Q U A N T I Z E D D E T E C T O R N E T WO R K S

Scientists have been debating the meaning of quantum mechanics for over a
century. This book for graduate students and researchers gets to the root of
the problem: the contextual nature of empirical truth, the laws of observation,
and how these impact on our understanding of quantum physics. Bridging the
gap between nonrelativistic quantum mechanics and quantum field theory, this
novel approach to quantum mechanics extends the standard formalism to cover
the observer and their apparatus. The author demystifies some of the aspects
of quantum mechanics that have traditionally been regarded as extraordinary,
such as wave–particle duality and quantum superposition, by emphasizing the
scientific principles. Including key experiments and worked examples throughout
to encourage the reader to focus on empirically sound concepts, this book avoids
metaphysical speculation and alerts the reader to the use of computer algebra to
explore quantum experiments of virtually limitless complexity.

g e o rg e j a ro s z k i e w i c z is a mathematical physicist recently retired from

the School of Mathematical Sciences, University of Nottingham. His research
interests are focused on the fundamental differences between quantum and
classical mechanics.
 CAMBRIDGE MONOGRAPHS ON MATHEMATICAL PHYSICS

General Editors: P. V. Landshoff, D. R. Nelson, S. Weinberg

S. J. Aarseth Gravitational N-Body Simulations: Tools and Algorithms†

J. Ambjørn, B. Durhuus and T. Jonsson Quantum Geometry: A Statistical Field Theory
Approach†
A. M. Anile Relativistic Fluids and Magneto-fluids: With Applications in Astrophysics and
Plasma Physics
J. A. de Azcárraga and J. M. Izquierdo Lie Groups, Lie Algebras, Cohomology and Some
Applications in Physics†
O. Babelon, D. Bernard and M. Talon Introduction to Classical Integrable Systems†
F. Bastianelli and P. van Nieuwenhuizen Path Integrals and Anomalies in Curved Space†
D. Baumann and L. McAllister Inflation and String Theory
V. Belinski and M. Henneaux The Cosmological Singularity†
V. Belinski and E. Verdaguer Gravitational Solitons†
J. Bernstein Kinetic Theory in the Expanding Universe†
G. F. Bertsch and R. A. Broglia Oscillations in Finite Quantum Systems†
N. D. Birrell and P. C. W. Davies Quantum Fields in Curved Space†
K. Bolejko, A. Krasiński, C. Hellaby and M.-N. Célérier Structures in the Universe by Exact
Methods: Formation, Evolution, Interactions
D. M. Brink Semi-Classical Methods for Nucleus-Nucleus Scattering†
M. Burgess Classical Covariant Fields†
E. A. Calzetta and B.-L. B. Hu Nonequilibrium Quantum Field Theory
S. Carlip Quantum Gravity in 2+1 Dimensions†
P. Cartier and C. DeWitt-Morette Functional Integration: Action and Symmetries†
J. C. Collins Renormalization: An Introduction to Renormalization, the Renormalization Group
and the Operator-Product Expansion†
P. D. B. Collins An Introduction to Regge Theory and High Energy Physics†
M. Creutz Quarks, Gluons and Lattices†
P. D. D’Eath Supersymmetric Quantum Cosmology†
J. Dereziński and C. Gérard Mathematics of Quantization and Quantum Fields
F. de Felice and D. Bini Classical Measurements in Curved Space-Times
F. de Felice and C. J. S Clarke Relativity on Curved Manifolds†
B. DeWitt Supermanifolds, 2nd edition†
P. G. O. Freund Introduction to Supersymmetry†
F. G. Friedlander The Wave Equation on a Curved Space-Time†
J. L. Friedman and N. Stergioulas Rotating Relativistic Stars
Y. Frishman and J. Sonnenschein Non-Perturbative Field Theory: From Two-Dimensional
Conformal Field Theory to QCD in Four Dimensions
J. A. Fuchs Affine Lie Algebras and Quantum Groups: An Introduction, with Applications in
Conformal Field Theory†
J. Fuchs and C. Schweigert Symmetries, Lie Algebras and Representations: A Graduate Course
for Physicists†
Y. Fujii and K. Maeda The Scalar-Tensor Theory of Gravitation†
J. A. H. Futterman, F. A. Handler, R. A. Matzner Scattering from Black Holes†
A. S. Galperin, E. A. Ivanov, V. I. Ogievetsky and E. S. Sokatchev Harmonic Superspace†
R. Gambini and J. Pullin Loops, Knots, Gauge Theories and Quantum Gravity†
T. Gannon Moonshine beyond the Monster: The Bridge Connecting Algebra, Modular Forms
and Physics†
A. Garcı́a-Dı́az Exact Solutions in Three-Dimensional Gravity
M. Göckeler and T. Schücker Differential Geometry, Gauge Theories, and Gravity†
C. Gómez, M. Ruiz-Altaba and G. Sierra Quantum Groups in Two-Dimensional Physics†
M. B. Green, J. H. Schwarz and E. Witten Superstring Theory Volume 1: Introduction
M. B. Green, J. H. Schwarz and E. Witten Superstring Theory Volume 2: Loop Amplitudes,
Anomalies and Phenomenology
V. N. Gribov The Theory of Complex Angular Momenta: Gribov Lectures on Theoretical Physics†
J. B. Griffiths and J. Podolský Exact Space-Times in Einstein’s General Relativity†
S. W. Hawking and G. F. R. Ellis The Large Scale Structure of Space-Time†
F. Iachello and A. Arima The Interacting Boson Model†
F. Iachello and P. van Isacker The Interacting Boson-Fermion Model†
C. Itzykson and J. M. Drouffe Statistical Field Theory Volume 1: From Brownian Motion to
Renormalization and Lattice Gauge Theory†
C. Itzykson and J. M. Drouffe Statistical Field Theory Volume 2: Strong Coupling, Monte Carlo
Methods, Conformal Field Theory and Random Systems†
G. Jaroszkiewicz Principles of Discrete Time Mechanics
G. Jaroszkiewicz Quantized Detector Networks
C. V. Johnson D-Branes†
P. S. Joshi Gravitational Collapse and Spacetime Singularities†
J. I. Kapusta and C. Gale Finite-Temperature Field Theory: Principles and Applications,
2nd edition†
V. E. Korepin, N. M. Bogoliubov and A. G. Izergin Quantum Inverse Scattering Method and
Correlation Functions†
J. Kroon Conformal Methods in General Relativity
M. Le Bellac Thermal Field Theory†
Y. Makeenko Methods of Contemporary Gauge Theory†
S. Mallik and S. Sarkar Hadrons at Finite Temperature
N. Manton and P. Sutcliffe Topological Solitons†
N. H. March Liquid Metals: Concepts and Theory†
I. Montvay and G. Münster Quantum Fields on a Lattice†
P. Nath Supersymmetry, Supergravity, and Unification
L. O’Raifeartaigh Group Structure of Gauge Theories†
T. Ortı́n Gravity and Strings, 2nd edition
A. M. Ozorio de Almeida Hamiltonian Systems: Chaos and Quantization†
M. Paranjape The Theory and Applications of Instanton Calculations
L. Parker and D. Toms Quantum Field Theory in Curved Spacetime: Quantized Fields and
Gravity
R. Penrose and W. Rindler Spinors and Space-Time Volume 1: Two-Spinor Calculus and
Relativistic Fields†
R. Penrose and W. Rindler Spinors and Space-Time Volume 2: Spinor and Twistor Methods in
Space-Time Geometry†
S. Pokorski Gauge Field Theories, 2nd edition†
J. Polchinski String Theory Volume 1: An Introduction to the Bosonic String†
J. Polchinski String Theory Volume 2: Superstring Theory and Beyond†
J. C. Polkinghorne Models of High Energy Processes†
V. N. Popov Functional Integrals and Collective Excitations†
L. V. Prokhorov and S. V. Shabanov Hamiltonian Mechanics of Gauge Systems
S. Raychaudhuri and K. Sridhar Particle Physics of Brane Worlds and Extra Dimensions
A. Recknagel and V. Schiomerus Boundary Conformal Field Theory and the Worldsheet
Approach to D-Branes
R. J. Rivers Path Integral Methods in Quantum Field Theory†
R. G. Roberts The Structure of the Proton: Deep Inelastic Scattering†
C. Rovelli Quantum Gravity†
W. C. Saslaw Gravitational Physics of Stellar and Galactic Systems†
R. N. Sen Causality, Measurement Theory and the Differentiable Structure of Space-Time
M. Shifman and A. Yung Supersymmetric Solitons
H. Stephani, D. Kramer, M. MacCallum, C. Hoenselaers and E. Herlt Exact Solutions of Einstein’s
Field Equations, 2nd edition†
J. Stewart Advanced General Relativity†
J. C. Taylor Gauge Theories of Weak Interactions†
T. Thiemann Modern Canonical Quantum General Relativity†
D. J. Toms The Schwinger Action Principle and Effective Action†
A. Vilenkin and E. P. S. Shellard Cosmic Strings and Other Topological Defects†
R. S. Ward and R. O. Wells, Jr Twistor Geometry and Field Theory†
E. J. Weinberg Classical Solutions in Quantum Field Theory: Solitons and Instantons in High
Energy Physics
J. R. Wilson and G. J. Mathews Relativistic Numerical Hydrodynamics†
†
Available in paperback
Quantized Detector Networks
The Theory of Observation

G E O RG E J A RO S Z K I E W I C Z
 University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre, New Delhi – 110025, India
79 Anson Road, #06–04/06, Singapore 079906

Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/9781107136236
DOI: 10.1017/9781316477182
© George Jaroszkiewicz 2017
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2017
Printed in the United Kingdom by Clays, St Ives plc
A catalogue record for this publication is available from the British Library.
Library of Congress Cataloging-in-Publication Data
Names: Jaroszkiewicz, George, author.
Title: Quantized detector networks : the theory of observation / George
Jaroszkiewicz (University of Nottingham (retired)).
Other titles: Cambridge monographs on mathematical physics.
Description: Cambridge, United Kingdom ; New York, NY : Cambridge University
Press, 2017. | Series: Cambridge monographs on mathematical physics
Identifiers: LCCN 2017028906 | ISBN 9781107136236 (hardback) |
ISBN 1107136237 (hardback)
Subjects: LCSH: Quantum theory.
Classification: LCC QC174.13 .J38 2017 | DDC 530.12–dc23 LC record
available at https://lccn.loc.gov/2017028906
ISBN 978-1-107-13623-6 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
 Contents

Preface page xv
Acronyms xviii

1 Introduction 1
1.1 Motivation 1
1.2 Physics, Not Metaphysics 3
1.3 A Brief History of Quantum Interpretation 6
1.4 Plan of This Book 9
1.5 Guidelines for Reading This Book 10
1.6 Terminology and Conventions 11

2 Questions and Answers 13

2.1 What Is Physics? 13
2.2 Physics and Time 14
2.3 Reduction versus Emergence 16
2.4 Peaceful Coexistence 17
2.5 Questions and Answer Sets 18
2.6 Answer Set Collapse 19
2.7 Incompatible Questions and Category Errors 21
2.8 Propositions 22
2.9 Negation, Context, Validation 23
2.10 Proof of Negation 24
2.11 A Heretical View of Reality 25
2.12 Generalized Propositions and Their Classification 25

3 Classical Bits 28
3.1 Binary Questions 28
3.2 Question Cardinality 28
3.3 Classical Binary Questions 29
3.4 Classical Bits 29
3.5 Signal Bits 30
3.6 Nodes 31
3.7 Dual Bits 32
3.8 The Interpretation of Bits and Their Duals 32
viii Contents

3.9 Matrix Representation 33

3.10 Classical Bit Operators 33
3.11 Labstates 34
3.12 Time and the Stages Concept 35
3.13 Stage Diagrams 38
3.14 Measurements and Observations 38
3.15 Transtemporal Identity 39
3.16 Typical Experiments 40
3.17 Rank-One Stochastic Evolution 42
3.18 Stochastic Bits 43
3.19 Left-Stochastic Matrices 44
3.20 Stochastic Jumps 44
3.21 Stochastic Questions 45

4 Quantum Bits 46
4.1 Quantum Bits 46
4.2 Preferred Bases 46
4.3 Qubit Properties 47
4.4 Qubit Operators 48
4.5 Signal Bit Operators 50
4.6 The Standard Born Interpretation 50
4.7 The Born Interpretation in QDN 53
4.8 Classical and Quantum Ensembles 53
4.9 Basis Transformations 54
4.10 The Preferred Basis Problem 55
4.11 Rank-One Qubit Evolution 57
4.12 Mixed Qubit States 58
4.13 Density Operators 58

5 Classical and Quantum Registers 60

5.1 Introduction 60
5.2 Labels versus Ordering 60
5.3 The Signal Basis Representation 61
5.4 Maximal Questions 62
5.5 Signality 63
5.6 The Economy of Success 64
5.7 Quantum Registers 65
5.8 The Computational Basis Representation 66
5.9 Register Operators 67
5.10 Classical Register Operators 68
5.11 The Signal Algebra 69
5.12 Signal Excitations 70
5.13 Signality Classes 71
 Contents ix

5.14 Binary Decomposition 72

5.15 Computational Basis Representation for Signal Operators 73

6 Classical Register Mechanics 75

6.1 Introduction 75
6.2 Classical Registers 76
6.3 Architecture 76
6.4 Permutation Flows 79
6.5 Signality-Conserving Flows 81
6.6 Evolution and Measurement 82
6.7 Random Initial States 83

7 Quantum Register Dynamics 84

7.1 Introduction 84
7.2 Persistence 84
7.3 Quantized Detector Networks 85
7.4 Persistence and Ensembles 88
7.5 Observers and Time 89
7.6 The Born Probability Rule 90
7.7 Principles of QDN Dynamics 92
7.8 Born Maps and Semi-unitarity 93
7.9 Application to Dynamics 94
7.10 The Signal Theorem 97
7.11 Null Evolution 100
7.12 Path Summations 102

8 Partial Observations 105

8.1 Introduction 105
8.2 Observables 105
8.3 Maximal Questions 107
8.4 Partial Questions 109
8.5 Partial Question Eigenvalues 110
8.6 Identity Classes 111
8.7 Needles in Haystacks 112

9 Mixed States and POVMs 114

9.1 Introduction 114
9.2 Sets and Measures 115
9.3 Hilbert Spaces 116
9.4 Operators and Observables 116
9.5 Trace 117
9.6 Projection-Valued Measure 117
9.7 Mixed States 118
x Contents

9.8 Partial Trace 119

9.9 Purification 120
9.10 Purity and Entropy 121
9.11 POVMs 122
9.12 Naimark’s Theorem 123
9.13 QDN and POVM Theory 124

10 Double-Slit Experiments 131

10.1 Introduction 131
10.2 Run Protocol 132
10.3 Baseball in the Dark 133
10.4 Observed Phenomena 134
10.5 A Wave-Mechanics Description 136
10.6 The QDN Account of the Double-Slit Experiment 138
10.7 Contextual Subspaces 141
10.8 The Sillitto–Wykes Variant 142
10.9 The Monitored Double-Slit Experiment 143

11 Modules 148
11.1 Modules 148
11.2 The Vacuum 149
11.3 The Wollaston Prism 150
11.4 The Newtonian Prism 152
11.5 Nonpolarizing Beam Splitters 157
11.6 Mirrors 160
11.7 Phase Changers 160
11.8 Polarization Rotators 160
11.9 Null Modules 161

12 Computerization and Computer Algebra 162

12.1 Introduction 162
12.2 Program MAIN 163
12.3 The Wollaston Interferometer 165
12.4 Going to the Large Scale 169
12.5 Prospects 171

13 Interferometers 172
13.1 Introduction 172
13.2 The Mach–Zehnder Interferometer 172
13.3 Brandt’s Network 174
13.4 The Two-Photon Interferometer 176
 Contents xi

14 Quantum Eraser Experiments 179

14.1 Introduction 179
14.2 Delayed-Choice Quantum Eraser 181
14.3 Which-Path Measure 185
14.4 Wheeler’s Double-Slit Delayed Choice Experiment 186
14.5 The Delayed Choice Interferometer 189
14.6 The Double-Slit Quantum Eraser 190
14.7 Concluding Remarks 197

15 Particle Decays 198

15.1 Introduction 198
15.2 One Species Decays 200
15.3 The Quantum Zeno Effect 205
15.4 Matrix Analysis 207
15.5 The Ammonium System 209
15.6 Kaon-type Decays 212

16 Nonlocality 217
16.1 Introduction 217
16.2 Active and Passive Transformations 221
16.3 Local Operations 225
16.4 Primary and Secondary Observers 226
16.5 Subregister Bases 228
16.6 Local and Remote Evolution 229
16.7 The No-Communication Theorem 229

17 Bell Inequalities 232

17.1 Introduction 232
17.2 The Stern–Gerlach Experiment 235
17.3 Circumventing the Simultaneity Problem 241
17.4 The Standard Quantum Calculation 244
17.5 The QDN Calculation 247

18 Change and Persistence 250

18.1 Introduction 250
18.2 Comparisons 255
18.3 Signal Correlation Measure of Change 259

19 Temporal Correlations 263

19.1 Introduction 263
19.2 Classical Bit Temporal Correlations 264
xii Contents

19.3 The Classical Leggett–Garg Inequality 265

19.4 Qubit Temporal Correlations 266
19.5 QDN Spin Correlation 267
19.6 The Leggett–Garg Correlation 268
19.7 Understanding the Leggett–Garg Prediction 268

20 The Franson Experiment 271

20.1 Introduction 271
20.2 The Franson Experiment 271
20.3 FRANSON-1: ΔT  τ2 273
20.4 FRANSON-2: τ1  ΔT 274
20.5 FRANSON-3: τ2  ΔT  τ1 277
20.6 Conclusions 278

21 Self-intervening Networks 279

21.1 Introduction 279
21.2 Experiment SI-1: Basic Self-intervention 284
21.3 Experiment SI-2: Double Self-intervention 285
21.4 Experiment SI-3: Interfering Single Self-intervention 286
21.5 Schrödinger’s Cat 287

22 Separability and Entanglement 291

22.1 Introduction 291
22.2 Quantum Registers 291
22.3 Splits 293
22.4 Partitions 294
22.5 Quantum Zipping 298

23 Causal Sets 299

23.1 Introduction 299
23.2 Causal Sets 302
23.3 QDN and Causal Sets 303
23.4 Quantum Causal Set Rules 304
23.5 Case Study 1: The Double-Slit Experiment 305
23.6 Case Study 2: The Monitored Double-Slit Experiment 306
23.7 Case Study 3: Module Causal Set Structure 307

24 Oscillators 309
24.1 Introduction 309
24.2 The Classical Oscillator Register 310
24.3 Quantization 314
24.4 Bosonic Operators 315
24.5 Quantum Register Oscillator Operators 317
 Contents xiii

24.6 Comparison with Quantum Field Theory 319

24.7 Fermionic Oscillators 320

25 Dynamical Theory of Observation 322

25.1 Introduction 322
25.2 Power Bits 323
25.3 Power Bit Operators 325
25.4 Matrix Representation 325
25.5 Special Operators 326
25.6 The Laboratory 328
25.7 The Universal Register 329
25.8 Contextual Vacua 330
25.9 Experiments 332
25.10 Quantization 335
25.11 The Elitzur–Vaidman Bomb-Tester Experiment 335
25.12 The Hardy Paradox Experiment 339
25.13 Implications and Comments 342

26 Conclusions 343

Appendix 345
A.1 QDN Notation 345
A.2 Lab Time and Frame Fields 346
A.3 Lab Time and Stages 346
A.4 Ensembles 348
A.5 Vector Spaces 350

References 357
Index 365
 Preface

In a routine optical telescope scan of a distant galaxy, astronomer Alice saw

nothing unusual. Her radio astronomer colleague Bob, however, reported intense
radio activity in that galaxy. Who had the true view of the galaxy?
This is the sort of question discussed in this book. If you said that Bob had
the “true” view of the galaxy, you would be quite normal. Normal, in the sense
of average, or typical, or even reasonable. But if you went on to read the rest
of this book and understand its main message, you might then give a different
answer to that question.
It is not a trick question, however. The “correct” answer is not “Alice has the
true view of the galaxy.” Neither is it “Neither of them” nor is it “Both of them.”
This preface is not the place to discuss possible alternative answers to the above
question; you should be able to work one out based on the principles discussed
in the main text of this book. Although the question is easy to state, the answer
we give in the last chapter is simple neither to explain nor to justify. It is best
discussed using a lot of words and rather sophisticated mathematical models
and technologies. These are introduced, developed, and applied after intensive
preliminary discussions of the issues concerned.
Our answer is intimately bound up with the laws of observation as they pertain
to quantum processes, the subject matter of this book. These laws are the
rules that underpin modern, empirically based perceptions of physical reality
(our term for the world of experience).1 It has taken over two thousand years
of philosophical, natural philosophical, and empirical inquiry into the physical
universe for some of these rules to be discovered, particularly the ones involving
quantum processes. These latter have been understood for only the last hundred
years or so, and what they mean is still an active subject of debate. The old
question of how many angels can dance on the head of a pin is nothing compared
with the question of what the wave function means in quantum mechanics.
The problem is that most rules of quantum mechanics are counter intuitive
and may even appear wrong and unphysical to the person in the street. But
then, quantum mechanics has appeared baffling ever since it was stumbled on by
Planck in 1900. But as with Alice and Bob above, appearances can be deceptive:

1
I don’t imply there are other forms of reality. I can say nothing about that.
xvi Preface

it is quantum mechanics that continues to give verified predictions, while our

classical intuition, experience, and expectations all continue to be confounded.
It is my intention that this book be of value to a wide spectrum of readers:
refined quantum theorists, philosophers of science, experimentalists, students,
and the well-educated person in the street. If nothing else, I want to provoke
readers in two ways: first, to question their own ingrained belief structures about
the world they live in and second, to be alerted to the fact that there are a
lot of speculative concepts and theories that are discussed by some scientists
as if they were scientifically meaningful, when in reality they have no empirical
basis whatsoever. I call these theories vacuous and they are dangerous, because
they equate theory with empirical evidence. They lure the unsuspecting into
unfounded quasi-religious modes of thinking that may be good science fiction but
have no place in science proper. A sound interpretation of quantum mechanics
really does matter, because that involves our perceptions of reality. After all
is said and done, that is really what distinguishes us from other creatures and
affects everything that we do.
This book is timely in two ways. First, after a century of attempts to under-
stand quantum theory from a classical, noncontextual perspective, intense activ-
ity into the neurological basis of human psychology has started to reveal a
starting truth: classical reality exists only in the mind. Every day, the results of
new laboratory experiments demonstrate more and more convincingly just how
misleading the classical perception of the world really is. We humans are now
being shown up for what we are: we live in illusory worlds of our imaginations,
constructed by our brains as they attempt to match the vast flood of sensory data
streaming in through our senses with patterns of preconditioned rationalization.
Second, there is intense interest and work worldwide in the development of
quantum computers. In this book I discuss the application of computer algebra
to quantum registers containing many quantum bits (qubits). Although my focus
in this book is on the description of quantum experiments, there will be a
great overlap between quantum computation and that focus, so that I expect
this book to be of some interest, and I hope value, to quantum computation
theorists.
A preface is the place for an author to give their thanks and acknowledgment.
Here are mine. First, I have been overawed by the fact that the underlying
theme of this book, which is the debate about the nature of perception, has
been going on for many centuries. Some of the views that I have read greatly
impressed me and I realized that there were individuals who saw through the fog
of conditioning that surrounds all of us. In particular, I greatly admire the works
of David Hume, whose views about the nature of perception I found remarkably
in tune with this book. As an undergraduate, I was greatly influenced by the
lectures of Nicholas Kemmer. Later, I encountered the brilliant works of Julian
Schwinger, and even had the great pleasure of meeting him and talking with him
in 1993 in Nottingham and London, during the bicentenary celebrations of the
 Preface xvii

birth of George Green. These extraordinarily generous individuals in particular

and many others influenced me in one way or another in the most significant of
ways.
Finally, I express my gratitude to all members of my extended family, past and
present. My parents gave me the opportunities to start writing and the arrival of
my granddaughter Julia gave me the motivation to continue when that writing
became tedious. Without them this book would not exist. In particular, my wife,
Malgorzata, has been immensely supportive and helpful in all my efforts; I thank
her for that and for her great patience.
 Acronyms

CA computer algebra
CBR computational basis representation
CM classical mechanics
CPT charge, parity, and time
CST causal set theory
CTC closed timelike curve
DS double-slit
EPR Einstein, Podolsky, and Rosen
eQDN extended quantized detector networks
FLS Feynman, Leighton, and Sands
GP generalized proposition
GPC general proposition classification
GR general relativity
HV hidden variables
LG Leggett–Garg
LSZ Lehmann–Symanzik–Zimmermann
MDS monitored double-slit
MS Misra and Sudarshan
NEO null evolution operator
ONB orthonormal basis
POVM positive operator-valued measure
PVM projection-valued measure
QDN quantized detector network
QED quantum electrodynamics
QFT quantum field theory
QM quantum mechanics
REC relative external context
RIC relative internal context
RQFT relativistic quantum field theory
SBR signal basis representation
SG Stern–Gerlach
SR special relativity
SUO system under observation
ToE Theory of Everything
 1
Introduction

1.1 Motivation
The aim of this book is to introduce, develop, and apply quantized detector
networks (QDN), an information-based, observer-centric approach to quantum
mechanics (QM). Six reasons motivating our development of QDN are the
following.

Avoidance of Metaphysical Speculation

There is such a variety of unproven (and unprovable) speculation concerning
the interpretation of QM that the subject of this book, QDN, may appear at
first sight to be yet another in this growing branch of metaphysics. In fact,
our motivation is precisely the opposite. QDN was intended from the outset to
reduce the level of metaphysics in the application of QM. To achieve this, our
strategy is to move the traditional focus of attention away from systems under
observation (SUOs) and toward the observers of those SUOs. In QDN, wave
functions represent not states of SUOs but states of apparatus. We shall call
such states labstates, to distinguish them from states of SUOs (which do have a
place in QDN). It is only labstates that observers can ever deal with directly.
In this respect, QDN is an attempt at a more laboratory-based description of
quantum physics than standard QM, focusing as much on how an experiment is
done as on the results of that experiment. For example, instead of talking about
the wave function of an electron, QDN talks about the labstate of the signal
detectors that allow us to say anything about that electron in the first place.
That is all, there is nothing deeper. Whether or not electrons actually “exist”
is thereby relegated to an inessential metaphysical issue that we can choose to
ignore.
It is important to understand what QDN does not say as much as what it does
say. QDN does not say that electrons do not “exist”. QDN merely asks for an
empirical definition of “existence”.
2 Introduction

QM Is Much More Than a Calculational Device

Many physicists hold the utilitarian view that the success of quantum theory
in predicting experimental data is sufficient for their purposes, and that further
inquiry is really the business of metaphysics and therefore outside the scope of
science proper. We cannot say that this is entirely unreasonable. However, we
take the view that QM represents such a radical departure from the principles
of classical mechanics (CM) that this pragmatical view cannot be all there is to
the subject.

New Interpretations Lead to New Experiments

The history of QM is littered with the debris of various interpretations and
paradigms. Planck’s quantization of energy (Planck, 1900b), Bohr’s radiation
damping veto (Bohr, 1913), de Broglie’s pilot waves (de Broglie, 1924), matrix
mechanics (Heisenberg, 1925), wave mechanics (Schrödinger, 1926), the Copen-
hagen interpretation (Heisenberg, 1930), Hidden Variables (Bohm, 1952), the
Relative State interpretation (Everett, 1957), decoherence (Zurek, 2002), and the
Multiverse (Deutsch, 1997) are some of the most discussed views of what quantum
theory means. We do not think any of these ideas are right or wrong in an absolute
sense, because we do not know what that means. It is clear, however, as we shall
argue later, that some interpretations are empirically vacuous, meaning that they
can be neither proved nor disproved according to scientific empirical protocols.
It is our inclination to avoid vacuous concepts and theories.
When examined in detail, each interpretation of QM has led to new insights
and questions that have motivated new experiments, resulting in advances in
physics.1

Strength in Diversity
There is no evidence for the view that there is, or even should be, a single, best
way to do QM. For instance, while we believe that quantum electrodynamics
(QED) can explain almost everything about the hydrogen atom, the Schrödinger
wave equation does a good enough job most of the time and far more econom-
ically. QDN is but another way to describe and formulate quantum physics; it
will have advantages in some contexts and disadvantages in others.

Logical Development
QDN is not intended to replace QM. On the contrary, QM works brilliantly and
it would be foolish to think we could do better. QDN is intended to enhance
QM in areas where little or no attention has been paid hitherto: the observer
and their apparatus. QDN seems to us to be a logical application of successful
principles that sooner or later would be developed further along the lines taken
in this book.

1
Not in every case, unfortunately.
 1.2 Physics, Not Metaphysics 3

Human Conditioning
There is now a great deal of empirical evidence for the proposition that humans
see the world not as it is but how they have been conditioned to see it (Halligan
and Oakley, 2000; Kahneman, 2011). This conditioning occurs because of various
forces: evolutionary, genetic, cultural, linguistic, familial, political, and so on. CM
is based on ordinary, everyday-life human experience. QM was developed recently,
only just over a hundred years ago, and only once sufficient technology had been
developed that could reveal the deficiencies in our classical conditioning. We have
not caught up with QM in our modes of thinking, and it is not obvious that we
ever will. It remains a natural tendency for even the most experienced quantum
theorist to explain things in familiar terms, such as space, time, and particles.
QDN can be seen as an attempt to steer the theorist away from those conditioned
thoughts whenever it is wise to do so.

1.2 Physics, Not Metaphysics

To motivate the QDN approach, we review in this chapter some of the conceptual
foundations of QM that are relevant to us here. Before we do that, however, a
few words of caution about the relationship between physics and metaphysics
seem advisable.
In this book, whenever we refer to metaphysics, we shall mean the study of
propositions and assertions that cannot be empirically tested, that is, cannot be
assigned a truth value relative to any physical context. In our view, it would
be a serious error to leave the interpretation of QM to metaphysicists and
philosophers. If they had something to contribute that was empirically testable,
then they would be physicists, not metaphysicists and philosophers.
However, while it is true that physics is properly concerned only with empir-
ically testable propositions, this does not mean that physicists should never
deal with abstract concepts. Physicists after all only do physics because of the
way they think and are motivated. Those processes are not currently regarded
as legitimate concerns in physics, but that is probably due to the status of
technology at this time.2
Moreover, physicists use mathematics all the time, because it is the best
language in which to formulate theories and propositions about the subject. But
all of that takes place in the theorist’s mind; theories do not exist by themselves.
Mathematics is based on axioms and postulates that are tested not empirically
but only for internal logical consistency.
Furthermore, the process of objectivization (the attribution of material exis-
tence and properties to recurring and persistent physical patterns and processes)

2
This may well change with advances in the neurosciences. Observers are fundamental to
physics, so how observers operate and fit into the grand scheme of things seems a
reasonable subject for investigation.
4 Introduction

is the only way that physics is done. Significant examples are the concepts of
SUO, observer, particle, atom, molecule, wave function, space, time, apparatus,
and so on. These are all mental constructs and essentially metaphysical in nature.
There is a paradox of sorts here involving the use of mathematics in physics. A
conventional, indeed perhaps ubiquitous, view about empirical physics is that the
empirically discovered laws of physics are independent of observers. Yet observers
are indispensable. Observers are needed first to formulate theories (which do not
exist except in the minds of theorists) and then to test those theories empirically.
It is a metaphysical proposition to assert or to believe the proposition that the
properties of objects “exist” independently of observers or observation. How
could we prove that proposition? We shall call this idea the realist interpretation
of physics, otherwise known as realism.
The paradox is that the realist interpretation has been remarkably successful in
classical physics, that is, the interpretation of phenomena according to the laws
of CM based on the Newtonian space-time paradigm or the Minkowski/general
relativistic spacetime paradigm. That success has led many physicists to believe
that this interpretation should apply to the whole of physics.
The paradox we refer to above has an additional twist: the classical realist
position simply does not work when applied to real physics experiments that go
beyond a certain basic level of sophistication. It just does not. The empirical
evidence against the classical realist position as the be-all and end-all is now
virtually unbreakable, although there will always be theorists who do not accept
that statement. We have in mind the Bell inequality-type experiments that are
discussed in Chapter 17. These have now established beyond much reasonable
doubt that classical intuition is deficient when it comes to quantum physics (by
which term we mean those experiments where quantum phases are significant).
Therefore, something has to give.
This book is an explanation of our interpretation of what quantum mechanics
really means, which is that it is a theory of observation. This involves the observer
as well as the observed.
The resolution of the above-mentioned paradox that led us to QDN was the
painful acceptance of the fact that physics is done by humans. This fact is painful
in that the current tendency in physics is to avoid reference to humans specifically,
and to marginalize observers as much as possible. There is a standard reason
given for this, often referred to as the principle of general covariance. This is the
assertion that the laws of physics are intrinsic and independent of observers.
The principle of general covariance is, when examined properly, a vacuous
proposition, but the paradox is that it works in CM, particularly in general
relativity (GR). Some theorists even employ it in their formulations of quantum
theory. Such efforts invariably end up as branches of mathematical metaphysics,
because QM is all about real world physics and experimentation. We think that
perhaps science has reached the point where observers should be factored much
more into the equations somehow. That agenda is not easy, and is what this book
 1.2 Physics, Not Metaphysics 5

is all about. What helps greatly here is the hard fact that, logically, there can be
no empirical evidence for the realist position or for the universality of the principle
of general covariance, because the very notion of empirical evidence necessarily
involves human observers. You cannot have your cake and eat it. You cannot use
observation to prove that the laws of physics are independent of observation.
If there is one thing that QM teaches us, then, it is the uncomfortable lesson
that not all intuition stands up. Sometimes, particularly when it comes to quan-
tum processes, the natural human interpretation of what is believed to be going
on is inconsistent, or just meaningless.
Our considered view, then, is that QM is not a theory describing objects per se
but a theory of entitlement. QM provides us with a set of rules that tells us what
we as observers are entitled to say in any particular context, and no more. The
good news is that this theory of entitlement is self-consistent and has definite
mathematical rules that have never yet been found to fail.
We illustrate the sort of scientific philosophy that guides us in this book with
the following, scenario A. Suppose a hundred, a thousand, a million observers
had independently measured some quantifiable property about some SUO S and
had come up with the same numerical value to within experimental error. Then
it would be natural to assert that S “has” that property. That is something we
do all the time. We say that our house “has” four bedrooms, that our car “is”
white, that John “is” tall, and so on. The logic seems inescapable. The world
around us “has” properties that we discover.
But when it comes to quantum physics, there is now enough evidence that
tells us that we cannot always rely on the above assertion. We cannot exclude
the possibility that the very processes used to “see” those presumed properties
of SUOs create, in some way, those very properties. If we take this as a warning,
then the only thing that we would be entitled to say about scenario A with real
confidence is only what we could be sure of: that a hundred, a thousand, a million
observers had independently performed some procedure and obtained the same
number to within experimental error.
This may sound limited, cautious, lacking in vision, and so on. But the history
of empirical quantum science over the last several decades has shown unam-
biguously that if we made the realist assertion that SUOs “have” measurable
properties, then sooner or later we would come to an empirically observed break-
down of some prediction based on that assertion. The theorist Wheeler went so
far as to formulate this principle of entitlement in the form of what he called the
participatory principle:

Stronger than the anthropic principle is what I might call the participatory
principle. According to it we could not even imagine a universe that did not
somewhere and for some stretch of time contain observers because the very
building materials of the universe are these acts of observer-participancy.
6 Introduction

You wouldn’t have the stuff out of which to build the universe otherwise.
This participatory principle takes for its foundation the absolutely central
point of the quantum: No elementary phenomenon is a phenomenon until
it is an observed (or registered) phenomenon.
(Wheeler, 1979)

So how should we interpret the empirical agreement of the observers’ mea-

surement on system S? The answer is to look only at what the facts tells us,
and in this case, they tell us no more than that a bunch of observers agree on
a measured value. The important point is that there is consistency (or repro-
ducibility) in their observations. It is consistency that is the key here and in
all empirical science (and indeed mathematics). Consistency requires observers,
for if we did not have observers, who is the judge of consistency? Therefore, we
should interpret the laws of physics not in terms of absolute properties of systems
under observation, but as consistency relationships between what observers do.
This means that we are proposing a psychological shift in emphasis, a change
in our perspective of what physics is about. In QDN, physics is not the study of
SUOs and the determination of their assumed properties, but the study of the
relationship between observers and their apparatus. That is the basic logic on
which QDN has been developed. QDN is essentially a theory of entitlement, a set
of rules for what observers can say with relative confidence about their apparatus
under contextually relevant circumstances.
We need to add one critical comment here. The QDN approach emphatically
does not assert or require us to believe that nothing exists outside of observation.
Indeed, were we to take such an attitude, we would have no way of calculating the
transition amplitudes needed to make predictions. As theorists, we are entitled
to model the information void , the unobserved regime between state preparation
and outcome detection, in any way that leads to successful predictions. To that
extent, QM really is like a black box of tricks, whose workings we can only
guess at.

1.3 A Brief History of Quantum Interpretation

Before 1900
To understand the impact QM had on physics, we should first understand the
principles of CM that most physicists accepted prior to 1900, the year in which
Planck stated his quantum hypothesis.
CM is a view of physical reality that is predicated on the way that humans
normally perceive and interact with their environment. Without any additional
equipment, a typical human interacts with their surroundings by vision, by touch,
by sound, by taste, and by smell. Several critical facts about each of these senses
conspire to lead the human brain to subconsciously create mental models of their
 1.3 A Brief History of Quantum Interpretation 7

environment, and it is these models that were used by theorists such as Newton,
Lagrange, and Hamilton as the basis on which to construct CM. Four of these
critical facts are the following.

Length and Time Scales

Humans are relatively enormous compared with atomic scales, so much so that
the existence of atoms was not established conclusively until just over a hundred
years ago. Matter therefore gives the impression of being continuous. Moreover,
the dynamical processes responsible for the stability of atoms and molecules have
time scales that are long enough to create the illusion that we shall refer to as
persistence, the idea that SUOs and observers exist as meaningful entities over
significant intervals of time and therefore can be objectified (that is, can be given
individual identities).

Daylight Is Bright
To the human eye, daylight is relatively bright. The human brain interprets
objects that it sees visually not in photonic, that is, discrete, terms but rather
as a continuous process. The net effect is to create the illusion that well-defined
objects or SUOs exist over extended time scales in continuous space and time.

Apparent Observer Independence

Different human observers looking visually at an object will generally agree that
they see the same thing, albeit with different relative positions and velocity
components. This belief leads to the principle of general covariance mentioned
above, the idea that intrinsic properties of SUOs are independent of observers
and frames of reference.

Zero Observational Cost

When humans see and hear objects, there is generally no noticeable effect back
on those objects due to that observation. In quantum physics this is referred to
as noninvasive measurability.

These facts and others, such as the way that the human brain processes infor-
mation, all conspire to give the impression that objects under observation “have”
physical properties independent of any observer. Moreover, these properties can
be measured nondestructively and understood by the rules of classical mechanics.
For example, in Newton’s laws of motion, there is no mention of any apparatus
or observer, apart from the implication that these laws are stated relative to
inertial frames of reference.

After 1900
Up to 1900, virtually all physicists thought about space, time, matter, and the
way physics was to be done in the above classical terms. The advent of special
relativity (SR) in 1905 in no way changed this perspective: special relativity is
8 Introduction

all about the differences between observers and how they see things classically,
and not about those things themselves. Indeed, theorists such as FitzGerald
(FitzGerald, 1889), Larmor (Larmor, 1897), and Lorentz (Lorentz, 1899) had
been working out many of the details of SR well before 1900, basing their work
on the Newtonian space-time paradigm. Their work and that of Einstein in his
landmark 1905 paper on SR (Einstein, 1905b) had nothing to do with what
Planck started in 1900. So little, in fact, that bridging the gap between SR and
QM remains perhaps the greatest challenge in mathematical physics.
Up to 1900, Planck had been trying to understand the empirical data from
experiments on black body cavities, using Newtonian physical principles and
Maxwell’s theory of electromagnetism. That approach led to the prediction that
the intensity of light emitted from a black body would grow with frequency,
but the data indicated otherwise. In 1900, in order to understand the data, he
approached the problem in a new way, departing from conventional principles.
To understand properly what Planck did not say in 1900, we need to under-
stand what Einstein did say in 1905. In that year, Einstein explained the pho-
toelectric effect in terms of particles of light (Einstein, 1905a). These particles
subsequently became known as photons (Lewis, 1926). What seemed paradoxical,
and still does to this day, is that photons are regarded as particles associated with
Maxwell’s theory of electromagnetism, a classical theory that predicts that light
is a wave process and not a particulate process.
Planck did not assert that light itself was quantized. What he said was that the
black body data could be explained if the assumed atomic oscillators lining the
walls of a black body cavity could absorb and/or emit electromagnetic radiation
only in well-defined amounts, known as quanta:

Let us consider a large number of monochromatically vibrating resonators

– N of frequency ν (per second), N  of frequency ν  , N  of frequency
ν  , . . . , with all N large number – which are at large distances apart and
are enclosed in a diathermic medium with light velocity c and bounded
by reflecting walls. Let the system contain a certain amount of energy, the
total energy Et (erg) which is present partly in the medium as travelling
radiation and partly in the resonators as vibrational energy. The question
is how in a stationary state this energy is distributed over the vibrations of
the resonators and over the various of the radiation present in the medium,
and what will be the temperature of the total system.
To answer this question we first of all consider the vibrations of the
resonators and assign to them arbitrary definite energies, for instance, an
energy E to the N resonators ν, E  to the N  resonators ν  , . . .. The sum

E + E  + E  + · · · = E0

must, of course, be less than Et . The remainder Et − E0 pertains then to

the radiation present in the medium. We must now give the distribution
 1.4 Plan of This Book 9

of the energy over the separate resonators of each group, first of all the
distribution of the energy E over the N resonators of frequency ν. If E
considered to be continuously divisible quantity, this distribution is possible
in infinitely many ways. We consider, however – this is the most essential
point of the whole calculation – E to be composed of a definite number of
equal parts and use thereto the constant of nature h = 6.55 × 10−27 erg·sec.
This constant multiplied by the common frequency ν of the resonators gives
us the energy element ε in erg, and dividing E by ε we get the number P of
energy elements which must be divided over the N resonators. If the ratio
is not an integer, we take for P an integer in the neighbourhood.
(Planck, 1900a)

Planck’s idea was altogether conceptually different from that of Einstein. We

can see the difference directly in the above quote from Planck’s paper: Planck
is concerned there with E0 , the energy associated with the oscillators, whereas
Einstein discussed Et − E0 , the energy of the radiation in the cavity. These are
very different things.
Einstein was a confirmed realist who never accepted Bohr’s interpretation of
QM, that is, the set of ideas generally referred to as the Copenhagen interpre-
tation. We have to side with Bohr here: the assertion that photons “exist” is
a vacuous one. How could we confirm that there are “particles of light” with-
out using detectors that are themselves subject to Planck’s quantum principle,
absorbing and/or emitting energy in discrete amounts only?
QDN can be understood as a return to Planck’s original position, away from
the objectification of quanta as “things” and toward the view that quanta char-
acterize processes of observation as they are in the real world of physics. QDN is
fully in accord with the principles laid down by Heisenberg, Born, and Bohr, but
not with the diametrically opposing views held by followers of HV or the Many
Worlds paradigm. We regard those as vacuous, reflecting the classical condition-
ing that humans are generally subjected to: no more than contextually incomplete
attempts to explain the nonclassical rules of quantum mechanics in realist terms.

1.4 Plan of This Book

The nature of the subject we are discussing makes a division into four themes
helpful. Each theme has its own particular characteristics and flavor.
The first theme, Basics, runs from Chapters 1 to 9. This theme builds up the
formalism of QDN from basics, starting with a review of what QDN is about, then
turning to the mathematics of classical bits, then quantum registers, and finally
the positive operator-valued measure (POVM) approach to quantum physics.
The second theme, Applications, runs from Chapters 10 to 20. In it, QDN is
applied to a number of experiments that show how QDN differs from standard
10 Introduction

quantum formalism. We include recent, exciting experiments, typically in quan-

tum optics, such as quantum eraser, delayed choice, and Bell and Leggett–Garg
inequality experiments.
The third theme, Prospects, runs from Chapters 21 to 26, and is more spec-
ulative than the previous themes. In it, we discuss the prospects for future
application of the QDN formalism, such as the possibility of constructing a
generalized theory of observation.
The final theme, Appendices, consists of material that stands alone but is
referred to at various places in the other themes.

1.5 Guidelines for Reading This Book

The subject of quantum mechanics and its interpretation has had a long and
tortuous history. There is now good empirical evidence for the statement that
quantum processes cannot be explained using classical modes of thinking.
It is inevitable that we should comment in this book on topics that may
appear metaphysical; that is unavoidable because there is no point in presenting a
detailed mathematical formalism (which we do in due course in this book) if we do
not say what we think it means in real-life terms. This book is aimed squarely at
representing what is actually done in a laboratory, not what is imagined is done.
Although the substantive mathematical formalism starts in Chapter 3, we
would advise the reader to go over the first two chapters, in order to gain a
feeling for what we are about in this book. The mathematics relating to those
concepts is developed in the main body of the text. Most of that mathematics will
be familiar and not reviewed in much detail, but where novelty is encountered,
mathematical structures are explained more fully.
An important aspect of our approach in this book is the concept of architec-
ture, by which we mean a verbal or diagrammatical description of the processes
involved in an experiment. Different experiments may have different architec-
tures. For example, an elementary particle scattering experiment will usually
involve an initial in state, an intermediate scattering regime, and a final out state.
That architecture is sufficient to describe electron–proton scattering at relatively
low energies, but if unstable particles are emitted in what appears to be the final
state, then their decay processes require the architecture to be extended in time
to include the detection of their decay products. Another important example of
architectural differences concerns the distinction between pure states and mixed
states in QM. Architecture is involved intimately in delicate subjects such as
Bell inequality experiments, where failure to appreciate details of architecture,
particularly those involving counterfactual arguments, can lead to confusion and
misleading conclusions to be arrived at.
Acronyms are used extensively throughout this book. We follow the general
convention that the first instance of an acronym in any chapter is defined at that
point and the acronym used thereafter in that chapter. For convenience, there is
a list of acronyms after the Preface.
 1.6 Terminology and Conventions 11

Where important concepts, paradigms (modes of thinking), and theories are

encountered, we will capitalize their first letters. For instance, we shall write
Absolute Time and not absolute time, Universe and not universe, and so on.

1.6 Terminology and Conventions

Basic mathematical concepts such as vector spaces and Hilbert spaces are
reviewed in the Appendix.

Maps, Operators, Transformations

If U and V are two vector spaces over the same field, then any process, linear or
otherwise, that takes any vector in U into V or back into U will be referred to
as a map, operator , or transformation, depending on context.

Retractions
If f is a function that maps a set A into a set B, we denote the set of image
points in B by f (A). In this book we encounter situations where f (A) is a proper
subset of B, which means that there are elements in B that are not images of
any element in A. If f is an injection, which means that there is a one-to-one
correspondence between elements in A and elements in f (A), then the retraction
f of f is the well-defined function with domain f (A) that takes elements from
f (A) back into A, given by f (f (a)) = a for any element a of A. We shall use
retractions a lot in this book.

The Two Worlds

As QDN was being developed, it became clear that the theory was dealing with
two separate worlds: the inner world of the system under observation and the
outer world of the observer and their apparatus. Linking these two worlds are
the quantum processes of interest. Therefore, it became necessary to devise a
notation that could readily distinguish between these two worlds, yet retain all
the characteristics of quantum mechanics.
Our notational solution to this problem is the following. We shall deal fre-
quently with a total Hilbert space, the tensor product H ⊗ Q of two Hilbert

A f
B
a b
f(a)
f(A)
f
Figure 1.1. If f : A → B is an injection and f (A) is a proper subset of B,
then there are elements such as b in B for which the retraction f of f is not
defined.
12 Introduction

spaces. The first Hilbert space H will contain the system states, the imagined
states of the SUO, and for those we shall use the conventional and familiar Dirac
bra-ket notation |Ψ, with inner product Φ|Ψ, and so on. The second Hilbert
space, Q, will be a quantum register containing the labstates, the quantum states
of the apparatus. For these labstates we shall use bold notation j and their duals
will be denoted j. Labstate inner products are denoted ij, and so on. Total states
are elements of the total Hilbert space and denoted by |Ψ, i) ≡ |Ψ ⊗ i, noting
the round bracket on the left-hand side of this definition. Inner products of total
states are given by

(Φ, i|Ψ, j) = Φ|Ψij. (1.1)

The use of bold font for labstates and overlined bold font for their duals is
consistent with the question and answer formalism described in the next chapter.
 2
Questions and Answers

2.1 What Is Physics?

Throughout this book we shall be concerned greatly with questions and answers.
The subject that an observer asks a question of will be denoted by a bold symbol
such as Ψ. On the other hand, the question that is asked of a subject will be
denoted by a barred bold symbol such as Q. Then the answer A that is obtained
when that question is asked of that subject will be denoted by A ≡ QΨ. This
notation will be used to explain what quantized detector networks (QDN) is all
about and in its mathematical implementation. Note that the left-right ordering
here is significant: ΨQ will mean something very different.
If I, the author, were asked the question Q ≡ What is physics?, my answer
QI would be
QI = Physics is the process whereby observers first formulate,
then ask, and then answer, empirical questions.
We need to clarify some points about this answer. By empirical, we mean accord-
ing to the established principles of science, whereby theorists make propositions
or assertions and then establish the truth values of those propositions relative
to experiments carried out by unbiased experimentalists. This is done to acquire
information about systems under observation (SUOs) in order to add to our
knowledge about the relative physical Universe (that part of the Universe empir-
ically accessible to us). We are motivated to do all this for numerous reasons, not
the least of which is basic curiosity. This is the view of physics taken in this book.
Physics is emphatically not a discipline that philosophers and metaphysicists
can meaningfully contribute to, because without empirical testing, any proposi-
tion has to be regarded as scientifically vacuous (empirically meaningless).
The questions relevant to physicists are posed through experiments and con-
sequently will be referred to as empirical questions. Other sorts of questions will
be referred to as theoretical questions.
14 Questions and Answers

Matters are never as simple as that. We will clarify presently what we mean
by observers, SUOs, and so on. At this point, we observe that the above answer
QI begs a number of questions that should have been asked and answered first.
Specifically, we should also ask the questions
A≡ Who is asking this question?
B≡ Why is this question being asked?
C≡ Who or what is being asked?
D≡ What will be done with the answer?

and perhaps more. It is the failure to ask questions A, B, and so on, that seems
to us to be the root cause of many if not all of the conceptual problems people
have in the interpretation of quantum mechanics (QM).
We could go further. The context, or circumstances, under which we ask our
questions is extremely important. For instance, physicists use experiments to
answer their empirical questions. The apparatus that they use is part of that
context: if the apparatus cannot be constructed, then the questions cannot even
be asked. When we look carefully at what we are doing when we do physics,
it soon becomes obvious that we have to take great care in how we ask our
questions.
We should say at this point that we do have a specific view about how questions
should be asked in physics, particularly quantum physics. This view is based
on a concept that we call contextual completeness, discussed in Section 2.12.
Contextual completeness is a measure, based on a simple algorithm, for deciding
whether any given assertion or proposition is a reasonable scientific proposition
or not. Specifically, we can establish whether it could be investigated according
to proper scientific protocols, or whether it is a vacuous proposition, meaning
that it cannot be empirically tested for a truth value. One of the surprising
features of our analysis is that it shows that there are several degrees of contextual
completeness, ranging from the completely vacuous (characteristic of metaphysics
and philosophy) to the fully contextually complete proposition required in QM.

2.2 Physics and Time

In this book, we take the view that it is necessary and good physics to discuss
the meaning of time, a central element in all branches of physics and certainly
too important to leave to philosophers and metaphysicists.
There are two contrasting natural philosophies of time, called Manifold Time
and Process Time (Encyclopaedia Britannica, 2000). Both of these philosophies
are inextricably woven into the fabric of science. Neither is right nor wrong; each
has value in the appropriate context. In this book, our focus will be mainly on
Process Time, as this is the time most relevant to how humans operate when
they do physics experiments.
 2.2 Physics and Time 15

Manifold Time
Manifold Time is the view that time is a “thing,” an objectivized parameter
modeled as a real geometrical dimension. An important development in Manifold
Time occurred in 1908, when Minkowski proposed that time could be mod-
eled geometrically, as one of the dimensions of spacetime, the four-dimensional
Lorentz-signature manifold of special relativity (SR), in which all past, present,
and future events are contained (Minkowski, 1908; Petkov, 2012). This model of
the universe is often referred to as the Block Universe (Price, 1997).
There are several significant factors commonly used by physicists to support
Manifold Time and the Block Universe concepts. Some of these are the following.

Relativity
Manifestly Block Universe theories such as SR and general relativity (GR) are
based on spacetime manifolds with Lorentzian signature metrics. These are dis-
cussed in the Appendix, Section A.2.

Reversibility
The laws of Newtonian classical mechanics excluding friction are generally time-
reversible. Lines in geometry do not carry any intrinsic direction. Therefore, if
time is reversible, it may be possible to model it geometrically.

Unitary Evolution
In QM, states of SUOs evolving without observational intervention are described
by Schrödinger unitary evolution, and this is regarded as reversible.

CPT Theorem
In high energy particle physics, the well-known charge, parity, time (CPT) rever-
sal theorem involves symmetry operations in spacetime that are consistent with
all experiments to date, particularly particle scattering amplitudes, which have
empirically confirmed crossing symmetries (Eden et al., 1966).

Differential Equations
Many differential equations in mathematical physics, such as Maxwell’s equations
for electrodynamics and Einstein’s field equations in GR, contain no dissipative
terms, leading to both advanced and retarded solutions to the field equations
with no natural (to the theory) reason to eliminate the classically unacceptable
advanced solutions.

Process Time
In contrast to Manifold Time, Process Time (Encyclopaedia Britannica, 2000)
is based on an acknowledgment that there are irreversible processes going on in
the Universe and that humans experience a feeling that there is a “moment of
16 Questions and Answers

the now.” Time is no longer viewed as a thing but as a process. This implies
the existence (in this context) of the systems or objects that are processing in
time. According to QDN, the observers of those processes are fundamental to
this discussion as well. Factors that we take here as supporting the Process Time
paradigm are the following.

Irreversibility
Thermodynamics, one of the fundamental branches of physics, has irreversibility
built into it.

The Born Rule

The Born rule in QM deals with probability, which is inherently asymmetric
regarding time.

Observers
Observers have memory and intention, which are asymmetric regarding the direc-
tion of time. Observers operate contextually in an irreversible way, acquiring
information as time progresses.

Hubble Expansion
The Universe appears to be expanding in an irreversible way from a definite
point in the past. The Block Universe model has no empirical justification for
postulating events earlier than that point. Indeed, the Steady State paradigm
(Bondi and Gold, 1948; Hoyle, 1948) that postulated an indefinite past has been
generally abandoned by cosmologists.

Contextual Incompleteness
The Block Universe paradigm is contextually incomplete with a generalized
propositional classification of zero (contextual completeness and generalized
propositional classification are discussed further on in this chapter). It is a
vacuous, metaphysical concept.

2.3 Reduction versus Emergence

The Manifold Time and Process Time paradigms can be reconciled if context is
taken into account. Manifold Time is based on a reductionist view of the laws
of physics, whereas Process Time is associated with emergence (or complexity).
Reductionism is the principle that all complex SUOs and indeed the universe can
all be explained by a relatively few basic laws. It is a bottom-up view of science.
Perhaps the purest form of reductionism is seen in elementary particle physics,
which presupposes that there is a single Lagrangian that can be used to describe
all phenomena by a Theory of Everything (ToE).
 2.4 Peaceful Coexistence 17

Emergence, on the other hand, is an acceptance that there are situations that
transcend reductionism. For example, life is a process, not a collection of atoms;
there is no mention of life in the standard model of high energy physics.
Situations in physics that fall out of the scope of reductionism and require
emergent thinking generally involve observers and large-scale processes typically
involving numbers of particles on the scale of Avogadro’s number and more, that
is, of the order 1026 . Examples in the physical sciences that require emergent
concepts are thermodynamics, gravitation, cosmology, and quantum physics. We
take the view in this book that it is a fundamental category error to believe that
reductionism alone can explain any of those concepts.
The question remains as to the status of reductionism in QDN. Does the
relatively successful reductionist Standard Model of particle physics have a place
in QDN?
Our answer is emphatically yes. Reductionism has a specific role in those
aspects of observation that do not involve apparatus explicitly, and that is in the
information void , the conceptual region between state preparation and outcome
detection. We shall be greatly preoccupied with the information void, for it is in
that region where the idealized Lagrangians of particle physics can be explored
and used to work out the quantum signal transition amplitudes that QDN needs.
Indeed, QDN has no way by itself of generating those amplitudes. In this respect,
QDN is analogous to scattering matrix (S-Matrix) theory1 in hadronic physics
(Eden et al., 1966), which deals with the architecture of scattering processes
but cannot account for the explicit details of amplitudes without introducing
Lagrangians.
An important and useful point here is that, by definition, we cannot “know”
what the information void is. We can only make models for the transmission of
information from preparation device to outcome detector. This is useful because it
means we are free in our choice of the models we use to calculate the transmission
amplitudes from source to detector across information voids. These models need
not be of the standard Lorentzian signature, spacetime manifold type that are
used in relativistic quantum field theory. They could be more complicated, such
as Snyder’s quantized spacetime algebra (Snyder, 1947a,b), or even manifesta-
tions of superstrings. At the end of the day, such models simply provide “black
box” recipes for the calculation of amplitudes that will have various degrees of
empirical correctness but should never be regarded as reflecting absolute truth
in any way. Of course, some models will be much better than others.

2.4 Peaceful Coexistence

Peaceful coexistence is a term often used to express the view that the two great
scientific paradigms of the twentieth century, quantum theory and relativity, can

1
S stands for scattering.
18 Questions and Answers

be reconciled without modification to either, and that neither is subordinate

to the other. The problems encountered in failed attempts to reconcile these
excellent theories arise because of a belief that they can each be explained in
reductionist terms. Our view is that each is an aspect of emergent physics, distin-
guished by their contexts. To date, these contexts appear to be nonintersecting.
General relativity (GR) holds as a classical description of processes involving
spacetime and matter on large scales, while QM is an observer-centric description
of processes where phase degrees of freedom are under the control of observers.
Denote the set of empirical contexts where GR appears necessary by G, and
the set of empirical contexts where QM appears necessary by Q. Then the only
place where a unified approach might be needed would be the intersection G ∩ Q,
assuming it is not empty. Currently, little is known empirically about such an
intersection. Specifically, there seems to be no way at this time of empirically
testing any of the proposed theories of quantum gravitation.
One area of great current interest is in the question of Beckenstein–Hawking
radiation from black holes. If detected, such radiation would “merely” serve to
confirm peaceful coexistence and not some incompatibility between GR and QM.
Black hole thermodynamics, as generally formulated, involves QM on a classical
curved background spacetime, is contextually incomplete (because observers are
inadequately treated), and is therefore not the fully quantized theory of gravita-
tion that many theorists would like to have.
We take the view in QDN that GR and QM are emergent descriptions in
physics involving different empirical contexts, so there is no clash of principles.

2.5 Questions and Answer Sets

In physics, questions are asked by observers of states of SUOs. These states will
typically be represented by elements of spaces such as Hilbert spaces. We shall
use the question and answer notation introduced at the start of this chapter.
We shall represent a given state of some classical or quantum SUO2 S by a bold
symbol such as Ψ and represent any question that we ask of that state by a barred
bold symbol such as Q. Then our convention is that the answer to question Q
asked of state Ψ will be denoted by QΨ and referred to as a contextual answer .
Now the whole point of asking a question is to reduce an observer’s uncertainty.
Before a question Q is asked of state Ψ, the observer (the entity asking the
question) will believe that the answer (still to be obtained) is one element of a
set of possible answers. We shall call this set an answer set and denote it by
A(Q|Ψ).
In principle, answer sets may be quite general. They could be sets of sentences,
sets of numbers, or sets of symbols. An answer set may even be empty or consist

2
By this we mean whether we have decided to use classical mechanics (CM) or QM in our
discussion of the experiment.
 2.6 Answer Set Collapse 19

of an infinite number of real or complex numbers. In this book, we shall focus

primarily on answer sets based on real or complex numbers.
A critical, implicit factor in any question and answer process is the role of
contextuality, that is, the circumstances under which a given proposition may
be true or false. Observers do not receive fully formed answers directly from
their equipment, such as “the particle landed here on the screen.” Rather, an
observer will look at a screen, either optically or via electronic equipment, and
notice for example that all pixels except one are white. That exceptional pixel
requires interpretation by the observer. A typical interpretation would be that it
represented the impact site of a particle. That would be valid, however, only if
the context of that particular experiment made that a reasonable inference. For
instance, the observer might already have determined that the exceptional pixel
in question was faulty and could not pick up any signals whatsoever, so did not
represent a particle impact.
Context has a critical role, perhaps the most critical role, in quantum physics.
In contrast, classical mechanics is based on the realist concept of noncontextual-
ity, which asserts that SUOs “have” physical properties regardless of how they
are observed or not observed.

2.6 Answer Set Collapse

Time plays a critical role in the concept of question and answer. Before we ask
a question Q of some state Ψ, the range of possible answers is generally two
or more. Having less than two possible answers in an answer set is of limited
practical value in general.
On the other hand, after we have obtained the answer value A ≡ QΨ, which
is some element of the initial answer set A(Q|Ψ), our uncertainty has completely
disappeared. That disappearance is accompanied by the disappearance, in our
minds, of the original answer set. The transition A(Q|Ψ) to the singleton set
{QΨ} is a completely natural result of the questioning and answering process.
Unfortunately, it has been objectified by some quantum theorists as the sudden
“collapse” of a quantum wave function or state and asserted as something to be
concerned about.
The reason for this concern has to do with the interpretation of the wave
function/quantum state vector and the role of time in QM. In general, questions
cannot always be asked nondestructively. For example, to determine a person’s
blood type, some blood has to be taken. One of the fundamental differences
between CM and QM is that in CM, it is generally asserted that questions can
be asked of states of SUOs without affecting those states in any way. That is not
always the case in QM.
The problem about wave function collapse, or state reduction, can be under-
stood as follows. Suppose we ask a question Q1 of a state Ψ of an SUO and
get the answer Q1 Ψ. There is no controversy about that, neither classically nor
20 Questions and Answers

quantum mechanically. Now suppose that we wanted to ask another question, Q2

of that state, after we have asked Q1 . There are three possibilities to consider.

Null Tests
The question Q1 could be completely nondestructive, giving an answer yet leaving
the original state unaltered. This scenario is assumed throughout CM, classical
questions generally being assumed completely nondestructive. In standard QM
this scenario is called a null test (Peres, 1995), being described by a state that
is an eigenstate of the observable concerned. In his famous book on QM, Dirac
discussed this scenario in terms of polarized light passing through two or more
polarizing crystals (Dirac, 1958). If an unpolarized monochromatic beam of light
is passed through a polarizing crystal, two polarized beams of light are observed
to emerge. If either of these beams is then passed through an identical crystal
with the same orientation as the first, that beam then goes through the second
crystal without further splitting.
A spectacular example of a null test was discussed by Newton in his astonishing
book Opticks published in 1704 (Newton, 1704). Newton showed that if a beam of
sunlight were passed through a prism P 1 , then it would be split into a set of sub-
beams of different colors, defining a spectrum. If that spectrum was then carefully
focussed onto a second prism P 2 , then P 2 would undo the action of P 1 , resulting
in a reconstituted beam emerging from P 2 that had the same spectral properties
as the initial incident beam. In essence, the combination P 1 followed by P 2 acts
as a null test on the original incident beam. This experiment is discussed in detail
in Section 11.4.

Partial Destruction
It could be the case that Q1 is partially destructive, also known as a measurement
of the first kind (Paris, 2012). After the answer Q1 Ψ has been obtained, the SUO
is then in some altered state Ψ different from the original state Ψ. Any attempt
now by the observer to ask question Q2 would then be extracting not the answer
Q2 Ψ but the answer Q2 Ψ .

Total Destruction
It could be the case that Q1 is totally destructive, also known as a measurement of
the second kind , or demolitive measurement (Paris, 2012). The process of asking
Q1 totally destroys the original state Ψ, so that Q2 could not even be asked.

In the real world, human observers get a lot of information about objects
around them optically, which is by and large a weakly destructive process that
gives the impression of being totally nondestructive. This has led to the classical
conditioning that regards the ordering of questions as not significant. In contrast,
QM is based on the empirical evidence that all observations that extract genuine
information are never nondestructive, Planck’s constant being a manifestation of
that fact. The temporal ordering of questions is significant in QM and is generally
 2.7 Incompatible Questions and Category Errors 21

discussed in terms of the commutation properties of the observables associated

with the questions involved.
Answer set collapse has everything to do with the observer’s state of informa-
tion and nothing to do with any changes in the state Ψ per se. What happens
to a state after a question has been asked is a separate issue and depends on the
experiment and on the dynamics assumed.
Standard QM does not in general discuss the third of the above three scenarios;
what happens to a quantum state when an experiment has finished and the
apparatus has been destroyed or decommissioned is considered outside the remit
of standard QM. In contrast, QDN has an agenda to consider such possibilities.
Such matters are discussed in Chapter 25.
Whether the processes of observation are destructive or not, answer set collapse
will still take place in general, illustrating the QDN view that QM is a theory of
observation and not of objects per se.

2.7 Incompatible Questions and Category Errors

Not all questions can be asked sensibly of all states. For instance, suppose Φ
represents the state of an apple resting on my desk. Then if Q is the question

Q ≡ How far is the Moon from the Earth? (2.1)

it is obvious that QΦ is meaningless. This is an example of an incompatible

question, a mismatch between the concepts used to define the state of the SUO
(in this case, the apple on the desk) and the concepts required to formulate the
question (in this case, the distance between the Moon and the Earth). Another
term for such a conceptual clash is category error .
Experimentalists generally have to spend a great deal of time and resources to
arrange for their questions and their states of SUOs to be compatible. Compat-
ibility is always contextual: a question Q that is compatible with states of one
SUO will usually be incompatible with states of some other SUO. For example,
a photon detector will click when placed in a laser beam, but will not click when
placed in an electron beam.
An important source of incompatibility may be due to the mechanical structure
and functionality of apparatus. This can come about in several ways. First, if a
detector has not yet been built, or cannot be constructed for some reason, then
any attempt to detect a signal is a trivial example of an incompatible question.
We may call this existential incompatibility. It is the source of a basic error in
the interpretation of quantum mechanics. Physics is an empirical subject. If an
experiment cannot actually be done in principle or in practice, then theoretical
conclusions should not be treated as empirical facts. For example, we have
theoretical reasons to believe that no apparatus can be constructed that can
detect simultaneously the exact position and exact momentum of an electron.
Experimentalists such as Afshar have attempted to circumvent such a conclusion
22 Questions and Answers

empirically in the case of two-path interferometry (Afshar, 2005), but the general
consensus at this time is that they have not succeeded (Jacques et al., 2008).3
Suppose a detector does exist in a laboratory. It could be faulty, so that any
attempt to detect a signal would be doomed to failure. Another possibility is that
a detector exists and has registered a signal that the observer knows about (such
as in a teleportation experiment), and then that detector has been deliberately
decommissioned in some way by the observer so as to prevent further signals
being generated. Any further attempts to detect a new signal automatically then
involve a incompatible question.
Finally, even if a detector exists and is working properly, its compatibility will
be contextual on the states being asked. For instance, a photon detector will not
detect neutrinos.
There is a class of questions that appear to be nominally physically based,
but which have an empty answer set under all known empirical circumstances.
This means that there is currently no way known of answering such questions
empirically. For example, the question Is time travel possible? is one such
question. We just don’t know the answer. We may refer to such questions as
speculative. There is a place for them in science, in that they can stimulate
reasonable discussion.
A more disturbing class of question is associated with vacuous theories, which
are theories such that none of their critical propositions can be tested by any
known practical means.4 Such theories are often designated as not even wrong
(Woit, 2006). Examples currently are the Multiverse paradigm, quantum gravity,
and string theory. What is disturbing is not so much that they cannot be tested,
but that some of their advocates claim that they represent real physics simply
because their apparent mathematical beauty (a subjective opinion) is regarded
by their authors as more important than a total lack of critical empirical evidence
at this time (2017) (Woit, 2006). Vacuous theories may well be consistent with
all currently known empirically established facts, but that is not enough to
validate them. A common characteristic is that although they have a generalized
proposition classification (GPC, discussed in Section 2.12) of zero or at best one,
many of their supporters claim that they represent valid quantum physics (which
requires a GPC of three). We disagree. They are speculations and we classify such
theories as mathematical metaphysics.

2.8 Propositions
In this book, a proposition is defined as any string of symbols that may be of
interest to an observer. For example, P 1 ≡ Today is Tuesday, P 2 ≡ XY Z,

3
We are not at all critical of Afshar’s attempts per se. Our view is that such attempts should
always be encouraged. That is what science is all about. There should always be someone
trying to upset the standard paradigms.
4
A critical proposition would be a specific claim made by that theory alone and no other,
that would, if confirmed, establish the scientific worth of that theory.
 2.9 Negation, Context, Validation 23

and so on. It is important to note that, by themselves, with no further context,

propositions such as P1 and P2 are neither true nor false.
Our view of physics will be based on the idea that observers formulate proposi-
tions about the Universe and then perform experiments to find out under which
contexts those propositions can be deemed to be true or false. We cannot at this
time explain why observers do such things, and simply take it as a given.5

2.9 Negation, Context, Validation

In two-valued classical logic, propositions are assumed to have truth values: given
proposition P , classical logic examines whether P is true or false.
To formalize this discussion, we introduce the validation function, denoted V.
Validation is really a question, which is why we put a bar over the symbol V.
Validation questions a given proposition and maps it into the binary set {0, 1},
where the value 0 denotes false and the value 1 denotes true. If proposition P is
false we write VP = 0, whereas if P is true we write VP = 1.
This leads to the concept of negation. Given a proposition P , there is often
an associated proposition, denoted ¬P , called the negation of P . For example,
given P1 is the proposition Today is Tuesday, then ¬P1 is the proposition
Today is not Tuesday.
In classical logic, the relationship between a proposition and its negation is
the rule
VP + V(¬P ) = 1. (2.2)
According to this rule, therefore, we need only establish the validity of a propo-
sition to immediately know the validity of its negation.
There are the following issues to discuss here.

Mathematics Is Not Physics

Eq. (2.2) is a mathematical rule. It is not physics. In physics, we have to be more
careful. A proposition such as P1 ≡ Today is Tuesday may have an obvious,
physically meaningful negation, but a proposition such as P 2 ≡ XY Z does not
have any obvious, physically meaningful negation (but of course, we could always
define it symbolically).

Classical Logic Is Insufficient

Classical binary logic deals with mutually exclusive answers, such as yes or else
no. Unfortunately, we cannot be sure that such logic can be applied to every
form of proposition in physics. For example, Jaynes put the case that classical
probability is an enhanced form of classical binary logic, because in probability
theory, answers are not definitely yes or else definitely no, but a probability of

5
The explanation that humans do physics in order to understand the universe does not
explain why they find themselves needing to do this. No other species seems to have this
ambition.
24 Questions and Answers

yes and a probability of no (Jaynes, 2003). Another example is QM, where the
answers are complex-valued amplitudes.

The Role of Context

Propositions in physics are often by themselves relatively meaningless. For
instance, the proposition Energy is always conserved is too glib and
potentially incorrect. What gives any proposition in physics a meaning and
the possibility of a truth value is context, by which we mean the circumstances
involved in establishing that truth value.

This last point is critical in QM. The rules of logic in QM have to be carefully
respected because they are not quite the same as those of CM. The reason has
to do with contextuality, which is not factored into classical logic in general. In
particular, we should not make the following error of physics. Suppose that P is
a physical proposition and we have constructed an apparatus A that allows us
to conclude that P is true. Then we may write

V(P , A) = 1. (2.3)

In other words, we have factored into our notation the context (apparatus A)
that is needed to empirically confirm the truth of P . In words, we would express
(2.3) as the statement that proposition P is true relative to context A.

Example 2.1 In CM, suppose P ≡ Energy is conserved, A1 is a closed

system, and A2 is an isolated system. Then by definition

V(P , A1 ) = 0, V(P , A2 ) = 1. (2.4)

2.10 Proof of Negation

Suppose that experiments showed that V(P , A) = 0. Does this imply that
V(¬P , A) = 1?
The answer is that in physics, we cannot assume that

V(P , A) + V(¬P , A) = 1, (2.5)

because the apparatus A that allows us to test for the truth of P might be com-
pletely useless to test for the truth of ¬P . For instance, suppose P is the propo-
sition Supersymmetric particles exist. To date (2017), experiments at the
Large Hadron Collider have found no sign of such things. That does not imply the
truth of the negation of P , which is ¬P ≡ Supersymmetric particles do not
exist. In order to prove empirically that such particles did not exist, we would
have to find apparatus A such that V(¬P , A ) = 1. It is most unlikely that such
apparatus could ever be constructed, because proof of a negation is very often
impossible (we cannot prove that fairies do not exist, for instance).
 2.12 Generalized Propositions and Their Classification 25

We conclude that the rule (2.2) that holds in classical two-valued logical cannot
always be replaced in physics by the rule

V(P , A) + V(¬P , A ) = 1, (2.6)

simply because A cannot always be constructed. The rule then in physics is to

say nothing about ¬P . Lack of evidence for a proposition is not evidence in favor
of its negation.

2.11 A Heretical View of Reality

Apparatus defines context, and propositions in physics can only be tested by
experiments based on apparatus. Therefore, we could logically take the view
that scientific propositions have no meaning beyond the experiments that test
them. What if the recent “discovery” of the Higgs particle was and will be the
only time in the history of the Universe that evidence for the Higgs particle was
ever found? We could then make the case for the assertion that the apparatus
and its context had created/defined the object of interest, rather than helped in
its “discovery.”
Of course, such thinking goes against the grain of the physicists’ human con-
ditioning, that generally imagines that the Universe is “out there”, waiting to be
explored. That is the same line of thinking that led to Plato’s theory of forms,
which asserts that mathematics “exists” in its own realm of reality and that math-
ematicians stumble upon pieces of it (discover preexisting truths) every so often.

2.12 Generalized Propositions and Their Classification

The discussion in the preceding sections suggests that empirical physics involves
a great deal of context that is frequently understated in discussions. In fact, the
situation is generally worse, as we now suggest, because only the apparatus has
been factored into the above discussion, with no mention of any observer. The
solution we have developed is to introduce the concept of generalized proposition
and a classification scheme for generalized propositions, as follows (Jaroszkiewicz,
2016).

Definition 2.2 A generalized proposition (GP) P is of the form

P = (P , Cint |O, Cext ), (2.7)

where P is a proposition, Cint is the relative internal context that an observer

O, defined by the relative external context Cext , can use to test P .

Relative internal context includes apparatus and the protocols (including

axioms and theory) used to test propositions. Relative external context is
generally classical information that establishes the relationship between the
26 Questions and Answers

observer and the wider environment/universe in which they are performing their
experiments.
A GP addresses the issues we raised in Section 2.1: ideally, any GP gives
information about who is making the proposition, what that proposition is,
and how the truth value of that proposition can be determined. This raises the
following two points.

Identification
We should always ask the question:
For whom is this GP relevant?
The answer is: for the observer involved and not necessarily for anyone or any-
thing else. It is, after all, an essential requirement of observation that an observer
understands their own relationship to their environment. Even in CM, frames of
reference have to be identified. No experiment makes sense if any aspect of the
GP concerned is unknown to the observer concerned.

Missing Context
What happens if some or all of the contextual information in a GP is missing?
We address this fundamental point now.

Definition 2.3 A GP is contextually complete if both relative internal

context and relative external context have been supplied.

The problem is, there are various degrees of contextual completeness. We

propose the following schema to classify any GP, allowing us a mechanism to
decide whether a proposition is scientifically useful or not.

Definition 2.4 Given a GP P ≡ (P, Cint |O, Cext ), its generalized proposition
classification (GPC) #P is given by

#P = α + 2β, (2.8)

where
 
0, Cint = ∅ 0, Cext = ∅
α= β= (2.9)
1, Cint = ∅, 1, Cext = ∅,
where the empty set symbol ∅ denotes missing context.

With this schema, we classify several classes of generalized proposition as

follows.

Metaphysical Propositions
These are completely contextually incomplete GPs of the form
(P , ∅|∅, ∅), (2.10)
 2.12 Generalized Propositions and Their Classification 27

where ∅ denotes a complete absence of any contextual information whatsoever.

Metaphysical propositions cannot be validated, so the term not even wrong
applies to them (Woit, 2006). A metaphysical proposition has a GPC of zero.

Mathematical Propositions
These are partially contextually complete GPs of the form

(P , A|∅, ∅), (2.11)

where A is a set of axioms relative to which the truth status of proposition P

can be established. A mathematical proposition has a GPC of 1.

Classical Propositions
These are partially contextually complete GPs of the form

(P , ∅|O, Cext ), (2.12)

where O is a primary observer described by relative external context Cext , such

as a statement of the rest frame of the observer. However, there is generally no
mention of the apparatus used in the experiment, so a classical proposition has
a GPC of 2. Such GPs are common throughout CM.

Complete Propositions
These are contextually complete GPs of the form

(P , A|O, Cext ), (2.13)

where A is a specification of the apparatus that can be used to establish the

relative truth status of the proposition P. It is not enough to construct any
apparatus; A has to be compatible with P , allowing the truth value of P to be
determined. It is not enough, either, to simply assert that such apparatus exists
or could be built.
A complete proposition is a GP that has a GPC of 3.

QDN is designed to deal with complete propositions: relative internal context

is explicitly modeled by quantum registers directly related to compatible appa-
ratus, with an endophysical observer assumed to be operating in a well-defined
laboratory running through a succession of temporal stages.
 3
Classical Bits

3.1 Binary Questions

A binary question is a yes or no question. A typical binary question Q will be
of the form Q ≡ Is it true that?, asked of some proposition P such as
P ≡ There is a signal in this detector.
Assuming compatibility of question and proposition, the answer QP will be
either one or zero, interpreted as yes and no, respectively.
The answer set associated with any binary question has two elements, denoted
0 (zero) and 1 (one). Usually the element 0 will be contextually interpreted as
no and element 1 will be interpreted as yes, but we could choose to interpret the
answer set elements the other way around.
We will base most of this book on the mathematical structures associated
with binary questions. Classical binary questions are discussed in this chapter
and their quantum analogues discussed in Chapter 4.

Example 3.1 A trial in an English or American court of law can be thought

of as a process that answers the binary question Q ≡ Is it true that? of the
proposition P ≡ This person committed that crime.

The negation ¬Q of a binary question Q ≡ Is it true that? is defined to be

¬Q ≡ Is it false that?.

3.2 Question Cardinality

The English and American legal systems are based on the ancient Roman prin-
ciple of in dubio pro reo (when in doubt, judge in favor of the accused). This
presumption of innocence gives a binary, or dichotomic, flavor to the proceedings:
assuming a verdict is reached, then verdicts in such courts can only be either
guilty, corresponding to yes, or else innocent, corresponding to no.
 3.4 Classical Bits 29

In Scotland, however, there is a third possible verdict, known as not proven.

This and other examples leads us to define the question cardinality #Q of a
question Q as the number of elements in its answer set, before that question
is asked. According to this definition, a rhetorical question has cardinality 1, a
binary question has cardinality 2, and a Scottish law trial answers a question
with cardinality 3.
The cardinality of a question is contextual, in that it depends on the compat-
ibility of a given question and a given proposition or state.

3.3 Classical Binary Questions

Binary questions (yes/no questions) are the fundamental building blocks of the
quantized detector network (QDN) approach to physics discussed in this book.
The reason has to do with the way experiments are conducted. Although modern
experimentalists acquire vast amounts of data electronically and interpret them
in terms of sophisticated theories, what goes on at the most basic level in any
experiment is that a number (possibly vast) of binary questions are answered.
For instance, the question of whether a photon detector has clicked or not is a
binary question.
Another reason for choosing to work with binary questions is the principle that
any empirical question, no matter how complicated, can always be expressed in
terms of some number (possibly infinite) of binary questions. This will be referred
to as the bitification principle. We shall return to this topic in Chapter 5 as it
leads directly to the concepts of classical and quantum registers. These are the
central mathematical constructs in terms of which the ideas of this book are
expressed.
For the rest of this chapter we shall review some of the properties of single
binary questions in classical mechanics. In that context, they are referred to as
classical bits, terminology that comes from the theory and application of classical
computation. When quantum rules are factored in, binary questions become
quantum bits, or qubits. These are discussed in the next chapter. A fundamental
difference (but not the only significant difference) between classical bits and
qubits is that the former have cardinality 2, whereas the latter have infinite car-
dinality in a particular sense to do with the interpretation of quantum mechanics.

3.4 Classical Bits

We turn now to the mathematical representation of binary questions known as
bits. A classical bit, or bit for short, is a set with two elements denoted 0 and 1.
The word bit is short for binary digit. The first recorded use of the word in
this context has been dated to 1947 in a Bell Labs memo written by John W.
Tukey. It first appeared in public in 1948 in Claude E. Shannon’s landmark paper
on information theory (Shannon, 1948). Bits are used extensively in classical
30 Classical Bits

computation and information theory and are central to our implementation of

quantized detector networks.
Bits represent the most elementary, useful form of data variable, one with only
two possible, mutually exclusive values. But that is a mathematical statement,
insufficient for our purposes. In the sciences, including the theory of computing,
a bit is generally more than a mathematical set with two elements; bits generally
have some associated context that gives some physical meaning or interpretation
to each of the two possible values. For example, the two elements of a bit in two-
valued logic might be thought of as true and false. With context, mathematical
data becomes physical information.

3.5 Signal Bits

The sort of experiment we are mainly interested in this book typically involves
one or more single-click detectors, each detector having two possible outcome
states. These states are associated with a classical bit as follows. If a detector is
found in its ground state, or no-click state, then that state is represented by the
bit state 0, whereas if that detector is found in its signal state, or click state,
then that state is represented by the bit state 1.
Single-click detectors should not be thought of as simple. Typical detectors
such as Geiger counters involve cascade processes that are irreversible and com-
plex. What is significant is that, like avalanches on mountainsides, they are
triggered by the smallest of effects, and that is where their value lies.
A signal bit B ≡ {0, 1; CB } is a set with two elements, denoted 0 and 1,
together with a context CB that gives each element contextual physical signifi-
cance relative to the observer involved. A signal bit state, or bit state for short,
is any one of the two elements, 0 or 1, of a signal bit.
Complexity rules the real world and experiments are vastly complex processes
involving many degrees of freedom. Signal bits are best thought of as equivalence
classes defined by a context with two clear alternatives. For example, in the
standard quantum mechanics (QM) description of the Stern–Gerlach (SG) exper-
iment, electrons with arbitrary momentum and position but with spin down are
identified with bit state 0 while electrons with arbitrary momentum and position
but with spin up are identified with bit state 1. Typically, these equivalence class
will not depend on the color of the observer’s shirt or other factors regarded by
the observer as inessential to the experiment. Such equivalence classes are defined
by the observer’s chosen criteria, which may appear subjective.
In the original SG experiment, the detecting screen was a photographic emul-
sion film, acting as a battery or register of signal detectors (Bernstein, 2010). At
the start of the experiment, the film was prepared by the observers by blowing
cigar smoke onto it. Then, after many silver atoms had passed through the main
magnet of the SG device and impinged onto the screen, the observers noticed that
there were two relatively crude but nevertheless separate spots where the bulk
 3.6 Nodes 31

of the atoms had landed. It was the observers’ decision to interpret these spots
as of empirical significance. Stern and Gerlach did not know that their observed
spots were a marker of electron spin. In 1922, electron spin was unknown; Stern
and Gerlach were trying to validate Bohr’s Old Quantum Mechanics theory of
the atom. The electron spin interpretation came into focus a few years later.
Classical bits are sets, not vector spaces, because there is no obvious math-
ematical or physical meaning to the multiplication of a bit state by a real or
complex number, or to the addition of two bit states. This is no longer the case
when we generalize bits to their probabilistic counterparts, where they are known
as stochastic bits (or s-bits) and to their quantum counterparts, where they are
known as quantum bits (or qubits). Nevertheless, it is useful and convenient
to represent bits via two-dimensional complex vector spaces. This allows us to
represent bit states as vectors and bit operators as matrices, but it should be
kept in mind that classical bit states cannot be added in principle. An exception
is in Boolean algebra, where rules such as 1 + 1 = 0 have a contextual meaning.

3.6 Nodes
QDN analyzes experiments in the simplest form possible, which is in terms of
binary questions and answers. We will show in later chapters how any given
apparatus is represented in QDN as a collection of binary questions and answers
forming a stage network , a collection of nodes connected by links across which
quantum information is transmitted.
Nodes come in two forms, external and internal .

External Nodes
External nodes correspond to physically existing equipment and come in two
varieties: sources (preparation devices) and detectors. Observers input contextual
information into stage networks via sources and extract signal information via
detectors.

Internal Nodes
These occur in the information void , the region of space and time where no
information is extracted. Internal nodes can be thought of as virtual detectors,
as they usually correspond to places in a network where a real detector could
have been placed, if the observer had so chosen.

Example 3.2 In the double-slit experiment, the two slits are identified with
internal nodes, while the source of the incoming beam and the detecting screen
are identified with external nodes.

Nodes are not necessarily localized in space. An example involving highly

nonlocalized nodes would be an apparatus for the measurement of particle
momentum.
32 Classical Bits

While it will always be easy to identify external nodes in any stage diagram,
that will not always be the case for the internal nodes. The rule for assigning
internal nodes is that any such node represents a potential opportunity for
information extraction, if the observer so chooses. So for instance in the double-
slit experiment example considered above, not only could the observer place
detectors at any of the two slits, but the observer could conceivably fill the space
between source and screen with a vast number of internal nodes, if that was
needed. In the limit of extremely large numbers, the QDN description would begin
to look more like quantum field theory. The art in QDN is to find the simplest
possible description of an experiment in terms of a limited number of nodes.

3.7 Dual Bits

In anticipation of subsequent developments, we introduce here the notion of a
dual bit. For every bit B ≡ {0, 1} we postulate the existence of another bit,
denoted B ≡ {0, 1}, referred to as the dual bit, or simply the dual, of bit B. The
two elements of a dual bit will be referred to as dual bit states.
Bits and their duals are related as follows: 0 is the dual of 0 and 1 is the dual
of 1. In anticipation of subsequent developments, we define a function ij from
the Cartesian product B × B into the set Z ≡ {0, 1} by the rule
ij ≡ δ ij , i, j = 0, 1, (3.1)
where i is an element of B, j is an element of B, and δ ij is the Kronecker delta.
This notation will be extended to classical registers, that is, collections of
bits. If B k ≡ {0k , 1k } is the kth bit in a register, then its dual B k is given by
B k ≡ {0k , 1k }. Then rule (3.1) becomes ik j k ≡ δ ij .
If k = l, then ik j l is undefined. The interpretation of this is that an observer
cannot expect to extract any information from one detector by looking at any
another detector. Exceptions can occur, provided the right context is in place,
such as charge conservation.

3.8 The Interpretation of Bits and Their Duals

In the previous chapter we discussed the role of questions and answers in physics,
and in this chapter started to link this to the notion of a detector. We now tie
in these two ideas with bits and their duals as follows. The two bit states 0, 1 in
a bit do not individually “have” absolute truth values per se: such truth values
are only contextual, relative to the questions asked of the associated detector.
The two questions that could be asked of a detector are
0 ≡ Is this the ground state?,
(3.2)
1 ≡ Is this the signal state?.
It should be now clear why we have chosen our bit state notation in the given
form. In our notation, the question i asked of the bit state j is written in the
 3.10 Classical Bit Operators 33

form ij, and then the answer is given by δ ij . If δ ij = 0 (the number zero), then
that means that the answer is no, whereas if δ ij = 1 (the number one), then that
means that the answer is yes. We shall refer to each side of expressions such as
(3.1) as a classical answer .

3.9 Matrix Representation

We introduce here a convenient matrix representation of bits and their duals.
The rule is that bit states are represented by two-component column matrices as
follows:
   
1 0
0= , 1= , (3.3)
R 0 R 1

where = denotes “is represented by.” For the dual bits, we have the row matrix
R
representation
   
0 = 1 0 , 1= 0 1 . (3.4)
R R

There is a small technical point concerning this representation that we clarify

now. The classical answer ij is a number, either zero or one. However, according
to the rules of matrix multiplication, the action of a two-dimensional row matrix
on a two-dimensional column matrix is a 1×1 matrix, not a number. For instance,
 
  0  
01 = 1 0 = 0 = 0. (3.5)
R 1
We resolve this problem by interpreting the left-hand side of (3.5) as the compo-
nent of the 1 × 1 matrix on the right-hand side. Henceforth we shall ignore this
technical point.

3.10 Classical Bit Operators

The process of asking a binary question i of a bit state j gives the answer ij,
which is a number (either zero or one). The ordering here is significant: the
binary question i is to the left and the bit state j is to the right. It turns out
to be useful to define objects known as transition bit operators, which have the
ordering interchanged, forming an object known as a dyadic. There are four such
operators, defined as T ij ≡ ij for i, j = 0, 1. The application rules of these
operators are as follows:

Action on Bit States

The action of T ij on bit state k is from the left and is given by

T ij k ≡ (ij)k ≡ i(jk) = δ jk i. (3.6)

34 Classical Bits

Action on Dual Bit States

The action of bit operator T ij on dual bit p is from the right and is given by
pT ij ≡ p(ij) ≡ (pi)j = δ pi j. (3.7)
The transition operators have the following matrix representations:
       
00 1 0 01 0 1 10 0 0 11 0 0
T = , T = , T = , T = . (3.8)
R 0 0 R 0 0 R 1 0 R 0 1

These four matrices form a basis for the four-dimensional vector space of complex
2 × 2 matrices, so we can use them to construct other useful matrices. There are
four bit operators, labelled I, F , D, and U here, that are occasionally useful.
They are defined and represented as follows.

The Bit Identity Operator I

This operator is defined as I ≡ T 00 + T 11 . It leaves bit elements unchanged, i.e.,
I0 = 0, I1 = 1. (3.9)

The Bit Flip Operator F

This operator is defined as F ≡ T 01 + T 01 . It switches bit elements, i.e.,
F 0 = 1, F 1 = 0. (3.10)
In quantum computation, F is known as the NOT gate and denoted X (Nielsen
and Chuang, 2000).

The Bit Down Operator D

This operator is defined as D ≡ T 00 + T 01 . It forces all bit states into the ground
state 0, i.e.,
D0 = 0, D1 = 0. (3.11)

The Bit Up Operator U

This operator is defined as U ≡ T 10 + T 11 . It forces all bit states into the signal
state 1, i.e.,
U 0 = 1, U 1 = 1. (3.12)

The four operators I, F , U , and D will be used in bit state dynamics, and
then it will be convenient to define O 1 ≡ I, O 2 ≡ F , O 3 ≡ D, and O 4 ≡ U .

3.11 Labstates
Our objective in this book is to interpret quantum mechanics via signal states
of apparatus instead of states of systems under observation (SUOs). In order to
keep this in mind, we shall use the term labstate whenever we refer to the former,
reserving the term system state for the latter.
 3.12 Time and the Stages Concept 35

In this book, we shall deal with three kinds of labstate associated with a
single detector: classical labstates, stochastic labstates, and quantum labstates.
Classical and stochastic labstates are discussed in this chapter, quantum labstates
are discussed in the next chapter. Each form of labstate has its own dynamical
evolution rules, which we shall discuss separately in some detail. Because only
one detector is involved in these preliminary discussions, such dynamics will
be referred to as rank one. If two detectors were involved, then we would be
discussing rank two dynamics, and so on.
Given a single detector, the observer would find it either in its signal ground
labstate 0 or in its signal labstate 1, assuming the detector existed, was not
faulty, and that the observer actually looked.

3.12 Time and the Stages Concept

Before we can discuss the QDN approach to dynamics, we need some more
precision in our modeling of the processes of observation, because detectors are
distributed not only in space but also in time. Time is a necessary ingredient in
our discussion. Once we start to incorporate that element into the discussion, we
are led naturally to the stage concept that underpins QDN.
In QDN, time is defined relative to an observer and is generally measured
in discrete steps called stages. This requires some explanation. It does not mean
that we have dispensed with time as conventionally modeled in standard physics,
that is, as a continuous real number–valued parameter via which velocities and
other temporal derivatives are calculated.
Contrary to what is implied in conventional formulations of quantum mechan-
ics, such as Schrödinger wave mechanics, the time in the laboratory required
to complete any observation of a signal state is always nonzero. There are in
fact no actual continuous time measurements possible in physics. Any references
to continuous time observations, such as in quantum Zeno (also referred to as
nondemolition) experiments (Itano et al., 1990), are to contextually incomplete
mathematical approximations that often have great validity and usefulness, but
only up to a point and under specific assumptions. Continuous time has much
the same status in experimental physics as the concept of temperature: a useful
and powerful emergent concept representing a great deal of contextuality, but
otherwise not an objective thing in its own right.
In QDN, the concept of events in continuous spacetime is replaced by the
concept of stage network.

Definition 3.3 A stage network is a conceptual collection of external

and internal nodes representing apparatus distributed over time and space
in a laboratory. Each node is connected by temporal links to other nodes
or to modules. Modules represent processes between nodes that influence
36 Classical Bits

the transmission of classical or quantum information. Nodes are indexed by

integers called labtimes. Each labtime is associated with a stage, the QDN
analog of a hypersurface of simultaneity in relativity.

Quantum state preparation occurs at the source nodes and generally takes
place over some contextually “small” or negligible interval of labtime (time as
measured in the laboratory). States then evolve undisturbed over temporal links
and are detected by the observer at the detectors, again over contextually “small”
or negligible intervals of labtime.
The power of QM in general is that detailed modeling of what actually goes
on at the nodes seems to be less significant than the detailed modeling of the
transition amplitudes evolving over temporal links. For example, Feynman dia-
grams are used in relativistic quantum field theory to calculate those amplitudes,
with virtually no modeling of the detection equipment that would be needed in
practice.
This view of quantum processes was taken to an extreme with the development
of the Multiverse paradigm (Deutsch, 1997). In that paradigm, only evolution
of the wave function for the Multiverse in the information void is asserted to be
significant. But because real, empirical information is extracted in the laboratory
at nodes only, it should not come as a surprise that the Multiverse concept turns
out to be empirically vacuous. So where does QDN stand in relation to these
nodes and links?
QDN is an attempt to investigate the nodal aspect of quantum physics more
than has been hitherto the case. As with scattering matrix (S-Matrix) theory
(Eden et al., 1966), QDN provides a framework for discussing the spatiotemporal
architecture of observation but does not provide the dynamical details of the
amplitudes involved.
Our aim in this book is to discuss a multi-detector approach to quantum
physics. By this we mean to discuss real experiments that involve perhaps many
preparation channels; large numbers of modules such as beam splitters, mirrors,
and suchlike; and batteries of outcome detectors. In such circumstances, we
naturally find ourselves encountering the dictates of special and general rela-
tivity (GR). To date there has been no empirical evidence that the principles
of relativity and of quantum mechanics are incompatible. There is in practice
“peaceful co-existence” between GR and QM, and that has to be respected in
QDN.
The grouping of nodes into sets called stages reflects classical causality, the
notion that an event can influence some events dynamically but not others. In
relativity, events outside each other’s light cones cannot be causally related. The
analogous concept in QDN is that nodes in the same stage cannot transmit or
receive quantum information from each other.
 3.12 Time and the Stages Concept 37

There is a subtlety here, however, to do with shielding. This needs some expla-
nation. Consider two detectors, A and B, that are, in relativistic parlance, time-
like separated. This means that one of the detectors, say B, is inside the forward
light cone of A, viewed from the conventional relativistic perspective. Therefore,
by standard causal physics, B could in principle be affected dynamically by
whatever was done at A. But suppose B was shielded in some way from any
dynamical effects from A (such as being placed inside a Faraday cage, in the
case of electromagnetic interactions). Then for all practical purposes, we could
regard A and B as if they were dynamically independent. It would not then
be inconsistent to assign them to the same QDN stage, even though they were
not on any hypersurface of simultaneity in physical spacetime. On this basis the
QDN definition of simultaneity is contextual.

Example 3.4 Consider an SG experiment where an electron passes through

an inhomogeneous magnetic field and is expected to land on one of two possible
sites on a screen. Provided there was no tampering with the screen after the
electron had passed through the magnetic field, then the observer could take
their time in looking at the two sites to see where the electron actually had
landed. The observer could in fact look at one site immediately after the
electron had passed though (that is, after it had been calculated to have
passed through), and then look at the other site 20 years later. The acts of
looking at the two sites would take place 20 years of real time apart, but
provided the screen had not been tampered with over those 20 years, the two
site examinations could legitimately be regarded as having taken place in the
same stage.
In fact, all experiments are conducted in this way. Signals are registered
irreversibly in detectors, and the observer generally looks at those memories
usually much later.

The stage concept is designed to reflect the inherent certainty/uncertainty

dichotomy in any experiment: an observer may be quite sure that a signal has
been detected in a detector (simply because they looked and found a signal),
but the actual exact laboratory time when the signal was triggered could be
quite uncertain. Indeed, it is a vacuous concept to imagine that signals trigger
instantaneously. How could that be proved? At best, approximate time intervals
of triggering could be determined.
The stage concept is naturally tuned in to the notion of wave-function collapse,
or state reduction. Nothing physical actually collapses when an observer looks
at a detector and finds a positive signal there: it is true that there will be some
quantifiable changes, both in the apparatus and in the observer’s information
store, but these changes are not manifestations of anything that happened in the
information void, merely interpreted as evidence that something had happened.
38 Classical Bits

S W B

Figure 3.1. A typical stage diagram. Dotted lines represent individual stages,
labeled by subscripts, the nth stage being denoted Σn . Circles represent nodes,
where information either enters or leaves the network, or passes on to other
nodes. Shaded circles represent actual outcome detectors. Boxes represent
modules such as the source (S), a Wollaston prism (W ), a phase-changer (φ),
a mirror (M ), and a beam splitter (B). Solid lines represent transmission in
the information void between nodes and modules.

3.13 Stage Diagrams

All experiments have a spacetime architecture that can be represented dia-
grammatically. In QDN, we use stage diagrams. These are simplified diagrams
showing the information flow between components of apparatus over the course
of an experimental run. Figure 3.1 is a typical stage diagram. Numbered circles
represent individual information gates or nodes, either real or virtual (explained
later). Shaded circles represent real detectors, that is, nodes that the observer
actually extracts signal information from. Boxes represent various modules, such
as mirrors and beam splitters. Modules are discussed in Chapter 11. Dotted lines
represent the stages, indexed by subscripts.

3.14 Measurements and Observations

To explain more fully the points we are making, we first need to pin down some
of the terms we shall use to describe any experiment.

Intervention
An intervention is a single act of information extraction from a set of detectors
at a single stage.

Run
A run is the complete process involved in a given sequence of actions, which
starts with initial labstate preparation and ends with final labstate detection.
 3.15 Transtemporal Identity 39

Experiment
An experiment consists of a given number of runs, each following an identical
protocol, or experimental procedure.

Runtime
Runtime is time required to perform a given run, as measured by the observer’s
laboratory clocks.

Measurement
A measurement is the statistical result of data accumulated over one or more
runs of a given experiment.

3.15 Transtemporal Identity

With the introduction of time, we come to an important question in the devel-
opment of our formalism: does an detector have a transtemporal identity (an
identity that persists over some interval of time), or is it something that exists
only at a specific time?
Such a question goes to the heart of an ancient debate concerning the nature
of reality. In The Way of Truth, a surviving fragment from a poem of the
ancient Greek philosopher Parmenides (ca. 520–450 BCE), it was argued that
change is impossible and that existence is timeless, uniform, and unchanging.
Parmenides also suggested that the world of appearance (or by our interpre-
tation, observation) is false and deceitful. He claimed that truth could not be
known through sensory perception and that only pure reason could lead to a
proper understanding of reality. In essence, this was an argument for not doing
experimental physics.
Parmenides’s ideas had an enduring effect on subsequent philosophy, physics,
and mathematics. One of his pupils, Zeno of Elea (ca. 490–430 BCE) took
Parmenides’s line of reasoning further and created a number of paradoxes about
motion, such as the race between the Tortoise and Achilles. Zeno’s paradoxes
could not be fully resolved until the modern mathematical understanding of the
limit concept was developed. Parmenides also denied the existence of nothingness,
or the void, which stimulated Leucippus to propose the existence of atoms.
In contrast to Parmenides, Heraclitus suggested that everything flows, nothing
stands still, and that change is the only constant.
We can make sense of some of the ideas of Parmenides, Leucippus, and Hera-
clitus provided we avoid introducing absolute truths and base our discussion on
contextuality, the proper basis for theories of observation of reality.
Any answer to the question of temporality determines the way in which dynam-
ics is thought about and represented. If SUOs and/or detectors have enduring
temporal identities, then it is reasonable to imagine that they “evolve” over time
in fixed spaces, changing perhaps their states but retaining sufficient attributes to
40 Classical Bits

justify giving them their particular identities. This is related to the phenomenon
of persistence, which depends on the time scales over which objects can be taken
reliably to have significance.
Such an approach is taken in conventional classical mechanics (CM), where
SUOs are represented by points moving about phase space, and in QM, where
state vectors move about in Hilbert space.
The alternative to this is to imagine, like Heraclitus, that everything changes,
nothing persists. According to this view, objects such as SUOs and apparatus
only appear to persist because sufficient patterns of mass and energy are repeated
sufficiently unchanged over certain time scales as judged by some observer, and it
is this that gives that observer the impression that “objects” exist in the universe.
We shall adopt the Heraclitian point of view, because we need to consider the
possibility that observers and apparatus can be created and destroyed. This will
mean changing, over time, the dimensions of the mathematical spaces used to
model processes of observation.
Therefore, when we discuss a rank-one labstate evolving from stage ΣM to
stage ΣN , we shall think of it as a succession of detectors, denoted, say, by
ΔM , ΔM +1 , . . . , ΔN . Each detector Δn is associated with a classical bit Bn and
its dual B n , and that detector exists only at stage Σn . But if persistence is
assumed, then the whole set {Δi : i = M, M + 1, . . . , N } of detectors may be
thought of as a single detector with an enduring, transtemporal identity existing
over the time interval [tM , tN ].
It is important not to mix different theoretical spaces associated with different
times. While we can ask the question in j n for M ≤ n ≤ N , we are not allowed
to ask the question in j m for n = m. The reason is obvious when stated in words:
we cannot observe today what does not yet exist or what used to exist. All actual
observations are done in process time. Past and future are inferred from the data
so acquired: archaeologists do not dig up the past – they dig up traces of the
past embedded in the present. Improper questions such as in j m , n = m, are not
defined mathematically either, in the same way that dual vectors (one-forms) are
defined only by their action on their individual, associated vector space.

3.16 Typical Experiments

A typical rank-one experiment of the classical type starts with a definite initial
labstate j M at initial stage ΣM . Usually we shall take M = 0, but this is not
essential. We shall not ask how this initial labstate was created but, in this
section, require it to be definite; that is, there is no element of probability here
other than certainty. There are therefore only two possible initial labstates in
any such experiment: j M can take the value 0 or the value 1.
Next, we imagine that the observer has no further interaction with “the”
detector until stage ΣM +1 , but by this time, something will have acted on it
to possibly change its signal state. We shall represent this by the action of one
 3.16 Typical Experiments 41

of the classical bit operators O k , k = 1, 2, 3, 4, defined in Section 3.10, i.e., we

suppose the transition
iM → iM +1 ≡ O kMM+1,M iM , (3.13)

where O kMM+1,M is given by


1 
1
O kMM+1,M ≡ aM +1 OkabM bM (3.14)
a=0 b=0

and OkabM are the ab components of the corresponding classical bit matrix. This
operator takes us from the two-dimensional vector space QM in which we have
embedded 0M and 1M into the two-dimensional vector space QM +1 in which we
have embedded 0M +1 and 1M +1 .
This process can be continued. Specifically, for any time n such that M ≤ n <
N we may write
in → in+1 ≡ O kn+1,n
n
in , (3.15)
from which we find
iM → iN ≡ O N,M iM , (3.16)
where the complete evolution operator O N,M is given by
O N,M ≡ O N,N −1 O N −1,N −2 . . . O M +1,M
1 1
  (3.17)
= aN O kN −1 O kN −2 . . . O kM ab bM .
a=0 b=0

If now at time N, the observer stepped in and looked at the detector, they would
normally ask the question 1N ≡ Is there a signal here? The answer is given
by 1N iN , where a value one represents an answer yes while a value zero represents
an answer no.
Each set of integers {kM , kM +1 , . . . , kN −1 } in (3.17) represents a specific “oper-
ator chain” of labstate dynamical changes. Each of the integers kn can be 1, 2, 3,
or 4, so there is in total 4N −M different operator chains. However, because the
multiplication of the operator matrices is closed, the net result is that there are
only four possible overall complete evolution operators, given by

1 
1
O kN,M ≡ k
aN Oab bM , k = 1, 2, 3, 4. (3.18)
a=0 b=0

This form of dynamics is relatively simple but there is a surprising aspect to

it: it is entirely deterministic, in that a given present labstate unambiguously
determines the labstate at any time in the future. However, if any one or more
of the N − M matrices in (3.17) is of D or U type, then the complete dynamics
is irreversible. What this means is that even if the observer knew the final
labstate and every detail of the evolution operator from initial to final times,
42 Classical Bits

but not the initial labstate, they could not say for sure what that initial state
was. Retrodiction is therefore impossible if a D or U transition occurs even
once.
The mathematical reason why the occurrence of just a single D or U transition
generates irreversibility is that each of these maps is two-to-one, and so their
matrix representations are singular.

3.17 Rank-One Stochastic Evolution

In the real world, classical certainty is an exception and we have to use concepts
of probability to discuss most situations. In this section we extend the ideas of
the previous section to incorporate this requirement.
There are two competing philosophies or schools of thought about probability:
the Frequentist school and the Bayesian school . These are in principle quite
different in their core philosophies of what probability means, although some
convergence of thinking appears to be taking place among the experts, and the
differences are at times too subtle to be of much significance to us here.
There is however one clear difference, analogous to the difference between the
CM noncontextual view of reality and the contextual QM view. Frequentists
talk about probabilities as if they were intrinsic to the events taking place, such
as a fair coin “having” a probability of 12 landing on a head. Bayesians require
a context to be supplied before they presume to make such an assertion. In an
absence of such context, Bayesians will make a prior , or educated starting guess,
such as 12 for the probability p of a head. A Bayesian would then throw the coin a
few times and make some outcome observations. With this new information, the
Bayesian would make an updated estimate of p based on a well-known formula
attributed to Bayes.
For Bayesians, probabilities are contextual. For example, a Bayesian who
had thrown a coin 10 times and not observed a single tail would calculate the
probability of the eleventh throw landing on a tail to be much less than 12 .
This is because the 10 observations (i.e., the 10 runs of the basic experiment)
had provided new information about the SUO (the coin) that could not be
discounted. The specific details of any calculation as to the likely outcome of the
eleventh throw would depend on the sort of assumptions made, such as whether
successive throws were truly independent. We shall not discuss those details
further here. Suffice it to say that prior information, or as we would put it, the
context of the eleventh throw, would have significant bearing on the probability
calculation. Also, how the coin was thrown would have an important bearing on
the probability outcome calculations and this is part of the context as well.
The pragmatic view of a person unfamiliar with the rules of probability would
be that the observation of ten successive heads is reasonably convincing evidence
for the hypothesis that the coin is in fact double-headed. On that basis, the
probability of getting a tail on the eleventh throw would be zero. A more careful
 3.18 Stochastic Bits 43

analysis based on particular “reasonable” assumptions about the prior gives a

probability of about 0.95 of getting a head on the eleventh throw.
We see here a similarity between our approach to observation and the Bayesian
approach to probability. In both cases, prior information or context is crucial to
the predictions. Our first principle of observation is that all truths in physics are
contextual. The equivalent principle in Bayesian statistics is that all probabilities
are conditional.
Probability in process physics has a different flavor compared with probability
in block world physics. In block world physics, probability has to be discussed in
terms of limits of ratios of large numbers of outcomes, presumably counted by
some unspecified exophysical observer over some number of runs, or repetitions,
that were embedded in the Block Universe. This is the Frequentist approach to
probability and is typical of the way probability is discussed in standard QM.
In Process Time, however, we may find ourselves in a situation where we
have no more than one opportunity to throw a coin. Real life is usually like
that and it is that aspect of physics that we are trying to develop. Under
such circumstances, the term propensity may be used rather than probability.
Propensity is a gambler’s view of probability as opposed to an accountant’s view.
We shall use the term “probability” to represent both kinds of concept.
Randomness and uncertainty enter into our discussion in two ways: the
observer may be uncertain as to the initial labstate of the detector, and also be
uncertain as to which dynamical operators are acting. We need to discuss both
aspects. We consider first random labstates.

3.18 Stochastic Bits

Previously, we represented the ground state 0 and signal state 1 by column
vectors, as in (3.3). Now suppose the observer was unsure as to which initial
labstate they had started with, to the extent that they could only assign a
probability of p for it to be in the ground state 0 and a probability q ≡ 1 − p to
be in its signal state 1. We shall represent such an uncertain labstate by
 
p
Ψ = p0 + q1 = . (3.19)
R q

Now suppose that the observer wanted to know if there was a signal in the
detector. They need to ask the question
1 ≡ Is the detector in its signal state? (3.20)
In our formalism, the answer is given by
 
  p
1Ψ = 0 1 = q, (3.21)
R q
where we interpret the right-hand side as a number. In this case, the answer is
not zero or one as in the classical case but interpreted as the probability of there
44 Classical Bits

being a signal in the detector. Likewise we find 0Ψ = p, which is the probability

that the detector would be found in its ground state.
The sum of the elements of the column vector (3.19) is unity. Such a vector
will be called a stochastic vector and the associated bit will be called a stochastic
bit, or s-bit. For such bits, we shall refer to expressions such as iΨ as stochastic
answers, as they are generally neither zero nor unity, and are interpreted as
probabilities.

3.19 Left-Stochastic Matrices

We turn now to the other possibility where randomness may occur: the dynamics
affecting an detector may be random. We suppose now that the observer is
uncertain as to which of the four possible evolutions I, F , D, or U actually
has occurred. Consider the operator S defined by
S ≡ pI + qF + rD + sU , (3.22)
where p, q, r, and s are probabilities summing to unity. With the matrices as
defined previously, we find
 
a b
S= , (3.23)
R 1−a 1−b

where a ≡ p + r and b ≡ q + r are in the interval [0, 1].

Such a matrix is a left-stochastic matrix , i.e., one that has the property that
each element lies in the interval [0, 1] and the sum of elements in each column is
unity.

3.20 Stochastic Jumps

We now consider stochastic labstates jumping under the influence of left-
stochastic operators, from stage to stage. The labstate at stage Σn is given by

1
Ψn = p0n 0n + p1n 1n = pin in , (3.24)
i=0
1
where 0  pin  1 and i=0 pin = 1. This state evolves to Ψn+1 ≡ S n+1,n Ψn ,
where S n+1,n is a left-stochastic operator given by

1 
1
S n+1,n ≡ in+1 S ij
n jn, (3.25)
i=0 j=0

where S ij
n are the components of the left-stochastic matrix Sn , where
 
an bn
Sn ≡ . (3.26)
1 − an 1 − bn
Then Ψn+1 is also a stochastic labstate. The determinant |Sn | of the stochastic
matrix Sn in (3.26) is given by |Sn | = an − bn , which means that stochastic
evolution is irreversible for an = bn .
 3.21 Stochastic Questions 45

Exercise 3.5 Prove that the product of any two left-stochastic matrices
is also a left-stochastic matrix. Prove also that the inverse of a nonsingular
left-stochastic matrix is also a left-stochastic matrix.

3.21 Stochastic Questions

In much the same way that we can ask a definite binary question about a
stochastic bit state, we may consider the possibility of asking stochastic binary
questions. By this we mean the following.
Suppose an observer has prepared a stochastic state Ψ of a system under
observation and could ask two different binary questions Q1 and Q2 in principle.
Suppose that for some reason outside their control, every time the observer asked
a question, there was a probability p that it was actually Q1 being asked and
not Q2 , and a probability q ≡ 1 − p that it was actually Q2 being asked and not
Q1 . Suppose further that the observer asked a large number of such questions,
not knowing precisely which question was being asked each time, but observing
the answer each time. Then the average of the observed answers would be given
by pQ1 Ψ + qQ2 Ψ, which we could use to define the stochastic binary question
pQ1 + qQ2 .
 4
Quantum Bits

4.1 Quantum Bits

We stressed previously that although classical bits (or equivalently, bits) are not
vectors, embedding them into a two-dimensional vector space Q gives several
advantages. The first advantage is that we can represent bit operators by either
dyadics or matrices, which allows an efficient encoding of bit dynamics in familiar
linear algebraic terms. The second advantage is that it allows us to generalize
bits to stochastic bits directly and efficiently.
A third advantage that is of fundamental importance in quantized detector
networks (QDN) and which we explore in this chapter is that we can generalize
bits to their quantum counterparts, known as quantum bits (or equivalently,
qubits).
A qubit is a complex two-dimensional Hilbert space, denoted Q. That is a
mathematical statement, but we need more. A signal qubit (Q, C) is a qubit Q
with an empirical context C that defines a preferred basis B ≡{0, 1} for that
particular Hilbert space Q.
Normally, we shall denote a signal qubit (Q, C) by Q whenever the context C
is understood and kept in mind. It is implicitly assumed that all of this discussion
is relative to some observer conducting experiments in a real laboratory.

4.2 Preferred Bases

Viewed in the right empirical context, a classical bit can be identified with a
preferred basis for a given signal qubit, as follows. First, we shall use the same
notation 0, 1 for the two elements of a given classical bit and for their embedding
in the associated qubit Q. In that embedding, we shall take the vectors 0 and 1
in Q to satisfy the “inner product” rule

ij = δ ij , i, j = 0, 1, (4.1)
 4.3 Qubit Properties 47

where 0 and 1 are the duals of 0 and 1, respectively, and δ ij is the Kronecker
delta. Then the relationship between the bit B ≡ {0, 1} and its quantum coun-
terpart Q is this: the elements of B are identified as the natural orthonormal
basis for Q, known as the preferred basis.
The existence of the preferred basis is fundamental to our approach to physics
and is in no way controversial. We do not claim that there is a unique or absolute
preferred frame in the Universe. Any preferred frame is preferred only by virtue
of the relative context associated with a given observer. By definition of what
is meant by “observer,” each observer is assumed always to know the empirical
context associated with each signal qubit that they use, being nothing other than
a mathematical representation of some detector in the observer’s laboratory. We
can be confident that there is always going to be such a local preferred basis,
because the idea that an observer could extract real information in an experiment
with no knowledge about their apparatus makes no sense whatsoever.
By definition, each detector has only two possible outcome states, known as
ground state and signal state. This contextually defines the preferred basis: 0
represents the ground (no signal) state of the apparatus and 1 represents the
signal state of the apparatus.
We emphasize the following point. There runs throughout QM a strand of
thinking, conditioned by experience with CM, that states of SUO have some sort
of existence of their own. According to this logic, such states do not need any
preferred bases for their mathematical representation. This line of thinking then
leads to the notion that observers are not needed either.
It is along such realist lines of thinking that Hidden Variables theory (Bohm,
1952), decoherence (in its original form) (Joos, 2012), and the Multiverse (Many
Worlds) (Deutsch, 1997) are based. The problem with those interpretations of
QM is that they are each contextually incomplete with a generalized proposition
classification1 of zero, meaning that those theories are empirically vacuous.

4.3 Qubit Properties

The power and mystery of QM stems from the possibility of creating states in the
laboratory that are represented by linear superpositions of preferred basis states.
Given the preferred basis, then any pure signal qubit state Ψ can be written in
the form
Ψ = α0 + β1, (4.2)
where α and β are complex numbers and 0, 1 are the two elements of the preferred
basis.
The fact that α and β are complex and not real is of fundamental significance
here. Recall that in our discussion of stochastic bits in Chapter 3, the components

1
Generalized propositions and their classification is discussed in Section 2.12.
48 Quantum Bits

of a stochastic bit state such as (3.19) are nonnegative real numbers representing
conditional probabilities. Because the components of a quantum bit state are
not even real, let alone nonnegative, we encounter here the first of several issues
in the interpretation of QM. This particular issue is generally regarded as being
resolved by Born’s interpretation of the above complex components as probability
amplitudes (Born, 1926), discussed in Section 4.6.
If we choose to use the matrix representation of bits (3.3), then we may write
 
α
Ψ= . (4.3)
R β

Given qubit state (4.2), we define its dual Ψ by

Ψ ≡ α0 + β1 = α∗ 0 + β ∗ 1, (4.4)

where α∗ and β ∗ are the complex conjugates of α and β, respectively. Then the
“inner product” ΨΨ is given, using linearity and (4.1), by

ΨΨ = (α0 + β1)(α0 + β1) = (α∗ 0 + β ∗ 1)(α0 + β1)

= α∗ α

00 + α∗ β

01 + β ∗ α

10 + β ∗ β

11 = |α|2 + |β|2 . (4.5)
1 0 0 1

A normalized signal qubit state is one for which ΨΨ = 1, that is, |α|2 + |β|2 = 1.
We shall deal extensively with normalized signal qubit states, as these are asso-
ciated with probability conservation in QM.
Normalized signal qubit states have been defined in terms of their components
relative to the preferred basis. However, we can discuss them more generally as
qubit states, that is, drop the observational context and think of them as just
elements of some qubit. A qubit is a Hilbert space, a concept that is independent
of basis and therefore does not require a preferred basis. We can discuss qubits
in this context in a more abstract way as follows.
Given any element Φ of a Hilbert space, then by definition it has a norm Φ
or length given by

Φ ≡ (Φ, Φ), ≥ 0, (4.6)

where (Φ, Φ) is the inner product of Φ with itself. This length is basis
independent.
A normalized qubit state therefore is an element of a complex two-dimensional
Hilbert space and has unit norm.

4.4 Qubit Operators

We saw in the section in Chapter 3 that there are only four bit operators, denoted
I, F , D, and U , that map bit states to bit states. In contrast, there is an infinite
number of qubit operators that map qubit states to qubit states. simply because
qubits are vector spaces.
 4.4 Qubit Operators 49

We define a qubit map to be any map from a qubit into itself. Specifically,
given such a map M , then for any element Ψ of qubit Q, the object M (Ψ) is
some element of Q. For example, the identity map I satisfies the rule I(Ψ) = Ψ
for any element Ψ of Q.
This definition of qubit map makes no reference to linearity, so a qubit map
need not be linear, as in the following example.

Example 4.1 Given a qubit Q with preferred basis {0, 1}, define the qubit
map M by M (Ψ) = 0 for every element Ψ of Q. Then M is a nonlinear map,
as, for example, we have M (0 + 1) = 0 but M (0) + M (1) = 0 + 0 = 20.

The map in the above example is not unphysical. It has the interpretation of
a resetting or preparation process that prepares a detector to be in its ground
state, ready to receive a signal, regardless of the state it is currently in. We shall
use such a process in later chapters.
Because we are concerned in this book with quantum processes, almost all
of the qubit maps we shall deal with will be linear. Linear qubit maps will be
called qubit operators. It is conventional in the case of linear operators to drop
the round brackets of the argument; that is, we shall write OΨ to mean O(Ψ)
whenever O is a linear operator. Then for linear operator O, for any elements
Ψ, Φ of Q, and for any complex numbers α, β, we have the rule
O{αΨ + βΦ} = αOΨ + βOΦ. (4.7)
In QM, linear operators are often associated with dynamical variables and so
additional mathematical structure is introduced. Given two qubit operators O 1
and O 2 over a qubit Q, we define the linear combination αO 1 + βO 2 of these
two operators to satisfy the rule
(αO 1 + βO 2 )Ψ ≡ (αO 1 Ψ) + (βO 2 Ψ), (4.8)
for arbitrary complex numbers α, β, and arbitrary elements Ψ of Q.2 Then we
state without proof that the set L(Q) of all qubit operators over a qubit Q has
all the properties of a four-dimensional complex Hilbert space (Paris, 2012).
The structure of the vector space L(Q) is relatively easily explored. We saw
in the previous chapter that given the preferred basis {0, 1} for Q, the four
elementary transition operators (ETOs) T ij ≡ ij : i, j = 0, 1, form a convenient
basis for L(Q), that is, any qubit operator O can be written in the form

1 
1
O= Oij T ij , (4.9)
i=0 j=0

where the coefficients Oij are complex.

2
Note that the + symbol on the left-hand side of (4.8) denotes operator addition, while the
+ symbol on the right-hand side denotes vector addition.
50 Quantum Bits

In addition to their vectorial additive properties, it is useful to define multi-

plication of qubit operators. We define the multiplicative product O 1 O 2 of two
qubit operators O 1 , O 2 by the rule (O 1 O 2 )Ψ ≡ O 1 {O 2 Ψ} for any element Ψ
of Q. The product of any two ETOs is also an ETO, that is, ETO multiplication
is closed. Specifically, the multiplication rule is
T ij T kl = δ jk T il . (4.10)
This ETO multiplication rule (4.10) is associative, that is to say, that
(T ab T cd )T ef = T ab (T cd T ef ), (4.11)
but not commutative, which means T ab T cd = T cd T ab in general.
These properties and the existence of the identity operator I mean that qubit
operators form a mathematical structure known as a unital associative algebra.

4.5 Signal Bit Operators

The space of operators O(Q) over qubit Q contains infinitely many elements.
Fortunately, the necessary existence, in our approach, of the preferred signal basis
{0, 1} singles out a very small number of special qubit operators that we shall
use extensively and refer to as signal bit operators. In addition to the identity
operator I and the zero operator Z (it maps any vector into the zero vector)
there are four important signal bit operators, defined as follows.

The Projection Operators

 are defined by
The qubit projection operators P , P
P ≡ T 00 = 00,  ≡ T 11 = 11.
P (4.12)

The Signal (Annihilation and Creation) Operators

 are in conventional parlance adjoints of each other
The signal operators A, A
and are defined by
A ≡ T 01 = 01,  ≡ T 10 = 10.
A (4.13)

 , A, and A
The four operators P , P  are by inspection just the four ETOs T ij
introduced earlier. The advantage in this new designation is mainly psychological:
the projection operators play one role in our formalism while the signal operators
play another, and it is very helpful to distinguish between them.
In this new notation, the multiplication rule (4.10) is best expressed in the
form of a table, Table 4.1, where the entries are the products LR, operator L
coming from the left-most column and R coming from the top-most row.

4.6 The Standard Born Interpretation

Although superficially qubits look similar to s-bits mathematically, being repre-
sentable by two component column matrices, they are very different objects as
 4.6 The Standard Born Interpretation 51

Table 4.1 Products of signal bit operators

L\R P 
P A 
A

P P Z A Z

P Z 
P Z 
A
A Z A Z P

A 
A Z 
P Z

far as physics is concerned. The components of an s-bit are real and interpreted
as conditional probabilities, whereas the components of a qubit relative to its
preferred basis are complex amplitudes and so cannot be probabilities. It was Max
Born who gave the empirically correct interpretation of such complex amplitudes
(Born, 1926), as follows.

The Born Interpretation

In QM, suppose H is a Hilbert space with inner product of elements ψ, φ denoted
by (ψ, φ). Consider two normalized elements Ψ and Φ in H that represent phys-
ical states of some system under observation. Then the conditional probability
P r(Ψ|Φ) of finding the system to be in state Ψ, given that the system was
prepared to be in state Φ, is given by

P r(Ψ|Φ) = |(Ψ, Φ)|2 . (4.14)

There is a lot of implicit contextual information not given in such a definition,

but physicists generally know what is meant and implied. Specifically, they would
understand that the formalism refers to a statistical analysis of a sequence of
runs. State Φ is prepared through one device at the start of each run, allowed to
evolve undisturbed by the observer over some intermediate space-time regime,
and then passed through another device that produces an outcome. Each outcome
occurs randomly from a range of potential outcomes, with a countable frequency
distribution over the ensemble of runs. In general, the ratios of observed outcome
frequencies, when a large number of runs is performed, conform excellently to
the probabilities predicted by the above Born rule.

Points to note are the following.

Symmetry
There is an inherent symmetry in the relationship between the two states in rule
(4.14), referred to as the microscopic reversibility of quantum processes (Bohm,
1952). This is because of the mathematical equality |(Ψ, Φ)| = |(Φ, Ψ)|, which
immediately leads to the physical prediction P r(Ψ|Φ) = P r(Φ|Ψ). This relation
has been confirmed empirically countless times.
52 Quantum Bits

Complex Amplitudes
The inner product (Ψ, Φ) is complex, which means it does not have the classical
interpretation of a probability. Physicists get around this by referring to such
expressions as “the amplitude for Φ to go to Ψ,” or similar terminology, thereby
endowing it with a touch of familiarity. The fact is, however, no one understands
precisely why complex amplitudes occur in QM and it remains one of the enduring
mysteries of the subject. Schwinger suggested that the appearance of complex
numbers in QM is associated with the existence of antiparticles (Schwinger,
1958). Other physicists have investigated the theoretical and empirical possibility
of replacing complex amplitudes in QM with hypercomplex (or quaternionic)
amplitudes (Adler, 1995; Procopio et al., 2016; Adler, 2016), but there is at this
time no empirical evidence that such a step is necessary.

Origin of the Born Rule

There have been attempts to derive the Born interpretation of the wave function
from basic principles, but so far none of these attempts has been satisfactory. An
interesting variant of such attempts is the idea that the Born rule is but the first
step in a possibly infinite hierarchy of terms, a so-called multiorder interference
rule that is a generalization of the above Born rule. Sorkin noted that for a two-slit
interference experiment, the standard Born interpretation gives for the probabil-
ity P rAB of a particle landing at a point on the detecting screen the formula

P rAB = P rA + P rB + IAB , (4.15)

where P rA is the probability when slit B is blocked off, P rB is the probability

when slit A is blocked off, and IA B is the so-called second-order interference
term. Sorkin considered a three-slit experiment with slits A, B, and C and looked
at the case for a generalization of the Born rule of the form

P rABC = P rAB + P rBC + P rAC − P rA − P rB − P rC + IABC , (4.16)

where IABC represents some novel third-order interference term not predicted
by the above Born rule (Sorkin, 1994). A recent experiment has virtually ruled
out such a term (Sinha et al., 2010).
Sorkin’s motivation in this discussion is directly opposite to ours in this book:
he explicitly states that he does not want to base the interpretation of his
generalized probabilities with “some undefined concept of ‘measurement made by
human observers,”’ and takes “the attitude that the ontology of QM is identical
to that of classical realism” (Sorkin, 1994).

Dynamics
The Born rule is usually applied to states of SUOs evolving in time. In such
a case, care has to be taken with the rules for the conservation of probability.
These rules are as contextual as anything else in an experiment. For instance,
if a state Φi is prepared at initial time ti , allowed to evolve undisturbed until
 4.8 Classical and Quantum Ensembles 53

time tf , then the amplitude to find the system under observation in state Ψf at
final time is given by (Ψf , Uf i Φi ), where Uf i is the unitary evolution operator
taking states from initial time to final time. In this scenario, total probability is
conserved. On the other hand, particle decay experiments may appear to involve
a loss of total probability, if the mathematical modeling is done in too basic a
fashion. For example, the Schrödinger wavefunction for a decaying particle SUO
is frequently asserted to be given by a function of the form
Ψ(t, x)  e−i(E−iΓ)t/ Φ(x), (4.17)
where Γ is a real constant related to the so-called half-life of the particle. We can
discuss such a scenario in QDN; the QDN approach to particle decay experiments
is covered in Chapter 15.

4.7 The Born Interpretation in QDN

We now consider the Born interpretation from the QDN perspective.
Given a normalized qubit state Ψ = α0 + β1, where |α|2 + |β|2 = 1, then the
conditional probability P r(0|Ψ) of the observer finding the associated detector
in its ground state 0 is given by the rule P r(0|Ψ) = |0Ψ|2 = |α|2 , while the
conditional probability P r(1|Ψ) of finding the detector in its signal state 1 is
P r(1|Ψ) = |1Ψ|2 = |β|2 . We shall discuss the generalization of this rule to
collections of detectors in the chapter on quantum register dynamics, Chapter 7.
There are two notational variants that we can use to discuss these probabilities.

Standard Expectation Value Notation

We may write
P r(0|Ψ) = |0Ψ|2 = (0Ψ)∗ (0Ψ) = (Ψ0)(0Ψ)
= Ψ(00)Ψ = ΨP Ψ, (4.18)
interpreted in words as “P r(0|Ψ) is the expectation value of the ground state
projection operator P , contextual on the prepared state Ψ.” Likewise, we have
the rule
 Ψ.
P r(1|Ψ) = Ψ P (4.19)

Density Operator Notation

Given a pure bit state Ψ, first define the density operator  ≡ ΨΨ. Then the
probabilities P r(0|Ψ), P r(1|Ψ) are given by the rules
P r(0|Ψ) = T r {P } ,  },
P r(1|Ψ) = T r{P (4.20)
where T r denotes the trace operation, discussed in Chapter 9.

4.8 Classical and Quantum Ensembles

The crucial differences between stochastic bits and qubits are not easy to see at
the rank-one level but one of them is this: in CM, a stochastic bit state represents
54 Quantum Bits

an observer’s epistemic uncertainty as to which bit state a detector is actually

in before they observe it, whereas in QM an observer can be certain of which
qubit state a quantized detector is in and it is only the future outcome of an
observation that is uncertain. Moreover, the uncertainty in the quantum case is
generally regarded as intrinsic, or aleatoric, uncertainty.
There are several other ways of saying much the same thing. We can discuss
state preparation, the processes that lead up to an observer having a contextual-
based belief about the signal state of their detector, before observation. A stochas-
tic bit state represents the observer’s uncertainty about the preparation pro-
cesses, whereas in the quantum case, the observer need have no such uncertainty
about the preparation of a qubit state. Such a qubit state is called a pure state.
The Born rule discussed above makes no reference to state preparation and it is
assumed that the qubit state is pure.
Another way of showing the difference between stochastic bits and qubits is
in terms of questions and answers. Given a stochastic bit of the form S ≡ a0 +
(1 − a)0, then iS, i = 0, 1, represents a stochastic answer, which is a probability.
On the other hand, given a qubit Ψ, then iΨ, i = 0, 1, represents a quantum
answer , which is a complex amplitude. Quantum answers have to be processed
according to the Born rule given above in order to extract outcome probabilities.
Yet another way of seeing the difference between a stochastic bit and a qubit
comes from the notion of ensemble. These are discussed in more detail in the
Appendix. An ensemble can be either a real collection of near identical systems
under observation, such as atoms in a crystal, or a hypothetical collection of
imagined alternative futures, only one of which is going to be realized. Stochastic
bits are generally associated with the former type of ensemble, while qubits are
associated with the latter. We shall refer to the former kind of ensemble as a
classical ensemble and refer to the latter kind as a quantum ensemble.
The differences between stochastic bits and qubits will become more obvious
when we deal with quantum registers, or collections of qubits, discussed in
Chapter 7.

4.9 Basis Transformations

Given a particular detector, then its associated physics provides the associated
empirical context: if they looked, the observer would recognize when that detec-
tor was in its ground state 0 and when it was in its signal state 1. If this were
not the case, we would have to ask what observation meant in this case.
This unambiguity about the context is reflected in the two questions an
observer could ask of any detector: Is this detector in its ground state?
and Is this detector in its signal state? A classical bit will always return
an unambiguous answer of yes or no, a stochastic bit returns a probability,
and a quantum bit returns a probability amplitude. This is encoded into the
vector space formalism by the fact that stochastic bits and qubits are linear
combinations of the classical answer states 0 and 1.
 4.10 The Preferred Basis Problem 55

We return now to the fact that the definition of a Hilbert space is basis
independent. Let us explore this further. Ignoring physical context and looking
at a qubit strictly as a mathematical vector space, we are entitled to change our
basis from the preferred basis. Consider therefore replacing each element i in our
original preferred basis with some new vector i , as follows. First, we note that our
preferred basis B is orthonormal, i.e., ij = δ ij , i, j = 0, 1. Anticipating physical
applications in later chapters, we shall preserve this relationship. Therefore,
we shall require i j  = δ ij , i, j = 0, 1. For a complex Hilbert space, such
transformations are called unitary.
Given our initial basis B ≡ {0, 1}, consider a unitary transformation B →
B  ≡ {0 , 1 } such that i j  = δij , for 0 ≤ i, j ≤ 1. For such a transformation we
may write

1
i → i ≡ U i = j  U ji , (4.21)
j=0

where the complex coefficients U ij satisfy the unitarity relations


1
U ij∗ U jk = δ ik , (4.22)
j=0

where U ij∗ is the complex conjugate of U ij . These unitarity relations guarantee

that orthonormality is preserved.
We may always write a unitary matrix in the form
 
α β
U= , (4.23)
γ δ
where the coefficients α, β, γ, and δ are complex. From (4.22) we deduce the
important relations
|α|2 + |β|2 = |γ|2 + |δ|2 = 1, αγ ∗ + βδ ∗ = 0. (4.24)

4.10 The Preferred Basis Problem

At this point we come across another problem related to the fact that a unitary
transformation of the elements of a Hilbert space H has an implied action on
the elements of its dual space H. This is because although the original definition
of the inner product (ψ, φ) of a Hilbert space H is defined as a map from the
Cartesian product H × H into the complex field, it can also be interpreted as
a mapping of the vector φ in H into the complex numbers by the action of a
one-form (a dual vector) ψ, which is an element of a different vector space, the
dual space H. The implied action of the above unitary operator U on elements
of the dual space is given by

1
i → i ≡ iU † = j  U ji∗ , (4.25)
j=0
56 Quantum Bits

where U † is the Hermitian conjugate operator. For finite-dimensional vector

spaces, the Hermitian conjugate operator is the same as the adjoint operator,
problems arising only with nonseparable Hilbert spaces (Streater and Wightman,
1964).
The point is that according to our interpretation of the dual vectors 0 and
1, they are associated with questions to be asked of answer states. Given the
transformation (4.25), we have to ask what “a linear combination of questions”
means. While the Born interpretation gives us a meaning for an answer to a
classical question asked of a linear combination of vectors, it is not immediately
obvious what the interpretation of an object such as u0+v1 is. Our current view is
that it is a mathematical artefact devoid of physical significance. What underpins
this view is that observers are always sure in their minds which questions they
are asking in a laboratory.
Another way of saying this is that we have not considered quantizing observers:
they are always regarded as classical and this policy will be maintained through-
out this book. This does not mean we shall not consider quantizing apparatus.
There will be situations where linear combinations of questions makes physical
sense. For instance, we saw in the previous chapter that we could interpret linear
combinations of questions of stochastic bits in terms of probabilities. On the
quantum side, not only will such a possibility be available to us, but there will
be a quantum side to this issue. Linear complex combinations of questions will
have a role in the information void, the regime between state preparation and
state outcome detection. In this regime, quantum rules apply and the concept of
observer is not meaningful.
We should add at this point that the linear combination of quantum questions
we have just referred to is not the same thing as a mixed quantum question.
A mixed quantum question would be the analogue of a mixed state in QM,
where an observer has an epistemic uncertainty as to which quantum state had
been prepared. A mixed quantum question would likewise involve an experiment
where an observer had an epistemic uncertainty as to which quantum question
was being asked. This is not the same thing as a linear combination of quantum
questions (which carries the implication that the observer itself is quantized).
One of the problems of interpretation that arises in the Multiverse paradigm
(Deutsch, 1999) is that observers and SUOs are described by the vacuous concept
of a wave function for the Universe. Since there is by assertion no primary
observer, it is not clear what superposition of different observers means, if any-
thing. In this respect, the original ideas of Everett’s “Relative State” interpre-
tation of QM seem less unattractive (Everett, 1957), because of the adjective
relative.3 It is not unreasonable to imagine that one observer A could describe
what other observers B and C are doing by a quantum state vector. On that

3
Everett does explicitly postulate an absolute wave function for the Universe, something
QDN cannot accept, because such a postulate has a GPC of zero.
 4.11 Rank-One Qubit Evolution 57

basis, relative to A, B and C are systems under observation and not observers.
QDN is fully compatible with that idea, but regards the Multiverse concept as
anathema.

4.11 Rank-One Qubit Evolution

As with bits and s-bits, we can consider the dynamical evolution of a single qubit
state. Given a normalized qubit state Ψn ≡ αn 0n +βn 1n in Qn at stage Σn where
|αn |2 + |βn |2 = 1, we consider a linear map U n+1,n from Qn to Qn+1 such that
normalization is preserved; i.e., we require
Ψn → Ψn+1 ≡ U n+1,n Ψn , Φn → Φn+1 ≡ Φn U †n+1,n , (4.26)
with Ψn+1 Ψn+1 = 1.
We come now to an important point. In the conventional theory of Hilbert
spaces and in its application to standard QM, unitary transformations are maps
from a given Hilbert space back into itself. In our situation this is no longer the
case. We are dealing with maps from a Hilbert space at time n to another Hilbert
space associated with time n + 1. For example, the transformation (4.26) gives
the dyadic representation

1
ij
U n+1,n = in+1 Un+1,n jn. (4.27)
i,j=0

This operator is one that not only preserves the norm but also preserves inner
products under the transformation from one Hilbert space to the next. Such
a transformation will be called a semi-unitary transformation, rather than a
unitary transformation, for the good reason that in a more general context, the
dimensions of the Hilbert spaces need not be the same. When the dimensions
of the Hilbert spaces are different, then we need to consider the retraction of
an evolution operator, rather than its inverse. We shall discuss this important
aspect of our dynamics in more detail later.
In standard QM, a dynamics that preserves magnitudes of inner products
(as opposed to inner products themselves) is usually realized via either unitary
transformations or anti-unitary transformations. Anti-unitary transformations
are subtle, having a great deal to do with the concept of time reversal in standard
QM. We shall not consider anti-unitary transformations further.
An important issue that impacts on our approach is that one way of discussing
time reversal in QM is to switch bra and ket vectors. Conventionally, this does
not seem to matter but it amounts to switching questions and answers in our
approach, which requires great care in the interpretation.
We may readily generalize our dynamical rule (4.26) to describe evolution from
initial stage SM to final stage SN , where N > M : the product of two semi-unitary
transformations is also a semi-unitary transformation. We find the rule
ΨM → ΨN = U N,M ΨM , (4.28)
58 Quantum Bits

where U N,M ≡ U N,N −1 U N −1,N −2 . . . U M +1,M . The conditional outcome prob-

abilities P r(iN |ΨM ) as measured at stage ΣN are then given by

P r(iN |ΨM ) ≡ |iN ΨN |2 , i = 0, 1. (4.29)

4.12 Mixed Qubit States

The uncertainty of s-bits is different from that associated with qubits. The former
is due to ignorance on the part of the observer and may be called classical (or
epistemic) uncertainty, while the latter is considered intrinsic and so referred to
as quantum uncertainty on that account.
It is possible to encounter situations where both types of probability occur
naturally. Whenever this happens, the labstates involved are no longer referred
to as pure but mixed . The following example illustrates what we mean.

Example 4.2 An observer is about to ask the question 1 of a single detector

but is not sure how the labstate was prepared. The information that they do
have, however, leads them to believe that the probability of the labstate (prior
to observation) being Ψ ≡ α0 + β1 is p, where |α|2 + |β|2 = 1, while the
probability of the labstate being Φ ≡ γ0 + δ1 is 1 − p, where |γ|2 + |δ|2 = 1.
What is the overall probability of finding a signal?
Solution
Consider a very large number N of runs. Of these, approximately pN will
involve the labstate Ψ. The signal outcome probability for each such run
is |β|2 according to the Born rule. Therefore the total number of runs that
involve the labstate Ψ with a signal outcome is approximately pN |β|2 . A
similar calculation for the labstate Φ gives the number of signal outcomes as
(1 − p)N |δ|2 . The total number of signal outcomes is therefore approximately
N p|β|2 + N (1 − p)|δ|2 ). In the limit N → ∞, the signal outcome probability
is therefore p|β|2 + (1 − p)|δ|2 (the answer).

4.13 Density Operators

An efficient and standard way of combining classical and quantum probability
is through the use of dyadics called density operators. Suppose we have a mixed
initial state consisting of k different possible states Ψa , a = 1, 2, . . . , k. The
density operator  is defined by the dyadic

k
≡ ω a Ψa Ψ a , (4.30)
a=1

where ω a is the probability that the initial state is actually Ψa . The expectation
value O of an observable O is then given by the standard rule
 4.13 Density Operators 59

O ≡ T r {O} , (4.31)

where T r denotes the trace operation, discussed in Chapter 9.

To demonstrate how this works, we apply this formalism to the example
considered above.

Example 4.3 With reference to the above example, we see k = 2 and

ω 1 = p, Ψ1 ≡ Ψ = α0 + β1
(4.32)
ω 2 = 1 − p, Ψ2 ≡ Φ = γ0 + δ1.
Hence the density matrix is

 = ω 1 Ψ1 Ψ1 + ω 2 Ψ2 Ψ2
= p{α0 + β1}{α∗ 0 + β ∗ 1} + (1 − p){γ0 + δ1}{γ ∗ 0 + δ ∗ 1}
(4.33)
= {p|α|2 + (1 − p)|γ|2 }00 + {pαβ ∗ + (1 − p)γδ ∗ }01
+{pβα∗ + (1 − p)δγ ∗ }10 + {p|β|2 + (1 − p)|δ|2 }11.

The observable to use is P ≡ 11, the projection operator associated with

the signal state 1. Then we find
  = {pβα∗ + (1 − p)δγ ∗ }10 + {p|β|2 + (1 − p)|δ|2 }11.
P (4.34)

Taking the trace then gives

 } = {p|β|2 + (1 − p)|δ|2 },
T r{P (4.35)

which agrees with the probability found above in Example 4.2.

 5
Classical and Quantum Registers

5.1 Introduction
In this chapter we extend the discussion of the previous chapter, from one
detector to an apparatus with an arbitrary number of nodes at any given stage.
Our labeling convention is followed throughout this book: subscripts always
label stages whereas superscripts always label nodes and modules (discussed
in Chapter 11). The number of nodes at stage Σn will be called the rank of the
apparatus at that stage and denoted rn . The ith node at stage Σn will be denoted
by in , which should not be confused with labstate vectors such as in , which are
always denoted in bold font.
Whenever we are using classical mechanics (CM), the collection of classical
nodes at stage Σn will be called a classical binary register , denoted Rn . In that
case, in is represented by classical bit Bni , so Rn is the Cartesian product of all
the bits at that stage, that is,
Rn ≡ Bn1 × Bn2 × · · · × Bnrn . (5.1)
The cardinality (number of elements) of the rank r classical binary register Rn
[r ]
is dn ≡ 2rn . If we wish to indicate the rank of Rn , we write Rn n .
On the other hand, whenever we are using quantum mechanics (QM), the
collection of nodes at stage Σn will be called a quantum register and denoted
[r ]
Qn or Qn n . In the quantum case, in is represented by qubit Qin . The quantum
register Qn is the tensor product of all the qubits at that stage, that is,
Qn ≡ Q1n ⊗ Q2n ⊗ · · · ⊗ Qrnn . (5.2)
Such a tensor product space is a Hilbert space of dimension dn ≡ 2 . rn

5.2 Labels versus Ordering

Superscripts are used in quantized detector networks (QDN) to identify individ-
ual nodes. Therefore, the standard left-right ordering in Cartesian products such
 5.3 The Signal Basis Representation 61

as (5.1) or the left-right ordering in tensor products such as (5.2) is redundant.

Provided we retain superscripts, we can drop the × symbol in Cartesian products
and the ⊗ symbol in tensor products. Moreover, we can dispense with the left-
right ordering rule in such products. Henceforth, we adopt the convention that
the Cartesian product B 1 ×B 2 can be represented unambiguously by either B 1 B 2
or B 2 B 1 . Moreover, the ordered element (i1 , j 2 ) of B 1 × B 2 can unambiguously
be represented by the notation i1 j 2 or j 2 i1 . A similar convention will be applied
to tensor products and to dual spaces.

Example 5.1 If Ψ1 is a vector in qubit Q1 and Φ2 is a vector in Q2 ,

then Ψ1 Φ2 = Φ2 Ψ1 ≡ Ψ1 ⊗ Φ2 is an element of the tensor product space
Q 1 Q2 = Q 2 Q1 ≡ Q1 ⊗ Q2 .

Example 5.2 The rank-two classical register R[2] ≡ B 1 B 2 contains four

classical states:

R[2] ≡ {01 02 , 11 02 , 01 12 , 11 12 }. (5.3)

These elements cannot be added together.

On the other hand, the corresponding rank-two quantum register Q[2] ≡
Q Q has a preferred basis B [2]  R[2] corresponding to the above four
1 2

classical states. Now, however, arbitrary linear combinations of the form

αi1 j 2 + βk1 l2 , for complex α, β and i, j, k, l = 0, 1, are allowed, giving new
elements in Q[2] . The four elements 01 02 , 11 02 , 01 12 , 11 12 form the preferred
basis for Q[2] , while their duals 01 02 , 11 02 , 01 12 , 11 12 form the preferred basis
for the dual space Q[2] .

5.3 The Signal Basis Representation

For any given rank, there are infinitely many more possible quantum states than
possible classical states. Representing these quantum states efficiently there-
fore requires suitable notation. In this section, we introduce a representation
that is based on observational context (what we can see). By this we mean it
utilizes the specific details of the detectors involved in the experiments being
discussed.
Throughout the rest of this chapter we discuss a collection of qubits seen at
a single stage, or instant of the observer’s time, so in this chapter we suppress
any reference to time. However, time will not be overlooked in general. In the
dynamical theory given in the next chapter, a temporal subscript n will be
introduced that is associated with every labstate and with other quantities such
as the rank of the quantum register. Our ultimate aim is to construct a general
theory of observation in which apparatus itself becomes a dynamical quantity.
62 Classical and Quantum Registers

Consider a rank-r quantum register Q[r] ≡ Q1 Q2 . . . Qr . Observational context

will inform us about the preferred basis B [r] for Q[r] . This consists of all possible
signal basis states of the apparatus. These will represent all possible classical
yes-no configurations of the detectors involved. Each element of this basis set
is a tensor product of the form i1 i2 i3 . . . ir , where the ia are all either the 0a
(ground state) or 1a (signal state) elements of the preferred basis B a for the ath
detector qubit Qa . When written in this form, B [r] will be referred to as the
signal basis representation (SBR).
By inspection, we see there are 2r distinct elements in B [r] and together they
constitute an orthonormal basis for the quantum register Q[r] . We can identify
the elements in the classical register R[r] with the elements of the preferred basis
B [r] of the quantum register Q[r] .

Example 5.3 The SBR for a rank-three quantum register has 23 = 8

elements and is given by
 
B [3] ≡ 01 02 03 , 11 02 03 , 01 12 03 , 11 12 03 , 01 02 13 , 11 02 13 , 01 12 13 , 11 12 13 .
(5.4)

To define orthonormality, we introduce the dual basis B [r] , consisting of ele-

ments of the form i1 i2 i3 . . . ir , where ia = 0a or ia = 1a for a = 1, 2, . . . , r. Inner
products in Q[r] are defined by the action of elements of B [r] on elements of B [r] ,
plus linearity:

i1 i2 i3 . . . ir j 1 j 2 j 3 . . . j r = (i1 j 1 )(i2 j 2 )(i3 j 3 ) . . . (ir j r )

1 1 2 2 3 3 r r
= δi j
δi j
δi j
. . . δi j
. (5.5)

This rule is interpreted as follows. If the right-hand side of (5.5) is zero, then the
elements i1 i2 i3 . . . ir and j 1 j 2 j 3 . . . j r of the Hilbert space Q[r] are orthogonal;
otherwise, they are the same element and it has length (norm) of one.

Example 5.4 According to our notation, indices keep track of factor qubit
spaces in a quantum register. Hence for a rank-four register, for example, we
would find

04 11 13 02 03 11 02 04 = (11 11 ) (02 02 ) (13 03 ) (04 04 ) = 1 × 1 × 0 × 1 = 0. (5.6)

 
 
 
 
1 1 0 1

5.4 Maximal Questions

Given an apparatus represented by r detectors at a given stage, the observer has
the freedom to ask any one of a number of questions. For instance, the observer
may look at a particular detector to ascertain its signal state and not look at any
 5.5 Signality 63

of the other detectors. Such a question that does not involve all of the detectors
will be called a partial question. Partial questions are discussed in Chapter 8.
A maximal question is one that asks a binary question of every detector in
a register. From (5.5), we see that the elements of the dual basis B [r] can be
interpreted as maximal questions, and there are 2r of them. It is important to
note that in the absence of any theory that quantizes observers and/or their
questions, arbitrary vectors in Q[r] are not in general physically meaningful
maximal questions; only the elements of the preferred dual basis B [r] are maximal
questions in the formulation of QDN being discussed at this point.
It is not impossible, however, to imagine a scenario where questions being asked
correspond to elements of Q[r] that are nontrivial linear combinations of preferred
basis elements. For instance, a given observer may be conducting an experiment
where there are two possible quantum outcomes, such that each outcome would
trigger a different maximal question of some state at a later stage. That would
correspond to quantization of apparatus and/or observers. This sort of scenario
is essentially the focus of Chapter 21.

Example 5.5 Given the rank-two quantum register Q[2] ≡ Q1 Q2 , the

preferred basis B [2] has four elements:

B [2] = {01 02 , 11 02 , 01 12 , 11 12 }. (5.7)

The maximal questions are all the elements of the dual basis

B [r] ≡{01 02 , 11 02 , 01 12 , 11 12 }. (5.8)

5.5 Signality
The signality of a basis state corresponds to the total “particle number” being
detected by the apparatus concerned, if any positive signal in any of the detectors
is interpreted as registering a particle in the conventional sense.

Definition 5.6 The signality of an element i1 i2 i3 · · · ia of the signal basis

r
B [r] is the sum a=1 ia .

Example 5.7 The signality of the element 06 12 11 05 03 14 in B [6] is 1 + 1 +

0 + 1 + 0 + 0 = 3.

The particle interpretation of signality has to be viewed cautiously for two

reasons. First, detectors register irreversible processes that may have nothing
to do with any identifiable, persistent “particle.” For example, a phonon is a
collective phenomenon, a quantum of excitation in a crystal that does not “exist”
64 Classical and Quantum Registers

in a reductionist sense. Second, it is possible in QDN to deal with labstates that

are superpositions of preferred basis states of differing signality. Classically, this
would have no particle interpretation but is part and parcel of QDN.
Dynamics may in many situations rule out certain superpositions of labstates
in a quantum register. For example, charge conservation rules out superpositions
of labstates corresponding to different numbers of electrons. Another example
is angular momentum: we are not allowed in conventional quantum mechanics
to superpose states of particles with different total spin number, conventionally
denoted by j. However, the general principle remains: superposition of labstates
of different signality is allowed in principle in QDN; only dynamical context rules
out certain mathematical possibilities.
These comments apply to the labstates. As far as the dual basis is concerned,
we are not at this stage entitled to create linear superpositions of any of the
basis elements in the dual basis set B [r] , as these elements represent questions
being asked by the observer. We have introduced no concept yet of a “quantized
observer,” whatever that might mean. Notwithstanding our comments just before
Example 5.5, all questions asked at this stage of this book have to be completely
classical. The point about QDN is that the labstates do not have this restriction
and can be superpositions of the classical-looking elements of the preferred basis.
This remark applies before the observer looks at a labstate.

5.6 The Economy of Success

A useful fact now emerges, a fact that underpins the success of CM (a moderately
simple theory) in describing a hideously complex reality. Of all the 2r states in
a rank-r classical register, only one of them has a truth value of one relative to
any given maximal question, and all the other 2r − 1 states have truth value
of zero, relative to that question. To see this, consider an arbitrary maximal
1 2 r
question i i . . . i asked of an arbitrary classical register state j 1 j 2 . . . j r . The
answer is
1 2 r 1 2 r 1 1 2 2 r r
i i . . . i j 1 j 2 . . . j r = (i j 1 ) (i j 2 ) . . . (i j r ) = δ i j
δi j
. . . δi j
, (5.9)

 
 
 
ir j r
δ i1 j 1 δ i2 j 2 δ

which is zero in general except for the particular case when ia = j a for each
possible value of a, and then the answer value is one. This result greatly simplifies
calculations in QDN.
This result contributes to the inherent economy of the particle description in
CM, as the following example illustrates. Suppose we wanted to describe a single
point particle in three-dimensional space. The conventional approach would be
to define a Cartesian coordinate system with coordinates x, y, and z along three
mutually orthogonal axes, relative to some chosen origin of coordinates. Then the
position of a particle P would be fully specified by giving only three numbers,
(xP , yP , zP ), referred to as the position coordinates of the particle. The economy
 5.7 Quantum Registers 65

of this approach is literally infinite: saying where the particle is immediately says
where the particle is not, which is every point in space apart from the one with
coordinates (xP , yP , zP ).
This economy of information comes about because of context: we have been
told that there is only one particle.
But suppose now we are presented with a situation where we were not told
beforehand how many particles there were in our laboratory. Then we would
have to look at every point in space if we wanted to know the total particle
number. This is not a trivial point but a hard fact in astrophysics and cosmology:
astronomers have to scan space thoroughly in order to make estimates of matter
and dark matter densities on cosmological scales.
In the quantum context, we may use signality to tell us how many signals
there are in our detector array. A labstate of signality one corresponds closely to
what can be thought of as a single particle state. This means that, if we knew
beforehand that we were dealing with a signality-one state, then it would suffice
to find a positive signal in just one detector in an array to be sure (by context)
that there were no positive signals anywhere else in that array.
It is an intriguing thought that when we specify the classical Cartesian coordi-
nates x, y, z of a point particle, we are not only answering the question where is
the particle? but also saying that it is not at any of the other points in space,
of which there is an uncountable number. Clearly, that is an impressive form of
economy, but one occurring only because of our contextual information.

5.7 Quantum Registers

We now extend the discussion from quantum labstates of a single detector to
quantum labstates of two or more detectors. These higher rank labstates are
elements of some quantum register Q[r] ≡ Q1 Q2 . . . Qr , where r  1 is the rank
of the register and Qi is the ith qubit in the register. We have suppressed the
tensor product symbol ⊗ here.
Tensor products of vector spaces are discussed in the Appendix. Unlike Carte-
sian products, tensor product spaces are genuine vector spaces, the quantum
register Q[r] being a complex vector space of dimension 2r . This is one of the
places where the fundamental differences between CM and QM manifest them-
selves. In the classical case, there is only a finite number of different possible
states in a classical register. For a system of r classical bits, each of which has
two possible mutually exclusive answer states, then the total number of integers
0 or 1 we would need in order to parametrize any classical state of the system
is just r. Any state in such a classical register can therefore be represented by
an r-tuple of the form {i1 , i2 , . . . , ir }, where ia = 0 or 1, for a = 1, 2, . . . r, and
there are just 2r different such r-tuples. In the quantum case, possible states are
elements of a finite-dimensional complex vector space of dimension 2r , which has
infinitely many elements.
66 Classical and Quantum Registers

Consider an arbitrary state Ψ in Q[2] . Given the above preferred basis, we can
always write
Ψ = z 1 01 02 + z 2 11 02 + z 3 01 12 + z 4 11 12 , (5.10)
where the z a are labeled complex numbers, not powers of z. The dual state Ψ is
given by
Ψ ≡ z 1 01 02 + z 2 11 02 + z 3 01 12 + z 4 11 12
= z 1∗ 01 02 + z 2∗ 11 02 + z 3∗ 01 12 + z 4∗ 11 12 , (5.11)
where z a∗ is the complex conjugate of z a . If the state is normalized, we have the
condition
ΨΨ = |z 1 |2 + |z 2 |2 + |z 3 |2 + |z 4 |2 = 1. (5.12)
If now we write z a = xa + iy a , where xa and y a are real, then the normalization
condition (5.12) is equivalent to saying that possible states rank-two quantum
register can be identified one-to-one with points on S 7 , the unit sphere in eight-
dimensional Euclidean space E8 .
The above analysis is for just two detectors. If we had in mind, say, a screen
consisting of a million detectors, or a neural network in a brain, the numbers
shoot up beyond imagination.1 Clearly, qubit register states have much more
structure than their classical counterparts. The exploration of this structure is
in a real sense in its infancy at this time. For example, there is still a great deal
to be understood about the physics of quantum entangled states in low rank
quantum registers. We discuss some aspects of this in Chapter 22.

5.8 The Computational Basis Representation

The SBR {i i i . . . ir : ia = 0, 1 : a = 1, 2, . . . , r} for elements of the preferred
1 2 3

basis B [r] is useful in some respects but less so in others. A frequently more
useful but equivalent notation for the elements of the preferred basis will be
called the computational basis representation (CBR) and is obtained as follows.
For each element i1 i2 i3 · · · ir of the SBR, there is precisely one element i of the
CBR, where i is an integer in the range [0, 2r − 1], given by the computational
basis map
i = i1 + 2i2 + 4i3 + · · · + 2r−1 ir , (5.13)
3
where i means the third element in the SBR and not i to the power of three.
The computational basis map can be inverted. Given an element p of the CBR,
where p is an integer in the range [0, 2r − 1], we can write
1  a [a]
r
p = p[1] + 2p[2] + 4p[3] + · · · + 2r−1 p[p] = 2 p , (5.14)
2 a=1

1
The average human brain has about a hundred billion neurons.
 5.9 Register Operators 67

where the symbol p[a] denotes the signal status (zero or one) of the ath detector
when the apparatus is in the labstate p. The coefficients p[a] will be referred to
as the binary components of the integer p and are discussed in further detail later
on in this chapter. They play an essential role in the formalism of QDN and in
many calculations.

Example 5.8 For the rank-three quantum register Q1 Q2 Q3 , the SBR of the
preferred basis B [3] is

B [3] = {01 02 03 , 11 02 03 , 01 12 03 , 11 12 03 , 01 02 13 , 11 02 13 , 01 12 13 , 11 12 13 }.
(5.15)
The CBR of B [3] is the set

B [3] = {0, 1, 2, 3, 4, 5, 6, 7}, (5.16)

where 0 = 01 02 03 , 1 = 11 02 03 , and so on.

The CBR generally has the advantage over the SBR of being more compact
and on that account is better suited in many but not all calculations. The dual
preferred basis B [r] can also be expressed in CBR terms, and then the inner
product relations (5.5) take the form
ij = δ ij , 0  i, j < 2r , (5.17)
which is very useful.
A disadvantage of the CBR is that it masks the signal properties of a given
state. For example, the CBR element 3 could represent the SBR element 11 12 for
a rank-two apparatus or the SBR element 11 12 03 04 05 of a rank-five apparatus
and so on. However, context will generally make it clear what is meant by a given
CBR expression.
The CBR is useful for representing linear operators over the register. These
will generally be denoted in blackboard font in QDN. For example, the classical
or quantum register identity operator I[r] is expressed in the CBR by
r
2 −1
I[r] = kk. (5.18)
k=0

More generally, if we know the action of a linear register operator on each

element of the signal basis, then we can represent that operator as a dyadic in
the computation representation, as we show in the next section.

5.9 Register Operators

A register operator is a function that maps elements of a classical or quantum
register into itself or another classical or quantum register. Linear register oper-
ators will be denoted in blackboard bold font, such as U, A, and so on. The same
68 Classical and Quantum Registers

labeling convention used for states and questions will be used for linear operators:
upper indices refer to qubits and, while lower labels indicate stages.
Suppose On,m is a linear register operator mapping states from an initial stage
Σm rank-rm classical or quantum register into a final stage Σn rank-rn classical
or quantum register. Suppose further that we are given its action on each element
im of the CBR, that is,
n −1
2r
On,m im = ji
On,m jn, i = 0, 1, . . . , 2rm − 1, (5.19)
j=0
ji
where the On,m are complex coefficients.
Now “multiply” each side from the right by im and sum over i:
m −1
2r m −1 2rn −1
2r 
On,m im im = ji
On,m j n im (5.20)
i=0 i=0 j=0

The left-hand side simplifies because of linearity:

m −1
2r
2rm −1 

On,m im im = On,m im im = On,m Im = On,m , (5.21)
i=0 i=0

using (5.18). Hence finally we arrive at the dyadic representation of a typical

register operator:
m −1 2rn −1
2r 
On,m = ji
j n On,m im . (5.22)
i=0 j=0

We define the retraction On,m of a register operator On,m by

m −1 2rn −1
2r 
On,m ≡ ji∗
im On,m jn, (5.23)
i=0 j=0
ij∗ ij
where On,m is the complex conjugate of On,m . A retraction as defined here is not
necessarily the equivalent of an inverse operator.

5.10 Classical Register Operators

We saw in Chapter 3 that there are only four classical bit operators C 1 ≡ I,
C 2 ≡ F , C 3 ≡ D , and C 4 ≡ U that map bit states to bit states. Likewise, a
classical register operator maps classical register states to classical register states.
We shall discuss a classical register operator Cn,m mapping register states from
a rank-rm classical register Rm to a rank-rn classical register Rn .
As before, the most economical way to discuss classical register operators is
via the CBR, as follows. For any element im in Rm , where i = 0, 1, 2, . . . , 2rm −1,
the operator Cn,m necessarily maps it into precisely one element in Rn , and not
into a linear combination of two or more, as in the quantum case. This means
that we can always write
 5.11 The Signal Algebra 69

n −1
2r
Cn,m im = ji
Cn,m jn, i = 0, 1, 2, . . . , 2rm − 1, (5.24)
j=0
ji
where for a given i, every element Cn,m is zero for every integer j in the interval
[0, 2 − 1] except for one, which has value one. Using the completeness property
rn

of the elements {im } we arrive at the dyadic representation

n −1 2rm −1
2r 
Cn,m = ji
j n Cn,m im . (5.25)
j=0 i=0
rm
This is a special case of (5.22) and there is a total of (2rn )2 different possible
such operators.
In this set of operators, there are two subsets that are of particular importance
in both classical register mechanics and quantum register mechanics. These are
the register projection operators and the register signal operators.

Register Projection Operators

Given a rank-r register, the observer may be interested in asking questions of an
individual bit or qubit and leaving all the other bits or qubits alone. The register
projection operators Pi and P i will be used in later chapters to construct partial
questions. These operators are defined by
Pi = I 1 I 2 . . . I i−1 P i I i+1 . . . I r ,
i = I 1 I 2 . . . I i−1 P
P  i I i+1 . . . I r ,

where I a is the bit identity operator for the ath bit or qubit and P i , P  i are
the bit projection operators for the ith bit or qubit, as discussed in the previous
chapter.

Register Signal Operators

Associated with a rank-r quantum register Q[r] are some important operators
connected to the physics of observation, and these will appear frequently through-
out the formalism. The most important of these are the r signal annihilation
 i . These
operators Ai , i = 1, 2, . . . , r and the related signal creation operators A
operators are defined in terms of the signal bit operators discussed in the previous
chapter, as follows:
Ai ≡ I 1 I 2 . . . I i−1 Ai I i+1 . . . I r ,
 i ≡ I 1 I 2 . . . I i−1 A
 i I i+1 . . . I r , (5.26)
A i = 1, 2, . . . , r,
where the superscripts on the right-hand side label the individual qubits in a
given rank-r register and we suppress the tensor product symbol ⊗.

5.11 The Signal Algebra

a
We define S , the ath signal set, to be the set of register operators
a , Aa , A
S a ≡ {Pa , P  a }, a = 1, 2, . . . , r. (5.27)
70 Classical and Quantum Registers

Table 5.1 The signal set algebra

Pa a
P Aa a
A

Pa Pa 0 Aa 0
a
P 0 a
P 0 a
A
Aa 0 Aa 0 Pa
a
A a
A 0 a
P 0

Then an extraordinarily useful fact is that any two elements from different signal
sets commute.
For a given signal set S a , we have the ath signal set algebra, as shown in
Table 5.1. For i = 1, 2, . . . , r we have
 
Ai Ai = A  iA i = 0,  i = Pi + P
Ai , A i = I[r] , (5.28)

where curly brackets denote an anticommutator, while for i = j, we have

 i, A
[Ai , Aj ] = [A  j ] = [Ai , A
 j ] = 0, (5.29)

where square brackets denote a commutator. Note that in the above, the symbol
0 represents the zero operator for the register concerned.
The signal set algebra gives QDN a particular flavor; parts of it are reminiscent
of a theory with fermions (anticommuting objects), while other parts have a
bosonic (commuting) character. At the signal level, however, we are dealing with
neither concept specifically; the signal algebra is determined by the physics of
observation as it relates to the apparatus and has its own logic that is distinct
from that of conventional particle physics.

5.12 Signal Excitations

The signal operators introduced above are used to construct signal states from
the signal ground state 0 as follows.
1. The signality one states are of the form  i 0 ≡ 2i−1 , 1 ≤ i ≤ r.
A
2. The signality two states are of the form 
Ai A j 0 ≡ 2i−1 + 2j−1 , 1 ≤ i < j ≤ r.
3. The signality k states are of the form

k
A  i2 . . . A
 i1 A  ik 0 ≡ 2ij −1 , (5.30)
j=1

for 1 ≤ i1 < i2 < · · · < ik ≤ r.

Remark 5.9 In the above, we underline expressions such as 2i−1 + 2j−1 to

indicate that this is not the vector sum of 2i−1 and 2j−1 but the vector in the
CBR corresponding to the sum of the integers 2i−1 and 2j−1 .
 5.13 Signality Classes 71

The same approach can be used to discuss maximal questions. Recall that these
are represented by the dual preferred basis elements of the form i, 0 ≤ i < 2r .
Then we can write

k
0Ai1 Ai2 . . . Aik ≡ 2ij −1 . (5.31)
j=1

Exercise 5.10 Prove that

Ai 0 = 0,  i = 0,
0A 1 ≤ i ≤ r. (5.32)

Exercise 5.11 Use the signal algebra to show that

ij = δ ij , 0 ≤ i, j, < 2r . (5.33)

5.13 Signality Classes

For each rank r quantum register Q[r] , we may write its preferred basis B [r] in
the form
 1 0, A
B [r] = {0, A  2 0, . . . , A
 r 0, A  2 0, . . . , A
 1A 2 . . . A
 1A  r 0}. (5.34)
For each element, the signality of that element is the number of signal creation
operators used to create it from the signal ground state.

Example 5.12 For a rank-11 quantum register, the preferred basis state
given by 851 in the CBR is given by 11 12 03 04 15 06 17 08 19 110 011 in the SBR,
from which we can read off its signality as 6. Equivalently, we can see that
this state is given by A  2A
 1A  5A
 7A
 9A
 10 0, which clearly has signality 6.

Such examples soon show the enormous advantage of using the CBR for large
rank registers. The CBR is particularly suited for computer algebra calculations.
By inspection, it is easy to see that the preferred basis B [r] separates into r + 1
disjoint signality classes:
signality zero B [r,0] ≡ {0},
signality one B [r,1] ≡ {A  a 0 : a = 1, 2, . . . , r},
signality two B [r,2]
≡ {A  aA b 0 : 1  a < b  r},
(5.35)
.. ..
. .
signality r B [r,r]
≡ {A  1A2 . . . A
 r 0}.
Then we may write the preferred basis as the union of all of these signality classes,
that is,

r
B [r] = B [r,a] . (5.36)
a=0
72 Classical and Quantum Registers

We define the signality count σ [r,a] as the number of elements in B [r,a] . Then
r!  
σ [r,a] = = ar (5.37)
a!(r − a)!
Hence the total signality σ [r] is given by

r 
r
r 
σ [r] ≡ σ [r,a] = a = 2a , (5.38)
i=0 a=0

as expected.
A significant feature of QDN is that the addition of register states of different
signality is permitted, unless the dynamics specifically rules it out. This will
happen, for example, whenever the signals are interpreted as electrically charged
particles. Under such circumstances, we expect conservation of total signality.

5.14 Binary Decomposition

Given a nonnegative integer k, the binary decomposition of k is the expansion
∞

k= k [a] 2a−1 , k = 0, 1, 2, . . . , (5.39)
a=1

where each of the binary coefficients k [a] is either zero or one. Table 5.2 shows
the binary decomposition up to k = 9.
From the table, we read off for example that
6 = 0 × 21−1 + 1 × 22−1 + 1 × 23−1 + 0 × 24−1 + 0 × · · · = 2 + 4 = 6.
For each k, the maximum a for which k [a] = 1 is the minimum rank of k,
denoted μ(k) and indicated by a box in Table 5.2. The minimum rank of k is the
rank of the “smallest” quantum register that contains the basis state k.

Table 5.2 The binary decomposition of the

integers up to nine

a → 1 2 3 4 ...
2a−1 → 1 2 4 8 ...

k↓ 0 0 0 0 0 ...
1 1 0 0 0 ...
2 0 1 0 0 ...
3 1 1 0 0 ...
4 0 0 1 0 ...
5 1 0 1 0 ...
6 0 1 1 0 ...
7 1 1 1 0 ...
8 0 0 0 1 ...
9 1 0 0 1 ...
 5.15 Computational Basis Representation for Signal Operators 73

Example 5.13 From Table 5.2 we read off

μ(1) = 1,
μ(2) = μ(3) = 2,
μ(4) = μ(5) = μ(6) = μ(7) = 3. (5.40)

We see μ(9) = 4 and 9[1] = 1, 9[2] = 0, 9[3] = 0, 9[4] = 1, so

9 = 1 × 21−1 + 0 × 21 + 0 × 23−1 + 1 × 24−1 = 1 + 8 = 9. (5.41)

5.15 Computational Basis Representation for Signal Operators

For a given rank-r register Q[r] , the preferred basis elements {k : k =
0, 1, 2, . . . , 2r − 1} are given in the CBR by

μ(k)
k= k [a] 2a−1 , k = 0, 1, 2, . . . , 2r − 1, (5.42)
a=1

where k [a] = 0 or 1 is the ath binary component of the integer k; i.e., we have
k = k [1] 21−1 + k [2] 22−1 + k [3] 23−1 + · · · + 2μ(k)−1 . (5.43)
Hence we may equate the CBR and SBR elements in the basis B [μk ]
 k[1]  k[2]  
1
k = k[1] k[2] . . . 1μk = A 2
A ... A μk 0. (5.44)
Now use the results
 ak = 0
A if k [a] = 1,
(5.45)
= k + 2a−1 if k [a] = 0.
Defining

k [a] ≡ 1 − k [a] , (5.46)
then we can readily show
 ak = 
A k [a] k + 2a−1 ,  a = k [a] k − 2a−1
kA
Aa k = k [a] k − 2a−1 , kAa = k [a] k + 2a−1 . (5.47)
Hence we must have
r
2 −1 r
2 −1
a
A = [a]
k k+2 a−1
k= k [a] kk − 2a−1 , (5.48)
k=0 k=0
r
2 −1 r
2 −1
Aa = k [a] k − 2a−1 k = 
k [a] kk + 2a−1 , (5.49)
k=0 k=0

which leads to
r
2 −1 r
2 −1
P =
a 
k [a] kk, a =
P k [a] kk. (5.50)
k=0 k=0
74 Classical and Quantum Registers

These results are consistent with

r
2 −1 r
2 −1
a a
P +P = (k[a] + 
k[a] )kk = kk = I[r] , a = 1, 2, . . . , r. (5.51)
k=0 k=0

Example 5.14 Using the above expansions, we find

Rank one
P = 00,  = 11
P
A = 01,  = 10.
A (5.52)
Rank two
P1 = 00 + 22, 1 = 11 + 33,
P
P2 = 00 + 11, 2 = 22 + 33,
P
A1 = 01 + 23,  1 = 10 + 32,
A
A2 = 02 + 13,  2 = 20 + 31.
A (5.53)
Rank three
P1 = 00 + 22 + 44 + 66, 1 = 11 + 33 + 55 + 77,
P
P2 = 00 + 11 + 44 + 55, 2 = 22 + 33 + 66 + 77,
P
P3 = 00 + 11 + 22 + 33, 3 = 44 + 55 + 66 + 77,
P
A1 = 01 + 23 + 45 + 67,  1 = 10 + 32 + 54 + 76,
A
A2 = 02 + 13 + 46 + 57,  2 = 20 + 31 + 64 + 75,
A
A3 = 04 + 15 + 26 + 37,  3 = 40 + 51 + 62 + 73.
A (5.54)

In general, the CBR for any signal operator in a rank-r register consists of a
sum of 2r−1 transition operators, all of which annihilate each other, including
themselves. Likewise, a product A  j of two different signal operators can be
 iA
r−2
expressed as a sum of 2 transition operators that mutually annihilate, and
so on. This process of representation can be continued until we arrive at the
saturation operator A 2 . . . A
 1A  r , which when applied to the signal ground state
creates the antithesis of the ground state, the fully saturated signal state 2r − 1 ≡
11 12 . . . 1r .
A particularly useful expression for the signal creation operators is obtained
by writing any of them in the form
 i = 2i−1 0 + Xi ,
A (5.55)
 2 −1 [i]
r
where the operator Xi ≡ k=1 k k + 2
i−1
k annihilates the signal ground
state. This expression can be used to greatly simplify calculations for those
experiments involving signality-one labstates, as in single-photon quantum optics
experiments.
 6
Classical Register Mechanics

6.1 Introduction
Classical mechanics (CM) is conventionally formulated according to partially
contextually complete Block Universe principles: an exophysical observer sitting
in some frame of reference looks in on a system under observation (SUO) and
assigns truth values to propositions about states of that SUO, but the details of
how these truth values are obtained are not given.
There is a curious aspect to this scenario: the contextual incompleteness of CM
is itself contextual. By this we mean that there are two scenarios, with different
general propositional classifications (GPCs).

The Experimentalist’s Perspective

From a classical experimentalist’s perspective, generalized propositions (GPs) in
CM take the form

P(P , ∅|∅, F ). (6.1)

Here P is some proposition of scientific interest, such as “The orbit of Jupiter

is an ellipse,” and F is relative external context describing the frame of reference
used to describe states of the SUO relative to the wider Universe external to that
SUO. This form of GP has a GPC of 2 according to the algorithm discussed in
Chapter 2.

Remark 6.1 Suppose an astronomer used a telescope T to test proposition

P . That would not upgrade the GP to one of the form P(P , T |∅, F ) with
a GPC of 3. The reason is that the use of a telescope is not essential in
this context. Historically, Kepler did not use a telescope to state his laws
of planetary motion: he used the data obtained by Tycho using naked eye
observations. On the other hand, the observation of the Hubble red shift using
telescopes fitted with spectrum analyzers would qualify for such an upgrade.
76 Classical Register Mechanics

The Theorist’s Perspective

Newtonian mechanics applied to the description of classical dynamically evolving
SUOs has the strength of mathematics: the classical laws of motion can be consid-
ered as a set of mathematical principles, axioms, and theorems, relative to which
propositions about classical states of SUOs can be tested. From this perspective,
the contextual information F about frames of reference becomes part of the
mathematical framework, and therefore should be included in relative internal
context. In this scenario, GPs take the form P(P , Laws of motion, F |∅, ∅), with
a GPC of 1, as befitting mathematical propositions.1

6.2 Classical Registers

We shall focus our attention on the second scenario discussed above, that of the
theorist’s perspective. Therefore, we shall not discuss apparatus or the observer.
Instead, we focus our attention on the mathematical structures that we employ
to model particles moving about in continuous space and time.
The first thing is to construct a model for classical space, objectifying it as a
vast collection of containers into which, and out of which, particles can move.
Now conventionally and until relatively recently, space has always been consid-
ered to be continuous. But that is an empirically vacuous assertion. Therefore,
we take the liberty of modeling space in discrete and finite terms. This leads us
to define a classical register R[r] of rank r, the Cartesian product of r classical
bits. We write
R[r] ≡ B 1 B 2 . . . B r , (6.2)
where the B i , i = 1, 2, . . . , r, are the individual labeled bits and we use bit labeling
to bypass the need for the Cartesian product symbol ×. We will consider classical
registers of sufficiently large rank so that they can model regions of classical
physical space over which particles can move. In this approach, particle motion
is discussed in terms of the tracking of signals from a vast collection of detectors
over time. A particularly useful feature of this approach is that signality need
not be conserved, which means that classical particle creation and annihilation
is readily incorporated into the formalism.

6.3 Architecture
In any such discussion it is important to establish the relevant spatiotempo-
ral architecture. There are two aspects of architecture relevant to the present
discussion. One has to do with the unspecified exophysical observer and their

1
It seems invalid to argue that in Newtonian mechanics, we have the laws of motion for
relative internal context and the observer’s frame for relative external context, so that
would appear to give a GPC of 3. That would be mixing theory and experiment to an
unacceptable degree. We have made the point before that mathematics is not physics.
 6.3 Architecture 77

apparatus, and the other is to do with the dynamics of the system under obser-
vation. We shall refer to these aspects as the external and internal architectures,
respectively.
External architecture involves time and persistence.

Time
In classical register mechanics, time is modeled as a succession of discrete stages,
ΣM , ΣM +1 , . . . , ΣN , representing the actions or lack of actions by an observer
in a physical laboratory. At stage Σn , the state of the SUO will be denoted Ψn ,
an element of a classical register Rn of rank rn .

Persistence
This refers to the question of identity of a given detector over a succession of
stages. In quantized detector networks (QDN), each detector is identified with
a single stage only. Therefore, there is no concept of “existential persistence” in
QDN. This applies as much to the observer and their apparatus as it does to the
SUOs that they observe. The first law of time (the dictum of Heraclitus), that
everything changes, applies here.
That does not rule out effective persistence, meaning that as the observer’s
time runs from stage to stage, their description of the apparatus seems to be
constant, in terms of the rank of the registers concerned and in terms of the
labeling of the detectors in those registers.
Effective persistence is the rule in ordinary experience, to the extent that
humans are strongly conditioned to believe that they are moving around in
time over a fixed spatial background. This belief carries over into conventional
descriptions of experiments, in which apparatus is assumed to persist while states
of SUOs evolve. What is really going on is that the first law of time always
applies, but the time scales for significant change in apparatus are generally so
great relative to those associated with the SUOs that the former time scales may
be taken to be infinite compared with the latter. This is a common assumption
made in high-energy particle physics scattering experiments, for instance.

Internal architecture depends on the assumed laws of mechanics, there being

three distinct types.

Time-Dependent Laws of Motion

The laws of mechanics for a given SUO could be contextual, changing in some
way from stage to stage, or they could be constant. An example of the former
scenario is demonstrated by phase diagrams in chemistry. These show under
which conditions of pressure and temperature a collection of molecules behaves
as a gas, a liquid, or a solid. Such behavior is a manifestation of emergent
processes driven by reductionist laws of physics, and the challenge is to explain
the emergent behavior in each phase using those reductionist laws.
78 Classical Register Mechanics

Autonomy or Not
An SUO could be autonomous, meaning that it has been effectively isolated
from its environment. Alternatively, the exophysical observer could arrange for
external agencies, such as electric and magnetic forces, to influence the dynamics
of an SUO.

Signality Conservation
Signality is taken in this chapter as a marker of particle number: a signality-
p state of a classical register at a given stage will be interpreted as a state
with p particles. Newtonian mechanics does not readily countenance changes of
particle number. If that has to be considered, then that can be readily modeled
by signality nonconserving register dynamics. The reason for this is that register
mechanics is more like a field theory than a particle theory.

For the rest of this chapter we shall restrict our attention to autonomous SUOs
moving over a succession of classical registers, each of which has the same rank r,
with stage-independent laws of dynamics, as these are generally of most interest.
In principle, there would be no problem in dealing with other forms of dynamics,
including those where the rank of the physical register changes with time. Indeed,
that is a common scenario in the quantum register dynamics we shall discuss in
the next chapter.
We could also deal with classical stochastic mechanics, which would incorpo-
rate Bayesian principles in a natural way, but for the rest of the chapter we shall
deal with deterministic laws of mechanics. In the following, the set of integers
{0, 1, 2, 4, . . . , 2r − 1} is denoted Z [r] .
We shall use the computational basis representation (CBR) Bn ≡ {kn : k ∈
Z [r] } to represent the 2r labstates in a rank r classical register Rn at stage Σn .
Consider the temporal evolution of a system from labstate Ψn in register Rn at
stage Σn to labstate Ψn+1 in register Rn+1 at stage Σn+1 . The dynamical rules
will be encoded into the expression
Ψn → Ψn+1 ≡ Cn+1,n Ψn , (6.3)
where Cn+1,n is a classical register operator. Such an operator maps any one of
the 2r labstates in Bn into precisely one of the 2r labstates in Bn+1 .
The reason for this constraint on any classical register operator comes from
the basic principles of CM: states in CM are well-defined, single-valued elements
of phase space at any given time. This translates in the present context to the
requirement that at any stage Σn , the classical state of an SUO is precisely
one of the 2r possible states of the register Rn . In consequence, any admissible
classical register operator acts on single elements of Rn and maps them into single
elements of Rn+1 . Note that this does not rule out nonsurjective or noninjective
operators; that is, there may be elements in Bn+1 that are not mapped into
(nonsurjective), and it could be the case that more than one element in Bn may
be mapped into the same element in Bn+1 (noninjective).
 6.4 Permutation Flows 79

Taking any initial state, there are 2r possible final states. Since there is a totalr
2
of 2r possible initial states, we immediately deduce that there is a total of (2r )
different possible classical register operators mapping from Rn into Rn+1 .
Even for low-rank registers, the number of possible operators can be impres-
sive. For a rank-one system, we deduce that there should be 22 = 4 different
possible such operators. Indeed, that is exactly what we saw in Chapter 3, where
we identified the four classical bit operators I, F , U , and D. The number of
possibilities grows rapidly with r. For example, there are 256 different operators
that can map a rank-two register into another rank-two register, and a total of
88 = 16,777,216 different operators mapping a rank-three register into another
rank-three register.
Given a potentially vast number of possible evolution operators, some criteria
need to be imposed in order to reduce the discussion to manageable and realistic
proportions. What comes to our assistance here is that most of the possible
evolution operators over a classical register will not be useful. Many of them
will correspond to irreversible and/or unphysical dynamical evolution and only
a small subset will be of interest. We need to find some principles to guide us in
our choice of evolution operator.
We turn to standard classical mechanics as discussed in Hamiltonian mechan-
ics. The first thing to note is that classical phase spaces are generally constant
in time. This corresponds to taking the rank of successive classical registers to
be constant, i.e., rn = rn+1 ≡ r for some integer r and for all n in the temporal
interval under discussion. We shall make this assumption from this point on.
Next, we recall that in standard CM of the Hamiltonian variety, Hamilton’s
equations of motion lead to Liouville’s theorem. This tells us that as we track a
small volume element along a classical trajectory,2 this volume remains constant
in magnitude though not necessarily constant in shape or orientation. This leads
to the idea that a system of many noninteracting particles moving along classical
trajectories in phase space behaves as an incompressible fluid, such a phenomenon
being referred to as a Hamiltonian flow.
An important characteristic of Hamiltonian flows is that flow lines never cross.
We shall encode this idea into our development of classical register mechanics.
There are two versions of this mechanics, one of which does not necessarily
conserve signality while the other does. We consider the first one now.

6.4 Permutation Flows

A classical rank-r register Rn contains 2r labstates denoted by kn , k ∈ Z [r] ,
in the CBR. Consider a permutation Pn of the integers k, such that under Pn ,
kn → kn+1 ≡ Pn (kn ) ∈ Z [r] . Define the evolution of the labstate kn over one

2
A classical trajectory is a phase space trajectory that satisfies Hamilton’s equations of
motion.
80 Classical Register Mechanics

stage by kn → kn+1 ≡ Pn (kn ). Such a process is reversible and will be referred

to as a permutation flow .

Example 6.2 Mathematicians often represent a permutation of, say, five

objects in the form (135)(24), which means the permutation 1 → 3 → 5 →
1, 2 → 4 → 2, and so on, where the notation a → b → c means that original
element a is replaced by, or goes to, element b, original element b is replaced
by element c, and so on. A group of elements within a given pair of brackets
is called a cycle.
Consider a rank-three classical register R[3] . This has a total of eight states
{k : k = 0, 1, 2, . . . , 7}. Under the permutation (0346)(1)(25)(7), the evolution
is given by

0 → 3 → 4 → 6 → 0, 1 → 1, 2 → 5 → 2, 7 → 7. (6.4)

This permutation flow does not conserve signality. The simplest proof of this
is to note that the signality zero state 0 evolves into a state with nonzero
signality.

There is a total of n! distinct permutations of n objects, so there are (2r )! pos-

sible distinct permutation flow processes. For large r, the number of permutation
r
flows is a rapidly decreasing fraction of the number (2r )2 of all possible forms
of rank-preserving classical register processes.
Permutation flows are restricted to constant rank registers, and even then, are
of debatable value in the following sense. Given two successive classical registers
Rn and Rn+1 of the same rank, the relationship between the labeling of states
in the two registers is contextual. By this we mean that the identification of
element 1n with element 1n+1 , 2n with 2n+1 , and so on, is up to the observer.
This is analogous to the parallel transport problem in general relativity, where
contextual information relating initial and final tangent spaces along a path is
required before a notion of parallelism can be established.

Example 6.3 Consider a permutation written as a transformation from R0

to R1 in the form

i0 → p(i)1 : i = 0, 1, 2, . . . , 2r − 1. (6.5)

Now suppose that we passively relabel the 2r states in R1 by the rule

p(i)1 → i1 . Then the permutation reduces to the “identity” transformation

i0 → i1 , i = 0, 1, 2, . . . , 2r − 1. (6.6)

This relabeling seems to have eliminated the significance of permutation

dynamics.
This apparent paradox is resolved by context: if we had no knowledge of the
original permutation (6.5) or of the physical meaning of the original states and
 6.5 Signality-Conserving Flows 81

the final states, then (6.6) would indeed be trivial. If we had such contextual
information, however, then the relabeling p(i)1 → i1 would have no physical
significance, and the real dynamics would be understood by the observer.

Permutation flows have the following features with analogues in CM.

Reversibility
Permutations form a group, which means that given a permutation P , then its
inverse P −1 always exists. Hence permutation flows are reversible.

Orbits
Any SUO evolving under stage-independent permutation dynamics will demon-
strate patterns known as orbits or cycles. Given a permutation of 2r objects,
there will exist subsets known as cycles of the objects being permuted such that
only elements within a given cycle replace each other under the permutation. This
is relevant here because we have chosen to discuss time-independent autonomous
systems, the evolution of which is given by repeated applications of the same
permutation. Therefore, the structure of the cycles does not change and so each
cycle is stable, consisting of the same p elements with a dynamical period p. For
example, the identity permutation gives a trivial form of mechanics where nothing
changes. It has 2r cycles each of period 1. At the other end of the spectrum, the
permutation denoted by (0 → 1 → 2 → · · · → 2r − 1 → 0) has no cycles except
itself and has period 2r .
Any physical register evolving under time-independent, autonomous permu-
tation mechanics must return to its initial labstate no later than after 2r time
steps. This is the analogue of the Poincaré recurrence theorem (Poincaré, 1890).

6.5 Signality-Conserving Flows

Most permutation flows will not conserve signality, as Example 6.2 shows. Sup-
pose now we have decided to identify signality with particle number. Then expe-
rience with Hamiltonian mechanics, where particle conservation is the general
rule, leads us to investigate signality-conserving flows.
We can readily identify the subset of the permutation flows that conserve
signality by using the signality classes discussed in the previous chapter. Suppose
Un+1,n is an evolution operator with the following characteristics:
1. Un+1,n 0n = 0n+1 : such an evolution is called calibrated .
2. Un+1,n permutes signality-1 states, that is, Un+1,n A  i 0n = A iP 0n+1 , for
n n+1
i = 1, 2, . . . , r and with iP the number into which i is transformed under the
permutation.
3. Likewise for each signality class, until finally, we have for the fully saturated
signal state Un+1,n 2r − 1n = 2r − 1n+1 .
Then clearly, signality is conserved under such a dynamical scheme.
82 Classical Register Mechanics

The total number of distinct permutations of r objects is r!, so there are

that many distinct forms of signal permutation dynamics for a rank-r classical
register. Since there are (2r )! distinct forms of permutation dynamics, the set of
signal permutation dynamics forms a rapidly decreasing fraction of the set of all
possible permutation dynamics.

6.6 Evolution and Measurement

Any experiment consists of several distinct phases. An important phase is the
process of measurement itself, which ends with the extraction of classical informa-
tion from an SUO. Typically this information will be in the form of real numbers,
and these can always be expressed in binary form, justifying our approach.
Context plays a vital role here. When, for example, an observer reports that a
particle has been observed at position x = 1.5, what they mean is that positive
signals have been detected at some normal detector or detectors associated with
the number x = 1.5. This assignment is based on the context of the experiment:
the observer will know on the basis of prior theoretical and empirical knowledge,
gained during the process of calibration (preparation of apparatus) what those
detectors mean in terms of the physics of the SUO concerned, and therefore,
what values of some measurable quantity those signals represent.
There is room here for error, in that the observer could associate the wrong
context to the signals being observed. Such a process occurred in the Mars
Climate Orbiter disaster in 1999, when there was a “failure to use metric units in
the coding of a ground software file” (Mars Climate Orbiter Mishap Investigation
Board, 1999). In the following, we assume all contexts have been interpreted
correctly by the observer.
So far we have discussed the evolution of labstates. For each run or repetition
of the experiment, this is modeled by the action of an evolution operator UN,0
mapping initial labstates at stage Σ0 into final labstates at stage ΣN . We need
now to discuss how numbers are extracted at the end of an experiment consisting
of a number of runs.
With reference to the position measurement discussed above, we model the
measurement process in terms of weighted relevant questions. What this means
is this. Suppose the final physical register RN has rank rN . Assuming the exper-
imentalist has established that each detector is working normally, then there will
be a total of 2rN possible normal labstates in this register. Therefore, the observer
could ask a total of 2rN maximal questions. These questions are represented by
the dual labstates k : k = 0, 1, . . . , 2rN − 1 . Given a final labstate ΨN , the
answer yes or no to each question kN ≡ is it true that ΨN = k? is repre-
sented by the number one or zero, respectively, and given by the answer kN ΨN .
Now the observer will generally have some theory as to what each answer
kN ΨN means physically. In many experiments, this will be some real number
X k . Therefore, the actual number XΨN obtained at time tN at end of a single
run of the experiment can be written in the form
XΨN = ΨN XN ΨN , (6.7)
 6.7 Random Initial States 83
2rN −1
where XN ≡ k=0 kN X k kN is an classical observable, a sum of dyadics
representing a weighted relevant question.
Two comments are relevant. First, despite appearances, this is still a classical
theory at this point. The final labstate ΨN is a single element in the final physical
register, RN , not a superposition of elements. Second, there is nothing in classical
mechanics that rules out weighted sums of dyadics. For any element in RN , all
the possible answers kN ΨN are zero except for one of them, so (6.7) returns a
physically sensible value for XΨN .
A further refinement, anticipating the possibility of random variations in the
initial state and the extension of these ideas to quantum mechanics, is to write
XΨN = T r{XN N }, (6.8)
where T r represents the familiar trace process and N is the dyadic ΨN ΨN
analogous to a pure state density operator in quantum mechanics.
We note that ΨN = UN,0 Ψ0 and ΨN = ΨN UN,0 , where the evolution
operator UN,0 maps elements of R0 into elements of RN and similarly for the
dual evolution operator UN,0 . In general, it will be true that
UN,0 UN,0 = I0 , (6.9)
the identity operator for R0 . However, because there is no requirement formally
in this approach for the rank rN of the final physical register RN to equal the
rank r0 of the initial physical register R0 , it is possible that UN,0 UN,0 does not
equal IN . This corresponds to irreversible dynamics. In the analogous quantum
formalism that we will discuss in the next chapter, such evolution operators are
referred to as semi-unitary, and UN,0 is the retraction of UN,0 .
Using (6.9) in (6.8), we may write XΨN = T r{XN UN,0 0 UN,0 }, where 0 is
the initial dyadic Ψ0 Ψ0 .

6.7 Random Initial States

Real experiments normally consist of a large number of repetitions or runs of a
basic process. However, it cannot always be guaranteed that the initial labstate is
always the same. In principle, we could start with any one of 2r0 initial labstates.
In such a case, a statistical approach can be taken.
Consider an extremely large number R of runs, such that there is a total
of R k
runs starting with initial labstate k0 , for k = 0, 1, . . . , 2r0 − 1. Clearly,
2r0 −1 k
k=0 R = R. Then in the limit of R tending to infinity, we would assign a
probability ω k ≡ limR→∞ Rk /R for the initial labstate to be in state k0 .
In such a scenario we define the initial density matrix
r0
2 −1
0 ≡ ω k k0 k0 , (6.10)
k=0

where k0 is any one of the 2r0 elements of the initial physical register R0 and
the ω k are probabilities summing to unity. The formalism outlined above then
gives the expectation values of operators.
 7
Quantum Register Dynamics

7.1 Introduction
In this chapter we move beyond the classical register scenario discussed in the
previous chapter, extending the discussion to experiments described by time-
dependent quantum registers of varying rank. We apply our previous discussion
of the signal basis representation (SBR), the computational basis representation
(CBR), signal operators, signal classes, and the CBR of signal operators to the
quantum case. Our discussion of dynamics covers persistence, that is, the stability
of apparatus, observers, and laboratory time, and the Born probability rule. We
state the principles of quantized detector network (QDN) dynamics and show
how they apply to the description of quantum experiments. We discuss the signal
theorem and path summations.

7.2 Persistence
Our first step is to clarify what QDN assumes about the evolution of appara-
tus in time, because this affects the modeling. QDN is designed to reflect the
behavior of apparatus in the real world and so it is not assumed in general that
a given observer’s apparatus is constant in time, even during a given run of an
experiment.
Although many experiments appear to be carried out with apparatus that
persists over any given run of the experiment, and indeed, perhaps over all the
runs of that experiment, that is just an incidental factor that reflects no more
than an economy in construction. In practice it is invariably easier and more
economical to use the same equipment over and over again rather than use it
once, throw it away after each run, and build a new version ready for the next
run. Lest this be thought of as a trivial point, it is nevertheless an integral
and costly feature of many experiments, involving maintenance and upgrading.
Indeed, in laboratories such as the Large Hadron Collider, actual run time is a
 7.3 Quantized Detector Networks 85

small fraction of total project time. A related, significant issue has to do with
the concept of ensemble, discussed in the Appendix.
Persistence has everything to do with time scale, specifically, the relative
laboratory time over which any piece of equipment can be meaningfully discussed
as such. The degree of persistence of apparatus is as contextual as anything else
in physics. If a run is relatively brief, say, over in very small fractions of a second,
as in high-energy particle scattering experiments, then in such an experiment,
the apparatus will behave as if it persists forever. On the other hand, in some
experiments, individual runs can involve enormous intervals of time between
state preparation and observation, as always happens in the case of astrophysical
observations of stars and galaxies. In such cases, light from a distant star may
be received by astronomers long after that star had ceased to exist as a star.
A necessary criterion for an experiment to be describable by QDN is that any
detector exists at least during a given stage. Indeed, we can take the definition
of a stage to be that interval of laboratory time over which a given detector can
be assumed to persist, that is, have an identity that has physical relevance in the
context of the experiment concerned.
In this respect, a stage is to be identified not with a moment or point in time
but with the interval of time over which information could be extracted by the
observer. Such an interval is of contextual temporal length in the laboratory,
being as short or as long as the observer requires to acquire a bit or “quantum”
of information. The temporal divide between two stages will be called a quantum
tick , or q-tick . It is necessarily a heuristic concept, because any attempt to
empirically measure a q-tick and assign a numerical value to a chronon, or
assumed fundamental unit of time, will be self-referential in some way or other.
We note that the concept of Planck time, commonly regarded as the shortest
meaningful interval of time in physics, is a theoretical construct that could
never be observed directly and has been criticized on such grounds (Meschini,
2007).

7.3 Quantized Detector Networks

In the projection-valued measure (PVM) formulation of standard quantum
mechanics (QM) (von Neumann, 1955; Peres, 1995), it is generally asserted that
state vectors of systems under observation (SUOs) evolve in Hilbert spaces of
fixed dimension. Any time dependence of the apparatus itself, such as externally
imposed time-dependent electromagnetic fields, is encoded into an explicit time
dependence of the Hamiltonian or whatever observables are involved. This
approach encodes the idea that experiments are done in fixed laboratories and
that information is extracted from states of SUOs that evolve unitarily in time.
This temporal architecture is known as the Schrödinger picture and is widely
used in QM. A variant but nominally equivalent temporal architecture is that of
the Heisenberg picture, in which the quantum states appear frozen in time but
86 Quantum Register Dynamics

now it is the observables of the theory that have unitary temporal evolution over
and above any intrinsic, explicit evolution.
In standard QM, these two related architectures were eventually recognized as
too limited. One reason is that in the PVM approach, the number of mutually
orthogonal possible outcome states associated with a given observable is the
number of its eigenstates, which is fixed and cannot exceed the dimension of
the Hilbert space concerned. The positive operator-valued measure (POVM)
formalism on the other hand was developed to deal with the possibility that
the number of observational outcome channels set up in a laboratory may be
different from the dimension of the Hilbert space used (Ludwig, 1983a,b; Kraus,
1983; Peres, 1995) and, indeed, may exceed that number, in the case of finite
dimensions.
In contrast, the QDN formalism assumes from the outset that the Hilbert space
representing outcome possibilities is always different from one stage to the next,
even if the dimensionality remains constant, and even if the detectors involved
in the experiment appear to persist over several or all stages of an experiment.
To be specific, let us denote the observer’s apparatus at any given stage Σn by
An . Some parts of An will correspond to actual, real detectors, while other parts
will be virtual, or potential, detectors. Whatever the case, An will be always be
modeled by a countable number rn of vertices at any stage of the relevant stage
diagram for that experiment. In this book, we denote the ith vertex at stage Σn
by in , where the superscript i runs from 1 to rn and labels all of the distinct
vertices in that stage. Note that in (a vertex) should not be confused with in
(bold font), our notation for a CBR element of the preferred basis at stage Σn .
Note also that detector vertices should not be confused with modules, which are
nondetecting parts of apparatus that sit in the information void between stages.
In the description of real experiments, rn will always be finite, in contrast to
the situation in QM, where the Hilbert space may be infinite dimensional. The
harmonic oscillator in one spatial dimension is an example where QM assumes
either that the position observable has a continuous spectrum or, equivalently,
that the Hamiltonian has a countable infinity of energy eigenstates. The harmonic
oscillator is discussed in Chapter 24. In the world of empirical reality, there are
no such harmonic oscillators, just SUOs for which such a description may be a
reasonable approximation.
The question as to whether rn is finite or not is central to many if not all of
the technical difficulties encountered in the refinement of standard QM known as
quantum field theory. The harmonic oscillator appears to be intimately involved
in many of these problems in one way or another. Although the mathematical
properties of the quantized harmonic oscillator play an essential role in account-
ing for the particle concept in free quantum field theory, those same properties
generate fundamental problems in interacting field theories. For instance, the
ultraviolet divergences encountered in most Feynman loop integrals are linked
to the unbounded energy spectrum of the standard QM oscillator, while infrared
 7.3 Quantized Detector Networks 87

divergences are linked to the assumed continuity of spacetime and the zero-point
energy of the quantized oscillator.
The issue of zero-point energy of the quantized oscillator was explicitly
addressed in quantum optics by Glauber in his landmark papers on photon
correlations (Glauber, 1963a,b,c). He pointed out that the irreversibility of
quantum detection processes involves the non-Hermitian positive energy electric
field operator E (+) (t, x) rather than the Hermitian electric field operator
E(t, x) ≡ E (+) (t, x) + E (−) (t, x). It is the vacuum expectation values (VEVs)
involving the E field operators that are affected by zero-point energy, whereas
VEVs involving the E (+) alone are not. Glauber’s conclusion was the following:

The electric field in the vacuum undergoes zero-point oscillations which, in the
correctly formulated theory, have nothing to do with the detection of photons.
(Glauber, 1963c)

According to Glauber, then, the standard quantum field theory approach to

quantum detection, if it involves VEVs of Hermitian operators, will be inad-
equate to model irreversible, local detection processes as they actually occur
in real laboratories. The reason the standard Lehmann–Symanzik–Zimmerman
(LSZ) (Lehmann et al., 1955) scattering formalism works is that the in and
out Hilbert spaces representing prepared and outcome states, respectively, are
applied at infinitely remote times in the past and future, respectively, in a
causally irreversible way. In between those two remote times, a quantum process
is technically in what we have called in earlier chapters the information void, a
laboratory regime during which no signal detection takes place. During such a
regime, unitary evolution can be assumed to take place without violating any
of the principles of QM. That is the essential reason why the standard model
Lagrangian, which respects CPT1 inversion symmetry and is used to work out
unitary evolution, is a good basis for particle scattering calculations. However,
finite time processes, which are generally not considered in high-energy physics,
would require the same review that Glauber carried out for photonic processes.
Given the apparatus An at stage Σn , its associated detectors, both real and
virtual, are represented by a set of signal qubits {Qin : i = 1, 2, . . . , rn }, with
qubit Qin being identified with vertex in . Here as elsewhere, upper indices label
individual qubits and detectors, while lower indices denote stages. The tensor
product Q1n Q2n . . . Qrnn , plus the information held by the observer about the
physical significance of those qubits, constitute a quantum register of rank rn .
This is a Hilbert space of dimension 2rn .
A fundamental property of any quantum register of rank greater than one is
that it contains entangled states as well as separable states. Separability and
entanglement are discussed in some detail in Chapter 22.

1
Charge, space (parity), and time.
88 Quantum Register Dynamics

Example 7.1 Given a rank-three quantum register Q1 Q2 Q3 , the state

Ψ ≡ 11 02 13 − 01 02 03 is partially separable relative to the chosen basis,
because we can write Ψ = {11 13 −01 03 }02 . On the other hand, Φ ≡ 11 02 13 −
01 12 03 cannot be written as a product, so is entangled relative to the chosen
basis.

In QDN, entanglement is regarded as contextual on the observer’s information

about their apparatus, and not as an intrinsic property of SUOs. QDN tries
to avoid terms such as “entangled photons,” but we reserve the right to use
such terminology occasionally, provided it does not mislead. The concept of an
entangled labstate is not only perfectly acceptable in QDN but actually essential
for the correct calculation of outcome signal detection probabilities in many
experiments.

7.4 Persistence and Ensembles

A conventional assumption in QM is that pure states of a system under obser-
vation may be represented by time-dependent elements of a fixed Hilbert space.
The chosen Hilbert space is usually assumed fixed for two reasons. First, there
is the conditioned belief that an SUO “exists” in time as a separate entity long
enough for the observer to study it. Another contributory factor is the persistence
of the apparatus, or the tendency of actual apparatus to exist in its original form
and functionality in a laboratory before and after its useful role has ended.
Most physics experiments deal with persistent apparatus. That is generally
arranged by the observer as a matter of economy: experimentalists generally do
not have the resources to scrap their apparatus at the end of each run and then
rebuild it in time for the next run.
There are situations, however, where persistence cannot be assumed. For
example, astronomers can catch light from a supernova shock wave only during
an extremely limited time, and that particular observation cannot be repeated
because the source of the signal has long gone. What helps the observers is the
vast numbers of photon signals that they manage to detect during that limited
window of opportunity.
A similar issue arises in quantum cosmology. The Universe is believed to be
expanding. On that account, any approach to quantum cosmology should take
the attendant irreversibility into consideration, and not treat the evolution of
the Universe in traditional QM terms as a typical SUO. The Universe is in an
ensemble of one, and not only contains the observer and the apparatus, but will
outlast both of them.
In QDN, individual detectors are never persistent. Each detector is assigned a
particular stage at which it operates as an detector, and outside that time, has
no role in the formalism. This is the QDN analogue of the concept of an event
in relativity. Some applications of QDN will for convenience assume persistence
 7.5 Observers and Time 89

of apparatus (Jaroszkiewicz, 2004), the only effect being to increase the number
of qubits used in the formalism.

7.5 Observers and Time

Observers generally come equipped with their own sense of time, and quantum
experiments are carried out relative to that time. Relativity teaches that there
are two observer-related time concepts with different properties; coordinate time
and proper time. In both special relativity (SR) and general relativity (GR), the
former time concept is used to label events in spacetime and is generally locally
integrable. This means that spacetime can be discussed in terms of coordinate
patches (Schutz, 1980). Within a given coordinate patch, events can be labeled by
spacetime coordinates in a path-independent way. When a particular coordinate
patch is related to clocks and rods in a specific laboratory, we shall refer to
that coordinate time as labtime. In GR, the description of labtime requires some
assumption about time-like foliations of spacetime and frame fields. On the other
hand, proper time is nonintegrable, which is to say that the proper time between
two events depends on the particular path taken between those events. In other
words, proper time is contextual.
In QDN, the time parameter associated with an experiment can normally be
identified with the proper time of an idealized inertial observer moving along a
time-like worldline, and for whom their laboratory appears to be at rest at all
times. In some situations, we may have to apply QDN to what we call interframe
experiments. These involve state preparation in one inertial frame and signal
detection in another. The Doppler effect in observational cosmology is an example
of interframe physics.
What are important in such situations in SR and GR are space-like hypersur-
faces: these are the analogues of the concept of stages in QDN. On a space-like
hypersurface, no two events can be causally connected; that is, no event can be
the cause of any other event on the same hypersurface of simultaneity. In QDN,
the analogous statement is that no two detectors on a given stage can affect each
other in any way. This could be because the detectors are really relatively space-
like separated, but the possibility exists, frequently in real experiments, that the
detectors are relatively time-like separated but shielded from each other.
In the real world, observers have finite existence: they come and go. Observers
and their apparatus are created at certain times and disappear at later times,
as seen by other observers in the wider universe. QDN as formulated here allows
for a discussion of different observers, each with their individual time param-
eters and lifetimes. The use of quantum registers also raises the possibility of
accounting for the origin of various temporally related concepts such as light
cones, time dilation, and other metric-based phenomena in terms of quantum
register dynamics. A useful way to discuss what is going on is in terms of causal
sets, the structures of which arise naturally within quantum register dynamics
(Eakins and Jaroszkiewicz, 2005). Causal sets are discussed in Chapter 23.
90 Quantum Register Dynamics

During their operational lifetimes, observers quantify their time in terms of real
numbers, usually read from clocks. Most clocks give only a crude estimate of the
passage of time, and as a result, the ordinary human perception of time as a one-
dimensional continuum is just a convenient approximation. The classical view
of time is that it is a continuum at all scales and for all phenomena. Certainly,
things appear consistent with that view in the ordinary world.
In QM, however, the situation is quite different. What matters in a quantum
experiment is information acquisition from the observer’s apparatus and this can
only ever be done in a discrete way, regardless of any theoretical assumption to
the contrary (Misra and Sudarshan, 1977). While an observer’s effective sense
of time can be modeled accurately as continuous, it is certainly the case that
an observer can look at a detector and determine its status in a discrete way
only. There are no truly continuous-in-time observations. It is important here to
distinguish between what experimentalists actually do in experiments and what
theorists imagine they do.
The discreteness of the information extraction process forms the basis of the
time concept in QDN. In general, a given observer will represent the state of
their apparatus (the labstate) at a finite sequence of their own (observer) times,
denoted by the integer n. These times will be referred to as stages. In QDN, a
pure labstate at stage Σn will be denoted by Ψn .
In QDN, stage Σn+1 is always regarded as definitely later than stage Σn . There
is no scope in QDN for the concept of closed time-like curve (CTC) found in some
GR spacetimes, such as the Gödel model (Gödel, 1949). Invariably, discussions
that do involve CTCs cannot accommodate quantum processes properly, at least
not those that involve probabilities.2
A final point on this topic: there is no need to assume that the laboratory time
interval between stages has a definite value or that a succession of stages involves
equal intervals of labtime.

7.6 The Born Probability Rule

One of the most significant attributes of quantum processes is the randomness
of quantum outcomes. Given identical state preparation, different runs of a
given experiment generally demonstrate controlled unpredictability. Controlled,
because the observer may know all about the range of possible outcomes and the
probabilities of those outcomes before one is actually observed, but unpredictable
because the observer cannot in general say beforehand which particular one of
those outcomes will occur in any particular run.

2
This is for the good reason that the probability concept requires an observer accumulating
information with an irreversible sense of time, something that a genuine CTC is
incompatible with.
 7.6 The Born Probability Rule 91

Remark 7.2 This subject is a minefield of issues that lurk unseen until
we stumble across them. Reference was made to “identical state preparation”
in the previous paragraph. That is clearly a vacuous concept. By definition,
different runs of a given experiment cannot have absolutely identical state
preparation: the Universe will have aged for sure during any two runs,
and there will be vast changes in the local environment of any apparatus
on atomic scales. Yet clearly, most of those changes will not influence the
outcome probabilities. As pointed out by Kraus, what appears significant
are equivalence classes of state preparation processes (Kraus, 1983). Exactly
where the line between influence and noninfluence of external factors in state
preparation should be drawn is one of those deep questions that no one knows
anything much about. The Heisenberg cut is a hypothetical line demarcating
the classical world of the observer and the quantum world of the SUO. We
propose the term Kraus cut to denote the hypothetical line between those
factors that have no influence on state preparation or outcome and those
that do. All we can do is to observe that some things are critical in state
preparation, and everything else seems unimportant. Much the same point is
emphasized in Peres (1995).

In practice the QM approach to probability works brilliantly and we use it in

QDN. The Born probability rule (Born, 1926) in QM states that if a final state
|Ψ is represented by a superposition of the form

d
|Ψ = Ψi |i, (7.1)
i=1

where the possible outcomes are represented by orthonormal vectors |i, i =

1, 2, . . . , d, in some Hilbert space, then the conditional probability Pi of outcome
|i is given by Pi = |i|Ψ|2 , if the final state is normalized to unity.
This rule is used in much the same way in QDN, as follows. Consider a
normalized pure labstate Ψn at stage Σn . This can always be expanded in terms
of the computational basis representation (CBR) of the preferred register basis
Bn at that stage in the form
n −1
2r
Ψn = Ψin in , (7.2)
i=0

where the coefficients Ψin are complex and

n −1
2r
|Ψin |2 = 1. (7.3)
i=0

In QDN, labstates are usually normalized to unity for convenience. Because the
preferred basis states form a complete orthonormal basis set, we may immediately
92 Quantum Register Dynamics

read off from (7.2) the various CBR conditional probabilities Pr(in |Ψn ), which
are given by the rule

Pr(in |Ψn ) ≡ |in Ψn |2 = |Ψin |2 , 0  i < 2rn . (7.4)

Pr(in |Ψn ) is the conditional (Bayesian) probability for the observer to find the
apparatus in signal state in at stage Σn , if the observer looked at their apparatus
at that time. These probabilities are conditional on the observer being sure, just
before they looked, that the labstate at stage Σn was Ψn .
There is no natural restriction in QDN to labstates that are eigenstates of
signality; i.e., superpositions of basis states from different signality classes are
permitted in principle. QDN is analogous in this respect to the extension of
Schrödinger wave mechanics to Fock space and to quantum field theory.

7.7 Principles of QDN Dynamics

We are now in a position to discuss the principles of labstate dynamics from
the perspective of a single observer. At stage Σn , this observer will hold in their
memory current information about their apparatus An , the associated quantum
register Qn , and the labstate Ψn (assumed pure here). An analogous statement
will hold for each stage in a finite sequence of stages running from some initial
stage Σ0 to some final stage ΣN , where N > 0. QDN does not assume observers
exist over unbounded intervals of time, so the formalism is applied only over a
finite number of stages.
We restrict attention in this chapter to pure labstates, that is, labstates fully
specified by single elements in a quantum register. A mixed-state, density matrix
approach to QDN dynamics is discussed in Chapter 9.
For the most basic sort of experiment, labstate preparation will be assumed to
have taken place by initial stage Σ0 and outcome detection is to take place at final
stage ΣN . For each integer n such that 0  n  N , the observer associates with
their apparatus An at that time a quantum register Qn . This register is the tensor
product Qn ≡ Q1n Q2n . . . Qrnn of a finite number rn of qubits, Q1n , Q2n , . . . , Qrnn ,
each qubit Qin representing the ith real or virtual detector in in An . The quantum
register Qn is a Hilbert space with preferred basis Bn consisting of the 2rn CBR
signal states.
There is no requirement in QDN or implication in our notation for the detector
represented by Qin+1 to be related in any obvious way to the detector represented
by Qin ; that is, we do not assume persistence. In other words, successive quantum
registers are understood as different Hilbert spaces, even if they have the same
rank, that is, if rn+1 = rn . This is one of the factors which makes QDN more
general in its scope than standard QM, although all the principles of QM are
incorporated in QDN.
At stage Σn , the observer describes the quantum state of their apparatus
at that time by a labstate Ψn , which is some normalized vector in Qn . Using
 7.8 Born Maps and Semi-unitarity 93

the CBR, this state can be written in the form (7.2) and normalized according
to (7.3).
A given run of an experiment will be described by the observer in terms of
a sequence {Ψn : 0  n  N } of normalized labstates, each element of which is
associated with a particular quantum register Qn , ending with state outcome
observation at the final stage ΣN .3 The question now is how successive labstates
relate to each other between times M and N .
Provided each run is prepared in the same way, and provided the apparatus
during eah run is controlled in the same way, we can discuss a typical labstate Ψn
as a representative for an ensemble of runs. Indeed, that is the only interpretation
that makes sense if an objectivized or hidden-variables interpretation of the wave
function is to be avoided. QDN follows the traditional QM view as emphasized
by Peres in this respect (Peres, 1995).
The dynamical transition from labstate Ψn to labstate Ψn+1 involves a map-
ping from quantum register Qn to quantum register Qn+1 that satisfies two
criteria fundamental to quantum mechanics: linearity and norm preservation.
This leads us to give a number of definitions and theorems that have proved
central to QDN.

7.8 Born Maps and Semi-unitarity

In the following, we restrict the discussion to Hilbert spaces of finite dimension.
This is in line with our general philosophy in QDN that there are no empirically
observable infinities in physics.

Definition 7.3 A Born map is a norm-preserving map from one Hilbert

space H to some other Hilbert space H ; if Ψ in H is mapped into Ψ ≡ B(Ψ)
in H by a Born map B, then (Ψ , Ψ ) = (Ψ, Ψ).

Born maps are used in QDN in order to preserve total probabilities (hence
the terminology), but unfortunately, their properties are insufficient to model all
quantum processes. Born maps are not necessarily linear, as can be seen from
the elementary example B(Ψ) = |Ψ|Φ for all Ψ in H, where Φ is a fixed element
of H normalized to unity and |Ψ| is the norm of Ψ in H. To go further, it is
necessary to impose linearity.
Henceforth, we adopt the rule that if U is a linear map, then we may drop the
parentheses and write U Ψ instead of U (Ψ).

Definition 7.4 A semi-unitary operator is a linear Born map. If U is such

a map, then for any elements ψ, φ in H and complex α, β, we may write
|αψ + βφ| = |αU ψ + βU φ| .

3
QDN allows partial observations to be made at intermediate stages.
94 Quantum Register Dynamics

The proofs of the following theorems are left to the reader.

Theorem 7.5 A semi-unitary operator from Hilbert space H to Hilbert space

H exists if and only if dim H  dim H .

Theorem 7.6 A semi-unitary operator from H to H is an injection; that

is, {U ψ = U φ} if and only if ψ = φ.

Theorem 7.7 If U is a semi-unitary operator from H to H , then its

retraction U exists; that is, we have U U = IH , where IH is the identity
operator over H.

Corollary 7.8 A semi-unitary operator preserves inner products and not

just norms. This means that an orthonormal basis set for Hilbert space H is
mapped by a semi-unitary operator into a mutually orthonormal set of vectors
in H with the same cardinality.

Theorem 7.9 If U is a semi-unitary operator from H to H and dim H =

dim H , then the retraction U of U is also a semi-unitary operator from H to
H. For such an operator, U U = IH and U U = IH .

Definition 7.10 An operator U satisfying the conditions of Theorem 7.9

will be called a unitary.

Remark 7.11 If U is a semi-unitary operator from H to H and dim H <

dim H , then U U = IH , simply because the retraction U is defined not over
H but over the proper subset U (H) of H .

7.9 Application to Dynamics

It is normally assumed in QDN that a labstate Ψn in Qn at stage Σn is mapped
into a labstate Ψn+1 in Qn+1 by some Born map Bn+1,n . Because Ψn+1 Ψn+1 =
Ψn Ψn under such a map, the Born rule used in conjunction with the signal bases
Bn and Bn+1 means that total probability is conserved. This is not the same
thing as conservation of signality, charge, particle number, or any other quantum
variable.
The following three scenarios are possible.
 7.9 Application to Dynamics 95

Bn+1,n Is Nonlinear
By Theorem 7.5, nonlinearity is unavoidable if the rank rn of Qn is greater than
the rank rn+1 of Qn+1 , but can arise even if this is not the case. Nonlinearity
here is interpreted as a marker of classical intervention by the observer. For
example, switching off any apparatus at stage Σn+1 would be modeled by the
Born map Bn+1,n (Ψn ) = 0n+1 for any state Ψn in Qn , where 0n+1 is the signal
ground state of the apparatus at stage Σn+1 . Another example is state reduction
due to observation; that is, if at stage Σn+1 the observer actually looks at the
apparatus and determines its signal status, then this would be modeled by the
nonlinear Born map Bn+1,n (Ψn ) = kn+1 , where now kn+1 is some element
of the CBR of the preferred basis Bn+1 , chosen randomly with a probability
weighting given by the Born rule. In this particular case, however, there are
actually two labstates associated with stage Σn+1 : Ψn+1 representing the state of
the apparatus immediately prior to state reduction and k representing the actual
observed outcome immediately after. None of this represents anything more than
mathematical modeling of the observer’s actual or potential knowledge about the
signal status of their apparatus. QDN makes no comment on whether anything
deeper than changes in apparatus signal status has occurred. In particular, any
speculation of superluminal information flow concerns correlations, which is a
phenomenon that occurs in all forms of mechanics and has everything to do with
what the observer did in the past.

Bn+1,n Is Linear and rn = rn+1

This scenario corresponds to unitary evolution in standard QM, and to reflect
this, we use the notation Bn (Ψn ) ≡ Un+1,n Ψn = Ψn+1 . From Theorem 7.9,
Un+1,n in this case satisfies the rules

Un+1,n Un+1,n = In , Un+1,n Un+1,n = In+1 , (7.5)

where Un+1,n is the retraction of Un+1,n . In such a case it is reasonable to call

Un+1,n unitary, being the formal analogue of a unitary operator in QM.

Bn+1,n Is Linear and rn < rn+1

In this case we use the same notation as in the second case above, i.e.,
Bn+1,n (Ψn ) ≡ Un+1,n Ψn = Ψn+1 , but now Un+1,n is properly semi-unitary
and only the first relation Un+1,n Un+1,n = In in (7.5) carries over. Such a
scenario arises in particle decay experiments, for example. These are discussed
in Chapter 15.

We cannot in general expect the rank rn of the quantum register Qn to be

constant with n, so if we wish to preserve probability and restrict the dynamical
evolution to be linear in the labstate, then we have to assume

rM  rM +1  · · ·  rn  · · ·  rN , (7.6)
96 Quantum Register Dynamics

where ΣM is the initial stage and ΣN is the final stage. From this, we can
appreciate that unless experimentalists are careful, their quantum registers will
grow irreversibly in rank. On the other hand, the particle decay experiments
discussed in Chapter 15 specifically require the rank to increase at each time step.
The use of Born maps means total probability is always conserved, even if lin-
earity is absent. In principle, therefore, QDN allows for a discussion of nonlinear
QM, still based on most of the familiar Hilbert space concepts used in QM. As
we have mentioned in the case of nonlinear Born maps, necessarily nonlinear pro-
cesses such as state preparation, state reduction, the switching on and off of appa-
ratus, and so on, which are outside the scope of unitary (Schrödinger) evolution
in standard QM, can all be discussed in QDN in terms of nonlinear Born maps.
Our interest will generally be in experiments based on linear quantum pro-
cesses, so (7.6) will be assumed. For such an experiment running from stage ΣM
to stage ΣN , for N > M , and knowing rn  rn+1 , then the labstate Ψn will
change according to the rule
Ψn → Ψn+1 ≡ Un+1,n Ψn , M  n < N, (7.7)
where Un+1,n is a semi-unitary operator (this terminology will be used from now
on even in the case where rn = rn+1 ).
The CBR at stages Σn and Σn+1 can be used to represent Un+1,n . Specifically,
we find
2rn+1
−1 2
rn
−1
j,i
Un+1,n = j n+1 Un+1,n in . (7.8)
j=0 i=0

This representation can be used to define a retraction operator Un+1,n , given by

2rn+1
−1 2
rn
−1
j,i∗
Un+1,n = in Un+1,n j n+1 , (7.9)
j=0 i=0
j,i∗ j,i
where Un+1,n is the complex conjugate of Un+1,n , if the semi-unitarity condition
holds, that is, if
2rn+1
−1 j,i j,k∗
Un+1,n Un+1,n = δ ik. (7.10)
j=0

A useful way of thinking about the semi-unitarity condition (7.10) and the
rules of semi-unitary operators is in terms of complex vectors as follows. Let
V be a finite-dimensional complex vector space with inner product a† · b. An
orthonormal d-subset of V is any set of elements {ai : i = 1, 2, . . . , d} of V
that are normalized to unity and mutually orthogonal; that is, we have the rule
ai† · aj = δ ij .
Now let V  be another finite-dimensional complex vector space with inner
product a† · b . The question of semi-unitarity reduces to the possibility of
finding an injection from any given orthonormal d-subset of V into at least one
 7.10 The Signal Theorem 97

orthonormal d-subset of V  . The above semi-unitarity theorems tell us that this

cannot be done if d > dim V  .

7.10 The Signal Theorem

The mathematical properties of semi-unitary operators and their relationship
to signal bases have an important bearing on the permitted physics of QDN
dynamics. Consider an experiment at stages Σn and Σn+1 and assume semi-
unitarity. At stage Σn the labstate Ψn is given by a superposition of signal states
from signal basis Bn ≡ {in : 0  i < 2rn }, while the labstate Ψn+1 is given as
a superposition of signal states from signal basis Bn+1 ≡ {in+1 : 0  i < 2rn+1 }.
Because of linearity, the crucial question as far as the dynamics is concerned is
how individual signal states evolve. Semi-unitarity imposes the following con-
straint, which we call the signal theorem:

Theorem 7.12 Two different preferred basis states in and j n in a signal

basis Bn cannot evolve by semi-unitary dynamics into labstates that have only
one preferred signal basis state in common.
Proof Take 0  i < j < 2rn . Suppose in evolves by semi-unitarity dynamics
into a labstate according to the rule

in → Un+1,n in = αkn+1 + φn+1 , (7.11)

while j n evolves according to the rule

j n → Un+1,n j n = βkn+1 + ψ n+1 . (7.12)

Here k is some integer in the semi-open interval [0, 2rn+1 ), α and β are non-
zero complex numbers, and φn+1 and ψ n+1 are elements in Qn+1 sharing no
signal states in common either with each other or with kn+1 in their CBR
basis expansions, which means

kn+1 φn+1 = kn+1 ψ n+1 = φn+1 ψ n+1 = 0. (7.13)

From Corollary 7.8, semi-unitarity preserves inner products and not just
norms, so we must have
0 = in j n = in Un+1,n Un+1,n j n = αkn+1 + φn+1 (βkn+1 + ψ n+1 ) (7.14)
= (α∗ kn+1 + φn+1 )(βkn+1 + ψ n+1 ) = α∗β, (7.15)

using (7.12). This establishes the theorem.

The signal theorem leads to the following important result for conventional
physics. Suppose an observer constructs an apparatus that, if prepared at stage
Σn to be in its signal ground state 0n , would evolve into 0n+1 . If the dynamics
is semi-unitary, then we may write
98 Quantum Register Dynamics

0n → Un+1,n 0n = 0n+1 . (7.16)

This condition models an important physical property expected of most labora-
tory apparatus; we would not expect equipment that was in its signal ground
state to spontaneously generate outcome signals subsequently, unless it was
interfered with by some external agency. Any apparatus that satisfies (7.16) will
be called calibrated between stages Σn and Σn+1 on that account. The analogue
of such a situation in Schwinger’s source theoretic approach to quantum field
theory (Schwinger, 1969) would be one where the external sources were switched
off during some interval of time, so that the vacuum (empty space) remained
unchanged during that time.
Suppose now that, given such a calibrated apparatus, the observer had instead
prepared at stage Σn some labstate Ψn of the CBR form
n −1
2r
Ψn = Ψi in , (7.17)
i=1

that is, a labstate with no signal ground component (note that the summation
(7.17) in runs from 1, not 0). Then for calibrated apparatus with semi-unitary
labstate evolution, the signal theorem tells us that there can be no signal ground
component in the labstate Ψn+1 at time n + 1, and so we may write
2rn+1
−1
Ψn → Un+1,n Ψn = Φj j n+1 , (7.18)
j=1

where
n −1
2r
j ji
Φ = Un+1,n Ψi . (7.19)
i=1

This is an important result, because it tells us that under normal circumstances,

calibrated apparatus does not normally fall into its signal ground state during
an experiment, unless forced to do so by an external agency, such as the observer
switching it off.

Example 7.13 Consider a calibrated rank-one apparatus evolving into

a rank-one apparatus between stages Σn and Σn+1 under semi-unitary
evolution. Then by a suitable choice of phase of basis elements, we may always
write

0n → Un+1,n 0n = 0n+1 ,
1n → Un+1,n 1n = 1n+1 , (7.20)

from which we conclude the dynamics is essentially trivial.

The following example is important, as it models what happens in various

quantum optics modules such as beam splitters and Wollaston prisms.
 7.10 The Signal Theorem 99

Example 7.14 Consider a calibrated rank-two apparatus evolving into

a rank-two apparatus between stages Σn and Σn+1 under semi-unitary
evolution. Then calibration means that we must have Un+1,n 0n = 0n+1 .
Suppose further that it is known that any signality-one labstate always evolves
into a signality-one labstate. Then we may write
 1 0n ≡ 1n → Un+1,n 1n = α1n+1 + β2n+1 ,
A n
 2 0n ≡ 2n → Un+1,n 2n = γ1n+1 + δ2n+1 ,
A (7.21)
n

where the complex coefficients α, β, γ, and δ satisfy the unitarity constraints

|α|2 + |β|2 = |γ|2 + |δ|2 = 1, α∗ γ + β ∗ δ = 0. (7.22)

Now consider the signality-two state A 2

1 A
n n 0n ≡ 3n . Since the apparatus is
calibrated, this state must necessarily evolve into a state that has no signal
ground component. Therefore, we may write

Un+1,n 3n = a1n+1 + b2n+1 + c3n+1 , (7.23)

where |a|2 + |b|2 + |c|2 = 1. Moreover, since the evolution is given as semi-
unitary, then inner products are preserved. Hence we deduce

α∗ a + β ∗ b = γ ∗ a + δ ∗ b = 0. (7.24)

Writing these relations in matrix form, we find

 ∗    
α β∗ a 0
∗ ∗ = . (7.25)
γ δ b 0
The 2 × 2 matrix on the left-hand side of this expression is necessarily
invertible, leading to the conclusion that a = b = 0 and therefore that

3n → Un+1,n 3n = 3n+1 , (7.26)

modulo some arbitrary phase.

The following application of the signal theorem is surprising and somewhat

counterintuitive, because what appears to be a trivial mathematical result rules
out an entire class of physics experiment.

Example 7.15 Suppose an experimentalist prepares a rank-two, signality-

one labstate of the form
 1 + βA
Ψn = (αA  2 )0n , (7.27)
n n

where |α|2 + |β|2 = 1. Suppose further that the dynamics is semi-unitary and
that the apparatus at stage Σn+1 is of rank three. Then by the signal theorem,
semi-unitary evolution such that
100 Quantum Register Dynamics

 1 0n → Un+1,n A
A  1 0n = (aA
 1 + bA
 2 )0n+1 , |a|2 + |b|2 = 1,
n n n+1 n+1
 0n → Un+1,n A
A 2  2 0n = (cA
 2 + dA
 3 )0n+1 , |c|2 + |d|2 = 1, (7.28)
n n n+1 n+1

is not possible.

This result tells us that a double-slit type of experiment where each slit has
only one quantum outcome site in common with the other cannot be physically
constructed. Experiments where two or more quantum outcome sites are in
common are possible, and then inevitable quantum interference terms will occur
in final state amplitudes. For example, in a standard double-slit experiment, every
site on the detector screen is affected by the presence of each of the two slits.
This result reinforces an important rule in QM: we cannot simply add pieces
of apparatus together and expect the result to conform to an addition of classical
expectations. A double-slit experiment where both slits are open is not equivalent
to two single-slit experiments run coincidentally and simultaneously.

7.11 Null Evolution

There is an interesting class of quantum process described by evolution operators
called null evolution operators, associated with the concept of a null test. Recall
that a null test is one that occurs between two or more stages but no information
is extracted. The phenomenon of persistence is associated with null evolution
operators

Persistence
To understand the action of a null evolution operator, consider the idealized
scenario of an initial labstate Ψn in a rank r quantum register Qn evolving into
labstate Ψn+1 in a quantum register Qn+1 of the same rank r. In the following,
we shall use the CBR at all stages.
Suppose we are given that
r
2 −1
Ψn = Ψin in . (7.29)
i=0

Now consider a particular evolution operator Nn+1,n defined by

r
2 −1
Nn+1,n ≡ j n+1 j n . (7.30)
j=0

Then under this evolution operator,

r
2 −1 r
2 −1 r
2 −1 r
2 −1
Ψn → Nn+1,n Ψn = j n+1 j n Ψin in = Ψin j n+1 j n in = Ψin in+1 .

 
j=0 i=0 i,j=0 i=0
δ ij
(7.31)
 7.11 Null Evolution 101

This is a labstate at stage Σn+1 with exactly the same coefficient profile, that is,
set of coefficients {Ψin }, as the initial labstate. It is reasonable in this context to
refer to this phenomenon as an example of persistence and refer to Ψn+1 as a
persistent image of Ψn .
Such an evolution operator will be referred to as a null evolution operator
(NEO). An NEO Nn+1,n makes sense only under particular circumstances: the
rank of the quantum register at stage Σn+1 must be the same as that at stage Σn
and there has to be a one-to-one identification of the elements of the preferred
bases.
This form of evolution demonstrates the two faces of time: on the one hand,
labtime (the time of the observer) goes on as normal, being collated with rel-
ative external context such as the expansion of the Universe. The labtime of a
given observer cannot be reversed relative to other observers, according to all
known current physics, although it can be slowed relative to the labtime of other
observers. On the other hand, some objects such as labstate profiles of persistent
labstates may appear to be indifferent to labtime.

Dynamical Null Tests

In Chapter 11 we discuss Newton’s famous experiment that showed how a light
beam incident on one prism would split into a spectrum of subbeams that could
subsequently be refocused onto a second prism and recombined once more into
a single beam. This is an example of a nontrivial null test. Because the action of
the first prism is nontrivial, and therefore requires nontrivial “undoing” by the
second prism, we refer to the overall process as an example of a dynamical null
test. The generic QDN description of such tests is based on the following.
Consider an labstate Ψn in initial quantum register Qn of rank r = rn , given
by (7.29). Now apply semi-unitary evolution from stage Σn to stage Σn+1 given
by evolution operator
2rn+1
−1 2
rn
−1
i,j
Un+1,n = in+1 Un+1,n jn, (7.32)
i=0 j=0
 
where dn ≡ dim Qn − 1, dn+1 ≡ dim Qn+1 − 1, and the coefficients Un+1,n i,j

satisfy the semi-unitary condition (7.10). According to our theorems on semi-

unitary operators, we require rn+1 , the rank of Qn+1 , to satisfy rn+1  rn .
The labstate Ψn+1 at stage Σn+1 is given by
2rn+1
−1 2
rn
−1
i,j
Ψn+1 = Un+1,n Ψn = in+1 Un+1,n Ψjn . (7.33)
i=0 j=0

Clearly, Ψn+1 will not be a persistent image of Ψn in general. Now consider

stage Σn+2 , and suppose that the rank rn+2 of Qn+2 is given by rn+2 = rn + p,
where rn is the rank of Qn and p  0. Then in terms of the qubits making up
the quantum register, we can write
102 Quantum Register Dynamics

Qn+2 ≡ Q1n+2 Q2n+2 . . . Qrn+2

n
Qrn+2
n +1
. . . Qrn+2
n +p
. (7.34)

 
 n+2
Q

Here Qn+2 is a subspace of Qn+2 of dimension equal to that of Qn . Note that, by

construction, the first 2rn elements of the CBR for Qn+1 involve signal excitations
only of the detectors associated with Q n+2 .
Now define the operator
n −1 2rn+1 −1
2r  j,i∗
Vn+2,n+1 ≡ in+2 Un+1,n j n+1 . (7.35)
i=0 j=0

Note that the upper limit on the summation over the index i is 2rn − 1, not
n+2 . Under
2rn+1 − 1. This operator effectively maps states in Un+1,n Qn into Q
evolution generated by Vn+2,n+1 we find
n −1
2r
Ψn+2 ≡ Vn+2,n+1 Ψn+1 = Ψin in+2 , (7.36)
i=0

which is a persistent image, in subspace Qn+2 , of the original labstate Ψn .

This process demonstrates the principle known as microscopic reversibility:
operator Vn+2,n+1 has effectively “undone” the action of Un+1,n on Ψn . It is
important to understand that as far as QDN is concerned, nothing has remained
unchanged: the observer changes from stage to stage and all labstates change
with those jumps. Microscopic reversibility is an illusion in a sense, but one with
significant empirical content.

7.12 Path Summations

The QDN formulation of dynamics has some of the hallmarks of the Feynman
path integral formulation of quantum mechanics (Feynman and Hibbs, 1965),
with some significant differences: in QDN, time is not continuous, the Hilbert
space changes at each intermediate time step and is assumed finite dimensional,
and there is no need to introduce a Lagrangian or Hamiltonian.
A typical run or repetition of a basic experiment will be assumed to start at
stage Σ0 and finish at a later stage ΣN , for N > 0. Given labstate preparation at
stage Σ0 , there will be semi-unitary evolution through a sequence of apparatus
stages {Σn : 0 < n < N }. At these intermediate stages, the observer does not
look at their detectors, which are therefore to be regarded as virtual. Outcome
detection takes place only at the final stage ΣN .
At the final stage ΣN , the observer looks at all of their detectors and works out
from rule (5.14) which element of the CBR of the preferred basis BN corresponds
to the observed set of signals, for that given run. The objective in practice is to
compare the statistical distribution of observed outcomes with the theoretically
derived conditional probability P r(kN |Ψ0 ) for each of the possible final state
signal basis outcomes kN , 0  k < 2rN .
 7.12 Path Summations 103

Semi-unitary evolution will be assumed to hold between stages Σ and ΣN ,

2r0 −1 i 0
i.e., condition (7.6) is valid. Given an initial labstate Ψ0 ≡ i=0 Ψ0 i0 , the next
labstate is given by Ψ1 = U1,0 Ψ0 , where U1,0 is semi-unitary, and so on, until
finally we may write

ΨN = UN,N −1 UN −1,N −2 . . . U1,0 Ψ0 , N > 0. (7.37)

Inserting a resolution of each evolution operator of the form (7.8), the final state
can be expressed in the form
N −1 2rN −1 −1
2r r0
−1
 2
j ,j j ,j j1 ,j0 j0
N N −1 N −1 N −2
Ψn = ... j N UN,N −1 UN −1,N −2 . . . U1,0 Ψ0 . (7.38)
jN =0 jN −1 =0 j0 =0

We may immediately read off from this expression the coefficient of the signal
basis vector iN . This gives the QDN analogue of the quantum mechanics Feyn-
man amplitude Φifinal |Ψinitial  for the initial state |Ψinitial  to go to a particular
final outcome state |Φifinal . In our case, what we are actually reading off is
A(iN |Ψ0 ), the amplitude for the labstate to propagate from its initial state Ψ0
and then be found in signal basis state i at stage ΣN . We find
2rN
−1 −1 2rN −2 −1

r0
2 −1
i,j j ,j j1 ,j0 j0
A(iN |Ψ0 ) = ... N −1
UN,N N −1 N −2
−1 UN −1,N −2 . . . U1,0 Ψ0 . (7.39)
jN −1 =0 jN −1 =0 j0 =0

The required conditional probabilities are obtained from the Born rule as
discussed above, and so we conclude
 2rN −1 −1 rN −2 −1 2r0 −1 2
  2   i,j 
 jN −1 ,jN −2 j1 ,j0 j0 
P r(iN |Ψ0 ) =  ... N −1
UN,N −1 UN −1,N −2 . . . U1,0 Ψ0  .
 
jN −1 =0 jN −2 =0 j0 =0
(7.40)
By writing the amplitude (7.39) in the form
2rN
−1 −1
i,j
A(iN |Ψ0 ) = UN,N −1 A(j N −1 |Ψ0 ), (7.41)
jN −1 =0

it is straightforward to use induction and the semi-unitary matrix conditions

(7.10) to prove that
N −1
2r
P r(iN |Ψ0 ) = 1, (7.42)
i=0

which means total probability is conserved, as expected.

Feynman derived his path integral for continuous time quantum mechanics by
discretizing time and then taking the limit of the discrete time interval going
to zero. Technical problems occur in the taking of this limit, and because of
these, the path integral in its original formulation (Feynman and Hibbs, 1965) is
generally regarded as ill-defined. However, it is an invaluable heuristic tool that
104 Quantum Register Dynamics

provides the best way to discuss the quantization of certain classical theories for
which other approaches prove inadequate. In QDN, time is discrete and in that
sense we follow Feynman’s lead while avoiding the pitfalls associated with the
continuum limit, which we do not take in QDN.
This completes our introduction to the QDN formalism. An obvious extension
is to include mixed labstates. These are discussed in Chapter 9.
 8
Partial Observations

8.1 Introduction
In this chapter we discuss the quantized detector network (QDN) approach to
partial observations, or the extraction, during an extended-in-time quantum pro-
cess, of only some of the quantum information embedded in a detector amplitude.
In order to deal with partial questions, we need first to discuss how QDN deals
with maximal questions, which means looking at all the detectors at a given stage.

8.2 Observables
Our preoccupation with detectors rather than systems under observation (SUOs)
is nothing new in quantum mechanics (QM). Indeed, the primary significance of
what could be observed was a guiding principle when Heisenberg formulated his
matrix mechanics approach to QM (Heisenberg, 1925). The conventional position
in QM is that the only important operators in the theory are the observables.
These correspond closely to the variables used in classical mechanics (CM) to
describe measurable quantities, such as energy, momentum, electric charge, and
so on.
In standard QM, observables are generally assumed to be self-adjoint operators
(Hermitian operators in the case of finite-dimensional Hilbert spaces) so that they
have real eigenvalues, and these are the QM analogues of CM variables. Apart
from that proviso, operators representing observables have few restrictions.
This generality raises significant questions. A particularly critical question
is whether a given “observable” makes empirical sense. There is no theorem
that proves that every self-adjoint operator corresponds to something that
can be actually observed in the laboratory. For example, in one-dimensional
wave mechanics, the operator pppx
ppp is self-adjoint and its classical analogue,
pppxppp, is a perfectly regular function over phase space. But we know of no
experiment that can measure such a quantity directly, whereas experiments to
measure p or x directly could be devised.
106 Partial Observations

The problem as we see it in this book is that the observable concept in standard
QM puts the cart before the horse: it is first implicitly assumed in QM that states
of SUOs can be created, and only then is the question of what can be done to and
on those states raised. QDN does things the other way around: the apparatus that
creates the states and detects the signals has to come first. Indeed, that is all that
is needed. “State preparation” defines the states contextually and the outcome
detectors define what information can be obtained. In QDN, “observables” are
nothing but the signals in final stage detectors plus the context that informs the
observer as to the meaning of those signals.

Transformations and Symmetries

In the twentieth century, great advances in quantum physics were made, driven by
relativistic transformation theory, leading to the Lorentz covariant formulation
of quantum fields known as relativistic quantum field theory (RQFT). That
theory is generally regarded as the best theory to describe the phenomenology of
elementary particles. However, there are some deep conceptual issues concerning
the interface between the classical world of the observer’s knowledge base and
the nonclassical behavior of labstates.
An important factor here is the relationship between different observers. In
the standard QM approach to observables, the transformation properties of
those observables are regarded as crucial. Indeed, Dirac’s approach to QM was
developed in part as an analogue of CM “transformation theory” (Dirac, 1958;
Goldstein, 1964; Leech, 1965).
In QDN the issue of transformation theory is avoided by the assertion that
a given laboratory needs no transformation. And if it is required to discuss the
relationship between two different laboratories (which we have to remind the
reader consist of atoms and molecules that cannot just pass through each other
without significant interaction, none of which is taken into account in standard
QM), then we simply regard the two laboratories as a single larger laboratory
defined by the context of the situation. For example, a Doppler shift experiment
is conventionally described in terms of a source of light moving relative to a
detector. In QDN terms, all of that can be described as one experiment in a
single laboratory, albeit one that changes from stage to stage. Laboratories in
QDN are not restricted to single inertial frames.
There is an interesting question that can posed here, and its resolution has
everything to do with the dominant principle of contextuality underpinning this
book: Given a distant source of light, such as a remote galaxy on the other side
of the Universe, how can astronomers observing light from that source consider
themselves to be part of a single laboratory that includes that galaxy (which might
have long ago been destroyed )?
The answer is context. A single detected photon, no matter how much “red
shifted” it was,1 could not carry any contextual information. But the actual

1
Once again, we have to resort to realist language to convey our meaning, though it is an
oversimplification to do so.
 8.3 Maximal Questions 107

described scenario gives the game away: how could we even say that there was a
galaxy acting as a source of light without having observed sufficient light from it
to establish that fact? That degree of observation thereby provides the context
that allows us to think of that remote galaxy and our detectors as part of the
same experiment.
In fact, such context is generally built up over relatively long periods of time
and depends on technology. Thousands of years ago, astronomers could only see
via their naked eyes a strange blob of diffuse light where M31, the Andromeda
Galaxy, is in the night sky. Then as telescopes were developed, followed by long-
exposure photography, sufficient observations were made to establish the context
that we know today. Only that relatively recent context allows us to say whether
a signal from Andromeda is red-shifted or blue-shifted light.2

8.3 Maximal Questions

In this section, dependence on the temporal index n is suppressed, as all questions
are asked at some given stage Σn .
Given a rank-r quantum register, an arbitrary pure labstate Ψ is of the general
form

r 
Ψ = Ψ0 0 +  i0 +
Ψi A  iA
Ψij A  j 0 + · · · + Ψ12...r A
 1A
2 . . . A
 r 0,
i=1 1i<jr
(8.1)

where we do not rule out superpositions of elements with different signality.

Labstates are generally normalized to unity, so the coefficients in (8.1) satisfy
the condition
  
r 
ΨΨ = Ψ0  2 + |Ψi |2 + |Ψij |2 + · · · + |Ψ12...r |2 = 1. (8.2)
i=1 1i<jr

Example 8.1 An arbitrary normalized labstate in a rank-two quantum

register is of the form
 1 + Ψ2 A
Ψ = {Ψ0 I[2] + Ψ1 A  2 + Ψ3 A
 1A
 2 }0, (8.3)

with |Ψ0 |2 + |Ψ1 |2 + |Ψ2 |2 + |Ψ3 |2 = 1.

The interpretation of these coefficients is based on the Born rule in standard

QM (Born, 1926): if the apparatus is in labstate (8.3) prior to the observer looking
at both detectors “simultaneously” (which is possible in QDN by definition),
then the probability of each detector being found in its ground state is |Ψ0 |2 ,
the probability of detector 1 being in its signal state and detector 2 being in its

2
In fact, Andromeda is moving toward our galaxy, so light from ordinary sources in
Andromeda should show a small blue shift.
108 Partial Observations

ground state is |Ψ1 |2 , the probability of detector 1 being in its ground state and
detector 2 being in its signal state is |Ψ2 |2 , and the probability of both detectors
being in their signal states is |Ψ3 |2 . Note that |Ψ1 |2 is not the probability that
there is a signal in detector 1. That probability is given by |Ψ1 |2 + |Ψ3 |2 .

Example 8.2 An observer prepares a pure labstate Φ in a rank-four quan-

tum register. Show that the probability P r(11 02 03 14 |Φ) that the observer
would find detectors 1 and 4 in their signal states and detectors 2 and 3 in
their ground states is given by

P r(11 02 03 14 |Φ) = ΦP 4 Φ.
1 P2 P3 P (8.4)

Solution In this case, we need to ask the maximal question 11 02 03 14 of the

labstate Φ, giving the amplitude A(11 02 01 14 |Φ) ≡ 11 02 03 14 Φ. Then the
Born rule gives

P r(11 02 03 14 |Φ) ≡ |A(11 02 03 14 |Φ)|2

= (11 02 03 14 Φ)∗ (11 02 03 14 Φ)
= (Φ11 02 03 14 )(11 02 03 14 Φ)
= Φ(11 02 03 14 11 02 03 14 )Φ
= Φ(11 11 )(02 02 )(03 03 )(14 14 )Φ

 
 
 
 
1
P P2 P3 4
P
 2
1
= ΦP P P P Φ. 3 4 (8.5)

But it is straightforward to show that for a rank-four register,

 1P 2P 3P
P 4 = P 4 ,
1 P2 P3 P (8.6)

where each of the operators on the right-hand side is a register operator. This
then gives the required result (8.4).

For a rank-r register, we shall call a register product of r distinct register

projection operators a maximal question. The above example illustrates how
any maximal question for a rank-r quantum register can be related to the tensor
product of r distinct bit projection operators, one for each detector in the register.
There are three points to note here. First, Eq. (8.6) holds only because the
left-hand side is a register operator, being the tensor product of r individual bit
operators, one for each detector in the register. Second, a maximal question can
be identified with a specific element of the preferred basis. Since there are 2r
elements in the latter, we deduce that there is a total of 2r distinct maximal
questions. The third point is a technical one: the product concepts on each side
of (8.6) are different. The left-hand side is the tensor product of bit operators,
the right-hand side is the product of register operators.
 8.4 Partial Questions 109

8.4 Partial Questions

The above example shows how to ask a specific question of each and every detec-
tor in a quantum register at a given stage. For a rank-r quantum register, any
maximal question involves a product of r distinct register projection operators.
For each detector i, 1  i  r, there are two related register operators, Pi and
i , which form a conjugate pair . Therefore there are exactly 2r distinct maximal
P
questions, as stated above.
In the real world, however, observers could choose to ask partial questions,
which involve looking at only some (or even none) of the detectors. The simplest
example of a partial question involves the normalization condition
ΨΨ = 1, (8.7)
because we can always write ΨΨ = ΨI Ψ, where I is the register identity
[r] [r]

operator. We may identify this operator with the question

What is the probability that every detector
is either in its ground state or signal state?
and call this a rank-zero partial question, because it involves looking at no (i.e.
zero) detectors.
Now suppose we wanted to ask a question involving just one detector, such as
the ath, where 1  a  r. Given (8.7), we insert the register identity operator
I[r] as before and use the property
a = I[r] ,
Pa + P a = 1, 2, . . . , r. (8.8)
Using this property and linearity, we find
a Ψ = 1.
ΨPa Ψ + Ψ P (8.9)
Each of the terms on the left-hand side is nonnegative. By inspection, ΨPa Ψ is
a Ψ
the probability that detector a would be found in its ground state, while Ψ P
is the probability that a would be found in its signal state, regardless of what
was going on in any of the other r − 1 detectors. We will refer to each of these
partial questions as a rank-one partial question.
This process can be extended naturally to higher rank partial questions, until
we reach rank r, which are the maximal questions we discussed in the previous
section.

Example 8.3 Given a labstate Ψ prepared in a rank-637 register, what is

the probability that if the observer looked only at detectors 99, 323, and 438,
they would find 99 and 323 each in its ground state and 438 in its signal state?
Solution
The required probability P r is given by the expectation value
438 Ψ.
P r = ΨP99 P323 P (8.10)
110 Partial Observations

It will be clear from the above that the set of all partial questions involves
expectation values of all possible products of the register projection operators.
This leads to the following theorem.

Theorem 8.4 For a rank-r classical or quantum register Q[r] , the number
of possible partial questions is 3r .
Proof To prove the theorem, we determine the number of partial questions
of each rank and then add up all those numbers.

There is only one rank zero partial question.

The observer could go to each of the r detectors one by one and ask one of
two possible questions of it: the two questions that can be asked at detector i
i . These are rank-one partial questions,
are given by the register projectors Pi , P
so we conclude that there is a total of 2r rank-one partial questions.

Assuming r > 1, the observer could now ask rank-two partial questions,
involving only two distinct detectors
 in the register. Given a rank-r register,
there is a total of r(r − 1) = 2r distinct pairs, and for each pair of choices, 22
alternative questions can be asked. For example, for the choice i < j, we can
i j i j i j i j
  four questions P P , P P , P P , and P P . Therefore, there is a total
ask the
2 r
of 2 2 such partial questions.
We can continue this argument until we reach the maximal questions, which
are rank-r partial questions. There is only one way of choosing r objects from
r objects, and 2r possible maximal questions to be asked of that choice. Hence
we find the grand total TQ of distinct partial question operators to be given by
  
TQ = 1 + 2 1r + 22 2r + · · · + 2r rr = (1 + 2)r = 3r , (8.11)

as asserted.

8.5 Partial Question Eigenvalues

Every partial question has the property that each preferred basis element is an
eigenstate of it, with an eigenvalue of either zero or one. Using the signal basis
representation (SBR), the eigenvalue in each case can be readily read off. For
example, in a rank-five register, the state 01 12 13 04 15 is an eigenstate of the
partial question operators P1 and P1 P4 , with eigenvalues 1 and 0, respectively.
On the other hand, using the computational basis representation (CBR) is not
so convenient here. The given state 01 12 13 04 15 has the CBR 22, so we see

P1 22 = 22, 4 22 = 0,
P1 P (8.12)

but the respective eigenvalues cannot now be directly read off. Since the CBR is
generally useful, we need to quantify the action of the partial questions on the
CBR. We do this as follows. The set {I[r] , P1 , . . . , P 2 . . . P
1 P r } of partial questions
 8.6 Identity Classes 111

Table 8.1 Question eigenvalues for a rank-two register

0 1 02 = 0 11 02 = 1 01 1 2 = 2 1 1 12 = 3

Q1 ≡ I[2] Y 1,0 = 1 Y 1,1 = 1 Y 1,2 = 1 Y 1,3 = 1

Q 2 ≡ P1 Y 2,0 = 1 Y 2,1 = 0 Y 2,2 = 1 Y 2,3 = 0

Q3 ≡ P1 Y 3,0 = 0 Y 3,1 = 1 Y 3,2 = 0 Y 3,3 = 1

Q 4 ≡ P2 Y 4,0 = 1 Y 4,1 = 1 Y 4,2 = 0 Y 4,3 = 0

Q5 ≡ P2 Y 5,0 = 0 Y 5,1 = 0 Y 5,2 = 1 Y 5,3 = 1

Q6 ≡ P1 P2 Y 6,0 =1 Y 6,1 =0 Y 6,2 =0 Y 6,3 =0

Q7 ≡P1 P2 Y 7,0 =0 Y 7,1 =1 Y 7,2 =0 Y 7,3 =0
Q8 2
≡ P1 P Y 8,0 =0 Y 8,1 =0 Y 8,2 =1 Y 8,3 =0
Q9 ≡P1 P
2 Y 9,0 =0 Y 9,1 =0 Y 9,2 =0 Y 9,3 =1

r
contains 3r elements. We define Q1 ≡ I[r] ,Q2 ≡ P1 , . . . , Q3 ≡ P 2 . . . P
1 P r . This
choice of labeling is arbitrary, there being no obvious way to order a complete
set of partial questions. Then for any partial question QP , P = 1, 2, . . . , 3r , and
any CBR element i, i = 0, 1, 2, . . . , 2r − 1, we write
QP i = Y P,i i, (8.13)
where the question eigenvalue Y P,i is either zero or unity.

Example 8.5 For a rank-two register, there are 32 = 9 distinct partial

questions and 22 = 4 distinct CBR elements. Table 8.1 shows the question
eigenvalues.

8.6 Identity Classes

A full partial question set can be divided into groups of operators that sum up,
in that group, to the register identity. Each such group will be called an identity
class. A rank-r register has 2r identity classes.

 register has two identity classes, given by C ≡

1,1
Example 8.6 Arank-one
 [1]  1 1
I and C ≡ P , P .
2,1

Example 8.7 A rank-two register has four identity classes, given by

 
C 1,2 ≡ I[2] ,
 
1 ,
C 2,2 ≡ P1 , P
 
2 ,
C 3,2 ≡ P2 , P
 
C 4,2 ≡ P1 P2 , P 2 , P
1 P2 , P1 P 2 .
1 P (8.14)
112 Partial Observations

In Table 8.1, the four identity classes are separated by horizontal lines. A given
identity class consists of partial questions of the same rank, so we shall refer to
each class by its rank. In the above example, C 1,1 and C 1,2 are rank-zero identity
classes; C 2,1 , C 2,2 , and C 3,2 are rank-one identity classes; and C 4,2 is a rank-two
identity class.
Identity classes are related to probability conservation. We shall find that if
we want to conserve probability in any calculation, we need to restrict partial
questions to the same identity class. It may then be necessary to relabel the
partial questions and their question eigenvalues with an extra label identifying
individual identity classes.

8.7 Needles in Haystacks

The classification of question rank and identity class sheds some light on how the
unimaginable complexity of the real world can be comprehended by intelligent
observers. Suppose an observer wanted to model the Universe by an enormously
large number N of qubits, giving a quantum register Q[N ] of rank 2N . Suppose
now that that observer was investigating their environment by asking a limited
number of partial questions. This is a typical scenario in empirical science:
resources are not infinite and experimentalists can only do so much. By inspection
of Table 8.1, we see an interesting pattern, one that would be repeated in the
general case. If we ask no questions, then that is represented by the rank-zero
identity class. We see from the question eigenvalues for such a question, denoted
Q1 in Table 8.1, that the answer for each possible labstate is one, meaning that
we can extract no new information about the system under observation (SUO)
from such a question. But that question has cost us nothing.
Moving on, we may now decide to ask rank-one questions, represented by Q2
and Q3 in Table 8.1. Now we start to get some real information about the state
of the SUO, but it is also starting to become expensive.
This process could continue, with greater rank questions being posed, with
more information coming out but at ever increasing cost.
There is an interesting trade-off. Looking at Table 8.1, we note that for a given
labstate, the maximal rank identity class partial questions all have answer zero
except one of them. This enormously simplifies the problem of finding a signal,
in that should we find an answer of one halfway through this process, we can
immediately stop, because we can be sure all the remaining answers are zero.
We can now appreciate what it means to do experiments and why they are
done. Experiments are careful arrangements of partial questions of the same
identity class, designed to eliminate as many zero-value answers and home in on
unity answers (those for which Y θ,i = 1), guided by theory. Such searches can
require planning to choose the right questions and the expenditure of enormous
resources as in the case of the search for the Higgs particle at the Large Hadron
Collider. That task was far more of a technical problem than searching for a
 8.7 Needles in Haystacks 113

needle in a haystack.3 At the end of the day, we note that the result of the search
was a single number, that is, one, which means yes, the Higgs particle exists
(according to our contextual interpretation of the data).
This line of thinking also encourages experimentalists not to give up but
to continue searches that appear to be unsuccessful. For instance, there is at
this time (2017) no direct empirical evidence for supersymmetric partners of
various particles such as electrons and photons. Because the number of zero-
value question eigenvalues is potentially vast, a lack of confirmation so far does
not mean that a yes answer does not exist.

3
Which can be done quickly with the right apparatus, such as a metal detector.
 9
Mixed States and POVMs

9.1 Introduction
As the decades of the twentieth century rolled on, quantum mechanics (QM)
became more and more sophisticated and mathematical. Understanding what
this theory means intuitively was confounded not only by nonclassical concepts
such as wave–particle duality and quantum interference, but also by issues to do
with observation. The Newtonian classical mechanics (CM) paradigm of reality,
wherein reductionist laws of physics describe observer-independent dynamics of
systems under observation (SUOs) with observer-independent properties, was
found to be inadequate. Quantum theorists were confronted with the measure-
ment problem, which attempts to understand, explain, and rationalize the laws
of QM that underpin the processes of observation that go on in the laboratory.
They are not the same as those of CM in several puzzling respects.
Historically, the first sign of the measurement problem was Planck’s quantiza-
tion of energy (Planck, 1900b) and the second was Bohr’s veto on radiation damp-
ing in hydrogen (Bohr, 1913). These occurred in the first quarter of the twentieth
century, a period in physics often referred to as Old Quantum Mechanics. Another
indicator that intuition was inadequate was Born’s interpretation of the squared
modulus of the Schrödinger wave function as a probability density (Born, 1926).1
That interpretation has everything to do with observers and observation, because
probability without an observer is a vacuous concept.
Eventually, the projection-valued measure (PVM) formalism emerged, cham-
pioned by von Neumann in an influential book on the mathematical formulation
of QM (von Neumann, 1955). Subsequently, pioneers such as Ludwig (Ludwig,
1983a,b) and Kraus (Kraus, 1974, 1983) refined the theory into the general pos-
itive operator-valued measure (POVM) formalism that we shall discuss and use.

1
Born appears to have at first taken the magnitude of the wave function as the probability
density, but corrected himself in time.
 9.2 Sets and Measures 115

The quantized detector network (QDN) approach to quantum experiments is

most naturally expressed in the PVM formalism, as QDN focuses on the individ-
ual detectors in the laboratory. However, the POVM formalism is more general
than the PVM formalism, giving a description of multiple detector processes
similar to QDN. This raises the question of how the two approaches, QDN and
POVM, are related. The aim of this chapter is to explore this relationship.
Before we review the PVM and POVM formalisms, we review some essential
mathematical concepts.

9.2 Sets and Measures

Most of the mathematics used in this book involves spaces, which are sets with
additional structure such as inner products (Howson, 1972).

Definition 9.1 A sigma-algebra on set A is a collection ΣA of subsets of A

such that
1. ΣA includes the empty subset ∅.
2. ΣA is closed under complement: if E is an element of ΣA , then so is its
complement in A, denoted E c . Since by property 1, the empty set ∅ belongs
to ΣA and its complement is A, then property 2 means that A itself is a
member of ΣA .
3. ΣA is closed under countable unions, which means that if we pick any
countable number E 1 , E 2 , . . . , E n of elements in ΣA , then their union
∪ni=1 E i is also in ΣA .
4. ΣA is closed under countable intersections, which means that if we pick
any finite number E 1 , E 2 , . . . , E n of elements of ΣA , then their intersection
∩ni=1 E i is also in ΣA . Note that this intersection could be empty, but that
possibility is covered by property 1.

Definition 9.2 The extended reals R∗ is the set of real numbers R and two
extra elements, +∞ and −∞. These latter two elements are referred to as plus
infinity and minus infinity, respectively, and are interpreted accordingly.

Definition 9.3 Given a sigma-algebra ΣA over set A, a measure on A assigns

an extended real number μ(E) to each element E of ΣA such that
1. Nonnegativity: for any element E in ΣA , μ(E)  0.
2. Measure of empty set: μ(∅) = 0.
3. Countable additivity: for any countable collection {E 1 , E 2 , . . .} of pairwise
disjoint elements of ΣA , meaning that E i ∩ E j = ∅ for i = j, then
∞  ∞
 
i
μ E = μ(E i ). (9.1)
i=1 i=1
116 Mixed States and POVMs

In applications to quantum physics, the set A referred to in the above defini-

tions will consist of all possible outcomes of an experiment; the sigma-algebra
will consist of all possible ways of grouping those outcomes; and the measure μ
will be the assignment of probabilities to the elements of ΣA .

9.3 Hilbert Spaces

Hilbert spaces are complex vector spaces with a suitable inner product concept. In
this chapter, we shall use the well-known Dirac bracket notation, following Paris
(2012). In most of this book we deal with finite-dimensional Hilbert spaces, as
the guiding philosophy of QDN is to model what goes on in the laboratory. No
infinities are ever actually encountered in the laboratory; at worst, a readout on
a counter goes off-scale.

Definition 9.4 An orthonormal basis (ONB) for a d-dimensional Hilbert

space H is a set of elements {|ψ n  : n = 1, 2, . . . , d} of H with the following
properties:
1. Orthonormality: ψ n |ψ m  = δ nm , 1  n, m  d, where δ nm is the
Kronecker delta.
2. Completeness/resolution of the identity:

d
|ψ n ψ n | = I H , (9.2)
n=1

where I H is the identity operator on H.

9.4 Operators and Observables

An operator is any rule that assigns an element of one Hilbert space to some
element either in the same Hilbert space or in another Hilbert space. There are
various kinds of operators that are important to us here. A linear operator O
from Hilbert space H1 to Hilbert space H2 is one such that if |Ψ and |Φ are
arbitrary elements of H1 and α and β are arbitrary complex numbers, then
O(α|Ψ + β|Φ) = αO|Ψ + βO|Φ, (9.3)
where we note that addition on the left-hand side of (9.3) is in H1 , while addition
on the right-hand side is in H2 .
The linear space of linear operators from H to H is denoted L(H) and is itself
a Hilbert space (Paris, 2012).
A positive operator O over Hilbert space H is one such that for any element
|Ψ of H, we have
Ψ|O|Ψ  0. (9.4)
A positive operator is self-adjoint.
 9.6 Projection-Valued Measure 117

An observable X is a self-adjoint operator that admits a discrete spectral

decomposition; i.e., we can write

d
X= xn P n , (9.5)
n=1

where the xn are real and the eigenvalues of X, and the projectors P n are given
by P n ≡ |xn xn |. Here the normalized eigenvectors |xn  satisfy the eigenvalue
equation
X|xn  = xn |xn  (9.6)
and form an ONB for H.2 Orthonormality then leads to the product rule

P n P m = δ nm P n (9.7)
and the completeness sum

d
P n = I H. (9.8)
n=1

9.5 Trace
If {|ψ n  : n = 1, 2, . . . , d} is an ONB for H, then the trace Tr{O} of an operator
O is defined by

d
Tr{O} ≡ ψ n |O|ψ n . (9.9)
n=1

The trace of an operator is independent of choice of ONB for H.

Given an arbitrary state |Ψ in H, then we can use any ONB to show that
Ψ|O|Ψ = Tr{O|ΨΨ|}. (9.10)
This result is crucial to the density operator and POVM formalisms widely
applied in QM.

9.6 Projection-Valued Measure

We are now in a position to discuss PVMs in standard QM.
Given an observable X, the probability Pr(xn |Ψ) of final outcome xn given
initial normalized state |Ψ is according to the Born rule (Born, 1926) given by
Pr(xn |Ψ) = |xn |U |Ψ|2 = Ψ|U † |xn xn |U |Ψ
= Ψ|U † P n U |Ψ = Tr{P n U U † }, (9.11)

2
This can always be arranged, even if some of the eigenvalues are degenerate.
118 Mixed States and POVMs

where U is the unitary evolution operator taking the initial state |Ψ to its final
state U |Ψ at the time of measurement and  is the initial density operator ,
defined by
 ≡ |ΨΨ| (9.12)
in this instance.
The expectation value XΨ of the observable X conditional on Ψ is given by

d d
XΨ ≡ xPr(xn |Ψ) = xn Tr{P n U U † }
n=1 d  n=1 

= Tr xn P n U U † = Tr{XU U † }. (9.13)
n=1

9.7 Mixed States

Suppose an observer carries out an experiment consisting of a large number of
runs but is unsure, at the start of each run, of the initial state. Suppose further
that the observer’s ignorance can be quantified into the statement that for each
run, the initial state is taken from a discrete probability space Ω, that is, a finite
distribution of possible states, each labeled by superscript κ, running from 1 to
some finite integer K, such that the probability of preparing state |Ψκ  is ω κ :
 

K
Ω ≡ ω κ , |Ψκ  : ω κ = 1, Ψκ |Ψκ  = 1 . (9.14)
κ=1

Here the possible initial states are normalized but not necessarily mutually
orthogonal. The number K of possibilities is arbitrary and can exceed the dimen-
sion d of HS . The probabilities ω κ are epistemic in character, that is, classical
probabilities. Such a random initial state is referred to as a mixed state when K >
1. When K = 1, the observer’s ignorance about the initial state is zero and the
corresponding unique element of the Hilbert space is referred to as a pure state.
The expectation value X of the observable X conditional on a mixed state
 is now

K
X = ω κ ψ κ |U † XU |ψ κ 
κ=1

K
= Tr{XU ω κ |ψ κ ψ κ | U † } = Tr{XU U † }, (9.15)
κ=1

 


where

K
≡ ω κ |ψ κ ψ κ | (9.16)
κ=1

is the appropriate generalization of the density matrix operator (9.12) to the

mixed state situation.
 9.8 Partial Trace 119

It is easy to show that a density operator is a positive operator and has unit
trace. This means that the eigenvalues of a density operator are nonnegative and
sum to unity.

9.8 Partial Trace

The above formalism is standard quantum theory as applied to a single Hilbert
space, typically the space of SUO states. We now extend the discussion to the
tensor product of two Hilbert spaces, HS and HA , where in anticipation of
our later needs, S will stand for SUO and A will stand for apparatus or, more
technically, ancilla.3 HS will have dimension d and HA will have dimension D.
Given two finite-dimensional Hilbert spaces HS , HA , not necessarily of the
same dimension, consider a state |ΨSA ) in the tensor product HS ⊗ HA . Note
that we shall henceforth use round brackets to denote states in such an SUO-
apparatus tensor product space, rather than angular brackets, and call them total
states. Then the associated density operator is defined by
SA ≡ |ΨSA )(ΨSA |. (9.17)
Now suppose X ∈ L(H ) is an observable over H . Then we can write
S S


d
X= xn P n . (9.18)
n=1

Now construct the operator Pn ≡ P n ⊗ I A in L(HS ⊗ HA ), where I A is the

identity operator over HA . Then
Pr(xn |ΨSA ) = (ΨSA |U † P n U |ΨSA ) = TrAB {Pn U SA U † }. (9.19)
Here the subscript AB reminds us we are taking the full trace, that is, in the
tensor product space HS ⊗ HA .
The concept of partial trace is related to the concept of partial question that
we discussed in the previous chapter. Partial traces are defined by constructing
ONBs for the component Hilbert spaces and summing only over some of them.
Suppose {|sm  : m = 1, 2, . . . , d} is an ONB for HS and {|an  : n = 1, 2, . . . , D}
is an ONB basis for HA . Then {|sm , an ) ≡ |sm  ⊗ |an  : m = 1, 2, . . . , d, n =
1, 2, . . . , D} is an ONB basis for HS ⊗ HA .
Suppose V is an operator over HS and W is an operator over HA . Then the
full trace TrSA {V ⊗ W } of the tensor product operator V ⊗ W is given by

d 
D
TrSA {V ⊗ W } ≡ (sm , an |V ⊗ W |sm , an )

m=1 n=1  D 

d 
= s |V |s 
m m
a |W |a  .
n n
(9.20)
m=1 n=1

3
In standard QM, ancillas are often treated as auxiliary, almost incidental aspects. In QDN,
they are essential and as important as SUOs.
120 Mixed States and POVMs

There are two partial traces available. TrS {V ⊗ W } is a partial trace over the
SUO degrees of freedom, giving
 d 
d 
TrS {V ⊗ W } = sm |V ⊗ W |sm  = sm |V |sm  W, (9.21)
m=1 m=1

and is an operator over H . Similarly, TrA {V ⊗ W } is a partial trace over the

A

apparatus degrees of freedom, giving

 D 
D 
TrA {V ⊗ W } = a |V ⊗ W |a  = V
n n
a |W |a  ,
n n
(9.22)
n=1 n=1

and is an operator over H . S

Given a density operator SA over HS ⊗ HA , then we define the partial traces
S ≡ TrA {SA }, A ≡ TrS {SA }. (9.23)
Then S is a density operator over HS and A is a density operator over HA ,
with
TrS {S } = TrA {A } = 1. (9.24)

Circularity
For a single Hilbert space H, the trace of operators A, B, . . . , Z satisfies the
circularity property
TrH {ABC . . . Z} = TrH {BC . . . ZA}. (9.25)
This property holds also for tensor products of Hilbert spaces. If R1 ,R2 , . . . , RN
are operators over HS ⊗ HA , then
TrSA {R1 R2 . . . RN } = TrSA {R2 R3 . . . RN R1 }. (9.26)
Circularity does not hold for partial traces.

9.9 Purification
Suppose HS is a Hilbert space of dimension d, representing states of some SUO.
Consider a density operator S over HS . From previous sections, we can find
a set {|ψ m  : m = 1 . . . d} of normalized eigenvectors of S with nonnegative
eigenvalues, that is,
S |ψ m  = λm |ψ m , m = 1, 2, . . . , d, λm  0. (9.27)
Then we can write

d
S
 = λm |ψ m ψ m |. (9.28)
m=1

We can assume orthonormality, that is, ψ m |ψ n  = δ mn .

 9.10 Purity and Entropy 121

Now construct another Hilbert space HA with dim HA = D  d, with an ONB

{θ : n = 1, 2, . . . , D}. Next, define a pure state |ψ SA ) in HS ⊗HA of the form
n


d √
|ψ SA ) ≡ λn |ψ n  ⊗ |θn . (9.29)
n=1

The density operator SA associated with this pure state is then given by

d √
SA ≡ |ψ SA )(ψ SA | = λm λn |ψ m ψ n | ⊗ |θm θn |. (9.30)
m,n=1

Now partially trace SA over HA :


D 
D 
d √
TrA {SA } ≡ θc |SA |θc  = θc | λm λn |ψ m ψ n | ⊗ |θn θn |θc 
c=1 c=1 m,n=1

D 
d √
= λm λn |ψ m ψ n |θc |θm θn |θc 
c=1 m,n=1

 
 
δ cm δ nc

d
= λm |ψ m ψ m |, (9.31)
m=1

that is,
TrA {SA } = S . (9.32)
In words, we can represent the density operator for a mixed state in one Hilbert
space by the density operator for a pure state in a larger Hilbert space. This fun-
damental result opens the door to Naimark’s theorem (below). The state |ψ SA )
is called a purification of S . There are infinitely many possible purifications of
a given density operator.

9.10 Purity and Entropy

Definition 9.5 Given a density operator  with spectrum of eigenvalues

{λk : k = 1, 2, . . . , d}, define the purity μ[] by

d
μ[] ≡ (λk )2 . (9.33)
k=1

It is straightforward to prove that for any density operator, d−1  μ[]  1.

Mixed states ignore correlation information encoded between an SUO and its
environment. Given a density operator , another measure of information loss
associated with  is the von Neumann entropy, defined by

S[] ≡ −Tr{ ln } = − λk ln λk . (9.34)
k
122 Mixed States and POVMs

Then it is straightforward to show that 0  S[]  ln d. Von Neumann entropy

is a monotonically decreasing function of purity and vice-versa (Paris, 2012). A
pure state has purity 1 and von Neumann entropy zero, while a maximally mixed
state has entropy ln d and purity 1/d.

9.11 POVMs
The Born rule for mixed states can be rewritten in a form that can be gener-
alized, leading to the POVM formalism, a more general approach to quantum
measurement than the PVM formalism.
In standard QM, given a density operator S on a d-dimensional SUO Hilbert
d
space HS and observable X ≡ m=1 xm P m , then the probability Pr(xm |S ) of
outcome xn is given by

Pr(xm |S ) ≡ Tr{P m U S U † P n }, (9.35)

where we use P n P n = P n .
Now suppose we can find a set {M n : n = 1, 2, . . . , N } of operators called
Kraus (or detection) operators such that

N
M n† M n = I S . (9.36)
n=1

Then the generalization of (9.35) is to assert that

Pr(y n |S ) ≡ Tr{M n S M n† } (9.37)

is the probability associated with detection outcome associated with M n , where

now the number N of possible outcomes is not necessarily equal to d ≡ dim HS .
We define the POVM elements {E n : n = 1, 2, . . . , N } of the probability (or
positive) operator-valued measure (POVM) by

E n ≡ M n† M n . (9.38)

Then (9.36) gives


N
En = I H .
S
(9.39)
n=1

The POVM operators E n are positive.

The detection operators M n are defined up to unitary transformations; that
is, given a detection operator M n , then for unitary operator V , M n ≡ V n M n
is a valid detection operator giving the same POVM element E n ≡ M n† M n as
M n , i.e.,

E n = E n . (9.40)

This scheme is called a generalized measurement.

 9.12 Naimark’s Theorem 123

9.12 Naimark’s Theorem

This theorem is critical in the formal justification and rationalization of QDN.
The theorem goes under the name of Naimark, but Naimark was not concerned
with application to quantum theory per se.
The theorem has two parts (Paris, 2012). We write it out here in its QDN
form, where S refers to the SUO and A refers to apparatus or ancilla.

Theorem 9.6 Part 1

Suppose we are given a POVM set {E n : n = 1, 2, . . . , N } on SUO Hilbert
space HS with finite dimension d. Then we know that the E n are positive
operators and

N
En = I S . (9.41)
n=1

Then there exists a Hilbert space HA of dimension at least N , and a pure state
|ω A  in HA such that
1. The density operator A ≡ |ω A ω A | ∈ L(HA ), the Hilbert space of linear
operator over HA .
2. There is some unitary evolution operator (in QDN a semi-unitary operator)
U ∈ L(HS ⊗ HA ) such that
U † U = U U † = I SA . (9.42)
3. There is a set {P n } of projectors over HA ,
such that
E n = T rA {I S ⊗ P n U I S ⊗ A U † }. (9.43)
Note that the U operator can encode any dynamical evolution in the
combined system.
Part 2
In this part, we derive an expression for the detection operators {M n }.
Suppose an initial mixed state is prepared, such that in HS ⊗ HA it is
described by density operator
0 SA ≡ 0 S ⊗ |ω0 A ω0 A | (9.44)
and then allowed to evolve under unitary evolution U , giving the final density
operator
SA = U 0 AB U † . (9.45)
Then a projective measurement to test for outcome |xn  of observable X is
made via the apparatus. The probability P r(xn |0 SA ) is given by the equivalent
of the Born rule, in this case
124 Mixed States and POVMs

Pr(xn |0 SA ) = TrSA {SA I S ⊗ P n }, (9.46)

where P n ≡ |xn xn |. Then

Pr(xn |0 SA ) = TrSA {U 0 SA U † I S ⊗ P n }

= TrSA {U 0 S ⊗ |ω0 A ω0 A |U † I S ⊗ |xn xn |}
= TrSA {0 S ⊗ |ω0 A ω0 A |U † I S ⊗ |xn xn |U }
= TrS {0 S ω0 A |U † I s ⊗ |xn xn |U |ω0 A }
= TrS {0 S ω0 A |U † |xn xn |U |ω0 A }
= TrS {xn |U |ω0 A 0 S ω0 A |U † |xn } (9.47)

 
 
Mn M n†

i.e.

Pr(xn |0 SA ) = TrS {M n 0 S M n† } = TrS {E n 0 S }, (9.48)

where E n ≡ M n† M n .

9.13 QDN and POVM Theory

We turn our attention now to the relationship between QDN and POVMs. This
is important because a generalized QDN-POVM formalism is used extensively in
our computer algebra (CA) implementation of QDN, program MAIN, discussed
in Chapter 12 and used in all our specific calculations.
The first step is to recognize that the quantum registers we are concerned
with model only the labstates, that is, the apparatus states, and say nothing
per se about the imagined SUO states. As we stressed previously, QDN was not
designed to give specific information about what happens in the information void.
Therefore, the quantum physics of SUOs has to be appended “by hand”. This
is achieved by introducing a stage-dependent Hilbert space HnS at each stage Σn
to contain the SUO “internal” states, and a separate quantum register Qn to
contain the signal state of the apparatus. Elements of HnS will be called system
states, elements of Qn are called labstates, and elements of Hn ≡ HnS ⊗ Qn will
be called total states.

Notation
In the following, Dirac’s bracket notation |ψn  is used for system states; our bold
notation in denotes computational basis representation (CBR) of labstates; and
modified Dirac brackets |Ψn ) = |ψn , in ) ≡ |ψn  ⊗ in describe total states. We
make an exception to our rule suppressing tensor product symbols in the case of
total states, in order to separate out system states and labstates.
By definition, system states are unobservable directly and there is no nat-
ural preferred basis for HnS .4 Therefore, we are free to choose any convenient

4
The principle of “general covariance,” or independence of frame of reference, is meaningful
as far as system states are involved, but vacuous as far as labstates are concerned.
 9.13 QDN and POVM Theory 125

orthonormal basis for HS . We shall denote elements of our choice by |αn ,

that is, with lowercase Greek labels, with the inner product rule αn |βn  ≡
δ αβ , and assume that HnS has finite dimension dn . For instance, dn = 2 if
Hn describes vertical and horizontal electromagnetic polarization eigenstates
(ignoring momenta and other attributes).
Summations occur frequently in the following, so we make the following sim-
plification:


dn 2 −1
rn

[n]
→ , (9.49)
α=1 i=1 αi

where rn is the rank of Qn .

Initial Total State

A pure initial total state |Ψ0 ) and its dual (Ψ0 | will take the general form


[0]

[0]
|Ψ0 ) ≡ ψ0αi |α0  ⊗ i0 , (Ψ0 | = ψ0αi∗ α0 | ⊗ i0 , (9.50)
i αi

where ψ0αi∗ is the complex conjugate of ψ0αi . If the initial total state is normalized
to unity, then we have


[0]
(Ψ0 |Ψ0 ) = |ψ0αi |2 = 1. (9.51)
αi

Final Total State

Assuming semi-unitary evolution, the final total state |ΨN ) is given by

|ΨN ) ≡ UN,0 |Ψ0 ), N  0, (9.52)

where ΣN is the final stage and the contextual evolution operator UN,0 is given by

 
[N ] [0]
αi,βj
UN,0 ≡ UN,0 |αN β0 | ⊗ iN j 0 . (9.53)
αi βj

The retraction operator U N,0 is given by

 
[N ] [0]
αi,βj∗
U N,0 = UN,0 |β0 αN | ⊗ j 0 iN . (9.54)
αi βj

The semi-unitary condition U N,0 UN,0 = I0 , where I0 is the identity operator for
the initial total space H0 , requires the conditions


[N ]
αi,βj∗ αi,γk
UN,0 UN,0 = δ βγ δ jk . (9.55)
αi
126 Mixed States and POVMs

QDN POVM Operators

The QDN Kraus operators are defined as
 
[N ] [0]
αi,βj
i
MN,0 ≡ iN UN,0 = UN,0 |αN β0 | ⊗ j 0 ,
α βj

i  
[N ] [0]
αi,βj∗
M N,0 ≡ U N,0 iN = UN,0 |β0 αN | ⊗ j 0 , (9.56)
α βj

giving the POVM operators

i
i
EN,0 ≡ M N,0 M⎧i
N,0 ⎫
[0] [0] ⎨ [N ]
   αi,βj∗ αi,δk ⎬
= U UN,0 |β0 δ0 | ⊗ j 0 k0 , i = 0, 1, 2, . . . , 2rN − 1.
⎩ α N,0 ⎭
βj δk
(9.57)
Using the semi-unitary conditions (9.36), we readily find

[0]
i
EN,0 = I0 , (9.58)
i

which is the QDN analogue of the standard POVM condition (9.39).

Interpretation
We may readily understand the QDN POVM formalism if we consider the pos-
sible questions that the observer could ask. Those questions can be only asked
about the signal status of the apparatus. If the final stage normalized total state
is |ΨN ) and the final stage apparatus has rank rN , then there is a grand total
of 2rN maximal questions, each of them equivalent to a projector of the form
INS
⊗ iN iN , for i = 0, 1, 2, . . . , 2rN − 1. The probability P (iN |Ψ0 ) that the final
labstate is iN is given by
P (iN |Ψ0 ) ≡ (ΨN |iN iN |ΨN ) = (Ψ0 |U N,0 iN iN UN,0 |Ψ0 )
i
= (Ψ0 |U N,0 iN iN UN,0 |Ψ0 ) = (Ψ0 |M N,0 MN,0
i
|Ψ0 ) (9.59)

 
 
 
i Mi i
M N,0 EN,0
 i
 N,0
= Tr EN,0 0 , (9.60)
where 0 ≡ |Ψ0 )(Ψ0 | is the initial stage density operator.
We saw in Chapter 8 that we can also ask partial questions, involving only a
subset of all the final state detectors. We can answer these questions as well,
because any partial question is equivalent to some combination of maximal
questions.

Example 9.7 Consider a rank-three quantum register. There are 23 = 8

maximal questions, corresponding to the eight projection operators associated
with the CBR. Specifically, we have Q0 ≡ P1 P2 P3 = 00, Q1 ≡ P 1 P2 P3 = 11,
 1 2 3
. . ., Q ≡ P P P = 77.
7
 9.13 QDN and POVM Theory 127

Suppose we wanted to ask the partial question P 2 , that is, look only at
detector 2 and ask whether it is in its signal state. Then we use the rule that
i = I, the register identity, to write
for i = 1, 2, ...,d, Pi + P
2 = IP
P 2 I = (P1 + P1 )P
2 (P3 + P3 )
2 P3 + P
= P1 P 2 P3 + P1 P
1 P 2 P
3 + P 2 P
1 P 3
= Q2 + Q3 + Q6 + Q7 . (9.61)

We conclude from this that knowledge of all the maximal questions and their
answers allows us to answer all partial questions.

Mixed Initial States

The above has assumed that the initial total state is a pure one. We can readily
extend the formalism to the case of mixed initial states. All we need do is replace
the pure density operator 0 ≡ |Ψ0 )(Ψ0 | with the appropriate mixed density
operator. This will have the generic form

0 ≡ ω κ |Ψκ0 )(Ψκ0 |, (9.62)
κ

where the ω are probabilities summing to unity and |Ψκ0 ) is the normalized total
κ

state occurring with probability ω κ in the mixture.

There is an interesting possibility here: not only could the randomness be
associated with the system states, but there could be uncertainty about the
apparatus. In other words, we should be prepared for the possibility that the
apparatus at any given stage is not determined beforehand but is created by
random processes. This raises deeper questions to do with the general theory of
observation that will certainly need to be addressed in the future. We discuss
some aspects of this topic in Chapter 21, on self-intervening networks.
There are three important aspects to our modeling. First, in common with
standard quantum modeling of pointer states, our initial apparatus register (at
stage Σ0 ) will be a rank-one quantum register Q1 referred to as a preparation
switch. This models the logic of state preparation: if the observer knows that an
SUO state |Ψ has been prepared, then that state is tensored with element 10 of
Q0 , whereas if such a state has not been prepared, then that state is tensored
with element 00 of Q0 . We see from this that QDN attaches significance to what
has not been done as much as to what has been done. This is reminiscent of
Renniger’s thought experiment, where an absence of observation has measurable
consequences (Renniger, 1953).
The second aspect concerns the interpretation of what is going on. In QDN,
only signal states of the apparatus are physically meaningful. The SUO and
therefore the mathematical representation of its states is a convenient fiction
encoding contextuality. Therefore, we do not need to treat HS on the same footing
as Qn . It may be convenient to employ the Schrödinger picture for evolution of
128 Mixed States and POVMs

SUO states and treat HS as stage independent.5 That is certainly not the case
for the quantum registers Qn .
The third aspect concerns the notion of purification that we discussed above for
standard QM. There it was pointed out that the dimension of the ancilla Hilbert
space HA had to be at least as great as the number of independent states in
the prepared mixture that was being purified, and that there was always an
infinite number of ancillas that could be employed. In the case of QDN, neither
of these comments applies. The analogue of the ancilla concept is the quantum
register modeling the apparatus, and the rank of that register is determined
by the number of detectors that the observer has constructed. That number is
independent of the dimensionality of HS . Moreover, the concept of preferred
basis does not apply to the standard ancilla concept, whereas it is relevant in
QDN. The moral here is that standard QM treats states of SUO as the primary
objects of interest, whereas QDN relegates them to auxiliary devices and elevates
the apparatus to be the only thing that matters. This is exactly in line with
Wheeler’s participatory principle, mentioned previously.
We start our analysis therefore with a statement as to what is known at initial
stage Σ0 . We will assume that a mixed SUO state has been prepared at stage
Σ0 ,6 such that the initial density matrix 0 SA is an element of L(HS ⊗ Q0 )
given by
1 ,
0 SA ≡ S ⊗ 10 10 = S ⊗ P (9.63)
0

where S is a density operator, an element of L(HS ).

We now imagine that the combined SUO-apparatus state evolves from stage
Σ0 to some final stage, ΣN , where N > 0. Consider quantum evolution from
Hilbert space HS ⊗ Qn at stage Σn to Hilbert space HS ⊗ Qn+1 at stage
Σn+1 , where HS is the SUO Hilbert space of dimension d and Qn , Qn+1 are
the quantum registers representing labstates of the apparatus. An ONB at Σn
is given by {|A, in ) ≡ |A ⊗ in : i = 0, 1, . . . , dn } with the orthonormalization
(A, in |B, j n ) = δ AB δ ij , while at Σn+1 we have ONB {|A, in+1 ) ≡ |A ⊗ in+1 :
i = 0, 1, . . . , dn+1 } with the orthonormalization (A, in+1 |B, j n+1 ) = δ AB δ ij .
Here dn ≡ 2rn − 1.
Pure quantum evolution is given by

d d
n+1

Un+1,n |A, in ) = U BA,ji

n+1,n |B, j n+1 ) : i = 0, 2, . . . , dn . (9.64)
B=1 j=0

5
In our computer algebra program MAIN, discussed in Chapter 12, we take the system space
to change from stage to stage.
6
States of SUOs are treated in QDN as convenient repositories of context. By their actions
in the state preparation process, an observer will in general be entitled to model some of
that context in terms of a density operator representing a mixed state of an SUO.
 9.13 QDN and POVM Theory 129

Using completeness, this gives the dyadic representation of Un+1,n


d 
dn+1 dn
Un+1,n = |B, j n+1 )U BA,ji
n+1,n (A, in |. (9.65)
A,B=1 j=0 i=0

The retraction operator U n+1,n is defined to be


d 
dn+1 dn
U n+1,n = |A, in )U BA,ji∗
n+1,n (B, j n+1 |. (9.66)
A,B=1 j=0 i=0

By definition, we have

U n+1,n Un+1,n = I S ⊗ I1 A , (9.67)

where I S is the identity over HS and I1 A is the identity over Q1 . From this, we
arrive at the semi-unitary conditions

d d
n+1

U CA,ja∗ CB,jb
n+1,n U n+1,n = δ
AB ab
δ . (9.68)
C=1 j=0

Now consider a pure total state


d 
d
|Ψ0 ) ≡ 0 |A
ΨA ⊗ 10 , |ΨA
0 | = 1,
2
(9.69)
A=1 A=1

evolving from state Σ0 to stage Σ1 , giving total state |Ψ1 ) ≡ U1,0 |Ψ0 ).
Continuing this process, we find the state at stage ΣN is given by |ΨN ) =
UN,0 |Ψ0 ), where UN,0 ≡ UN,N −1 UN −1,N −2 . . . U1,0 .
Suppose now the observer asks a partial question QθN of the state of the
apparatus at stage ΣN . The expectation value E[QθN |Ψ0 ] of the answer is given by

d
E[QθN |Ψ0 ] ≡ (ΨN |I S ⊗ QθN |ΨN ) = 0 Ψ0 A|E N |B,
ΨA∗ B θ
(9.70)
A,B=1

where

E θN ≡ 10 U N,0 I S ⊗ QθN UN,0 10 . (9.71)

If now the initial state is a mixed state, such that the initial SUO state |Ψα
0  is
given with probability ω α , then the expectation value is given by

E[QθN |0 ] = TrS {A

0 E N },
θ
(9.72)

where

0 ≡
A ω α |Ψα
0 Ψ0 |.
α
(9.73)
α

This is the QDN generalization of the standard POVM formalism.

130 Mixed States and POVMs

We note that, provided we sum θ over all elements of an identity class Θ only,7
then
 
E θN = 10 U N,0 I S ⊗ QθN UN,0 10 = 10 U N,0 I S ⊗ IN UN,0 10 = I S . (9.74)
θ in Θ θ in Θ

Hence
 
 
E[QN |0 ] = TrS
θ
0 A
EN θ
= TrS {0 A I S } = 1. (9.75)
θ in Θ θ in Θ

The interpretation of the expectation values E[QN θ |0 ] therefore is consistent

with probability, provided we restrict θ to a single identity class.

7
Identity classes are discussed in the previous chapter.
 10
Double-Slit Experiments

10.1 Introduction
In this chapter we show how the quantized detector network (QDN) formal-
ism describes the double-slit (DS) experiment. This is arguably the simplest
experiment that demonstrates quantum affects such as wave–particle duality and
quantum interference. It continues to be the focus of much debate and experiment
(Mardari, 2005), because theoretical modeling of what is going on reflects current
understanding of quantum physics and hence physical reality. We will apply QDN
to two variants: the original DS experiment and the monitored DS experiment,
where an attempt is made to determine the imagined path of the particle.
The DS experiment is widely acknowledged by physicists to be of importance
to the understanding of quantum mechanics (QM). So much so that in 2002,
the single electron version, first performed by Merli, Missiroli, and Pozzi (Merli
et al., 1976), was voted by readers of Physics World to be “the most beautiful
experiment in physics” (Rosa, 2012).
The DS experiment can be discussed in terms of three stages, shown in
Figure 10.1. By the end of the preparation stage, Σ0 , a monochromatic beam of
light or particles has been prepared by a source P , such as a laser. The beam
emerges from point O and then passes through an information void V1 to the
first stage, Σ1 , which consists of a wall or barrier W . This wall has two openings
denoted A and B that allow parts of the beam to pass through into another
information void V2 and onto the second and final stage Σ2 , which consists of a
detecting screen S.
The screen S is in general some material that can absorb and record particle
impacts. In reality, any screen will consist of a finite number of signal detectors,
such as photosensitive molecules, but the typical QM modeling is done as if
there were a continuum of sites on the screen, such as C, that could register
particles.
132 Double-Slit Experiments

V
D
A
C z
O
P
B
S
E
W

Figure 10.1. The DS experiment. C is a typical detector site in the detecting

screen S.

10.2 Run Protocol

In addition to the above geometrical architecture, there are some important
protocol features that need to be clarified.

Runs
Any DS experiment will consist of a large, possibly enormous, number of runs,
or repetitions of a basic protocol.

Statistics
The conclusions of the experiment are based on a statistical analysis of the data
averaged over all valid runs.

Preparation Stage
Each run consists of the observer establishing the start of that run, indicated by
stage Σ0 in Figure 10.1. This means that the observer will have reliable contextual
information that some previously agreed-on procedure has been carried out at
device P . If anything occurs to prevent confidence in that information, then that
run is discarded.

Outcome Stage
At the end of each run, corresponding to stage Σ2 in Figure 10.1, the observer
looks at every accessible point/detector on the screen S and records whether
each detector has a signal or not. The data from each run are entirely classical
at this point, being in the form of a vast number of bits of yes/no values, each
value coming from a given detector on the screen S.

Information Void
It is a critical feature of any quantum experiment, apart from the quantum
Zeno-type experiments discussed in Chapter 15, that between stages Σ0 and
 10.3 Baseball in the Dark 133

Σ2 , no attempt is made by the observer to extract any information from either

information void V1 or V2 , or at stage Σ1 .

Reset
At the end of each run, the detectors in screen S are reset to their ground states,
ready for the next run, which takes place as if no other run had ever occurred.

10.3 Baseball in the Dark

It is significant that in the above description of protocol, Section 10.2, no reference
is made to particles, beams, interactions, waves, or any such classical mechanics
(CM) imagery. What has been described is only what the observer actually does
in the laboratory, not what theorists imagine they are doing. Indeed, we not
pushing this point too far to remind the reader that even the above “objective”
account is contextual, being human-centric. From any other species’ point of
view, nothing of importance would be going on in the laboratory. A dog, for
example, would be more concerned where the observer kept their sandwiches. Of
course, when we discuss such experiments in practice, it is most convenient to
objectify procedures in familiar terms, so we will usually talk about preparing
a beam of particles at P and allowing that beam to pass through a double slit.
We may even be caught out referring to photons impacting on a screen. But all
of that is to be read as a convenience, not as a statement of belief that there are
objective things known as photons.
When discussing quantum processes, we should be wary of invoking undue
mental imagery. In this respect, a helpful analogy is to imagine the DS experiment
as “baseball played in the dark.” Suppose we were asked to describe a game
of baseball or cricket played not in broad daylight but at night in pitch black
conditions and with no sound. Now the game would look very different from what
it would look like during the day. The pitcher or bowler would be the analogue
of the preparation device P ; the batter or batsman would be the analogue of one
of the openings, A or B, in the wall W ; and the fielders would be the detectors
in the screen S. P would have some contextual information that they should
throw the ball in a certain direction. Suppose they did that. If the ball reached
the batter, it would be struck in some random direction. The odds of any of
the fielders catching that ball at night (the analogue of a quantum experiment)
would be quite different from the odds during daytime (the analogue of a classical
experiment), because in daylight, fielders would have constant visual information
as to the current position of the ball. At night, they would be faced with a real
information void.
The above description of baseball in the dark is a classically based attempt to
convey some of the attributes of observation; it cannot adequately account for
all the nonclassical attributes of quantum processes. Any experimentalist who
has done real DS experiments would possibly find the description of protocol
134 Double-Slit Experiments

in Section 10.2 simplistic and perhaps misleading. What we have described is a

highly idealized version of rather complex procedures that involve a great deal of
beam calibration, noise suppression, timing protocols, shielding, detector physics,
and complex electronics. Indeed, major experiments such as the search for the
Higgs particle at the Large Hadron Collider take data during a small fraction
of the duration of the experiment, the rest of the time being spent in planning,
funding, construction, and calibration of the apparatus. It is a common feature
of experimental physics doctorates that constructing the apparatus takes two or
more years in preparation, and then there is a frantic race to take enough data
to justify submitting a thesis.
Perhaps the best way to think of these issues is to accept that theory does not
describe reality directly but deals with equivalence classes of processes (Kraus,
1983). For example, what we mean by a “double-slit experiment” is a theoretical
model of the equivalence class of activities in the laboratory that each has the
essential features outlined in Section 10.2, disregarding many contexts such as
whether the observer is male or female, wears a hat or not, and so on.
What is truly remarkable is that when we overlook these issues in much in the
same way as friction is overlooked in Newtonian mechanics, then there emerges
from the overall complexity of any experiment some simple theoretical rules as
to what is going on. That is really how QM was discovered. Moreover, these
rules have great applicability and in the case of QM, work particularly well and
far better than CM in experiments such as the DS experiment. QDN should be
seen in this light: it will give the essential architecture of an experiment but
not a detailed description of the “friction” encountered in quantum experiments,
unless that is called for.

10.4 Observed Phenomena

The DS experiment reveals a number of phenomena that continue to puzzle
physicists, because those phenomena cannot be fully explained according to the
principles and ideology of CM.

Line-of-Sight Violation
With reference to Figure 10.1, suppose opening B is blocked off but otherwise the
experiment runs as described above. According to CM, particles passing through
opening A should pass more or less undisturbed onto position D on the screen
S. Of course, the opening at A would deflect some of the particles, since the
wall W consists of atoms, so it is to be expected that at the edges of opening
A, forces will act on any beam particles passing nearby. However, the fraction of
all the particles passing through either opening that comes close to the edges of
the slits is expected to be relatively small, so we expect to find a relatively large
distribution of particles around position D on the screen S with relatively few
to either side.
 10.4 Observed Phenomena 135

What is observed is quite different. There is indeed a broad peak found centered
on D, but it extends further over the screen than expected from naive CM
expectations. Let us call the probability profile of this broad peak P A (z), where
z is the position coordinate of points along the screen as shown in Figure 10.1.
If now instead of B we blocked off opening A, then we would get a similar broad
distribution P B (z), this time centered on E, on the line of sight from O to B.

Interference
What is astonishing is that if now we ran the experiment with both A and B open,
then we would not find a simple distribution P A (z) + P B (z) as expected from
CM principles. Instead, we would find a distribution P A (z) + P B (z) + I AB (z),
where I AB is known as an interference term. It is this interference term that
causes all the fuss.

Self-interference
It was originally believed that a theoretical particle–wave conflict could be
avoided if the explanation of the interference term I AB was that there were
interactions of the particles coming from opening A that were somehow
interacting with particles coming from opening B, thereby disrupting the basic
addition of P A to P B on the detecting screen.
But the mystery was only deepened when DS experiments were done such that
the rate of particles falling on the screen was extremely low (Taylor, 1909). So
low in fact that there would be one (or even fewer than one)1 on average landing
on the screen S per run. It was found that the interference term occurred in the
statistical analysis when a large number of runs was performed, even when only
one particle could possibly have passed through at a time. The point is that the
aforementioned picture of clouds of particles interfering with each other cannot
be valid here when only one particle passes through at a time. The paradox is
that it is then hard if not impossible to understand why the interference term
should occur in the analysis of many separate, uncorrelated runs. When QM
is interpreted in terms of particle–waves interfering with each other, the low-
intensity interference pattern is generally referred to as self-interference, a term
commonly attributed to Dirac in his famous book on QM (Dirac, 1958).
Not all theorists subscribe to this view (Mardari, 2005). From the QDN per-
spective, the term self-interference is a dangerous one to take seriously, as it
invokes a confusing picture of a classical particle that behaves unlike a classical
particle. We do not have to believe in particles: the only things that we can be
sure of are signals in detectors.

1
This is deep. An average of fewer than one particle per run would mean, of course, that
during some runs, no particles were recorded as having landed on the screen. An average of
fewer than one particle per run vindicates the view that what matters is what the observer
does, not what we think they do. Observers push buttons and look at screens. They do not
know for sure what is happening beyond that description.
136 Double-Slit Experiments

Path Indeterminacy
The interference term IAB requires both openings A and B to be present. If any
attempt is made to block off one or another opening, then I AB disappears. Now
according to CM, every time a detector registers a signal at S, that is evidence
for a particle having traveled from source P to that detector. According to CM
principles, that particle had to have traveled along a continuous path from O
through either A or B. If true, it should be possible to establish which of the two
openings it was. But any attempt to do this appears to destroy I AB .
It is as if the observer has two mutually exclusive choices: either have no
knowledge about which path was taken, and then P A , P B , and the interference
term occur, or know which path was taken, such as through A, but then the inter-
ference term and P B disappear and only P A is observed. In QM, the principle
that there is this exclusive choice is generally referred to as complementarity.
It may be possible to trade off information, with the observer having only
a probability estimate for each of these alternatives. In that case, the overall
distribution on the screen would depend on that probability estimate in some
way. However, predicting that distribution would undoubtedly require the most
careful analysis of context. Experiments along such lines have been attempted,
such as that of Afshar (Afshar, 2005). The results of such experiments remain
controversial (Kastner, 2005).

Wave–Particle Duality
We shall see in the next section that physicists can get a good theoretical handle
on the experiment by applying the particle–wave concept inherent to Schrödinger
wave mechanics. The DS experiment touches on particle-like attributes because
the detectors in the screen S respond in a discrete yes/no way characteristic of
particle impacts, while on the other hand the distribution P A +P B +I AB is char-
acteristic of wave dynamical processes. The mystery only arises when theorists
suppose that there are objects with particle and wave properties simultaneously
and fail to recognize empirical context as a critical factor in the experiment. In
actuality, each aspect (particle or wave) is significant within its own particular
empirical context, and there are no real paradoxes in the laboratory. The so-
called wave–particle paradox occurs only because of the way humans generally
choose to interpret their experiments.

10.5 A Wave-Mechanics Description

In this section we discuss the DS scenario using standard non-relativistic QM
for a wave–particle of nonzero mass m propagating in three spatial dimensions.
The standard theory assumes that apart from the production process in the
source P , the interaction with the wall W , and the screen S, the quantum state
representing the dynamics propagates in the information void regions V1 and V2
according to the Schrödinger–Dirac equation
 10.5 A Wave-Mechanics Description 137

d 
i |Ψ, t = H|Ψ, t, (10.1)
dt
 is the free particle Hamiltonian given by
where H

H = p · p . (10.2)
2m
Relative to a standard improperly normalized particle position basis
{|x, x ∈ E3 }, the state vector |Ψ, t is given by
%
|Ψ, t = d3 xΨ(x, t)|x. (10.3)

The wave-function part Ψ(x, t) of the solution to Eq. (10.1) is readily found by
standard methods to be given by
%
Ψ(x, t) = d3 yF (x, t; y, t0 )Ψ(y, t0 ), t > t0 , (10.4)

&
where the propagator F (x, t; y, t0 ) ≡ x|e−iH(t−t0 )/ |y is given by (Feynman
and Hibbs, 1965)
 3/2  '
−im im(x − y)2
F (x, t; y, t0 ) = exp , t > t0 . (10.5)
2π(t − t0 ) 2(t − t0 )
This propagator will be valid for events (x, t) and (y, t0 ) in the same information
void region. We shall consider what happens in region V1 and then in region V2 .
In the following, we take standard Cartesian coordinates x ≡ (x, y, z) with
origin at O in Figure 10.1, x-axis along the beam direction, y-axis transverse
to the beam, and z-axis in the direction from B to A. For simplicity, we shall
suppress any transverse effects and assume that openings A and B are almost
point-like, with coordinates xA = (d, 0, a) and y B = (d, 0, −a), respectively,
where d is the distance of the wall W from the opening O and a is positive.
We shall consider what happens at a point C on the detecting screen S, with
coordinates xC = (d + D, 0, c), where D is the distance of the screen S from the
wall W .
For the rest of this section we use the notation x1 ≡ (x1 , t1 ), y2 ≡ (y 2 , t2 ),
and so on.

Region V1
We imagine that at the first stage Σ0 , a normalized pulse is emitted from O,
characterized by wave function Φ(x0 ). At a given event with coordinates x1 on
the wall W , the wave function Ψ(x1 ) after propagation through V1 is given by
%
Ψ(x1 ) = d3 x0 F (x1 ; x0 )Φ(x0 ), t 1 > t0 . (10.6)

Conservation of probability requires that

% %
d3 x1 |Ψ(x1 )|2 = d3 x0 |Φ(x0 )|2 , (10.7)
138 Double-Slit Experiments

which is satisfied by virtue of the relation

%
d3 x1 F ∗ (x1 ; y0 )F (x1 ; x0 ) = δ 3 (x0 − y 0 ). (10.8)

Region V2
Equation (10.6) gives the wave function impacting on the wall W on the side
facing the first information void V1 . We need an expression for the wave function
on the other side of W that acts as an initial wave function propagating into
information void V2 and hitting the screen S. Stage Σ1 can be thought of in this
respect as a preparation stage.
To this end we introduce shape functions GA (x1 ) and GB (x1 ) that characterize
the openings A and B. Feynman and Hibbs (1965) discuss this calculation where
a Gaussian shape function is assumed.
 A The prepared  wave function on the V2
B
side of the wall is then given by G (x1 ) + G (x1 ) Ψ(x1 ).
Propagation through V2 follows the same pattern as through V1 . The final
stage wave function Ψ(x1 ) for an event on S at stage Σ2 is given by
%
 
Ψ(x2 ) = d3 x1 F (x2 ; x1 ) GA (x1 ) + GB (x1 ) Ψ(x1 ). (10.9)

What matters here is the squared modulus |Ψ(x2 )|2 , which according to the
Born interpretation (Born, 1926) is the probability density relevant to outcome
detection on the screen S. From (10.9) we find
|Ψ(x2 )|2 = P A (x2 ) + P B (x2 ) + I AB (x2 ), (10.10)
where
%
A
P (x2 ) = d3 x1 d3 y 1 F (x2 ; x1 )F ∗ (x2 ; y1 )GA (x1 )G∗A (y1 )Ψ(x1 )Ψ∗ (y1 ),
%
P B (x2 ) =d3 x1 d3 y 1 F (x2 ; x1 )F ∗ (x2 ; y1 )GB (x1 )G∗B (y1 )Ψ(x1 )Ψ∗ (y1 ),
%  A '
∗ G (x1 )G∗B (y1 )+
AB 3 3
I (x2 ) = d x1 d y 1 F (x2 ; x1 )F (x2 ; y1 ) Ψ(x1 )Ψ∗ (y1 ).
GB (x1 )G∗A (y1 )
(10.11)
In these integrals, we may use (10.6) to work out the outcome probabilities
from a knowledge of the detectors in the screen S and the characteristics of
the preparation device P ; that is, we should be able to specify the initial wave
function Φ(x0 ) reasonably well.
Blocking off opening B corresponds to setting GB to zero, and then we see
from (10.11) that the interference term and P B vanish. A similar remark applies
the blocking off of opening A.

10.6 The QDN Account of the Double-Slit Experiment

We are now in position to describe the DS experiment via QDN. Applying
our bitification process, we introduce qubits at all those sites where significant
 10.6 The QDN Account of the Double-Slit Experiment 139

12
22
A 1
1 32
P 10 B 42 S
21 52

K2
S0 S1 S2

Figure 10.2. The DS experiment.

information could in principle be extracted. This excludes the information void

regions V1 and V2 , and all parts of the wall W apart from the two openings A
and B. The QDN architecture is given by Figure 10.2. We model the detecting
screen as a (possibly vast) number K of detectors and now discuss each run,
stage by stage.

Stage Σ0
The initial normalized labstate Ψ0 prepared by apparatus P by stage Σ0 is
denoted in QDN by

Ψ0 = 10 , (10.12)

where 10 is the signal state element of the one-qubit quantum computational

basis representation (CBR) {00 , 10 }. Essentially, Ψ0 carries the information that
the run should proceed, analogous to the proposition “go for burn,” validated by
Mission Control prior to moon rocket engine ignition, after all safety checks had
been completed (Apollo Program Office, 1969). We shall refer to such a one-qubit
register as a preparation switch.
Note that we are ignoring the internal spin state of the electromagnetic radi-
ation, as polarization is not a factor in this version of the experiment. However,
such effects can be included easily if required.

Stage Σ1
The QDN description of Ψ1 , the labstate at stage Σ1 , has a dual role. On
the one hand it represents what the observer could find (statistically) if they
looked at either or both openings A and B at that stage. Actually, doing this
would be part of the complex calibration processes involved in setting up the
experiment in the first place. On the other hand, Ψ1 should be regarded as the
initial labstate for propagation to the next stage, Σ2 , landing on the screen S.
If this latter alternative is chosen, then the observer should make no attempt to
extract information from Ψ1 .
The rules of QM lead us to assert that
 1 + α2 A
Ψ1 = U1,0 Ψ0 = (α1 A  2 )01 , (10.13)
1 1
140 Double-Slit Experiments

where α1 and α2 are complex numbers satisfying the normalization condition

|α1 |2 + |α2 |2 = 1.
Here U2,1 is a semi-unitary operator taking normalized labstates from Q0 to
Q1 ≡ Q11 Q21 . We shall comment on the nature of this operator presently. At
this point, we note that signality is preserved in the transition from stage Σ0
to stage Σ1 , since the initial labstate Ψ0 ≡ A  1 00 has signality one, and by
0
inspection of (10.13), Ψ1 also has signality one. We rule out (by hand) terms
 1A
proportional to the signality-zero state, 01 , or to the signality-two state A 2
1 1 01 ,
as these represent dynamics different from the one that we wish to explore here.
When charged particles such as electrons are involved in the DS experiment,
charge conservation rules out changes of signality. For bosons such as photons,
it is quite possible to encounter signality nonconservation, such as in the case
of parametric down-conversion (Burnham and Weinberg, 1970; Klyshko et al.,
1970).

Stage Σ2
The transition from stage Σ1 to stage Σ2 is handled as follows. We assert that
 a 01 , a = 1, 2, on
there is a semi-unitary operator U2,1 such that for each term A 1
the right-hand side of equation (10.13),

K
 a 01 =
U2,1 A j,a  j
U2,1 A2 02 , a = 1, 2, (10.14)
1
j=1

assuming there are K detector sites on the detecting screen S. This process
 U2,1presently. Conserva-
preserves signality. We shall comment on the nature of
j,a
tion of probability requires the complex coefficients U2,1 to satisfy the semi-
unitarity rule

K
U j,b∗ j,a ab
2,1 U2,1 = δ , (10.15)
j=1

where U j,b∗ j,b

2,1 is the complex conjugate of U2,1 . We note that Eq. (10.15) is the
QDN analogue of Eq. (10.8).
The linearity rules of QM now give the relationship between the initial labstate
Ψ0 and the final labstate Ψ2 to be

2 
K
j,a  j
Ψ2 = U2,1 U1,0 Ψ0 = αa U2,1 A2 02 . (10.16)
a=1 j=1

It can be readily checked, using the normalization condition and the above semi-
unitarity rules that Ψ2 Ψ2 = 1.
We can now readily calculate all outcome probabilities, by choosing any of
the 3N maximal or partial questions. For example, the conditional probability
Pr(k2 |Ψ0 ) that the kth detector at stage Σ2 would be in its signal state is given
by Pr(k2 |Ψ0 ) = Ψ2 Pk Ψ2 , which readily evaluates to the value
2
 10.7 Contextual Subspaces 141


2
Pr(k2 |Ψ0 ) = αa∗ αb U k,a∗ k,b
2,1 U2,1 , k = 1, 2, . . . , K. (10.17)
a,b=1

Exercise 10.1 Prove (10.17) and show that total probability is conserved,
that is,

K
Pr(k2 |Ψ0 ) = 1. (10.18)
k=1

10.7 Contextual Subspaces

Suppose Un+1,n is the semi-unitary evolution operator from quantum register
Qn at stage Σn to quantum register Qn+1 at stage Σn+1 . Suppose further that
we have a complete specification of the action of Un+1,n , in the form of the rules
for CBR element evolution given by
2rn+1
−1 j,i
Un+1,n in = Un+1,n j n+1 , rn+1 = rank Qn+1 . (10.19)
j=0

Then using completeness, we find the dyadic representation

n −1 2rn+1 −1
2r  j,i
Un+1,n = j n+1 Un+1,n in . (10.20)
i=0 j=0

Then we have the rule

Un+1,n Un+1,n = In , (10.21)
where Un+1,n is the retraction of Un+1,n and In is the identity operator over Qn .
At this point, we are confronted with what appears to be a serious problem;
we do not have all the information that allows us to construct the full evolution
operator U2,1 in the DS experiment. The number of elements in the initial
preferred basis B1 for Q1 is four, but of these, only two have signality one,
 1 01 and A
that is, A  2 01 . The only specific information we have is given by the
1 1
relations (10.14) for those two elements of the preferred basis.
Fortunately, this problem is easily circumvented by the observation that for a
DS experiment, an observer will not in general be interested in the complete quan-
tum registers Qn and Qn+1 but only in the subspaces spanned by the signality-
one elements of their respective preferred bases. This is really the meaning of the
term self-interference.
This leads us to define the notion of contextual basis and contextual subspace.

Definition 10.2 In a given experiment, a contextual basis Bnc is a subset of

the preferred basis Bn for the quantum register Qn at stage Σn , the elements
of Bnc being dictated by the context of the experiment.
142 Double-Slit Experiments

Definition 10.3 A contextual subspace is a subspace Qcn of a quantum

register Qn , the preferred basis for Qcn being a given contextual subset Bnc of
the preferred basis Bn for Qn

In the case of the DS experiment, the contextual bases B1c , B2c are given by all
the respective signality-one states, so we have
   
B1c ≡ A  1 01 , A
 2 01 , B c ≡ A i 02 : i = 1, 2, . . . , K . (10.22)
1 1 2 2

These define the contextual subspaces Qc1 and Qc2 . From (10.14), we can construct
the contextual evolution operator Uc2,1 and its retraction Uc2,1 . These are given
in the CBR by

2 
K
j,a j−1 a−1

2 
K
Uc2,1 ≡ U2,1 22 21 , Uc2,1 ≡ U j,a∗ a−1 j−1
2,1 22 21 (10.23)
a=1 j=1 a=1 j=1

and satisfy the required relation Uc2,1 Uc2,1 = Ic1 , where Ic1 is the stage-Σ1 contextual
identity
Ic1 = 11 11 + 21 21 . (10.24)

Remark 10.4 The real world of experience and the world of the theorist’s
imagination are each far too complex to understand fully. Experiments
attempt to limit the complexity of the former by focussing on a limited
number of detectors. Contextual subspaces implement that strategy as closely
as possible, limiting the amount of complexity that the theorist needs to face.

Usually it will be clear by context when we are dealing with contextual sub-
spaces, so we shall usually drop the superscript c in our notation.

10.8 The Sillitto–Wykes Variant

The DS experiment was identified by Feynman, Leighton, and Sands (FLS) as the
fundamental experiment to understand. They wrote that “it contains the only
mystery” (Feynman et al., 1966). From the beginning of QM, experimentalists
sought to probe this mystery deeper, such as greatly reducing the light intensity
(Taylor, 1909) in the case of electromagnetic waves. Technology finally made
possible experiments where one electron came through the device at a time (Merli
et al., 1976).
Another variant was performed by Sillitto and Wykes, who arranged for the
two openings, A and B, to be opened and closed so that only one was open
at a time (Sillitto and Wykes, 1972). Nevertheless, an interference pattern was
observed. This seems at first sight impossible to understand if we think in terms
of particles.
 10.9 The Monitored Double-Slit Experiment 143

12

A 22
11
32

P 10 42 S
B 52
21

K2

S0 S1 S2
Figure 10.3. The QDN explanation of the Sillitto-Wykes experiment result
is that stage Σ1 need not coincide with a hyperplane of simultaneity in the
laboratory, or indeed, in any other frame of reference.

We can readily explain the result of the Sillitto–Wykes experiment in QDN

by pointing out that time is handled in QDN in terms of stages. In Figure 10.2,
we note that the openings A, B are associated with stage Σ1 , not a specific
laboratory time. We have stressed previously that stages are analogues of space-
like hypersurfaces in conventional relativity, but with an important proviso:
shielding can play a role. In the case of the Sillitto–Wykes experiment, a more
appropriate diagram is Figure 10.3. In that representation, stage Σ1 is not
necessarily physically space-like everywhere. What matters is that the optical
paths (in the case of photons) or trajectory channels (in the case of electrons)
from P through openings A and B are such that wave trains from the two slits
subsequently “intersect and interfere” at the screen S. Whether A and B are
open “at the same labtime” turns out to be irrelevant. Indeed, that is as it
should be, because it is not possible anyway to determine when any “particle”
passes through either hole without destroying the interference pattern on S.

10.9 The Monitored Double-Slit Experiment

A fundamental question raised by the DS experiment concerns the particle inter-
pretation of quanta. Speaking classically, if a particle such as an electron is
released from source P and lands on screen S, then “it stands to reason” that
that particle must have passed through either opening A or opening B. That is,
after all, what is meant by a “particle.”
But this is just an appeal to intuition. We have emphasized that our interpre-
tation of QM is that it is really a theory of entitlement: QM tells us what we
are entitled to say in a given context, and no more and no less. Therefore, if we
have not monitored through which slit an electron has gone, we are not entitled
to say it had to have gone through one opening for sure. We do not even have to
144 Double-Slit Experiments

think of that as having happened in any way; physicists are not in the business
of believing in unverified propositions.
It is that lack of entitlement that really upsets people, because it destroys the
comfortable belief that real things are going on, even when we are not observing
them. Physicists, being people, felt compelled to find out whether the path of
electron could be determined when an interference pattern on screen S was found.
The monitored DS experiment was devised in order to investigate this issue.

The Feynman–Leighton–Sands Discussion

We shall first discuss the treatment given by FLS in Feynman et al. (1966). The
apparatus is given in Figure 10.4.
The apparatus is the same as for the original DS experiment except for three
items. L is a source of light and DA and DB are photon detectors. The idea is
that if an electron passes through opening A, then there is a possibility that it
will interact with light from the source L, causing a signal in DA . A signal in
DA may therefore be an indicator that the electron has passed through opening
A and not B. A similar remark applies if the electron passes through opening B.
The analysis goes as follows. First, consider the case when the light source is
absent. If opening B is closed off and an electron passes through opening A and
lands on the screen at C, then the amplitude is φ1 . Conversely, if opening A is
closed off and an electron passes through opening B and lands on C, then the
amplitude is φ2 . If A and B are both open, then the amplitude at C is φ1 + φ2 .
This is just the QM description of the original DS experiment.
When the light source L is present and both slits are open, then the situation
needs some careful analysis. Suppose an electron has been detected at C and

V V

DA

A C
z
P L

B
DB S
W

Figure 10.4. The monitored DS experiment.

 10.9 The Monitored Double-Slit Experiment 145

a photon detected at DA . Then two alternative paths contribute to the overall

amplitude A(C|P ).

Path A
The electron passes through slit A and interacts with the light, with an amplitude
at C written as aφ1 , where a is a factor representing the electron–electromagnetic
field interaction.

Path B
The electron passes through slit B and interacts with the light, with an amplitude
at C written as bφ2 , where b is a factor representing the electron–electromagnetic
field interaction.

The coefficients a and b are expected to be different, from the geometry

of the situation. The overall amplitude at C, according to Feynman’s sum of
paths prescription, is therefore the sum of the two contributions, and so given
by A(C, D A |P ) = aφ1 + bφ2 . Assuming suitable normalization, the probability
Pr(C, DA |P ) for an electron to land on C and a photon to be registered at DA ,
conditional on the electron being fired from P , is therefore given by
Pr(C, DA |P ) = |A(C, DA |P )|2 = |aφ1 + bφ2 |2 . (10.25)
If, on the other hand, an electron has been detected at C and a photon
detected at DB , then a similar argument (using symmetry) gives Pr(C, DB |P ) =
|aφ2 + bφ1 |2 .
The probability Pr(C, DA or DB |P ) of an electron landing at C and a photon
being detected at either DA or DB is the sum of the two probabilities, not the
squared modulus of the sum of the two amplitudes (a point stressed by FLS),
and is therefore given by
Pr(C, DA or DB |P ) = |aφ1 + bφ2 |2 + |aφ2 + bφ1 |2 , (10.26)
which is Eq. (3.10) in Feynman et al. (1966).

The QDN Discussion

The relevant QDN version of Figure 10.4 is Figure 10.5. In the latter figure,
we have two extra nodes, labeled A2 and B2 , added to stage Σ2 detectors in
the screen S. These new nodes represent the detectors DA and DB shown in
Figure 10.4.
The QDN analysis in this case goes by the three stages discussed in the original
scenario. Nothing is different by stage Σ1 , so Eq. (10.13) is still valid. However,
the jump to stage Σ2 has to treated more carefully. We deal with each opening
separately.
In the following, a = 1 corresponds to opening A and a = 2 corresponds to
opening B. We expect the following dynamics, which is more general than that
considered in Feynman et al. (1966):
146 Double-Slit Experiments

A2
B2
A 1 12
1

P 10 22 S
B
21 32

K2

S0 S1 S2

Figure 10.5. The QDN stage diagram for the monitored DS experiment.


K
 a 01 =
U2,1 A  j + V j,a A
(U j,a A j A
A j,a  j  B j A
A2 A2 + X j,a A A B
2 2 +W 2 2 A2 )02 . (10.27)
1
j=1

 
2

 
 
 
(a) (b) (c) (d)

The four terms on the right-hand side represent the following alternatives:
(a) U j,a is the signality-one amplitude that an electron has passed through
opening a and has landed on the screen at site j and no signal is detected in
either DA or DB .
(b) V j,a is the signality-two amplitude that an electron has passed through
opening a and has landed on the screen at site j and a signal is detected in
DA and not in DB .
(c) W j,a is the signality-two amplitude that an electron has passed through
opening a and has landed on the screen at site j and a signal is detected in
DB and not in DA .
(d ) X j,a is the signality-three amplitude that an electron has passed through
opening a and has landed on the screen at site j and a signal is detected in
DA and a signal is detected in DB . This possibility is not considered in the
calculation of FLS.
The labstate Ψ2 is given by

2 
K
 j + V j,a A
αa (U j,a A j A
A j,a  j  B j A
A2 A2 + X j,a A A B
Ψ2 = 2 2 2 +W 2 2 A2 )02 . (10.28)
a=1 j=1

From this we can readily determine all the various probabilities of interest by
asking the appropriate questions. For example, if we ask for the probability
that the electron has landed on the kth detector site, and there is a signal in
DA , and there is no signal in DB , we need to calculate the expectation value
k P
Ψ2 P A B
2 2 P2 Ψ2 . To evaluate this, we first note that the signal algebra gives


2
P A B
k P kAA
2 2 P2 Ψ2 = αa V k,a A 2 2 02 , (10.29)
a=1
 10.9 The Monitored Double-Slit Experiment 147

and then we readily find

 2 
 
k P
A B  
Ψ2 P 2 2 P2 Ψ2 = α V | = α1 V k,1 + α2 V k,2 |2 .
a k,a 2
(10.30)
 
a=1

k PA P
Similarly, the probability Ψ2 P B
2 2 2 Ψ2 , that the electron landed on the kth
screen site and there was no signal in DA and there was a signal in DB , is
given by
k PA P
Ψ2 P B
2 2 2 Ψ2 = |α W + α2 W k,2 |2 .
1 k,1
(10.31)

We recover the FLS result in their notation if we use symmetry, setting α1 = α2

and

V k,1 = aφ1 , V k,2 = bφ2 , W k,1 = aφ2 , W k,2 = bφ1 . (10.32)

The QDN calculation allows greater flexibility in the architecture of the exper-
iment than that assumed in the FLS calculation, but the price is that more
assumptions have to be made about the various amplitudes.
 11
Modules

11.1 Modules
In this chapter, we discuss important modules such as Wollaston prisms, Newto-
nian prisms, beam splitters, phase shifters, mirrors, and null modules. Modules
are pieces of apparatus placed at various strategic positions in the informa-
tion void and are designed to influence labstate amplitudes at signal detectors.
Modules are therefore a critical component in the quantized detector network
(QDN) approach to quantum mechanics (QM).
Modules are not classified as either real or virtual detectors, because no infor-
mation is extracted from them. Their function is solely to influence quantum
amplitude propagation between stages in specific ways. Even empty space (the
vacuum) can be regarded as a module, for the propagation of signals through
nominally “empty” space is a fundamental subject in its own right. A particularly
important example illustrating how nontrivial that can be is the Hubble–Doppler
red shift of light from distant galaxies.
In the QDN description of the double-slit (DS) experiment, shown in Figure
10.2, the stage Σ1 detectors labeled 11 and 21 may be considered as on the V2
side of the wall W , where V2 is the information void between stages Σ1 and Σ2 .
Viewed in this way, the wall may be interpreted as a part of the appara-
tus that is positioned in the information void region V1 so as to influence the
signals obtained at detectors 11 and 21 . Virtually all experiments have similar
components of apparatus in the information void. We shall call any such piece of
apparatus a module. Modules are necessary to the architecture of the experiment,
are situated in the information void, and therefore are not detectors.
Examples of modules are Stern–Gerlach (SG) inhomogeneous field magnets,
Wollaston prisms, mirrors, phase changers, and so on. Even empty space, other-
wise known as the vacuum, should be regarded as a module.
Each module has its own physical properties. These include the number of
detectors feeding amplitudes into that module and the number of detectors that
 11.2 The Vacuum 149

amplitudes are fed out from. We shall discuss a number of modules relevant to
the experiments discussed in this book, starting with the vacuum.

11.2 The Vacuum

In this book, we make a distinction between the information void concept and
the vacuum of elementary particle physics. The differences are subtle.
Each concept has the characteristic that no signal detectors are associated
with it. While it is commonplace to talk about “observers in empty space” or
“detectors in the vacuum,” such expressions are manifestly inconsistent if taken
literally, even in classical mechanics (CM). What distinguishes the information
void concept from the vacuum concept is that the latter is a mathematical
objectivization of the former, based on some input context. For instance, if
we believe that the information void has some structure or physical properties
associated with a vacuum, then we may choose to believe in or set up some sort
of mathematical model of empty space, such as a three-dimensional manifold
with a Euclidean metric, as in Newtonian mechanics, or a general relativistic
(GR) spacetime with a metric, or think of it in operator terms, as in Snyder’s
noncommuting spacetime (Snyder, 1947a,b), or even assign a quantum state
vector to it, as in Fock space and relativistic quantum field theory (RQFT).
We are here faced with an important question as to the status of the vacuum:
is it part of the relative external context (REC) that defines the observer, or is
it part of the relative internal context (RIC) that defines the apparatus?
On the one hand, we have argued in earlier chapters that real observers are
always endophysical, that is, are sitting inside the physical Universe. According
to GR, the physical space that we image observers to be sitting in will have
some physical attributes, such as metrical structure, curvature, mass, and energy
densities. In this context, empty space plays a classical, auxiliary role as part
of REC.
On the other hand, the vacuum concept will play a fundamental role in the
calculation of quantum signal propagation amplitudes, such as black hole physics,
where the classical background spacetime structure plays an important, even
crucial, role (Birrell and Davies, 1982; DeWitt, 1975). In that context, the vacuum
contributes to the RIC.
Elementary particle physics has been very successful in modeling the vacuum
as having special relativistic (SR) symmetries and certain physical properties
such as zero electric charge density, and so on. The standard model of particle
physics appears not to need or use any concepts associated with the so-called
quantum gravity program, a conjectural program based on the belief that the
equations of GR and those of QM should be unified. The jury is out on all
programs of research that attempt to give a detailed model of the vacuum, such
as quantum gravity, string theory, and noncommutative geometry such as that
of Snyder (Snyder, 1947a,b).
150 Modules

0
1

Figure 11.1. Amplitude propagation across the information void.

The question of what the best model of the “internal” vacuum of the RIC
could be is a crucial one, because there are many contexts where the answer will
influence the calculation of the quantum signal amplitudes of relevance to QDN.
For example, we may well need to take spacetime curvature into account if we are
sending signals across vast distances. Such a scenario is found in all observations
in astronomy and cosmology.
We have indicated previously that QDN has nothing to say about the infor-
mation void or the vacuum, precisely because QDN models what is happening
at the signal detectors. Therefore, we will need to bring in standard physics
theory, such as relativistic quantum field theory, to help us write down signal
amplitudes between those detectors. Should that appear to make QDN rather
limited, we point out that standard theories do have their limitations as well,
such as the appearance of infinite renormalization constants in quantum field
theory. We interpret that as the same problem seen from the opposite direction.
It will probably be only by a judicious combination of QDN (or whatever should
replace it) on the apparatus side of the physics coin and of standard relativistic
quantum field theory on the reductionist side that we will find a better approach
to empirical physics than with just either alone.
In QDN notation, signal amplitude propagation across the information void is
represented by featureless lines, as in Figure 11.1.

11.3 The Wollaston Prism

A Wollaston prism is a quantum optics module that splits up a beam of light
into two orthogonally polarized beams. Figure 11.2 is a schematic QDN diagram
of such a device.
Consider an initial total state
 
|Ψ0 ) ≡ α|s10  + β|s20  ⊗ 10 , (11.1)

where 10 ≡ A  1 00 and|si  is a stage-Σ0 polarization state vector in two-

0 0
dimensional photon polarization Hilbert space H0 . Here i = 1 or 2 represents
either of two orthogonal polarizations such as horizontal and vertical. The
coefficients α and β are complex numbers. If the initial state is normalized to
unity, then |α| 2 + |β| 2 = 1.
 11.3 The Wollaston Prism 151

Figure 11.2. The Wollaston prism.

The contextual evolution operator1 U1,0 from stage Σ0 to stage Σ1 is assumed

semi-unitary, mapping from a two-dimensional contextual subspace of the initial
four-dimensional Hilbert space H0 ⊗ Q10 to the final eight-dimensional one H1 ⊗
Q1 , where H1 is a copy of H0 and Q1 ≡ Q11 Q21 . The assumed rules for the
Wollaston prism are
U1,0 {|s10  ⊗ 10 } = |s11  ⊗ 11 , U1,0 {|s20  ⊗ 10 } = |s21  ⊗ 21 , (11.2)
where 11 ≡  1 01
A
and 21 ≡  2 01 .
A
1 1
Using contextual completeness we have
U1,0 = |s11 s10 | ⊗ 11 10 + |s21 s20 | ⊗ 21 10 (11.3)
and its retraction
U1,0 = |s10 s11 | ⊗ 10 11 + |s20 s21 | ⊗ 10 21 . (11.4)
These operators satisfy the semi-unitary relation
U1,0 U1,0 = {|s10 s10 | + |s20 s20 |} ⊗ 10 10 = I0H ⊗ Ic0 , (11.5)
the identity operator for the initial contextual total Hilbert space H0 ⊗ Qc0 .
There are two generalized Kraus matrices associated with stage Σ1 , given by
M 11,0 ≡ 11 U1,0 = |s11 s10 | ⊗ 10 ,
M 21,0 ≡ 21 U1,0 = |s21 s20 | ⊗ 10 , (11.6)
with retractions
M 11,0 = |s10 s11 | ⊗ 10 ,
M 21,0 = |s20 s21 | ⊗ 10 . (11.7)
From these, the generalized POVM operators associated with Σ1 are given by
E 11,0 ≡ M 11,0 M 11,0 = |s10 s10 | ⊗ 10 10 ,
E 21,0 ≡ M 21,0 M 21,0 = |s20 s20 | ⊗ 10 10 . (11.8)

1
As stated in a previous chapter, we drop the superscript “c” denoting “contextual.”
152 Modules

In this particular case, these POVMs satisfy the relations E i1,0 E j1,0 = δ ij E i1,0 (no
sum over i) and
 
E 11,0 + E 21,0 = |s10 s10 | + |s20 s20 | ⊗ 10 10 = IH ⊗ Ic0 , (11.9)

the identity operator for the initial contextual Hilbert space H ⊗ Qc0 . From these
operators we find the conditional outcome rates
   
Pr(11 |Ψ0 ) ≡ T r E 11,0 0 = |α|2 , Pr(21 |Ψ0 ) ≡ T r E 21,0 0 = |β|2 , (11.10)

assuming complete efficiency. Here 0 ≡ |Ψ0 )(Ψ0 |.

11.4 The Newtonian Prism

Newton’s researches in optics revealed features of light that demonstrate fun-
damental properties of relevance to us here (Newton, 1704). In this section we
discuss two of his observations with prisms.

The Splitting of Light

Newton found that a beam of white light incident on a prism P 1 would be
split into a spectrum, a set of emerging rays each of a different color, according
to anyone looking at it.2 If any one of those single-color component subbeams
were in turn passed through another prism P 2 , no further splitting occurred;
Figure 11.3(a). From this, Newton concluded that the primary colors in white
light were associated with properties of that light, and not induced by the prism.
Figure 11.3(b) shows the same process as described by QDN.
Significantly, Newton took the view that the perception of color itself was
a sensation induced in the mind by the processes of visual observation. His
reasoning is based on empirical evidence: he noticed that superposing certain
primary colors in the spectrum emerging from a prism created colors such as
purple, in the mind of the observer, that were not contained in the original
spectrum. In that respect he could reasonably be regarded as having a deeper
view of observation than that generally associated with classical mechanics (CM),
which pays no lip service to the observer and their subjective perceptions.
Newton’s idea was vindicated subsequently by the development of the theory
of color vision (the colors that humans believe they “see”), principally by Young
and Helmholtz. They proposed that the human eye uses three distinct types of
receptor (that is, detector). The signal information from these detectors is then
processed to trigger the color sensations that we think we see.
The QDN modeling is based on the tensor product total Hilbert space Hn
at stage Σn defined as Hn ≡ HnEM ⊗ Qn , where HnEM is the standard RQFT

2
Note that this expression is based on the ancient and misleading paradigm that observers
look “at” objects.
 11.4 The Newtonian Prism 153

S P i P i

P P
S K K

Figure 11.3. (a) Newtonian prism P 1 splits an incoming beam of white light
from source S (the Sun) into subbeams associated with primary colors. Prism
P 2 does not split any one of those subbeams further. (b) The QDN schematic
of the same process.

Hilbert space containing free photon states3 and Qn is the apparatus quantum
register at that stage. Relevant states in HnEM are constructed in terms of a
suitable orthonormal set {|in  : i = 1, 2, . . . K}, with orthonormality condition
in |jn  = δ ij .
With reference to Figure 11.3(b), the stage-Σ0 normalized incident state |Ψ0 )
is defined as

K 
K
|Ψ0 ) ≡ ψ0i |i0  ⊗ 10 , |ψ0i |2 = 1. (11.11)
i=1 i=1

The action of the first prism P 1 in Figure 11.3 is taken to be


K
|Ψ0 ) →1 |Ψ1 ) =  i 01 ,
ψ0i |i1  ⊗ A (11.12)
1
P
i=1

which models the splitting up of the original beam into its primary colour
constituents.

Three points to note here are the following.

Index Labels
The electromagnetic state indices i0 and i1 have a temporal subscript that labels
stages only and does not affect the “value” of the index. So, for example, we take
i1 = i0 ≡ i.

Discreteness
Contrary to what we might anticipate from experience with RQFT, the elec-
tromagnetic index in in the above is discrete, not continuous, because we can

3
Any reference to photons is for convenience. The “basis” states here form a discrete set and
therefore are not plane wave solutions but approximate them suitably.
154 Modules

only ever have a discrete set of detectors. This overturns the usual approach
to RQFT, which presupposes a spectrum of photon states with a continuous
frequency range. We do not need to deal with any supposed continuity in HnEM ,
because the only parts of it that we can ever deal with are found at the finite,
discrete detector end. Everything done here is contextual: it is the presupposition
that an unobserved (and unobservable) continuum of states should play a role
in, for example, Feynman graph integrals that contributes to the divergence of
renormalization constants in RQFT. In the laboratory, all signals are finite.
The effect of the first prism P 1 on the incident beam is modeled by the action
of a contextual semi-unitary evolution operator U1,0 , given by

K
U1,0 ≡  i 01 10 ,
|i1 i0 | ⊗ A (11.13)
1
i=1

where K is the number of subbeams in the emerging spectrum that is relevant to

the discussion. Newton chose K to be 7, referring to the subbeams as red, orange,
yellow, green, blue, indigo, and violet (Newton, 1704). This choice is dependent
on the observer, for if we had equipment that could detect infrared or ultraviolet
light, we would have a different value for K.

Coherence versus Incoherence

When physical process are affected by either constructive or destructive inter-
ference of amplitude waves, there are two factors to take into account: the
amplitudes should have the same frequency (more or less), and if so, they should
be coherent. In the case of white light, neither of these factors is in play because by
definition, white light consists of waves of many different frequencies, and these
are necessarily incoherent. That does not mean that we cannot describe such
situations with our QDN formalism. Newton’s recombination experiment is a
demonstration of incoherent superposition but not of constructive or destructive
interference.

With reference to Figure 11.3(a), we see that only one chosen subbeam, labeled
i in Figure 11.3(b), enters the second prism P 2 , all the other subbeams being
blocked off. Then we may write
 
 i 01 ≡ |i2  ⊗ A
U2,1 |i1  ⊗ A  i 02 . (11.14)
1 2

But what about the other, blocked subbeams?

There is a significant point here that we will be able to address more carefully
in Chapter 25 and that has to do with information extraction versus decommis-
sioning. By the term decommissioning, we mean the blocking off of signals and
the elimination of the corresponding detectors from further consideration. We can
see this in Figure 11.3(a), where all except one of the spectral subbeams emerging
from P 1 are blocked off, with the remaining one allowed to pass through prism P 2 .
The QDN approach to this issue is two-fold. In Figure 11.3(b), all of the
stage-Σ1 detectors except i1 are shown shaded. This indicates that those
 11.4 The Newtonian Prism 155

components are either blocked off or observed irreversibly, with consequent

information extraction. This blocking off/information extraction plays no further
role in the experiment, and so is represented in the diagram by null tests taken
from stage Σ1 to stage Σ2 . On the other hand, Figure 11.3(b) indicates that i1
sends a signal through the prism P 2 that is subsequently observed at i2 , which
is now shown shaded in the figure.
Since the evolution from stage Σ1 to stage Σ2 is essentially trivial (which is
what null tests are in practical terms), we may also write
 
 j 01 ≡ |j2  ⊗ A
U2,1 |j1  ⊗ A  j 02 , j = i. (11.15)
1 2

Hence we deduce

K
U2,1 =  j 02 01 Aj + |i2 i1 | ⊗ A
|j2 j1 | ⊗ A  i 02 01 Ai . (11.16)

 2

1

 2 
1
j=1,
j=i null tests through P2

Although the two terms on the right-hand side of this expression look similar,
they are very different contextually. A practical difference would be that |j2  is
an exact copy (in H2EM ) of |j1  (in H1EM ), for j = i, whereas |i2  would not be
an exact copy of |i1  because of, say, attenuation effects as that subbeam passed
through P 2 .
The point here is that context underpins the significance of everything, and
that context is determined by the observer. If they decide to block off all sub-
beams except the one labeled i, and then actually do that, then that action
(not the decision alone) automatically brings in empirically significant context
that involves separate dynamics and apparatus. It is the essence of Wheeler’s
participatory principle quoted on Chapter 1 that action be actually carried out
when discussing quantum processes.
We see from this that the freedom that the observer has to choose how to
rearrange apparatus creates a contextuality. This is emphasized in the next part
of our description of the Newtonian prism, the recombination of a split spectrum
back into the original beam.

The Recombination of Light

In his remarkable book Opticks, Newton discusses an extraordinary variant of the
prism experiment shown in Figure 11.3(a) (Newton, 1704). He inverted prism P 2
and placed a lens between the two prisms. The lens refocused the spectrum emerg-
ing from P 1 onto P 2 , as shown in Figure 11.4(a). The result was to recombine the
spectrum, back into the original form of a beam of white light emerging from P 2 .
Certainly, the recombined beam would not be precisely the same as the original,
in that there would be some attenuation due to passing through glass, but as in
the case of friction in mechanics, such effects can be regarded as secondary and
not relevant to the main discussion.
156 Modules

S P LP

K
S L
P P

Figure 11.4. (a) Newton used lens L and prism P 2 to recombine the spectrum
from prism P 1 back into a beam of white light. (b) In the QDN schematic of
the same process, the lens and second prism constitute module LP .

The QDN stage diagram for Newton’s recombination experiment is shown in

Figure 11.4(b), with the following analysis.

Stage Σ0
The initial labstate is

K 
K
|Ψ0 ) ≡ ψ0i |i0  ⊗ 10 , |ψ0i |2 = 1. (11.17)
i=1 i=1

Stage Σ0 → Σ1
The evolution operator U1,0 that models the action of prism P 1 is given by

K
c
U1,0 =  i 01 10 .
|i1 i0 | ⊗ A (11.18)
1
i=1

Hence the labstate |Ψ1 ) emerging from prism P 1 is given by


K
|Ψ1 ) ≡ U1,0 |Ψ0 ) =  i 01 .
ψ0i |i1  ⊗ A (11.19)
1
i=1

Stage Σ1 → Σ2
The evolution operator U2,1 that models the undoing action of the module LP
consisting of the lens L and the prism P 2 is given, by inspection, by

K
U2,1 =  1 02 01 Ai .
|i2 i1 | ⊗ A (11.20)
2 1
i=1

Here, we have chosen to represent the action of LP as refocusing the spectrum

 1 02 . The final outcome total state |Ψ2 ) is therefore
from P 1 onto labstate 12 ≡ A 2
given by
 11.5 Nonpolarizing Beam Splitters 157


K
|Ψ2 ) ≡ U2,1 |Ψ1 ) = ψ0i |i2  ⊗ 12 . (11.21)
i=1

Comparing this with the original labstate (11.17), we see that the final total
state |Ψ2 ) is a persistent image of |Ψ0 ). Therefore, the combination of prism P 1
followed by prism P 2 is equivalent to a null test of the original beam.
It might be believed that we have encountered an example of semi-unitary
evolution where the initial quantum register had a larger rank than the final
quantum register. In fact, that is not the case. The action of the lens between
the prisms is designed to focus on a one-dimensional subspace of the stage-Σ1
register, a subspace containing the image of the original beam. It is that subspace
alone that is then mapped by the lens and second prism into the one-dimensional
subspace representing the outgoing recombined beam.
When Newton’s recombination experiment is examined in fine detail, only then
can it be appreciated how extraordinarily difficult it is to model what happens
with some degree of correctness. The reader is invited to model this experiment
using only quantum electrodynamics (QED) in its standard formulation, and
they will then see what we mean.

11.5 Nonpolarizing Beam Splitters

Despite the implication in its name, a beam splitter will in general have two
input channels and two outcome channels, as shown in Figure 11.5. In fact, it
is perhaps best to think of a beam splitter as a specific example in quantum
optics of a more general two-two particle scattering process that satisfies certain
properties, such as unitarity and various conservation laws, such as conservation
of energy, momentum, and charge.
In applications, we shall be interested in coherent, signality-one, normalized
initial labstates |Ψ0 ) involving both channels 10 and 20 , of the generic form
 1 00 + ψ 2 |s2  ⊗ A
|Ψ0 ) ≡ ψ 1 |s10  ⊗ A  2 00 , (11.22)
0 0 0

Figure 11.5. The beam splitter.

158 Modules

where |ψ 1 |2 + |ψ 2 |2 = 1, and |si0 , i = 1, 2, are initial, normalized one-photon

states, not necessarily orthogonal. In this scenario, the polarization properties
associated with each input channel are unspecified, so the discussion here is
rather general in that respect.
The dynamics is specified by stating the beam splitter rules for each input
channel. In general, semi-unitary evolution applies (if we ignore inefficiency and
dissipation), so we write
 
U1,0 |s10  ⊗ A 1 00 = α|s1  ⊗ A  1 01 + β|s1  ⊗ A
 2 01 ,
0 1 1 1 1
 
U1,0 |s20  ⊗ A 2 00 = γ|s2  ⊗ A 1 01 + δ|s2  ⊗ A
 2 01 , (11.23)
0 1 1 1 1

where the beam splitter coefficients satisfy the semi-unitary relations

|α|2 + |β|2 = |γ|2 + |δ|2 = 1, α∗ γ + β ∗ δ = 0. (11.24)
It will be useful now and later to define the matrix of coefficients
 
α β
B≡ . (11.25)
γ δ
Then the semi-unitary relations (11.24) tell us that B is in fact a unitary matrix.
It is easy to prove from (11.24) that |α| = |δ| and |β| = |γ|, so we may write
α ≡ ueiA , β ≡ veiB , γ ≡ veiC , and δ ≡ ueiD , where |u|2 + |v|2 = 1 and the
phases A, B, C, and D are real. Now define the phase E by 2E ≡ A + D. Then
it is straightforward to show that B can be written in the form
 
a −b∗
B ≡ eiE , (11.26)
b a∗
where |a|2 + |b|2 = 1 and the phase arguments of a and b are linear combinations
of A, B, and C. We shall refer to the right-hand side of equation (11.26) as the
standard form of a 2 × 2 unitary matrix.
Usually, the phase E in the standard form may be ignored. As for the complex
coefficients a and b, these are related by Fresnel’s equations for the reflection and
transmission of light through optical media. Suppose a monochromatic beam of
light passes through medium μ1 and is incident on a plane surface boundary
of medium μ2 . Generally, it will split into a reflected part that goes back into
medium μ1 and a transmitted part that moves into μ2 . If the speed of light c1
in μ1 is greater than the speed of light c2 in μ2 , then the reflected part will be
out of phase with the incident beam by π, that is, equivalent to a sign change.
There is no such change in the phase of the transmitted part. If, conversely, c1 is
less than c2 , then there are no phase changes in either reflected or transmitted
parts.
With this information there are two forms we may choose for B:
Form 1  
t −r
B= , (11.27)
r t
 11.5 Nonpolarizing Beam Splitters 159

where t and r are real and satisfy the equation t2 + r2 = 1.

 
Form 2 t ir
B= , (11.28)
ir t
where t and r are real and satisfy the equation t2 + r2 = 1.

In stage diagrams such as Figure 11.5, where the square representing the beam
splitter is rotated as shown to face the two input ports 10 and 20 , we shall
adopt the convention that 21 and 11 are the transmitted and reflected beams,
respectively, associated with input port 10 , while they are the reflected and
transmitted beams, respectively, associated with input port 20 . This convention
is used with Form 2 above throughout this book in the encoding of our computer
algebra program MAIN, discussed in the next chapter.
In many experiments discussed in the√literature, Form 2 is assumed, with
t and r taken equal, that is, t = r = 1/ 2. In our computer algebra program
MAIN discussed in the next chapter, we do not do this. We use Form 2, but it is
generally most instructive to leave the transmission and reflection coefficients
arbitrary up to the required constraints. For instance, ti and ri will be the
transmission and reflection coefficients for beam splitter B i and will be arbitrary,
apart from the condition that (ti )2 + (ri )2 = 1.

Signality-Two Input
On occasion, the possibility arises that a signality-two labstate or its equivalent
is incident on a beam splitter. We shall refer to this as beam splitter saturation.
The dynamics for such a scenario has to be decided contextually. One possibility
is that the two photons (speaking loosely) do not emerge and are absorbed by
the device. In such a case, the output channels of the beam splitter remain in
their ground states. On the other hand, we could decide to model what happens
as a case of transparency, such that the output labstate is a signality-two state
with both output detectors registering a signal.

Exercise 11.1 Suppose a calibrated beam splitter satisfies the semi-unitary

dynamics
U1,0 00 = 01 ,
 1 00 = α A
U1,0 A  1 01 + β A
 2 01 ,
0 1 1
 2 00 = γ A
U1,0 A  1 01 + δ A
 2 01 , (11.29)
0 1 1

where α, β, γ, and δ satisfy the semi-unitary relations (11.24).

Show that if
 1A
U1,0 A 2 1 2 1 2
0 0 00 = a01 + bA1 01 + cA1 01 + dA1 A1 01 , (11.30)

where |a|2 + |b|2 + |c|2 + |d|2 = 1 and U1,0 is semi-unitary, then a = b = c = 0

and |d|2 = 1.
160 Modules

11.6 Mirrors
Mirrors are important modules in many experiments. The basic action of a mirror
is to deflect electromagnetic radiation, that is, change path direction.
Depending on the physics, mirrors can also change the phase of electromagnetic
waves. According to Fresnel’s laws of optics, light incident from air on a mirror
undergoes a phase shift of π, if reflecting from the front surface of a mirror, while
there is no phase shift on rear surface reflection.
Naturally occurring mirrors can have a severe influence on signal detection
amplitudes. For example, television signals received at an aerial are built up
from the superposition of all the signals that have followed separate paths from
the transmitting station to that aerial. If a transmitted signal has one path
that goes in line of sight from transmitter to detector and another path that
goes from transmitter, is reflected from the surface of the sea, and then arrives
at the detector, then destructive interference can occur, thereby degrading the
overall detected signal (Laven et al., 1970). Moreover, if the sea level changes
due to tidal action, then the interference varies during the day and cannot be
eliminated easily.
In stage diagrams, mirrors are labeled M .

11.7 Phase Changers

Mirrors are a specific example of more general modules referred to as phase chang-
ers. Such modules are fundamental to experiments where quantum interference
is used to explore material properties, such as in interferometry.
Although phase changers can be used in any context, an important scenario
involves electromagnetic signals, because in classical optics, Maxwell’s equations
lead to the conclusion that light is a wave process involving transversely oscillat-
ing electric and magnetic waves. Phase changer modules other than mirrors will
be labeled by the phase angle involved.

11.8 Polarization Rotators

Electromagnetic waves have transverse electric and magnetic field polarization
degrees of freedom, and this plays a significant role in many experiments. Gen-
erally, the convention is to define the polarization of a plane polarized electro-
magnetic wave as that of the electric field component. By default, the magnetic
polarization is orthogonal to the electric field polarization.
We shall encounter some experiments where a polarized electromagnetic wave
passes through a module that turns the wave’s polarization plane by a known
angle. Such a module will be labeled Rθ in stage diagrams, where θ is the angle
of rotation.
 11.9 Null Modules 161

11.9 Null Modules

A null module is any process that essentially does nothing to a signal. Because of
this, a signal that is passed through a succession of null modules can be thought
of as “on hold” until such time as the observer decides to look.
Null modules are used to model the concept of persistence, discussed in
Chapter 18. On that account, null modules should not be considered trivial in
the sense of identity operators, the action of which does nothing observable.
In contrast, null modules require the right context and reflect a fundamental
property of physics: that structures can persist over time in an observable sense.
Given a rank-r quantum register Qn at stage Σn , a computational basis repre-
sentation of a complete null operator (one that acts on the whole register) Nn+1,n
that maps into stage Σn+1 register Qn+1 of the same rank is
r
2 −1
Nn+1,n = in+1 in . (11.31)
i=0
 12
Computerization and Computer Algebra

12.1 Introduction
The quantized detector network (QDN) formalism was designed from the outset
to deal with detector networks of arbitrary complexity and rank. However, there
is a price to be paid. A qubit register of rank r is a Hilbert space of dimension 2r ,
so the dimensionality of quantized detector networks grows exponentially with
rank. For example, a relatively small system of, say, 10 detectors involves a
Hilbert space of dimension 210 = 1, 024.
Even such a relatively small system cannot be dealt with easily by manual cal-
culation, because quantum mechanics (QM) involves entangled states. Labstates
in QDN are far more complex (in the sense of having far more mathematical
structure) than the corresponding states in a classical register of the same rank.
Some of the mathematical entanglement and separability properties of quantum
labstates are discussed in Chapters 22 and 23.
The quantum entanglement structure of labstates poses a ubiquitous and
serious problem, for both theorists and experimentalists. At this time, there
is significant interest in quantum entanglement, particularly regarding its use
in quantum computing, but theoretical understanding of entanglement is still
surprisingly limited.
On the experimental side, quantum computers with 2000 qubits are currently
being developed (D-Wave Systems, 2016). There is scope here for the application
of QDN to networks of rank going into the many thousands. For those sorts of
systems under observation (SUOs), quantum entanglement makes calculations
by hand far too laborious to be of any practical use.
Fortunately, three factors come to our aid here, making the application of QDN
to large-rank networks a potentially viable proposition.
 12.2 Program MAIN 163

Modularization
We saw in previous chapters that QDN deals with discrete aspects of observa-
tion, despite the fact that standard QM and relativistic quantum field theory
(RQFT) deal with continuous degrees of freedom. This discreteness comes in
three forms, referred to us as stages, nodes, and modules, and all of that is due
to three inescapable physical facts. First, all real observations take time and
involve discrete signals in finite numbers of detectors.1 Second, real apparatus is
constructed from atoms, not continua. Third, these atoms form finite numbers
of well-characterized modules, as discussed in the previous chapter.

Contextuality
A critically helpful fact here is that in any particular experiment, the contextual
subspaces that the observer needs to deal with will usually be of significantly
lower dimensions than that of the full quantum registers involved. For example,
while the register ground state 0n at stage Σn is an indispensable component
of the QDN formalism, there are few situations that will involve the completely
saturated labstate 2r − 1n ≡ A 2
1 A r
n n . . . An 0n . This will certainly be the case
when the rank r is large, say, of the order of hundreds or even possibly thousands.

Computerization
While QDN calculations can often be done by hand for small rank calculations
involving a small number of stages, the use of computer algebra (CA) software,
such as Maple, Mathlab, and Mathematica, allows problems with much greater
rank registers running over many more stages to be dealt with. All but the
simplest network calculations in this book were done using a computer algebra
program developed by us called Program MAIN, for example.

12.2 Program MAIN

In this chapter we discuss the application of CA to QDN. We illustrate this
approach to QDN by describing a typical experiment, referred to here as the
Wollaston interferometer (WI), shown in Figure 12.1.
The aim in this experiment is to investigate the polarization structure of an
unpolarized monochromatic beam of light. From source S, the beam is first passed
through a Wollaston prism W that splits it into two orthogonally polarized
components 11 and 21 as shown. Component 11 has “internal” polarization
represented by ket |s11  while component 21 has polarization state |s21 , orthogonal
to |s11 . Component 11 is then deflected by mirror M to 12 and then toward one
input channel of beam splitter B. Component 21 , meanwhile, is passed through
polarization rotator Rθ that rotates |s21  into cos(θ)|s22 +sin(θ)|s12 , before passing

1
Technically, the signals are not discrete or continuous. It is the observer who interprets
whatever they find at a detector as a discrete bit of information.
164 Computerization and Computer Algebra

M B

S W Rq
Figure 12.1. The Wollaston interferometer. A monochromatic, unpolarized
beam of light from source S is passed through Wollaston prism W . Emerging
polarized beam 11 is deflected by mirror M onto beam splitter B, while
emerging polarized beam 21 has its polarization rotated by angle θ through
module Rθ , before hitting B. Signal detection is at detectors 13 and 23 .

into the other input channel of B. The two beam splitter outcome channels are
monitored by final stage detectors 13 and 23 . The angle θ is chosen by the observer
before the preparation switch 10 is thrown.
If the transmission t and reflection r coefficients of the beam splitter B are
known, then the observed outcome frequencies at the final stage detectors will
depend on θ, and this gives information about the relative magnitudes of the two
polarizations in the initial beam.
For applying QDN to the WI and all other experiments we have a single CA
program, referred to here as MAIN . This program consists of two sections, A
and B. In Section A, we input the details of the particular experiment we want
to calculate, such as how many stages the experiment involves, the rank of each
stage quantum register, and so on.
In practice, MAIN carries full details of all the experiments we have investi-
gated, with a path parameter that allows us to select any experiment of current
interest. For example, the path parameter for WI is 44.
Given the path parameter, MAIN jumps to Section A, initializing the variables
from the relevant data there. Then MAIN jumps to Section B, automatically
calculating everything of interest, such as the outcome probabilities at each final
labstate detector.
The power of this approach is that all the work is done in encoding Section A.
In general, this does not take long and is straightforward, given a stage diagram
for the process and the module rules discussed in the previous chapters. Section B
is universal, that is, is the same for all experiments. This gives a very economical
and powerful approach indeed, allowing a great range of questions to be posed
and answered rapidly. The only real theoretical problem is in deciding the stage-
to-stage evolution operators Un+1,n in the encoding of Section A. This part of
 12.3 The Wollaston Interferometer 165

the process is currently where the computer does not help and the theorist has to
decide what goes in. However, it should be possible to automate this part as well,
if stage diagrams could be encoded first. We comment on this further, below.

12.3 The Wollaston Interferometer

We now describe the steps followed for a typical experiment, in this case the WI.
This example illustrates the fact that MAIN can deal with total states, that is,
entangled quantum states of the form

|Ψn ) ≡ |san  ⊗ ian , (12.1)
a

where |san 
is an “internal” SUO state, carrying information such as polarization,
and ian is a computational basis representation (CBR) labstate element carrying
information about the signal state of the detectors.

Encoding
Our QDN notation is transcribed into program MAIN as follows. Although the
observer’s focus is the signality of the signal detector amplitudes, the internal
calculations of MAIN are written in terms of the CBR. The reason is that
i ,
labstates of signality greater than one are nonlinear in the signal operators An
whereas the CBR is transparent to signality.
Our encoding takes the following form:

in → a[i, n], in → A[i, n],

|spn  → s[p, n], spn | → S[p, n]. (12.2)

Here the symbols a, A, s, and S were arbitrarily chosen, with the rule that
lowercase letters represent quantum “ket” states and capital letters represent
their duals.
The power of the CA approach comes in at this point in that we can use
differentiation to implement “inner product” rules. For example,
∂a[j, n] ∂s[q, n]
in j n → = δ ij , spn |sqn  → = δ pq , (12.3)
∂a[i, n] ∂s[p, n]
which is readily encoded in CA.2
In the following, the asterisk symbol ∗ denotes ordinary commutative multipli-
cation in the CA program. It is an important and helpful fact that our encoding
does not need any noncommutative product operation, although CA is capable
of handling those when required.

2
Differentiation is a good example of the dichotomy between the continuous and the discrete.
Mathematicians frequently deal with continuous variables, but mathematicians themselves
operate on discrete principles: all of mathematics is done line by line, symbol by symbol, in
a discrete way, very much like the way experimentalists have to operate in the real world.
166 Computerization and Computer Algebra

If rn is the rank of the stage-Σn qubit register and dn is the dimension of the
internal SUO Hilbert space, also at that stage, then total states will be of the
form
n −1 d
2r  n

|Ψn ) ≡ n |sn  ⊗ in
cp,i p

i=0 p=1
n −1 d
(12.4)
2r  n

→ Ψ[n] ≡ c[p, i, n] ∗ s[p, n] ∗ a[i, n],

i=0 p=1

while their duals are of the form

n −1 d
2r  n

(Φn | ≡ Dnp,i spn | ⊗ in

i=0 p=1
n −1 d
(12.5)
2r  n

→ Φ[n] ≡ D[p, i, n] ∗ S[p, n] ∗ A[i, n],

i=0 p=1

where the cp,i p,i

n and Dn are complex coefficients (readily handled in CA).
Inner products are evaluated by the CA prescription
n −1 d
2r  n
∂ 2 Φ[n] ∂ 2 Ψ[n]
(Φn |Ψn ) → ∗
i=0 p=1
∂S[p, n]∂A[i, n] ∂s[p, n]∂a[i, n]
n −1 d
(12.6)
2r  n

= D[p, i, n] ∗ c[p, i, n].

i=0 p=1

Draw a Stage Diagram

Having decided to model a particular experiment, we need to have a clear
overview of the experiment’s space-time architecture. That is best done graph-
ically. Any quantum experiment can be described diagrammatically by a stage
diagram, that is, a representation of the detectors and modules at each stage,
with different stages linked by amplitude transmission lines.
Currently, MAIN has no facility for automatically transcribing such diagrams
into code, but we envisage the development of graphical interface software that
would make this part of the programming straightforward. The reason for this
optimism is that there is only a finite variety of modules, and their properties
can be well specified and catalogued mathematically, as discussed in Chapter 11.
Figure 12.1 is the relevant diagram for the WI calculation.

Transcription Phase
Given the stage diagram, the next step is to read from it the following information
and encode it into MAIN:
1. N , the number of stages, not counting the initial stage, which is labeled
Σ0 . There is no limit to the size of N apart from the capacity limits of the
 12.3 The Wollaston Interferometer 167

computer. In our case, a personal computer with 32 gigabytes random access

memory and a 240 gigabyte hard disc was quite sufficient to deal with all the
experiments discussed in this book. For our WI calculation, N = 3.
2. r[0], r[1], . . . , r[N ], the rank at each stage. From Figure 12.1 we have r[0] = 1,
r[1] = r[2] = r[3] = 2.
3. d[0], d[1], . . . , d[N ], the dimensions of the “internal” SUO Hilbert space at
each stage. In the case of the WI, we have two orthonormal polarizations of
light, |s1n , |s2n , to factor in. These are often taken as horizontal and vertical
polarizations, but in principle could be circular polarizations if required. Hence
we take d[0] = d[1] = d[2] = 2.
4. The initial labstate: for the WI being discussed, we take |Ψ0 ) ≡ {c1 |s10  +
c2 |s20 } ⊗ 10 , and its dual (Ψ0 | ≡ {c1∗ s10 | + c2∗ s20 |} ⊗ 10 , where ci∗ is the
complex conjugate of ci . The coefficients ci characterize the two incident beam
polarizations and satisfy |c1 |2 + |c2 |2 = 1.
These states are encoded into MAIN as
|Ψ0 ) → ψ[0] ≡ (c[1] ∗ s[1, 0] + c[2] ∗ s[2, 0]) ∗ a[1, 0],
(Ψ0 | → Ψ[0] ≡ (C[1]) ∗ S[1, 0] + C[2] ∗ S[2, 0]) ∗ A[1, 0]. (12.7)
MAIN checks at each stage that normalization is preserved, so that at any
stage Σn , we have
(Ψn |Ψn ) = (Ψ0 |Ψ0 ) → C[1] ∗ c[1] + C[2] ∗ c[2]. (12.8)
5. From Figure 12.1 and the rules of the modules concerned (covered in the
previous chapter) we write (on paper)

Stage Σ0 → Σ1
 
 1 00 = |s1  ⊗ A
U1,0 |s10  ⊗ A  1 01 ,
0 1 1
 
 1 00 = |s2  ⊗ A
U1,0 |s20  ⊗ A  2 01 . (12.9)
0 1 1

Stage Σ1 → Σ2
 
 1 01 = |s1  ⊗ A
U2,1 |s11  ⊗ A  1 02 , (12.10)
1 2 2
 
 2 01 = (cos θ|s2  + sin θ|s1 ) ⊗ A
U2,1 |s21  ⊗ A  2 02 . (12.11)
1 2 2 2

Stage Σ2 → Σ3
   
 1 02 = |s1  ⊗ tA
U3,2 |s12  ⊗ A  2 03 + irA
 1 03 ,
2 2 3 3
   
 2 02 = |s1  ⊗ tA
U3,2 |s12  ⊗ A  1 03 + irA
 2 03 ,
2 2 3 3
   
 2 02 = |s2  ⊗ tA
U3,2 |s22  ⊗ A  1 03 + irA
 2 03 . (12.12)
2 2 3 3

Here t and r are beam splitter parameters, and we use (11.28).

168 Computerization and Computer Algebra

That is all the theoretical input needed. The next step is to transcribe it
into Section A code.
6. To transcribe the above information, we first use contextual completeness to
write down U1,0 , U2,1 , and so on, and then use the rules (12.2). For instance,
from (12.9) this gives
 1 01 s1 | ⊗ 01 A1 + |s2  ⊗ A
U1,0 = |s11  ⊗ A  2 01 s2 | ⊗ 01 A1
1 0 0 1 1 0 0
= |s11 s10 | ⊗ 11 10 + |s21 s20 | ⊗ 21 ⊗ 10
→ U [1, 0] ≡ s[1, 1] ∗ S[1, 0] ∗ a[1, 1] ∗ A[1, 0]+
s[2, 1] ∗ S[2, 0] ∗ a[2, 1] ∗ A[1, 0], (12.13)

and similarly for U2,1 and U3,2 .

This is all the input we need to provide program MAIN for it to answer all
possible maximal questions about this particular experiment.

Evaluation Phase
Once Section A has been initialized, Section B goes into action, employing several
key procedures (subroutines):
1. Given U [n + 1, n] and U [n + 2, n + 1], then the operator Un+2,n ≡ Un+1,n+1
Un+1,n is given by the rule Un+2,n → U [n + 2, n], where

−1 d[n+1]
2r[n+1]  ∂ 2 U [n + 2, n + 1] ∂ 2 U [n + 1, n]
U [n + 2, n] ≡ ∗ .
i=0 p=1
∂S[p, n + 1]∂A[i, n + 1] ∂s[p, n + 1]∂a[i, n + 1]
(12.14)
The overall contextual evolution operator U [N, 0] is evaluated by iterating
this process.
2. The retraction operator U N,0 is calculated from U [N, 0] by complex conjuga-
tion of coefficients and the interchange s ↔ S, a ↔ A. Complex conjugation
is readily handled in CA.
3. The detector POVM operators EN,0 i
→ E[i, N ] discussed in Chapter 9 are
calculated by the rule


d[N ]
∂ 2 U [N, 0] ∂ 2 U [N, 0]
E[i, N ] ≡ ∗ . (12.15)
p=1
∂S[p, N ]∂A[i, N ] ∂s[p, N ]∂a[i, N ]

These should satisfy the rule

−1
2r[N −1
]

d[0] 2r[0]
E[i, N ] = s[p, N ] ∗ S[p, N ] a[i, 0] ∗ A[i, 0], (12.16)
i=0 p=1 i=0

which is a representation of I0 , the contextual identity operator at stage Σ0 .

In MAIN, this is used as a useful check on the evolution operators U [n + 1, n]
written down in Section A.
 12.4 Going to the Large Scale 169

4. The outcome probabilities P (iN |Ψ0 ), i = 0, 1, 2, . . . , 2r[N ] − 1 are given by the

rule (9.59) and are readily evaluated in MAIN, again using differentiation in
the taking of inner products and operator action on state vectors. This gives
the answers to all the 2rN maximal questions. Because large-rank stages may
have excessive dimensions, the program gives a listing only of the nonzero
answers to the maximal questions. This listing will generally be significantly
smaller than the full potential set of answers, reflecting the fact that experi-
ments generally deal with contextual subspaces of quantum registers, not the
complete registers.
Points to note are the following.

Answers
Answers can be either in numerical form or in the far more useful algebraic form.
For example, the output listing for the Wollaston interferometer is
 1 03 |Ψ0 ) = |c1 |2 r2 + |c2 |2 t2 + i(c2∗ c1 − c1∗ c2 )rt sin θ,
P (A 3
 2 03 |Ψ0 ) = |c1 |2 t2 + |c2 |2 r2 − i(c2∗ c1 − c1∗ c2 )rt sin θ.
P (A (12.17)
3

These sum to unity, as expected. The implication here is that MAIN has found
 1A
the other two potential probabilities P (03 |Ψ0 ) and P (A 2
3 3 03 |Ψ0 ) to be zero.

Signal Decomposition
The CBR is excellent for CA but not so helpful with understanding signality.
To assist in the interpretation, MAIN converts CBR probabilities into a listing
of total signal probability for individual final-stage detectors, excluding any
that have probabilities of zero. This gives valuable insight into the patterns
of information flow. With some experience, it becomes clear why the photon
concept has become so popular: signality is conserved in many quantum optics
experiments unless some specific module such as a nonlinear crystal that creates
photons is used, as in parametric down-conversion.

Partial Questions
MAIN answers the full set of maximal questions. We have stated before that this
set of answers allows all partial questions to be answered. Therefore, CA gives
us the possibility of answering all possible questions in the context of QDN.
This becomes important in several complex experiments, such as the two-photon
interferometer of Horne, Shimony, and Zeilinger (Horne et al., 1989) and the
double-slit quantum eraser of Walborn et al. (Walborn et al., 2002), experiments
covered in detail in Chapter 14.

12.4 Going to the Large Scale

There is little doubt in our mind that the greatest challenge facing quantum
physics lies not with the reductionist laws of physics, which have been determined
170 Computerization and Computer Algebra

empirically to be excellent, but with the relatively unexplored laws of complexity

and emergence. Even the simplest of experiments, such as the double-slit (DS)
experiment discussed in Chapter 10, poses a serious challenge to QDN, simply
because the dimensions of quantum registers grow exponentially with rank. With
reference to Figure 10.2, suppose we wanted to model a DS experiment with K
detector sites on screen. Three possible approaches that we could take are the
following.

Approach 1: The Theoretical Approach

Here K is regarded as fixed but undetermined, and treated algebraically by the
theorist on paper.

Approach 2: The Brute Force CA Approach

Here K is given a specific value, such as K = 20 in a CA program such as MAIN,
and a relatively laborious calculation is done by a computer.

Approach 3: The Symbolic Approach

Here K is encoded symbolically and treated as an algebraic parameter char-
acterizing the experiment. CA does in principle have the potential to handle
this approach, but it requires more sophisticated programming, because that
would essentially be an attempt to simulate the way the human theorist handles
abstract modeling. A good example is that theorists can with training readily
discuss infinite-dimensional Hilbert spaces on paper, but CA cannot easily handle
infinities directly.

Approach 1 works for the relatively simple DS experiment but is impractical

for other large-rank networks, prompting the development of the CA approach.
Approach 2 has definite limitations. A CA simulation of the DS experiment
using MAIN was done with a range of values of K and timed. The results are
given in the following table:

Rank r Tr (seconds)
10 0.137
12 0.515
14 2.609
16 11.73
18 40.92
20 192.3
22 1284

Analysis of this table suggests that the time Tr for MAIN to complete a DS
simulation with a rank-r screen register is approximately of the form Tr = A +
B2r , where A and B are constants. It is clear that even a rank-100 simulation
could not be completed on this basis.
 12.5 Prospects 171

Approach 3 should be made to work if contextual subspaces are incorporated

adequately (which they are not, currently, in our program MAIN).

Three conclusions can be drawn from this analysis:

1. Computerization seems inevitable.
2. Computerization will be very useful in a broad range of problems, but some
problems may remain intractable.
3. Intelligent use of contextual subspaces will be necessary for problems where
the rank is too large for the brute force approach to work.

12.5 Prospects
Our current CA program, MAIN, requires the user to transcribe “by hand” the
information contained in a stage diagram into CA form. This process could be,
and perhaps should be, fully automated, opening the door to potentially vast
applications. Ideally, this would utilize the currently accessible technology of
touch screens, whereby the user used electronic graphics pens to sketch stage
diagrams in freehand. These would then be automatically transcribed into the
information currently fed in by hand in Section A of MAIN.
We envisage the development of user-friendly graphical software that would
allow the user to select all the necessary modules that were of interest, using
palettes of symbols representing real and virtual detectors, modules of all sorts,
and lines representing transmission through various media. With the right infor-
mation, it should be possible to encode different material properties, so that
quantized detector networks representing crystals, glasses, and other complex
systems such as neural networks and retinas could be dealt with in a systematic
and coherent way.
There is little doubt in our mind that such an approach to complex prob-
lems involving quantum mechanical effects in engineering, materials science, and
medical science will one day become commonplace.
 13
Interferometers

13.1 Introduction
The discussion of modules and computerization in previous chapters gives us the
means to consider empirically useful quantized detector networks (QDN). In this
chapter we shall focus on a particular class of network known as interferometers.
Interferometers play a crucial role in quantum mechanics (QM) because they
demonstrate the “paradox” of wave–particle duality in a direct way. On the one
hand, discrete signals are detected by the observer and those signals are usually
interpreted as particles or quanta. On the other hand, the observed frequencies
built up over many runs show effects interpreted as due to the interference
of waves.
In the previous chapter, we used computer algebra (CA) to discuss the Wol-
laston interferometer (WI), a relatively simple interferometer. We shall use the
same approach in the following discussion of the Mach–Zehnder interferometer
(MZI), apparatus that allows the observer to investigate optical transmission
through various materials.

13.2 The Mach–Zehnder Interferometer

The basic structure of an MZI is given in Figure 13.1. A source S of light sends
a monochromatic, unpolarized beam 10 into one input channel of beam splitter
B 1 . Output channel 11 is deflected by mirror M 1 onto a module labeled φ that
contains some medium under investigation, such as a crystal or liquid. The net
effect is to create a phase change in that deflected beam by an amount φ, and
then that modified beam, 12 , is passed into beam splitter B 2 . The second output
channel, 21 , from B 1 , meanwhile, is deflected by mirror M 2 into beam splitter
B 2 . The two deflected beams, 12 and 22 , that pass into B 2 interfere and are
finally monitored by detectors 13 and 23 .
Our analysis of the MZI follows the CA approach used for the WI in the
previous chapter.
 13.2 The Mach–Zehnder Interferometer 173

M f B

S B M
Figure 13.1. The Mach–Zehnder interferometer: S is a source of monochro-
matic, unpolarized light, B 1 and B 2 are beam splitters, M 1 and M 2 are
mirrors, and φ contains a medium that changes the phase of the light by φ.

Parameters
For this experiment, the number of stages N is N = 3. From Figure 13.1, the
rank at each stage is seen to be r0 = 1, r1 = r2 = r3 = 2. Since polarization is not
a factor in this version of the experiment, we take the dimensions of the internal
(photonic) space at the four stages to be given by d0 = d1 = d2 = d3 = 1.

The Initial State

Since we are not interested here with polarization, the initial total state |Ψ0 ) is
 1 00 = |s0  ⊗ 10 , where |s0  is a normalized photon state.
given by |Ψ0 ) ≡ |s0  ⊗ A 0
From Figure 13.1 and the known properties of the modules, we write
Stage Σ0 to Stage Σ1
 
 1 00 = t1 |s1  ⊗ A
U1,0 |s0  ⊗ A  2 01 + ir1 |s1  ⊗ A
 1 01
0 1 1

≡ |s1  ⊗ (t1 21 + ir1 11 ). (13.1)

Stage Σ1 to Stage Σ2
 
 1 01 = eiφ |s2  ⊗ A
U2,1 |s1  ⊗ A  1 02 ≡ eiφ |s2  ⊗ 12 ,
1 2
 
 2 01 = |s2  ⊗ A
U2,1 |s1  ⊗ A  2 02 ≡ |s2  ⊗ 22 . (13.2)
1 2

Stage Σ2 to Stage Σ3
 
 1 02 = t2 |s3  ⊗ A
U3,2 |s2  ⊗ A  2 03 + ir2 |s3  ⊗ A
 1 00 ≡ |s3  ⊗ (t2 23 + ir2 13 ),
2 3 3
 
 2 02 = ir2 |s3  ⊗ A
U3,2 |s2  ⊗ A  2 03 + t|s3  ⊗ A 1 00 ≡ |s3  ⊗ (ir2 23 + t13 ).
2 3 3
(13.3)

Here we have ignored any phase changes at the mirrors and characterized each
beam splitter separately. In the above, all superscripts are labels, not powers.
174 Interferometers

This is all the input information required for our CA program MAIN, as
discussed in the previous chapter. The nonzero conditional outcome probabilities
are found to be
 1 03 |Ψ0 ) = −2r1 r2 t1 t2 cos(φ) + (r1 r2 )2 + (t1 t2 )2 ,
Pr(A 3
 2 03 |Ψ0 ) = +2r1 r2 t1 t2 cos(φ) + (r1 t2 )2 + (t1 r2 )2 .
Pr(A (13.4)
3

These sum up to unity as required, given that (ti )2 + (ri )2 = 1, for i = 1, 2.

The significance here is that the outcome probabilities are affected by the
phase change module. By altering the path length in that module, and other
parameters such as temperature and density of the medium in that module,
significant information about that medium can be extracted.

13.3 Brandt’s Network

The next example is a quantum optics network discussed by Brandt (Brandt,
1999) in terms of conventional positive operator-valued measure operators
(POVMs) and shown in Figure 13.2. A source S prepares a monochromatic
unpolarized beam of light 10 that is split by Wollaston prism W into two
orthogonally polarized components 11 and 21 . One component 11 is then passed
into beam splitter B 1 and thereby split into two components 12 and 22 with no
change in polarization. Component 12 is subsequently observed at detector 13 ,
while component 22 is passed into beam splitter B 2 . Meanwhile, component 21
emerging from the Wollaston prism W has its polarization turned by π/2 at
module R. The resulting beam 32 is then passed into beam splitter B 2 , where it
interferes with 22 , with subsequent detection at detectors 23 and 33 .

B B

S W R
Figure 13.2. Brandt’s network: source S prepares an unpolarized, monochro-
matic beam of light that passes through Wollaston prism W . This splits the
beam into two orthogonally polarized components. These are passed through
beam splitters B 1 and B 2 as shown. Module R rotates the polarization of one of
the polarized beams into that of the other, prior to it being passed through B 2 .
 13.3 Brandt’s Network 175

Brandt’s analysis was in terms of nonorthogonal system under observation

(SUO) state vectors. Our analysis avoids nonorthogonality issues directly. The
initial state is given by
 1 00 = (α|s1  + β|s2 ) ⊗ 10 ,
|Ψ0 ) ≡ (α|s10  + β|s20 ) ⊗ A (13.5)
0 0 0

where |s10 , |s20 

denote orthogonal photon polarization states, and α and β are
complex coefficients satisfying |α|2 + |β|2 = 1. By inspection of Figure 13.2 we
take N = 3, r0 = 1, r1 = 2, r2 = r3 = 3, d0 = d1 = d2 = d3 = 2. The dynamics
is given by the following rules.

Stage Σ0 to Stage Σ1
 
 1 00 ≡ |s1  ⊗ A
U1,0 |s10  ⊗ A  1 01 = |s1  ⊗ 11 ,
0 1 1 1
 
U1,0 |s0  ⊗ A
2  00 ≡ |s  ⊗ A
1 2  01 = |s2  ⊗ 21 .
2
(13.6)
0 1 1 1

Stage Σ1 to Stage Σ2
   
 1 01 ≡ t1 |s1  ⊗ A
U2,1 |s11  ⊗ A  1 01 + ir1 |s1  ⊗ A
 2 02 = |s1  ⊗ t1 11 + ir1 22 ,
1 2 2 2 2 2
 
 2 01 ≡ −|s1  ⊗ A
U2,1 |s21  ⊗ A  3 02 = −|s1  ⊗ 42 , (13.7)
1 2 2 2

where (t1 )2 + (r1 )2 = 1. The second equation in (13.7) represents a rotation of

the photon polarization vector |s21  by − 12 π into −|s12  as it passes through the
module labeled R in Figure 13.2 (the sign change follows the convention used in
Brandt (1999)).

Stage Σ2 to Stage Σ3
 
 2 02 ≡ |s1  ⊗ A
U3,2 |s12  ⊗ A  2 03 = |s1  ⊗ 13 ,
2 3 3 3
 
 2 02 ≡ t2 |s1  ⊗ A
U3,2 |s12  ⊗ A  3 03 + ir2 |s1  ⊗ A
 2 03
2 3 3 3 3

 = t |s3  ⊗ 43 + ir |s3  ⊗ 23 ,
2 1 2 1

U3,2 |s12  ⊗  3 02
A  03 + t2 |s1  ⊗ A
≡ ir |s3  ⊗ A
2 1 3  2 03
2 3 3 3

= ir2 |s13  ⊗ 43 + t2 |s13  ⊗ 23 , (13.8)

where (t2 )2 + (r2 )2 = 1.
With this information transcribed into Section A of the CA program MAIN,
we find the following nonzero outcome probabilities for the Brandt network:
 1 03 |Ψ0 ) = |α|2 (t1 )2 ,
Pr(A 3
 2 03 |Ψ0 ) = |α|2 (r1 r2 )2 − (α∗ β + αβ ∗ )r1 r2 t2 + |β|2 (t2 )2 ,
Pr(A 3
 3 03 |Ψ0 ) = |α|2 (r1 t2 )2 + (α∗ β + αβ ∗ )r1 r2 t2 + |β|2 (r2 )2 ,
Pr(A (13.9)
3

assuming perfect efficiency and wave-train overlap. When the reflection and
transmission coefficients are chosen as by Brandt (Brandt, 1999), these rates
agree with his precisely.
176 Interferometers

f
S C B B

Figure 13.3. The HSZ two-photon interferometer.

13.4 The Two-Photon Interferometer

Up to now, we have restricted attention to signality-one processes. It is important
to test our QDN approach on more subtle situations. With this in mind, in this
section we discuss the application of QDN to a signality-two process discussed
by Horne, Shimony, and Zeilinger (HSZ) (Horne et al., 1989). The relevant stage
diagram is given in Figure 13.3.
A source S creates a preparation switch 10 that by stage Σ1 has generated a
two-photon entangled state via module C. As discussed in Horne et al. (1989),
module C creates four separate components that are processed as follows. Com-
ponent 11 passes through phase changer φ1 and then enters beam splitter B 1 ,
while component 21 passes through phase changer φ2 and then enters beam
splitter B 2 . Components 31 and 41 pass directly on to beam splitters B 2 and
B 1 , respectively, as shown. The observer monitors signals at stage Σ3 as shown.
The aim of the experiment is to investigate photon signal pair correlations in the
detectors.

Parameters
We take N = 3, d[0] = d[1] = d[2] = d[3] = 1, r[0] = 1, r[1] = r[2] = r[3] = 4.

Initial State
The preparation switch is given by
 1 00 .
|Ψ, 0) = |s10  ⊗ A (13.10)
0

Evolution
Stage Σ0 to Σ1
   
 1 00 ≡ √1 |s1  ⊗ A
U1,0 |s10  ⊗ A  3 01 + A
 1A  4 01 = √1 |s1  ⊗ {51 + 10 } .
 2A
0 1 1 1 1 1 1 1
2 2
(13.11)
 13.4 The Two-Photon Interferometer 177

Stage Σ1 to Σ2
 
U2,1 |s11  ⊗ A 3
 1A iφ1 1 3
 1A iφ1 1
1 1 01 ≡ e |s2  ⊗ A 2 2 02 = e |s2  ⊗ 52 , (13.12)
 
U2,1 |s11  ⊗ A 4
 2A 2
4
 2A 2
1 1 01 ≡ e |s12  ⊗ A |s12  ⊗ 102 .
iφ iφ
2 2 02 = e (13.13)

Stage Σ2 to Σ3
    
U3,2 |s12  ⊗ A 3
 1A 1 4 1 3  1 + ir2 A
2
2 2 02 = |s3  ⊗ t A3 + ir A3 t2 A
1
3 3
 
= |s13  ⊗ t1 t2 93 + ir1 t2 53 + ir2 t1 103 − r1 r2 63 ,
    
U3,2 |s12  ⊗ A 4
 2A 2 2 2 1  3 + ir1 A
4
2 2 02 = |s3  ⊗ t A3 + ir A3 t1 A
1
3 3
 
= |s13  ⊗ t1 t2 63 + ir2 t1 53 + ir1 t2 103 − r1 r2 93 .
(13.14)

With this information, program MAIN gives the following nonzero correlation
probabilities:

 1A
3 2 4 1 1 2 2 1 2 1 2
Pr(A 3 3 03 |Ψ0 ) = Pr(A3 A3 03 |Ψ0 ) = (r t ) + (r t ) + r r t t cos(φ − φ ),
1 2 1 2 1 2
2 2
3
 2A 1 4 1 1 2 2 1 2 1 2
Pr(A 3 3 03 |Ψ0 ) = Pr(A3 A3 03 |Ψ0 ) = (r r ) + (t t ) − r r t t cos(φ − φ ).
1 2 1 2 1 2
2 2
(13.15)

These agree with the results given

√ in Horne et al. (1989), assuming no losses and
1 1 2 2
taking t = r = t = r = 1/ 2.

The following comments are relevant.

Partial Questions
The above signal pair correlations show dependency on the phases φ1 , φ2 , whereas
as pointed out by HSZ, the answers to the partial questions involving single
detectors only have no such dependence. For instance, the probability Pr(13 |Ψ0 )
that stage Σ3 detector 13 has fired is given by

3
 1A 1 4 1
Pr(13 |Ψ0 ) = Pr(A 3 3 03 |Ψ0 ) + Pr(A3 A3 03 |Ψ0 ) = , (13.16)
2
which is independent of the phases. The same result holds for the other single
detectors.

Nonlocality
The QDN formalism shows that the phase change φ1 applied to component 11
could be applied to component 31 instead with no change in physical predictions;
this is evident from the fact that the phase factor exp(iφ1 ) multiplies the product
 1A
A 3 2
2 2 in Eq. (13.12). A similar remark applies to the other phase change φ . Since
the component beams are spatially separated when this phase change is applied,
this means that quantum states are inherently nonlocal in character.
178 Interferometers

From such considerations, we conclude that physical space is not as simple as

it seems from a classical perspective. Perhaps a better way of saying this is that
the classical model of physical space as a three-dimensional continuum of spatial
position parameters is a good classical model but contextually incomplete as far
as quantum mechanics is concerned.

Probabilities versus Rates

As with all calculations in QDN, normalization to unity is an idealization that
avoids empirically significant but theoretically marginal considerations to do with
flux factors, particle production rates, and such like. Perhaps the best way to
deal with these issues is to recognize that what is calculated represents idealized
situations. The more empirically useful interpretation of stated probabilities is
that they are best case rates, that is, predicted relative average signal rates during
the time when incoming wave trains are long enough to intersect and interfere,
with no inefficiencies or extraneous losses in detection.
 14
Quantum Eraser Experiments

14.1 Introduction
In recent years, numerous quantum optics experiments have demonstrated novel
quantum effects based on quantum amplitude superposition. Some of these
empirical observations raise disturbing questions about the nature of reality,
particularly concerning quantum nonlocality. In this chapter we discuss some
experiments that suggest (to us, at least) that the information void, that
uncharted regime between signal preparation and outcome detection, is not well
modeled by classical spacetime. Something strange seems to be lurking there.
For the purposes of exposition, the sequence of experiments we discuss in this
chapter does not follow the historical sequence of those experiments, We discuss
first the delayed-choice quantum eraser experiment of Kim et al. (Kim et al.,
2000), referred to here as KIM.1 This leads us to define a heuristic measure
of which-path information that is central to the theme of this chapter. This
leads to a discussion of Wheeler’s thought experiments on the conundrum of
delayed choice in physics, particularly evident in the phenomenon of galactic
lensing (Wheeler, 1983), and its empirical implementation, the delayed-choice
interferometer experiment of Jacques et al., referred to as JACQUES. Our final
experiment in this sequence is the double-slit quantum eraser of Walborn et al.
(Walborn et al., 2002), referred to as WALBORN. That experiment gives a
significant challenge for quantized detector networks (QDN) to reproduce its
empirical results.
Such experiments have led to suggestions that interference patterns formed
by particles impacting on a screen may be influenced in some way by decisions
made long after those particles had landed on that screen. Our objective in
this chapter is to show by a detailed QDN analysis of those experiments that

1
With profound apologies to co-authors, we will employ the convention that several
experiments are referred to by the capitalized surname of the first named author.
180 Quantum Eraser Experiments

quantum principles do not support these suggestions. Nevertheless, something

bizarre seems to be there.
There are two complementary aspects in these experiments that have deeply
worried physicists. The first involves spatial nonlocality and is the point of
Wheeler’s concerns. Historically, it has been a great source of debate among
physicists. It motivated Einstein to write to Born thus:2

So clearly that you consider my attitude towards statistical quantum

mechanics to be strange and archaic . . .
. . . I cannot make a case for my attitude in physics which you would
consider at all reasonable. I admit, of course, that there is a considerable
amount of validity in the statistical approach which you were the first to
recognise clearly as necessary given the framework of the existing formalism.
I cannot seriously believe in it because the theory cannot be reconciled with
the idea that physics should represent a reality in time and space, free from
spooky action at a distance.
(Born–Einstein Letters, 1971, p. 158)

The second aspect involves temporal nonlocality: the temporal sequence of

actions taken in the laboratory at different places seems to be irrelevant in certain
experiments. We regard that as a vindication of the stages concept. The empirical
evidence from experiments such as WALBORN, who specifically investigated
the temporal aspects, is that the passage of “detector time” is synonymous with
quantum information acquisition (QIA) occurring in a sequence of stages (Eakins
and Jaroszkiewicz, 2005). Stages have rules that are different in some aspects
from those of classical information acquisition, and this accounts for some of the
apparent strangeness of quantum mechanics.
The rules governing QIA are those of quantum mechanics (QM) and indeed
conform with all known physics. For example, QIA never violates the light-cone
constraint of relativity that classical information cannot be acquired between
space-like intervals. Underpinning this is the so-called no-communication theorem
in QM, discussed in Section 16.7. This states that the actions taken by an observer
on a substate of a total state of a system under observation (SUO) cannot be
detected by another observer of another substate of that total state.
One way of understanding this is to note that quantum correlations appearing
to violate the principle of Einstein locality always require observations to be
completed before those correlations can be defined by observers, and this comple-
tion necessarily always takes place in a classically consistent matter. Currently,

2
The specific English translation of what Einstein wrote here has been disputed: mysterious
rather than spooky has been suggested by some authorities as closer to what Einstein
intended to say.
 14.2 Delayed-Choice Quantum Eraser 181

there is no empirical evidence for, or theoretical necessity of, the contextually

incomplete notion that information can flow backward in time, as suggested by
Cramer’s transactional interpretation of QM (Cramer, 1986), or for the existence
of closed time-like curves (CTCs), as found in some relativistic spacetimes such
as that of Gödel (Gödel, 1949).
On the other hand, quantum information can be shielded against the effects
of decoherence and preserved in a state of stasis for arbitrarily long periods of
laboratory time. QDN uses the concept of null test to encode this phenomenon,
as demonstrated in our discussion of particle decay experiments in Chapter 15. In
the delayed-choice experiments discussed in the present chapter, suitably shielded
observations involving separate detectors can be taken in apparently random
order relative to laboratory time without affecting correlations.

14.2 Delayed-Choice Quantum Eraser

We turn first to the delayed-choice quantum eraser . The architecture of an
experiment proposed in Kim et al. (2000), and referred to here as KIM, is shown
in Figure 14.1. In that stage diagram, we leave out dotted lines indicating stages
and use subscripts for this purpose. In the following description, we use the term
photon for convenience only.
A source S produces an initial total state 10 that is a superposition of two
coherent photon pairs. By stage Σ1 these are split by suitable apparatus into
four components 11 , 21 , 31 , and 41 as shown. Components 11 and 31 are from
different initial pairs and are passed onto a detecting screen DS, consisting of
detectors 52 , 62 , . . . , K2 , where it is supposed that they interfere, just as in the
original double-slit experiment. Our analysis will show that the situation is more
subtle than that.

B S
K K

Figure 14.1. KIM, a proposed delayed-choice quantum eraser experiment (Kim

et al., 2000).
182 Quantum Eraser Experiments

Meanwhile, components 21 and 41 are each passed through beam splitters

B , B 2 as shown. The transmitted components from these beam splitters are
1

sent on to detectors 13 and 43 as shown, while the reflected components 22 and

32 are passed through beam splitter B 3 . The outcome channels of this beam
splitter are monitored by detectors 23 and 33 .
Analysis of this arrangement suggests that choices made by the experimentalist
at B 3 can influence the signal patterns seen on DS, even though the signals in
that screen had been captured earlier at stage Σ2 . Kim et al.’s argument is that
correlations between signals in 13 and DS or between signals in 43 and DS give
which-path information, whereas correlations between signals in either 23 and
DS, and 33 and DS do not. Therefore, there should be no interference terms on
DS when there are signals in 13 or 44 , whereas interference terms are expected
when 13 or 33 register a signal.
We proceed with our stage analysis as follows. From Figure 14.1 we read off
the following parameters for our CA program MAIN: N = 3, r0 = 1, r1 = 4,
r2 = r3 = K, d0 = d1 = d2 = d3 = 1, where K is chosen to be large enough so as
to show what happens on the screen DS. In MAIN a value K = 6 was sufficient
to show all the essential features of the experiment.
As with the double-slit experiment discussed in Chapter 10, we represent our
 1 00 .
initial preparation switch at stage Σ0 by the total state |Ψ0 ) = |s0  ⊗ A 0

Stage Σ0 to Stage Σ1
The jump to stage Σ1 is defined by the production of two correlated photon
pairs. Photon polarization is ignored here, but the formalism can readily deal
with any situation where this is not the case. Therefore, MAIN assumes a one-
dimensional internal Hilbert space in this experiment with normalized basis |sn 
for the internal degrees of freedom at any given stage Σn .
The semi-unitary transformation from Σ0 → Σ1 is given by
   
U1,0 |s0  ⊗ A 1 00 = |s1  ⊗ αA 2
 1A 3 4
0 1 1 + β A1 A1 01 , (14.1)

where |α|2 + |β|2 = 1. As we stated above we assume that photon spin is not
relevant to the issues being explored here.

Stage Σ1 to Stage Σ2
For this jump we have
  
K
U2,1 |s1  ⊗ A  2 01 = |s2  ⊗
 1A  i {t1 A
C iA  1 + ir1 A
 2 }02 ,
1 1 2 2 2
i=5

  
K
U2,1  3 4
|s1  ⊗ A1 A1 01 = |s2  ⊗  i {t2 A
Di A  4 + ir2 A
 3 }02 . (14.2)
2 2 2
i=5

Here, one component beam from each pair is focused onto the detecting screen
DS, while the other component is channeled onto either beam-splitter B 1 or B 2 ,
 14.2 Delayed-Choice Quantum Eraser 183

as shown. The {C i } coefficients represent the amplitudes for landing on the screen
DS from 11 , while the {Di } are from 31 . The coefficients ti , ri are characteristic
transmission and reflection parameters associated with beam splitter B i , i =
1, 2, 3, and satisfy the rule (ti )2√
+(ri )2 = 1. It is useful not to set these parameters
to the conventional value 1/ 2 but to keep them open and available to be
changed. It is in these parameters that we encode the observer’s freedom of
choice in this particular experiment.

Stage Σ2 to Stage Σ3
The final transition from stage Σ2 to Σ3 involves four terms, each of which
involves null tests in one way or another. Most significantly for this discussion, it
is supposed that the DS screen detectors have registered their signals irreversibly
by stage Σ2 . Therefore, this information is carried to stage Σ3 by null tests, shown
by the horizontal lines on the right-hand side of Figure 14.1.
We write, for i = 5, 6, . . . , K:
 
U3,2 |s2  ⊗ A i A 1 02 = |s3  ⊗ A 1
i A
2 2 3 3 03 ,
 
U3,2 |s2  ⊗ A i A2 i 3 3 3 2
2 2 02 = |s3  ⊗ A3 {t A3 + ir A3 }03
 
U3,2 |s2  ⊗ A i A3 i 3 2 3 3
2 2 02 = |s3  ⊗ A3 {t A3 + ir A3 }03 (14.3)
 
U3,2 |s2  ⊗ A i A4 i 4
2 2 02 = |s3  ⊗ A3 A3 03 .

This is all the information needed for our CA program MAIN to answer all
possible maximal questions. MAIN gives four sets of nonzero maximal question
answers, equivalent to two-site correlations:
 1A
Pr(A i
3 3 03 |Ψ0 ) = |t C α| ,
1 i 2

 2A
Pr(A i
3 3 03 |Ψ0 ) = |r r C α − ir t D β| ,
1 3 i 2 3 i 2

 3A
Pr(A i
3 3 03 |Ψ0 ) = |r t C α + ir r D β| ,
1 3 i 2 3 i 2

 A
Pr(A 4  03 |Ψ0 ) = |t D β| ,
i 2 i 2
i = 5, 6, . . . , K. (14.4)
3 3

These give the total sum


4 
K 
K 
K
i A
Pr(A j
3 3 03 |Ψ0 ) = |α| |C i |2 + |β|2 |Di |2 ,
2
(14.5)
i=1 j=5 j=5 j=5

which is consistent with probability conservation if we take the semi-unitarity

K K
j=5 |C | = j=5 |D | = 1 into account. Note also that semi-
i 2 i 2
conditions
K
unitarity requires j=5 C i∗ Di = 0, where C i∗ is the complex conjugate of C i .
There are several observations to be made about these results.
1. The only genuine interference found in our analysis occurs in the two-signal
correlation probabilities Pr(A i
 2A 3 i
3 3 03 |Ψ0 ) and Pr(A3 A3 03 |Ψ0 ).
i i i
2. The parameters t , r for beam splitter B represent places in the apparatus
where the experimentalist could make changes, either before or after signals
184 Quantum Eraser Experiments

have been registered on the screen DS during any given run of the experiment.
In other words, choices can be made at B 1 , B 2 , and B 3 that affect various
incidence rates. The question is, does any change made by the experimentalist
at any beam splitter affect anything that has been measured before that
change was made? In particular, can any change in B 3 affect what has already
happened on the screen?
By inspection of (14.3), we see that no change in t3 or r3 , subject to (t3 )2 +
(r ) = 1, has any effect whatsoever on Pr(A i
 1A 4 i
3 3 03 |Ψ0 ) or Pr(A3 A3 03 |Ψ0 ).
3 2

These coincidence rates actually involve signal detection completed during

earlier stages. The conclusion therefore is that any suggestion that delayed
choice can erase information irreversibly acquired in the past is incorrect and
misleading.
3. By inspection, it is true that changes in t3 and r3 affect Pr(A i
 2A
3 3 03 |Ψ0 ) and
Pr(A i
 3A
3 3 03 |Ψ0 ). However no acausality is involved, because a coincidence rate
is undefined until signals from both detectors involved have been counted.
Pr(A i
 2A 3 i
3 3 03 |Ψ0 ) and Pr(A3 A3 03 |Ψ0 ) cannot be measured until after the choice
3 3
of t and r has been made.
Suggestions that events in stage Σ3 could influence events in earlier stages
do not take into account the crucial role of post-selection 3 in such experiments.
The proper way to understand what is happening is to view the role of the
four detectors 13 , 23 , 33 , and 43 as post-selection apparatus for the processing
of data already accumulated on the screen DS.
4. Program MAIN allows for partial questions to be answered. If we look at the
individual total counting rates at each of the four detectors i3 , i = 1, 2, 3, 4,
we find

K
 1 03 |Ψ0 ) ≡
Pr(A  1A
Pr(A i
3 3 03 |Ψ0 ) = |t α| ,
1 2
3
i=5

K
 2 03 |Ψ0 ) ≡
Pr(A  2A
Pr(A i
3 3 03 |Ψ0 ) = |r r α| + |t r β| ,
1 3 2 3 2 2
3
i=5

K
 3 03 |Ψ0 ) ≡
Pr(A  3A
Pr(A i
3 3 03 |Ψ0 ) = |r t α| + |r r β| ,
1 3 2 2 3 2
3
i=5

K
 4 03 |Ψ0 ) ≡
Pr(A  4A
Pr(A i
3 3 03 |Ψ0 ) = |t2 β| ,
2
3 (14.6)
i=5

using the semi-unitarity of the {C i } and {Di } coefficients. Again, changes

made at B 3 would have no effect on Pr(A  1 03 |Ψ0 ) or Pr(A
 4 03 |Ψ0 ) but would
3 3

3
In ordinary usage, the term post-selection means selection occurring after some given
process. In probability theory, given an event A, then the probability of some event B
conditional on that given is a conditional probability, denoted P (B|A). Either way, the
concern is with what happens after something has been done or occurred.
 14.3 Which-Path Measure 185

affect Pr(A  3 03 |Ψ0 ). These probabilities sum to unity as

 2 03 |Ψ0 ) and Pr(A
3 3
expected.
5. If we look at the individual relative probabilities for a given detector i3 , i = 5,
6, . . . , K, on the screen DS, we find

4
 i 03 |Ψ0 ) ≡ Pr
Pr(A j A
Pr(A i
3 3 03 |Ψ0 ) = |αC | + |βD | ,
i 2 i 2
3 (14.7)
j=1

which shows no interference on DS. No change at any of the beam splitters

affects the single detector rates observed on the screen DS. This is essentially
the same phenomenon, discussed in Section 13.4, found in the two-photon
interferometer experiment of Horne, Shimony, and Zeilinger (Horne et al.,
1989).
We should ask: Given that signals on the detector screen DS came from the
double slits, why is there no interference on that screen? Surely there should
be such interference.
The answer is deep. This experiment is not just a double-slit experiment.
The mere fact that a photon has gone off toward the beam splitters B 1 or
B 2 and the observer could in principle find out which beam splitter it was is
enough to provide which-path information. It is that contextual information
that destroys any possibility of an interference pattern on DS. In fact, the
beam splitters are not needed for this. It is enough to have 21 and 41 as iden-
tifiable potential sites for signal detection to destroy any chance of interference
on DS.

14.3 Which-Path Measure

The double-slit and delayed-choice eraser experiments discussed up to this point
in this book belong to an important class of experiment that, to use colloquial
terminology, provide partial or complete information about which path a photon
had taken in its journey from initial to final stages. We shall discuss in the next
section another example, Wheeler’s delayed choice experiment.
Each of these experiments carries with it contextual attributes arising from
the experimental setup that determine the extent to which photon paths can be
determined from the data or not. For example, the double-slit experiment with
both slits open gives zero information about which slit a particular detected
photon came from. On the other hand, the same setup with one of the slits
blocked up gives total information as to where any of the detected photons
originated.
It is of interest therefore to find some measure or parameter Φ that is char-
acteristic of any given experimental setup and that gives us an indication as to
how much which-path information we could extract from the experiment. There
is no precise theory for this known to us at present, so in the absence of any
deeper analysis, our choice is to define Φ heuristically as the total probability of
186 Quantum Eraser Experiments

determining with certainty full path information from a single detected “indi-
cator” photon, that is, from a single signal in a specific set of detectors. We
define

ΦDS ≡ P rob(full path information|any single indicator detector fires). (14.8)

For example, in the case of the double-slit experiment discussed in Chapter 10,
the detecting screen contains the indicator detectors. Then with both slits open,
we find ΦDS = 0. When one of the slits is open, then ΦDS = 1. In the case of
the delayed choice quantum eraser discussed above, the indicator detectors are
13 and 43 , so we define

ΦDC ≡ Pr(A  1 03 |Ψ0 ) + Pr(A

 4 03 |Ψ0 ) = |t1 α|2 + |t2 β|2 .
3 3
√
In the conventional symmetric situation, |α| = |β| = t1 = t2 1/ 2, giving ΦDC =
1/2, as we should expect. When t1 = t2 = 0, transmission to 13 or 43 cannot
occur, so there is normally interference for sure, giving ΦDC = 0, meaning no
path information can be established.
There is a pathology in this last case, because t1 = t2 = 0 gives Pr(A  2 03 |Ψ0 ) =
3
3 2 
|r α| + |t β| and Pr(A3 03 |Ψ0 ) = |t α| + |r β| . If now, in addition to setting
3 2 3 3 2 3 2

t1 = t2 = 0, the experimentalist had set r3 = 0, then a single photon would be

detected at 23 or 33 and it would clear which path had been taken by the photons.
But overall, this is equivalent to having no beam splitters, so this scenario is of
no value here.

14.4 Wheeler’s Double-Slit Delayed Choice Experiment

The physicist John Wheeler gave a theoretical discussion of a quantum interfer-
ence experiment that has stimulated a great deal of puzzlement and controversy
concerning the nature of reality (Wheeler, 1983). It is our considered judgment
that his concern reflects the inadequacy of our classical views about reality.
There are two forms of his experiment that are commonly discussed: one is a
modified Mach–Zehnder interferometer (MZI) and the other is a modified double-
slit (DS) experiment. We shall refer to these as WHEELER-1 and WHEELER-2,
respectively.

WHEELER-1
Figure 13.1 is relevant to WHEELER-1. Specifically, WHEELER-1 is concerned
with stage-Σ1 nodes 11 and 21 and their relationship to stage-Σ2 nodes 12 and 22 .
Suppose the laboratory time between stage Σ1 and stage Σ2 is significantly long
enough for anything done to one of the beams, at say module φ, to be long after
the action of beam splitter B 1 in any reasonable sense of the word. Then there
appears to be a clash of particle and wave concepts.
To be specific, consider a pulse of light 10 coming from a distant galaxy billions
of years ago at stage Σ0 being split by gravitational lensing (equivalent to B 1 )
 14.4 Wheeler’s Double-Slit Delayed Choice Experiment 187

into components 11 and 21 . These two components follow very different paths
across the Universe until an observer here on Earth observes stage-Σ3 signals at
detectors 13 or 23 . Now two kinds of process could affect outcome probabilities at
those detectors: (1) there could be some galactic cloud affecting, say, the optical
path from 11 to 12 , equivalent to module φ, and (2) the observer could choose to
do something significant at B 2 so that either an interference pattern is built up
or no such pattern is built up.
The conventional picture here is of a particle known as a photon traveling
through space. Intuitively, we would like to think of such objects. But it seems
incredible that the presence of some disruptive process φ on the potential path
11 → 12 could affect a photon traveling from 21 to 22 . Yet that is just what the
photon paradigm requires us to contemplate, particularly if the observer decides
to arrange the apparatus at B 2 to detect an interference pattern at the detectors
13 and 23 , or not.
Of this scenario, Wheeler wrote:

We get up in the morning and spend the day in meditation whether

to observe by “which route” or to observe interference between “both
routes.” When night comes and the telescope is at last usable we leave
the half-silvered mirror out or put it in, according to our choice. The
monochromatizing filter placed over the telescope makes the counting rate
low. We may have to wait an hour for the first photon. When it triggers a
counter, we discover “by which route” it came with the one arrangement;
or by the other, what the relative phase is of the waves associated with the
passage of the photon from source to receptor “by both routes” – perhaps
50,000 light years apart as they pass the lensing galaxy. But the photon has
already passed that galaxy billions of years before we made our decision.
This is the sense in which, in a loose way of speaking, what the photon
shall have done after it has already done it. In actuality it is wrong to
talk about the “route” of the photon. For a proper way of speaking we
recall once more that it makes no sense to talk of the phenomenon until
it has been brought to a close by an irreversible act of amplification. “No
elementary phenomenon is a phenomenon until it is a registered (observed)
phenomenon.” (Wheeler, 1983)

The problem as we see it lies not with the quantum physics but with mental
imagery associated with the particle-wave concept. There is, for instance, a clas-
sically conditioned vacuous assertion glaringly obvious in Wheeler’s statement
above: that “the photon has already passed that galaxy billions of years before
we made our decision.” Proof? There can be none, but of course, we are all
strongly conditioned to believe this counterfactual, vacuous statement and many
like it. According to quantum physics principles, as we see them in this book, we
have no direct empirical entitlement to make that assertion. In quantum physics,
188 Quantum Eraser Experiments

we have to resist the constant temptation to think classically in those places

where such thinking is unwarranted.

WHEELER-2
Wheeler’s delayed choice experiment can be discussed as a double-slit experiment
with a modified screen. Some of the detectors on the screen can receive quantum
signals from both slits, while the others can receive a signal from only one of
the slits. The interest in this arrangement comes from the possibility that the
observer can decide in principle which detectors receive which signal(s) long after
light has left the two slits.
An idealized version of WHEELER-2 is shown in Figure 14.2. The details are
much the same as the double-slit experiment studied in Chapter 10, but with the
difference that now there are three groups of detectors on the screen. Detectors
12 to R2 can receive a quantum amplitude from 11 only, (R + 1)2 to (R + S)2
can receive quantum amplitudes from 11 and from 21 , and (R + S + 1)2 to K2
can receive an amplitude from 21 only. Here, K = R + S + T .
Following Wheeler, we imagine that the experimentalist can shuffle the values
of R, S, and T during any given run, after they are sure that light has left the two
slits 11 and 22 and before any impact on the final detecting screen at stage Σ2 .
Any actual experiment would require analysis of the final data, post-selecting
signals corresponding to equivalent values of R, S, and T .

S
R S

R S

Figure 14.2. Idealization of Wheeler’s delayed-choice experiment, WHEELER-2.

 14.5 The Delayed Choice Interferometer 189

For this discussion, we do not need CA assistance. Photon polarization is not

an issue in this particular instance either, so the initial labstate |Ψ0 ) is given by
 1 00 . The stage dynamics goes as follows.
|Ψ0 ) ≡ |s0  ⊗ A 0

Stage Σ0 to Stage Σ1
The evolution of the preparation switch is given by
   
 1 00 = |s1  ⊗ αA
U1,0 |s0  ⊗ A  1 + βA
 2 01 , (14.9)
0 1 1

where |α|2 + |β|2 = 1.

Stage Σ1 to Stage Σ2
The evolution of the two separate components at stage Σ1 is given by
 R 
   
R+S
 01 = |s2  ⊗
U2,1 |s1  ⊗ A1
RA
i  02 +
i
S A
i,1  02 ,
i
1 2 2
i=1 i=R+1

 
 
Region R
 Region S

  
R+S K
 2 01 = |s2  ⊗
U2,1 |s1  ⊗ A  i 02
S i,2 A +  i 02
T iA . (14.10)
1 2 2
i=R+1 i=R+S+1

 
 
Region S Region T

Here the coefficients {Ri }, {S i,1 , S i,2 }, and {T i } satisfy the semi-unitary condi-
tions

R 
R+S 
K 
R+S 
R+S
|Ri |2 + |S i,1 |2 = |T i |2 + |S i,2 |2 = 1, S i,1∗ S i,2 = 0.
i=1 i=R+1 i=R+S+1 i=R+1 i=R+1
(14.11)
The outcome probabilities are found to be
Pr(Ai2 02 |Ψ0 ) = |αRi |2 , 1iR
= |αS i,1 + βS i,2 |2 , R<iR+S (14.12)
= |βT i |2 R + S < i  K ≡ R + S + T.
These probabilities sum to unity as required.
From this we find the which-path parameter ΦW 2 for W 2 to be the sum of the
probabilities over regions R and T , that is,

R 
K
ΦW 2 = |α|2 |Ri |2 + |β|2 |T i |2 . (14.13)
i=1 i=R+S+1

This reduces to unity when S = 0 as expected and to zero when both R and T
are zero.

14.5 The Delayed Choice Interferometer

A delayed choice experiment that confirms the predictions of QM was done by
Jacques et al. with a Mach–Zehnder interferometer (Jacques et al., 2007), the
190 Quantum Eraser Experiments

relevant stage diagram being the same as Figure 13.1. In this experiment, referred
to here as JACQUES, the final beam-splitter B 2 could be removed while the light
was on its way from the first beam-splitter B 1 .
Although this experiment is of type WHEELER-1, it is also equivalent to the
above WHEELER-2 scenario with K = 2, because a Mach–Zehnder experiment
represented by stage diagram Figure 13.1 is equivalent to a double-slit experiment
with just two detectors in the final stage-detecting screen. In JACQUES, the
configuration with the second beam splitter B 2 removed corresponds to taking
R = T = 1, S = 0 in WHEELER-2, while that with B 2 in operation corresponds
to R = T = 0, S = 2.

14.6 The Double-Slit Quantum Eraser

The above discussed experiments do not involved photon spin specifically. The
experiment we discuss next requires a careful analysis of spin.
Prior to the delayed-choice quantum eraser experiment of Jacques et al.
(Jacques et al., 2007), the double-slit quantum eraser experiment of Walborn
et al. (Walborn et al., 2002) had demonstrated the empirical validity of the
stage concept in quantum mechanics. The Walborn et al. experiment consists
of three subexperiments, referred to here as WALBORN-1, WALBORN-2, and
WALBORN-3.

WALBORN-1: No Which-Way Information

The first subexperiment, WALBORN-1, is shown in Figure 14.3. Source S pre-
pares a spinless photon pair 10 , which is then split, with photon 21 passing
onto a double-slit (22 , 32 ), and then onto a screen with final stage Σ3 detectors
23 , 34 , . . . , K. Meanwhile, 11 is passed onto a detector 13 . Coincidence measure-
ments are taken involving screen impacts and 13 detection, with no polarization
involved.
The initial total state |Ψ0 ) is given by |Ψ0 ) ≡ |s0  ⊗ A 1 00 .
0

Figure 14.3. WALBORN-1: zero which-path information.

 14.6 The Double-Slit Quantum Eraser 191

Stage Σ0 to Stage Σ1
Evolution from Σ0 → Σ1 is given by
 
 1 00 = |s1  ⊗ A
U1,0 |s0  ⊗ A 2
 1A
0 1 1 01 . (14.14)

Stage Σ1 to Stage Σ2
Evolution from Σ1 → Σ2 is given by
   
U2,1 |s1  ⊗ A  2 01 = |s2  ⊗ αA
 1A  2 + βA
 1A  3 02 ,
 1A (14.15)
1 1 2 2 2 2

where |α|2 + |β|2 = 1.

Stage Σ2 to Stage Σ3
The final stage transition Σ2 → Σ3 is given by
  
K
U3,2 |s2  ⊗ A  2 02 = |s3  ⊗
 1A  1A
C iA i
2 2 3 3 03 ,
i=2
  
K
U3,2 |s2  ⊗ A 3
 1A  1A
i
2 2 02 = |s3  ⊗ Di A 3 3 03 , (14.16)
i=2

where the screen consists of K −1 detectors and the {C i }, {Di } coefficients satisfy
the usual semi-unitarity conditions.
The coincidence rates Pr(A i
 1A
3 3 03 |Ψ0 ) are given by

 1A
Pr(A i
3 3 03 |Ψ0 ) = |αC + βD | ,
i i 2
i = 2, 3, . . . , K, (14.17)

and these are the same as the single site detection rates Pr(A  i 03 |Ψ0 ). These
3
results demonstrate double-slit interference because the detection of the 13 signal
provides no which-way information, so that this version of the experiment is
equivalent to a standard double-slit experiment.

WALBORN-2: Creation of Which-Path Information

WALBORN-1 is now reconsidered with some modifications, shown schematically
in Figure 14.4. Walborn et al. placed two quarter-wavelength polarizers P 2
and P 3 in front of 23 and 33 as shown (Walborn et al., 2002). Each polarizer
alters the beam it acts on in a way that distinguishes it from the other beam.
The consequence is that each signal observed on the detecting screen contains
information about the path taken. Hence no interference should be observed on
the screen.
To understand the action of the P 2 and P 3 modules, we need to consider three
sets of orthonormalized photon polarization bases, and the fact that we have a
two-spin photonic internal space.
192 Quantum Eraser Experiments

S P

Figure 14.4. WALBORN-2: creation of which-path information.

Spin Bases
The first orthonormal photon spin basis B HV ≡ {|H, |V } consists of a hori-
zontal polarization state |H and a vertical polarization state |V ; the second
photon spin basis B LR ≡ {|R, |L} consists of a right circularly polarized state
|R and a left circularly polarized state |L; and the third photon spin basis is
B W P ≡ {|P , |N }, a conceptual “which-path” basis defined by Walborn et al.
(Walborn et al., 2002). For clarity, our |P  and |N  states are the same as |+
and |− states, respectively, used in Walborn et al. (2002).
These bases are related as follows:
1 1−i
|H = √ {|P  + |N } , |R = {|P  + i|N } ,
2 2
(14.18)
1 1−i
|V  = √ {|P  − |N } , |L = {i|P  + |N } .
2 2

Photonic Space Basis

The photonic internal spin space involves symmetrized spin states of two photons
denoted p and s by Walborn et al., where the p beam is passed into module X
while the s beam is passed onto the double slit. Therefore, we have to consider
tensor products of the form |H p |H s , |H p |V s , and so on, where we drop the
usual tensor product symbol ⊗. For clarity, we keep a tensor product symbol
between “inner” (SUO) states and “outer” (detector) labstates. In our CA pro-
gram MAIN, we use the following four states, {s[i, n] : i = 1, 2, 3, 4}, to serve
as an orthonormal basis for the internal photonic degrees of freedom, defined at
stage Σn by
|Hnp |Hns  → s[1, n], |Hnp |Vns  → s[2, n],
(14.19)
|Vnp |Hns  → s[3, n], |Vnp |Vns  → s[4, n].
 14.6 The Double-Slit Quantum Eraser 193

Double-Slit Polarizers
The two modules P 2 and P 3 have the following active4 actions on their respective
input beams:
⎧ 2 s
⎨ P |H2  →2 |Ls3  = √12 {|H3s  + i|V3s } ,
P
beam 22 (14.20)
⎩ P 2 |V2s  → i|R3s  = √1 {i|H3s  + |V3s } ,
2 P 2
⎧ 3 s
⎨ P |H2  →3 |R3s  = √12 {|H3s  − i|V3s } ,
P
beam 32 (14.21)
⎩ P 3 |V2s  → −i|Ls3  = √1 {−i|H3s  + |V3s } .
3P 2

The stage dynamics is as follows.

Stage Σ0 to Stage Σ1
⎧ ⎫
⎪
⎨ ⎪
⎬
1  1A 2 01
U1,0 |Ψ0 ) = √ |H1p |V1s  + |V1p |H1s  ⊗ A . (14.22)
⎩
 
 ⎪
2⎪ ⎭ 1
1 
s[2,1] s[3,1] a[20 +21 ,1]≡a[3,1]

We show in (14.22) how the various terms are transcribed into the notation used
in program MAIN.

Stage Σ1 to Stage Σ2
   
U2,1 |H1p |V1s  ⊗ A  2 01 = √1 |H p |V s  ⊗ A
 1A  2 02 + A
 1A  3 02 ,
 1A
1 1 2 2 2 2 2 2
2
  1  
2
 1A 1 2 1 3
U2,1 |V1p |H1s  ⊗ A 1 1 01 = √ |V2 |H2  ⊗ A2 A2 02 + A2 A2 02 .
p s
(14.23)
2

Stage Σ2 to Stage Σ3
Here the modules P 2 and P 3 take effect:
 
U3,2 |H2p |V2s  ⊗ A  2 02 = √1 {i|H p |H s  + |H p |V s } A
 1A 2
 1A
2 2 3 3 3 3 3 3 03 ,
2
  1
2
 1A 1 2
U3,2 |V2p |H2s  ⊗ A 2 2 02 = √ {|V3 |H3  + i|V3  ⊗ |V3 } A3 A3 03 ,
p s p s
2
  1
3
 1A 1 3
U3,2 |H2p |V2s  ⊗ A 2 2 02 = √ {−i|H3 |H3  + |H3  ⊗ |V3 } A3 A3 03 ,
p s p s
2
 
U3,2 |V2p |H2s  ⊗ A  3 02 = √1 {|V p |H s  − i|V p  ⊗ |V s } A
 1A  1A
3
2 2 3 3 3 3 3 3 03 . (14.24)
2
The point here, as stressed by Walborn et al., is that all four photon polariza-
tion states on the right-hand side of (14.24) are mutually orthogonal. Therefore,

4
Here “active” means that the changes are physically observable.
194 Quantum Eraser Experiments

no interference is to be expected in any subsequent pattern of signals. To confirm

that QDN gives such a conclusion, we need to evolve the total state to the final
stage.

Stage Σ3 to Stage Σ4
No polarizations are affected on passage through the double slit, so we have
  K
U4,3 |si3  ⊗ A  2 03 =
 1A  1A
C j |si4  ⊗ A j
3 3 4 4 04 ,
j=2
  
K
U4,3 |si3  ⊗ A 3
 1A  1A
Dj |si4  ⊗ A j
3 3 03 = 4 4 04 , (14.25)
j=2

for i = 1, 2, 3, 4. Here the {C j }, {Dj } coefficients satisfy the usual semi-unitarity

relations.

The above information was encoded into program MAIN, with the following
results for the nonzero maximal question answers:
  1 1 j2
Pr A j
 1A
4 4 04 |Ψ0 = |C | + |D | ,
j 2
j = 2, 3, . . . , K. (14.26)
2 2
This confirms that QDN reproduces the empirical results of Walborn et al.
(Walborn et al., 2002). Essentially, placing P 2 and P 3 in front of their respective
slits destroys the lack of which-way information observed in the unpolarized
double-slit experiment WALBORN-1  discussed
 above.
 
MAIN also confirms that Pr A  j 04 |Ψ0 = Pr A j
 1A
4 4 4 04 |Ψ0 , j = 2, 3, . . . , K,
that is, that the single signal pattern on the screen shows no interference.

WALBORN-3: Erasure of Which-Path Information

The variants WALBORN-1 and WALBORN-2 confirm standard QM expecta-
tions: interference occurs in WALBORN-1 because no which-path information is
available, while the modules P 2 and P 3 in WALBORN-2 provide such informa-
tion by labeling the two beams passing through the double slit, and so there is
no interference. The essence of WALBORN-3 is that the module labeled X in
Figure 14.5 counteracts the effects of P 2 and P 3 so that interference in correla-
tions is now observed. What is incomprehensible from a classical perspective is
that the action of X is nonlocal relative to P 2 and P 3 . This is the point made
by Wheeler: P 1 and P 2 could be operating on one side of the Universe and X
on the other, but the interference destroyed in WALBORN-2 is restored by the
action of module X.
The QDN analysis of the total state evolution for WALBORN-3 is identical to
that for WALBORN-2 up to stage Σ2 . However, at stage Σ3 , the most suitable
internal polarization basis to use is B W P rather than B HV , because X essentially
filters out the two elements |P  and |N  of that basis.
 14.6 The Double-Slit Quantum Eraser 195

S P

Figure 14.5. WALBORN-3: erasure of which-path information.

The combined actions of P 2 , P 3 , and X are now given by

⎧
⎨ P 2 |H2s  → |Ls3  = (1−i)
2 {i|P3s  + |N3s } ,
P2
beam 22 (14.27)
⎩ P 2 |V2s  → i|R3s  = (i+1) {|P3s  + i|N3s } ,
P2 2
⎧
⎨ P 3 |H2s  → |R3s  = (1−i) 2 {|P3s  + i|N3s } ,
P3
beam 32 (14.28)
⎩ P 3 |V2s  → −i|Ls3  = − (i+1)
2 {i|P3  + |N3 } .
s s
3P

Module X has the following effects on the beam from 12 :

⎧
⎨ X|H2p  → √1 {|P3p  + |N3p }
X 2
beam 12 (14.29)
⎩ X|V2p  → √12 {|P3p  − |N3p } .
X

Although this looks like a passive basis change, module X has the property of
splitting the beam into two separately observable components and this is critical
to the experiment. Module X is really an active transformation, with stage Σ3
output beams denoted 13 and 23 .

Stage Σ2 to Stage Σ3
Here the pattern of information flow is intricate, requiring great care in the
programming. The difficulty of correctly encoding what the experimentalists had
done turned out to generate the most significant test of the QDN formalism: a
single sign error would easily invalidate the whole calculation. This underlines
what we said at the end of Chapter 12, that there is an obvious need for a more
sophisticated process of transcribing stage diagrams into CA code. We find the
following transition rules:
196 Quantum Eraser Experiments
 
   1A
{|P3p |P3s  + i|P3p |N3s } A 3
2
 1A (1+i) 3 3 03 +
U3,2 |H2p |V2s  ⊗ A 2 2 02 = √ ,
2 2
{|N p |P p  + i|N p |N s } A  3 03
 2A
3 3 3 3 3 3
(14.30)
 
   1A
{i|P3p |P3s  + |P3p |N3s } A 4
p  1 3 (1+i) 3 3 03 +
U3,2 |H2 |V2  ⊗ A2 A2 02 = − 2√2
s
,
{i|N3p |P3p  + |N3p |N3s } A 2A4
3 3 03
(14.31)
 
  {i|P p
|P s
 + |P p
|N s
}  1A
A  3 03 −
p  1A
U3,2 |V2 |H2s  ⊗ A 2 (1−i)
√
3 3 3 3 3 3
2 2 02 = 2 2 ,
{i|N p |P p  + |N p |N s } A 2A 3 03
3 3 3 3 3 3
(14.32)
 
   1A
{|P3p |P3s  + i|P3p |N3s } A 4
 3 02 =
 1A (1−i) 3 3 03 −
U3,2 |V2p |H2s  ⊗ A √ .
2 2 2 2
{|N3p |P3p  + i|N3p |N3s } A 4
 2A
3 3 03
(14.33)

Stage Σ3 to Stage Σ4
The evolution is given by (14.25), taking into account that detector 24 is not
associated now with the double-slit detecting screen but is part of the X module
output channel detectors.

The above information when fed into program MAIN gives the results:
  1 1 j 2 i  j j∗ 
Pr A j
 1A
4 4 04 = |C | + |D | + C D − C j∗ Dj ,
j 2
4 4 4
  1 1 i  j j∗ 
Pr A j
 2A
4 4 04 = |C | + |D | − C D − C j∗ Dj ,
j 2 j 2
j = 3, 4, . . . , K,
4 4 4
(14.34)

demonstrating the interference observed by Walborn et al. Moreover, the two

alternative output channels, 14 and 24 , show the out-of-phase interference
referred to by Walborn et al. as fringe and antifringe, respectively.
QDN does indeed simulate the empirical results of Walborn et al. The following
comments are relevant.

The No-Signaling Theorem Is Vindicated

The action of module X is entirely local, being applied only to beam 12 . The
question is, could anything done by X be observed at the possibly remote
screen consisting of detectors 34 , 54 , . . . , K4 alone? The answer is emphatically
no. Specifically, we find the single detector outcome probabilities to be given as
     
Pr A  j 04 ≡ Pr A  1Aj 2 j
4 4 4 04 + Pr A4 A4 04
1 1 (14.35)
= |C j |2 + |Dj |2 , j = 3, 4, . . . , K,
4 4
that is, showing no interference terms.
 14.7 Concluding Remarks 197

The Stage Concept Is a Valid Model of Empirical Time

Most significantly, Walborn et al. repeated their experiment with the screen and
p photon detection order reversed with significant time differences and found
no change in the results (Walborn et al., 2002). This is strong evidence for the
validity of the stages concept in such quantum process.

14.7 Concluding Remarks

Our analysis supports the notion that QM never actually needs to involve any
acausality in order to account for empirical data. We should be worried if it did,
for then our entire view of what probability and information represent would
need drastic revision.
Detailed QDN analysis reveals the basic fact that interference phenomena
involve a lack of information about quantum states, and not specifically about
particles per se. Conceptual problems arise when our classical conditioning is
relied on too much. We would like to believe in photons as particles and we
would like to believe that time runs continuously. Both concepts have their uses,
but QM seems to require a generalization of both. In the case of the former,
experiments tell us that we have to deal with interference of amplitudes, not
particles. In the case of the latter, we cannot expect quantum processes to evolve
strictly according to an integrable timetable, such as coordinate time, or even
the physical time in a laboratory. What is important is whether or not quantum
information has been extracted. If it has been placed “on hold,” as can be seen
in our analysis of the delayed-choice eraser and the double-slit eraser, then it can
remain in a stage that could in principle persist until the end of the Universe.
 15
Particle Decays

15.1 Introduction
In this chapter we discuss experiments where the run architecture is significantly
different from that of standard in-out experiments such as particle scattering.
We apply the quantized detector network (QDN) formalism to particle decays,
the ammonium molecular system, Kaon-type regeneration decay experiments,
and quantum Zeno experiments. In all of these experiments, the problem is the
modeling of time, which conventionally is taken to be continuous. In QDN, time
is treated in terms of stages, which are discrete. We show how the QDN formalism
deals with such experiments.
In standard quantum mechanics (QM), time is assumed to be continuous. That
is a legacy from classical mechanics (CM), which does not concern itself in general
with the processes of observation. CM assumes systems under observation (SUOs)
“have” physical properties that are independent of how they are observed. In
contrast, QM cannot be considered without a discussion of the processes of
observation. On close inspection of any process of observation, as it is actually
carried out in the laboratory and not how it is modeled theoretically, the continuity
of time does not look quite so obvious.
The problem is that there are two mutually exclusive views about the nature
of observation in physics. These were discussed in detail by Misra and Sudarshan
(MS) in an influential paper on the quantum Zeno effect (Misra and Sudarshan,
1977). On the one hand, no known principle forbids the continuity of time, so
the axioms of QM are stated implicitly in terms of continuous time. When the
Schrödinger equation is postulated to be one of them (Peres, 1995), temporal
continuity is assumed explicitly. On the other hand, it is an empirical fact that
no experiment can actually monitor any SUO in a truly continuous way. All
references to continuous time measurements are invariably based on statistical
modeling of complex processes, with the continuity of time having much the same
status as that of temperature. Such effective parameters are extremely useful in
 15.1 Introduction 199

physics, but their status as model-dependent, emergent attributes of SUOs, and

the apparatus used to observe them, should always be kept in mind.
The best that could be done toward simulating temporal continuity in physics
would be to perform a sequence of experiments with a carefully prescribed
decreasing measurement time scale, such as occurs in experiments investigating
the phenomenon known as the quantum Zeno effect (Itano et al., 1990).
MS analyzed particle decay processes and asked certain questions about them
not normally investigated in quantum mechanics. Three of these questions were
referred to as P , Q, and R and this convention will be followed here.

P (t|Ψ)
This question asks for the probability that an unstable system prepared at time
zero in state Ψ has decayed sometime during the interval [0, t].

Q(t|Ψ)
This question asks for the probability that the prepared state has not decayed
during this interval.

R(t1 , t|Ψ)
This question asks for the probability that the state has not decayed during the
interval [0, t1 ], where 0 < t1 < t, and has decayed during the interval [t1 , t].

Here we come across an example where mathematics and logic cannot be used
to explore physics. We pointed out in Section 2.10 that the validation1 of the
negation ¬P of a physical proposition P cannot always be undertaken by the
same apparatus that is used to validate P . The point is that suppose we had used
apparatus AP to answer MS’s question P (t|Ψ) by looking for decay products of
an unstable SUO and had found no such decay products over any given interval
of time. We could not conclude that the SUO was absolutely stable; there could
be decay products that our apparatus could not detect. At best we could only
say that the SUO was stable relative to AP .
In physics, therefore, we cannot simply assert Q(t|Ψ) = 1 − P (t|Ψ) as an
empirical fact, because as we stated in Chapter 2, what are important in physics
are generalized propositions, and these require full specification of apparatus. In
order to answer Q(t|Ψ), we would have to use apparatus AQ , which could be
very different from AP .
Likewise, in order to answer MS’s question R(t1 , t|Ψ), we would have to use
apparatus AR .
The point made by MS is not quite the same as what we have just made. Our
concern is about apparatus, theirs was about time. MS stressed that P (t|Ψ),
Q(t|Ψ), and R(t1 , t|Ψ) are not what quantum mechanics normally calculates,

1
Our convention is that validation means the attempt to establish the truth of a proposition.
200 Particle Decays

which is the probability distribution of the time at which decay occurs, denoted
by T . The difference as they saw it is that the P , Q, and R questions involve
a continuous set of observations (according to the standard paradigm), or the
nearest practical equivalent of it, during each run of the experiment, whereas
T involves a set of repeated runs, each with a one-off observation at a different
time to determine whether the particle has decayed or not by that time. Because
P , Q, and R involve an experimental architecture different from T , it should be
expected that empirical differences might be observed.
Note that the observations referred to by MS to can have negative outcomes;
i.e., a failure to detect an expected particle decay in an experiments counts as an
observation. The correct statement of such an observation is not that the particle
is stable, but that that particular experiment has failed to detect any decay.
MS emphasized the limitations of QM, stressing that although it works
excellently in many situations, QM does not readily give a complete picture of
experiments probing questions such as P , Q, and R. They concluded that “there
is no standard and detailed theory for the actual coupling between quantum
systems and the classical measuring apparatus” (Misra and Sudarshan, 1977).
We fully agree. QDN is a relatively simplified attempt to move toward such a
theory.
Our first task in this chapter is to apply QDN to the simplest idealized decay
process, a particle decaying via a single channel. The quantum Zeno effect is
then discussed. That effect demonstrates that the answer as to whether a system
decays while it is being monitored or whether it remains in its initial state
depends on the experimental context, i.e., the details of the apparatus and
the measurement protocol involved. We follow this by applying QDN to more
complex phenomena such as the ammonium molecule and neutral Kaon decay.
We show how QDN can readily provide the empirical architecture to describe
the spectacular phenomenon of Kaon decay regeneration, originally discussed by
Gell-Mann and Pais in standard QM (Gell-Mann and Pais, 1955).
It will be shown that for all of these phenomena, QDN incorporates probability
conservation at all levels of the discussion and therefore does not require the
introduction of any ad hoc imaginary terms in any energies or the use of non-
Hermitian Hamiltonians.

15.2 One Species Decays

In this subsection, we apply QDN to the description of what in standard ter-
minology would be called the decay of an unstable particle, the initial state X
of which can decay into some multiparticle state Y . Our aim is to show that
the QDN account of such processes readily conserves total probability at all
stages. Because the essence of such processes lies in the temporal architecture,
the momenta and other attributes of the particles involved will be ignored here,
the discussion being designed to illuminate the basic principles of the formalism
 15.2 One Species Decays 201

only. Should such aspects be required, the formalism can readily deal with them
by introducing an “internal” Hilbert space associated with the SUO states.
The run architecture follows the pattern used throughout this book, with
labstate preparation for each run being completed by an initial stage denoted as
Σ0 , and referred to as stage zero. All subsequent stages in that run are counted
from stage zero, so Σ1 is the first stage (after stage zero).
By stage Σ0 of any given run, the observer will have contextual evidence that
they have prepared an X-particle state, in the language of standard QM. This
is represented in QDN by the normalized labstate Ψ0 ≡ A  X 00 , which we have
0
previously designated a preparation switch.
Now consider the first stage, Σ1 , at which the observer has the means to
detect any decay. Suppose by that stage, the labstate is now represented by Ψ1
and given by
 X + βA
Ψ1 ≡ U1,0 Ψ0 = (αA  Y1 )01 , |α|2 + |β|2 = 1. (15.1)
1 1

Here the first term on the right-hand side (RHS) represents the possibility that
the particle has not decayed, whereas the second term, involving Y , represents
the possibility that a decay has occurred.
It is part of the underlying philosophy of QDN that the term in Y in Eq. (15.1)
does not model specific details of the Y state. It models a yes/no possibility that
something has happened. It is an example of a virtual detector rather than a real
detector. A virtual detector is informational in character, not necessarily directly
identifiable with a specific, real detector in the laboratory, although such things
are necessary to establish the context for the labstate Y .
To clarify this point further, suppose that the multiparticle state Y consisted
of N identifiable particles. We have not modeled here the stage Σ1 labstate
Y 1AY 1AY 1 . . . A
 Y 1 01 , where for example A Y 1
1 2 3 N i
by a term in (15.1) such as β A 1 1 1 1 1
would create a signal in a specific detector for decay component particle Y i at
stage Σ1 . We could do that, if we wanted to, however. That would undoubtedly
add to the complexity of a problem that already has some degree of complexity
in its architecture, so that is a scenario where a computer algebra approach to
QDN would be most suitable.
The modeling of the labstate Ψ1 given by (15.1) does however include some
desirable features that we have put in “by hand.” We exclude from the RHS of
(15.1) the possibility that we find no signals whatsoever at stage Σ1 ; that is, we
exclude the signal ground state 01 . This means that the apparatus is what we
have referred to before as calibrated .
In the same spirit, we exclude the signality-two state A XA  Y 1 01 , on the grounds
1 1
that any run with a labstate consisting of the original particle and its decay prod-
uct would be discounted by the observer as contaminated by external influences
(as happens in real experiments).
From (15.1), the amplitude A(X1 |X0 ) for the particle not to have decayed by
stage Σ1 is given by
202 Particle Decays

A(X1 |X0 ) ≡ 01 AX
1 Ψ1 = α, (15.2)
while the amplitude A(Y1 |X0 ) for the particle to have made the transition to
state Y by stage Σ1 is given by
A(Y1 |X0 ) ≡ 01 AY1 1 Ψ1 = β. (15.3)
 X , AY ] = 0, and so on.
Here we have used the fact that [A n n
From the above, we see that
|A(X1 |X0 )|2 + |A(Y1 |X0 )|2 = 1, (15.4)
so total probability is conserved.
The above probabilities can also be calculated directly as expectation values
of partial questions. For the probability Pr(X1 |X0 ) of no decay by stage Σ1 , we
find Pr(X1 |X0 ) ≡ Ψ1 P X Ψ1 = |α|2 , while the probability Pr(Y1 |X0 ) of decay
1
into Y by stage Σ1 is given by Pr(Y1 |X0 ) ≡ Ψ1 P Y1 Ψ1 = |β|2 . Note that these
1
are contextual probabilities: we noted at the previous section that these are
statements valid only relative to the detectors used, which are assumed suitable
for what they are supposed to detect.
On the RHS of (15.3), the label is Y1 ; that is, the decay state label is itself
labeled by a temporal subscript, in this case the number 1, which is the stage
Σ1 at which the amplitude is calculated for. This label of a label is significant.
It registers the fact that when a detector is triggered, it does so irreversibly. The
stage at which this happens is a crucial feature of the analysis, being directly
related to the measurement issues discussed by MS (Misra and Sudarshan, 1977).
Our architecture is based on monitoring the state of the SUO as much as possible,
that is, attempting to perform as good an approximation to continuous-in-time
monitoring as our equipment allows.
The above process conserves signality one, so the dynamics can be discussed
economically in terms of the evolution of the signal operators rather than the
labstates. For instance, evolution from stage Σ0 to stage Σ1 can be given in the
form
A  X U1,0 = αA
 X → U1,0 A  X + βA
 Y1 , (15.5)
0 0 1 1

where U1,0 is a semi-unitary operator satisfying the rule U1,0 U1,0 = I0 , with
I0 being the identity for the initial lab register Q0 ≡ QX 0 and U1,0 being the
retraction of U1,0 .
Process (15.5) involves a change in rank, since Q1 ≡ QX Y1
1 Q1 , but not in
signality. Because dim Q1 > dim Q0 , the evolution operator is properly semi-
unitary, that is, satisfies the condition U1,0 U1,0 = I1 , which is a statement of
irreversibility relative to the observer. This is a critical feature of the experiments
discussed in this chapter, apart for the ammonium molecule, and is the reason
for the apparent loss of probability in conventional Schrödinger wave mechanics
descriptions of unstable particles. In those descriptions, a common strategy is
to consider only the Hilbert space of the original SUO and add an imaginary
 15.2 One Species Decays 203

term −iΓ to energies, thereby forcing wave functions to fall off with time, with
the interpretation that this represents particle decay. Where this probability loss
goes is left unstated.
In relativistic quantum field theory, such as quantum electrodynamics (QED),
the architecture is usually different. There, the Hilbert space is big enough to
accommodate particles (such as muons) and their decay products. Decays are
treated as scattering problems, with initial undecayed particles coming in at
remote negative (infinite past) time and decay products going out at remote
positive (infinite future) time. The extraction of decay lifetimes then is usually
done in a heuristic manner, usually involving manipulation with symbols that are
nominally divergent, such as dividing an amplitude by the four-volume measure
of Minkowski spacetime in order to determine a flux decay rate. Our ambition
in QDN is to avoid such manipulations while retaining probability conservation.

The QDN description of the next stage of the process, from stage Σ1 to stage
Σ2 , is more subtle and involves a null test. Considering the labstate of the above
decay process at stage Σ1 , there are now two terms to consider.

No Decay
The first term on the RHS in (15.1), αA  X 01 , corresponds to a no decay outcome
1
by stage Σ1 . This potential outcome can now be regarded as a preparation, at
stage Σ1 , of an initial X state that could subsequently decay into a Y state
or not, with the same dynamical characteristics as for the first temporal link
of the run, held between stages Σ0 and Σ1 . If the measured laboratory time
interval τ10 ≡ t1 − t0 between stages Σ0 and Σ1 is the same, within experimental
uncertainty, as the measured laboratory time interval τ21 ≡ t2 − t1 between
stages Σ1 and Σ2 , and so on for subsequent links, then spatial and temporal
homogeneity may be assumed, if the apparatus has been set up in the laboratory
carefully enough. This will be a physically reasonable assumption in the absence
of gravitational fields and in the presence of suitable apparatus.

Decay
The second term, β A  Y1 01 , in (15.1) corresponds to decay having occurred during
1
the first time interval. Such an outcome is irreversible in this example, but this
is not an inevitable assumption in general. Situations where the Y state could
revert back to the X state are more complicated but of empirical interest, such
as in the ammonium maser and Kaon and B meson decay. These scenarios are
discussed later sections in this chapter.

Assuming homogeneity, the next stage of the evolution is given by

 X → U2,1 A
A  X U2,1 = αA X + βA
 Y2 , (15.6)
1 1 2 2
 1 → U2,1 A
AY  1 U2,1 = A
Y  1.
Y
(15.7)
1 1 2

Equation (15.7) is justified as follows. The decay term in (15.1), proportional

 Y1 at stage Σ1 , corresponds to the possibility of detecting a decay product
to A 1
204 Particle Decays

state Y at that time. Now there is nothing that requires this information to be
extracted precisely at that stage. The experimentalist could choose, or indeed
be required, to delay information extraction until some later stage, effectively
placing the decay product observation “on hold.” The inherent irreversibility of
signal detectors means that, as a rule, such signals on hold are not lost.
As stated above, this may be represented in QM by passing a state through
a null test, which does not alter it. In QDN this is represented by Eq. (15.7).
Essentially, quantum information about a decay is isolated from the rest of the
experiment and passed forward in time until it is physically extracted.
The lab register Q2 at stage Σ2 has rank three, being the tensor product
Q2 ≡ Q X Y1 Y2
2 Q2 Q2 . Semi-unitary evolution from stage Σ0 to stage Σ2 is still of
signality one and is given by
 X → U2,1 U1,0 A
A  X U1,0 U2,1 = α2 A
 X + αβ A
 Y2 + β A
 Y1 , (15.8)
0 0 2 2 2

with the various probabilities being read off as the squared moduli of the corre-
sponding terms.
The temporal architecture of this process is given in Figure 15.1. It will be
apparent from a close inspection of (15.8) that what appears to look like a space-
time description with a specific arrow of time is being built up, with a memory
of the change of rank of the lab register at stage Σ1 being propagated forward in
time to stage Σ2 . This is represented by the contribution involving A  Y1 , which
2
is interpreted as a potential decay process that may have occurred by stage Σ1
and contributing to the overall labstate amplitude at stage Σ2 .

X Y

X Y

Y Y

Figure 15.1. The temporal architecture of a single X particle decay experi-

ment.
 15.3 The Quantum Zeno Effect 205

Subsequently the process continues in an analogous fashion, with the rank

of the lab register increasing by one at each time step. By stage Σn , assuming
homogeneity, the dynamics is given by

n
 X → Un,0 A
A  X Un,0 = αn A
X + β  Yi ,
αi−1 A (15.9)
0 0 n n
i=1

where Un,0 ≡ Un,n−1 Un−1,n−2 . . . U1,0 is semi-unitary and satisfies the constraint
Un,0 Un,0 = I0 .
The amplitude A(X, n|X, 0) that the original state has not decayed by stage
Σn can be immediately read off the RHS of (15.9) and is given by

A(X, n|X, 0) = 0n AX X
n Un,0 A0 00 = α .
n
(15.10)

The probability Pr(X, n|X, 0) of no decay by stage Σn is the squared modulus

of this amplitude, so provided β = 0, this probability falls monotonically with
increasing n, consistent with our expectations of particle decay. Specifically, if
1
we write α = eiθ− 2 Γτ , where θ is some real phase, Γ is a characteristic lifetime
associated with the decay, and τ is the effective time between successive stages,
then we have

Pr(X, n|X, 0) = e−Γnτ , (15.11)

which is the exponential decay form expected with such phenomena.

Commentary
Figure 15.1 does not reveal the full complexity of what is going on. That will be
appreciated by the observation that the labstate has signality one at every stage.
This means that at any stage, either the original state has not decayed or it has
decayed once, either at the stage being examined or prior to that stage.
This stage diagram reinforces the message that the Block Universe picture
of reality is too simplistic, because that picture is a classical record of what
was actually observed and cannot include the future of whatever “now” is being
discussed (which is stage Σ0 here), unless the vacuous assumption is made that
the future is single valued and predetermined. That would not be compatible
with quantum principles as we know them, however. Figure 15.1 refers to the
future of stage Σ0 and the probability outcomes predicted for the observer by
QM, for that stage only; it is not a valid stage diagram for any process time stage
after stage Σ0 . The contextuality of stage diagrams underlines the message that
physics is contextual, never absolute.

15.3 The Quantum Zeno Effect

The discussion at this point calls for some care with limits, because there arises
the theoretical possibility of encountering the so-called quantum Zeno effect, in
which a carefully monitored state of an unstable SUO appears not to change.
206 Particle Decays

In the following, we will assume that the parameter α in the one-particle decay
discussed above satisfies |α| < 1, because the case |α| = 1 corresponds to a
stable particle, which is of no interest here.
Consider the physics of particle decay. Calculated probabilities should be
functions of labtime, the clock time used by the observer in the laboratory.
Labtime is not assumed here to be a continuous variable on the microscopic
level. Instead, it is linked to the scale of time associated with successive stages,
and this is determined by the apparatus used. The temporal subscript n in our
concept of stages will, when it is so arranged, correspond to a physical interval of
time τ , where τ is some reasonably well-defined time scale characteristic of the
apparatus.
Certainly, stages need not be strictly regulated in terms of being equally spaced
out in time. But in the sort of experiments relevant to this chapter, there will
be such an interval τ , and it will be typically a minute fraction of a second, but
certainly nowhere near the Planck time scale of 10−44 second. Indeed, the conjec-
ture that there is such a Planck time scale remains conjectural and has received
some meaningful criticism (Meschini, 2007). The smallest interval currently that
has been measured empirically is of the order 10−23 second, which is on the
shortest hadronic resonance scale, comparable with the time light takes to cross
a proton diameter. More realistic measurement scales that could be involved in
our discussion directly would probably be electromagnetic in origin, in the range
of 10−9 to 10−18 second. For instance, the shortest controllable time is about 10
attoseconds, that is, about 10−17 second (Koke et al., 2010). Experimentalists
would generally have a good understanding of what their relevant τ is.
Suppose first that we have some reason to believe that we can relate the
transition amplitude α to the characteristic time τ by the rule |α|2 ≡ e−Γτ ,
where Γ is a characteristic inverse time introduced to satisfy this relation. Then
the survival probability P (tn ) is given by P (t) ≡ Pr(X, n|X, 0) = e−Γt , which
is the usual exponential decay formula. No imaginary term proportional to Γ
in any supposed Hamiltonian or energy has been introduced in order to obtain
exponential decay.
A subtlety may arise here, however. Exponential decay implies that |α|2 is an
analytic function of τ with a Taylor expansion of the form
 
|α|2 = 1 − Γτ + O τ 2 , (15.12)
i.e., one with a nonzero linear term. Under such circumstances, the standard
result limn→∞ (1 − x/n)n = e−x leads to the exponential decay law. The possibil-
ity remains, however, that the dynamics of the apparatus is such that the linear
term in (15.12) is zero, so that the actual expansion is of the form
 
|α|2 = 1 − γτ 2 + O τ 3 , (15.13)
where γ is a positive constant (Itano et al., 1990). Then in the limit n → ∞,
where nτ ≡ t is held fixed, the result is given by
 15.4 Matrix Analysis 207
  n
lim 1 − γτ 2 + O τ 3 = 1, (15.14)
n→∞, nτ =t fixed

which gives rise to the quantum Zeno effect scenario. An expansion of the ampli-
tude of the form a = 1+iμτ +ντ 2 +O(τ 3 ) is consistent with (15.13), for example,
if μ is real and μ2 + ν + ν ∗ < 0.
To understand properly what is going on, it is necessary to appreciate that
there are two competing limits being considered: one where an SUO is being
repeatedly observed over an increasingly large macroscopic laboratory time scale
t ≡ nτ , and another one where more and more observations are being taken
in succession, each separated on a time scale τ that is being brought as close
to zero as possible by the experimentalist. In each case, the limit cannot be
achieved in the laboratory. The result is that in such experiments, the specific
properties of the apparatus and the experimental protocol may play a decisive
role in determining the results. If the apparatus is such that (15.12) holds, then
exponential decay will be observed, whereas if the apparatus behaves according
to the rule (15.13), or any reasonable variant of it, then approximations to the
quantum Zeno effect should be observed.
From the QDN perspective, the quantum Zeno effect can be understood from
the architecture of decay observation as follows. Looking at Figure 15.1, we see
that there is one channel, denoted by circles with an X, that runs across all
stages. That channel is the “no decay” channel. If during a run involving a great
number of stages the net probability of any of the other outcomes being detected
is sufficiently low, then it would appear that the original system was stable.
However, given enough stages with a fixed duration τ between each, the decay
outcomes would eventually win out. The quantum Zeno effect therefore relies on
having as brief a duration τ as possible and finding the critical time scale over
which the apparent effect could be observed.
Another way of understanding the quantum Zeno effect is in terms of envir-
onment. For instance, a free neutron will decay with a mean lifetime of about
880 seconds, whereas inside a nucleus, neutrons are generally stable. We can
understand the quantum Zeno effect as the effect of the detection environment
on an otherwise unstable system.

15.4 Matrix Analysis

The single-particle decay scenario discussed above can be discussed efficiently in
terms of semi-unitary matrices.

Definition 15.1 A semi-unitary matrix M is an m × n complex matrix such

that M † M = In , where In is the n × n identity matrix.

Exercise 15.2 Prove that no semi-unitary matrix exists if m < n.

208 Particle Decays

Consider the X decay scenario discussed above. If the initial labstate Ψ0 is

represented by the 1 × 1 column vector [Ψ0 ] ≡ [1], then the action of U1,0 acting
on that labstate Ψ0 given by(15.1)
 may be represented by the action of the 2 × 1
α
semi-unitary matrix U1,0 ≡ acting on [Ψ0 ], giving
β
   
α α
[Ψ1 ] ≡ U1,0 [Ψ0 ] = [1] = . (15.15)
β β
The two required transition amplitudes at stage Σ1 are just the two components
of this vector.
Continuing this process to the next stage, we deduce that the labstate [Ψ2 ] at
⎡ ⎤
α 0
stage Σ2 is represented by the action of the semi-unitary matrix U2,1 ≡ ⎣β 0⎦
0 1
on [Ψ1 ], giving
⎡ ⎤ ⎡ 2⎤
α 0   α
α
[Ψ2 ] ≡ U2,1 [Ψ1 ] = ⎣β 0⎦ = ⎣αβ ⎦ . (15.16)
β
0 1 β
For n > 1, the relevant semi-unitary matrix is an (n + 1) × n matrix
given by
⎡ ⎤
α θ Tn−1
Un+1,n = ⎣ β θ Tn−1 ⎦ , (15.17)
θ n−1 In−1

where θ n is a column of n zeros, θ Tn is its transpose, and In is the n × n identity

matrix. This leads to the final state [Ψn ] at stage Σn :
⎡ ⎤
αn
⎢βα n−1 ⎥
⎢ ⎥
⎢ .. ⎥
[Ψn ] = Un,n−1 Un−1,n−2 . . . U1,0 [Ψ0 ] = ⎢ . ⎥ . (15.18)
⎢ ⎥
⎣ βα ⎦
β

The squared modulus of the first component of this column vector gives the
same survival probability |α|2n as before. It is also easy to read off all the other
transition amplitudes and from them determine discrete time versions of the
P , Q, and R functions discussed by MS (Misra and Sudarshan, 1977).
Although the QDN analysis gives results that look formally like the standard
decay result, the scenario involved is equivalent to that discussed by MS; namely,
there is a constant questioning (or its discrete equivalent) by the apparatus as
to whether decay has taken place or not. In this case the results are simple. For
Kaon and B meson decays, the results are more complicated.
 15.5 The Ammonium System 209

15.5 The Ammonium System

In order to understand the QDN approach to neutral Kaon decay, discussed in
the next section, it will be necessary to review first how stable systems such as
the ammonium molecule are dealt with in our formalism.
The ammonium molecule consists of three hydrogen atoms and one nitrogen
atom. If we ignore molecular rotation and translation as inessential to this
argument, then we can think of the three hydrogen atoms as defining a plane in
three dimensions. Then the nitrogen can be found on either side of this plane.
What is observed is consistent with the classical explanation that the nitrogen
oscillates from one side of this plane to the other and back with a characteristic
frequency. It is this behavior that is the focus of our discussion here.

The Standard QM Account

With the above assumptions about neglecting rotation and translation, a simple
but effective model of the ammonium molecule is described quantum mechani-
cally as a superposition of two orthonormal states, |X and |Y , each of which
represents one of the two possible position states of the single nitrogen atom
relative to the plane defined by the three hydrogen atoms. These two states form
a basis for a two-dimensional Hilbert space HAM , in other words, a qubit. It is
interesting to note that we have encountered here a naturally occurring preferred
basis for a qubit, that is, one that is dictated not by detector apparatus but by
the assumed geometry of the SUO.
An account of the nonrelativistic quantum theory for ammonium is given by
Feynman et al. (1966) so we give only a simplified brief resume here to set the
scene.
It is most convenient to use a matrix representation for the states and the
Hamiltonian operator. We define the preferred basis representation
   
1 0
|X = , |Y  = . (15.19)
R 0 R 1

Then relative to this representation, the Hamiltonian for the system is given by
the Hermitian matrix
 
e f
H= , (15.20)
R f∗ e
where e is real and f can be complex. When f is zero, the two states are
degenerate energy eigenstates and so are stable. This possibility is of no in interest
here, so we shall assume that f = |f |eiφ , where |f | is nonzero and φ is a constant
phase.
The two eigenvalues of H are E ± ≡ e ± |f | with corresponding normalized
energy eigenstates
 
1 eiφ
|± = √ , (15.21)
R 2 ±1
210 Particle Decays

where we have set the arbitrary phases to zero for convenience. Hence an arbitrary
normalized solution to the Schrödinger equation
d
i |Ψ, t = H|Ψ, t (15.22)
dt
has matrix representation
−
|Ψ, t = ae−iE |+ + be−iE
+
t/ t/
|−, (15.23)
R

where |a|2 + |b|2 = 1.

If now we calculate the probability Pr(X, t|Ψ, 0) ≡ X|Ψ, t that the SUO be
found in state |X at time t > 0 we find

Pr(X, t|Ψ, 0) = 1
2 + |ab| cos(β − α + 2|f |t/), (15.24)

where a = |a|eiα , b = |b|eiβ . Similarly, we find

Pr(Y, t|Ψ, 0) = 1
2 − |ab| cos(β − α + 2|f |t/). (15.25)

These probabilities successfully model our expectations. First, if f is zero, then

these probabilities are fixed. Second, if the initial state is prepared in an energy
eigenstate to begin with, which means setting either a or b to zero, then the
probability of finding the SUO in state |X is the same as that of finding it in
state |Y .

The QDN Account

In the QDN description, the temporal architecture is given by Figure 15.2. It is
assumed that there are two different detectable signal states, X, Y , with signal
operators A  Y , respectively, evolving from stage Σn to stage Σn+1 according
X, A
n n
to the rule
 X 0n = {AA
Un+1,n A  X + BA Y }0n+1 ,
n n+1 n+1
  
Un+1,n An 0n = {C An+1 + DAYn+1 }0n+1 ,
Y X
(15.26)

where Un+1,n is a semi-unitary operator and A, B, C, and D are constants deter-

mined by the dynamics of the situation. Semi-unitarity requires the constraints

X X X X

Y Y Y Y

Figure 15.2. The temporal architecture of the ammonium molecule, with the
two orthogonal states denoted X and Y .
 15.5 The Ammonium System 211

|A|2 + |B|2 = |C|2 + |D|2 = 1, A∗ C + B ∗ D = 0, which is equivalent to standard

unitarity in quantum mechanics in this case, because the rank of the quantum
register is constant from stage to stage. All other states will be disregarded on
the basis that there are no dynamical channels between them and states X, Y .
From (15.26) we can find a dyadic representation for Un+1,n in the form
Un+1,n = ΦTn+1 U Φn , (15.27)
where
 
 X 0n+1 , A
 Y 0n+1 ], 0n AX
ΦTn+1 ≡ [A Φn ≡ n
, (15.28)
n+1 n+1
0n AYn
and U is the unitary matrix
 
A C
U≡ . (15.29)
B D

The retraction Un+1,n of Un+1,n is given by

Un+1,n = ΦTn U † Φn+1 (15.30)
and satisfies the relation Un+1,n Un+1,n = Icn , where Icn is the contextual identity
operator
 X 0n 0n AX + A
Icn ≡ A  Y 0n 0n AY (15.31)
n n n n

 X 0n , A
for the contextual subspace Qcn with orthonormal basis {A  Y 0n }.
n n
The form (15.27) is particularly suitable for finding the evolution operator
UN,0 taking states of the SUO from stage Σ0 to some final stage ΣN . We find
UN,0 = ΦTN U N Φ0 , N = 0, 1, 2, . . . (15.32)
N
The problem therefore reduces to finding U , which we do as follows.
As discussed in Section 11.5, a unitary matrix U such as (15.29) can always
be put in standard form, that is,
 
a −b∗
U = eiη , (15.33)
b a∗
where η is real and a and b satisfy the condition |a|2 + |b|2 = 1.

Exercise 15.3 Find expressions for η, a, and b in terms of A, B, C, and D.

We now state without proof that matrix U can be written in the form
U = eiη V ΛV † , (15.34)
where matrix V is a unitary matrix given by
 
u −v ∗
V ≡ , (15.35)
v u∗
212 Particle Decays

where |u|2 + |v|2 | = 1 and Λ is the diagonal matrix

 + 
λ 0
Λ≡ . (15.36)
0 λ−
Here λ+ , λ− are the two eigenvalues of U and so are the roots of the equation
λ2 − (a + a∗ )λ + 1 = 0. These roots are complex conjugates of each other and
have magnitude one, so we can write λ+ = eiθ , λ− = e−iθ for some real angle θ.
The relations between u, v and a, b are
a = |u|2 eiθ + |v|2 e−iθ , b = u∗ v(eiθ − e−iθ ), (15.37)
noting that these are nonlinear in u and v. Indeed, u and v are not unique: we can
multiply each by an element of the unit circle without changing relations (15.37).
The significance of the form (15.34) is that it is now easy to evaluate powers
of U . Specifically, we find
Un,0 = eiηn ΦTn V Λn V † Φ0 . (15.38)
Applying this to the evolution of the signal operators then gives, modulo a phase
factor,
 X → {|u|2 einθ + |v|2 e−inθ }A
A  X + u∗ v{einθ − e−inθ }A
Y ,
0 n
  n
 Y → uv ∗ {einθ − e−inθ }A
A  X + |u|2 e−inθ + |v|2 einθ AY . (15.39)
0 n n

Hence we find the conditional probabilities

Pr (X, n|X, 0) = Pr(Y, n|Y, 0) = |u|4 + |v|4 + 2|u|2 |v|2 cos (2nθ) ,
Pr(Y, n|X, 0) = Pr(X, n|Y, 0) = 4|u|2 |v|2 sin2 (nθ) , (15.40)
which agrees with the QM expressions (15.24) and (15.25) when the parameters
u, v, and θ are chosen suitably.
It was noted by Itano et al. (1990) that a survival probability of the form
P (τ ) ∼ 1 − γτ 2 + O(τ 3 ) would be needed to make observations of the quantum
Zeno effect viable. The above calculation of the ammonium survival probabilities
is compatible with this, as can be seen from the expansion
Pr (X, n|X, 0) = |u|4 + |v|4 + 2|u|2 |v|2 cos (2nθ)
 
∼ 1 − 4|u|2 |v|2 n2 θ2 + O n4 θ4 . (15.41)
Therefore, it is predicted that the quantum Zeno effect (or at least behavior
analogous to it) should be observable in the ammonium system, if the right
experimental conditions are set up. As with the particle decays discussed in the
previous section, it would be necessary to ensure that the two limits, t → ∞,
τ → 0, were carefully balanced.

15.6 Kaon-type Decays

The explanation by Gell-Mann and Pais (Gell-Mann and Pais, 1955) of the
phenomenon of regeneration in neutral Kaon decay was a successful application of
 15.6 Kaon-type Decays 213

QM to particle physics. In the standard calculation (Feynman et al., 1966), a non-

Hermitian Hamiltonian is used to introduce the two decay parameters needed to
describe the observations. We will show that QDN readily reproduces the results
of the Gell-Mann and Pais calculation while conserving total probability and
without the introduction of any complex energies.
The analysis of the Kaon system is more complicated than the single-particle
decay process discussed above, involving the interplay of two distinct neutral
Kaons, the K 0 and its antiparticle, the K̄ 0 . In that respect, our discussion in
the previous section of the ammonium molecule is useful. The QDN discussion
of neutral Kaon decay goes as follows (Jaroszkiewicz, 2008b).
Consider three different particle states, X, Y , and Z, making transitions
between each other in the specific way described below. An important example
of such behavior in particle physics involves the neutral Kaons, with X
0
representing a K 0 meson, Y representing a K meson, and Z representing
their various decay products. Kaon decay is remarkable for the phenomenon of
regeneration, in which the Kaon survival probabilities fall and then rise with
time. A similar phenomenon has been observed in B meson decay (Karyotakis
and de Monchenault, 2002).
As before, attention can be restricted to signality-one states. The dynamics is
described by the transition rules
A X → αA X + βA Y + γA  Z n+1 , (15.42)
n n+1 n+1 n+1
  
An → uAn+1 + v An+1 + wAZ
Y X Y  n+1
n+1 , (15.43)
 Za → A
A  Z a , a = 1, 2, . . . , n (15.44)
n n+1

where semi-unitarity requires the transition coefficients to satisfy the constraints

|α|2 + |β|2 + γ|2 = |u|2 + |v|2 + |w|2 = 1, α∗ u + β ∗ v + γ ∗ w = 0. (15.45)
The above process is a combination of the decay and oscillation processes
discussed in previous sections.
The temporal architecture is given by Figure 15.3: the dynamics given by
(15.42)–(15.44) rules out transitions from Z states to either X or Y states.
Therefore, once a Z state is created, it remains a Z state, so there is an irreversible

X
X
X
X Y
Y
Y Z
Y Z
Z Z
Z
Z
S S S S

Figure 15.3. The temporal architecture of Kaon decay.

214 Particle Decays

flow from the X and Y states and so these eventually disappear. Before that
occurs, however, there will be back-and-forth transitions between the X and Y
states that give rise to the phenomenon of regeneration.
In actual Kaon decay experiments, pure K 0 states can be prepared via the
0
strong interaction process π − +p → K 0 +Λ, while pure K states can be prepared
0
via the process π + + p → K + + K + p. In our notation, these preparations
 X 0n and A
correspond to initial labstates A  Y 00 , respectively. In practice, super-
0 0
0
positions of K 0 and K states may be difficult to prepare directly, but the analysis
of Gell-Mann and Pais shows that such states could in principle be prepared
indirectly (Gell-Mann and Pais, 1955). Therefore, labstates corresponding to X
and Y superpositions are physically meaningful and will be used in the following
analysis.  
Consider an initial labstate of the form Ψ0 ≡ x0 A  X + y0 A
 Y 00 , where
0 0
|x0 |2 + |y0 |2 = 1. Matrix methods are appropriate here. The dynamics of the
system will be discussed in terms of the initial column vector Ψ0 ≡ [ x0 y0 ]T ,
equivalent to the statement that each run of the experiment starts with the
rank-two lab register Q0 ≡ QX Y
0 Q0 . The dynamical rules (15.42)–(15.44) map
X Y Z1
labstates in Q0 into Q1 ≡ Q1 Q1 Q1 , so there is a change of rank from two to
three. The transition is represented by the semi-unitary matrix
⎡ ⎤
α u
U1,0 ≡ ⎣β v ⎦ , (15.46)
γ w
which subsequently generalizes to
⎡ ⎤
α u 0Tn
⎢ ⎥
⎢β v 0Tn ⎥
Un+1,n ≡⎢
⎢
⎥,
⎥ n > 0, (15.47)
⎣γ w 0Tn ⎦
0n 0n In
where In is the n×n identity matrix and 0n is a column of n zeros. The observer’s
detector array increases rank by one over each time step. The state at stage
Σn is represented by a column vector Ψn with n + 2 components, given by
Ψn = Un,n−1 Un−1,n−2 . . . U2,1 U1,0 Ψ0 . Overall probability is conserved, because
of the semi-unitarity of the transition operators.
As before, the key to unraveling the dynamics is linearity, which is guaranteed
by the use of semi-unitary evolution operators. Suppose the state Ψn at time n
is represented by

Ψn = [xn , yn , znn , . . . , zn1 ]T , (15.48)

where the components xn and yn are such that xn = λn x0 and yn = λn y0 , where

λ is some complex number to be determined. Such states will be referred to as
eigenmodes. They are not eigenstates of any physical operator, but their first two
 15.6 Kaon-type Decays 215

components, xn and yn , behave as if they were. The dynamics gives the relations
n+1
xn+1 = αxn + uyn = λxn , yn+1 = βxn + vyn = λyn , and zn+1 = γxn + wyn .
Experimentalists will be interested principally in survival probabilities for the
X and Y states, so the dynamics of Z states will be ignored here; i.e., the behavior
of the components znk for k < n will not be discussed. Clearly, however, the QDN
formalism is capable of giving much more specific details about the process than
just the X and Y survival probabilities.
It will be seen from the above that λ is an eigenvalue of the matrix
 
α u
,
β v
which means that in principle there are two solutions, λ+ and λ− , for the
eigenmode values, given by λ± = 12 {α + v ± (α − v)2 + 4βu}. It is expected
that these will not be mutual complex conjugates in actual experiments, because
if they were, the analysis could not explain observed Kaon physics. Therefore, the
coefficients α, β, u, and v will be such that the above two eigenmode values are
complex and of different magnitude and phase, giving rise to two decay channels
with different lifetimes, as happens in neutral Kaon decay. In the quantum
mechanics analysis of neutral Kaon decays, Gell-Mann and Pais described the
neutral Kaons as superpositions of two hypothetical particles known as K10
and K20 , which are charge-parity eigenstates and have different decay lifetimes
(Gell-Mann and Pais, 1955). The K10 decays to a two-pion state with a lifetime
of about 0.9 × 10−10 second, while the K20 decays to a three-pion state with a
lifetime of about 0.5 × 10−7 second.
Semi-unitarity guarantees that

|xn+1 |2 + |yn+1 |2 + |zn+1 | = |xn |2 + |yn |2 ,

n+1 2
(15.49)

and so it can be deduced that

|zn+1 |
n+1 2
|λ|2 = 1 − < 1, n = 0, 1, 2, . . . , (15.50)
|xn | + |yn |2
2

given |xn |2 + |yn |2 > 0. From this and the conditions xn = λn x0 , yn = λn y0 , the
eigenmode values can be written in the form λ2 ≡ λ+ = ρ1 eiθ1 , λ2 ≡ λ− = ρ2 eiθ2 ,
where 0 < ρ1 , ρ2 < 1 and θ1 and θ2 are real. The eigenmodes at time t = 0
corresponding to λ1 and λ2 will be denoted by Λ1,0 and Λ2,0 respectively, i.e.
Λ1,0 = [ a1 b1 ]T , Λ2,0 = [ a2 b2 ]T , and then the evolution rules give
⎡ ⎤ ⎡ ⎤
λn1 a1 λn2 a2
⎢ λn1 b1 ⎥ ⎢ λn2 b2 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
Λ1,n = ⎢ cn,n ⎥ , Λ2,n = ⎢ dn,n ⎥ , (15.51)
⎢ . ⎥ ⎢ . ⎥
⎣ .. ⎦ ⎣ .. ⎦
c1,n d1,n
216 Particle Decays

where the coefficients {ck,n } , {dk,n } can be determined from the dynamics. The
initial modes Λ1,0 and Λ2,0 are linearly independent provided λ1 and λ2 are
different. Given that, then any initial labstate Ψ0 can be expressed uniquely as
a normalized linear combination of Λ1,0 and Λ2,0 , i.e., Ψ0 = μ1 Λ1,0 + μ2 Λ2,0 , for
some coefficients μ1 and μ2 . This is the analogue of the decompositions
√
|K 0  = {|K10  + |K20 }/ 2,
0 √
|K  = {|K10  − |K20 }/ 2 (15.52)

in the Gell-Mann and Pais approach.

From this, the amplitude A(X, n|Ψ, 0) to find an X signal at time n is given by

A (X, n|Ψ, 0) = μ1 a1 λn1 + μ2 a2 λn2 , (15.53)

so that the survival probability for X is given by

Pr (X, n|Ψ, 0) = |μ1 |2 |a1 |2 ρ2n

1 + |μ2 | |a2 | ρ2
2 2 2n

+ 2ρn1 ρn2 Re{μ∗1 μ2 a∗1 a2 e−i(θ1 −θ2 ) }, (15.54)

and similarly for Pr (Y, n|Ψ, 0).

There is scope here for various limits to be considered, as discussed in the
single-channel decay analysis, such that either particle decay is seen or the
quantum Zeno effect appears to hold over limited time spans. If we are justified
on empirical grounds in writing ρn1 ≡ e−Γ1 t/2 , ρn2 ≡ e−Γ2 t/2 , where t ≡ nτ
and Γ1 , Γ2 correspond to long and short lifetime decay parameters, respectively,
then the various constants can always be chosen to get full agreement with the
standard Kaon survival intensity functions

I(K 0 ) = (e−Γ1 t + e−Γ2 t + 2e−(Γ1 +Γ2 )t/2 cos Δmt)/4,

0
I(K ) = (e−Γ1 t + e−Γ2 t − 2e−(Γ1 +Γ2 )t/2 cos Δmt)/4, (15.55)

for pure K 0 decays. Here Δm is proportional to the proposed mass difference

between the hypothetical K10 and K20 “particles,” which are each charge-parity
eigenstates and are supposed to have charge-parity–conserving decay channels.
From the QDN approach, such objects need not exist. Instead, they are regarded
0
as manifestations of different possible superpositions of K 0 and K labstates,
each of which is physically realizable via the strong interactions, as mentioned
above. Conversely, the apparatus dynamics may be such that quantum Zeno-type
effects are observed instead of long-term decays. Again, this will depend on the
details of the experiment chosen.
 16
Nonlocality

16.1 Introduction
Our concern in this chapter is locality in quantum mechanics (QM). Locality is
a heuristic physics principle based on the following propositions.

No Action-at-a-Distance
All the evidence points to the principle that physical actions, taken within
restricted (localized) regions of space and time by observers or other agencies
such as systems under observation (SUOs), do not cause instantly observable
effects on other SUOs at large distances. This does not apply to mathematical/
metaphysical concepts such as quantum wave functions or correlations, as these
are conceptual objects (Scarani et al., 2000). Statements about instantaneous
wave function collapse are vacuous (have no empirical significance) and are
therefore not an issue of significance in physics. Such statements are an issue
to theorists who objectivize wave functions, as in Hidden Variables (HV) theory.
Action-at-a-distance is generally regarded as anathema by most physicists. For
example, Newton’s law of universal gravitation is well known for mathematically
encoding action-at-a-distance. There is direct evidence, however, in the form of
a letter written by Newton to Bentley, that Newton believed that gravity acting
“at a distance through a vacuum without the mediation of anything else” was
an absurdity (Newton, 2006).
The no action-at-a-distance principle is encoded in quantized detector net-
works (QDN) by the requirement that labstate preparation and consequent signal
detection never occur at the same stage.

Causal Transmission
All physically observable consequences of local actions taken by an observer or
SUO are transmitted by identifiable physical processes, such as electromagnetic
waves or neutrinos. There is no such thing as magic or action-at-a-distance.
218 Nonlocality

In QDN this proposition is taken into account implicitly in the labstate out-
come amplitudes at each stage, as these model how information is propagated
from stage to stage. When necessary, the information void can be modeled
as if there were fields and/or particles propagating through it, giving scope
for different mathematical models, such as Euclidean space, curved spacetime,
noncommuting spacetimes, and so on. The structure of the information void
is essentially a discussion of whatever modules have to be taken into account
between labstate preparation and signal detection.

No Superluminosity
According to the standard principles of relativity, the speed of transmission of
any observed physical effect is never greater than the local speed of light, as
measured in a standard localized laboratory. To date, no particles or signals that
travel faster than the speed of light (tachyons) have been observed.
A necessary, but not sufficient, condition for QM to respect the no-
superluminosity principle is that the no-communication theorem holds. This
theorem in standard QM and in QDN is discussed at the end of this chapter.
The no-communication theorem is insufficient in relativistic QM because it
does not mention the speed of light. In relativistic quantum field theory (RQFT),
the theorem is replaced by the above-mentioned no-superluminosity principle
directly, which involves lightcones. These have a special place in physics, having
both emergent and reductionist aspects.
Lightcones are obviously emergent structures because they define macroscopic
subsets of spacetime consisting of events that are all either time-like or space-like
relative to a given event (identified with the vertex of a given lightcone). On the
other hand, lightcones are intimately involved in the reductionist formulation of
RQFT, such as the postulate that operators representing observables at relatively
spacelike intervals have to commute. Interestingly, RQFT places no such restric-
tion on unobservables such as Dirac fields, which obey anticommutation relations.
In Chapter 24, we discuss the construction of fermionic quantum fields from
a QDN perspective, based on Jordan and Wigner’s nonlocal quantum register
approach, a manifestly emergent formulation (Jordan and Wigner, 1928).
QDN is not a reductionist approach to QM, as it deals principally with appa-
ratus sitting on the interface between relative internal context (the world of
SUO states) and relative external context (the environment in which that appa-
ratus and the observer are situated). The precise relationship between QDN
and lightcone structure is not clear at this time. This is consistent with the
general situation at this time that a proper quantum theory of spacetime, so-
called quantum gravity, has not been established. All attempts to do so from a
reductionist approach have failed, to date.1 It is not even clear at this time what
has to be “quantized.”

1
Failure from the point of view of proving empirically vacuous.
 16.1 Introduction 219

We shall show in the last section of this chapter that QDN does obey the
no-communication theorem, which is certainly necessary for lightcone physics
to work. However, that theorem alone is not sufficient to establish lightcone
structure.

Shielding
QDN actually goes further than the no-superluminosity principle, in that it allows
for shielding. This is the empirical possibility of bypassing lightcone causality, the
standard relativistic assumption that if event B is inside the forward lightcone
of event A, then A can be the location of processes that “cause” effects to
be observed at B. In practice it is possible, and often necessary, to materially
isolate detectors so that whatever happens at one detector does not affect others,
even when they are time-like separated and could in principle interfere with each
other. Lest this be thought contrary to standard physics, we point out that all
experiments are done on this basis. For example, neutrino detectors are located
in deep mines, to filter out noise.
The shielding concept is really what underpins the stages concept. Our discus-
sion of the double-slit experiments in Chapter 10 illustrates the point well: when
looking at a screen for signals, an observer can do so over extended periods of real
laboratory time, provided the screen is not interfered with by processes external
to the experiment. The final stage of such a run, then, need not be identified
with a definite instant of labtime. This is a form of loss of absolute simultaneity,
different from the one discussed in special relativity (SR).

The above criteria are based on current empirical evidence and are subject
to constant empirical reexamination: experimentalists continue to search for
tachyons, for example. The hard fact is, however, that there have been no
observed violations of the locality principle to date. Therefore, since QDN is
greatly concerned with signal preparation and outcome detection, the locality
principle is one that should be respected by QDN. There are several aspects to
be discussed here regarding this.

Observers and Their Apparatus Are Essential

Any discussion of locality or its breakdown (referred as nonlocality) is meaning-
less if there is no mention of the observers involved, their apparatus, and the
protocols of observation employed. All of these are necessary in order to define
what is meant by the phrase “instantly observable” in the above.

The Information Void Cannot Be an Absolute Void

The information void refers to an absence of detectors, relative to a given
observer, but makes no claims about the reality, structure, or otherwise of
“empty space.” Indeed, what an information void is to one observer may be
a seething mass of detectors to another. For example, an observer looking to
detect neutrinos faces immense technical difficulties, while an observer looking
to detect sunlight need only open their eyes.
220 Nonlocality

Apparatus Nonlocality
A critical attribute of observation that cannot be ignored here concerns the spa-
tial distribution of the observer’s apparatus. No apparatus is perfectly localized
in space or in time. All equipment consists of vast numbers of atoms, which
have spatial extent, and all observations take time. We have in previous chapters
formalized this latter fact, that observations take time, into the stage concept.

Correlations
Apparatus nonlocality plays a critical role in quantum correlations. The super-
luminal transmission of certain types of information, interpreted as correlations,
has been investigated and speeds in excess of 10,000 times the speed of light
reported (Scarani et al., 2000). One of our aims in this book is to demystify
such phenomena. QDN interprets quantum nonlocality as originating from the
fact that apparatus is invariably nonlocal, as are the processes of extracting
information from it, rather than reflecting strange, nonclassical properties of
SUOs. Since apparatus has to be constructed before a quantum state or wave
function can be given any meaning, or a correlation measured, it is then obvious
that nonlocality is built into quantum physics from the word go, in the form of
correlations arising from empirical context.

The Interpretation of Relativity

A particular problem with nonlocality arises from the principles of physics embed-
ded in special and general relativity (GR). Relativity holds a strategic position in
physics. It has passed all its tests and its principles cannot be dismissed, according
to all current empirical evidence. It is as good in its domain of applicability as
QM is in its domain.
Before we go further, we should clarify what we mean by relativity. There really
are two different discussions going on here: the gravitational and the relational.
The former concerns spacetime curvature, while the latter concerns relationships
between observers.

Gravitational Side of Relativity

Einstein’s field equations in GR couple spacetime curvature to the stress energy
tensor. Although phrased in local differential form (thereby giving GR a reduc-
tionist flavor), Einstein’s equations are really aspects of emergent physics that
belong to the relative external context side of any Heisenberg cut: observers
will usually think of themselves as moving along time-like worldlines in a GR
spacetime, and that is an emergent concept.

Relational Side of Relativity

This side of relativity can be seen as an attempt to formulate a theory of observers
in classical mechanics. However, it is emphatically not a theory of observation
 16.2 Active and Passive Transformations 221

per se: there are no prescriptions in SR or GR about apparatus, for instance,

beyond how it is related to relative external context. This aspect of relativity
gives the rules relating the classical data held by one observer with the classical
data held by another, and no more. Put in these terms, we can appreciate why
relativity and QM are not contradictory or incompatible. They are frameworks
discussing different aspects of observation.

We may summarize these comments by saying that relativity deals with relative
external context (REC), whereas QM discusses relative internal context (RIC).
Two historical category errors, the quantum gravity program and the Multiverse
paradigm, appear to have been made here. Quantum gravity attempts to extend
GR into RIC, while Multiverse attempts to extend QM into REC. Both attempts
appear to be empirically vacuous at this time.
Our division of empirical context into REC and RIC is not clear cut but is
identified in QDN with Heisenberg’s cut:

The dividing line between the system to be observed and the measuring
apparatus is immediately defined by the nature of the problem but it obviously
signifies no discontinuity of the physical process. For this reason there must,
within certain limits, exist complete freedom in choosing the position of the
dividing line. (Heisenberg, 1952)

Relativity impacts on QM because there is a principle in relativity, known as

Einstein locality or the principle of local causes (Peres, 1995), that crosses the
line between REC and RIC and has a direct impact on the sort of information
that an observer can extract from their apparatus (which we remind the reader
has nonlocality built into it even before an experiment starts). This principle
asserts that

events occurring in a given spacetime region are independent of external

parameters that may be controlled, at the same moment, by agents located in
distant spacetime regions. (Peres, 1995)

The Einstein principle affects the QDN formalism because relativity asserts
that detectors that are outside each other’s light cones cannot causally influ-
ence each other, yet QDN may give amplitude effects between those detectors.
Explaining how the classical Einstein locality principle can survive in QM, and
indeed in QDN, is a major challenge.

16.2 Active and Passive Transformations

Before we go further, we need to clarify a fundamental point about trans-
formations, as it concerns the relationship between mathematics and physics.
222 Nonlocality

In general, transformations come in two types, called active and passive. We

illustrate the difference between these by considering some set Θ ≡ {θa } of
objects, labeled by an index a.

Active Transformations
An active transformation is something done to the original set Θ, replacing it
with a new set, Θ ≡ {θa }. We represent an active transformation by the rule
Θ → Θ = Θ, (16.1)
where the arrow → means “is replaced by.”An active transformation implies the
existence of some observer (that is, mathematician, experimentalist, or theorist)
who is making a specific change in, or of, a set of objects. Moreover, there is an
implicit assumption that some observer (who may be a “superobserver” playing
the role of a god) has a memory of the original set and can compare it with the
new set.

Passive Transformations
On the other hand, a passive transformation is merely a relabeling of the elements
in a set, with no actual change in the contextually significant properties of those
elements. For such a transformation, we write

Θ → Θ = Θ, (16.2)

where Θ ≡ {θa }. A passive transformation therefore concerns how a set is
described, which, again, implies the existence of some observer (that is, mathe-
matician, experimentalist, or theorist) who is changing their way of describing a
given set.

Example 16.1 Consider a d-dimensional real vector space V with orthonor-

mal basis {ei : i = 1, 2, . . . , d}. Then an arbitrary vector v in V may be written
in the form v = v i ei , where the components {v i } are real and the summation
convention is used.
An active transformation of v is defined by a change in the components
of v but not in the basis: v i is replaced by an arbitrary v i , with the basis
vectors remaining the same. Hence for an active transformation, we write
v → v  ≡ v i ei = v.
On the other hand, a passive transformation leaves the vector v unchanged
but the basis vectors are changed for a new set. For example, a change of
basis from {ei } to {ei }, defined by the invertible linear relations ei ≡ C ij ej ,
gives v → v  ≡ v j ej = v. In this case, the new coefficients {v j } are given by
v j ≡ v i C ij .

Active transformations are generally the only ones of interest to physicists,

because they concern the real, physical world, whereas passive transformations
are done in the mind. There are several kinds of active transformation in physics.
 16.2 Active and Passive Transformations 223

Construction of Apparatus
The construction of apparatus is the severest form of active transformation, as
this creates physical context. We discuss this form of active transformation in
Chapter 25.

Changes in State Preparation

A common form of investigation involves an observer sitting in a fixed laboratory
and making active changes in prepared states, keeping the detectors unchanged.
For example, switching on an electric field will often be associated with the accel-
eration of a charged particle in a scattering experiment. In terms of generalized
propositions, we may describe such a change by the rule
(S, D|Ω, F ) → (S  , D|Ω, F ), (16.3)
where S is a proposition about a prepared state, D is the detection apparatus,
Ω is the observer, and F is the relative external context (the frame of reference)
that defines the observer.

Changes in Outcome Detection

Another common form of investigation involves active changes in detection appa-
ratus, keeping prepared states and relative external context unchanged. For
example, the main magnetic field axis of a Stern–Gerlach experiment may be
rotated to a new orientation.
We may describe such changes by the rule
(S, D|Ω, F ) → (S, D  |Ω, F ). (16.4)
It is commonly assumed that in some cases such as spatial rotations, trans-
formations (16.3) and (16.4) are related by a change in sign in the parameters
involved. That is a matter for empirical validation and not an absolute truth.
The overthrow of parity in 1956 in the Wu–Ambler experiment demonstrated
emphatically that symmetry and logic must not be treated as equivalent to
empirical truths (Wu et al., 1957).

Interframe Experiments
An interframe experiment is essentially an experiment involving two observers:
one of these is associated with state preparation and the other is involved with
outcome detection. Such experiments explore perhaps the most spectacular and
deepest issues in physics. Examples are the Doppler shift, the Unruh effect
(Unruh, 1976), and indeed, the notion that more than one observer is a meaning-
ful topic in physics. Questions about standardization of physics protocols (units
and such like), observation of observers, the constancy of physical “constants,”
Early Universe physics, and so on, come flying at us immediately.
Such experiments may be represented symbolically by
(S, A|Ω, F ) → (S  , D |Ω , F  ), (16.5)
224 Nonlocality

where A is the apparatus that creates the state S, relative to observer Ω and D
is the detecting apparatus relative to observer Ω .
We place a prime on the transformed state/proposition, S  , on the right-hand
side in (16.5), for two reasons. First, what the detecting observer Ω detects will
appear to have properties different from those of the state that the preparing
observer Ω believes they have prepared. Second, that a quantum state means
the same thing to different observers is a notion that needs to be questioned
in several respects, concerning the standardization of physics and the exchange
of context. In QDN, for instance, we do not accept that quantum states are
objective “things.”
To illustrate our concerns, consider the detection by observer Ω of a single
photon signal. On what basis could Ω believe that they had detected a signal
sent from a distant galaxy? The only plausible scenario is that the observer had
already received sufficient contextual information about that galaxy to formulate
and justify such a belief.

We could take the view that such an exchange of sufficient contextual infor-
mation from the source observer to detecting observer is equivalent to having
a single observer, encompassing both source and detectors. While reasonable in
most cases, that point of view seems bizarre as far as intergalactic processes are
concerned.
Active transformations play a role in mathematics, where they are associated
with functions, maps, and operations. A function can be regarded as a form of
active transformation: given a set Θ, a function f maps elements of that set into
some other set, Θ . We can even think of this as defining a “mathematical arrow
of time.”
Passive transformations have played a role in physics in some important situ-
ations.

Symmetries in the Void

By the fact that there are no detectors in the information void, it is permissible
to model signal propagation through it in any theoretically useful way, such
as through Minkowski spacetime. It may then be useful to consider passive
transformations in that spacetime and explore the empirical consequences. For
example, relativistic quantum field theory, which is used to calculate the dynam-
ics of quantum fields in the information void, is generally assumed to Lorentz
covariant, meaning that it does not matter which inertial frame is used to do the
calculations.

Space-Time Symmetries
Passive coordinate transformations played a critical role in the development of
modern physics. Aristotelian physics was firmly based on the proposition that
the Earth is an absolute frame of reference. After the works of Galileo and
Newton, observers discussed physics more carefully. The observer’s frame of
 16.3 Local Operations 225

reference now became identified as an important ingredient. For convenience,

the important frames were usually taken to be inertial frames. It was pointed
out by Bishop Berkeley early on in the development of Newtonian mechanics that
Newton’s laws of motion are invariant to (unchanged by) any passive Galilean
transformation of an observer’s inertial frame of reference (Berkeley, 1721). Later,
this idea was extended to special relativity, leading to the notion that the laws
of physics (excluding gravitation) are invariant to passive Lorentz and Poincaré
transformations of inertial frames of reference.
We describe this concept in generalized proposition form by

(L, ∅|Ω, F ) ≡ (L, ∅|Ω , F  ), (16.6)

where L are Newton’s laws of motion in the nonrelativistic case or the laws of
relativistic mechanics in the relativistic case, and frames F and F  are related
by either a Galilean or Poincaré transformation, as appropriate.
We note the absence of any relative internal context in (16.6), typical of both
nonrelativistic classical mechanics and relativistic mechanics. This contextual
incompleteness gives the generalized proposition classification of such theories
as two.

16.3 Local Operations

In this and following sections, we pin down our meaning of local operation. In
line with the discussion in the previous section, such an operation is always an
active one, taking place over at least one stage and possibly more. We consider
[r]
an active physical operation Ln+1,n on a rank-r apparatus An at stage Σn , an
[r]
operation that affects a number p of detectors in An and leaves the remaining
q ≡ r − p detectors unaffected in a specific way, to be explained below. The
affected detectors and their corresponding signal qubits will be called relatively
local, while the unaffected detectors and their corresponding signal qubits will
be called relatively remote. By unaffected, we mean that no possible partial
measurements on the remote detectors alone would detect any changes due to
Ln+1,n .2
Our approach is to split the quantum register Qn at stage Σn into two sub-
[L] [R] [L] [R]
registers Qn and Qn , such that Qn = Qn Qn . Splitting a quantum register
is a purely mathematical operation, discussed in detail in Chapter 22, although
the motivation for doing this comes entirely from the physics of the experiment.
[L]
Here Qn is the rank-p tensor product Q1n Q2n . . . Qpn of the local signal qubits
[R]
and is therefore referred to as the local subregister , while Qn is the rank q ≡ r−p

2
Note that the language here is imprecise. Experiments to detect changes in the remote
detectors would actually involve two ensembles of runs, comparing partial measurements
on apparatus evolving with the action of Ln+1,n with partial measurements on apparatus
evolving without it.
226 Nonlocality

tensor product Qp+1n Qn

p+2
. . . Qrn of the remote signal qubits and is therefore
referred to as the remote subregister .
Note that the labeling of detectors is arbitrary in principle, so we are always
entitled to order local and remote qubits in the way given above.

Example 16.2 An observer has prepared a labstate Ψ in a rank-five

quantum register Q[5] ≡ Q1 Q2 Q3 Q4 Q5 , given in the computational basis
representation (CBR) by

Ψ = ac24 + ad29 + ae20 + bc10 + bd15 + be6, (16.7)

where each underlined number in bold, such as 24, represents a single element
of the CBR, and a, b, c, d, and e are complex coefficients. Show that this state
is separable relative to the split Q[5] = Q[L] Q[R] , where Q[L] ≡ Q2 Q5 and
Q[R] ≡ Q1 Q3 Q4 .
Solution
The CBR is not best suited to discuss splits, so we need to find the equivalent
of the signal basis representation (SBR) of the state. We translate the above
CBR basis vectors into SBR counterparts as follows. By inspection, the binary
decomposition of the integers 6, 10, 15, 20, 24, and 29 is

6 = 2+4, 10 = 2+8, 15 = 1+2+4+8, 20 = 4+16, 24 = 8+16, 29 = 1+4+8+16.

(16.8)
 2A
Then for example we have 6 = A  3 0, and so on. Hence we can write

Ψ = acA  5 0 + adA
 4A  3A
 1A  4A
 5 0 + aeA 5 0 +
 3A
 2A
+ bcA  4 0 + bdA  2A
 1A  3A
 4 0 + beA  3 0.
 2A (16.9)

By inspection, this can be operator factorized into the form

 5 + bA
Ψ = (aA  2 )(cA
 4 + dA  3A
 1A  4 + eA
 3 )0. (16.10)

This can now be interpreted as the tensor product of two subregister states;
 5 +bA
that is, we may write Ψ = Ψ[L] Ψ[R] , where Ψ[L] ≡ (aA  2 )0[L] is in Q[L] ≡
2 5 [R]  4  1 3 4 
Q Q , Ψ ≡ (cA + dA A A + eA )0 is in Q ≡ Q Q Q , 0 = 0[L] 0[R] ,
3 [R] [R] 1 3 4

and we define signal creation operators for each subregister accordingly.

16.4 Primary and Secondary Observers

There is an issue here, the physical implications of which are deep to say the least
and that underpins many debates about the nature of reality. In the above section
we considered splitting a quantum register into a local subregister and a remote
subregister. What we have in mind, of course, is to associate different “observers”
with each of these subregisters, as this is of interest in various branches of physics
 16.4 Primary and Secondary Observers 227

and information theory. Typically we would call the local observer Alice, and the
remote observer Bob.
What needs to be addressed is the following: if Alice and Bob have no connec-
tion, meaning that they have no channels of communication between them, then
for whom is the combined scenario Alice and Bob meaningful? We have argued
elsewhere in this book that truth values are contextual. So if we want to discuss
Alice and Bob together, we have to specify the context in which we are doing so.
The only answer that makes empirical sense is that there must be (implicitly,
if not explicitly) some third observer Carol who has the contextual information
to know about Alice and Bob and what they are observing and the outcomes
that they have found.
In CM, such an overseeing observer is generally not specified, a factor that
contributes to the essential contextual incompleteness of that discipline. In QM,
we cannot allow such contextual incompleteness. Whatever is asserted must have
some empirical basis for its truth values. Quantum entanglement runs directly
into this issue.
This chain of reasoning leads us to the primary observer concept. A primary
observer is the overseer and custodian of all relevant context: the buck stops with
a primary observer, there is nothing behind them, in the given context. So when
we discuss Alice and Bob as local and remote observers respectively, they are by
implication not primary observers. We will refer to them as secondary observers,
or subobservers.
A fundamental difference in QM between a primary observer and any sec-
ondary observers is that dimensions of Hilbert spaces do not follow an additive
rule: if Alice thinks she is dealing with a p-dimensional Hilbert space and Bob
thinks he is dealing with a q-dimensional space, then primary observer Carol,
who is overseeing Alice and Bob, is dealing with a pq-dimensional space, not a
p + q-dimensional Hilbert space. What is additive is qubit register rank.

Exercise 16.3 If Alice models her experiment with a rank-a qubit quantum
register and Bob models his with a rank-b qubit quantum register, prove that
if Carol models both experiments by a rank c = a + b qubit quantum register
such that the dimension of Carol’s quantum register equals the sum of the
dimensions of Alice and Bob’s registers, then a = b = 1.

There is an interesting question here: is it possible for a primary observer to

observe themselves completely, that is, know everything about themselves? Our
resolution of this is that it is not possible to do this in a complete way: real
empirical information always comes in discrete form and real observers have
finite capacity to store information. According to Sen (Sen, 2010), a finite set
cannot be mapped bijectively to any proper subset (but an infinite set can be
so mapped). Therefore, a real observer cannot faithfully observe themselves in a
complete way.
228 Nonlocality

An observer can observe themselves partially, however: we do this every time

we look in a mirror. What we see is only part of ourselves (our face, clothes, and
so on), and that information is stored not in those parts but in other parts that
serve as memory (our brains). In essence, part of us plays the role of an SUO
and other parts play the role of a primary observer.

16.5 Subregister Bases

Contextuality is equivalent in many respects to information. Given the split of
[L]
the quantum register Qn into the tensor product of a local subregister Qn and
[R]
a remote subregister Qn , discussed in Section 16.3, the question of basis set for
each subregister in the split arises. It turns out that QDN gives an immediate
answer as follows.
As we saw in previous chapters, context gives a preferred basis set for every
signal qubit component of a quantum register. Factoring such a register into
two subregisters, each consisting of an integral number of signal qubits, does
not change this. A preferred basis is determined by physical context, whereas a
split is a purely mathematical operation. Therefore, each signal qubit retains its
preferred basis during a split.
Suppose we have a detector array consisting of r = p + q detectors, where p
and q are at least one and we intend to split it into the tensor product of a local
[L] [R]
subregister Qn consisting of p detector qubits and a remote subregister Qn of
q detector qubits. Now in principle there is no natural way to order any array of
detectors, because they are located in three-dimensional space, and there is no
natural ordering in such a set. We are free to label the original array in any way
that we want. In the context of a split such as we envisage, it seems sensible to
do it as follows. The p detectors that we will include in the local subregister will
be labeled 1 to p and the remote detectors will be labeled p + 1 to r when we
discuss the original register, but from 1 to q when we are discussing the remote
subregister as a separate entity.
We now have an obvious way of organizing each of our subregister bases. We
can define CBRs and SBRs for each. Moreover, we can define signal projection
operators Pin , Pi , Ai , A
 i for each separately, just as if they were independent
n n n
registers (as indeed, they could be).
A natural question is: given a basis element i[L] n in the CBR for the local
register, where 0 ≤ i < 2 and a basis element j n in the CBR for the remote
p [R]

register, where 0 ≤ j < 2q , to what element kn of the CBR for the original
register does the tensor product i[L] n ⊗ j n correspond? It is easy to show that
[R]

with the ordering described above, we have the rule k = i + 2p j.

We may readily construct operators that act on elements of a given subregister
and not on elements in another subregister. For example, if U[L] is a subregister
operator that acts over Q[L] , then relative to the original register Q ≡ Q[L] Q[R] ,
its action is equivalent to that of the operator U[L] ⊗I[R] , where I[R] is the identity
operator for Q[R] .
 16.7 The No-Communication Theorem 229

16.6 Local and Remote Evolution

When a primary observer discusses operations performed by local and remote
secondary observers, such actions are never implemented instantaneously but
take laboratory time, so should be discussed in terms of stage to stage dynamics.
Suppose we split a register Qn into local and remote subregisters as above, with
their respective CBR bases. Consider further a process of evolution from stage
Σm to stage Σn , for n > m, with the local secondary observer Alice performing
[L]
an action U n,m on her apparatus and remote secondary observer Bob performing
[R]
an action U n,m on his apparatus. The question is, how does Carol, the primary
observer, describe both of these processes?
For simplicity, we shall assume that the rank of each subregister is stage
independent. Following our discussion above on subregister bases, we can write
L −1
2r R −1
2r
U [L]
n,m ≡ i[L] [L]i,j [L]
n Un,m j m , U [R]
n,m ≡ i[R] [R]i,j [R]
n Un,m j m . (16.11)
i,j=0 i,j=0

Then Carol can describe what Alice and Bob are doing by the evolution operator
L −1 2rR −1
2r 
Un,m ≡ U [L]
n,m ⊗ U [R]
n,m = n ⊗ kn Un,m Un,m j m ⊗ lm .
i[L] [R] [L]i,j [R]k,l [L] [R]
(16.12)
i,j=0 k,l=0

It is straightforward to rewrite this expression in terms of the CBR for Carol’s

quantum register. Likewise, all subregister operators acting on local or remote
labstates can be readily rewritten in terms of the global register (that is, from
Carol’s perspective).

Two important conclusions can be drawn from this analysis. (1) It is consistent
to apply QM to parts of the universe, while ignoring the rest, even though all of
it is subject to the laws of QM, and (2) it is the possibility of isolating apparatus
that gives rise to the SUO concept in the first place

16.7 The No-Communication Theorem

We have stressed the critical significance of the no-communication concept in
QM. We shall discuss it in simplistic terms in both standard QM and in QDN.

The QM Account
Consider two observers, Alice and Bob. Alice will be our local observer, conduct-
ing active transformations on locally prepared signal states, while Bob will be
our remote observer, conducting observations at remote detectors. The aim is to
see if anything Alice does can affect what Bob observes.
First, we put ourselves in the position of a primary observer, Carol, who has an
overview of what Alice and Bob do. They are now to be regarded as secondary
observers, relative to Carol, although each of them believes themself to be a
230 Nonlocality

primary observer. Suppose Alice models the states she can prepare by elements
of a Hilbert space denoted HA , and suppose Bob models the states he observes
by elements of a Hilbert space HB . We will suppose that Carol has enough
information to model the combined system by the tensor product space HC ≡
HA ⊗ HB .
In the following we shall deal with pure states only, assuming that mixed states
present no exceptional concerns. This is on account of the fact that mixed states
involve no more than the extra complication of classical probabilities, and these
do not interfere in the way that entangled quantum states do.
Suppose Carol arranges for an entangled state to be prepared, described by a
normalized element in HC given by
|ΨC ) ≡ α|ψ A  ⊗ |ψ B  + β|φA  ⊗ |φB , (16.13)
where |ψ A  and |φA  are normalized elements in HA , |ψ B  and |φB  are nor-
malized elements in HB , and |α|2 + |β|2 = 1. The actions of Alice and Bob are
respectively as follows.

Alice
Alice performs a local active operation U A on states in HA that she has access
to, changing them according to the prescription
|ψ A  → |ψ A  ≡ U A |ψ A , |φA  → |φA  ≡ U A |φA . (16.14)

Bob
Bob performs a measurement of an observable OB on states in HB to which he
has access.

Carol
From Carol’s perspective, a holistic account has to be given, in terms of states
in, and operators over, the total Hilbert space HC . From her perspective, Bob’s
observable corresponds to the operator OC ≡ I A ⊗ OB , where I A is the identity
operator over HA , while Alice’s local operator corresponds to the operator U C ≡
U A ⊗ I B , where I B is the identity operator over HB .
Before Alice performs her operation, Carol calculates that Bob’s expectation
value OB  will be given by
OB  ≡ (ΨC |OC |ΨC )
= |α|2 ψ B |OB |ψ B  + |β|2 φB |O|φB 
+ α∗ βψ A |φA ψ B |O|φB  + αβ ∗ φA |ψ A φB |OB |ψ B , (16.15)
where we note the presence of interference terms.
After Alice performs her operation, Carol calculates that Bob’s expectation

value OB  will be given by OB  ≡ (Ψ C |OC |ΨC ), where now

|ΨC ) ≡ U C |ΨC ) = α|ψ A  ⊗ |ψ B  + β|φ A  ⊗ |φB . (16.16)
 16.7 The No-Communication Theorem 231

Then we find

OB  = |α|2 ψ B |OB |ψ B  + |β|2 φB |O|φB 

+ α∗ βψ A |φA ψ B |O|φB  + αβ ∗ φA |ψ A φB |OB |ψ B . (16.17)

Comparing (16.15) and (16.17), it is easy to see that OB  = OB , because
ψ A |φA  = ψ A |U A† U A |φA  = ψ A |φA .
The prediction, therefore, is that no active unitary transformations performed
by Alice would affect Carol’s calculation of Bob’s measured expectation value.
In other words, Alice could not transmit any signals to Bob using entanglement.

The QDN Account

QDN has no problem in fully accommodating the no-communication theorem,
because as discussed in previous sections, it is straightforward to define concepts
of local and remote subregisters that can model all the necessary requirements
for the no-communication theorem to hold. Indeed, we encounter such processes
in our discussion of the quantum eraser experiments in Chapter 14.
 17
Bell Inequalities

17.1 Introduction
From the dawn of the quantum age in 1900, there has been an ongoing debate,
often fierce and hostile, between the supporters of the classical world view and
those of the quantum world view. This debate eventually led to remarkable exper-
iments, the convincing and repeated results of which are claimed by quantum
theorists to support the predictions of quantum mechanics (QM) and not those of
classical mechanics (CM). However, old paradigms tend to linger on the shelves of
science long past their sell-by dates and there are extant schools of theorists that
strive to find loopholes in the above-mentioned experiments. That is a legitimate
activity up to a point, but the divisive nature of the debate requires commentary,
which is the subject matter of this chapter.
QM was discovered only by advanced technology, so it stands to reason that
any test of QM will require even more advanced technology. In the previous
chapter we stressed the differences between active and passive transformations.
The experiments discussed here rely on such technically difficult active trans-
formations that the Hidden Variables (HV) theorists (for that is really what
they are) opposing QM are frequently able to think of objections, known as
loopholes, to the empirical protocols employed by the experimentalists. Some
of these loopholes are reasonable but many are not. Closing those loopholes
convincingly is an ongoing important activity in experimental quantum physics.
A significant feature of the experiments discussed in this chapter is that they
go beyond a certain point of complexity in terms of the number of classical and
quantum degrees of freedom involved. Historically, before that point had been
reached, classical interpretations of quantum wave functions appeared viable and
perhaps even attractive, such as that proposed by Bohm (Bohm, 1952). We
shall refer to the collective of such interpretations as the hidden variables (HV)
paradigm and the point in question as the Heisenberg point.
 17.1 Introduction 233

It seems obvious with hindsight that the Heisenberg point was reached almost
immediately after Schrödinger introduced his wave mechanics formulation of QM
in 1926 (Schrödinger, 1926). Certainly, the wave function Ψ(x, t) for a single-
particle system under observation (SUO) can be visualized as an objective wave
of sorts in physical space. By the term physical space, we mean the three-
dimensional space P3 (Λ) of extension, position, and distance, associated with
a real laboratory Λ,1 a concept that would surely have been understandable to
Aristotle, Galileo, Newton, Hamilton, and Lagrange. But the Schrödinger wave
function Ψ(x1 , x2 , . . . , xN ; t) for an SUO consisting of N particles is actually a
time-dependent complex function over a real 3N -dimensional space C 3N , known
as configuration space, which is quite different conceptually from P3 (Λ).
Humans are remarkably stable creatures, both physically and mentally. Good
health is measured in years and mental outlook is measured in decades. The phe-
nomenon of persistence, which we have attributed as the origin of objectivization,
seems to us responsible for the classical world view. We see objects apparently
unchanged over significant stretches of time, and we come to believe that those
objects have real identities. In fact, we are generally strongly conditioned men-
tally to think in such classical terms. We imagine ourselves as sitting in physical
space P3 (Λ) and we try to relate all experience to it. We may refer to this as
mental persistence, or equivalently, classical conditioning.
Absolute physical space and absolute time (Newton, 1687)2 form the con-
ceptual foundations on which Newtonian mechanics was built. We shall refer
to this mathematical model as space-time, noting the hyphen between “space”
and “time.” This hyphen is important: it marks the recognition that observers
operate in their physical space with a process view of time rather than a manifold
or Block Universe perspective. Indeed, QM seems to us empirically meaningful
only in the space-time perspective, simply because probability, as we know it,
makes no sense otherwise.3
Not only does the space-time model give a remarkably good account of many
physical phenomena such as planetary orbits, but it conforms excellently to our
inherited classical conditioning: we think in terms of objects moving around
physical space with the passage of absolute time.
The advent of relativity did little to change this in practice, for the basic
reason that the speed of light is so great in relative terms that the space-time
model is a good one for all practical purposes. Indeed, its replacement, the Block

1
We follow here Schwinger’s statement, quoted in Chapter 24, that space and time are
contextually defined by apparatus.
2
We refer the reader to the Principia for Newton’s defining comments on what he meant by
“absolute space” and “absolute time”.
3
There will be theorists who interpret QM and probability in terms of abstract
mathematical structures over Block Universe manifolds, with operator norms, C ∗ algebras,
and such like. Those approaches to QM equate empirical physics with mathematical physics
and generally neglect observers and apparatus, thereby usually having a generalized
propositional classification of one.
234 Bell Inequalities

Universe manifold M4 (four-dimensional Minkowski spacetime), came into the

orbit of theorists’ attention only relatively recently, in 1908 (Minkowski, 1908).
It is too soon for human conditioning to have evolved (if it ever will) to the
point where the world around us is interpreted naturally according to the M4
spacetime paradigm rather than the Newtonian space-time paradigm.
When Schrödinger introduced his wave function in 1926, therefore, classical
conditioning ensured that attempts would be made to understand his theory in
classical mechanical terms. Indeed, Schrödinger himself originally interpreted his
wave function in realist terms, but soon realized that the configuration space
argument required a revision in that interpretation.
Such attempts have persisted: there remain groups of theorists who have
the agendas of either refuting relativity (Dingle, 1967) or refuting QM (Bohm,
1952), or refuting both (Kracklauer, 2002). A theme common to these agendas
is contextual incompleteness, with little or no attention being paid to the details
of relative internal context (discussed in Section 2.12). A significant point to
make here is that none of these theorists has reported experiments that they
have actually done, so nullius in verba 4 and Hitchens’ razor 5 can legitimately
be applied here.
Fortunately, the matter goes beyond talking-shop physics because of the
remarkable contribution of the theorist J. S. Bell. In 1964, Bell discussed an HV
model of a two spin-half SUO such as a two-electron state (Bell, 1964). He based
his model on a number of reasonable assumptions about the nature of classical
reality and discovered the possibility of testing those assumptions. The signifi-
cance of this is that Bell opened the door to empirical tests of quantum principles,
because these give predictions different from those based on classical principles.
To understand his ideas, we need to review some concepts.

Parameters
Parameters are used in the description of apparatus and observers, and they
are to be found in relative internal context and relative external context. Every
classical model, including that of Bell, relies on parameters, such as the masses
of the particles being observed, orientation of apparatus relative to the labora-
tory, and so on. Parameters may be physical constants determined by previous
experiments, such as particle masses, or they may be classical degrees of freedom
under the control of the observer, such as orientation angles of a Stern–Gerlach
(SG) main magnetization field.
Parameters should not be confused with dynamical variables, which are theo-
retical constructs related to states of SUOs. Parameters are generally expressed
in terms of classical real numbers, associated with agreed systems of units. Some
parameters are specific, meaning that they are assumed to be exact to within

4
Take no one’s word for it.
5
That which is asserted without proof can be dismissed without proof.
 17.2 The Stern–Gerlach Experiment 235

measurement errors, or statistical in nature. An example of a specific parameter

is the mass of the free electron in vacuo, 9.1938291 . . . × 10−31 kg. An example
of a statistical parameter would be the temperature of the laboratory in which
an experiment was being carried out.

Variables
Variables have to do with states of SUOs. As the word suggests, variables change
over the course of an experimental run. The point of doing experiments is to
determine to what extent observers understand those changes.

Model Limits
A good model will make predictions about the range over which the variables
can go, for a fixed set Θ of parameters. For example, Newtonian mechanics and
Newton’s law of gravitation predicts how far the Earth could go from the Sun
in its annual orbit, given the known parameters, such as the mass of the Earth
and of the Sun, and the current position and velocity of the Earth relative to
the Sun.
Bell showed, in a CM model of two spin-half particles based on “reasonable”
classical assumptions such as locality, that there was a limit Λ to a certain
empirically measurable function F of the parameters Θ,6 given in the form of an
inequality F (Θ)  Λ. Such inequalities are now universally referred to as Bell
inequalities (Bell, 1988). They are in focus in this chapter, because they provide
an empirical test of CM predictions versus those of QM.

In the following sections we shall first discuss the SG experiment from a

classical perspective. Then we shall show how a classical Bell-type inequality
can be derived. Then we shall show how standard QM predicts the possibility
of a nonclassical violation of this inequality. Then we shall give the quantized
detector network (QDN) account of the violation.

17.2 The Stern–Gerlach Experiment

Along with the double-slit experiment, the SG experiment serves as a standard
test of QM principles. In this section we discuss the latter experiment from the
perspective of HV theorists applying standard CM principles.
The stage diagram Figure 17.1 shows the SG architecture. Observer Ted oper-
ates a preparation device T that directs a beam of particles7 10 toward an SG
module S a containing a strong inhomogeneous magnetic field aligned principally

6
By empirically measurable, we mean fixing the parameters and then performing an
experiment to calculate a value for F .
7
We emphasize again that such descriptions express a classical interpretation of what
happens in the laboratory: Ted pushes buttons in preparation device T and Alice looks at
signals on the detector screen. Neither Ted nor Alice see “particles” in the way spectators
observe baseballs or cricket balls.
236 Bell Inequalities

T S

Figure 17.1. Stage diagram of the SG experiment.

along a fixed direction a. This direction is the significant parameter in this

experiment. Interestingly, while the classical explanation of what is going on
requires a knowledge of the magnetic field in module S a , that turns out not to
be directly relevant to the outcome of the experiment; what is important here is
the forest, not the trees.
It was observed by Stern and Gerlach (Gerlach and Stern, 1922a,b) that on
a detecting screen on the opposite side of S a to the source T , signals were
−
observed in two principal areas, or spots labeled A+
1 and A1 in Figure 17.1. These
spots will henceforth be assumed to be disjoint, that is, not overlapping, because
experience suggests that this can always be arranged to a good approximation
in real experiments.
The standard CM interpretation of these empirical observations is that par-
ticles from T have entered the magnetic field in S a and that, by virtue of
interaction with that field, some of them have been deflected into region A+ 1
while the rest have been deflected into region A− 1 . The standard CM theory is
based on the following assumptions.

Extended Charged Particles

Each particle passing through S a has nontrivial electromagnetic structure, mean-
ing that it is not a point but an extended system of electric charge density that
is swirling around in such a way as to create a time-dependent magnetic dipole
moment, denoted μ.

Electrodynamic Forces
As it passes through the SG module S a , each particle’s instantaneous magnetic
dipole moment μ interacts with the inhomogeneous magnetic field B in S a , an
interaction modeled by a term proportional to μ · B in the classical Hamiltonian.
This generates an additional contribution to the standard Lorentz force qv × B
on that particle, and it is this additional force that is interpreted classically as
responsible for what Stern and Gerlach observed. Here q is the electric charge of
the particle, v is its instantaneous velocity, and B is the effective magnetic field
in which the particle is moving.
 17.2 The Stern–Gerlach Experiment 237

The fact that Stern and Gerlach observed two regions, labeled by us A+ 1
and A− 1 in Figure 17.1, is indisputable. Therefore, a CM calculation should
account for that splitting. We assume that that can be done. There are
good examples of such bifurcations in CM, such as comets either falling into
elliptic captured orbits around the sun or entering the solar system once and
then leaving forever on hyperbolic orbits. Another example is from the statistics
of single car accidents on icy roads: cars will veer off a road either to the left or
to the right.
Lacking more detailed information, particularly about any hidden variables
that could contribute additional forces in the SG experiment, we simply assume
that if we could complete such calculations, they would show the necessary
− 8
bifurcation into either A+1 or A1 .

Deterministic Outcomes
The classical electromagnetic forces guiding each particle through module S a
are deterministic, meaning the following. Suppose that at the start of a run,
one of the particles, #1, in 10 has an initial position r 0 and initial velocity v 0 .
Subsequently, it enters S a and ends up somewhere on the detecting screen. Now
a typical beam will consist of a vast number of particles. Suppose another one
of the particles, #2, started off in 10 at exactly the same initial position, the
same initial velocity, and with the same internal degrees of freedom (HVs), as
#1, although at some other time. Then it would subsequently follow exactly the
same spatial path and would end up in exactly the same spatial position on the
screen as #1.
Consider now a beam of many particles from 10 passing through the same
apparatus. There will be a spread of initial positions and initial momenta, with
most particle velocities approximately along the same common direction toward
S a . There may also be some additional hidden variables, such as those involved
with the internal charge structure of the particles. Suppose there are N particles
in such a beam. Then the ith particle starts in 10 with a set of initial variables
denoted θ i . This includes any hidden variables.
Now according to the deterministic principle outlined above, that particle will
−
certainly end up in A+ 1 or else certainly in A1 , assuming perfect transmission,
that is, with no outside interference and with completely efficient detection. It is
important to note that for given θ i , which of the two outcomes occurs is not the
question; what matters is the asserted fact that one of them will be definitely
forced to occur by the deterministic nature of the mechanics. Moreover, this is
not a random outcome: which of the two sites it will be is predetermined by θi .

8
The assumption is made that cases where the particle would end up in the middle of the
−
detecting screen between A+1 and A1 constitute a tiny proportion of the whole and can be
neglected. There must be, for instance, relatively few, if any, comets that are on genuinely
parabolic trajectories.
238 Bell Inequalities

It is also important to note that the question of whether we could calculate that
outcome is irrelevant to the discussion.9
The classical deterministic paradigm outlined
 above  leads naturally to the
following assertion. Consider the set Θ ≡ θ 1 , θ 2 , . . . , θ N of all the initial HVs
associated with a beam in a given run. Then this set can be regarded as the
union Θ = A+ ∪ A− of two disjoint subsets A+ and A− , where A+ is the subset
−
of HVs that send a beam particle into A+ 1 and A is the subset of HVs that send
−
a beam particle into A1 .

Classical Counterfactuality
So far, nothing controversial has been written. To understand the next step,
we need to make a small diversion. There is a piece of logic (or metaphysics,
if you prefer) that is at the heart of the problem in “understanding” QM. It is
the essential point on which Einstein, Podolsky, and Rosen pitched their famous
argument against standard QM (Einstein et al., 1935). It goes by the name of
counterfactuality. A basic definition is the following.

Definition 17.1 If P and Q are two propositions (that is, statements) and
P is known to be false, then the statement P implies Q, written P ⇒ Q, is
a counterfactual (statement) if it is logically true, despite the fact that P is
false.

Example 17.2 Let P be the proposition My computer is broken today

and Q is the proposition I cannot do any typing today. Then it is true
that P ⇒ Q. But actually, at the time of writing this, P is false: my computer
is working and I have just typed out these words.

There are serious questions about the above definition, particularly in view of
the fact that we cannot avoid employing counterfactual reasoning in physics. The
most obvious problem is that Definition 17.1 is contextually incomplete: there
is no reference in it to any observer for whom the “truth” concept is valid, nor
is there any statement of a method by which a “truth” value could be
established. This is not hair-splitting: it matters in physics. That is why we
have qualified the word truth in that definition with the adjective logical. Logic
is not physics.
When it comes to physics, CM adopts a modified form of counterfactual-
ity, referred to here as the principle of classical counterfactuality, also known

9
It is remarkable that QM does not even attempt any such calculation, because not only is it
regarded as a vacuous enterprise, but there is also no mechanism in QM to deal with
individual outcomes. QM is after all a theory about the statistics of observation. HV
theorists, on the other hand, do not consider it contradictory to assert (1) that there are
variables they cannot know anything about, but (2) if they did know about them, they
could predict everything about their behaviour, in principle.
 17.2 The Stern–Gerlach Experiment 239

as counterfactual definiteness. In brief, this principle states that counterfac-

tual statements can be empirically meaningful in physics, that it is meaning-
ful to assume SUOs exist and have properties that are independent of actual
observation.
Classical counterfactuality is used by observers in all experiments, even those
that are described by QM. This is because state preparation is based on prior
established context: how do we know that a beam of particles is entering an
SG apparatus? Such knowledge arises only because we believe in the constancy,
internal consistency, and reliability of the laws of physics that we have established
over centuries, and because we have previously prepared such a beam many times
and tested it for its properties, a process called calibration.
It is just a fact of life that when it comes to actually using such a beam in a
complete run in an experiment, we can no longer test it. If we were to interrupt
the beam before it entered the apparatus, in order to check that the beam is
what we think it is, then the rest of the run could not subsequently take place.
We literally cannot have our cake and eat it.
This is perhaps the only place in science where metaphysics is critical. It is a
self-evidently vacuous assertion to say that we cannot complete an uninterrupted
run if we constantly interrupt it to check on the state of the apparatus. It is a
quasi-religious belief throughout all of science that calibration allows us to make
certain assumptions without the need to check them: I have pressed this button
and I am confident that a beam of electrons is now on its way into an SG module.
It is on such a basis that when each run of the real experiment starts, we do
not check the beam any more but rely on classical counterfactuality: that what
we are doing in the beam preparation stage means what we think it means.
Classical counterfactuality is an article of faith in the laws of persistence and
consistency, and in a universe that does not play tricks on us by arbitrarily
changing its laws. It is implicit in the protocol of every experiment. The point
about Bell inequality experiments is that they show that on the quantum level,
classical counterfactuality is a false principle. That surely is enough to worry
anyone, because it means we do not really “understand” physical reality: it is
not completely described by a classical world view, only most of the time.

The relevance of classical counterfactuality to the SG experiment is as follows.

Suppose that the same, identical (at the HV level) beam had been sent into
the given SG apparatus but with one crucial difference: the direction of the
magnetic field had previously been altered by the observer and was now along
some different direction b and not a. According to classical counterfactuality, it is
legitimate to discuss both scenarios “simultaneously.” Of course, that is fanciful,
for it is not possible to go back and alter the past, as far as we know.
Classical counterfactuality and the principle of CM, however, allow us to
imagine the possibility of a different past. We are permitted to decompose Θ
as the union Θ = B + ∪ B − , where B + is the subset of HVs that would have sent
a beam particle into B1+ and B − is the subset of HVs that would have sent a
240 Bell Inequalities

beam particle into B1− , if the magnetic field axis had been b and not a. Here B1+
and B1− are the two regions on the detecting screen observed with module S b .
Classical counterfactuality allows us to go further and to make the decompo-
sition

Θ = (A+ ∩ B + ) ∪ (A+ ∩ B − ) ∪ (A− ∩ B + ) ∪ (A− ∩ B − ). (17.1)

This is a proposition rather different from the assertion Θ = A+ ∪A− = B + ∪B − :

we could at least perform each experiment S a and S b separately, at different
times, with magnetization direction well defined each time, whereas none of the
four terms on the right-hand side of (17.1) could be tested directly. For instance,
the first term, A+ ∩ B + , is the set of HV that would send a beam particle into
A+1 if the magnetization axis were along direction a, and would send the same
beam particle into B1+ if the magnetization axis were along direction b instead
of a.
In the real world it is physically not possible to have both directions a and b
in the same run.
That is not considered a problem by theorists who accept the CM paradigm,
but is a point of view inconsistent with Wheeler’s participatory principle, stated
in Chapter 1. It is on such points of interpretation and natural philosophy that
rest the irreconcilable differences between CM and QM. These very different,
incompatible visions of physical reality are what this debate is all about.

Counterfactual Probabilities
HV theorists address the problem of the meaning of the counterfactual intersec-
tion A+ ∩ B + by the following argument. Consider a very large number N a of
particles, prepared by Ted in a standard way, passed through S a . By counting the
number N (A+ 1 ) that land in A1 , we can estimate the probability PCM (A ) 
+ +
+ a + −
N (A1 )/N that a particle would land in A1 and not in A1 . Likewise, if the
axis were b, then we can estimate the probability PCM (B + )  N (B1+ )/N b that
a particle would land in B1+ and not in B1− .
Note that we are now discussing a probability measure on the set Θ of hidden
variables associated with a beam prepared in a standard way by Ted.
According to CM, the rules of classical probability apply, so the counterfactual
probability PCM (A+ ∩B + ) that a particle would land in A+ 1 if the magnetization
axis were a, but would have landed in B1+ if the magnetization axis were b, is
given by the rule

PCM (A+ ∩ B + ) = PCM (A+ )PCM (B + ). (17.2)

HV theorists simply cannot escape this assertion, unless they are prepared to
introduce novel possibilities such as nonlocality, contextual effects, and such like,
and these generally make their line of argument unappealing.
The problem is, we cannot do any experiments to measure PCM (A+ ∩ B + )
directly, or any of the other counterfactual probabilities directly, because as
 17.3 Circumventing the Simultaneity Problem 241

stated, only one magnetization axis can be set up at a time. We shall call this
the simultaneity problem.

17.3 Circumventing the Simultaneity Problem

An ingenious way of getting around the simultaneity problem is to use two
electrons or photons at a time, as follows.

Spin-Zero Two-Electron States

In both CM and QM, electrons are imagined as having an internal degree of
freedom called spin, associated with angular momentum. There is great evidence
for the empirical validity of this idea.
CM and QM treat spin differently, however: the former puts no restriction on
the direction of an electron’s spin axis or the magnitude of the electron’s angular
momentum, whereas the latter requires external context (that is, the parameters
associated with the apparatus being used) and predicts the quantization of
angular momentum in units of , the reduced Planck’s constant.10 Despite their
fundamental differences, however, both CM and QM theorists are happy to
discuss a two-electron state with total angular momentum of zero.
This means the following in both paradigms: if a spin-zero state of two electrons
is prepared and subsequently one of the electrons is passed through S a and
observed to land in region A+ 1 , then if the other electron was passed through
a
another, identical11 apparatus S , then that second electron would land in region
− +
A1 and not in A1 . Likewise, for every electron that landed in A− 1 , its partner
+
electron would land in A1 .
There is a useful fact about electrons that we can exploit: electrons repel each
other. If we fired a beam of two-electron states along a given direction, we should
expect that beam to spread out due to this repulsion. This would then allow us
to make observations on separate electrons.

We are in position now to circumvent the simultaneity problem as follows.

Alice, Bob, and Ted

Imagine now three experimentalists, Alice, Bob, and Ted, with apparatus shown
schematically in Figure 17.2(a). We will call this the enhanced Stern–Gerlach
a
(eSG) experiment. In eSG, Bob has an identical copy S of Alice’s SG module
S a , except Bob’s module is displaced in the laboratory so as not to overlap Alice’s
module in their common physical space P3 (Λ).
By stage Σ0 , Ted has prepared a beam of two-electron, spin-zero states. By
stage Σ1 , the beam has split by electric repulsion into subbeams 11 and 21 .

10
The angular momentum of an electron state is generally discussed in terms of /2.
11
Empirically identical, apart from being spatially displaced, so not identical in the sense of
Leibniz.
242 Bell Inequalities

A A

S S

A A
T T
A B

S S

A B

Figure 17.2. The enhanced SG experiment on two-electron, spinless states.

Subbeam 11 then enters Alice’s SG module S a , while subbeam 21 enters Bob’s

a
module S . By conservation of angular momentum (which holds in both CM and
+
QM), whenever Bob observes an electron in region A2 he can be sure that Alice
has observed her electron in region A−
2 , and so on.

It is important to appreciate that this statement is not about statistics; it

is about what happens in each individual run. The conservation of angular
momentum (as well as that of electric charge, energy, linear momentum, and
other classically motivated conserved quantities) takes place at the emergent,
process time level of observation. It is an extraordinary and deep fact that in QM,
there are classical concepts that can be relied on. The neutrino was discovered
precisely because the conservation of energy applies at the individual-run level in
beta decay, as proposed by Pauli in 1930, not at the statistical level, as proposed
by Bohr.12

Now we come to the circumvention of simultaneity problem. The trick here is

that Bob need not set the main magnetic field axis in his module to be in the
same direction a as that set by Alice in her module S a . Suppose Bob sets his

12
At the time, beta decay experiments could not account for a discrepancy between the
energy going into a beta decay process and that going out. Pauli proposed that there was
an unobserved particle, now known as the neutrino and subsequently detected by Cowan,
Reines, and collaborators in 1956 (Cowan et al., 1956), carrying off the energy discrepancy.
 17.3 Circumventing the Simultaneity Problem 243

b
field axis to some different direction b (we will now refer to his module as S
a
rather than S ). The relevant stage diagram is now Figure 17.2(b). Then we can
give an operational definition of PrCM (A+ ∩ B + ) and the other counterfactual
probabilities as follows.
Consider a large number N of runs in the eSG experiment. Suppose Alice
and Bob count their respective outcomes, recording the time of each. Afterward,
they come together and compare data sets and times. Then for every outcome in
−
which Alice found her electron in A+ 2 and Bob found his electron in B 2 , in the
same run, then that is counted as originating from the subset A+ ∩ B + , as if all
that information came from the original DS experiment, with a single electron.
Essentially, the second electron observed by Bob is like a ghost version of the
electron observed by Alice, and both are observed in the same run. Of course,
care is taken to take the opposite spins into account. Given the total count, an
estimate for the counterfactual probability PrCM (A+ ∩ B + ) follows immediately.
With this procedure, it is clear that total probability is conserved, that is, we
have

PrCM (A+ ∩ B + ) + PrCM (A+ ∩ B − ) + PrCM (A− ∩ B + ) + PrCM (A− ∩ B − ) = 1.

(17.3)

Bob & Carol & Ted & Alice

To derive the appropriate Bell inequality, the analysis requires the above dis-
cussion to be extended to a third direction, c, with observer Carol, giving the
decomposition Θ = C + ∩ C − , where C + is that subset of Θ that would send an
electron into C + , and similarly for C − . Classical counterfactuality also entitles us
to make the decomposition

Θ = (A+ ∩ B + ∩ C + ) ∪ (A+ ∩ B + ∩ C − ) ∪ (A+ ∩ B − ∩ C + ) ∪

(A+ ∩ B − ∩ C − ) ∪ (A− ∩ B + ∩ C + ) ∪ (A− ∩ B + ∩ C − ) ∪
(A− ∩ B − ∩ C + ) ∪ (A− ∩ B − ∩ C − ). (17.4)

A Bell Inequality
We are now in position to construct a Bell inequality for this experiment. The
first step is to direct the discussion from the set Θ of HV (which by definition we
have no knowledge about), to outcome probabilities, for which we have empirical
data. Suppose we passed a beam with a large number N of electrons through our
SG device, and repeated that run a large number of times. By counting electron
impacts on the detector screen, we would then determine outcome frequencies,
and finally we would be in position to discuss probabilities.
Given a probability measure P over Θ, then for any subsets A, B of Θ, we
have the rule

P (A ∪ B) = P (A) + P (B) − P (A ∩ B). (17.5)

244 Bell Inequalities

With this rule and what we know about the subsets involved, we can deduce the
following relations:

PCM (A+ ∩ B − ) = PCM (A+ ∩ B − ∩ C + ) + PCM (A+ ∩ B − ∩ C − ), (17.6)

− − − −
PCM (B ∩ C ) = PCM (A ∩ B ∩ C ) + PCM (A ∩ B ∩ C ),
+ + + +
(17.7)
− − − −
PCM (A ∩ C ) = PCM (A ∩ B ∩ C ) + PCM (A ∩ B ∩ C ).
+ + + +
(17.8)

Every term in these equations is a probability, so is nonnegative. Adding (17.6)

to (17.7) and subtracting (17.8), we exploit this nonnegativity directly to find
the Bell inequality

PCM (A+ ∩ B − ) + PCM (B + ∩ C − )  PCM (A+ ∩ C − ). (17.9)

This and similar inequalities are the focus of interest in numerous experiments.

Exercise 17.3 Use (17.5) and other relevant information to prove (17.6),
(17.7), and (17.8). Hence prove (17.9).

It has to be pointed out that there are more serious questions about this
inequality than those we raised about PCM (A+ ∩ B + ). We used two-spin states
to circumvent that latter problem, but that approach cannot deal with three
simultaneous magnetization axes. Essentially, we have to perform separate subex-
periments to determine the empirical values PEM P (A+ ∩ B − ), PEM P (B + ∩
C − ), and PEM P (A+ ∩ C − ) separately, ensuring that standardization across all
three subexperiments makes the test of the inequality (17.9) beyond reasonable
doubt.
We note in passing that (17.9) looks like the triangle inequality d(a, b) +
d(b, c)  d(a, c) defined for a metric space, where d(a, b) is the “distance” between
elements a and b of the metric space.

17.4 The Standard Quantum Calculation

We give now a brief account of the standard QM calculation used to test the
inequality (17.9). We will use the nonrelativistic Pauli electron theory, but taking
account only of the internal spin degree of freedom of each particle, assumed
to be an electron. This is modeled by a qubit Hilbert space Q, no different
in mathematical structure from the qubits used to construct QDN quantum
registers.

Calibration of Single Electron States

With reference to Figure 17.1, consider an uncalibrated beam of electrons sent by
Ted into an SG calibration module S k where k = (0, 0, 1) in standard Cartesian
coordinates defined previously by Ted. As discussed above, this beam will split
into two subbeams, denoted K1+ and K1− . An electron state entering K1+ will be
 17.4 The Standard Quantum Calculation 245

denoted | + k, while one entering K1+ will be denoted | − k. These two states
form a calibration basis for Q and satisfy the orthonormality conditions
+k| + k = −k| − k = 1, +k| − k = 0. (17.10)
In contrast to QDN, standard QM tends to work with a fixed Hilbert space over
any given run. Neither of these approaches is incorrect: they encode the same
information differently. We will use the calibration basis as a reference to describe
all state vectors and other bases in the following discussion.
Having calibrated his module T , Ted now uses it to create a beam of normalized
particle states, |Ψ, at stage Σ0 , given by
|Ψ = u| + k + v| − k, (17.11)
where |u|2 + |v|2 = 1.13
With reference to Figure 17.1, now suppose a beam of particles represented by
such a state is subsequently sent through SG module S a , where the main mag-
netic field direction a is given by a = (sin θ cos ψ, sin θ sin ψ, cos θ), where θ and
ψ are standard spherical polar coordinates relative to the standard Cartesians
referred to hitherto. Any beam passing through S a will in turn be split into two
−
components, denoted A+ 1 and A1 , as discussed above. In standard QM, each
of these components will be associated with orthogonal, normalized quantum
outcome states | + a and | − a, respectively, and these can be used to form an
orthonormal preferred basis for Q, associated with S a .
Standard quantum theory gives the following relations between the calibration
basis {| + k, | − k} and the preferred basis {| + a, | − a} (the outcome states),
up to inessential arbitrary overall phase factors:
sin θ sin θ
| + k = √ | + a + √ | − a,
2 − 2 cos θ 2 + 2 cos θ
(17.12)
(1 − cos θ) (1 + cos θ)
| − k = √ | + a − √ | − a.
2 − 2 cos θ 2 + 2 cos θ
From this we can find the amplitude A(+a| + k) ≡ +a| + k for outcome
state | + a given initial state |+k, and so on. Hence we can find the conditional
probabilities. We find, for example,
Pr(+a| + k) ≡ |A(+a| + k)|2 = cos2 ( 12 θ). (17.13)

Two-Spin States
Disregarding all factors inessential to the present discussion, a normalized two
half-spin state |Φ of zero total spin zero is given by
|Φ = √1 |
2
+ ka  ⊗ | − kb  − √1 |
2
− ka  ⊗ | + kb , (17.14)

13
The reader will appreciate by now how difficult it is to describe such a process without
using suggestive, misleading language.
246 Bell Inequalities

where the superscripts label the two spin-half particles.14 We now imagine that
particle a is sent through Alice’s module S a and particle b is sent through Bob’s
module S b .
Using (17.12), we may write
| + ka  = αa | + a + β a | − a, | + kb  = αb | + b + β b | − b,
(17.15)
| − ka  = γ a | + a + δ a | − a, | − kb  = γ b | + b + δ b | − b,
where again, superscripts label particles and
sin θa sin θb
αa ≡ √ , αb ≡ √ , (17.16)
2 − 2 cos θa 2 − 2 cos θb
and so on. Note that here {| + a, | − a} is a preferred basis for Qa , the qubit
associated with Alice’s electron, and {| + b, | − b} is a preferred basis for Qb ,
the qubit associated with Bob’s electron.
Given (17.15), we readily find
{(αa γ b − γ a αb )| + a ⊗ | + b + (αa δ b − γ a β b )| + a ⊗ | − b
|Φ = √1 .
2 +(β a γ b − δ a αb )| − a ⊗ | + b + (β a δ b − δ a β b )| − a ⊗ | − b}
(17.17)

The Bell Inequality

Now recall that the focus of attention here is the classical Bell inequality (17.9).
Considering the CM single particle counterfactual probability PCM (A+ , B − ), this
translates into the QM outcome probability PQM (+a, +b|Φ) in the extended SG
experiment (involving both Alice and Bob). This means that the classical Bell
inequality (17.9) is replaced by the assertion that

PQM (+a, +b|Φ) + PQM (+b, +c|Φ) − PQM (+a, +c|Φ)  0. (17.18)

The first term in this expression is just the squared modulus of the coefficient of
the tensor product term | + a ⊗ | + b in (17.17). Hence we deduce

PQM (+a, +b|Φ) = 12 |αa γ b − γ a αb |2 . (17.19)

We can simplify this expression by using rotational symmetry, orienting our

Cartesian coordinates along the direction of vector a. Then we find

PQM (+a, +b|Φ) = 1

2 sin2 ( 12 θab ), (17.20)

where θab is the angle between a and b. A similar calculation for the other two
terms in (17.15) gives

sin2 ( 12 θab ) + sin2 ( 12 θbc ) − sin2 ( 12 θac )  0. (17.21)

14
We can ignore the fact that electrons are identical and obey Fermi–Dirac statistics,
because the spreading of the beam prior to the electrons entering either SG(a) or else
SG(b) has introduced a form of classical labelling.
 17.5 The QDN Calculation 247

q bc
qab

Figure 17.3. Plot of the function f (θab , θbc ) over a suitable domain.

The problem is that we can find vectors a, b, and c for which (17.21) is wrong.
For example, take these vectors to lie in a plane, with θab = π/3, θbc = π/3,
θac = θab + θbc = 2π/3. Then

sin2 ( 12 π3 ) + sin2 ( 12 π3 ) − sin2 ( 12 2π

3 )=
1
4 + 1
4 − 3
4 = − 41 . (17.22)

In Figure 17.3 we show a plot of the function f (θab , θbc ) ≡ sin2 ( 12 θab ) +
sin2 ( 12 θbc ) − sin2 ( 12 θab + 12 θbc ) over a range of possibilities. There are two distinct
regions where the function value is negative, while the HV calculation predicts
that there should be no such regions.
There have been many experiments related to the one discussed here that
have shown violations of Bell’s inequalities, a frequently quoted one being that
of Aspect and others using photons (Aspect et al., 1982).
There is now not much doubt among the majority of physicists that classical
counterfactuality has been shown empirically to be a false principle in physics.
It remains an excellent principle as far as relative external context (the wider
Universe) is concerned: we can usually go to work and remain confident that our
house will still be there when we get back in the evening.
There remains a relatively small group of committed HV theorists who continue
to probe this issue and have come up with classically based possible explanations
for the observed empirical violations of Bell’s inequalities. Experimentalists con-
tinue to test these loopholes, and have reached the point where the HV classically
motivated “explanations” seem more unpalatable than the quantum theory they
are trying to circumvent.

17.5 The QDN Calculation

In this section we apply QDN to the eSG scenario discussed above. The relevant
stage diagram is Figure 17.4.
248 Bell Inequalities

Sa Sb

Figure 17.4. The QDN stage diagram of the enhanced SG experiment.

It was found that the total state at stage Σ1 , corresponding to the entangled
state (17.14) needs to be described as a state in the tensor product of a four-
dimensional internal spin space and a rank-four quantum register.

Stage Σ0
The preparation switch stage Σ0 creates a beam of spin-zero, two electron states,
 1 00 . Here, |s1  represents a normal-
represented by the total state |Ψ0 ) ≡ |s10  ⊗ A 0 0
ized state in the one-dimensional Hilbert space of two-electron spin zero states.

Stage Σ0 to Stage Σ1
By the first stage, Σ1 , the beam has split into two entangled spin-half subbeams.
This stage is before any of the subbeams enter their respective SG apparatus.
Bitification requires each of these sub-beams to be associated with two signal
qubits,15 so we require a rank-four quantum register at that stage, as stated
above.
The dynamics is given by the rule
 
  1 | + k a
 ⊗ | − k b
 ⊗  1A
A  3−
1  00 = √
U1,0 |s0  ⊗ A1 1 1 1 1
01 . (17.23)
0
2 | − k1  ⊗ | + k1  ⊗ A
a b  2A4
1 1

Stage Σ1 to Stage Σ2
Using (17.15) and from Figure 17.4 we have
   
U2,1 | + ka1  ⊗ | − kb1  ⊗ A  3 01 = α a | + a 2  ⊗ A
 1A  1 + β a | − a2  ⊗ A
2
1 1 2 2
 
b  3 b 
× γ | + b2  ⊗ A2 + δ | − b2  ⊗ A42 02 ,
   
U2,1 | − ka1  ⊗ | + kb1  ⊗ A  4 01 = γ a | + a 2  ⊗ A
 2A  1 + δ a | − a2  ⊗ A
2
1 1 2 2
 
b 
× α | + b2  ⊗ A2 + β | − b2  ⊗ A
3 b  4 02 .
2

(17.24)

15
This is on account of the fact that entanglement can be detected, given the right apparatus.
 17.5 The QDN Calculation 249

This is all that is needed to run our computer algebra program MAIN, which
gives us much more information than we need here. Our interest originated in
PCM (+a, −b). Context tells us that we need to calculate the probability that
detectors 12 and 32 are each in their respective signal state at stage Σ2 . Program
MAIN gives us the answer
 1A
Pr(A 3
2 2 02 |Ψ0 ) = 2 |α γ − γ α | ,
1 a b a b 2
(17.25)

which is precisely (17.19), the result of the standard QM calculation. The same
argument applies to the other terms in the Bell inequality (17.18).
Our conclusion is that QDN gives the same results as standard QM, and hence
the same prediction of violations of Bell inequalities.
 18
Change and Persistence

18.1 Introduction
Our interest in this chapter is in change, which is another way of discussing
persistence, or the apparent endurance over intervals of time of spatially extended
structures.
Persistence is the phenomenon that underpins all human activity. Given the
rule that we may think of as the first law of time, the dictum of Heraclitus
that everything changes, persistence is our name for those remarkable processes
that appear to circumvent that law and give stability to our lives. We should be
interested, for instance, in the fact that when we wake up each morning, we feel
that we are the same individuals who went to sleep the previous night. Indeed,
without persistence in one form or another, nothing would make sense, including
logic, rational thought, and mathematics, for these depend on comparisons of
standards and rules that persist in memory with information that we acquire in
process time.
Persistence is necessary for physics to make sense. It is emphatically not a
metaphysical topic but of the greatest relevance to science, including quantum
mechanics (QM). Persistence is implicit in Wheeler’s dictum that only acts of
observation are meaningful in physics, because observers and their apparatus
have to endure long enough to make observations. In quantized detector networks
(QDN), the persistence of observers and their apparatus long enough to perform
experiments should be regarded as axiomatic.
The subject of persistence is a deep and complex one, and will probably never
be fully understood. Several reasons contribute to this.

Observers Themselves are Subject to Change

Contrary to the general implicit belief that the laws of physics transcend the
subjective behavior of observers, it is our thesis that this is a vacuous proposition.
It cannot be proved; it is just an assertion, albeit a very useful one. A more
 18.1 Introduction 251

pragmatic view is the one held throughout this book, that the laws of physics are
contextual to observers. Without observers, science means nothing in a literal
sense. But according to the first law of time, observers themselves change over
time. This includes their memories and the conditioning that they have; these
are carried over from one stage to another. If the processes transporting those
memories and belief structures change those memories and belief structures,
then the laws of physics as understood by the observers at the time change as
well. Five hundred years ago, science knew little of Newton’s laws of motion; two
hundred years ago, science did not know that there was such a thing as quantum
mechanics.
We deal with the variability of observers by recognizing that truth values of
propositions, including the laws of physics, are contextual to observers at the
time of observation only. That the laws of physics themselves may change as the
Universe expands is not a fanciful idea: the possibility that the so-called constants
of nature have changed has been looked at by theorists such as Dirac (1938b)
and Magueijo (2003). A complete theory of observation would take into account
dynamical interaction between systems under observation, observers, and the
Universe that contains those observers.

Time Scales Are Important

Give the contextuality of persistence, what are the factors that contribute to that
contextuality? One of them is time scale, for the time scales over which objects
are said to persist is a crucial factor in creating the illusion. Indeed, persistence
may be thought of as a comparison of two time scales: the first is associated with
the SUOs being investigated, and the second is the time scale that the observer
regards as significant for their purposes.

Example 18.1 A scientist investigating the flight of insects needs to employ

video equipment that can adequately capture the motion of insect wings. There
will be no point in using a video frame speed of 25 frames per second if an
insect under observation flaps its wings 100 times per second.

Sheer Complexity
Another crucial factor associated with persistence is that observers and SUOs
are hideously complex phenomena. Indeed, the concepts of science are attempts
to organize that complexity into sufficiently simple forms that the brain can
interpret relatively quickly and easily.

Example 18.2 The human eye has a detecting screen known as the
retina, over which are detectors known as rods and cones. There are perhaps
120 million rods and 6 million cones in the typical human eye. When light from
suitable sources strikes the retina, many of these detectors will not register
252 Change and Persistence

a signal. Of those that do, complex chemical processes in the rods and cones
occur, resulting in signals being passed into specialized nerve layers below the
retina. These raw signals from the rods and cones are then processed in those
nerve layers before more complex signals are passed further on into the brain,
where much more complex processes of comparison and pattern recognition
must be occurring. The templates that the brain uses in these comparisons
appear to be relatively stable themselves, so much so that the brain does its
best to match the incoming signals to what it, the brain, has already prepared
and expected for. When this process fails in some way, optical illusions can
occur, such as a failure to recognize individuals whom we have met before.
The simplicity that we think we see around us is a total illusion.

Quantum State Fragility

A critical factor contributing to persistence is atomic stability: atoms are gener-
ally stable, at least long enough to create the illusion of persistence. This brings
us to perhaps the greatest enigma in physics, one that is necessary for persistence
in the first place and one that could not be resolved by classical mechanics (CM).
According to CM principles, an accelerating electric charge should dissipate
energy by electromagnetic radiation. This led to the classical prediction that
hydrogen atoms should be unstable, contrary to empirical evidence.
In order to bypass this prediction, Bohr constructed a model of hydrogen based
on an unexplained veto of such radiation (Bohr, 1913). More refined QM, such as
wave mechanics and quantum field theory, account for many structural aspects
of atoms, but a completely satisfactory, reductionist explanation of atomic sta-
bility does not exist. For instance, the postulate of normalized wave functions in
Schrödinger mechanics that leads to quantized atomic energy levels in hydrogen
is manifestly an emergent one based on an implicit assumption of a persistent
exophysical observer dealing only with normalized states. In interacting quantum
field theories too, there is generally no proof that the vacuum, or state of lowest
energy, exists; such a state is normally postulated.
The stability of atoms helps create the illusions of persistence. But below that
classical layer of illusion is a seething mass of change. Of relevance here is the
idea that the classical world view is good when quantum phases are so random
and disorganized that a classical average approach can be relied on. The greatest
developments in QM have occurred principally in the theoretical understanding
and empirical control of quantum phase. For example, Feynman’s path integral
formulation of QM is based on a specific mathematical recipe for adding quantum
phases associated with different dynamical paths. On the empirical side, the
various quantum optics experiments discussed in the present book demonstrate
the point excellently.

Our ambition to understand persistence will be limited. We shall discuss the

persistence of labstates in a carefully controlled environment. Specifically, given
 18.1 Introduction 253

an initial labstate Ψm at stage Σm , suppose it is allowed to evolve through

a quantized detector network to labstate Ψn at a later1 stage Σn , under the
evolution given by Ψn = Un,m Ψm . A natural question to ask is: how different
are states Ψm and Ψn ?

This question is not simple to answer, for several reasons.

The Parallel Transport Problem

An analogous problem arises in general relativity (GR), where there is a need
to make a comparison of directions at different places. Specifically, given two
points P and Q in a GR spacetime manifold, there is a tangent space associated
with each point, denoted TP for the tangent space at P , and TQ for that at Q.
Suppose we pick a vector v P in TP and vector v Q in TQ . The question is: how can
we compare v P and v Q ? The problem is that these are vectors in two different
vector spaces. Somehow, a method of “bringing them together” has to be found,
so that a proper comparison can be made.
A solution to this problem was found by the mathematicians Christoffel, Levi-
Civita, and others, who developed Riemannian (differential) geometry in the
nineteenth and early twentieth centuries. Their approach was to introduce the
notion of a connection, a rule for relating basis vectors in TP to those in TQ .
Such a rule involves parallel transport, which means defining what it means to
transport a vector from one tangent space to another in a manifold “without
changing direction.”
There is no unique connection over a manifold. Fortunately for Einstein, who
from about 1911 onward was developing GR, there is a natural and unique con-
nection for any manifold that has a metric, or distance, rule. The so-called metric
connection can be derived from the metric tensor of the spacetime concerned,
and GR deals specifically with manifolds with Lorentzian signature metrics. The
metric connection in GR has been found over the last century to give excellent
empirical predictions, such as the precession of the orbit of the planet Mercury,
the gravitational deflection of light, and galactic lensing. Accordingly, it is the
connection in general use to this day, although alternatives such as those with
torsion are occasionally proposed.
In QDN, an important feature that assists us here is that empirical context
gives a preferred basis Bm for the quantum register Qm at stage Σm and likewise
for stage Σn . Therefore, the parallel transport problem seems to have a natural
resolution.
However, QDN does not insist on the constancy of quantum register dimensions
from stage to stage. When discussing persistence, we need to ensure that the
dimensions of the registers concerned are the same, that is, dim Qn+1 = dim Qn .
Otherwise, we would be trying to relate labstates in vector spaces that were

1
In this context, stage Σn later than stage Σm always means n > m.
254 Change and Persistence

inherently dissimilar. This dimensionality issue does not arise in the case of
Riemannian geometry because the dimension of each tangent space in a manifold
is precisely the same as the dimension of the manifold over which they are defined.
We shall henceforth assume that when an observer is measuring the change
in a state of an SUO, it will be under carefully controlled conditions such that
dim Qn = dim Qm ≡ r. For instance, we might be interested in some branch
of macroscopic quantum mechanics, such as superfluidity or superconductivity.
Changes in quantum register dimension in that context would correspond to
fluctuations in the numbers of electrons or other particles in the SUO, which
would be a manifest breakdown of persistence.

Do We Know What We Are Doing?

One possible reason for the difficulty we have in understanding persistence is that
perhaps we really don’t understand the problem. It may be the case that QM
is the wrong branch of physics needed to understand persistence. A comparison
of Ψm and Ψn is a comparison of quantum states: these do not represent a
reality that “exists” in a classical sense. As emphasized before in this book,
quantum states represent the statistics of empirical context and are not snapshots
of reality, while the persistence that we see around us is an illusion based on a
near instantaneous comparison of signals received on our retinas and patterns in
our memories. Whether QM is right way to discuss that process of comparison
is debatable. The assumption that it is leads to vacuous concepts such as “wave
functions for the Universe.”
This issue has long been discussed in QM, examples being Ehrenfest’s theorem
(Ehrenfest, 1927) and decoherence theory (Zurek, 2002). These are viewed by us
as attempts to bridge the gap known as the Heisenberg cut, from the relative
internal contextual (RIC) side where QM is good and the relative external
context (REC) where a classical description is good. We view these attempts as
misguided, because REC is the domain proper of emergent physics, and we should
not expect any reductionist approach to “explain” emergence. We note that
despite initial hopes that decoherence would “explain” why the world around us
appears classical, these expectations have not been met and have been criticized
(Kastner, 2016).

As in other situations, the resolution of our problem comes from looking at

what goes on in the laboratory, not in our theories. When all is said and done,
the only things that matter in physics are the signals received in our detectors.
That is where we can understand persistence and change. We shall continue our
investigation into persistence, therefore, along the lines of comparing changes in
labstate outcomes.

Correlation of Preferred Basis States

Given a computational basis representation (CBR) Bm ≡ {im : 0  i < 2r } for
Qm and a CBR Bn ≡ {in : 0  i < 2r } for Qn , we need to relate the individual
 18.2 Comparisons 255

elements of each preferred basis. Clearly, there will be no use attempting to

relate these basis states if the quantum registers have unrelated contexts. For
instance, if the stage Σm apparatus is observing photon signals and stage Σn is
observing electron signals, then there may be no natural relationships between
them. Therefore, the discussion requires us to correlate contexts.
Assuming this is the case, then since the labeling of CBR elements is arbitrary,
we can assume that the ordering of elements in each preferred basis is correlated
contextually. This means for instance that 0m is correlated to 0n , that 1m is
correlated to 1n , and so on.

The Null Evolution Operator

Given the above provisos, an important operator will be the null evolution
operator Nn,m , defined by
r
2 −1
Nn,m ≡ in im . (18.1)
i=0

This will have the critical role, rather like the metric connection in GR, of defining
a concept of parallel transport in QDN, as we see from the following argument.
Given an initial labstate
r
2 −1
Ψm ≡ Ψim im , (18.2)
i=0

we find
r
2 −1
Ψ̃n ≡ Nn,m Ψm = Ψim in , (18.3)
i=0

which is a carbon copy of Ψm but at stage Σn .

18.2 Comparisons
The discussion now reduces to a comparison of the naturally evolved labstate
Ψn ≡ Un,m Ψm and the persistent image state Ψ̃n ≡ Nn,m Ψm . The problem
is that there is no natural measure of difference between two quantum register
states in the same register, no measure of distance, that survives all criticism.
The following are some possibilities.

The Born Measure of Similarity

In standard QM Hilbert space theory, the inner product (Φ, Ψ) of two normalized
state vectors in the same Hilbert space has an empirical significance: if (Φ, Ψ) = 0,
the two states are regarded as totally different. More generally, the square mod-
ulus P (Φ|Ψ) ≡ |(Φ|Ψ)|2 of this amplitude has the Born interpretation as the
conditional probability of a positive answer if tested for state Φ given prepared
state Ψ.
256 Change and Persistence

We may, with some justification, refer to P (Φ|Ψ) as the Born measure of

similarity.
The problem with the Born measure of similarity is that it can be far too crude
for quantum register states, as the following example illustrates.

9
Example 18.3 Consider a rank-billion quantum register Q[10 ] . The states
Ψ≡A 2 . . . A
 1A  109 0 and Φ ≡ A
 999999999 A 2 . . . A
 1A  999999999 0 have zero inner
product, but differ in only one signal out of a billion. By any conventional,
heuristic measure of similarity, these two states would be regarded as very
similar though not identical, but the Born measure of similarity gives them as
totally different.

We conclude that the Born measure of similarity is not good enough, in the
context of this chapter and the next, as a measure of similarity.

The Hamming Measure of Dissimilarity

The problem of comparing two quantum register states in the same quantum
register has an interesting parallel in the world of computing, cryptography, and
information science. In those subjects, a frequent problem is to compare two
strings of symbols, such as S 1 ≡ αQ55{7 and S 2 ≡ aP 56[7. There are two possi-
ble cases: either the strings have the same number of elements (six in this case),
or they have different numbers of elements. We shall discuss only the equal case.

Equal-Length Strings
In his study of the transmission of information over telephone networks, Ham-
ming discussed the problem of identifying and then correcting errors in transmis-
sion (Hamming, 1950). This scenario is very much like a quantum experiment,
with essentially the same architecture, including an information void. There are
observers (the speaker and the listener ) and their preparation devices and final
state detectors (the telephones). The information void here consists of extensive
transmission lines and modules, such as telephone exchanges, signal amplifiers,
and so on.
Suppose S ≡ s1 s2 . . . sr is the message that is sent (the prepared state) and
T ≡ t1 t2 . . . tr is the message that is actually received (the outcome state). A
natural question is: how different are these two strings? Hamming devised a
geometrical method of quantifying the difference between two character strings
of equal length. He defined a distance, or metric, dH (S, T ) between S and T
according to the rule
dH (S, T ) ≡ number of matched pairs (si , ti ), i = 1, 2, . . . , r, for which si = ti .
(18.4)

Example 18.4 For S ≡ αQ55{7 and T ≡ aP 56[7 we find dH (S, T ) = 4,

as these strings of length six differ everywhere except in their third and sixth
elements.
 18.2 Comparisons 257

The interpretation of the Hamming distance between two equal length strings
is that it is the minimum number of single character replacements in one string
that would convert it into the other. In Example 18.4, we can convert S into T
by the four replacements α → a, Q → P , 5 → 6, and {→ [.

Exercise 18.5 Prove that the Hamming metric is a true metric in the sense
of a metric space.

Suppose we prepare a labstate Ψm in a rank-r quantum register Qm at stage

Σm , and pass it through an information void, until it is received as outcome Ψn
in an identical rank quantum register Qn at stage Σn , where n > m.2 The two
registers are copies of each other, having the same rank. Moreover, their preferred
bases are contextually correlated: computational basis element im means the
same to the speaker (the observer at stage Σm ) as in means to the listener (the
observer at stage Σn ). In other words, speaker and listener understand the same
language.
Taking account of the parallel transport problem discussed above, we need
to compare like with like. We choose to compare the persistent image Ψ̃n ≡
Nn,m Ψm in Qn of the original state with the transmitted state Ψn ≡ Un,m Ψm ,
also in Qn .
The problem is compounded here by superposition: neither Ψn nor Ψ̃n are
classical, but thay are complex superpositions of classical information (which is
essentially what preferred basis elements are). This is a factor that Hamming did
not face.
In the following, we drop the temporal index, as it does not play an essen-
tial role.

Hamming Distance between Preferred Basis Elements

In the simplest case, case, suppose Ψ = i and Ψ̃ = j, where i and j are
elements of the computational basis representation (CBR) for Q. Each basis
element carries classical information, that is, corresponds to a signal state that
could actually be observed in a run. We calculate the Hamming distance dH (i, j)
between these two elements by first converting each integer into its associated
binary string and then working out the Hamming distance using the rule given
in (18.4).
That means applying the process of binary decomposition. Specifically, we
write

i = i[1] + i[2] 2 + i[3] 22 + · · · + i[r] 2r−1 ,

j = j [1] + j [2] 2 + j [3] 22 + · · · + j [r] 2r−1 , (18.5)

2
Note that the transmitted state is not an actual outcome (which is classical information),
but a quantum state immediately before it is looked at for an outcome, and is therefore best
thought of as a different form of information, referred to by us as quantum information.
258 Change and Persistence

Table 18.1 The Hamming distance between elements of a

rank-three quantum register basis

000 100 010 110 001 101 011 111

0 ≡ 000 0 1 1 2 1 2 2 3
1 ≡ 100 1 0 2 1 2 1 3 2
2 ≡ 010 1 2 0 1 2 3 1 2
3 ≡ 110 2 1 1 0 3 2 2 1
4 ≡ 001 1 2 2 3 0 1 1 2
5 ≡ 101 2 1 3 2 1 0 2 1
6 ≡ 011 2 3 1 2 1 2 0 1
7 ≡ 111 3 2 2 1 2 1 1 0

where the coefficients i[k] , j [k] , k = 1, 2, . . . , r, are each either zero or one. Next,
we form the strings S i ≡ i[1] i[2] . . . i[r] , S j ≡ j [1] j [2] . . . j [r] . Finally, we calculate
the Hamming distance dH (S i , S j ).
Table 18.1 gives the Hamming distance between pairs of elements in the
preferred basis for a rank-three quantum register.
We note the following:
1. The maximally saturated state 7 ≡ 111 is furthest away in Hamming distance
terms from the ground state 0 ≡ 111,
2. All elements of the same signality σ are a Hamming distance σ from the signal
ground state.
3. For any two register basis elements, the Hamming distance between them
defines what can be thought of as a relative signality: if any one of these
elements were chosen to be the new ground state, then the other state would
have signality equal to its Hamming distance from that new ground state. This
underlines the fact that the ground state in a QDN register is not intrinsic to
the apparatus but defined contextually by the observer. The choice of signal
ground state is not dictated by lowest energy state.
The power of Hamming’s approach is that his distance rule is a genuine metric,
so that all the theorems of metric spaces can be applied. Lest this appear trivial,
we should consider that quantum registers modeling real situations may have
immense rank, such as that modeling a superconducting quantum interferometer.
In such situations, the Hamming metric distance approach may well prove useful.
The above approach is classical, in that it covers classical register states com-
pletely. QM, however, involves superpositions of register basis states, so we are
faced with the problem of defining the equivalent of a Hamming distance between
arbitrary, normalized quantum register states in the same register.

The Hamming Operator

For a given rank-r quantum register Q with CBR {i : i = 0, 1, 2, . . . , 2r − 1} we
define the Hamming operator H by
iHj ≡ dH (i, j), (18.6)
 18.3 Signal Correlation Measure of Change 259

where dH (i, j) is the Hamming distance between elements i and j. Then using
completeness, we have the representation
r
2 −1
H= idH (i, j)j. (18.7)
i,j=0

We would in the first instance like to use this operator to attempt several
definitions of “distance” between more general quantum register states. Examples
are the following.

Quantum Hamming Measure of Dissimilarity

2r −1 2r −1
Given the CBR expansions Φ ≡ i=0 Φi i and Ψ ≡ i=0 Ψi i we define the
quantum Hamming measure of dissimilarity H1 (Φ, Ψ) as the magnitude of the
matrix element ΦHΨ, that is,
 r 
 2 
 −1 i∗ 

H1 (Φ, Ψ) ≡  j
Φ dH (i, j)Ψ  . (18.8)
i,j=0 

A variant proposal is to define the quantum Hamming measure of dissimilarity

H2 as
r
2 −1 r
2 −1
H2 (Φ, Ψ) ≡ Pr(i|Φ)dH (i, j) Pr(j|Ψ) = |Φi |2 dH (i, j)|Ψj |2 . (18.9)
i,j=0 i,j=0

Neither (18.8) nor (18.9) is a genuine metric, because they do not return zero
in general when Φ = Ψ. However, that is not necessarily a bad thing. After all,
quantum states are not objective things. The fact that neither H1 (Ψ, Ψ) nor
H2 (Ψ, Ψ) is not zero in general reflects the fact that labstates are not classical.
If they were, they would be represented by a single element of the CBR, and in
that case, both measures of dissimilarity vanish.

18.3 Signal Correlation Measure of Change

Suppose an observer is able to prepare and investigate, separately, two labstates
Ψ and Φ in a rank-r quantum register Q. This means that the observer can exam-
ine each of these labstates separately and measure the answer to any maximal or
partial question about either of these states. Another way of saying this is that
the observer can determine, empirically, the frequencies of finding either labstate
in a given preferred basis state. In reality, that is all that an observer can do.
With such information, the observer can meaningfully discuss differences in
signal outcome probabilities, and it is on that basis that we now proceed. First
we review briefly the concept of correlation.

Correlations in Statistics
The problem of comparing two data sets occurs frequently in statistics. Suppose
we have two sets of data, X and Y , each consisting of n real numbers, such that
260 Change and Persistence

the ith element of X is xi and the ith element of Y is y i . The question often
arises: how “close” are the two sequences {x1 , x2 , . . . , xn } and {y 1 , y 2 , . . . , y n }?
This question is a variant of the one originally addressed by Hamming. Statisti-
cians have developed the concept of correlation in a variety of forms that attempt
to give a number, usually between −1 and +1, that gives the degree of similarity,
dependence, or correspondence between the two data sets (or sequences). A
well-known correlation coefficient is the Pearson correlation coefficient, which
in the case we are discussing is known as a sample correlation coefficient CX,Y
defined by
n
(xi − x)(y i − y)
CX,Y ≡ / i=1 n , (18.10)
n
j=1 (x − x) k=1 (y − y)
j 2 k 2

where x and y are the respective sample averages. If a sample correlation coeffi-
cient has value +1, 0, or −1, the two samples are said to be perfectly correlated,
uncorrelated, or perfectly anticorrelated, respectively.

Detector Correlations
In our case, our interest is directly related to detector physics. The “data” values
in which we are interested, that is, the analogues of the xi and y i values in the
above, are truth values, suitably modified to give correlations. What greatly helps
here is that the values are binary. Suppose we look at two detectors and compare
their signal status. If they both register a signal, or if they both register ground
(no signal), then we can say they are perfectly correlated. Otherwise, they are
perfectly anticorrelated.
More commonly, we might take many such joint readings. Suppose the prob-
ability of finding perfect correlation is p. Therefore, the probability of anticorre-
lation is 1 − p. Assigning a value +1 for every case of perfect correlation and a
value of −1 for every case of perfect anticorrelation, we deduce that the average
correlation C will be given by C = 2p − 1.

Single Qubit Temporal Correlation

Quantum registers in practice may be vast, and the description of change and
persistence of labstates will be accordingly complicated. It is reasonable to
focus attention on the simplest SUO first, in order to appreciate what we are
faced with.
Consider an experiment that is dealing with an SUO represented by a single
qubit. Suppose that, by stage Σ0 , the observer has prepared a labstate Ψ0 in the
rank-one quantum register Q0 given by the CBR expression
Ψ0 ≡ α00 + β10 , (18.11)
where |α|2 + |β|2 = 1.
Now suppose the labstate Ψ0 is allowed to evolve undisturbed to stage Σ1
under semi-unitary evolution, into labstate Ψ1 in a rank-one quantum register
 18.3 Signal Correlation Measure of Change 261

Q1 . Then we write Ψ1 ≡ U1,0 Ψ0 , where the evolution operator U1,0 is semi-

unitary. Taking the matrix representation of this operator in standard form,
shown in Eq. (11.26), with the phase E set to zero, we have the rules

U1,0 00 = a01 − b∗ 11 ,
U1,0 10 = b01 + a∗ 11 , (18.12)

where |a|2 + |b|2 = 1.

We now come to a critical point, one that impinges on the quantum Hamming
measures of dissimilarity discussed above. Consider a persistent array of detectors
from stage Σm to stage Σn , where n > m, such that detector im at stage Σm is
regarded as the same as in at stage Σn . Given the labstates Ψm and Ψn , there
are two very different quantities that could be measured:

Correlation of Probabilities
i
Suppose the observer measured Pm , the probability of a positive signal in im
i
and, separately, Pn , the probability of a signal in in . These two quantities give
important information about what goes on at detector i at two separate stages.
But there is no direct correlation between those two pieces of information. Every
observation of im at stage Σm cannot possibly be influenced by what could
happen at in at stage Σn , and vice-versa, simply because each observation in
any given run necessarily precludes the other observation, in that run.
The argument goes as follows. If the observer looks at stage Σm to see what
the signal status of detector im is, then that stops the run and so in cannot be
investigated during that run. Likewise, if the observer decides to see what the
signal status is of detector in at stage Σn , that means they cannot have looked
at im at stage Σm , because that would have stopped that run immediately. In
brief, a comparison of the signal status im and in in any given run is ruled out.
i
The only information the observer has are the two probabilities Pm and Pni , and
these have been measured separately over many different runs.
If Pm
i
≡ 1 − Pmi
, Pni ≡ 1 − Pni , we define the correlation of probabilities Pn,mi

as follows:
i
Pn,m ≡ Pni Pm
i
+ Pni Pm
i
− Pni Pm
i
− Pni Pm
i

= (2Pni − 1)(2Pm
i
− 1) = Cni Cm
i
, (18.13)

where Cni and Cm

i
are detector i correlations at stage Σn and Σm , respectively.

Probability of Correlation
What the observer is really after is the probability of finding a signal at in
knowing for sure that there was a signal in im . This is then a genuine temporal
i
correlation, denoted Cn,m . In the following, we simplify the discussion by taking
the register to consist of just one detector.
262 Change and Persistence

Definition 18.6 Given initial state Ψm ≡ α0m +β1m , where |α|2 +|β|2 = 1,
and the evolution operator is Un,m , the temporal correlation Cn,m is given by

Cn,m ≡ Pr(0m |Ψm ) Pr(0n |0m ) + Pr(1m |Ψm ) Pr(1n |1m )

− Pr(1m |Ψm ) Pr(0n |1m ) − Pr(0m |Ψm ) Pr(1n |0m )
= 2 Pr(0m |Ψm ) Pr(0n |0m ) + 2 Pr(1m |Ψm ) Pr(1n |10 )n ,
= 2|α|2 |0n Un,m 0m |2 + 2|β|2 |1n Un,m 1m |2 − 1. (18.14)

We point out that the probabilities Pr(0m |Ψm ) ≡ |α|2 and Pr(1m |Ψm ) ≡ |β|2
are found during the calibration process, prior to the experiment proper starting.

From (18.12), we readily find that

Cn,m = 2|a|2 − 1, (18.15)

which, significantly, is independent of the components α, β, of the initial labstate.

This result makes sense: if the dynamics is trivial, in that the only effect is to
multiply the initial labstate by some complex phase factor, then |a| = 1 and the
correlation is perfect, that is, Cn,m = 1. If, conversely, the dynamics flips the
initial labstate to an orthogonal labstate, then the correlation is −1.
We will use the above result (18.15) in the next chapter on the Leggett–Garg
inequalities.
The generalization of these calculations to higher rank registers will undoubt-
edly be more complicated but, in principle, amenable to the same logic as above:
in all cases, the observer should decide what it is that they can actually measure,
and then the formalism will give unambiguous predictions.
A final point we need to make is that we expect our apparatus to be large rank
and calibrated. Then the observables will be (say) the up and down components
of spin, and interest will be in spin correlations. The QDN analysis in such a
case assigns two qubits, one for spin up and one for spin down. This effectively
doubles the size of the quantum register. If we labeled spin up by index 2i, then
spin down would be indexed by 2i + 1, and our temporal correlations would
involve signals in detectors 2i and 2i + 1.
 19
Temporal Correlations

19.1 Introduction
Our aim in this chapter is to extend the Bell inequality discussion in Chapter 17
to its temporal analogue, known as the Leggett–Garg (LG) inequality.
Bell inequalities involve spatial nonlocality, that is, signal observations dis-
tributed over space. It was shown by Leggett and Garg that analogous inequalities
involving temporal nonlocality can be formulated (Leggett and Garg, 1985). We
shall discuss one of these, known as the LG inequality.
We saw in Chapter 17 that Bell inequalities are based on certain classical
mechanics (CM) assumptions about the nature of reality. Likewise, the LG
inequality is based on two CM principles that are not incorporated into quantum
mechanics (QM).

Definition 19.1 The principle of macrorealism asserts that if a macroscopic

(that is, large-scale) system under observation (SUO) can be observed to be in
one of two or more macroscopically distinct states, then it will always be in one
or another of those states at any given time, not in a quantum superposition
of those states, even when it is not being observed.

This principle is a foundational principle in CM but is incompatible with QM

in at least two ways. First, it is vacuous, as it asserts the truth of a proposition
in the absence of empirical validation: how can we define the concept of always,
without introducing counterfactuality? Second, it is violated in quantum theory,
for instance, in path integral calculations.
This principle in embedded in the nexus of issues explored in this book. The
quantized detector network (QDN) version of it takes the form “If any SUO can
be observed in any number of possible states, then it will be observed in one of
them in any run.” This is a near tautology. The classical version says the same
thing but makes an additional assertion about something going on in the absence
of observation, which is the vacuous element that QM cannot accept.
264 Temporal Correlations

Definition 19.2 The principle of noninvasive measurability asserts that the

actual state an SUO is in can always be determined cost-free, that is, without
having any effect on that state or on its subsequent dynamical evolution.

This principle is subtle. We experience its apparent validity all the time in our
ordinary lives: we look at objects and they appear not to change by those acts
of observation. However, quantum physics tells us otherwise, for all information
comes to us via quantum processes, and those involve quanta.1 Newton’s third
law, action and reaction are equal and opposite and act on different bodies, really
does have its quantum counterpart: no observation leaves the observed completely
unchanged.

The LG inequality involves temporal correlations, which measure and compare

changes in observables. We shall focus our attention on the temporal correlations
of the signal states of dynamically evolving bits and qubits.

19.2 Classical Bit Temporal Correlations

Without loss of content or generality, we shall restrict the discussion in this
section to an SUO consisting of N interacting classical bits. In such discussions,
it is not necessary to say what these bits represent physically. What matters
is that if the observer chose to look at any one of these bits, that bit would be
observed to be in precisely one of two possible states. We will refer to these states
as up and down. A useful mental image is that the up state represents a raised
flag denoting a signal, while the down state represents a lowered flag denoting
an absence of a signal.
Consider an experiment where at initial stage Σ0 , all N of these bits are
definitely in the up state. We then allow the state of the SUO to evolve until
stage Σ1 and then we look at the state of the SUO at that stage.
Suppose that at stage Σ1 we find a total of N up bits in the up state and
N down ≡ N − N up bits in the down state. A natural question is: how different is
the state of the SUO at stage Σ1 compared with its state at Σ0 ?
There are many ways to answer this question, that is, to define temporal
change. Before we can do that, we need to clarify a metaphysical point, concerning
the notions of permanence and identity.

Change Is Contextual
Up to now, our exposition of QDN has emphasized that everything changes.
But on reflection, that is a vacuous inconsistency, for change can be measured

1
We have not been much concerned with Planck’s constant  so far in this book, but it
comes in here. Changes occurring to states of SUO when they are observed are quantified
by that unit of action, which is relatively negligible on our ordinary real-life (emergent)
scales of measurement.
 19.3 The Classical Leggett–Garg Inequality 265

only by observers. But those observers themselves change, according to their

own subjective experience. Therefore, any change in an SUO has to be carefully
distinguished from the changes going on naturally in the laboratory.
Leggett and Garg chose temporal correlations. We discussed these for single
qubits in the previous chapter.

Definition 19.3 If a classical bit is in the same state at stage Σn as it was

at an earlier stage Σm , then the states are said to be perfectly correlated, with
a correlation Cn,m = +1. If, conversely, a classical bit is in a different state at
stage Σn compared with its state at stage Σm , then the two states are said to
be perfectly anticorrelated, with a correlation Cn,m = −1.

Our interest will be in the average correlation C 1,0 . From the information given
above, we find
N up N down 2N up
C 1,0 = − = − 1. (19.1)
N N N
By inspection, it is easy to see that the average correlation satisfies the constraints
−1  C 1,0  1.

19.3 The Classical Leggett–Garg Inequality

The LG inequality is based on a three-stage experiment, as follows. An SUO
consisting of r classical bits, each in its up state at stage Σ0 , evolves to stage Σ1 ,
and the average correlation C 1,0 is measured. The state is then allowed to evolve
to stage Σ2 , where two new average correlations are measured. One of these is
C 2,1 and the other is C 2,0 .
We note that (1) each bit is followed from stage Σ0 to stage Σ1 and then
finally to stage Σ2 and its up or down state is observed noninvasively and
recorded at each stage, and (2) according to the CM principles of macrorealism
and noninvasive measurability, the observer will have all the information needed
to calculate these three ensemble average quantities precisely and without error.
Because the dynamics is assumed deterministic, only one run is required. If on
the other hand, the dynamics is classical stochastic, or the initial state is random,
then the argument can adjusted to take probabilities into account, with exactly
the same result.
Given these three correlations, we define the LG correlation K by

K ≡ C 1,0 + C 2,1 − C 2,0 . (19.2)

We derive the LG inequality −3  K  1 as follows.

Given an SUO of r bits, noninvasive measurability allows us to track any or all
of the bits in the SUO over the three stages, without interfering in the dynamics.
Suppose we tracked the ith bit in the SUO over the three stages Σ0 , Σ1 , and Σ2 .
266 Temporal Correlations

Table 19.1 Calculation of the possible values of the

LG correlation K i for the ith bit

signal status at Σ0 up up up up
signal status at Σ1 up up down down
signal status at Σ2 up down up down
i
C1,0 1 1 −1 −1
i
C2,1 1 −1 −1 1
i
C2,0 1 −1 1 −1
Ki 1 1 −3 1

Table 19.1 shows the possible up or down states of that bit at each stage, the
correlations associated with those bit states, and the LG correlation for that bit.
By inspection of Table 19.1, we see that the LG correlation K i ≡ C1,0 i i
+ C2,1
− C2,0 for the ith bit lies in the interval [−3, 1]. This is perfectly general, being
1

valid for any interaction whatsoever between the bits in that SUO. Since this is
true for any bit in the SUO, we conclude that the average K satisfies the same
condition. Hence we arrive at the LG condition
− 3  K  1. (19.3)
There is no way that the LG inequality could ever be violated classically, given
the principles of macrorealism and noninvasive measurability.

19.4 Qubit Temporal Correlations

Extending the above classical bit discussion to the quantum case involves sig-
nificant differences. Specifically, each bit in the ensemble is replaced by a qubit,
and then all of the qubits are tensored together to create a quantum register of
rank r.
As will be appreciated by now, such registers are exceedingly complicated
structures. We will assume in the first instance that the qubits in the quantum
register do not interact with each other, but can interact with other elements of
their environment. This assumption means that we can meaningfully discuss the
evolution of a single qubit. We shall follow the evolution of a typical single qubit
in a noninteracting ensemble from initial stage Σ0 to intermediate stage Σ1 and
then to final stage Σ2 .
We shall discuss the correlation C1,0 of two signal observations, at stage Σ0
and stage Σ1 . Assuming that the semi-unitary operator acting on the chosen
qubit is in the standard form
 
a1 −b∗1
U1,0 = , (19.4)
b1 a∗1
where |a1 |2 + |b1 |2 = 1 and we ignore an overall phase, we use the results of the
previous chapter to find
 19.5 QDN Spin Correlation 267

C0,1 = |a1 |2 − |b1 |2 = 2|a1 |2 − 1. (19.5)

As noted previously, this result is independent of the initial state of the qubit.
Now consider a further evolution from stage Σ1 to stage Σ2 with evolution
operator
 
a2 −b∗2
U2,1 = , (19.6)
b2 a∗2
2 2
where |a2 | + |b2 | = 1. Then C2,1 = 2|a2 |2 − 1.
According to QM principles, evolution from stage Σ0 to stage Σ2 is given by
the evolution operator U2,0 = U2,1 U0,1 . We find
    
a2 −b∗2 a1 −b∗1 a1 a2 − b1 b∗2 −a∗1 b∗2 − b∗1 a2
U2,0 = = , (19.7)
b2 a∗2 b1 a∗1 a1 b2 + b1 a∗2 a∗1 a∗2 − b∗1 b2
from which we deduce
C2,0 = 2|a1 a2 − b1 b∗2 |2 − 1. (19.8)

19.5 QDN Spin Correlation

In this section we show how QDN deals with spin correlation. The first thing
to note in the above calculation is that according to QDN principles, a single
persistent qubit representing a calibrated detector would show no changes in its
signal status, by definition. Because we wish to discuss an actual spin, such as
that of a proton, which would show changes in its signal status, bitification tells
us that we need two detector qubits, one for proton spin up and the other for
proton spin down. QDN qubits do not model angular momentum or spin per se
but yes/no logic.
Suppose then that we want to investigate temporal spin correlation for a single
proton. Following the discussion in previous sections, we will have a three-stage
process, represented by Figure 19.1.
The initial labstate Ψ0 is taken to be given by
 1 + βA
Ψ0 ≡ (αA  2 )00 , (19.9)
0 0

where 10 detects spin up and 20 detects spin down.

U U

Figure 19.1. Stage diagram for a proton temporal spin correlation.

268 Temporal Correlations

We encounter here an example of a contextual subspace. Our rank-two quan-

tum register has four basis states, but only two of these are needed to describe
the spin dynamics. For example, semi-unitary evolution from stage Σ0 to stage
Σ1 is given by
 1 00 = a 1 A
U1,0 A  1 0 1 − b∗ A
2
0 1 1 1 01 ,
 2 00 = b1 A
U1,0 A  1 01 + a∗ A2
0 1 1 1 01 , (19.10)
and similarly for evolution to stage Σ2 .
It is not hard to see that in this case, the required temporal correlation
coefficient C1,0 is given by C1,0 = 2|a1 |2 − 1, exactly as above. Our conclusion
is that, as elsewhere, QDN reproduces the standard QM results without any
difficulty. Note that the bitification process appears to make QDN quantum
registers unduly larger than the standard QM counterparts. That is not seen
as anything other than an advantage, because it permits the modeling of more
situations, such as labstates of signality greater than just one.

19.6 The Leggett–Garg Correlation

From the above correlations, using Definition 19.2 we find
K = 2|a1 |2 + 2|a2 |2 − 2|a1 a2 − b1 b∗2 |2 − 1. (19.11)
Now the parameters of an experiment, such as a1 , a2 , . . . are under the control
of the experimentalist. It is reasonable therefore to assume that the parameters
a1 , b1 , a2 , and b2 can be chosen to be whatever we wish, subject to the unitarity
constraints |a1 |2 + |b1 |2 = |a2 |2 + |b2 |2 = 1.
Consider the reparametrization
a1 = cos θ1 , b1 = sin θ1 eiφ1 ,
a2 = cos θ2 , b2 = sin θ2 eiφ2 , (19.12)
where θ1 , θ2 , φ1 , and φ2 are real. Now take φ1 = φ2 . Plotting K as θ1 and θ2
range from −π to π each gives Figure 19.2.
It is clear that there are values of the parameters where K exceeds the classical
limit +1. In fact, the maximum value of K over the region shown is 1.5, in
clear violation of the Leggett–Garg upper bound of one. This is an entirely non-
classical result that is the temporal analogue of the violation of Bell inequalities
in experiments such as those of Aspect (Aspect et al., 1982) and others.

19.7 Understanding the Leggett–Garg Prediction

The predicted QM violation of the Leggett–Garg inequality is another demonstra-
tion of the thesis running throughout this book: that empirical truth is contextual
and that QM is best regarded as a statement of the laws of entitlement, of what
we can legitimately say about SUOs, rather than an explicit description of those
SUOs.
 19.7 Understanding the Leggett–Garg Prediction 269

1.5

K
–1

–2

–3 –2
–3 –2
–1 0 2
0
q2
1
q1 2 3

Figure 19.2. Values of K versus angle parameters θ1 , θ2 .

In view of the great predictive success of QM, it would be unwise to believe

that we can replace it. At best, we can enhance it. But that does not answer
the question, what is the origin of the discrepancy between the classical LG
inequality and the predicted QM violation of it?
Our view is that there must be hidden assumptions somewhere that, if iden-
tified, would explain the discrepancy. Since QM has never yet been proved
empirically incorrect while classical mechanics has, there must be something
that we have missed in the way we think classically. It is not QM that is wrong,
but our preconditioned classical way of thinking about physical reality that is
not quite right.
The answer is given clearly by Wheeler’s participatory principle that if we have
not actually done something, we should not make any undue assumption about it.
With this in mind, consider the LG correlation K discussed above. It is calcu-
lated from three separate correlations that classically could be calculated from a
single run. This is because of the assumed principle of noninvasive measurability
that is one of the standard assumptions of classical mechanics.
But, according to the requirements of quantum experimentation, the corre-
lation C2,0 cannot involve any observation at stage Σ1 . The evolution operator
U2,0 has to be applied on the strict understanding that no attempt is made to
extract information between initial stage Σ0 and final stage Σ2 .
The essential fact is that the principle of noninvasive measurability cannot be
assumed to hold in QM (although as discussed in Chapter 25, there are some
experiments that can be described in such terms). Classically, our normal human
conditioning is to think automatically that the stage-Σ1 signals observed in the
measurement of the correlation C1,0 should play a role in C2,0 . But how can
they? They are not observed when the SUO evolves undisturbed from stage Σ0
to stage Σ2 .
In order to calculate the LG quantity K, we would have to perform three
separate “subexperiments,” one for each of the three correlations involved. These
270 Temporal Correlations

are separate experiments with separate contexts. Classically, we could believe

that we could get away with just one experiment. Quantum mechanically, we
have to recognize context is critical and perform three subexperiments.
The explanation then is that in the C2,0 subexperiment, there is a lack of which-
path information at stage time Σ1 , analogous to the architecture of the double-slit
experiment, where an interference pattern on a screen is observed provided no
attempt is made to determine through which slit the particle had gone.
 20
The Franson Experiment

20.1 Introduction
Experimental violations of the Bell and Leggett–Garg inequalities studied in
Chapters 17 and 19, respectively, show that quantum states of systems under
observation (SUOs) cannot be interpreted classically. Interpretations that claim
to do this require contextually incomplete modifications of classical principles,
such as a breakdown of classical locality. Such modifications merely serve to make
standard quantum mechanics (QM) more appealing.
However, that is only one side of the observer–SUO fence. In this chapter,
we start to explore the other side of that fence, where the observer and their
apparatus live. Our focus is a thought experiment proposed by Franson (Franson,
1989), which suggests that apparatus cannot always be treated classically. Three
scenarios are discussed, each involving different time scales, with corresponding
different outcomes.

20.2 The Franson Experiment

In the following, we refer to “photons” as if they were actual particles, but that
is only a convenient and intuitive way of describing what at the end of the day
are just clicks in detectors. There are experiments, for example, where it is not
reasonable to think of the sources of photons (such as certain crystals) as point
like (Paul, 2004).
Franson’s proposed experiment FRANSON has parameters that can alter out-
comes significantly. There are three significant choices of parameters, leading to
three distinct scenarios labeled FRANSON-1, FRANSON-2, and FRANSON-3.
The basic architecture consists of a coherent pair of photons, produced at a
localized source S, sent in opposite directions toward a pair of separated Mach–
Zehnder interferometers, as shown in Figure 20.1. Each photon passes through
its own interferometer and, depending on path taken, can suffer a change in
272 The Franson Experiment

M j M S M j M

B B B B

Figure 20.1. FRANSON-1: the Franson experiment for ΔT < τ2  τ1 . S is

the photon pair source, the B i are beam splitters, the M i are mirrors, and the
φi are phase changers.

phase φ and a time delay ΔT . The time delay ΔT is assumed the same for each
interferometer in any given run. The phase changes φ1 , φ2 associated with the
different interferometers can be altered by the observer, but are fixed before and
during each run.
The experiment hinges on the relationship between three characteristic times
(Franson, 1989).

Coherence Time τ1
In quantum optics experiments, a finite electromagnetic wave train of length L
moving at the speed of light c takes a time τ ≡ L/c to pass a given point. Such
a time is known as a coherence time. In the Franson experiment, the coherence
time τ1 is that associated with the production of the photon pair by stage Σ1 .
It is a characteristic of the photon pair source S and of the collimation proce-
dures applied subsequently. While τ1 cannot be altered, it can be determined
empirically. We shall assume that τ1 is the same for each photon.

Emission Time τ2
The second characteristic time is τ2 , the effective time interval within which
both photons in a pair can be said to have been emitted. This can be measured
during calibration by coincidence observations of detectors 13 and 23 with all
beam-splitters removed. It is assumed τ2 can be determined empirically and
that τ2  τ1 . This last inequality is crucial to FRANSON because when this
inequality holds, the observer has no way of knowing when a photon pair was
created during the relatively long time interval τ1 .1 It is this lack of knowledge
that leads to quantum interference in the FRANSON-3 scenario discussed below.
The spectacular aspect of FRANSON is that unlike the double-slit experiment,

1
Of course, “photon creation” is a vacuous metaphysical picture extrapolated from
observations done after the source has been triggered. There is no evidence, based on the
given apparatus, for the belief that anything has been “created” at S.
 20.3 FRANSON-1: ΔT  τ2 273

where the observer does not know from which point in space a photon came, here
the observer does not know at which point in time the photon pair was produced.

Travel Time Difference ΔT

The third characteristic time is ΔT , the time difference between a photon trav-
eling along the short arm of its interferometer and along its long arm. This time
is adjustable but fixed during each run and is assumed the same for each of the
two interferometers in that run.
There are three scenarios we shall discuss: FRANSON-1, for which ΔT  τ2 ;
FRANSON-2, for which τ1  ΔT ; and FRANSON-3, for which τ2  ΔT  τ1 .
In Franson’s original analysis (Franson, 1989), photon spin did not play a role.
Therefore, in all scenarios considered here, photon spin is assumed fixed once a
given photon pair has been created.

20.3 FRANSON-1: ΔT  τ2
The relevant figure for this scenario is Figure 20.1. We discuss the experiment in
terms of its stages, as follows.

Stage Σ0
The initial total state is
 1 00 ,
|Ψ0 ) ≡ |s0  ⊗ A (20.1)
0

where |s0  represents the initial source spin state.

Stage Σ0 to Stage Σ1
By stage Σ1 , the initial state has split into a correlated pair of photons moving
in opposite directions. The creation of this pair by this stage is represented by
the action of the contextual evolution operator U1,0 :
 1A
|Ψ1 ) ≡ U1,0 |Ψ0 ) = |s1  ⊗ A 2
1 1 01 , (20.2)

where |s1  represents the combined spin state of the photon pair at this stage.

Stage Σ1 to Stage Σ2
The stage-Σ1 photons pass through beam splitters B 1 and B 2 as shown. One
output channel from each beam splitter leads directly to a final beam splitter,
while the other output channel is deflected by a mirror through a phase changer
before being deflected onto that final beam splitter. The four beam splitters B i ,
i = 1, 2, 3, 4, are parametrized by real transmission and reflection coefficients ti ,
ri respectively, according to the prescription given by Eq. (11.28).
274 The Franson Experiment

The labstate evolution is given by

    
U2,1 |s1  ⊗ A 2
 1A 1 1 1 iφ1  3  2 + ir2 eiφ2 A
 4 02
1 1 01 = |s2  ⊗ t A2 + ir e A2 t2 A 2 2
 1

t1 t2 32 + ir1 t2 eiφ 62 +
= |s2  ⊗ 2 1 2 ,
ir2 t1 eiφ 92 − r1 r2 ei(φ +φ ) 122
(20.3)
where φ1 and φ2 are total phase change factors due to the increased path length
of the long arms of the interferometers and phase-shift plates introduced in those
long arms by the observer. In the last line in (20.3), we show the computation
basis representation (CBR) of the expression in the previous line.

Stage Σ2 to Stage Σ3
There are four terms to consider in the transition from Σ2 to Σ3 :
    
U3,2 |s2  ⊗ A 2
 1A 3 1 3 3  2 + ir4 A 4 03 ,
2 2 02 = |s3  ⊗ t A3 + ir A3 t4 A 3 3
    
U3,2 |s2  ⊗ A  2 02 = |s3  ⊗ t3 A
 3A  3 + ir3 A
1 t 4 2
A + ir 4 4
A
2 2 3 3 3 3 03 ,
    
U3,2 |s2  ⊗ A 4
 1A 3 1 3 3  4 + ir4 A 2 03 ,
2 2 02 = |s3  ⊗ t A3 + ir A3 t4 A 3 3
    
U3,2 |s2  ⊗ A 4
 3A 3 3 3 1  4 + ir4 A 2 03 . (20.4)
2 2 02 = |s3  ⊗ t A3 + ir A3 t4 A 3 3

This is all the information needed for our computer algebra program MAIN
to evaluate the answers to all maximal questions. These answers turn out to
be complicated, long polynomials in the ti and ri parameters, so are not √ listed
here. However, setting them to the empirically useful value ti = ri = 1/ 2, i =
1, 2, 3, 4, as assumed by Franson (Franson, 1989), gives the relative coincidence
rates
     
Pr A  1A 2 03 |Ψ0 = sin2 1 φ1 sin2 1 φ2 ,
3 3 2 2
     
Pr A  A 
3 3 03 |Ψ0 = cos
3 2 2 1 2
1
2φ sin 12 φ2 ,
     
Pr A  1A4
3 3 03 |Ψ0 = sin
2 1 1 2 1 2
2 φ cos 2 φ ,
     
Pr A  3A4
3 3 03 |Ψ0 = cos
2 1 1
2φ cos2 12 φ2 . (20.5)

Each of these rates shows angular dependence due to independent “photon

self-interference” within each separate interferometer. This form of interference
will be referred to as local. There are no global interference effects involving both
interferometers and no post selection of data is required.

20.4 FRANSON-2: τ1  ΔT
In this variant of the Franson experiment, the photon wave trains 32 , 42 reflected
at B 1 and B 2 , respectively, travel along the long arms of their respective inter-
ferometers at the speed of light or less, depending on the medium through which
 20.4 FRANSON-2: τ1  ΔT 275

M j M M j M

B S B

B B B B

Figure 20.2. FRANSON-2: the Franson experiment for τ2  τ1  ΔT .

they move. Because now the travel time difference ΔT is very much greater than
the coherence time τ1 , these wave trains arrive at B 3 and B 4 long after the
transmitted wave trains 12 and 22 have impinged on B 3 and B 4 , respectively. In
consequence, no local or global interference can take place. In fact, the observer
can now obtain total information concerning the timing of each coincidence
outcome in every run of the experiment and know precisely what path was taken
by each photon.
Under this circumstance, the four original detectors 13 , 23 , 33 , and 43 used
in FRANSON-1 now have to be regarded as eight separate detectors, i3 , i =
1, 2, . . . , 8, as shown in Figure 20.2. The first four of these register photon clicks
from short-path photons, while the last four signal clicks from those that have
traveled the long paths. This information is specific to each photon and does not
involve any photon pairs.
Significantly, the final-stage quantum register involved in this scenario and
the next one, FRANSON-3, is 256-dimensional. However, our computer algebra
program MAIN has no difficulty dealing with this because it has been encoded
to process only the relevant contextual Hilbert spaces, and these are of greatly
reduced dimensions.
This demonstrates a fundamental point about apparatus. In the conventional
usage of apparatus, experimentalists tend to regard their equipment as having
some sort of “transtemporal” identity, or persistence. In Figure 20.2, for example,
beam splitters B 3 and B 4 would most likely persist in the laboratory as material
objects, during the long interval ΔT between their interaction with wave-trains
12 and 22 and with the delayed wave-trains 32 and 42 . Even classically, however,
this need not be the case. It is conceivable that ΔT could be so long, such as
several years, that the beam splitters could be destroyed and rebuilt at leisure
between the observation of any short-arm photons and any long-arm photons.
276 The Franson Experiment

Whatever the actuality in the laboratory, from a quantum point of view, the
beam splitters receiving short- and long-arm photons should be considered as
completely separate pieces of equipment in this scenario (but not in the next).
In other words, apparatus and how it is used is time dependent. The analysis in
the next section shows that the rules for doing this can be quite nonclassical and
appear to violate the ordinary rules of causality.
For the FRANSON-2 scenario, τ1  ΔT , the dynamics follows the same rules
as in the previous section up to the transition from stage Σ2 to stage Σ3 . At this
point, the transformation rules have to take into account the possibility that the
observer could know the timings of all events. The rules for this transition are now
    
U3,2 |s2  ⊗ A 2
 1A 3 1 3 3  2 + ir4 A 4 03 ,
2 2 02 = |s3  ⊗ t A3 + ir A3 t4 A 3 3
    
U3,2 |s2  ⊗ A 2
 3A 3 5 3 7  2 + ir4 A 4 03 ,
2 2 02 = |s3  ⊗ t A3 + ir A3 t4 A 3 3
    
U3,2 |s2  ⊗ A  4 02 = |s3  ⊗ t3 A
 1A  1 + ir3 A
3 t 4 6
A + ir 4 8
A
2 2 3 3 3 3 03 ,
    
U3,2 |s2  ⊗ A 4
 3A 3 5 3 7  6 + ir4 A 8 03 .
2 2 02 = |s3  ⊗ t A3 + ir A3 t4 A 3 3
(20.6)
which should be compared with (20.4).
In this scenario, we find sixteen nonzero coincidence rates, each of the form
Pr(A i A j
3 3 03 |Ψ0 ), where i = 1, 3, 5, 7 and j = 2, 4, 6, 8. All of them are independent
of φ1 and of φ2 . For example, Pr(A  1A6
3 3 |Ψ0 ) = (t√
1 2 3 4 2
r t t ) , and so on. In the case of
i i
symmetrical beam-splitters, where t = r = 1/ 2, all 16 rates are equal to 1/16.
For FRANSON-2, the detectors behave in a manner consistent with the notion
that photons are classical-like particles propagating along definite paths.
We should comment on the choice of stage diagram for FRANSON-2, as
this impacts significantly on the encoding of program MAIN. Our choice of
introducing detectors 53 , 63 , 73 , and 83 seems inevitable but puzzling, because it
could be claimed that in a real experiment, there would only be four final stage
detectors, not eight. However, our observation above that the long-arm photon
signals could occur perhaps many years after the short-arm signals addresses that
point: the observer will have real evidence that one set of signals has arrived
long before the other set. The fact that the same atoms could persist in the
configuration of detectors 13 , 23 , 33 , and 43 until they did service as detectors
53 , 63 , 73 , and 83 is irrelevant.
A related point is that, for simplicity, we chose to model all of the eight final-
stage detectors to be associated with stage Σ3 , but that is not necessary. It would
perhaps have been more logical to assign the long-arm detectors to a separate,
later stage, but that would have been an inessential complication in programming
that would make no difference to the calculations or to the conclusions drawn
from them.
FRANSON-2 is an important illustration that the observer–apparatus side of
quantum physics can contain real surprises. The next variant, FRANSON-3, is
even more surprising.
 20.5 FRANSON-3: τ2  ΔT  τ1 277

M j M M j M

B S B

B B B B

Figure 20.3. FRANSON-3: the Franson experiment for τ2  ΔT  τ1 .

20.5 FRANSON-3: τ2  ΔT  τ1
This is the actual scenario discussed by Franson (Franson, 1989). In the following,
S stands for “short path” and L for “long path.” The fundamental change
induced by the observer’s setting of ΔT such that τ2  ΔT  τ1 is that,
unlike the previous scenario, the observer cannot now use individual times of
detector clicks to establish which of the coincidences S–S or L–L has occurred
in a given run of the experiment. This presumes that a given run lasts at least
as long as the coherence time, and that coincidence observations are being made
at times greater than a critical time τc ≡ τs + ΔT into any given run. Here τs is
the earliest time that a wave train following an S path could take to a stage Σ3
detector. Coincidence observations of interference involving any L path at any
time earlier than τc into a given run will not occur, simply because any wave
train taking such a path would still be on its way to stage Σ3 detectors.
FRANSON-3 is discussed and modeled assuming observations are made at
times later than τc . In that regime, it will then not be possible for the observer to
use detection times to distinguish between S–S correlations and L–L correlations.
However, the observer will be able to filter out the two separate S–L correlations.
The relevant diagram is Figure 20.3, which is identical to Figure 20.2 except
now 53 is replaced by (3 ∨ 5)3 , 73 is replaced by (1 ∨ 7)3 , 63 is replaced by (4 ∨ 6)3 ,
and 83 is replaced by (2 ∨ 8)3 , where for example 3 ∨ 5 means “3 or 5.” Which
alternative is taken depends on the contextual information available in principle
to the observer.
The dynamics for this scenario is identical to that for the previous one, except
for the last equation in (20.6), which is replaced by
    
 3A
U3,2 |s2  ⊗ A 4 3 3 3 1  4 + ir4 A
 2 03 .
2 2 02 = |s3  ⊗ t A3 + ir A3 t4 A 3 3 (20.7)
278 The Franson Experiment

With this amended information, program MAIN gives the following nonzero
coincidence rates:
     
Pr A 2
 1A 3 4
3 3 03 |Ψ0 = Pr A3 A3 03 |Ψ0 = 4 cos
1 2 1 1 1 2
2 φ + 2 φ ,
     
Pr A  2A
3 1 4
3 3 03 |Ψ0 = Pr A3 A3 03 |Ψ0 = 4 sin
1 2 1 1 1 2
2φ + 2φ , (20.8)
i i
√ eight others that each have value 1/16, for the symmetric case t = r =
and
1/ 2, i = 1, 2, 3, 4.
These results are a spectacular demonstration of temporal nonlocality as well
as spatial non-locality: interference is occurring between S–S and L–L outcomes.
Note that in actual FRANSON-3 type experiments, the observer would have
to measure the times at which coincidence clicks were obtained during each run
and then post-select, that is, filter out, those coincidences corresponding to the
{S–S, L–L} processes and those corresponding to {S–L, L–S}.

20.6 Conclusions
Since the publication of Franson’s original paper, there has been great interest
in empirical confirmation of FRANSON-3 predictions. While there is still some
room for debate concerning the interpretation of the experiment, the results
of Kwiat et al. (1993) vindicate Franson’s prediction, which corresponds to
2
 1A
Pr(A 3 3 03 |Ψ0 ) = 4 cos ( 2 φ + 2 φ ) in our approach.
1 2 1 1 1 2

Assuming the quantum theoretical interpretation of this experiment is correct,

then there is an extraordinary lesson to be learned, not about systems under
observation in particular, but about the rules concerning the use of apparatus
and how these can differ spectacularly from those expected classically. The inter-
ference of the S–S and L–L amplitudes in FRANSON-3 cannot be envisaged
in a classical way to occur locally in time. Any attempt to think about such
interference in terms of photons as actual particles would lead to bizarre concepts
that would never be acceptable conventionally. The conclusion is that a two-
photon state is not equivalent under all circumstances to a state with two separate
but entangled photons. A more recent quantum optics experiment with similar
conclusions has been reported by Kim (Kim, 2003).
The weight of evidence points to the conclusion that quantum outcome
amplitudes are dynamically correlated with contextual information held by the
observer. When some information is absent, then quantum interference can occur.
This supports the position of Heisenberg and Bohr concerning the fundamental
principles and interpretation of quantum physics. Quantum optics experiments
such as FRANSON are providing more and more evidence that QM is not just
a theory of SUOs but also a fundamental perspective on the laws of observation
in physics. It is our considered view that the surface of those laws has only been
scratched to date.
 21
Self-intervening Networks

21.1 Introduction
In this chapter, we discuss some of the differences between apparatus and systems
under observation (SUOs). The discussion is phrased in terms of self–intervening
networks, wherein partial quantum outcomes in an early stage of an apparatus
can alter future configurations of that apparatus.
To explain what we mean, we introduce the following classification of experi-
ments. Note that reference to apparatus here includes real and virtual detectors,
and real and virtual modules. Virtual detectors and modules are mathematical
fictions that are introduced into the formalism for convenience.

Type-Zero Experiments
Many important experiments involve signal propagation through “empty space.”
We classify these as type zero (T0). The feature of T0 experiments qualifying
for this classification is that the observer has no control of the modules in the
information void. Examples are experiments in astrophysics, where the observer
can choose which source to observe and how they observe it, but has no influence
on the physical properties of whatever lies between source and detector. If there
are gas clouds between source and detector, the observer can only recognize that
fact after the experimental data are analyzed. An important modern variant of
T0 experiments involves map location via the Global Positioning System (GPS):
signals sent from Earth to geostationary satellites and received by mobile devices
back on Earth have traveled through Earth’s atmosphere and through empty
space, where special relativistic and general relativistic effects have to be taken
into account (Ashby, 2002).
Before the rise of experimental science, most “experiments” were based on
visual observations, which are principally of T0 classification. The advent of the
telescope and the microscope did not change things in this respect. Indeed, it
has been suggested that science did not progress in antiquity precisely because
of the philosophical principle that the only valid experiments could be of T0, as
280 Self-intervening Networks

these do not introduce the artificiality of constructed modules into observation,

which (it was argued) would give a false perspective on “reality.” The logic of
that argument is superb, but it is a dead-end principle in science.
Several sciences such as geology, archaeology, and, indeed, cosmology started
off with periods of basic observation rather than experimentation. During
such initial stages of these sciences, the observations could be classified as
T0 experiments. In mechanics, the apocryphal dropping by Galileo of two
spheres of different mass from the Leaning Tower of Pisa can be classified as of
type T0.

Type-One Experiments
Many quantum experiments involve time-independent modules, introduced by
an observer into an information void, which appear to persist unchanged in the
laboratory over the course of each run of an experiment. By this we mean that
for each run of such an experiment, the apparatus that prepares the initial state,
shields it from the environment during that run, and detects the outcome state
is considered fixed during each stage of a run. In other words, such apparatus
persists over time scales significantly greater than the time scale of each run. We
classify such experiments as type one (T1). These differ from T0 experiments in
that the observer explicitly introduces real modules, such as beam splitters and
mirrors, into the information void.
The illusion of persistence is a convenient mental device, because all experi-
ments are done in process time and everything changes. However, the concept
of persistence is a potent and necessary one that is needed to make sense of
what science is all about. It is formalized in quantized detector networks (QDN)
by the use of null evolution, as discussed in Section 7.11 and Chapter 18. In
standard quantum mechanics (QM), T1 experiments are described in terms of
time-independent Hamiltonian operators in Schrödinger equations.
In practice, all experiments have some degree of T1 behavior. For instance,
the very action of constructing apparatus before an experiment starts carries the
hidden assumption that the constructed apparatus will persist long enough to be
useful. That assumption has some amusing aspects. For example, no one would
consider constructing parts of their apparatus out of ice, unless the environment
made that feasible, as in the Antarctic.
In quantum optics, the double-slit (DS) experiment, Mach–Zehnder interfer-
ometer experiments, and quantum eraser experiments (Walborn et al., 2002) are
all of this type. So too are high-energy particle-scattering experiments such as
those conducted at the Large Hadron Collider at CERN.
After T0, T1 experiments are technically the easiest to construct, and many
historical experiments were of this type. T1 experiments allow the focus of
attention to be entirely on the dynamical evolution of the states of an SUO,
because such states are conventionally regarded as the only objects of interest in
physics.
 21.1 Introduction 281

Type-Two Experiments
An important class of experiment, referred to here as type two (T2), involves
some time dependence in the apparatus that is controlled by factors external
to the apparatus, either by the observer directly or by factors external to the
laboratory. When controlled by the observer, such experiments require a more
advanced technological level than in T1 experiments, such as advanced electronics
and computerization. In standard quantum mechanics (QM), T2 experiments
are described in terms of time-dependent Hamiltonian operators in Schrödinger
equations.
Spin-echo magnetic resonance experiments are of this type, because the exper-
imentalist arranges for certain magnetic fields to be rotated precisely, while addi-
tionally, the sample environment introduces random external influences related to
local temperature. An example of random changes controlled by the experimen-
talist are the delayed-choice experiments such as that of Jacques et al. (2007),
discussed in Section 14.5, where carefully arranged random changes are made
during each run. QM typically describes such experiments via time-dependent
Hamiltonian that may have stochastic elements.
T2 experiments are always going to be more interesting than T1 experiments
for two reasons: T1 experiments can be regarded as limiting cases of T2 exper-
iments where time dependence is turned off, and because T2 experiments have
the potential to reveal more information about the dynamics of systems under
observation than T1. Schwinger’s source theory in relativistic quantum field
theory shows that, in principle, T2 experiments allow for the extraction of all
possible information about quantum systems (Schwinger, 1969, 1998a,b,c).
T1 and T2 experiments may be collectively labeled as exophysical, because they
involve classical apparatus interventions that are external in origin to the SUO.
In such experiments, the apparatus is classically well defined at each instant
of time during each run, even in those situations where it changes randomly.
Therefore, a classical Block Universe (Price, 1997) account of apparatus during
each run of a T1 or T2 experiment may be possible.

Type-Three Experiments
In this chapter, we explore a third type of quantum experiment, referred to as
type three (T3). In such an experiment, the apparatus is modified internally by
the quantum dynamics of the SUO, rather than externally by the observer or the
environment.
There is an interesting point to be made here about experiments conducted
at high-energy particle-scattering laboratories such as the Large Hadron Collider
(LHC). Such experiments have fixed sources and detectors and do not have real,
physical modules in the interaction region. From that perspective they are T0
experiments. However, the observers can arrange for their particle beams to inter-
sect and allow particle collisions, so from that perspective, such experiments are
of type T1. However, the interactions that occur are entirely quantum processes,
282 Self-intervening Networks

so they are of also type T3 in an essential way. Indeed, that is the whole point
of such experiments.
In Chapter 25 we discuss such experiments as processes where virtual modules
are created in the information void by the dynamics of the SUOs. For exam-
ple, two-particle to two-particle scattering amplitudes discussed in high-energy
particle physics have all the characteristics of beam splitters.

Type-Four Experiments
In recent years, there has been interest in experiments that contain T2 and
T3 aspects. We have in mind experiments of the Eliztur–Vaidman bomb-tester
variety, discussed in Chapter 25. Such experiments can be classified as type-four
(T4) experiments
A question that we address toward the end of this chapter is whether T3
experiments can always be given a classical Block Universe account or whether
something analogous to superpositions of different apparatus has to be envisaged.
This is not to be confused with the superposition of labstates that we have
discussed up to now.
This question is related to the rules of quantum information extraction as they
are currently understood in QM. These rules state that quantum interference can
occur in the absence of classical which-path information, the most well-known
example of this being the DS experiment.
The question here is what precisely does a lack of which-path information
mean? If such as thing as a photon actually passed through one of the slits,
would it leave any trace in principle? Even if it did, it would have to be essentially
unobservable, at least far below the scales of classical mechanical detection. In
that case, an observer of the interference pattern would indeed be unaware of
such an interaction. In essence, such a scenario would be welcomed by supporters
of the Hidden Variables (HV) interpretation of empirical physics.
This supposition seems wrong to us for the following reason. In the DS exper-
iment, there are three cases: (1) a photon actually goes through one of the slits
but leaves no trace whatsoever; (2) a photon actually goes through one of the
slits and leaves a trace that cannot be detected; or (3) we avoid the use of the
photon concept.
Our inclination is to discount case 1 on the grounds that any assertion that
something “exists” or has occurred, but no evidence for it can ever be found, is
vacuous and can be dismissed by Hitchens’ razor.1
As for case 2, the idea that it is the mere lack of suitable photon-slit interaction
detection technology itself that produces interference patterns on a remote screen
seems be wrong on two counts. First, there is now sufficient evidence against the

1
“What can be asserted without evidence can be dismissed without evidence,” a modern
version of the Latin proverb “Quod gratis asseritur, gratis negatur ” (what is freely asserted
can be freely deserted).
 21.1 Introduction 283

notion that photons are particles in the conventional sense (Paul, 2004). Second,
the idea that interference patterns could occur when one observer was unaware
of actual which-path information while another observer was aware of it seems
inconsistent.2 It seems likely that there has to be something deeper in the origin
of quantum interference. We do not believe it lies in the hidden variables direction
because of its contextual incompleteness.
Neutron interference experiments (Greenberger and YaSin, 1989) explore this
question in that they are moving toward larger scales of interaction between
SUOs and apparatus. In their experiment, the movement of mirrors involved
in their quantum erasure scheme involves macroscopic numbers of atoms and
molecules (Becker, 1998). In this case, the dynamical effects of the impact of
a particle on a mirror is reversed by a second impact. What is amazing is the
idea that all possible traces of the first impact could be completely erased, even
though there could (in principle) be time for information from the first impact
to be dissipated into the environment, thereby rendering the process irreversible.
This raises the question of what irreversibility actually means. Is a process
irreversible in some absolute sense, or is it contextual and dependent on the
observer? We have previously suggested that there are no absolutes in physics.
Therefore, we should examine the possibility that what is irreversible to one
observer need not be irreversible to another. This does not seem inconsistent or
unreasonable, given that irreversibility and observers are emergent processes, the
laws of which have not yet been understood in any significant way.
We focus exclusively on linear quantum evolution, i.e., quantum state evolution
conforming to the principles of QM as discussed for example in Peres (1995),
rather than appeal to any form of nonlinear quantum mechanics to generate
self-intervention effects. We explore a number of type-three thought experiments
involving photons, which act as either quantum or classical objects at various
times. As quantum objects they pass through beam-splitters and suffer random
outcomes as a result. As classical objects they are used to trigger the switching
on or off of macroscopic apparatus, a switching that determines the subsequent
quantum evolution of other photons.
We shall not discuss the nature of photons per se, except to say that they
are referred to as particles for convenience only: our ideology and formalism
treats them as signals in detectors (Jaroszkiewicz, 2008a). Everything is idealized
here, it being assumed that all detectors operate with 100 percent efficiency and
that photon polarizations and wavelengths can be adjusted wherever necessary
to make the scenarios discussed here physically realizable. The experiments we
discuss are not necessarily based on photons: other particles such as electrons
could be used in principle.

2
There is a possible loophole here: our comment involves two observers with access to
different contextual information. What that means physically is a nontrivial question for
any theory of observation. In the absence of any proper theory of such a scenario, we can at
best express our opinion on the matter.
284 Self-intervening Networks

S B

Figure 21.1. If detected, photon 22 triggers the switching-on of mirror M that

deflects 12 into 43 rather than 13 .

21.2 Experiment SI-1: Basic Self-intervention

To illustrate the sort of experiment we are interested in, we start with the
experiment shown schematically in Figure 21.1. By stage Σ1 , a correlated, nonen-
tangled two-photon state Ψ1 ≡ A 2
 1A
1 1 01 is created by source S. Channel 21 is
subsequently passed through beam splitter B into channels 22 or 32 . If 22 is
detected physically at stage Σ2 , then a macroscopic mechanism triggers mirror
M to be swung into place so as to deflect 12 onto 43 rather than 13 . The spin
of the photons is neglected in this analysis but could easily be included in the
discussion and encoded into our computer algebra program MAIN if it were
required.
The labstate Ψ2 at stage Σ2 is given by
 1 (tA
Ψ2 = A  2 + irA
 3 )02 , (21.1)
2 2 2

where t and r are beam splitter B parameters satisfying t2 + r2 = 1.

Note that in Figure 21.1, detector 22 is shown shaded, indicating that the
observer looks at it and definitely ascertains its signal status, thereby extracting
classical information. If there is no signal there, then the mirror M is not swung
into place to intercept the channel from 12 , and so a signal is certainly registered
in 13 . On the other hand, if there is a signal in 22 , mirror M swings into place
and deflects channel 12 into detector 43 .
Semi-unitary evolution from stage Σ2 to stage Σ3 is given by
 1A
U3,2 A 2 4 2  1A
3 1 3
2 2 02 = A3 A3 03 , U3,2 A 2 2 02 = A3 A3 03 . (21.2)

Note that in the program MAIN, null evolution carries 22 to 23 and 32 to 33 , and
the program reports detector status at stage Σ3 . The predicted nonzero outcome
probabilities are as expected:

Pr(A 3
 1A 2 4
3 3 03 |Ψ0 ) = r , Pr(A3 A3 03 |Ψ0 ) = t .
2 2
(21.3)
 21.3 Experiment SI-2: Double Self-intervention 285

There is nothing unexpected in these results, but there is an unusual confluence

of classical and quantum interaction: on the quantum side of the experiment,
mirror M is activated in a classically deterministic way only if and when quantum
outcome 22 occurs, and that is unpredictable.
One interpretation of this experiment is that it is a form of delayed choice
experiment: the mirror M is activated only if and when 22 has been detected,
and this is after 12 is on its way. The difference between this experiment and
previous analysis of delayed choice is that the “choice” is made via quantum
processes occurring within the experiment itself, which motivates our chapter
title “Self-Intervening Networks.”

21.3 Experiment SI-2: Double Self-intervention

The next variant experiment, SI-2, is shown in Figure 21.2. Source S creates a
correlated photon pair 11 21 by stage Σ1 . These are directed into beam splitters
B 1 and B 2 as shown. B 2 is used as a trigger. If a signal is detected at 32 , that
triggers the classical switching-on of beam splitter B 3 , which intercepts 12 and
channels it into 53 or 63 instead of it passing on to 13 . Conversely, if a signal
is observed at 42 , beam splitter B 4 is switched on, rather than B 3 . Now 22 is
channeled by B 4 into 73 or 83 rather than passing on to 23 .
The dynamics is worked out by the methods outlined in previous experiments.
The final stage rank is now eight, which presents no problem for program MAIN.
Nonzero correlations are found to be

B
B

S
B

Figure 21.2. Experiment SI-2: in this scheme, no quantum interference occurs,

because complete which-path information can be worked out from the signal
pattern in each run.
286 Self-intervening Networks

 2A
Pr(A 3  1A4
3 3 03 |Ψ0 ) = (r r ) , Pr(A 3 3 03 |Ψ0 ) = (t t ) ,
1 2 2 1 2 2

 A
Pr(A   A 
3 3 03 |Ψ0 ) = (t r r ) , Pr(A 3 3 03 |Ψ0 ) = (t r t ) ,
3 5 1 2 3 2 3 6 1 2 3 2 (21.4)
 4A
Pr(A  7 03 |Ψ0 ) = (r1 t2 r4 )2 ,  4A
Pr(A  8 03 |Ψ0 ) = (r1 t2 t4 )2 .
3 3 3 3

Individual detector signal probabilities can be worked out from these correlations.
For example, Pr(13 ), the probability that detector 13 will be in its signal state
rather than its ground state at the final stage Σ3 , is given by

Pr(13 ) ≡ Pr(A 3
 2A 3 5 3 6
3 3 03 |Ψ0 ) + Pr(A3 A3 03 |Ψ0 ) + Pr(A3 A3 03 |Ψ0 ) = (r ) , (21.5)
2 2

and so on.
There are no interference effects predicted in this experiment.

21.4 Experiment SI-3: Interfering Single Self-intervention

The third scenario, SI-3, is shown in Figure 21.3. In this case, the initial photon
pair is passed through a pair of beam splitters B 1 and B 2 exactly as in experiment
SI-2. The difference lies in the next stage. Channels 12 and 22 are sent off over
sufficiently long optical paths so as to allow interference between channels 32
and 42 in beam splitter B 3 . If a signal is registered in 43 , then beam splitter
B 4 is swung into place to intercept 13 and 23 , so that they can interfere and be
observed at 54 and 64 . If B 4 is not triggered, then 13 and 23 are channeled on to
14 and 24 , respectively.
There is a point about beam splitter B 3 that needs attention. This experiment
deals with signality-two amplitudes, so there is an issue at beam splitter B 3 in
that the possibility exists that a photon comes in from 32 and another from 42 .
This is a case of beam splitter saturation, referred to in Chapter 11. Exercise
11.1 shows that if B 3 is calibrated, and if the evolution is semi-unitary such that

S B B

Figure 21.3. Interference at beam splitter B 3 can trigger beam splitter B 4 .

 21.5 Schrödinger’s Cat 287

signality one is conserved, then the only scenario permitted is that a signal is
detected in 34 and in 44 . In that case, even though beam splitter B 4 is triggered,
no signals are detected at 54 or 64 .
The beam splitter saturation scenario shows that there is a fundamental dif-
ference between photon self-interference, which is a signality-one process, and
two-photon scattering, which is a signality-two process. The former is a mani-
festation of the quantum superposition principle, whereas the latter depends on
the details of photon–matter coupling. For instance, the lowest order amplitude
for photon–photon scattering in quantum electrodynamics is proportional to
α2 , where α is the fine structure constant. We mention in passing that the
correct calculation of the photon–photon scattering cross section requires delicate
treatment of divergent integrals, in order to bring the predictions in line with
cosmological observations (Liang and Czarnecki, 2011). This is a good example
of the need to factor in detector physics properly in relativistic quantum field
theory.
The dynamics for experiment SI-3 is handled in the usual way in program
MAIN, with the following predictions:
Pr(A 2
 1A 3
 1A
4 4 04 |Ψ0 ) = (t t ) , Pr(A 4 4 04 |Ψ0 ) = (t r r ) ,
1 2 2 1 2 3 2

Pr(A 5
 4A 2 3
4 4 04 |Ψ0 ) = (t r t r + r t r t ) , Pr(A4 A4 04 |Ψ0 ) = (r t t ) , (21.6)
1 2 3 4 1 2 3 4 2 1 2 3 2

Pr(A 6
 4A 3 4
4 4 04 |Ψ0 ) = (t r t r − r t r t ) , Pr(A4 A4 04 |Ψ0 ) = (r r ) .
1 2 3 4 1 2 3 4 2 1 2 2

In this variant experiment, two of the outcome correlations, Pr(A 5

 4A
4 4 04 |Ψ0 )
 4 6
and Pr(A4 A4 04 |Ψ0 ), show interference that has essential contributions from
beam splitters B 1 and B 2 in a manner that seems impossible to explain in terms
of photons as classical particles.
In all experiments where quantum interference takes place, there inevitably has
to be some which-path uncertainty somewhere. This does not occur in variant
experiments SI-1 or SI-2 but does occur in SI-3. Program MAIN shows that the
interference at 54 and 64 is photon self-interference and occurs only when a single
photon enters B 3 and triggers B 4 . Whenever that happens, the observer cannot
say whether it was actually 32 or 42 that passed into B 3 , and that is the origin of
the missing which-path information that is responsible for the interference effect
at 54 and 64 . The signality-two beam splitter saturation possibility triggers B 4
but does not lead to any signals at 54 or 64 .
Experiment SI-3 involves apparatus change that produces interference and can
be identified as doing so: after each run, it will be clear whether B 4 was triggered,
and the observer will then be sure that it was that beam splitter that created
the interference.

21.5 Schrödinger’s Cat

The Schrödinger’s cat scenario is a well-known illustration of the dangers of tak-
ing a too-literal view of what a quantum wave function represents (Schrödinger,
288 Self-intervening Networks

1935). In brief, the architecture is this. A living cat is placed in a sealed, isolated
box along with a flask of poison and a mechanism that contains a radioactive
sample. The mechanism will release the poison if it is triggered by the decay of
any one of the atoms in the sample. An observer external to the box closes the
box and leaves it alone for 1 hour. At that time, the observer opens the box and
ascertains whether or not the cat is still alive or now dead.
The issue here is that according to standard quantum wave mechanics, the
wave function for the combined system of cat plus radioactive sample will develop
from a separable state of a {live cat and undecayed sample} to a entangled
superposition of a {live cat and undecayed sample} with a {dead cat and decayed
sample}. This prompts questions, such as “How can a cat exist in a superposition
of a live state and a dead state?” and so on, which have, over the decades since
Schrödinger’s article, generated a plethora of confusing views about the nature
of reality.
This scenario is one we can label T3. It is a complex example of a self-
intervening network, presented in simple terms.
For us, such questions are vacuous. All of them. If context is taken properly
into account, then no such questions need be raised. It suffices to look at what
the external observer can do, what they are actually doing, and when they
actually do it. The QDN prescription is clear. The observer prepares the combined
{cat/box/poison sample} SUO at stage Σ0 . On the basis of their contextual
information, such as the decay probability associated with the radioactive sample,
the observer can, just after closing the box, use standard quantum theory to
estimate the probability of the cat being alive when the box is opened at stage
Σ1 , that is, 1 hour later. What the observer eventually believes has actually
happened inside the box depends on the classical information that is allowed to
accumulate there during the hour. For instance, if a decay occurs, not only could
it trigger the poison to be released to kill the cat, but the lab time at which this
occurred could be registered. When the observer finally opened the box, if they
found the cat dead, then the registered time of death would be available.
It is such examples that make it clear that “the” information void is not
an absolute concept. It is contextual. In this case, the external observer has no
information about what is going on inside the box during the given hour that the
experiment runs. Therefore, a quantum description has to suffice. If after opening
the box, sufficient classical information is recovered from it to enable the observer
to reconstruct events inside that box during that hour, then a rather classical
description can be given.
At no point is the observer empirically justified in asserting that the cat is
in a quantum superposition of alive and dead. The observer may be justified in
imagining a mathematical description of the state of the system in such terms.
That is not the same as objective reality.
There are two further points to make here.
 21.5 Schrödinger’s Cat 289

A Cat Is a Living Process

It is rather fanciful to think of a cat in terms of a pure quantum state, albeit with
a vast number of degrees of freedom. A cat is a living process and therefore not
amenable to a reductionist description. In particular, a cat needs an extensive
environment to maintain itself: an atmosphere to breathe in, a source of warmth,
and so on.
Suppose it was argued that such an environment could be provided inside
the box. Yes, but then the observer could not allow that environment to be
contaminated by any factors external to the box. What about gravity? What
about temperature? What about the expansion of the Universe? Can we really
arrange for total shielding of the inside of the box from the outside?
Reductionism in high-energy particle physics seems to work because elemen-
tary particle interactions appear to be amenable to shielding during scattering
experiments, but even this has been questioned. For instance, it is often imagined
that electrons are elementary point particles with a well-identified physical mass
in vacuo, represented by a simple pole in their momentum space propagator. But
radiative corrections due to “soft” photons (associated with the low-frequency
end of the electromagnetic spectrum) appear to turn that pole into a branch
cut (in the complex plane), with the consequence that the concept of “electron
mass” is problematical. On that basis, an electron cannot be imagined as a single
particle, but an extraordinarily complex system of a charge surrounded by an
indeterminate cloud of photons.3
Our point is that a highly complex emergent concept such as a living cat might
not be amenable to any form of quantum state description, because quantum
processes manifest themselves only when phases are carefully controlled. The
same comment applies to the so-called quantum mind program of investigation
by Hameroff and others, in which brain function is modeled as a form of quantum
computation (Hameroff, 1999).
One way of seeing why such a program would not work would be if we described
the brain as affected by contextual overcompleteness, that is, subject to so many
external, extraneous contextual factors that cannot be eliminated, such as tem-
perature variations, pressure fluctuations, and so on, that phase control becomes
meaningless and decoherence dominates.

Life and Death Are Not Classic Binary Alternatives

Zero and one (or their equivalents) are regarded as the only two possible alter-
native states of a classical bit. This degree of simplicity is so powerful and useful
that we tend to apply it to many complex situations: we won the war or we

3
It is such considerations that make it obvious to us that elementary particle physics is
really a branch of emergent physics and should be approached as such, rather than as a
reductionist theory from which emergence could be understood.
290 Self-intervening Networks

lost it; he is a good driver or he is a bad driver; the cat is alive or it is dead.
In reality, wars may be lost by both sides, a man may drive well in his village
but be reckless on the motorway; a cat may be in a vegetative state that is not
dead but is not really worth calling life.
All of these concerns should be addressed before simplistic arguments based
on contextually incomplete views of quantum mechanics are given.
 22
Separability and Entanglement

22.1 Introduction
In this chapter, we discuss the structure of signal amplitudes in terms of the
separability and entanglement of quantum register states. We review the concepts
of subregisters, splits, partitions, factorizable Hilbert spaces, the separability of
states, separations, and entanglements.
All Hilbert spaces in this chapter are taken to be complex and finite dimen-
sional, and denoted by capital calligraphic Latin letters, such as H. We do not
restrict the discussion to quantum bits, but ultimately, it is those in which we
are most interested.
Suppose X and Y are two finite-dimensional Hilbert spaces with a surjective
linear map U from X to Y.1 Suppose further that U has the property that it
preserves norm, that is, if for any element x in X we have U xY = xX ,
√
where xX ≡ (x, x)X is the Hilbert space “length” of x in X and yY ≡
√
(y, y)Y is the Hilbert space “length” of y in Y. Then U is called an isometric
isomorphism. Under these conditions, it is necessarily the case that X and Y have
the same dimension and, as far as basic Hilbert space properties are concerned,
are identical. We denote this form of equality by X  Y.

22.2 Quantum Registers

Hilbert spaces per se are central to quantized detector networks (QDN) but that
is on the mathematical side. Physics brings in empirical context that adds an
extra flavor to the discussion, requiring the mathematics of tensor products and
quantum registers. We define a quantum register as the tensor product of two or
more finite-dimensional Hilbert spaces, referred to as subregisters. In principle,

1
Here surjective means that U X = Y, where U X is the image of X in Y under the linear
map U . This condition is not imposed in our definition of semi-unitary operators (which are
linear maps).
292 Separability and Entanglement

subregisters can have any dimension, including one. In the following, we shall
rule out one-dimensional Hilbert spaces as being of no empirical interest to us.

Example 22.1 Let U, V, and W be three complex, finite-dimensional

Hilbert spaces of dimensions 3, 2, and 4, respectively. Consider the tensor
product X defined as

X ≡ U (1) ⊗ W ⊗ U (2) ⊗ V, (22.1)

where U (1) and U (2) are copies of U. Then X is a complex Hilbert space of
dimension 3 × 4 × 3 × 2 = 72.

The number of subregisters in a given quantum register H is called the rank of

that register and denoted rankH. In Example 22.1, X has rank four. A rank-one
Hilbert space will be called an atom.
The dimension of a rank-r register must have at least r non-trivial factors.
For example, a rank-two register of dimension 63 is the tensor product of a
7-dimensional atom with a nine-dimensional atom, or the tensor product of a
21-dimensional atom with a 3-dimensional atom.
The subregister concept, and by implication, that of rank, is contextual,
because it is possible to encounter situations where a subregister could be
thought of as a register itself, with its own subregisters. In other words, atoms
can have constituents.

Example 22.2 Positronium is generally described as an unstable bound

state of an electron and a positron. When viewed as a single particle,
positronium comes in two forms: para-positronium and ortho-positronium. The
former has a particle spin angular momentum classification j = 0 (spin zero),
described by a singlet spin state in a one-dimensional Hilbert space Hpara ,
while the latter has spin j = 1 (spin one) described by spin states in a three-
dimensional Hilbert space Hortho .
On the other hand, a nonrelativistic analysis of positronium in terms of
its two constituents would lead to its spin being modeled in terms of a four-
dimensional Hilbert space Helectron ⊗ Hpositron , a rank-two tensor product of
an electron spin space Helectron , and a positron spin space Hpositron .
Mathematically, we may write

Hpara ⊕ Hortho  Helectron ⊗ Hpositron , (22.2)

where the symbol ⊕ denotes a direct sum of vector spaces. The direct sum of
two Hilbert spaces is a Hilbert space with dimension equal to the sum of the
two Hilbert spaces being summed. The group theory of such bound states and
their decomposition leads in this case to the rule 1 ⊕ 3  2 ⊗ 2 (Lichtenberg,
1970).
 22.3 Splits 293

Throughout this chapter, we adhere to our convention that subscripts label

stages while superscripts label subregisters. Stages concern dynamics, which is
not the focus in the present chapter, so we avoid subscripts in this chapter.
As we have seen before in this book, the ordering of subregisters in a tensor
product will not usually be regarded as significant; i.e., Ha ⊗ Hb will mean the
same mathematically as Hb ⊗ Ha , as well as physically. What is important is the
fact that each subregister can be identified, meaning that labels are physically
significant.
The reason for this is based on the physics of observation. For instance, given
two detectors Da and Db , we may represent them by two Hilbert spaces Ha
and Hb , respectively. There will in general be no natural way of ordering these
detectors, and so there is no logical reason to order Ha and Hb in a tensor
product. What matters is the labeling, which can be regarded as part of the
empirical contextual information always available to observers.
With these comments in mind, we henceforth suppress the tensor product
symbol ⊗, so that Ha Hb and Hb Ha both mean Ha ⊗ Hb .
Given two Hilbert spaces Ha and Hb of dimension da and db , respectively, the
tensor product Ha Hb is a Hilbert space of rank two and dimension da db . We
define H[ab] as a rank-one Hilbert space (that is, an atom) of dimension da db
that is isometrically isomorphic to Ha Hb , that is,

H[ab]  Ha Hb . (22.3)

Specifically, H[ab] is the same mathematically as Ha Hb , but its physical context,

that it is a tensor product, is now ignored.
Likewise, given three Hilbert spaces Ha , Hb , Hc , we define H[abc] to be an
atom isometrically isomorphic to Ha Hb Hc , and so on for higher rank registers.
Note the trivial identity H[a]  Ha .
As we have indicated previously, the order of superscripts in the above is not
significant, so H[ab] = H[ba] , and so on.

22.3 Splits
A split is any convenient way of grouping the subregisters in a quantum register
into two or more factor registers, or atoms, each of which is regarded for the
purposes of that split as a Hilbert space of rank one, that is, a Hilbert space not
itself split into two or more factor registers. For large-rank quantum registers,
very many different splits will be possible.

Example 22.3 The rank-three register Ha Hb Hc may be split in five distinct

ways:

Ha Hb Hc  Ha H[bc]  Hb H[ac]  Hc H[ab]  H[abc] . (22.4)

Note that one of these ways is the original register itself.

294 Separability and Entanglement

The number of ways of splitting a rank-n quantum register is the same as the
number of ways Bn of partitioning a given set of cardinality n, a historically
important topic in combinatorial mathematics. The Bn are called Bell numbers 2
and satisfy many curious and interesting relations in diverse fields, such as
probability, game theory, and number theory. For instance,

n
n!
Bn+1 = Bk , B0 = 1, (22.5)
k!(n − k)!
k=0

from which we find the sequence {Bn , n = 1, 2, . . .} = {1, 2, 5, 15, 52, . . .}. An
explicit formula for Bn is given by Dobinski’s formula,
∞
1  kn
Bn = . (22.6)
e k!
k=0

22.4 Partitions
Consider a set {Ha : a = 1, 2, . . . , r} of Hilbert spaces, denoting the dimension
of Ha by da . Then the rank-r quantum register H[r] ≡ H1 H2 . . . Hr is a vector
space of dimension d[r] ≡ d1 d2 . . . dr that contains both entangled and separable
states.
The classification of states in such a register into separable or entangled types
is too limited for us, so we introduce the more useful concepts of separations and
entanglements. We explain our terminology starting with the separable sets.

Separations
For any two subregisters Ha , Hb of the quantum register H[r] , such that a = b,
we define the rank-two separation Hab to be the subset of the tensor product
Ha Hb consisting of all separable elements in that tensor product, that is,
 
Hab ≡ φa ψ b : φa ∈ Ha , ψ b ∈ Hb . (22.7)
By definition, Hab includes the zero vector 0ab of the tensor product Ha Hb . Note
that Hab = Hba .
The separation concept readily generalizes to higher rank separations as fol-
lows. Pick an integer k in the interval [1, r] and then select k different elements
1 2 k
a1 , a2 , . . . , ak of this interval. Then the rank-k separation Ha a ...a is the subset
1 2 k
of Ha Ha . . . Ha given by
1 2 k
 1 2 k i i

Ha a ...a ≡ ψ a ψ a . . . ψ a : ψ a ∈ Ha , 1  i  k . (22.8)

Every element of a rank-k separation has k factors. A rank-one separation of a

subregister is by definition equal to that subregister, and is therefore a Hilbert
space in its own right. Separations of rank greater than one, however, cannot

2
After E. T. Bell (Bell, 1938).
 22.4 Partitions 295

be Hilbert spaces because they do not contain entangled states, which all tensor
products of rank two or more necessarily do.

Rank-Two Entanglements
We can now construct the entanglements, which are defined in terms of com-
plements. Starting with the lowest rank possible, we define the rank-two entan-
glement Hab to be the complement of the separation Hab in the tensor product
Ha Hb , that is,
Hab ≡ Ha Hb − Hab . (22.9)
Note that Hab cannot be a vector space because it does not contain the zero
vector.
The original tensor product space considered as a set is therefore the union
Ha Hb = Hab ∪ Hab (22.10)
of the set Hab of all separable states and the set Hab of all entangled states.
These two subsets are disjoint and neither is a vector space.

Separation Products
The generalization of the above to larger rank entanglements is straightforward,
but first it will be useful to extend our notation to include the concept of
separation product.
Suppose A and B are subsets of Hilbert spaces Ha and Hb , respectively, where
a = b. We define the separation product A • B to be the subset of Ha Hb given by
A • B ≡ {ψφ : ψ ∈ A, φ ∈ B} . (22.11)
Properties of the separation product are:
1. The separation product is symmetric, that is, A • B = B • A. This means
that the separation product is not equivalent to the Cartesian product A × B,
which is the set of all ordered pairs of elements.
2. Hab = Ha • Hb . Note that this is not the same thing as Ha Hb , which in our
notation is the tensor product of Ha and Hb .
3. The separation product is associative, commutative, and cumulative, i.e.
 a   
H • Hb • Hc = Ha • Hb • Hc ≡ Habc
Hab • Hc = Habc . (22.12)
The separation product can also be defined to include entanglements. For
example,
 
Hab • Hc = φψ : φ ∈ Hab , ψ ∈ Hc . (22.13)

Significantly, while the separation product of two separations is a separation, the

separation product of two entanglements is not an entanglement, that is,
Hab • Hcd = Habcd . (22.14)
296 Separability and Entanglement

A further notational simplification is to use a single H symbol, incorporating

the separation product symbol • with indices directly, as the following example
illustrates.

Example 22.4 Given Hilbert spaces Ha , Hb , . . . , Hi , we may write

Hcd•hi•abef g ≡ Hab • Hcd • Hef g • Hhi . (22.15)

Other expansions are possible, given that separations such as Hef g can be
expanded further as separation products.

Associativity of the separation product applies to both separations and entan-

glements, as can be readily proved.

Higher Rank Entanglements

We can now define higher rank entanglements, such as Habc , Habcd , and so on.
These are defined in terms of complements, in the same way that Hab was defined.

Example 22.5 Consider the rank-three tensor product Ha Hb Hc and the

following disjoint subsets: Habc , Ha•bc , Hb•ac , Hc•ab . These are all separable
in one way or another. For instance, Habc is completely separable, while the
other three subsets referred to are partially separable. If we remove all those
subsets from Ha Hb Hc , then what is left will be completely entangled, which
is what we want to define. Hence we define the rank-three entanglement Habc
as the complement

Habc ≡ Ha Hb Hc − Habc ∪ Ha•bc ∪ Hb•ac ∪ Hc•ab . (22.16)

We will refer to a set such as Habc as a rank-three entanglement, and so

on. In general, higher rank entanglements such as Habcd in the above require
a deal of filtering out of separations and cross-entanglements from the original
tensor product Hilbert space for their definition to be possible, which accounts
partly for the fact that entanglements are generally not as conceptually simple
or intuitive as separations. From this we can appreciate just how complicated
the entanglement structure of qubit registers can be.

Exercise 22.6 Show that the rank-four entanglement Habcd is given by

Habcd ≡ Ha Hb Hc Hd
−Habcd ∪ Hab•cd ∪ Hac•bd ∪ Had•bc ∪ Hbc•ad ∪ Hbd•ac ∪ Hcd•ab
∪Hab•cd ∪ Hac•bd ∪ Had•bc ∪ Ha•bcd ∪ Hb•acd ∪ Hc•abd ∪ Hd•abc .
(22.17)
 22.4 Partitions 297

Entanglement Partitions
The decomposition of a quantum register H into the union of disjoint separations
and entanglements will be called an entanglement partition of H, and each
element of that partition will be referred to as a partition element. We can now
list the various entanglement partitions that we have encountered, along with
the number of partition elements in each:
number of
rank register
partition elements
1 Ha 1 = B1
(22.18)
2 Ha Hb 2 = B2
3 Ha Hb Hc 5 = B3
4 Ha Hb Hc Hd 15 = B4
A calculation gives 52 partition elements in the entanglement partition of a
rank-five quantum register, which we recognize as B5 , the 5th Bell number.
The conjecture, then, is that the entanglement partition of a rank-r quantum
register has Br partition elements. This was proved by Eakins (Eakins, 2004),
and shows that the entanglement structure of large-rank quantum registers will
be too complicated to deal with without computer assistance.
The relationships between separations and entanglements are subtle, as are the
relationships between splits and entanglement partitions. Although the number
of partitions in the entanglement partition of a rank-r quantum register is the
same as the number of splits and given by the Bell numbers, splits and partitions
cannot coincide for r > 1. Every factor register in a split is a vector space,
whereas no partition element is a vector space. Both splits and partitions are
essential and unavoidable features in quantum mechanics and hence in the QDN
account of quantum measurement and causal set structure.
A final simplification in this line of investigation is to use the above superscript
notation to label the various elements of entanglements and separations. So, for
example, Ψabc•de•f gh denotes a state in the partition element Habc•de•f gh and so
on. Translated into direct terms, this means the following. First, this state is an
element of the rank-eight quantum register

H[8] ≡ Ha Hb Hc Hd He Hf Hg Hh . (22.19)

Second, we may write this state in the factorized form

Ψabc•de•f gh = ψ a ψ b ψ c ψ de ψ f gh , (22.20)

where ψ a ∈ Ha , ψ b ∈ Hb , ψ c ∈ Hc , ψ de ∈ Hde , and ψ f gh ∈ Hf gh .

Exercise 22.7 Investigate the superposition (vector addition) properties of

separations and entanglements.
298 Separability and Entanglement

22.5 Quantum Zipping

We may avoid the material presented in this chapter when dealing with most
situations discussed so far in this book, because those tend to involve relatively
low-rank quantum registers with subregisters of relatively low dimension. How-
ever, we envisage the application of QDN or its analogues to large-rank, large-
dimensionality contexts. Then it will be necessary to have some organization
principles available. The split, partition, separation, and entanglement concepts
discussed above appear unavoidable in this respect.
Consider the inner product between two pure labstates in a quantum register
H. Each state will come from a specific partition. Depending on the details of the
two partitions concerned, this inner product may or may not factorize. This is
because factor states can only take inner products in combinations that lie in the
same factor Hilbert space of some split of the H, a process we refer to as quantum
zipping. Using the notation for splits, partitions, separations, and entanglements
given in the previous chapter, the following example illustrates the point.

Example 22.8 Consider the rank-eight quantum register H[1...8] . By

inspection, the inner product of the states Ψ123•456•78 and Φ145•23•678 takes
the factorized form

Φ145•23•678 Ψ123•456•78 = (φ1 ψ 1 )(φ23 ψ 2 ψ 3 )(φ4 φ5 φ678 ψ 456•78 ), (22.21)

which cannot be simplified further. Figure 22.1 illustrates the factorization

structure of this inner product. The numbers 1 to 8 represent the individual
atoms of the quantum register, not necessarily detectors.

This example is reasonably simple, as the two labstates involved can be ordered
as shown. In general, more complex patters will occur, and then the correspond-
ing zip diagrams will be more complicated.
The separation and entanglement structure of labstates should be of impor-
tance in any experiment. A significant application of these concepts is to causal
sets, studied in the next chapter.

Figure 22.1. Quantum zipping.

 23
Causal Sets

23.1 Introduction
In this chapter we apply the concepts presented in Chapter 22 to a topic that
has been of much theoretical interest in recent decades: causal set theory (CST).
It is interesting and useful to review the status of CST relative to other
foundational topics in mathematical physics that have been and remain of prime
significance to empirical physics, namely, special relativity (SR), general relativ-
ity (GR), and quantum mechanics (QM). Those great frameworks are built on
foundations that presuppose the continuity of space and time.
Despite this, it is an empirical fact that we are surrounded by discreteness in
the form of atoms, so it is natural to see how that discreteness fits in with
the continuity of space and time. We have attempted in this book to show
that the application of discrete principles to QM can go far beyond the basic
Planck quantum of action. We have gone as far as discretizing empirical time,
in the form of the stage concept. However, as far as relativity in either of its
manifestations is concerned, the case for any form of discreteness has not yet been
established empirically, although great efforts are being made in that direction
in the program known as quantum gravity. We have particularly in mind the
discretization approaches to GR of Regge (Regge, 1961) and Penrose (Penrose,
1971).
Regge’s approach is a classically based discretization of the Riemannian man-
ifold concept, so has at best a generalized propositional classification (GPC)
of two.1 Penrose’s spin network paradigm started as a quantum description
of spacetime and has led to much work in the area of spin networks in the
hope of unifying GR and QM. However, workers in spin networks and related
branches of quantum gravity have focued their attention on the mathematical

1
When Regge’s approach is used to regularize Feynman path integrals in quantum gravity,
its GPC drops to one, as it then becomes an exercise in mathematical physics.
300 Causal Sets

complexities and hardly any on the role of observers and their apparatus, leading
to a GPC classification of quantum gravity as one, that is, as mathematical
physics and not physics. Claims made that such programs represent empirical
physics have not been empirically substantiated to date; quantum gravity at this
time remains empirically vacuous, despite the best efforts of many theorists over
many decades.2
In various attempts to account for the existence of space, time, and matter
in the Universe, physicists often adopt one of two opposing viewpoints. These
may be labeled bottom-up and top-down, reflecting the basic difference between
reductionist physics and emergent physics. QM makes an appearance in both
approaches because while it really models the emergent side of observation, its
mathematical structure has a strong reductionist flavor.
A number of bottom-up approaches to cosmology proceed from the assertion
that at its most basic level, the Universe can be represented by a vast collection of
discrete events embedded in some sort of mathematical space. For example, the
pregeometric approach asserts that conventional classical time and space and the
classical reality that we appear to experience all emerge on macroscopic scales
due to the complex connections between certain fundamental microscopic, unob-
served and unobservable, pregeometric entities. This approach was pioneered by
Wheeler (Wheeler, 1980) and has received attention more recently by Stuckey
(Stuckey, 1999).
A bottom-up approach to cosmology that we can relate to in quantized detector
networks (QDN) is the causal set hypothesis, which asserts that spacetime itself
is discrete at the most fundamental level. In CST models it is postulated that
classical, discrete events are generated either randomly or through some agency,
though neither the nature of these events nor the mechanism generating them
is explained or discussed in detail. QDN comes in cost-free at this point. In
this chapter we show how the particular mathematical properties and dynamical
principles of QDN generates causal set structures naturally, without any extra
assumptions.
Although it seems natural to generate causal sets by the discretization of a
pseudo-Riemannian spacetime manifold of fixed dimension, as is done in lattice
gauge theories and Regge’s approach to GR (Regge, 1961), causal set events
need not in principle be regarded as embedded in some background spacetime
of fixed dimension d. One view taken by some theorists is that conventional
(i.e., physical) spacetime emerges in some appropriate limit as a consequence
of the causal set relations between discrete events. If so, then it is reasonable
to expect that in the correct continuum limit, metric structure should emerge
naturally (Brightwell and Gregory, 1991). An even more intriguing hypothesis

2
This is not the same thing as QM over classical curved GR background spacetimes, for
which there is some empirical evidence, such gravitational lensing and the need to correct
GPS signals because of gravitational curvature and time dilation (Ashby, 2002).
 23.1 Introduction 301

is the suggestion that the dimension of this emergent spacetime might be scale
dependent (Bombelli et al., 1987), making the model potentially compatible with
GR in four spacetime dimensions, string theory and p-brane cosmology, and
higher dimensional Kaluza–Klein theories. Such a hypothesis fits in with the
general philosophy taken in this book that physics is contextual: different effective
theories may explain different experiments.

The Origin Problem

One of the problems we have in accepting any pregeometric approach is what
we may call the origin problem. This is the problem of explaining where any
foundational concept comes from. It seems scientifically inconsistent to try to
explain empirical physics on the basis of unverifiable hypotheses.
Mathematics came to terms with the origin problem in set theory by admitting
that there are concepts, such as those of set, integer, infinity, and infinitesimal,
that cannot be explained further and have to be taken as primitive concepts that
cannot be derived from deeper principles and have to be treated as axiomatic.
We have taken the same approach in our concept of primary observer. It is
not inconsistent to base our foundations on such a concept, because we our-
selves are primary observers: we can point to real, physical examples of the
concept.
Our approach to causal set structure then is predicated on the concept of a
primary observer. Any causal set structure being discussed will be contextual to
the primary observer involved and their apparatus, so we do not conjecture any
absolute pregeometric structure.
It turns out that the separation and entanglement properties of quantum
register labstates discussed in the previous chapter have enough structure in
them to provide the necessary causal set attributes of interest here. A novel
feature of our approach is the occurrence of two distinct but interleaved causal
set structures, in contrast to the one normally postulated in CST. One of these
causal sets arises from the entanglement and separation properties of labstates
themselves, while the other arises from the split and partition properties of
the quantum registers and operators representing modules. In our view, these
two distinct causal sets correspond to the two distinct classes of information
transmission observed in physics. One of these involves nonlocal quantum
correlations in signal labstates and does not respect Einstein locality. The
other involves classical information transmission equivalent to local operations
carried out by the observer on apparatus, and that does respect Einstein
locality.
In the next section we review classical CST. Then we show how the separation
and entanglement concepts discussed in the previous chapter lead to a natural
definition of the concepts of families, parents, and siblings used in CST. Then
we show how two sorts of causal set structure arise in QDN, one of which is
associated with labstates and the other with modules.
302 Causal Sets

23.2 Causal Sets

A number of authors (Bombelli et al., 1987; Brightwell and Gregory, 1991;
Markopoulou, 2000; Requardt, 1999; Ridout and Sorkin, 2000) have discussed the
idea that spacetime could be discussed in terms of causal sets. In the causal set
paradigm, the universe is envisaged as a set C ≡ {x, y, . . .} of objects (or events)
that may have a particular binary relationship among themselves denoted by the
symbol ≺, which may be taken to be a mathematical representation of a temporal
ordering. For any two different elements x, y, if neither of the relations x ≺ y nor
y ≺ x holds, then x and y are said to be relatively spacelike, causally independent,
or incomparable (Howson, 1972). The objects in C are usually assumed to be the
ultimate description of spacetime, which in the causal set hypothesis is often
postulated to be discrete (Ridout and Sorkin, 2000). Minkowski spacetime is an
example of a causal set with a continuum of elements (Brightwell and Gregory,
1991), with the possibility of extending the relationship ≺ to include the concept
of null or lightlike relationships.
The causal set paradigm supposes that for given elements x, y, z of the causal
set C, the following relations hold:
∀ x, y, z ∈ C, x ≺ y and y ≺ z ⇒ x ≺ z (transitivity)
∀x, y ∈ C, x≺y⇒y⊀x (asymmetry) (23.1)
∀x ∈ C, x ⊀ x. (irreflexivity)
A causal set may be represented by a Hasse diagram such as shown in Figure
23.1 (Howson, 1972). In a typical Hasse diagram, the events are shown as labeled
circles or spots and the relation ≺ as a solid line or link between the events, with
the “temporal” ordering running from bottom to top. Stage diagrams in QDN
are a form of Hasse diagram, with the added complication of including modules.
Hasse diagrams can be discussed in familial terms. For example, in Figure 23.1,
event 10 is the ancestor of all subsequent events; 22 and 32 are the parents of
their child 23 ; 32 and 42 are siblings; and so on.
One method of generating a causal set is via a process of “sequential growth”
(Ridout and Sorkin, 2000). At each step of the growth process a new element is
created at random, and the causal set is developed by considering the relations

Figure 23.1. A typical Hasse diagram.

 23.3 QDN and Causal Sets 303

between this new event and those already in existence. Specifically, either the
new event y may be related to another event x as x ≺ y, or x and y are said
to be unrelated. Thus the ordering of the events in the causal set is as defined
by the symbol ≺, and it is by a succession of these orderings, i.e., the growth
of the causet, that constitutes the passage of time. The relation x ≺ y is hence
interpreted as the statement: “y is to the future of x.” Further, the set of causal
sets that may be constructed from a given number of events can be represented
by a Hasse diagram of Hasse diagrams (Ridout and Sorkin, 2000).
The importance of CST is that in the large-scale limit of very many events,
causal sets may yield all the properties of continuous spacetimes, such as metrics,
manifold structure, and even dimensionality, all of which should be determined
by the dynamics (Bombelli et al., 1987). For example, it should be possible to use
the causal order of the set to determine the topology of the manifold into which
the causet is embedded (Brightwell and Gregory, 1991). This is the converse
of the usual procedure of using the properties of the manifold and metric to
determine the lightcones of the spacetime, from which the causal order may in
turn be inferred.
Distance may be introduced into the analysis of causal sets by considering the
length of paths between events (Bombelli et al., 1987; Brightwell and Gregory,
1991). A maximal chain is a set {a1 , a2 , ..., an } of elements in a causal set C such
that, for 1 ≤ i ≤ n, we have ai ≺ ai+1 and there is no other element b in C such
that ai ≺ b ≺ ai+1 . We may define the path length of such a chain as n − 1.
The distance d(x, y) between comparable (Howson, 1972) elements x, y in C may
then be defined as the maximum length of path between them, i.e., the “longest
route” allowed by the topology of the causet to get from x to y. This implements
Riemann’s notion that ultimately, distance is a counting process (Bombelli et al.,
1987). For incomparable elements, it should be possible to use the binary relation
≺ to provide an analogous definition of distance, in much the same way that light
signals may be used in special relativity to determine distances between space-like
separated events.
In a similar way, “volume” and “area” in the spacetime may be defined in
terms of numbers of events within a specified distance. Likewise, it should be
possible to give estimates of dimension in terms of average lengths of path in a
given volume. An attractive feature of causal sets is the possibility that different
spatial dimensions might emerge on different physical scales (Bombelli et al.,
1987), whereas in conventional theory, higher dimensions generally have to be
put in by hand.

23.3 QDN and Causal Sets

Our approach to causal sets in QDN differs from the above in that we do
not assume any structure to the information void, such as spacetime per se.
304 Causal Sets

Our discrete set structure arises from the separation and entanglement properties
of labstates and modules.
Furthermore, in the QDN approach, various relations assumed in the “sequen-
tial growth” mentioned above must be interpreted carefully. In quantum physics,
past, present, and future can never have the status they have in Block Universe
models. At best we can only talk about conditional probabilities, such as asking
for the probability of a possible future stage if we assumed we were in a given
present stage. This corresponds directly to the meaning of the Born interpretation
of probability in QM, where all probabilities are conditional.
Another point is that the causal set relations discussed by Sorkin and others
imply that the various elements a, b, . . . , z have an independent existence outside
the relations themselves and that these relations merely reflect some existing
attributes. This is a vacuous Block Universe perspective that QDN does not
accept.
A further criticism of classical CST comes from dynamics. In some diagrams,
relatively space-like events with no previous causal connection are permitted
to be the parents of the same event. The question arises then, given two such
unrelated events at a given “time” n, how any information from either of them
could ever coincide, that is, be brought together to be used to create any mutual
descendants. The only sensible answer is that there must be some external agency
organizing the flow of that information. But the whole point of standard CST,
however, is that there is no external space in which these events are embedded, or
any external “memory,” observer, or information store correlating such informa-
tion. It is not clear then how such processes could be encoded into the dynamics.
In our QDN approach, however, we have no such issue: the primary observer is
such an agency, manipulating apparatus in order to produce such processes.

23.4 Quantum Causal Set Rules

Stage diagrams are greatly useful in understanding the architecture of appa-
ratus. Similarly, causal set diagrams are greatly useful in understanding the
pattern of causality as labstates “flow” through that apparatus. With reference to
Figure 23.2, the conventions in causal set diagrams are as follows.

Time
Stages go up the page, starting from stage Σ0 .

Separations
These are represented by squares surrounding single integers identifying separa-
tions. As with detectors on a given stage in a stage diagram, labeling is arbitrary.
Our convention then is to label atoms left to right.

Entanglements
A rank-r entanglement is represented by a left to right rectangle equivalent to r
adjacent squares, labeled by the atoms involved in the entanglement.
 23.5 Case Study 1: The Double-Slit Experiment 305

Figure 23.2. The causal set diagram for the DS experiment.

Lines
These connect elements of the causal set as required from the modules involved.

23.5 Case Study 1: The Double-Slit Experiment

We are now in a position to reexamine the QDN analysis of the double-slit (DS)
experiment given in Chapter 10 and analyze its causal set structure.
Looking at Figure 10.2, we would get the impression that it is a Hasse diagram
as it stands and so the causal set structure of the experiment would appear to be
obvious. This is not what we are after, however, because the causal set structure
we are interested in involves labstates and not detectors. These are two separate
concepts.
Our point is this. The stage Σ1 labstate given in Eq. (10.13) is a signality-
one state with two identifiable components α1 A  1 01 and α2 A
 2 01 added together,
1 1
giving the impression that this is not an entangled state. If we return to a more
explicit signal basis representation, the entanglement at stage Σ1 becomes clear.
If instead of (10.13) we write

Ψ1 = α1 111 021 + α2 011 121 , (23.2)

then it is clear that Ψ1 is actually an entangled state, an element of the entan-

[12]
glement Q12 1 , one of the partition elements of the quantum register Q1 ≡
Q11 Q21 .
Note that in the above, the superscripts refer to the two slits, labeled 1 and 2.

Remark 23.1 In standard QM, the state at stage Σ1 in the DS experiment

would also usually be written in nonentangled form |Ψ1  = α1 |slit 1 +
α2 |slit 2, masking the fact that it is actually an entangled state (contextual
on the observer’s ability to look at both slits simultaneously). This is a far
more subtle point than it appears and goes to the heart of the difference
between classically conditioned thinking and the sort required to make sense
of quantum processes.
306 Causal Sets

K K

Figure 23.3. The causal set diagram for the monitored DS experiment.

The stage Σ2 labstate requires some commentary also, because Figure 10.2 is
misleading in the same respect as at stage Σ1 . The final stage labstate Ψ2 is an
element of the entanglement Q123...K
2 , where K is the number of detectors in the
detecting screen S. The final labstate may be written as a sum of computational
basis representation states:
 2 
K 
a j,a
Ψ2 = α U2,1 2j−1 2 . (23.3)
j=1 a=1

With this analysis of the entanglement structure of the DS experiment, the appro-
priate QDN causal set diagram is given by Figure 23.2. Temporally successive
stages run bottom to top. Separations are in square boxes, and entanglements
are in rectangular boxes, labeled by the atoms involved.

23.6 Case Study 2: The Monitored Double-Slit Experiment

The interference pattern observed in the DS experiment disappears when which-
path information is available to the observer, that is, when it is possible to identify
which path a photon had taken from source to detector. In this situation, known
as the monitored double-slit experiment, the labstate at stage Σ1 can no longer be
considered entangled. In consequence, there is doubling of final screen elements
in the relevant causal set diagram, as shown in Figure 23.3. While the two slit
elements at stage Σ1 are not considered entangled once which-path information
is obtained post each run, the labstate for stage Σ1 may be considered entangled
from the perspective of the observer at stage Σ0 , that is, before any which-path
information is obtained at stage Σ2 . This underlines the fact that such a causal
set diagram changes with context.
It should be noted that the two stage-Σ1 separations 11 and 21 in the monitored
DS experiment are physically disjoint, in that they start outcome branches that
do not occur in the same run. They are classically distinct. The stage-Σ1 labstate
is not the tensor product of elements from these two separations; neither is it a
superposition of those two separations.
A significant point to make here is that Figure 23.3 is not a Block Universe
diagram.
 23.7 Case Study 3: Module Causal Set Structure 307

23.7 Case Study 3: Module Causal Set Structure

We mentioned above that QDN leads naturally to another causal set structure,
one associated with the modules in a given experiment. For a given experiment,
this causal set structure is associated with the links between real and virtual
detector nodes, and on that account will be referred to as the dual causal set.
It is at this point that our stage diagram convention of representing modules by
boxes, rather than the circles associated with the detectors, becomes useful.
Given a conventional stage diagram, such as Figure 13.3, the two-photon
interferometer, the first step is to fill in the missing module boxes. These occur
when signals propagate through the information void directly. In Figure 23.4
we show how Figure 13.3 is amended: the new modules V 1 and V 2 represent
information void propagation.
The next step is to remove all clear circles (that is, except final stage detectors),
join up the module boxes according to how they interact with the real and
virtual detectors, and then readjust the resulting diagram into a vertical Hasse
form. Figure 23.4 gives Figure 23.5, the module Hasse diagram for the two

f
S C B B

Figure 23.4. The amended stage diagram for the two photon interferometer.

B B

f f V V

Figure 23.5. Two-photon interferometer module causal set diagram.

308 Causal Sets

photon interferometer. This casual set diagram is entirely classical: there are
no entanglements.
The reader is invited to analyze the other experiments discussed in this book in
terms of their labstate and apparatus causal set structures. It will become appar-
ent that there are significant complexities in even relatively simple experiments.
 24
Oscillators

24.1 Introduction
Our aim in this chapter is to show how quantized detector networks (QDN)
describe the quantized one-dimensional bosonic (harmonic) oscillator. The adjec-
tive bosonic here refers to the fact that in the standard quantum theory of
this system under observation (SUO), the phase space operators of position
and momentum satisfy commutation properties, in contrast to the fermionic
oscillator, whose corresponding variables satisfy anticommutation properties and
which is studied at the end of this chapter.
The one-dimensionality referred to here is somewhat misleading: it refers to
the classical mechanics (CM) theory of a single particle moving in one spatial
dimension under the influence of an attractive quadratic force potential. In the
quantized version of the same SUO, there are infinitely many degrees of freedom,
requiring the use of an infinite-rank quantum register. This is one of the few
occasions in QDN in which we refer to the concept of infinity.
Because the rank of the oscillator’s quantum register is infinite, we find that the
QDN representation comes with a tremendous amount of mathematical overkill
and redundancy. Specifically, the Hilbert space dimension of such a quantum reg-
ister is far bigger than that actually needed to describe a quantized bosonic oscil-
lator, the vast bulk of states in the register being what we shall call transbosonic.
As a Hilbert space, the infinite-dimension quantum register is nonseparable,
which means that it has no denumerable (that is, countable) basis, unlike the
Hilbert space HHO of energy eigenstates used to describe the standard QM
bosonic oscillator. While the improper position coordinate basis {|x : x real} is
nondenumerable, the normalized energy eigenstate basis for HHO is countable.
We will show how the QDN labstates that are the analogues of standard QM
oscillator states are restricted to a subspace of the register of measure zero and
remain there from stage to stage.
The degree of “overkill” in the QDN description of the oscillator is no different
in essence to that encountered in the quantum field theory (QFT) description
310 Oscillators

of the oscillator. QFT is a many particle theory that allows for the possibility
of more than one oscillator to be excited at the same time. This is in contrast
to Heisenberg’s matrix mechanics (Heisenberg, 1925) and Schrödinger’s wave
mechanics (Schrödinger, 1926), which are one-particle descriptions of the oscilla-
tor. Being based on signal detector principles, QDN can readily describe arbitrary
numbers of oscillator-type signals, over and above quantum superpositions of one-
particle signal states. This underlines the point that QDN looks much more like
a halfway house to QFT rather than another representation of standard QM.
The transbosonic states referred to above are really a reflection of that fact.

24.2 The Classical Oscillator Register

The first step in the QDN description of the harmonic oscillator is to define an
infinite-rank classical bit register R[∞] . This is an infinite, countable collection of
classical bits, each bit being labeled by a distinct nonnegative integer n running
from zero to infinity. We shall call R[∞] the classical oscillator register and write
R[∞] ≡ B 0 B 1 B 2 . . ., where B i is the ith classical bit.
Here we adopt the convention that R[∞] is the Cartesian product of the binary
sets B 0 , B 1 , and so on.
Our QDN approach to the harmonic oscillator raises a significant issue. Up
to this point, we have argued that no experiment deals with actual infinities, so
QDN has exploited that fact. Indeed, finiteness seems to be a strength of QDN
rather than a weakness. We should explain why we now introduce an infinite-rank
quantum register to describe one of the simplest of SUOs, the harmonic oscillator.
We have found that whenever such issues arise in QDN, the answers are to
be found by looking at what happens in the laboratory and what the observer
actually does. QDN discusses detectors, not the supposed objects being detected.
There are two contrasting facts about apparatus, however. The first fact is that
relative to any real observer, all apparatus comes in discrete packages, at any
empirical level.1 At the highest empirical level, which we may call the emergent
level, a detector has a single objectivized identity, which is as a detector.2 In
QDN we model this by a single binary detector. At the lowest empirical level,
which we may call the reductionist level, a real detector is described as a vast
but countable number of discrete components called molecules and atoms. At
that level, it need not be recognizable as a detector at all.
The second fact is that despite the countability associated with the emergent
and reductionist levels, there are three scenarios where infinity makes an
appearance.

1
By this we mean what is observable, not what is theorized.
2
If we view reductionism as the counterbalance to emergence, then the highest empirical level
corresponds to the lowest reductionist level, where minimal mathematical details are given.
 24.2 The Classical Oscillator Register 311

Modeling Space Is Not Easy

The great mathematical physicist Schwinger gave the following succinct state-
ment on how he viewed QM, apparatus, space, and time:

The mathematical machinery of quantum mechanics is a symbolic expression

of the laws of atomic measurement, abstracted from the specific properties of
individual techniques of measurement. In particular, the space-time manifold
that is the background of any quantum-mechanical description is an idealiza-
tion of the function of a measurement apparatus to define a macroscopic frame
of reference. (Schwinger, 1958)

This is in accord with the principles of QDN. If we wish to model space-

time itself, rather than just apparatus that is in space-time (a rather different
proposition altogether), then we are faced with the concept of indefiniteness
rather than infinity. By this we mean that there no natural, obvious, or observable
limits to space and time as far as any observer is concerned. Even if we discretized
space and time coordinates, thereby eliminating continuity, how many points
would we include in our modeling?

Apparatus Depends on Continuous Parameters

Continuity cannot be eliminated even when we have very simple apparatus. For
example, in the Stern–Gerlach experiment, the orientation of the main magnetic
field is parametrized by three angles in a continuum of angles, and there is no
natural way of discretizing any of those.

Quantum Process in the Information Void

QDN uses, but does not derive from reductionist principles, any transition ampli-
tudes from stage to stage: it is designed to work with the architecture of processes
and how those amplitudes are related to observable signals. The calculation of
such amplitudes will usually (but not invariably) involve working models of empty
space (the vacuum) that depend on continuity.
Every classical bit B i has two elements, 0i and 1i , representing “no signal ” and
“signal ” respectively. A rank-r classical register has 2r elements, so R[∞] contains
an infinite number of classical states. Representing these requires dealing with
infinite sets.
The mathematical difficulty here is that these sets are nondenumerable, that
is, are not countable. The paradox is that while the bits making up the register
are each of finite cardinality, and the bits themselves can be counted (our labeling
proves this), the number of states is not countable. Mathematically, this is the
same phenomenon that allows mathematicians to express every rational real
number in terms of a recurring decimal expansion, but none of the irrationals.
312 Oscillators

The Signal Basis Representation

The most natural representation of a classical state in R[∞] is the signal basis
representation (SBR), which we have met before. An arbitrary state Ψ in R[∞]
can be defined in the form
∞
0
Ψ ≡ i0 i 1 i2 . . . = ia , ia = 0a or 1a for a = 0, 1, 2, . . . (24.1)
a=0

Every such state therefore corresponds to a unique binary sequence SΨ ≡

{i0 , i1 , i2 , . . .}, consisting of an infinite string of ones and zeros.

Example 24.1 The register state

Ψ ≡ 00 01 12 13 04 15 06 17 18 09 110 111 012 113 114 015 116 . . . (24.2)

corresponds to the binary sequence

SΨ ≡ {0, 0, 1, 1, 0, 1, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, . . .}. (24.3)

The nondenumerability of R[∞] creates a potential problem when we come

to quantization, because the Hilbert space Q[∞] corresponding to R[∞] is an
infinite tensor product. Such Hilbert spaces are always nonseparable (Streater
and Wightman, 1964), which means that they have no countable basis. This con-
trasts with the fact that the Hilbert space of the standard QM quantized bosonic
oscillator has a complete, countable basis set, consisting of energy eigenstates.
Fortunately, in the QDN analysis for the harmonic oscillator, we can restrict
our attention to a set of special operators, referred to here as bosonic operators
over the infinite-rank quantum register Q[∞] , which have the merit that, in phys-
ical applications dealing with a single oscillator, the problems of nonseparability
can be avoided. A similar phenomenon occurs in relativistic quantum field theory
(Streater and Wightman, 1964; Klauder and Sudarshan, 1968).
To understand how this comes about, we first classify each state in R[∞] as one
of three possible types. Two of these types form countable subsets of the register,
while the third type forms a nondenumerable subset. These types correspond,
roughly speaking, to the integers, the rationals, and the irrationals, respectively,
in the real number system. This can be seen by the following heuristic arguments.

Finite Countable States

The first type, the set of all finite countable states in R[∞] , consists of
states associated with binary sequences that consist of zeros after some given
finite element J, which depends on the sequence. For example, the state
10 11 02 13 04 05 06 07 08 09 010 . . . is finite countable (J = 4), whereas the state
corresponding to the infinitely recurring sequence

10 01 12 03 14 05 16 07 18 09 110 . . . (24.4)
 24.2 The Classical Oscillator Register 313

is not finite countable. For a finite countable state i0 i1 i2 . . . iJ 0J+1 0J+2 . . . a

modification of the computational basis map (5.14) maps this state to the integer
i0 + 2i1 + 22 i2 + · · · + 2J ij , which is finite.

Recurring Sequence States

The second type, the recurring sequence states in R[∞] , consists of those
sequences that would be finite countable sequences but for the fact that the
infinite string of zeros after J is replaced by some nontrivial recurring binary
sequence. Recurring sequence states cannot be classified by finite integers using
the computational map (5.14). However, we can use another map, defined by
i1 i2
i 0 i1 i2 . . . → i 0 + 1
+ 2 + ··· (24.5)
2 2
to map such states into the interval [0, 2]. We shall call this the continuum map.
It is easy to see that, in fact, all states in R[∞] can be mapped into the interval
[0, 2] via the continuum map. The signal ground state 00 01 02 . . . maps into 0,
while the fully occupied state 10 11 12 . . . maps into 2. All other states necessarily
map into the open interval (0, 2).
It is not hard to see that finite countable states and recurring sequence states
map into the rationals via the continuum map, but not in a one-to-one way. For
example, the finite countable state 10 01 02 03 . . . maps into the number 1 by the
continuum map, which is also the value mapped from the recurring sequence
state 00 11 12 13 . . .

Remark 24.2 In standard decimal-based arithmetic, it is generally asserted

that the infinitely recurring decimal 0.9̇ ≡ 0.9999 . . . is “equal” to the number
1. While the register states 10 01 02 03 . . . and 00 11 12 13 . . . map to the same
value 1 via the continuum map, physically, these are two very different states.
This underlines the difference between pure mathematics (0.9̇ = 1) and physics
(10 01 02 03 . . . = 00 11 12 13 . . .).
Not all pure mathematicians would agree that 0.9̇ = 1 is an absolute
equality; some would argue that the difference 1 − 0.9̇ is an infinitesimal.
But not every mathematician accepts the concept of infinitesimals as sound.

Irrational Sequence States

The problem with nondenumerability arises because of the existence of the third
type of infinite binary sequence. This consists of all those binary sequences that
are not recurring, such as
{1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, . . .}, (24.6)
an example based on the successive digits in the decimal representation of π.
There are infinitely many such sequences and they cannot be counted, as they
correspond to the irrationals, which is easy to prove.
314 Oscillators

Our conclusion is, therefore, that states in R[∞] cannot be classified by the
integers alone. Instead, we may use the sequence corresponding to each state as
an index. Specifically, if S is the binary sequence
S ≡ {sa : sa = 0 or 1 for a = 0, 1, 2, . . .} , (24.7)
then the corresponding state S in R [∞]
is given uniquely by the expression
∞
0
S ≡ s 0 s1 s2 . . . = sa . (24.8)
a=0

Remark 24.3 The zero sequence Z ≡ {0, 0, 0, . . .} corresponds to the signal

ground state, denoted 00 01 02 . . . in the occupation representation and 0 in
the computation representation. This state is not the QDN analogue of the
conventional oscillator ground state |0.

Given two binary sequences S, T corresponding to register states S and T ,

respectively, their “inner product” ST is defined in the obvious way, viz.
 ∞  ∞  ∞
0 0 0
ST = s a
tb ≡ sa ta ≡ δ ST , (24.9)
a=0 b=0 a=0

where the generalized Kronecker symbol δ ST takes the value unity if and only if
the binary sequences S and T are identical; otherwise, it is zero.

Definition 24.4 Two binary sequences S ≡ {sa : a = 0, 1, 2, . . .} , T ≡

{ta : a = 0, 1, 2, . . .} are identical if and only if sa = ta for all nonnegative
integers, that is, for a = 0, 1, 2, . . .

We take it as an axiom that, given two binary sequences S and T , the product
1
∞
sa ta is 1 if the sequences are identical and 0 otherwise.
a=0

24.3 Quantization
Quantization amounts to associating the set of states in R[∞] as a (preferred)
basis B [∞] for a nonseparable Hilbert space Q[∞] . States in Q[∞] will be called
quantum register states and are of the form

Ψ= Ψ (S) S, (24.10)
S

where the summation is over all possible infinite binary sequences S and the
coefficients Ψ (S) are complex. The Hilbert space inner product is defined in the
obvious way, that is,
  
 
ΨΦ = Ψ(S)S Φ(T )T
S T

= Ψ(S)Φ(T )

ST = Ψ∗ (S) Φ (S) . (24.11)
S T δ ST S
 24.4 Bosonic Operators 315

As discussed above, finite countable sequences can be mapped into the integers
via the computational map (5.14). For a sequence S that maps into integer n,
we can use the notation n rather than S to denote the corresponding quantum
register state.

24.4 Bosonic Operators

The quantum register states in QDN required to represent the standard discrete
set of quantized bosonic oscillator energy eigenstates {|n : n = 0, 1, 2, 3, . . .}
form a subset of the finite countable states. We shall call any normalized element
of this subset a bosonic state. To identify this subset, we need to filter out of the
set of all finite countable states those states that are redundant. To do this, we
first define the bosonic projection operators as follows.
Each component B a of the classical oscillator register R[∞] is a classical
bit with two elements. Quantization starts with the interpretation of the two
bit states 0a , 1a in B a as the two preferred basis quantum outcomes states
of a quantum bit, Qa , representing the two states, ground and signal, of a
detector. Each quantum bit Qa is a two-dimensional
 Hilbert
 space with its own
set of projection and signal operators P a , P a , Aa , A
 a satisfying Table 24.1, a
generalization of Table 4.1.
The quantized bosonic register Q[∞] is the infinite-dimensional Hilbert space
given by the commuting tensor product of all the individual qubits, i.e.,
Q[∞] ≡ Q0 ⊗ Q1 ⊗ Q2 ⊗ · · · = Q0 Q1 Q2 . . . (24.12)
Here and below we shall drop the tensor product symbol ⊗, it being understood
we are dealing with commuting tensor products.

Definition 24.5 The set of bosonic filter operators {P a : a = 0, 1, 2, . . .} is

B
defined by the commuting tensor product
⎧ ⎫
⎨0∞ ⎬ a
a ≡
P P b
P  a P a+1 . . . , a = 0, 1, 2, . . . (24.13)
 = P 0 P 1 . . . P a−1 P
B
⎩ ⎭
b=a

Each bosonic projection operator P a defines a one-dimensional Hilbert sub-

B
space of Q [∞]
with basis {2 } in the computational basis representation. Eigen-
a

states of these operators with eigenvalue +1 will be called bosonic eigenstates.

Table 24.1 Products of signal bit operators

Pa a
P Aa a
A

Pa Pa 0 Aa 0
a
P 0 a
P 0 a
A
Aa 0 Aa 0 Pa
a
A a
A 0 a
P 0
316 Oscillators

Theorem 24.6 The bosonic filter operators satisfy the multiplicative rule

P b
a P ab  a
B B = δ PB , a, b ∈ N ≡ {0, 1, 2, . . .}. (24.14)

The element 2n in Q[∞] corresponds to the energy eigenstate |n in the stan-
dard quantum theory of the oscillator.
The bosonic filter operator a should not be confused with the signal projec-
P
1 B a
 ≡
tion operators P a ∞ b
I P  , where I b is the identity operator for qubit
b=a

Q . The difference is based on logic: an eigenstate of P

b a will return a positive
signal if detector a is examined, regardless of whatever signal state any other
a will return a positive signal in detector
detector is in, whereas an eigenstate of PB
a only if all the other detectors are each in their signal ground state.
Given the bosonic projection operators Pa , the next step is to define the bosonic
identity operator
∞

IB ≡ a .
P (24.15)
B
a=0

This operator satisfies the idempotency condition required of any projection

operator, that is,
IB IB = I B (24.16)
and plays the role of a “bosonic filter,” passing only those states and operators
associated with the quantum register that have the desired properties associated
with the harmonic oscillator.

Definition 24.7 A quantum register operator O is bosonic if and only if it

commutes with the bosonic identity, i.e.,

O bosonic ⇔ [O, IB ] = 0. (24.17)

Definition 24.8 A bosonic state is defined to be any vector in Q[∞] that

is an eigenstate of IB with eigenvalue +1. All other states in Q[∞] will be
referred to as transbosonic.

The set of all bosonic states is denoted QB and is the QDN analogue of the
Hilbert space of quantum oscillator states in standard QM.
Examples of transbosonic states are the signal ground state 0 and all those
finite countable elements p of the computational basis B where p is not some
power of two. In fact, almost all elements in the quantum register are trans-
bosonic. It can be readily verified that linear superpositions of bosonic states are
always bosonic states, while the linear superposition of any transbosonic state
with any other state in the register is always transbosonic.
 24.5 Quantum Register Oscillator Operators 317

The importance of the bosonic operators is that when they are applied to
bosonic states, the result is always a bosonic state, which can be easily proved.

24.5 Quantum Register Oscillator Operators

We now in a position to discuss how we map the standard quantum oscillator
into the quantum register.
In the standard QM of the bosonic oscillator, the most important operators
are the ladder operators a and a† . These satisfy the commutation relation
 
a, a† = 2I, (24.18)
where I is the identity operator in the standard oscillator Hilbert space H and we
take Planck’s constant  and the oscillator constant ω to be unity. These ladder
operators have the representations
∞

a= |n (n + 1) 2n + 1|,
n=0
∞
a† = |n + 1 (n + 1) 2n|, (24.19)
n=0

where the states |n, n = 1, 2, 3, . . . are the usual orthonormal excited states of
the oscillator ground state |0. The key to the QDN description is the observation
that these states are identified one-to-one with the bosonic states 2n discussed
above, namely,
|n ↔ 2n , n = 0, 1, 2, . . . (24.20)
We note that
|0 ↔ 10 01 02 03 . . . = 1, |1 ↔ 00 11 02 03 . . . = 2, (24.21)
and so on. A particularly significant observation is that the quantum register
signal ground state 0 is not the oscillator ground state |0. This is one of the
reasons we felt it necessary to introduce nonstandard notation for labstates:
keeping a clear distinction between those and conventional QM states is more
than a matter of notion but reflects deeper interpretational issues.
To find a quantum register representation of the ladder operators, we first
introduce some auxiliary notation. We define the register operators
∞  ⎧ ⎫ ⎧ ⎫
0 ⎨0 ∞ ⎬ ⎨0 ∞ ⎬ a
P≡ P a , Aa ≡ I b Aa , A a ≡ Ib A  . (24.22)
⎩ ⎭ ⎩ ⎭
a=0 b=a b=a

None of these operators is bosonic.

Exercise 24.9 Prove that the operators P, Aa , and A a do not commute

with the bosonic identity IB defined by Eq. (24.15).
318 Oscillators

The operators A  a can be used to generate bosonic states from the signal ground
state 0, which is transbosonic.3 Specifically, we have
 a 0,
2a = A a = 0, 1, 2, . . . (24.23)
With these definitions we construct the operators
a ≡ A
B  a PAa+1 ,  a+1 PAa .
Ba ≡ A (24.24)
Remarkably, these operators are bosonic, as can be readily proved from the fact
that
Aa IB = Aa . (24.25)
We find for a = 0, 1, 2, . . .
 '
∞
Ba ≡ ⊗ P b Aa A  a+1 = P 0 P 1 . . . P a−1 Aa A a+1 P a+2 P a+3 . . .
 b=a,a+1
'
∞
a ≡
B ⊗ Pb A  a Aa+1 = P 0 P 1 . . . P a−1 A a Aa+1 P a+1 P a+3 . . .
b=a,a+1
(24.26)
These operators satisfy the relations
 a Bb = δ ab Pa+1 ,
B  b = δ ab Pa .
Ba B (24.27)
Then the standard ladder operators a, a† take the quantum register
representation
∞

a ↔ aB ≡ 2 (n + 1)Bn ,
n=0
∞
a† ↔ a†B ≡ n.
2 (n + 1)B (24.28)
n=0

These operators are bosonic and have the commutation relation

2 3
aB , a†B = 2IB . (24.29)
Because these register ladder operators commute with the bosonic identity IB ,
any states that they create from bosonic states are also bosonic. We find for
example
 n
 † n
a a†B
|n ≡ √ |0 ↔ √ 1 = 2n , n = 0, 1, 2, . . . (24.30)
2n n! 2n n!
Note that aB annihilates the bosonic ground state 1, giving the zero vector 0 in
the quantum register Q[∞] , not the signal ground state 0.

3
The signal ground state 0 is not the same as the oscillator ground state |0, which is
identified with the quantum register state 20 = 1 in the computational basis representation
and 10 01 02 03 . . . in the occupation basis representation.
 24.6 Comparison with Quantum Field Theory 319

 ≡ 1 a† a +
The standard QM bosonic oscillator Hamiltonian operator H 1
has
2 2
quantum register representation
∞ 
 
 ↔ H B ≡ 1 a† aB + 1 IB =
H n+
1
PnB . (24.31)
2 B 2 2
n=0

This operator commutes with the bosonic identity IB , and therefore, any states
in the quantum register that start off bosonic at initial time remain bosonic, if
they evolve under this Hamiltonian.

24.6 Comparison with Quantum Field Theory

Several factors indicate that the QDN formalism is a halfway house between fixed-
particle number Schrödinger–Dirac quantum mechanics and the multiparticle
formalism of quantum field theory (QFT). First, the QDN strategy focuses
not on particles but on signals. There is no intrinsic requirement to conserve
signality except in those experiments where there is a physical reason, such
as the conservation of some quantum number, such as electric charge. Second,
the existence of transbosonic states in the QDN register Q[∞] is a clear marker
that the QDN formalism can accommodate the equivalent of QFT multiparticle
states.
An important similarity between QDN and QFT is the existence of the Fock
vacuum state |0 in QFT and the signal ground state 0 in QDN. We recall that
in QFT, Fock space F is defined by expansions of the form

F ≡ {|0} ⊕ H ⊕ S(H ⊗ H) ⊕ · · · , (24.32)

that is, as the infinite direct sum of appropriately symmetrized tensor products
of copies of a single-particle Hilbert space H. Looking at the above presented
QDN version of the bosonic oscillator, we see that the QDN signal ground
state 0 corresponds to the QFT vacuum |0; the QDN bosonic state space
QB corresponds to H; and the transbosonic states in QDN correspond to the
multiparticle states in QFT.
The QDN formalism may be interpreted as an attempt to encode Fock’s vision
of QFT into a mathematical formalism that is based on detectors rather than
SUOs. In the next chapter, we go further in this respect by extending the QDN
formalism to allow for the possibility of constructing the signal ground state
0 itself through the creation of apparatus in the laboratory from a state of
nonexistence, denoted ∅, that corresponds to the information void that we have
focused on in other chapters. The information void represents the state of a
laboratory in which there are no detectors whatsoever. However, this certainly
does not mean that there is no observer or laboratory.
320 Oscillators

24.7 Fermionic Oscillators

In this section we extend the above discussion of the bosonic oscillator to the
fermionic oscillator, that is, a system under observation (SUO) that requires
anticommuting degrees of freedom in its classical (Martin, 1959b; Casalbuoni,
1976a) and quantum formulations (Candlin, 1956; Martin, 1959a; Casalbuoni,
1976b).
Not long after the discovery of quantum mechanics by Heisenberg and
Schrödinger, Jordan and Wigner showed how to describe fermions in quantum
register terms (Jordan and Wigner, 1928; Bjorken and Drell, 1965). Their
construction of local fermionic quantum field operators requires tensor product
contributions from all of the qubits in a quantum register. In a QDN approach to
fermionic quantum fields (Eakins and Jaroszkiewicz, 2005), their techniques were
used to describe fermionic quantum fields using an infinite-rank quantum register
associated with a net of detectors distributed throughout all of physical space.
Because the Jordan–Wigner construction requires nontrivial contributions from
all qubits in the register, fermionic fields are manifestly and inherently nonlocal
in QDN.
In contrast with the bosonic oscillator studied above, we may restrict attention
to a finite-rank quantum register Q[N ] ≡ Q1 Q2 . . . QN .
The most convenient basis here is the signal basis representation. We follow the
approach outlined in Bjorken and Drell (1965) for the construction of fermionic
operators.
We use all the bit operators discussed previously and introduce a new one,
denoted σ a , for each qubit a = 1, 2, . . . , N in the register, defined by
 a,
σa ≡ P a − P a = 1, 2, . . . , N. (24.33)

a , a = 1, 2, . . . , N :
Next, we define a set of nonlocal operators, αa , α

α1 ≡ A1 I 2 I 3 . . . I N ,
αa ≡ σ 1 σ 2 . . . σ a−1 Aa I a+1 . . . I N ,
1
α ≡A 1I 2I 3 . . . I N ,
a
α ≡ σ 1 σ 2 . . . σ a−1 A a I a+1 . . . I N , a = 1, 2, . . . , N. (24.34)

These are the required “fermionic field operators.” They satisfy the anticommu-
tation relations

αa αb + αb αa = 0, a α
α b + α
b α
a = 0,
b + α
αa α b αa = δ ab I, 1  a, b,  N. (24.35)

where I ≡ I 1 I 2 . . . I N is the register identity operator.

Exercise 24.10 Use the signal bit algebra listed in Table 4.1 to prove
(24.35).
 24.7 Fermionic Oscillators 321

b operators have the following properties, which are easy to prove:

The αa , α

αa 0 = 0,
a 0 = 0,
α
 α
α  0 = 0, a = 1, 2, . . . , N.
a a
(24.36)

This means that we can reconstruct all the relevant structure of fermionic field
theory, the operators α a and αa playing the role of fermionic creation and
annihilation operators.
A significant feature of the above anticommutation relations is that they are
achieved via the use of the nonlocal operators and do not invoke Grassmannian
(anticommuting) numbers to do so.
 25
Dynamical Theory of Observation

25.1 Introduction
In this chapter we discuss how quantized detector network (QDN) theory can
be extended to cover the creation and decommissioning of apparatus, from the
perspective of observers and their laboratories. This extension will be referred to
as extended QDN. It allows us to discuss bit power sets, laboratories, the universal
register, contextual vacua, and the creation of quantized detector apparatus.
This extended formalism is used to describe the Elitzur–Vaidman bomb-tester
experiment and the Hardy paradox experiment.
A central concept running through this book is that of the observer, the
enigmatic “I ” of I think therefore I am. The problem is that, despite the many
triumphs of quantum mechanics (QM), the physics of the observer and observa-
tion is still not well understood.
Regardless of how observers are defined and whether classical or quantum prin-
ciples are involved, physicists generally believe that classical information in some
form is extracted from systems under observation (SUOs) in actual physics exper-
iments. In all branches of science, their language reflects this belief: experimental-
ists talk of measuring an electron’s spin or the mass of a new particle, and so on.
The conceptual issues in quantum mechanics such as wave–particle duality,
quantum interference, and nonlocality gave the first indication that all might
not be well with this perspective. We need only to look at the photon concept to
appreciate some of the problems with the idea that photons are particles (Paul,
2004). There are experiments where a photon (that is, a signal) is detected from
a crystal, but the atom “from whence it came” cannot be identified, suggesting
that a cooperative process is involved, rather like the phonon concept in the
physics of crystals.
The problem we have with the photon-as-particle concept is the gap in logic.
If the particle interpretation is taken literally, then obvious questions about
the physical structure or its equivalent of such particles spring up. Does such
 25.2 Power Bits 323

a particle have a “surface”? If it has a surface, what is that surface made of?
These and other simple (minded) questions quickly lead to the conclusion that
the particle concept is a convenient and economical objectification of context. In
other words, a useful illusion.
Some particle concepts appear better in this respect than others. For instance,
fermion number is conserved in astrophysical processes, but photon number is
not. Photons from the sun cannot be said to have traveled from the deep interior,
whereas we could say so about neutrinos.
So much for the signals that observers detect. What about the observers them-
selves? It is surely too great an assumption to think of them as a mere auxiliary
phenomenon in physics. Without observers, physics is a vacuous subject.
What seems to be missing is a comprehensive dynamical theory of observation
that would treat observers and SUOs more on the same footing. In such a
theory, observers would be subject to the same laws of physics as the SUOs
that they were observing. Such a theory would be capable of accounting for
the creation and annihilation of observers and their apparatus, as well as states
of SUOs, because in the real world, that is what goes on and nothing lasts
forever.
We do not have such a theory. That may come only once the physics of
emergence has been much better understood. However, we can make a start
with what we have at present in what we think may be a useful direction
(Jaroszkiewicz, 2010). In this chapter, we develop an extension of the binary
truth value concepts explored in earlier chapters, aimed at describing the possible
states of apparatus in a more general form than having just two states, yes or no.

25.2 Power Bits

We have up to this point identified the two possible normal signal states, ground
and signal, of a functioning binary detector as the two elements of a bit, denoted
0 and 1, respectively. As we have mentioned, however, bits are not vector spaces
and there seems to be no meaning to the addition of bit state 0 to bit state 1,
or even of the multiplication of a bit state by a real or complex number.
There is in fact a way of defining bit state addition, of a kind, in terms of
set theory. We recall that the power set P(S) of a set is the set of all possible
subsets of S including the empty set ∅ and S itself. Recall also that the number of
elements in the power set of a set of cardinality c is 2c . Therefore, we expect the
power set P(B) of a bit B to have four distinct, nonempty elements. These are

P(B) = {0 , 1 , 2 , 3 },
∼ ∼ ∼ ∼
(25.1)

where we define

0 ≡ {0}, ∼
∼
1 ≡ {1}, ∼
2 ≡ {0, 1}, ∼
3 ≡ {∅}. (25.2)
324 Dynamical Theory of Observation

In this scheme, the set {∅} is a nontrivial element of P(B) and counts as one
element of the power set. We shall refer to the power set of a bit as a power bit.
In this chapter, we shall work in terms of the elements of P(B) rather than with
the elements of B itself, identifying elements ∼ 0 and ∼1 of P(B) as synonymous
with bit states 0 and 1 of B, respectively. The value of using P(B) rather than B
itself is that the elements of the former are sets, so we can use the set properties
of union and intersection to make propositions about power bits equivalent to
those in the algebra of sets (Howson, 1972).

Interpretation
Before we proceed further, however, we need to resolve the following problem:
the power set P(B) of a bit B appears to have too many elements for a physical
interpretation. Logic suggests that only the elements ∼0 and ∼
1 of the power set are
actually needed in an experiment. What could the elements ∼ 2 and ∼
3 represent?
It has been our experience that many such questions in QDN are answered
by looking at the situation in question and identifying any hidden assumptions.
Consider an experimentalist who is going into a laboratory in order to determine
the signal status of a certain detector. Even before they could get an answer, one
glaring fact has to be true: that the detector actually exists.
Now whether or not a detector actually exists in a laboratory (“existence”
being defined as a detector actually being found and recognized as such by an
observer coming into that laboratory) is surely an empirical question, the answer
to which the observer cannot assume is “yes, it exists,” before they enter the
laboratory. This then is where one of our two extra elements of the power set
P(B) finds a natural interpretation. We interpret the element ∼ 3 ≡ {∅} as a state
of physical nonexistence, in the laboratory, of the detector in question. We shall
call it the void state of a detector.
It will undoubtedly seem strange to assign a state in our formalism to a
condition of nonexistence of a detector. But if we want to construct a theory
where apparatus can be created or destroyed, then that is precisely what we
have to do.
As for the remaining element, ∼2 ≡ {0, 1}, of the power set P(B), the ambiguity
in the elements it has (0 and 1) fits in nicely with an obvious physical condition
that a detector could be in: a state where the apparatus exists physically but is
not operating normally, to the extent that it is not registering either a ground
state or a signal state. For instance, such a detector could be faulty. Or it could
have been deliberately taken off-line (or decommissioned). We shall call such a
condition of a detector the faulty state, the off-line state, or the decommissioned
state.
Given the above four states of a detector, we can find four natural power bit
∼
questions, denoted i , i = 0, 1, 2, 3. These are explained in Table 25.1, along with
the answers they induce when asked of each of the four power bit states.
 25.4 Matrix Representation 325

Table 25.1 The extended QDN questions and answers

Question in words Symbol 0

∼
1
∼
2
∼
3
∼

∼
Is this detector in its ground state? 0 1 0 0 0
∼
Is this detector in its signal state? 1 0 1 0 0
∼
Is this detector off-line? 2 0 0 1 0
∼
Is this detector in its void state? 3 0 0 0 1

We may summarize these answers by the rule

∼
i j = δ ij , 0  i, j  3. (25.3)
∼

25.3 Power Bit Operators

A power bit operator is any mapping from the power set P(B) back into the
power set. Given an element ∼i of P(B) and a power bit operator ···
O , then we
O ∼i . There is a total of 44 = 256
denote the value of the operator’s action on ∼i by ···
different bit operators and only a few will be of use to us.

25.4 Matrix Representation

A useful way of representing power bit operators is via matrices. The elements
i of P(B) may be represented by column matrices [∼i ] given by
∼

⎡ ⎤ ⎡ ⎤
1 0
⎢0⎥ ⎢1⎥
0  [0
∼ ∼
] ≡⎢ ⎥
⎣0⎦ , 1  [1
∼ ∼
] ≡⎢ ⎥
⎣0⎦ , and so on. (25.4)
0 0
O on ∼i by the action of a power bit
We represent the action of bit operator ···
matrix [O
···
] on a column matrix [∼i ], such that

O i  [O
··· ∼
][ i ] ≡ [O
··· ∼
i ].
··· ∼
(25.5)
∼
In this matrix representation the dual elements i are represented by the row
matrices
∼   ∼  
0 1 0 0 0 , 1 0 1 0 0 ,
∼   ∼  
2 0 0 1 0 , 3 0 0 0 1 . (25.6)
This is consistent with the question and answer relations (25.3).
We can use the power bit matrix representation to define the operational
∼
meaning of the dyadics ∼i j . These can then serve as formal basis elements in
326 Dynamical Theory of Observation

the expansion of power bit operators. Given a power bit operator defined by
3 ∼
O = i=0 ···
(25.5), we can write it as the formal (dyadic) expression ··· O ∼i i .
Products of power bit operators are defined in the natural way: given power bit
operators M
···
,N
···
, we define their “product” M N by its action on any element ∼i
··· ···

of the power set P(B) according to the rule (M N)i ≡ M

··· ··· ∼ ···
{N i }. This product
··· ∼

rule is associative but not commutative, which can be seen from the matrix
representation.

25.5 Special Operators

In the following, we have to take into account that a bit power set does not
contain a zero element, because such a set is not a vector space. The void element
3 is not physically meaningless quantity, either: it represents a definite state in
∼

a laboratory. Therefore, the bit operators we discuss in this section cannot map
elements of a bit power set to any state other than ∼ 0, ∼
1, ∼
2 , or ∼
3.
The operators we define below have been constructed on the basis of what
seems reasonable to us. The following bit operators seem physically reasonable,
practically useful, and unavoidable in the kind of formalism we are aiming for.

Identity ···
I
This operator maps every element back into itself, i.e., ···
I ∼i = ∼i . Its matrix
I ]ij = δij .
elements are given by the Kronecker delta, that is, [···

Annihilator ···
Z
This operator maps any element ∼i of the power set P(B) into the void state
3 ≡ {∅}, that is, ···
∼
Z ∼i = ∼
3 , i = 0, 1, 2, 3. Its matrix representation is
⎡ ⎤
0 0 0 0
⎢0 0 0 0⎥
[Z ] =⎢
⎣0
⎥. (25.7)
··· 0 0 0⎦
1 1 1 1
The annihilator has a fundamental role in our theory: it represents the process of
destroying and removing from the laboratory all traces of an already physically
existing detector.

P and ···
The Power Bit Signal Projection Operators ··· P
These operators have the action
P0 = ∼
··· ∼
0, P1 = ∼
··· ∼
3, P2 = ∼
··· ∼
3, P3 = ∼
··· ∼
3,
 0 = 3,  1 = 1,  2 = 3,  3 = 3, (25.8)
P
··· ∼ ∼
P
··· ∼ ∼
P
··· ∼ ∼
P
··· ∼ ∼
 25.5 Special Operators 327

so their matrix representations are

⎡ ⎤ ⎡ ⎤
1 0 0 0 0 0 0 0
⎢ 0 0 0 0⎥ ⎢0 1 0 0⎥
[P ] =⎢ ⎥ ] =⎢ ⎥.
··· ⎣ 0 0 0 0⎦ , [···
P ⎣0 0 0 0⎦
(25.9)
0 1 1 1 1 0 1 1

These operators are idempotent, namely, ···

P ···
P = ··· P
P , ···
P  = P and power bit
··· ···

orthogonal , namely, ···  =P

P ···
P  P = Z . In this context, the annihilator Z plays
··· ··· ··· ···

the role of a zero element.

A, ···
The Power Bit Signal Creation and Annihilation Operators ··· 
A
These operators are defined principally by their action on the normal signal states
0 and ∼
∼
1:

A0 = ∼
··· ∼
3, A1 = ∼
··· ∼
0, A2 = ∼
··· ∼
3, A3 = ∼
··· ∼
3,
 0 = 1,  1 = 3,  2 = 3,  3 = 3, (25.10)
A
··· ∼ ∼
A
··· ∼ ∼
A
··· ∼ ∼
A
··· ∼ ∼

which gives the matrix representations

⎡ ⎤ ⎡ ⎤
0 1 0 0 0 0 0 0
⎢ 0 0 0 0⎥ ⎢1 0 0 0⎥
[A ] =⎢ ⎥  ] =⎢ ⎥.
··· ⎣ 0 0 0 0⎦ , [···
A ⎣0 0 0 0⎦
(25.11)
1 0 1 1 0 1 1 1

These operators are power bit nilpotent, that is, ···

A··· AA
A = ···  = Z , with Z once
··· ··· ∼

again playing the role of a zero element.

 , A, and A
P , ···
The product rules for the operators ··· P  are given in Table 25.2.
··· ···

Comparison with Table 4.1 shows that these tables are isomorphic, provided the
Z in Table 25.2.
zero operator Z in Table 4.1 is identified with the annihilator ···

Construction Operator ···

C
This operator acts on every element ∼i of power bit set and sets it to the signal
ground state in readiness for observation, i.e., ···
C ∼i = ∼
0 , i = 0, 1, 2, 3. There are

Table 25.2 The product table for the

special operators

P &
P A &
A
··· ··· ··· ···

P
···
P
···
Z
··· ···
A Z
···
&
P Z &
P Z &
A
··· ··· ··· ··· ···
A
···
Z
···
A
··· ···
Z P
···
&
A &
A Z &
P Z
··· ··· ··· ··· ···
328 Dynamical Theory of Observation

two scenarios. If a detector is in its void state, then it does not exist, so the
action of the construction operator represents the physical construction of that
detector, set up in its ground state in the laboratory, prior to any experiment.
It is assumed that facilities exist in the laboratory for this. Alternatively, if the
detector already exists, then the construction operator resets it to its ground
state if it is normal or repairs it and sets it to its ground state if it is faulty.
The construction operator is represented by the matrix
⎡ ⎤
1 1 1 1
⎢ 0 0 0 0⎥
[C
···
] =⎢
⎣ 0 0 0 0⎦ .
⎥ (25.12)
0 0 0 0

Decommissioning Operator D
···

This operator represents the action of decommissioning an already existing detec-

2 . This operator does not reset states
tor, setting it into its decommissioned state ∼
0 , 1 , or ∼
∼ ∼
2 to the void state ∼
3 because in the real world, there will invariably be
some remaining information in the form of debris that will inform the observer
that apparatus has been decommissioned. This is an important feature of our
discussion toward the end of this chapter of the Elitzur–Vaidman bomb-tester
experiment and Hardy’s paradox experiment.
The decommissioning operator is represented by the matrix
⎡ ⎤
0 0 0 0
⎢0 0 0 0⎥
[D
···
]= ⎢
⎣1 1 1 0⎦ .
⎥ (25.13)
0 0 0 1

25.6 The Laboratory

In extended QDN it is assumed that an observer exists in a physical environment
referred to as the laboratory, Λ. This will have the facilities for the construction
or introduction of apparatus consisting of a number of detectors. At any given
stage Σn , the observer will associate a state known as a generalized labstate to
the collection of detectors at that time. This state could be a pure state or a
mixed state. We shall restrict our attention in this chapter to pure labstates for
reasons of space.
Extended QDN is based on what really happens in laboratories. Suppose an
observer walks into a laboratory, intending to look at a given detector. There are
the following four classical scenarios.

No Detector
The detector in question does not exist. The observer will see this; it is an
empirical fact that has physical significance. This absence is represented by
3.
extended labstate ∼
 25.7 The Universal Register 329

Faulty Detector
The detector in question exists physically, but is faulty, off-line, or decommis-
sioned. This situation is represented by extended labstate ∼2.

Ground State
The detector in question exists physically, is in perfect working order, and is in
its ground state. This is represented by extended labstate ∼0.

Signal State
The detector in question exists physically, is in perfect working order, and is in
1.
its signal state. This is represented by extended labstate ∼

25.7 The Universal Register

The power set approach to detectors allows us to think of an absence of a
detector in Λ as an observable fact that can be representable mathematically.
The state corresponding to an absent detector is represented by the element ∼3
of its associated power set. We can do this for any number of absent detectors.
Therefore, we can represent a complete absence of any detectors whatsoever by
an infinite collection of such elements.

Remark 25.1 This is one of the very few places in QDN that we permit an
infinity to enter the discussion; it costs us nothing in material terms and is a
purely theoretical construct that nevertheless represents a very real state of
existence of a laboratory, before any detectors have been constructed in it.

Such a condition corresponds to an observer without any apparatus, i.e., an

empty laboratory. We denote this labstate by the symbol ∼ ∅ and call it the
information void , or just the void. It represents a potential for the existence
of any detector or detectors, relative to a given observer.
∅ , that does not mean that
If the observer’s laboratory Λ is in its void state ∼
the laboratory Λ or the observer do not exist, or that there are no systems under
observation in Λ. It means simply that the observer has no current means of
acquiring any information. An empty laboratory devoid of any detectors is a
physically meaningful concept, but one with no interesting empirical content.
The information void can be thought of as one element in an infinite set called
the universal register Ω, the Cartesian product of an infinite number of bit power
sets. We write
∞
0 ∞
0
∅
∼
≡ 3
∼
α
, ∈ Ω ≡ P(B α ), (25.14)
α α

where the index α could in principle be discrete, continuous, or a combination

of both.
330 Dynamical Theory of Observation

The cardinality of the universal register is a measure of how many power sets
could belong to it. That depends on the imagination, and may be assumed to be
infinity. Precisely what sort of Cantorian cardinality it should be is not clear, but
that is a vacuous concept anyway. If we thought in terms of Λ sitting in continuous
space, then we expect at least the cardinality, c, of the continuum. However, that
is a metaphysical statement, because there would not be enough energy in the
universe to create a continuum of detectors.1 What helps us immeasurably here
is that real observers can only ever deal with finite numbers of detectors. We
can generally ignore all nonexistent potential detectors, but keep in mind that
an absence of a detector can be significant in some circumstances, such as in the
Renniger thought experiment (Renniger, 1953).
The product notation in (25.14) reflects the relationship between collections
of power sets and the tensor products of qubit spaces that we encounter in the
quantum version of this approach, discussed later. In these products, ordering is
not significant, since labels keep track of the various terms. An arbitrary classical
1∞ α
labstate Ψ∼
n in the universal register Ωn at stage Σn will be of the form α ψn , ∼

n is one of the four elements of P(Bn ) ≡ {0

where ψ α , 1 α , 2 α , 3 α }.
α α
∼ ∼n ∼n ∼n ∼n

Operators acting on universal register states will be denoted in blackboard

Oα
bold font with three dots below, and act as follows. If ··· n is a power bit operator
1 ∞
acting on elements of P(Bn ), then O n ≡ α ···
α α
O n acts on an arbitrary classical
1∞ α ···

state Ψ
∼n
≡ α ψ n in Ωn according to the rule
∼

∞ ⎧ ∞ ⎫
0 ⎨0 ⎬ ∞
0
OnΨ
∼n
= Oα ψ βn ≡ Oα α
n ψn . (25.15)
··· ··· n ⎩ ∼ ⎭ ··· ∼
α β α
1∞
For every classical register state Ψ
∼
≡ α ψ α there will be a corresponding
∼
∼ 1∞ ∼ ∼
dual register state Ψ ≡ α ψ α , where ψ α is dual to ψ α . Classical register states
∼

including the void satisfy the orthonormality condition

∞  ∞ ∞ ∞
∼ 0∼ 0 0 ∼ 0 α α
ΦΨ ∼
≡ φ α
ψ β
= φ α α
ψ = δφ ψ . (25.16)
∼ ∼
α β α α

Classical register states Φ

∼
, Ψ
∼
that differ in at least one bit power set element
∼
therefore satisfy the rule ΦΨ
∼
= 0.

25.8 Contextual Vacua

In conventional classical mechanics or Schrödinger–Dirac quantum mechanics,
empty space is generally not represented by any specific mathematical object.

1
In other words, the Universe could not observe itself completely.
 25.8 Contextual Vacua 331

In relativistic quantum field theory (RQFT), however, empty space is represented

by the vacuum, a normalized vector in an infinite-dimensional Hilbert space. It
has physical properties such as zero total momentum, zero total electric charge,
and so on, which although bland are physically significant attributes nevertheless.
In RQFT, particle states are represented by the application of particle creation
operators to the vacuum.
In our approach we encounter an analogous concept. Starting with an infor-
mation void ∼ ∅ n , we represent the construction of a collection of detectors in
the laboratory Λn at stage Σn by the application of a corresponding number of
C in acting on that information void.
construction operators ···

Example 25.2 The construction of detector in in its signal ground state

C in ∼
in a previously empty laboratory is represented by the element ··· ∅ n of the
universal register Ωn at stage Σn , where
⎧ ⎫
⎨0∞ ⎬
C i
≡ I j
Ci . (25.17)
··· n
⎩ ··· ⎭ ··· n
n
j=i

[r]
More generally, an extended labstate ∼ 0 n consisting of a number r of detectors
each in its signal ground state at stage Σn is given by
0 [r] ≡ ···
∼n
C 1n ···
C 2n . . . ···
C rn ∼
∅n , (25.18)

where without loss of generality we label the detectors involved from 1 to r. Such
a state will be said to be a rank -r ground state, or contextual vacuum state.
We can now draw an analogy between the vacuum of RQFT and the rank-r
ground states in our formalism. The physical three-dimensional space of con-
ventional physics would correspond to a contextual vacuum of extremely large
rank, if physical space were relevant to the experiment. This would be the case
for discussions involving particle scattering or gravitation, for example. For many
experiments, however, such as the Stern–Gerlach experiment and quantum optics
networks, physical space would be considered part of the relative external context
and therefore could be ignored for the purposes of those experiments. It all
depends on what the observer is trying to do.
In the real world there is more than one observer, so a theory of observation
should take account of that fact. That is readily done in extended QDN. For
example, the mutual contextual vacuum for two distinct observers A and B for
which some commonality of time had been established would be represented by
an element in Ωn of the form
0 A,B ≡ ···
∼n
C A,1 C A,2
n ··· C A,r
n . . . ··· n
A
C B,1 ···
··· n
C B,2 C B,r
n . . . ··· n
B
∅n ,
∼
(25.19)

and so on. If subsequent dynamics was such that the detectors of observer A
never sent signals to those of observer B and vice versa, then to all intents and
332 Dynamical Theory of Observation

purposes we could discuss each observer as if we were primary and they were
separate secondary observers. If on the other hand some signals did pass between
them, then that would be equivalent to having only one secondary observer.
If no commonality of time or other context has been established between
primary observers, then there can be no physical meaning to (25.19). This is
an important point in cosmology, where there are frequent discussions about
multiple universe “bubbles” beyond the limits of observation. The mere fact
that astronomers have received light from extremely distant galaxies establishes
a context between the signal sources in those galaxies and the detectors used
by the astronomers and validates the use of general relativity (GR) all the way
to those regions of spacetime. If, on the other hand, no such signals have been
received, then there is no such context. Therefore, relative to astronomers today,
the universe beyond the horizon of observation can be meaningfully represented
only by an information void, not a spatial vacuum. Something may be there, but
we should not discuss it as if we had access to any form of information about it,
such as its spacetime structure.
Much the same concern must be raised about black hole physics. That requires
careful analysis of the contextual relation between observers outside the critical
(Schwarzschild) radius and those who were assumed to be inside it, if any such
context can be found. An event horizon is a boundary between observers on one
side and observers on the other. For all intents and purposes, observers on one side
of such a horizon have to regard those on the other side as in an information void.
This raises deep questions and concerns about the use of coordinate patches
in general relativity and black hole physics that may be related to the problems
of “quantizing gravity.” We have asserted throughout this book that quantum
mechanics is much more like a theory of entitlement than about the structure of
SUOs. With this in mind, we should always ask the question: how do we know
this or that? Consider GR. It is standard to postulate a line element without
asking what could have been the source of that spacetime structure, or whether
there are physically enigmatic possibilities, such as closed time-like curves, as in
Gödel’s spacetime (Gödel, 1949). There may be event horizons in our solutions to
Einstein’s equations. Our point is this: what empirical entitlement do we have to
discuss two regions of spacetime separated by an event horizon, such as happens
in the case of the Schwarzschild metric? Contextual completeness should tell
us that such a scenario is vacuous. Trying to quantize any theory under such
circumstances seems to be asking for trouble.

25.9 Experiments
Long before any experiment can begin, an observer starts off with a laboratory Λ
∅ . Then at some stage long before any runs can be taken, specific
in its void state ∼
apparatus consisting of a finite number r of detectors has to be constructed
in Λ. We will assume without loss of generality that these are all functioning
 25.9 Experiments 333

normally and in their ground state, so the labstate at that point is given by the
equivalent of the right-hand side of (25.18). All of this is necessary before state
preparation.

Modules
We have so far discussed only the detectors. Similar comments can be made about
the modules that sit in the information void. The difference is that labstates are
directly associated with detectors, both real or virtual, whereas modules are
almost invariably classical in their representation, at least in the formulation of
QDN that we have discussed in this book. Therefore, we have not invested much
effort in constructing a dynamical theory of modules, although that is expected
in a more complete theory.
There is an intriguing relationship here with Feynman diagrams in RQFT.
Consider quantum electrodynamics (QED) and the simplest Feynman diagram
for electron–proton scattering, Figure 25.1. In QED, the particle source and final
state detectors are taken to be at remote past and future times, respectively, and
propagation is in the information void (the vacuum). However, the quantum fields
are interacting in QED. This particular Feynman diagram can be interpreted as
a process where a virtual module (the photon, represented by the wavy line in
Figure 25.1) is created not by the observer, but by the quantum process itself.
This module has many of the characteristics of a beam splitter, discussed in
Chapter 11. Because this module is created via the quantum dynamics in the
information void, it is reasonable to classify the experiment as type T3, as defined
in Chapter 21.
Care has to be taken not to take Feynman diagrams too literally. They rep-
resent individual terms in infinite perturbation expansions or in asymptotic
expansions of complete amplitudes. Those complete amplitudes will satisfy all
the requirements of scattering matrix theory (S-matrix) (Eden et al., 1966), and
they will behave as virtual quantum modules.

The Contextual Register

According to what we said earlier, external context involving detectors in their
3 can be ignored; they do not exist in any physical sense. Therefore,
void state ∼
we need only discuss those detectors that are in states ∼
0, ∼
1 , or ∼
2 . A further sim-
plification is that in real experiments, observers generally filter out observations

Figure 25.1. Lowest order Feynman diagram for electron–proton scattering.

334 Dynamical Theory of Observation

from faulty detectors (assuming these have been identified) by postselecting only
those labstates that contain the normal bit states ∼ 0 or ∼1 . We shall confine our
attention to such normal labstates until we deal with applications to quantum
mechanics.
Given this condition, we can restrict our discussion at any given stage Σn to
the physical register R ∼ n
, a subset of the universal register Ωn . If there are rn
working detectors in the laboratory at stage Σn , then R ∼
n consists of 2
rn
normal
states. This is the extended QDN version of the classical registers Rn discussed in
earlier chapters. We may represent these normal states with the extended version
of the computational basis and signal basis representations discussed previously,
For example, a typical normal extended basis labstate is of the form

∼
i n ≡ ∼i 1n ∼i 2n . . . ∼i rnn , (25.20)

where ij = 0 or 1 for j = 1, 2, . . . , rn , and

in ≡ i1n + 2i2n + · · · + 2rn −1 irnn . (25.21)

The physical register R ∼

n represents all those detectors in the laboratory Λ at
stage Σn that exist relative to the observer and are not faulty.
Virtually all of the concepts, such as signality, encountered with binary regis-
ters are encountered with physical registers, with some enhancements that are
particular to enhanced QDN. One such is that the power bit operators ··· 
A and ···
A
defined in (25.10) are not adjoints of each other, whereas the bit operators A
and A defined in (4.13) are mutual adjoints. It was in anticipation of that fact
that we chose the hat (circumflex) notation for A and not A† .

Physical Register Signal Operators

Given a rank-r physical register R ∼
n we define the physical register signal oper-
ators
⎧ ⎫ ⎧ ⎫
⎨0∞ ⎬ ⎨0∞ ⎬
A i
≡ I j
A i
, Ai ≡ I j i ,
A 1  i  r. (25.22)
··· n
⎩ ··· ⎭ ···
n n ··· n
⎩ ··· ⎭ ··· n
n
j=i j=i

There are several interesting processes that can be described by these extended
A in to the rank-
register operators. For example, an application of the operator ···
[r]
0 n ≡ ···
r contextual vacuum ∼ C 1n ···
C 2n . . . ···
C rn ∼
∅ n gives a rank-(r − 1) contextual
vacuum:

···
C 1n ···
A in ··· C 2n . . . ···
C rn ∼
∅ n = ···
C 1n ···
C 2n . . . ···
C i−1 C i+1
n ··· C rn ∼
n . . . ··· ∅n . (25.23)

This is a nonzero labstate in Ωn but is not an element of the original physical

[r]
register R
∼
n . What has happened is analogous to the convention qubit register

result Ain 0n = 0. In the extended QDN case, we do not get zero but the equivalent
 25.11 The Elitzur–Vaidman Bomb-Tester Experiment 335

[r]
A in on the signal ground state (contextual vacuum) ∼
of it: the action of ··· 0 n of
R
∼
[r]
maps it into the ground state of a different physical register, one of rank
r − 1, a state orthogonal to every state in R
∼
[r]
.

25.10 Quantization
The signal operators and the corresponding projection operators ···  i can
P
P in , ··· n
be used to discuss all the quantum processes covered in this book, and will
be used in the discussions on the bomb-tester and Hardy paradox below. An
additional feature we make use of is that we can now factor into the mathematical
C in and the decommissioning operators
representation the construction operators ···
D in explicitly.
···

In the next two sections we show how the extended QDN formalism describes
experiments where the apparatus changes in one way or another during the
experiment. In particular, the faulty/off-line state plays an important role in
these experiments.

25.11 The Elitzur–Vaidman Bomb-Tester Experiment

In this experiment, a stockpile of active (A) and dud (D) bombs is analyzed,
one by one, in order to find as many unexploded active bombs as possible.
The approach follows that discussed in Elitzur and Vaidman (1993). The stage
diagram shown in Figure 25.2 is a modified Mach–Zehnder interferometer. B 1
and B 2 are beam splitters, M is a mirror, and T is the actual bomb-testing
triggering device that explodes the bomb if it is an active bomb and a signal
could have been detected at 21 .2

M B

B T

Figure 25.2. The bomb-tester network.

2
One of the reasons this experiment is significant and to some extent baffling is the strange
mix of classical and quantum counterfactuality. Analyzing it in terms of which-path
information seems the best approach, because such information is not controversial and its
extent can be established directly from a stage diagram in general.
336 Dynamical Theory of Observation

The process goes as follows. Source S prepares a single photon state at 10 ,

which is then passed into beam splitter B 1 . Transmitted channel 21 feeds onto
the bomb tester module T while reflected channel 31 is deflected by mirror M
toward beam splitter B 2 .
The action at module T is critical and is based on a specific assumption that
sits on the knife edge between the classical and quantum worlds. For each run
of the experiment, a bomb is taken from the untested stockpile of bombs (which
includes active and dud bombs) and is placed in contact with module T . If the
bomb is a dud (D), then 21 is not prevented from being deflected onto 52 , and
thence onto beam splitter B 2 , where it can interfere with 42 .
If on the other hand the bomb is active (A), then either channel 31 is realized
and the bomb does not explode, or channel 31 is not realized, and then certainly
the bomb explodes. In this latter eventuality, channel 52 is blocked.
This experiment involves a randomly changing apparatus network, because the
bomb being tested in a given run is replaced by a new, randomly active or dud
bomb from the unused stockpile. In addition, the network is self-intervening in
the case of an active bomb coinciding with realized channel 21 . This experiment
is a combination of T2 and T3 experiments (these are discussed in Chapter 21).
The experiment looks like T2 in that the probability of a bomb being active or
dud in each run introduces an explicit time dependence in the experiment as a
whole, but that probability is determined by factors external to the apparatus.
It has also the characteristics of a T3 experiment, in that beam splitter B 1 has
random outcomes, and these determine whether module T is triggered. Therefore,
we may classify this experiment as T4.
There are two distinct networks to consider separately, because whether the
bomb is active or dud is part of relative external context and not a quantum
process requiring extended QDN. Therefore, we deal with each of the networks
separately, using a pure labstate description for each initially. Then we use a
density matrix approach to consider the combined experiment.
Our formulation of this scenario differs in two respects from those considered
previously in this book. First, we shall use extended QDN. Second, we shall
introduce persistence, in that all real and virtual detectors will be created before
stage Σ0 and persist to the final stage.

Creation of the Apparatus

Prior to the creation of the apparatus, well before stage Σ0 , the signal state of
∅ in the universal register Ω.3 We may take the
the laboratory Λ is its void state ∼
apparatus to be created in its ground state (equivalent to a contextual vacuum)

3
The choice of the words its and the in this sentence is deliberate. A void state is contextual,
whereas we can assume that there is just one enormous universal register that describes the
totality of possible contextual existence and nonexistence that could ever be imagined,
under all possible circumstances. This latter assumption costs us nothing.
 25.11 The Elitzur–Vaidman Bomb-Tester Experiment 337

before stage Σ0 , that is, before the start of each run, so we would normally expect
the apparatus to exist and be in its quiescent, no-signal state,

∼
0 0 ≡ ···
C 10 ···
C 20 ···
C 30 ···
C 40 ···
C 50 ···
C 60 ···
C 70 ∼
∅0 . (25.24)

This deserves some explanation, because sites 10 , 21 , 31 , 42 , and 52 are not real
detectors. So the question arises, what does it mean to “create” them?
The answer is given by Schwinger, quoted in Chapter 24. The modules S,
B 1 , B 2 , M , and T really do “exist” relative to the observer, and it is that
very existence that creates, as it were, the context for virtual detector sites
10 , 21 , 31 , 42 , and 52 to be meaningful. We may refer to this as “Schwinger
counterfactual reality.” It is not metaphysics but actually part and parcel of the
process of running any experiment, as we have commented before. The observer
would have had to place real detectors in those sites in order to calibrate the given
modules, before the actual runs had started (and then taken those calibration
detectors away). Otherwise, there could be no degree of confidence in what the
observer believed about the labstates being studied.
In addition to the creation of real and virtual detector sites, the preparation
device S prepares a photon signal in 10 , so the initial extended labstate Ψ ∼
0 is

given by Ψ A 1 0 0 . This is the initial extended labstate for each of the two
0 ≡ ···
∼ 0∼
networks we now consider.
In the following, it is useful to label the respective transmission coefficient and
reflection coefficient associated with each beam splitter separately, in order to
see how information is flowing through the networks. We work in extended QDN
throughout.

Runs with a Dud Bomb

Stage Σ0 to Stage Σ1
By inspection of Figure 25.2 we have
Ψ ≡ ···
U 1,0 Ψ A 2 + ir1 A
= (t1 ···  3 )0 1 . (25.25)
∼1 ∼0 1 ··· 1 ∼

Stage Σ1 to Stage Σ2
In this scenario, the bomb is a dud, so any signal entering module T from port
21 is transmitted directly to 52 . Taking the mirror M not to change any phase,
we then have
1 5  4 )0 2 .
Ψ
∼
2 ≡ ···
U 2,1 Ψ
∼
A 2 + ir1 ···
1 = (t ··· A 2 ∼ (25.26)

Stage Σ2 to Stage Σ3
Using
 5 0 2 = (t2 A
U 3,2 ···
A  7 + ir2 A
 6 )0 3 , (25.27)
··· 2∼ ··· 3 ··· 3 ∼

A 4 02
U 3,2 ··· = (t 2
A6 + ir 2
A 7 )0 3 , (25.28)
··· 2∼ ··· 3 ··· 3 ∼
338 Dynamical Theory of Observation

we end up with the final state outcome probabilities

Pr(63 |Ψ
∼
0 ) = |t r + r t | ,
1 2 1 2 2
Pr(73 |Ψ
∼
0 ) = |t t − r r | .
1 2 1 2 2
(25.29)

The significance of this result is√ that for matched symmetric beam splitters, for
which t1 = r1 = t2 = r2 = 1/ 2, the prediction is that Pr(73 |Ψ ∼
0 ) = 0. The
conclusion, that for symmetrical beam splitters a dud bomb never coincides with
a positive signal in detector 73 , is critical to the point of this experiment.

Runs with an Active Bomb

In this scenario, the bomb is active and could explode if triggered at module
T . We shall take the decommissioning of 52 as a marker of that explosion: the
observer will be able to see real debris in the laboratory if the bomb explodes.
Everything is the same as in the previous scenario with a dud bomb, up to
stage Σ1 .

Stage Σ1 to Stage Σ2
In this scenario, the bomb explodes and decommissions the device T if a signal
enters from 21 . Hence we may write
Ψ ≡ ···
U 2,1 Ψ = (t1 ··· A 4 )0 2 .
D 52 + ir1 ··· (25.30)
∼2 ∼1 2 ∼

Stage Σ2 to Stage Σ3
The debris created by any bomb exploding at stage Σ2 will certainly be trans-
mitted on to stage Σ3 , so we have

···
D 52 ∼
U 3,2 ··· D 53 ∼
0 2 = ··· 03 ,
A 4 0 2 = (t2 A
U 3,2 ···  6 + ir2 A
 7 )0 3 , (25.31)
··· 2∼ ··· 3 ··· 3 ∼

giving the final state

Ψ D 3∼
1 5
3 = t ··· A 6 + ir2 A
0 3 + ir1 (t2 ···  7 )0 3 . (25.32)
∼ 3 ··· 3 ∼

We conclude that in this scenario, the following probabilities hold:

probability of an explosion : (t1 )2 ,
probability of no explosion and signal in 63 : (r1 t2 )2 , (25.33)
probability of no explosion and signal in 73 : (r1 r2 )2 .
In the symmetric case, we conclude that the probability of no explosion of an
active bomb and a signal in detector 73 is 14 .
It is this outcome that allows the observer to find an unexploded, active bomb,
because detector 73 does not trigger in the symmetric beam splitter scenario when
the bomb in question is a dud.

Random Testing
Unfortunately, the observer does not know before each run whether a particular
bomb is active or a dud. Consider a sequence of runs such that there is a (classical)
 25.12 The Hardy Paradox Experiment 339

probability ω A of encountering an active bomb and a probability ω D ≡ 1 − ω A

of a dud. In this case we take a density matrix approach. At the nth stage, for
n = 0, 1, 2, 3, we define the extended density matrix
∼ ∼
ρ n ≡ ωA Ψ
∼n
A A
Ψn + ω D Ψ
∼n
D D
Ψn . (25.34)
···

A D
Here Ψ∼n
and Ψ∼n
are the extended labstates for the active and dud scenarios,
respectively, at stage Σn . It could be thought that this must be incorrect at stage
Σ0 , because the observer has no information as to which scenario is in place. But
in fact, that information could be established in principle before stage Σ1 by, for
example, triggering the bomb directly. It would explode in the active case and
A
that would inform the observer that the initial state should have been Ψ ∼n
in the
now aborted run. A similar remark holds for the dud case.
The overall probability of triggering detector 73 is given by
 
7 ,
P
Pr(73 ) = T r ρ 3 ··· (25.35)
3
···

 7 is the signal projection operator for detector 73 . We find

P
where ··· 3

Pr(73 ) = ω D (t1 t2 − r1 r2 )2 + ω A (r1 r2 )2 . (25.36)

In the symmetric case, we conclude that whenever detector 73 registers a signal,

the bomb is active and can be stockpiled accordingly. A single sweep of the
original stockpile of active and dud bombs will identify one-quarter of the active
bombs without exploding them (the rest go up in smoke). Moreover, whenever
detector 63 has triggered, there has been no explosion, but the observer cannot
be sure which scenario has occurred. Therefore, that bomb can be retested.
In this way, the observer should be able to find up to one-third of the active
bombs, the rest having exploded.
This thought experiment has been the motivation for real experiments that
simulate in some way the action of module T in the above experiment, a class of
experiment known as interaction-free measurement (Kwiat et al., 1994).

25.12 The Hardy Paradox Experiment

The Elitzur–Vaidman bomb-tester experiment discussed above may be inter-
preted as a simplified form of double-slit experiment, where the screen has only
two sites and one of the slits can be blocked off or not, depending on whether a
bomb is active or dud. This blocking off occurs in a classical way, because the
uncertainty as to whether the bomb is active or dud is not intrinsic to the nature
of the bomb but reflects the observer’s ignorance of the nature of the bomb.
A variant of the bomb-tester experiment is known as the Hardy paradox
experiment (Hardy, 1992). In this variant, the blocking-off of a slit occurs in an
intrinsically random way because it is governed by quantum processes, contrasted
340 Dynamical Theory of Observation

B R B

B B

Figure 25.3. The Hardy paradox experiment.

to the bomb-tester experiment analyzed above, where the observer’s ignorance

about the nature of the bomb plays a role.
The Hardy paradox experiment shown in Figure 25.3 consists of an electron–
positron pair 11 , 21 produced by a source S. These particles are then deflected
by magnets (not shown) to pass through the equivalent of two coupled Mach–
Zehnder-type networks, as shown in Figure 25.3. Output ports 32 and 42 of beam
splitters B 2 and B 1 , respectively, are arranged to intersect in a confined region,
R, where electron–positron annihilation may occur.
There are two possibilities if both particles are in region R at the same stage:
either (1) they annihilate into a photon pair, or (2) they pass onto their respective
beam splitters B 3 and B 4 as shown.

Case (1)
In case (1) and in conventional terminology, the annihilation effectively blocks
off one of the entry ports in each of those beam splitters. This is equivalent
to a virtual module in interaction region R, which would otherwise be a true
information void. In this case, the photons produced by the annihilation are
detected in 33 and 43 .

Case (2)
In case (2), the virtual detectors 32 and 42 transmit their amplitudes onto beam
splitters B 3 and B 4 , respectively, as shown in Figure 25.3. Interference occurs in
each of these beam splitters. In the symmetric case, a signal will be detected at
63 and not in 54 , and a signal will be detected in 73 and not in 83 .

The Hardy paradox experiment is intrinsically a pure quantum experiment and

we can discuss it via pure labstates alone. An important point is that electron–
positron annihilation is a well-known quantum process that occurs in nature,
whereas the detonation mechanism of the Elitzur–Vaidman bomb-tester is left
unspecified.
 25.12 The Hardy Paradox Experiment 341

As with the bomb-tester experiment above, we start our analysis of the

Hardy paradox experiment by first defining the contextual vacuum ∼ 00 ≡
C 1n ···
C 2n . . . ···
C 8n ∼
∅ and then the initial state is given by Ψ  1 00 .
0 ≡ ···
A
··· ∼ 0∼
0
The dynamics then goes as follows.

Stage Σ0 to Stage Σ1
The electron–positron pair produced by source S is split by magnetic fields ready
to be deflected into beam splitters B 1 and B 2 as shown, giving extended labstate
Ψ 1A
≡ ···
A  2 01 .
∼1 1 ··· 1 ∼

Stage Σ1 to Stage Σ2

1 1  4 )(t2 A
 2 + ir2 A
 3 )0 2 .
Ψ
∼
2 ≡ ···
U 2,1 Ψ
∼
A 2 + ir1 ···
1 = (t ··· A 2 ··· 2 ··· 2 ∼
(25.37)

Stage Σ2 to Stage Σ3
The only novel factor concerns potential electron–positron annihilation, involving
32 and 42 . We will explore the possibility of partial annihilation, by taking the
evolution for this term to be given by
A4A
U 3,2 ···  3 0 2 = αD 4 D 3 0 3 + β(t4 A
 8 + ir4 A
 7 )(t3 A
 5 + ir3 A
 6 )0 3 , (25.38)
··· 2 ··· 2 ∼ ··· 3 ··· 3 ∼ ··· 3 ··· 3 ··· 3 ··· 3 ∼

where |α|2 + |β|2 = 1. These coefficients reflect the fact that electron–positron
annihilation is a quantum process and does not always occur. Note the intro-
duction of the destruction operators here, signifying that photons have been
produced (these can be regarded as the debris from the annihilation of the virtual
electron–positron detectors at 33 and 43 ).
The relevant information when fed into program MAIN gives the following
correlations for the symmetric beam splitter case:
Pr(33 , 43 |10 ) = 14 |α|2 ,
Pr(53 , 73 |10 ) = Pr(53 , 83 |10 ) = Pr(63 , 83 |10 ) = 16 |1
1
− β|2 , (25.39)
Pr(63 , 73 |10 ) = 16 |3
1 2
+ β| .
When β is set to one, corresponding to no annihilation in region R, the predic-
tion is that the only outcome is that both 63 and 73 show a signal. However,
when annihilation is permitted, α = 0 and the outcomes show characteristics
of interaction-free measurement. This is deduced from conservation of electric
charge. The argument goes as follows. If a signal is detected at, say, 53 , then
annihilation cannot have occurred, because signality is conserved in this experi-
ment; annihilation would give a signal at 33 and at 43 , and so a signal elsewhere
would be ruled out. But a signal in 53 arises only if something is affecting the
signal from 32 , and the only factor that could affect that could be 42 . Therefore,
any signal in 53 is an “interaction-free” detection of the particle at 42 . A similar
remark holds for any signal observed at 83 .
342 Dynamical Theory of Observation

25.13 Implications and Comments

The application of our extended QDN formalism to the Elitzur–Vaidman bomb-
tester and Hardy paradox experiments demonstrates that the concept of faulty
or decommissioned states has physical significance.
An important point about this chapter is that the phenomena discussed show
that specific reductionist details of interactions seem not to be relevant to the
overall picture. For instance, in the bomb-tester experiment we have not said
anything at all about what sort of module T really is. Likewise, we have not
done any specific quantum electrodynamical calculation in the Hardy paradox.
What seems important is the architecture of the apparatus involved, and that is
an emergent aspect of these experiments.
Such considerations are what have led us to the conclusion that quantum
mechanics is really a theory of contextuality. That the rules are unfamiliar
and counterintuitive in many cases is a commentary not on reality, but on the
inadequacy and limitations of our classical conditioning.
 26
Conclusions

In the Preface to this book, we started with Alice and Bob having very different
views of their observations. Alice used an optical telescope and reported nothing
unusual about a distant galaxy that she could see. Bob, on the other hand,
detected intense radio activity in that galaxy. Our question was: who has the
true view of that galaxy?
We advised the reader that the answer is not Alice. Neither is it Bob. The
answer is not both of them, nor is it neither of them. It is not a trick question
either. So what is the answer that this book would supply?
Our answer will come presently.
While this book has set forth a definite mathematical perspective on empirical
physics, it has contained quite a lot of commentary that might be disparaged as
metaphysics or philosophy. Such commentary is generally frowned on in science,
because it has no empirical content. It is vacuous.
In our defense, we point out the obvious: science is not a robotic activity; it is
carried out by humans and these are driven by their metaphysical, philosophical,
and emotional imperatives. For instance, the hard-core scientific view that quan-
tum theory needs no interpretation but only application is itself a philosophy.
It is no more than a conditioned response, based on opinion and subscription
to current scientific norms. It is, indeed, a philosophy of how to do quantum
mechanics. So we all do it, in one way or another. Our concern in this book is
that the way that we do it should be based soundly on scientific principles. The
quantized detector approach that we describe and use in this book is our attempt
to do just that.
There is an excellent paper on all of this that has the provocative title “Quan-
tum Theory Needs No Interpretation,” by Fuchs and Peres (2000). It seems on
the face of it to dismiss any sort of “interpretation” of quantum mechanics. In
fact, close reading of it confirms (to us at least) the agenda that we have set out
in this book. Yes, quantum mechanics needs no interpretation, when it is being
applied to processes that take place in the information void.
344 Conclusions

But in our view, that is only half the story. We cannot exclude the observer and
what they are doing. The information void is contextual: it is defined by whatever
observer is involved, and by the apparatus that they are using. Schwinger said
just that (see his quote on this in Chapter 24). It has long seemed obvious to us
that the idea expressed by Schwinger really points to an extension of what physics
means. Physics cannot be just the study of systems under observation (SUOs).
The relationship of observers to those SUOs and to the Universe in which both
observers and SUOs are embedded should surely be just as important an issue
in physics as the perceived properties of SUOs.
That naturally leads to the realization that the physics of emergence should
be at least as important in physics as the reductionist agenda is currently. Our
explanation as to why it is not so regarded currently is that it is hard, much
harder than standard reductionist approaches to physics, which absolutely rely
on and utilize emergent concepts, but place them firmly outside the scope of
investigation.
Fortunately, human opinions on what is significant carry no weight in real
science: nature has a tendency to tell us when we are being complacent. Markers
of this are evident throughout quantum mechanics: the randomness of quantum
outcomes, the nonlocality of quantum correlations, the violations of Bell and
Leggett–Garg inequalities, quantum interference, state reduction, the Kochen–
Specker theorem (Kochen and Specker, 1967), entanglement, and more.
That brings us back to Alice and Bob. It was not a trick question: it was not
a proper question in the first place, because it is contextually incomplete.
As given, there was no indication for whom this question could have a truth
value. If you say that it is you, the reader, then our response is: what sort of
“truth” is it that deals with an invented scenario? Alice and Bob don’t exist. They
were invented in the Preface for the purposes of demonstrating a fundamental
point.
It may seem a glib, cheap point, but it is not. It points to the fact that “truth”
does not exist in a vacuum. It requires an observer for whom that “truth” is
empirically meaningful.
Contextual incompleteness is invariably the main ingredient in a number of
theoretical disciplines that purport to “explain” reality, but in fact are as fictional
in their scientific content as our image of Alice and Bob is. That is the essential
point behind the article of Fuchs and Peres, mentioned above. If we could, we
would retitle their paper: quantum theory needs no interpretation, it needs to be
extended to cover the observer. That is what this book is all about.
A final comment about Alice and Bob. Our answer is that each of them is
stating the truth, but only relative to their individual contexts. In that sense
they are both making contextually true statements. But that is not the same as
saying they both have the “true” view of the galaxy they are observing: there is
no “true” view of that galaxy in an absolute sense. Reality is far too deep for
that.
 Appendix

A.1 QDN Notation

Symbol Meaning
× Cartesian product, direct product
⊗ tensor product
Σn stage n
An apparatus at stage Σn
rn number of real or virtual detectors in
An = the rank of An
in the ith real or virtual detector in An
Qin the qubit representing in , i = 1, 2, . . . , rn
Q n ≡ Q n Q n . . . Qn
1 2 rn
the quantum register at stage Σn : a rank-rn
tensor product
In the observer’s information about Qn
Hn ≡ (Qn , In ) Heisenberg net at stage Σn
2 rn
dimension of Qn ≡ dim Qn
Bni the preferred basis for Qin , i = 1, 2, . . . , r
Rn ≡ Bn Bn . . . Bn
rn 1 2 rn
the preferred basis for Qn : a Cartesian product
kn kth element of computational basis Bn ,
k = 0, 1, 2, . . . , 2rn − 1
kn dual of kn
Ψn pure labstate at stage Σn : an element of Qn
Ain ith signal destruction operator at stage Σn

An i
ith signal creation operator at stage Σn
Pin ≡ Ain A i , ith no-signal projection operator
n
i ≡ A
P  i Ai ith signal projection operator
n n n
i i i i
Sn ≡ {Pn , Pn , An , An } ith signal set
i

mn ≡ im j n
Tij transition operator
346 Appendix

A.2 Lab Time and Frame Fields

In general relativity (GR), spacetime is modeled as a four-dimensional manifold
with a Lorentz signature metric. In general, GR spacetimes cannot be covered
by a single coordinate patch, particularly if there are closed time-like curves
(CTCs) as in the case of the Gödel metric (Gödel, 1949). In spacetimes with
CTCs, a global temporal foliation cannot be constructed, which means that a
global laboratory perspective cannot be contemplated in such cases.
The principles of quantized detector networks (QDN) are well suited to deal
with such issues. First, QDN is an endophysical approach to empirical physics,
meaning that it does not attempt a global (exophysical) description of the
Universe. In addition, finiteness is the order of the day, which means that
actual physics laboratories are regarded as of finite extent and duration, and
that no infinities are measurable. QDN does not normally attempt to discuss
systems under observation (SUOs) in terms of an infinite number of real or
virtual detectors. The QDN discussion of the bosonic and fermionic oscillators
in Chapter 24 are given to illustrate the remarkable theoretical properties of
infinite-rank quantum registers. In applications to real SUOs, QDN invariably
involves finite-rank quantum registers.
In GR, a relatively localized laboratory description typically involves a complex
of four frame fields, {eμ : μ = 0, 1, 2, 3} constituting the laboratory frame, or
coordinate patch adapted to cover a real physical laboratory. These basis vectors
are usually chosen to satisfy the orthonormality relations
g(eμ , eν ) = ημν , (A.1)
where g is the metric tensor and the ημν are the components of a 4 × 4 matrix
displaying the Lorentz signature of a Minkowski metric:
⎡ ⎤
1 0 0 0
⎢0 −1 0 0⎥
[ημν ] ≡ ⎢
⎣0 0 −1 0 ⎦ .
⎥ (A.2)
0 0 0 −1
The significant frame field is e0 , which is time-like and indicates the temporal
foliation that dictates clock time over the laboratory. The three other frame fields
are space-like and lie in the space-like hypersurfaces of relative simultaneity that
the observer has set up in their laboratory using some chosen synchronization
protocol.
QDN requires such a framework in order for quantum principles to be appli-
cable: space and time have different roles in QM as it is encountered in the
laboratory. The concept of stage is intimately linked to the existence of such a
framework.

A.3 Lab Time and Stages

Consider an experiment of the Stern–Gerlach type, wherein a beam of particles
is passed through a beam splitter feeding onto two detectors denoted A and B.
 A.3 Lab Time and Stages 347

Suppose each run consists of a beam prepared at labtime reset to t = 0 with

the observer looking at each detector separately, at labtime TA in the case of
detector A, and labtime TB in the case of detector B. We are interested here in
the possibility that TB is very much greater than TA . For instance, suppose TA
is of the order of a millionth of a second and TB is a million years. Then it is an
empirical fact that the two times are not simultaneous relative to the laboratory
concerned, and then the question arises as to whether the observation at A could
causally impact on the observation at B.
In the framework of special relativity, the answer is determined by the lightcone
structure of Minkowski spacetime. If B is outside the forward lightcone centered
on A, then Einstein causality tells us that A could not influence B. In that
situation, an inertial frame could always be found in which these two events were
simultaneous, relative to that frame. Since probabilities are related to outcome
frequencies, which is a counting process of signals, we would expect the same
outcomes in such a frame as in the original frame where TA and TB were vastly
different.
We expect the same results would hold if B was inside the forward lightcone
of A but adequate shielding was in place. In modern electronics, the problem of
undesirable signal interference due to one detector affecting another is known as
crosstalk . Shielding is our term for the elimination of crosstalk. It is a problem
but one that in principle can be overcome. Indeed, it could be argued that the
very existence of physically distinct persistent SUOs co-existing at the same
time is evidence for shielding. The fact that atoms are generally stable and can
be regarded as distinct is direct evidence for that.
This line of reasoning leads to the interesting idea that space itself is a man-
ifestation of shielding. Consider two hydrogen atoms. According to standard
nonrelativistic QM, the combined SUO consists of two identical protons and
two identical electrons. If the atoms were separated by, say, 2 angstroms, we
might be tempted to describe their combined wave function in terms of a two-
proton, two-electron wave function, properly antisymmetrized on account of the
indistinguishability and fermionic nature of the constituents. On the other hand,
if the two atoms were, say, a light year apart, no one would ever think of these
atoms as anything other than two separate SUOs, each described by a one-proton,
one-electron wave function (in standard nonrelativistic QM).
This raises the interesting question: which comes first, space or shielding?
According to Schwinger, quoted in Chapter 24, the space-time is an idealization
of empirical context, so the implication is that the emergent concept of shielding
comes before we can define the reductionist concept of space.

The above arguments tells us that labtime simultaneity or strict lightcone

causality is not essential in signal detection; the important criterion is that
different signal detectors should not interfere causally with, or have the pos-
sibility of interfering causally with, other signal detectors in the experiment.
It is this condition, referred to as shielding, that defines our concept of stage.
Signal detectors that are looked at in such a way constitute a single stage in an
348 Appendix

experiment. A stage is an collection of detectors that are looked at by an observer

in such a way that no subset of those detectors can causally influence any other
subset.
A stage is therefore a contextual classification of detectors that collectively
plays the same role in QDN as events on a space-like hyperplane of “simultaneity”
in some time-like foliation in general relativity.
The concept of stage undermines the contextually incomplete Block Universe
concept: stages require the existence of observers, and the experimental protocol
(context) to be given explicitly, something that the Block Universe has no place
for. The “Consistent Histories” approach to QM is an interesting development of
QM in that it deals with empirical propositions at multiple different observation
times (Griffiths, 1984), and these could be regarded as stages.

A.4 Ensembles
The ensemble concept arises in physics for two reasons. First, no experiment
normally validates a generalized proposition in a single run, even in classical
physics, which deals with generalized propositions with a generalized proposition
classification (GPC) of 2. Such propositions are based on the classical assertion
that SUOs have precise qualities that can be quantified by exact measurement.
Regardless of their view of that assertion, experimentalists will know that, in
reality, experimental errors and inaccuracies are always present. To overcome
this, multiple runs of the basic experimental protocol are generally performed and
then averages and other statistical quantities established from the accumulated
data. These multiple runs constitute ensembles of one kind or another.
The second reason for the use of ensembles is that quantum physics asserts
that the outcome of any given run in an experiment is a random variable, so that
statistical analysis based on the Born rule is required as a matter of principle
and policy.
Ensembles come in several varieties, each with its particular spatiotemporal
architecture, and each characterized by generally unstated, implicit context.
Whichever kind of ensemble is chosen depends on several factors usually outside
the experimentalist’s control, such as limited resources and time. The following
are two important kinds of ensemble.

Spatial Ensembles
A spatial ensemble is a physical collection in a given laboratory of multiple,
mutually isolated (from each other) copies of a given SUO, such as atoms in
a crystal, such that a basic run is performed on each copy once. The statistics
for the experiment is then established by collecting the outcome data from each
copy and assuming that outcome frequencies can be attributed to probabilities
associated with the original SUO.
 A.4 Ensembles 349

In experiments based on relatively localized laboratories, wherein physical

conditions are relatively homogeneous, this ensemble concept is usually reason-
able. However, three problems may arise. First, different copies may actually
interact with each other. This is the case in magnetic resonance experiments, for
instance, where an interaction between nuclear spins and their neighboring spin
environment is all the point. The second problem is that inhomogeneities in the
laboratory environment may invalidate the above logic. For instance, a spatial
ensemble carrying out an Unruh-type experiment in an accelerating laboratory
(in contrast to a freely falling one) will almost certainly display laboratory
inhomogeneities in what looks like a local gravitational field. The third problem
is one of economics. Some experiments cannot be based on the idealized spatial
ensemble concept simply because each individual run may be too costly or too
big in spatial terms to duplicate in any laboratory. An example is the Large
Hadron Collider: there is only one particle accelerator and it is very big and very
expensive.
In the case of the Large Hadron Collider, there is a modification of the spa-
tial ensemble concept that works excellently: multiple copies of the same SUO
(protons) are contained in a single circulating beam. Assumptions are then made
that during the very brief time of interaction involved, each proton behaves as if
it was isolated from the other protons in that beam and would interact with only
one other proton in the opposing beam. The beam statistics of the Large Hadron
Collider, are impressive: in a given run, each beam consists of 2,808 bunches of
protons, and each bunch contains about 1011 protons.

Temporal Ensembles
Some experiments are too costly to perform via spatial ensembles, so the standard
alternative is to use a temporal ensemble. In such an architecture, multiple
runs of the same basic protocol are implemented in temporal succession using
the same apparatus each time. In principle, an ideal temporal ensemble should
be equivalent to an ideal spatial ensemble, but that is an assertion that can
be challenged on the basis of cosmological evidence. It is now believed that the
universe is expanding in an irreversible way, relative to all endophysical observers.
Therefore, the environment around any laboratory is not quite the same during
any given run of an experiment as any other run. Of course, such discrepancies
are minute and could be laughed at as generating a pointless debate, were it not
for one glaring fact: the expansion does have observable effects, namely, the red
shift of light from distant galaxies.
The issue here is the relative scale of times: a comparison of the typical time
τ to complete a given run, the interval T to the next run, and the age A of
the Universe estimated from the observed Hubble constant. When τ and T are
both negligible compared with A, as is almost always the case, then temporal
ensembles should be as good as spatial ensembles.
350 Appendix

The impact of cosmological expansion should not be too lightly dismissed. It

has been speculated by scientists such as Dirac that the gravitational constant G
and perhaps even the speed of light c may change over cosmological time scales
(Dirac, 1938a). Therefore, any discussion of observed physical properties such as
electron mass and other properties should take such issues into account if the
context merits it.
We note that Peres did not consider temporal ensembles to be proper ensembles
in QM (Peres, 1995), but the fact is that many experiments are indeed carried
out via such ensembles.

A.5 Vector Spaces

A vector space (V, F) over a field F such as the real numbers R or complex
numbers C is a set V of elements V ≡ {a, b, . . .}, known as vectors, with the
following properties:
(i) There is a binary map  : V × V → V such that, for any elements a, b ∈ V ,
the object a  b ∈ V . This is called addition of vectors, or just vector
addition. The elements of V are called vectors.
Vector addition is commutative, i.e.,
a  b = b  a, ∀a, b ∈ V. (A.3)
Vector addition is associative, i.e.,
a  (b  c) = (a  b)  c, ∀a, b, c ∈ V. (A.4)
(ii) There is a unique element in V , known as the zero vector, denoted by 0V ,
such that
a  0V = a, ∀a ∈ V. (A.5)
(iii) For every vector a, there exists an additive inverse, denoted by −a, such
that
a  (−a) = 0. (A.6)
These properties mean that V is an abelian group under vector addition.
(iv) For any a ∈ V , λ ∈ F, then the object λa ∈ V . This is known as multiplica-
tion by a scalar, or just scalar multiplication. In this context, the elements
of F are called scalars.
Scalar multiplication satisfies the property
λ (μa) = (λμ) a, λ, μ ∈ F, a ∈ V. (A.7)
(v) Scalar multiplication is distributive, i.e.,
λ (a  b) = (λa)  (λb) ,
(A.8)
(λ + μ) a = (λa)  (μa)
 A.5 Vector Spaces 351

It is standard practice to use V to mean (V, F). The ground field F is generally
understood, but it is important to know whether it is R or C. In the former case
we say V is a real vector space, while in the latter case we say V is a complex
vector space. The space of three-vectors used to represent position in physical
space is a real vector space, while the Hilbert space of quantum state vectors is
a complex vector space.
It is easy to show that, ∀λ ∈ F, λ0V = 0V . Likewise, if 0F is the zero element
in F, then 0F a = 0V , ∀a ∈ V . It is important not to confuse the scalar zero 0F
and the vector zero 0V .
In practice, we do not bother to use a different symbol for vector addition, ,
in order to distinguish addition in the field, +. Henceforth, the same symbol, +,
will be used for both.
We often modify the notation involving the additive inverse −b, writing

a + (−b) = a − b, (A.9)

thereby suggesting a new binary process known as subtraction. This is not nec-
essary, but it is useful and should always be interpreted in terms of the addition
of vectors.

What is immensely astounding is that the above theory, which may appear no
more than a mathematician’s game, seems necessary to describe empirically val-
idated quantum physics, with the additional surprise that the field F is required
to be C and not R.

Subspaces of a Vector Space

Suppose U is a subset of a vector space V over some field F. If U is a vector
space over F, using the same rules for vector addition and scalar multiplication
as for V , then we say U is a subspace of V . Every subspace of V necessarily
contains the zero vector 0V of V .
Given two subspaces U1 , U2 of V , then the intersection U1 ∩ U2 is the set of
elements common to U1 and U2 , and is also a subspace of V .

Spanning Sets
Suppose S ≡ {v 1 , v 2 , . . . , v k } is a set of vectors in some vector space V with
ground field F. Let M be the set of all vectors of the form x1 v 1 +x2 v 2 +· · ·+xk v k ,
where x1 , x2 , . . . , xk ∈ F. Then M is a subspace of V , spanned by S. We write

M ≡ [v 1 , v 2 , . . . , v k ] . (A.10)

S is called a spanning set for M .

Linear Independence
An expression of the form x1 v 1 +x2 v 2 +· · ·+xk v k , where v 1 , v 2 , . . . , v k ∈ V and
x1 , x2 , . . . , xk ∈ F is called a linear combination of the vectors v 1 , v 2 , . . . , v k .
352 Appendix

If the xi are not all zero, then it is called a nontrivial linear combination.
Otherwise it is called trivial.
A set of vectors {v 1 , v 2 , . . . , v k } is linearly dependent if there exists a nontrivial
linear combination equal to the zero vector 0V . In other words, {v 1 , v 2 , . . . , v k }
is linearly dependent if the equation

x 1 v 1 + x 2 v 2 + · · · + xk v k = 0 V (A.11)

has a solution for which at least one of the xi is nonzero.

A set of vectors {v 1 , v 2 , . . . , v k } is linearly independent if the only solution to
Eq. (A.11) is x1 = x2 = · · · = xk = 0F .

Theorem A.1 The nonzero vectors v 1 , v 2 , . . . , v n ∈ V are linearly dependent

if and only if one of the vectors v k is a linear combination of the preceding ones
v 1 , v 2 , . . . , v k−1 .

A single nonzero vector v is necessarily independent, since the xv = 0V if and

only if x = 0F .

A linearly independent spanning set is called a basis. (A.12)

Theorem A.2 Any vector space that has a finite spanning set contains a basis.

A vector space is finite dimensional if it has a finite basis (i.e., one consisting
of a finite number of vectors). Hence, every vector space spanned by a finite
spanning set is finite dimensional.

Theorem A.3 If V is a finite-dimensional vector space with basis e1 , e2 , . . . , en

then every vector v in V can be expressed in one and only one way as a linear
combination

v = x 1 e 1 + x 2 e 2 + · · · + xn e n = x i e i (A.13)

using the summation convention.

All bases of a finite-dimensional vector space have the same number of ele-
ments. The dimension dim V of a finite-dimensional vector space V is the number
of elements of a basis.

Linear Transformations
Let U and V be two vector spaces, not necessarily of the same dimension, over
the same field F. A linear transformation (or linear mapping) T of U into V is
a mapping that assigns to every u ∈ U a unique vector T (u) ∈ V , such that
T (u1 + u2 ) = T (u1 ) + T (u2 ) , ∀u1 , u2 ∈ U,
(A.14)
T (λu) = λT (u) , ∀u ∈ U, ∀λ ∈ F.
T (u) is the image of u under T .
The set of all linear mappings of U into V is denoted L (U, V ) .
 A.5 Vector Spaces 353

Given a linear transformation T (U, V ), the set of all vectors u ∈ U such that

T u = 0V (A.15)

is called the kernel of T and written ker T .

The set of all vectors T (u), u ∈ U is called the image of U under T , and is
denoted by T (U ) .
The following is a critical theorem in QDN.

Theorem A.4 (Tropper, 1969) If T ∈ L (U, V ), then ker T is a subspace of

U and T (U ) is a subspace of V , such that

dim ker T + dim T (U ) = dim U. (A.16)

If dim ker T is called the nullity of T and dim T (U ) is called the rank of T ,
then the above theorem can be stated as

nullity of T + rank of T = dim U. (A.17)

Linear Functionals
A linear functional f˜ is a linear mapping of a vector space V into its ground field
F, such that

f˜ (αu + βv) = αf˜(u) + β f˜ (v) ∈ F, ∀α, β ∈ F, ∀u, v ∈ V. (A.18)

Note that summation on the left-hand side is in V , while summation on the

right-hand side is in F. Denote the set of all linear functionals over V by L (V, F).
When F = R, then f˜ is a real-valued linear functional, whereas if F = C then
˜
f is a complex-valued linear functional.

One-Forms
Given any two linear functionals f˜, g̃ ∈ L (V, F), define the linear combination
αf˜ + βg̃, where α,β ∈ F by the rule
 
αf˜ + βg̃ (v) ≡ αf˜(v) + βg̃ (v) , α, β ∈ F, v ∈ V. (A.19)

With this rule, L (V, F) is itself a vector space, known as the dual vector space
and denoted by V ∗ . Elements of this vector space will be called one-forms.
The convention we shall follow as much as possible is that vectors will be
represented by symbols in bold, such as v, while one-forms will be denoted by
symbols with a tilde, or a bar, such as ω̃ or ω.
The one-form/vector relation is employed in QDN in our representation of
questions and answers, as discussed in Chapter 2.
An important fact is that when V is finite dimensional, then V ∗ has the same
dimension, namely, dim V ∗ = dim V .
354 Appendix

Dual Basis
Suppose V is an n-dimensional vector space with basis B (V ) ≡ {ea : a = 1,
2, . . . , n}. Then an arbitrary vector v ∈ V can be written in the form

n
v= v a ea , (A.20)
a=1

where the {v i } are known as the components of v relative to the basis B (V ).

Given B (V ), we can always find a basis B (V ∗ ) ≡ {ẽa : a = 1, 2, . . . , n} for the
dual space V ∗ such that ẽa (eb ) = δ ab . We call B ∗ (V ) the conjugate basis. This
greatly simplifies calculations.

Bracket Notation
Given a vector v ∈ V and one-form ω̃ ∈ V ∗ , the bracket notation ω̃, v ≡ ω̃ (v)
is often used. In quantum mechanics, vectors are often written as kets, such as
|ψ, and dual vectors as bra-vectors, such as φ|, a notation used extensively by
Dirac (Dirac, 1958). Then the “inner product” φ|ψ is called the bracket of |ψ
and φ|.
Quantum mechanics vector spaces use a complex-valued field and the following
rule is imposed: φ|ψ∗ = ψ|φ.

Tensor Product Spaces

Suppose U and V are vector spaces over the same field F. It is possible for U and
V to be copies of the same vector space, but not necessarily so. If they were, we
would simple label them U 1 and U 2 , respectively. Significantly, U and V need
not have the same dimension.
If u ∈ U and v ∈ V , then the direct product u ⊗ v is identified with (u, v), an
element of the Cartesian product space U × V .

Example A.5 Consider the vector space of all real 2 × 2 matrices M (2, R) and
the vector space of all real 3×3 matrices M (3, R) . Then elements of the Cartesian
product M (2, R) × M (3, R) are given by Kronecker products of matrices. For
example, if A ≡ [Aab ] ∈ M (2, R) and B ≡ [Bij ] ∈ M (3, R), then the Kronecker
product A ⊗ B ≡ [Cai,bj ] is an array with double matrix indices, such that

Cai,bj ≡ Aab Bij , a, b = 1, 2, i, j = 1, 2, 3. (A.21)

Unfortunately, direct product vector spaces are not vector spaces themselves,
which can be readily seen by considering linear combinations of arbitrary
elements.
Because we need vector addition to represent superposition in QM, and we find
ourselves dealing with tensor products, we overcome this problem by extending
U × V to a larger space, the tensor product of U and V , denoted by U ⊗ V. This
tensor product space is defined to contain all linear combinations of elements of
the Cartesian product space U × V and satisfies all the axioms of a vector space
 A.5 Vector Spaces 355

over the common ground field F. Elements of U ⊗ V are either of the form u ⊗ v
(which is in U × V ) or linear combinations of such direct products, which may
or may not be in U × V .
Elements of U ⊗ V that are of the form u ⊗ v for u ∈ U , v ∈ V are called
separable. Elements of U ⊗ V that are not separable are called entangled.
Entanglement is of great significance in quantum mechanics. The physically
observable properties of entangled quantum states lie at the heart of the problems
with the interpretation of quantum mechanics.
As with ordinary arithmetic, the tensor product operation ⊗ takes precedence
over the vector summation operation + in U ⊗V , so we may leave out the brackets
and just write

(u1 ⊗ v 1 ) + (u2 ⊗ v 1 ) ≡ u1 ⊗ v 1 + u2 ⊗ v 2 ∈ U ⊗ V. (A.22)

Note that U ⊗ V is a vector space over the field F common to U and V .

Multiplication by a scalar can be considered in several ways:

λ {u ⊗ v} = (λu) ⊗ v = u ⊗ (λv) = λu ⊗ v. (A.23)

Similarly,

λ {u1 ⊗ v 1 + u2 ⊗ v 2 } = λu1 ⊗ v 1 + λu2 ⊗ v 2 . (A.24)

Denote the zero vectors in U , V , and U ⊗V by 0U , 0V , and 0U ⊗V , respectively.

Then for any u ∈ U , v ∈ V , we have

0U ⊗ v = u ⊗ 0V = 0U ⊗V . (A.25)

Likewise, if 0F is the zero element of the field F, then

0F (u ⊗ v) = 0U ⊗V . (A.26)

Rank
The rank of a tensor product space is the number of vector spaces in the product.
For example, U ⊗ V is a rank-two tensor product space; U 1 ⊗ U 2 ⊗ · · · ⊗ U n is
of rank n.
Elements of a given tensor product have the rank of that tensor product space.
Hence scalars have rank zero, while vectors and one-forms have rank one.
Given a number of vectors spaces V 1 , V 2 , . . . , V r , then the rank-r tensor
product space V 1 ⊗ V 2 ⊗ · · · ⊗ V r has dimension equal to the product of all
the individual dimensions, that is,
    
dim V 1 ⊗ V 2 ⊗ · · · ⊗ V r = dim V 1 dim V 2 . . . {dim V r } . (A.27)

Likewise, if V 1∗ , V 2∗ , . . . , V s∗ are dual vector spaces, we define the rank-s

tensor product space V 1∗ ⊗ V 2∗ ⊗ · · · ⊗ V s∗ in an analogous way, and similarly
for mixed tensor product spaces such as V 1 ⊗ V 2∗ ⊗ · · · .
356 Appendix

Separable Bases
If {ua : a = 1, 2, . . . , dim U } is a basis for U and {v b : b = 1, 2, . . . , dim V } is a
basis for V , then a frequently useful basis for U ⊗ V is given by {ua ⊗ v b : a = 1,
2, . . . , dim U , b = 1, 2, . . . , dim V }. Every element of this basis is a separable
element of the tensor product space U ⊗ V , so we call this a separable basis.
From this we immediately conclude that

dim {U ⊗ V } = {dim U } · {dim V } . (A.28)

This result generalizes to higher rank tensor product spaces.

Hilbert spaces
Hilbert spaces are finite or infinite dimensional vector spaces with a complete
inner product (Streater and Wightman, 1964).
 References

Adler, S. L. 1995. Quaternionic Quantum Mechanics and Quantum Fields. International

Series of Monographs on Physics, 88. Oxford University Press.
Adler, S. L. 2016. Does the Peres Experiment Using Photons Test for Hyper-complex
(Quaternionic) Quantum Theories? arXiv:quant-ph/1604.04950, 1–5.
Afshar, S. S. 2005. Violation of the Principle of Complementarity, and Its Implications.
Pages 229–244 of: The Nature of Light: What Is a Photon? Proceedings of SPIE,
no. 5866.
Apollo Program Office. 1969. Apollo 11 (AS-506) Mission. Mission Operation Report,
1–109.
Ashby, Neil. 2002. Relativity and the Global Positioning System. Physics Today, May,
41–47.
Aspect, A., Grangier, P., and Roger, G. 1982. Experimental Realization of Einstein–
Podolsky–Rosen–Bohm Gedankenexperiment: A New Violation of Bell’s Inequali-
ties. Phys. Rev. Lett., 49, 91–94.
Becker, L. 1998. A New Form of Quantum Interference Restoring Experiment. Phys.
Lett. A, 249, 19–24.
Bell, E. T. 1938. The Iterated Exponential Integers. Ann. Math., 39(3), 539–557.
Bell, J. S. 1964. On the Einstein–Podolsky–Rosen paradox. Physics, 1, 195–200.
Bell, J. S. 1988. Speakable and Unspeakable in Quantum Mechanics. Cambridge
University Press.
Berkeley, G. 1721. De Motu or Sive de Motus Principio & Natura et de Causa
Communicationis Motuum [The Principle and Nature of Motion and the Cause
of the Communication of Motions]. Translated by A. A. Luce Pages 253–276 of:
Michael R. Ayers, George Berkeley: Philosophical Works. London: Everyman, 1993.
Bernstein, J. 2010. The Stern–Gerlach Experiment. asXiv.org[physics.hist.ph],
arXiv:1007.2435, 1–16.
Birrell, N., and Davies, P. 1982. Quantum Fields in Curved Space. Cambridge University
Press.
Bjorken, J. D., and Drell, S. D. 1965. Relativistic Quantum Fields. McGraw-Hill.
Bohm, D. 1952. A Suggested Interpretation of the Quantum Theory in Terms of
“Hidden Variables,” I and II. Phys. Rev., 85, 166–193.
Bohr, N. 1913. On the Constitution of Atoms and Molecules. Philos. Mag., 26(1), 1–24.
Bombelli, L., Lee, J., Meyer, D., and Sorkin, R. 1987. Space-Time as a Causal Set.
Phys. Rev. Lett., 59(5), 521–524.
Bondi, H., and Gold, T. 1948. The Steady-State Theory of the Expanding Universe.
MNRAS, 108, 252–270.
Born, M. 1926. Zur Quantenmechanik der Stossvorgänge [The Quantum Mechanics of
the Impact Process (Collision Processes)]. Zeitschrift fur Physik, 37, 863–867.
The Born–Einstein Letters. 1971. Trans. Irene Born. Macmillan.
358 References

Brandt, H. E. 1999. Positive Operator Valued Measure in Quantum Information

Processing. Am. J. Phys., 67(5), 434–439.
Brightwell, G., and Gregory, R. 1991. Structure of Random Discrete Spacetime. Phys.
Rev. Lett., 66(3), 260–263.
Burnham, D. C., and Weinberg, D. L. 1970. Observation of Simultaneity in Parametric
Production of Optical Photon Pairs. Phys. Rev. Lett., 25, 84–86.
Candlin, D. J. 1956. On Sums over Trajectories for Systems with Fermi Statistics.
Nuovo Cimento, 4(2), 231–239.
Casalbuoni, R. 1976a. The Classical Mechanics for Bose–Fermi Systems. Nuovo Cimento
A Series 11, 33(3), 389–431.
Casalbuoni, R. 1976b. On the Quantization of Systems with Anticommuting Variables.
Nuovo Cimento A Series 11, 33(1), 115–125.
Cowan, C. L. Jr., Reines, F., Harrison, F. B., et al. 1956. Detection of the Free Neutrino:
A Confirmation. Science, 124(3212), 103–104.
Cramer, J. G. 1986. The Transactional Interpretation of Quantum Mechanics. Rev.
Mod. Phys., 58(3), 647–688.
D-Wave Systems. 2016. The D-Wave 2000Q Quantum Computer. D-Wave Systems,
1–12.
de Broglie, L. 1924. Recherches sur la Théorie des Quanta [Researches on the Quantum
Theory]. Ph.D. thesis, Faculty of Sciences at Paris University.
Deutsch, D. 1997. The Fabric of Reality. Penguin Press.
Deutsch, D. 1999. Quantum Theory of Probability and Decisions. Proc. R. Soc, 455,
3129–3137.
DeWitt, B. S. 1975. Quantum Field Theory in Curved Spacetime. Physics Reports C,
19(6), 295–357.
Dingle, H. 1967. The Case against Special Relativity. Nature, 216, 119–122.
Dirac, P. A. M. 1938a. Classical Theory of Radiating Electrons. Proc. Roy. Soc. A,
167, 148–169.
Dirac, P. A. M. 1938b. A New Basis for Cosmology. Proc. Roy. Soc (London) A,
165(921), 199–208.
Dirac, P. A. M. 1958. The Principles of Quantum Mechanics. Clarendon Press.
Eakins, J. 2004. Classical and Quantum Causality in Quantum Field Theory, or, “The
Quantum Universe.” Ph.D. thesis, University of Nottingham.
Eakins, J., and Jaroszkiewicz, G. 2005. A Quantum Computational Approach to the
Quantum Universe. Pages 1–51 of: Albert Reimer (ed.), New Developments in
Quantum Cosmology Research. Horizons in World Physics, vol. 247. New York:
Nova Science.
Eden, R. J., Landshoff, P. V., Olive, D. I., and Polkinghorne, J. C. 1966. The Analytic
S-Matrix. Cambridge University Press.
Ehrenfest, P. 1927. Bemerkung über die angebote Gültigkeit der klassischen Mechanik
Innerhalb der Quantenmechanik. Zeitschrift für Physik, 45(7–8), 455–457.
Einstein, A. 1905a. Über einen die Erzeugung und Verwandlung des Lichtes betreffenden
heuristischen Gesichtspunkt [Concerning an Heuristic Point of View Toward
the Emission and Transformation of Light]. Annalen der Physik, 17, 132–148.
Translation into English American Journal of Physics, 33(5), May 1965.
Einstein, A. 1905b. Zur Electrodynamik Bewgter Körper. Annalen der Physik, 17,
891–921. On the Electrodynamics of Moving Bodies, translation in The Principle
of Relativity, Dover Publications.
Einstein, A., Podolsky, B., and Rosen, N. 1935. Can Quantum-Mechanical Description
of Physical Reality Be Considered Complete? Phys. Rev., 47, 777–780.
 References 359

Elitzur, A. C., and Vaidman, L. 1993. Quantum-Mechanical Interaction-Free Measure-

ments. Found. Phys., 23, 987–997.
Encyclopaedia Britannica. 2000. Time. CD Rom. Britannica.co.uk.
Everett, H. 1957. “Relative State” Formulation of Quantum Mechanics. Rev. Mod.
Phys., 29(3), 454–462.
Feynman, R. P., and Hibbs, A. R. 1965. Quantum Mechanics and Path Integrals. New
York: McGraw-Hill.
Feynman, R. P., Leighton, R. B., and Sands, M. 1966. The Feynman Lectures on
Physics: Quantum Mechanics. Vol. III. Addison-Wesley.
FitzGerald, G. F. 1889. The Ether and the Earth’s Atmosphere. Science, 13, 390.
Franson, J. D. 1989. Bell Inequality for Position and Time. Phys. Rev. Lett., 62(19),
2205–2208.
Fuchs, C. A., and Peres, A. 2000. Quantum Theory Needs No Interpretation. Physics
Today, March, 70–71.
Gell-Mann, M., and Pais, A. 1955. Behavior of Neutral Particles under Charge
Conjugation. Phys. Rev., 97, 1387–1389.
Gerlach, W., and Stern, O. 1922a. Das magnetische Moment des Silberatoms. Zeits.
Phys., 9, 353–355.
Gerlach, W., and Stern, O. 1922b. Der experimentelle Nachweis der Richtungsquan-
telung im Magnetfeld. Zeits. Phys., 9, 349–352.
Glauber, R. J. 1963a. Coherent and Incoherent States of the Radiation Field. Phys.
Rev., 131(6), 2766 –2788.
Glauber, R. J. 1963b. Photon Correlations. Phys. Rev. Lett., 10(3), 84–86.
Glauber, R. J. 1963c. The Quantum Theory of Optical Coherence. Physical Review,
130(6), 2529–2539.
Gödel, K. 1949. An Example of a New Type of Cosmological Solution of Einstein’s
Field Equations of Gravity. Rev. Mod. Phys., 21(3), 447–450.
Goldstein, H. 1964. Classical Mechanics. Addison-Wesley.
Greenberger, D., and YaSin, A. 1989. “Haunted” Measurements in Quantum Theory.
Found. Phys., 19(6), 679–704.
Griffiths, R. B. 1984. Consistent Histories and the Interpretation of Quantum Mechan-
ics. J. Stat. Phys., 36(12), 219–272.
Halligan, Peter, and Oakley, David. 2000. Greatest Myth of All. New Scientist, 18
November, 34–39.
Hameroff, S. 1999. Quantum Computation in Brain Microtubules? The Penrose–
Hameroff “Orch-OR” Model of Consciousness. www.u.arizona.edu/∼hameroff/
royal.html, 1–30.
Hamming, R. W. 1950. Error Detecting and Error Correcting Codes. Bell Syst. Tech.
J., 29(2), 147–160.
Hardy, L. 1992. Quantum Mechanics, Local Realistic Theories and Lorentz-Invariant
Realistic Theories. Phys. Rev. Lett., 68(20), 2981–2984.
Heisenberg, W. 1925. Über Quantentheoretische Umdeutung Kinematischer und Mech-
anischer Beziehungen [Quantum-Theoretical Reinterpretation of Kinematic and
Mechanical Relations]. Zeits. Physik A Hadrons and Nuclei, 33(1), 879 –893.
Heisenberg, W. 1930. The Physical Principles of the Quantum Theory. Dover Edition,
1949 ed. University of Chicago Press.
Heisenberg, W. 1952. Questions of Principle in Modern Physics. Pages 41–52 of:
Philosophic Problems in Nuclear Science. London: Faber and Faber.
Horne, M. A., Shimony, A., and Zeilinger, A. 1989. Two-Particle Interferometry. Phys.
Rev. Lett., 62(19), 2209–2212.
360 References

Howson, A. G. 1972. A Handbook of Terms Used in Algebra and Analysis. Cambridge

University Press.
Hoyle, F. 1948. A New Model for the Expanding Universe. MNRAS, 108, 372–382.
Itano, W. M., Heinzen, D. J., Bollinger, J. J., and Wineland, D. J. 1990. Quantum Zeno
Effect. Phys. Rev. A, 41(5), 2295–2300.
Jacques, V., Wu, E., Grosshans, F., Treussart, F., Grangier, P., Aspect, A., and Roch,
J.-F. 2007. Experimental Realization of Wheeler’s Delayed-Choice Gedankenexper-
iment. Science, 315(5814), 966–968. www.arxiv.org/abs/quant-ph/0610241.
Jacques, V., Lai, N. D., Zheng, D., Chauvat, F., Treussart, F., Grangier, P., and
Roch, J.-F. 2008. Illustration of Quantum Complementarity Using Single Photons
Interfering on a Grating. New. J. Phys., 10, 123009.
Jaroszkiewicz, G. 2004. Quantum Register Physics. arXiv:quant-ph/0409094.
Jaroszkiewicz, G. 2008a. Quantized Detector Networks: A Review of Recent Develop-
ments. Int. J. Mod. Phys. B, 22(3), 123–188.
Jaroszkiewicz, G. 2008b. Quantized Detector Networks, Particle Decays and the
Quantum Zeno Effect. J. Phys. A: Math. Theor., 41(9), 095301.
Jaroszkiewicz, G. 2010. Towards a Dynamical Theory of Observation. Proc. Roy. Soc.
A, 466(2124), 3715–3739.
Jaroszkiewicz, G. 2016. Principles of Empiricism and the Interpretation of Quantum
Mechanics. Pages 139–173 of: M. Dugic, R. Kastner, and G. Jaroszkiewicz (eds.),
Quantum Structural Studies. World Scientific.
Jaynes, E. T. 2003. Probability Theory: The Logic of Science. Cambridge University
Press.
Joos, E. 2012. Decoherence. www.decoherence.de.
Jordan, P., and Wigner, E. P. 1928. Über Das Paulische Äquivalenzverbot. Zeitschrift
für Physik, 47, 631–651.
Kahneman, D. 2011. Thinking, Fast and Slow. Allen Lane.
Karyotakis, Y., and de Monchenault, G. H. 2002. A Violation of CP Symmetry in B
Meson Decays. Europhysics News, May/June, 89–93.
Kastner, R. E. 2005. Why the Afshar Experiment Does Not Refute Complementarity.
Stud. Hist. Philos. M. P., 36(4), 649–658.
Kastner, R. E. 2016. The Illusory Appeal of Decoherence in the Everettian Picture:
Affirming the Consequent. arXiv:1603.04845 [quant-ph], 1–7.
Kim, Y., Yu, R., Kulik, S., Shih, Y., and Scully, M. 2000. A Delayed Choice Quantum
Eraser. Phys. Rev. Lett., 84, 1–5. arXiv:quant-ph/9903047.
Kim, Yoon-Ho. 2003. Two-Photon Interference Without Bunching Two Photons. Phys.
Lett. A, 315, 352–355.
Klauder, J. R., and Sudarshan, E. C. G. 1968. Fundamentals of Quantum Optics. W.A.
Benjamin.
Klyshko, D. N., Penin, A. N., and Polkovnikov, B. F. 1970. Parametric Luminescence
and Light Scattering by Polaritons. JETP Lett., 11(1), 5–8.
Kochen, S., and Specker, E. 1967. The Problem of Hidden Variables in Quantum
Mechanics. J. Math. Mechanics, 17, 59–87.
Koke, S., Grebing, C., Frei, H., Anderson, A., Assion, A., and Steinmeyer, G. 2010.
Direct Frequency Comb Synthesis with Arbitrary Offset and Shot-Noise-Limited
Phase Noise. Nature Photonics, 4, 463–465.
Kracklauer, A. F. 2002. Time Contortions in Modern Physics. arXiv:quant-ph/0206164,
1–7. Draft for: Proceedings, The Nature of Time: Geometry, Physics and Percep-
tion; May 21–24, 2002, Tatranska Lomnica, Slovak Republic.
 References 361

Kraus, K. 1974. Operations and Effects in the Hilbert Space Formulation of Quantum
Theory. Berlin: Springer, pages 206–229.
Kraus, K. 1983. States, Effects, and Operations: Fundamental Notions of Quantum
Theory. Lecture Notes in Physics (190). Berlin: Springer-Verlag.
Kwiat, P. G., Steinberg, A. M., and Chiao, R. Y. 1993. High-Visibility Interference
in a Bell-Inequality Experiment for Energy and Time. Physical Review A, 47(4),
R2472–R2475.
Kwiat, P., Weinfurter, H., Herzog, T., Zeilinger, A., and Kasevich, M. 1994. Exper-
imental Realization of “Interaction-Free” Measurements. In: K. V. Laurikainen,
C. Montonen, and Sunnarborg (eds.), Symposium on the Foundations of Modern
Physics 1994. Editions Frontières.
Larmor, J. J. 1897. On a Dynamical Theory of the Electric and Luminiferous Medium,
Part 3, Relations with Material Media. Phil. Trans. Roy. Soc., 190, 205–300.
Laven, P. A., Taplin, D. W., and Bell, C. P. 1970. Television Reception over Sea Paths:
The Effect of the Tide. BBC Research Department Report, 1–25.
Leech, J. W. 1965. Classical Mechanics. Methuen and Co.
Leggett, A. J., and Garg, A. 1985. Quantum Mechanics versus Macroscopic Realism:
Is the Flux There When Nobody Looks? Phys. Rev. Lett., 54(9), 857–860.
Lehmann, H., Symanzik, K., and Zimmermann, W. 1955. Zur Formulierung Quan-
tisierter Feldtheorien. Il Nuovo Cimento, 1(1), 205–225.
Lewis, G. N. 1926. The Conservation of Photons. Nature, 118, 874–875.
Liang, Y., and Czarnecki, A. 2011. Photon–Photon Scattering: A Tutorial.
arXiv:1111.6126v2 [hep-ph], 1–9.
Lichtenberg, D. B. 1970. Unitary Symmetry and Elementary Particles. Academic
Press.
Lorentz, H. A. 1899. Simplified Theory of Electrical and Optical Phenomena in Moving
Systems. Proc. Acad. Science Amsterdam, I, 427–443.
Ludwig, G. 1983a. Foundations of Quantum Mechanics I. New York: Springer.
Ludwig, G. 1983b. Foundations of Quantum Mechanics II. New York: Springer.
Magueijo, J. 2003. New Varying Speed of Light Theories. Rept. Prog. Phys., 66(11),
2025–2068.
Mardari, G. N. 2005. What Is a Quantum Really Like? In: G. Adenier, A. Y. Khren-
nikov, and T. M. Nieuwenhuizen (eds.), Quantum Theory: Reconsideration of
Foundations–3. AIP Conference Proceedings, vol. 81.
Markopoulou, F. 2000. Quantum Causal Histories. Class. Quant. Grav., 17,
2059–2072.
Mars Climate Orbiter Mishap Investigation Board. 1999. Phase I Report. 1–44.
Martin, J. L. 1959a. The Feynman Principle for a Fermi System. Proc. Roy. Soc., A251,
543–549.
Martin, J. L. 1959b. Generalized Classical Dynamics and the “Classical Analogue” of
a Fermi Oscillator. Proc. Roy. Soc., A251, 536–542.
Merli, P. G., Missiroli, G. F., and Pozzi, G. 1976. On the Statistical Aspect of Electron
Interference Phenomena. Am. J. Phys., 44(3), 306–307.
Meschini, D. 2007. Planck-Scale Physics: Facts and Beliefs. Found. Science, 12(4),
277–294.
Minkowski, H. 1908. Space and Time. A translation of an Address delivered at the 80th
Assembly of German Natural Scientists and Physicians, at Cologne, 21 September,
75–91. Reprinted in H. A. Lorentz, A. Einstein, H. Minkowski, and H. Weyl, The
Principle of Relativity. Dover.
362 References

Misra, B., and Sudarshan, E. C. G. 1977. The Zeno’s Paradox in Quantum Theory.
J. Math. Phys., 18(4), 756–763.
Newton, I. 1687. The Principia (Philosophiae Naturalis Principia Mathematica).
University of California Press, 1999. New translation by I. B. Cohen and Anne
Whitman. University of California Press, 1999.
Newton, I. 1704. Opticks. 1952 ed. Dover Publications.
Newton, I. 2006. Original letter from Isaac Newton to Richard Bentley. The Newton
Project (online).
Nielsen, M. A., and Chuang, I. L. 2000. Quantum Computation and Quantum
Information. Cambridge University Press.
Paris, M. G. A. 2012. The Modern Tools of Quantum Mechanics (A Tutorial on
Quantum States, Measurement, and Operations). Eur. Phys. J. Special Topics,
203, 61–86.
Paul, H. 2004. Introduction to Quantum Optics. Cambridge University Press. Translated
from the German Photonen. Eine Einführung in die Quantenoptik, 2. Auflage
(1999).
Penrose, R. 1971. Angular Momentum: An Approach to Combinatorial Spacetime. In:
T. Bastin (ed.), Quantum Theory and Beyond. Cambridge University Press.
Peres, A. 1995. Quantum Theory: Concepts and Methods. Kluwer Academic.
Petkov, V. 2012. Space and Time: Minkowski’s Papers on Relativity. Minkowski
Institute Press, pages 1–37.
Planck, M. 1900a. On the Theory of the Energy Distribution Law of the Normal
Spectrum. Verhandl. Dtsch. Phys. Ges., 2(17), 237–245.
Planck, M. 1900b. Ueber eine Verbesserung der Wien’schen Spectralgleichung [On an
Improvement of Wein’s Equation for the Spectrum]. Verhandl. Dtsch. Phys. Ges.,
2(13), 202–204.
Poincaré, H. 1890. Sur Le Problème Des Trois Corps et Les Équations de la Dynamique.
Acta. Math., 13, 1–270.
Price, H. 1997. Time’s Arrow. Oxford University Press.
Procopio, L. M., et al. 2016. Experimental Test of Hyper-Complex Quantum Theories.
arXiv:1602.01624v2 [quant-ph].
Regge, T. 1961. General Relativity Without Coordinates. Il Nuovo Cimento, 19(3),
558–571.
Renniger, M. 1953. Zum Wellen-Korpuskel-Dualismus. Zeits. für Physik, 136,
251–261.
Requardt, M. 1999. Space-Time as an Orderparameter Manifold in Random Networks
and the Emergence of Physical Points. gr-qc/99023031, 1–40.
Ridout, D. P., and Sorkin, R. D. 2000. A Classical Sequential Growth Dynamics for
Causal Sets. Phys. Rev., D61, 024002. arXiv: gr-qc/9904062.
Rosa, R. 2012. The Merli–Missiroli–Pozzi Two-Slit Electron-Interference Experiment.
Phys. Perspect., 14, 178–195.
Scarani, V., Tittel, W., Zbinden, H., and Gisin, N. 2000. The Speed of Quantum
Information and the Preferred Frame: Analysis of Experimental Data. Phys. Lett.,
A276, 1–7.
Schrödinger, E. 1926. Quantisierung als Eigenwertproblem (Erste Mitteilung). Ann.
Phys., 384(4), 361–376.
Schrödinger, E. 1935. The Present Situation in Quantum Mechanics. Naturwis-
senschaften, 23(48), 807–812. Original title: Die gegenw artige Situation in der
 References 363

Quantenmechanik, reprinted in Quantum Theory and Measurement, edited by J. A.

Wheeler and W. H. Zurek, Princeton University Press, 1983.
Schutz, B. 1980. Geometrical Methods of Mathematical Physics. Cambridge University
Press.
Schwinger, J. 1958. Spin, Statistics and the TCP theorem. Proc. N. A. S., 44,
223–228.
Schwinger, J. 1969. Particles and Sources. Gordon and Breach.
Schwinger, J. 1998a. Particles, Sources, and Fields. Advanced Books Classics. Reading,
MA: Perseus.
Schwinger, J. 1998b. Particles, Sources, and Fields. Advanced Books Classics, vol. 2.
Reading, MA: Perseus.
Schwinger, J. 1998c. Particles, Sources, and Fields. Advanced Books Classics, vol. III.
Reading, MA: Perseus.
Sen, R. N. 2010. Causality, Measurement Theory and the Differentiable Structure
of Space-Time. Cambridge Monographs on Mathematical Physics. Cambridge
University Press.
Shannon, C. E. 1948. A Mathematical Theory of Communication. Bell Syst. Tech. J.,
27(July, October), 379–423, 623–656.
Sillitto, R. M., and Wykes, C. 1972. An Interference Experiment with Light Beams
Modulated in Anti-Phase by an Electro-Optic Shutter. Phys. Lett. A, 39(4),
333–334.
Sinha, U., et al. 2010. Ruling Out Multi-Order Interference in Quantum Mechanics.
Science, 329, 418 –421.
Snyder, H. S. 1947a. The Electromagnetic Field in Quantized Space-Time. Phys. Rev.,
72(1), 68–71.
Snyder, H. S. 1947b. Quantized Space-Time. Phys. Rev., 71(1), 38–41.
Sorkin, R. D. 1994. Quantum Mechanics as Quantum Measure Theory. Mod. Phys. Lett.
A, 9, 3119–3128.
Streater, R. F., and Wightman, A. S. 1964. PCT, Spin and Statistics, and All That.
W.A. Benjamin.
Stuckey, M. 1999. Pregeometry and the Trans-Temporal Object. In: R. Buccheri, V.
Di Gesu, and M. Saniga (eds.), Studies of the Structure of Time: from Physics to
Psycho(Patho)Logy, 121–128. Dordrecht: Kluwer.
Taylor, G. I. 1909. Interference Fringes with Feeble Light. Proc. Camb. Philos. Soc.,
15, 114–115.
Tropper, A. M. 1969. Linear Algebra. Thomas Nelson and Sons.
Unruh, W. G. 1976. Notes on Black-Hole Evaporation. Phys. Rev. D, 14(4),
870–892.
von Neumann, J. 1955. The Mathematical Foundations of Quantum Mechanics.
Princeton University Press. Originally published as Mathematische Grundlagen
der Quantenmechanik, Berlin: Springer, 1932.
Walborn, S. P., Cunha, M. O. Terra, Pádua, S., and Monken, C. H. 2002. Double–Slit
Quantum Eraser. Phys. Rev., 65, 033818 1–6.
Wheeler, J. A. 1979. From the Big Bang to the Big Crunch. Cosmic Search Magazine,
1(4). Interview with J. A. Wheeler.
Wheeler, J. A. 1980. Pregeometry: Motivations and Prospects. Pages 1–11 of:
A. R. Marlow (ed.), Quantum Theory and Gravitation. New York: Academic
Press.
364 References

Wheeler, J. A. 1983. Quantum Theory and Measurement. Princeton Series in Physics.

Princeton University Press, pages 182–213.
Woit, P. 2006. Not Even Wrong: The Failure of String Theory and the Search for Unity
in Physical Law. Basic Books.
Wu, C. S., Ambler, E., Hayward, R. W., Hoppes, D. D., and Hudson, R. P. 1957.
Experimental Test of Parity Conservation in Beta Decay. Phys. Rev., 105(4),
1413–1415.
Zurek, W. 2002. Decoherence and the Transition from Quantum to Classical–Revisited.
Los Alamos Science, (27), 2–24.
 Index

action-at-a-distance, 217 Born–Einstein letters, 180

aleatoric uncertainty, 54 bosonic filter operator, 315
ammonium, 198, 209 bosonic identity operator, 316
ancilla, 119, 123 bosonic operator, 312
annihilator, 326 bosonic oscillator, 309
answer set, 18, 28 bosonic projection operators, 315
collapse, 19 bosonic quantum register operator, 316
antifringe, 196 bosonic state, 315, 316
antiparticles, 52 bottom-up viewpoint, 300
antiunitary transformation, 57 Brandt’s network, 174
Apollo Mission Control, 139
architecture, 10, 76 calibrated apparatus, 98, 201
asymmetry, 302 calibrated detector, 267
atom, 292, 293 calibrated evolution, 81
autonomous, meaning of, 78 calibration, 239
Avogadro’s number, 17 basis, 245
cardinality, 60
baseball in the night, 133 category error, 21
basis transformations, 54 causal set, 89, 298, 299, 302
Bayesian school, 42 ancestor, 302
beam splitter, 148, 157 child, 302
Bell, J. S., 234 hypothesis, 300
Bell inequality, 4, 235, 271 parent, 302
Bell numbers, 294 sibling, 302
beta decay, 242 causally independent elements, 302
binary components, 67 change and persistence, 250
binary decomposition, 72 charge, parity, time (CPT) theorem, 15
binary digit, 29 chronon, 85
binary question, 28, 29 circularity, 120
binary sequence, 312 classical answer, 33
bit, 29, 46 classical binary register, 60
classical, 264 classical bit, 29
down operator, 34 classical causality, 36
flip operator, 34 classical conditioning, 233
identity operator, 34 classical ensemble, 54
operators, 46 classical information, 257
power set, 322 classical labstate, 35
state, 30 classical logic, 23
up operator, 34 classical observable, 83
bitification principle, 29 classical oscillator register, 310
black body, 8 classical proposition, 27
Block Universe, 15, 75, 205, 281 classical register, 32, 76
Boolean algebra, 31 operator, 68, 78
Born interpretation, 50 classical trajectory, 79
Born map, 93 closed timelike curves, 181, 346
Born measure of similarity, 256 coherence time, 272
366 Index

color vision theory, 152 Einstein, Albert, 8

complementarity, 136 Einstein locality, 180, 301
complete proposition, 27 electric field operator, 87
completeness, 116 elementary transition operator (ETO),
complex numbers, 52 49
computational basis Elitzur–Vaidman bomb-tester, 322, 335
map, 66 emergence, 16
representation, 66, 78, 110, 257 emergent level, 310
computer algebra, 163 emission time, 272
computerization, 163 empirical questions, 13
configuration space, 233 empty space, 148
conjugate pair, 109 endophysics, 346
“Consistent Histories” approach, 348 enhanced Stern–Gerlach experiment, 241
construction operator, 327 ensemble, 348
contextual answer, 18 entangled state, 87
contextual basis, 141 entanglement, 162, 291, 294, 295
contextual completeness, 14, 26 partition, 297
contextual identity, 142 epistemic uncertainty, 58, 118
contextual overcompleteness, 289 ETO, see elementary transition operator
contextual subspace, 141, 142, 163 event, in space and time, 88
contextual vacuum, 322, 331 existential incompatibility, 21
contextuality, 19, 163 expectation value, 118
continuum map, 313 experiment, 39
coordinate time, 89 double-slit, 280, 306, 322
Copenhagen interpretation, 2, 9 FRANSON, 271
correlation, 259 FRANSON-1, 273
of probabilities, 261 FRANSON-2, 274
countable additivity, 115 FRANSON-3, 277
counterfactual definiteness, 239 JACQUES, 189
counterfactuality, 238 KIM, 181
CPT theorem, see charge, parity, time WALBORN-1, 190
(CPT) theorem WALBORN-2, 191
crosstalk, 347 WALBORN-3, 194
cycle, 80, 81 WHEELER-1, 186
WHEELER-2, 188
decoherence, 2 exponential decay, 205
decommissioned detector, 22 extended QDN, 322
decommissioned state, 324 extended reals, 115
decommissioning, 154 external architecture, 77
operator, 328 external node, 31
delayed-choice interferometer, 179
delayed-choice quantum eraser, 179, 181 faulty detector, 22, 329
demolitive measurement, 20 faulty state, 324
density operator, 58, 118 fermionic operator, 320
detection operators, 122 fermionic oscillator, 309, 320
detectors, 31 Feynman, R. P., 103
deterministic outcomes, 237 fine structure constant, 287
direct sum, 292 finite countable states, 312
Dobinski’s formula, 294 first law of time, 77, 250
Doppler effect, 89 Fitzgerald, George F., 8
double-slit experiment, 131, 280 Fock space, 149, 319
monitored, 144, 306 frame fields, 89, 346
dual bit, 32 Franson experiment, 271
state, 32 Frequentist school, 42
dual causal set, 307 Fresnel’s equations, 158
dual qubit state, 48 fringe, 196
dual vector space, 353 full trace, 119
dyadic object, 33, 46 fully occupied state, 313
dynamical null test, 101 fully saturated labstate, 81
 Index 367

galactic lensing, 179, 187 laboratory, 328

general covariance, 4, 7 frame, 346
generalized labstate, 328 labstate, 1, 12, 34
generalized measurement, 122 normalized, 107
generalized proposition, 25 labtime, 36, 89
Global Positioning System, 279 Lagrange, Joseph-Louis, 7
ground field, 351 Large Hadron Collider, 281
ground state, 47 Larmor, Joseph, 8
Leaning Tower of Pisa, 280
half-life, 53 left-stochastic matrix, 44
Hamilton’s equations, 79 Leggett–Garg correlation, 265
Hamilton, William Rowan, 7 Leggett–Garg inequality, 263, 265, 271
Hamiltonian flow, 79 Leucippus, 39
Hamming distance, 257 lightcone, 218
Hamming measure of similarity, 256 causality, 219
Hamming operator, 258 lightlike, 302
Hardy paradox, 322 linear functional, 353
experiment, 339 linear operator, 116
Hasse diagram, 302 linear transformation, 352
Heisenberg cut, 91, 254 linearity, 93
Heisenberg picture, 85 Liouville’s theorem, 79
Heisenberg point, 232 local subregister, 225
Heraclitus, 39 locality, 217
dictum of, 250 Lorentz, Hendrik A., 8
hidden variables, 2, 232 Lorentz signature metric, 346
Higgs particle, 25
Hitchens’s razor, 234, 282 Mach–Zehnder interferometer, 172
Hume, David, xvi macrorealism, 263
MAIN, 164, 174
identity class, 111 Manifold Time, 14, 15
identity map, 49 map, 11
in dubio pro reo, 28 Mars Climate orbiter, 82
incomparable events, 302 mathematical metaphysics, 22
incompatible question, 21 mathematical proposition, 27
information void, 6, 17, 31, 87, 148, matrix mechanics, 2, 105, 310
149, 329 matrix representation, 33
injection, 11, 94 maximal chain, 303
inner product, 48 maximal question, 63, 71, 105, 107, 108, 110
interaction-free measurement, 339 measure, 115
interference term, 135 measurement, 39
interferometer, 172 of the first kind, 20
interframe experiments, 89 problem, 114
internal architecture, 77 of the second kind, 20
internal node, 31 metaphysical proposition, 27
intervention, 38 metaphysics, 3
irreflexivity, 302 microscopic reversibility, 51, 102
isometric isomorphism, 291 minimum rank, 72
minus infinity, 115
Kaon decay, 212 mirror, 148, 160
Kaon regeneration, 198 mixed initial states, 127
Kemmer, Nicholas, xvi mixed labstates, 58
Kepler, Johannes, 75 mixed states, 118
kernel, 353 modularization, 163
ket, 354 module, 35, 148
Kraus cut, 91 monitored double-slit experiment, 144, 306
Kraus matrix, 151 multiorder interference, 52
Kraus operator, 122 Multiverse, 2, 56
368 Index

Naimark’s theorem, 123 path

negation, 23, 28 integral, 102, 103
proof of, 24 summation, 102
Newton’s third law, 264 peaceful co-existence, 17, 36
Newton, Isaac, 7 Pearson correlation coefficient, 260
Newtonian prism, 148, 152 permutation flow, 80
no-communication theorem, 180, 218, 229 persistence, 7, 40, 77, 84, 88, 100, 101,
no-signaling theorem, 196 250, 280
node, 31 effective, 77
noncontextuality, 19 persistent image, 101
nondemolition, 35 phase changer, 160
nondestructive test, 20 phase diagram, 77
noninvasive measurability, 264 phase shifter, 148
noninvasive measureability, 7 phonon, 63
nonlinear Born map, 95 photoelectric, 8
nonlocality photon, 8
spatial, 180 photon–photon scattering, 287
temporal, 180 physical register, 334
nonnegativity, 115 physical space, 233
nonseparable, 309 physics, 13
norm, 48
Planck, Max, 6, 8
preservation, 93
Planck time, 85, 206
normal bit states, 334
plus infinity, 115
normalized qubit, 48
Poincare recurrence theorem, 81
not even wrong, 22
positive operator, 116
NOT gate, 34
positive operator-valued measure, 86,
not proven, 29
114
null evolution, 100
positronium, 292
null evolution operator, 100, 101
postselection, 184
null module, 148, 161
null test, 20, 100 POVM, 86, 122
nullity, 353 elements, 122
nullius in verba, 234 operators, 168
power bit, 323, 324
objectivization, 3 matrix, 325
observable, 86, 105, 117 nilpotent, 327
off-line state, 324 operator, 325
Old Quantum Mechanics, 114 orthogonality, 327
one forms, 353 power set, 323
operator, 11, 116 preferred basis, 46, 47, 62
orbit, 81 pregeometry, 300
origin problem, 301 preparation switch, 127, 139
ortho-positronium, 292 prior, 42
orthonormal basis, 116 prior information, 42
orthonormal d-subset, 96 Process Time, 14, 43
Program MAIN, 163
para-positronium, 292 projection operators, 50
parameters, 234 projection-valued measure, 85, 114
Parmenides, 39 projectors, 117
partial destruction, 20 propagator, 137
partial observation, 105 propensity, 43
partial question, 63, 169 proper time, 89
eigenvalue, 110 proposition, 22
partial trace, 119 pure labstate, 107
participatory principle, 5, 240 pure state, 54, 118
particle decay, 198 purification, 120
partition, 291 purity, 121
element, 297 PVM, 85
 Index 369

q-tick, 85 rods and cones, 251

quantum answer, 54 run, 38, 132
quantum bit, 29, 31, 46 runtime, 39
quantum correlation, 220
quantum cosmology, 88 s-bit, 31
quantum electrodynamics, 2, 287, 333 S-Matrix, 17, 333
quantum ensemble, 54 saturation
quantum erasure, 179 (of beam splitter), 159
quantum field theory, 86, 309 operator, 74
quantum gravity, 299 Schrödinger picture, 85
quantum Hamming measure of dissimilarity, Schrödinger wave function, 233
259 Schrödinger’s cat, 287
quantum information, 257 Schwinger, Julian, xvi
quantum interference, 131 Schwinger counterfactual reality, 337
quantum labstate, 35 Schwinger source theory, 98
quantum register, 60, 65, 291 Scottish law trial, 29
quantum tick, 85 self-interference, 135, 287
quantum Zeno effect, 35, 198, 205 self-intervening network, 279
quantum zipping, 298 semi-unitary operator, 93
qubit, 29, 31, 46 separable basis, 356
map, 49 separable state, 87
operator, 49 separation, 291, 294
question product, 295
and answer notation, 18 sequential growth, 302
cardinality, 29 Shannon, C. E., 29
eigenvalue, 111 shape function, 138
shielding, 37, 89, 347
rank, 353 sigma-algebra, 115
of apparatus, 60 signal
of register, 292 basis representation, 61, 62, 66, 110, 312
rank-one dynamics, 35 bit, 30
rank-one partial question, 109 operator, 50
rank-two dynamics, 35 bit state, 30
rank-zero partial question, 109, 110 decomposition, 169
realism, 4 operator, 50
recurring sequence states, 313 projection operators, 326
red shift, 148 qubit, 46
reductionism, 16 rate, 178
register set, 69
operator, 67 algebra, 70
projection operators, 69 state, 47
signal operators, 69 theorem, 84, 97
relative external context, 25, 75, 149 signality, 63, 71, 78
relative internal context, 25, 76, 149 class, 71, 81
Relative State interpretation, 2, 56 conserving flows, 81
relatively local, 225 count, 72
relatively remote, 225 Sillitto–Wykes experiment, 142, 143
relatively spacelike, 302 simultaneity problem, 241
relativistic quantum field theory, 106, 287, source, 31
331, 333 theory, 98, 281
relativity, 299 space-like hypersurfaces, 89
remote subregister, 226 space-time, Newtonian, 233
Renniger thought experiment, 127 spacetime, 15
resolution, of the identity, 116 spatial ensemble, 348
retina, 251 special operator, 326
retraction, 11, 68, 94, 95 special relativity, 7
reversibility, 81 spectral decomposition, 117
rhetorical question, 29 spectrum, 152
370 Index

speculative questions, 22 transition bit operators, 33

spin, 241 transitivity, 302
split, 291, 293 transtemporal identity, 39
spooky action at a distance, 180 truth values, 23
stage, 35, 61, 85 Tukey, J. W., 29
concept, 180 two-photon interferometer, 176
diagram, 38, 166 Tycho, 75
network, 31, 35 type four experiment, 282, 336
standard form, 158 type one experiment, 280
Standard Model, 17 type three experiment, 281
state reduction, 19, 95 type two experiment, 281
Steady State paradigm, 16 type zero experiment, 279
Stern–Gerlach experiment, 235
stochastic answer, 44 unital associative algebra, 50
stochastic binary question, 45 unitary operator, 94
stochastic bit, 31, 44, 46, 47 universal register, 322, 329
stochastic jump, 44
stochastic labstate, 35
vacuous proposition, 14
stochastic vector, 44
vacuous theory, 22
subregister, 291, 292
vacuum, 148, 149, 331
subspace, 351
expectation values, 87
surjective linear map, 291
validation function, 23
system state, 12, 34
variables, 235
vector addition, 350
tachyon, 218 vector space, 350
temporal correlation, 262, 263, 265 virtual detector, 201, 279
temporal ensemble, 349 virtual module, 279
temporality, 39 von Neumann entropy, 121
theoretical questions, 13
Theory of Everything, 16
time, 14, 77 wave function collapse, 19
scale, 85 wave mechanics, 2
time-like foliation, 89 wave–particle duality, 131
top-down viewpoint, 300 wave-particle duality, 136
total Hilbert space, 11 which-path, 182
total signality, 72 measure, 179, 185
total state, 12, 119, 165 Wollaston interferometer, 163
trace, 117 Wollaston prism, 148, 150
transactional interpretation, 181
transbosonic state, 309, 316 Zeno of Elea, 39
transformation, 11 zero sequence, 314