expanding cosmos, where the
redshift reduces the intensity of starlight by only a few percent compared to a static
cosmos. But the darkness of the night sky is nevertheless qualitatively important: it
shows that the stars are in disequilibrium with the surrounding space. How did this
disequilibrium originate? and how is it maintained? A plausible answer is that cosmo-
logical expansion provides the means for maintaining disequilibrium. Qualitatively, there-
fore, the expansion of the universe continues to be a key feature in the resolution of
Olbers' paradox.

2. 4 The finite age solution

A study of the history of the riddle of cosmic darkness discloses a curious inhibition in the
thoughts of astronomers.... Astronomers in the eighteenth and nineteenth century recoiled from
the monstrous idea that when we look out in space we look back in time to the creation of the
universe.... Many ofthese astronomers were ordained ministers active in the affairs ofthe society
and dependent on patronage. Few felt impelled to affront society with a quixotic attack on the
Book of Genesis and the Mosaic chronology. 3

By dropping the assumption of (local) conservation of mass/energy, the steady-state

theory shows that cosmological expansion is logically independent of any notion of a
finite age for the cosmos. Therefore, the cosmos might expand or evolve without
necessarily being of finite age. But if matter is conserved, one would expect an expand-
ing universe to have a finite age. For if the universe is expanding, it must have been a
little smaller yesterday, and still smaller the day before. So one could identify the age
of the universe with the time elapsed since the beginning of the expansion. (Such an
identification is admittedly a little simple-minded, for the beginning of the expansion
may not be a beginning of time.)
Olbers' paradox can also be resolved by supposing that the universe has a finite age.
Let us consider the riddle first in the form: what happened to the missing starlight?
In a static universe of finite age filled homogeneously with constant luminosity stars, the stars
act as localized heat sources in a cold space. Gradually, as the starlight accumulates, the
temperature of space rises, until eventually equilibrium is reached when every star is
absorbing energy from the surrounding space at the same rate as it emits (this is the situation
Olbers wished to avoid). As this condition is approached, the entropy of the system assumes
its constant maximum value. On the other hand, if the universe expands faster than t113, then
however long the stars burn this equilibrium will never be reached, and there will always be
a temperature gradient around the stars.4
 COSMOLOGICAL TIME 195

Davies gives an estimate 1027 s for the 'relaxation' time to reach equilibrium in a
static universe. Since this is enormously longer than any cosmological time scale, Davies
concludes that the expansion of the universe is irrelevant, except to maintain disequi-
librium. On the other hand, Harrison gives an estimate for the current intensity of
starlight as some 10-13 of that anticipated by Olbers' paradox. If we were to put the two
figures together we would arrive at an estimate of 1017 s, or 1010 years for the age of the
universe, necessary to account for the estimated darkness of the night sky.
The finite age solution also applies to Olbers' paradox in the form: what happened
to the missing stars? Here one must take into account the finite speed of light together
with the luminous lifetime of stars, and the occultation of the distant stars by the nearer
stars. A plausible solution was expressed poetically by Edgar Allen Poe in 1845:
the only way we could comprehend the voids which our telescopes frod in innumerable
directions, would be by supposing the distance of the invisible background so immense that
no ray from it has yet been able to reach us at all.

In a universe with a finite lifetime, of course, the light from the 'invisible background'
might never reach us.

2.5 Summary

Harrison gives an eloquent summary:

A dark night sky has two interpretations. According to the first, stars cover the sky, and this
interpretation is true in the steady-state universe. The solution of the riddle is that the light
from the distant background stars has been diluted and redshifted by expansion. According
to the second, stars fail to cover the sky, and this interpretation is true in the big-bang
universes. The solution of the riddle in this case is that the distant background stars are
invisible because their light is in transit and has not reached us (and may never reach us) ....
Through the starless gaps of the night sky we look far out in space and far back in time to
the early universe .... We see what our immediate forebears feared to see: the creation of the
universe written across the heavens. The realization that the big bang covers the dark night
sky, a big bang that was once incandescent and is now redshifted by expansion into an
infrared gloom, is a fitting climax to an old and celebrated riddle.

A couple of things need to be added to this conclusion. The expansion of the

universe provides a basis for the disequilibrium required on both interpretations. If
matter is conserved, expansion suggestst a finite age. So, on any interpretation, expan-
sion of the cosmos seems to be the key feature in accounting for the darkness of the
night sky.

t The notion of a 'fmite age' for the universe cannot be disconnected from the cosmological model being
used - one doesn't really have in mind a world suddenly created by God (complete with an external
measure of time), in which one can only wait for a distant starlight catastrophe!
196 CHAPTER VII

3 Cosmological expansion and redshifts

The simple observation of the darkness of the night sky strongly suggests that the cosmos
expands, and also perhaps that it has a finite age. We now examine these issues more
systematically.
How should one go about testing the hypothesis that the cosmos expands? Expan-
sion of the universe means that all the stars, or at least most of them, are moving away
from us.
Of course, we have no reason to suppose that all the stars have a particular antipathy
for us. (This is sometimes called the Copernican Principle of Modesty: we do not reside
in an especially privileged or underprivileged place in the universe.) So it must be the
case that all stars are moving away from each other. In the standard Hubble-Bubble
analogy, the galaxies are like dots distributed on the surface of an expanding balloon.
The limitation of the analogy is that the universe does not expand into anything - the
universe, by definition, includes all that there is.
So if the universe expands, then an observer located anywhere would find that light
from other stars would be redshifted,5 due to relative motion.

3.1 Hubble's law

Consequently the frequency of light vibrations diminishes with increasing distance from the
origin of the coordinates. The lines in the spectra pf very distant stars or nebulae must therefore
be systematically displaced towards the red, giving rise to a spurious positive radial velocity. 6

De Sitter's observation was inspired by his theoretical analysis of solutions of the

Hilbert-Einstein equations. Wirtz in 1918-1924 was the first to try to obtain velocity-
distance relationships by applying the Doppler formula to redshifts. As more data on
redshifts were assembled, Hubble was able to fit a linear relationship to the data.
According to Hubble's law, the further away a galaxy is, the faster it is moving away. If
v is the recession velocity of a galaxy, and d its distance, then

v=Hd (1)

where H is called the Hubble's constant.

Hubble did not directly measure the velocity of recession of distant galaxies.
Rather, he measured the amount by which the light from them was redshifted. (This is
done by identifying a series of lines, due to the more common elements, that have all
been shifted by the same amount.) The recession-velocity interpretation of the redshift
is widely, though not universally, accepted, but it is undeniable that light from other
stars is redshifted. Rather more problematic is the fact that Hubble could not have
measured the distance of the distant galaxies directly. He deduced the distance by
measuring the brightness (or apparent luminosity) and comparing it with the brightness
of certain other galaxies called standard candles.
This procedure is very indirect, and Burbidge has pointed out that had Hubble
applied the modern values for both these quantities, to his sample of galaxies, he would
 COSMOLOGICAL TIME 197

not have obtained anything like his law! Thus, it is better to speak of the redshift-
apparent luminosity relationship rather than a velocity-distance relationship.
In general, Hubble's 'constant' is expected to vary with the epoch, over cosmolog-
ical time scales. Variations due to difficulties in measurement are another matter: the
measured value of Hubble's 'constant' seems to have changed by an order of magnitude
each decade! The current accepted value of Hubble's constant is l00h km Sl Mpct,
where h denotes an 'uncertainty factor' with a value somewhere between 0.5 and 1.

3. 2 Other interpretations of the redshift

The redshift of a given galaxy may be an indication of its recession velocity; but, a
redshift may not relate to recession velocity in all cases.
Quasars provide an instance where the recession-velocity interpretation of redshift
has been seriously challenged. Quasars, or quasi-stellar objects (QSO's) are strange
objects which may not infrequently have very high redshifts7 (redshift factor z = 4). H
these objects obey Hubble's law, they must be very far off indeed. In that case, from a
knowledge of their apparent luminosity, one can conclude that they must be pumping
out enormous amounts of energy, compared to the average galaxy. The surprising thing
is that the luminosity of these objects sometimes varies so rapidly that one infers that
these objects are rather compact - no larger than the solar system.
According to the 'mainstream' line of thinking, these objects are active galactic
nuclei, inhabited by supermassive black holes which ingest stars at a frightening rate to
spew out enormous amounts of energy. The question of the mechanism of energy
production in QSO's is important only if one applies Hubble's law to QSO's; other lines
of thought have met with more than scientific resistance. In a scandalous incident, H.
Arp, who has harped on the apparent association of quasars with other low redshift
galaxies, was denied telescope time. 8 According to Arp, the probability of these
associations arising by chance might be some 10-200; according to some others this
probability is actually 1, because of statistical selection effects, and the question of
apparent association of QSO's with low redshift galaxies is unimportant. The debate on
quasars will no doubt continue, though this appears to be a unique case of a difference
of opinion which spans 200 orders of magnitude !
To summarize, the observation of redshifts is a further indication that the cosmos
expands. However, Hubble's law, though very plausible, may need to be applied with
some discrimination.

4 The Friedmann models

The expansion of the universe is also a feature of the simplest cosmological models
formulated using GRT.

4. 1 The cosmological principle

The most attractive feature of the Friedmann-LeMaitre-Robertson-Walker (FLRW)

198 CHAPTER VII

models is that they can be derived from the Hilbert-Einstein equations only by assuming
the cosmological principle (homogeneity and isotropy).
Isotropy is just the simplest assumption one can make about the cosmos: the cosmos
looks about the same in all directions. More mathematically, at any instant, the group
of isometries, which leaves a point fixed, should be just the group of rotations, SO(3),
corresponding to orthogonal matrices of determinant 1.
Homogeneity means that any point of 3-space is like any other point. Mathemati-
cally, at any instant, the group of isometries acts transitively, so that one can isometri-
cally map any point to any other point. Now, GRT relates the geometry of spacetime
to the distribution of matter, and on almost any scale from the nuclear to galactic, the
distribution of matter is known to be inhomogeneous. But homogeneity suggests itself
naturally to the Euclidean imagination, and does not contradict the existence of
'small-scale' structures like planetary systems, stars, galaxies, or even clusters of galax-
ies. The idea is that, on a cosmological scale, one can ignore the observed discreteness,
just as in the continuum treatment of a fluid, one may ignore the molecular structure
of the fluid. Of course, if future observations were to confirm the existence oflarge-scale
structure, in the form of say 'clumpiness' in the matter distribution or huge voids
between superclusters of galaxies, then it would become necessary to rethink the
assumption of homogeneity.

4. 2 The FLRW line element

By assumption, the 3-spaces t = const. are homogeneous and isotropic. Since any point
of 3-space has been assumed to be like any other point, it is very plausible that the
Gaussian curvature K is everywhere constant. (This follows from Schur's theorem, if
the dimension of the space is ~ 3.) For a space with constant Gaussian curvature K,
there exist coordinates i so that the metric takes the form

(2a)

If K ;f 0, we may make the transformationi -+ i IV'IKT, and rewrite the metric as

(2b)

where k = 1, and R = IKI- V2 is the radius of curvature, and the Gaussian curvature
K = k1R2. The advantage of writing the metric in this form is that one can include the
case k = 0 by taking R constant. By isotropy, we may use the usual curvature coordi-
nates, to obtain the FLRW line element:
 COSMOLOGICAL TIME 199

(2)

The 'scale factor' R=R(t) may be loosely thought of as the 'radius of the universe' at
any instant. Here r is not the 'distance' from the origin, but is the curvature coordinate:
a sphere with radius of curvature r has a surface area of 4n'r2, though not usually the
volume (4/3 )3lr3 The curvature parameter k = 0, -1 or + 1. This parameter decides
the type of the cosmos; k= 1 corresponds to the case of recontraction, and k= -1 to the
ever-expanding model.
Homogeneity and isotropy also imply that the matter tensor T",v in the Hilbert-Ein-
stein equations

G",v = 8;r G T",v (3)

has three equal eigenvalues, corresponding to the pressure p, and one eigenvalue
corresponding to the energy density E. At the present (matter-dominated) epoch the
pressure (of the cosmological fluid) may be neglected in comparison with energy
density, so that (3) reduces to the rather simple equation

R? = 2GM_ k (4)
R

whereM (= E 4;rR3/3) maybe loosely thought of as the 'mass' of the universe, subject
to the realization that anything like 'mass' cannot be easily defined in a curved
spacetime, which is not asymptotically flat.
One can also carry out the entire analysis including the cosmological constant, A,
in the Hilbert-Einstein equations (3). This constant was first introduced by Einstein
into the field equations in order to be able to obtain a static cosmological solution to
the field equations. It performs the role of a negative pressure, and could give a
cosmological model which agreed with the idea of the cosmos generally accepted at that
time. Later on, Einstein called it 'the greatest blunder of my life'. This constant was
abandoned for a long time, until recent work on inflation, which has sought to revive it.
In addition to homogeneity and isotropy, the above derivation of the FLRW line
element also assumed that time is a parameter, in assuming that spacetime may be split
into a family of space like 3-spaces (foliation). Dropping this assumption ( 8) leads to
cosmologies like the GOdel cosmology, which admit a tilt in the arrow of time.

4.3 Identification with observable quantities

In the context of the FLRW models, it is possible to identify the theoretical analogues
of the basic observational quantities of interest. Thus, the Hubble constant H === R IR,
while the deceleration parameter q === - Rlf1R2. The deceleration parameter measures
the rate at which the expansion of the universe is slowing down; it is even harder to
estimate than Hubble's constant.
200 CHAPfER VII

The critical density for recontraction is given by

po =
3 {R}2
87tG R .
(5)

The estimated density of the universe is usually expressed as a fraction of the critical
density, using the density parameter o.=plpo. This parameter, too, is hard to estimate,
and its current value, based on the density of luminous matter, is ::::::O.l.
The curvature parameter k is related to the density parameter:

< < (6)

k = 0 according as 0. = 1.
> >

In principle, one can distinguish between the three Friedmann cosmologies using
geometry, or the sign of the Riemannian curvature. (The Riemannian curva-
ture =k1R2 = H(o.-l).) The sum of the interior angles ofageodesic triangle would be
respectively greater than, equal to, or less than Jr, corresponding to (6). Currently
measurements of this kind are not feasible; but there are reasons to suspect that the
luminous matter is only a small fraction of the matter in the universe ( 7), so that the
above estimate for the density parameter may be incorrect.
The observed value of the Hubble constant also provides an estimate for the age
of the universe, or rather the expansion age, lIH =lOho- 1 Gyr. One can also hope to
infer the physical age of the universe from radioactive dating of minerals, meteorites,
or lunar rocks. This gives galactic ages in the range 12.6 Gyr to 19.6 Gyr. Yet another
method is to estimate, using stellar evolution theory, the age of the oldest star clusters,
like the globular clusters. This suggests an age of at least 15 Gyr.
With the FLRW models, it is possible to obtain the age, t, of the universe, more
precisely, by integrating the Friedmann differential equation (4):

1
to = Ho II y ] -21 dy
0 [1-0.0 + (0. )
0
(7)

where a subscript 0 denotes the value at the present epoch, and it has been implicitly
assumed that the universe has been in the matter-dominated state for most of its life.
The relationship to < HO-l always holds. In the flat case (k=O, 0.0= 1), (7) yields
to = ~HO = 13 Gyr, for ho = 0.5, a value which seems to be below the observational
lower limits.
To reconcile this disagreement, one may call into question one or all of the
following:
(i) omission of the cosmological constant (this would permit HO-l < to),
(ii) radiometric dating methods (physical constants may vary over cosmological time
scales),
(iii) the measurement of Hubble's constant (which has only shown a tendency to
 COSMOLOGICAL TIME 201

increase in recent times),

(iv) the assumption of homogeneity and isotropy.
In a way (iv) represents the most conservative possibility. Is there any way to test
homogeneity and isotropy?

5 The cosmic microwave background

A matter-conserving expanding universe must have been smaller yesterday, and so

there must have been a time, in the beginning long long ago, when all the matter in the
universe, all the stars and galaxies and all the dust between them, must have been
crushed together into a small space. This would have caused an almighty explosion -
the big bang - the biggest known bang, which set off the expansion of the universe.
In the early phase of the universe the radiation from the explosion would have
interacted strongly with matter. As the universe expanded and cooled, matter became
decoupled from radiation. Radiation from this epoch of last scattering, though
redshifted into the far infra-red, should still be observable, and should give information
on the surface of last scattering.
On the basis of this reasoning, Gamow had predicted that relic radiation from the
big bang would be observable as background radiation. Since every line of sight
stretches back to the big bang, this radiation would seemingly come from all directions.
Gamow calculated that this radiation would appear as black body radiation from a black
body with temperature 60K The cosmic background microwave radiation (CMBR) was
actually observed by Penzias and Wilson in 1967. The spectrum is rather close to that
of a black body at 2.750K Deviations from the black-body spectrum are an active area
of experimental study.
Because of its high degree of isotropy, the 30K CMBR has almost universally been
interpreted as relic radiation from the primordial cosmic fireball, sounding the death
knell of the steady-state theory. Anisotropies are observed - prominent among them
is the dipole anisotropy, which is consistent with a motion of the local group of galaxies
at 600 km Sl relative to the rest frame defined by the CMBR. In a manner of speaking,
the resembles the aether, so that the dipole anisotropy corresponds to the observation
of an aether drift.
There are other ways of interpreting the dipole anisotropy - it could conceivably
arise due to departures from homogeneity and isotropy in the primeval mass density.
But in that case one would expect a comparable quadrupole anisotropy which is not, in
fact, observed. ll
Such expectations are based on the classic work of Sachs and Wolfe9 on linear
perturbations of homogeneous cosmological models, and the effect of these on the
CMBR. Recent rethinking on the problem1o suggests that the analysis carried out by
Sachs and Wolfe (and by others who claimed gauge invariance) in fact depends on the
choice of a specific gauge without a clear physical or geometric meaning. Therefore,
the question of interpreting anisotropies in the CMBR remains open. In particular, one
cannot yet rule out the possibility that these may well be due to large-scale departures
from homogeneity and/or isotropy.
202 CHAPTER VII

6 The beginning of time

6. 1 The initial singularity

If the universe is expanding and originated with a big bang, a question naturally arises:
was the big-bang really the beginning of time? Currently, the very fashionable, but
'appallingly speculative', trends are to talk of quantum gravity during the first 10-32 s
after the big bang, or inflation or topological defects like cosmic strings. However, it is
possible to consider matters in a more conservative way, from the point of view of
classical general relativity.
Att=O, all the FLRW models have a 'singularity'. Should one take this 'singularity'
seriously, as the beginning of time, or is this an artefact of the process of mathematical
and physical modeling? Note that this question is being asked from the point of view
of classical relativity theory. Thus, the question is, for example, 'Would the 'singularity'
disappear with another choice of coordinates, or with small departures from spherical
symmetry?' Could rotation, for instance, forestall a singularity?

6. 2 Singularities and black holes

To understand some of the problems here, let us go back to the first solution of the
Hilbert-Einstein equations, due to Schwarzschild. This solution is derived only by
assuming spherical symmetry and asymptotic flatness: that far away from the central
mass, spacetime is approximately flat. The solution corresponds to a spacetime with the
line element (in curvature coordinates):

This line element has two sorts of peculiarities. One occurs at r=2m (= 2Gmlc 2) and
is due to a bad choice of the coordinate system. 12 The other at r=O is believed to be a
genuine 'singularity'. The surface at r=2m only acts like a one-way membrane at which
time stops.
Laplace derived the value r=2Gmlc 2, by considering (in the Newtonian theory) the
mass for which the escape velocity would equal c, the velocity for 'particles oflight'. But
on the Newtonian theory the surface r=2Gmlc2 does not act like a one-way membrane:
a particle coming from infinity may enter the surface and leave it.
The Schwarzschild solution shows that the point-mass model of matter used in
Newtonian physics creates problems in general relativity: there ought to be a one-way
membrane around a point mass. On the other hand, one may show that, for the usual
case of a finite density of matter, such as the density of matter in a star, both singularities
are absent.
But if a star with mass beyond the Chandrasekhar limit begins to collapse, the
one-way membrane may be eventually exposed. This is a black hole. For an observer
on the star (poor fellow!) this collapse would take only a finite amount of time, whereas
for a 'static' observer at infinity the star would take an infinite amount of time to collapse
into the black hole.
 COSMOLOGICAL TIME 203

Two questions now arise. (1) Could the formation of a black hole be prevented by
small departures from spherical symmetry or by rotation? (2) After a black hole has
formed, will the matter inside continue to collapse? In other words, could a genuine
'singularity' be formed?
Penrose's singularity theorem was initially motivated by these questions. Having
defined the notion of 'singularity', the theorem proves that a singularity is inevitable
under certain circumstances. Thus, the answer to question (1) above is no, while that
to question (2) is yes. Hawking observed that Penrose's theorem could be time-reversed
to apply also to the singularity at the beginning of the FLRW models. The conclusion
seems to be that the real universe did originate in a singularity.

6. 3 Cosmological singularities

A coordinate-free way to define a 'singularity' is as follows. A spacetime has a singularity

if it is geodesically incomplete. Null geodesics, for instance, may be thought of as the
paths of light rays, whereas timelike geodesics may be thought of ~ the world-lines of
particles of matter. Therefore, a point beyond which a geodesic cannot be extended
may be thought of as the beginning or end of time for a particle.
In the Friedmann models, every line of sight extended backwards intersects the
initial singularity. But the Hawking-Penrose theorems prove the existence of a singu-
larity under general conditions which do not involve an assumption of spherical sym-
metry, or absence of rotation etc.
The original conditions were: validity of the equations of relativity, positivity of
energy, absence of closed timelike curves, and a timelike convergence condition, which
could be loosely interpreted to mean that there must be enough matter around to
initiate a collapse. Equally, the location of the singularity, or the manner in which the
singularity is approached, may be quite general. For instance, it is possible that the
singularity is approached chaotically through a series of 'bounces', so that, following
evolution in reverse, the cosmos may go through a series of expansion-contraction cycles
before a beginning of time is encountered.

6. 4 Interpretation of singularities: shocks

The d-function must have had a very sad childhood since neither mathematicians norphysicists
recognized it as belonging to their domain. If mathematicians used it, it was as an intuitive
physical notion with no mathematical reality.... On the other hand, physicists usually consid-
ered the d-function, orpoint mass, as a purely mathematical idealization which did not exist in
nature.
Lutzen (1982)
The Prehistory of the Dirac d-Function, p 110

One assumption, however, which is barely dealt with, substantively affects the interpre-
tation of the singularity. The assumption concerns the smoothness of the metric tensor
- it is usually assumed to be at least C1 (once continuously differentiable). Non-tech-
nically, the issue is a bit confusing, since the conclusion too seems to concern smooth-
ness - a singularity suggests a 'blow up' or some sort of breakdown in smoothness.
The usual technicalities, however, tend to obscure the underlying physical issues.
(i) In the absence of a relativistic kinetic theory, how legitimate is the continuum model
204 CHAPTER VII

of matter in GRT? (ii) Does a singularity refer to a beginning of time (or equivalently
to a breakdown of physical laws)? or does it merely refer to the breakdown of smooth
solutions, as in the case of a classical shock wave?
Continuum models of the matter tensor are simply based on the analogy of
compressible-fluid flow. In this analogy, the effect of an explosion is modeled by a shock
wave. There is a thin layer of fluid across which there is a sudden (or discontinuous)
change in pressure, density etc. In the old differential calculus, a discontinuous function
was not differentiable, so physical laws given by differential equations seemed to fail at
a discontinuity. Thus, the belief has been that in 'reality' there are no discontinuities.
Nevertheless, discontinuous models are in use, though they have been usually
regarded as a simplification for a situation which is 'really' smooth. The 'real' situation
is presumably modeled by kinetic theory, which is unavailable in relativity. An alterna-
tive procedure for discontinuous models is to use the integral formulation of physical
(conservation) laws, but this procedure does not work in all cases of interest. More
recently, it has been possible to extend the continuum approach to a discontinuity, and
model the matter tensor by a Dirac 0 function, or even permit discontinuities in the
metric tensor. (The mathematical problem here is identical with the renormalization
problem of quantum field theory, or the Thomson lamp considered in Chapter I.)
In GRT such discontinuous models have an ambiguous status. On the one hand,
there is no well-established alternative to GRT: neither a kinetic nor a quantum theory.
On the other hand, under extreme stress, there may be a real failure of assumed
smoothness. There are no compelling grounds for believing that Nature is smooth, that
Nature obeys classical differential calculus and has been kind enough to have made
every function infinitely differentiable (if not analytic). The smoothness of Nature is
especially in doubt when one is trying to prove the existence of a singularity!
Hawking and Ellis13 put forward rather weak arguments against such an interpretation:
that there is a failure not of GRT but of their tools of analysis. Their assertion that the
'physical justification' for discontinuous models is that they are 'mathematical idealizations
of C2 or C'" solutions .. .' rather begs the question. They go on to admit: 'Although we are
not quite able to prove it we believe that the singularities must be such that the metric cannot
be extended to be even a distributional solution of the Einstein equations ... .'

6. 5 Shocks and hyperbolicity

There are good reasons why Hawking and Ellis fail to prove the existence of a singularity
in this sense: shocks are an intrinsic feature of hyperbolicity, in the sense that any
nonlinear hyperbolic equation could be expected to admit shock solutions. (We recall
that the Hilbert-Einstein equations of GRT were motivated by the notion of 'locality'
or hyperbolicity.)
Lax14 gives a generic example of the case of a single first order equation, arising from
a conservation law. Initial data for this equation may be so prescribed that characteristics
intersect. (These characteristics are the counterparts of null geodesics.) The intersection of
characteristics means that the solution cannot be (uniquely) extended beyond a certain
point, if one continues to assume smoothness. The interpretation is that smoothness must
break down. It must be emphasized that this phenomenon (of inextendable smooth
solutions) is a general feature of (hyperbolic) wave motion, and one can easily see this in
the breaking of waves on a seashore, or hear it in the blast of a firecracker.
 COSMOWGICAL TIME 205

The conclusion is that the interpretation of the notion of 'singularity' may be

fundamentally defective: the intersection of characteristics, and the finite-time break-
down of smooth solutions, is a characteristic feature ofhyperbolicity, and no fundamen-
tal conclusions, such as the existence of a beginning or end of time, may be drawn from
the Hawking-Penrose theorems. On the other hand, with the aid of new mathematical
techniques, using products with the Dirac 6, it may be possible to go 'across a singularity'
so to say, and to deal with 'singularities' mathematically, in the continuum approach. 15
Thus, the expansion of the universe at most implies a dense initial state for the
universe, it does not necessarily imply a beginning of time. We note in passing that
Tipler's no-return theorem (Chapter IV) fails for the same reasons.

7 The end of time: dark matter and rotation of galaxies

Symmetrically to the question about a beginning of time, one can ask about an end of
time. Will the cosmos go on expanding indefinitely? Which of the FLRW models fits
the real universe most closely? The last question can be decided by estimating the
amount of matter in the cosmos. The straightforward approach is to estimate the
amount of matter in the universe from the visible galaxies, leading to Q ,.,. 0.1.
But, one of the assumptions that one must guard against, in cosmology, is that the
visible matter represents all the matter in the cosmos. This remains true even if we
enlarge our wavelength base to include the entire electromagnetic spectrum, and not
just the optically visible part, to define 'visible'. There are other lines of evidence which
suggest that dark matter may dominate the matter density in the cosmos.
One striking line of evidence comes from the dynamics of rotating galaxies,
especially the spiral galaxies.16 On the Newtonian theory, assuming that all the stars
move in circular orbits, the equality of the gravitational and centrifugal forces implies

(9)

where G is the Newtonian gravitational constant, M(r) is the mass of the galaxy out to
radius r, v(r) is the orbital velocity at radius r, and k is a constant dependent on the
geometry of the mass distribution: for a sphere k= 1, for a disk k=2In. Therefore, (9)
implies that

(10)

In the solar system, where all the mass is essentially in the sun, M(r) is a constant
outside the sun. It follows from (10) that orbital velocities of planets decrease as r~.
Mercury, with r=0.39 A.u. orbits with v =47.9 km Sl, whereas Pluto, 100 times more
distant, withr=39 A.U., orbits with one-tenth the velocity v =4.7 km Sl.
On the other hand, for most galaxies, v(r) increases, with increasing r, and flattens
206 CHAPTER VII

out to a more-or-Iess constant value at large distances from the galactic centre. This is
strange because most of the luminous mass is concentrated near the centre of the galaxy,
and decreases as one moves out from the centre.
Even stranger is the observation that the flatness and the general form of the
rotation curve (graph of vCr) against r) is essentially independent of galaxy morphology,
though the 'amplitude' (i.e., the particular constant value achieved by vCr)) may vary.
In some instances, where it has been possible to carry out velocity measurements out
to some three times the optical radius of the galaxy, the rotation curves continue to be
flat.
One can account for these strange circumstances in a number of ways. One
possibility is to suppose that the flat rotation curves are due to the presence of invisible
dark matter. In that case, one must suppose that the fraction of the dark matter is
independent of galaxy luminosity, hence galaxy mass and radius, and also that it is
independent of galaxy morphology. While the mass density, M(r)lr3, decreases with
increasing r, the total mass within radius r, M(r), increases linearly with increasing r, so
that the actual mass distribution is wildly different from the light distribution in the
galaxy.
Even when the galaxy seems like a disk, considerations of stability suggest that the
dark matter may sometimes be distributed in the form of a halo. Taking the dark matter
into account, the total mass of the galaxy may be some 100 times the mass inferred from
the luminous material, and, even to several times the optical radius, the mass may
remain some 100 to 1000 times the mean density of matter in the universe, as inferred
from the luminous matter.
The other line of thinking, which suggests the existence of dark matter, relates to
the inflationary scenario, where 'fine tuning' seems to require that the actual mass of
the universe is much larger than that inferred from the luminous matter.
As to the form of this dark matter, there are many hypotheses ranging from
planet-sized black dwarfs to massive neutrinos, gravitinos, photinos, axions, mini and
maxi blac~ holes, and monopoles. Difficulties with the usual picture of galaxy formation
suggest that the dark matter is non-baryonic. The observation of a quadrupole anisot-
ropy from COBE has been interpreted as favouring cold dark matter (ref 12). But this
is an area of active speculation, for there are difficulties with the usual picture of galaxy
formation, in the standard Friedmann cosmologies.

8 The age of the universe and closed timelike curves

Another assumption that one must guard against in cosmology is that the universe has
an age at all. A universe may be homogeneous, spherically symmetric, and stationary,
but may not have a finite age because the existence of closed timelike curves (or a tilt
in the cosmological arrow of time) precludes the possibility of defining a global notion
of time or an age for the cosmos. In fact, we saw (in Chapter VIA) that with any 'cyclicity'
in the global structure of time, a binary earlier-later relation (or the notion of a time
asymmetry) is inadequate.
The notion of a tilting cosmological 'arrow' of time, or a boomerang of time, was
 COSMOWGICAL TIME

brought out vividly by the great metamathematician Kurt GOdel,17 who first con-
structed a cosmological model with some remarkable features.
Most simple cosmological models have an 'absolute' time coordinate because of
the existence of a one parameter family of3-spaces, everywhere orthogonal to the world
lines of matter. In the GOdel universe, though one can define a notion of past and future
at any point, such an 'absolute' time coordinate is absent, because the universe rotates.
This rotation is not rotation about an axis, as in the case of the earth, the moon, the
sun or the galaxy. The GOdel universe rotates in a very peculiar manner. The sense in
which this universe rotates is perhaps best brought out by studying the apparatus needed
to determine whether the universe rotates.
The idea is that of Foucault, who demonstrated that the earth rotates, even though
we do not directly perceive this rotation. Foucault's idea was to suspend the bob of a
pendulum from a support that was free to rotate. He constructed a massive pendulum,
and the whole apparatus was fixed to two posts.
The pendulum was set in motion between the two posts. In the absence of any
'external forces', the pendulum would keep swinging this way due to inertia. Hence a
Foucault pendulum is also known as the compass of inertia. In the actual experiment
performed by Foucault, it was observed that the plane of the pendulum swung around
in about 24 hours. To put matters in another way, the pendulum kept swinging in a fixed
plane, but the earth (and the two posts fixed in it) moved around, completing one
rotation.
Using a Foucault pendulum, rotation of the universe can be measured, in principle,
from within the universe. There is, however, a slight problem similar to the one faced
by Archimedes when he boasted that he could lift the Earth. Where should one fix the
posts?
More precisely, with respect to what should the posts remain fixed? A rough answer
is to say that the posts should remain fixed with respect to the fixed stars. One may
proceed as follows. Matter in the universe is perpetually in motion. The moon moves
around the earth, the earth around the sun, the sun around the centre of the galaxy, and
the galaxies themselves keep colliding and moving away from each other. But, in
studying the motion of the earth-moon system around the sun, we can neglect the
motion of the moon around the earth. To a better approximation, in studying the motion
of the sun around the galaxy, we can ignore the motions of the planets around the sun.
In this manner, one can go from galaxy to clusters of galaxies to superclusters to the
bulk matter in the universe. The pillars should be fixed with respect to the bulk matter
in the universe.
To summarize, the GOdel universe rotates in the sense that the compass of inertia
(Foucault pendulum) rotates with respect to the bulk matter in the universe. Rotation
in this sense is sometimes called vorticity. In a homogeneous cosmological model with
vorticity, the cosmological arrow of time must tilt.
In the case of the GOdel universe, the arrow of time tilts so much that it starts
pointing backwards and takes a full circle, like a car with its steering wheel locked on a
left tum. So the existence of the GOdel universe is connected with its other peculiar
feature - the existence of closed time like curves.
A timelike geodesic may be regarded as representing the history of a material
particle. The world-lines of matter in the GOdel universe are equidistant, open, of
infinite length, and no world-line ever approaches its preceding points. At every point
on a world-line, it is possible to distinguish between future-pointing and past-pointing
208 CHAPTER VII

vectors. But if one uses this notion to order two points P and Q, so that P is later to Q
on the world line, then there also exists a time like curve on which Q is earlier than P.
The Godel universe does not provide an externally consistent description of the
cosmos. This is not because of the existence of closed timelike curves, but because the
Godel universe does not expand: the world lines of matter are equidistant. According
to Chandrasekhar's caiculations,18 the radius of the closed timelike geodesics in the
GOdel universe is larger than the estimated radius of the universe. Other cosmological
models with closed timelike curves are known, but exact, cosmological solutions which
both rotate and expand are hard to come by.
Nevertheless, the thought of closed timelike curves seems to make some physicists
nervous for some reason. Initially, this kind of cosmos was dismissed on the grounds
that it was fantastic! Recent work on quantum gravity, however, suggests that such
models are extremely common - almost any perturbation of a spacetime results in a
model with closed timelike curves. We have already seen (Chapter VB) that there is no
question of any logical absurdities such as the grandfather paradox: for example, the
Cauchy problem in a spacetime with closed timelike curves is well posed.

9 Problems about time: a cosmological perspective

This completes our summary of what is known at the present time concerning time. Our
knowledge is certainly adequate to allow us to formulate a great many questions. Some
important groups of questions are the following.

9. 1 Is there a proper clock?

We saw in Chapter IIIB how relativity defined equal intervals of time by postulating the
constancy of the speed of light. So, conceptually, a proper clock may be constructed by
bouncing a photon between two mirrors which stay a constant distance apart. In
Newtonian physics, one could have fixed a rigid rod between the mirrors to ensure that
they stay a constant distance apart. But, in relativity, there is no clear concept of a rigid
body (Chapter IlIA). So the best one can do is to suppose that the mirrors stay parallel
to each other.
Leaving aside conceptual questions about the meaning of 'parallel', can one
actually carry out such a construction? That is, can one construct an inertial (freely
falling) frame, with two parallel mirrors, which will remain parallel for substantial
periods of time? Near the Earth, for instance, geodesics which seem locally parallel may
intersect if extended far enough. So, does a proper clock exist for long enough a period
oftime?
There are many physical clocks, such as atomic clocks, which closely approximate
a proper clock. But can one be sure that physical clocks will remain proper clocks for
all time? That is, can one be sure that one lives in a cosmos where physical constants
will remain constant as the cosmos evolves?
The Dirac cosmological model, based on the large number coincidences, is an
example of a model where physical constants vary with time. One can use the same ideas
but modify the cosmology so that physical 'constants' which vary are different from those
 COSMOLOGICAL TIME 209

considered by Dirac. The Dirac model expands as Rex tVl, so that the issue of the
transparency or opacity of the universe would also have to be re-examined, if this model
were to be taken seriously.

9. 2 Related asymmetries?

In Chapters IV, V, and VI we examined the physical basis of a distinction or asymmetry

between the past and the future. A priori there are several such asymmetries.
(1) Entropy increases (thermodynamic arrow of time),
(2) retarded radiation dominates (electromagnetic arrow of time),
(3) the wave function collapses (quantum mechanical arrow), and
(4) the cosmos expands (cosmological arrow).
Clearly, it is not as if there are four different types of pasts and futures, though
philanderers and day dreamers may like to think so! How are all these asymmetries
related? That is, is the relationship between these asymmetries necessary or is it
contingent? In the recontraction phase of a closed cosmos, could entropy go on
increasing? Indeed, Tolman had argued, long ago, that every cycle of a closed
Friedmann model would only result in an increase of entropy. Many others, such as
Wbeeler-Feynman and Popper, have sought to base the electromagnetic arrow of time
upon the thermodynamic arrow.
As explained in the introduction, we have argued in the opposite direction. There
are adequate reasons to believe that the thermodynamic arrow of time cannot stand on
its own, though the dominance of retarded interactions may provide a basis for thermo-
dynamic time asymmetry. The nature of interactions depends, as we have seen, upon
the nature of the cosmos, so that both the thermodynamic and the electrodynamic
arrows of time depend upon the cosmological arrow of time.
But this is only a partial answer. There are other apparent asymmetries which
remain unexplained.
(5) Matter dominates over anti-matter,
(6) the K-meson decays asymmetrically (CP-violation).
According to the Stiickelberg-Feynman view of an anti-particle as a particle
traveling into the past, the observed baryon asymmetry provides a time-asymmetry.
Most particles, like retarded photons, travel into the future, while some, like advanced
photons, travel into the past.
Do we live in a baryon symmetric cosmos? Can the apparent dominance of matter
over anti-matter be linked to the dominance of retarded radiation over advanced?
Similarly, the K-meson decays in one direction:

+
or
~..o
A - ... 1- + I' + _
- + vJ.l.

The reverse transition is never observed. This is usually interpreted as a CP-violating

process. But, in view of the CPT theorem, one could interpret this as a T-violation. In
that case, the K-meson decay provides a time asymmetry. How does this asymmetry
relate to the expansion of the cosmos? or to the increase of entropy? It has been argued
210 CHAPTER. VII

that this is not a T-violating process because it is not possible to use the K-meson decay
to bring about an entropy decrease. So, perhaps, the K-meson decay is irrelevant to the
whole discussion.

9. 3 Imperfect asymmetries?

Are the observed asymmetries perfect? That is, is a spontaneous local decrease of
entropy impossible? This is clearly linked to the following question. Is the observed
electromagnetic radiation entirely retarded? Or, as suggested by Partridge's experi-
ment (Chapter VB), and our analysis, do small amounts of advanced radiation exist?
While one can assert that advanced interactions would bring about a spontaneous
entropy decrease, the relationship of one imperfect asymmetry to another is not
generally clear. Small amounts of anti-matter certainly do exist, but what is their
relevance to the discussion?
The cosmological analogue to the tilt in the arrow of time relates to the existence
of rotation in the cosmos. At one level rotation can be regarded as a straightforward
inductive generalization. As E. T. Whittaker19 asks,
The earth and all members of the solar system rotate on their axes, the satellites revolve
round the planets, the planets revolve around the Sun, and the Sun himself is a member of
the galaxy or Milky way system which revolves in a very remarkable way. How did all these
rotary motions come into being?
But as mentioned earlier, rotation of the cosmos is usually ruled out, using the
analysis of Sachs and Wolfe, due to the absence of a quadrupole anisotropy in the
CMBR: vorticity would produce a dipole anisotropy, whereas 'pure rotation' would
produce a quadrupole anisotropy. As also mentioned earlier, that analysis is not quite
sound, so it may be better to leave this question open, especially due to the very
suggestive similarity in the form of the rotation curves of galaxies. If rotation does exist,
and results in inhomogeneities or anisotropy, the assumption of a linear absorber
response would also become invalid.
A theoretical point in favour of universal rotation is that rotation of the universe
would help in galaxy formation. In big bang cosmologies, the time available for galaxy
formation is very limited. On the one hand, the formation of protogalaxies or in-
homogeneities could not have begun before the matter decoupled from the radiation.
On the other hand, it is believed that the CMBR comes to us from a very early epoch,
and its extreme isotropy suggests that galaxy formation must have been essentially
complete by then. So it would be advantageous to have a mechanism that could generate
and maintain inhomogeneities.20 At this time, however, the existence of cosmic rotation
remains a matter of pure conjecture.

9. 4 Local asymmetries?

The Poincare recurrence theorem tells us that, with time-symmetric evolution, the
thermodynamic arrow of time cannot but be 'local' in character: entropy cannot go on
increasing, but must 'eventually' decrease. The GOdel universe provides an example
where a local distinction between past- and future-pointing vectors is possible at any
point, but this distinction cannot be extended to a global distinction between past and
future. Is there enough matter to halt cosmological expansion and reverse the cosmo-
 COSMOLOGICAL TIME 211

logical arrow? Does the (hypothesized) tilt in the electromagnetic arrow of time remain
constant with time, or does it go on increasing so that it 'eventually' reverses its
direction? That is, could the amount of advanced radiation go on increasing until
eventually it dominates?

9. 5 Is asymmetry adequate?

We have seen that an imperfect asymmetry, or a tilt in the arrow of time, calls into
question the very concept of asymmetry. An asymmetry is a binary earlier-later rela-
tionship, and such a relationship is formally inadequate to describe the various struc-
tures of time that may arise globally. Given the peculiar nature of the logic on which
quantum probabilities are based, and in view of the intrinsic non-uniqueness of solu-
tions when the arrow of time tilts, there is a strong case for a microphysically structured
time. An asymmetry may be inadequate in this situation as well.
Thus, at both ends of the scale, the very question of a time asymmetry involves an
oversimplification. Therefore, should the whole problem of time asymmetry be recon-
ceptualized?
To conclude, only the future can provide answers to these questions about the
nature of time.
 Towards a consistent
E. model of time
= : ;;;::::;:::;:::: : ::::::::;:::::::::;;::: :.::: ::::::~:: :
VIII MUNDANE TIME

The problem of consistency with everyday experience

We might compare our traditional idea of Time to a rope that seems whole and unaltered as it
passes through our world but is now frayed and insecure at each end far away from us. But if it
will no longer do for stars and atoms, it is still good enough for man .... many scientists (perhaps
most of them) cling all the more tenaciously to the classical concept of Time, so far as our life
in this world is concemed, because it has been so successfully challenged ... on the largest and
smallest scales. They feel strongly that nobody must startpicking at the middle ofthe rope, where
it still seems to hold.... I can imagine them crying out to each other with Yeats: Things fall apart;
the centre cannot hold.
J. B. Priestleyl

ABSTRACf. We bring out the difference between the notion of time in physics and the notion oftime in
daily use. In the mundane notion of time, the past is linear and the future branches: the past is decided,
the future is not. We argue that the mundane view cannot be ignored because both views pertain to time
in the real world. If the mundane view is accepted, the branching of time suggests a failure of physical laws,
but rejection of the mundane view implies rejection of the belief that one may choose between physical
theories by putting them to test.

Quantum indeterminism fails to resolve the conflict, because mere indeterminism is quite different from
mundane choice, and this approach begs the question by implicitly assuming mundane choice in the
measurement postulate. The thermodynamic argument from complexity, and the argument via chaos also
fail: if unpredictable events are regarded as undecided, the argument may be time-reversed to conclude
that the past is not decided.

A microphysical tilt in the arrow of time not only provides the most conservative solution to the consistency
problem with time, within physics, but a decided past and an undecided future, as in mundane time, are
consistent with this hypothesis. Unlike mere indeterminism, future data may now have some bearing on
evolution, so that 'purpose' dermed as future data resembles mundane intention. Certain problems remain
unresolved. These concern, for example, the availability of choice at the macrophysical level, by the
selective amplification of future contingents, while past contingents remain confmed to microphysics.
These problems all seem to involve deep questions concerning the characterization of living organisms in
physics.

1 Introduction

A. RISTOTLE is often accused of having ignored the empirically obvious: he

ftapparently theorized about the number of teeth that women had, without both-
ering to count the number of teeth in his wife's mouth! Our study of time, ranging
from microphysics to cosmology, should not fall into the same trap of discussing deep
theoretical issues while ignoring the empirically obvious. Time is not merely a
theoretical issue of concern to physicists; time is also an aspect of everyday experi-
ence.
What have we learnt from our study of time in physics? Can we now relate the time of
physics to the time of mundane experience? The notion of time in physics still seems rather
216 CHAPlER VIII

far off from the notion of time in use in daily life. Should we, then, regard the mundane
experience of time as a hallucination? Or should we suitably modify physics to incor-
porate the mundane notion of time? Or should we continue to keep the two in separate
water-tight compartments?

2 Mundane time: linear past and branching future

2. 1 Reformulation of the A -series

Let us begin by formalizing the notion of time derived from daily experience. As we
saw in Chapters I and VI, beliefs about time, including the notion of the 'dynamic
Fig. 1: Mundane time

/
On the mundane view, time is be-
lieved to be linear towards the past,
and branching towards the future.
There is only one pas~ and the pos-
sibilities actually realized are repre-
sented by the thick line. The arrows
represent the several possible fu-
tures, while the thin lines represent
the might-have-been possibilities,
now excluded, presumably due to
human choices.
IFuture I

temporal' or the 'flow of time' or the 'A-series', may be formalized in terms of the
structure of time. The logical structure of (mundane) time resembles a tree (Fig. 1).
The figure summarizes the following everyday beliefs about time. Each branch of
the tree represents a possible future. We worry about possible courses of action, and
struggle to reach a decision because our present decisions and actions will result in one
or the other of the future possibilities being realized. In contrast the past is linear, not
subject to change: there is a unique way back into the past, corresponding to those
possibilities that were actually realized. One and only one thing happened in the past.
It is no use worrying about the past which is not open to change. What has happened
has happened, and nothing we do now can alter it.

2. 2 Ignorance, past and future

The mundane belief about the difference between past and future is some-
times described colloquially in terms of 'knowledge': the past is known, the
future unknown. But it is only a half-truth that the past is believed to be known and
the future unknown. Historians and geologists thrive because so many things about the
past are unknown. It would be better to say: the future is more unknown than the past.
As we saw in Chapter IV, the entropy law formalizes this assertion.
 MUNDANE TIME 217

But a more accurate description would be that the past is unknown in a qualitatively
different way from the future. A murder may have been committed, and the murderer
may be unknown; but the identity of the murderer is decided, and not open to change.
The past, even if not known, is decided, even though knowledge of the past may
not be. Past events are unaffected by current actions. One may not know the result of
an examination, even after the result has been declared, but current effort can hardly
change the result (except in a very corrupt society).
The future, on the other hand, is not yet decided. Assuming that an assertion
concerning a future event is either true or false leads to paradoxes like Aristotle's sea
battle (Chapter I). Future possibilities are contingent upon current choices and actions.
The result of an examination in the future depends upon one's performance. One may
not know this result either, but one believes one's choices about current activities have
some bearing on the performance and the result.
Thus, in everyday life, one proceeds on the belief that ignorance of the past pertains
to lack of knowledge about something already decided, whereas ignorance of the future
pertains to lack of knowledge about something not yet decided (and hence intrinsically
unknowable). In the above figure, the past is linear because it is decided; the future
branches because it is yet to be decided.

2. 3 Epistemic and ontic contingents

The difference between an eventt the truth (value) of which is decided but unknown
and an event the truth of which is undecided hence unknown has been responsible for
considerable confusion. This motivates the introduction of the following terminology.
An event, the truth of which is decided but unknown will be called an epistemic
contingent. In the mundane picture of time, uncertainties or possibilities which refer
to the past correspond to 'epistemic contingents'; one or the other of the possibilities
has definitely occurred, in principle, though we may not always know which.
An event the truth of which is undecided, hence unknown, will be called an ontic
contingent. In the mundane picture of time, uncertainties or possibilities referring to
the future typically correspond to 'ontic contingents'; we do not know (with certainty),
even in principle, which possibility will occur. One cannot assert that the occurrence of
the sea battle tomorrow is, in principle, true or false, as of now. Thus, in the mundane
picture, future events are ontically contingent.
We consider later some arguments which suggest that there is no real difference
between ~pistemic and ontic contingents.

t I will consistently abbreviate 'the truth-value of a proposition concerning an event' to 'truth-value of an

event' or 'truth of an event'.
218 CHAPrER VIII

3 Comparison with time in physics

3. 1 The B-series from relativity

Let us now compare the picture of past-linear, future-branching, mundane time with
the time of physics.
The problem with the measurement of time (Chapter II,IIIA, liB) was resolved
in relativity by using the B-series view of time. In relativity it makes sense only to speak
of entire world-lines, and a world-line embraces both future and past history. But it is
meaningful to speak of entire world-lines only if the entire future is also decided. Thus,
time is logically a line in relativity, as in classical mechanics, since both past and future
are decided. (More generally, the differential equations of physics are meaningful only
if one supposes that, locally, time has the topology of the real line. )
In the traditional view, thermodynamics enables one to distinguish between the
head and tail of this line by providing an 'arrow of time': one can use the entropy law
to distinguish between a decided past and a decided future.

3. 2 Limitations of the arrow simile

We have already considered some of the more technical defects in the thermodynamic
approach to the arrow of time. These technical defects pertain to the standard ways of
resolving the recurrence and reversibility paradoxes, even overlooking the difficulty in
establishing the variants of the ergodic hypotheses that are needed to derive the
entropy law from dynamics. But there is an additional problem, inherent in the arrow
simile, which persists even if one assumes that the entropy law may be derived from
dynamics, or that such a derivation is inessential.
Given the information-theoretic meaning of entropy (Chapter IV), the entropy law
guarantees that one is surer of the past than of the future. Thus, the entropy law provides a
quantitative difference between past and future; it does not provide a qualitative difference.
Accordingto the entropy law, both past and future are contingent, but they areboth epistemica11y
contingent This does not agree with the mundane picture where future contingents are ontic.

3. 3 Branching of time and the failure ofphysical laws

There is a good reason why thermodynamic contingents cannot be ontic. For if the past
is taken as decided and the future is allowed to be undecided (ontically contingent), or
if time is linear to the past but branches into the future, this would mean that the entire
past is inadequate, in principle, to decide the future.
Now, in the traditional view of physics, a physical law decides the future states of a
system given the past states. Typically, the current state is adequate to decide both past
and future states. Therefore, if the past (including the present) is decided, and the future
is not decided, then physical laws must fail. Thus, the traditional view of physics is not
compatible with the picture of a decided past and an undecided future. According to
the traditional view of physics, the mundane notion of a decided past and undecided
future (past-linear, future-branching time) forces the conclusion that physical laws must
somehow fail.
 MUNDANE TIME 219

4 Need for consistency

4. 1 The problem of consistency

Thus, even granting that an arrow of time can be defined in physics, the picture of time
as an arrow does not agree with the mundane picture of time as future branching but
past linear. There is an even more direct conflict: accepting the mundane picture of
time seems to lead to the conclusion that physical laws must somehow fail. What, if
anything, should one do about it?
There are a number of possibilities.
(1) Ignore the contradiction.
(2) Dismiss the mundane experience of time as irrelevant.
(3) Suppose that physical laws somehow fail for human beings.
(4) Assert that quantum mechanics resolves the contradiction.
(5) Deny that there is any contradiction.
(6) Assert that thermodynamics resolves the contradiction.
(7) Assert that chaos resolves the contradiction.
Each one of these possibilities is considered below, and found to be inadequate.

4. 2 Building a consistent model of experience

4. 2. 1 The two-time theory. The first possibility is, of course, to ignore the difference
between an arrow and a branch. A physicist can use the notion of the arrow of time
while a physicist, and continue to use the usual notion of a past-linear, future-branching
tree while a human being!
The problem is that both models of time pertain to the real world. One would like
to have a consistent model of the real world, rather than a two-time theory: one for
physics and one for everyday life. Generally, any physical theory is consistent with a part
of experience. The whole idea of putting physical theories to test is to ensure consistency
with all experience and not just a part of it.
An example of an alternative way to achieve consistency through theoretical
analysis is provided by the theory of relativity, which reconciled the differing notions of
time in classical mechanics and electrodynamics. This example also shows the point-
lessness of hanging on to mutually inconsistent bits of knowledge.

4. 2. 2 Mundane time and assumptions about time in physics. Though the need for
consistency seems clear enough, the two-time theory does operate in practice. Mundane
time, being part of everyday experience, is plausible.
Dumett's paradox of the dancing chief (Chapter I) demonstrates how the plausi-
bility of mundane time creeps into an argument. The chiefs argument seems ridiculous,
though it is formally identical with the more plausible argument for a distinction
between fatalism and determinism. Since the two arguments are formally identical, the
difference between the two must arise from an additional implicit assumption concern-
ing a factual or physical issue: the chiefs argument implicitly assumes an open past,
while the earlier argument assumes an open future. Since both arguments are formally
identical, the ridiculousness of one and the plausibility of the other both concern only
220 CHAPrnR VIII

the additional implicit assumption, which, of course, is the assumption of mundane time.

The plausibility of mundane time has even been used to arrive at postulates
concerning the nature of time in physics:
However, the existence of such [closed timelike] curves would seem to lead to the possibility
of logical paradoxes: for one could imagine that with a suitable rocketship one could travel
around such a curve and arriving back before one's departure, one could prevent oneself
from setting out in the first place. Of course there is a contradiction only if one assumes a
simple notion of free will; but this is not something which can be dropped lightly since the
whole of our philosophy of science is based on the assumption that one is free to perform
any experiment.
The curious thing about the above quote from Hawking and Ellis2 is that it leads
directly from considerations of 'free will' (Le., mundane time) to an assumption about
the nature of time in physical theory. The passage continues.
It might be possible to form a theory in which the concept of free will was modified... but
one would be much more ready to believe that spacetime satisfies the chronology condition:
namely that there are no closed timelike curves. [First emphasis mine, second original]
It is beside the point that the chronology condition may now be unnecessary for proving
the existence of a singularity. What Hawking and Ellis seem to have in mind is a 'folding
around' of mundane time to bring the future branching into conflict with past linearity,
while overlooking the difference between mundane time and the superlinear time of
physics.
Similarly, using the past-linearity of mundane time, Popper3 seeks to exclude the
GOdel cosmology by making axiomatic the possibility of preserving records indefinitely.
The argument is, then, that along a closed world-line every record would have to be
destroyed to ensure consistency. It is beside the point that the world-lines of matter in
the Godel universe are not closed and that the Godel universe can be excluded on other
physical grounds. Popper's record-postulate, based on mundane time, excludes even
the oscillatory Friedmann model, and seeks to limit by postulate the amount of matter
in the universe! This postulate, based on mundane experience, again supposes that there
is no difference between mundane time and time in physics.
We have seen that Popper's pond paradox seeks to exclude advanced actions on
the grounds that short of a conspiracy, there could be no explanation for a convergent
ripple. Since advanced actions do not, in principle, admit explanations from the past
('causal explanations'), this amounts to imposing a physical condition (absence of
advanced interactions) on the basis of the requirement, plausible with mundane
time, that all phenomena should admit an explanation from past causes.
The situation is similar with the other paradoxes of advanced action, like the
grandfather paradox and the Wheeler-Feynman paradox (Chapter VA). While these
have been put forward as paradoxes of advanced action, they use the availability of
advanced action only to bring the past within the scope of mundane choice, thereby
making the past apparently branch, in conflict with the notion of past linearity. These
paradoxes seek to eliminate advanced action by demonstrating the incompatibility
between (i) advanced action, (ii) mundane choice, inherent in mundane time, and
(iii) the rigid connections between past and future provided by physical laws (in the
absence of advanced actions). By overlooking (through familiarity) the prima facie
contradiction between mundane time and physical laws, these paradoxes sacrifice the
important escape route provided by advanced actions.
 MUNDANE TIME 221

To summarize, the two-time theory operates in practice by bringing out the conflict
between mundane time and any new notion of time in physics, while overlooking the
conflict between mundane time and the superlinear time of physics.

4. 3 Mundane experience as hallucination: the problem of intentional choice

4. 3. 1 The flat earth and parallax. The next possibility is to dismiss mundane experi-
ence as irrelevant. There are situations where mundane experience must be regarded
as a hallucination from the point of view of physical theory.
The earth seems perfectly flat. We now know that it is round. So we must
suppose that for centuries people believed the earth was flat on account of a mass
hallucination. Parallax provides another example, where distant objects seem to
move relative to nearer object, even though no such relative motion takes place.
Many more interesting examples of this sort can be provided. So why can't the
mundane experience of time be regarded as a hallucination?
The B-theory view does precisely this. It dismisses the notion of a 'flow of time'
as a hallucination.
Now it seems clear enough that mundane experience is based on human
perception, and human perception is limited. Phenomena on the scale of the
cosmos, or phenomena at the microphysical scale of electrons, are not normally
amenable to direct and unaided human perception, though such phenomena may
be amenable to perception with the necessary intervention of both sophisticated
instruments and theory. Therefore it may well be that time, on the cosmological scale
or the microphysical scale, is totally different from the time of mundane experience.
So the mundane experience of time may well be dismissed as irrelevant to phenom-
ena at the two ends of the scale.

4. 3. 2 The problem of intentional choice. But at the usual mesophysicallevel one can-
not dismiss the mundane notion of time so easily. One reason is that the very idea that
physical theories may be put to test derives its credibility from the mundane picture of
time. One believes one can choose to perform an experiment with the intention of testing
a physical theory. One can conceive of different possibilities that may falsify a given
theory. The distinction between valid and invalid physical theories, as also the distinc-
tion between physics and metaphysics provided by the criterion of refutability, rests on
the belief in mundane choice, or the intentional choice available with mundane time.
But, if the logical linearity of time in physics were to be taken seriously, the belief
in intentional choice would have to be discarded: one would have to suppose that all
such 'choices', 'intentions' and 'conceptions' were decided at the time of the big bang
(even though they may not be predictable). It may, for instance, have been decided
long ago that a certain physical theory, though valid, would be rejected without ever
being put to test. So, even if mundane time is a hallucination, the arrow of time in
physics is not good enough to justify the belief that physical theories may be validated
by putting them to test.
222 CHAPTER VIII

4. 4 The complete problem of consistency

We can now formulate the complete problem of consistency between mundane time
and the superlinear time of physics. If one accepts the mundane picture of time, then
according to the arguments of 3.3 one is forced to the conclusion that physical laws
somehow faiI.t On the other hand, if one rejects the picture of mundane time, then
according to the arguments of 4.3.2 one must concede that the validity of physical laws
is suspect or illusory.

4. 5 Do physical laws somehow fail for human beings?

A tempting way out of the above paradox is to sharpen the two-time theory as follows:
mundane time applies to human beings, physical laws don't. This was the belief at the
time of Newton when some people thought that divine intervention was necessary to
maintain the planetary system, and that humans had divine souls which were exempt
from physical laws.
Today, many people would prefer the more modest view that human beings are
somehow special, and hence at least partly exempt from physical laws. As it stands, this
way out is naive and vague for the following reasons. As Boltzmann pointed out long
ago, even the most superficial observation shows that the laws of physics apply equally
to animate and inanimate matter. When a cashiered cashier jumps off the roof of an
eleven-storied building, the theory of gravitation well describes his predictable crash
to the ground. But we believe the cashier had a choice before jumping off. Of course,
the theory of gravitation applies equally to the cashiered cashier before he jumped off,
else he might well have already floated off into the heavens! The most detailed testing
has failed to show up any violations, by human beings, of even the suspect entropy law.
In the absence of any evidence, if one wishes to continue to provide humans with
partial exemption from physical laws, one must at least identify those aspects of physics
that are purported to fail. If human beings are not in the domain of validity of physical
laws, one must delineate the actual domain of validity of physics.
In the: absence of any evidence or identification or delineation, this point of view
is too vague to be refuted, and need not be considered further.

5 Consistency through broken time

Most of the remaining ways, listed in 4.1, of resolving the problem of consistency
between physical time and mundane time, may be put into the category of 'broken time'.
The idea is to permit 'choice' by denying the existence of any direct, logical (causally

t Indeed, the suggested conclusion is that physics may eventually be need to be reformulated using a
mathematical theory other than the calculus.
 MUNDANE TIME 223

deterministic) connection between one event and the next. By 'breaking the connection'
between past and future, a decided past may coexist with an undecided future. The
major difficulty is that the connection must be 'broken' without denying the validity of
any part of physics.

5. 1 Ontically broken time: quantum indeterminism

For the purposes ofthis subsection (5.1) we will disregard the structured time interpre-
tation, and consider only the usual Copenhagen interpretation of quantum mechanics
(Chapter VIA). We recall that, according to the Copenhagen interpretation, a measure-
ment of electron spin may result in an outcome which is not, in principle, accurately
predictable.
Let us further suppose, for this subsection, that the world, at the macrophysical
level, behaves in a way which is as unpredictable as the outcome of a measurement of
electron spin.

5. 1. 1 Chance vs. choice. Our temporary hypothesis certainly eliminates determin-

ism, by assumption. Nevertheless, it does not permit choice of the kind implicit in the
mundane picture of time.
As an example of such unpredictability, consider a slot machine which delivers
either ice-cream or chocolates in a way that is in principle completely unpredictable.
By assumption, what one eats is a matter of chance, but does one have a choice?
In practice, one could perhaps go on putting coins in the machine until it delivers
the item one wants. But this is fallacious. The choice available, in practice, exists only
with the mundane picture of time!
A more accurate picture of the world, according to the Copenhagen interpretation,
is provided by a giant machine which slams one down in front of either ice cream or
chocolates, and rams one of these down one's throat. What one eats becomes a matter
of chance, but one has no choice in the matter.
There is, thus, a difference between the intentional choice available with mundane
time, and the unpredictability arising from the chance permitted by the Copenhagen
interpretation.

5. 1. 2 Occasionalism. By breaking any connection between one instant and the next,
one obtains ontically broken time. A world in which time is ontically broken at the
macrophysicallevel has been described graphically by Ghazali4
One might have to say 'I do not know what there is at present in my house; 1 only know that
1 left a book in my house, but perhaps by now it is a horse which has dirtied the library
with its urine and excrement...' Again, on seeing a stranger, one might have to acknowledge:
'It may be that he was one of the fruits in the market which has been changed into a man,
and that this is that man.'

Such a world is completely indeterministic, but it is easy to see that, in such a world,
not only does all choice evaporate, but the very notion of the person exercising the
choice becomes suspect. So, regardless of the tradition of theological debate between
free will and determinism, mere absence of determinism cannot be equated with the
existence of choice.
224 CHAFI'ER VIII

Quantum indeterminism does not go to such an extreme: electrons don't change

into protons, the measurement process only reduces the wavefunction of the electron,
and the transition probabilities are fixed by quantum mechanics. But, despite indeter-
minism, there is still no choice, for there is nC? way of making one future (or one reduced
state) more probable than another.

5. 1. 3 The transition to macrophysics. The classical axiomatic framework for quan-

tum mechanics visualizes a two-tier system for modeling the world. In that framework,
macrophysical objects obey classical mechanics, and quantum mechanics is limited to
microphysical systems. This postulate was not accidental, but was motivated by the
correspondence principle, and the indeterminacy relations. The measuring apparatus,
being macrophysical, is subject to classical mechanics.
On the other hand, to obtain the macrophysical choice offered by mundane time,
we temporarily supposed that quantum indeterminism applied at the macrophysical
level. Actually there is a problem of transition to macrophysics. If this problem is
bypassed, paradoxes like Schrodinger's cat would have to be taken literally.
Let us suppose that the problem of transition to macrophysics has somehow been
resolved by developing an alternative framework. Nevertheless, there is the additional
problem that the SchrOdinger equation decides the future and past states of the system,
once its present state is decided. Quantum indeterminism materializes only when an
act of measurement disturbs the evolution according to the SchrOdinger equation. The
act of measurement involves a coupling between measuring apparatus or 'observer' and
the system. Problems arise if both are taken to be macrophysical.
How one now divides the world between 'system' and 'observer' is a matter of deep
speculation. Wigner, for instance, has suggested that physical laws do not apply to
human consciousness, which acts as 'the observer outside the system'. This creates all
sorts of problems. For instance, if a man and a frog 'observe' each other simultaneously,
whose wavefunction will be reduced?

5. 1. 4 Quantum 'choice' and measurement. The other peculiar feature of the 'choice'
offered by an indeterministic world of the quantum mechanical type is that the choice
is effected only through an act of measurement! The quantum 'choice' materializes
only when 'an observer', outside the system, 'chooses' to measure or observe the
system.
The freedom of the observer to 'choose' to carry out an act of measurement with
the 'intention' to measure the spin of an electron in thex rather thany-direction, already
assumes the intentional choice available with mundane time. The measurement postulate
(Chapter VIB) fixes the probability of an outcome only after the measurement process
has been specified. It is tacitly understood that the observer 'chooses' the specification
of the measurement process. Recall, for instance, Bell's locality postulate (with revised
emphasis): the outcome of measuring the spin of electron 2 cannot depend upon the
direction in which we choose to measure the spin of electron 1.

5. 1. 5 Summary of quantum indeterminism and mundane time. To summarize, the

traditional approach to choice through quantum indeterminism offers a poor form of
choice, at the expense of incurring a severe form of the logical fallacy of begging the
question: the approach implicitly assumes the choice available with mundane time, and
the conclusion provides rather less 'choice'. Thus, one only succeeds in deriving a
 MUNDANE TIME 225

conclusion weaker than the assumption, and that too subject to a number of technical
obscurities!

5. 2 Epistemically broken time: the argument from complexity

The other kind of approach is to suggest that it is not the physical laws per se which fail.
It is merely not possible to use them to predict the evolution of complex systems. In
view of the great trouble in using quantum indeterminism to model the intentional
choice of mundane time, let us revert to classical mechanics. For the remainder of 5
we will assume that the physical laws relevant to macrophysical systems are those of
classical mechanics, as traditionally understood.
The basic approach of epistemically broken time may be understood by means of
an analogy. In a mystery novel, the murderer is already decided, since the book has been
written. But to a reader on p 5 the identity of the murderer is unknown. It is this lack
of knowledge which makes mystery novels (and life) interesting.

5. 2. 1 The complexity of macrophysical systems. In this approach one looks upon a

human being not as an individual but as a huge conglomerate of molecules. This
naturally makes a human being very complex, though no more complex perhaps than a
huge piece of stone! Ideally one should imagine the constituent molecules to be moving
about as haphazardly as the molecules constituting a gas in a box.

5. 2. 2 Thermodynamics and stochastic evolution. The argument now is that one sim-
ply cannot hope to be able to integrate all the many-particle equations of motion.
Therefore, one must give up the picture of deterministic evolution, and replace it by a
picture of stochastic or Markovian evolution, as in the Ehrenfest model (Chapter IV):
the system, as it evolves, hops from one state to another. One cannot hope to trace the
evolution in detail, but one can hope to calculate the probabilities of transition from
one macro-state to another.

5. 2. 3 Comparison with quantum indeterminism. Thermodynamic unpredictability is

in one respect similar to quantum indeterminism: while the future is unpredictable, the
transition probabilities are decided and immune to human intervention. Neither
provides any way by which human action can make one future more probable than
another.
The difference between thermodynamic unpredictability and quantum indetermin-
ism is the following.
(i) Evolution is stochastic rather than deterministic, as in quantum mechanics; and
(ii) the future is decided in principle, though it is not actually predictable.
The more technical differences between thermodynamic probabilities and quantum
probabilities were considered in Chapter VIB, where they were related to the structure
of time: thermodynamics assumes a linear time, whereas quantum mechanics assumes
different structures of time as convenient - a real line for the Schrodinger equation, a
structured time for the Hilbert space, and mundane time for the measurement postu-
late.
In one way, thermodynamic unpredictability seems stronger than quantum inde-
terminism, since the evolution is no longer deterministic: the future is not strictly
226 CHAPrER VIII

decided by the past, only probabilities for the future are decided by the past. But the
stochasticity of evolution is a bit ambivalent, since thermodynamics does not seek to
replace the underlying laws of evolution which are deterministic. This is brought out by
(ii), which means that future contingents are epistemic: according to thermodynamics
the future is decided though impossible to calculate. That is, the argument from
complexity erases any qualitative distinction between ignorance of the past and igno-
rance of the future.

5.2.4 Laplace's demon. Laplace's 'intelligence', nowadays known as Laplace's

demon (Chapter II), well illustrates the distinction between what we have called
epistemic and ontic contingents. The demon is designed to overpower epistemic
contingents: being a super-observer, a super-scientist and a super-computer rolled into
one, all ignorance due to complexity disappeared before the demon.
Therefore, the distinction between past and future also disappears. The other
technical disabilities of the thermodynamic argument from complexity lead to the same
conclusion.

5. 3 Epistemically broken time: the argument from chaos

Some of the disabilities of the argument from complexity are removed by the argument
from chaos, given by Popper, to exorcise Laplace's demon (which, as we have observed,
stands in the way of the criterion of falsifiability). In the first place, the complexity of
the system is measured not by size alone, but by the nature of its dynamics.. The dynamics
of complex systems, like human beings, is believed to be chaotic.
A dynamical system is chaotic if it is ergodic (Chapter IV), and has a positive
Uapunov exponent. For a chaotic system, a small difference in initihl conditions
explodes exponentially towards the future.
For a chaotic dynamical system, prediction is impossible, unless the initial condi-
tions are known exactly. As we saws in Chapter II, while very accurate knowledge may
be possible, exact knowledge of initial conditions would be impossible even for
Laplace's demon, since the demon has no supernatural powers: its powers are limited
to those of an ordinary scientist, extended to the highest imaginable degree.

5. 3. 1 Predictability and contingents. Therefore, Popper argues, even if the future is

decided and only epistemically contingent, it is epistemically contingent in principle.
Popper's next argument is that is futile to attempt to distinguish between a future which
is ontically contingent, and a future which is in-principle epistemically contingent. That
is, it is meaningless to distinguish between events that are decided but in-principle
unpredictable, and events that are undecided. Having erased the distinction between
epistemic and ontic contingents, the conflict between mundane time and superlinear
time also disappears, and there is no need to suppose that any physical law actually fails.

5. 3. 2 Reversibility of chaos. Popper's approach is robust and full of common sense.

But it must contend with the time symmetry of physical laws. Let us reverse Popper's
argument and apply it to the past.
 MUNDANE TIME 227

Chaotic systems are just as chaotic towards the past. From an inexact knowledge of
the current state, Laplace's demon would be unable to retrodict the past quite as much
as it would be unable to predict the future.

5. 3. 3 Unpredictability of the past. One can now apply the second part of Popper's
argument: it is meaningless to speak of an event being decided when one cannot, in
principle, calculate it. The conclusion is that the past is not decided. The dancing chief
in Dumett's paradox (Chapter I) was right after all. It does not help to point out that
one remembers the past. In the first place, memory or records cannot, in principle, be
exact, for exactly the same reason that observation of the present cannot be exact. In
the second place, nothing in the mundane experience of memory suggests that Laplace's
demon could be allowed to have a record of the entire past.
Therefore, if Popper's argument is valid, the entire past must be treated as actually
contingent on current information. The result of an experiment performed two minutes
ago, for instance, may by now have changed. Thus, Popper's argument is useless, since
it establishes the possibility of (logical) falsifiability only by devaluing experimentation,
the means of (empirical) falsifiability. The argument leads to an undecided or open
future only at the cost of an equally open past. Therefore, such a view disagrees with
the mundane view of time.

6 Consistency through tilting time

6. 1 Tilting time and consistency within physics

For the moment, let us forget about the problem of mundane time. The case for tilting
time has already been argued out within physics. The problem of time measurement
leads to the linear time of relativity (Chapter II, IlIA, IIIB). Thermodynamics fails to
provide an arrow of time (Chapter IV), and the introduction of the electromagnetic
field only seems to create more problems (Chapter VA). Taking cognizance of history-
dependence does provide a time asymmetry (Chapter VB), but does not help with the
problem of radiative damping. With the absorber theory of radiation, the problem of
radiative damping is naturally related to cosmology, and suggests the existence of a tilt
in the arrow of time. Moreover, a tilt in the arrow of time provides the structured time
necessary for quantum mechanics (Chapter VIB).
Two points are worth emphasizing. In the first place, there are problems with the
notion of time, purely within physics, since, for example, no satisfactory arrow of time
can be defined, if orthodox beliefs in physics are accepted. Secondly, the solution we
have proposed suggests a 'minimal' change. We have presented in detail the point of
view that a physical theory consist primarily of its differential equations and its
mathematical formalism. The proposal for a tilt in the arrow of time only carries the
relativistic postulate to its logical conclusion, while leaving largely unaffected the
mathematical formulation q.m. Strictly speaking, the proposal does not even involve
any new hypothesis; rather, it discards the unsound, though traditional, hypothesis of
'causality' or perfect time asymmetry.
228 CHAPTER VIII

6. 2 Preliminary consistency with mundane time

Next let us consider the problem of consistency between tilting time and mundane time.
With pure history dependence, a decided past results in a decided future.
With tilting time, on the contrary, a decided past is consistent with an undecided
future. Thus, the hypothesis of a tilt in the arrow of time, is prima facie consistent with
mundane time and offers a consistent resolution of many of the paradoxes and
problems concerning the notion of time in physics.

6. 3 Remaining problems of consistency

A few points remain. It is necessary to ensure that the indeterminism arising from tilting
time is not 'random' or occasionalistic, and that it permits choice. It is necessary to
explain how such a choice extends to macrophysics, when the tilt in the arrow of time
has been assumed to be microphysical. Finally, it is necessary to decide whether one
must forever live with a decided past!
The proposals below are not intended to answer any of these questions fully and
satisfactorily, but they are intended to make a beginning.

6. 4 The problem of intentionality

6. 4. 1 Temporal assumptions underlying epistemically broken time. An implicit and

unstated assumption which underlies the idea of epistemically broken time is that the
future is not amenable to direct observation. This idea derives its strength from the time
of mundane experience. The essential role of this implicit assumption may be seen in
the mystery-novel analogy: the mystery survives only so long as one follows the 'rules'
and does not peek at the last page. H one does not respect this assumption, then a future
event which is actually decided but cannot be calculated or predicted from present data
may yet be directly 'guessed' or observed.

6.4.2 Explanation from final causes. Now this assumption - that information does
not propagate from future to present though it does propagate from past to present -
is an assumption about the nature of time in the real world. It is a physical assumption.
But, as we saw in Chapter VA, it is an assumption which is justified on the basis of
philosophy.
So the moment we allow the possibility of anticipatory phenomena, or a small 'tilt
in the arrow of time', we must be prepared for some philosophical re-adjustment. The
counter-examples in Chapter VB show that anticipatory phenomena cannot, in princi-
ple, be explained from past causes. Popper's pond paradox is immediately resolved if
we discard the metaphysical assumption that all phenomena must admit an explanation
in terms of past causes.
The idea of an explanation from final causes fell into disrepute from the time of Francis
Bacon, and because of its association with the teleological argument for the existence of
God. But Francis Bacon lived a long time ago, and science has changed very much since
then. Purely history-dependent phenomena can be explained only from past data; purely
anticipatory phenomena can be explained only from future data For the realistic case of
 MUNDANE TIME 229

mixed phenomena, implied by the hypothesis of a tilt in the arrow of time, one may
expect that both past and future data will be required. But the example in Chapter VB
shows that if past and future data are arbitrarily prescribed then neither past data nor
future data nor both are, in general, adequate. However, in some cases at least, prescription
of both past and future data determines the intermediate evolution.

6. 4. 3 Intention or 'purpose' as future data. Let us use the word 'purpose' to denote
future data that are necessary, but not necessarily sufficient, to determine the complete
evolution. With a tilt in the arrow of time a 'purpose' is the time-symmetric counterpart
of the notion of 'cause' or past data. Neither 'cause' nor 'purpose' is, in general,
adequate to 'explain' phenomena or predict intermediate evolution, but both together
may be. We compare below this notion of 'purpose' with the more conventional notion
of intention.

6. 4. 4 Intentionality with mundane time. Let us take another look at the conventional
notion of intentionality with mundane time. In the first place intention concerns future
possibilities. In the simplest case (like picking up a glass to drink water), an intention
results in the realization of one or the other of future possibilities. Thus intention, like
'purpose', partly determines or constrains the future. In a more complex situation
(where the glass is slippery), mundane intention may make a future event more
probable.
In the second place, supposing that intention is determined by past data leads to
difficulties. 'Psychological determinism' may be true in a number of cases, but it leads
to difficulties if it is extended indefinitely to include even creative acts or the intention
relevant to the creation of alternative theories. For if intention decides the future and
intention is decided by the past, then the future is decided by the past.

6. 4. 5 Differences between the two notions. Two differences persist between mundane
intentionality and intention as 'purpose' or final data. One is that the notion of intention
conventionally applies only to humans or at least to living organisms. The second
difference is the following.
Mundane intenti"on tends to assume a deterministic connection between present
and future. If action A will lead to future consequence B, one chooses action A because
one intends consequence B. In the absence of any logical connection between A and B,
one may hope for B, but one can hardly intend B. The availability of choice eliminates
deterministic connections and constrains intention through intervening contingents.
That is, choice constrains intention. On the other hand, the more rigid the connections
between phst and future the less the choice that is possible. Thus, intentionality
constrains choice. So, in the conventional picture, there is a competition or 'trade-off'
between intention and choice: one is possible only to the extent that the other is not. In
the long run, this competition becomes a contradiction.
This 'long term' should not be seen anthropocentrically, in relation to the human
life span; rather, it should be seen in terms of the frequency with which choice is
available or exercised.
The situation might be different with tilting time. Greater purposiveness suggests
greater choice, and less predictability from history or habit.
230 CHAPTER VIII

6. 5 The problem of macrophysical choice

6. 5. 1 Advanced interactions as perturbations. One does not expect a microphysical

tilt in the arrow of time to result in macrophysical choice. Consider _a moving rocket.
Suppose it emits one advanced photon. An electron or two might acquire energy and
move off erratically. But one would not expect the motion of the entire rocket to be
affected. ( Wheeler and Feynman argued analogously that acausal effects due to
preacceleration would remain confined to microphysics.)
In fact, at the level of single photons there is no difference between the emission
of an advanced photon and absorption of a retarded photon. Thus, emission of a single
photon would merely act as a small perturbation, and one can assert with a fair degree
of confidence that the motion of the rocket would remain unaffected.

6. 5. 2 Deep history dependence. There could be some special cases where the future
evolution of the rocket might be affected. If the rocket is being guided by a computer
so delicate that its registers are at the molecular level, and the advanced photon affects
the value in a crucial register, one could expect a noticeable change.
How would one recognize such special cases? An actual case of this sort concerns
the evolution of life itself. This process is guided by genes: macromolecules which store
the information necessary for cell reproduction in their very structure. A single high-
energy photon, such as an X-ray or y-ray, may alter the structure of the gene: some
portion may be deleted and substituted by a different sequence, resulting in a different
conformation. Given the deep history dependence of biological evolution, such a
mutation could result in a noticeable change, at a future time, in the organism or its
progeny.

6. 5. 3 Anticipatory phenomena (choice) and Maxwell's demon. To find macrophysical

examples of anticipatory behaviour, one must alternatively look for a situation where a
cloud of advanced photons is emitted. Thermodynamically, one would expect this to
result in a spontaneous local decrease in entropy, i.e., the spontaneous generation or
creation of information.
(Thus, purely history-dependent interactions provide the time-asymmetric dynam-
ics necessary for an increase of entropy; symmetrically, purely anticipatory systems
provide the time-asymmetric dynamics necessary for a decrease of entropy. Hence,
anticipatory phenomena would, like Maxwell's demon, tend to 'sort' rather than 'mix'.)
Therefore, one would expect macrophysical choice, arising from anticipatory behavi-
our, to be restricted to macrophysical systems which exhibit spontaneous 'sorting' or
entropy decrease.

6. 5. 4 Ampliatory self-organization. How would one recognize such systems? A spon-

taneous local decrease in entropy may easily be masked by spontaneous entropy-in-
creasing processes in the neighbourhood: one should not expect any evident violation
of the entropy law. Such a local entropy decrease would be masked even more effec-
tively by energy-intensive locally entropy decreasing processes in the neighbourhood.
A living organism is a typical example of such a neighborhood: living organisms
consume energy and are (locally) self-organizing in the manner of a refrigerator. How
 MUNDANE TIME 231

should one isolate, in practice, a local entropy-decreasing process which also tends to
decrease global entropy, but is immersed in a sea of local entropy-decreasing processes
which tend to increase global entropy? Can living organisms amplify self-organization?
I will not pursue this question any further, but will end with the suggestion that
there is some reason to suspect that macrophysical choice arising from anticipatory
behaviour is confined to living organisms.

6. 6 The problem ofpast contingents: structured time vs. past linearity

Finally, there is the fact that tilting time may give rise to permanent or past ontic
contingents, in disagreement with mundane time.
With structured time, time may branch and collapse in both directions. There is
no qualitative distinction between past and future: both contain ontic contingents. Can
one expect a quantitative distinction? Can one hope to have a situation wherein past
contingents remain confined to microphysics, whereas future contingents are selec-
tively amplified to become macrophysical?
This question is tied to the preceding one. If the amplification of anticipatory
behaviour and future contingents, to macrophysics, is a property tied to living organ-
isms, then this question would take us far beyond the scope of the present exposition.

7 Conclusions

In the mundane view of time, the past is decided, but the future is not. It is necessary
to make the time of physics consistent with the time of daily experience. The mundane
view of time could possibly be rejected at the microphysical or the cosmological level;
but at the macrophysical level the mundane view cannot be entirely rejected as a
hallucination because it provides the basis for the belief that one may choose between
physical theories by putting them to empirical test.
Quantum indeterminism, in the classical sense, cannot be used to reconstruct
mundane time because (a) mere indeterminism or chance is irrelevant to choice, (b) the
indeterminism arises from the measurement process which implicitly assumes the
intentional choice available with mundane time, so that the approach begs the question.
Equating unpredictability with contingency, as in the thermodynamic argument, or
the argument via chaos, is inadequate: this approach may be reversed and applied to
the past to result in a completely undecided past. In addition, this approach assumes
purely retarded interactions through the extra-physical beliefs that (a) explanations
must be solely from the past ('causal explanations'), and (b) the future, unlike the past,
is not knowable, except through prediction.
Not only does the hypothesis of a tilt in the arrow of time provide a solution to the
problem of time in physics and structured time in quantum mechanics, but a decided
past is consistent with an undecided future on this hypothesis. While the solution is
conservative from the point of view of physics, it requires some philosophical readjust-
ment, since some extra-physical beliefs need to be abandoned. Like mundane intention,
purpose, defined as future data, may be chosen independently of the past and may serve
to influence future evolution. For consistency with mundane time, future contingents
232 CHAPTERVIll

must be selectively amplified to macrophysics, while past contingents remain confined

to microphysics. The problem of how (or whether) this comes about remains open
because it seems to involve deep questions concerning living organisms.
Notes and References

Chapter I: Philosophical Time

1. According to Paul Davies, in The Physics of Time Asymmetry, Surrey Univ. Press, London, 1974, the
only remaining questions about time are philosophical in nature, and nothing qualitatively new can
emerge from a study of the physics of time.
2. R. M. Gale, The Philosophy of Time, Macmillan, London, 1968.
3. F. Waismann, in Gale, Ref. 2.
4. Not every language has three tenses. The Hopi language is tenseless. See B. Whorf Language, Thought
and Reality, MIT Press, Cambridge, Mass., 1950.
5. Often an order relation is required to be reflexive as well. The usual way of changing from < to S could
well be followed here by replacing 'earlier than' with 'earlier than or simultaneous with'. However, the
reflexivity of the temporal relation is usually reserved for describing cyclic time where an event may be
earlier than itself.
6. Ref. 2, p 67.
7. J.M.E. McTaggart [1908], reprinted in Gale, Ref. 2.
8. According to the serialist argument of J.W. Dunne, inAn Experiment with Time, Faber, London, 1942, the
regress is benign, and an infinite sequence of times provides the natural way of depicting time. Such claims
are contrary to the requirement of simplicity (Occam's razor). Moreover, I find the serialist argument quite
irrelevant to the claimed empirical evidence of precognitive dreams developed further by Priestley, Ref. lB.
9. It is not clear how much of this impression is due to language. As pointed out earlier, the Hopi language
has no tenses so that for the Hopi time does not move!
10. AN. Prior, Past, Present and Future, Clarendon, Oxford, 1967.
11. M. Dumett, 'Bringing about the past', Philosophical Review, 73, 1964. Reprinted in Gale, Ref. 2.
12. S.W. Hawking and G.F.R. Ellis, The Large Scale Structure ofSpace Time, University Press, Cambridge,
1974.
13. N. Rescher and A Urquhart, Temporal Logic, Springer, Wien, 1971.
14. N. Rescher,Many-ValuedLogic, McGraw Hill, New York, 1969.
15. A detailed account of Zeno's paradoxes of motion, with possible relevance to quantum mechanics, may
be found in A Griinbaum, Modem Science and Zeno's Paradoxes, Allen and Unwin, London, 1968.
16. A discretized formulation of Newton's laws, on the other hand, assumes that points of time may be
well-ordered, like the natural numbers and unlike the rationals.
17. Such as those implicit in the definition of (J. 15, (J being the Heaviside function and 15 the Dirac delta,
using a formal Fourier transform to express the product as a (divergent) convolution integral.
18. J. B. Priestley, Man and Tune, Aldus Books, London, 1964.
Additional Notes: Chapter I
19. A more recent survey of the ancient and medieval philosophy of time may be found in R. Sorabji, Time,
Creation and the Continuum, Duckworth, London, 1983.
234 NOTES: Chapter II: Newton's Time

20. Quotations from Sri Har~a use the translation of Khandana Khanda Khadya by Ganganath Jha, Vol.
II, Sri Satguru Publications, New Delhi, [1911], 2nd ed. 1986. At one point where high accuracy seemed
absolutely essential I have used the more literal 'self reference' in place of Jha's 'vicious circle'. I have
also modified Jha's translation of 11.1.39 from The Nyaya Sutras of Gautam, Vol II, reprint, Motilal
Baursidass, New Delhi, 1986. Similarly, the translations of Pfu;rini and Pataiijali incline towards the
more literal. The translation of Vyasa is abridged from J. H. Woods, The Yoga System of Patanjali... ,
Harvard Oriental Series, Vol 17, Cambridge, Mass., 1927, p 287-88.

Chapter II: Newton's Time

1. A. K. Raychaudhuri, Classical Mechanics, Clarendon, Oxford, 1984.
2. For more on Newton's bucket and Mach's principle, see F. Hoyle and J. V. NarIikar, Proc. R. Soc.,
A273, 1 (1963).
3. For a brief description of Milne's theory see G. J. Whitrow, Natural Philosophy of Time, 2nd ed.,
(Oxford: Univ. Press, 1980). The original account may be found in E. A. Milne, Z. Astrophys., 6, 1-95
(1933); Proc. R. Soc., A158, 324 (1937).
4. K. R. Popper, Realism and the Aim of Science. Postscript to the Logic of Scientific Discovery, Vol. I,
Hutchinson, London, 1982.
5. For the current status of Dirac's cosmology'arid the large number hypothesis, see the papers by
L. Halpern, pp 547-560, and J. V. NiUIikar, pp 561-573 in A. O. Barut, A. Van der Merwe and J.-P
Vigier (eds), Quantum, Space and Time - the quest continues, Part III: Papers dedicated to P. A. M.
Dirac (Cambridge: Univ. Press, 1983-84). For the original account see P. A. M. Dirac, Nature 139, 323
(1937); Proc. R. Soc., A 165, 199 (1938); Proc. R. Soc., A 365, 19 (1979); A Eddington, Proc. Camb.
Phil. Soc., 27,15 (1931).
6. K. R. Popper, The Open Universe. An argument for indeterminism. Postscript to the Logic of Scientific
Discovery, Vol. II, Hutchinson, London, 1982.
7. Many original articles on chaos may be found in Hao Bai-Lin, Chaos II (Singapore: World Scientific,
1992). Specific studies on chaos in biological systems may be found in G. Mayer-Kress, Dimensions &
Entropies in Chaotic Systems (Berlin: Springer, 1986).
8. For the conditions under which a differential equation admits a unique global solution see any book on
the subject. Easy to read is G. F. Simmons, Differential Equations (N.Y.: McGraw Hill, 1973).
9. Quoted by E. T. Whittaker, History ofthe Theories ofAether and Electricity Vol. I: The Classical Theories,
Thomas Nelson and Sons, London, 1951.
10. These quotations are taken from the biography of Newton in the Dictionary of Scientific Biography, C.
C. Gillispie, Editor in Chief, Charles Scribner's Sons, New York, 1981.

Chapter IlIA: The Michelson-Morley Experiment

1. H. A. Lorentz et al, The Principle of Relativity, Dover, 1952.
2. E. T. Whittaker, [1910], History of the Theories ofAether and Electricity Vol. I: The Classical Theories,
Thomas Nelson and Sons, London, 1951.
3. J. C. Maxwell [1879], Nature, 29 Jan 1880, 315.
NOTES: Chapter IIIB: Einstein's Time 235

4. A. Fresne~ Annales de Chimie et de Physique, 9, 57-&) (1818).

5. G. G. Stokes, Philosophical Magazine, 27, 9-15 (1845); 28, 76-81 (1846).
6. A. A. Michelson, Amer. J. Sci., 22, 120-9 (1881).
7. Jonathan Powers, Philosophy and the New Physics, Methuen, London, 1987.
8. A. A. Michelson and E. W. Morley, Philosophical Magazine, S.5, 24,449-63 (1887), Amer. J. Sci. 3,
333-45 (1887); E. W. Morley and D. C. Miller, Philosophical Magazine S.6, 8, 753-4 (1904); Phil.Mag.
680-5 (1905) and Proc. Amer. Acad. Arts Sci., 41, 321-7 (1905); D. C. Miller, Rev. Mod. Phys., 5, 203-42
(1933); see also Phys. Rev., 19(4), 407-8 (1922).
9. Conference on the Michelson-Morley Experiments held at Mount Wilson Observatory, Pasadena,
California, Feb. 4 and 51927, ApJ., 68(5), 341-402 (1928).
10. R. S. Shankland, S. W. McCuskey, F. C. Leone, and G. Kuert~ Rev. Mod. Phys., 27, 167 (1955); J.L.
Synge, The Scientific Proceedings of the Royal Dublin Society, 26, 45-54 (1952-4).
11. W. Kaufmann, Ann. der Phys., 19, 487 (1906).
12. A. Pais, 'Subtle is the Lord .. .' (Oxford: Univ. Press, 1982).
13. J. L. Synge, Relativity: the Special Theory, North Holland, Amsterdam, 1956, p 161-2.
14. Ref. 7.
15. R.S. Shankland, Amer. J. Phys., 31, 47 (1962);32, 16 (1964);41, 895(1973)

Chapter IIIB: Einstein's Time

1. Jagdish Mehra, 'Einstein's "First Paper" " Science Today, 10, 125 (1971).
2. HA. Lorentz et ~ The Principle of Relativity, Dover, New York, 1952, p 4.
3. Ref. 2, pp 5-6.
4. See, e.g., J.R. Reitz and FJ. Milford, Foundations of Electromagnetic Theory, India Book House 1978,
p397.
5. Ref. 2, p 22.
6. Ref. 2, p 18.
7. Ref. 2, p 24.
8. Abraham Pais, 'Subtle is the Lord .. .', (Oxford: University Press, 1982) p 167.
9. H. Poincare, Science and Hypothesis [1902], Eng. Trans., Dover, New York, 1952, p 111.
10. H. Poincare, The Value ofScience [1905]. Eng. Trans. by G.B. Halstead, 1913. Reprinted, Dover 1958,
p98.
11. Ref. 9, p 171.
12. Ref. 9, pp 169-70.
13. Ref. 9, p 141.
14. Ref. 10, p 99.
15. Ref. 9, p 217.
236 NOTES: Chapter IIIB: Einstein's Time

16. Ref. 9, pp 160-61.

17. Ref. 9, pp 171-72.
18. Ref. 10, p 99.
19. Ref. 10, p 103.
20. Ref. 10, p 108.
21. Ref. 10, p 107.
22. Ref. 9, p 243.
23. Ref. 9, pp 175-76.
24. Ref. 10, p 98. See also Ref. 9, pp 243-4.
25. Ref. 10, p 104.
26. Ref. 10, p 102.
27. Ref. 9, p 175.
28. Ref. 10, p 104.
29. Ref. 10, p 104.
30. Ref. 10, p 32; petito principi, and post hoc (after this) ergo (hence) propter hoc (because of this) are
names of classical logical fallacies, like the vicious circle.
31. Ref. 10, p 27.
32. Ref. 10, p 35.
33. Ref. 10, p 35.
34. Ref. 10, p 36.
35. Ref. 10, p 99.
36. Ref. 10, p 107.
37. Jagdish Mehra, Einstein, Hilbert and The Theory of Gravitation (Dordrecht: D. Reide~ 1974).
38. Ref. 8, p 134.
39. Ref. 8, p 168.
40. Ref. 8, p 168.
41. Ref. 8, p 172.
42. Ref. 8, p 172.
43. Ref. 10, p 100.
44. Ref. 10, p 99.
45. Ref. 10, p 98.
46. Ref. 9, p 175.
47. Ref. 8, p 128.
48. Ref. 8, p 164.
 NOTES: Chapter IV: Thermodynamic Time 237

49. Ref. 8, p 166.

50. Ref. 8, p 166.
51. Ref. 8, p 170.

Chapter IV: Thermodynamic Time

1. R. B. Ash, Infonnation Theory, Wiley-Interscience, New York, 1965.
2. S. W. Hawking and G.F.R. Ellis, The Large Scale Strncture of Spacetime (Cambridge: University Press,
1973). The relevance of singularities to recurrence may be found in F. J. Tipler, 'General relativity and
the eternal return', in Tipler (ed.), Essays in General Relativity, Academic Press, New York, 1981. See
also his article in Nature, 280, 203-5 (1979).
3. D. ter Haar, Elements of Statistical Mechanics, Reinhart, New York, (1954), p 342; H. L. Frisch, Phys.
Rev., 104, 1 (1956).
4. P.C. W. Davies, The Physics of Time Asymmetry, Surrey Univ. Press, London, 1974.
5. M. Kac, Probability and Related Topics in Physical Sciences, Wiley Interscience, New York, 1959.
6. N. Dunford and J.T. Schwartz, Linear Operators, Part I, Interscience, New York, 1958.
7. W. Rudin,FunctionaIAnalysis, McGraw Hill, New York, 1973.
8. L.D. Landau and E.M. Lifshitz, Statistical Physics, Pergamon, Oxford, 1958.
9.1.1. Gihman and A.V. Skorohod, The Theory of Stochastic Processes, Springer, Berlin, 1974.
10. L. Brillouin, Science and Infonnation Theory, Academic Press, New York, 1962.
11. R. J ancel, Foundations of Classical and Quantum Statistical Mechanics, Pergamon, Oxford, 1963; V. I.
Arnold and A. Avez, Ergodic Problems of Classical Mechanics, Benjamin, New York, 1967.
12. The biographies of Clausius and Boltzmann are based upon the Dictionary of Scientific Biography, C.C.
Gillespie, Ed. in Chief, Charles Scribner's Sons, New York, 1981. More details may be found in, e.g.,
S.G. Brush, The Kind of Motion We Call Heat, North Holland, Amsterdam, 1976.

Chapter VA: The Electromagnetic Field

1. M. Hesse, Forces and Fields, Westport, CT, 1973.
2. Ref. 1, p 78.
3. Ref. 1, p 168.
4. H. Poincare [1902], Science and Hypothesis (N.Y.: Dover, 1952), p 160-61.
5. Max Born, Einstein's Theory of Relativity, Tr. Broce, London, 1924, p 132.
6. Nyayavarttika, Tr. Ganganath Jha, 1919, Indian Thought, pp 238-48; In: K. H. Potter, Encyclopaedia of
Indian Philosophy, Vol. II, Motilal Banarsidass, Delhi, pp 334-5, 1977. The ftrst part of the quotation
refers to the purva palqa or the opponent's position, and is actually a translation of Sutra IV.2.24. See,
The Nyaya Sutras of Gautam, Tr. Ganganath Jha, Vol. IV, Motilal Banarsidass, New Delhi, reprint
1984.
7. Ref. 1, p 160.
238 NOTES: Chapter VB: Electromagnetic Time

8. Ref. 1, p 168.
9. Ref. 4, p 169.
10. In quantum-field theory, with the associated renormalization procedure, the field is regarded as
necessary; but this could well be a case of bad mathematics masquerading as good physics. Specifically,
the renormalization procedure can be made rigorous, using the Hahn-Banach theorem, but the usual
procedure does not quite 'work', since the non-uniqueness of the Hahn-Banach extension cannot be
removed for a variety of otherwise plausible Lagrangians. See, e.g., C.K.Raju, J. Phys. A: Math. Gen.,
16,3739 (1983); Hadronic J. Suppl., 1, 352 (1986).
11. This does not mean that a uniformly accelerated charge does not radiate. This point has caused some
confusion especially in the relativistic case, e.g., a freely falling charge need not radiate, while a particle
at rest in a gravitational field should. See, e.g., H.D. Zeh, The Direction of Time, Springer, Berlin, 1989.

12. K. R. Popper, Nature, 171, 538 (1956); 178,382 (1956); 179, 297 (1957); 181, 402 (1958); Brit. J. Phil.
Sci., 8, 151 (1957) ; The Open Universe, Hutchinson, London, 1982.
13. J.L. Friedmann and M.S. Morris, Phys. Rev. Lett., 66(4),1991; M.S. Morris and K.S. Thome, Amed.
Phys., 56, 395 (1988); M.S. Morris, K.S. Thome, and U. Yurtserver, Phys. Rev., D 42,1915 (1990), and
references cited therein.
14. JA. Wheeler and R.P. Feynman, Rev. Mod. Phys., 21, 425 (1949).
15. P.C.W. Davies, The Physics of Time Asymmetry, Surrey Univ. Press, London, 1974.
16. F. Rohrlich, Classical Charged Particles, Addison-Wesley, Reading, Mass., 1965.
17. G.N. Plass, Rev. Mod. Phys., 33, 37-62 (1961); or Ref. 15.
18. Ref. 16, p 135, P 140.
19. C.J. Eliezer, Rev. Mod. Phys., 19, 147 (1947).
20. cited in Ref. 17.
21. M. Dresden, HA. Kramers: Between Tradition and Revolution, Springer, New York, 1987.
22. Ref. 15.

Chapter VB: Electromagnetic Time

1. J.L. Synge, Proc. R. Soc., A177, 118-139 (1940); R.D. Driver, Phys. Rev., 178, 2051-57 (1969); D.K.
Hsing, Phys. Rev., D16, 974-82 (1977). In more than one dimension, the equations are not only more
complicated, but are of a different type, being 'mixed' or 'neutral'.
2. H. Van Dam and E.P. Wigner, Phys. Rev., 138B, 1576 (1965); 142,838 (1966). These equations were
considered in the context of relativistic point dynamics.
3. For more details, see, e.g., R.D. Driver, Introduction to Differential and Delay Equations (Springer,
Berlin, 1977); L.E. EI'sgol'ts, Introduction to the Theory of Differential Equations with Deviating
Arguments, (Tr.) RJ. McLaughlin (Holden-Day, San Francisco, 1966). The theorem does not apply to
equation (3) which admits unbounded delays.
4. P.A.M. Dirac, Proc. R. Soc., A167, 148 (1938), equations (55) and (56). Essentially the same derivation
may be found in, e.g., F. Rohrlich, Classical Charged Particles (Addison-Wesley, Reading, Mass., 1965)
p 142, who manages to avoid another Taylor expansion used by Dirac.
NOTES: Chapter VB: Electromagnetic Time 239

5. L. Page, Phys. Rev., 12, 371 (1918); 24,296 (1924); A. Einstein, L. Infeld and B. Hoffmann, Ann. Math.,
39,65 (1938); H. P. Robertson, Ann. Math., 39,101 (1938); L. Infeld, Phys. Rev., 53, 836 (1938); A.
Eddington and G. L. Clark, Proc. R. Soc., A166, 465(1938); A. Einstein and L. Infeld, Can. J. Math., 1,
209(1949); G. L.Clark, Proc. R. Soc. (Edinb.), A64, 49 (1954); B. Bertott~ Nuovo Cim., 12, 226 (1954).
A more detailed list may be found in the article by P. Havas in Statistical Mechanics of Equilibrium and
Non-Equilibrium, ed. J. Meixmer (Amsterdam: North Holland, 1965). More recently, somewhat similar
approximations have been attempted by L. P. Grishchuk and S.M. Kopejkin, in: J. Kovalevsky and VA.
Brumberg (eds), Relativity in Celestial Mechanics and Astronomy, IAU (1986) pp 19-34; V.I. Zhdanov,
J. Phys. A: Math. Gen., 24, 5011-27 (1991).
6. One may prove this using the standard theory of linear differential equations, and proceeding in steps
from past data, x = 'P on [-r, 0], say.
7. In his book cited in Ref. 3, p 22.
8. E. Hairer, S.P. Nrsett, and G. Wanner, Solving Ordinary Differential Equations, Springer Series in
Computational Mathematics, Vol. 8 (Springer, Berlin, 1987).
9. An additional problem is that discontinuities at the ends of delay intervals, which are smoothed out for
retarded equations are not smoothed out for mixed-type equations, as noted by P. H~ in Schulman
(Ref. 15), while purely advanced equations progressively lose smoothness. Numerical solution of delay
equations is, of course, possible, in some cases, and the figures obtained here have used numerical
techniques, with adaptive step-size control, and discontinuity-handling.
10. C.J. Eliezer, Proc. R. Soc., A194, 543 (1948); W.B. Bonnor, Proc. R. Soc, A337, 591 (1974); T.C. Mo
and C.H. Papas, Phys. Rev., D 4, 3566 (1971); J.C. Herera, Phys. Rev., DIS, 453 (1977).
11. J. Huschilt and W.E. Bayles, Phys. Rev., D 9, 2479 (1974); E. Conay, Phys. Lett., A 125,155 (1987); E.
Conay, Int. J. Theor. Phys., 29,1427 (1990).
12. A. Schild, Phys. Rev., 131,2762 (1963).
13. C. M. Anderssen and H. C. von Baeyer, Phys. Rev., D 5, 802 (1972); Phys. Rev., D 5, 2470 (1972).
14. R. D. Driver, Phys. Rev., D 19, 1098 (1979).
15. L. S. Schulman, J. Math. Phys., 15, 295-8 (1974).
16. K. L. Cooke and D. W. Krumme, J. Math. Anal. Appl., 24, 372-87 (1968).
17. A discontinuity develops because the characteristics of a nonlinear hyperbolic partial differential
equation may intersect. Such a phenomenon is well known in the theory of shocks in fluids. While one
can deny that the Euler equations provide a true model of fluid flow, one can hardly deny
'hyperbolicity', which more or less amounts to a belief in finite propagation speeds, and well-posedness
of the Cauchy problem; see, e.g. Bers et al, Partial Differential Equations (New York: Wiley-Intersci-
ence, 1968). The Hawking-Penrose theorems, for example, may be interpreted in this way.
18. J. L. Friedmann and M.S. Morris, Phys. Rev. Lett., 66(4) (1991).
19. The assumption (i) might seem a bit dubious if stated honestly: Physicists (i.e., human beings) are free
to perform any experiment, and physical laws are no 'serious' constraint. We argue later on, in Chapter
VIII, that the precise assumption here is an assumption about the nature of time, derived from
mundane experience.
20. A. Einstein and W. Ritz, Phys. Z., 10, 323 (1909).
240 NOTES: Chapter VIA: Bell and Non-Locality

21. A possible cosmological reason for this choice has been put forward by D.W. Sciama, Proc. R. Soc.,
A273, 484 (1963). The argument is that if Fin c# 0, it could diverge, as in Olbers' paradox. For a
resolution of Olbers' paradox, see Chapter VII.
22. J. A. Wheeler and R.P Feynman, Rev. Mod. Phys., 17, 157 (1945); 21, 425 (1949).
23. J. G. Cramer, Rev. Mod. Phys., 58, 647 (1986).
24. C. K. Raju, J. Phys. A: Math. Nucl. Gen., 13, 3303-17 (1980) .
25. For example, Sciama, Ref. 21, has slightly reformulated the argument: the sum of the retarded fields
in equation (29) is an incoherent sum, and hence vanishes, whereas the sum in equation (33) is a
coherent sum and compensates by just the right amount to produce anti-damping. That is, the
conclusion rests on time-asymmetric (thermodynamic) assumptions similar to those in Popper's pond
paradox.
26. J. E. Hogarth, Proc. R. Soc., A267, 365-383 (1962).
27. P. C. W. Davies, 'Is the universe transparent or opaque', J. Phys. A: Gen. Phys., 5,1722-37 (1974).
28. F. Hoyle and J. V. Narlikar, Proc. R. Soc., A277, 1-23 (1964).
29. The hypothesis of finite size seems ad hoc, and does not seem to augur well for quantum mechanical
generalizations. However, an interpretation of quantum mechanics was proposed on this basis [C. K.
Raju, Int. J. Theor. Phys., 20, 681-96 (1981)], and it was proved that [CO K. Raju, J. Phys. A, 16, 3551
(1983); Hadronic J. Suppl., 1 (1986)] one could modify the propagators of quantum field theory, on the
same hypothesis, to obtain a Lorentz covariant finite field theory for all polynomiallagrangians.
30. R. B. Partridge, Nature, 244, 263-65 (1973).
31. Ref. 24.
32. M. L. Heron and D.T. Pegg, J. Phys. A: Math. Nucl. Gen., 7, 1975-9 (1974).

Chapter VIA: Bell and Non-Locality

1. N. Rescher and A. Vrquhart, Temporal Logic (Wien: Springer, 1971).

2. W. H. Newton-Smith, The Structure of Time, Routledge and Kegan Paul, London, 1974.
3. Rescher, Ref. 1, tries to relate the V-calculus to the measurement of time by beginning with a set T of
'instants' and a metric (or pseudo-metric) on it. This induces a V-relation that is irreflexive, asymmet-
ric, and transitive, provided one can find two 'instants' to, t1, such that Utot1. Such a V-relation,
obtained by globalizing the past and future at one instant of time, turns out to be counter-intuitive and
also counter-physical when the local (physical) earlier-later relationships are not globally consistent.
4. For instance, in the one-dimensional case of two like charges without radiation damp,ing or any other
external forces, people have attempted to prove the uniqueness of solutions provided the charges stay
separated by a distance which is 'a large multiple of the radius of the universe'. We recall that the
higher-dimensional case is qualitatively different. See, e.g., R. D. Driver, Phys. Rev., 19, 1098-1107
(1979).
5. The paradigm is an observable with non-degenerate discrete (preferablyfmite) spectrum. The difficulty
is that von Neumann's collapse postulate does not trivially generalise to the continuous spectrum case,
as he had thought.
6. Strictly, the terminology of spin detection or spin-meters may be preferable.
NOTES: Chapter VIA: Bell and Non-Locality 241

1. EPR and other papers on measurement theory are reproduced in J.A. Wheeler and W.H. Zurek (eds),
Quantum Theory and Measurement (Princeton: Univ. Press, 1983). The original reference is: A.
Einstein, B. Podolsky, N. Rosen, Phys. Rev., 47, m (1935).
8. D. Bohm, Quantum Theory, Prentice-Hall, New Jersey, 1951.
9. S. Kochen and E. Specker, J. Math. Mech., 17, 59 (1967). Reprinted along with other relevant papers,
such as Gleason's in C. A. Hooker (ed) The Logico-Algebraic Approach to Quantum Mechanics, D.
Reidel, Dordrecht, 1983.
10. J. M. Jauch and C. Piron, Helv. Phys. Acta 36, Fa1 (1963).
11. G.W. Mackey, Mathematical Foundations of Quantum Mechanics, Benjamin, New York, 1963.
12. A. M. Gleason [1956] reproduced in Hooker, Ref. 9.
13. A detailed account of these no-hidden-variable theorems and the resulting constraints on hidden
variable theories may be found in F. J. Belinfante, A Survey of Hidden Variable Theories, Pergamon
Press, Oxford, 1913. See also Bell's 1966 paper, Ref. 15 or 16 below.
14. D. Bohm, 'Quantum theory in terms of "hidden" variables I,' Phys. Rev., 85, 166-119 (1952); 'A
suggested interpretation of the quantum theory in terms of "hidden" variables II,' Phys. Rev., 85, 180-93
(1952).
15. J. S. Bell, Rev. Mod. Phys., 38, 458-62 (1966). This paper was prepared prior to the more famous paper
inPhysics, 1,195 (1964).
16. J. S. Bell, Speakable and Unspeakable in Quantum Mechanics (Cambridge: Univ. Press 1987) pp 45-46.
The book contains a collection of all the papers of Bell on the philosophy of quantum mechanics.
11. e.g. J. F. Clauser and A. Shimony, Rep. Prog. Phys., 41, 1881 (1918); A. Aspect, P. Grangier and G.
Roger, Phys. Rev. Lett., 47, 460 1981); 49, 91 (1982); A. Aspect, J. Dalibard, and G. Roger Phys. Rev.
Lett., 49, 1804 (1982). W. Perrie et al, Phys. Rev. Lett., 54,1190 (1985).
18. G. Zukav, The Dancing Wu-Li Masters (N.Y.: Flamingo, 1981). The book is, otherwise, a well-written
popular introduction.There have been more serious attempts at generalization to telepathy, telekinesis
etc., by some physicists such as o. Costa de Beauregard.
19. o. Pemose and I.e. Percival, Proc. Phys. Soc., 79, 605 (1962).
20. Bell 1966, Ref. 15, or chapter 1 of Ref. 16.
21. J. Bub, The Interpretation of Quantum Mechanics (N.Y.: Benjamin, 1914).
22. E. P. Wigner, Amer. J. Phys., 33,1005-9 (1910).
23. S. Freedman and E. P. Wigner, Found. Phys., 3, 451 (1913).
24. J. S. Bell, Ref. 16.
25. Some of the issues are considered, for instance, in R.K. Clifton, M.L.G. Redhead, and J. N. Butterfield,
Found. Phs., 21,149-183 (1991). An easier exposition may be found in Michael Redhead, Incomplete-
ness, Nonlocality and Realism, Clarendon, Oxford, 1981.
26. According to Barut (personal communication), Bell disagreed, though I do not know the nature of the
disagreement. Bell had considered a very similar example to motivate his proof.
21. A. O. Barut and P. Meystre, Phys. Lett., IDS-A, 458-62, 1984; A. O. Barut, Hadronic J. Suppl., 1,
(1986).
28. A. Garg and D. Mermin, Phys. Rev., D3S, 383 (1987).
242 NOlES: Chapter VIB: Quantum-Mechanical Time

29. 'local realism', as defined by Garg and Mermin, Ref. 28, refers to the averaging procedure for a hidden
variable theory.
30. e.g., A Fine, Found. Phys., 21, 365 (1991).
31. e.g., E. Santos, Found. Phys., 21, 221 (1991).
32. G.C. Hegerfeldt, Phys. Rev., DI0, 3320 (1974); G. C. Hegerfeldt and S.N.M. RUijsenaars, Phys. Rev.,
D22, 377 (1980). This has been made almost a starting point for stochastic quantization by E.
Prugovecki and members of his school. See also, G.C. Hegerfeldt, Phys. Rev. Lett., 72, 596 (1994).
33. C. K. Raju, J. Phys. A: Math. Gen., 16, 3739-53 (1983).
34. See, e.g., R D. Richtmeyer, Principles of Mathematical Physics, Vol. 1. Springer, Berlin (1982), or M.
Taylor, Pseudo-Differential Operators, N.Y., Academic (1974).
35. As a leading expert, Dafermos, remarked in private conversation, the right way to defme hyperbolicity
is to put in more and more physics!
36. See, e.g., P.C.W. Davies, The Physics of Time Asymmetry, Surrey Univ. Press, London, 1974.
37. J. B. Hartle, in: Proceedings of the Fifth Marcel Grossman Meeting, D. Blair et al (eds), R. Ruffmi
(series ed.), Wiley Eastern, Singapore, 1989, pp 107-24.
38. e.g., D. T. Pegg, J. Phys. A: Math., Gen., 24, 3031-40 (1991).
39. AJ. Leggett and A Garg, Phys. Rev. Lett., 55, 857 (1985). AJ. Leggett, 'Quantum mechanics at the
macroscopic level.' In: J. de Boer, E. Dal and O. Ultbed (eds), The Lessons of Quantum Theory: Niels
Bohr Centenary Symposium, Elsevier, Amsterdam, 1986, pp 35-58, and in: G. Grinstein, and G.
Mezenko (eds) , Directions in Condensed Matter Physics: Memorial Volume in Honor of Shang-keng Ma,
World Scientific, Singapore, 1986, pp 185-248.
40. S. Foster and A Elby, Found. Phys., 21, 773-80 (1991).
41. First noted by W. Ehrenberg and RW. Siday, Proc. Phys. Soc. (Lond.) 862, 8 (1949), and by Y.
Aharonov and D. Bohm, Phys. Rev., 115, 485 (1959). An exhaustive review may be found in M. Peshkin
and A Tonomura, The Aharonov-Bohm Effect, Lecture Notes in Physics No. 340, Springer, Berlin,
1989.
42. Article by M. Peshkin, p 4, in the book cited in Ref. 41.

Chapter VIB: Quantum-Mechanical Time

1. Though we do not enter into this question here, a structured time is also closer to the time of daily
experience.
2. For those who have only vaguely heard of measure theory, the significance of Borel sets is that it always
makes sense to speak of the measure of such a set with respect to a regular Borel measure! The
significance of a (fmite) regular Borel measure is that it corresponds (both ways) to an integral which
will (i) integrate continuous functions, and (ii) provide a fmite value for the integral of a continuous
function which 'vanishes at infmity', i.e., is less than the proverbial epsilon outside a compact set. The
formal definition of a Borel set in a topological space X is that it is an element of Bx, the smallest
a-algebra which contains all open sets.
3. P. R Halmos, Introduction to Hilbert space and the Theory of Spectral Multiplicity, Chelsea, New York,
1951. The whole book is devoted to the proof of this theorem.
NOTES: Chapter VIR: Quantum-Mechanical Time 243

4. From Dirac's classic treatise on Quantum Mechanics, as cited by J.S. Bell, Speakable and Unspeakable
in Quantum Mechanics (Cambridge: University Press, 1987), p 40.
5. For the difficulties in extending von Neumann's collapse postulate to observables with continuous
spectra, and for a new formulation see M.D. Srinivas, 'Collapse postulate for observables with
continuous spectra', Commun. Math. Phys., 71, 131-58 (1980).
6. E. P. Wigner, Phys. Rev., 40, 749 (1932).
7. E. P. Wigner, in: Perspectives in Quantum Theory, W. Yourgrau and A. van der Merwe (eds), MIT
Press, Cambridge, Mass., 1971.
8. The other possibility, pursued for instance in stochastic quantization, is to change the axiomatic basis
of q.m., so that joint probability distributions do exist even for canonically conjugate variables.
9. For the measure-theoretic foundations of probability theory, as enunciated by Kolmogorov, see, for
instance, P. Halmos, Measure Theory, D. Van Nostrand, New York, 1950. To change from sets to
sentences, see, e.g., J. R. Lucas, The Foundations of Probability, Clarendon, Oxford, 1957.
10. The logico-algebraic approach was initiated by G. Birkhoff and J. von Neumann, Ann. Math., 37,823
(1936). For further developments of this theory, see V.S. Varadarajan, Geometry of Quantum Theory,
Vol. 1, Van Nostrand, Princeton, NJ., 1968, and C. A. Hooker (ed) The Logico-AlgebraicApproach to
Quantum Mechanics, D. Reidel, Dordrecht, 1975.
11. A 'subspace' will always mean a closed linear manifold.
12. One has to be a little careful in applying the analogy to implication. The sentential connective ~ should
not be confused with the relation of implication. Thus, it makes sense to speak ofA~(B~A), but not
of as (bSa).
13. See G. Kalmbach, Hilbert Lattices, World Scientific, Singapore, 1987, for a counter-example, which
however, does not apply to probability measures.
14. Those interested in probing further may like to be discouraged by the text of S. Lang, Algebra,
Addison-Wesley, Reading, Mass., 1965.
15. C. K. Raju, 'Quantum mechanics and the microphysical structure of time' [1990]. In: R. Nair (ed) In
Search of Quantum Reality (to appear).
16. This is true directly for charged particles, and indirectly, at least, for all particles which interact in any
way with charged particles.
17. The whole debate over 'hidden variables' is related to the metaphysical assumption (of linear time)
that 'reality is unambiguous'. This assumption breaks down with branching time.
18. The terminology derives from temporal logic.
19. See, e.g., K.R. Apt (ed), Logics and Models of Concumnt Systems, Springer, Berlin, 1984, NATO ASI
Series in Computers and Systems Science, Vol. 13. The moral of the story is that one should link
non-locality to CSP rather than ESP!
20. The usual definition of a random variable as a measurable jUnction assumes the possibility of such an
in-principle specification.
21. N. Rescher and A. Urquhart, Temporal Logic (Wien: Springer, 1971).
22. This does not lead to any paradoxes, as discussed in detail in Chapter VB. In terms of the analogy to
the formally similar but-more pragmatic CSP, the ALT construct of OCCAM is non-deterministic, and
the contingent 'present' is what accounts for the current practical difficulty in constructing a debugger
for parallel processors. Incidentally, in India, this difficulty has been currently resolved by abandoning
244 NOTES: Chapter VII: Cosmological Time

CSP and adopting a queuing model in the PARAS system software developed by C-DAC for use with
parallel FORTRAN, and C, which have some constructs resembling the PAR and ALT of OCCAM.
The empirical significance of a contingent past is discussed later on.
23. Traditionally, a proposition is what a sentence expresses or means. Formally, 'meaning' may be deemed
to have been assigned to a sentencep by specifying the set of 'worlds', Ip I, in which the sentence is true,
so that one knows when the sentence is true and when it is false. Since we shall be using different notions
of 'world', we should have, strictly, distinguished between statements, truth-functional propositions and
quasi truth-functional propositions. We shall, however, not retain this difference explicitly, in the hope
that the type of entity being considered would be clear from the context.
24. Bas C. Van Fraassen, in: Hooker (ed) Contemporary Research in Foundations and Philosophy of
Quantum Theory, D. Reide~ Dordrecht, 1973, p 84.
25. Strictly speaking, a state should be identified with a convex combination of the probability measures
induced by each q.t.f. world in the collection. Thus, to each q.t.f. world C/) in a state, one must also assign
a number t(C/), O<t(C/) < 1, such that I t(C/) = 1, the sum being taken over all q.t.f. worlds in the state.
26. The selection function is, in fact, a projection: fp 0 fp = fp , and could, under the circumstance (25),
quite literally be identified with a projection operator in Hilbert space.
27. W. L. Harper, R. Stalnaker, and G. Pearce (eds), Ifs, D. Reidel, Dordrecht, 1981.
28. See Harper et ~ Ref. 27.
29. I will use statements and propositions interchangeably; see note 23.
30. Ref. 15, Theorems 1, and 2.
31. Ref. 15, Theorem 5, equivalence of (i) and (v).
32. Ref. 15.
33. J.C.T. Poo~ (reproduced in Hooker, Ref. 24 above). In the next paper in Hooker (ed), Pool also
considers the possible significance of semi-modularity.
34. J. G. Cramers, Rev. Mod. Phys., 58, 647 (1986).
35. See van Fraassen, Ref. 24.
36. As a specific example, consider van Fraassen's assumption (2), in Ref. 24, corresponding to Pool's
Axiom 1.3, in Ref. 33. This says that cp .. OJ implies p" q. That is, if the necessity of p implies
the necessity of q, then the possibility ofp implies the possibility of q. I do not know of any traditional
modal logic in which this is true.
37. C. K. Raju, Int. J. Theor. Phys., 20, 681 (1980).

Chapter VII: Cosmological Time

1. D.W. Sciama, In: Early Evolution ofthe Universe andlts Present Stnlcture, G.O. Abell and G. Chincarini
(eds), IAU, 1983, pp 493-506.
2. This paradox has a long history, dating back to centuries before Olbers. See, e.g. E. R. Harrison,
Darkness at Night: A Riddle of the Universe (Cambridge, Mass.: Harvard Univ. Press, 1987). The name
Olbers' paradox seems to have been popularised by H. Bondi in his classic text on Cosmology, 2nd ed.
(Cambridge: Univ. Press, 1960).
NOTES: Chapter VII: Cosmological Time 245

3. E. R. Harrison, in: Proc.lAU No. 139, Galactic and Extragalactic Background Radiation, S. Bowyer and
Ch. Lienert (eds), Kluwer Academic, Dordrecht, 1989, pp 3-17.
4. P.C.W. Davies, The Physics of Time Asymmetry, Univ. of Surrey Press, London, 1974; H. Bondi and T.
Gold, Mon. Not. R. Astr. Soc., 108, 252 (1948).
5. For a clear exposition of the general relativistic redshift, see J.L. Synge, Relativity, the General Theory
(Amsterdam, Elsevier, 1966). Briefly, every redshift, including the gravitational redshift, may be
regarded as a Doppler shift.
6. W. de Sitter, Mon. Not. R. Astr. Soc., 78,3 (1917).
7. One has to distinguish between an emission and absorption redshift. Both tend to be similar, though not
identical. This is strange because one would expect the absorption redshift, due to absorption by
intergalactic dust, to have a range of values all the way up to the emission redshift.
8. H. Arp, personal communication (1988).
9. R. K. Sachs and A. M. Wolfe, 'Perturbations of model and angular variations of the microwave
background', Ap. J., 147, 73-90 (1967).
10. W.R. Stoeger, G.F.R. Ellis and B.G. Schmidt, 'Gauge-invariant variation in the redshift and cosmic
background radiation anisotropies', Gen.Rel. Grav.,23, 1169-94 (1991).
11. A 154,uK (5 ppm) anisotropy in the CMBR was detected by COBE (Cosmic Background Explorer).
G. I. Smoot, C.L. Bennett, A. Kogut et at, Ap. J., 316, L1 (1992). For the interpretation of this
anisotropy as favouring cold dark matter, see C.L. Bennett et at, ApJ., 316, L7 (1992); R.L. Davis,
H.M. Hodges, G.I. Smoot et at, Phys. Rev. Lett., 69, 1856 (1992).
12. The surface r = 2m has area W, but has infinite volume; curvature coordinates may not be used while
going across the surface r = 2m.
13. S. W. Hawking and G .F.R. Ellis, The Large Scale Structure ofSpacetime (Cambridge: Univ. Press, 1974,
1987) pp 286-7.
14. P.D. Lax, Hyperbolic Systems of Conservation Laws, American Mathematical Society Publications,
Providence, Rhode Island, 1973. The example is that of the flow of a compressible fluid in a tube with
the piston being pushed out with constant acceleration. This now-standard example may also be found
in, e.g., R. D. Richtmyer, Principles ofAdvanced Mathematical Physics, Vol. I, Springer, Berlin, 1978,
pp 388-90. The situation in GRT is analogous: a collapse, once initiated, must continuously accelerate,
leading to a finite-time 'blow up' of smooth solutions.
15. For a review of the techniques and pitfalls here, see C. K. Raju, 'Distributional matter tensors in
relativity', Proceedings of the Fifth Marcel Grossman Meeting, D.G. Blair et al eds, Wiley Eastern,
Singapore, 1989) p 421. The background may be found in J. Phys. A: Math. Gen., 15, 381-96 (1982); 15,
1785-97 (1982).
16. Vera C. Rubin, 'What is the matter in spiral galaxies?' In: Large-Scale Structure of the Universe, V. C.
Rubin and G.V. Coyne (eds) (Princeton: Univ. Press, 1980).
17. K. Gode~ Rev. Mod. Phys., 21, 447-50 (1948).
18. S. Chandrasekhar, in: The Nature of Time, (ed) T. Gold, Cornell Univ. Press, Ithaca, 1967, pp 68-74.
19. Cited by S. Chandrasekhar, Ellipsoidal Figures of Equilibrium, Yale Univ. Press, New Haven, 1968.
20. As already noted by GOdel, on the speculation that galaxies may have formed due to inhomogeneities
resulting from rotation, it is not essential that the axes of rotation should be preferentially oriented like
those of the planets.
246 No'IES: Chapter VIII: Mundane Time

Chapter VIll: Mundane Time

1. J. B. Preistley, Man and Time, Aldus Books, London, 1964, p 97.

2. S.w. Hawking and G.F. R. Ellis, The Large Scale Structure ofSpacetime (Cambridge: Univ. Press, 1974),
p 189.
3. K.R. Popper, The Open Universe, Hutchinson, London, 1982.
4. R. Sorabji, Time, Creation and the Continuum, Duckworth, London, 1983, p 299.
5. In addition to the arguments given earlier, Popper argues, in Ref. 3, that in the case of relativity, the
information available to an observer must come from the past null cone. Therefore, complete Cauchy
data can never be available to Laplace's demon. This argument is correct as it stands, but nothing
forces Laplace's demon to solve a Cauchy problem rather than a characteristic initial value problem.
For the latter problem the required data lies on the past null cone. Therefore, the suggestion that there
is an added source of uncertainity in relativity is incorrect.
INDEX

A-series problem, 123

change essential to time, 19 advanced potentials, 107
A-determinations to B-statements, 16 physically characterized, 107
explained, 15-16 aether
formalized using structure of time, 216 diversity in meanings of, 65
see also McTaggart drift and dipole anisotropy in CMBR, 201
related to change of logic, 20 intangibility, 52
in classical mechanics, 43 origin in theories of action by contact, 50
A-stability, 122 age of the universe
aberration, stellar, 51 does it have one? 206
see also Stokes, Fresnel, Michelson-Morley expansion age, and physical age, 200
Abraham-Lorentz mode~ 109, 114 obtained by integrating Friedmann
structure dependent terms, 110 equation, 200
electromagnetic electron, 110 reconciling the disagreement, 200
absolute time Aharonov-Bohm effect, 146, 158
and foliation of spacetime, 199, 207 and locality, 157
absorber response, 129 explained, 156
absorber theory stray field problem, 156
empirical tests of, 130 Airy, 53
see also Partridge's experiment anticipatory phenomena, 118
see also Heron and Pegg, Pegg, Wheeler- and Maxwell's demon, 230
Feynman, Hogarth, Hoyle-Narlikar impossibility of explanation from past
conditions for consistency of causes,123,228
retarded radiation, 130 time asymmetry of, 123
Raju, 130, 191 see also advanced o.d.e.
accessibility and access relation, 178, 184 Aquinas, Thomas, 36, 50
action at a distance, 126 Arago, F., 52
and elliptic equations, 104 argument from chaos, 4, 46-47, 226
equivalent to field theory, 111
assumes retarded interaction, 4
how does it operate? 103
effect of denying retarded interactions, 4, 228
its rejection motivated aether, 103
time-symmetry forces past contingents, 4, 47,
why is it 'spooky'? 103
see also non-locality, ellipticity, Schwarzschild- 226-227
Tetrode-Fokker action argument from complexity, 7, 226
action by contact and Laplace's demon, 226
and chains of causes, 103 complexity of human beings, 225
and introduction of aether, 50 stated,44
see also locality, hyperbolicity transition probabilities decided, 225
active galactic nuclei, 197 Aristotle
Adler, Friedrich, 57 accused of ignoring the obvious, 215
admissibility, 182 action by contact, 36, 50, 103
advanced action causes for motion, 33
and teleology, 228 definition of time, 13
and temporal assumptions in the mystery- modalities, 22-25
novel analogy, 228 paradox of ever-shrinking present, 12
as a perturbation, 181, 230 sea battle, 21
makes prediction impossible, 123-124 Arp, Halton, 197
needed to balance action and reaction, 111 arrow of time, 6, 139
not logically impossible, 126 see also thermodynamic arrow etc.
paradoxes of, 6, 108 Aspect, Alain 155, 183
Popper's pond as illustration of two-time Ash, R.B., 83
theory, 220 atomism, 26
purpose as final data, 229 Boltzmann on, 101
see also two-time theory, advanced o.d.e., Einstein on, 49
anticipatory phenomena Augustine, 5,14-15
advanced o.d.e., 118-119 biographical sketch, 13
non-existence of solution to past-value rejection of cyclic time, 13
 248 INDEX

time as subjective, 14 Boole, G., 166

automorphisms, of a probability space, 90 Borel measure, 97
rel(Ular, 163
B-series,6 BoreT set, 163, 176, 181
see also A-series Bom,M.
explained, 16 on Poincare, 73
flow of time as hallucination, 221 on ellipticity and action at a distance, 104
implied by relativistic notion of Boscovich, R., 105
simultaneity, 71, SO, 190, 192, 218 bouncing photon, 79
in classical mechanics, 43 used to define proper clock, 70
see also McTaggart Boyle, R., 50
Bacon, Francis, 158 law, 54
on influence at a distance, 103 Bradler, Thomas, 51
Banach-Alaogulu theorem, 97 Brilloum, L. 91
Barrow, Isaac, 26, 39-40, 44 de Broglie, Louis, 148
Barut A.O. and P. Meystre, 155, 158 broken time, 223
beginning of time see also epistemically broken time, ontically
and singularities, 203 broken time, evolutionary law (stochastic),
distinguished from singularity, 205 occasionalism
was it the big bang? 194, 202 Bub,J.,155
Bell,J.S. Burbidge, G.
survey of no-hidden-variable theorems, 150 on Hubble's law, 196
Bell locality, 146
and locality of quantum field theory, 157 caloric theory, 98
distinguished from simple correlatIOn, 152 Carnap, Rudolph, 40
see also locality Carnot's theorem, 98
postulate stated, 153 Carroll, Lewis, 29, 92
thermodynamic locality irrelevant to, 154 Cauchy problem, 125, 208
Bell's inequalities Chandrasekhar, S., 208
detector efficiency, 6, 155 limit 202
summary of experiments, 153 change'
'delayed choice', 153-154 and motion, 13
and quantum correlation function, 152 essential to time, 18
stated, 153 related to A-series, 19
Bergson, H., 27 Russell's definition and its criticism, 18
Besso, M., 56 chaos
Bhartrhari,14 see argument from chaos
big bang, 201 Liapunov exponent, 46, 226
cosmological model, 130 Chapman-Kolmogorov equation, 86
Birkhoff, G. and J. von Neumann, 172 Cheseaux, J.P. Loys de, 193
algebraic approach to q.m. explained, 165-67 choice
physical interpretation of lattice a wrong and occasionalism, 3
question, 181 and unpredictability, 223
relation to Hilbert-space axiomatics, 165 see also mundane time, falsifiability
body trade-off with intention, 229
what is it? 34 Clauser, J.F. 153
as mass point, 43 Clausius, Rudolph, 83
see also matter tensor biographical sketch, 98-99
Bohm, David, clocks
non-locality of his hidden-variable needed to define force, 38
theory,150,152 physical clocks and proper clocks, global
reformulation of EPR paradox, 148 relationship, 208
see also Aharonov-Bohm effect undefined in classical mechanics, 38
Bohr, N., 114 closed timelike curves, 47
Boltzmann, Ludwig, 83, 87, 95 CMBR
biographical sketch, 99-100 properties summarized, 201
on physical laws and humans, 222 dipole anisotropy, 201
transport equation, 100 quadrupole amsotropy, 201, 206, 210
H-theorem,loo COBE
constant, 99 detection of quadrupole anisotropy, 206
Bondi, Hermann, 194 coherence,55,108,123
 INDEX 249

commutation relations, 162 is objectivity possible? 14

compact tOp'ol<>&ical space, 89, 96, 97 not necessary, 14
compass of inertia, 'lff7 operational, 13
see also Foucault pendulum density parameter, 200
conservation law, 204 Descartes,Rene,66,I03
contingents contiguous action, 36, 50
and conditionals, 179 plenum, 51
do ontic and epistemic contingents differ? 217 str~t line motion as natural, 33
epistemic, 217 determmism
epistemic equated with ontic, 226 distinguished from fatalism, 21
irreducible, 177 see also Dumett's paradox
ontic, 217 hard, 125
past contingents a necessary consequence of of classical mechanics, 43
equating epistemic and ontic contingents, 227 _psycholo&i:ca1, 229
thermodynamic contingents are epistemic, 218 differentiability
Cooke K.L. and D.W. Krumme, 125 and singularities and shock waves, 204
coordinates differential equations, ordinary, 118, 190
generalized, 93 mixed-type deviating arguments, 119
Copenhagen interpretation, 183, 223 with advanced deviating arguments, 6
and choice, 223 with retarded deviating argument, 118
and instrumentalism, 148 time-symmetry of, 12(J
summarized, 148 functional, 119
Copernicus, 148 see also retarded, advanced, mixed-type o.d.e.
principle of modesty, 196 differential equations, partial
cosmological arrow of time, 192 see evolutionary law, hyperbolicity, ellipticity
and closed timelike curves, 2!J7 Diodorus Cronus, 22
tilt in, 2!J7, 210 see Master Argument
cosmological principle, 198 Dirac, PA.M., 113-114, 121-122, 127
cosmological model cosmology, 42, 2(J8
baryon symmetric, 2!J9 approach to divergence difficulties, 111
see also Einstein-de Sitter, Dirac, FLRW, childhood of delta function, 2(J3
Gtidel definition of radiation field, 112
cosmology derivation of radiative damping, 160
speculation in, 191 elimination of runaway solutions, 113
which theory should one use? 192 preacceleration as evidence for finite size of
counterfactuals, 162 electron, 114
CP-violation, 143,2!J9 radiative damping and theory of positron, 112
CPT theorem, SO, 143,2!J9 direct action
Crick, Francis, 17 electrodynamics, 190
CSP, 175 divergence
curvature parameter, 200 of self-field, 109
cyclic time, 140 division ring, 172
anathema against, 44 Driver, R.D., 125
not representable by a binary earlier-later Dumett's paradox, 21
relationship, 142 as illustration of two-time theo~, 219-22!J
Augustine's rejection of, 13 indeterminism and unpredictability, 227
no distinction between fatalism and
determinism, 22
Dahlquist barrier, 122 Dunford, N. and J.T. Schwartz, 97
Dallanacirya, 15 dynamical system
dark matter, 2!J5 abstract, 90
distribution compared to distribution of
luminous matter, 2!J6 Ehrenfest, P and T., 86
existence inferred from rotation curves of model, 86-87, 225
galaxies,2!J6 refutation of reversibility paradox, 90
see also galaxies, rotation curves of Einstein, A, 59-60, 72, 110, 148, 150
Davies, Paul, 5, 114 and cosmological constant, 199
estimate of relaxation time for cosmos, 195 and the Miclielson-Morley experiment, 50
deceleration parameter, 199 controversy with Ritz, 126
definition of time first paper, 59
a linguistic matter, 15 on atomism, 49
250
on Besso's death, 71
on the discovery of relativity, 59
on Miller's experiment, 56
Einstein-de Sitter cosmological mode~ 130
Einstein-Podolsky-Rosen paradox, 103, 150
and 'clumsy fingers', 149
assumed locality, 152
Bohm's formulation, 148-149
electromagnetic arrow of time, 191
defmed,107
electromagnetic mass, 110
Eliezer, CJ., 114
Eliot, T.S., 16
ellipticity
of Laplace and Poisson equation, 104, 160
EI'sgol'ts, L.E., 122
entropy, 190
and equivalence of transformations, 98
and H-function, 100
as extension of thermodynamic entropy, 84
as used by Maxwell, 98
as used by Tait, 98
Boltzmann, 84
coarse-grained, 85
coarse-grained and appt:oach to
equilibrium, 86
etymolo~, 98
fme-gramed, 85
meaning of, 82-83, 218
spontaneous decrease of, 210
entropy law, 6, 92, 102, 216, 222
and cycles of the closed Friedmann mode~ 209
classical statement, 81
distinguished from mundane time
asymmetry, 217
distinguishes between a decided past and
decided future, 218
motivationJor H-theorem, 83
regarded as tautology by Boltzmann, 101
related to expectation and memory, 83
restated, 83
stated, 81
epistemically broken time, 4, 217, 226-227
mystery-novel analogy, 47, 49, 225
see also argument from chaos, argument from
complexity
equations of motion, see evolutionary law
equivalence principle, 41
result of a historical accident? 192
ergodic hypothesis, 92, 218
classical, 90
in terms of averages, 93
ergodic theorem
individual, 94
mean, 94
ergodicity, 190
Boltzmann on its origin, 101
for a finite Markov chain, 90
for a Markov chain, 86
see also metric transitivity, mixing
quasi-, 90
INDEX
Everett, H. and JA. Wheeler, 183
evolutionary law
and smoothness of nature, 204
equilibrium and unitary evolution in Hilbert
space,183
given by functional differential equations, 190
given by o.d.e.'s, 190
~ven by p.d.e.'s, 190
if humans are exempt, what is the domain of
validity? 222
Markovian, 190
stochastic, 190, 225-226
with a tilt in the arrow of time, 135, 174
with retarded action, 117, 134-135
falsifiabili
and ae~er, 65
and Laplace's demon, 226
assumes mundane time, 2
empirical, 2
explained, 39
logical, 2
temporal assumptions for logical contradict
those for empirical, 227
field
and incompressible aether, 104
equivalence of retarded and advanced
representations, 126
divergences of, 109
derived from Lienard-Wiechert potentials,
109,134
infinite number of degrees of freedom
reinterpreted as history-dependence, 122, 190
indefinitely extended contact, 159
is it real? 102
mechanical model of, 104
operates on 'contact' like aether, 104
relativistically covariant, unlike aether, 104
rigidity (ellipticity)of Coulomb field, 104
FIZeau's expenment, 53, 58
flow of time
as illusory, 20
see also A-series
FLRW models, 130, 197
and the real universe, 205
critical density for recontraction, 200
empirical determination of curvature
parameter, 200
foliation of space time assumed, 199
identification of observable quantities, 199
force
centrifu~al, 35-36
difficulties with the notion, 103
what is it? 35
when is an external force absent? 37
Foucault pendulum, 207
and rotation of the earth, 207
and rotation of the universe, 207
'free will', 3, 22, 24,148,220,223
Freedman, S. and E.P. Wigner,155
 INDEX 251

Fresnel, A. H-theorem
difficulties in his theory, 54 assumptions in Boltzmann's proof, 99
drag coefficient, 53 consequence of shuffling, 90
refraction in moving bodies, 52 informal statement, 83
theory of aberration, 52 use of Newton's laws questioned, 85
future Haa~R, 114
absolute, 71 Hablcht, K., 72
as existing, 80 Hadamard, J., 46
as logically dependent on the present, 142 Halley, Edmund, 193
as non-existent, 12, 14 Hamilton's equations, 90, 93
connection with past and the continuity of Hamiltonian
time, 223 classical, 93
logically different from the past, 21 classical and quantum, 173
mundane future not quite unknown, 81 Harrison, E.R, 195
not qualitatively different from past, 81 Hausdorff space, 97
future absorber, 129 Hawking, S.W. and G.P.R Ellis, 22
distributional solutions of Hilbert-Einstein
equations, 204
motivation for chronology condition, 220
r -space, see phase space Hawking-Penrose theorem, 89, %, 203, 205
galaxies, rotation curves of, 205 Heaviside, Oliver
flatness, 206 jump conditions, 55
independent of galaxy morphology, 206 Heaviside-Lorentz law, 61, 62, 109, 135
galaxy formation Hegerfeldt, G.C., 157
in big bang cosmologies, 210 Heisenberg uncertainty principle, 148
Gale, RM., 14, 29 and Schwartz inequallty, 163
Galileo Popper's analogue in classical mechanics, 46
neutral motions, 34 Heron, M.L. and D.T. Pegg
parabolic trajectories, 42 proposed experiment, 131
rectilinear inertia, 34 absence of plcosecond switch, 132
representation of time by a line, 26 Herschel, William, 193
Gamow, George, 201 Hesse, Mary, 103
Garg, A and D. Mermin, 155 hidden-variable theory
Gauss, c.P., 111 and 'clumsy fingers', 147
curvature, 198 example of a local Bell-non-Iocal theory, 155
Gautam,12 is to q.m. as classical mechanics is to
geodesics thermodynamics, 149
and characteristics, 204 non-contextual, 150
in Godel's universe, 207 non-locality of q.m. as justification for Bell
geodesic hypothesis, 192 non-locality of, 154
Hilbert, David, 72
geodesic triangle, 200
Hilbert-Einstein equations, 160, 1%, 198-199,
Ghazili, 223 202
Gibbs, W., 91 motivated by locality, 204
Gleason's theorem, 164 Hilbert-space axioms for q.m., 161-163
and context dependence, 150 arising from change of logic, 161
stated,150 history dependence
Gleason, AM., 154 entropy law and time asymmetry of, 85
GOOel cosmological model, 207, 210, 220 incorrectly modeled by instantaneous system
earlier-later relationship in, 208 with additional degrees of freedom, 122
not externally consistent, 208 impossibility of retrodiction with, 120-121
grandfather paradox, 6, 125, 208 of biological evolution, 230
resolution through hard determinism, 125 see also retarded o.d.e.
stated,108 Hoare, C.AR., 175
Green, G., 55 Hoek,M.,53
GRT Hogarth, J.E., 191
most satisfactory, despite drawbacks, 192 conditions for consistency of retarded
not concerned with structure of time or radiation, 129
matter, 192 homeomorphism, 89, 90, 97
Grtinbaum, A, 20, 29 from Hamiltonian evolution, 93
Gudder, S., 168, 172 homogeneity
 252 INDEX

effect on matter tensor, 199 isotropy

how does one test it? 201 of 3-space, 198
of 3-space, 198 of space, 63
ofspace,63 effect on matter tensor, 199
Hooke, Robert, 37, 51 effects of anisotropy, 210
Hoyle, F., and J.V. Narlikar, 6, 36,191
external inconsistency of their absorber Jauch, J.M., and C. Piron, 171-172
theory, 129
see also Hogarth, absorber theory, steady- Piron's representation theorem, 150, 171
state theory irreducible CROC, 172
Hubble, E.,
law, 196 K-meson decay, 209
constant and its variation with time, 196, 197 Kac, Marc, 95
and age of the universe, 200 Kandda,16
Hume, David, 103, 105 Kaflt, Immanuel, 105
hyperbolic p.d.e. Kaufmann, W., 56
intersection of characteristics of, 204-205 Kepler, 63, 159
hyperbolici!}' Omar Khayyam, 12
and locality, 160 Kochen, S. and E. Specker' theorem, 150
of evolutionary p.d.e.'s of physics, 160 Kuhn, Thomas, 40
of wave equation, 160
shocks as an intrinsic feature of, 204 Laplace, 44-45
'calculation of black-hole radius', 202
implication demon, 226
and order relation in a poset, 166 equation, 104, 160
counterfactual, 179 transform, 114
in 2-valued logic, 23 Laplace's demon, 4, 45
in 3-valued lo~c, 24 exorcism of, 46
possible defImtion in q.t.f.logic, 25 assumptions underlying exorcism, 47
strong, 179 Napoleon's remark, 45
strong and weak, 179 overpowers epistemic contingents, 226
indecomposability Larmor formula, 106, 108
see metric transitivity lattice
indeterminism compatibility, distributivity and
mere indeterminism irrelevant to mundane orthomodularity, 170
time, 3 compatibility and commutativity, 169-170
and mixed-type o.d.e., 124 Hilbert, 172
contrasted With unpredictability, 4, 223 objections to, 172, 176, 181
equated with unpredictability, 225-226 of projections on a Hilbert space, 165
consequences of this equation, 227 of subspaces of 3-space, 166
see also quantum indeterminism order relation and Implication, 166
inertial frame, 36 orthogonality and disJointness, 167
inflation, 199,206 laws of motion
instant as a self-contained ehysical theory, 5
as a ~eometrical point, 12 compared with Galilean physics, 42
as imtial conditions, 43-44 fail as defInitions, 39, 41
as a plane of simultaneity, 71 falsifIable when combined with law of
as a spacelike hypersurface in a foliated gravitation, 41-42
spacetime, 198 frame-independence of force, 64
as a state, 177 Galilean invariance, 63
see also possible world semantics impossibility of determining initial
structure of, 140-141 conditions, 46
q.tJ. logic and structure of, 142 instantaneous formulation, 34, 44
instrumentalism, 148 not falsifIable, 40
intention, 221 only relative velocities meaningful, 64
as future data, 229 practical scheme for measuring time, 42
see also mundane intention stated,34
invariant function, 94 stated as defInitions, 38
invariant subspace, 94 straight line motion as 'natural', 63
involution, 172 velocity-dependent forces, 62
isometry, 97, 198 Lax, P.D., 204
 INDEX 253

Leb~e integral, 162 see also Abraham-Lorentz

Leibmz, G., 103 force law, see Heaviside-Lorentz
on indivisibility, 105 purely electromagnetic electron, 62
Lienard-Wiechert potential, 133 rejection of Stokes' theory, 60
see also retarded, advanced solutions theory of the electron, 62
Liouville's theorem, 90 transform, 61, 71
classical version, 85, 88 Lorentz-Dirac equation, 112, 122
hypotheses interpreted, 89 alternatives to, 122
hypotheses enumerated, 89, 96 Loschmidt, J., 87
proof,97 Lukasiewicz, Jan, 23
stated, 89 Lutzen, J., 203
loc::1!ts diverse, unconnected formulations, 146 ,u-space, 84
aether failed but Newton succeeded, 159 Mach, Ernst, 36
and Bell's inequalities, 6 Mackey, G.W., 150
and hyperbolicity, 160 Maggie, William, .57 .
and hyperbolicity of Hilbert-Einstein many-body equations of motion
equations, 204 see evolutionary law
and 10000000bility and contact, 146 many-worlds interpretation of quantum-
and propagated action, 152 mechanics, 182
and the contrapositive of contact, 159 relation to formalism of q.m., 183
as a fuzzy and metaphysical demand, 146 Markov chain, 86, 92
as opposite of voodoo principle, 150 ergodicity,90
assumes localizability, 157 Markov-Kakutani theorem, 97
Bell's notion of, 7 Markovian evolution, 91, 225
see also Bell locality mass
contradiction between relativity and aether difficulty of definition in a non-asymptotically
shows fuzziness of, 159 flat spacetime, 199
does classical physics support it? 146 Master Argument
does it admit a uni<j.ue formulation? 146 resolution with q.t.f.logic, 24
see also hyperboliClty stated, 22
in classical mechanics, 7 matter tensor
inconsistent with localizability, 157 absence of a statistical continuum model in
irrelevant thermodynamic considerations in GRT, 192, 204
macrophysical intuition of, 146 can it be assumed to be smooth at a
irrelevant to witchcraft, 158 sinItuIaritY? 203
is the intuitive notion reliable? 146 MaxweI("J.e.! 54, 106
see also non-locality and reflection and refraction, 55
thermodynamic, 154 on entropy, 98
logic on FresnCl and Stokes theories, 55
cannot alone decide nature of physical equations, 61, 160
reality,23 Maxwell's demon, 90-91, 99
change required by mundane time and and sorting and mixing, 230
q.m., 142 exorcism of, 91
Lemmon's 'minimal' tense logic, 141 Maxwell-Boltzmann distribution, 99-100
Lukasiewicz's three-valued, 23 Mayer,Julius, 99
modal,l86 McTaggart, J.M.E., 18-19
quantum, 168 A-tIleorist answer to paradox, 24
quasi truth-functional, 24 B-series resolution of paradox, 20
temporal, 1, 175 B-serlesview of time, 16
time measurement, simultaneity, and need for biographical sketch, 17
change oflogic, 142 paradox, 18-19
truth-table for two-valued logic, 23 measure
two-valued, 22 conserved, 93
Lorentz, H. A., 60 . measurement of time, 139
calculation of self action, 109 related to simultaneity, 69, 79, 190
contraction of rods due to increase of inter- and velocity dependent forces, 79
molecullir forces, 60 as basis of an measurement, 6, 79
contraction of rods due to shrinking difficulty in Newtonian mechanics, 64, 79
electrons, 61 impossibility of, 14
electron as extended body, 61 inq.m., 143
254
is it possible in relativity? 208
length-measurement requires simultaneity, 70
not possible when cosmos is in
thermodynamic equilibrium, 92
measurement postulate,
stated for an observable with discrete
spectrum, 164
assumes mundane time, 3,143
for an observable with continuous
spectrum, 164
paradigmatic, 164
measurement process, 176
Mehra, Jagdisb, 59
metric transitivity, 90, 94
characterized by invariant functions, 94
dermed, 90
Koopmans' operator, 94
difficulty in physical characterization, 92
Michelson, Alliert and E. Morley
confirmation of Fresnel drag, 53
Michelson-Morley experiment, 5, 71, 74, 79
conclusion of, 56
difficulties encountered, 56
moral of, 57
objective of, 55
prmciple, 55
Miller, D.C.
observation of aether drift, 56
Milne, E., 39
mixed potentials, 107
mixed-type o.d.e.'s
non-existence of solutions of past and future
value problem, 124
mixed-type p.d.e., 160
distinguished from p.d.e.'s with mixed
deviating arguments, 160
see also tilt in the arrow of time
mixing, 90, 190 '
as a time-asymmetric assumption, 6
contrasted with sorting, 91
defined, 91
proof that it implies metric transitivity, 95
related to metnc transitivity, 91
related to quasi-ergodicity, 90
modal interpretation of q.m., 183, 186
modalities, 22-25
q.tJ., In, 184
molecular chaos, 84, 100
motion
beliefs about natural state of, 33
mundane intention
its effect on the future, 221, 229
assumes deterministic or stochastic
connections between past and future, 229
cannot be entirely decided by past, 229
competition with choice, 229
mundane time, 7, 144, 190
and falsifiability, 2, 140, 221
and global and microphysical structure of
time, 221
decided past and undecided future incompati-
INDEX
ble with evolutionary law, 218
explained, 1, 139-40,216-217
is It a hallucination? 221
is it relevant to physics? 2
mundane future as ontically con~ent, 217
mundane past as epistemically contmgent, 217
mundane past believed to be qualitatively
different from future, 217
need for consistency with time in physics, 219
no conflict with physics, 3
possible contradiction with other
experience, 33
suggests change of logic for physics, 142
see also twotime theory
mundane time, paradox of, 2-3, 222
escape routes enumerated, 219
resolution by equating ontic and
epistemic contingents, 226
Nagaoka, M., 74
Napoleon, 45
Narlikar, J.V.
see Hoyle, F.
Navier-Stokes equations, 160
Neumann problem, 54
von Neumann, J., 150-151, 164
no hidden-variables, 149
Newton, Isaac
and action at a distance, 103
and Descartes, 37
and Galileo, 34
back calculated law of gravitation, 37
bucket,35
see also laws of motion
on absolute time, 39
physical laws and humans, 222
static universe must be infinite, 193
transition from cyclic to superlinear time, 44
Newtonian paradigln, 43-44, 118
non-locality
as a tag, 7, 158
force of locality requirement comes from
macrophysical intuition, 145
irrelevant to voodoo principle, 150
of classical mechanics, 159
OCCAM,175
ALT construct, its formal indeterminism and
classical random variables, 175
Occam's razor, 65, 83
occasionalism, 3
Olbers' paradox, 193
Bondi's expansion solution, 194
Cheseaux's absorption solution and
tlawin,l93
Halley's form of, 193
Poe on, 195
redshift solution in a big bang cosmology, 194
significance of disequilibrium, 194
 INDEX 255

where are the missing stars? 193 response to Miller's result, 56

where is the missing starlight? 193 Plass, G.N., 114
why are stars visible? 193 transformation, 114
Olbers, Henrich Wilhelm, 193 Poe, Edgar Allen, 195
ontically broken time, 3, 217, 223, '226-227 Poincar~, Henri, 5, 60, 72, 135
operationalism, 13 analysis of time did not go far enough, 173
order relation and implication, 179 on accwnulation of hypotheses, 73
orthocomplement, 166 ad hoc cohesive forces, 110
aether and the form of the equations of
Pais, Abraham, 56, 58, 72 evolution, 65, 105, 116
'Poincar~ needed a third hypothesis', 73-74 aether as substratwn for absolute motion, 65
'Poincar~ never understood relativity', 75 analysis of classical mechanics, 64
'Whittaker lacked J.>~~ insight', 73 attractiveness of Lorentz's theory, 67
'Born did not acqwt elf well', 73 balancing action and reaction, 111
overlooked 'fourth hypothesis', 73 classical mechanics as an approximation to
Panini, 14,83 relativity, 74
paradoxes: clocks deCided by principle of simplicity, 69
see advanced action, Dwnett, EPR, grandfa- democracy of clocks, 38
ther, Master Argwnent, McTaggart, mundane formulation of principle of relativity, 64
time, Popper's pond, present:ever-shrinking, increase of inertia with velocity, 68
recurrence, reversibility, Schrodinger's cat, local time and clock synchronization, 69
sea battle, Thomson lamp, Wheeler and local time as physical, 69
Feynman, Zeno Lorentz's theory contradicts physical
parallel computing principles, 67-68
formal semantics of, 174 mechanical models redundant, 104
partial differential equations moving observers would use local time, 68
hyperbolic, 125 on the nature of the evolutionary law, 74
evolutionary equations asswne real-line recurrence theorem, 87,210
topology for time, 139 redundancy of aether as a means of
Partridge's experiment, 131, 210 mechamcal explanations, 66
Pegg's proposal for a dynamic absorber, 131 rejection of absolute motion, 66-67
past the new mechanics, 68
absolute, 71 time-measurement a matter of definition, 69
as existing, 80 time-measurement related to simultaneity, 69
as logically dependent on the present, 142 two me~ of 'aether', 65
as non-existent, 12, 14 velocity of light as an absolute limit, 68, 74
connection with future and continuity of Poisson, 45
time, 223 bracket, 93
contingent,I26,175 equation,I04
not qualitatively different from future, 81 see also wave equation
not quite known, 81 pond paradox
past absorber, 129 resolution through time asymmetry of
past-value problem advanced o.d.e., 122
existence theorem for retarded o.d.e., 119 stated, 108
Pataiijal~ 12 Pool, J.C.T., 181, 186
crow paradox, 27 Popper, Karl, 2, 4, 39, 65, 226
Pegg, D.T., 131 account of Laplace's demon, 45
Penrose, o. and I.C. Percival, 154 and Kuhn, 40
Penrose, R., 203 biographical sketch, 40
Penzias, AA. and R.W. Wilson, 201 exorcism of Laplace's demon, 46-47
phasespace,85,89,149 see also falsifiability
phase space: intersection oftrajectories in, 119 no dist~ction between epistemic and ontic
not possible with Hamiltonian evolution, 89 contmgents, '226
future collapse with retarded o.d.e., 121 pond paradox, 108, 124, 228
future branching with advanced o.d.e., 123 pond paradox resolution, 122
photon record postulate, 220
mean free path of, 129 thermodynamic arrow of time as basic, 209
Piron's theorem, 172 possible-world semantics, 140
see also Jauch-Piron physical existence of possible worlds, 142
Planck, Max, 72 possible world as an lUStant of time, 141-142
compressible aether, 54 preacceleration, 113
 256 INDEX

and systematic propagation of information to and active galactic nuclei, 197

the past, 230 Arpon, 197
effect of order of Taylor-series expansion, 122
Eliezer's objection to Dirac's procedure, 114 radiative dampin&, 110
invalidity of Taylor series expansion used to Dirac's denvatlon, 112
derive Schott term, 121 necessity of advanced potentials, 6, 111-112,
present 122
as contin&ent, 175 rejection of higher order terms, 112
as non-eXIstent, 12 random variable
as punctal, 12 as a measurable function, 167
ever-shrinking, 12 as a projection valued-measure, 167
specious, 12 as an isomorphism between a-algebras, 167
see also instant Raychaudhuri, A.K., 35
Priestley, J.B. Rayleigh
on McTaggart, 17 experiment to detect absolute motion, 61
on mundane time, 215 realism and hidden variable theories
'principle of causality' see structure of time
and physics texts, 5 recurrence paradox, 6, 89,190
principle of micro-reversibility, 84, 99-100 Boltzmann's reaction, 101
Prior, A.N., 20 field as escape route, 6, 96, 102
probability hypotheses enumerated, 96
on a a-algebra, 165 in long-lived cosmos, 95
on a logic of sentences, 165 refutation in Ehrenfest model, 91
probability interpretation, 162 refutation questioned, 92
and operator representation of dynamical resolved by history-dependence, 119
vanables, 167 singularities as escape route, 96
projection postulate stochastic evolution as escape route, 96
see measurement postulate unboundedness as escape route, 96
projection-valued measure recurrence theorem
see spectral measure hyPotheses interpreted, 89
sunplified proof, 88
quantum gravity statement and proof, 88
absence of internally consistent recurrence time, 102
formulation, 192 assumption of metric transitivity needed for
see also grandfather paradox calculation, 92
quantum indeterminism calculation questioned, 92
and occasionalism, 224 what does it mean? 92
and the chocolate-icecream machine, 223 redshift
begs the question, 224 and apparent luminosity, 197
does not imply mundane time, 3 consequence of cosmological expansion, 196
if it applies to macrophysics what happens to quasars, 197
measurement? 224 refutabilitY.
measurement assumes mundane time, 224 see falsifiability
see also measurement postulate Reichenbach, H., 24
requires an act of measurement, 224 relativity, special
transition probabilities immune to origin in time-measurement, 5
intervention, 223 see also B-series, Einstein, GRT, Lorentz,
quantum logic, 168, 172 Poincare
why at the microphysical level alone? 173 Respighi, L., 53
relation to Hilbert space, 166 retardation vector, 134
quantum probabilities retarded action
non-eXIStence of joint distributions, 165 equations of motion with, 117
and failure of the distributive law, 165 see also evolutionarr.law
and single systems, 151 retrodiction impOSSible, 123
quantum-mechanical arrow of time, 209 retarded o.d.e.
quasi truth-functional logic see also differential equations
and quantum logic, 175, 180 discontinuities in derivatives of solution of, 122
evades Master Argument, 24 invalidity of Taylor-series approximation in
resolution of sea battle, 24 powers of delay, 122
semantic interpretation, 24 numerical solution of, 122
quasars, 197 retarded proper time, 117
 INDEX 257

retarded solutions distinguished from a beginning oftime, 203-

physical characterization, 107 204
reversibility paradox, 6, 87 smoothness of the matter tensor, 203-204
Boltzmann on, 101 De Sitter, W., 196
forces subsidiary assumptions, 87 SO(3),198
refutation in Ehrenfest mode~ 90 Solovine, M., 72
resolved through time asymmetry of retarded Sommerfeld radiation condition, 126-127
o.d.e,l20 sorting, 6, 91, 190, 230
reversible process, 84 spectral measure, 162-163
Riemannian curvature, 200 spectral theorem, 162
Riesz representation theorem, 94 spectral multiplicity theorem, 163
Ritz, W., 126 spectrum, 162, 164
Rohrlich, F., 114 speed of light
rotation of cosmos, 208 measured using Jupiter's moons, 51
and galactic rotation, 210 postulated constancy of, SO, 79
and galaxy formation, 210 SQUIDs, 152
see also CMBR (quadrupole anisotropy), Sri Harsa, 15, 18, 29
Godel cosmological model biographical sketch, 19
Russell, Bertrand standard candles, 196
B-series as basic, 20 state
defmition of change, 18 as density matrix, 164
onZeno,27 as element of Hilbert space, 162
resemblance to Teniel's caricature, 17 stationary
Markov chain, 86
Sachs, R.K. and A. M. Wolfe, 210 stochastic process and Schrodinger
analysis not gauge invariant, 201 equation, 183
Schott term, 112 steady-state cosmological mode~ 130
numerical robustness of, 122 Stern-Gerlach experiment, 147
origin in an invalid Taylor expansion, 121-122 Stokes, G. G.
Schrodinger, E., 148 difficulty with his theory, 54
Schrodinger equation, 143-144 theory of aberration, 54
and equilibnum, 183 Stone's theorem, 163
in Hilbert space, 163 structure oftime, 1, 4, 6, 142
SchrOdinger's cat, 143, 145, 147, 151, 224 and definiteness of reality in hidden variable
and SQUIDs, 152 theories, 145
Schulman, L.S., 125 and quantum logic, 4, 162
Schur's (lemma) theorem, 198 cyclic, 144
Schwarzschild solution, 202 defined, 139
Schwarzschild-Tetrode-Fokker action, 111, formalization in temporal logic, 141, 161
126-127 formalization in the U-calculus, 140, 161
Sciama, D. W., 191 implied by a tilt in the arrow of time, 161
sea battle, 1, 20, 217 possible difference between classical and
see also Aristotle quantum mechanics, 142
resolution in three-valued system, 23 related to change of logic and q.m., 161-162
selection function, 178, 182, 184 related to topology, 140
sentence calculus used to formalize A-series, 216
axioms for, 166 structured time interpretation of q.m., 161
Boolean algebra, 165 access relation, 178
truth table lor connectives, 23 admissible worlds in, 182
separable topological space, 97,162 avoidance of objections to lattice
serialism, 19 structure, 176
Shankland, R.S., 58 branching of time, and ambiguous reality, 174
shufflin~ 84 branching time and semantics of CSP, 175
of pomts in r -space, 85 closest world and selection function, 182
of points in }l-space, 84 condition for lattice structure, 181-182
related to ergodicity, 90 definition of order relation, 180
Siddhanta Muktavall, 16 derivation of a quantum logic from structured
singularity, 96, 203 time,l80
and intersection of characteristics, 204 failure of the distributive law, 181
as beginning of time, 203 incompatibility, joint and repeated
assumptions, 203, 220 measurements, 178
258
interpretation of measurement postulate, 177-
178
lattice as a technical convenience, 181
like relativity, can q.m. be related to an intrin-
sic shortcoming of classical mechanics? 173
measurable and quasi-measurable
propositions, 178
non-locality of distinguished from Bell non-
locality, 145
order relation and strong implication for
measurable propositions, 179
path-integral analogy, 174
propositions distinguished from those of
Birkhoff-von Neumann and Jauch-Piron
theories, 176
sentences and propositions in, 176
states and probability measures, 177
technical difficulties in, 180
Stuckelberg-Feynman, 209
superlinear time, 4
defined, 2, 142
how did it evolve? 44
Synge,J.L.
on measurement using proper clocks, 70
on Miller's experiment, 56-57
T-symmetry
see time symmetry
T-violation,209
Tait, P.G., 98-99
teleology
and God's purpose, 228
see also advanced action
thermodynamic arrow of time
believed independent of cosmology, 6
defined, 83
difficulties with, 218
effect of fluctuations, 95
not fundamental, 6
relation to other arrows, 209
used as fundamental, 128
thermodynamic equilibrium, 82-84
effect on measurement of time, 93
thermodynamic laws
ClaUSIUS' couplet on, 98
see also entropy law
Thomson lamp paradox, 28, 204
tilt in the arrow oftime, 4, 7, 108, 160, 174, 191
and transfer of macrophysical information to
the past, 230
as mirumal change, 227
claims summarized, 174
evolutionary law with, 174
~eneric non-uniqueness of solutions, 174
unplies non-locality, 144
implies quantum logic, ISO
im{llies structure of time, 161
is tIme asymmetry conceptually adequate? 211
prima facie consistency with mundane
time,228
recapitulation, 143-144,227
INDEX
resulting non-locality distinguished from Bell
non-Iocality,145
time
as the real line, 2, 44,139-140,190,218
beginning of, 192
end of, 192
fission-fusion, 140
parallel time streams, 140
time asymmetry
and binary earlier-later relationship, 211
arrow simile inadequate for mundane
time,218
is a binary earlier-later relationship
Idoballyadequate? 7, 142, 192, 206
different tyPes of earlier-later relations, 140
does phySICS ignore the empirically
obvious? 215
probabilistic and intermediate, 91
local, in Godel universe, 208
logical,21
matter-anti-matter asymmetry as, 143
physical, SO, 102
SlDall departures from, see tilt in the arrow of
time
see also U-calculus
time-s~etry
explained, 80
is It reasonable? SO
see also CPT theorem
Tipler, F. J., 96, 205
Tolman, R.C., 209
topology of time
ldobaI,l92
focally like that of the real line, 139
transactional interpretation of q.m., 175, 183
transitive
group action, 198
order relation, 16, 141, 142
trapezoidal rule, 122
Trouton, F. and H.R. Noble
experiment to detect absolute motion, 61
two-slit diffraction experiment, 146, 151
wave-theory explanation, 146
with 'single' electrons, 147
two-time theory, 219
chronology condition as illustration, 220
Dumett's paradox as illustration, 220
how does It operate in practice? 219
Popper's pond as illustration, 220
Popper's record postulate as illustration, 220
used to exclude novel views of time, 220-221
U-calculus, 140
two-valued logic inadequate, 142
density and completeness, 142
existence of instants of time questioned, 141
semantic interpretation of F, P, G, H, 141
U-relation inadequate, 142
U-relation as an earlier-later relation, 140
Udyotkara, 105
universe, logical, 177, 184
 INDEX 259

Van Dam, H. and E.P. Wigner, 119 quantum probabilities, 165

varying G, 42 reformulation of Bell's inequalities, 155
Vatsyayana, 18 time-symmetry reduced to Newtonian
velocity distance relationship, 196 paradigm, 119
reinterpreted, 197 Wirtz, C., 196
Voltaire, 37 Wittgenstein, Ludwig, 15, 176, 192
Vyasa, 14, 26 world
'closest',I77-178
admissible, 177, 182
Waismann, Friedrich, 15 impossible, 177
wave equation, 106, 160 see also possible-world semantics
Weyl, Hermann, 20 q.t.f., 177
Wheeler-Feynman absorber theory, 6, 209 q.tJ. and microphysical reality, 177
see also absorber theory true, 177
assumption of time-symmetry questioned, 129
begs the question, 128
explained, 127-128 Young, Thomas, 52
-Davies, 191 explanation of stellar aberration, 52
preacceleration, 230
Wheeler-Feynman paradox
resolution through hard determinism, 125 Zeno's paradoxes of motion
resolution using mixed o.d.e., 126 Achilles,27
stated,108 apply to continuous and discrete time, 26
their resolution, 125 chocolate paradox, 26
Whittaker, E.T., 72-73, 75, 210 dichotomy, 28
Wiener, Norbert, 17 not resolved by the continuum, 26
Wigner, E. P. Pataiijali's crow paradox, 27
friend,l48 stadium, 27
psycho-physical parallelism, 151,224 Zermelo, E., 87
 Fundamental Theories of Physics
64. M. Evans and J.-P. Vigier: The Enigmatic Photon. Volume 1: The Field B(3). 1994
ISBN 0-7923-3049-8
65. C.K. Raju: Time: Towards a Constistent Theory. 1994 ISBN 0-7923-3103-6

KLUWER ACADEMIC PUBLISHERS - DORDRECHT I BOSTON I LONDON