Fortin, Sebastian - Holik, Federico - Lombardi, Olimpia - López, Cristian - Quantum Worlds - Perspectives On The Ontology of Quantum Mechanics-Cambridge University Press (2019)

QUANTUM WORLDS
Quantum theory underpins much of modern physics, and its implications draw the
attention of industry, academia, and public funding agencies. However there are
many unsettled conceptual and philosophical problems in the interpretation of
quantum mechanics, which are a matter of extensive debate. These hotly debated
topics include the meaning of the wave function, the nature of the quantum objects,
the role of the observer, the nonlocality of the quantum world, and the emergence
of classicality from the quantum domain. Containing chapters written by eminent
researchers from the fields of physics and philosophy, this book provides interdis-
ciplinary, comprehensive, and up-to-date perspectives of the problems related to
the interpretation of quantum theory. It is ideal for academic researchers in physics
and philosophy who are working on the ontology of quantum mechanics.
olimpia lombardi is a principal researcher at the National Scientific and

Technical Research Council (In Spanish: Consejo Nacional de Investigaciones
Científicas y Técnicas [CONICET]). She is the director of the research group in
the philosophy of physics and philosophy of chemistry at the University of Buenos
Aires, and has been awarded grants from the Foundational Questions Institute and
John Templeton Foundation.
sebastian fortin and federico holik are research fellows at the National
Scientific and Technical Research Council (CONICET). cristian lόpez is a
PhD student at the University of Buenos Aires and the University of Lausanne. All
are members of the group headed by Olimpia Lombardi and collaborated in
organizing the international workshop Identity, indistinguishability and non-locality
in quantum physics (Buenos Aires, 2017) on which this volume is based.
QUANTUM WORLDS
Perspectives on the Ontology of Quantum Mechanics
Edited by
OLIMPIA LOMBARDI
University of Buenos Aires
SEBASTIAN FORTIN
CRISTIAN LÓPEZ
FEDERICO HOLIK
National University of La Plata
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/9781108473477
DOI: 10.1017/9781108562218
© Cambridge University Press 2019
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 2019
Printed in the United Kingdom by TJ International Ltd, Padstow Cornwall
A catalogue record for this publication is available from the British Library.
Library of Congress Cataloging-in-Publication Data
Names: Lombardi, Olimpia, editor. | Fortin, Sebastian, 1979– editor. | López, Cristian, editor. |
Holik, Federico, editor.
Title: Quantum worlds : perspectives on the ontology of quantum mechanics / edited by
Olimpia Lombardi (Universidad de Buenos Aires, Argentina), Sebastian Fortin
(Universidad de Buenos Aires, Argentina),
Cristian López (Universidad de Buenos Aires, Argentina), Federico Holik
(Universidad Nacional de La Plata, Argentina).
Description: Cambridge ; New York, NY : Cambridge University Press, 2019. |
Includes bibliographical references and index.
Identifiers: LCCN 2018045102 | ISBN 9781108473477 (hardback)
Subjects: LCSH: Quantum theory.
Classification: LCC QC174.12 .Q385 2019 | DDC 530.12–dc23
LC record available at https://lccn.loc.gov/2018045102
ISBN 978-1-108-47347-7 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
List of Contributors page vii

Preface xi
Introduction 1
Part I Ontology from Different Interpretations of
Quantum Mechanics 7
1 Ontology for Relativistic Collapse Theories 9
wayne c. myrvold
2 The Modal-Hamiltonian Interpretation: Measurement,

Invariance, and Ontology 32
olimpia lombardi
3 Quantum Mechanics and Perspectivalism 51

dennis dieks
4 Quantum Physics Grounded on Bohmian Mechanics 71

nino zanghı̀
5 Ontology of the Wave Function and the Many-Worlds Interpretation 93

lev vaidman
6 Generalized Contexts for Quantum Histories 107

marcelo losada, leonardo vanni, and roberto laura
Part II Realism, Wave Function, and Primitive Ontology 119

7 What Is the Quantum Face of Realism? 121
james ladyman
8 To Be a Realist about Quantum Theory 133

hans halvorson
v
vi Contents
9 Locality and Wave Function Realism 164

alyssa ney
Part III Individuality, Distinguishability, and Locality 183

10 Making Sense of Nonindividuals in Quantum Mechanics 185
jonas r. b. arenhart, otávio bueno, and décio krause
11 From Quantum to Classical Physics: The Role of Distinguishability 205

ruth kastner
12 Individuality and the Account of Nonlocality: The Case for the

Particle Ontology in Quantum Physics 222
michael esfeld
13 Beyond Loophole-Free Experiments: A Search for Nonergodicity 245

alejandro a. hnilo
Part IV Symmetries and Structure in Quantum Mechanics 267

14 Spacetime Symmetries in Quantum Mechanics 269
cristian lópez and olimpia lombardi
15 Symmetry, Structure, and Emergent Subsystems 294

nathan harshman
16 Majorization, across the (Quantum) Universe 323

guido bellomo and gustavo m. bosyk
Part V The Relationship between the Quantum Ontology and

the Classical World 343
17 A Closed-System Approach to Decoherence 345
sebastian fortin and olimpia lombardi
18 A Logical Approach to the Quantum-to-Classical Transition 360

sebastian fortin, manuel gadella, federico holik, and
marcelo losada
19 Quantum Mechanics and Molecular Structure:

The Case of Optical Isomers 379
juan camilo martı́nez gonzález, jesús jaimes arriaga, and
sebastian fortin
Index 393
Contributors
Jonas R. B. Arenhart
Federal University of Santa Catarina
Guido Bellomo
University of Buenos Aires – CONICET
Gustavo M. Bosyk
National University of La Plata – CONICET
Otávio Bueno
University of Miami
Dennis Dieks
Utrecht University
Michael Esfeld
University of Lausanne
Sebastian Fortin
Manuel Gadella
University of Valladolid
Hans Halvorson
Princeton University
vii
viii List of Contributors
Nathan Harshman
American University
Alejandro Hnilo
CITEDEF – CONICET
Federico Holik
National University of La Plata – CONICET
Jesús Jaimes Arriaga

Ruth Kastner
University of Maryland
Décio Krause
Federal University of Santa Catarina
James Ladyman
University of Bristol
Roberto Laura
National University of Rosario
Olimpia Lombardi
Cristian López
University of Buenos Aires – University of Lausanne – CONICET
Marcelo Losada
National University of Rosario – CONICET
Juan Camilo Martínez González

Wayne C. Myrvold
University of Western Ontario
List of Contributors ix
Alyssa Ney
University of California, Davis
Lev Vaidman
Tel Aviv University
Leonardo Vanni
Nino Zanghì
University of Genova
Preface
Developing a research group in philosophy of physics is not an easy task in

Argentina, the southernmost country in South America. In addition to language
barriers and the lack of a tradition on the matter, distance is the main obstacle to
attend the best academic meetings and to visit renowned research centers. For this
reason, the Grant 57919 that we were awarded by the John Templeton Foundation
(JTF) represented an invaluable support to our work: It allowed us to develop an
intense activity manifested in publications and participation in specialized confer-
ences, which gave a qualitative boost to our research development.
In the context of this JTF grant, we organized the International Workshop
Identity, indistinguishability and non-locality in quantum physics, held in Buenos
Aires from June 26 to June 29, 2017. We were proud to welcome some of the most
salient international specialists on the interpretation of quantum mechanics, who
kindly accepted our invitation to participate in the workshop and to contribute to
the present volume: Dennis Dieks, Michael Esfeld, Hans Halvorson, Nathan
Harshman, Alejandro Hnilo, Ruth Kastner, Décio Krause, James Ladyman, Wayne
Myrvold, Alyssa Ney, Lev Vaidman, and Nino Zanghì. Our first acknowledgment
is for them. Those who want to know about the workshop can access to the link
www.filoexactas.exactas.uba.ar/project-ontology/workshop.html, which contains
the videos of the full talks and of the lively final discussion.
Of course, our most special acknowledgment goes to the John Templeton
Foundation, which made possible the successful development of our project. But
since institutions do not exist without the people who embody them, we want to
particularly thank Alexander Arnold for his continued support during the three
years of work.
We also want to acknowledge the Academia Nacional de Ciencias Exactas y
Naturales, which supplied the venue for the meeting. However, the workshop
would not have been successful without the essential assistance of the members
of the Group of Philosophy of Science led by Olimpia Lombardi and based both in
xi
xii Preface
the Faculty of Exact and Natural Sciences and in the Faculty of Philosophy and
Letters of the University of Buenos Aires: Hernán Accorinti, Guido Bellomo,
Martín Bosyk, Mariana Córdoba, María José Ferreira Ruiz, Manuel Herrera, Jesús
Jaimes Arriaga, Martín Labarca, Marcelo Losada, Juan Camilo Martínez González,
Erick Rubio, Leonardo Vanni, and Alfio Zambon. Their strong commitment and
unlimited enthusiasm made the organization of the meeting an enjoyable task.
Although the presentations from the workshop were the basis for this book, we
are also grateful to Jonas Arenhart, Otávio Bueno, and Manuel Gadella, who
graciously accepted the invitation of the editors to contribute to the present volume
in different chapters. Last, but not least, we want to express our gratitude to
Cambridge University Press, in the person of Simon Capelin, Editorial Director
(physical sciences) and Sarah Lambert, Editorial Assistant, for their support and
assistance during all the stages of this project.
Introduction
In its original meaning, ‘ontology’ is the study of what there is – not only of what
entities exist but also of the very structure of reality. For the most part of the history
of philosophy, ontology was the core of metaphysics, perhaps the major branch of
philosophy. Nowadays, however, the word has different meanings and nuances. In
the analytic tradition, for instance, ontology is the study not only of what there is,
but also of the most general features of and the relations among what there is. This
study commonly starts out from our intuitions about reality or from an a priori
reasoning. Yet, another, increasingly growing sense of ontology has to do with
reality itself in relation to our best scientific theories: When one asks for “the
ontology of” a certain scientific theory, the question is about what reality would be
like if the theory were true. Although this second meaning does not exactly match
the etymology of the word (from Ancient Greek: on, what is; logos, discourse,
account), the meaning drift is completely natural in the light of the fact that, at least
after the Renaissance, scientific knowledge was crucial with respect to how the
structure of reality and the nature of its entities were conceived.
Quantum mechanics is probably the most successful and the least understood
physical theory that we have ever had. Even though this claim has become almost a
cliché, its frequent repetition does not make it less true. Indeed, after almost a
century of its first formulations, quantum mechanics is still posing unsolved
puzzles with respect to our understanding of the microscopic world. Of course,
numerous important results have been obtained during years of research, and many
of them are relevant to the foundations and the interpretation of the theory.
However, it is not completely clear yet how the ontology of the theory is, in
particular, how reality would be if quantum mechanics were true. Not only does
the question remain in force just as in the first decades of the twentieth century, but
a century of philosophical and scientific discussions has brought to light that
quantum reality is far more complex than originally supposed. The numerous
and varied perspectives developed up to the present time just manifest this
1
2 Quantum Worlds
complexity. If each different perspective tells us a different story about the

quantum realm, then the current variety of perspectives point out to the many
“quantum worlds” we have come to conceive to date.
The aim of this volume is, precisely, to present this variety of “quantum worlds”
in the most unbiased way. The different perspectives on the ontology of quantum
mechanics that this volume compiles rely on different metaphysical commitments,
diverse formal tools, diverging aims, or even disparate readings of the theory’s
formalism. All this not only makes manifest how rich and puzzling quantum
mechanics is for our understanding of the physical world, but also how bridges
between philosophy and physics can be built in order to make progress in such
understanding. To unfold the wide variety of perspectives in an organized way, the
volume is structured in five parts.
Part I, “Ontology from Different Interpretations of Quantum Mechanics,”
groups the chapters focusing on particular interpretations already proposed in the
literature on the matter. This part opens with the chapter, “Ontology for Collapse
Theories,” where Wayne Myrvold claims that the natural ontology for a collapse
theory is a distributional ontology, according to which dynamical quantities, such
as charge or mass within a specified region, do not take on precise values, but
rather have associated with them a distribution of values; this chapter discusses the
extension of such a picture to the context of a relativistic spacetime. The second
chapter, “The Modal-Hamiltonian Interpretation: Measurement, Invariance, and
Ontology” by Olimpia Lombardi, draws the attention to this interpretation by
exposing it in a conceptually clear and concise way, stressing its advantages both
for dealing with the traditional interpretive problems of quantum mechanics and
for supplying a physically meaningful account of relevant aspects of the theory.
The main aim of the third chapter, “Quantum Mechanics and Perspectivalism” by
Dennis Dieks, is to argue for a perspectival noncollapse interpretation that, by
assigning relational or perspectival states, makes it possible to reconcile universal
unitary evolution, and the resulting omnipresence of entangled states, with the
occurrence of definite values of physical quantities. In the fourth chapter, “Quan-
tum Physics Grounded on Bohmian Mechanics,” Nino Zanghì rejects the inter-
pretations based on the concepts of measurement and observer; according to him,
the theory must be based on a clear primitive ontology that provides the spatio-
temporal histories of its basic entities. In the fifth chapter, “Ontology of the Wave
Function and the Many-Worlds Interpretation,” Lev Vaidman undertakes a vigor-
ous defense of this interpretation by claiming that what we see is only a tiny part of
what there is: There are multiple parallel worlds similar to ours, and our experi-
ences supervene on the wave function of the universe. In the sixth and closing
chapter of the first part, “Generalized Contexts for Quantum Histories,” Marcelo
Losada, Leonardo Vanni, and Roberto Laura propose a formalism of generalized
Introduction 3
contexts for quantum histories, in which the contexts of properties at each time
must satisfy compatibility conditions given by commutation relations in the Hei-
senberg representation; any family of histories satisfying these conditions is
organized in a distributive lattice with well-defined probabilities obtained by a
natural generalization of the Born rule.
Part II, “Realism, Wave function, and Primitive Ontology,” is devoted to the
question of realism in the quantum domain in general and, in particular, of realism
regarding the wave function. In the first chapter, “What Is the Quantum Face of
Realism?,” James Ladyman explores the interaction between different forms of
realism and different forms of quantum physics, showing the tension between usual
arguments for scientific realism in the philosophy of science literature and the
invocation of realism in certain interpretations of quantum mechanics. In the second
chapter of this part, “To Be a Realist about Quantum Theory,” Hans Halvorson takes
a closer look at the distinction between realist and antirealist views of the quantum
state, and argues that this binary classification should be reconceived as a continuum
of different views about which properties of the quantum state are representationally
significant. The final chapter of the second part, “Locality and Wave Function
Realism” by Alyssa Ney, advocates for wave function realism, according to which
the fundamental quantum entity is the wave function, understood as a scalar field on a
high-dimensional space with the structure of a configuration space; according to her,
this kind of realism is an attempt to explain nonlocal influences, instead of taking
them as brute facts of the world.
In Part III, “Individuality, Distinguishability, and Locality,” the ontological
problems related to the identity and nature of quantum particles are addressed.
This part begins with the chapter by Jonas Arenhart, Otávio Bueno, and Décio
Krause, “Making Sense of Nonindividuals in Quantum Mechanics,” which focuses
on a very specific question: Assuming that quantum theories deal with “particles”
of some kind, what kind of entities can such particles be? The authors respond that
quantum entities are nonindividuals and that a metaphysics of nonindividuals
requires a system of logic where the basic items have no identity. In the second
chapter of this part, entitled “From Quantum to Classical Physics: The Role of
Distinguishability,” Ruth Kastner reviews the derivations of the classical and the
quantum statistics in order to argue that a form of separability is a key feature of
the quantum-to-classical transition; on this basis, she considers the question of
what allows separability to serve as a form of distinguishability in the classical
limit. The third chapter, “Individuality and the Account of Nonlocality: The Case
for the Particle Ontology in Quantum Physics” by Michael Esfeld, examines
different solutions to the measurement problem to conclude that the particle
ontology of Bohmian mechanics provides the least deviation from the ontology
of classical mechanics that is necessary so as to accommodate quantum physics,
4 Quantum Worlds
both in the case of quantum mechanics and in the case of quantum field theory. This
third part closes with the chapter by Alejandro Hnilo, “Beyond Loophole-Free
Experiments: A Search for Nonergodicity,” where he analyzes the experiments
designed to measure violations of Bell’s inequalities and argues that, besides locality
and realism, the measurement of the inequalities implicitly assumes the ergodic
hypothesis; therefore, in order to save the validity of local realism in nature, it is
necessary to search for evidence of nonergodic dynamics in Bell’s experiments.
The chapters composing Part IV, “Symmetries and Structure in Quantum
Mechanics,” deal with structural features of the quantum theory. The first chapter
of this part, “Spacetime Symmetries in Quantum Mechanics” by Cristian López
and Olimpia Lombardi, stresses the relevance of symmetries to interpretation; on
this basis, the authors consider the behavior of nonrelativistic quantum mechanics
under the Galilean group and critically analyze the widely-accepted view about the
invariance of the Schrödinger equation under time reversal. In the second chapter,
“Symmetry, Structure, and Emergent Subsystems,” Nathan Harshman focuses on
the particular structures called irreducible representations of symmetry groups, in
order to explore the connections between the mathematical units of symmetry
embodied by those irreducible representations and the conceptual units of reality
that form the basis for the interpretation of quantum theories. Finally, the third
chapter of this fourth part, “Majorization, across the (Quantum) Universe” by
Guido Bellomo and Gustavo Bosyk, reviews the wide applicability of majorization
in the quantum realm and stresses that such applicability emerges as a consequence
of deep connections among majorization, partially ordered probability vectors,
unitary matrices, and the probabilistic structure of quantum mechanics.
The chapters in the fifth and final part of this volume, “The Relationship
between the Quantum Ontology and the Classical World,” address the classical
limit of quantum mechanics from different perspectives. In the first chapter,
“A Closed-System Approach to Decoherence,” Sebastian Fortin and Olimpia
Lombardi argue that the conceptual difficulties of the orthodox approach to
decoherence are the result of its open-system perspective; so, they propose a
closed-system approach that not only solves or dissolves the problems of the
orthodox approach, but is also compatible with a top-down view of quantum
mechanics. In the second chapter of this part, “A Logical Approach to the
Quantum-to-Classical Transition,” Sebastian Fortin, Manuel Gadella, Federico
Holik, and Marcelo Losada present a logical approach to the emergence of
classicality based on the Heisenberg picture, which describes how the logical
structure of the elementary properties of a quantum system becomes classical
when the classical limit is reached. In the chapter, “Quantum Mechanics and
Molecular Structure: The Case of Optical Isomers,” which closes the last part
and the volume, Juan Camilo Martínez González, Jesús Jaimes Arriaga, and
Introduction 5
Sebastian Fortin address the difficulty of giving a quantum explanation of chirality,

that is, of the difference between the members of a pair of optical isomers or
enantiomers; according to them, the solution of the problem requires an interpret-
ation of quantum mechanics that conceives measurement as a breaking-symmetry
process.
As this brief review shows, the plurality of perspectives and “quantum worlds”
collected by this volume is an excellent opportunity not only to show how alive
the debate on the longstanding puzzles of quantum mechanics remains, but
also to present an updated state of affairs regarding our understanding of the
quantum reality.
Part I
Ontology from Different Interpretations of
Quantum Mechanics
1
Ontology for Relativistic Collapse Theories
wayne c. myrvold
1.1 Introduction
Dynamical collapse theories, such as the Ghirardi–Rimini–Weber (GRW) theory
(Ghirardi, Rimini, and Weber 1986), the Continuous Spontaneous Localization
theory, or CSL (Pearle 1989, Ghirardi, Pearle, and Rimini 1990), Quantum
Mechanics with Universal Position Localization, or QMUPL (Diósi 1989), and
their respective relativistic extensions (Dove 1996, Dove and Squires 1996,
Tumulka 2006, Bedingham 2011a, b, Pearle 2015), modify the usual deterministic,
unitary quantum dynamics such as to produce something like the textbook collapse
process. See Bassi and Ghirardi (2003), Bassi et al. (2013), and Ghirardi (2016) for
overviews.
If some sort of dynamical collapse theory is correct, what might the world be
like? Can a theory of that sort be a quantum state monist theory, or must such
theories supplement the quantum state ontology with additional beables? In a
previous work (Myrvold 2018), I defended quantum state monism. The view
defended involves a natural extension of the usual eigenstate-eigenvalue link,
which provides a sufficient condition for a quantum state to be one in which a
system has a definite value of some dynamical variable, namely, that the quantum
state be an eigenstate of that variable. The usual eigenstate-eigenvalue link leaves
open the question of what to say about states that are not eigenstates. A state that is
not an eigenstate of some dynamical variable, but is very close to an eigenstate,
exhibits behaviour that closely approximates that of the eigenstate. In accordance
with a proposal of Ghirardi, Grassi, and Pearle (1990), in such a case the quantity
may be treated as if it were definite. However specification of the quantities that are
definite or near-definite does not exhaustively specify the condition of the physical
world, as there are matters of fact about such things as the spread of values of a
dynamical variable in a given state. The natural ontology for a collapse theory is a
distributional ontology along the lines advocated by Philip Pearle (2009). On such
9
10 Wayne C. Myrvold
an account, dynamical quantities such as charge or mass within a specified region

do not take on precise values, but, rather, have associated with them a distribution
of values.
This chapter discusses the extension of such a picture to the context of a
relativistic spacetime. This will not be presumed to be Minkowski spacetime, as
we will want to consider curved spacetimes; furthermore, we do not wish to
exclude the possibility of discrete spacetimes. What the spacetimes we will
consider have in common is a relativistic causal structure. In particular, we will
focus attention on spacetimes in which, for any spacetime point p, there are
temporally extended processes that go on at spacelike separation from p, prohibited
by the spacetime structure from either causally influencing p or being influenced
by it.
In Section 1.2 I give a brief recapitulation of the argument, presented in more
detail in Myrvold (2018), for distributional ontology. This is based on a principle
that, I claim, ought to be respected by any project of seeking to draw ontological
conclusions from nonfundamental physical theories, a principle that I call the
Principle of Metaphysical Continuity, outlined in Section 1.2.1. This principle
permits us to draw conclusions about ontology for nonideal collapse theories – that
is, collapse theories that yield, not exact eigenstates of the dynamical quantities one
would like to be definite, but close approximations to them – from ontological
considerations regarding ideal collapse theories. Section 1.3 presents a fairly
general schema for collapse theories in a relativistic spacetime. Finally, in Section
1.4, we turn to the identification of local beables for theories of that sort.
1.2 The Case for Distributional Ontology

1.2.1 Ontology for Nonfundamental Theories
If we want to know something about the make-up of the world, we can do no better
than to look to our best scientific theories. Doing this poses a prima facie problem,
however, as there is not and never has been a convincing candidate on the table for
a complete and fundamental physical theory.
One reaction to this fact might be take on the task of giving an account of what
the world would be like if such-and-such physical theory were a complete and
fundamental theory. On this view, metaphysics is a subgenre of fiction, though
stripped of plot and character and, indeed, of everything that motivates us to read
fiction. This strikes me as an uninteresting enterprise, except insofar as consider-
ations of unrealistic theories yields insights regarding the ontology of the actual
world. For instance: Though we currently lack a theory that incorporates both
quantum and gravitational phenomena, one could, perhaps, investigate the
Ontology for Relativistic Collapse Theories 11
structure of a world in which there is no gravitation and in which the standard

model of quantum field theory is exactly correct. However such a world would be a
lonely and boring place, as it would contain no stars or planets, and, since virtually
all nuclei heavier than helium are formed in stars, would contain little in the way of
chemical reactions. We could not ask what it would be like to be a denizen of a
world like that, because a world like that would contain no life.
Another reaction might be to abandon all ontological inquiry as hopeless, on the
grounds that we can expect future theories to generate radical ontological shifts.
This strikes me as overly pessimistic. The objects dealt with in classical physics do
after all exist, even if their behaviour is not exactly what classical physics would
lead one to expect. Any theory that can lay claim to the title of a viable successor
theory to our current theories is obliged to recover the empirical successes of our
current theories, and as long as we resist the temptation to draw stronger conclu-
sions from our current physical theories than we have warrant for, there are reasons
for optimism that those conclusions will weather the storms of future theory
change.
This sort of attitude recommends due caution in our metaphysical musings. The
evidence we have concerning our physical theories warrants only the conclusions that
they hold to a good approximation within their domains of applicability, and that any
viable successor theory will have to entail something like the current theories within
those domains. If our current theories have metaphysical consequences that are
sensitive to the precise details of the theory, consequences that would not hold if
the theory were slightly different, then we have no warrant for taking those conse-
quences to hold of our world. Our metaphysical conclusions should satisfy a Principle
of Metaphysical Continuity: they should be robust under small perturbations of
theory. This is a principle that we will put to work, in Section 1.2.3.
1.2.2 The Requirement of Local Beables

Consider a region of spacetime that is bounded in both time and space, say, the
spatial region inside your office, during some specified hour of time. Of the things
that are true of that bounded spacetime region, some are local to that region: They
refer only to intrinsic properties of that region. These are to be contrasted with
things that involve relations to states of affairs outside the region, or either implicit
or explicit reference to things outside.
For example, on the usual way of thinking about things, if your office, during
the hour we are considering, contains a cabinet-shaped piece of steel, this is a local
fact about that spacetime region. If the proposition that the spacetime region under
consideration contains an object of that sort is true, its truth is compatible with
completely arbitrary states of affairs outside the region, and its truth value cannot
12 Wayne C. Myrvold
be changed by goings-on outside the region unless those goings-on have an effect
on local matters within the region. By contrast, if the contents of your office are
approximately 150 million kilometers from the nearest star, this fact is clearly a
fact about relations between the things in your office and the world outside of it.
A symptom of this fact is that it can be changed by making changes outside your
office that do not affect anything within it.
By a local beable, I will mean something that is, in this sense, local to a
bounded spacetime region. The ontology of a physical theory might contain both
local and nonlocal beables. If it is the case that, for an arbitrarily fine covering of
spacetime with open sets, the full ontology of the theory supervenes on beables
that are local to elements of that covering, we will say that the ontology satisfies
the condition of separability (see Myrvold 2011 for further discussion).
Quantum state realism entails rejection of separability. It does not follow
that there are no local beables. For one thing, there could be local beables
postulated in addition to the quantum state. Additionally, some aspects of the
quantum state – in particular, the reduced state that is the restriction of the state to
observables pertaining to a bounded spacetime region – might be counted as local
beables.
Need there be any local beables at all? If we are willing to countenance a
rejection of separability, might we not go all the way and accept a radically holistic
view in which there are no beables intrinsic to any region short of the whole of
spacetime?
The difficulty with this is that, if the theory is meant to be one that is in principle
comprehensive, it must have room for such things as experimental apparatus that is
subject to local manipulations and whose experimental readouts are, presumably,
matters of fact local to the laboratory. In the absence of things like these, the theory
runs the risk of undermining its own evidential base (see Maudlin 2007 for a lucid
discussion of this point).
A brief comment, before we continue. What it means to say that a structure
found within a physical theory plays the role of spacetime for that theory is that it
has the appropriate connections with dynamics. In speaking of spacetime, I will
always mean that structure that plays the role in the theory of affording spatio-
temporal relations, such as distances, temporal intervals, causal connectability and
the like, distances and temporal intervals and causal relations that are relevant to
the dynamics. It is necessary to say this because it has been claimed that quantum
theory motivates the introduction of a so-called fundamental space, or fundamental
arena, a high-dimensional space that would be such that quantum states involve
nothing more than assignments of local beables to points in that fundamental space
(see Albert 1996 and the various contributions to Ney and Albert 2013). In a
quantum theory, even if such a space can be found, that space is not the structure
on which the distances, temporal intervals, and causal relations relevant to the
dynamics are defined. For that reason, such a space, even if it were to exist, is not
spacetime in the sense of the word being used in this chapter. Hence, even if a
fundamental space of the sort sought by Albert and others did exist, a quantum
state realist ontology violates separability as we are using the term.
1.2.3 Ontology for Ideal Collapse Theories

According to the textbook collapse postulate, after an experiment the quantum
state of the system subjected to the experiment is an eigenstate of the observable
whose value has been obtained. Naively, one might expect a dynamical collapse
theory to be like that. There are good reasons for thinking that this is an unattain-
able goal. If, however, we could have a theory like that – a theory that yielded
eigenstates of an appropriate dynamical variable – then, I claim, there would be no
problem of ontology for the theory once we have settled on a choice of dynamical
variable to collapse to eigenstates of (a suitable choice seems to be that of a
smeared mass density, as advocated by Ghirardi, Grassi, and Benatti 1995). That
there is any question about the ontology for a collapse theory is an aspect of what
has been called the tails problem (first flagged as an issue by Shimony 1991 and by
Albert and Loewer 1991), which stems from the fact that collapse theories do not
lead to eigenstates of familiar dynamical quantities.
Consider a quantum theory on a discrete spacetime, one on which space
consists of elementary cells of size vastly smaller than the scales on which we
deal with things. Suppose we had a collapse theory that tended to suppress
superpositions of distinct mass densities smeared over regions (which could
consist of a great many of these elementary cells) of order 105 cm, small on a
human scale, but large compared to atomic dimensions. Suppose that our collapse
theory induced collapse, within a finite time, to eigenstates of the operators
corresponding to total mass within regions of this size, and that states that are
not eigenstates of these mass operators could persist only for a minuscule fraction
of a second.
On such a theory, for every region of space of sufficient size, the quantum state
would, most of the time, be an eigenstate corresponding to a definite mass within
that region. Hence, by the eigenstate-eigenvalue link, there would be a matter of
fact about the amount of mass within that region. Thus, a possible state of the room
in which I am sitting would be one in which there was a desk-shaped region of
higher mass density than its surroundings. Provided that these regions of high mass
density exhibited the right sort of dynamical behaviour, there would be no problem
in identifying them with desks, chairs, and laboratory equipment, and there would
be no problem of ontology for collapse theories.
14 Wayne C. Myrvold
1.2.4 Distributional Ontology

Prospects are dim for a viable collapse theory that yields precise eigenstates of total
mass in any bounded region, or indeed, precise eigenstates of any local beables.
A collapse theory can, however, yield close approximations to eigenstates of
appropriate local beables, such as mass smeared over sufficiently large regions.
Whether the dynamics is linear, unitary, and deterministic, as in the Schrödinger
equation, or nonunitary and stochastic, initial states that are close to each other, in
Hilbert space norm, evolve in approximately the same way. Thus, a state that is
close to being an eigenstate of a given dynamical quantity will evolve in approxi-
mately the same way as the eigenstate that it is close to.
If we accept (as we should; see Myrvold 2018 for a fuller discussion, also Albert
2015: 127ff.) that to be a physical body is nothing more and nothing less than to
have a certain place in a network of dynamical and causal relations of an appropri-
ate sort, and if we accept (as we should) that there would be no problem of
interpretation of an ideal collapse theory that yielded eigenstates of the right sort
of dynamical quantities, then, by the Principle of Metaphysical Continuity, we
should accept that regions of space whose states are very near to eigenstates of total
mass can serve as physical objects just as well as would regions of space in exact
eigenstates of total mass.
Considerations such as this have led to a proposed modification of the
eigenstate-eigenvalue link, as follows:
if one wishes to attribute objective properties to individual systems one has to accept that
such an attribution is legitimate even when the mean value of the projection operator on the
eigenmanifold associated to the eigenvalue corresponding to the attributed property is not
exactly equal to 1, but is extremely close to it.
(Ghirardi, Grassi, and Pearle 1990: 1298)
This modification has been dubbed, by Clifton and Monton (1999), the fuzzy
link.
To say that we can ascribe a property to a system when the quantum state is such
that its variance is negligibly small requires that there be a matter of fact about
what the variance is. Considerations of this sort suggest a revision of the way we
think about dynamical quantities, along the lines advanced by Pearle (2009). On
this view, dynamical variables typically do not take on sharp values as they would
classically. What they have, instead, is a distribution associated with them. These
distributions, though having the formal characteristics of probability distributions,
are to be thought of not as a probability distribution over precise but unknown
possessed values, but as reflecting a physical, ontological lack of determinacy
about what the value is. A limiting case would be the classical case, in which the
distribution is a delta function.
On this view, the value of every dynamical variable is distributional. A collapse

theory will tend to narrow the spread of the distributions of some of these
quantities. When the distribution is sufficiently narrow, things will be almost
exactly as if the quantity has a precise value, and under such circumstances, we
can treat the variable as if it does possess a precise value. In seeking objects that
behave like our familiar macroscopic objects, it is to those variables that we should
direct our attention. But the spread-out distributions of the other variables are no
less part of physical reality.
1.2.5 Primitive Ontology as an Alternative?

Given a family of operators M ^ ðxÞ, corresponding to a smeared mass density
centered at the point x, for any quantum state ψ, one can define a function mðxÞ,
whose value at the point x is equal to the expectation value of M ^ ðxÞ in state ψ.
When Ghirardi et al. (1995) introduced the smeared mass density as a basis for the
ontology of collapse theories, their proposal was an application of the fuzzy
eigenstate-eigenvalue link. They argued that the quantity mðxÞ behaves like a mass
density when – and only when – the variance of M ^ ðxÞ is sufficiently small as to be
negligible, in which case the mass density is said to be objective. When this
condition is not satisfied, the quantity mðxÞ, though well defined, cannot be
interpreted as a mass density, as other systems do not behave as if a quantity of
mass corresponding to mðxÞ is present. In some later works (Ghirardi and Grassi
1996, Ghirardi 1997a, b) the mass density is said to be ‘accessible’ if its variance is
sufficiently small (this shift is attributed by Ghirardi and Grassi 1996: fn. 5 to a
conversation with S. Goldstein).
There is, at least apparently, a rival interpretation of mðxÞ. On this view,
introduced by Goldstein (1998) and discussed extensively by Allori et al. (2008),
a mass density equal at every point to the expectation value of M ^ ðxÞ is posited as
additional, primitive ontology over and above the quantum state.
The quantities mðxÞ are well defined for any quantum state. However, since in
situations in which the objectivity, or accessibility, condition is not satisfied – that
is, situations in which the variance of M ^ ðxÞ is not small enough to be neglected –
other objects do not respond as if a mass density equal to mðxÞ is present, mðxÞ acts
like a mass density only when the accessibility condition is satisfied. Something
that does not act like a mass density is not a mass density. Thus, on the supposed
rival interpretation, despite what is said, a mass density is present only when there
is a mass density present on the original, quantum state monist proposal, that is,
when the accessibility condition is satisfied. The proposal to take mðxÞ as add-
itional, primitive ontology does not present a genuine alternative to the original
proposal of Ghirardi et al. (1995).
16 Wayne C. Myrvold
1.3 A Schema for Relativistic Collapse Theories

1.3.1 Relativistic Spacetimes
We assume a spacetime equipped with a causal order, that is, a relation of causal
precedence, assumed to be transitive and antisymmetric (that is, if p causally
precedes q, then q does not causally precede p). Two spacetime points are said
to be causally unconnected if they stand in no causal order, that is, if neither p q
nor q p obtains. Because no point is in the causal past of itself, the relation of
being causally unconnected is reflexive. That it is symmetric follows straightfor-
wardly from its definition. Two distinct points that are causally unconnected are
said to be spacelike separated.
In Galilean spacetime, the relation of being causally unconnected is transi-
tive, and therefore, is an equivalence relation, and the spacetime can be
portioned into equivalence classes of simultaneity. In Minkowski spacetime,
on the other hand, for any two points p, q that are spacelike separated from
each other, there are other points r that are spacelike separated from p but not
from q. Define a relativistic spacetime as one in which, for any spacelike
separated p, q, there exists a point r that is spacelike separated from p, such
that r q.
A causal curve is a curve such that, for any pair of distinct points p, q, either
p q or q p. A Cauchy surface is a set of spacetime points that is intersected
exactly once by every inextendible causal curve. A spacetime that contains Cauchy
surfaces is said to be globally hyperbolic.
We will, in what follows, assume a globally hyperbolic relativistic spacetime.
We can define the relation between Cauchy surfaces: If σ, σ0 are two Cauchy
surfaces, then σ σ0 when no part of one if no part of σ0 is in the causal past of σ.
This relation is reflexive and transitive, and hence is a partial order on Cauchy
surfaces.
1.3.2 Collapse Theories in Relativistic Spacetime

A collapse theory modifies the deterministic, unitary evolution so as to produce
something like the textbook collapse. Gisin (1989) has demonstrated, on the
assumption that the evolution is Markovian (meaning that future states depend
only on the present state and not on any details about the past that are not
reflected in the present), that any deterministic, nonlinear dynamics for quantum
states that does not respect a certain linearity condition permits signalling – if two
spatially separated systems are in an entangled state, a choice of experiment on
one can influence probabilities of outcomes of experiments performed on
the other.
The relevant condition is the following:

Linearity. Let T be a dynamical map on the set of pure states of a system. Let
fψ i ; i ¼ 1; . . . ; ng and φj ; j ¼ 1; . . . ; m be sets of pure states such that, for some
non-negative weights {xi}, {yj},
X
n X
m
xi ψ i ¼ yj φj :
i¼1 j¼1
Then
X
n X
m
xi T ðψ i Þ ¼ yj T φj :
i¼1 j¼1
As Kent (2005) has argued, although violations of linearity permit signalling,

this need not be superluminal signalling. Nonetheless, ordinary quantum mech-
anics does not allow operations on one system to be used for signalling to another
system, unless there is an interaction term between the two systems in the
Hamiltonian, and we will assume that the no-signalling condition holds, and
hence that the evolution is linear. This means that (unsurprisingly) a theory that
produces collapse must be a theory with indeterministic dynamics.
It is convenient to work within what may be called the stochastic Tomonaga–
Schwinger picture. The usual Tomonaga–Schwinger picture (see Schweber
1961: 419–422 for an introduction) is an extension of the interaction picture
to a relativistic spacetime. One divides the Lagrangian density into two parts
(typically regarded as the free Lagrangian density and the interaction Lagrangian),
L ðxÞ ¼ L 0 ðxÞ þ L 1 ðxÞ: (1.1)
The operators representing observables are Heisenberg-picture operators for the
free theory. We utilize, however, evolving state vectors; with each Cauchy surface
σ is associated a state vector jψ ðσ Þi. Evolution from a surface σ to another, σ0 ,
differing by a small deformation δσ about a point x, satisfies the Tomonaga-
Schwinger equation:
δjψ ðσÞi
iℏc ¼ H 1 ðxÞjψ ðσÞi: (1.2)
δσ
Integration of this equation yields, for any Cauchy surfaces σ and σ0 , a unitary
mapping from jψ ðσ Þi to jψ ðσ 0 Þi.
We wish to modify this equation so as to produce collapse. On the stochastic
Tomonaga–Schwinger picture, we work with Heisenberg-picture operators that are
solutions to the standard field-theoretic equations, for free or interacting fields. The
18 Wayne C. Myrvold
difference between states on different Cauchy surfaces is due to the stochastic

modifications to the usual evolution. We will assume that the new dynamics is
Markovian: that is, that if σ σ0 , the set of possible states on σ0 and their
respective probabilities are determined by jψ ðσ Þi, and not by other facts about
the history leading up to that state.
Given two Cauchy surfaces, σ, σ0 , with σ σ0 , and a state vector jψ ðσ Þi, there
will be some state vector jψ ðσ 0 Þi, but what this vector will be is not determined by
jψ ðσ Þi and the dynamics. Instead, there will be some set of alternatives
n o
ψ ðσ 0 Þ , which we take to be indexed by a parameter γ that takes on values
γ
in a set Γ. We expect our theories to specify, given two surfaces σ, σ0 , with σ σ0 ,

n o
and the state vector jψ ðσ Þi, the set of alternatives ψ ðσ 0 Þ ; γ 2 Γ , and a
γ
probability distribution over the possible values of γ. Suppose that, with respect
to some background measure μ, this probability distribution is represented by a
density function pðγÞ.
With this apparatus in place, we can define a mixed state ρ ðσ0 ; σÞ as the
weighted average over the possibilities for the state on σ0 , given jψ ðσ Þi.
ð

ρ ðσ ; σÞ ¼ pðγÞ ψγ ðσ0 Þ ψγ ðσ0 Þ dμ:
0
(1.3)
γ
This would be the state used by someone who knows the state on σ and
the possible state transitions from jψ ðσ Þi to jψ ðσ 0 Þi and their respective
probabilities, but does not know the outcome of the process that occurs between
σ and σ0 .
Gisin’s proof, mentioned earlier, generalizes to stochastic theories. If we take
T to be the mapping that takes a pure state on σ to a mixed state ρ ðσ0 ; σÞ, no-
signalling entails that this map must satisfy the linearity condition (Simon,
Bužek, and Gisin 2001, Bassi and Hejazi 2015), and from this, together with
the condition that, applied to subsystems in entangled states, the mapping
extends to a positive map on the state space of the wider system, it entails that
the map from the state on σ to the mixed state ρ ðσ0 ; σÞ be a completely
positive map.
We will therefore take the mapping from a pure state on σ to the mixed state
ρ ðσ0 ; σÞ to be a nonselective completely positive map, which is a mixture of
0
selective completely positive maps that takes us from jψ ðσ Þi to ψ γ ðσ Þ . This
entails that there is a set of operators K γ ; γ 2 Γ , which we will call evolution
operators, such that, for some γ,

jψ ðσ 0 Þi ¼ ψ γ ðσ 0 Þ ¼ K γ jψ ðσ Þi= K γ jψ ðσ Þi , (1.4)
with probabilities for which state is realized given by

ð
Prðγ 2 ΔÞ ¼ pðγÞ dμ: (1.5)
Δ
The linearity condition entails that

2
pðγÞ ¼ K γ jψðσÞi : (1.6)
Any probabilities that depart from these would lead to signalling. The condition
that p always be normalized is the condition that
ð
K †γ K γ dμ ¼ 1: (1.7)
Γ
The evolutions should also satisfy the semi-group property, which requires that, for
Cauchy surfaces σ σ0 σ00 , the possible evolutions from σ to σ00 be the compos-
itions of evolutions from σ to σ0 with evolutions from σ0 to σ00 .
The theory of quantum dynamical semi-groups is well studied (see Bassi and
Ghirardi 2003 or Alicki and Lendi 2007 for an introduction). Provided that the
evolution satisfies an appropriate continuity condition, the mixed-state density oper-
ators on Cauchy surfaces σ to the future of some surface σ0 will satisfy a Lindblad
equation. We consider the change in ρ ðσ; σ0 Þ as we pass from one surface σ to
another σ0 differing by a small deformation about a point x on σ with spacetime
volume δσ. Let H ðxÞ be the Hamiltonian density, that is, the component of the
energy-momentum density along the normal to σ at x. For a Lindblad-type evolution,
there is also a countable set fLα ðxÞg of operators, such that the change δρ satisfies
!
δρ 1 X 1 X † X
† †
¼ ½H ðxÞ; ρ þ Lα ðxÞρ Lα ðxÞ L ðxÞLα ðxÞρ þ ρ Lα ðxÞLα ðxÞ :
δσ iℏc α
2 α α α
(1.8)
Consider two Cauchy surfaces σ, σ0 , with σ σ0 , that coincide everywhere

except on the boundaries of two bounded regions δ and δ0 (see Figure 1.1). The
evolution from σ to σ0 through δ [ δ0 must equal the composition of the evolution
through δ and the evolution through δ0 , in either order. The necessary and sufficient
s¢
d d¢
s
0 0
Figure 1.1 Cauchy surfaces σ, σ , with σ σ , that coincide everywhere except on
the boundaries of two bounded regions δ and δ0 .
20 Wayne C. Myrvold
condition for this is that evolution operators corresponding to spacelike separated

regions commute.
Moreover, for computing probabilities for the results of experiments on the
overlap of σ and σ0 , it should not matter whether jψ ðσ Þi or ρ ðσ0 ; σÞ is used.
Someone located in the overlap, who knows the state on σ and believes collapse
will occur between σ and σ0 , but does not know (because it occurs at spacelike
separation) what the outcome of that collapse is, should be able to use jψðσÞi or
ρ ðσ0 ; σÞ for computing probabilities of results of experiments that he or she is about
to undertake, and these should yield the same probabilities for outcomes of those
experiments. The necessary and sufficient condition for this is that evolution
operators that implement evolution through a given spacetime region δ should
commute with operators representing observables at spacelike separation from δ.
These conditions give us a rather general schema for a quantum theory with
stochastic dynamics on a relativistic spacetime. It includes, as a special case,
deterministic, unitary evolution, in which case the set of evolution operators
pertaining to any region of spacetime is a singleton set. Concrete theories will fill
in the details, specifying in particular what the sets of evolution operators are.
1.4 Beables for Relativistic Collapse Theories

1.4.1 Intrinsic and Extrinsic States of a Spacetime Region
Consider a bounded spacelike region α, that is common to Cauchy surfaces
fσ; σ0 ; σ00 ; . . .g. In the stochastic Tomonaga–Schwinger picture, there will be
quantum states ρðσÞ, ρðσ0 Þ, ρðσ00 Þ, . . . Each of these states yields probabilities of
outcomes of experiments to the future of its Cauchy surface, conditional on events,
including any collapses, to the past of the Cauchy surface. For each of these states,
we can consider the reduced state that consists of the restriction of the state to
observables in the forward domain of dependence of α. Call these reduced states
ρα ðσÞ, ρα ðσ0 Þ, etc. If the evolution between two surfaces σ and σ0 is purely unitary,
then ρα ðσÞ will coincide with ρα ðσ0 Þ. If, however, collapse occurs between σ and
σ0 , then they need not coincide (see Figure 1.2).
Since the reduced state ρα ðσÞ is conditioned on any collapses to the past of σ,
including any that are spacelike separated from α, it should be clear that, though it
is associated with the region α, ρα ðσÞ cannot in general be regarded as a beable
local to α. Thus, if the reduced states ρα ðσÞ and ρα ðσ0 Þ differ, they do not offer
competing accounts of intrinsic properties of the region α.
The intrinsic state of a bounded spacelike region α must be conditioned only on
collapses to the past of α. We can define this state by a limiting procedure.
Consider a sequence of Cauchy surfaces fσ1 ; σ2 ; . . . ; σn ; . . .g, that is such that α
s²
s¢
s
a
s²
Figure 1.2 Bounded spacelike region α, that is common to Cauchy surfaces

σ, σ0 , σ00 .
is contained as a common part of all σn and for all n, σnþ1 σn , and the sequence
converges on the past light cone of α (that is, the set of points that are to the past of
all σn is precisely the causal past of α). Define the past light-cone state of α as the
limit, if it exists, of ρα ðσn Þ, as n increases indefinitely. Though a state derived from
a Cauchy surface with events to its past that are spacelike separated from α cannot
be regarded as the intrinsic state of α, its past light-cone state can.
1.4.2 Compatibility of Extrinsic States

Maudlin (1996: 301–302) raised the question of consistency of state assign-
ments derived from different hypersurfaces passing through a given region. If
two hypersurfaces σ, σ0 , having a region α in common, yielded reduced states
that were orthogonal to each other, yielding conflicting definite (probability
equal to unity) predictions for the outcome of some experiment, this would be
problematic.
The question arises: Do the conditions on collapse dynamics outlined previously
guarantee that the differing extrinsic state assignments obtained from different
Cauchy surfaces are not in outright conflict with each other? It can be shown (see
Myrvold 2003: 489, 2016: 255–257) that these conditions suffice to guarantee that
the states ρα ðσÞ and ρα ðσ0 Þ are not orthogonal.
In fact, a stronger sense of compatibility obtains. The question of the compati-
bility of reduced states derived from states on different Cauchy surfaces is essen-
tially the same as that addressed by Brun, Finkelstein, and Mermin (2002). They
demonstrate that state assignments that can represent information about a system
available to different observers are compatible, in the sense that they have overlap-
ping support.
The context in which Brun et al. work is that of finite-dimensional Hilbert
spaces. However, essentially the same conclusion holds in a setting appropriate to
quantum field theory. In this context we cannot assume a finite-dimensional Hilbert
space, nor can we assume that the mixed state of a bounded region α obtained from
a pure state on a Cauchy surface containing α admits of a decomposition into pure
22 Wayne C. Myrvold
states. We must take care to formulate the condition in a manner that is independ-
ent of assumptions such as these.
We assume a von Neumann algebra RðαÞ, whose self-adjoint elements represent
the bounded observables pertaining to the forward domain of dependence of α. Let
ρ be a normal state of RðαÞ (that is, a completely additive state). We define the
support projection for ρ as the orthogonal complement of the union of all projec-
tions in RðαÞ to which ρ assigns expectation value zero.
With these definitions in hand, it can be shown that, given a set fσ; σ0 ; σ00 ; . . .g
of Cauchy surfaces containing α, then on the assumption that there is a Cauchy
surface containing α that is nowhere to the past of any of them (which, in
particular, will always be the case for any finite set of Cauchy surfaces), the
corresponding set of states fρα ðσÞ; ρα ðσ0 Þ; ρα ðσ00 Þ; . . .g have nonzero overlapping
support. See Appendix for details.
1.4.3 Local Beables for Collapse Theories

Suppose that we have a collapse theory that yields near-eigenstates of an appropri-
ate dynamical quantity. For example: The natural extension to the relativistic
context of a mass density would be the components of the stress-energy tensor.
Assume that we have an appropriate relativistically invariant smearing function
(see Bedingham 2011a, b) and formulate smeared operators T^ μν ðxÞ, representing a
smeared stress-energy density centered at the point x. For any state ρ, we can define

T μν ðxÞ ¼ T^ μν ðxÞ ρ : (1.9)
The 00-component of this is the relativistic analogue of the mass density that has
been proposed as an appropriate ontology for nonrelativistic collapse theories.
For a bounded region α contained in distinct Cauchy surfaces σ and σ0 , the
reduced states ρα ðσÞ and ρα ðσ0 Þ may yield differing values for T μν ðxÞ, with x
within α. But, obviously, these do not yield rival accounts of local beables within
α, as they are defined via the extrinsic states ρα ðσÞ and ρα ðσ0 Þ, which are not
themselves local beables.
Where then, may we find local beables for a relativistic collapse theory? There
are, in the literature, two proposals for extending the fuzzy link to a relativistic
context. One is what might be called the agreement criterion, formulated by
Ghirardi, Grassi, and Pearle (1991):
We think that the appropriate attitude is the following: when considering a local observable
A with its associated support we say that an individual system has the objective property a
(a being an eigenvalue of A), only when the mean value of Pa is extremely close to one,
when evaluated on all spacelike hypersurfaces containing the support of A.
(Ghirardi et al. 1991: 1310).
This means that, for x within α, T μν ðxÞ will be regarded as representing an

objective property if and only if the accessibility criterion is satisfied in ρα ðσÞ for
every Cauchy surface containing α.
The other criterion is the past light cone criterion, formulated by Ghirardi and
Grassi (1994: 419, see also Ghirardi 1996: 336, 1999: 139, 2000: 1364). On this
criterion, a system is said to possess the property A = a when the expectation value
of Pa is extremely close to one, evaluated on the past light-cone state.
If the criterion for property attribution were an exact eigenstate-eigenvalue
link – that is, if we were ascribing a property A = a only when the expectation
value of Pa is exactly equal to one – then the two would be equivalent. The past
light-cone state of a region α is an eigenstate of an observable A pertaining to α,
with eigenvalue a, if and only if the state on every Cauchy surface containing α is.
On the fuzzy link, the agreement criterion entails the past light-cone criterion, but
the past light-cone criterion does not guarantee satisfaction of the agreement
criterion; it only entails that the agreement criterion will hold with high probability.
If the property attribution criterion is meant to supply local beables, then it is
clear that what is wanted is the past light-cone criterion and not the agreement
criterion. The agreement criterion makes reference to events at spacelike separation
from the region in question. Moreover, as Ghirardi, Grassi, Butterfield, and
Fleming (1993: 358) have shown, a choice regarding experiments performed at
spacelike separation from α can affect the probability that the agreement criterion is
satisfied. Consider a case of two spin-½ particles, located in world-tubes A and B.
We take initial conditions on a Cauchy surface σ0 , and suppose that the particle in
A is undisturbed in the interval between σ0 and some later Cauchy surface σ1 . Let α
be a spacelike slice of A between σ0 and σ1 . Let β0 , β1 be the intersections of B with
σ0 and σ1 , respectively. We will take σ0 and σ1 such that β0 and β1 are to the past
and future, respectively, of α (see Figure 1.3).
Suppose that the state of the pair of particles on σ0 is
pffiffiffiffiffiffiffiffiffiffiffi pffiffi
jψðσ0 Þi ¼ 1 ε jþiA jþiB þ ε jiA jiB , (1.10)
where jþi and ji are spin eigenstates in some designated direction (say, the z-
direction), and ε is an extremely small number, small enough that the state jψðσ0 Þi
is sufficiently close to an eigenstate of spin for the particle in A that the accessibil-
ity criterion is satisfied. Thus, on the past light-cone criterion, we ascribe + spin in
the z-direction to the particle in A as a possessed property.
Suppose that Bob, located near B, has a choice of whether to perform a spin
experiment on B. Suppose that, if he does not, the pair of spin-½ systems is
effectively isolated from outside interference and that in that case our collapse
theory assigns, for some δ smaller than ε, probability 1 δ that the state will
remain undisturbed in the interval between σ0 and σ1 , in which case the agreement
24 Wayne C. Myrvold
Figure 1.3 Two spin-½ particles, located in world-tubes A and B. The initial
conditions are taken on a Cauchy surface σ0 , and the particle in A is undisturbed
in the interval between σ0 and some later Cauchy surface σ1 . α is a spacelike slice
of A between σ0 and σ1 . β0 , β1 are the intersections of B with σ0 and σ1 ,
respectively, such that they are to the past and future, respectively, of α.
criterion for property attribution is satisfied. If Bob chooses to do a spin experiment

on the particle in B, there is probability ε that he will obtain the result ji. If he
does, then the state of the combined system on a Cauchy surface σ2 that includes α
and runs to the future of Bob’s experiment will not be a state in which the particle
in A is close to a + eigenstate for spin-z; on the contrary, it will close to a –
eigenstate for spin-z. In such an eventuality, the agreement criterion for ascribing
“spin-z = +” to the particle in A is not satisfied.
Now, if the threshold for satisfaction of the accessibility criterion is stringent
enough – say, 10–40, as suggested by Pearle (1997) – then the probability of
disagreement between the past light-cone criterion and the agreement criterion is
sufficiently low as to be negligible, whether or not Bob chooses to do an experi-
ment. However, it is still true that the value of this negligibly low probability
depends on Bob’s choice regarding his experiment, and hence, if we were to apply
the agreement criterion for outcome attribution, this would require acceptance that
the theory exhibits parameter dependence, albeit a very weak parameter depend-
ence (see Ghirardi et al. 1993 for discussion). If, however, we adopt the past light-
cone criterion, then (as noted already by Ghirardi and Grassi 1994), there is no
parameter dependence at all, not even very weak dependence. The conclusion to be
drawn is that local beables for a relativistic collapse theory are to be identified
according to the past light-cone criterion.
As mentioned previously, Ghirardi et al. (1995) have argued that a theory on

which, at the macroscopic scales, a smeared mass density is almost always near-
definite yields an adequate picture of the world. Combined with the past light-cone
criterion, this gives a past light-cone matter density ontology, which is discussed
by Tumulka (2007) and in more detail, by Bedingham et al. (2014).
1.5 Conclusion
There is a sensible ontology for collapse theories in a relativistic context. More-
over, considerations of what it takes for a theory to represent a world that contains,
among other things, objects like our experimental apparatus, to be thought of a
local beables, determine the form that this ontology takes. It is one on which all
dynamical quantities are distributional in character. In spite of this distributional
character, dynamical quantities may have effectively precise values (in the sense
that they behave, to a high degree of approximation, as if they have precise values);
it is the goal of a collapse theory to ensure that the properties of macroscopic
objects almost always have this character. Beables local to a bounded spacetime
region are to be evaluated via the past light-cone state of that region.
Appendix
We consider a finite set of Cauchy surfaces fσ1 ; σ2 ; . . . ; σn g, all containing an open

subset α. Our goal is to show that, given the conditions on a relativistic collapse
theory, the states ρα ðσ1 Þ, ρα ðσ2 Þ, . . . , ρα ðσn Þ have common support.
We assume a Hilbert space that contains vectors jψ ðσ 1 Þi, jψ ðσ 2 Þi, . . . , jψ ðσ n Þi.
We assume also that, if σ σ0 , there exists K γ such that

jψ ðσ 0 Þi ¼ K γ jψ ðσ Þi= K γ jψ ðσ Þi : (A1.1)
If σ σ , and α is in the overlap of σ and σ , then the region between σ and σ0 is
0 0
spacelike separated from α. Therefore, K γ commutes with all self-adjoint elements

of RðαÞ.
The restrictions of the states on the Cauchy surfaces σ1 , σ2 , . . . , σn are states
(which will typically be mixed states) of RðαÞ, ρα ðσ1 Þ, ρα ðσ2 Þ, . . . , ρα ðσn Þ. We do
not assume that these are represented by density operators in RðαÞ or that they are
mixtures of pure states of RðαÞ, as this is not needed in what follows.
As mentioned in the chapter text, the projector onto the null space of any state
ρ of RðαÞ is the union of all projections P that have zero expectation value in ρ,
and the support projection of ρ is the orthogonal complement of the projector
onto the null space. We will call the null space and the support of ρ, Null½ρ,
and Supp½ρ.
Lemma 1. Let σ and σ0 be Cauchy surfaces containing a common open subset α. If
σ σ0 , then for any positive operator E 2 RðαÞ, if hψ ðσ ÞjE jψ ðσ Þi ¼ 0, then
hψ ðσ 0 ÞjE jψ ðσ 0 Þi ¼ 0.
Proof. Any positive operator E has a square root E 1=2 . Suppose that
hψ ðσ ÞjEjψ ðσ Þi ¼ 0: (A1.2)
Therefore, since
2
hψ ðσ ÞjEjψ ðσ Þi ¼ E1=2 jψ ðσ Þi , (A1.3)
26
then
E1=2 jψ ðσ Þi ¼ 0: (A1.4)
For some K γ that commutes with all self-adjoint E 2 RðαÞ,

jψ ðσ 0 Þi ¼ K γ jψ ðσ Þi= K γ jψ ðσ Þi : (A1.5)
Therefore,

E1=2 jψ ðσ 0 Þi ¼ E1=2 K γ jψ ðσ Þi= K γ jψ ðσ Þi ¼ K γ E1=2 jψ ðσ Þi= K γ jψ ðσ Þi ¼ 0 (A1.6)
and so
hψ ðσ 0 ÞjE jψ ðσ 0 Þi ¼ 0 (A1.7)
□.
Lemma 1 gives us a relation between the supports of ρα ðσÞ and ρα ðσ0 Þ when
σ σ0 .
Proposition 1. Let σ and σ0 be Cauchy surfaces containing a common
open subset α. If σ σ0 , then Null½ρα ðσÞ Null½ρα ðσ0 Þ; equivalently,
Supp½ρα ðσ0 Þ Supp½ρα ðσÞ.
Proof. This is immediate from Lemma 1.

From this follows the result concerning overlapping support of
fρα ðσ 1 Þ; ρα ðσ 2 Þ; . . . ; ρα ðσ n Þg.
Proposition 2. Let σ1 , σ2 , . . . , σn be Cauchy surfaces sharing a common open
subset α. Then the supports of ρα ðσ1 Þ, ρα ðσ2 Þ, . . . , ρα ðσn Þ have nonzero intersection.
Proof. We can construct a Cauchy surface σþ that contains α and is such that
σi σþ for each i, by taking the least upper bound of the set fσ1 ; σ2 ; . . . ; σn g under
the ordering . Consider, now, the state ρα ðσþ Þ. By Proposition 1,
Supp½ρα ðσþ Þ Supp½ρα ðσi Þ for all i, and hence the support of σþ lies within the
intersection of the supports of ρα ðσ1 Þ, ρα ðσ2 Þ, . . . , ρα ðσn Þ.
Remark I. The restriction to a finite set is unnecessary; the result holds for an infinite
set of Cauchy surfaces provided that there is a Cauchy surface that is the upper
bound of all of them. Furthermore, if there is a future light-cone state that is the limit
of an increasing (in the ordering ) set of Cauchy surfaces that converges on the
future light cone of α, then the support of the future light-cone states is the intersec-
tion of the supports of all ρα ðσÞ for all σ containing α.
Remark II. A quantum field is an assignment of a field operator φ ^ ðxÞ to each point of
spacetime. In standard quantum field theories on Minkowksi spacetime, it is assumed
that there is a unitary representation of the group of spacetime translations, with
infinitesimal generators Pμ that satisfy the spectrum condition:
For any future-directed timelike vector a, the spectrum of Pa is in Rþ
This ensures positivity of the energy, with respect to any reference frame.
28 Wayne C. Myrvold
We assume a unique vacuum state that is invariant under all spacetime symmet-
ries. Define the standard Hilbert space of the theory as the closure in norm of the
set of all vectors that can be obtained by operating on the vacuum state with
operators constructed from standard fields. It follows from the Reeh–Schlieder
theorem that, for any state ρ that is analytic in energy, for any α that is such that the
set of points spacelike separated from α contains an open set, the null space of ρ is
empty. If ρα ðσ1 Þ, ρα ðσ2 Þ, . . . , ρα ðσn Þ are all states in the standard Hilbert space of
the theory that are analytic in the energy, each of their null spaces consists solely of
the zero vector, and hence Proposition 2 holds trivially.
The proposition is less trivial for theories that introduce nonstandard fields and
whose states go beyond the standard Hilbert space, as do the relativistic versions of
CSL due to Bedingham (2011a, b) and Pearle (2015). It can be shown, for any
indeterministic theory formulated within the framework sketched in Section 1.3
that is set in Minkowski spacetime, it is necessary to go beyond the standard
Hilbert space (see Myrvold 2017).
Acknowledgments
Many thanks are due to the organizers of the workshop Identity, indistinguish-
ability and non-locality in quantum physics (Buenos Aires, June 2017) and to the
participants in that workshop, for their helpful comments. I would also like to
thank Philip Pearle for comments and advice. I am grateful to Graham and Gale
Wright, who generously sponsor the Graham and Gale Wright Distinguished
Scholar Award at the University of Western Ontario, for financial support of
this work.
References
Albert, D. Z. (1996). “Elementary quantum metaphysics,” pp. 277–284 in J. T. Cushing,
A. Fine, and S. Goldstein (eds.), Bohmian Mechanics and Quantum Mechanics: An
Appraisal. Dordrecht: Kluwer Academic Publishers.
Albert, D. Z. (2015). After Physics. Cambridge, MA: Harvard University Press.
Albert, D. Z. and Loewer, B. (1991). “Wanted dead or alive: Two attempts to solve
Schrödinger’s paradox,” pp. 278–285 in A. Fine, M. Forbes, and L. Wessels (eds.),
PSA 1990: Proceedings of the 1990 Biennial Meeting of the Philosophy of Science
Association, Volume One: Contributed Papers. East Lansing, MI: Philosophy of
Science Association.
Alicki, R. and Lendi, K. (2007). Quantum Dynamical Semigroups and Applications, 2nd
ed. Berlin: Springer.
Allori, V., Goldstein, S., Tumulka, R., and Zanghì, N. (2008). “On the common structure
of Bohmian mechanics and the Ghirardi-Rimini-Weber theory,” The British Journal
for the Philosophy of Science, 59: 353–389.
Bassi, A. and Ghirardi, G. C. (2003). “Dynamical reduction models,” Physics Reports,

379: 257–426.
Bassi, A. and Hejazi, K. (2015). “No-faster-than-light-signaling implies linear evolution.
A re-derivation.” European Journal of Physics, 36: 055027.
Bassi, A., Lochan, K., Satin, S., Singh, T., and Ulbricht, H. (2013). “Models of wave-
function collapse, underlying theories, and experimental tests,” Reviews of Modern
Physics, 85: 471–527.
Bedingham, D. (2011a). “Relativistic state reduction model,” Journal of Physics: Confer-
ence Series, 306: 012034.
Bedingham, D. (2011b). “Relativistic state reduction dynamics,” Foundations of Physics,
41: 686–704.
Bedingham, D., Dürr, D., Ghirardi, G. C., Goldstein, S., Tumulka, R., and Zanghì, N.
(2014). “Matter density and relativistic models of wave function collapse,” Journal of
Statistical Physics, 154: 623–631.
Brun, T., Finkelstein, J., and Mermin, N. D. (2002). “How much state assignments can
differ,” Physical Review A, 65: 032315.
Clifton, R. and Monton, B. (1999). “Losing your marbles in wavefunction collapse
theories,” The British Journal for the Philosophy of Science, 50: 697–717.
Diósi, L. (1989). “Models for universal reduction of macroscopic quantum fluctuations,”
Physical Review A, 40: 1165–1174.
Dove, C. (1996). Explicit Wavefunction Collapse and Quantum Measurement. PhD thesis.
Durham: Department of Mathematical Sciences, University of Durham.
Dove, C. and Squires, E. J. (1996). “A local model of explicit wavefunction collapse,”
arXiv:quant-ph/9605047.
Ghirardi, G. C. (1996). “Properties and events in a relativistic context: Revisiting the
dynamical reduction program,” Foundations of Physics Letters, 9: 313–355.
Ghirardi, G. C. (1997a). “Quantum dynamical reduction and reality: Replacing probability
densities with densities in real space,” Erkenntnis, 45: 349–365.
Ghirardi, G. C. (1997b). “Macroscopic reality and the dynamical reduction program,”
pp. 221–240 in M. L. D. Chiara, K. Doets, D. Mundici, and J. V. Benthem (eds.),
Structures and Norms in Science: Volume Two of the Tenth International Congress of
Logic, Methodology, and Philosophy of Science. Dordrecht: Kluwer Academic
Publishers.
Ghirardi, G. C. (1999). “Some lessons from relativistic reduction models,” pp. 117–152 in
H.-P. Breuer and F. Petruccione (eds.), Open Systems and Measurement in Relativis-
tic Quantum Theory: Proceedings of the Workshop Held at the Istituto Italiano per gli
Studi Filosofici, Naples, April 3, 1998. Berlin: Springer.
Ghirardi, G. C. (2000). “Local measurements of nonlocal observables and the relativistic
reduction process,” Foundations of Physics, 30: 1337–1385.
Ghirardi, G. C. (2016). “Collapse theories,” E. N. Zalta (ed.), The Stanford Encyclopedia
of Philosophy (Spring 2016 Edition), https://plato.stanford.edu/archives/spr2016/
entries/qm-collapse/
Ghirardi, G. C. and Grassi, R. (1994). “Outcome predictions and property attribution: The
EPR argument reconsidered,” Studies in History and Philosophy of Science, 25:
397–423.
Ghirardi, G. C. and Grassi, R. (1996). “Bohm’s theory versus dynamical reduction,”
pp. 353–377 in J. T. Cushing, A. Fine, and S. Goldstein (eds.), Bohmian Mechanics
and Quantum Theory: An Appraisal. Dordrecht: Kluwer Academic Publishers.
Ghirardi, G. C., Grassi, R., and Benatti, F. (1995). “Describing the macroscopic world: Closing
the circle within the dynamical reduction program,” Foundations of Physics, 25: 5–38.
30 Wayne C. Myrvold
Ghirardi, G. C., Grassi, R., Butterfield, J., and Fleming, G. N. (1993). “Parameter depend-
ence and outcome dependence in dynamical models for state vector reduction,”
Foundations of Physics, 23: 341–364.
Ghirardi, G. C., Grassi, R., and Pearle, P. (1990). “Relativistic dynamical reduction
models: General framework and examples,” Foundations of Physics, 20: 1271–1316.
Ghirardi, G. C., Grassi, R., and Pearle, P. (1991). “Relativistic dynamical reduction models
and nonlocality,” pp. 109–123 in P. Lahti and P. Mittelstaedt (eds.), Symposium on
the Foundations of Modern Physics 1990. Singapore: World Scientific.
Ghirardi, G. C., Pearle, P., and Rimini, A. (1990). “Markov processes in Hilbert space and
continuous spontaneous localization of systems of identical particles,” Physical
Review A, 42: 78–89.
Ghirardi, G., Rimini, A., and Weber, T. (1986). “Unified dynamics for microscopic and
macroscopic systems,” Physical Review D, 34: 470–491.
Gisin, N. (1989). “Stochastic quantum dynamics and relativity,” Helvetica Physica Acta,
62: 363–371.
Goldstein, S. (1998). “Quantum theory without observers – Part two,” Physics Today, 51:
38–42.
Kent, A. (2005). “Nonlinearity without superluminality,” Physical Review A, 72:
012108.
Maudlin, T. (1996). “Space-time in the quantum world,” pp. 285–307 in J. T. Cushing,
A. Fine, and S. Goldstein (eds.), Bohmian Mechanics and Quantum Theory: An
Appraisal. Dordrecht: Kluwer Academic Publishers.
Maudlin, T. (2007). “Completeness, supervenience and ontology,” Journal of Physics A:
Mathematical and Theoretical, 40: 3151–3171.
Myrvold, W. C. (2003). “Relativistic quantum becoming,” The British Journal for the
Philosophy of Science, 53: 475–500.
Myrvold, W. C. (2011). “Nonseparability, classical and quantum,” The British Journal for
the Philosophy of Science, 62: 417–432.
Myrvold, W. C. (2016). “Lessons of Bell’s theorem: Nonlocality, yes; action at a distance,
not necessarily,” pp. 238–260 in S. Gao and M. Bell (eds.), Quantum Nonlocality and
Reality: 50 Years of Bell’s Theorem. Cambridge: Cambridge University Press.
Myrvold, W. C. (2017). “Relativistic Markovian dynamical collapse theories must employ
nonstandard degrees of freedom,” Physical Review A, 96: 062116.
Myrvold, W. C. (2018). “Ontology for collapse theories,” pp. 99–126 in S. Gao (ed.),
Collapse of the Wave Function: Models, Ontology, Origin, and Implications. Cam-
bridge: Cambridge University Press.
Ney, A. and Albert, D. Z. (eds.) (2013). The Wave Function: Essays on the Metaphysics of
Quantum Mechanics. Oxford: Oxford University Press.
Pearle, P. (1989). “Combining stochastic dynamical state-vector reduction with spontan-
eous localization,” Physical Review A, 39: 2277–2289.
Pearle, P. (1997). “Tales and tails and stuff and nonsense,” pp. 143–156 in R. S. Cohen, M.
Horne, and J. Stachel (eds.), Experimental Metaphysics: Quantum Mechanical Stud-
ies for Abner Shimony, Volume One. Dordrecht: Kluwer Academic Publishers.
Pearle, P. (2009). “How stands collapse II,” pp. 257–292 in W. C. Myrvold and J. Christian
(eds.), Quantum Reality, Relativistic Causality, and Closing the Epistemic Circle:
Essays in Honour of Abner Shimony. Berlin: Springer.
Pearle, P. (2015). “Relativistic dynamical collapse model,” Physical Review D, 91:
105012.
Schweber, S. (1961). An Introduction to Relativistic Quantum Field Theory. New York:
Harper & Row.
Shimony, A. (1991). “Desiderata for a modified quantum mechanics,” pp. 49–59 in A.

Fine, M. Forbes, and L. Wessels (eds.), PSA 1990: Proceedings of the 1990 Biennial
Meeting of the Philosophy of Science Association, Volume Two: Symposia and Invited
Papers. East Lansing, MI: Philosophy of Science Association. Reprinted in (1993),
Search for a Naturalistic Worldview, Volume II: Natural Science and Metaphysics.
Cambridge: Cambridge University Press.
Simon, C., Bužek, V., and Gisin, N. (2001). “No-signalling condition and quantum
dynamics,” Physical Review Letters, 87: 170405.
Tumulka, R. (2006). “A relativistic version of the Ghirardi-Rimini-Weber model,” Journal
of Statistical Physics, 125: 825–844.
Tumulka, R. (2007). “The unromantic pictures of quantum theory,” Journal of Physics A:
Mathematical and Theoretical, 40: 3245–3273.
2
The Modal-Hamiltonian Interpretation: Measurement,
Invariance, and Ontology
olimpia lombardi
2.1 Introduction
In the seventies, Bas van Fraassen (1972, 1974) proposed an approach to quantum
mechanics different than those of the best known interpretations. According to
him, although the quantum state always evolves unitarily (with no collapse), it is a
modal element of the theory: It describes not what is the case but what may be the
case. This idea led several authors since the eighties to propose the so-called modal
interpretations (Kochen 1985, Dieks 1988, 1989, Vermaas and Dieks 1995, Dieks
and Vermaas 1998, Bacciagaluppi and Dickson 1999, Bene and Dieks 2002), that
is, realist, noncollapse interpretations of the standard formalism of quantum mech-
anics, according to which the quantum state assigns probabilities to the possible
values of all the properties of the system. But since the contextuality of quantum
mechanics (Kochen and Specker 1967) implies that it is not possible to consist-
ently assign definite values to all the properties of a quantum system at a single
time, it is necessary to pick out, from the set of all observables of a quantum system
the subset of definite-valued properties. The different modal interpretations differ
from each other mainly with respect to their rule of definite-value ascription (see
Lombardi and Dieks 2017 and references therein).
Like most interpretations of quantum mechanics, the traditional modal
interpretations were specifically designed to solve the measurement problem. In
fact, they successfully reached this goal in the case of ideal measurements.
However, a series of articles of the nineties (Albert and Loewer 1990, 1991,
1993, Elby 1993, Ruetsche 1995) showed that those traditional approaches based
on the modal views did not pick out the right properties for the apparatus in
nonideal measurements, that is, in measurements that do not introduce a perfect
correlation between the possible states of the measured system and the possible
states of the measuring apparatus. As ideal measurements can never be achieved in
practice, this shortcoming was considered a “silver bullet” for killing modal
32
The Modal-Hamiltonian Interpretation: Measurement, Invariance, and Ontology 33
interpretations (Harvey Brown, cited in Bacciagaluppi and Hemmo 1996). This

explains the decline of the interest in modal interpretations since the end of the
nineties.
What was not sufficiently noticed in the nineties was the fact that the difficulties
of those original modal interpretations to deal with nonideal measurements was
due not to their modal nature, but to the fact that their rule of definite-value
ascription made the set of definite-valued observables to depend on the instantan-
eous state of the system. An author who did notice this was Jeffrey Bub, whose
preference for Bohmian mechanics in those days can be understood in this context.
In fact, if Bohmian mechanics is conceived as a member of the modal family
whose definite-valued observables are defined by the position observable (Bub
1997), it turns out to be a natural alternative given the difficulties of the original
modal interpretations.
Bub showed that the shortcomings of the original modal interpretations can be
overcome by making the rule of definite-value ascription independent of the
system’s state and only dependent on an observable of the system. This was
certainly an important step. Nevertheless, it was not sufficient to rehabilitate modal
interpretations in the eyes of most philosophers of physics. What was not realized
at that time is that position is not the only observable that can be appealed to in
order to define the state-independent rule of definite-value ascription of a modal
interpretation. It is in this point that the modal-Hamiltonian interpretation (MHI;
Castagnino and Lombardi 2008, Lombardi and Castagnino 2008) entered the
scene: The MHI endows the Hamiltonian of the quantum system with the role of
selecting its definite-valued observables. With this strategy, it not only solves the
problems of the original modal interpretations, but can also be successfully applied
to many physical situations. However, perhaps due to the shadow of doubt that still
covers the entire modal interpretation project, the MHI did not receive a serious
attention by the community of the philosophers of physics. The present chapter
intends to contribute toward modifying this situation by introducing the MHI in a
conceptually clear and concise way, stressing its advantages both for facing the
traditional interpretive problems of quantum mechanics and for supplying a phys-
ically meaningful account of relevant aspects of the theory.
For this purpose, the chapter is organized as follows. In Section 2.2, the
two main interpretive postulates of the MHI will be introduced, emphasizing
the role played by the Hamiltonian in them. In Section 2.3, the measurement
problem is addressed from the MHI perspective; in particular, it will be argued
that, beyond the formal von Neumann model, quantum measurement is a
symmetry-breaking process that renders empirically accessible an otherwise
inaccessible observable of the system. Section 2.4 will be devoted to assessing
the MHI from the viewpoint of the invariances of the theory, in particular, of the
34 Olimpia Lombardi
Galilei group. Finally, in Section 2.5, the ontological picture suggested by the MHI
will be described, stressing how that picture supplies a conceptually clear solution
to some traditional interpretive problems of quantum mechanics.
2.2 The Modal-Hamiltonian Interpretation

In this section, we shall introduce the MHI without discussing its advantages over
other proposals. The arguments in its favor will become clear in the following
sections, where we will argue for its physical relevance and we will apply it to
solve some traditional interpretive challenges.
By adopting an algebraic perspective, the MHI defines a quantum system S
as a pair ðO; H Þ such that (i) O is a space of self-adjoint operators representing
the observables of the system, (ii) H 2 O is the time-independent Hamiltonian
0 0
of the system S, and (iii) if ρ0 2 O (where O is the dual space of O) is the
initial state of S, it evolves according to the Schrödinger equation. Here we will
assume that the space O is a C*-algebra, which can be represented in terms of a
Hilbert space H (Gelfand-Naimark-Segal [GNS] theorem). In this particular
0 0
case, O ¼ O and, therefore, O and O are represented by H ⊗H . Nevertheless,
O may be a different *-algebra, under the necessary conditions for its
representation.
In this algebraic framework, the observables that constitute the quantum system
are the basic elements of the theory, and the states are secondary elements, defined
in terms of the basic ones. The adoption of an algebraic perspective is not a merely
formal decision. As we will see in Section 2.5, when the logical priority of
observables over states is transferred to the ontological domain, the space of
observables turns out to embody the representation of the elemental items of the
ontology, and this has relevant interpretive consequences.
A quantum system so defined can be decomposed into parts in many ways;
however, not any decomposition will lead to parts which are, in turn, quantum
systems. The expression “tensor product structure” (TPS) is used to call any
partition of a closed system S, represented in the Hilbert space H ¼ H A ⊗H B ,
into parts SA and SB represented in H A and H B , respectively. Quantum systems
admit a variety of TPSs, each one leading to a different entanglement between their
parts. However, there is a particular TPS that is invariant under time evolution: The
TPS is dynamically invariant when there is no interaction between the parts
(Harshman and Wickramasekara 2007a, b). In other words, in the dynamically
invariant case the components’ behaviors are dynamically independent from each
other; each one evolves unitarily according to the Schrödinger equation. On this
basis, according to the MHI, a quantum system can be split into subsystems when
there is no interaction among the subsystems.
Composite systems postulate: A quantum system S : ðO; H Þ, with initial state

0
ρ0 2 O , is composite
when it can be partitioned into two quantum systems
S1 : O1 ; H 1 and S2 : O2 ; H 2 , such that (i) O ¼ O1 ⊗O2 and (ii)
H ¼ H 1 ⊗I 2 þ I 1 ⊗H 2 (where I 1 and I 2 are the identity operators in the correspond-
ing tensor product spaces). In this case, we say that S1 and S2 are subsystems of the
composite system S ¼ S1 þ S2 . If the system is not composite, it is elemental.
With respect to the definite-valued observables, the basic idea of the MHI is that
the Hamiltonian of the system, with its own symmetries, defines the subset of the
observables that acquire definite actual values. The group of the transformations
that leave the Hamiltonian invariant is usually called Schrödinger group (Tinkham
1964). In turn, each symmetry of the Hamiltonian leads to an energy degeneracy.
The degeneracies with origin in symmetries are called “normal” or “systematic,”
and those that have no obvious origin in symmetries are called “accidental”
(Cohen-Tannoudji, Diu, and Lalöe 1977). However, a deeper study usually shows
either that the accidental degeneracy is not exact or else that a hidden symmetry in
the Hamiltonian can be found that explains the degeneracy. For example, the
degeneracy in the hydrogen atom of states of different angular momentum l but
the same principal quantum number n arises from a four-dimensional rotational
symmetry of the Hamiltonian in momentum space (Fock 1935). For this reason it is
assumed that once all the symmetries of the Hamiltonian have been considered, a
basis for the Hilbert space of the system is obtained and the “good quantum
numbers” are well defined.
Once these symmetry considerations are taken into account, the basic idea
of the MHI can be expressed by the classical Latin maxim Ubi lex non
distinguit, nec nos distinguere debemus (where the law does not distinguish,
neither ought we to distinguish). The Hamiltonian of the system, with its
symmetries, is what determines which observables acquire definite values.
This means that any observable whose eigenvalues would distinguish among
eigenvectors corresponding to a single degenerate eigenvalue of the Hamilto-
nian does not acquire definite value, because such an acquisition would
introduce in the system an asymmetry not contained in the Hamiltonian. Once
this idea is understood, the rule of definite-value ascription can be formulated
in a very simple way:
Actualization rule: Given an elemental quantum system S : ðO; H Þ, the actual
definite-valued observables of S are H, and all the observables commuting with H
and having, at least, the same symmetries as H.
The justification for selecting the Hamiltonian as the preferred observable

ultimately lies in the physical relevance of the MHI and in its ability to solve
interpretive difficulties. These issues will be the content of the following sections.
36 Olimpia Lombardi
2.3 The Modal-Hamiltonian View of Quantum Measurement

2.3.1 Measurement and Correlations
In general, the quantum measurement problem is presented in terms of the von
Neumann model, without framing it in the context of the measurement practices.
But the purpose of a quantum measurement is not to discover the preexisting value
of a system’s observable, but to reconstruct, at least partially, the state of the
measured system. Therefore, the following distinction is in order:
• Single measurement: It is a single process, in which the reading of the pointer is

registered. A single measurement, considered in isolation, does not yet supply
relevant information about the state of a measured system.
• Frequency measurement: It is a repetition of identical single measurements,
whose purpose is to obtain the certain coefficients of the measured system’s
state on the basis of the frequencies of the pointer readings in many single
measurements.
A frequency measurement supplies relevant information about the state of the

system, but is not yet sufficient to completely identify such a state. In order to
reconstruct the state of the measured system it is necessary to perform a collection
of frequency measurements with different experimental arrangements.
The von Neumann model addresses the quantum measurement problem in the
framework of the single measurement. This is completely reasonable to the extent
that, if we do not have an adequate explanation of the single case, we cannot
account for the results obtained by the repetition of single cases. Nevertheless, it
should not be forgotten that a single measurement is always an element of a
measurement procedure by means of which, finally, frequencies are to be obtained.
Let us begin, then, by the single measurement. If, as in the original modal
interpretations, the rule of definite-value ascription depends on the instantaneous
state of the system, it is not surprising that it does not supply the expected result
in nonideal measurements: When the state of the composite system measured
system+apparatus does not introduce a perfect correlation between the eigenstates
of the measured system’s observable and the eigenstates of the apparatus’ pointer,
it is not difficult to see that the pointer will not belong to the context of definite-
valued observables. By contrast, if the rule of definite-value ascription does not
depend on the instantaneous state of the system, this problem does not arise. It can
be proved that the MHI explains the definite value of the measurement apparatus’
observable both in the ideal and in the nonideal single measurements (for a formal
demonstration and physical examples, see Lombardi and Castagnino 2008: section
6; for the account of consecutive measurements, see Ardenghi, Lombardi, and
Narvaja 2013).
While definite records of the apparatus’ pointer are obtained even in nonideal
situations, one can legitimately ask whether all nonideal measurements are equally
unsatisfactory. The MHI supplies a clear criterion to distinguish between reliable
and nonreliable frequency measurements (for a detailed explanation and applica-
tions, see Lombardi and Castagnino 2008: section 6). In the former case, the
coefficients of the measured system’s state can be computed on the basis of the
frequencies of the pointer’s readings is spite of the imperfect correlation; in the
latter case, the same computation would give inaccurate results (for a presentation
of the reliability criterion in informational terms, see Lombardi, Fortin, and López
2015). Albert and Loewer (1990, 1993) were right in claiming that the ideal
measurement is a situation that can never be achieved in practice: The interaction
between measured system and measurement apparatus never introduces a com-
pletely perfect correlation; in spite of this, physicists usually perform successful
measurements. The MHI account of the quantum measurement shows that perfect
correlation is not a necessary condition for “good” measurements: The coefficients
of the system’s state at the beginning of the process can be approximately obtained
even when the correlation is not perfect, if the reliability condition is satisfied.
Nevertheless, both in the reliable and in the nonreliable case, a definite reading of
the apparatus’ pointer is obtained in each single measurement.
2.3.2 Measurement and Symmetries

In the von Neumann model of a single measurement, the observable A of the
measured system S, whose eigenstates will be correlated with those of the pointer
P, is considered in formal terms and deprived of any physical content. Then, the
interaction between S and the measuring apparatus M is only endowed with the
role of introducing the correlation between A and P. However, the analysis of
physical situations of measurement shows that there are further aspects to be
considered beyond correlations.
Let us consider the free hydrogen atom, characterized by the Coulombic
interaction between nucleus and electron. Since the Hamiltonian is degenerate
due to its space-rotation invariance, the hydrogen atom is described in terms of
the basis fjn; l; ml ig defined by the complete set of commuting observables
(CSCO) H; L2 ; Lz . Nevertheless, that space-rotation invariance makes the selec-
tion of Lz a completely arbitrary decision: Given that space is isotropic, we can
choose Lx or Ly to obtain an equally legitimate description of the free atom. The
arbitrariness in the selection of the z-direction is manifested in spectroscopy by the
fact that the spectral lines of the free hydrogen atom give no experimental evidence
about the values of Lz : We have no empirical access to the number ml of the free
atom. The MHI agrees with those experimental results, because it does not assign a
38 Olimpia Lombardi
definite value to Lz ; the definite value of Lz would break the symmetry of the
Hamiltonian of the free hydrogen atom in a completely arbitrary way.
If we want to have empirical access to Lz , we need to apply a magnetic field B
along the z-axis, which breaks the isotropy of space and, as a consequence, the
space-rotation symmetry of the atom’s Hamiltonian. In this case, the symmetry
breaking removes the energy degeneracy in ml : Now Lz is not arbitrarily chosen
but selected by the direction of the magnetic field. However this, in turn, implies
that the atom is no longer free: The Hamiltonian of the new system includes the
magnetic interaction. As a consequence, the original degeneracy of the ð2l þ 1Þ-
fold multiplet of fixed n and l is now removed, and the energy levels turn out to be
displaced by an amount Δωnlml , which is also function of ml : This is the manifest-
ation of the so-called Zeeman effect. This means that the Hamiltonian, with
eigenvalues ωnlml , is now nondegenerate: It constitutes by itself the CSCO fH g
that defines the preferred basis fjn; l; ml ig. According to the MHI’s rule of definite-
value ascription, in this case H and all the observables commuting with H are
definite-valued: Since this is the case for L2 and Lz , in the physical conditions
leading to the Zeeman effect, both observables acquire definite values.
Besides the free hydrogen atom and the Zeeman effect, the MHI was applied to
many other physical situations, leading to the results expected from a physical
viewpoint; e.g., the free-particle with spin, the harmonic oscillator, the fine struc-
ture of atoms, the Born-Oppenheimer approximation (see Lombardi and Castag-
nino 2008: section 5). Recently, the interpretation was applied to solve the problem
of optical isomerism (Fortin, Lombardi, and Martínez González 2018), which is
considered one of the deepest problems for the foundations of molecular
chemistry.
All those physical situations show that we have no empirical access to the
observables that are generators of the symmetries of the system’s Hamiltonian;
and, in the context of measurement, the observable A of the measured system S
may be one of those observables. This is also the case in the Stern–Gerlach
experiment, where Sz is a generator of the space-rotation symmetry of
H spin ¼ k S2 ; it is the interaction with the magnetic field B ¼ Bz that breaks the
isotropy of space by privileging the z-direction and, as a consequence, breaks the
space-rotation symmetry of H spin . Therefore, when the observable A to be meas-
ured is a generator of a symmetry of the Hamiltonian of S, the interaction with the
apparatus M must not only establish a correlation between A and the pointer P, but
also must break that symmetry. Therefore, from a physical viewpoint, measure-
ment can be conceived as a process that breaks the symmetries of the Hamiltonian
of the system to be measured and, in this way, turns an otherwise nondefinite-
valued observable into a definite-valued and empirically accessible observable.
This means that the formal von Neumann model of quantum measurement must be
complemented by a physical model, in terms of which, measurement is a symmetry

breaking process that renders a symmetry generator of the system’s Hamiltonian
empirically accessible.
2.4 The Modal-Hamiltonian Interpretation and the Role of Symmetries

2.4.1 The MHI and the Galilei Group
In contrast with the great interest of physicists in the symmetries of physical
theories, the discussion on this topic has been scarce in the field of quantum
mechanics (Lévi-Leblond 1974). This situation is reflected in the field of the
interpretation of the theory, where the relevance of the Galilei group the
symmetry group of nonrelativistic quantum mechanics is rarely discussed in
the impressive amount of literature on the subject. This is a serious shortcoming in
the foundational context, because the fact that a theory is invariant under a group
does not guarantee the same property for its interpretations, to the extent that, in
general, they add interpretive assumptions to its formal structure. The MHI, on the
contrary, addresses the issue of the role and meaning of the Galilei transformations
in the interpretation: the study of whether and under what conditions the MHI
satisfies the physical constraints imposed by the Galilei group leads to interesting
consequences.
As it is well known, the invariance of the fundamental law of a theory under its
symmetry group implies that the behavior of the system is not altered by the
application of the transformation: In terms of the passive interpretation of symmet-
ries, the original and the transformed reference frames are equivalent. In the
particular case of nonrelativistic quantum mechanics, the application of a Galilei
transformation does not introduce a modification in the physical situation, but only
expresses a change of the perspective from which the system is described.
Harvey Brown, Mauricio Suárez, and Guido Bacciagaluppi (1998) correctly
pointed out that any interpretation that selects the set of the definite-valued
observables of a quantum system is committed to explaining how that set is
transformed under the Galilei group. This question is particularly pressing for
realist interpretations of quantum mechanics, which conceive a definite-valued
observable as a physical property that objectively acquires an actual definite value
among all its possible values: The actualization of one of the possible values has to
be an objective fact. Therefore, to the extent that the theory preserves its invariance
under the Galilei group, the set of the definite-valued observables of a system
should be left invariant by the Galilei transformations. From a realist viewpoint, it
would be unacceptable that such a set changed as the mere result of a change in the
perspective from which the system is described. The MHI meets the challenge and
40 Olimpia Lombardi
overcomes it successfully. In fact, it can be proved that, in the situations in which

the Schrödinger equation remains invariant under the group, the set of definite-
valued observables picked out by the modal-Hamiltonian rule of definite-value
ascription also remains invariant (see Ardenghi, Castagnino, and Lombardi 2009,
Lombardi, Castagnino, and Ardenghi 2010).
However the argument can also be developed in the opposite direction. Instead
of starting by the interpretation and considering its behavior under the transform-
ations of the relevant group, one can begin by the group of symmetry and ask for
the constraints that it imposes on interpretation. In the case of nonrelativistic
quantum mechanics, the objectivity of the definite-valued observables must be
preserved by making them invariant under the Galilei group. The natural way to
reach this goal is to appeal to the Casimir operators of the Galilei group, which by
definition are the operators invariant under all the transformations of the Galilei
group. If the interpretation has to select a Galilei-invariant set of definite-valued
observables, the members of such a set must be the Casimir operators of the group
or functions of them. The central extension of the Galilei group has three Casimir
operators which, as such, commute with all the generators of the group: They are
the mass operator M, the squared-spin operator S2 , and the internal energy operator
W ¼ H P2 =2mW. The eigenvalues of the Casimir operators label the irreducible
representations of the group; so, in each irreducible representation, the Casimir
operators are multiples of the identity: M ¼ mI, where m is the mass;
S2 ¼ sðs þ 1ÞI, where s is the eigenvalue of the spin S; and W ¼ wI, where w is
the scalar internal energy.
This result, which places the Casimir operators of the group in the center of the
stage, may seem to disagree with an interpretation such as the MHI, which endows
the Hamiltonian with the leading role. The definite-valuedness of the mass operator
M and the squared-spin operator S2 are compatible with the MHI rule of definite-
value ascription, because both commute with H and do not break its symmetries
(they are multiples of the identity). But the Hamiltonian is not a Casimir operator
of the Galilei group; the Casimir operator is the internal energy. Nevertheless, the
disagreement is only apparent. The Hamiltonian is the sum of the internal energy
and the kinetic energy of the system. But the kinetic energy can be disregarded:
When the system is described in a reference frame at rest with respect to its center
of mass, then the kinetic energy turns out to be zero and the Hamiltonian is
identified with the internal energy. This means that the internal energy is the
magnitude that carries the physically meaningful structure of the energy spectrum,
whereas the kinetic energy represents an energy shift that is physically nonrelevant
and merely relative to the reference frame used for the description.
Summing up, the modal-Hamiltonian interpretation can be reformulated in an
explicitly invariant form, according to which the definite-valued observables of a
quantum system are (i) the observables C i represented by the Casimir operators of
the Galilei group in the corresponding irreducible representation, and (ii) all the
observables commuting with the Ci and having, at least, the same symmetries as
the Ci (Lombardi, Castagnino, and Ardenghi 2010). Therefore, the interpretation
should be more precisely referred to by the name “modal-Casimir interpretation,”
although in the case of nonrelativistic quantum mechanics the original name is also
adequate.
2.4.2 Interpretation and Symmetry

Now we can come back to the question about the constraints that the Galilei group
imposes on the interpretation of quantum mechanics, but now independently of the
MHI. Let us recall that the application of a transformation belonging to the
symmetry group of a theory does not introduce a modification in the physical
situation, but only expresses a change of the perspective from which the system is
described. This leads to the natural idea, expressed by a wide spectrum of authors
(e.g., Minkowski 1923, Weyl 1952, Auyang 1995, Nozick 2001), that the invari-
ance under the relevant group is a mark of objectivity.
On the other hand, as a consequence of the Kochen-Specker theorem (1967), it
is necessary to pick out, from the set of all observables of a quantum system, the
subset of observables that may have definite values. In turn, from a realist
viewpoint, the fact that certain observables acquire an actual definite value is an
objective fact in the behavior of the system; therefore, the set of definite-valued
observables selected by a realist interpretation must be also Galilei-invariant. But
the Galilei-invariant observables are always functions of the Casimir operators of
the Galilei group. As a consequence, one is led to the conclusion that any realist
interpretation that intends to preserve the objectivity of the set of the definite-
valued observables may not stand very far from the MHI (Lombardi and Fortin
2015).
The invariance of the Schrödinger equation holds for the case of isolated
systems, that is, in the case that there are no external fields applied on the system.
Since, in nonrelativistic physics, fields are not quantized, the effect of external
fields on the system has to be accounted for by its Hamiltonian: The potentials
have to modify the form of the Hamiltonian because it is the only observable
involved in the time-evolution law. As a consequence, in the presence of external
fields, the Hamiltonian is no longer the generator of time-displacements; it only
retains its role as the generator of the dynamical evolution (see Laue 1996,
Ballentine 1998). In turn, since the Hamiltonian includes the action of the fields,
the result of the action of the Galilei transformation on it must be computed in each
case, and the Galilei invariance of the Schrödinger equation can no longer be
42 Olimpia Lombardi
guaranteed. This fact suggests the possibility of generalizing the idea of relying on
symmetry groups in two senses.
It cannot be expected that relativistic quantum mechanics be invariant under
the Galilei group, given the fact that it includes the action of electric and
magnetic fields described by a theory that is not Galilei invariant, but Poincaré
invariant. In turn, in quantum field theory, fields are quantum items, not
external fields acting on a quantum system; as a consequence, the generators
of the Poincaré group do not need to be reinterpreted in the presence of external
factors. These facts lead to generalize the group-based interpretive ideas: The
realist interpretation, expressed in terms of the Casimir operators of the Galilean
group in nonrelativistic quantum mechanics, can be transferred to the relativistic
domain by changing the symmetry group accordingly – the definite-valued
observables of a system in relativistic quantum mechanics and in quantum field
theory would be those represented by the Casimir operators of the Poincaré
group. Since the mass operator M and the squared-spin operator S2 are the only
Casimir operators of the Poincaré group, they would always be definite-valued
observables. This conclusion agrees with a usual physical assumption: Elemen-
tal particles always have definite values of mass and spin, and those values are
precisely what define their different kinds. Moreover, the classical limit of
relativistic theories manifests the limit of the corresponding Casimir operators
(see Ardenghi, Castagnino, and Lombardi 2011): There is a meaningful limiting
relation between the observables that acquire definite values according to
relativistic theories and those that acquire definite values according to nonrela-
tivistic quantum mechanics.
These group-based interpretive ideas can be further generalized in a second
sense. If invariance is a mark of objectivity, there is no reason to focus only on
spacetime global symmetries. Internal or gauge symmetries should also be con-
sidered as relevant in the definition of objectivity and, as a consequence, in the
identification of the definite-valued observables of the system. For instance, in
relativistic quantum mechanics a gauge symmetry is what identifies the charge as
an objective quantity. Therefore, a realist interpretation can be extended to the
gauge symmetries of the theory: The observables represented by operators invari-
ant under those symmetries are also definite-valued observables according to the
theory.
In summary, besides its wide applicability in the nonrelativistic quantum
domain, the MHI opens the way for a general interpretive strategy, valid for any
realistic view of quantum theories – the definite-valued observables of a system,
whose behavior is governed by a certain theory, are the observables invariant under
all the transformations corresponding to the symmetries of the theory, both exter-
nal and internal.
2.5 A Modal Ontology of Properties

2.5.1 The Structure of the Ontology
Traditionally, the interpretations of quantum mechanics concentrate their efforts on
the interpretive challenges of the theory. For instance, they focus on searching a
solution of the measurement problem without falling beyond the limitations
imposed by the no-go quantum theorems. Due to their difficulty, these tasks
usually lead people to disregard ontological issues, in particular, the questions
about the nature of the items referred to by quantum mechanics. The MHI has
tackled the ontological questions from the very beginning.
As explained in Section 2.2, the MHI adopts an algebraic perspective. This
decision about the formalism is not confined to the formal domain, but rather has
relevant consequences about the structure of the ontology referred to by quantum
mechanics, in particular, about the basic categories of such an ontology. In fact,
when the logical priority of observables over states is transferred to the ontological
domain, the space of observables turns out to embody the representation of the
elemental items of the ontology – observables (mathematically represented by self-
adjoint operators) ontologically represent type-properties, and the values of the
observables (mathematically represented by the eigenvalues of the corresponding
operators) ontologically represent the possible case-properties corresponding to
those type-properties. Among the possible case-properties of a type-property, only
one acquires a definite value (Lombardi and Castagnino 2008: section 8).
In this modal ontology of properties, a quantum system is a bundle of properties:
type-properties with their corresponding case-properties. The notion of bundle of
properties is a well-known idea in contemporary metaphysics: Philosophers of the
empiricist tradition have preferred to replace the traditional picture of properties
“stuck” to an underlying and unobservable substance by an ontological realm
where individuals are nothing but bundles of properties. Properties have meta-
physical priority over individuals; therefore, they are the fundamental items of the
ontology. However, the view of bundles of properties that is appropriate for
quantum mechanics does not agree with the “bundle theory” of twentieth-century
analytic metaphysics concerning two aspects.
In the first place, according to the traditional versions of the bundle theory, an
individual is the confluence of certain case-properties, under the assumption that
the corresponding type-properties are all determined in terms of actual definite
values. For instance, a particular ball is the confluence of a definite position, say,
on the chair; a definite shape, say, round; a definite color, say, white; etc. The ball
is the bundle of those actual case-properties. In general, bundle theories identify
individuals with bundles of actual properties. By contrast, in the framework of the
MHI, a system is identified by its space of observables, which defines all the
44 Olimpia Lombardi
admissible type-properties with their corresponding possible case-properties.

Therefore, a quantum system is a bundle of possible case-properties; it inhabits
the realm of possibility and manifests itself only partially in the realm of actuality.
This ontological interpretation embodies a possibilist conception of possibility,
as opposed to an actualist view, which reduces possibility to actuality. According
to possibilism, possibility is an ontologically irreducible feature of reality. Possible
items possibilia constitute a basic ontological category (see Menzel 2007). In
other words, possibility is a way in which reality manifests itself, a way independ-
ent of and not less real than actuality. The reality of possibilia is manifested by the
fact that they may produce definite effects on actual reality even if they never
become actual (e.g., “non-interacting experiments” of Elitzur and Vaidman 1993,
Vaidman 1994).
The second specific aspect of this quantum-bundle view is related to the way in
which bundles are conceived. In the traditional versions of the bundle theory, the
claim is that individuals are bundles of properties; therefore, it is necessary to find
what confers individuality to individuals. In general, the task is fulfilled by some
subset of the bundle’s properties, together with some further principle that ensures
that no other individual must possess that subset and that preserves the identity of
the individual through change. By contrast, due to the indistinguishability of
“identical particles,” quantum mechanics poses a serious challenge to the notion
of individual, either in the substratum-properties picture or in the bundle picture
(see French and Krause 2006 and references therein). The identification of the
complexions resulting from the permutations of identical particles makes the
notion of individual run into trouble.
The MHI tackles the problem by endorsing the idea that quantum systems are
not individuals – they are strictly bundles, and there is no principle that permits
them to be subsumed under the ontological category of individual. Regrettably,
this ontological picture is not properly captured by any formal theory whose
elemental symbols are variables of individual. An ontology populated by bundles
of possible properties cries for a “logics of predicates,” in the spirit of the “calculus
of relations” proposed by Alfred Tarski (1941), where individual variables are
absent.
2.5.2 One Ontology, Many Solutions

Quantum mechanics poses different ontological problems – contextuality prevents
the simultaneous assignment of determinate values to all the properties of the
system, nonseparability seems to undermine the independent existence of nonin-
teracting systems, and indistinguishability challenges the traditional category of
individual. The usual strategies focus only on one of these problems: They design
an interpretation to solve it, disregarding the remaining difficulties. With its

ontology of possible properties, the MHI aspires to provide a “global” approach,
which solves most problems in terms of a single ontology.
The Kochen-Specker theorem expresses the impossibility of ascribing actual
case-properties to all the type-properties of the system in a noncontradictory
manner. The classical idea of a bundle of actual properties does not work in the
quantum ontology. But this is not a difficulty for the MHI, which conceives the
quantum system as the bundle of all the possible type-properties with their
corresponding case-properties, as defined by the space of observables. This onto-
logical view is immune to the challenge represented by the Kochen-Specker
theorem, because this theorem imposes no restriction on possibilities (see da Costa,
Lombardi, and Lastiri 2013).
Quantum nonseparability is the consequence of the nonfactorization of entan-
gled states. When the states are assigned to individual systems that interacted in the
past, the difficulty is to explain the correlations between the values of observables
belonging to noninteracting systems, which typically are separated in space. The
assumption of collapse leads to understand nonseparability as nonlocality, at risk
of falling into the “spooky action at a distance” reported by Albert Einstein.
Without collapse, nonseparability seems to imply a kind of holism, in the sense
that quantum systems are not composed by what are commonly conceived as their
subsystems; but this idea can hardly be compatibilized with the view of systems as
individuals, that is, entities that preserve their identity through change. For the
MHI, the interpretation of nonseparability as holism does not represent a difficulty.
Since quantum systems are strictly bundles and not individuals, there is no
principle of individuality that preserves the individuality of the component systems
in the composite system (see da Costa and Lombardi 2014). The composite system
is a single bundle, where the identity of the components is not retained. Therefore,
the new bundle-system acts and reacts as a whole – there are not subsystems whose
state nonseparability must be explained or whose correlations seem to imply
instantaneous action at a distance.
The same idea of the “dissolution” of component bundles in the composite
bundle is what allows the MHI to face the problem of indistinguishability. In the
discussions about the indistinguishability of “identical particles,” the problem is
usually formulated in terms of the possible combinations (complexions) that can be
obtained in the distribution of particles over possible states. The problem is, then,
to explain why a permutation of particles does not lead to different complexions in
quantum statistics. This feature is introduced in the formalism as a restriction on
nonsymmetric states, but the strategy has an unavoidable ad hoc flavor in the
context of the theory. According to the MHI ontology, when a bundle is the result
of the combination of identical bundles, it can be expected that the result does not
46 Olimpia Lombardi
depend on the order in which the original identical bundles are considered; the
combination of identical bundles must be commutative. This commutativity is
manifested by the fact that the observables that constitute the resulting bundle-
system are represented by operators symmetric with respect to the permutation of
the indices coming from the original identical bundles. Here symmetry is not an ad
hoc assumption but a consequence of an ontological feature. When the expectation
values of these symmetric observables are computed, only the symmetric part of
the state has an effect. The nonsymmetric part is superfluous, because it plays no
role in the physically measurable magnitudes (see details in da Costa et al. 2013).
Therefore, symmetrization is not the result of an ad hoc strategy, but is due to
ontological reasons: The symmetry properties of states are a consequence of the
symmetry of the observables of the whole composite system, which is, in turn, a
consequence of the ontological picture supplied by the interpretation. In other
words, from the perspective given by the modal-Hamiltonian interpretation,
indistinguishability is not a relation between particles manifested in statistics, but
rather an internal symmetry of a single bundle of properties.
In summary, according to the MHI, the talk of individual entities and their
interactions can be retained only in a metaphorical sense. In fact, even the number
of particles is represented by an observable, and superpositions of different particle
numbers are theoretically possible. This fact, puzzling from an ontology populated
by individuals, involves no mystery in an ontology of properties: If quantum
systems are bundles of possible properties, the particle picture, with a definite
number of particles, is only a contextual picture valid exclusively when the number
of particles satisfies the constraints of the rule of definite-value ascription. In other
cases, wave packets may remain narrow and more or less localized during a
relatively long time. In this way, particle-like behavior can temporarily emerge –
wave packets can represent approximately definite positions and can follow an
approximately definite trajectory (see Lombardi and Dieks 2016). Moreover, the
MHI has proved to be compatible with the theory of decoherence (Lombardi 2010,
Lombardi, Ardenghi, Fortin, and Castagnino 2011, Lombardi, Ardenghi, Fortin,
and Narvaja 2011). Nevertheless, those particular situations do not undermine the
fact that quantum systems are nonindividual bundles of properties.
2.6 Conclusions and Perspectives

The MHI has been developed and successfully articulated in many directions since
its first presentation in 2008. Of course, this does not mean that any interpretive
question about quantum mechanics has already been solved. Nevertheless, given
the results obtained up to this moment, it deserves to be taken into account and
further explored.
There are several issues that can still be faced from this interpretive frame-
work. A very interesting question is that related to the interpretation of external
fields in a theory that, as quantum mechanics, does not treat fields as quantized
entities. In particular, the Aharonov-Bohm effect is worthy of being analyzed
from an ontology-of-properties view. Another topic to be examined is how the
MHI is in resonance with a closed-system view of decoherence (Castagnino and
Lombardi 2004, 2005a, b, Castagnino, Laura, and Lombardi 2007, Castagnino,
Fortin, and Lombardi 2010, 2014), according to which decoherence is a process
relative to the selected partition of a closed system and how this leads to a top-
down view of quantum mechanics based on an algebraic view that turns
entanglement and discord also into relative phenomena (for initial ideas, see
Lombardi, Fortin, and Castagnino 2012, Fortin and Lombardi 2014, Lombardi
and Fortin 2016). Finally, the natural subsequent interpretive step consists in
extending the MHI to quantum field theory, not only regarding the definite-
valued observables, but also with respect to the ontology referred to by the
theory. In particular, an ontology-of-properties view seems to favor a field view
in the debate on fields vs. particles, but without representing an obstacle to
explaining the emergence of the nonrelativistic quantum ontology. These differ-
ent problems will guide the future research on the further development and
extrapolation of the MHI.
Acknowledgments
I am grateful to the participants of the workshop Identity, indistinguishability and
non-locality in quantum physics (Buenos Aires, June 2017) for their useful com-
ments. This work was made possible through the support of Grant 57919 from the
John Templeton Foundation and Grant PICT-2014–2812 from the National
Agency of Scientific and Technological Promotion of Argentina.
References
Albert, D. and Loewer, B. (1990). “Wanted dead or alive: Two attempts to solve Schrö-
dinger’s paradox,” pp. 277–285 in A. Fine, M. Forbes, and L. Wessels (eds.),
Proceedings of the PSA 1990, Vol. 1. East Lansing, MI: Philosophy of Science
Association.
Albert, D. and Loewer, B. (1991). “Some alleged solutions to the measurement problem,”
Synthese, 88: 87–98.
Albert, D. and Loewer, B. (1993). “Non-ideal measurements,” Foundations of Physics
Letters, 6: 297–305.
Ardenghi, J. S., Castagnino, M., and Lombardi, O. (2009). “Quantum mechanics: Modal
interpretation and Galilean transformations,” Foundations of Physics, 39: 1023–1045.
48 Olimpia Lombardi
Ardenghi, J. S., Castagnino, M., and Lombardi, O. (2011). “Modal-Hamiltonian interpret-

ation of quantum mechanics and Casimir operators: The road to quantum field
theory,” International Journal of Theoretical Physics, 50: 774–791.
Ardenghi, J. S., Lombardi, O., and Narvaja, M. (2013). “Modal interpretations and
consecutive measurements,” pp. 207–217 in V. Karakostas and D. Dieks (eds.), EPSA
2011: Perspectives and Foundational Problems in Philosophy of Science. Dordrecht:
Springer.
Auyang, S. Y. (1995). How Is Quantum Field Theory Possible?. Oxford: Oxford Univer-
sity Press.
Bacciagaluppi, G. and Dickson, M. (1999). “Dynamics for modal interpretations,” Foun-
dations of Physics, 29: 1165–1201.
Bacciagaluppi, G. and Hemmo, M. (1996). “Modal interpretations, decoherence and
measurements,” Studies in History and Philosophy of Modern Physics, 27: 239–277.
Ballentine, L. (1998). Quantum Mechanics: A Modern Development. Singapore: World
Scientific.
Bene, G. and Dieks, D. (2002). “A perspectival version of the modal interpretation of quantum
mechanics and the origin of macroscopic behavior,” Foundations of Physics, 32: 645–671.
Brown, H., Suárez, M., and Bacciagaluppi, G. (1998). “Are ‘sharp values’ of observables
always objective elements of reality?,” pp. 289–306 in D. Dieks and P. E. Vermaas
(eds.), The Modal Interpretation of Quantum Mechanics. Dordrecht: Kluwer Aca-
demic Publishers.
Bub, J. (1997). Interpreting the Quantum World. Cambridge: Cambridge University Press.
Castagnino, M., Fortin, S., and Lombardi, O. (2010). “Is the decoherence of a system the result
of its interaction with the environment?,” Modern Physics Letters A, 25: 1431–1439.
Castagnino, M., Fortin, S., and Lombardi, O. (2014). “Decoherence: A closed-system
approach,” Brazilian Journal of Physics, 44: 138–153.
Castagnino, M., Laura, R., and Lombardi, O. (2007). “A general conceptual framework for
decoherence in closed and open systems,” Philosophy of Science, 74: 968–980.
Castagnino, M. and Lombardi, O. (2004). “Self-induced decoherence: A new approach,”
Studies in History and Philosophy of Modern Physics, 35: 73–107.
Castagnino, M. and Lombardi, O. (2005a). “Self-induced decoherence and the classical
limit of quantum mechanics,” Philosophy of Science, 72: 764–776.
Castagnino, M. and Lombardi, O. (2005b). “Decoherence time in self-induced decoher-
ence,” Physical Review A, 72: #012102.
Castagnino, M. and Lombardi, O. (2008). “The role of the Hamiltonian in the interpret-
ation of quantum mechanics,” Journal of Physics, Conferences Series, 28: 012014.
Cohen-Tannoudji, C., Diu, B., and Lalöe, F. (1977). Quantum Mechanics. New York: John
Wiley & Sons.
da Costa, N. and Lombardi, O. (2014). “Quantum mechanics: Ontology without individ-
uals,” Foundations of Physics, 44: 1246–1257.
da Costa, N., Lombardi, O., and Lastiri, M. (2013). “A modal ontology of properties for
quantum mechanics,” Synthese, 190: 3671–3693.
Dieks, D. (1988). “The formalism of quantum theory: An objective description of reality?,”
Annalen der Physik, 7: 174–190.
Dieks, D. (1989). “Quantum mechanics without the projection postulate and its realistic
interpretation,” Foundations of Physics, 38: 1397–1423.
Dieks, D. and Vermaas, P. (eds.). (1998). The Modal Interpretation of Quantum Mechan-
ics. Dordrecht: Kluwer Academic Publishers.
Elby, A. (1993). “Why ‘modal’ interpretations of quantum mechanics don’t solve the
measurement problem,” Foundations of Physics Letters, 6: 5–19.
Elitzur, A. C. and Vaidman, L. (1993). “Quantum mechanical interaction-free measure-

ments,” Foundations of Physics, 23: 987–997.
Fock, F. (1935). “Zur Theorie des Wasserstoff Atoms.” Zeitschrift für Physik, 98:
145–154.
Fortin, S. and Lombardi, O. (2014). “Partial traces in decoherence and in interpretation:
What do reduced states refer to?,” Foundations of Physics, 44: 426–446.
Fortin, S., Lombardi, O., and Martínez González, J. C. (2018). “A new application of the
modal-Hamiltonian interpretation of quantum mechanics: The problem of optical
isomerism,” Studies in History and Philosophy of Modern Physics, 62: 123–135.
French, S. and Krause, D. (2006). Identity in Physics: A Historical, Philosophical and
Formal Analysis. Oxford: Oxford University Press.
Harshman, N. L. and Wickramasekara, S. (2007a). “Galilean and dynamical invariance of
entanglement in particle scattering,” Physical Review Letters, 98: 080406.
Harshman, N. L. and Wickramasekara, S. (2007b). “Tensor product structures, entangle-
ment, and particle scattering,” Open Systems and Information Dynamics, 14:
341–351.
Kochen, S. (1985). “A new interpretation of quantum mechanics,” pp. 151–169 in
P. Mittelstaedt and P. Lahti (eds.), Symposium on the Foundations of Modern Physics
1985. Singapore: World Scientific.
Kochen, S. and Specker, E. (1967). “The problem of hidden variables in quantum
mechanics,” Journal of Mathematics and Mechanics, 17: 59–87.
Laue, H. (1996). “Space and time translations commute, don’t they?,” American Journal of
Physics, 64: 1203–1205.
Lévi-Leblond, J. M. (1974). “The pedagogical role and epistemological significance of
group theory in quantum mechanics,” Nuovo Cimento, 4: 99–143.
Lombardi, O. (2010). “The central role of the Hamiltonian in quantum mechanics:
Decoherence and interpretation,” Manuscrito, 33: 307–349.
Lombardi, O., Ardenghi, J. S., Fortin, S., and Castagnino, M. (2011). “Compatibility
between environment-induced decoherence and the modal-Hamiltonian interpretation
of quantum mechanics,” Philosophy of Science, 78: 1024–1036.
Lombardi, O., Ardenghi, J. S., Fortin, S., and Narvaja, M. (2011). “Foundations of
quantum mechanics: Decoherence and interpretation,” International Journal of
Modern Physics D, 20: 861–875.
Lombardi, O. and Castagnino, M. (2008). “A modal-Hamiltonian interpretation of quan-
tum mechanics,” Studies in History and Philosophy of Modern Physics, 39: 380–443.
Lombardi, O., Castagnino, M., and Ardenghi, J. S. (2010). “The modal-Hamiltonian
interpretation and the Galilean covariance of quantum mechanics,” Studies in History
and Philosophy of Modern Physics, 41: 93–103.
Lombardi, O. and Dieks, D. (2017). “Modal interpretations of quantum mechanics,” in
E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Spring 2017 Edition).
http://plato.stanford.edu/archives/spr2014/entries/qm-modal/
Lombardi, O. and Dieks, D. (2016). “Particles in a quantum ontology of properties,”
pp. 123–143 in T. Bigaj and C. Wüthrich (eds.), Metaphysics in Contemporary
Physics. Leiden: Brill.
Lombardi, O. and Fortin, S. (2015). “The role of symmetry in the interpretation of quantum
mechanics,” Electronic Journal of Theoretical Physics, 12: 255–272.
Lombardi, O. and Fortin, S. (2016). “A top-down view of the classical limit of quantum
mechanics,” pp. 435–468 in R. Kastner, J. Jeknić-Dugić, and G. Jaroszkiewicz (eds.),
Quantum Structural Studies: Classical Emergence from the Quantum Level. Singa-
pore: World Scientific.
50 Olimpia Lombardi
Lombardi, O., Fortin, S., and Castagnino, M. (2012). “The problem of identifying the
system and the environment in the phenomenon of decoherence,” pp. 161–174 in
H. W. de Regt, S. Hartmann, and S. Okasha (eds.), EPSA Philosophy of Science:
Amsterdam 2009. Berlin: Springer.
Lombardi, O., Fortin, S., and López, C. (2015). “Measurement, interpretation and infor-
mation,” Entropy, 17: 7310–7330.
Menzel, C. (2007). “Actualism,” in E. N. Zalta (ed.), The Stanford Encyclopedia of
Philosophy (Spring 2007 Edition), http://plato.stanford.edu/archives/spr2007/entries/
actualism/
Minkowski, H. (1923). “Space and time,” pp. 75–91 in W. Perrett and G. B. Jeffrey (eds.),
The Principle of Relativity: A Collection of Original Memoirs on the Special and
General Theory of Relativity. New York: Dover.
Nozick, R. (2001). Invariances: The Structure of the Objective World. Harvard: Harvard
University Press.
Ruetsche, L. (1995). “Measurement error and the Albert-Loewer problem,” Foundations of
Physics Letters, 8: 327–344.
Tarski, A. (1941). “On the calculus of relations,” The Journal of Symbolic Logic, 6: 73–89.
Tinkham, M. (1964). Group Theory and Quantum Mechanics. New York: McGraw-Hill.
Vaidman, L. (1994). “On the paradoxical aspects of new quantum experiments,”
pp. 211–217 in Proceedings of 1994 the Biennial Meeting of the Philosophy of
Science Association, Vol. 1, East Lansing, MI: Philosophy of Science Association.
Van Fraassen, B. C. (1972). “A formal approach to the philosophy of science,”
pp. 303–366 in R. Colodny (ed.), Paradigms and Paradoxes: The Philosophical
Challenge of the Quantum Domain. Pittsburgh: University of Pittsburgh Press.
Van Fraassen, B. C. (1974). “The Einstein-Podolsky-Rosen paradox,” Synthese, 29:
291–309.
Vermaas, P. and Dieks, D. (1995). “The modal interpretation of quantum mechanics and its
generalization to density operators,” Foundations of Physics, 25: 145–158.
Weyl, H. (1952). Symmetry. Princeton, NJ: Princeton University Press.
3
Quantum Mechanics and Perspectivalism
dennis dieks
3.1 Introduction: Early Hints of Noncollapse and Perspectivalism

In introductions to quantum mechanics it is standard to introduce “collapses of the
wave function” in order to avoid the occurrence of superpositions of states
associated with different macroscopic properties. The paradigm case is the quan-
tum mechanical treatment of measurement: If the interaction between a quantum
system and a measuring device is described by means of unitary Schrödinger
evolution, the composite system of object plus device will generally end up in an
entangled state that is not an eigenstate of the measured observable, but rather a
superposition of such states. However, successful measurements end with the
realization of exactly one of the possible outcomes, so it appears plausible that at
some stage during the measurement interaction unitary evolution is suspended and
a collapse to one of the terms in the superposition takes place.
However, experimental research of the last few decades has undermined this
motivation for the introduction of collapses. “Schrödinger cat states,” i.e., super-
positions of distinguishable quantum states of mesoscopic or even practically
macroscopic physical systems, are now routinely prepared in the laboratory, and
interference between the different terms in the superpositions have abundantly
been verified (see e.g., Johnson et al. 2017 for a sample of recent developments).
This lends inductive support to the hypothesis that superpositions never really
collapse, but are merely difficult to detect in everyday situations. In such situations
huge numbers of environmental degrees of freedom come into play, so that the
mechanism of decoherence may be invoked as an explanation for the practical
unobservability of interference between macroscopically different states under
standard conditions. This line of thought leads in the direction of noncollapse
interpretations of quantum mechanics.
The evidence against collapses has not yet affected the textbook tradition, which
has not questioned the status of collapses as a mechanism of evolution alongside
51
52 Dennis Dieks
unitary Schrödinger dynamics. However, the relevant views of the pioneers of

quantum mechanics were not at all clear-cut. The locus classicus for the introduc-
tion and discussion of collapses is chapter 6 of von Neumann’s (1932) Mathemat-
ical Foundations of Quantum Mechanics. In this chapter von Neumann underlines
the fundamental difference between collapses occurring in measurements and
unitary evolution, but connects this difference to the distinction between, on the
one hand, the experience of an observer and, on the other hand, external descrip-
tions (in which the observer is treated in the same way as the other physical
systems involved in the measurement interaction). In the external description von
Neumann assumes unitary evolution, with superpositions (also involving the
observer) as an inevitable consequence. Nevertheless, von Neumann states that
the content of the observer’s “subjective experience” corresponds to only one
single term in the superposition.
So the distinction between collapses and unitary evolution for von Neumann is
not a distinction between two competing and potentially conflicting physical
interaction mechanisms on the same level of description, but rather concerns what
can be said in relation to two different points of view an idea taken up and
developed by London and Bauer (discussed later).
Niels Bohr also took the position that the standard rules of quantum mechanics
apply even to measuring devices and other macroscopic objects, so that strictly
speaking these cannot be characterized by sets of precise values of classical
quantities (even though folklore has it that Bohr assumed that quantum mechanics
does not apply to the macroscopic world, see Dieks 2017 for an extensive discus-
sion). Thus, in 1935 Bohr wrote that “a purely classical account of the measuring
apparatus still implies the necessity of latitudes corresponding to the uncertainty
relations. If spatial dimensions and time intervals are sufficiently large, this
involves no limitation” (Bohr 1935). And in 1948 he commented in the same vein:
“We may to a very high degree of approximation disregard the quantum character
of the measuring instruments if they are sufficiently heavy” (Bohr 1948). Although
for Bohr there is thus no difference of principle between macro and micro objects,
he does assign a special role to the observer and to the “conditions of measure-
ment” (the specific experimental setup, chosen by the observer). However, he does
not acknowledge a sui generis measurement dynamics but rather refers to the
specific epistemological vantage point of the observer, who can only communicate
what he finds by using definite values of classical quantities (the paradigm case
being the assignment of a definite value of either position or momentum,
depending on the chosen kind of measuring device).
According to Bohr, the object that is being measured and the measuring device
form, in each individual case, one insoluble whole so that “an independent reality
in the ordinary physical sense can not be ascribed to the phenomena” (Bohr 1928).
Quantum Mechanics and Perspectivalism 53
The properties of a quantum system according to Bohr only become well defined in
the context of the system’s coupling to a measuring device which points in the
direction of a relational nature of physical properties.
A more formal analysis of quantum measurements, close to von Neumann’s
account, was given by London and Bauer in their 1939 booklet on the Theory of
Observation in Quantum Mechanics. London and Bauer consider three interacting
systems: x, the object system, y, a measuring device, and z, the observer. As a
result of the unitary evolution of the combined object-device system, an entangled
P
state will result: k ck jxik jyik . When the observer reads off the result of the
measurement, a similar unitary evolution of the x,y,z system takes place, so that
P
the final state becomes: jΨi ¼ k ck jxik jyik jzik . London and Bauer (1939: 41–42)
comment:
“Objectively” that is, for us who consider as “object” the combined system x,y,z the
situation seems little changed compared to what we just met when we were only
considering apparatus and object. . . . The observer has a completely different viewpoint:
for him it is only the object x and the apparatus y which belong to the external world, to
that what he calls “objective.” By contrast, he has with himself relations of a very special
character: he has at his disposal a characteristic and quite familiar faculty which we can call
the “faculty of introspection.” For he can immediately give an account of his own state. By
virtue of this “immanent knowledge” he attributes to himself the right to create his own
objectivity,
P namely, to cut the chain of statistical correlations expressed by
c
k k jx i k jy i k zik by stating “I am in the state jzik ,” or more simply “I see yk ” or even
j
directly “X ¼ yk .” [Here X stands for the observable whose value is measured by the
apparatus.]
It is clear from this quote and the further context that London and Bauer
believed that there is a role for human consciousness in bringing about a definite
measurement outcome even though they also assumed, like von Neumann, that
“from the outside” the observer, including his consciousness, can be described in a
physicalist way, by unitary quantum evolution (see Jammer 1974). The appeal to
consciousness can hardly be considered satisfactory, though: It appears to invoke a
deus ex machina, devised for the express purpose of reconciling unitary evolution
with definite measurement results. More generally, the hypothesis that the definite-
ness of the physical world only arises as the result of the intervention of (human?)
consciousness does not sit well with the method of physics.
Although certain elements of London and Bauer’s solution are therefore hard to
accept, the suggestion that it should somehow be possible to reconcile universal
unitary evolution and the resulting omnipresence of entangled states, with the
occurrence of definite values of physical quantities, appears plausible. Indeed,
the theoretical framework of quantum mechanics itself (as opposed to modifica-
tions of the quantum formalism, as in the Ghirardi–Rimini–Weber [GRW] theory)
54 Dennis Dieks
does not in a natural way leave room for another dynamical process beside unitary
evolution; e.g., there is no time scale or scale of complexity at which this alterna-
tive evolution could set in. As already mentioned, empirical results support this
verdict. Accordingly, in the next sections we shall investigate whether the early
intuitions about the universality of unitary evolution, excluding collapse as a
dynamical process, can be salvaged in a purely physicalist way. We shall argue
that “perspectival” noncollapse interpretations capture the intuitions behind the
London and Bauer and von Neumann analyses, without an appeal to consciousness
or human observers.
3.2 Relational Aspects of Noncollapse Interpretations

The common feature of noncollapse interpretations is that they single out unitary
evolution (Schrödinger evolution or one of its relativistic generalizations) as the
only way that quantum states evolve in time. Consequently, entangled quantum
states generally result after interactions, even after interactions with macroscopic
objects like measuring devices. Of course, the task is to reconcile this with the
definite states of affairs encountered in experience.
There are several proposals for such a reconciliation. The best known is
probably the many-worlds interpretation, according to which each individual term
in a superposition that results from a measurement-like interaction represents an
actual state of affairs, characterized by definite values of some set of observables.
In this approach there are many actual states of affairs, worlds, or “branches,”
living together in a “super-universe.” The experience of any individual observer is
restricted to one single branch within this super-universe. In other words, the
experienced world is the part of the super-universe that is accessible from the
observer’s perspective (a relational aspect of the scheme, which is the reason that
Everett in 1957 first introduced it as the “relative state” formulation of quantum
mechanics).
A second category of interpretations, modal interpretations, holds that there is
only one actual reality, so that all except one of the “branches” of the total
entangled state do not correspond to actual worlds but rather to unrealized possi-
bilities “modalities.” Some of these modal interpretations make the assumption
that there is one a priori preferred observable (or set of commuting observables)
that is always definite-valued in each physical system, others assume that the set of
definite-valued quantities depends on the form of the quantum state and can
therefore change over time (see Bub 1997, Dieks and Vermaas 1998, Lombardi
and Dieks 2017 for overviews). To the first category belongs the Bohm (1952)
interpretation, in which position is always definite, and the modal-Hamiltonian
interpretation (Lombardi and Castagnino 2008), according to which energy plays a
privileged role. An example of the second category is the proposal according to

which the bi-orthogonal (Schmidt) decomposition of the total state determines the
definite quantities of partial systems (namely, the quantities represented by the
projection operators projecting on the basis vectors that diagonalize the partial
system’s density matrix; see Vermaas and Dieks 1995); another proposal is to
make decoherence responsible for the selection of definite quantities, in the same
way as is now standard in many-worlds accounts.
Other noncollapse approaches are the consistent-histories interpretation
(Griffiths 2017) and Rovelli’s relational interpretation (Laudisa and Rovelli
2013). Interestingly, the latter interpretation posits from the outset that the dynam-
ical properties of any physical system are purely relational and only become
definite with respect to some other system when an interaction between the two
systems (in the formalism described by unitary quantum evolution) correlates the
systems (so that there is an “exchange of information” between them).
However, relational features also have a natural place in most of the other just-
mentioned noncollapse interpretations (although not in all of them), as can be
illustrated by further considering the situation discussed by London and Bauer
(1939: section 1) which essentially is the well-known “Wigner’s Friend”
thought experiment.
Suppose that an experimentalist (our friend, who is a perfect observer) performs
a quantum measurement within a hermetically sealed room. Let us say that the spin
of a spin-½ particle is measured in a previously fixed direction, and that the
experimentalist notes the outcome (either +½ or ½). After some time we, who
are outside the room, will be sure that the experiment is over and that our friend
will have observed a definite result. Yet, we possess no certainty about the
outcome. In a classical context we would therefore represent the state of the room
and its contents by an ignorance mixture over states: There are two possibilities
(“up” and “down”), both with probability ½.
However, in unitary quantum mechanics the situation is different in an important
respect. According to the von Neumann measurement scheme, the final situation of
the room after the experiment, including a record of the friend’s observation, will
be given by a linear superposition of terms, each containing a definite spin state of
the particle coupled to a state of our experimentalist in which he is aware of the
spin value he found. For us outside, this superposition is the correct theoretical
description of the room and its contents, and this (coherent) superposition is
different from an (incoherent) ignorance mixture over different possible states.
As mentioned in the previous section, experience supports the ascription of this
superposed state: Experiments with Schrödinger cat states demonstrate that we
need the superposition to do justice to the experimental facts. For example, if we
are going to measure the projection operator jΨihΨj (where jΨi stands for the
56 Dennis Dieks
superposed state of the room and its contents), the formalism tells us that we shall
find the result “1” with certainty; this is different from what a mixed state would
predict. Experiment confirms predictions of this kind.
But we also possess robust experience about what happens when we watch an
experiment while finding ourselves inside a closed laboratory room: There will be
a definite outcome. It therefore seems inevitable to accept that during the experi-
ment our friend becomes aware of exactly one spin value. As stated by London and
Bauer, our friend will be justified in saying either “the spin is up” or “the spin is
down” after the experiment. The dilemma is that we, on the outside, can only
derive an “improper mixture” as a state for the particle spin and that well-known
arguments forbid us to think that this mixture represents our ignorance about the
actually realized spin-eigenstate (indeed, if the spin state actually was one of the up
or down eigenstates, it would follow that the total system of room and its contents
had to be an ignorance mixture as well, which conflicts with the premise
supported both theoretically and empirically that the total state is a
superposition).
Our proposed perspectival way out of this dilemma is to ascribe more than one state
to the same physical system. In the case under discussion, with respect to us,
representing the outside point of view, the contents of the laboratory room are
correctly described by an entangled pure state so that we should ascribe improper
mixtures (obtained by “partial tracing”) to the inside observer, the measuring device
and the spin particle. But with respect to the inside observer (or with respect to the
measuring device in the room) the particle spin is definite-valued. So the inside
observer assigns a state to his environment that appropriately reflects this definiteness.
This line of thought leads to the idea of assigning relational or perspectival
states, i.e., states of a physical system A from the perspective of a physical system
B. This step creates room for the possibility that the state and physical properties of
a system A are different in relation to different “reference systems” B. As suggested
by the examples, this move may make it possible to reconcile the unitary evolution
during a quantum measurement with the occurrence of definite outcomes. The
properties associated with the superposition and the definite outcomes, respect-
ively, would relate to two different perspectives the idea already suggested by
von Neumann and by London and Bauer. Of course, we should avoid the earlier
problems associated with consciousness. The different perspectives, and different
relational states, should therefore be defined in purely physical terms.
The idea as just formulated was tentative: We spoke in a loose way of “states,”
thinking of wave functions (or vectors in Hilbert space) without specifying what
the attribution of quantum states to physical systems means on the level of physical
quantities, i.e., in terms of physical properties of the systems concerned. In fact,
this physical meaning is interpretation-dependent.
In the many-worlds interpretation the perspectival character of quantum states,

for the Wigner’s Friend–type of scenario that we just discussed, translates into the
following physical account. When the measurement interactions within the her-
metically sealed room have completely ended, the contents in the room have split
into two copies: one in which the outcome +½ has been realized and observed and
one with the outcome ½. However, we as external observers can still verify the
superposed state by measuring an observable like jΨihΨj, so that for us the two
“worlds” inside the room still form one whole. Apparently, the splitting (branch-
ing) of worlds that happens in measurements cannot be a global process, extending
over the whole universe at once, but must be a local splitting that propagates with
the further physical interactions that take place (see Bacciagaluppi 2002). There-
fore, although we know (if we reason in terms of the many-worlds interpretation)
that there are two copies of our friend inside, each having observed one particular
outcome, we still consider the room plus its contents as represented by the coherent
superposition that corresponds to the definite value “1” of the physical quantity
represented by the observable jΨihΨj. So here we encounter a perspectivalism on
the level of physical properties: There exists a definite spin value for the internal
observer but not for his external colleague.
The same type of story can be told in those modal interpretations in which the
definite-valued physical properties of systems are defined by their quantum states
(one detailed proposal for how to define physical properties from the quantum state
can be found in Bene and Dieks 2002 and Dieks 2009). The main difference with
the many-worlds account is that now the interactions within the room do not lead
to two worlds but to only one, with either spin up or spin down. Additionally, in
this case there is a definite spin value for the internal observer whereas from the
outside it is rather the observable jΨihΨj (and observables commuting with it) that
is definite-valued, which conflicts with the attribution of a value to the spin even
though outside observers may be aware that for their inside counterpart there is
such a value.
Rovelli’s relational interpretation, which takes part of its inspiration from
Heisenberg’s heuristics in the early days of quantum mechanics, says that a
quantity of system B only becomes definite for A when an interaction (a measure-
ment) occurs between A and B (Rovelli 1996). In our Wigner’s Friend scenario,
this again leads to the verdict that the internal interactions in the laboratory room
lead to a situation in which the spin is definite with respect to an internal observer
but not for an external one. Only when (and if ) external observers enter the room
and interact with the spin system does the spin become definite for them as well.
In all these cases we obtain accounts that are similar to the London and Bauer
analysis, but with the important distinction that nonphysical features do not enter
the story. It should be noted that the relational properties introduced here are
58 Dennis Dieks
intended to possess an ontological status: It is not the case that for an outside
observer the internal spin values are definite but unknown. The proposal is that the
spin really is indeterminate with respect to the world outside the laboratory room.
This perspectivalism with respect to properties does not seem an inevitable
feature of all noncollapse interpretations, however. In particular, those interpret-
ations of quantum mechanics in which it is assumed that there exists an a priori
given set of preferred observables that is always definite in all physical systems,
at all times, and in all circumstances are by construction at odds with the
introduction of a definiteness that is merely relative. The Bohm interpretation is
a case in point. According to this interpretation all physical systems are composed
of particles that always possess a definite position, as a monadic attribute inde-
pendent of any perspective. So in our sealed-room experiment the instantaneous
situation inside is characterized by the positions of all particles in the room, and
this description is also valid with respect to the outside world even though an
outside observer will usually lack information about the exact values of the
positions. Thus for an external observer there exists one definite outcome of the
experiment inside, corresponding to one definite particle configuration. The out-
come of any measurement on the room as a whole that the outside observer might
perform again corresponds to a definite configuration of particles with well-defined
positions. The fact that this value is not what we would classically expect (for
example, when we measure jΨihΨj) is explained by the Bohm theory via the
nonclassical measurement interaction between the external observer’s measuring
device and the room. The quantum states that in perspectival schemes encode
information about which physical properties are definite, in the Bohm types of
interpretations only play a role in the dynamics of a fixed set of quantities, so that
the possibility of relational properties or perspectivalism does not suggest itself.
However, it has recently been argued that all interpretations of this unitary kind,
characterized by definite and unique (i.e., one-world) outcomes at the end of each
successful experiment even though the total quantum state always evolves unitar-
ily, cannot be consistent (Araújo 2016, Frauchiger and Renner 2016). This argu-
ment is relevant for our theme, and we shall discuss it in some detail.
3.3 Unitarity and Consistency

In a recent paper, Frauchiger and Renner (2016) consider a sophisticated version of
the Wigner’s Friend experiment in which there are two friends, each in her own
room, with a private information channel between them. Outside the two rooms are
Wigner and an Assistant. The experiment consists of a series of four measure-
ments, performed by the individual friends, the Assistant and Wigner, respectively.
In the room of Friend 1 a quantum coin has been prepared in a superposition of
pffiffiffi pffiffiffiffiffiffiffiffi
“heads” and “tails”: 1= 3 jhi þ 2=3jt i. The experiment starts when Friend
1 measures her coin and finds either heads (probability 1/3) or tails (probability 2/3).
pffiffiffi1 then prepares a qubit in the state j0i if her outcome was h, and in the state
Friend
1= 2 ðj0i þ j1iÞ if the outcome was t, and sends this qubit via the private channel
between the rooms to Friend 2. When Friend 2 receives the qubit, he subjects it to a
measurement of an observable that has the eigenstates j0i and j1i. As in the thought
experiment of Section 3.2, the external observers subsequently measure “global”
observables on the respective rooms; this is first done by the Assistant (on the room
of Friend 1) and then by Wigner (on the room of Friend 2).
Frauchiger and Renner claim, via a rather complicated line of reasoning (Araújo
2016 has given a concise version of the argument), that any interpretation of
quantum mechanics that assigns unique outcomes to these measurements “in one
single world,” while using only unitary evolution for the dynamics of the quantum
state (also during the measurements), will lead to an inconsistent assignment of
values to the measurement outcomes. If this conclusion is correct, there are
significant implications for the question of which unitary interpretations of quan-
tum mechanics are possible. The theories that are excluded according to Frauchiger
and Renner are theories that “rule out the occurrence of more than one single
outcome if an experimenter measures a system once” (2016: 2). If this is right,
accepting many worlds would seem inevitable. In fact, Frauchiger and Renner
themselves conclude that “the result proved here forces us to reject a single-world
description of physical reality” (2016: 3).
However, we should not be too quick when we interpret this statement. As
Frauchiger and Renner make clear, they use their “single-world assumption” to
ensure that all measurement outcomes are context-independent. In particular, what
they use in their proof is a compatibility condition between different “stories” of a
measurement: If one experimenter’s story is that an experiment has outcome t,
every other experimenter’s story of the same event must also contain this same
outcome t (Frauchiger and Renner 2016: 7). This is, first of all, a denial of the
possibility of perspectivalism. As we shall further discuss in a moment, perspec-
tival interpretations will be able to escape the conclusion of the Frauchiger-Renner
(F-R) argument. Therefore, we claim that the F-R argument can be taken to lend
support to perspectivalism as one of the remaining consistent possibilities.
The details and the domain of validity of the F-R proof are not completely
transparent and uncontroversial. Indeed, there is at least one nonperspectival
single-world interpretation, namely the Bohm interpretation, whose consistency
is usually taken for granted. This consistency is confirmed by a result of Sudbery
(2017), who has concretely constructed a series of outcomes for the F-R thought
experiment as predicted by a modal interpretation of the Bohm type. According to
Sudbery, there is an unjustified step in Frauchiger and Renner’s reasoning, because
60 Dennis Dieks
they do not fully take into account that in unitary interpretations only the total
(noncollapsed) state can be used for predicting the probabilities of results obtained
by the Assistant and Wigner (The bone of contention is statement 4 in Araújo’s
(2016) reconstruction of the F-R inconsistency, in which Friend 1 argues that her
coin measurement result is only compatible with one single later result obtained by
Wigner in the final measurement. However, in unitary interpretations previous
measurement results do not always play a role in the computation of probabilities
for future events. Indeed, a calculation on the basis of the total uncollapsed
quantum state, as given by Araújo (2016: 4), indicates that Wigner may find either
one of two possible outcomes, with equal probabilities, even given the previous
result of Friend 1 this contradicts the assumption made by Frauchiger and
Renner).
The situation becomes more transparent when we make use of an elegant
version of the F-R thought experiment recently proposed by Bub (2017). Bub
replaces Friend 1 by Alice and Friend 2 by Bob; Alice and Bob find themselves at a
great distance from each other. Alice has a quantum coin which she subjects to a
measurement of the observable pffiffiffiA with eigenstates
pffiffiffiffiffiffiffiffi jhiA , jt iA ; the coin has been
prepared in the initial state 1= 3 jhiA þ 2=3jt iA . Alice
pthen ffiffiffi prepares a qubit in
the state j0iB if her outcome is h and in the state 1= 2 j0iB þ j1iB if her
outcome is t. She subsequently sends this qubit to Bob this is the only
“interaction” between Alice and Bob. After Bob has received the qubit, he subjects
it to a measurement of the observable B with eigenstates j0iB , j1iB .
In accordance with the philosophy of noncollapse interpretations, we assume
that Alice and Bob obtain definite outcomes for their measurements, but that
the total system of Alice, Bob, their devices and environments, and the coin
and the qubit, can nevertheless be described by the uncollapsed quantum state,
namely:
1
jΨi ¼ pffiffiffi jhiA j0iB þ jt iA j0iB þ jt iA j1iB (3.1)
3
For ease of notation, the quantum states of Alice and Bob themselves, plus the
measuring devices used by them, and even the states of the environments that have
become correlated to them, have here all been included in the states
jhiA , jt iA , j0iB , j1iB (so that these states no longer simply refer to the coin and the
qubit, respectively, but to extremely complicated many-particles systems!).
Now we consider two external observers, also at a great distance from each other,
who take over the roles of Wigner and his Assistant, and are going to perform
measurements on Alice and Bob (and their entire experimental setups), respectively.
The external observer pffiffiwho
ffi focuses on Alice measures an
pffiffiobservable
ffi X with
eigen-
states jfailiA ¼ 1= 2 jhiA þ jtiA and jokiA ¼ 1= 2 jhiA jt iA , and the
observer dealing with Bob andpffiffiBob’s

ffi entire experiment
measures
pffiffithe
ffi observable Y
with eigenstates jfailiB ¼ 1= 2 j0iB þ j1iB and jokiB ¼ 1= 2 j0iB j1iB .
A F-R contradiction now arises in the following manner (Bub 2017: 3). The
state jΨi can alternatively be expressed in the following forms:
1 1
jΨi ¼ pffiffiffiffiffi jokiA jokiB pffiffiffiffiffi jokiA jfailiB þ
12 12
rffiffiffi (3.2)
1 3
þ pffiffiffiffiffi jfailiA jokiB þ jfailiA jfailiB
12 4
rffiffiffi
2 1
jΨi ¼ jfailiA j0iB þ pffiffiffi jt iA j1iB (3.3)
3 3
rffiffiffi
1 2
jΨi ¼ pffiffiffi jhiA j0iB þ jt i jfailiB (3.4)
3 3 A
From Eq. (3.2), we see that the outcome fok; okg in a joint measurement of X and
Y has a nonzero probability: This outcome will be realized in roughly 1/12 th of all
cases if the experiments are repeated many times. From Eq. (3.3) we calculate that
the pair fok; 0g has zero probability as a measurement outcome, so fok; 1g is the
only possible pair of values for the observables X, B in the cases in which X has the
value ok. However, from Eq. (3.4) we conclude that the pair ft; okg has zero
probability, so h is the only possible value for the observable A if Y has the value
ok and A and Y are measured together. So this would apparently lead to the pair of
values fh; 1g as the only possibility for the observables A and B, if X and Y are
jointly measured with the result fok; okg. But this pair of values has zero prob-
ability in the state jΨi so it is not a possible pair of measurement outcomes for
Alice and Bob in that state. So although the outcome fok; okg for X and Y is
certainly possible, the (seemingly) necessarily associated outcome fh; 1g for A and
B is not this seems an inconsistency.
In this inconsistency argument there is a silent use of nonperspectivalism
conditions. For example, if Bob’s measurement outcome is 1 from the perspective
of the X measurement, it is assumed that this outcome also has to be 1 as judged
from the perspective of the Y observer. However this assumption does not sit well
with what the quantum formulas show us: The relative state of Bob with respect to
the Y outcome “ok” is not j1iB , but jokiB (see Eq. (3.5)).
To see what is wrong with the inconsistency argument from a perspectival point
of view that closely follows the quantum formalism, it is helpful to note that the
states in Eqs. (3.1), (3.2), (3.3), and (3.4) are all states of Alice and Bob, including
their devices and environments, but without the external observers. In a consistent
noncollapse interpretation we must also include the external observer states in the
62 Dennis Dieks
total state if we want to discuss the measurements of the observables X and Y. If we

denote by joi and jf i the external states corresponding to the measurement results
“ok” and “fail,” respectively, we find for the final state, in obvious notation:
1 1
pffiffiffiffiffi jokiA jokiB joiX joiY pffiffiffiffiffi jokiA jfailiB joiX jf iY þ
12 12
rffiffiffi (3.5)
1 3
pffiffiffiffiffi jfailiA jokiB jf iX joiY þ jfailiA jfailiB jf iX jf iY
12 4
From this equation, and its counterparts for when only X or Y is measured, we read
off that the relative state of Alice and Bob with respect to Alice’s external observer
in state joiX is jokiA j1iB ; with respect to Bob’s external observer in state joiY it is
jhiA jokiB . Both these state assignments refer to the situation in which only X or Y is
measured.
However, the state of Alice and Bob relative to the combined external observers
state joiX joiY is:
1
jokiA jokiB ¼ jhiA jt iA j0iB j1iB (3.6)
2
This is an entangled state in which neither the coin toss nor the qubit measurement
has a definite result it is not the state jhij1i that was argued to be present in
Bub’s version of the F-R argument. This illustrates the fact that in the case of an
entangled state between two systems, the perspectives of an external observer who
measures one system and an observer who measures the other can generally not be
glued together to give us the perspective of the system that consists of both
observers. In fact, as we see from Eq. (3.5), in the final quantum state not only
Alice and Bob, but also the external observers have become entangled with each
other this should already make us suspicious of combining partial viewpoints
into a whole, as it is well known that entanglement may entail nonclassical holistic
features.
So, perspectival views, which make the assignment of properties dependent on
the relative quantum state, are able to escape the inconsistency argument just
discussed by denying that the X-perspective and the Y-perspective can be simply
juxtaposed to form the XY-perspective.
A further point to note is that the quantities measured by the external observers
do not commute with A and B, respectively. So the measurement by the observer
near to Alice introduces a “context” that is different from Alice’s one, and similarly
for the external observation near Bob. This reinforces the notion that the combined
XY perspective need not agree with the measurement outcomes initially found by
Alice and Bob (compare Fortin and Lombardi 2019). Indeed, the correct “Alice
and Bob state” from the XY-perspective, given by Eq. (3.6), does not show one
definite combination of results of Alice’s A and Bob’s B measurement but contains

all of them as possibilities.
The Frauchiger and Renner argument, in Bub’s formulation, therefore does not
threaten perspectival one-world interpretations with unitary dynamics. However,
we should wonder whether nonperspectival unitary schemes, like the Bohm
theory, will also be able to escape inconsistency, and if so, exactly how they do
so. When we again follow the steps in the measurement procedure of the thought
experiment, we can conclude that the X result “ok” can only occur if Bob had
measured “1”. In the context of the Bohm interpretation, this means that the
particle configuration of X ends up at a point of configuration space that is
compatible with the state joiX only if Bob’s configuration is in a part of configur-
ation space compatible with j1iB . The same conclusion can be drawn with respect
to Y and A for a measurement series in which Y is measured before X: If the Y
measurement is performed first and the result “ok” is registered, A must have seen
“heads”. Now, X and Y are at space-like separation from each other, and this might
seem to imply that it cannot make a difference to the state of Alice and Bob what
the order in time is of the X and Y measurements. If the X measurement with result
“ok” is the earlier one, Bob must have been in state j1iB before the external
measurements started; if Y is measured first, Alice must have been in state jhiA
before the start of the external measurements. Therefore, if the time order is
immaterial, Alice and Bob together will with certainty have been in the state
fh; 1g in the cases in which the X,Y measurements have ended with the result
fok; okg. But this is in contradiction with what the unitary formalism predicts: Eq.
(3.1) shows that the pair of outcomes fh; 1g is impossible. So we have an
inconsistency, and the Bohm interpretation and other nonperspectival unitary
interpretations seem to be in trouble.
However, in the Bohm theory the existence of a preferred reference frame that
defines a universal time is assumed (see for more on the justification of this
assumption in Section 3.4). This makes it possible to discuss the stages of the
experiment in their objectively correct temporal order. Assume that after Alice’s
and Bob’s “in-the-room” measurements (the first by Alice, the second by Bob), the
external observer near to Alice measures first, after which the observer close to
Bob performs the second measurement. The first external measurement will disturb
the configuration of particles making up Alice and her environment, so Alice’s
state will be changed. If the result of the X measurement is “ok”, Bob’s internal
result must have been “1” (Bob is far away, but the possibility of this inference is
not strange, because the total state is entangled, which entails correlations between
Alice and Bob). This “1” will remain unchanged until the second external meas-
urement, of Y. This second measurement will change Bob’s result. Now, in this
story it is not true that the outcome “ok” of Y is only compatible with Alice’s initial
64 Dennis Dieks
outcome “heads”, as used in the inconsistency argument: the previous measure-

ment of X has blocked this conclusion, as inspection of the initial state given in Eq.
(3.1), and its evolution under the external measurements, shows. In fact, the initial
outcomes combination {t, 1} will give rise to the later outcomes {ok, ok} in 1/4 th
of the cases, which is exactly what is needed to achieve complete consistency:
since {t, 1} will occur in 1/3 rd of the cases, {ok, ok} will be found in 1/12 th of all
cases.
So also Bohm, and possibly other nonperspectival schemes, are able to
escape the inconsistency argument. In the perspectival schemes the key was
that two perspectives cannot always be simply combined into one global
perspective; because of this, we were allowed to speak about the X perspective
and the Y perspective without specifying the temporal order of the X and Y
measurements. The threat of inconsistency was avoided by blocking the
composition of the two perspectives into one whole. In the nonperspectival
scheme the existence of a preferred frame of reference comes to the rescue and
protects us against inconsistency: We can follow the interactions and the
changes produced by them step by step in their unique real-time order, so
that no ambiguity arises about which measurement comes first and about what
the actual configuration is at each instant. The issue of combining descriptions
from different perspectives accordingly does not arise, and this is enough to
tell one consistent story.
The difference between the perspectival and nonperspectival unitary accounts,
and the apparent connection between perspectivalism and Lorentz invariance,
suggests that there is a link between perspectivalism and relativity. Perspectivalism
seems able to avoid inconsistencies without introducing a privileged frame of
reference. On the other hand, the introduction of such a privileged frame in
Bohm-like interpretations now appears as a ploy to eliminate the threat of incon-
sistencies without adopting perspectivalism.
3.4 Relativity
The diagnosis of the previous section is confirmed when we directly study the
consequences of relativity for interpretations of quantum mechanics. In particular,
when we attempt to combine special relativity with unitary interpretational
schemes, new hints of perspectivalism emerge. As mentioned, the Bohm interpret-
ation has difficulties in accommodating Lorentz invariance. Bohmians have there-
fore generally accepted the existence of a preferred inertial frame in which the
equations assume their standard form a frame resembling the ether frame of
prerelativistic electrodynamics. Accepting such a privileged frame in the context of
what we know about special relativity and Minkowski spacetime is, of course, not
something to be done lightly; it must be a response to a problem of principle.

Indeed, it can be mathematically proved that no unitary interpretation scheme that
attributes always definite positions to particles (as in the Bohm theory) can satisfy
the requirement that the same probability rules apply equally to all hyperplanes in
Minkowski spacetime (Berndl, Dür, Goldstein, and Zanghì 1996).
The idea of this theorem (and of similar proofs) is that intersecting hyper-
planes should carry properties and probabilities in a coherent way, which means
that they should give agreeing verdicts about the physical conditions at the
spacetime points where they (the hyperplanes) intersect. The proofs demonstrate
that this meshing of hyperplanes is impossible to achieve with properties that are
hyperplane independent. The no-go results can be generalized to encompass
nonperspectival unitary interpretations that attribute other definite properties
than position, and to unitary interpretations that work with sets of properties
that change in time (Dickson and Clifton 1998). A general proof along these
lines was given by Myrvold (2002a). Myrvold shows, for the case of two
systems that are (approximately) localized during some time interval, that it is
impossible to have a joint probability distribution of definite properties along
four intersecting hyperplanes such that this joint distribution returns the Born
probabilities on each hyperplane. An essential assumption in the proof (Myrvold
2002a: 1777) is that the properties of the considered systems are what Myrvold
calls local: The value of quantity A of system S at spacetime point p (a point
lying on more than one hyperplane) must be well defined regardless of the
hyperplane to which p is taken to belong and regardless of which other systems
are present in the universe. It turns out that such local properties cannot obey the
Born probability rule on each and every hyperplane. The assumption that the
Born rule only holds in a preferred frame of reference is one way of responding
to this no-go result.
The argument has been given a new twist by Leegwater (2018), who argues that
“unitary single-outcome quantum mechanics” cannot be “relativistic,” where a
theory is called relativistic if all inertial systems have the same status with respect
to the formulation of the dynamic equations of the theory (i.e., what usually is
called Lorentz or relativistic invariance). Like Frauchiger and Renner, Leegwater
considers a variation on the Wigner’s friend thought experiment: There are three
laboratory rooms, at spacelike distances from each other, each with a friend inside
and a Wigner-like observer outside. In each of the lab rooms there is also a spin-½
particle, and the experiment starts in a state in which the three particles (one in each
room) have been prepared in a so-called GHZ-state (Greenberger, Horne, Shi-
mony, and Zeilinger 1990). The description of the thought experiment in an
initially chosen inertial rest frame is assumed to be as follows: At a certain instant
the three friends inside their respective rooms simultaneously measure the spins of
66 Dennis Dieks
their particles, in a certain direction; thereafter, at a second instant, each of the three
outside observers performs a measurement on his or her room. This measurement
is of a “whole-room” observable, like in the Frauchiger-Renner thought experi-
ment discussed in the previous section. As a result of the internal measurements by
the three friends the whole system, consisting of the rooms and their contents, has
ended up in an entangled GHZ-state. Leegwater is able to show that this entails that
the assumption that the standard rules of quantum mechanics apply to each of three
differently chosen sets of simultaneity hyperplanes, gives rise to a GHZ-
contradiction: The different possible measurement outcomes (all +1 or 1) cannot
be consistently chosen such that each measurement has the same outcome irre-
spective of the simultaneity hyperplane on which it is considered to be situated
(and so that all hyperplanes mesh). As in the original GHZ-argument (Greenberger
et al. 1990), the contradiction is algebraic and does not involve the violation of
probabilistic (Bell) inequalities.
One way of responding to these results is the introduction of a preferred inertial
system (a privileged perspective!), corresponding to a state of absolute rest,
perhaps defined with respect to an ether. This response is certainly against the
spirit of special relativity, in particular because the macroscopic predictions of
quantum mechanics are such that they make the preferred frame undetectable.
Although this violation of relativistic invariance does not constitute an inconsist-
ency, it certainly is attractive to investigate whether there exist other routes to
escape the no-go theorems. Now, as we have seen, a crucial assumption in these
theorems is that properties of systems are monadic, independent of the presence of
other systems and independent of the hyperplane on which they are considered.
This suggests that a transition to relational or perspectival properties offers an
alternative way out.
In fact, that unitary evolution in Minkowski spacetime leads naturally to a
hyperplane-dependent account of quantum states if one describes measurements
by effective collapses has been noted in the literature before (see e.g. Dieks 1985,
Fleming 1996, Myrvold 2002b). The new light that we propose to cast on these and
similar results comes from not thinking in terms of collapses, and of a dependence
on hyperplanes or foliations of Minkowski spacetime as such, but instead of
interpreting them as consequences of the perspectival character of physical prop-
erties: that the properties of a system are defined with respect to other systems.
What we take the considerations in the previous and present sections to suggest is
that it makes a difference whether we view the physical properties of a system from
one or another system – or from one or another temporal stage in the evolution of a
system. In the case of the (more-or-less) localized systems that figure in the
relativistic no-go theorems that we briefly discussed, this automatically leads to
property ascriptions that are different on the various hyperplanes that are
considered. As a result, the meshing conditions on which the theorems hinge no

longer apply.
3.5 Concluding Remarks

If unitary evolution is accepted as basic in quantum mechanics and is combined
with the requirement that results of experiments are to be definite and situated in
one single world, this naturally leads to a picture in which physical systems have
properties that are relational or perspective dependent. As we have seen in Section
3.3, perspectivalism makes it possible to escape arguments saying that interpret-
ations of unitary quantum mechanics in terms of one single world are inconsistent.
Moreover, perspectivalism removes obstacles to the possibility of formulating
interpretational schemes that respect Lorentz invariance by making the introduc-
tion of a preferred inertial frame of reference superfluous (Section 3.4).
The single world that results from perspectivalism is evidently much more
complicated than the world we are used to in classical physics: There are more
than one valid descriptions of what we usually think of as one physical situation.
This reminds of the many-worlds interpretation. There are important differences,
though, between a multiplicity of worlds and the multiplicity of descriptions in
perspectivalism. According to the single-world perspectivalism that we have
sketched, only one of the initially possible results of a measurement becomes
actual from the perspective of the observer, whereas in the many-worlds interpret-
ation all possibilities are equally realized. So the multiplication of realities that
takes place in many-worlds is avoided in perspectivalism. It is of course true that
perspectivalism sports a multiplicity of its own, namely of different points of view
within a single world. But this multiplicity seems unavoidable in the many-worlds
interpretation as well, in each individual branch. For example, in the relativistic
meshing argument of Myrvold (2002a), a situation is discussed in which no
measurements occur: The argument is about two freely evolving localized systems
as described from a number of different inertial frames. Since no measurements are
taking place during the considered process, the inconsistency argument goes
through in exactly the same way in every single branch of the many-worlds
super-universe: There is no splitting during the time interval considered in the
proof of the theorem. So even in the many-worlds interpretation the introduction of
perspectival properties (in each single branch) seems unavoidable in order to avoid
inconsistencies. Another case to be considered is the Wigner’s friend experiment:
When the measurement in the hermetically sealed room has been performed, an
outside observer will still have to work with the superposition of the two branches.
So the splitting of worlds assumed by the many-worlds interpretation must remain
confined to the interior of the room, as mentioned in Section 3.2. In this situation it
68 Dennis Dieks
is natural to make the description of the measurement and its result perspective
dependent: For the two friend-branches inside the room there is a definite outcome,
but this is not so for the external observer. So perspectivalism as a consequence of
holding fast to unitarity and Lorentz invariance seems more basic than the further
choice of interpreting measurements in terms of many worlds; even the many-
worlds interpretation must be committed to perspectivalism. But perspectivalism
on its own is already sufficient to evade the anti-single-world arguments of Section
3.3, so for this purpose we do not need the further assumption of many worlds.
Finally, the introduction of perspectivalism opens the door to several new
questions. In everyday circumstances we do not notice consequences of
perspectivalism, so we need an account of how perspectival effects are washed
out in the classical limit. It is to be expected that decoherence plays an important
role here, as alluded to in the Introduction however, this has to be further worked
out (compare Bene and Dieks 2002). Further, there is the question of how the
different perspectives hang together; for example, in Section 3.3 it was shown that
perspectives of distant observers cannot be simply combined in the case of
entanglement, which may be seen as a nonlocal aspect of perspectivalism. By
contrast, it has been suggested in the literature that perspectivalism makes it
possible to give a purely local description of events in situations of the Einstein-
Podolsky-Rosen type, and several tentative proposals have been made in order to
substantiate this (Rovelli 1996, Bene and Dieks 2002, Smerlak and Rovelli 2007,
Dieks 2009, Laudisa and Rovelli 2013). These and other questions constitute
largely uncharted territory that needs further exploration.
References
Araújo, M. (2016). “If your interpretation of quantum mechanics has a single world but no
collapse, you have a problem,” http://mateusaraujo.info/2016/06/20/if-your-interpret
ation-of-quantum-mechanics-has-a-single-world-but-no-collapse-you-have-a-problem/
Bacciagaluppi, G. (2002). “Remarks on space-time and locality in Everett’s interpretation,”
pp. 105–122 in T. Placek and J. Butterfield (eds.), Non-Locality and Modality.
Dordrecht: Springer.
Bene, G. and Dieks, D. (2002). “A perspectival version of the modal interpretation of
quantum mechanics and the origin of macroscopic behavior,” Foundations of Physics,
32: 645–671.
Berndl, K., Dür, D., Goldstein, S., and Zanghì, N. (1996). “Nonlocality, Lorentz invari-
ance, and Bohmian quantum theory,” Physical Review A, 53: 2062–2073.
Bohm, D. (1952). “A suggested interpretation of the quantum theory in terms of ‘hidden’
variables, I and II,” Physical Review, 85: 166–193.
Bohr, N. (1928). “The quantum postulate and the recent development of atomic theory,”
Nature, 121: 580–590.
Bohr, N. (1935). “Can quantum-mechanical description of physical reality be considered
complete?,” Physical Review, 48: 696–702.
Bohr, N. (1948). “On the notions of causality and complementarity,” Dialectica, 2:

312–319.
Bub, J. (2017). “Why Bohr was (mostly) right,” arXiv:1711.01604v1 [quant-ph].
Dickson, M. and Clifton, R. (1998). “Lorentz invariance in modal interpretations,”
pp. 9–47 in D. Dieks and P. Vermaas (eds.), The Modal Interpretation of Quantum
Mechanics. Dordrecht: Kluwer Academic Publishers.
Dieks, D. (1985). “On the covariant description of wave function collapse,” Physics Letters
A, 108: 379–383.
Dieks, D. (2009). “Objectivity in perspective: relationism in the interpretation of quantum
mechanics,” Foundations of Physics, 39: 760–775.
Dieks, D. (2017). “Niels Bohr and the formalism of quantum mechanics,” pp. 303–333 in
J. Faye and H. J. Folse (eds.), Niels Bohr and the Philosophy of Physics Twenty-
First-Century Perspectives. London and New York: Bloomsbury Academic.
Dieks, D. and Vermaas, P. (eds.). (1998). The Modal Interpretation of Quantum Mechan-
ics. Dordrecht: Kluwer Academic Publishers.
Everett, H. (1957). “‘Relative state’ formulation of quantum mechanics,” Reviews of
Modern Physics, 29: 454–462.
Fleming, G. (1996). “Just how radical is hyperplane dependence?,” pp. 11–28 in R. Clifton
(ed.), Perspectives on Quantum Reality. Dordrecht: Kluwer Academic Publishers.
Fortin, S. and Lombardi, O. (2019). “Wigner and his many friends: a new no-go result?,”
http://philsci-archive.pitt.edu/id/eprint/15552.
Frauchiger, D. and Renner, R. (2016). “Single-world interpretations of quantum theory
cannot be self-consistent,” arXiv:1604.07422v1 [quant-ph].
Greenberger, D., Horne, M., Shimony, A., and Zeilinger, A. (1990). “Bell’s theorem
without inequalities,” American Journal of Physics, 58: 1131–1143.
Griffiths, R. (2017). “The consistent histories approach to quantum mechanics,” in E. N.
Zalta (ed.), The Stanford Encyclopedia of Philosophy (Spring 2017 Edition). https://
plato.stanford.edu/archives/spr2017/entries/qm-consistent-histories/
Jammer, M. (1974). The Philosophy of Quantum Mechanics. New York: Wiley & Sons.
Johnson, K., Wong-Campos, J., Neyenhuis, B., Mizrahi, J., and Monroe, C. (2017).
“Ultrafast creation of large Schrödinger cat states of an atom,” Nature Communi-
cations, 8: Article 697. doi:10.1038/s41467-017-00682-6.
Laudisa, F. and Rovelli, C. (2013). “Relational quantum mechanics,” in E. N. Zalta (ed.),
The Stanford Encyclopedia of Philosophy (Summer 2013 Edition). https://plato
.stanford.edu/archives/sum2013/entries/qm-relational/
Leegwater, G. (2018). “When Greenberger, Horne and Zeilinger meet Wigner’s Friend,”
arXiv:1811.02442 [quant-ph].
Lombardi, O. and Castagnino, M. (2008). “A modal-Hamiltonian interpretation of
quantum mechanics,” Studies in History and Philosophy of Modern Physics, 39:
380–443.
Lombardi, O. and Dieks, D. (2017). “Modal interpretations of quantum mechanics,” in E.
N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Spring 2017 Edition).
https://plato.stanford.edu/archives/spr2017/entries/qm-modal/
London, F. and Bauer, E. (1939). La Théorie de l’Observation en Mécanique Quantique.
Paris: Hermann. English translation, pp. 217–259 in J. A. Wheeler and W. H. Zurek
(eds.), 1983, Quantum Theory and Measurement. Princeton: Princeton University
Press.
Myrvold, W. (2002a). “Modal interpretations and relativity,” Foundations of Physics, 32:
1773–1784.
70 Dennis Dieks
Myrvold, W. (2002b). “On peaceful coexistence: is the collapse postulate incompatible

with relativity?,” Studies in History and Philosophy of Modern Physics, 33: 435–466.
Rovelli, C. (1996). “Relational quantum mechanics,” International Journal of Theoretical
Physics, 35: 1637–1678.
Smerlak, M. and Rovelli, C. (2007). “Relational EPR,” Foundations of Physics, 37:
427–445.
Sudbery, A. (2017). “Single-world theory of the extended Wigner’s friend experiment,”
Vermaas, P. and Dieks, D. (1995). “The modal interpretation of quantum mechanics and its
generalization to density operators,” Foundations of Physics, 25: 145–158.
von Neumann, J. (1932). Mathematische Grundlagen der Quantenmechanik. Berlin:
Springer.
4
Quantum Physics Grounded on Bohmian Mechanics
nino zanghı̀
4.1 Copenhagen and the Measurement Problem

Quantum mechanics is one of the greatest intellectual achievements of the twenti-
eth century. Its laws govern the atomic and subatomic world and reverberate on a
myriad of phenomena, from the formation of crystals to superconductivity, from
the properties of low-temperature fluids to the emission spectrum of a burning
candle. However, as it is usually presented in textbooks, quantum mechanics is
basically a set of rules for calculating probability distributions of the results of any
experiment (in the domain of validity of quantum mechanics). As such, it does not
directly provide us with a description of reality. A description of reality should tell
us what there is in the world and how it behaves.
Whereas there is an almost general agreement on the correctness of the formal-
ism, the description of the reality that emerges from it remains controversial. It has
even been doubted whether such a description of reality should conform to the
rules of ordinary logic – and if any description is, after all, truly desirable. It has
also been argued that quantum theory forces us to abandon the reality of an
external world that exists objectively and independently of the human mind.
It is widely believed that between the end of the nineteenth and the beginning of
the twentieth century, physics underwent a radical change. Experimental know-
ledge about the atomic and subatomic world that was accumulating during that
period challenged the overall conceptual framework that physics had developed
from Galileo and Newton onward. Thus the idea was born that not only would it be
necessary to develop new theories that would replace classical mechanics and
classical electromagnetism, but that it was also necessary to abandon the classical
ideal according to which the laws of physics govern an external world
objectively given.
In the nineteen-twenties, the Danish physicist Niels Bohr, at that time probably
the most authoritative and influential quantum physicist, began to defend the idea
71
72 Nino Zanghì
that traditional scientific realism was childish and nonscientific, and he proposed
what it is still called the Copenhagen interpretation of quantum mechanics. On the
basis of this doctrine, the physical laws no longer have to do with the question of
how the world is made, but with our ability to know it, which is intrinsically
limited: The quantum mechanics of Copenhagen refuses in principle to provide a
consistent history of what happens to microscopic objects. From the point of view
of Copenhagen, reality is divided into two worlds, the microscopic and the
macroscopic, the classical and the quantum, the world regulated by classical
logic and the one regulated by quantum logic. Although it is not clear where the
boundary between these two worlds lies and how this duality can be compatible
with the fact that apples and chairs consist of electrons and other particles, the
Copenhagen doctrine has become orthodoxy. That is to say, it has become not only
the majority viewpoint among physicists, but also the dogma.
In more recent years, a version of quantum mechanics based on information
theory has grown in popularity. It is a dress that seems new, packed on the wave of
the theoretical and experimental development of quantum information and quan-
tum computation, but in reality it is a used dress, which was already tailored in
Copenhagen. Already in 1952, Schrödinger warned against the idea of reducing
quantum mechanics to a simple representation of our knowledge (Schrödinger
1995).
In spite of the pragmatic tribute reserved to the dogma, the peculiar role of the
observer in the formulation of the theory has always puzzled many physicists, as
can be seen for example, from the following considerations by Richard Feynman:
This is all very confusing, especially when we consider that even though we may
consistently consider ourselves to be the outside observer when we look at the rest of
the world, the rest of the world is at the same time observing us, and that often we agree on
what we see in each other. Does this then mean that my observations become real only
when I observe an observer observing something as it happens? This is a horrible
viewpoint. Do you seriously entertain the idea that without the observer there is no
reality? Which observer? Any observer? Is a fly an observer? Is a star an observer? Was
there no reality in the universe before 10⁹ B.C. when life began? Or are you the observer?
Then there is no reality to the world after you are dead? I know a number of otherwise
respectable physicists who have bought life insurance.
(Feynman, Morinigo, and Wagner 2003: 14)
Feynman is putting his finger on the most commonly cited conceptual difficulties
that plague quantum mechanics – the measurement problem, or what amounts to
more or less the same thing, the paradox of Schrödinger’s cat. The problem can be
rephrased as follows: Suppose that the wave function of any individual system
provides a complete description of that system. When we analyze the process of
measurement in quantum mechanical terms, we find that the after-measurement
Quantum Physics Grounded on Bohmian Mechanics 73
wave function for system and apparatus arising from the Schrödinger equation
typically involves a superposition over terms corresponding to what we would like
to regard as the various possible results of the measurement, e.g., different pointer
orientations. It is difficult to discern in this description of the after-measurement
situation the actual result of the measurement, e.g., some specific pointer orienta-
tion. In brief, quantum mechanics does not account for the obvious fact that
measurements do have results.
Bohr considered that philosophy was very important to understand quantum
mechanics and introduced the notion of complementarity, a many-purpose notion
good for solving the wave-particle duality, the measurement problem, and indeed,
all interpretative problems of quantum mechanics. This attitude sustained the idea
that with the problem of measurement we are facing a purely philosophical
problem. This idea was then nurtured and nourished by a sort of naive realism
about the operators – the idea that in quantum mechanics the observables and the
properties acquire a radically new, highly nonclassical meaning, reflected in the
noncommutative structure of the algebra of quantum observables.
4.2 Noncommutativity
The Hilbert space of quantum states is a vector space with a scalar product rule,
and it would be surprising if the operators on this space did not play an important
role in the formulation of quantum theory. And indeed, it is obviously so: The
temporal evolution of the states is given by a unitary operator that is generated by a
self-adjoint operator, the Hamiltonian. Not only are the temporal translations
governed by a self-adjoint operator, but so also are all the other symmetries of
the system. For example, the momentum operator is the generator of spatial
translations and the angular momentum operators govern the change of states as
a consequence of a rotation of the physical space. In quantum field theory, the
operators of creation and annihilation, operators that transform the state of a system
with a certain number of particles into another with a different number of particles,
play an extremely important role, as basic bricks of the Hamiltonian. In brief, linear
operators play an important role in quantum mechanics. And the main algebraic
feature of the operators is not to commute. So far, everything is clear and nothing is
mysterious.
The mystery arises when it is argued that the association of quantum observ-
ables with self-adjoint operators is to be considered a direct generalization of the
notion of classical observables and that quantum theory should not be conceptually
more problematic than classical physics once this fact is appreciated.
The classical observables – for a particle system, their positions, their momenta
and the functions of these variables, that is, functions on phase space – form a
74 Nino Zanghì
commutative algebra. It is generally taken to be the essence of quantization, the

procedure that converts a classical theory into a quantum theory, what makes the
corresponding operators to correspond to the positions and momenta and therefore
to all the functions of these variables. Thus the quantization leads to a noncom-
mutative algebra of observables, of which the usual examples are provided by
matrices and linear operators. In this way it seems perfectly natural that classical
observables are functions on phase space and quantum observables are self-adjoint
operators.
But in all this there is much less than meets the eye. What does it mean to
measure a quantum observable, a self-adjoint operator? It seems rather clear that
this should be specified – without such a specification it cannot have any meaning.
Therefore, one should be careful and use a more cautious terminology by saying
that, in quantum mechanics, the observables are associated with self-adjoint
operators, as it is difficult to see what more can be understood than an association.
What could it mean, an identification of the observables – considered as having, in
some way, an independent meaning as regards observation or measurement – with
a mathematical abstraction such as that of self-adjoint operators?
Note that it is important to insist on association rather than identification in
quantum theory, but not in classical theory, because in this case we begin with a
rather clear notion of observable (or property) which is well captured by the notion
of a function on the phase space, the state space of complete descriptions. If the
state of the system were observed, the value of the observable would of course be
given by this function of the phase point, but the observable might be observed by
itself, yielding only a partial specification of the state. In other words, measuring a
classical observable means determining to which level surface of the correspond-
ing function the state of the system, the phase point – which is at any time definite
though probably unknow – belongs. In the quantum realm the analogous notion
could be that of function on Hilbert space, not that of self-adjoint operator. But we
don’t measure the wave function (the nonmeasurability of the wave function is
related to the impossibility of copying the wave function. This question arises
sometimes in the form, “Can one clone the wave function? [Ghirardi personal
communication; see Wooters and Zurek 1982, Ghirardi and Weber 1983]) so that
functions on Hilbert space are not physically measurable, and thus do not define
“observables.”
4.3 Contextuality
A milestone in the foundations of quantum mechanics is Bell’s nonlocality analy-
sis (Bell 1964). It has a by-product that is interesting in itself: The incompatibility
of Bell’s inequality with the predictions of quantum mechanics is a demonstration
of the impossibility that self-adjoint operators can be interpreted as a way of

representing the properties of an object.
There are other demonstrations of this impossibility in addition to that of Bell in
1964. The best known are due to von Neumann (1932), Gleason (1957), and
Kochen and Specker (1967). These theorems are usually called “impossibility
theorems for hidden variables,” a name that seems to suggest that what the
theorems show would be the impossibility of completing the quantum mechanics
through additional variables that determine, together with the quantum state, the
result of the experiments. But this interpretation of the theorems is completely
erroneous and is clearly disproved by the existence of a theory like Bohmian
mechanics (discussed later), in which the additional variables are simply the
positions of the particles that make up an object, and the results of the experiments
turn out to be determined by the complete specification of the quantum state and
particle positions (discussed later). A more appropriate term for these theorems
would perhaps be “theorems of impossibility of naive realism about self-adjoint
operators” (Daumer, Dürr, Goldstein, and Zanghì 1996).
A substantial difference between Bell’s theorem and other impossibility the-
orems is, that while Bell’s theorem stems from a question with a clear physical
meaning – is the locality principle compatible with quantum mechanics? – it is not
at all clear to what question the theorems of impossibility for hidden variables
correspond. If Bohr’s warning that the fundamental lesson of quantum mechanics
is to recognize the impossibility of a sharp division between the behavior of atoms
and the interaction with the measurement apparatus is kept in mind, it would not be
hard to recognize that the hypotheses on which the theorems are based are in clear
contrast with the experimental principles of quantum mechanics.
But the halo of mystery is hard to erase, and this has led to the formation of a
new myth: contextuality. According to contextuality, in quantum mechanics prop-
erties acquire a new, highly nonclassical meaning; quantum properties depend on
other compatible properties – if they exist – that are measured at the same time. All
this sounds very mysterious, but if one remembers the active role of quantum
measurements, the mystery will disappear. Let us consider, for example, the
operator A that commutes with the operators B and C which do not commute
among them. What is often called the result of A in an experiment for the
measurement of A and B together, usually differs from the result of A in an
experiment for the measurement of A and C together, because these experiments
are different from each other in the sense that they act on the state of the system in a
very different way. The misleading reference to measurement, with the associated
naive realism about operators, makes the context appear much more mysterious
than it is. The same terminology is misleading and fails to transmit with due force
the peremptory character of what it brings: Properties that are uniquely contextual
76 Nino Zanghì
are not properties at all, do not exist, and their inadequacy to carry out the role of
properties is in the strongest sense possible.
In short, contextuality means nothing more than the fact that the result of an
experiment depends on the experiment itself, and this applies equally to both
classical physics and quantum physics: For any experiment, be it classical or
quantum, it would be a mistake to assume that any device involved in the experi-
ment plays only a passive role, unless the experiment is not the genuine measure-
ment of a property of the system, in which case the result is determined by the
initial condition of the system only. In classical physics, it is traditionally assumed
that it is in principle possible to measure any property without sensitively disturb-
ing the measured object, but this is false in quantum mechanics – and should be
questioned in classical physics, too.
4.4 The Classical Variables of Bohr

It is useful to go back to Bohr’s solution of the measurement problem. For Bohr, it
is in principle impossible to formulate the fundamental concepts of quantum
mechanics without using classical mechanics. To put it in the words of Landau
and Lifshitz
quantum mechanics occupies a very particular position in the realm of physical theories: it
contains classical mechanics as a limiting case, and at the same time needs this limit case
for its foundation.
(Landau and Lifshitz 1958)
So, the orthodox vision ends up providing a response to the problem of measure-
ment that many orthodoxy enthusiasts still struggle to accept: the wave function
does not provide a complete representation of the state of affairs of the world; in
addition, you need to specify the values of classic variables – which for conveni-
ence will be named here “Z-variables.”
According to Bohr, the Z-variables are precisely those that establish quantum
mechanics and make the quantum formalism coherent and applicable to the
study of the phenomena we observe in the laboratory or in nature. In other
words, according to the orthodox interpretation, the complete description of a
state of affairs of the world is given by the pair ðΨ; Z Þ, where Ψ is the wave
function and the Z are in some sense macroscopic classical variables. So,
although according to the orthodox view Ψ does not represent anything real –
thus the famous motto “there is no quantum world, there is only an abstract
quantum description” – the role of the wave function is to govern the statistics
of the Z-variables that indeed do represent what is to be considered real, or at
least “concrete.”
As pointed out by John Bell (1990) in the original formulation of the Copen-
hagen interpretation, the sense in which the complete description of a state of
affairs of the world is given by the pair ðΨ; Z Þ is not clearly specified. In addition,
the dynamics of the pair is not specified in a clear and unambiguous way:
Sometimes the dynamics of the wave function is given by the Schrödinger
equation and sometimes the dynamics of the macroscopic variables is that fixed
by the laws of classical mechanics and classical electromagnetism. However, when
the classical variables interact with the quantum variables, the dynamical laws
change: The wave function no longer evolves according to the Schrödinger
equation, instead it evolves according to the collapse rule and the Z-variables
undergo random leaps that are statistically governed by the wave function.
The difficulty raised by Feynman in the passage quoted in Section 4.1 might be
so rephrased: Where is the borderline between what is classical and what is
quantum? When can we treat an object as classical and when must we treat it as
quantum? In other words, the distinction between microscopic and macroscopic, as
well as that between the classical world and the quantum world, lacks a precise
definition and introduces a fundamental ambiguity that cannot have any place in
any theory that claims physical accuracy.
4.5 The Classical Variables of Bohm

If we must anyhow pay the price of incompleteness, why should we also pay the
surcharge of fuzziness and ambiguity? Why not supplement the quantum descrip-
tion provided by the wave function with variables that are well defined on all scales
and not simply on the macroscopic one? This was the conviction of Albert
Einstein:
I am, in fact, rather firmly convinced that the essentially statistical character of
contemporary quantum theory is solely to be ascribed to the fact that this (theory)
operates with an incomplete description of physical systems.
(Einstein, in Schilpp 1949: 666)
Indeed, if the completion is achieved in what is really the most obvious way – by
simply including the positions of the particles of a quantum system as part of the
state description of that system, allowing these positions to evolve in the most
natural way – one arrives at the theory developed by David Bohm (1952). This
theory is nowadays known as Bohmian mechanics, de Broglie-Bohm’s theory, the
wave-pilot theory, or the causal interpretation of quantum mechanics.
In the theory proposed by Bohm, a particle system is described in part by its
wave function Ψ, which evolves, as usual, according to the Schrödinger equation.
However, the wave function only provides a partial description of the system. This
78 Nino Zanghì
description is complemented by the specification of the real positions of the

particles Q ¼ ðQ1 ; . . . ; QN Þ. Their evolution is governed by a guiding equation
that expresses the velocities of the particles in terms of the wave function (dis-
cussed later). Thus, in Bohmian mechanics the configuration Q of a particle system
evolves according to a movement that is somehow “choreographed” by the wave
function. In Bohmian mechanics, the complete state description of a system is
provided by the pair ðΨ; QÞ. The entire quantum formalism, including the uncer-
tainty principle, quantum randomness, the quantum statistics for identical particles,
and the role of operators as observables, emerges from an analysis of the dynam-
ical system ðΨ; QÞ.
With a theory such as the Bohmian mechanics, in which the description of the
situation after a measurement includes, in addition to the wave function, the values
of the variables that record the result, there is no problem of measurement. In
Bohmian mechanics the pointers of the measurement devices always have a well-
defined orientation: The particles that form the pointer of an apparatus always
have, according to this theory, a well-defined configuration, and the macroscopic
appearance of such a configuration is precisely that of a pointer that points in a
well-defined direction.
As such, Bohmian mechanics is a counterexample to the claim that quantum
mechanics forces us to abandon the idea of an objective external world, which
exists independently of the human mind. It is a “realistic” quantum theory, and
since in its formulation no reference is made to “observers,” it is also a “quantum
theory without observers.” For historical reasons it has been called a “hidden
variables theory.” The existence of Bohmian mechanics shows that many of the
radical epistemological consequences, which usually many physicists and philoso-
phers have drawn from quantum mechanics, are groundless. It shows that there is
no need for contradictory notions such as “complementarity,” that there is no need
to imagine an object as if it were simultaneously in different places, or that the
physical quantities have values that are somehow “blurred,” that is, intrinsically
undefined. Finally, it shows that there is no need for human consciousness to
intervene in physical processes, for example, to collapse the wave function.
Bohmian mechanics solves all the paradoxes of quantum mechanics, eliminating
oddities and mysteries. It is important to stress that the Q-variables of a Bohmian
theory need not be configurations of particles. These variables may represent
geometry, field or string configurations, or whatever is needed to describe
nature best.
Starting from 1952, Bohm’s theory has been investigated and refined. Various
ways have been proposed to extend Bohmian mechanics to quantum field theory.
One of them (Bohm 1952), for bosons (i.e., for the quantum fields of force), is
based on field configurations on the three-dimensional physical space that evolve
over time guided by a wave functional according to a similar dimensional infinity

of the guiding equation. Another proposal (Dürr, Goldstein, Tumulka, and Zanghì
2004) is based on Bell’s seminal work (1986) and assigns trajectories to electrons
and to any sort of particles that intervene in a given field quantum theory; however,
in contrast to the original Bohmian mechanics, this proposal involves a stochastic
dynamics according to which particles can be created and destroyed.
4.6 The Equations of Motion

For a nonrelativistic system of N particles, Bohmian mechanics is defined by two
equations of motion: the Schrödinger equation
∂Ψ
iℏ ¼ HΨ (4.1)
∂t
for the wave function Ψ, and the guiding equation
dQk ℏ Ψ∗ rk Ψ
¼ Im ∗ ðQ1 ; . . . ; QN Þ, k ¼ 1, . . . , N (4.2)
dt mk Ψ Ψ
for the configuration Q ¼ ðQ1 ; . . . ; QN Þ of the particles. The guiding equation is
indeed the simplest law of evolution of the first order for the position of the
particles, compatible with the Galilean symmetry of the Schrödinger evolution
and with time reversal (Dürr, Goldstein, and Zanghì 1992).
A few clarifications on the symbols that appear in Eqs. (4.1) and (4.2): ℏ
denotes, as usual, the Planck constant divided by 2π; H is the usual nonrelativistic
(Schrödinger) Hamiltonian, which contains the particle masses mk , k ¼ 1, . . . , N,
and for a system of spinless particles has the simple form
XN
ℏ2 2
H¼ r þU (4.3)
k¼1
2mk k
with U the corresponding classical potential energy; rk is the gradient with respect
to the coordinates of the k-th particle; for a complex-valued wave function Ψ, Ψ∗
denotes its complex conjugate; if Ψ has spinorial values, the products in the
numerator and in the denominator of the guiding equation should be understood
as scalar products in the spinor space; if external magnetic fields are present, the
gradient should be understood as a covariant derivative, involving the potential
vector; “Im” means taking the imaginary part of a complex number.
To gain familiarity with the guiding equation, it is useful to consider the
extremely simple case of a single free particle of mass m guided by a monochro-
matic wave with wave vector k (and thus with wavelength λ ¼ 2π=k), ideally
80 Nino Zanghì
approximated by the plane wave Ψðq; tÞ / eiðk:qωtÞ , ω ¼ ℏk2 =ð2mÞ. The l.h.s. of
the guiding equation is the velocity v of the particle. A simple calculation shows
that the right-hand side r.h.s. of the guiding equation is ðℏ=mÞk. Thus the guiding
equation of Bohmian mechanics turns out to be precisely the relation p ¼ ℏk
which de Broglie proposed in late 1923 and which quickly led Schrödinger, during
the end of 1925 and the beginning of 1926, to the discovery of his wave equation.
Now consider the guiding equation in the case of two particles. The wave
function Ψ ¼ Ψðq1 ; q2 ; tÞ generates the possible speeds of the two particles, i.e.,
ℏ Ψ∗ ðq1 ; q2 ; tÞr1 Ψðq1 ; q2 ; tÞ
v1 ¼ Im (4.4)
m1 Ψ∗ ðq1 ; q2 ; tÞΨðq1 ; q2 ; t Þ
for particle 1 and
ℏ Ψ∗ ðq1 ; q2 ; tÞr2 Ψðq1 ; q2 ; tÞ
v2 ¼ Im (4.5)
m2 Ψ∗ ðq1 ; q2 ; tÞΨðq1 ; q2 ; t Þ
for particle 2. These formulas show that the velocity of a particle, at a certain
instant of time, depends in general on where, at the same time, the other particle
is. An exception is the case in which the wave function factorizes, i.e., is of the
form Ψ ¼ Ψðq1 ; t ÞΨðq2 ; t Þ. In this case, the velocity of particle 1 does not
depend on the position q2 of particle 2, and vice versa. However, for a general
wave function, i.e., for an entangled quantum state, there is a corresponding
entanglement of the velocities which persists, without attenuating in any way,
even when the distance between the two particles is very large. This important
property of the guiding law shows that the velocity of a particle can depend on
the position of the other even when the module of the wave function is very
small (which, in fact, is what happens when the distance between the two
particles is very large).
Thus, in Bohmian mechanics, the law that governs the motion of particles in
physical space makes manifest the most dramatic effect of quantum mechanics,
quantum non locality – the fact that physical events can mutually influence each
other more quickly than the speed of light, even at arbitrarily large distances,
without this mutual influence being mediated by physical fields (such as, for
example, the electromagnetic or gravitational field) or by particles (or energy or
signals) that can somehow travel from one event to another. Far from being a
defect, it is a remarkable merit of the theory, since, as Bell has shown,
nonlocality is a basic property of nature. One could say that Bohmian mechanics
does nothing but to inherit and to make explicit the nonlocal character implicit in
the notion, common to all the formulations and interpretations of quantum
mechanics, of a wave function on the configuration space of a system with many
particles.
Many objections have been and continue to be raised against Bohmian

mechanics. Some of them arise from the lack of understanding that the structure
of Bohmian mechanics cannot be interpreted in classical terms, for example, in
terms of force or in terms of conservation laws, such as energy or momentum,
or in terms of fields that propagate in physical space. Moreover, the wave
function, which is part of the description of the state of a Bohmian universe
(and therefore part of the reality of this universe) is not the sort of field on the
physical space to which we are accustomed in classical physics, but a field on
the space of all possible configurations, and the role of the wave function in
Bohmian mechanics is that of determining the velocities of the particles, and
therefore, of making possible the formulation of a law of motion for stuff in
physical space.
4.7 Bohm’s Original Formulation

Unfortunately, Bohm chose the conceptual framework of classical mechanics to
formulate his theory, which contributed, at least in part, to some of the objections
we have just mentioned. Let us see how Bohmian mechanics can be dressed in
classical clothes.
If the two sides of the guiding Eq. (4.2) are derived with respect to time (and in
calculating the derivatives with respect to the time of the r.h.s., one uses the
Schrödinger equation), one obtains the equation
d 2 Qk
mk ¼ rk U rk W (4.6)
dt 2
where W is a certain function of the positions of all the particles, which is
determined uniquely by the square of the wave function in a rather complicated
way that, for the purposes of the present discussion, it is not necessary to make
explicit; we just mention an important property of this function: If the wave
function is multiplied by a constant, the value of W does not change. Bohm called
the function W quantum potential (energy).
Eq. (4.4) has a Newtonian form: Its left-hand side (l.h.s.) is the second derivative
with respect to the time of the position of the k-th particle, i.e., its acceleration,
multiplied by the mass of the particle. Therefore, it has the structure of the first
member of the classical Newton equation (mass acceleration). Now consider the
r.h.s. of the equation: U is the potential energy of N interacting particles that
appears in the Schrödinger equation, a function that has exactly the same form as in
the corresponding classical situation. Thus the first term on the right is actually the
force, derived from the potential energy U, which in the corresponding classical
situation would act on the k-th particle. If we interpret the second term as a force
82 Nino Zanghì
derived from the quantum potential W, the r.h.s. of Eq. (4.6) can be interpreted as a
force – a force of classical origin added to a force of quantum origin.
However, this formulation of Bohmian mechanics in classical clothes has a cost.
The most obvious cost is an increase in complexity: The quantum potential is
neither simple nor natural. Moreover, it does not seem very satisfactory to think
that the quantum revolution is reduced to the understanding that, after all, nature is
classical, with an additional force term, all in all fairly ad hoc, the term that
corresponds to the quantum potential.
Furthermore, Bohmian mechanics is not simply a reformulation of quantum
mechanics in classical terms with an additional force term. In Bohmian mechan-
ics, the velocities are not independent of the positions, as in the classical case, but
are constrained by the guiding equation. Despite the apparent form of the second
order of Newton’s quantum equation, its solutions are not characterized by
positions and velocities at some initial time, but only by the initial positions,
because the initial speeds are determined by the guiding law at the initial time. In
other words, having derived the members of the guiding equation with respect to
time, and thus obtained the quantum Newton equation, does not change the
nature of the theory; at most it makes it more complicated and less transparent.
It is as if in classical mechanics we derived both members of the Newton
equation and decreed that the equation of the third order thus obtained should
be considered the classical law of motion. In this way we would not obtain a
genuinely different theory, that is, a third-order theory, with positions, velocities,
and accelerations as initial independent conditions, because the initial acceler-
ations would in any case be bound by the Newton equation to be functions of
initial positions and velocities.
Because the dynamics of Bohmian mechanics is completely defined by the
Schrödinger equation and the guiding equation, there is no need for further axioms
involving a quantum potential. Therefore the quantum potential, together with the
quantum Newton equation in which it appears, should not be considered funda-
mental. The correct way to look at Bohmian mechanics is as a first-order theory, in
which the fundamental quantities are the positions of the particles, whose dynam-
ics is specified by the guiding equation. Second-order concepts, such as acceler-
ation and force, work and energy, play no fundamental role in the theory. The
artificiality suggested by the quantum potential is the price you pay when dressing
a highly nonclassical theory with classical clothes. This does not mean that these
second-order concepts cannot play any role in Bohmian mechanics. These are
emerging notions, which are fundamental for the theory to which Bohmian
mechanics converges in the classical limit – that is, Newtonian mechanics: When
the quantum force is negligible, one has in effect classical behavior (Allori, Dürr,
Goldstein, and Zanghì 2002).
4.8 The Classical Variables of Bohr out of the Classical Variables of Bohm
The statistical character of quantum theory was first fully acknowledged in
1926 by Max Born, shortly after Schrödinger discovered his famous equation.
Born interpreted Schrödinger’s function in a statistical sense and postulated that
the configuration Q of a quantum system is random, with probability distribution
given by the density jΨðqÞj2 . Under the influence of the growing consensus in
favor of the Copenhagen interpretation, the density jΨðqÞj2 began to be considered
the probability of finding the configuration Q, whatever it was measured, rather
than the probability that the configuration was really Q, a notion that was believed
to be meaningless. In accordance with these quantum probabilities, measurements
performed on a quantum system with a defined wave function typically provide
random results.
The density jΨðqÞj2 takes on a particular importance in Bohmian mechanics. As
an elementary consequence of the Schrödinger equation and the guiding equation,
it is equivariant, in the sense that these equations are compatible with respect to the
distribution jΨðqÞj2 . More precisely, this means that if the configuration Q of a
system is random, with distribution jΨðqÞj2 at some time, then this will also be true
for any other time. This distribution is therefore called the quantum equilibrium
distribution.
A Bohmian universe, although deterministic, evolves in such a way that an
appearance of randomness emerges, precisely as described by quantum formalism.
In order to understand how this comes about, one should realize that, in a world
governed by Bohmian mechanics, the measuring apparatuses are also made of
particles. In a Bohmian universe, the measurement apparatus, tables, chairs, cats,
and other objects of our daily experience are simply agglomerations of particles,
described by the positions they occupy in physical space, and whose evolution is
governed by Bohmian mechanics.
The following theorem is crucial:
Theorem 1 Bohmian mechanics provides the same predictions of quantum mech-

anics (assuming the same Schrödinger equation for both) for the results of any
experiment carried out on a system with a wave function ψ, if at the beginning of
the experiment the configuration Q of the largest system required for the analysis of the
experiment is random, with probability density given by the quantum equilibrium
distribution, i.e., the probability distribution that quantum mechanics prescribes for Q.
The “largest system required for the analysis of the experiment” means the
composite system that includes the system on which the experiment is performed,
as well as the measuring apparatus and all the other instruments used in the
execution of the experiment, together with all the other systems that have signifi-
cant interaction with them during the experiment.
84 Nino Zanghì
If we denote the configuration of the particles forming the measured system by

X, and the configuration of all the rest with Y, we have the natural decomposition
Q ¼ ðX; Y Þ. For example, we can think that X represents the positions of the
electrons, protons, and neutrons that form a silver atom, while Y represents the
positions of the electrons, protons, and neutrons that make up the magnets and
the screen of a Stern-Gerlach apparatus.
The assumption that the measured system has a well-defined initial wave
function ψ ðxÞ can be guaranteed by assuming that the initial wave function of
the large system is of the product form ΨðqÞ ¼ ψ ðxÞΦ0 ðyÞ. This condition guaran-
tees the initial independence of the system to be measured and the apparatuses and
instruments used in the experiment. In particular, we can think that the function
Φ0 ðyÞ represents the initial READY state of the apparatus. The assumption that Q
is initially in quantum equilibrium means that the initial randomness of Q ¼ ðX; Y Þ
is governed by the probability density jψ ðxÞj2 jΦ0 ðyÞj2 , that is, the probability
distribution prescribed by quantum mechanics for the initial configuration.
Precisely because of this last assumption, the proof of the theorem is trivial,
being nothing but an immediate consequence of the equivariance property. In fact,
the initial configuration is transformed, by means of the guiding equation of the
large system, in the final configuration at the end of the experiment. But, on the
basis of the hypothesis of quantum equilibrium – that the initial configuration of
the large system is random, with distribution given by jΨðqÞj2 – the final configur-
ation of the large system, which includes, in particular, the orientation of the
measuring devices, will be distributed according to the distribution jΨðqÞj2 at the
final time. Thus, the macroscopic variable Z describing the result of the experi-
ment, which is a function Z ¼ f ðQÞ of the final configuration Q of the system and
apparatus, have in Bohmian mechanics and in quantum mechanics exactly the
same statistical distribution. So, whenever the predictions of quantum mechanics
for the results of a certain experiment are unambiguous – ambiguity in orthodox
formulation arises, inter alia, from the absence of a well-defined microscopic
dynamics – these predictions must necessarily coincide with those of Bohmian
mechanics. It is important to be clear that, as far as the measured system is
concerned, we are not just talking about position measurements, but about any
measurement or experiment, for example, a spin measurement. Theorem 1 estab-
lishes that Bohmian mechanics provides the same predictions of quantum mech-
anics for the measurement of any quantum observable.
Both the statement of the theorem and its proof contain some implicit assump-
tions that we want to bring to light and comment. First, we assumed that the result
of any measurement or experiment can be, at least potentially, registered in terms
of a macroscopic configuration Z ¼ f ðQÞ. However, this is in complete agreement
with the experimental practice and with any interpretation of quantum mechanics,
including Bohr’s interpretation, in which the macroscopic configurations of objects

(and their classical behavior) are assigned a privileged status. A more delicate
assumption concerns the initial assignment of a wave function ψ ¼ ψ ðxÞ to the
system. Whereas in ordinary quantum mechanics one can always appeal to an
external observer who somehow “prepares” the initial wave function, in a theory in
which observers play no privileged role, and in particular, in a strictly deterministic
theory such as Bohmian mechanics, the situation seems more problematic.
Nevertheless, this problem has been addressed and solved (Dürr et al. 1992).
The main ingredients for its resolution are substantially three. The first ingredient,
of a technical nature, concerns the clarification of the notion of random system in a
deterministic theory (a clarification of the conditions according to which the
identity of a particular subsystem and the time in which it is identified are reflected
in the external environment of the system, and therefore, in the last instance, in the
initial conditions). The second ingredient concerns the ways in which, in Bohmian
mechanics, it is possible to assign a wave function to a subsystem of a larger
system – what in Dürr et al. (1992) has been called the conditional wave function.
Finally, the third ingredient concerns precisely the clarification of one of the crucial
assumptions of the theorem: The hypothesis of quantum equilibrium, i.e., the
assumption that, if the wave function is Ψ ¼ ψ ðqÞ, then the configuration Q is
random with probability distribution jΨðqÞj2 . Justifying this hypothesis of quantum
equilibrium is in fact a rather delicate matter, a problem that has been investigated
in a very detailed manner (see Dürr et al. 1992) by showing that the probabilities
for the positions given by the quantum equilibrium distribution jΨðqÞj2 emerge
naturally from an equilibrium analysis of the deterministic dynamical system
defined by Bohmian mechanics, in roughly the same way that the Maxwellian
velocity distribution emerges from the equilibrium analysis of a gas.
4.9 Classical Variables and Operators

Since orthodox quantum theory provides us with a vast class of observables and
properties in addition to particle positions, it would seem, at first glance, that this
theory is much richer than Bohmian mechanics, which seems to relate exclusively
to positions. It turns out, however, that the quantum observables, represented by
self-adjoint operators, arise from an analysis of quantum experiments. The key
theorem is the following (Dürr, Goldstein, and Zanghì 2004):
Theorem 2 Consider an experiment carried out on a system with a wave function,

and in addition to the hypotheses of Theorem 1, assume that the experiment is
reproducible. Then, under minimal technical conditions, there exists a self-adjoint
operator A that regulates the statistics of the results Z ¼ f ðQÞ of the experiment in the
86 Nino Zanghì
manner prescribed by quantum mechanics. In particular, the average value of the

results is given by the usual formula hZ i ¼ hψ; Aψ i.
The condition of reproducibility means that the quantum experiment should give
the same result if immediately repeated (after having brought the apparatus back to
its READY state). It should be noted that if it is not assumed that the experiment is
reproducible, the statistics of the experiment is given by a more general operator
structure than that of the self-adjoint operators, namely that of generalized quantum
observables that has been introduced to describe quantum measurements in par-
ticular situations encountered in quantum optics, and more generally, in the theory
of open quantum systems (these are the positive-operator-valued measures
[POVM], also used in quantum information). Thus, Bohmian mechanics provides
an immediate understanding of this generalization of the notion of quantum
observable and a clarification of the type of idealization involved in the notion of
“operator as observable.”
The Stern-Gerlach experiment is particularly illuminating to clarify the content
of Theorem 2. In the Stern-Gerach experiment, as a consequence of the interaction
with the magnetic field, the parts of the wave function that are in the different
autospaces of the relevant spin operator (for example, the component along z)
become spatially separated and the particle (the silver atom), which moves
according to the guiding equation, ends up being in the support of only one of
these parts. The end result (“up” or “down”) is therefore a function of the final
position of the particle, which is revealed on the screen. Of this position, what we
can predict is only that it is random and distributed according to jΨðqÞj2 at the final
time. By calibrating the results of the experiment with numerical values, for
example +1 for a top detection and 1 for a bottom detection, it is not difficult
to show (by solving the Pauli-Schrödinger equation and the guiding equation for
this situation) that the probability distribution of these values is expressed in terms
of the usual quantum spin operators – the Pauli matrices.
It is important to observe that, since the results of a Stern-Gerlach experiment
depend not only on the position and initial wave function of the particle, but also
on the choice of the different possible magnetic fields that could be used to
measure the same spin operator, this experiment is not a genuine measurement in
the literal sense; that is, it does not reveal a preexisting value associated with the
spin operator. Indeed, there is nothing mysterious or nonclassical about the non-
existence of such values associated with the operators. Bell said that (for Bohmian
mechanics) spin is not real. Perhaps he should have said, “even spin is not real,”
not simply because among all the quantum observables, the spin is considered to
be the paradigmatic quantum observable, but also because spin is treated in
orthodox quantum mechanics in a fairly similar way to position, as a “degree of
freedom,” a discrete index that completes the continuous degrees of freedom

corresponding to the position. Be that as it may, its fundamental significance is
this: Unlike the position, spin is not primitive, i.e., it should not be added to the
state description as a real degree of freedom, analogous to the real positions of the
particles. Speaking roughly, spin is simply in the wave function. At the same time,
as we have just shown, spin measurements are completely clear and simply reflect
the role played by the spin wave functions in the particle motion description.
Moreover, Theorem 2 provides a clarification of the content of the impossibility
theorems for hidden variables. Let us look at the obvious correspondence between
the experiments and the Z-variables that represent the results of the experiments
themselves (assuming, as for the spin, that the measurement instrument does not
introduce any external randomness and therefore that the Z-statistics is given by a
self-adjoint operator on the Hilbert space of the system on which the experiment is
performed): EXPERIMENT!Z. Moreover, since Theorem 2 establishes that a
(reproducible) experiment corresponds to a self-adjoint operator that describes the
statistics of the results, there is a correspondence between these variables and the
self-adjoint operators: Z!OPERATOR. But there is no reason to expect this
correspondence to be invertible, that is, that the correspondence OPERATOR!Z
exists, which is the premise of the theorems of nonexistence of hidden variables.
Therefore, the widespread idea that in a realist quantum theory all quantum
observables should possess real and preexisting values was ab initio an idea not as
reasonable as it might seem, but rather the fruit of the prejudice of having taken the
operators too seriously – an attitude almost implied in the words ‘observable’ or
‘property.’
4.10 The Collapse of the Wave Function

According to the quantum formalism, an ideal quantum measurement performed
on a quantum system causes a jump or “collapse” of its wave function into one of
the eigenstates of the measured observable. But while in orthodox quantum
mechanics collapse is simply superimposed on the unitary evolution of the wave
function, without a specification of the circumstances in which it can legitimately
be invoked – and this ambiguity is nothing but another facet of the problem of
measurement – Bohmian mechanics incorporates both the unitary evolution and
the collapse of the wave function as appropriate.
This claim might seem puzzling, since in Bohmian mechanics the wave function
evolves unitarily according to the Schrödinger equation. However, since any
observation implies interaction, a system under observation cannot be an isolated
system, but must be a subsystem of a larger system that is isolated (for example,
88 Nino Zanghì
the entire universe). Additionally, there is no a priori reason that a subsystem of a

Bohmian universe is itself a Bohmian system – a system governed by its wave
function that evolves according to its Schrödinger equation – even if the system is
to some extent isolated. In truth, it is not even clear a priori what should be
interpreted by the wave function of a subsystem of a Bohmian universe.
Recall the comments after Theorem 1: The configuration Q of the largest
system – the Bohmian universe – naturally separates into X, the configuration of
the subsystem, and Y, the configuration of his environment. Let
Ψ ¼ ΨðqÞ ¼ Ψðx; yÞ be the wave function of the universe. According to Bohmian
mechanics, this universe is completely described by Ψ, which evolves according to
the Schrödinger equation, and by the X and Y configurations. There is therefore a
fairly natural choice of what should be considered the wave function of the
subsystem: It is the conditional wave function ψ ðxÞ ¼ Ψðx; Y Þ, which is obtained
by inserting the actual configuration of the environment in the wave function of the
universe. (Note that orthodox quantum mechanics lacks the resources necessary to
define the conditional wave function. From an orthodox point of view it is not at all
clear what should be understood by wave function of a subsystem). Consider now
how the conditional wave function depends on time, ψ ðx; tÞ ¼ Ψðx; Y ðt Þ; t Þ. It is
not difficult to show that ψ ðx; tÞ obeys the Schrödinger equation for the subsystem
when this is appropriately decoupled from its environment, and that using the
hypothesis of quantum equilibrium, it undergoes, when the appropriate conditions
are realized, processes of collapse, that is, random transformations in complete
agreement with the rules of the quantum formalism. In Bohmian mechanics, the
rule of the collapse of the wave function is a theorem and not an axiom as in
orthodox quantum mechanics.
4.11 The Paradox of Schrödinger’s Cat

In Bohmian mechanics the cat is alive or dead at any time, regardless of whether it
is observed or not. Even if its wave function is in a superposition
1
ψ ¼ pffiffiffi ½ψ dead ðxÞ þ ψ alive ðxÞ (4.7)
2
the cat is an agglomeration of particles, described by their actual configuration X,
which determines its microscopic and macroscopic properties at each moment –
first of all, its property of being alive or being dead. The two wave functions of the
superposition discussed earlier have macroscopically disjoined supports, say M alive
and M dead , which create a natural partition of the configuration space of the cat
M ¼ M alive [ M dead ; the actual configuration X of the cat is either in M alive or in
M dead , i.e., the cat is alive or is dead.
If at certain time we have that X 2 M dead (as it results from appropriate condi-
tions of the external environment Y, for example, from the temperature value of the
cat measured with a thermometer), the relevant wave function for the subsequent
temporal evolution of the configuration of the cat is ψ dead ðxÞ. In other words, the
guiding law provides the same temporal evolution for the cat configuration
whether we use the whole wave function ψ, or just its ψ dead ðxÞ. Using only the
latter, that is, applying the rule of collapse, is therefore, in Bohmian mechanics,
only a practical matter that does not change anything of what actually happens in
the world – the actual history of the cat and, more generally, of the physical
systems that, like cats, populate the world.
The following question arises: Can a dead cat become alive in the future? What
is it that forbids the X configuration, originally in M dead , to enter at a certain
moment of the future in the region M alive ? Strictly speaking, nothing. In Bohmian
mechanics, such an event is possible, although the probability of its occurrence is
frighteningly small, ridiculously small. However, the mere possibility that such an
event is realized does not involve any interpretative problem for the theory. On the
contrary, it only emphasizes that the explanatory structure of Bohmian mechanics
is quite analogous to that of classical statistical mechanics.
In fact, an event of this type, extraordinary and highly improbable, is also
possible in the dynamical scheme provided by classical mechanics, as it is possible
that all the air in this room spontaneously leaves the window and I die from
suffocation, and if it were possible to imagine a cat such that the particles that
compose it were governed by the laws of classical mechanics, this hypothetical
dead cat could return alive. Events of this type are possible, but highly improbable
because they would entail a decrease in the entropy of the universe. The explan-
ation of the impossibility for the cat to return alive is, in Bohmian mechanics,
exactly of the same type: It is guaranteed by simple entropic reasons, that is, by the
typical behavior of the physical systems according to the second law of
thermodynamics.
The difference with the classical case concerns only how the macroscopic
thermodynamic laws are stabilized by the underlying microscopic dynamics. In
the quantum case, quantum entanglement plays a very important role in the process
of stabilizing macroscopic properties of physical systems and makes (if ever
needed) it extremely unlikely that dead cats can resurrect. In fact, even if the
regions M alive and M dead were not macroscopically disjoint, but only sufficiently
disjoint, the interaction of the system with the external environment and the
consequent formation of entangled states in superpositions containing many terms,
would produce a partition of the configuration space of the larger system (which
includes the starting system and its external environment) in distinct regions, each
in a two-way correspondence with a term of the superposition of the wave function
90 Nino Zanghì
of the largest system. These regions would be macroscopically disjoint, and the
greater the number of external environment systems that come into play, the
greater the macroscopic separation that is achieved, and thus the greater the degree
of irreversibility of the process.
4.12 Relativity
Many of the objections to Bohmian mechanics – if not the overwhelming major-
ity – are very weak and often arise from a gross misunderstanding of the theory.
Some of them arise from the lack of understanding that Bohmian mechanics should
be considered, mathematically and conceptually, as a theory profoundly different
from Newtonian mechanics.
The most serious objection raised against Bohmian mechanics is that this theory
does not account for those phenomena, such as the creation and destruction of
particles, which are characteristic of the quantum theory of fields. Actually, this is
not, in itself, an objection to the Bohmian mechanics, but simply the observation
that quantum field theory, for what concerns what happens in the physical world,
explains much more than the nonrelativistic theory, be it in the orthodox or the
Bohmian form. However, this objection has the merit of highlighting how import-
ant and necessary finding an adequate Bohmian version of quantum field theory is.
We have already mentioned proposals in this direction, involving a stochastic
dynamics according to which particles can be created and destroyed (Dürr, Gold-
stein, Tumulka, and Zanghì 2004).
An objection in some way connected to the previous one is that Bohmian
mechanics cannot be made invariant under Lorentz transformations, with what,
presumably, it is meant that it is not possible to find any Bohmian theory – a theory
that could be considered as a natural extension of Bohmian mechanics – that is
invariant under Lorentz transformations. If it were correct, this objection should be
seriously considered. However, it is not supported by any argument that makes
plausible the alleged impossibility of finding an extension of the Bohmian mech-
anics invariant under Lorentz transformations. The reason for this widespread
belief is the manifest nonlocality of Bohmian mechanics. But, as Bell has shown,
nonlocality is a fact established by the experiments.
Moreover, as regards the equally widespread belief that conventional quantum
theories would have no difficulty in incorporating Einstein’s relativity, whereas
Bohmian mechanics would have it, there is, even in this case, much less than meets
the eye. One should always bear in mind that the empirical content of conventional
quantum mechanics is based on (1) the unitary evolution of the state vector (or the
equivalent unitary evolution in the Heisenberg representation) and (2) the collapse
or reduction of the wave function (or any other equivalent artifact that allows the
effect of observations and measurements to be incorporated into the theory). But

the Lorentz invariance of the latter is rarely considered – most of the empirical
content of conventional relativistic quantum mechanics concerns the scattering
regime. If this were done, the tension between Lorentz invariance and quantum
nonlocality would be immediately apparent. On the other hand, various approaches
for constructing an invariant Bohmian theory for Lorentz transformations have
been proposed, some models have been formulated and discussed (Berndl, Dürr,
Goldstein, and Zanghì 1996), and the subtleties involved in the task of constructing
a relativistic version of Bohmian mechanics have been thoroughly analyzed (Dürr,
Goldstein, Norsen, Struyve, and Zanghì 2014).
4.13 What Is a Bohmian Theory?

In the structure of a Bohmian theory, one can see some very general features that
are, in fact, common to all the precise and serious formulations of quantum
mechanics that are not based on vague and imprecise concepts such as measure
or observer:
1. The theory must be based on a clear ontology describing what the theory is
fundamentally about – what has been named the primitive ontology (Allori,
Goldstein, Tumulka, and Zanghì 2008) – that is, the type of basic entities (like
the particles in Bohmian mechanics) that are the building blocks for anything
else in the world, including tables, chairs, cats, and measuring devices. The
primitive ontology may include geometry, particle, field, or string configur-
ations, or whatever is needed out to best describe nature.
2. There must be a quantum state vector, a wave function, which evolves (at least
approximately) unitarily and whose role is to generate the temporal evolution
(which does not need to be deterministic) of the variables that describe the
primitive ontology.
3. The empirical relevance of the theory should emerge from its providing a notion
of typical spatiotemporal histories, presumably specified by a measure of typic-
ality on the set of all possible histories of the primitive ontology of the theory.
4. The predictions must agree (at least approximately) with those of orthodox
quantum theory – in the cases that the latter ones are unambiguous.
In brief, a Bohmian theory is simply a quantum theory with a coherent ontology.
References
Allori, V., Dürr, D., Goldstein, S., and Zanghì, N. (2002). “Seven steps towards the
classical world,” Journal of Optics B, 4: 482–488.
92 Nino Zanghì
of Bohmian mechanics and the Ghirardi-Rimini-Weber theory,” British Journal for
Bell, J. S. (1964). “On the Einstein-Podolsky-Rosen paradox,” Physics, 1: 195–200.
Reprinted in Bell (1987).
Bell. J. S. (1986). “Beables for quantum field theory,” Physics Reports, 137: 49–54.
Reprinted in Bell (1987).
Bell. J. S. (1987). Speakable and Unspeakable in Quantum Mechanics. Cambridge:
Cambridge University Press.
Bell. J. S. (1990). “Against measurement,” Physics World, 3: 33–40.
Berndl, K., Dürr, S., Goldstein, S., and Zanghì, N. (1996). “Nonlocality, Lorentz invari-
ance, and Bohmian quantum theory,” Physical Review A, 53: 2062–2073.
variables: Parts I and II,” Physical Review, 85: 166–193.
Daumer, M., Dürr, D., Goldstein, S., and Zanghì, N. (1996). “Naive realism about
operators,” Erkenntnis, 45: 379–397.
Dürr, D., Goldstein, S., Norsen, T., Struyve, W., and Zanghì, N. (2014). “Can Bohmian
mechanics be made relativistic?,” Proceedings of the Royal Society A, 470: 20130699.
Dürr, D., Goldstein, S., Tumulka, R., and Zanghì, N. (2004). “Bohmian mechanics and
quantum field theory,” Physical Review Letters, 93: 090402.
Dürr, D., Goldstein, S., and Zanghì, N. (1992). “Quantum equilibrium and the origin of
absolute uncertainty,” Journal of Statistical Physics, 67: 843–907.
Dürr, D., Goldstein, S., and Zanghì, N. (2004). “Quantum equilibrium and the role of
operators as observables in quantum theory,” Journal of Statistical Physics, 116:
959–1055.
Feynman, R., Morinigo, F., and Wagner, W. (2003). Feynman Lectures On Gravitation.
Boca Raton: CRC Press.
Ghirardi, G. C. and Weber, T. (1983). “Quantum mechanics and faster-than-light commu-
nication: Methodological considerations,” Nuovo Cimento, 78B: 9–20.
Gleason, A. M. (1957). “Measures on the closed subspaces of a Hilbert space,” Journal of
Mathematics and Mechanics, 6: 885–893.
Kochen, S. and Specker, E. P. (1967). “The problem of hidden variables in quantum
Landau, L. D. and Lifshitz, E. M. (1958). Quantum Mechanics: Non-relativistic Theory.
J. B. Sykes and J. S. Bell (trans.). Oxford and New York: Pergamon Press.
Schilpp, P. A. (ed.). (1949). Albert Einstein, Philosopher-Scientist. Evanston, IL: The
Library of Living Philosophers.
Schrödinger, E. (1995). The Interpretation of Quantum Mechanics. Dublin Seminars
(1949–1955) and Other Unpublished Essays. M. Bitbol (ed.). Woodbridge: Ox
Bow Press.
von Neumann, J. (1932/1955). Mathematische Grundlagen der Quantenmechanik. Berlin:
Springer Verlag. R. T. Beyer (trans.), Mathematical Foundations of Quantum Mech-
anics. Princeton, NJ: Princeton University Press.
Wooters, W. K. and Zurek, W. H. (1982). “A single quantum cannot be cloned,” Nature,
299: 802–803.
5
Ontology of the Wave Function and the
Many-Worlds Interpretation
lev vaidman
5.1 Introduction
Quantum theory is about a century old, but as the existence of this volume shows, we
are far from consensus about its interpretation. Science does not develop in a straight
line. For a decade, quantum theory had no real basis, only phenomenological
equations found by Bohr, who made many of us believe until today that quantum
mechanics cannot be understood. The relativistic generalizations of the Schrödinger
equation, however, provide a complete, elegant physical theory that is fully consist-
ent with experimental data with precision of up to 10 significant digits. The situation
today is much better than at the time of Lord Kelvin’s speech in 1900, in which he
argued that physics is almost finished except for solving “two clouds,” which were
later to become the theory of relativity and quantum theory. This error, and the fact
that quantum equations describe several outcomes for quantum measurements
although we always see just one, are probably the main reasons why contemporary
physicists are reluctant to state that physics is close to being finished.
To deal with this second problem, we either have to add something to the wave
equation, but no proposal attractive enough to reach consensus has been found, or
to admit that what we see is only a tiny part of what is and that there are multiple
parallel worlds similar to ours. I find that this last option is the only reasonable one,
and I hope it will reach consensus in a foreseeable future.
5.2 A Toy Model of Classical Mechanics

Before discussing quantum mechanics, which is clearly not simple since we are
very far from consensus about its interpretation, I would like to discuss a toy model
of classical mechanics. It is an implementation of Laplace Universe:
We ought to regard the present state of the universe as the effect of its antecedent state and
as the cause of the state that is to follow. An intelligence knowing all the forces acting in
93
94 Lev Vaidman
nature at a given instant, as well as the momentary positions of all things in the universe,
would be able to comprehend in one single formula the motions of the largest bodies as
well as the lightest atoms in the world, provided that its intellect were sufficiently powerful
to subject all data to analysis; to it nothing would be uncertain, the future as well as the past
would be present to its eyes. The perfection that the human mind has been able to give to
astronomy affords but a feeble outline of such an intelligence.
(Laplace 1820/1951)
I assume that the clouds Lord Kelvin talked about do not exist. Newton’s laws and
Maxwell’s equations are somewhat different, such that they provide a consistent
theory for microscopic particles moving on continuous trajectories, which, using
the methods of statistical mechanics, explains well all experimental data. It is also
an assumption about different experimental data, because we know that actually
observed data cannot be explained by a classical model.
In the model we consider, the three-dimensional space is given. There are
particles moving in well-defined trajectories and fields spread out in space. Par-
ticles create fields that propagate in space and change the motion of other particles
present in locations with nonvanishing fields. The laws of creation and propagation
of fields explain the existence of stable rigid objects and everything else (including
ourselves) that we experience with our senses. The behavior of objects is deter-
ministic; free will is an illusion.
Like in actual physics, our model can be presented in a different way. There is a
point in the configuration space of all particles and the configuration of fields
fulfilling some global equation of extremal action. The global laws provide the
same solution for trajectories. Both explanations are acceptable, but I feel that it is
the first presentation, with fields locally acting on particles moving on trajectories
in three-dimensional space, that is a more convincing explanation of the world.
The existence of a global mathematical representation is important, but it hides the
local causal story, which is what is considered as an explanation of motion of
micro systems as well as our behavior.
It seems to me that in a counterfactual universe with successful classical physics
as described earlier, there will be no philosophical controversy in how to describe
reality. Particle trajectories governed by local forces through fields in three dimen-
sions would be a clear consensus.
5.3 Ontology of Collapsed Wave Function

Our world is not classical. Numerous experiments (e.g., particle interference)
contradict this picture. Moreover, Bell-type correlations show that no classical-
type local theory, i.e., a theory that locally predicts a single outcome for each
possible experiment, can reproduce observed correlations. We have a new theory,
Ontology of the Wave Function and the Many-Worlds Interpretation 95
quantum mechanics. It is true that the community of physicists working in the field
of foundations of quantum mechanics has not reached any consensus over its
interpretation. Many of them (like me) are certain that their favorite view is a
satisfactory (or even an excellent) solution, but each separate group is a small
minority, so the message of the community as a whole is that currently there is no
good solution. There are many, also outside the foundations community, who feel
that we need the correct interpretation, but that it is different from all current
proposals. However, the majority of physicists really think that the problem does
not exist and that textbook quantum mechanics is satisfactory. It tells us that every
time we perform a quantum measurement there is a collapse of the quantum wave
function and that the collapsed wave function well describes all that we see around
us. Von Neumann proved that the tough question of when exactly collapse occurs
needs not to be answered, because wherever we put the cut between classical and
quantum, we will observe no contradiction with our experience. A vague statement
according to which all “macroscopic” objects are “well” localized provides a
satisfactory criterion.
Von Neumann collapse is ad hoc without any concrete mechanism. In physical
collapse theories, such as those of Pearle (1976), Ghirardi, Rimini, and Weber
(1986), Diosi (1987), and Penrose (1996), the collapsed wave function is not
completely identical to that of von Neumann, but it is very close; so proponents
of physical collapse theories also consider the collapsed wave function as a
satisfactory description of what we see. Apparently it is the promotion of the wave
function ontology in configuration space by Albert (2013) that led to strong
criticism. Maudlin (2013) understandably complained: How can a mathematical
object in high dimensional space represent our experience in three dimensions?
The key to answering this question is the understanding that our experience
supervenes on macroscopic objects. We do not directly experience the electron
wave function in the atoms of our body. Parts of our body and our neurons are
macroscopic bodies. We needed configuration space because of entanglement.
Whereas for describing classical particles we had a choice between one point in
3N dimensional space or N points in three-dimensional space, in the quantum case
with entanglement the second option does not exist, entanglement requires multi-
dimensional space. Complete descriptions of all particles separately do not provide
a complete description of entangled particles. But entanglement of quantum
systems does not exist for macroscopic systems. In von Neumann’s approach, it
is absent by definition, and in physical collapse theories, the mechanism removes
entanglement of macroscopic objects very quickly. Without entanglement, we can
describe all objects in three dimensions. To summarize, the analysis of the process
of obtaining experience from our senses when the universe is described by a
collapsed wave function of a textbook or by the collapsed wave function of a
96 Lev Vaidman
physical collapse proposal provides an explanation similar to that of classical

physics: (Macroscopic) objects moving in three-dimensional space locally trigger
sensory organs of observers living in this three-dimensional space.
Current (quantum) physical theory based on minimal action does not always
provide a simple local explanation of the behavior of particles as in our gedanken
classical model. The basic concept of quantum theory is that referring to potentials,
which are not measurable locally. Classical physics also uses potentials, but they
are just auxiliary tools, helping to solve various problems more efficiently. In
classical physics, the operational meaning of potentials is that their derivatives
provide observable forces. In quantum mechanics, the Aharonov-Bohm effect
teaches us that potentials provide a direct physical effect: A particle moving in a
field-free region behaves differently depending on the potentials present there
(Aharonov and Bohm 1959). At least for myself, I resolved the difficulty by
finding a local explanation of the Aharonov-Bohm effect based on considering
the source of the potential to be quantum and taking into account entanglement
between the electron and the source (Vaidman 2012a). This explanation removed
the proof that quantum theory based on direct action of local fields cannot exist.
I opened the way to such a theory, which, if constructed, will be conceptually more
satisfactory than the present one.
5.4 Against Collapse

The nonlocality of the Aharonov-Bohm effect makes a physical explanation more
sophisticated, but the main ugly scar of quantum mechanics is collapse. I want to
believe that today’s physics by and large correctly explains everything we see
around us. For this belief I need two ingredients. First, that predictions of the
theory correspond to what we see, i.e., to be confirmed by experimental results.
Second, that the theory is elegant enough to believe that this is the description of
nature. We have had incredible success with the first part. In all cases we can
calculate and measure, there is a complete agreement, and sometimes with astro-
nomical precision of more than 10 digits. There are no “clouds” similar to those
seen by Lord Kelvin in classical physics. There is also incredible success of the
theory. All basic laws of physics can be written on a t-shirt. Today’s quantum
mechanics (even with potentials) is a good theory. It is deterministic and it does not
have action at a distance. Only the collapse spoils it. It is the only random process
in physics.
Collapse is sometimes considered to be in “peaceful coexistence” with special
relativity. Indeed, we cannot send superluminal signals using the collapse process.
But it is an action at a distance. A particle described by a superposition of wave
packets in two separate locations leads to action at a distance when the presence of
a particle in one location is measured. The complete quantum description of a

region of space with the second wave packet before the measurement is a mixed
state. It has the following operational meaning: Everyone, everywhere, has a well-
defined betting strategy on the result of a measurement testing the presence of the
particle in this place. Measurement in the location of the first wave packet changes
the situation immediately. Obviously so if the wave function is an ontic entity,
because it is changed by a remote action, but also if it is only an epistemic concept.
There is a change in operational meaning. If I and you are in the location of the first
wave packet, you are ready to bet with me on the result of the remote measurement,
but this changes immediately after you observe that I made a measurement on the
first wave packet.
Contrary to a widespread belief, there is no evidence for collapse of the wave
function. It is true that for every quantum experiment we observe only one
outcome, while a theory without collapse has a wave function corresponding to
several outcomes. But to be considered evidence, we need for the theory to predict
a different experience in case the collapse takes place or not. The theory does not
predict this. In fact, looking on a wave function of quantum mechanics without
collapse it is not easy to make any prediction about experience. We need to add
some postulates to connect the ontology with our experience.
5.5 Connecting Ontology with our Experience

Let us start the analysis with an interpretation of quantum mechanics where the
connection is simple. In Bohmian mechanics (see, e.g., Goldstein 2017), the
postulate is that experience supervenes on Bohmian positions of particles and
the wave function is only a pilot wave of these particles. The Bohmian picture of
positions of all particles is similar to the gedanken classical theory I sketched
earlier. All particles provide a familiar picture of what we see around us drawn
in three dimensions in a pointillist style. Bohmian particles, according to
Bohmian mechanics postulates, are distributed in the locations of a nonvanish-
ing wave function according to the Born Rule. But the size of the atoms, and
thus the size of the wave packets of elementary particles, is so small that the
difference between locations of Bohmian points and centers of particle’s wave
packets is inessential. Thus, considering the centers of the wave functions of
particles in the collapsed wave function also provides a familiar picture. The fact
that most of the elementary particles are entangled and do not have a pure state
does not spoil this picture. Their mixed states are still well localized on the
atomic scale, so the expectation values of their positions are not far from those
corresponding to Bohmian positions.
98 Lev Vaidman
Bohmian particles describe a world that looks like the one we observe and so
also does the collapsed wave function. Because the wave functions of macroscopic
objects are well localized, the picture drawn by the expectation values of the
position vectors of all particles of these objects also provides a familiar picture.
The particles are very close to each other, so we do not observe the points. We
observe the smoothed picture of everyday objects: tables, cats, people, etc. – the
collapsed wave function provides a familiar picture of the world. The theory has a
tacit assumption, a postulate: Our everyday experience supervenes on the collapsed
wave function. Since the picture drawn by the particles and the picture we draw
based on our experience are so similar, we usually forget that we make an
assumption connecting the formalism with experience. The theory is supposed to
describe our experience. In this theory there is only one picture that looks like the
world we see, we assume that there is only one world, so naturally we connect
them. But the postulate is needed, because there can be other options: We can also
imagine the presence of Bohmian particles in a theory that makes collapses, and
attaches experience to Bohmian particles. It is a different rule, although it seems to
provide essentially the same experience.
5.6 Connecting Ontology with Experience in the Framework of the MWI

In quantum mechanics without collapse we must add a postulate to connect to our
experience, because mathematics does not provide a (unique) picture correspond-
ing to what we see around us.
My postulate will be as follows. The universe with a noncollapsing wave
function corresponds to multiple experiences. Each experience should correspond
to at least one world, the definition of the concept of a world does not allow
multiple simultaneous experiences of a person in a world. To connect the wave
function of the universe to our experience, we first need to decompose the wave
function of the universe into a superposition of wave functions of worlds.
X
jΨiUNIVERSE ¼ αi jψ i iWORLD (5.1)
i
The wave function of a world jψ i i must be of a particular type. Probably the most
informative definition is that this is the type of wave function that might appear in
the textbook as “collapsed wave function of the universe.” Without relying on a
textbook definition, this type of wave function can be defined by the property that
all macroscopic objects must be well localized. Both definitions are vague. We
usually require rigor and precision in our theories. However, it must be so when we
consider exact sciences. The many-worlds interpretation (MWI) has two clearly
separated parts: (i) a precise physical theory of evolution of the wave function of
the universe jΨiUNIVERSE and (ii) the connection of this universal wave function to
our experience(s) (see Vaidman 2002). In classical theory, in Bohmian theory, and
in quantum mechanics with collapse, the separation between the two parts of the
theory was not emphasized because the second part was very simple: The connec-
tion is natural and obvious in contrast to the MWI, where the second part is
significant. But I believe that the connection implicitly taken in single-world
theories can be directly adopted to the MWI, and vagueness of the splitting of
the worlds is much less problematic than the vagueness of the answer to the
question: When does collapse occur? Our experiences cannot be described in terms
of exact science. So, it is understandable and acceptable that the concepts
belonging to the second part of the theory are not rigorously defined. The wave
function of a world has the following form
! ! ! ! ! !
jψ i iWORLD ¼ ψ 1CM r 1CM φ1rel r1i r1j ψ 2CM r 2CM φ2rel r2i r 2j . . .
!CM M ! !

ψMCM r M φrel rMi rMj ΦREST (5.2)
!

It is a product state of wave functions ψ kCM r kCM of centers of mass of macro-
scopic
! objects, times entangled states of relative coordinates of these objects
!
φkrel rki rkj , times the wave function ΦREST of the remaining particles that are
not part of macroscopic objects. The terms “macroscopic” and “well localized”
might be chosen as more or less “fine grained,” so the decomposition Eq. (5.1) is
only approximately defined.
It seems to me that I can stop here. Textbook quantum theory is a well-
established and well-tested theory that well explains everything we see around
us. However, it includes the unphysical process of collapse, which makes it very
difficult to believe that it is true. I remove the collapse and use the same postulate
of connection to our experience. Now, the theory is a good physical theory
(deterministic, no action at a distance, and no ad hoc rules). The postulate of the
correspondence of the experience with the wave function of the type of the
universal wave function of a collapsed world makes the experience, by fiat,
identical to that of an experience of an observer living in a universe with collapsing
wave functions. However, I know that this picture is (still) not in a consensus.
I need to persuade the community that it is consistent.
5.7 Naïve Criticism

The most naïve criticism is that we do not have experimental evidence for the
MWI. How can all these numerous worlds be present in the same space as our
world without us noticing their presence? Equations tell us that there is no
interaction between objects in different worlds. Moreover, we see in the laboratory
100 Lev Vaidman
that, for micro systems, there is no scattering between wave packets of the same
electron. Also note that we have no experimental evidence for the existence of
collapse. Testing MWI versus collapse theories will require a quantum experiment
up to a stage that the collapse proponents accept that collapse must have happened
and then undoing the experiment (Deutsch 1986, Vaidman 1998). If we get the
original state every time we perform this procedure, it is a proof that collapse did
not take place. Until today there is no sign of a collapse, but we are very, very far
from a decisive experiment. If collapse proponents will claim that collapse happens
only after we write the result of a quantum experiment in a notebook, it can never
be tested. Of course, it is the MWI that cannot be proved. If collapse exists at an
early stage, it will be observed (although it will not be easy to persuade opponents
that the explanation of the signal is not a failure of experimentalists to prevent
decoherence).
Slightly less naïve criticism is the problem of preferred basis. Mathematically,
one can decompose the wave function of the universe into a superposition of
orthogonal components, not just as in Eq. (5.1), but in many other ways that will
not provide a familiar world’s picture in every branch. So, critics might say that the
proposal is circular: I define by fiat what I want to explain. First, a simple definition
that is confirmed by observation sounds to me like a legitimate strategy. But there
is also a more specific answer. The basis of the decomposition is indeed preferred.
Until now I have not mentioned time evolution. Everything was considered at a
particular moment. But we cannot experience anything at zero time. We need an
order of 0.1 seconds to identify our experience. Thus, the world needs some finite
time to be defined. The world has to be stable, at least on the scale of seconds.
Locality of interactions in nature ensures that only the decomposition of wave
functions corresponding to well-localized macroscopic objects can be stable.
A quantum state describing the superposition of a macroscopic object in separate
locations with a particular phase evolves almost immediately into a mixture that
has a large component with a different phase. This obvious fact is analyzed in
numerous papers using the buzzword ‘decoherence.’
5.8 The Probability Problem

The most difficult issue is probability. The idea of parallel worlds is relatively new
for humanity, so language and philosophy describing this situation is not well
developed. Lewis (1986) provided some insight, but the philosophical issues
related to plurality of worlds are far from clear.
The MWI that I advocate has no intrinsic uncertainty, randomness, or objective
chance, which makes the concept of probability difficult. In the collapse theory,
probability has a clear meaning: The event happened while it could also not have
happened. In the MWI, such meaning does not exist. The event happened in this
world, but there is no alternative, it could not be otherwise. I want to say that the
traditional concepts of probability are not applicable in the framework of the MWI
when we consider the outcomes of quantum experiments to be performed. Every
time we perform a quantum experiment and it seems to us that a single result is
obtained, all possible outcomes are obtained, each in a different world. There is no
meaning to the question: In which world will I end up? In some sense I will be in
all. In no sense will I be in a particular one.
Still, we have to explain our experience of apparent random behavior and the
frequency pattern of the results of quantum experiments. My claim remains that
there is no difference between my experience if I live in one of the MWI worlds
and my experience if I live in the only world of the universe with collapses on
every measurement. How do we reconcile the difference between the existence of a
probability concept and identity of experience? To avoid that difficulty, there are
proposals to introduce uncertainty in the MWI and provide the meaning for the
probability that I will end up in a world with a particular outcome (Saunders and
Wallace 2008). In my view, adding uncertainty to the theory spoils it. The wave
function of the universe is supposed to be the whole physical ontology, and it does
not have any pointer moving from one world to another.
But do we really have a problem here? The fact that there is no meaning for the
probability of the result of a future measurement does not contradict the claim of
identical experiences. The standard approach to probability is to consider events
that will happen, but testing probability claims relies on records of experiments in
the past – frequencies of the outcomes of repeated identical measurements per-
formed in the past. So even in the framework of collapse theory, probability
assignments are confirmed or refuted by our experiences in the past. Thus, the
difficulty of the MWI to introduce the concept of probability for future outcomes is
not relevant.
Assignment of probability for future experiment relies on an additional assump-
tion, even in the framework of a collapse theory. We assume that nature will not
change its laws. Another way to make predictions about future measurement is to
assume that, after performing the measurement, the frequency of the measurement
results fits our probability assignment. This approach may also be applied in the
framework of the MWI. We expect that within the world, the frequencies of the
results of past measurements will correspond to the probability assignment.
5.9 Probability of Self-Location

The counterpart of the Born Rule in the framework of the MWI is sometimes
named the Born-Vaidman Rule:
102 Lev Vaidman
The probability of self-location of an observer in a particular world is proportional

to the measure of existence of that world.
The (somewhat controversial) term ‘measure of existence’ (Vaidman 1998, Grois-
man, Hallakoun, and Vaidman 2013) is just the square of the amplitude of the
coefficient of the corresponding wave function in the decomposition of the univer-
sal wave function Eq. (5.1). This rule explains the correspondence of experimental
data of quantum experiments with the statistics predicted by the Born Rule.
Probability is a very controversial philosophical concept. Frequency of particular
outcomes in experimental records is just one approach. Another approach is readi-
ness to put bets on the results of experiments according to their probability of
happening. Since in the MWI all possible outcomes happen, it is difficult to
understand this aspect of probability. I proposed a gedanken experiment which
allows sensible betting also in the framework of the MWI, tightening the connection
of the Born-Vaidman rule with the concept of probability. In my proposal, after
arranging a quantum experiment, the observer takes a sleeping pill (Vaidman 1998).
During the observer’s sleep, the experiment is performed and the observer is moved
to various rooms with identical interiors, according to the results of the experiment.
When the observer is awakened, the observer understands that there are several
copies of the observer in different rooms corresponding to the results of the experi-
ment; however, the observer does not know who he or she is, and thus the observer
does not know the result of the experiment in the observer’s world. The observer
might be given the wave function of the universe; the observer still has ignorance,
which allows the observer to bet. Awakened descendants of the experimentalist
preparing the experiment have a genuine ignorance concept of probability, similar to
that of an experimentalist in the collapsing universe. Only the questions are differ-
ent. The latter asks: What will be the result of the experiment? The former asks: In
which world (defined by the result of the experiment) am I?
Note that all descendants have the same information and thus provide identical
assignments for probabilities of different outcomes. This allows them to define a
rational betting strategy for an experimentalist before the experiment. Instead of a
probability postulate, he or she has the caring principle (Vaidman 1998, Greaves
2004):
The experimentalist cares about his descendants in proportion to their measure of
existence.
The justification of this principle for our betting example is as follows. Every
descendant will have a genuine probability concept and would like to have a bet
according to a particular probability assignment. Since the descendants will get the
reward of the bet, the experimentalist, naturally caring for his or her descendants,
has a rational reason to put the bet for the results of the experiment. Tappenden
(2011) suggested (and I think it is a reasonable approach) that there is no need to
perform the complicated procedure with the sleeping pill. It is enough that one can
imagine performing such a procedure to justify the betting assignment of the
experimentalist who believes in the MWI. So, identical experiences lead to identi-
cal behavior, although the argumentation is different, which is not surprising in
view of different world views.
5.10 Deriving the Born Rule

The betting assignments are according to the Born Rule or to the Born-Vaidman
Rule. The rules are postulates that are added to the formalism of quantum
mechanics of evolution of the wave function. In my view, the addition of a
postulate is well justified through its confirmation by our observations. However,
I do not want to leave unnoticed a large and increasing effort to derive the Born
Rule. Deutsch (1999) started a program (extended by Wallace 2012) to derive the
proper betting behavior in the framework of the MWI based on some postulates
of Decision Theory. The program encountered criticisms of apparently circular
definitions. The Deutsch-Wallace proof is complicated, and it is not simple to
understand what exactly is proved and what is assumed. I also suggested a proof
(Vaidman 2012b). My work was triggered by Deutsch’s (1999) paper, but the
proof is apparently different. It is based on symmetry and relativistic causality.
Apart from the fact that in the framework of collapse theory we have to postulate
relativistic causality, while in the framework of the MWI it is part of the physical
theory, my proof is applicable to both cases (McQueen and Vaidman 2018). Very
recently, Sebens and Carroll (2018) proposed yet another proof. They say it
follows my approach of self-locating uncertainty and relies on the assumption
that actions on the environment cannot change outcomes of local measurements.
This assumption is not very different from my relativistic causality; however, it
seems to me that their proof does not hold. It is based on permutation symmetry
and relies on a metaphysical approach to personal identity, the justification for
which I cannot understand. Sebens and Carroll manipulate the concept of prob-
ability of an observer to be in a particular world when splitting has already
happened, but the wave function of the observer is identical in both worlds.
Indeed, both the mental states and the (wave functions of the relative variables of
the) bodies of the observers in the two worlds are identical. The worlds are
distinguished by macroscopic differences of other objects. In this case there is no
matter of fact about what is the world, out of the two worlds, in which the
observer is present, because the observer is present in both (see details in Kent
2015, and in McQueen and Vaidman 2018).
104 Lev Vaidman
Putting aside the correctness of the proofs of the Born Rule, I argue that none of
them can be considered as unconditioned on any assumption. For the Born Rule in
the framework of collapse theories, we need an additional postulate about collapse.
It is not part of the standard formalism. A priori it need not follow any laws of
standard physical theory. We need to postulate some assumption – in my proof it is
the impossibility of superluminal signaling. In the framework of the MWI we also
need an assumption. The physics part, the evolution of the wave function of the
universe, has to be supplemented by some law connecting it to our experience.
Here, the assumption may be considered natural and minimal: Everything, includ-
ing our experiences, supervenes on the wave function of the universe. Then,
physical laws governing the evolution of the wave function are relevant to our
experience, too; so we might claim that no additional assumption was made. Still,
in all cases there is a tacit assumption that probability (or the illusion of probabil-
ity) depends on the wave function.
5.11 Conclusions
Today’s physics is quantum theory and it enjoys unprecedented success
explaining all observed phenomena. There are questions that do not have good
answers (yet). Quantum gravity, dark matter, dark energy. . . it might happen
that resolving these questions will require new revolutionary ideas. However,
quantum mechanics apparently will remain the theory explaining electromag-
netic interactions – the interactions that are responsible for almost everything
we see in everyday life. Unphysical features of collapse are the main reasons for
doubts that this is the final theory of nature. But actually, there is no evidence
for collapse. Apparently, it is just the difficult philosophical consequences of no
collapse that prevents consensus about quantum theory without collapse and
about the existence of multiple worlds. It took time before people were ready to
accept that the Earth is not the center of the universe. We also need time to
accept that we are not unique and that there are many similar copies of us. We
need time to establish the connection between the well-established mathemat-
ical part of the theory and our experience. It is an unusual situation, which we
did not encounter in old physical theories. Philosophers should play an import-
ant role in this project because it requires a dramatic change in our world view.
Observing the rapidly increasing number of publications related to the MWI in
the philosophical literature makes me optimistic. I am not sure that the large
effort to find a derivation of the Born Rule is justified (I doubt that the MWI has
significant advantage here), but this activity leads to accepting the legitimacy of
the MWI by physicists, and I believe that its advantage as a physical theory will
bring it to consensus.
Acknowledgments
I benefited greatly from numerous discussions with Kelvin McQueen, David
Albert, and the participants of the workshop Identity, indistinguishability and
non-locality in quantum physics (Buenos Aires, June 2017). This work has been
supported in part by the Israel Science Foundation Grant No. 1311/14.
References
Aharonov, Y. and Bohm, D. (1959). “Significance of electromagnetic potentials in the
quantum theory,” Physical Review, 115: 485–491.
Albert, D. Z. (2013). “Wave function realism,” pp. 52–57 in A. Ney and D. Z. Albert
(eds.), The Wave Function: Essays on the Metaphysics of Quantum Mechanics.
Oxford: Oxford University Press.
Deutsch, D. (1986). “Three experimental implications of the Everett interpretation,”
pp. 204–214 in R. Penrose and C. J. Isham (eds.), Quantum Concepts of Space and
Time. Oxford: The Clarendon Press.
Deutsch, D. (1999). “Quantum theory of probability and decisions,” Proceedings of the
Royal Society of London A, 455: 3129–3137.
Diósi, L. (1987). “A universal master equation for the gravitational violation of quantum
mechanics,” Physics Letters A, 120: 377–381.
Ghirardi, G. C., Rimini, A., and Weber, T. (1986). “Unified dynamics for microscopic and
Goldstein, S. (2017). “Bohmian mechanics,” in E. N. Zalta (ed.), The Stanford Encyclo-
pedia of Philosophy (Summer 2017 Edition), https://plato.stanford.edu/archives/
sum2017/entries/qm-bohm/
Greaves, H. (2004). “Understanding Deutsch’s probability in a deterministic multiverse,”
Groisman, B., Hallakoun, N., and Vaidman, L. (2013). “The measure of existence of a
quantum world and the sleeping beauty problem,” Analysis, 73: 695–706.
Kent, A. (2015). “Does it make sense to speak of self-location uncertainty in the universal
wave-function? Remarks on Sebens and Carroll,” Foundations of Physics, 45:
211–217.
Laplace, P. (1820/1951). “Essai Philosophique sur les Probabilités,” Introduction to
Théorie Analytique des Probabilités. Paris: V Courcier. F. W. Truscott and F. L.
Emory (trans.), A Philosophical Essay on Probabilities. New York: Dover.
Lewis, D. (1986). On the Plurality of Worlds. Oxford: Basil Blackwell.
Maudlin, T. (2013). “The nature of the quantum state,” pp. 126–153 in A. Ney and D. Z.
Albert (eds.), The Wave Function: Essays on the Metaphysics of Quantum Mechanics.
McQueen, K. and Vaidman, L. (2018). “In defence of the self-location uncertainty account
of probability in the many-worlds interpretation,” Studies in History and Philosophy
of Modern Physics, in press, https://doi.org/10.1016/j.shpsb.2018.10.003.
Pearle, P. (1976). “Reduction of statevector by a nonlinear Schrödinger equation,” Phys-
ical Review D, 13: 857–868.
Penrose, R. (1996). “On gravity’s role in quantum state reduction,” General Relativity and
Gravitation, 28: 581–600.
Saunders S. and Wallace D. (2008). “Branching and Uncertainty,” The British Journal for
106 Lev Vaidman
Sebens, C. T. and Carroll, C. M. (2018). “Self-locating uncertainty and the origin of

probability in Everettian quantum mechanics,” The British Journal for the Philosophy
of Science, 69: 25–74.
Tappenden, P. (2011). “Evidence and uncertainty in Everett’s multiverse,” The British
Journal for the Philosophy of Science, 62: 99–123.
Vaidman, L. (1998). “On schizophrenic experiences of the neutron or why we should
believe in the many-worlds interpretation of quantum theory,” International Studies
in the Philosophy of Science, 12: 245–261.
Vaidman, L. (2002). “Many-worlds interpretation of quantum mechanics,” in E. N. Zalta
(ed.), The Stanford Encyclopedia of Philosophy (Fall 2016 Edition), https://plato
.stanford.edu/archives/fall2016/entries/qm-manyworlds/
Vaidman, L. (2012a). “Role of potentials in the Aharonov-Bohm effect,” Physical Review
A, R86: 040101.
Vaidman, L. (2012b). “Probability in the many-worlds interpretation of quantum mechan-
ics,” pp. 299–311 in Y. Ben-Menahem and M. Hemmo (eds.), Probability in Physics,
The Frontiers Collection. Berlin/Heidelberg: Springer-Verlag.
Wallace, D. (2012). The Emergent Multiverse: Quantum Theory According to the Everett
Interpretation. Oxford: Oxford University Press.
6
Generalized Contexts for Quantum Histories
marcelo losada, leonardo vanni, and roberto laura
6.1 Introduction
In the standard approach to quantum mechanics, there is no way to compute the
probability for expressions involving properties at different times. These probabil-
ities can be useful to relate a property of a microscopic system, before the
measurement process, to the value of the pointer variable of the macroscopic
apparatus after the measurement. Moreover, in the double-slit experiment, it is
important to have a suitable language to describe through which slit the particle
detected on a photographic plate has passed.
The theory of consistent histories has been introduced by Griffiths (1984),
Omnès (1988), and Gell-Mann and Hartle (1990), defining the notion of history
as a sequence of properties at different times. The probability for a history was
defined in this theory by an expression motivated by the path integral formalism,
but with no direct relation to the usual Born Rule. For a valid description of a
quantum system, the histories with well-defined probabilities should belong to a
family satisfying a state-dependent consistency condition.
In this theory, measurement is considered as a quantum interaction between the
measured microscopic system and the measuring macroscopic apparatus, and there
is no collapse postulate. Thus, the theory of consistent histories appeared to some
as a strong candidate for the realization of a quantum theory in which the act of
measurement would not have the distinguished role assigned in ordinary quantum
mechanics.
However, an important problem of this theory is that it does not provide us with
a single family of consistent histories, and the choice of different families may give
different descriptions for the time evolution of the same physical system. More-
over, different families of consistent histories can provide the prediction or the
retrodiction of contrary properties. If this is the case, it seems that the future or the
107
108 Marcelo Losada, Leonardo Vanni, and Roberto Laura
past of the quantum system could depend on the choice of the universe of discourse
(see Dowker and Kent 1996, Laloë 2001, Okon and Sudarsky 2014).
In this chapter we present a summary of our formalism of generalized contexts
for quantum histories (Laura and Vanni 2009, Losada, Vanni, and Laura 2013), in
which the ordinary contexts of properties for each different time should satisfy
compatibility conditions given by commutation relations in the Heisenberg
representation. A family of histories satisfying these compatibility conditions is
organized in a distributive lattice having well-defined probabilities obtained by a
natural generalization of the Born Rule of ordinary quantum mechanics.
In Section 6.2, our formalism is introduced for the case of histories in classical
mechanics. In Section 6.3, the formalism of generalized contexts for quantum
histories is presented in Section 6.4, it is applied to quantum measurements and to
the description of the double-slit experiment, with and without measurement
apparatuses. The conclusions are presented in Section 6.5.
6.2 Classical Histories

6.2.1 States and Properties in Classical Mechanics
In classical mechanics, the states of a physical system are represented by points in
the phase space Γ (i.e., the space of generalized coordinates and momenta). Each
property p of the system is represented by a subset Cp of the phase space (C p Γ).
A classical system in a state represented by a point x 2 Γ has a property p
represented by the set Cp Γ if x 2 C p . A partition of the phase space is obtained
by considering a collection of subsets C j of the phase space, where j 2 σ, and σ is a
set of indexes. This is a disjoint collection of subsets covering the phase space, i.e.,
Ci \ C j ¼ ϕ for i 6¼ j and [j2σ C j ¼ Γ.
The system’s properties, represented by all possible unions of the sets of the
partition, with the order relation represented by the inclusion (), form a Boolean
lattice (i.e., it is orthocomplemented and distributive). The conjunction (∧) and the
disjunction (∨) of two properties are represented respectively by the intersection
and the union of the corresponding subsets. The null element of the lattice of
properties is represented by the empty set, and the universal element by the phase
space Γ. The complement p of a property p, which is represented by a subset Cp , is
represented by the set Cp ¼ Γ C p . The set of classical properties pj correspond-
ing to the partition Cj (j 2 σ) of the phase space Γ are the atomic properties of the
lattice.
Anticipating what is going to be done in quantum mechanics, we use the term
classical context to denote the Boolean lattice generated by the atomic properties
pj (j 2 σ) of the system. It is always possible to combine different classical contexts
Generalized Contexts for Quantum Histories 109
in a single, more refined one. Let us consider two sets of atomic properties, pj
represented by the sets C j , (j 2 σ), and pμ , represented by the sets C μ and (μ 2 σ 0 ),
generating two different classical contexts. A more refined partition of the phase
space is obtained with the sets C j \ Cμ , representing a new set of atomic properties.
This new set generates a classical context containing the two previous ones. As an
important consequence, there is no restriction for the properties that can be
included in a classical context. This is not the case for quantum properties, as we
will discuss in Section 6.3.
6.2.2 Probabilities for Properties at a Single Time

In some cases there is no precision about the point representing the state at a given
time t. Therefore, there is no certainty about the system having or not having a
property, but only about the probability of having it. In these cases it is necessary
to appeal to what is known as a probability distribution.
Ð It is represented by a
function ρt : Γ⟶R, nonnegative and normalized ( Γ ρt ðxÞdx ¼ 1).
By means of the density function ρt , it is possible to define a probability on the
set of all properties. Classical statistical mechanics gives the following expression
for the probability of a property p represented by the set C p at time t:
ð
Pr t ðpÞ ¼ ρt ðxÞdx, (6.1)
Cp
Ð
which satisfies the Kolmogorov axioms:Ð (i) Pr t ðC Þ ¼ C ρt ðxÞdx 0 for any
property represented by C, (ii) Pr t ðΓÞ ¼ Γ ρt ðxÞdx ¼ 1, and (iii) if C \ C 0 ¼ ϕ,
then Pr t ðC [ C 0 Þ ¼ Pr t ðCÞ þ Pr t ðC 0 Þ.
The time evolution of the state is determined by the Hamilton equations. A state
represented by the point x 2 Γ at the time t evolves into a state represented by the
point x0 ¼ St0 t x, where St0 t : Γ ! Γ is invertible (S1 t 0 t ¼ Stt0 ) and volume preserv-
ing. If ρt ðxÞ is the state probability density at time t, the probability density at
time t 0 is given by

ρt0 ðxÞ ¼ ρt S1
t0 t x : (6.2)
This last equation can be used to obtain the probability Pr t ðpÞ of property p at time
t given in Eq. (6.1) in terms of the probability density at a reference time t0 . By
considering Eq. (6.2) with t 0 equal to an arbitrary fixed time t 0 in Eq. (6.1), we
obtain
ð ð
Pr t ðpÞ ¼ ρt ðxÞdx ¼ ρt0 ðxÞdx: (6.3)
Cp Cp, 0 St0 t Cp
We notice that the probability for the property p at time t can be expressed in two
different forms: either with a time-dependent probability density ρt ðxÞ together
with a time-independent set C p representing the property p, or with a time-
independent density ρt0 ðxÞ together with a time-dependent set Cp, 0 St0 t C p repre-
senting the property p. By anticipating what is found in quantum mechanics, we
will use the terms Schrödinger and Heisenberg representations to name the first
and the second forms of expressing the same probability for property p at time t.
6.2.3 Probabilities for Properties at Different Times

In what follows, we will present a formalism suitable for including properties at
different times in a probabilistic description of a classical system (Losada et al.
2013). As we showed in the previous subsection, the probabilities Prðp1 ; t 1 Þ for a
property p1 at time t 1 and Prðp2 ; t 2 Þ for a property p2 at a different time t 2 can both
be written in the Heisenberg representation corresponding to a single fixed time t 0 :
ð
Prðp1 , t 1 Þ ¼ ρt0 ðxÞdx , (6.4a)
C 1, 0 St0 t1 C1
ð
Prðp2 , t 2 Þ ¼ ρt0 ðxÞdx, (6.4b)
C2, 0 St0 t2 C 2
where C 1 and C2 (C 1, 0 and C 2, 0 ) are the subsets of the phase space corresponding
to the properties p1 and p2 in the Schrödinger (Heisenberg) representation.
The previous equations strongly suggest the following definitions for the prob-
abilities corresponding to the conjunction and the disjunction of the properties p1 at
time t 1 and p2 at time t 2 :
ð
Pr½ðp1 , t 1 Þ∧ðp2 , t 2 Þ ρt0 ðxÞdx, (6.5)
C 1, 0 \ C2, 0
ð
Pr½ðp1 , t 1 Þ∨ðp2 , t 2 Þ ρt0 ðxÞdx: (6.6)
C 1, 0 [ C 2, 0
In these expressions, C 1, 0 \ C2, 0 can be interpreted as the Heisenberg

representation for the conjunction ðp1 ; t 1 Þ∧ðp2 ; t 2 Þ, while C 1, 0 [ C 2, 0 can be
considered as the Heisenberg representation for the disjunction ðp1 ; t 1 Þ∨ðp2 ; t 2 Þ.
Classical histories involving properties at two times can be obtained starting
from two classical contexts of properties at times t 1 and t 2 . A classical context of
properties at time t1 , generated by atomic properties pj1 (j 2 σ), with Schrödinger
(Heisenberg) representations given by the partition C j1 (C j1, 0 St0 t1 C j1 ) of the
phase space, can be combined with a classical context of properties at time t 2 ,
generated by atomic properties pμ2 (μ 2 σ ∗ ), with Schrödinger (Heisenberg) repre-

sentations given by the partition C μ2 (Cμ2, 0 St0 t2 C μ2 ) of the phase space. The sets
μ
Cjμ j
0 C 1, 0 \ C 2, 0 , with ðj; μÞ 2 σ σ ∗ , form a partition of the phase space (i.e.,
j0 μ0 S
they satisfy C 0 \ C0 ¼ ϕ if ðj; μÞ 6¼ ð j0 ; μ0 Þ and jμ Cjμ
jμ
0 ¼ Γ). They represent the
j μ
histories “property p1 at time t 1 and property p2 at time t 2 ,” and they are the atomic
elements generating a distributive and orthocomplemented lattice of classical
histories for the two times t1 and t 2 , with well-defined probabilities defined by
Eq (6.5) and Eq. (6.6). The generalization to a lattice of classical histories at n
times t 1 < t2 < . . . < t n can easily be obtained.
6.3 Quantum Histories

6.3.1 States, Properties, Probabilities, and Contexts in Quantum Mechanics
In quantum mechanics, states are represented by vectors of a Hilbert space H or,
more generally, by density operators. Each property p of the system is represented
by a subspace H p of the Hilbert space H or, equivalently, by a projector Πp ,
such that H p ¼ Πp H . The time evolution between times t 0 and t of a density
operator ρ, representing
i a state
is given by ρt ¼ U ðt; t0 Þρt0 U 1 ðt; t 0 Þ, where
U ðt; t0 Þ ¼ exp ħ H ðt t 0 Þ , being H the Hamiltonian operator for the quantum
system. In the Schrödinger representation, the probability at time
t for a property p
in a state ρt is given by the Born Rule, Prðp; t Þ ¼ Tr ρt Πp . In the Heisenberg
representation, the same probability is given by Prðp; t Þ ¼ Tr ρt0 Πp, 0 , where
Πp, 0 ¼ U 1 ðt; t 0 ÞΠp U ðt; t 0 Þ.
The lattice structure for the properties of a quantum system is built from the
order relation () corresponding to the inclusion of the Hilbert subspaces.
For two properties p and p0, p p0 if H p H p0 . The conjunction p∧p0 of two

properties is represented by H p∧p0 ¼ Inf H p ; H p0 ¼ H p \ H p0 . The disjunction

p∨p0 is represented by H p∨p0 ¼ Sup H p ; H p0 ¼ H p þ H p0 . The complement p of
⊥
a property p is represented by H p ¼ H p . With this structure, the set of all
possible quantum properties forms an orthocomplemented and nondistributive
lattice, and the Born Rule does not yield, in general, well-defined probabilities.
An orthocomplemented and distributive lattice with well-defined probabilities is
obtained if the universe of discourse is restricted to a single quantum context.
A quantum context is generated by a projective decomposition of the Hilbert space
into mutually orthogonal subspaces H i ¼ Πi H (i 2 σ). The projectors Πi satisfy
P
Πi Πj ¼ δij Πj for all i, j 2 σ, and j2σ Πj ¼ I, where I is the identity operator. Any
property p of the quantum context is represented by a projector of the form
P
Πp ¼ j2σ p σ Πj .
In the previous section, we have shown that it is always possible to include

two different classical contexts in a single, more refined one. This is not, in
general, the case for quantum contexts. Two different quantum contexts gener-
ated by atomic properties represented by projective
decompositions Πj (j 2 σ) and
Πμ (μ 2 σ ∗ ) are said to be compatible if Πj ; Πμ ¼ 0 for all ðj; uÞ 2 σ σ ∗ . If
this is the case, the properties of the two contexts are included in a single context
generated by the properties represented by the projectors Πjμ Πj Πμ with
ðj; μÞ 2 σ σ ∗ .
6.3.2 Generalized Contexts for Quantum Histories

Now we will extend the notion of quantum contexts to consider the probabilities
for expressions involving properties at different times.
Let us consider at time t 0 a system in a state represented by the density
operator ρt0 , a context of properties C 1 at the time t 1 , and a context of properties
C 2 at the time t 2 . The atomic properties pk11 for the context C 1 are represented by
the projective decomposition with projectors Πk11 (k1 2 σ 1 ), while the atomic
properties pk22 for the context C 2 are represented by projectors Πk22 (k 2 2 σ 2 ).
The Heisenberg representation with reference time t 0 of these projectors is
given by
Πk11, 0 U 1 ðt 1 ; t 0 ÞΠk11 U ðt 1 ; t 0 Þ, Πk22, 0 U 1 ðt 2 ; t 0 ÞΠk22 U ðt 2 ; t 0 Þ: (6.7)
By definition, C 1 and C 2 are said to be compatible quantum contexts (Laura and

Vanni 2009, Losada et al. 2013), if the corresponding projectors commute in the
Heisenberg representation, i.e., if
h i
Πk11, 0 ; Πk22, 0 ¼ 0 for all ðk 1 ; k 2 Þ 2 σ 1 σ 2 : (6.8)
If C 1 at t 1 and C 2 at t 2 are compatible contexts, they generate what we call a

generalized context of histories at two times, with atomic histories “pk11 at time t 1
and pk22 at time t 2 ,” having the Heisenberg representation given by the projectors
Πk01 k2 Πk11, 0 Πk22, 0 , for ðk 1 ; k2 Þ 2 σ 1 σ 2 . These projectors provide a decompos-
ition of the Hilbert space H into mutually orthogonal subspaces and, therefore, the
atomic histories generate an orthocomplemented and distributive lattice of two
times histories. The Born Rule in Heisenberg representation provides well-defined
probabilities on this lattice. For the atomic histories we have

Pr pk11 at time t 1 and pk22 at time t 2 Tr ρt0 Πk01 k2 : (6.9)
The formalism presented for the case of two times can be easily generalized to
include histories for sequences of n times.
6.4 Results Obtained with the Formalism of Generalized Contexts

In axiomatic quantum theories, the states are considered functionals acting on the
space of observables and, therefore, they appear after the observables in a some-
how subordinate position (Laura and Castagnino 1998, Castagnino, Id Betan,
Laura, and Liotta 2002). Quantum histories play the role of the observables of
ordinary quantum theory and, as a consequence, it seems reasonable that the
allowed sets of histories satisfy state-independent conditions. The consistency
conditions of the theory of consistent histories produce state-dependent families
of consistent histories (Griffiths 1984, Omnès 1988, Gell-Mann and Hartle 1990).
On the contrary, the compatibility conditions of the generalized context formalism
given in Eq. (6.8) are state independent.
The state-independent compatibility conditions of the generalized-context
formalism produce an important difference with respect to the theory of
consistent histories. Each quantum history has a Heisenberg representation
given by a projection operator and each valid set of quantum histories is
generated by a projective decomposition of the Hilbert space. As a conse-
quence, a generalized context of quantum histories has the logical structure of
a distributive orthocomplemented lattice of subspaces of the Hilbert space,
i.e., the same logical structure of the quantum properties of an ordinary
context. It is because of this logical structure that in our formalism there is
no place for the retrodiction of contrary properties (Losada and Laura 2014b),
which is a problem for the theory of consistent histories (Okon and Sudarsky
2014).
In the opinion of some authors, the theory of consistent histories allows too
many histories and some of them are difficult to interpret (Dowker and Kent 1996,
Laloë 2001). We have proven that our compatibility conditions, given by the
commutation of the projectors representing the properties translated to a common
time, are equivalent to the consistency conditions imposed on all possible states of
the system (Losada and Laura 2014a). Therefore, the formalism of generalized
contexts imposes more restrictions than the theory of consistent histories for the
valid families of quantum histories, and it allows fewer families of histories (see
discussion in Losada and Lombardi 2018). The fact that the universe of discourse
about a physical system depends on a choice of a family of histories is a problem
for quantum theory. The freedom of choice of our formalism is smaller than in the
theory of consistent histories, but not enough to single out a realist interpretation of
quantum mechanics. We are working in the direction of endowing the formalism
with interpretive content.
In what follows, we will introduce a brief description of two physically relevant
applications of our formalism, presented in our previous works (Losada et al. 2013,
Vanni and Laura 2013).
6.4.1 Quantum Measurements

A nonideal measurement of an observable Q of a system S is an interaction during
the time interval ðt 1 ; t 2 Þ of the measured system S with the measuring apparatus A.
It is represented in the Hilbert space H S ⊗H A by a unitary transformation
U ¼ U ðt2 ; t 1 Þ satisfying
U
jqi i ja0 i ! j ϕi i j ai i, (6.10)
where j qi i is an eigenvector of the observable Q with eigenvalue qi , j a0 i is the
initial reference state of the apparatus, and j ai i is the state of the apparatus with the
value ai of the pointer variable. The state of the composite system at time t 1 is
P
j ψ 1 i ¼j φ1 i j a0 i, where j φ1 i ¼ i ci j qi i.
The formalism of generalized contexts can provide a description of the process
involving the possible values qi of the observable Q of system S at time t 1 and the
possible pointer values aj of the apparatus A at time t 2 . These properties are
represented by the projectors

Πqi ¼ jqi ihqi j⊗I A , Πaj ¼ I S ⊗aj ihaj , (6.11)
and satisfy
the compatibility
conditions when translated to the common time t 1 ,
i.e., U 1 Πaj U; Πqi ¼ 0. Therefore, the generalized context formalism allows
computing the conditional probability

Pr ðqi ; t 1 Þ∧ aj ; t 2 hψ j U 1 Πaj UΠqi j ψ 1 i
Pr qi ; t 1 jaj ; t 2 ¼ ¼ 1 ¼ δij : (6.12)
Pr aj ; t 2 hψ 1 j U 1 Πaj U j ψ 1 i
For the composite system prepared in the state j ψ 1 i ¼j φ1 i j a0 i, this result can be
interpreted by saying that if the apparatus’ pointer variable has the value aj after the
measurement, the system S had the property Q ¼ qj before the measurement. More
details of the application of this formalism to the logic of quantum measurements
can be found in our previous works (Vanni and Laura 2013, Losada, Vanni, and
Laura 2016).
6.4.2 The Double-Slit Experiment

We also applied the generalized-context formalism to describe the double-slit
experiment (Losada et al. 2013). A particle in a state represented by a wave packet,
coming from left to right, passes through a double slit at time t 1 . The particle
reaches a vertical zone located to the right of the double slit at a later time t 2 .
We can attempt to give a description of the process involving through which slit
the particle has passed at time t 1 , that appears to be in some region of the vertical
zone at a later time t 2 . As we assume no measurement apparatus, the Hilbert space
to describe this process is the Hilbert space of the particle (H ¼ H particle ). The
projectors representing the particle located in each slit at time t 1 are
ð ð
Πt 1
u
d rjrihrj,
3
Πt1
d
d3 rjr ihrj , (6.13)
Vu Vd
where V u (V d ) is the volume of the upper (lower) slit, and jr i is a generalized
eigenvector of the position operator of the particle with generalized eigenvalue r .
For the later time t 2 , the projectors corresponding to the particle in small regions of
the vertical zone to the right of the double slit are
ð
Πt 2
n
d 3 rjrihrj, (6.14)
Vn
where V n is the volume of the small region of the vertical zone labelled by the
index n. We proved that the properties represented by the projectors Eq. (6.13) and
Eq. (6.14), translated to the common time t 1 , are represented by noncommuting
projectors, i.e.,
h i h i
Πut1 ; U 1 Πnt2 U 6¼ 0, Πdt1 ; U 1 Πnt2 U 6¼ 0, (6.15)
where U ¼ U ðt 2 ; t 1 Þ ¼ eiH 0 ðt2 t1 Þ=ℏ is the unitary evolution generated by the free-
particle Hamiltonian H 0 ¼ p2 =2m. Therefore, our formalism shows the well-
known fact that it is not possible to provide a description of the quantum process
suitable to specify through which slit the particle passed before reaching a region
of the vertical zone.
We also considered a modified double-slit experiment with an ideal measure-
ment apparatus A located in the slits zone, interacting with the particle during the
short time interval ½t 1 ; t 1 þ Δ1 , and with its pointer variable indicating au (ad ) if
the particle is detected passing through the upper (lower) slit. A second ideal
measurement apparatus B is located in the vertical zone to the right of the double
slit, interacting with the particle in the short time interval ½t 2 ; t 2 þ Δ2 , and with a
pointer variable indicating bn if the particle is detected in the small zone labelled by
the index n of the vertical zone. The Hilbert space for the description of this
process is the tensor product of the Hilbert space of the particle and the two Hilbert
spaces of the detectors, i.e., H ¼ H particle ⊗H A ⊗H B . The unitary time evolution is
assumed to be dominated by the interaction between the particle and apparatus A in
the short time interval ½t 1 ; t 1 þ Δ1 , by the free evolution in the time interval
½t1 þ Δ1 ; t 2 , and by the interaction of the particle and apparatus B in the time
interval ½t 2 ; t 2 þ Δ2 .
The possible pointer indications of the apparatus A at time t 1 þ Δ1 are repre-
sented by the projectors
Πat1uþΔ1 I particle ⊗jau ihau j⊗I B , (6.16a)
Πat1dþΔ1 I particle ⊗jad ihad j⊗I B , (6.16b)
and the possible indications of apparatus B at time t 2 þ Δ2 are represented by the

projectors
Πbt2nþΔ2 I particle ⊗I A ⊗jbn ihbn j: (6.17)
We also proved that the properties corresponding to the projectors Eq. (6.16) and
Eq. (6.17), translated to a common time, are represented by commuting projectors.
Therefore, within the generalized-context formalism there is a generalized context
for the composite system’s history that includes the fact that the particle is
measured to pass through a definite slit at a certain time and the fact that the
particle is measured in a definite region of the vertical plane at a later time. The
corresponding conditional probabilities give the expected noninterference pattern.
6.5 Conclusions
We have presented our formalism of generalized contexts for quantum histories. It
was successfully applied to describe the logic of quantum measurements (Vanni
and Laura 2013, Losada et al. 2016) and the results of the double-slit experiment
with and without measurement apparatuses (Losada et al. 2013). It was also
suitable to give a discussion of the decay process (Losada and Laura 2013) and
to provide a deduction of the complementarity principle for the case of the Mach-
Zehnder interferometer (Vanni and Laura 2010).
The compatibility conditions of our formalism impose stronger conditions on
the allowed families of histories than the conditions imposed by the theory of
consistent histories (Losada and Laura 2014a). Therefore, the number of universes
of discourse (families of histories) allowed by our formalism is reduced. Two
important consequences of our formalism are that different families of histories
would not give retrodictions or predictions of contrary properties (Losada and
Laura 2014b), and that any allowed family of histories is organized in a distributive
and orthocomplemented lattice.
However, our formalism is not in position to provide a full interpretation of
quantum mechanics. There are no different allowed families of histories predicting
or retrodicting contrary properties, but there is still the freedom of choice of
different generalized contexts. The formalism in itself gives no indication about
which family should be privileged in a description of the time evolution of a
system. It seems that stronger conditions should be added to our formalism in order
to satisfy a realist perspective. In order to endow this formalism with realist
interpretive content, it is necessary to associate it with a specific interpretation that
is able to be consistently combined with the compatibility condition – but this is

still a work in progress.
Acknowledgments
This work was made possible through the support of Grant 57919 from the John
Templeton Foundation and Grant PICT-2014–2812 from the National Agency of
Scientific and Technological Promotion of Argentina.
References
Castagnino, M., Id Betan, R., Laura, R., Liotta, R. (2002). “Quantum decay processes and
Gamov states,” Journal of Physics A, 35: 6055–6074.
Dowker, F. and Kent, A. (1996). “On the consistent histories approach to quantum
mechanics,” Journal of Statistical Physics, 82: 1575–1646.
Gell-Mann, M. and Hartle, J. B. (1990). “Quantum mechanics in the light of quantum
cosmology,” pp. 425–458 in W. Zurek (ed.), Complexity, Entropy and the Physics of
Information. Reading: Addison-Wesley.
Griffiths, R. (1984). “Consistent histories and the interpretation of quantum mechanics,”
Journal of Statistical Physics, 36: 219–272.
Laloë, F. (2001). “Do we really understand quantum mechanics? Strange correlations,
paradoxes, and theorems,” American Journal of Physics, 69: 655–701.
Laura, R. and Castagnino, M. (1998). “Functional approach for quantum systems with
continuous spectrum,” Physical Review E, 57: 3948–3961.
Laura, R. and Vanni, L. (2009). “Time translation of quantum properties,” Foundations of
Physics, 39: 160–173.
Losada, M. and Laura, R. (2013). “The formalism of generalized contexts and decay
processes,” International Journal of Theoretical Physics, 52: 1289–1299.
Losada, M. and Laura, L. (2014a). “Generalized contexts and consistent histories in
quantum mechanics,” Annals of Physics, 344: 263–274.
Losada, M. and Laura, R. (2014b). “Quantum histories without contrary inferences,”
Annals of Physics, 351: 418–425.
Losada, M. and Lombardi, O. (2018). “Histories in quantum mechanics: distinguishing
between formalism and interpretation,” European Journal for Philosophy of Science,
8: 367–394.
Losada, M., Vanni, L., and Laura, R. (2013). “Probabilities for time-dependent properties
in classical and quantum mechanics,” Physical Review A, 87: 052128.
Losada, M., Vanni, L., and Laura, R. (2016). “The measurement process in the generalized
contexts formalism for quantum histories,” International Journal of Theoretical
Physics, 55: 817–824.
Okon, E. and Sudarsky, D. (2014). “On the consistency of the consistent histories approach
to quantum mechanics,” Foundations of Physics, 44: 19–33.
Omnès, R. (1988). “Logical reformulation of quantum mechanics. I. Foundations,” Jour-
nal of Statistical Physics, 53: 893–932.
Vanni, L. and Laura, R. (2010). “Deducción del principio de complementariedad en la
teoría cuántica,” Epistemología e Historia de la Ciencia. Selección de trabajos de las
XX Jornadas, 16: 647–656.
Vanni, L. and Laura, R. (2013). “The logic of quantum measurements,” International
Journal of Theoretical Physics, 52: 2386–2394.
Part II
Realism, Wave Function, and Primitive Ontology
7
What Is the Quantum Face of Realism?
james ladyman
7.1 Introduction: Realisms and Theories of the Quantum

The main argument of this paper is that there are many forms of realism and many
forms of quantum physics, and that the interaction of the two is apt to confuse (this
section begins to make this case below). The second argument (which reinforces the
first and is presented in the next section) is that there is a considerable tension between
the arguments for scientific realism in the philosophy of science literature and the
invocation of realism as a reason for adopting revisionary interpretations of quantum
mechanics that are popular among philosophers. The third argument (in the third
section) is that scientific realists ought anyway to consider quantum physics as a whole,
where this includes much more than nonrelativistic many-particle quantum mechanics
(NRMPQM) and that doing so does not lend any support to some of the most popular
realist interpretations of quantum mechanics among philosophers. The fourth argument
(made briefly in the last section) is that there is a form of scientific realism that is
compatible with accepting the revolution in our understanding of matter wrought by
quantum physics, and that despite the objections of many philosophers and some
physicists, the revisions to our conception of matter ought not to be undone.
The history of quantum physics teaches us several important things about
realism and the quantum.
(1) There are many different forms of realism that are often run together in
debates about the interpretation of quantum mechanics (see the exchange between
Deutsch and Ladyman in Saunders et al. 2010). Often the triumph of the Copen-
hagen interpretation over Einstein’s “common sense” realism is associated with
general and strong forms of antirealism (Maudlin 2018), and not without justifica-
tion, given some of the weird things that Bohr and others have said about the nature
of reality in the light of quantum physics (for example, Maudlin is among the authors
of Daumer et al. 2006 who strongly object to Wheeler’s and Zeilinger’s ideas about
reality and information being somehow inseparable). However, as Folse (1985)
121
122 James Ladyman
documents, Bohr often made realist claims to go alongside the much quoted anti-
realist ones. For example, in his Nobel Prize lecture Bohr expressed his belief in
atoms and our knowledge of their microscopic constituents. Similarly, presumably
most, if not all, the physicists who immediately applied quantum mechanics to
problems in chemistry and solid-state physics, believed in atoms and electrons. The
fact that many physicists learned to invoke Copenhagen as a way of avoiding
worrying about philosophy does not imply their adherence to any detailed positive
interpretation, and certainly not to anything but a very localized antirealism.
A much discussed survey of physicists, asking them to which interpretation of
quantum mechanics they subscribe, has results as follows for the most discussed
interpretations: 42% Copenhagen, 18% Everett, 0% de Broglie-Bohm, and 9%
Objective Collapse (see Schlosshauer, Kofler, and Zeilinger 2013). Sean Carroll
(2013) says that the lack of agreement makes this the most embarrassing graph in
physics. However, such lack of agreement is not unusual in the history of science.
(It would be interesting to know what they would have said if asked first whether
they accepted metaphysical and scientific realism as defined below.) The physicists
who expressed their support for the Copenhagen interpretation in the poll surely
would not deny that there is a supermassive black hole at the center of the Milky
Way. In this way, quantum orthodoxy sits alongside scientific realism, and it is
plausible that many of those who opted for Copenhagen did so because for many
practicing physicists the measurement problem is irrelevant to their working lives,
and their adherence to Copenhagen orthodoxy goes as far as instrumentalism about
wave functions and collapse when pressed (the idea of wave function collapse as a
real process has nothing to do with Bohr, but is largely due to von Neumann; see
Howard 1985). It is because of the confusion and conflation surrounding the idea
of the Copenhagen orthodoxy that pragmatic attitudes to the interpretation of
quantum mechanics, by which is meant solving the measurement problem, are
taken to be tantamount to idealism or other forms of antirealism about the world in
general and even physics in particular (there is more discussion below of the
different ideas associated with the Copenhagen interpretation).
To bring more precision to the discussion consider the following:
Metaphysical realism. The (physical) world is (largely) independent of our

beliefs and desires.
Scientific realism. Our best scientific theories should be taken literally, as
talking about unobservable entities and processes, and as such successfully
refer, and are approximately true.
These definitions are representative of those in canonical discussions of scientific

realism (see Psillos 1999). Of course, this definition of scientific realism is very
What Is the Quantum Face of Realism? 123
vague without an account of reference and truth, and much ink has been spilled
about these topics without much agreement, and different forms of scientific
realism have proliferated. However, even those who defend the strongest forms
of scientific realism are not required to think that all theoretical terms successfully
refer. For example, a scientific realist may think superstrings have not yet been
shown to exist. Structural realism arose from these discussions and various other
variants including, most relevantly for the present discussion, entity realism.
Nancy Cartwright and Ian Hacking, among others, argued for this view, and it is
the fairly mainstream position in the scientific realism debate that seems to clearly
take the side of the scientific realist where it counts most for scientific practice.
Entity realism. Unobservable entities that are interacted with in the laboratory
should be taken as real, even if the theories that describe them are not taken to
be true.
It should be noted that entity realists are skeptical about claims to completeness
and fundamentality (which Section 7.3 argues are particularly dubious in the case
of quantum mechanics). Entity realism is clearly compatible with an instrumental-
ist view of the wave function and the problem of collapse. The measurement
problem might be supposed to force the adherent of scientific realism to commit
to more, but it is argued that it does not in Section 7.3.
(2) No way was found in practice to distinguish in advance among physical
situations to which quantum physics assigns only probabilities, but which have
different observable outcomes. Paradigmatically, radioactive decay is still con-
sidered random for all intents and purposes in atomic and nuclear physics. John
Stuart Mill (1843) defined determinism as the claim that for any situation there is a
description of it, such that for any other situation satisfying the same description,
the same future will unfold. This may sound overly epistemic to current ears but it
is the way the notion of determinism is applied to the world in scientific practice.
Classical phenomena pass Mill’s test insofar as there is a level of precision of
initial conditions that makes it possible to specify in advance, for example, whether
a fair coin will land heads or tails. Effective classical deterministic behavior of
macroscopic bodies emerges somehow, even though quantum systems fail Mill’s
test as far as anybody has been able to determine. This is a remarkable fact when
one considers the exponential advances and growth in measurement technology
since the discovery of radioactivity more than a century ago. When physicists first
debated the uncertainty relation, they were interested in whether it was a constraint
on practice, and Einstein argued it was not (Bohr 1949). The hidden variables
theorists lost the scientific battle long ago in this sense. Despite the unsoundness of
the no-go theorems for any form of hidden variable, the idea of Bohr – that
124 James Ladyman
quantum mechanics was complete for all practical purposes – has only been
vindicated by subsequent developments.
De Broglie, and later Bohm, showed that the idea that the world is fundamen-
tally random is not logically required by NRMPQM. It is possible to attribute
underlying deterministic trajectories to quantum particles moving in accordance
with the evolution of the wave function. However, since such hidden variables
must be nonlocal in the strong sense, namely that a preferred frame is required in
which it is unambiguously true that the adjustment of a measurement setting
changes the trajectories of particles at spacelike separation, they require revision-
ary physics for no immediate practical use just because they are hidden. These are
unlikely to have been received well by those who had learned to accept relativity,
even if their possibility had not been deliberately or mistakenly ignored by most
physicists. Cushing (1998) speculates that Bell’s theorem could have become
known in the Twenties, and sees this in terms of people accepting that quantum
mechanics is nonlocal. This is highly tendentious, as discussed later, but even
granting it, he must admit that the hidden variables would not have been used in
practice for the problems people went on to solve in all the areas of physics to
which quantum mechanics was applied. He assumes that Dirac’s operator formal-
ism would still have been available to apply quantum mechanics to problems.
Scientific realist literature abounds with the call to take the practice of science
seriously and not to only consider theoretical possibilities. Any form of scientific
realism that does not take quantum randomness at face value is at least out of step
with the practice of science.
Cushing discusses the Forman thesis linking the repudiation of the law of
causality by German quantum theorists with the neoromantic disdain for science
in Weimar culture (see Forman 1971, 1984; for a critique, see Kragh 2002: chapter
10). However, note that Peirce, in the nineteenth century, rejected determinism and
accepted brute randomness in nature within his philosophy, which was based on
his appreciation and study of science (see Ladyman and Ross 2013). Similarly,
both Cushing and Forman offer no explanation of why British physicists such as
Dirac and Mott readily embraced the new quantum mechanics. More importantly,
there is an equivocation on causality in this context, for while Kant and others took
it to be the same thing as determinism in the sense of there being necessary and
sufficient antecedent events for everything that happens, in the twentieth century,
probabilistic causality became the norm as a result of the extension of causal
modeling to the behavioral and social sciences.
(3) Quantum mechanics was successfully applied in chemistry and solid state
physics in the first months and years of its existence. It was also more or less
immediately extended to relativistic physics and to field theory. All of its
seemingly impossible implications that have been tested have been confirmed. It is
applied throughout the rest of physics, and quantum statistical mechanics and
semiclassical domains have been discovered. However, it has not been successfully
applied to gravity itself (though quantum field theory [QFT] has been applied to the
effects of gravity in the form of Hawking radiation) and so is not a complete theory
of reality. In philosophical discussions of physics, people often consider models in
which there is only a spacetime and some fields or particles of some particle kind.
This generates the idea of the ontology of the theory, as if it purported to be an
account of the whole of reality rather than just an aspect of it. For example,
diffraction and Stern-Gerlach experiments on beams of particles or individual ones,
and EPR-type experiments on entangled systems, are a tiny proportion of the
applications of quantum physics to reality. Any viable form of quantum realism
must not conflate NRMPQM with quantum physics, and it must be apt for phenom-
ena other than familiar, simple experiments that are not representative of quantum
physics, though they are legitimately used to understand the conceptual and math-
ematical foundations of the theory and how it represents physical systems.
7.2 Scientific Realism and the Interpretation of NRMPQM:

The Case of Bohm Theory
As yet, the only grounds for adopting Bohm theory are philosophical, namely
because it is a causal and realist version of quantum theory that solves the
measurement problem and eliminates in principle puzzles about the weird behavior
of quantum systems. However, this section argues that to adopt Bohm theory is to
cut the philosophical ground from under the feet of the realist. Much of what is said
is relevant to other examples of alternatives to quantum orthodoxy, but the case of
relativistic dynamical collapse theories is in many ways very different, insofar as
they make predictions that differ from those of quantum mechanics.
In discussions in the philosophy of science, scientific realism is usually
articulated in terms of the theories that the relevant scientific community has
accepted. The question is whether the scientific community adopting some theory
of unobservables compels belief in their existence. It is standardly assumed on all
sides that theory choice is rational and that the methods scientists use to solve the
underdetermination problem in practice are appropriate. The mere theoretical
availability of empirically equivalent alternatives to accepted theories is not taken
to be sufficient for skepticism about the latter. In philosophy of science, those who
offer sociological explanations for theory choice in science are usually antirealists,
and the invocation of theoretical empirical equivalence of a rival research program
as a reason for skepticism about the orthodoxy is commonly associated with
126 James Ladyman
constructive empiricism (van Fraassen 1980), which is the most discussed alterna-
tive to realism. It is ironic then that the positive case for Bohmian mechanics
involves the combination of an insistence on a realist interpretation of quantum
mechanics based on the theoretical empirical equivalence of the formalism, with
the claim that the reasons why Bohmian mechanics was not accepted by the
scientific community are extrascientific. Hence, the aforementioned discussion of
the Forman thesis by Cushing, and the emphasis on how discussion of the
measurement problem, and the work of de Broglie, Bohm, and Everett was
supressed. (Bohm’s ideas were not well known until after the work of Bell, and
the current relative prominence of many-worlds interpretations is due to the revival
of Everett’s ideas by others. Saunders et al. 2010 includes discussion of all this and
a very comprehensive relevant citation.)
Realist responses to the underdetermination problem usually aim at rationally
reconstructing theory choice, for example, in the case of wave versus particle
theories of light, special relativity versus Lorentzian contraction, and so on. The
theoretical virtues of novel predictive success, non-ad hocness, explanatory power,
simplicity, coherence with background metaphysics, and so on, are discussed to
explain why mere empirical equivalence is not sufficient to make the community’s
choice arbitrary and why the chosen theory scientifically preferable. In this context,
historical theories of confirmation and the idea of progressive versus degenerating
research programs are much discussed following the work of Popper, Lakatos, and
Musgrave. Among the theoretical vices are the negation of each of the virtues just
described, such as ad hocness and ontological profligacy. Another theoretical vice
that is often discussed is parasitism, which is when a theory requires another theory
for its formulation. An extreme example is the theory that says “the world is as if
QED”; it generates no predictions at all without the input of QED. The wave and
particle theories of light were not like this. Ray optics fits well with the mechanics
of particles, and diffraction fits well with waves.
The main virtue of Bohm theory for many of its current defenders is that it offers
the best prospects for realism about the quantum world. This is because Bohm
theory posits definite values for all physical quantities at all times, eliminating the
apparent indeterminacy in the values of physical quantities prior to measurement.
The Bohmian mechanics usually discussed is such that all particles have well-
defined trajectories at all times and the evolution of all quantum systems is entirely
deterministic and causal. Thus particles can be individuated by their spatiotem-
poral properties (although of course these will not be accessible to experiment). It
is also virtuous in cohering with background ideas of causality and determinism,
and of the nature of particles. However, it is important to note that, as Harvey
Brown and others (1996) argue, the “particles” of Bohm theory are not those of
classical thought. The dynamics of the theory is such that the properties normally
associated with particles like mass, charge, and so on, are in fact, all inherent in the
quantum wave function and not in the particles. It seems that the particles only have
position. Apart from any worries we might have about the intelligibility of this
notion of particle, it seems that they have none of the features of classical particles
other than point position; hence, there would seem to be little referential continuity
for ‘particle’ available to the realist. Furthermore, if the trajectories are enough to
individuate particles in Bohm theory, what makes the difference between an “empty”
trajectory and an “occupied” one? Since none of the physical properties ascribed to
the particle inheres in points of the trajectory, giving content to the claim that there is
actually a “particle” there would seem to require some notion of the raw stuff of the
particle; in other words, haeccities seem to be needed to make the ontology of Bohm
theory intelligible after all. This is without even beginning the discussion of the
nature of the wave function in Bohm theory (Brown et al. 1996 also argue that the
action–reaction principle is violated by Bohm theory).
Bohm theory is often criticized for its alleged ad hocness and lack of simplicity
relative to the standard formalism. These complaints aside, more problematic is that it
is nonlocal in the strong sense that, if Bohm theory is correct, then physicists will
have to rewrite their textbooks so that a Newtonian or Galilean spacetime with a
global time coordinate underlies the Lorentz invariant phenomena of electromagnet-
ism. When such theories are proposed as alternatives to relativity, they are sometimes
linked to cosmic conspiracies because they propose far-fetched hidden machinations.
Similarly, instantaneous action at a distance as a physical process has been widely
regarded as unscientific since the beginnings of field theory in the mid-nineteenth
century, but Bohm theory makes it a pervasive feature of the world, and it requires
that the change of setting on the apparatus on one side has a physical effect on the
trajectory of the particle on the other side. Unsurprisingly, Bohmians always argue
that quantum mechanics requires action at a distance anyway, but this is not correct.
Bell’s theorem does not show that there is an objective causal asymmetry between the
two wings of the Aspect experiment, only that the attribution of possessed values, or
counterfactual definiteness, or some other condition, is incompatible with locality.
There is no proof of Bell’s theorem without what is often termed a “realist”
assumption of some kind, but note that none of these realist assumptions are required
by either metaphysical or scientific realism.
The so-called pessimistic meta-induction argument against scientific realism
would be reinforced if Bohm theory is adopted by most scientists for two reasons:
(a) It would support the arguments of those who argue that we cannot learn
metaphysical lessons from science, because the orthodoxy among physicists
would have been quite wrong about the world since at least 1935, insofar as it
has been widely held and taught that quantum phenomena are genuinely
128 James Ladyman
indeterministic – that electrons and other “particles” are really neither waves
nor particles, that matter has new properties like spin, and so on.
(b) The acceptance of relativity theory as a theory of the nature of space and time,
such that there is no absolute simultaneity or privileged coordinate system, and
such that the Lorentz invariance of Maxwell’s equations has been taken to be
not merely empirical but also reflective of fundamental symmetries in the
nature of reality, would have to be regarded as quite mistaken. The more cases
there are of the ontological interpretation of scientific theories being later
abandoned, the more compelling the meta-induction becomes.
Overall, skepticism about scientific realism will be encouraged if Bohm theory is

adopted, because it will have to be accepted that most scientists have been wrong
about most of the most important metaphysical implications of twentieth-century
physics for most of the twentieth century.
Furthermore, the positive arguments for realism will also be undermined if
Bohm theory replaces standard quantum theory. Consider the most straightforward
defense of realism by inference to the best explanation: according to the realist,
since the quantum theory of spin has been so empirically successful, we ought to
believe that the best explanation of this fact is that there really are subatomic
particles with spin states. By contrast, Bohm theory denies that spin is a funda-
mental property, reducing it to some aspect of particle trajectories and their
interaction with the wave function. Can we explain the importance of spin in
physics and chemistry if Bohmian mechanics is true? Is it plausible that quantum
chemistry and quantum field theory would have been developed without realism
about spin and other quantum numbers? As argued previously, we are owed more
than an appeal to the in principle empirical equivalence of the theories.
Similarly, the no-miracles argument is undermined by the empirical success of
relativity theory if we are to interpret its ascription of deep structure to spacetime as
merely empirically useful and not to be believed. The more cases there are of great
empirical success where a realistic construal of the theory is not available, the
weaker the realist’s argument that the history and practice of science is only
intelligible if scientific realism is adopted.
Bohm theory requires the plausibility of a counterfactual history. In the light of
the discussion of underdetermination and how realists appeal to the role of theories
in heuristics, the onus is on the defender of the alternative theory to give a detailed
account of a plausible counterfactual history that leads to equally productive
physics. Cushing’s case depends only on the high-level theoretical equivalence
of Bohmian and quantum mechanics. The situation is even worse when it comes to
interacting QFT and the standard model, because no fully worked out Bohmian
alternatives exist.
According to van Fraassen, the big problem with scientific realism is that it adds
metaphysics to scientific theories for no empirical gain, and this seems to be true in
the case of Bohm theory. In this context we might ask: What has the interpretation
of quantum mechanics ever done for us? To which the answer is a lot, most
obviously, it has more or less directly given us Bell’s theorem (Bohm), quantum
computation and information processing (Deutsch), and weak values (Aharonov).
As with the theoretical virtues, there is no support for any one interpretation
from consideration of fecundity for physics, except perhaps for the Copenhagen
interpretation, insofar as the orthodoxy, such as it was and is, has undoubtedly
accompanied the most extraordinary scientific success. Quantum physics has been
outstandingly virtuous on all criteria other than that of cohering with background
metaphysics. For all the theory’s indeterminism, when it is taken to the relativistic
domain, it is as predictively accurate in practice as any determinist could reason-
ably demand. It does not tell us what the fundamental nature of being is, in terms of
something we can think of as objectively and separately located, with intrinsic
properties in space and time. However, it is not a complete theory of the world, as
discussed in the next section.
7.3 Realism about What?

There is no lost age of ontological purity in science. Classical physics is not just point
particle mechanics, and the existence of ethers and other weird kinds of matter was
posited throughout the history of physics to explain the phenomena of electricity, heat,
light, and magnetism as well as in chemistry. No theory we have ever had in physics
has had a settled metaphysical interpretation or been a plausible theory of the whole
world. As argued in Section 7.1, not being a realist about NRMPQM is completely
compatible with metaphysical realism, entity realism, and scientific realism. More-
over, it has not been our best physical theory since very soon after its inception. For
this reason it is bizarre to seek an interpretation of it as if it were complete. As pointed
out in the last section, the true object of any kind of scientific realism that goes beyond
effective entity realism is our best physics as a whole. Yet realists arguing about
NRMPQM often espouse instrumentalism about QFT because of its mathematical
inexactitudes and the fact that it does not present a clear conceptual framework
(outside of algebraic QFT, which some philosophers think is apt to study accordingly).
Several points that are not always noted are relevant to the measurement
problem in this context:
(1) The time evolution of open systems is not unitary.
(2) There is no empirical warrant for applying quantum theory to macroscopic
objects outside of very special circumstances.
130 James Ladyman
(3) There many cases in physics and science generally in which “more is
different.”
For all these reasons it seems that the measurement problem does not compel a
choice between Everett, Bohm, and dynamical collapse. Maudlin (2018) quotes
Lakatos, saying that Bohr and his associates brought about the “defeat of reason
within modern physics.” However, all the revisionist interpretations of NRMPQM
are ultimately parasitic, in practice as well as in theory, on the great empirical
success of standard quantum physics, which they have not yet matched, even in
principle and with the benefit of hindsight. They may say there is no collapse but
they all have some notion of “effective collapse.” In the two-slit experiment, no
matter what the mechanism for detecting which slit a particle went through,
determining the position always has the same effect on the determinacy of the
momentum in accordance with the uncertainty relation.
7.4 Against Counterrevolutionary Conservativism

It is not exaggerating to describe the replacement of classical mechanics and field
theory by quantum physics as a revolution in physics (in the colloquial, not the
Kuhnian theoretical sense) because so much changed. In particular, the distinction
between matter and radiation broke down, and everyday matter is understood to be
stable and apparently continuous and solid because of the quantum nature of the
atom and how its parts interact electromagnetically. Very small things do not
behave like big things made small. This should not now be so hard to swallow.
Major scientific advances almost always require conceptual innovation and the
abandonment of the kind of assumptions we get from common sense and the
manifest image. Invariably, things are more complicated than they first seem. As
argued in the previous sections, rejecting the realist rivals to standard quantum
physics does not require abandoning metaphysical or scientific realism. Indeed,
forms of scientific realism based on the ideas of structure and pattern rather than of
material particles (Ladyman and Ross 2007), which are compatible with accepting
the revolution in our understanding of matter wrought by quantum physics, are
specifically motivated by our scientific understanding of what matter is like. From
a naturalistic point of view, revisionary physics should be a last resort.
In spite of the pressure to adopt a realist interpretation of NRMPQM, saying “a
plague on all your houses” is compatible with both realism and reason. Wave
functions do an excellent job of representing the physical state of systems that in
practice are sufficiently separable from the rest of the world to have an effective
state of their own. There should be no ontology of the wave function because wave
functions are representations of, for example, the momentum and spin of particles,
not extraphysical properties of particles. Realists insist that the “beables” of

quantum mechanics be specified. Prima facie they are the pure states of the theory,
which can only be attributed to systems separately if they are not entangled.
Schrödinger time evolution only applies to systems that are effectively isolated,
and does not apply to the whole universe or to open systems. As yet, nobody
knows where to make the final cut.
References
Bohr, N. (1949). “Discussions with Einstein on epistemological problems in atomic
physics,” pp. 200–241 in P. A. Schilpp, Albert Einstein: Philosopher-Scientist.
Evanston: The Library of Living Philosophers.
Brown, H., Elby, A., and Weingard, R. (1996). “Cause and effect in the pilot-wave
interpetation of quantum mechanics,” pp. 309–319 in J. T. Cushing, A. Fine, and
S. Goldstein (eds.), Bohmian Mechanics and Quantum Theory: An Appraisal.
Dordrecht: Kluwer Academic Publishers.
Carroll, S. (2013). “The most embarrassing graph in modern physics,” www.prepost
erousuniverse.com/blog/2013/01/17/the-most-embarrassing-graph-in-modern-physics/
Cushing, J. T. (1998). Philosophical Concepts in Physics: The Historical Relation between
Philosophy and Scientific Theories. New York: Cambridge University Press.
Daumer, M., Dürr, D., Goldstein, S., Maudlin, T., Tumulka, R., and Zanghì, N. (2006).
“The message of the quantum?,” pp. 129–132 in A. Bassi, D. Dürr, T. Weber, and
N. Zanghì (eds.), Quantum Mechanics: Are There Quantum Jumps? and On the
Present Status of Quantum Mechanics, AIP Conference Proceedings. College Park,
Maryland: American Institute of Physics.
Folse, H. J. (1985). The Philosophy of Niels Bohr: The Framework of Complementarity.
Amsterdam-Oxford: North-Holland.
Forman, P. (1971). “Weimar culture, causality, and quantum theory: Adaptation by
German physicists and mathematicians to a hostile environment,” Historical Studies
in the Physical Sciences, 3: 1–115.
Forman, P. (1984). “Kausalität, Anschaullichkeit, and Individualität, or how cultural
values prescribed the character and lessons ascribed to quantum mechanics,”
pp. 333–347 in N. Stehr and V. Meja (eds.), Society and Knowledge. New Bruns-
wick-London: Transaction Books.
Howard, D. (1985). “Einstein on locality and separability,” Studies in History and Phil-
osophy of Science, 16: 171–201.
Kragh, H. (2002). Quantum Generations: A History of Physics in the Twentieth Century.
Princeton: Princeton University Press.
Ladyman, J. and Ross, D. (2007). Every Thing Must Go. Metaphysics Naturalized. Oxford-
New York: Oxford University Press.
Ladyman, J. and Ross, D. (2013). “The world in the data,” pp. 108–150 in D. Ross,
J. Ladyman, and H. Kincaid (eds.), Scientific Metaphysics. Oxford: Oxford University
Press.
Maudlin, T. (2018). “The defeat of reason,” Boston Review, http://bostonreview.net/sci
ence-nature-philosophy-religion/tim-maudlin-defeat-reason
Mill, J. S. (1843). “System of logic,” pp. 1963–1991 in J. M. Robson (ed.), The Collected
Works of John Stuart Mill. Toronto: University of Toronto Press, London: Routledge
and Kegan Paul.
132 James Ladyman
Psillos, S. (1999). Scientific Realism: How Science Tracks Truth. London: Routledge.
Saunders, S., Barrett, J., Kent, A., and Wallace, D. (2010). Many Worlds? Everett,
Quantum Theory, and Reality. Oxford: Oxford University Press.
Schlosshauer, M., Kofler, J., and Zeilinger, A. (2013). “A snapshot of foundational
attitudes toward quantum mechanics,” Studies in History and Philosophy of Modern
Physics, 44: 222–230.
van Fraassen, B. (1980). Scientific Image. Oxford: Oxford University Press.
8
To Be a Realist about Quantum Theory
hans halvorson
8.1 Introduction
There is a story that some philosophers have been going around telling. It goes
something like this:
The pioneers of quantum mechanics – Bohr, Heisenberg, Dirac, et al. – simply abandoned hope
of providing a realist theory of the microworld. Instead, these physicists settled for a
calculational recipe, or statistical algorithm, for predicting the results of measurements. In
short, Bohr et al. held an antirealist or operationalist or instrumentalist view of quantum theory.
Implicit in this story is a contrast with the “traditional aspirations of science” to

describe an observer-independent reality. Having built up a sense of looming crisis
for science, the storyteller then introduces us to the heroes, those who would stay
true to the traditional aspirations of science.
As the 20th century moved into its second half, there arose a generation of renegade
physicists with the courage to stand up against antirealism and operationalism. These
valiant men – David Bohm, Hugh Everett, John Bell – renewed the call for a realist
theory of the microworld.
This kind of story can be very appealing. It is the age-old “good guys versus bad guys”
or “us versus them” motif. And those “ist” words make it easy to distinguish the good
guys from the bad, sort of like the white and black hats of the classic westerns.
The story is brought into clearer focus by talking about the quantum wave
function. What divides the realists from the antirealists, it is said, is their respective
attitudes toward the wave function: Antirealists treat it as “just a bookkeeping
device,” whereas realists believe it has “ontological status.” Witness the faux-
historical account of Roger Penrose:
It was part of the Copenhagen interpretation of quantum mechanics to take this latter
viewpoint, and according to various other schools of thought also, ψ is to be regarded as a
calculational convenience with no ontological status other than to be part of the state of
133
134 Hans Halvorson
mind of the experimenter or theoretician, so that the actual results of observation can be
probabilistically assessed.
(Penrose 2016: 198)
I suppose that Penrose can be forgiven for oversimplifying matters, as well as for
propagating the myth of the “Copenhagen interpretation” (see Howard 2004).
After all, there can be great value in simple fictional tales if they get readers
interested in the issues.
I also imagine that Sean Carroll is aiming to generate some heat – rather more
heat than light – when he poses the following dilemma about the wave function:
The simplest possibility is that the quantum wave function isn’t a bookkeeping device at
all . . .; the wave function simply represents reality directly.
(Carroll 2017: 167)
This seemingly simple dilemma – ontological status: yes or no, – is a fine device
for popular science writing, which should not demand too much from the reader.
But is it really the right place to locate a pivot point? Is the question, “Ought I to
commit ontologically to the wave function?” the right one to be asking?
Popular science writers are not the only ones to have located a fulcrum at this point.
In fact, some philosophers say that if you are a scientific realist, then you are logically
compelled to accept the Everett interpretation. I am thinking of this kind of argument:
If you’re a realist about quantum theory, then you must grant ontological status to the
quantum state. If you grant ontological status to the quantum state, and if quantum
mechanics is true, then unitary dynamics is universal. Under these conditions, realism
and unitary dynamics, you have two options: either you accept the completeness of
quantum theory, or you don’t. And if you accept the completeness of quantum theory,
then the Everett interpretation is true.
In short, we are told that the following implication holds:

Realism þ Pure QM ) Everett
Notice how much work realism is supposed to do in this implication!
You might accuse me of caricature, and I am sure I have left out much of the
nuance in this argument. And yet, Everettians regularly gesture in this direction. For
example, Wallace (2013) claims that the Everett interpretation “is really just quan-
tum mechanics itself understood in a conventionally realist fashion,” and that “there
is one pure interpretation which purports to be realist in a completely conventional
sense: the Everett interpretation” (Wallace 2008). Similarly, Saunders claims that if
we don’t think of the wave function as a measure of our ignorance, then
the only other serious alternative (to realists) is quantum state realism, the view that the
quantum state is physically real, changing in time according to the unitary equations and,
somehow, also in accordance with the measurement postulates.
(Saunders 2010: 4)
To Be a Realist about Quantum Theory 135
In short, if you are a good realist, then you will say that the quantum state is
physically real, and from there it is a short step to the Everett interpretation.
There is something strange about this sort of argument. The notion of “realism”
is doing so much of the work – and yet, nobody has told us what it means. How
could the “if realism, then Everett” argument be valid when “realism” has not been
defined clearly? And how could the argument be convincing when realism has not
been motivated, except through its undeniable emotional appeal?
In this chapter, I will take a closer look at the distinction between realist and
antirealist views of the quantum state. I will argue that this binary classification
should be reconceived as a continuum of different views about which properties of
the quantum state are representationally significant. What is more, the extreme
cases – all or none – are simply absurd and should be rejected by all parties. In
other words, no sane person should advocate extreme realism or antirealism about
the quantum state. And if we focus on the reasonable views, it is no longer clear
who counts as a realist and who counts as an antirealist. Among those taking a
more reasonable intermediate view, we find figures such as Bohr and Carnap – in
stark opposition to the stories we have been told.
8.2 Extremists
Suppose that you were asked to list historical figures on two sheets of paper: On
the first sheet, you are supposed to list realists (about the quantum state), and on
the second sheet you are supposed to list antirealists (about the quantum
state). Suppose that you are asked to sort through all of the big names of quantum
theory – Bohr, Heisenberg, Dirac, Bohm, Everett, etc.
I imagine that this task would be difficult, and the outcome might be controver-
sial. For almost none of these people ever explicitly said, “I am a realist” or “I am
an antirealist” or “the wave function has ontological status” or anything like that.
You would have to do quite a bit of interpretative work before you could justify
assigning a person to one of the lists. You would have to assess that person’s
attitude toward the quantum state by studying their behavior and utterances with
respect to it. For example, if person X makes free use of the collapse postulate, with
no proposed physical mechanism, then you might surmise that X is either a mind-
body dualist, or an operationalist about the quantum state, or both. In other words,
an operationalist stance might serve as the best explanation for X’s utterances and
behavior.
The task of sorting people into realist and antirealist would be simpler
for contemporary figures, who seem happy to embrace one of these two labels. For
example, Sean Carroll and Lev Vaidman will tell you, with great passion, that the
wave function is just as real as – in fact, more real than! – a rock, or a tree, or your
spouse. In contrast, Carlo Rovelli speaks of the wave function as Laplace spoke of
136 Hans Halvorson
God: Je n’avais pas besoin de cette hypothèse-là. These are just a few examples
among the many philosophers and physicists who have openly labeled themselves
as realist or antirealist about the quantum state. Self-identified state realists include
Esfeld, Goldstein, Ney, Saunders, Wallace, Zanghì, etc. Self-identified state anti-
realists include Bub, Fuchs, Healey, Peres, etc. The battle lines have been clearly
drawn, but what is at stake?
The right-wing extremists say: Quantum wave functions are things. That view is
silly. The left-wing extremists say: Quantum wave functions are just bookkeeping
devices. That view is just as silly.
8.3 Right-Wing Extremists

One might think that the litmus test for realism about quantum theory could be
posed as:
Do you believe that the wave function (more generally, the quantum state)
exists?
Or, as Callender (2015) puts it,
Is the quantum state part of the furniture of the world?
So, when Carroll (2017: 142) says that, “the basic stuff of reality is a quantum
wave function,” he is declaring his allegiance to wave function realism.
But what is this wave function thingy? Should I be thinking about it like
I think about chairs or tables? No, say the philosophers; you have to be a bit
more sophisticated about it. The preferred ontological reading of the wave
function is as a field, on analogy to things like the magnetic field that surrounds
the earth. Thus, to push the ontological picture further, things are represented
by points in the domain of that field, and the properties of those things are the
values of that field. What then are the things according to this ontological
view? Some philosophers say that the wave function is a field on configuration
space (Albert 1996, Ney 2012, North 2013), so that the things are points of
configuration space. Others say that the wave function is a multi-field on
physical space (Forrest 1988, Belot 2012, Chen 2017), so that the things are
spacetime points.
These straightforwardly ontological views have been subjected to many criti-
cisms (see Wallace and Timpson 2010, Belot 2012). Here I want to raise another
kind of objection. Or rather, I want to make a request of the ψ-field theorists:
Would you please describe your theory clearly, including its states, properties, and
the relationship between them? To my mind, the attraction of ψ-field theories is
due in large extent to the vague realist associations that they conjure up in our
heads: The wave function is a thing with a definite shape! I wager that such theories
are plausible only to the extent that it is unclear what they are really saying.
For starters, in quantum theory, the primary theoretical role of the wave function
ψ is as a state. The ψ-field theorists ask us to change our point of view. Instead of
thinking of ψ as a state, we are to think of ψ as a field configuration. There are
numerous problems with this proposal.
In classical physical theories, the word “state” is shorthand for “a maximally
consistent list of properties that could be possessed by the system simultaneously,”
or equivalently, “an assignment of properties to objects.” In that case, there are two
possible things we could mean by the sentence “the state σ exists.” First, we could
mean that the list of properties exists. But this list is an abstract mathematical
object, which would exist whether or not the corresponding theory is true. So, in
this first sense, “σ exists” is not interesting from the point of view of physics.
Second, we could use “σ exists” as an obscure shorthand for “σ is the actual state,”
which, in turn, is shorthand for saying that certain other objects have certain
properties. Thus, in this second case, “σ exists” is cashed out in terms that do
not refer to σ at all. In philosophers’ lingo, “σ exists” is grounded in facts about
other objects, and so is not really about σ at all.
Now, the defender of quantum state realism might simply say: “That was
classical physics. In quantum physics, the state takes on a new role.” I certainly
accept that quantum physics changes some of the ways we talk about the physical
world. But I am not so sure that it makes sense to reify states. According to the
normal senses of “object” and “state,” we affirm that objects can be in states. Thus,
if states are objects, then states themselves can be in states. But then, to be
consistent, we should reify the states of those states, and these new states will
have their own states, ad infinitum. In short, if you run roughshod over the
grammatical rules governing the word “state,” then you can expect some strange
results.
To continue that line of thought, we assume that things can be counted. In other
words, it makes sense to ask: How many things are there? But then, if states were
things, it would make sense to ask: How many states are there? But now I am
completely puzzled. According to quantum theory, the universe has an infinite
number of potential states, but only one actual state. What in the world would
explain the absence of all the intermediate possibilities? Why couldn’t there have
been 17 states? And what’s more, why do physicists never raise as an empirical
question: How many states are there? The reason is simple: Physicists do not treat
states as they do things, not even in the extended sense where fields also count as
things.
I hope that by this stage you are at least partially convinced that it does violence
to the logic of physical theories to talk about states as if they were things. But then
138 Hans Halvorson
you should agree that the role of a wave function is not to denote an object.
Moreover, if ψ does not denote a physical object, then the properties of ψ do not
directly represent the properties of a physical object. Granted, we should be careful
with this latter claim. Even in classical physics, the properties of a state can
represent, albeit indirectly, the properties of a physical object. For example, for a
classical particle, “being in subset Δ of statespace” is a property of states that
represents a corresponding property of the relevant particle. Nonetheless, there are
two different types of things here – the particle, which is a concrete physical object,
and its state, which is an abstract mathematical object. The latter tells us about the
former, but should not be conflated with it.
In classical theories, there is also a sharp distinction between instantaneous
configurations and states. If a configuration is represented by a point in the
manifold M, then a state is represented by a point in the cotangent bundle
T ∗ M. In many scenarios, T ∗ M looks like a Cartesian product M M, where
the first coordinate gives instantaneous configuration, and the second compon-
ent gives momentum. In every case, there is a projection mapping
π : T ∗ M ! M, and the preimage of any particular configuration q 2 M is an
infinite subset of T ∗ M. But now, if ψ is both a state and a field configuration,
then it is unclear where it lives. Does ψ live in the space M of configurations,
or does it live in the space T ∗ M of states? How can it do both jobs at the
same time?
These considerations show that the ψ-field view stretches the logic of classical
physics beyond the breaking point. To treat ψ as representing a field configuration
is to disregard its primary theoretical role as a state. Or, at the very least, to treat it
thus would obscure the difference, central to classical theories, between configur-
ations and states.
If that is not enough trouble for ψ-field views, we can also ask them to give an
account of the properties that are possessed by this thing, the ψ-field. Recall that in
a classical theory with statespace S, properties are typically represented by subsets
of statespace S. (We might require that these subsets be measurable or something
like that. But that point will not matter in this discussion.) Then, we say that the
system has property E S just in case it is in state σ 2 E.
Now, ψ-field theorists would like us to think of quantum theory on the model of
a classical field theory. In this case, the statespace would be the space C ∞ ðX Þ of
smooth complex-valued solutions to some field equation, and subsets of C ∞ ðX Þ
would represent properties that the system can possess. (Let us ignore here the fact
that classical field theories typically use the space of real-value functions.) For
example, for any field state f 2 C ∞ ðX Þ, the singleton set ff g represents the
property of being in state f , and its complement C ∞ ðX Þ∖ff g represents the property
of not being in state f .
For the purpose of performing certain calculations, a classical field theorist

might complete C∞ ðX Þ relative to some norm, obtaining a Hilbert space, such as
the space L2 ðX Þ of (equivalence classes of ) square-integrable functions on X. The
elements in L2 ðX Þ are no longer smooth functions, and in fact, they are not really
functions at Ðall – they are equivalence classes of functions under the relation: f g
just in case j f g j dμ ¼ 0.
In contrast, for a quantum theorist, L2 ðX Þ is simply one instance of, or one
representation of, a Hilbert space H of countably infinite dimension. Any two
Hilbert spaces of the same dimension are isomorphic, so it does not matter (for the
physics) which one we choose. The states of the system are represented not by
points in L2 ðX Þ, but by rays. And the properties of the system are represented not
by subsets of L2 ðX Þ but by closed subspaces. Thus, in short, while L2 ðX Þ is used by
both the classical field theorist and the quantum mechanic, it is used in completely
different ways in the two cases. For those of us who believe that the Hilbert space
formalism is intended to represent reality, we could say that it represents reality in
a very different way than a classical field theory does.
The ψ-field views ask us to forget the differences between quantum mechanics
and classical field theories. But it will not be easy to forget these differences
without doing violence to the representational role of the various pieces of the
Hilbert space formalism. A classical theory comes with representatives (subsets of
statespace) for many properties that are not represented in the Hilbert space
formalism. The ψ-field theorist wants to lay claim to all these properties – for this
seems to provide the coveted “god’s-eye view” of reality. In effect, the ψ-field
picture is designed to make us feel like we have evaded the Kochen-Specker
theorem. For, if physical properties are represented by subsets of L2 ðX Þ, or by
the mathematical properties of a function ψ 2 L2 ðX Þ, then each such property is
either definitely possessed or definitely not possessed when the system is in state ψ.
This view is intended to hide (or ignore? or deny?) the fact that a quantum state
does not answer all questions about which properties are possessed.
What is more, the ψ-field view only follows classical physics as far as giving an
instantaneous snapshot of the possessed properties. As soon as it comes to drawing
inferences about the system, it imposes ad hoc rules to block fallacious inferences.
For example, in a classical field theory, if σ and σ 0 are distinct field states, then
knowing that the system is in state σ permits you to assert that the system is not in
state σ 0 . Or in probabilistic terms, the probability of σ 0 conditional on σ is 0. If you
carry that inference rule over directly into quantum theory, then you will make
false predictions. A Gaussian function ψ centered at 0 is a different field state than
a Gaussian function ψ 0 centered at 0:01. Thus, on a classical picture, the property E
of “being in state ψ” is inconsistent with the property E0 of “being in state ψ 0 ,” and
PrðE 0 jE Þ ¼ 0. But quantum theory says that PrðE0 jE Þ 1.
140 Hans Halvorson
Of course, ψ-field theorists are too clever to fall into the trap of carrying
classical inference rules into the quantum domain. Although they purport to view
the ψ-field classically, they stop short when it comes to reasoning about it. For the
purposes of reasoning and making predictions, they turn to the Hilbert space
formalism to guide them. Thus, we might summarize the attitude of these ψ-field
views in a phrase: You can look at the world from the god’s-eye point of view; just
don’t reason about it as god would.
8.4 Left-Wing Extremists

At one extreme, we have people telling us that the wave function is part of the
furniture of reality. At the opposite extreme, we have people telling us that the
wave function is “a mere calculational device” (Rovelli 2016: 1229), and that “it is
mistaken to view the universal wave-function as a beable” (Healey 2015). This
second group of extremists is a curious bunch. They protest loudly against the
wave function, producing elaborate (and interesting) arguments against its onto-
logical status. And yet, they cannot seem to live without it. In their books and
articles, they accord a privileged role to the wave function. When they want to say
something true about a quantum system, they consult the wave function before
anything else. It makes one wonder: If they do not believe in the wave function,
then why do they grant it a special role in their representations of reality?
The practice in physics, followed by realists and antirealists alike, is that each
classically described “preparation” or “experimental setup” may be represented
by a unique quantum state. In fact, the ability to associate quantum states to
classically described experiments is one of the skills that displays mastery of
quantum theory. Once an experiment has been adequately described, then there is
no remaining latitude for idiosyncratic or subjective state assignment. There is
just one correct state, as will be borne out by checking the statistics of measure-
ment outcomes. The fact that physicists have correctness standards for quantum
state assignments strongly suggests that they grant these states some sort of
representational role.
Healey (2017) makes exactly this point, and he uses it to make an argument for
quantum state objectivism, i.e., the belief that there are objectively correct ascrip-
tions of quantum states to physical systems. But isn’t this sort of state objectivism
strictly inconsistent with state antirealism? If the quantum state is not real, then
how could one be wrong about the quantum state? In order to answer this question,
Healey engages in subtle reasoning about how objective correctness can be disen-
tangled from the correspondence theory of truth and about how the meaning of the
quantum state can be accounted for in an inferential theory of content. This just
goes to show that matters are not as simple as they initially appeared to be.
Healey’s subtlety is laudable, but sometimes it verges on doublespeak. For

example, Healey hangs much on the distinction between ascribing a state ψ to a
thing X, and describing X with ψ.
Pragmatists agree with QBists [quantum Bayesians] that quantum theory should not be
thought to offer a description or representation of physical reality: in particular, to ascribe a
quantum state is not to describe physical reality.
(Healey 2016: on line, emphasis added)
What are we supposed to be doing when we “ascribe” a quantum state? If ψ has no

representational role whatsoever, then why speak of “ascribing” it to a physical
object or situation? Why not just speak of “using” the wave function – as one uses
a computer or a hammer – to get a job done?
In the English language, the word “ascribe” involves a subject postulating a
relation between two objects: S ascribes Y to X. More is true: In normal conversa-
tion, to ascribe Y to X involves judging that there is a preexisting relation between
Y and X. For example, “He ascribed Jane’s short temper to her upset stomach.” In
this way, ascribing is different than using: I can use Y to do something to X without
making any judgment about the relation between Y and X. These considerations
show that the word “ascribe” is tantalizingly close to other words – such as
“describe” – that connote the existence of a representational relation, exactly the
sort of thing that Healey wishes to deny. To consistently carry out his pragmatist
program, Healey should use a different word than “ascribe.”
Here is what I think is really going on here. The phrase “Y describes X” is rather
vague; and being vague, it can be thought to license all sorts of inferences about the
relationship between Y and X. When people say that “Y describes X,” they tend to
import a lot of baggage that goes far beyond the simple existence of a representa-
tional relation between Y and X. In fact, it seems all too easy to fall into the mistake
of thinking that Y describes X only if Y is similar to X. Of course that is not true:
The phrase “is over 6 feet tall” describes Goliath, but this phrase is not similar to
Goliath.
That temptation to assume similarity is all the more difficult to resist when the
first argument of “Y describes X” is a geometrical object, such as a wave function.
The reason we fall into this trap, I assume, is because we do frequently use
geometric objects as pictorial representations. For example, I might draw a rect-
angle on a piece of a paper and say, “This rectangle describes the shape of my
desk.” In this case, the rectangle on the paper is indeed similar to the desk in a well-
defined mathematical sense.
Healey, Rovelli, and other self-proclaimed antirealists have surrendered too
much to their opponents. They have allowed their opponents to define words like
“ontological status” and “describes.” Then, because Healey and Rovelli reject the
142 Hans Halvorson
implications that come along with this particular definition of “describes,” they are
forced to say that the quantum state does not describe at all. Thus, Healey and
Rovelli lay themselves open to the charge of antirealism – which, of course, carries
highly negative connotations. To be an antirealist implies a sort of failure of
courage – it implies a sort of retreat. Ergo, Healey and Rovelli are seen as making
less bold assertions about reality than their realist counterparts are making.
8.5 The State as Directly Representing

Are you a realist about the quantum state? We have already seen that this question
cannot be paraphrased as: Do you believe that the quantum state exists? So what
could the question mean? According to Wallace and Timpson,
Traditionally realist interpretations . . . read the quantum state literally, as itself standing
directly for a part of the ontology of the theory.
(Wallace and Timpson 2010: 703)
In fact, Wallace (2018b) locates the crucial divide between “representational” and
“non-representational” views of the quantum state. Thus the shift is signaled from
the material mode of speech (does the state exist?) to the formal mode of speech
(does the state represent?). In particular, Wallace and Timpson claim that realism
involves commitment to both literal and direct representation. Thus, Carroll utters
the shibboleth when he says, “the wave function simply represents reality directly”
(Carroll 2017: 167). But what work is the word “directly” doing here? I am led to
think that the task of representing must be a bit like getting to work, where you
have to take the right turns in order to follow the most direct route. So what are the
instructions for following the direct route to representation?
When a person says that Y represents X, then that typically signals that the
person endorses some inferences of the form
(†): If Y has property ϕ, then X has property ϕ0 ,
where ϕ ↦ ϕ0 is some particular association of properties (the details of which need
not detain us). Let us call (†) a property transfer rule. For example, if I say that a
certain map represents Buenos Aires, then I mean that some facts about Buenos
Aires can be inferred from facts about the map.
What then is the force of insisting that Y does not merely represent X, but that it
represents X directly? I suspect that the word “directly” is supposed to signal
endorsement of quite liberal use of property transfer rules. But just how liberal?
The key question to keep in mind is: Which specification of permitted property-
transfer inferences corresponds most closely to the notion of “direct representa-
tion” that is favored by realists such as Carroll, Wallace, Saunders, and Timpson?
When they say that “ψ directly represents reality,” what exactly are they saying
about the relation between ψ and the world?
Consider first the proposal:
(DR1) Y directly represents X just in case every property of Y is also a property of X.
This proposal is logically consistent, but also absurd. One of the properties that X
has is being identical to X. Thus, according to DR1, if Y directly represents X then
Y ¼ X. Could Wallace and Timpson possible intend this? Does Carroll mean to
say our universe is a subset of R3n C? If so, then scientific realism is truly a
radical point of view. The wave function is an abstract mathematical object. Thus,
if the universe is a wave function, then the universe is an abstract mathematical
object. Perhaps mathematicians will applaud this conclusion, because then pure
mathematics tells us everything there is to be known about the universe.
I suspect that the realists do not mean their direct representation claim in the
sense of DR1. Let us try a more reasonable proposal.
(DR2) Y directly represents X just in case each mathematical property of Y corres-
ponds to some physical property of X.
Here we need some precise account of the “mathematical properties” of Y.
According to standard set-theoretic foundations of mathematics, the mathematical
properties of Y are precisely those properties that can be described in the language
of Zermelo-Fraenkel (ZF) set theory. Thus, for example, the mathematical proper-
ties of Y would include its size (cardinality). In contrast, arbitrary predicates in
natural language do not pick out mathematical properties of Y. For example, “is an
abstract object” cannot be articulated in ZF set theory, and so would not count as a
mathematical property of Y. Thus, DR2 does not say that “anything goes” in terms
of the representationally significant properties of Y.
Even so, DR2 is still implausibly profligate in the number of representationally
significant properties it assigns to the wave function. In particular, for each
definable name c in ZF set theory, there is a definable predicate Θc given by
Θc ðSÞ $ c 2 S:
Among these definable set names, we have ∅, f∅g, and so on. Now, a wave
function is a function ψ : A ! B, with domain set d 0 f ¼ A and codomain set
d1 f ¼ B. Thus, for any definable name c, it makes sense to ask whether
Θc ðd0 ψ Þ, i.e., whether c is contained in the domain of ψ.
Imagine now the following scenario. Two physicists, Jack and Jill, are arguing
about whose wave function is a better representation of the universe. The funny
thing is, Jack and Jill’s wave functions are both Gaussians, centered on 0, and with
the same standard deviation. If you ask Jack to draw a picture of his wave function,
then he draws a Gaussian centered at 0. If you ask Jill to draw a picture of her wave
144 Hans Halvorson
function, then she also draws a Gaussian centered at 0. They agree that this picture
is a correct representation of their respective wave functions. They also agree that
their wave functions are written in the configuration space basis, and that the origin
0 represents the same point in the universe. It seems that there is nothing left for
them to disagree about.
And yet, Jack and Jill insist that their wave functions cannot both be correct.
According to Jack, the correct wave function ψ has the property that Θ∅ ðd 0 ψ Þ, that
is, the empty set is an element of the domain of the wave function. According to
Jill, the correct wave function ψ 0 does not have that property. They both believe
that Θ∅ corresponds to a genuine physical property. Jack asserts that this property
is instantiated, and Jill asserts that it is not.
Jack and Jill would fail their quantum mechanics course. They do not under-
stand how the theory works. In using the formalism of quantum theory to represent
reality, we do not care about these fine-grained set theoretic differences. If
two wave functions have the same shape, then we consider them to be the same.
If two wave functions can be described via the same equation, then we take them to
be identical. But what is this notion of same shape that we are using here? How can
we tell when two wave functions are the same, at least for the purpose of doing
physics?
At this point, we might want to lay down the ace card of recent philosophy of
science: the notion of isomorphism. Can’t we just say that two wave functions are
representationally equivalent just in case they are isomorphic? In this case, we
could then propose the following criterion for direct representation:
(DR3) Y directly represents X just in case Y and X are isomorphic.
This proposal sounds a lot more plausible than the previous two – especially
because the word “isomorphism” is simultaneously precise (within certain fixed
contexts) and flexible (since it means different things in different contexts). But
that is precisely the problem with DR3: the phrase “Y is isomorphic to X” is no
better defined than the phrase “Y directly represents X.”
In mathematics, isomorphism is a category-relative concept. If you hand me two
mathematical objects and ask, “Are they isomorphic?” then I should reply by asking
“Which category do they belong to?” For example, two mathematical objects can be
isomorphic qua groups, but nonisomorphic qua topological spaces. Thus, it makes no
sense to say that a mathematical object is isomorphic to the world tout court. In order
to make sense, we would first have to specify a relevant type (or category) of
mathematical objects. For example, one might say that the world is isomorphic to a
topological space Y, as shorthand for saying that the world has topological structure,
and is in this sense isomorphic to Y. But if you give me a concrete mathematical object
A and say that the world is isomorphic to A, then I have no idea what you are saying.
So, if we want to say that the world is isomorphic to a wave function ψ, then we
need to say what category of mathematical objects we take ψ to belong to. And that
is not going to be easy, for ψ is not a group, or a topological space, or a
differentiable manifold, or any other of the standard types of mathematical struc-
ture. There is no category of wave functions; and, there is no nontrivial notion of
isomorphism between wave functions. It will not help to say that ψ and ψ 0 are
isomorphic wave functions just in case there is a unitary symmetry U such that
Uψ ¼ ψ 0 , for in that case, all wave functions would be isomorphic. The closest we
come to finding a home for ψ is in the category of Hilbert spaces: ψ is an element
of a Hilbert space, which is an object in the category of Hilbert spaces. But that will
not help, because we do not want to say that the world is isomorphic to the Hilbert
space H, but that it is isomorphic to a particular wave function ψ.
There are numerous other problems with analyses of representation in terms of
isomorphism, some of which are discussed in a recent article by Frigg and Nguyen
(2016). We mention two further problems here, each of which might be taken to
deliver a fatal blow to the account. First, an isomorphism is a function between two
mathematical objects, and the world is not a mathematical object. In fact, as
pointed out long ago by Reichenbach (1965), the only grip we have on the
structure of the world is by means of our representations.
Second, our account of representational significance should mesh with our
account of theoretical equivalence, and many philosophers of science hold views
of theoretical equivalence according to which equivalent theories need not have
isomorphic models. For example, Halvorson (2012) labels this view as “the model
isomorphism criterion of theoretical equivalence,” and he argues that it must be
rejected. However, if the model isomorphism criterion of theoretical equivalence is
rejected, then we must also reject the claim that representation entails isomorphism
between the world and one of the theory’s models. We can argue as follows: If two
theories T and T 0 are equivalent, and if T is representationally adequate, then T 0 is
also representationally adequate. But if the models of T are not isomorphic to the
models of T 0 , then it cannot be the case that the world is isomorphic to a model of
T and also to a model of T 0 . Therefore, to say that T is representationally adequate
does not entail that the world is isomorphic to one of the models of T.
8.6 Representationally Significant Properties

As we have seen, isomorphism-based analyses of representation have difficulty
explaining how wave functions represent – because there is no obvious candidate
notion of “isomorphism” for wave functions. Perhaps, however, we can attack this
problem from the other side. Having a notion of isomorphism in place gives us a
criterion for identifying representationally significant properties:
146 Hans Halvorson
A property ϕ is representationally significant just in case ϕ is invariant under

isomorphism.
But of course, we need not already have a notion of isomorphism in place to
choose the representationally significant properties. We can simply say what those
properties are.
As we know, it would be disastrous to propose that all mathematical proper-
ties of a wave function are representationally significant. For example, the
property of “having domain that contains the element f∅; f∅gg” is a perfectly
good mathematical property that a wave function either possesses or does not
possess. But nobody, to my knowledge, has ever proposed that this mathemat-
ical property represents a bona fide physical property. In practice, we simply do
not care whether we use a wave function ψ that has that property or a similar
wave function ψ 0 that lacks that property. Many of these set-theoretically
definable properties of a wave function are routinely ignored as “surplus
mathematical structure.”
In my experience, physicists cannot usually say explicitly which properties of ψ
are representationally significant. However, we can determine which properties of
ψ they care about by watching what they do. If they treat two wave functions ψ and
ψ 0 as interchangeable, then their behavior suggests that they accord no representa-
tional significance to properties that separate these two functions. Here we say that
a property Θ separates ψ and ψ 0 just in case Θðψ Þ and ¬Θðψ 0 Þ.
The art of discriminating between wave functions is not so unlike the fabled art
of “chicken sexing.” The skilled chicken-sexer has the ability to judge reliably
whether two chicks are of the same sex. But if you ask what criteria he or she is
using, the chicken-sexer will be at a loss for words. In the same way, the skilled
quantum mechanic has the ability to judge whether two wave functions are
representationally equivalent. And he or she displays his or her judgment of
representational equivalence by his or her disinterest in the question: Which of
these two wave functions provides the correct representation of reality?
I am not sure that it would be possible to give a fully explicit account of the
equivalence relation of “representational equivalence” for wave functions. None-
theless, there are certain sufficient conditions for representational equivalence that
are uncontroversial.
First, two wave functions are representationally equivalent if one is a complex
multiple of the other, i.e., if they lie in the same ray in Hilbert space. Thus, if a
property Θ of wave functions is not invariant under this relation (of lying in the
same ray) then Θ is not representationally significant. For example, consider the
property Θ given by
Θðψ Þ $ ðψ ð0Þ ¼ 1Þ:
Clearly there are two functions ψ and ψ 0 such that ψ ψ 0 , but Θðψ Þ and ¬Θðψ 0 Þ.
Therefore, Θ is not a representationally significant property of wave functions.
Second, wave functions are not actually functions at all. In fact, the space of
square integrable functions on configuration space is not a Hilbert space. Instead,
to define a positive-definite inner product, one has to take equivalence classes of
functions relative to the equivalence relation of “agreeing except on a set of
measure zero.” But now consider the property Θ defined by:

Θðψ Þ $ jψ ð0Þj2 ¼ 1 :
Again, there are two functions ψ and ψ 0 such that ψ ψ 0 , but Θðψ Þ and ¬Θðψ 0 Þ.
Therefore, Θ is not a representationally significant property of wave functions.
This is not to say that there are no representationally significant properties of
wave functions. For example, consider the property
ð
Θðψ Þ $ j ψ ðxÞ j dμðxÞ ¼ 1:
Δ
This property Θ can be shown to be invariant under the equivalence relations mentioned
previously. Indeed, practitioners of quantum theory know exactly what this property is:
It is the property ½Q 2 Δ of being located in the region Δ. What other invariant
properties are there? Can we give some sort of systematic description of them?
As mentioned before, the Hilbert space formalism is normally taken to represent
properties by means of the subspaces of the statespace. Let us think about how this
works in the case of the space L2 ðX Þ of (equivalence classes of ) wave functions.
What does a subspace of L2 ðX Þ look like? Some subspaces correspond to proper-
ties of functions. For example, consider the property
Θðψ Þ ψ has support in the region Δ:
It is not difficult to see that the set of functions satisfying Θ forms a closed
subspace of L2 ðX Þ. But not every subspace of L2 ðX Þ has such an interpretation
in terms of straightforwardly geometric features of functions. For example, let
U : L2 ðK Þ ! L2 ðX Þ be the unitary isomorphism between the momentum-space
and position-space representation of wave functions. Now begin by defining the
same sort of subspace, but relative to the momentum-space representation. That is,
let E be the subspace of L2 ðK Þ consisting of functions with support in Δ. The
natural interpretation of E is: having momentum value in the set Δ. Then U ðE Þ is a
subspace of L2 ðX Þ, and hence, represents a quantum-theoretic property Θ. But this
property Θ does not manifest itself as a natural property of functions on the original
configuration space X. Indeed, it is not clear that it would be possible to express Θ
without making reference to the isomorphism between L2 ðK Þ and L2 ðX Þ.
148 Hans Halvorson
We have here a nice concrete example of an issue that philosophers have

been discussing in the abstract – the issue of abundant versus sparse views of
properties (see Bricker 1996). The Hilbert space formalism gives a special
version of the sparse view of properties: Not every subset of L2 ðX Þ corres-
ponds to a natural property. One might think initially that this sparse view
makes life difficult by preventing us from saying certain things. For example,
as Wallace (2012) points out, this sparse view entails that “has a definite value
of energy” fails to pick out a property (a consequence which he finds to be
unacceptable).
However, there is an obvious problem with trying to take an abundant view of
the properties of quantum theory, i.e., with taking every subset of L2 ðX Þ to pick out
a physical property. The problem is that there are too many such subsets, and their
physical interpretation is unclear. Nonetheless, the Hilbert space formalism pro-
vides a method for identifying those subsets of L2 ðX Þ that represent physical
properties. In particular, we have the following result:
(SQ) Let H be an abstract Hilbert space of countably infinite dimension. Then each
subspace of H is of the form U 1 ½Z 2 Δ, where U : H ! L2 ðRÞ is a unitary
isomorphism, Δ is a Borel subset of R, and ½Z 2 Δ is the subspace of functions
with support in Δ.
(This result is part of the folklore of functional analysis, and may be reconstructed
from the results in chapter 9 of the book by Kadison and Ringrose [1991].) Here,
we think of L2 ðRÞ as wave functions of some particular dynamical variable Z,
which could be position (along some axis), or momentum (along some axis), or
energy, or . . . In this case, U 1 ½Z 2 Δ is the subspace of wave functions where the
value of Z lies in Δ. In other words, U 1 ½Z 2 Δ and ½Z 2 Δ represent the same
property – only, this property’s physical interpretation is more perspicuous in the
latter case.
Thus, there is a nontrivial question about which properties of functions (i.e.,
subsets of L2 ðX Þ) represent bona fide, or “natural,” physical properties. Take an
arbitrary mathematical predicate of functions, such as
Θðψ Þ ψ is a smooth ði:e:; infinitely differentiableÞ function,
which seems to be quite natural, at least from a mathematical point of view. But
why suppose that Θ represents a natural physical property? What criteria should
we use to sort out the genuine predicates from the spurious predicates? Some might
suggest an operationalist criterion:
(operationalist) A predicate Θ of wave functions represents a natural physical property if
and only if there is a measurement that would verify whether an object’s state ψ has
property Θ.
But that criterion is too imprecise. And, in any case, the operationalist criterion is
stricter than quantum theory’s own criterion, which countenances many natural
properties that cannot be operationally detected.
The language of quantum theory, represented via the Hilbert space formalism,
comes with a vocabulary, including a list of predicates.
(QM properties) A predicate Θ of wave functions represents a natural physical
property if and only if the set fψ 2 L2 ðX Þ j Θðψ Þg is a subspace of L2 ðX Þ.
By this result, the previous criterion can be restated as follows:
(QM properties) A predicate Θ of wave functions represents a natural physical
property if and only if there is a dynamical variable Z, and a measurable Δ R, such
that Θðψ Þ if and only if ψ lies in the subspace ½Z 2 Δ.
These predicates can then be taken as giving quantum theory’s preferred account of
natural properties. In short, the natural properties are precisely those picked out by
saying that a quantity Z has value in a certain range.
So, we return to the original question: If Θðψ Þ is the predicate “ψ is a
smooth function,” then does Θ pick out a physical property of wave functions?
Quantum theory answers this question by saying: Θ represents a physical
property only if there is some quantity Z such that that Θ picks out the subspace
½Z 2 Δ.
When we talk about giving a “physical interpretation” to a subset E of state-
space, the demand is not that E be given an operational interpretation, as, e.g.,
corresponding to some measurement operation. Instead, we are simply asking that
the mathematical object E be describable in words that have some antecedent
physical meaning. It is simply the demand that we understand what the formalism
purports to represent.
8.7 Reading the State Literally

Recall that Wallace and Timpson say that a quantum state realist does two things:
(1) he or she believes that the state represents reality directly, and (2) he or she
reads the state literally. As we saw, there are various ways of cashing out “Y
directly represents X.” If you push the notion to the extreme, where Y ¼ X, you
will end up saying stupid things; however, as soon as you start to nuance this
notion, you start to sound less like a full-blooded realist.
So can we find a firm foothold for realism in the second criterion? Is it the
commitment to a “literal reading of the state” that sets the quantum state realists
apart from their antirealist counterparts? Here we have tapped into a central vein in
philosophical discussions of scientific realism. For example, 40 odd years ago, van
Fraassen described scientific realism as the belief that
150 Hans Halvorson
The aim of science is to give us a literally true story of what the world is like; and the
proper form of acceptance of a theory is to believe that it is true.
(van Fraassen 1976: 623, emphasis added)
The debates of the last 40 years seem not to have brought into question the
connection between realism and literalism. In a recent authoritative account of
scientific realism, Chakravartty reasserts the connection:
Semantically, realism is committed to a literal interpretation of scientific claims about
the world.
(Chakravartty 2017: on line, emphasis added)
But something fishy must be going on here. The idea that a scientific theory is a set
of claims (i.e., sentences) fell out of favor about 40 years ago. Nowadays, most
philosophers of science say that a scientific theory consists of a collection of
models, plus some claim to the effect that one of these models represents the
world. But if a theory is a collection of models, then how am I supposed to read it
literally? Nor can this problem be brushed away by adopting a different view of
scientific theories. For better or worse, the theories of mathematical physics
involve collections of mathematical models, such as Lorentzian manifolds, Hilbert
spaces, etc. So how then are we supposed to read these theories literally?
The answer, in short, seems to be: To read a theory literally is to take one of its
models M as a reliable guide to features of the world. But now we are right back to
where we were when considering analyses of “Y directly represents X.” If I am a
literalist about M, then which features of M should I take to be representationally
significant? The simple answer “all features of M” leads immediately to absurdity.
The answer “all mathematical features of M” also leads to a bizarre and untenable
picture. Thus, we are thrown back on a more piecemeal approach, where one has to
know how to interpret the model M, which means being able to distinguish its
representationally significant properties from the insignificant ones.
Indeed, learning how to use a physical theory requires that learning the art of
“reading claims off” of a model. Consider, for example, the general theory of
relativity (GTR), where a model M is a Lorentzian manifold. What might it look
like to read M literally? Well, GTR claims that at each point p 2 M, there is a four-
dimensional tangent space T p . And living on top of T p there is an infinite tower of
ðm; nÞ tensors, for all natural numbers m and n. Are these things I have just said
among the “scientific claims” of GTR? If I am a realist about GTR, then am
I committed to these claims? Should I envision an infinitely extended tangent space
T p of four dimensions sitting on the tip of my nose, and indeed, a different such
tangent space for each instant of time? Are these tangent spaces “part of the
furniture of the world”? If this is what it means to be a realist about GTR, then
Einstein himself was no realist.
To make the point more clearly, GTR entails that

For each point p 2 M, there is an open neighborhood O of p, and a coordinate chart
ϕ : O ! R4 .
These coordinate charts are just as much elements of a model of GTR as a wave
function is an element of a model of QM. Thus, if literalism demands commitment
to the wave function ψ, then it also demands commitment to the coordinate chart ϕ.
If quantum state realism is just a “literal reading of QM,” then coordinate chart
realism is just a “literal reading of GTR.”
If you do not think that GTR involves a commitment to an ontology of tangent
spaces, coordinate charts, etc., then I can only agree: Not every true statement,
made within the language of a theory, is one of the “scientific claims” of that
theory. To say that a model M accurately represents the physical world does not
mean that every mathematical thing in M represents a physical thing. Realism,
according to Chakravartty, Timpson, Wallace, van Fraassen, et al., requires com-
mitment to the scientific claims of a theory, interpreted literally. But you cannot
interpret a mathematical object literally. That simply does not make sense. The
demand for literal interpretation only makes sense after we have used the formal-
ism to express claims in a language that we understand.
Here we have to lay some blame at the door of the semantic view of theories.
The semantic view of theories plus realism suggests the idea that one ought to
interpret models literally – an idea that can lead to absurd consequences if not
further nuanced. A model’s elements need not all play the same representational
role. For example, suppose that I make a map of Princeton University, on which
I draw several buildings. Suppose that I also draw a picture of a compass in the
lower right hand corner of my map – to indicate its orientation. Now, I am a
realist about the geography of Princeton, and I believe that my map is a faithful
representation of it. But that does not mean that I believe there is a huge compass
lying on the ground just outside of the university. Nor would I say that the
compass on the map is “just a bookkeeping device” or that it “has no representa-
tional role.” The compass does have a representational role: It represents a claim
about how my map is related to the actual town of Princeton. And if this compass
can be said to have a representational role, then so can a wave function. (For an
illuminating investigation of the notion of “literal interpretation,” see Hirsch
2017).
8.8 Spacetime State Realism

The most recent development in the realist ontology program is the proposal to
upgrade wave function realism to “spacetime state realism” (see Wallace and
152 Hans Halvorson
Timpson 2010; Wallace 2018a). But does this technical maneuver dodge the
various philosophical problems that confront wave function realism? In order to
press the question further, we need to sketch the idea behind spacetime state
realism.
Let us begin with the simplest (and least interesting) case of spacetime state
realism – the case where spacetime consists of a single point. In this case, we
represent a quantum system by means of a C∗ -algebra A of observables (For
an account of this formalism, see Ruetsche 2011). The important point is that
A is closed under operations of addition, multiplication, and conjugation
A ↦ A∗ . Moreover, there is a preferred multiplicative unit I 2 A, the identity
operator. The prototypical case of a C ∗ -algebra is the algebra of n n complex
matrices.
We need a few definitions. An operator A 2 A is said to be self-adjoint just in
case A∗ ¼ A, and A is said to be positive just in case A ¼ B∗ B for some operator
B 2 A. A function ω : A ! C is said to be a linear functional just in case
ωðcA þ BÞ ¼ cωðAÞ þ ωðBÞ for all A, B 2 A and c 2 C. A linear functional ω is
said to be positive just in case ωðAÞ 0 for every positive operator A 2 A.
A positive linear functional ω is said to be a state just in case ωðI Þ ¼ 1. We will
use Σ ðAÞ to denote the space of states of A.
We can formulate quantum mechanics in the language of C∗ -algebras just as
well as we can in the language of Hilbert spaces. Indeed, the self-adjoint operators
in A represent observables (or more accurately, quantities), and the elements of
Σ ðAÞ represent physical states. As a particular case in point, if A is the algebra of
2 2 matrices, then the self-adjoint operators are simply the Hermitian matrices,
and the states on A correspond one-to-one with density operators on C2 via the
equation
ωðAÞ ¼ TrðW ω AÞ:
With these definitions in hand, we can state Wallace and Timpson’s proposal quite
simply:
For a system represented by the algebra A, the properties correspond one-to-one with the
states in Σ ðAÞ.
This proposal can be made more picturesque and plausible if you think of a “field
of states,” where each point p in spacetime is assigned a state ωp . And if you feel
that this is just empty mathematics, then it might help to think of the typical case,
where ωp is represented concretely by a density operator W p . Then the field
p ↦ W p of density operators starts to look more like a classical field configuration,
where some mathematical object, such as a tensor, is assigned to each point in
space. The only mathematical difference is that W p is a complex matrix instead of a
tensor. But as Wallace and Timpson point out, the relative unfamiliarity of
complex matrices such as W p should not rule them out as legitimate values of a
physical field.
To this point, I agree with Wallace and Timpson. What bothers me is not the
difference between tensors and complex matrices. What bothers me is the confla-
tion of the various theoretical roles of states, quantities, and properties. The typical
job of states is to assign values to quantities. So, if we ask states also to serve as
values of quantities, then the job of states will be to assign states. In order to try to
keep things straight in our heads, we might try to declare some “types.” First, the
standard way of thinking of states is that they are of type Q ! V, where Q is the
quantity type, and V is the value type. But now, Wallace and Timpson tell us that
states are also of type V. In this case, states would be both of type Q ! V and of
type V, resulting in a type confusion.
What’s more, we typically ask a physical theory to provide some sort of “state-
to-property” link. For example, the so-called orthodox interpretation of quantum
theory proposes the eigenstate-eigenvalue link:
(EE link) A property E of the system is possessed in state ψ just in case Eψ ¼ ψ.
Wallace and Timpson also propose a state-to-property link. However, their prop-
erties are of the form “being in state W,” and so their proposal reduces to:
(WT link) A system has property W when it is in state W
Or perhaps it would be better to say:
(WT link) A system has the property of being in state W just in case it is in state W.
I suppose this claim is true. But I did not need to learn any physics to draw that
conclusion. This is nothing more than a disquotational theory of truth.
Is it possible that Wallace and Timpson’s proposal only trivializes in the trivial
case – where spacetime consists of a single point? Perhaps their proposal is only
meant to give an interesting picture in the case where we associate a different
algebra of observables AðOÞ to each region O of spacetime. In that case, their
recipe would yield a much richer structure, something like a co-presheaf of states
(see Swanson 2018). But I do not see any reason to think that this additional
mathematical structure can undo the conflation of states and properties that already
occurs at the level of individual algebras.
Finally, even if you can get past these other worries, there is a worry that the
Wallace-Timpson proposal shows too much. Indeed, there is a case to be made that
any reasonable generalized probability theory can be formulated in the framework
of C ∗ -algebras. In that case, it would seem that the Wallace-Timpson proposal
yields a realistic physical ontology for any reasonable generalized probability
theory. In other words, it is realism on the cheap.
154 Hans Halvorson
8.9 The Wave Function as Symbol

We began our discussion with the dilemma: Either the quantum state has onto-
logical status, or it does not. We saw that this dilemma cannot be taken seriously,
because a state is not a candidate for existence or nonexistence in the physical
sense. Thus the original ontological dilemma was transformed into a representa-
tional one: Either the quantum state represents reality, or it does not. But then we
discovered that “represents” can be understood in many different senses – and in
the most extremely realistic sense of “represents,” no sane person would say that
the quantum state represents the world. Thus, the disagreement between realists
and antirealists – where it is not simply a matter of emotional associations with
words – boils down to different stories about how to use the quantum state to
represent reality.
It is ironic then, that early interpreters of quantum theory – such as Bohr and
Carnap – are often assumed to be operationalists about the quantum state. That
could not be further from the truth. Both Bohr and Carnap explicitly say that the
wave function is not merely a calculational device. Presumably, somebody ran a
word-search on Bohr and Carnap’s writings, and having found no hits for “ψ
represents reality directly” or “ψ has ontological status,” they concluded that these
guys must have been antirealists.
There is another possibility that we should take seriously. What if Bohr and Carnap
intentionally exercise caution with words like ontological status and direct represen-
tation, because those words might lead to a misunderstanding of how quantum theory
works? Perhaps Bohr and Carnap were groping their way, if ever so haltingly, toward
a more articulate account of how the wave function represents reality.
8.9.1 Bohr
Analytic philosophers have been quick to categorize Bohr as an operationalist
about the wave function, citing statements like this one:
the symbolic aspect of Schrödinger’s wave functions appears immediately from the use of
a multidimensional coordinate space, essential for their representation in the case of atomic
systems with several electrons.
(Bohr 1932: 370)
Faye, for example, seems to think that Bohr’s use of “symbolic” is code for
“should not be taken literally.”
Thus [for Bohr], the state vector is symbolic. Here “symbolic” means that the state vector’s
representational function should not be taken literally but be considered a tool for the
calculation of probabilities of observables.
(Faye 2014: on line)
Faye’s confusion here is understandable. We analytic philosophers of science tend

to associate “symbolic” with “nonreferential” or “uninterpreted.” In particular,
with regard to a sentence X in a formal calculus, to say that “X is symbolic” means
precisely that X in uninterpreted, and so lacks a truth value. In other words, when
we hear “symbol,” we immediately think “does not purport to describe reality.”
However, the considerations of previous sections show that this usage of
“symbolic” does not make much sense when X is a mathematical object, such as
a wave function. Nor would it make much sense to attribute this usage of
“symbolic” to Bohr, who did not use words in exactly the way that analytic
philosophers of science have come to use them. When Bohr uses “symbolic,”
I assume that his meaning draws on a his peculiar educational background, which
was heavy on continental figures such as Kant, Goethe, Hegel, and Helmholtz.
I assume that his meaning was also shaped by his interactions with continental-
type philosophers such as Ernst Cassirer, and mathematicians such as his brother
Harald. Thus, when Bohr says something is “symbolic,” we should not immedi-
ately conclude that he means it in the sense of the uninterpreted predicate calculus.
Indeed, one of Bohr’s students, Christian Møller, asked him explicitly what he
meant by calling the wave function “symbolic.” In a 1928 letter, Bohr replies in so
many (!) words:
Regarding the question discussed in your letter about what was meant, when I in my article
in Naturwissenschaften, emphasized so strongly the quantum-theoretical method’s
symbolic character, I am naturally in complete agreement with you that every
description of natural phenomena must be based on symbols. I merely sought to
emphasize the fact, that this circumstance – that in quantum theory, we typically use the
same symbols we use in the classical theory – doesn’t justify our ignoring the large
difference between these theories, and in particular necessitates the greatest caution in
the use of the intuitive concepts [anskuelsformer] to which the classical symbols are
connected. Naturally, one doesn’t easily run this danger with the matrix formulation,
where the calculation rules, which diverge so greatly from the previously standard
algebraic ones, hold quantum theory’s special nature before our eyes. Furthermore, to
use the word “symbolic” for non-commutative algebra is a way of speaking that goes back
long before quantum theory, and which has entered into standard mathematical
terminology. When one thinks about the wave theory, it is precisely its “visualizability”
[anskuelighed] which is simultaneously its strength and its snare, and here by emphasizing
the approach’s [behandlingens] symbolic character, I was trying to bring to mind the
differences – required by the quantum postulate – from classical theories, which are hardly
ever sufficiently heeded.
(Bohr 1928, original in Danish)
As is typical with reading Bohr, one does not feel that the situation has been greatly
clarified. However, one thing is clear: Bohr does not intend to single out the
quantum state for operational treatment. If Bohr is an antirealist about the quantum
156 Hans Halvorson
state, then he is an antirealist about all of mathematical physics. For Bohr, all
mathematical representation is “symbolic,” whether observable or unobservable
aspects of reality are being represented. Among the symbolic representations of
physics, he would include the F ab of Maxwell’s equations, the gab of general
relativity, as well as functions representing the trajectories of material bodies
through spacetime. Bohr’s point might be summed up simply by saying that
mathematical objects are not sentences, and so they cannot “be read literally.”
To understand Bohr’s use of “symbolic,” it might also help to look at a
philosopher whose career ran in parallel with his. In fact, it is well known that
Bohr interacted extensively with Ernst Cassirer when the latter was composing his
book Determinismus und Indeterminismus in der Modernen Physik, first published
in 1937. Whether there is a more substantial overlap in their usage of “symbolic”
will have to await more detailed historical investigations.
Nonetheless, it is clear that there are many common themes in the views of Bohr
and Cassirer (see e.g., Pringe 2014). One such common theme is giving careful
thought to the way that mathematical objects can be used to represent the physical
world. In putting forward his views on this issue, Cassirer is clear that “symbolic”
should not be opposed to “representational.” The interesting question is not
whether something is representational, but rather how it represents. In particular,
Cassirer believes that the development of mathematics and physics in the ninteenth
century provides a particularly clear demonstration of the need to expand the
notion of representation beyond a simplistic “similarity of content” account.
Mathematicians and physicists were first to gain a clear awareness of this symbolic character
of their basic implements. . . . In place of the vague demand for a similarity of content
between image and thing, we now find expressed a highly complex logical relation, a general
intellectual condition, which the basic concepts of physical knowledge must satisfy.
(Cassirer 1955: 75)
For the former, more narrow, use of symbols, Cassirer uses the word Darstellings-
funktion. For the latter, more general, use of symbols, Cassirer uses the word
Bedeutungsfunktion. Thus, to relate back to our earlier analysis of “Y represents
X,” we might think that Darstellungsfunktion picks out a kind of representational
relation that licenses many inferences about X from Y, especially inferences having
to do with spatiotemporal properties. The paradigm case, of course, of such
representations are the directly geometric. In contrast, Bedeutungsfunktion picks
out a more general kind of representation relation that does not imply geometric
similarity between X and Y.
Bohr does not avail himself of Cassirer’s classification of symbolic forms.
However, he often does speak of things being “unvisualizable” (uanskuelig) –
opening a door to the deep dark recesses of the Kantian tradition. Bohr’s notion of
representing something to visual intuition doubtless overlaps in important ways

with Cassirer’s notion of Darstellungsfunktion. And if there is any coherence in
Cassirer’s idea of moving toward Bedeutungsfunktion, then Bohr may be blazing
the same trail. In particular, when Bohr says that subatomic processes cannot be
visualized, he should not be taken as saying that quantum theory is nonrepresenta-
tional. Instead, Bohr might be groping his way toward a more nuanced account of
how mathematics can be used to represent physical reality.
8.9.2 Carnap
We began the chapter with a story about how the early interpreters of quantum
theory were operationalists. That story is often neatly combined with another story
that post-Quinean analytic philosophers love to tell: the story about how silly and
stupid the logical positivists were. According to this story, the logical positivists
viewed scientific theories as “mere calculi” for deriving predictions. Thus, the
story concludes, it is no surprise that Bohr et al. were operationalists about the
quantum state, given that operationalism had so thoroughly infected the prevailing
view of scientific theories.
If you have ever read a serious historical account of the origins of quantum
theory, you know that the first story is mostly propaganda. None of the pioneers of
quantum theory – Bohr, Heisenberg, Dirac, etc. – was a crass operationalist. And if
you have ever read a serious historical account of twentieth-century philosophy,
you also know that the second story is largely Quinean propaganda. In fact, Carnap
himself was a vocal critic of operationalism – long before he felt the pressure of
Quine’s critiques of the positivist program.
Some, especially philosophers, go so far as even to contend that these modern theories,
since they are not intuitively understandable, are not at all theories about nature but “mere
formalistic constructions”, “mere calculi”. But this is a fundamental misunderstanding of
the function of a physical theory.
(Carnap 1939: 210)
Notice how Carnap feels the same pressure that Bohr and Cassirer feel – the
pressure that the new theories of physics are not “intuitively understandable.”
Moreover, like Bohr and Cassirer, he refuses to take the breakdown of intuitive
understandability (or anskuelighed, or Darstellbarkeit) to demand a retreat to
operationalism. Instead, Carnap – like Bohr and Cassirer – asks us to think harder
about how our theories purport to represent physical reality.
Like Bohr, Carnap insists that the representational status of the quantum wave
function is not all that different from the situation of the symbols of classical
mathematical physics.
158 Hans Halvorson
If we demand from a modern physicist an answer to the question what he means by the
symbol “ψ” of his calculus, and are astonished that he cannot give an answer, we ought to
realize that the situation was already the same in classical physics. There the physicist
could not tell us what he meant by the symbol “E” in Maxwell’s equations. . . .Thus the
physicist, although he cannot give us a translation into everyday language, understands the
symbol “ψ” and the laws of quantum mechanics. He possesses the kind of understanding
which alone is essential in the field of knowledge and science.
(Carnap 1939: 210–211)
Interestingly, the words of Carnap here are echoed – quite unintentionally, I am

sure – by Wallace and Timpson.
. . .it’s not as if we really have an intuitive grasp of what an electric or magnetic field is,
other than indirectly and by means of instrumental considerations. . . .Thus, it seems that
we gain a basic understanding of the electromagnetic field by seeing it as a property of
spatial regions, and our further understanding must be mediated by reflecting on its role in
the theory. . . .beyond that there doesn’t seem to be much further to be grasped.
(Wallace and Timpson 2010: 700)
We might just add that the concept of spatial regions does not provide us with a
truly Archimedian reference point – for these regions themselves are understood in
a mediated way, via their description in physical theory.
At this point, it should be thoroughly unclear how the views of Bohr, Cassirer,
and Carnap differ from some of the more moderate and reasonable quantum state
realists. To one such view we now turn.
8.10 The Nomological View

According to the cutting-edge survey of Chen (2018), there are three versions of
wave function realism – the two ψ-field views and a “nomological view,” where
the wave function represents a law of nature (Goldstein and Zanghì 2013, Esfeld
2014, Miller 2014, Callender 2015, Esfeld and Deckert 2017). Thus, if we were to
regiment the nomological view, we might say that the wave function plays the
theoretical role of a proposition, or perhaps of a rule for generating propositions.
The theoretical role of propositions is, of course, quite different than the theoretical
role of names or even variables, both of which are used to denote existing things.
Thus, only by stretching the word “ontological” beyond the breaking point could
we say that the nomological view is ontological. No matter what view of laws we
take, a law is not a thing and is not in the domain of quantification of a physical
theory. Thus, according to the nomological view, the wave function is not a beable.
Why then should the nomological view of the wave function be called a realist
view if it does not treat the wave function as corresponding to an existing thing?
Presumably, nomologists would say that what makes their view realist is that the
propositions encoded in ψ are objectively true, i.e., they correspond to reality. But
what then are these propositions that are encoded in ψ? Of course, Bohmians have
an answer ready at hand: ψ encodes propositions about the trajectories of particles.
Notice that the specific Bohmian answer is not implicit in the very idea that ψ
encodes true propositions. Even a rank operationalist will say that ψ encodes true
propositions – about the probabilities of measurement outcomes. Only we might
question whether these propositions are “objectively true,” because probabilities of
measurement outcomes are indexed by measurements, and the latter has yet to be
objectively defined.
So what makes the nomological view realist? Is it simply that ψ encodes
objectively true propositions? Or is it that ψ encodes true propositions about
particle trajectories? I would be loath to accept the second answer, because it
would make realism hostage to one idiosyncratic ontological picture, viz., a
particle ontology. Surely one can be a realist and have some sort of gunky
ontology, or a field ontology. So, it seems that realist-making feature of the
nomological view is merely its commitment to the idea that ψ represents object-
ively real features of the world. But now, if that is enough to make a view realist,
then Healey’s view is also a realist view. For Healey says that each physical
situation is correctly represented by at most one quantum state. Healey and the
nomologists agree that ψ represents objectively real features of the world.
Nor can we say that the nomological view is more realist than Healey’s because
it takes ψ to be a direct representation of reality. The representation relation
posited by the nomological view is every bit as indirect and nuanced as that
posited by Healey (or by Bohr for that matter). Indeed, the nomological view
includes an intricate translation scheme from mathematical properties of ψ to
various meaningful physical statements, some of which are about occurent states
of affairs, and some of which are about how things will change as time progresses.
Thus, in terms of how ψ represents, the nomological view is closer to the views of
Healey, Bohr, and Carnap than it is to ψ-field views. The nomologists may be
horrified to hear this, for they take great pride in being realists. But recall that
Bohm often emphasized that his point of view was not so radically different from
Bohr’s. He even offered his point of view as a clarification of Bohr’s. Perhaps then
the nomological view could be thought of as an attempt to clearly articulate some
of the things that Bohr was trying to say about the wave function.
8.11 Conclusion
The primary aim of this chapter was to investigate the meaning of realism about
quantum theory, and in particular, realism about the quantum state. We found that,
for the most part, these phrases are empty of substantive content. They are emotive
160 Hans Halvorson
catch phrases that are meant to muster the troops – and perhaps to sell books. But
please don’t get me wrong. I am not saying that there are no substantive questions
about how to interpret the quantum state. First of all, dissolving the antirealism/
realism distinction does not solve the measurement problem. There is still the
thorny issue of why it appears to us that measurements have outcomes. Second,
there are genuine disagreements about how to use quantum states – even if these
disagreements do not correlate directly with a distinction between “real” and
“not real.”
First, there is a genuine question of how to think of the relation of quantum
states to physical situations. (For simplicity, I will suppose that a physical situation
is picked out by an ordinary language description, for example, by the sorts of
instructions that one might give to an engineer or to a postdoc in the lab.) At one
extreme, we have objectivists who think that each such situation corresponds to a
unique, correct quantum state. At the opposite extreme, we have the Quantum
Bayesians who propose no correctness standards whatsoever between physical
situations and quantum states. For these QBists, a quantum state just is a person’s
point of view – it is neither correct nor incorrect, appropriate or inappropriate.
Between these two extremes, we have views like Rovelli’s, where each physical
situation can be described equally by at least two quantum states, depending on
one’s choice of a direction of time. Some people also think that Bohr was a
nonobjectivist about quantum states (see Zinkernagel 2016). However, I find that
view hard to square with Bohr’s repeated pronouncements of the “objectivity of
the quantum-mechanical description.”
I propose that we stop talking about the ill-defined notion of quantum state
realism, and that we instead start talking about these sorts of questions, e.g.,
whether quantum theory comes with objective standards for the ascription of states
to physical situations. First of all, what role do physical situations, described in
ordinary language, play in this debate? Could we replace “physical situation” with
something more neutral and description-free, such as “object” or “system”? The
problem with that suggestion is that the bare notion of an object or a system cannot
give us any sort of standard for comparison. For example, we might say:
“According to Healey, for each object X, there is a unique correct quantum state.”
But how does Healey individuate objects? If he has different standards for indi-
viduating objects than Rovelli has, then their apparently diverging views might in
fact agree. Thus, the question of appropriate use of quantum states requires a
target, or standard of reference, on which all parties antecedently agree. The notion
of a “physical situation” is supposed to offer a plausible standard of reference.
I have already suggested a shift from the ontological question: Do states exist?
to the representational question: How do states represent? Now I am suggesting
that this representational question be given a normative reading: What are the rules
governing the use of quantum states? That, I believe, is the real issue at stake,
although it is masked by emotionally charged words such as “ontological status.”
There is a second question, closely related to the first one. Should we apply
unitary dynamics without exception? Some people say yes (e.g., Bohm, Everett,
Wallace), and others say no (e.g., Ghirardi-Rimini-Weber [GRW], Rovelli,
Healey). But even this disagreement is not as clear-cut as it may seem. Even those
who believe in the universal validity of unitary dynamics allow themselves to use
“effective states.” The “true state,” they say, follows unitary dynamics. But for
calculational purposes, there can be great advantages to using the effective state.
I am no verificationist, and so I do not propose that we collapse the distinction
between real and effective states. Nonetheless, I am interested here in the rules for
using states, i.e., for deciding whether one ought to use the state that results from
unitary evolution or whether one is permitted to use the state that results from
application of the projection postulate. Or, to put it in explicitly representational
language: The question is whether the state that results from unitary evolution is
the only one that is “apt” to one’s situation or whether the state resulting from the
projection postulate might also be “apt” to one’s situation. Interestingly, all parties
seem to agree that the state resulting from the projection postulate is “apt” in some
sense. Even the most fervent anticollapsers will tell you that the projected state is
correct for all practical purposes. Then they will remind you that it is not the “real”
state. But I would then ask: not the real state of what? We are back again to the
question of how to identify the target X of our representation via a quantum state.
Acknowledgments
I thank Eddy Chen for guidance about wave function realism, to Catherina Juel for
help translating Bohr’s letter to Møller, and to Tom Ryckman for sending a
preprint of (Ryckman 2017), which got me interested in Cassirer’s view.
References
A. Fine, and S. Goldstein (eds.), Boston Studies in the Philosophy of Science:
Bohmian Mechanics and Quantum Theory: An Appraisal, Vol. 184. Dordrecht:
Springer.
Belot, G. (2012). “Quantum states for primitive ontologists,” European Journal for
Bohr, N. (1928). Letter to Christian Møller, 14 June. Archive for the History of Quantum
Physics, 1898–1950, Library of the American Philosophical Society.
Bohr, N. (1932). “Faraday lecture: Chemistry and the quantum theory of atomic consti-
tution,” Journal of the Chemical Society, 349–384.
162 Hans Halvorson
Bricker, P. (1996). “Properties,” pp. 469–473 in The Encyclopedia of Philosophy, Supple-

ment. New York: Simon & Schuster.
Callender, C. (2015). “One world, one beable,” Synthese, 192: 3153–3177.
Carnap, R. (1939). “Foundations of logic and mathematics,” pp. 139–213 in O. Neurath,
R. Carnap, and C. Morris (eds.), International Encyclopedia of Unified Science.
Chicago: University of Chicago Press.
Carroll, S. (2017). The Big Picture: On the Origins of Life, Meaning, and the Universe
Itself. New York: Dutton.
Cassirer, E. (1955). The Philosophy of Symbolic Forms: Language. Vol. 1. New Haven:
Yale University Press.
Chakravartty, A. (2017). “Scientific realism,” in E. N. Zalta (ed.), The Stanford Encyclo-
pedia of Philosophy (Summer 2017 Edition). https://plato.stanford.edu/archives/
sum2017/entries/scientific-realism/
Chen, E. (2017). “Our fundamental physical space: An essay on the metaphysics of the
wave function,” Journal of Philosophy, 114: 333–365.
Chen, E. (2018). “Realism about the wave function,” Philosophy Compass, forthcoming.
Esfeld, M. (2014). “Quantum Humeanism, or: Physicalism without properties,” The
Philosophical Quarterly, 64: 453–470.
Esfeld, M. and Deckert, D.-A. (2017). A Minimalist Ontology of the Natural World. New
York: Routledge.
Faye, J. (2014). “Copenhagen interpretation of quantum mechanics,” in E. N. Zalta (ed.),
The Stanford Encyclopedia of Philosophy (Fall 2014 Edition). https://plato.stanford
.edu/archives/fall2014/entries/qm-copenhagen/
Forrest, P. (1988). Quantum Metaphysics. Oxford: Blackwell.
Frigg, R. and Nguyen, J. (2016). “Scientific representation,” in E. N. Zalta (ed.), The
Stanford Encyclopedia of Philosophy (Winter 2016 Edition), https://plato.stanford
.edu/archives/win2016/entries/scientific-representation/
Goldstein, S. and Zanghì, N. (2013). “Reality and the role of the wave function in quantum
theory,” pp. 91–109 in A. Ney and D. Z. Albert (eds.), The Wave Function: Essays on
the Metaphysics of Quantum Mechanics. Oxford: Oxford University Press.
Halvorson, H. (2012). “What scientific theories could not be,” Philosophy of Science, 79:
183–206.
Healey, R. (2015). “How pragmatism reconciles quantum mechanics with relativity,” 3am
Interview. www.3ammagazine.com/3am/how-pragmatism-reconciles-quantum-mech
anics-with-relativity-etc/
Healey, R. (2016). “Quantum-Bayesian and pragmatist views of quantum theory,” in E. N.
Zalta (ed.), The Stanford Encyclopedia of Philosophy (Spring 2017 Edition), https://
plato.stanford.edu/archives/spr2017/entries/quantum-bayesian/
Healey, R. (2017). The Quantum Revolution in Philosophy. Oxford: Oxford University
Press.
Hirsch, R. J. (2017). “From Representation to Reality: Essays on the Nature of Scientific
Exegesis,” PhD thesis, Princeton University. https://catalog.princeton.edu/catalog/
10690088
Howard, D. (2004). “Who invented the “Copenhagen interpretation”? A study in myth-
ology,” Philosophy of Science, 71: 669–682.
Kadison, R. V. and Ringrose, J. R. (1991). Fundamentals of the Theory of Operator
Algebras. Providence: American Mathematical Society.
Miller, E. (2014). “Quantum entanglement, Bohmian mechanics, and Humean superve-
nience,” Australasian Journal of Philosophy, 92: 567–583.
Ney, A. (2012). “The status of our ordinary three dimensions in a quantum universe,”
Noûs, 46: 525–560.
North, J. (2013). “The structure of a quantum world,” pp. 184–202 in A. Ney and D. Z.
Albert (eds.), The Wave Function: Essays on the Metaphysics of Quantum Mechanics.
Penrose, R.. (2016). Fashion, Faith, and Fantasy in the New Physics of the Universe.
Pringe, H. (2014). “Cassirer and Bohr on intuitive and symbolic knowledge in quantum
physics,” Theoria. Revista de Teoría, Historia y Fundamentos de la Ciencia, 29:
417–429.
Reichenbach, H. (1965). The Theory of Relativity and a Priori Knowledge. Berkeley:
University of California Press.
Rovelli, C. (2016). “An argument against the realistic interpretation of the wave function,”
Ruetsche, L. (2011). Interpreting Quantum Theories. Oxford: Oxford University Press.
Ryckman, T. (2017). “Cassirer and Dirac on the symbolic method in quantum mechanics:
A confluence of opposites,” Journal for the History of Analytic Philosophy, 6:
194–224.
Saunders, S. (2010). “Many worlds? An introduction,” pp. 1–49 in S. Saunders, J. Barrett,
A. Kent, and D. Wallace (eds.), Many Worlds? Everett, Quantum Theory, and Reality.
Swanson, N. (2018). “How to be a relativistic spacetime state realist,” The British Journal
for the Philosophy of Science, forthcoming.
van Fraassen, B. C. (1976). “To save the phenomena,” The Journal of Philosophy, 73:
623–632.
Wallace, D. (2008). “The quantum measurement problem: State of play,” pp. 16–98 in
D. Rickles (ed.), The Ashgate Companion to the New Philosophy of Physics. Farn-
ham: Ashgate.
Wallace, D. (2012). The Emergent Multiverse. Oxford: Oxford University Press.
Wallace, D. (2013). “The Everett interpretation,” pp. 460–488 in R. Batterman (ed.), The
Oxford Handbook of the Philosophy of Physics. Oxford: Oxford University Press.
Wallace, D. (2018a). “Against wave function realism,” in B. Weslake and S. Dasgupta
(eds.), Current Controversies in the Philosophy of Science. London: Routledge.
Wallace, D. (2018b). “Quantum theory as a framework, and its implications for the
quantum measurement problem,” in S. French and J. Saatsi (eds.), Scientific Realism
and the Quantum. Oxford: Oxford University Press.
Wallace, D. and Timpson, C. (2010). “Quantum mechanics on spacetime I: Spacetime state
realism,” The British Journal for the Philosophy of Science, 61: 697–727.
Zinkernagel, H. (2016). “Niels Bohr on the wave function and the classical/quantum
divide,” Studies in History and Philosophy of Modern Physics, 53: 9–19.
9
Locality and Wave Function Realism
alyssa ney
9.1 Introduction
Wave function realism is a framework for interpreting quantum theories. Applied to
nonrelativistic versions of quantum mechanics, wave function realism yields a
metaphysics according to which the central, fundamental object is the quantum
wave function, understood as a field on a high-dimensional space with the structure
of a classical configuration space, perhaps supplemented with additional degrees of
freedom to capture spin and other variables. Particles and other low-dimensional
objects are understood by the wave function realist to be ontically derivative objects,
constituted ultimately out of wave function stuff. For more sophisticated relativistic
quantum theories and quantum field theories, the framework recommends a suitable
relativistic extension of this metaphysics: a field in whatever high-dimensional
space is capable of capturing the full range of pure quantum states.
The case for such high-dimensional field interpretations varies from one frame-
work proponent to another, but a recurrent theme is wave function realism’s ability
to provide ontologies for quantum theories that have some intuitively nice meta-
physical features. For example, one may note the fact that quantum entanglement
threatens to force a fundamentally nonseparable metaphysics on the interpreter or,
what is to some (Howard 1985, and, he argues, Einstein) worse, a fundamentally
nonlocal metaphysics. However, these defects may be seen to drop away in the
higher-dimensional interpretations preferred by the wave function realist. For her,
what initially appear to be distinct entities possessing primitive relations and
communicating instantaneously across distant regions of space are revealed to be
manifestations of a single object, fundamentally possessing only intrinsic features
and acting locally on a high-dimensional space. This motivation for wave function
realism is, as I shall explain later, more compelling than that suggested by others,
who have argued that one should adopt such a framework simply because it is the
sort of thing that is most naturally read off the physics.
164
Locality and Wave Function Realism 165
Although some have challenged the wave function realist’s claim to provide a
separable metaphysics for quantum theories (e.g., Myrvold 2015, Lewis 2016),
I would say it is the claim that wave function realism provides a local metaphysics
that is more difficult and less straightforward (I rebut the concerns about separabil-
ity in Ney 2019b). And so, this is what I wish to examine in the present chapter.
I will introduce and distinguish several senses in which a metaphysics for physics
may be local, starting with two notions made use of by Bell (1964, 1976). From
there, we may evaluate in which sense or senses, if any, the wave function realist’s
metaphysics are local; and what, after all, is the virtue of having interpretations for
quantum theories that are local in that (those) sense(s). I will focus on the
nonrelativistic case. Ney (2019b) examines the extension to relativistic theories.
9.2 Wave Function Realism and its Competitors

It is worth noting at the outset that the interpretational question to which wave
function realism is intended to provide an answer is to a large extent orthogonal to
other interpretational questions, for example, that of what is the most promising
approach to addressing the measurement problem for quantum theories. The
measurement problem is the problem of how systems that are indeterminate with
respect to one or another variable may evolve into states of what appear to be
determinate values of such variables upon measurement. How can indeterminacy
appear to evolve into determinacy, given a dynamics that seems to make such
evolution impossible?
There are many approaches available to solving the measurement problem,
ranging from collapse theories, to the postulation of hidden variables, to the appeal
to relative states (or many worlds). In this article I focus on what are termed psi-
ontic or objectivist approaches to solving the measurement problem, because my
interest is in what might be the mind-independent metaphysics of quantum theor-
ies. This is not to reject the existence or interest of psi-epistemic or subjectivist
approaches, which deny that the role of the quantum state is to represent some
mind-independent reality. However, the measurement problem raises the question:
Should the quantum dynamics be supplemented, modified, or left alone? By
contrast, wave function realism and its competitor frameworks are addressed at
the question of what such dynamics describe: What is the ontology of a theory with
a dynamics like that? In principle, each rival metaphysical framework may be
applied to any of the dynamical proposals aimed at solving the measurement
problem, although some combinations are more natural than others. I will not
address all of the possibilities here, though I will provide brief overviews of wave
function realism and its competitors. (Wave function realism is more natural as an
interpretation of collapse theories like that of Ghirardi, Rimini, and Weber [GRW]
166 Alyssa Ney
and many worlds. The competitor primitive ontology approach is more naturally
combined with hidden variable theories like Bohmian mechanics).
As mentioned, wave function realism is a framework for the interpretation of
quantum theories in which the wave function is the central ontological item and is
interpreted as a field on a high-dimensional space that (for the nonrelativistic case)
is assumed to have the structure of a classical configuration space. For hidden
variable theories, this wave function is supplemented with additional ontology,
e.g., for Bohmian mechanics, a single particle evolving in a way determined by the
motion of the wave function. Wave function realists thus take wave function
representations literally and straightforwardly. As Albert has written
The sorts of physical objects that wave functions are, on this way of thinking, are (plainly)
fields – which is to say that they are the sorts of objects whose states one specifies by
specifying the values of some set of numbers at every point in the space where they live,
the sorts of objects whose states one specifies (in this case) by specifying the values of two
numbers (one of which is usually referred to as an amplitude, and the other as a phase) at
every point in the universe’s so-called configuration space.
(Albert 1996: 278)
As has been mentioned and will be discussed in more detail later, this simple
interpretation of the wave function has the advantage of providing a metaphysics
for quantum theories that is fundamentally separable and local. However, other
interpreters challenge this reading.
The primitive ontology approach of Dürr, Goldstein, Zanghì, Allori, and Tumulka
(Dürr, Goldstein, and Zanghì 1992, Allori et al. 2008, Goldstein and Zanghì 2013)
insists that the ontology of quantum theories consists primarily of entities in ordinary
space or spacetime, for example, particles for Bohmian mechanics and matter
density fields for collapse theories or many-worlds approaches. On the primitive
ontology approach, the wave function is interpreted as real, but not an element of a
quantum theory’s primitive ontology: It is not what any physical theory is primar-
ily about, not what constitutes the matter in the theory. (See Ney and Phillips 2013
for a detailed examination and critique of the notion of a primitive ontology).
Instead, the wave function plays some other role, to guide the behavior of the
matter, and so it is something more like a law, broadly speaking.
Those adopting the primitive ontology framework (and some of the other
approaches I describe later in the chapter) complain about the use of the name
‘wave function realism’ to apply solely to views according to which the wave
function is a physical field on a higher-dimensional space, claiming they too are
realists about the wave function, taking it to be a real, mind-independent element of
quantum ontology. In a sense, this complaint is fair, but by now the terminology
has become so entrenched, I will continue to use it. And anyway, in defense of the
terminology, one thing that distinguishes the status of the wave function on the
wave function realist view from how it is viewed in primitive ontology or other psi-
ontic approaches to interpretation is that, for the wave function realist, the wave
function is real in the classical sense of being res- or thing-like; it is a substance,
re-al. This contrasts with its status on other interpretational approaches, in which it
occupies one or another distinct ontological category; rather than being res, it
viewed as law, property, or a pattern of relations.
To many, the primitive ontology framework has appeal over wave function
realism in providing metaphysics for quantum theories that are intuitive in certain
respects. According to most applications of the approach, the fundamental entities
of the theory inhabit our familiar space or spacetime, and the macroscopic objects
we observe may be built out of these basic constituents (particles or matter fields)
in straightforward ways. (For a critique, see Ney and Phillips 2013.) An exception
is the flash ontology offered as a primitive ontology interpretation of some collapse
approaches; this ontology is surprisingly sparse and unfamiliar. Moreover, these
approaches hold out the promise of a separable metaphysics. The features of
composite objects are determined by the features of their smaller constituent
particles or field values at individual spacetime points. For example, for Bohmian
mechanics, one may argue that there are no facts about joint states of the particles
that fail to be determined by the states of individual particles. Facts about quantum
entanglement do fail to be determined by facts about the states of particles taken
individually or together, but on a primitive ontology approach to Bohmian mech-
anics, entanglement is a feature of the wave function, not the particles. And so, one
could say that the matter ontology of Bohmian mechanics is perfectly separable.
The same may be said for the matter density ontology that the primitive ontology
view attaches to collapse theories.
Nonetheless, such metaphysics are not local. As Bell showed
In a theory in which parameters are added to quantum mechanics to determine the results of
individual measurements, without changing the statistical predictions, there must be a
mechanism whereby the setting of one measuring device can influence the reading of
another instrument, however remote. Moreover, the signal involved must propagate
instantaneously, so that such a theory could not be Lorentz invariant.
(1964/1987: 199)
In situations characterized by the presence of quantum entanglement, measure-

ments made on one entity at one location can have immediate influence on that
entity’s entangled partner at a spatially distant region. This is so in situations
evolving according to both collapse and noncollapse dynamics. Even for many-
worlds versions of the primitive ontology approach, although locality is often
claimed, as Lewis (2016) notes, this locality is only available in the higher-
dimensional space that the wave function inhabits. Once the wave function is
168 Alyssa Ney
relegated to some other nonprimitive status, as something unfield-like and

unmatter-like, this advantage is lost.
A distinct class of approaches to the interpretation of quantum theories seeks to
obtain a local metaphysics, but at the cost of rejecting separability. This is
characteristic of the approaches of Howard and Teller from the eighties. Howard
(1985) argues that what seem to be entangled pairs of objects in distant regions of
space should not be viewed after all as numerically distinct entities. In the extreme
case in which every putative entity is entangled with any other, the view becomes a
version of monism. There is only one thing. According to Howard, there is no
instantaneous action between spatially separated entities because there are not
multiple entities after all. He argues that if faced with the choice of preserving
separability or locality, Einstein too would have chosen to reject separability in
order to maintain locality (Howard 1985: 197).
Teller (1986) offers a distinct nonseparable approach, allowing that entangled
pairs are fundamentally distinct entities while claiming that entanglement forces us
to admit the existence of irreducible relations between pairs that do not reduce to
any intrinsic features of the individuals constituting those pairs. He argues (1989)
that this allows one to avoid nonlocality, because the view rejects the general claim
that correlations between objects must ever be explained in terms of more funda-
mental features of these objects, such as the causal relations between them.
A more recent development of the idea that quantum entanglement should be
interpreted as characterizing a world with fundamental relations not reducible to
features of their relata is ontic structural realism, which has been advocated in a
variety of forms (Ladyman 1998, Esfeld 2004, Ladyman and Ross 2007, French
2014). Such a framework for interpretation provides a metaphysics that is mani-
festly nonseparable. However ontic structural realists typically resist the claim of
Teller that the existence of irreducible relations allows one to avoid the conse-
quence of nonlocality. (Thanks to Michael Esfeld regarding this point.)
Similarly, another interpretative framework, the spacetime state realism advo-
cated by Wallace and Timpson (2010) and Myrvold (2015), aims neither at
providing metaphysics for quantum theories that are either separable or local. On
this view, even in nonrelativistic quantum mechanics, the wave function charac-
terizes highly abstract features of spacetime regions, where the features of com-
posite regions do not generally reduce to features of their constituents. What
happens at one region can instantaneously affect what happens at another. In both
cases, that of the ontic structural realist and of the spacetime state realist, the view
is motivated not by the fact that it provides an intuitive metaphysics with various
attractive and natural features such as separability or locality, but instead by the
fact, inter alia, that it stays truer to the way physics represents the world rather than
our expectations about what an ontology for physics should look like.
9.3 Entanglement and Separability

So far, I have said that wave function realism is a framework aimed at providing
metaphysics for quantum theories that are separable, but it would be good to have a
straightforward definition of separability with which to work. We may initially
consider the following:
A metaphysics is separable if and only if (i) it includes an ontology of objects or properties

instantiated at distinct regions, and (ii) when any objects or properties are instantiated at
distinct regions R1 and R2, all facts about the composite region R1[R2 are determined by
the facts about the objects and properties instantiated at its subregions.
The first clause is needed in order to rule out monistic metaphysics in which there
is only one thing or one spatial location. Separability implies that there are distinct
objects, or at the very least (if one prefers a field metaphysics), distinct field values
instantiated at distinct locations in space.
One drawback to this definition is that, as it stands, it requires a separable
metaphysics to be a Humean metaphysics. Since it speaks of all facts being
determined by facts about what occurs at individual spatial regions, it makes
separability require that all facts about dispositions, counterfactuals, causation,
and laws be determined by what occurs at individual spatial regions. Loewer
(1996) defends wave function realism explicitly for its ability to provide an
interpretation of quantum theories compatible with Humean supervenience. One
might avoid this implication by modifying the second clause of the definition to
state only that the categorical, nondispositional, or non-nomic facts are determined
by the facts about individuals at subregions. We then have the following:
A metaphysics is separable if and only if (i) it includes an ontology of objects or properties
instantiated at distinct regions, and (ii) when any objects or properties are instantiated at
distinct regions R1 and R2, all categorical facts about the composite region R1[R2 are
determined by the facts about the objects and properties instantiated at its subregions.
Some might object that the matters of concern when we discuss entanglement
relations are dispositional – this electron would be measured spin up were its z-spin
to be measured. And so we really want a definition of separability that also requires
dispositional features to reduce to localized facts about individual spatial regions.
But although these are some of the features of interest, entanglement can appear to
force on us as well, the violation of even this weaker account of separability. For,
unless one adopts the Copenhagen-ish view that we can only talk sensibly about
the results of measurements or the features of systems when they are in eigenstates,
it is the occurrent and categorical spin states of entangled pairs as well, not merely
how they would behave upon measurement, that appears to be determined only
jointly, not individually, by objects at distinct spatial regions.
170 Alyssa Ney
With this definition of separability in hand, we may see how the wave function
realist may claim to provide interpretations of quantum theories that recognize the
phenomenon of quantum entanglement without committing to fundamental non-
separability. To illustrate, consider the EPRB state, in which a pair of atoms is
entangled with respect to their z-spin. Suppose our atoms are created in the
singlet state
ψS ¼ 1=√2 j z-upiA j z-downiB 1=√2 j z-downiA j z-upiB
and are then sent in opposite directions toward two Stern-Gerlach magnets, which
will bend them up or down in accordance with their z-spin toward two respective
measurement screens. Consider four locations between the magnets and screen
with the following labels:
R1: where atom A goes at time t should it get deflected up
R2: where atom A goes at t should it get deflected down
R3: where atom B goes at t should it get deflected up
R4: where atom B goes at t should it get deflected down
At time t, the atoms will be an entangled state of position:
ψx ¼ 1=√2 j R1iA j R4iB 1=√2 j R2iA j R3iB
And so, there are facts at t about properties instantiated at the joint regions R1[R4
and R2[R3 that are not determined by any facts local to their subregions, e.g.,
there is an atom at R1 if and only if there is one at R4. There is an atom at R2 if and
only if there is one at R3. We thus have a violation of separability.
The wave function realist argues that what appears as nonseparability arises
because what we are seeing is a three-dimensional manifestation of a more
fundamental and higher-dimensional metaphysics that is entirely separable. The
individual atoms A and B are ultimately constituted out of a field – the quantum
wave function. This field is spread not over our familiar three-dimensional space in
which there are the four locations R1, R2, R3, and R4, but instead over a space
with the structure of a classical configuration space. This space instead contains
(for example) regions we may suggestively label R13, R14, R23, R24. The wave
function has amplitude at these regions corresponding to the Born Rule probabil-
ities for the quantum state. So, in the present case, given the quantum state ψx, the
wave function will have nonzero amplitude only at the two locations R14 and R23
and it will have an amplitude of ½ at each of these locations. Spin states will
correspond to additional degrees of freedom. For a system initially appearing to
have N particles then, the dimensionality of the wave function’s space is posited to
be at least 3N.
The wave function realist’s proposed higher-dimensional metaphysics is thus
entirely separable. All categorical features are determined by features of the wave
function instantiated at individual regions in its space. (The wave function has
phase values in addition to the amplitude values highlighted in this discussion.)

That the wave function takes a particular shape across the joint region R14[R23,
for example, is entirely determined by its features at the individual regions R14
and R23.
So far, what we have been discussing is wave function realism as applied to a
nonrelativistic quantum mechanics without hidden variables. But as has been men-
tioned, wave function realism is only a framework for interpretation: The metaphysics
it entails will vary in details depending on the details of the theory of which it is applied
to be an interpretation. If we are interested in the interpretation of a hidden variables
theory, then in addition to the wave function, there will also exist some entity in the
high-dimensional space corresponding to these variables. In quantum field theories in
which particle number fails to be conserved, the dimensionality of the space will
instead be determined by the number of basis states of the quantum field. In summary,
the wave function realist’s metaphysics depends on the quantum theory one wants an
interpretation of. It will consist (at least) of the following:
• A background space with the structure of a classical configuration space

• The wave function, a field on that space, characterizable in terms of an assignment of
amplitude and phase values and evolving according to the dynamics of the theory,
e.g., the Schrödinger equation, perhaps supplemented with a collapse mechanism
This metaphysics is postulated to be separable. There are properties instantiated at
distinct spatial regions. But there are no facts about the wave function’s categorical
features at composite regions that are not determined by facts local to these
regions’ respective subregions.
The question I now want to raise is whether this metaphysics is also local.
9.4 Concepts of Locality

The concepts of locality most frequently invoked when purported violations of
locality brought about by quantum entanglement are under discussion are those
highlighted by Bell. Wiseman (2014) has argued that Bell really had two different
accounts of what locality may come to in physics. The first is the notion of locality
invoked in his 1964 paper “On the Einstein-Podolsky-Rosen paradox”:
. . . the requirement of locality, or more precisely that the result of a measurement on one
system be unaffected by operations on a distant system with which it has interacted in
the past.
(Bell 1964: 195)
This is equivalent to what Shimony called “parameter independence.” Applied to

the situation described in the previous section, it is the principle that the
172 Alyssa Ney
probabilities for the results of a measurement on atom B are independent of what

we choose to do to what is at time t the spacelike separated atom A, including what
measurements we choose to perform on it.
In his 1976 paper, “The theory of local beables,” however, following Wiseman,
we may see Bell invoking a distinct principle that he calls “local causality”:
Let A be localized in a space-time region 1. Let B be a second beable in a second region

2 separated from 1 in a spacelike way. . . Now my intuitive notion of local causality is that
events in 2 should not be ‘causes’ of events in 1, and vice versa.
(Bell 1976: 13)
By “beable,” Bell simply means entity, something that is real. This is a stronger
principle than the early “locality” principle from 1964. It states not only that the
probabilities for the results of a measurement on one system are independent of
how we may manipulate another system at a spacelike separation from it, but also
that these probabilities are independent of the actual measurement results we find
when we measure that other system. Wiseman argues that it is local causality that
Bell took to be the primary locality principle of interest from at least 1976 on. And
it is what he argued must be violated if quantum theory is correct.
Arguably, neither of these interpretations ‘locality’ suffices to explicate the
sense in which the metaphysics of the wave function realist is claimed to be
local. For the principles invoked by Bell both concern the existence of causal
relations in spacetime – the second one especially explicitly by invoking facts
about certain events exhibiting spacelike separation. Yet there is no spacetime
interval defined on the space that the wave function is said to inhabit, nor is the
space of the wave function the space in which light propagates, and so there is
no sensible notion of spacelike separation in the wave function realist’s funda-
mental metaphysics to let us settle the issue of whether these senses of
‘locality’ obtain. To explain the way in which wave function realism may be
claimed to involve a local metaphysics, we must move to a concept of locality
that makes sense in the context of the high-dimensional space of the wave
function realist.
Sometimes, when Bell discusses his principle of local causality, he states it in
broader terms than we just saw:
What is held sacred is the principle of ‘local causality’ – or ‘no action at a distance’.
(Bell 1981: 46)
This principle is generalizable so as to be of use by the wave function realist; and

so, we may say:
A metaphysics is local if and only if it contains no instantaneous action across spatial
distances.
Or perhaps:
A metaphysics is local if and only if it contains no instantaneous and unmediated action
across spatial distances,
where ‘spatial’ refers to whatever is the spatial background of the metaphysics. It

may be our familiar three-dimensional space or spacetime. But it may also be the
high-dimensional spatial background of the wave function realist.
Can we be more precise? In his philosophy of physics textbook, Marc Lange
provides an account which seems aimed at capturing such a notion:
Spatiotemporal locality: For any event E, any finite temporal interval τ > 0, and any finite
distance δ > 0, there is a complete set of causes of E such that for each event C in this set,
there is a location at which it occurs that is separated by a distance no greater than δ from a
location at which E occurs, and there is a moment at which C occurs at the former location
that is separated by an interval no greater than τ from a moment at which E occurs at the
latter location.
(Lange 2002: 15)
But as Myrvold has noted (personal communication), without a metric, it is not

clear how the wave function realist can make use of this account. Perhaps no more
precision can be achieved for the sense of locality claimed by the wave function
realist than with the intuitive definition of no instantaneous (unmediated) action
across spatial distances.
To see how the wave function realist’s metaphysics generally satisfies this form
of locality, one should be careful about distinguishing the situation for the inter-
pretation of different quantum theories, including those with and without collapse.
In the case of Everettian quantum theories without collapse, the wave function
simply evolves unitarily in accordance with the Schrödinger equation or its
relativistic variant. The wave function spreads out and may interfere with itself
as waves do. But at no point does an event at one point in the space influence an
event somewhere else.
For collapse theories like GRW, the wave function may evolve unitarily, but
from time to time there is a spontaneous collapse. This involves the entire wave
function undergoing a hit, which may be represented mathematically by the
multiplication of the quantum state by a Gaussian function localized on a particular
region of the space. In this case, it is not correct to say that what happens in one
region of the wave function’s space acts immediately to influence what happens in
another. Rather, in these models, collapses are not caused by anything about the
state of the wave function at the previous time, but occur spontaneously. One could
say that there are facts about the wave function at the time prior to collapse that
determine how likely it is that the hit is localized at one point rather than another.
The probability of the collapse being localized at one point rather than another is
174 Alyssa Ney
given by the Born Rule probabilities, which are associated with the amplitude
squared of the wave function at the different points in its space. But there is still no
reason to say that the amplitude of the wave function at one distant region R causes
a collapse to be localized at another region R’ of the space instantaneously. Even if
the wave function later becomes more peaked around R’, the collapse is not
something that takes place at R’, but is rather something that happens across the
entire space. So there is not really a localized effect that may be influenced by some
distant cause. The evolution of the wave function through collapse may be jerky
and discontinuous, but it does not result in nonlocal action.
Finally, in the case we are interpreting noncollapse theories with hidden
variables such as Bohmian mechanics, the wave function behaves identically to
how it does in the nonrelativistic Everettian model. However, in this case, there
will be some additional ontology, such as a particle (the so-called marvelous point)
that moves around the wave function’s space in a way described by the theory’s
guidance equation
dQk ℏ ψ ∂k ψ
¼ Im * ðQ1 ; . . . ; QN Þ
dt mk ψ*ψ
In this case, the behavior of this additional ontology, the particle, is determined by
the state of the wave function in the neighborhood of the place in the high-
dimensional space it occupies, and so, there is no threat of nonlocal action. Despite
this fact, Bohmian mechanics is a quantum theory (or solution to the measurement
problem) that combines rather poorly with the interpretational framework of wave
function realism. After all, the very motivation for adopting Bohmian mechanics
(at least as presented in Dürr et al. 1992) depends on an argument that wave
function realists should not accept, namely that quantum theories are not theories
about the behavior of the wave function but rather of something else, matter in
three-dimensional space or spacetime Regarding this point, see Section 5 of the
paper by Alyssa Ney and Ian Phillips (2013). But if one is worried about nonlocal
action, a wave function realist interpretation of Bohmian mechanics could be
of help.
9.5 Fundamental and Derivative

So we have seen that wave function realism is a framework most naturally
combined with Everettian and collapse versions of quantum mechanics – theories
in which the wave function ontology is not supplemented with additional variables.
When applied as an interpretation of these theories, wave function realism may
yield a metaphysics that is local in its distinctive, (relatively) fundamental high-
dimensional framework. But we may also ask what becomes of the nonlocality that
appeared in the lower-dimensional spacetime framework. Is there still not an issue

of nonlocality there according to the wave function realist?
The answer to this question turns first on whether the wave function realist
accepts the existence of three-dimensional space and spacetime, or of such low-
dimensional facts. As has been mentioned, the wave function realist typically does,
although he or she will also claim that the low-dimensional facts require no
addition to being beyond the high-dimensional framework he or she posits. The
low-dimensional facts are derivative, i.e., metaphysically explained (grounded) by
the behavior of the wave function.
One response the wave function realist may give to the question of whether
there is nonlocal action in the derivative low-dimensional framework is that no, in
that framework, there are correlations between spacelike separated events, but no
genuine causal interaction, because all such correlations have a deeper explanation
in terms of the behavior of the wave function. The dynamical explanation for these
correlations thus undercuts any causal explanation that may be provided by the
existence of spatially distant events such as measurements on one half of an
entangled pair.
I tried out such a line of reasoning in earlier work (Ney 2019a), as do Ismael and
Schaffer (2018). However, I now find such a position unsatisfactory. The reason is
that if one wants to argue in this way that there is no immediate causation across
spatial distances because such causal relations are undercut or screened off by the
behavior of the wave function, then one must similarly do so for all other causal
relations in the low-dimensional framework. For there will always be a wave
function dynamical explanation available at the more fundamental level. So, unless
we are to be causal nihilists about what happens in ordinary spacetime, the
behavior of the wave function does not undercut the reality of nonlocal action in
spacetime. What the wave function realist can offer is a more fundamental explan-
ation of in virtue of what that derivative nonlocal action obtains, one that may give
a more satisfying picture of what makes things happen in our world than one that
contains unexplained nonlocal action. But providing this explanation does not
remove low-dimensional facts about nonlocal action.
9.6 Motivating a Local Metaphysics

Assuming the wave function realist can provide a local interpretation of at least
some versions of quantum theories in at least some sense of ‘local’, we may ask,
why should one care?
I will start with some empirical considerations and move toward some that are
more a priori. I have already noted that wave function realism is not successful at
securing the conceptions of locality used by Bell. And yet, these are the senses of
176 Alyssa Ney
locality that some would say are mainly at issue when one worries about the
incompatibility of relativity and quantum mechanics. But perhaps there is more
one can say, and the fact that wave function realism provides a metaphysics local
in its own space may help alleviate some of the concerns arising about nonlocality
in spacetime.
In his recent book Quantum Ontology, Peter Lewis (2016) states the reason why
nonlocal action is in tension with relativity in the following way. Suppose one
allows that there exists at least some instantaneous action at a distance. Then there
is some one time at which one event influences another at a spatial distance from it.
For example, something happening right here, right now depends on the simultan-
eous mass of a distant star. But according to special relativity, there are no absolute
facts about which spatially distant events are simultaneous with which others. So,
in Lewis’s example, there is no fact about the mass of the star right now. Thus, this
action at a distance is ill-defined according to relativity. Thus, it would seem,
according to relativity, there cannot be action at a distance.
However, what looks puzzling, ill-defined, or brute from the perspective of a
nonfundamental metaphysics may be revealed as expected and explained in terms
of a more fundamental metaphysics. To the extent that wave function realism
supports a derivative ontology, it will yield an account of which spacetime
configurations exist and are causally related in that derivative ontology. So, at
least Lewis’s concern about the conflict between special relativity and quantum
mechanics seems avoided if one adopts wave function realism. This is not to say
that other issues, which I would concede are more basic, are not avoided, namely a
conflict with Lorentz covariance.
Another important feature of local theories is articulated by Einstein, who, in a
famous paper from 1948, argued that local metaphysics seem to be required for the
possibility of physical theories:
For the relative independence of spatially distant things (A and B), this idea is
characteristic: an external influence on A has no immediate effect on B; this is known as
the ‘principle of local action’, which is applied consistently only in field theory. The
complete suspension of this basic principle would make impossible the idea of the
existence of (quasi-)closed systems and, thereby, the establishment of empirically
testable laws in the sense familiar to us.
(Einstein 1948: 321–322)
The point seems straightforward enough. If what is nearby and observable may be
affected by objects that are spatially distant, then without full knowledge of the
occupants of the total spacetime manifold, how are we to make predictions about
how the objects we observe will behave? Locality appears required to allow us to
formulate testable empirical theories.
Now this point of Einstein’s is itself contestable. In conversation, Myrvold has

questioned it, claiming that even in classical physics we are very comfortable
writing down and testing laws knowing full well that there are spatially distant
objects affecting the behavior of local objects. His example is the astrophysicist’s
description of the motion of Jupiter’s moons. The Sun being 480 million miles
away, Einstein’s reasoning would lead one to believe that the physicist would need
to reject its influence, modeling the behavior of the moon solely in terms of nearby
factors. But this would produce wildly wrong results. This is a clear case in which
the assumption of spatially distant influences is essential, not an obstacle to the
formulation and testing of physical laws. Of course what Einstein rejects is that an
external influence on A has an immediate effect on B, and one might respond to
Myrvold by arguing that relativistic modeling will reject that the Sun’s influence is
immediate. But, astronomical phenomena are modeled quite well by Newtonian
physics, according to which gravitational influence is unmediated and instantan-
eous. Einstein seems wrong that physics simply cannot be done when we assume
there are nonlocal influences and build these into our models.
Myrvold is right to object to hyperbole in Einstein’s defense of locality in
physics, but I do not believe this undermines a weaker defense of locality as an
assumption guiding the formulation of tractable and testable physical theories. For
in the case of the Sun and Jupiter’s moons, the physics works because we are
considering the influence of just a few large bodies at a distance away. Things
would devolve quite quickly if the modeling of Jupiter’s moons needed to take into
account immediate and significant influences from many or all distant bodies. So
perhaps this is what Einstein is concerned about, thinking of widespread effects
from quantum entanglement that would massively complicate physics, perhaps
leading to intractability. And so, for physics “in the sense familiar to us” to work,
we discount immediate influence in general and for the most part where this is
justified. Note that with advances in computational modeling, physics can take into
account a much larger number of distant bodies successfully. However this does
not undermine Einstein’s point, as in the age of big data, we are moving away from
physics in the sense familiar to Einstein circa 1948.
So I propose we can make use of the weak point that we have good inductive
reason to believe that physics of the kind that is already familiar to us, which
involves modeling systems based on the assumption of (mostly) local influence,
is a way of developing inductively successful theories. But can this justification
for a local metaphysics be used to generate any support for wave function
realism? I am afraid it cannot – the point is, after all, that the testing of laws
depends on our ability to manipulate and observe what is happening at a
confined region of space, isolating objects from outside influences. Of course
the space in which human beings’ manipulations and observations take place is
178 Alyssa Ney
not the high-dimensional space of the wave function. Thus, it seems, Einstein’s
defense of locality justifies a local metaphysics in three-dimensional space or
spacetime, the framework in which we interact with objects, but not a local wave
function metaphysics.
Perhaps another case to be made for local interpretations of physical theories
may be found in the work of Allori. Allori (2013) defends another view she finds in
Einstein, that “the whole of science is nothing more than a refinement of our
everyday thinking.” She elaborates:
The scientific image typically starts close to the manifest image, gradually departing from it
if not successful to adequately reproduce the experimental findings. The scientific image is
not necessarily close to the manifest image, because with gradual departure after gradual
departure we can get pretty far away. . . The point, though, is that the scientist will typically
tend to make minimal and not very radical changes to a previously accepted theoretical
framework.
(Allori 2013: 61)
One might then say that since our prescientific thinking and subsequent physical
theories postulated local and separable metaphysics, our quantum theories should,
if possible, do so as well.
To be clear, Allori is herself not making this point to argue for the local
metaphysics of wave function realism. She is using the point to argue for her
preferred primitive ontology view, because she believes that all previous (i.e.,
nonquantum) physical theories also possess a primitive ontology. But one might
hope that her point extends to make a case for a local wave function metaphysics
as well.
Unfortunately, I do not think it does. Because wave function realism also rejects
as fundamental a three-dimensional spatial background, replacing it with an
unfamiliar, high-dimensional background, it is not really so plausible to argue that
this local metaphysics is closer to the manifest image and classical theories than
one that would jettison one or both of separability and locality, but retain the low-
dimensional spatial background of our experience. If we agree with Allori that
minimal departures should, where possible, be preferred, the move to higher
dimensions is very far from a minimal departure.
Finally, we may move to consider more purely a priori reasons in support of a
local metaphysics. Some of these were brought to bear in the eighteenth century as
natural philosophers struggled with Newton’s characterization of gravitational
forces as acting immediately across spatial distances. Newton himself sometimes
claimed that action at a distance is impossible, for example:
The cause of gravity is what I do not pretend to know and therefore would take more time
to consider of it. . . That gravity should be innate, inherent, and essential to matter, so that
one body may act upon another at a distance through a vacuum, without the mediation of
anything else, by and through which their action and force may be conveyed from one to
another, is to me so great an absurdity that I believe no man who has in philosophical
matters a competent faculty of thinking can ever fall into it.
(Newton, letter to Bentley, in Bentley 1838: 202)
To claim nonlocality is absurd is not thereby to offer an argument against it. Nor to
my knowledge did Newton ever offer a clear argument for why action at a distance
is absurd; however, we do find something in the work of Clarke in his correspond-
ence with Leibniz:
That one body should attract another without any intermediate means, is not a miracle, but
a contradiction: for ‘tis supposing something to act where it is not. But the means by which
two bodies attract each other, may be invisible and intangible, and of a different nature
from mechanism; and yet, acting regularly and constantly, may well be called natural. . .
(Clarke, fourth letter to Leibniz, in Alexander 1956: 53)
Clarke, like Newton, supposes that gravity must act locally, even if the means by which
it does so may be invisible. And the reason why this must be so is for something to act,
it must be located where it acts. Otherwise, it would not be it itself that is so acting, but
something else, or nothing at all. There is something, I believe, that is sensible about
this point and it explains at least one reason why nonlocal action strikes us as deeply
unintuitive and worse, incoherent. And it is, finally, a consideration that may be
brought to be bear in support of wave function realism’s local metaphysics.
9.7 Intuitions
It is my view that the best case the wave function realist has for developing a
distinctive local metaphysics comes from such conceptual considerations and
intuitions. But one might question whether it is at all desirable to have an
interpretation of quantum theories that conforms to our intuitions. Ladyman and
Ross (2007) criticize such interpretational projects, calling them “domestications
of science.” My project is openly one of the domestication of a large part of
physics. It is my attitude that quantum theories stand very much in need of
domestication to the scientific community and greater public (this is not to deny
that the project of domestication has already been carried out to a large extent by
the work of those providing clear solutions to the measurement problem).
Following out interpretations that are compatible with our intuitions may be useful
for a number of reasons. I will now mention three benefits that such an interpret-
ation may bring. All are unabashedly pragmatic.
First, an interpretation of a physical theory, by providing one with a clear
account of what the world is like according to the theory, benefits students and
180 Alyssa Ney
scientists in allowing them a clearer handle on the theory with which they are
working. Although it is not possible to understand our best scientific theories
without having a handle on the mathematics used to state it, a clear metaphysics
to supplement the mathematics can be instrumental in seeing more clearly what the
theory says, allowing one to more easily learn and use it. As an example, the
special theory of relativity, before it is supplemented with the clear interpretation
of a four-dimensional Minkowski spacetime, can seem to lead to paradoxes in
measurements that are difficult to comprehend – like the paradox of the train and
the tunnel or the twins paradox. These are not genuine paradoxes; there is no such
inconsistency in the theory, but this is much easier to comprehend when one grasps
the theory not purely through the predictions the mathematics produces, but
supplements it with a picture of entities spread out in four-dimensional spacetime,
for which facts about elapsed time or spatial distance fail to be absolute. I believe
something similar can come to pass for quantum theories. Once supplemented with
a clear metaphysics, what looks paradoxical or surprising becomes clear and
natural and easier to use. And there is no reason why distinct interpretations cannot
produce alternative accounts that are useful in this respect.
Second, an interpretation says things that go beyond what the theory on its own
says, and in this respect, interpretations can be fruitful in generating new specula-
tions or predictions that can then extend the theoretical power of the theory. Should
one adopt the wave function metaphysics and its attendant higher dimensions, one
can begin to ask more questions about the structure and contents of this higher-
dimensional space and learn more facts about it that would simply not be discussed
without attention to this question of interpretation.
Third, for myself and many other former physics students, the reason we chose
physics as a focus of study was to learn about the fundamental nature of reality.
Without an interpretation, physics does not provide this. Under the influence of
Copenhagen, Mermin’s “Shut up and calculate!”, and Feynman’s “I think I can
safely say no one understands quantum mechanics,” students often come to
quantum theories puzzled about what they say about the world, but then they are
told not to ask such questions because the theory is impossible to understand. This
is disappointing, and it drives students out of the field. Not all physics students care
about questions of interpretation and the deep issue of the nature of reality, but for
those that do, it is worth having serious work on interpretation that can give them
what they are looking for. We need more, not fewer students of physics.
I do not want to leave the reader with the sense that anything I am saying
challenges the idea that we should not at the same time work on interpretations that
challenge our thinking. In fact, all of the interpretations of quantum theories that
are available have aspects of unintuitiveness – this is simply unavoidable in the
interpretation of quantum theories. In addition, this is what is so exhilarating about
the study of these theories – how they challenge what we previously thought was
obvious. What is being suggested in this last section, however, is that there is
nothing problematic about trying to fit these startling aspects of the world into a
picture we can understand.
9.8 Conclusion
The wave function realist need not deny that there is a clear sense of locality in
which our world contains nonlocal influences. This is the sense of local causality
taken up by Bell from 1976 onwards. The question is whether one should take this
to be a brute fact about our world or should attempt to provide explanations in
terms of an underlying metaphysics. Wave function realism is such an attempt at
explanation. The virtues of having interpretative options that provide such an
explanation justify the exploration and development of this framework that should
be pursued alongside others.
References
A. Fine, and S. Goldstein (eds.), Boston Studies in the Philosophy of Science:
Bohmian Mechanics and Quantum Theory: An Appraisal, Vol. 184. Dordrecht:
Springer.
Alexander, H. (ed.) (1956). The Leibniz-Clarke Correspondence. Manchester: Manchester
University Press.
Allori, V. (2013). “Primitive ontology and the structure of fundamental physical theories,”
pp. 58–75 in A. Ney and D. Z. Albert (eds.), The Wave Function: Essays on the
Metaphysics of Quantum Mechanics. Oxford: Oxford University Press.
Bell, J. (1964). “On the Einstein-Podolsky-Rosen paradox,” Physics, 1: 195–200.
Bell, J. (1976). “A theory of local beables,” Epistemological Letters, 9: 11–24.
Bell, J. (1981). “Bertlmann’s socks and the nature of reality,” Journal de Physique
Colloques, 42: 41–62.
Bentley, R. (1838). The Works of Richard Bentley, A. Dyce (ed.). London: Francis
Macpherson.
Dürr, D., Goldstein, S., and Zanghì, N. (1992). “Quantum equilibrium and the origin of
absolute uncertainty,” Journal of Statistical Physics, 67: 1–75.
Einstein, A. (1948). “Quanten-mechanik und Wirklichkeit,” Dialectica, 2: 320–324.
Esfeld, M. (2004). “Quantum entanglement and a metaphysics of relations,” Studies in
History and Philosophy of Modern Physics, 35: 601–617.
French, S. (2014). The Structure of the World. Oxford: Oxford University Press.
Goldstein, S. and Zanghì, N. (2013). “Reality and the role of the wave function in quantum
theory,” pp. 91–109 in A. Ney and D. Z. Albert (eds.), The Wave Function: Essays on
the Metaphysics of Quantum Mechanics. Oxford: Oxford University Press.
182 Alyssa Ney
Howard, D. (1985). “Einstein on locality and separability,” Studies in History and Phil-
osophy of Science, 16: 171–201.
Ismael, J. and Schaffer, J. (2018). “Quantum holism: Nonseparability as common ground,”
Synthese, https://link.springer.com/article/10.1007/s11229-016-1201-2.
Ladyman, J. (1998). “What is structural realism?,” Studies in History and Philosophy of
Science, 29: 409–424.
Ladyman, J. and Ross, D. (2007). Every Thing Must Go. Oxford: Oxford University Press.
Lange, M. (2002). An Introduction to the Philosophy of Physics. Oxford: Blackwell.
Lewis, P. (2016). Quantum Ontology. New York: Oxford University Press.
Loewer, B. (1996). “Humean supervenience,” Philosophical Topics, 24: 101–127.
Myrvold, W. (2015). “What is a wave function?,” Synthese, 192: 3247–3274.
Ney, A. (2019a). “Separability, locality, and higher dimensions in quantum mechanics,”
forthcoming in B. Weslake and S. Dasgupta (eds.), Current Controversies in Phil-
osophy of Science. London: Routledge.
Ney, A. (2019b). “Wave function realism in a relativistic setting,” forthcoming in D. Glick,
G. Darby, and A. Marmodoro (eds.). The Foundation of Reality: Fundamental, Space,
and Time. Oxford: Oxford University Press.
Ney, A. and Phillips, K. (2013). “Does an adequate physical theory demand a primitive
ontology?,” Philosophy of Science, 80: 454–474.
Teller, P. (1986). “Relational holism and quantum mechanics,” British Journal for the
Teller, P. (1989). “Relativity, relational holism, and the Bell inequalities,” pp. 208–223 in
J. T. Cushing and E. McMullin (eds.), Philosophical Consequences of Quantum
Theory: Reflections on Bell’s Theorem. Notre Dame: University of Notre Dame Press.
Wallace, D. and Timpson, C. (2010). “Quantum mechanics on spacetime I: Spacetime state
realism,” British Journal for the Philosophy of Science, 61: 697–727.
Wiseman, H. (2014). “Two Bell’s theorems of John Bell,” Journal of Physics A, 47:
424001.
Part III
Individuality, Distinguishability, and Locality
10
Making Sense of Nonindividuals in Quantum Mechanics
jonas r. b. arenhart, otávio bueno, and décio krause
“It is only a slight exaggeration to say that good physics has at times
been spoiled by poor philosophy.”
(Heisenberg 1998: 211)
10.1 Motivation
As the epigraph by Heisenberg suggests, physics and philosophy may both benefit
from a constructive exchange in which one may enlighten the other. Physics can
illuminate philosophy, and philosophy can illuminate physics. Of course, some
may think that philosophy has nothing to contribute to physics (see Weinberg
1992), and although we shall not provide a detailed defense of why we take
philosophy to be relevant for science in general, we want to defend the relevance
of ontology, as a field of metaphysics, to physics and to what physics is about. We
stress, in this work, through a case study, the way in which ontology, as a
philosophical field, can engage with physics, particularly in clearing the ground
for the understanding of the nature of physical reality.
Ontology is concerned with what exists and with what kinds of things exist.
Although this description may sound abstract and far from the concerns of physics,
the relation between ontology and physics is a close one. Of course, we are not
claiming that physics cannot be successful without ontology. If that were the case,
ontology would be required for physics, and it is not. However, physicists work
with ontological problems all the time. For instance, when it is claimed that, in
general relativity, space and time are no longer independent and that a new kind of
entity is required, spacetime, this is a physical move with significant ontological
consequences. It directly affects how the furniture of the world looks.
Physicists need not be concerned with ontological problems raised by physics,
just as one need not be familiar with the Peano axioms in order to be able to use
arithmetical operations. Nevertheless, ontology is part of the enterprise, shared by
185
186 Jonas R. B. Arenhart, Otávio Bueno, and Décio Krause
most physicists, of obtaining information about how the world works and what it is
made of. What kinds of things are there? Particles, fields, space, time? What are
they like? Answering questions like these is part of the articulation of an under-
standing of physical reality. As a result, the furniture of the world is involved in
such understanding. Ignoring those questions and their importance may prevent
one from getting closer to the most fundamental problems.
In this chapter, we will not focus on such general questions, but rather on a very
specific case study: Assuming that quantum theories deal with “particles” of some
kind (point particles in orthodox nonrelativistic quantum mechanics, field excita-
tions in quantum field theories), what kind of entity can such particles be? One
possible answer, the one we shall examine here, is that they are not the usual kind
of object found in daily life – individuals. Rather, we follow a suggestion by Erwin
Schrödinger (among others, as will become clear later), according to which
quantum mechanics poses a revolutionary kind of entity – nonindividuals. While
physics, as a scientific field, is not concerned with whether entities posited by a
specific physical theory are individuals or not, answering this question is part of the
quest for a better understanding of physical reality. Here lies, in large measure, the
relevance of ontology.
10.2 Introduction
There is little doubt that quantum entities are difficult to categorize. Quantum
mechanics introduces so many oddities that it is easier to state what quantum entities
are not than to affirm what they are. (We use ‘entity’ here as a term that is neutral
regarding whether the things that are referred to have well-defined identity condi-
tions or not). According to some of the first creators of quantum theory, quantum
entities are nonindividuals. This view is now known as the Received View on
quantum non-individuality (henceforth, for the sake of brevity, “Received View”;
see French and Krause 2006: chapter 3, for further historical details on this view).
In a section aptly called “A particle is not an individual,” Erwin Schrödinger
(1998) advanced one of the formulations of the Received View. One passage is
worth quoting in full:
This essay deals with the elementary particle, more particularly with a certain feature that
this concept has acquired – or rather lost – in quantum mechanics. I mean this: that the
elementary particle is not an individual; it cannot be identified, it lacks “sameness” . . . In
technical language it is covered by saying that the particles “obey” a new fangled statistics,
either Bose-Einstein or Fermi-Dirac statistics. The implication, far from obvious, is that the
unsuspected epithet “this” is not quite properly applicable to, say, an electron, except with
caution, in a restricted sense, and sometimes not at all.
(Schrödinger 1998: 197)
Making Sense of Nonindividuals in Quantum Mechanics 187
Several significant points are made in this passage. It is noted that quantum
particles (i) are not individuals, (ii) cannot be identified, (iii) lack “sameness,”
and (iv) cannot be referred to by the use of “this,” at least not typically. Of course,
it is not clear, by considering this quotation alone, what Schrödinger’s conception
of identification, individuality, and sameness ultimately is, nor is it specified what
the proper relations among these concepts are. But a central feature of his view
becomes salient in another important passage. He notes:
I beg to emphasize this and I beg you to believe it: it is not a question of our being able to
ascertain the identity in some instances and not being able to do so in others. It is beyond
doubt that the question of “sameness”, of identity, really and truly has no meaning.
(Schrödinger 1996: 121–122)
Here, it is emphasized that the very question of the identity of quantum entities, the
question of their “sameness,” has no meaning. As a result, the difference between
these entities, provided their sameness is meaningless, has no meaning either. One
still needs to examine, of course, what exactly is the relation between the lack of
sameness (or identity) of quantum entities, on the one hand, and their lack of
individuality, on the other. It seems that Schrödinger takes them all to be conceptu-
ally the same: to “lose” one’s individuality just is to lose one’s identity. On his
view, the question of the identity of quantum particles in general makes no sense.
The proper understanding of the relations between these concepts, and the kind of
view that results from them in the context of quantum particles, is the topic of this
chapter.
These issues were also central to another contributor to the development of
quantum theory. In a classical passage, in which the issues of identity and individu-
ality were prominent, Hermann Weyl points out:
. . . the possibility that one of the identical twins Mike and Ike is in the quantum state E1
and the other in the quantum state E2 does not include two differentiable cases which are
permuted on permuting Mike and Ike; it is impossible for either of these individuals to
retain his identity so that one of them will always be able to say ‘I’m Mike’ and the other
‘I’m Ike’. Even in principle one cannot demand an alibi of an electron!
(Weyl 1950: 241)
The questions of discernibility and of an “alibi” of a quantum particle are clearly

posed. Once quantum particles, such as electrons, are in an entangled state, it
cannot be determined which particle is in which state. In other words, it cannot be
settled which particle is which. There is nothing – no property, no special ingredi-
ent – that could act as an alibi to discern electrons. In this respect, it is their
indiscernibility rather than their identity that should take center stage. Differently
from what Schrödinger suggests, perhaps identity need not lose its meaning,
provided that indiscernible things can still be numerically distinct (or identical).
As will becomes clear, to articulate this proposal it is required that identity and
indiscernibility be distinguished. In classical logic and standard mathematics,
identity is formulated in terms of indiscernibility. So, in order to keep one and
change the other, one needs to resist this identification and clearly separate the two
notions. (We will return to this later).
As these quotations illustrate, when it comes to the investigation of the nature of
quantum entities, various possibilities are open. One can examine the commonal-
ities between the conceptions underlying Schrödinger’s and Weyl’s approaches or
pursue their differences. A major feature that is common to both is that they seem
to suggest that something is lost by quantum entities: something that marks a
difference between quantum entities and classical entities.
In this chapter, we address the articulation of the Received View and the
conception of nonindividuality that it attributes to quantum entities. As we discuss
in Section 10.3, the conception can be formulated in distinct ways, some more
radical, others more conservative, at least with regard to the role of the concept of
identity as used in quantum theories. The main issue turns on the behavior of
identity and its relation with individuality. Central to the Received View is the
claim that identity makes no sense, a claim that, as just noted, Schrödinger seems
to have favored. We discuss, in Section 10.4, how to make metaphysical sense of
that idea. The bare claim that identity makes no sense should be accompanied by
an account of how this view entails that particles are not individuals. In Section
10.5, we discuss the formal consequences of the idea, and apply the Received
View to suggest a revision of classical logic. In Section 10.6, we draw some
consequences of this case study to the significance of research in the foundations of
physics.
10.3 The Received View

Common to the claims of both Schrödinger and Weyl quoted earlier is an import-
ant point: What is responsible for the strange metaphysical behavior of quantum
particles is the statistics they obey. Behind this trait one finds encapsulated the so-
called permutation symmetry (PS). According to PS, quantum states should be
symmetric (or antisymmetric) with regard to the permutation of labels of particles.
As a result, if we are to represent a system composed by two particles x1 and x2 , so
that one of the particles is in a region A and the other is in a region B, it cannot be
determined which particle is in which region. (A qualification is in order: Assum-
ing that the underlying “space” is Newtonian and thus, mathematically, its top-
ology is Hausdorff, it follows that two separate points can always be discerned by
disjoint open balls centered on the points [see Krause 2019]. Attention to the
interface between the mathematical framework and the physical setup is
important). In this case, nonsymmetric wave functions, Ψ A ðx1 Þ Ψ B ðx2 Þ or

Ψ A ðx2 Þ Ψ B ðx1 Þ, are unable to describe the situation alone; what is needed is a
superposition of both of them:
Ψ AB ¼ Ψ A ðx1 Þ Ψ B ðx2 Þ Ψ A ðx2 Þ Ψ B ðx1 Þ, except for a normalization factor
(10.1)
Thus, the permutation of A with B results either in the same state Ψ AB in the case of
bosons or in the state Ψ AB in the fermions’ case (see French and Krause 2006:
chapter 4, for a discussion of the physical aspects of this situation and an examin-
ation of how far the metaphysics can go). More importantly, the square of the
resulting wave function, which gives us the relevant probabilities, is preserved
(since jΨ AB j2 ¼ jΨ AB j2 ). Hence, if A stands for an arbitrary observable and P is a
permutation operator, then
hΨ AB jAjΨ AB i ¼ hP Ψ AB jAjP Ψ AB i: (10.2)
As it turns out, this is as far as one can go based on the quotes given previously.
Any additional step will break the shared agreement, given that different metaphysical
conclusions will be drawn in light of the same physical fact.
In both of his claims stated earlier, Schrödinger seems to identify “sameness”
and identity, so that the fact that one cannot attribute sameness to the particles also
means that one cannot attribute identity to them. Individuality is lost as a result of
the lack of sameness. Given that it makes no sense to state that one particle is the
same as the other and given that it is not possible to refer to a particle as “this” one,
particles are no longer individuals.
Of course, the issue is more complex than these considerations suggest. If it were
possible to determine that there is one particle and then another, it would certainly
make sense to state that they are different. But this is not quite what Schrödinger
claims. At this point, an additional ingredient should be added to make clear what
Schrödinger’s conception of individuality ultimately is. With regard to the typical
principle of individuality of the metaphysicians, which accounts for what an entity is
in contrast to others, Schrödinger advances a particular spacetime principle of
individuation, one which accounts for the individuality of an item in terms of its
spatiotemporal position (see French and Krause 2006: chapter 1, for details).
In discussing the individuality of familiar objects, Schrödinger (1998: 204)
claims that science has taken for granted the permanence of pieces of matter, and
this is what accounts for the identity and individuality of objects. This is mani-
fested in one’s confidence when the identity of familiar objects becomes an issue:
When a familiar object reenters our ken, it is usually recognized as a continuation of
previous appearances, as being the same thing. The relative permanence of individual
pieces of matter is the most momentous feature of both everyday life and scientific
experience. If a familiar article, say an earthenware jug, disappears from your room, you
are quite sure somebody must have taken it away. If after a time it reappears, you may
doubt whether it really is the same one – breakable objects in such circumstances are often
not. You may not be able to decide the issue, but you will have no doubt that the doubtful
sameness has an indisputable meaning – that there is an unambiguous answer to
your query.
Compare the view articulated in this passage with the one Schrödinger advanced
earlier when he claimed that the notion of identity makes no sense for quantum
entities (see the quotation from Schrödinger 1996: 121–122, in the previous
section). While ordinary objects typically are supposed to have well-defined
identity conditions, which allows one to answer questions about their identity over
time (even if, in some cases, one may be unable to decide the issue), for quantum
objects such questions do not even make sense. As a result, there is simply no fact
of the matter regarding the individuality (as well as the identity or sameness) of
quantum particles. In fact, in the case of quantum particles, situations involving
distinct observations of an object through time generate problems that prevent the
individuality of the items in question from making sense. As Schrödinger notes
Even if you observe a similar particle a very short time later at a spot very near to the first,
and even if you have every reason to assume a causal connection between the first and the
second observation, there is no true, unambiguous meaning in the assertion that it is the
same particle you have observed in the two cases. The circumstances may be such that they
render it highly convenient and desirable to express oneself so, but it is only an
abbreviation of speech; for there are other cases where the “sameness” becomes entirely
meaningless . . .
Schrödinger highlights the need for identity in order to claim that an entity is
observed in distinct places at distinct times. It is then just one step to add that
without the possibility that a particle observed at one instant of time t 1 is the same
as a particle observed at a later time t 2 , individuality is lost. Given that, on this
view, it makes no sense to state that those particles are the same (or different),
identity loses its meaning. As we have already noted, identity, individuality, and
sameness are taken as conceptually the same by Schrödinger.
Based on these considerations, a straightforward version of the Received View
emerges. Quantum particles are not individuals, given that they have no well-
determined trajectories in spacetime, and it is not possible to identify distinct
detections of an entity as being detections of the same entity (we will return to
this view in the next section and will provide additional details there).
However, this is not the only way to articulate the Received View via spacetime
continuity. Another form is to keep the restriction that quantum entities fail to have
well-defined spatiotemporal trajectories (as is the case in most versions of the

theory), and thus insist that these entities are nonindividuals in this sense, but not
connect this lack of individuality with a lack of identity. That is, one could keep a
form of spacetime nonindividuality, but separate individuality from the logical
relation of identity (see the suggestion in French and Krause 2006: 153 and
Arenhart 2017). Perhaps Schrödinger could be interpreted as suggesting this view:
Identity does not apply to quantum particles, but this is no violation of the Principle
of Identity of Indiscernibles (PII), according to which, some quality always
discerns numerically distinct items. Since, on this view, PII is supposed to apply
only to individuals, which quantum particles are not, such particles do not provide
a violation of the principle. Along these lines, another version of the Received
View could be advanced, along with a distinct conception of nonindividual.
However another option, which is less revisionary regarding the role of identity
than Schrödinger’s, can still be pursued. One can take literally the claim that
quantum particles are nonindividuals but resist to follow Schrödinger in making
the further claim that identity, as a logical relation, loses its meaning. In fact, if
Weyl’s “alibi” is taken to refer to a principle of individuality, then it makes sense to
claim that, even though no alibi is available for quantum particles, this lack of alibi
needs not be connected with the lack of meaning of identity (hence, identity would
be metaphysically deflated, as recommended by Bueno 2014). In order to do that, it
is enough that one resists binding so tightly an item’s individuality with its identity.
Let us explore this option further, given that pursuing this route provides an
additional (alternative) version of the Received View. (See also Arenhart 2017
for a discussion of alternative formulations of the Received View, which do not
involve abandoning identity).
One can take the alibi Weyl refers to as involving a property that distinguishes a
particle bearing it from any other particle. It is always possible to differentiate a
classical particle (that is, a particle described by classical mechanics) from another
particle by at least one property. Of course, it cannot be a state-independent property
(given that particles of the same kind share such properties), but at least their
spatiotemporal location distinguishes them. On this view, no two classical particles
occupy the same location at the same time, due to a principle of impenetrability. This
means that numerically distinct particles have their difference grounded in a prop-
erty that accounts for their numerical diversity and their individuality.
This trait leads to the validity of PII in classical mechanics. According to this
principle, as noted previously, numerically distinct particles are always discernible
by some quality. Entities are individuals due to the fact that once two of them are
present, there is always some property that accounts for their numerical difference:
This is their alibi. In light of PII, an alibi is always available in classical mechanics.
As a result, classical particles are indeed individuals.
In contrast, quantum particles have no alibi – nothing that accounts for their
individuality. Not even spatiotemporal location can be employed to this effect. Due
to the permutation symmetry, quantum particles are indiscernible by their proper-
ties, including both state-dependent and state-independent ones. Hence, the version
of PII presented earlier, according to which there is always some property that
accounts for the numerical diversity of particles, fails in quantum mechanics. The
result is clear: As Weyl noted, there is no alibi for quantum entities (see French and
Krause 2006: chapter 4, for further discussion).
It could be argued that, if properties are unable to account for the numerical
difference of quantum particles, perhaps some relations could do that, such as the
relation “to have spin opposite to” in a given spatial direction. But this proposal is
still unable to account for the particles’ individuality. After all, if x has spin
opposite to y, y also has spin opposite to x (the relation is symmetric). While no
particle has spin opposite to itself (the relation is irreflexive), there is no quantum
mechanical fact of the matter to determine which of x or y has spin up in a given
direction, and which has spin down in that same direction. Thus, those relations,
called weakly discerning relations, in principle can account for the numerical
diversity of the particles (although whether they do account for that is still
debatable; see French and Krause 2006: chapter 4). Despite that, they are unable
to provide an alibi for the particles in question, because weakly discerning relations
are unable to individuate such particles. Accounting for the particles’ numerical
diversity (if at all) is the closest one can get in quantum mechanics to discernibility
(see Muller and Saunders 2008, and the discussion in Lowe 2016).
However, if weakly discerning relations are implemented in a mathematical
context whose underlying set theory is ZFC (Zermelo-Fraenkel set theory with the
axiom of choice), as is the case of Muller and Saunders (2008), all entities become
fully discernible and identifiable in virtue of the resources of set theory alone (we
will return to this point and provide the argument later). Thus, there is a tension
between the motivation for the introduction of weakly discerning relations and the
adopted set-theoretic framework.
In principle, if the option of maintaining that identity holds for quantum
particles can be fully worked out, one could claim that they are different or
identical, without thereby implying that they are individuals. What is required, as
we have been suggesting, is that their individuality be grounded in some kind of
alibi (in Weyl’s sense) that is not formulated in terms of identity.
There are additional possibilities to articulate alibis (that is, principles of
individuality) without requiring the removal of identity (see Arenhart 2017). It is
enough that the content of identity be deflated from the metaphysical content that
would be required if identity also played the role of a principle of individuality.
(For a defense that identity should be deflated, see again Bueno 2014). As will
become clear later (when a formal approach to identity is discussed), identity can
be thought of as something very minimal and without much metaphysical content,
just in terms of two features: reflexivity (that is, every object is identical to itself )
and substitutivity (if x is identical to y, then if x is F, so is y). One could add some
metaphysical content to identity, so that it can be used as a principle of
individuality. But that changes identity by making it more substantive than it
needs to be. Schrödinger, of course, does not seem to follow this path since he
appears to keep identity and individuality very closely connected. As a result,
insisting on nonindividuality requires abandoning identity, at least for quantum
entities. In what follows, we investigate the prospects for a Schrödingerian
approach to nonindividuality.
10.4 Making Sense of Losing Identity

If one is to pursue the option that seems to be suggested by Schrödinger – namely,
that individuality and identity go together, and that one cannot have nonindividuals
without abandoning also the relation of identity – it is crucial to explain in detail
why identity should go hand in hand with individuality regarding quantum
particles.
One of the possible ways of doing that consists in exploring the relation between
identity and individuality through the notion of haecceity, as it is done, for
instance, by French and Krause (2006). Basically, a haecceity is a nonqualitative
property uniquely instantiated by an object, something like an individual essence
possessed by a single individual. Each individual has its own essence, which, of
course, accounts for its individuality (see further discussion in Lowe 2003).
Being nonqualitative, a haecceity does not count as a quality able to discern among
two individuals. That is, two individuals may share every qualitative property, but still
not be the same individual, due to the fact that they have distinct haecceities. As a
result, PII (restricted to qualitative properties) may fail and, despite that, individuality
is still saved in light of a haecceity. In this sense, having a haecceity is what French and
Krause (2006: chapter 1) call a “Transcendental Individuality” principle: that which
confers individuality over and above an item’s qualities.
This point has an important formal counterpart. If the underlying mathematical
framework in which quantum theories are formulated is that of ZFC, one should
consider, in particular, the resulting models – the structures in which quantum
mechanics holds. Within these structures, it is possible that certain entities cannot
be discerned based on the resources of quantum theories alone. However, outside
such structures, it is possible to discern the entities in question, and this can be
thought of as a formal expression of transcendental individuality (clearly, the
relevant structures cannot be rigid).
The problem, however, is not to account for an item’s individuality, but rather
for its nonindividuality. How can haecceity achieve that? The answer seems to be:
through the notion of identity. As French and Krause put it
. . . the idea is apparently simple: regarded in haecceistic terms, “Transcendental
Individuality” can be understood as the identity of an object with itself; that is, ' a ¼ a' .
We shall then defend the claim that the notion of non-individuality can be captured in the
quantum context by formal systems in which self-identity is not always well-defined, so
that the reflexive law of identity, namely, 8xðx ¼ xÞ, is not valid in general.
(French and Krause 2006: 13–14)
That is, a haecceity may be formally represented by self-identity. Plato’s individu-

ality, if it were attributed by a haecceity, would consist in his bearing the property
of being identical with Plato. This, of course, is a nonqualitative property, and it is
able to connect identity (as a logical concept) and individuality (as a metaphysical
concept). This quote also provides the basic idea to make sense of nonindividuality
within a framework that takes into account the Schrödingerian claims that quantum
entities are not individuals and that identity makes no sense for them. In order to
accommodate metaphysically the idea that identity has no meaning for quantum
particles, it is enough that the reflexive law of identity fails or does not hold in
general. Thus, not everything is self-identical. In light of this connection between
self-identity and haecceity, those entities for which the law fails are nonindivi-
duals: They lack a haecceity, which is the individuation principle.
French and Krause acknowledge explicitly that the connection between identity
and individuality is particularly tight:
We are supposing a strong relationship between individuality and identity . . . for we have
characterized ‘non-individuals’ as those entities for which the relation of self-identity
a ¼ a does not make sense.
(French and Krause 2006: 248)
This is only one of the possible ways to accommodate metaphysically the combin-
ation of nonindividuality and the loss of identity. This proposal allows one to make
a good case for the failure of identity, given that the relation between individuality
and identity is very clearly established in this approach. However, in addition to
burdening identity with the role of attributing individuality, there is another
disadvantage of adopting this approach to nonindividuality: It takes us very far
from the Schrödingerian ideas with which we started. Of course, it allows us to
make sense of the claim that identity and difference do not apply to quantum
entities. But the lack of haecceity arguably was not what Schrödinger had in mind
in his discussion of identity and identification of quantum particles. Rather, as
discussed previously, he seems to favor an account of individuality framed expli-
citly in terms of spatiotemporal trajectories.
The following two conditions seem to articulate better the conception of an

entity being individuated by its spatiotemporal position, in the sense that an entity
satisfying these two conditions should be counted as an individual (see also Bueno
2014 and Bueno 2019):
(A) Identity conditions: an individual has (clearly determined) identity conditions.

(B) Persistence conditions: an individual persists over time (despite changes).
Note that these minimal conditions are satisfied by what are typically considered
individuals (such as, chairs, cherries, or chariots). In particular, as Schrödinger
emphasizes, an earthenware jug would be an individual according to this approach,
and so are classical particles, given that their well-determined trajectories ground
both their persistence and their identity conditions (see French and Krause 2006:
chapter 2). Quantum particles in Bohmian mechanics are also individuals
according to this characterization: They have trajectories attributed by hidden
variables.
It should be noted that there are at least two ways of satisfying persistence
condition (B):
(B.1) Essential traits: as long as certain essential traits (or necessary properties) of
an individual are preserved, the individual remains in existence.
(B.2) Closest continuers: given an individual i that satisfies condition (A), at each
moment of time the closest continuer individual to i (the one that shares most
properties with i) is taken to be i (Nozick 1981: chapter 1).
Of course, a haecceity could be an essential trait, and in this way, haecceities could
be used to account for the permanence of an individual. Given that we have already
suggested avoiding a theory of haecceities to account for individuality and to frame
an approach to nonindividuality, we favor the less metaphysically committing
option (B.2). The idea is that an individual persists through a sequence of closest
continuers, which, taken together, account for the permanence of an individual
over time despite the changes it undergoes.
Given this theory of individuality, formulated by the conjunction of conditions
(A) and (B.2), for something to be a nonindividual, three options emerge: condi-
tion (A) can be violated; condition (B.2) can be undermined, or both conditions can
fail. Quantum entities, as the discussion of Schrödinger’s view indicates, violate
both conditions. This is a Humean point: There appears to be no causal connection
that would allow one to determine that similar objects detected in different
moments in time are, in fact, the same. In the quantum case, consider some
quantum entities that have no continuous trajectory. One cannot look for a
quantum mechanical justification to connect two observations of two such entities
through a single trajectory. Nothing in the theory allows us to do that (unless one is
a Bohmian). As a result, as we have seen, Schrödinger claimed that identity makes
no sense for those entities, given that there is no fact of the matter to determine
whether the two observations correspond to the same entity or not. The question of
the identity of the observed entities ends up being entirely ungrounded.
This accounts for both the nonindividuality of the particles and the fact that
identity does not apply to them. This metaphysical picture is closer to what
Schrödinger had in mind, it seems, and it is less inflated than the one first suggested
by French and Krause (2006), which proceeds through the concept of haecceity.
However, both approaches require a corresponding rejection of the overall validity
of identity. One of the ways to accommodate such a limitation of identity is
through logics that restrict identity, the so-called nonreflexive logics. We turn to
them now.
10.5 The Formal Approach to Identity

It is important to be clear about what identity is, particularly when it is stated that
quantum objects lack identity. Throughout this chapter, we have been using the
term “identity” in the sense of what is typically called standard identity (or simply
“identity,” for short) as conceptualized by classical logic and standard
mathematics.
But there is a pretheoretical conception of identity (let us call it a numerical
identity, for lack of a better word). This conception states that every object is
identical just to itself and to nothing else. According to this informal view, when it
is coupled with the tight connection between identity and individuality just dis-
cussed, it follows from the fact that all objects have identity that they are individ-
uals. So, under this interpretation, the informal notion of identity is closely related
to that of individuality.
The informal view of identity discussed earlier is generally thought to be encapsu-
lated in classical logic or in other systems of logic that share the basic features of
identity. Let us focus on classical logic, for the sake of brevity. Classical logic takes
identity as a binary relation between objects of the domain. Identity statements are
usually written as a ¼ b (or as a 6¼ b, depending on the case), in order to express that
the objects denoted by a and by b are the same, they are identical (or are not the same,
they are different). This intuitively means that there are no two objects, but just one,
which can be named (described) by such expressions. A typical case is the famous
identity statement ‘Hesperus = Phosphorus’ (Frege 1960). True statements such as
this make reference to the identity of the objects of the domain, and ‘Hesperus’ and
‘Phosphorus’ denote the same object. One should consider both the syntactic char-
acterization of the notion of identity (given by the binary predicate ‘=’) and the
semantic characterization in which the identity of the objects of the domain of

interpretation is at issue (we denote the domain by D).
Let us consider the semantics first. The identity of D is taken to be the set
I D ¼ fhx; xi : x 2 Dg. This clearly presupposes that the identity of the elements of
D is well determined, and the metamathematical framework is consistent with this
fact. If one assumes a standard semantics, that is, a semantics built in a standard set
theory such as ZFC, then this assumption is met, given that the identity of all sets is
presupposed (we will return to this point later). I D is a set. According to Cantor, it
is thus “a collection of definite and distinct objects of our intuition or of our
thought” (Cantor 1915/1955: 85). This informal, circular idea of a set (of course, it
is not a definition), which accounts for sets in terms of collections, is couched in
terms of numerical identity.
The problem concerns the syntactic side. Is it possible to axiomatize a first-order
logic having a primitive binary predicate for identity having the set I D as its sole
interpretation? That is, is it possible to provide a definition (or a group of
postulates) such that the identity predicate has just I D as a model? The answer,
we argue, is negative. Let us see why.
To begin with, it is important not to confuse numerical identity with the notion
of identity in standard logic and mathematics. Arguably, it is primarily the latter
that can be rigorously dealt with. Suppose that the language L under consideration
is first-order. Two cases emerge. First, L contains just a finite set of primitive
predicates. In this case, we can “define” identity by exhausting all predicates. An
example suffices: Suppose that the predicates are two monadic predicates P and Q
and a binary predicate R. Thus, a ¼ b can be “defined” as follows:
a ¼ b $ ðPa $ PbÞ∧ðQa $ QbÞ∧8xððRxa $ RxbÞ∧ðRax $ RbxÞÞ: (10.3)
The problem with this “definition” is that clearly there can be additional predicates
not belonging to the language that could distinguish a and b, not to mention the
possibility of there being some kind of haecceity that achieves that (as we noted in
the beginning of the previous section when we made the point about the “formal
counterpart” regarding haecceity). In fact, Eq. (10.3) stands only for the
indiscernibility of a and b regarding the predicates of L.
Second, usually first-order languages introduce identity as a primitive binary
predicate ‘=’. In this case, the standard formulation makes use of two postulates,
namely
(R) Reflexivity: 8xðx ¼ xÞ

(S) Substitutivity: x ¼ y ! ðαx ! α½y=xÞ, where x and y are individual variables,
αx is a formula having x free, and α[y/x] results from the substitution of y for x in
some free occurrences of x, in which x and y are distinct variables.
(Note the use of identity in the very formulation of the substitutivity rule: The
variables x and y need to be distinct, that is, not identical). From these postulates, it
follows that identity is symmetric and transitive. Thus, it is an equivalence relation
as well as a congruence relation due to the presence of substitutivity. Logicians say
that identity is the finest congruence over the domain in the sense that if ffi is
another congruence, then a ¼ b entails a ffi b, for all a and b.
However things are not so easy. Postulates (R) and (S) cannot guarantee that the
interpretation of the predicate ‘=’ is the set ID. In fact, it can be shown that a
congruence, other than identity, can be defined over the domain that also models
the predicate of identity (da Costa and Bueno 2009, Krause and Arenhart 2018). In
other words, from the point of view of L, it cannot be known whether one is
working with a structure where ‘=’ is interpreted as the identity of the domain D,
namely, the set I D , or in terms of another structure that has the defined congruence
as the interpretation of syntactic identity. These structures are elementary
equivalent.
Leaving first-order languages behind, higher-order languages should then be
considered. It suffices to consider L as a second-order language (the generalization
to other higher-order languages is immediate). In this case, identity can be
(allegedly) “defined” in terms of indiscernibility (indistinguishability) by what is
called Leibniz law, namely:
x ¼ y if, and only if, 8F ðFx $ FyÞ, (10.4)
where x and y are variables for individuals, and F is a variable for properties of
individuals. The right side of the biconditional expresses the indiscernibility of x
and y, and it states that the objects that stand for x and y have the same properties
(hence they also share all relations).
The problem now is with the semantics. Suppose that the domain is the none-
mpty set D ¼ f1; 2; 3; 4; 5g and that our second-order language has three monadic
predicate constants – P, Q, R – and two individual constants – a and b. Consider
the following interpretation: 1 is assigned to a and 2 to b. Furthermore, the
extensions of the predicates are interpreted as the following sets: A ¼ f1; 2; 3g,
B ¼ f1; 2; 4g, and C ¼ f1; 2; 5g. Thus, since 1 and 2 belong to all sets, it follows
that a and b have all properties in common. In other words, the right side of
Eq. (10.4) holds, despite the fact that 1 6¼ 2.
The only way of guaranteeing that Eq. (10.4) will have its full intuitive meaning
is to add all subsets of D to the semantics, that is, to consider what Church calls
principal interpretations (Church 1956: 307). But then, as is well known, com-
pleteness is lost.
As these considerations make clear, identity is not a simple concept when one
tries to provide a rigorous account of the intuitive idea. But from a logical point of
view, this is what classical logic presents us with. Based on this theory of identity,
which is called classical theory of identity (CTI), we can consider stronger
systems, such as various set theories.
As is well known, there are several nonequivalent set theories with distinct
properties and which yield significantly different, and even incompatible, the-
orems. For instance, ZFC includes the axiom of choice; Quine-Rosser’s NF
(New Foundations) system does not: It is incompatible with this axiom (Forster
2014). In ZFC, if consistent, there is no Russell set, namely R ¼ fx : x 2 = xg, but in
some paraconsistent set theories, this set is legitimate (da Costa, Krause, and
Bueno 2007). It can be proved, in ZFC, that there are sets that are not Lebesgue
measurable, but in “Solovay set theory” all sets are Lebesgue measurable (Mait-
land Wright 1973). What is remarkable is that all these set theories invoke the same
theory of identity (CTI). Thus, our considerations apply to all of them.
It is undeniable that set theory is the most widely used basis for standard
mathematics, that is, the part of mathematics that can be developed in theories
such as ZFC. This is also the mathematics that underlies quantum theories. In fact,
it is unclear what kind of quantum mechanics could be developed in a system such
as NF, given its incompatibility with the axiom of choice (AC). After all, AC is
necessary for the usual mathematical formulation of quantum mechanics, so that it
can be guaranteed, for instance, that the relevant Hilbert spaces have a basis (of
course, quantum mechanics can be developed in many different ways that need not
rely on von Neumann’s approach; see Styer et al. 2002).
It is a remarkable fact, we noted, that in all of these set-theoretic frameworks, all
objects are individuals, in the sense that all of them have identity. In other words,
given any objects (that is, any sets; the case of Urelemente will be mentioned
soon), there is always a way to distinguish them, if not effectively, at least in
principle. The proof is immediate. Given a certain object a, which is either a set or
a Urelement, the postulates of a set theory enable us to form the set {a}, the
singleton of a (as is well known, there are pure set theories, containing only sets,
and impure set theories, systems that also include atoms – the Urelemente in the
original Cantor’s terminology. These atoms are not sets but can be elements of
sets). Define the “property” Id a ðxÞ≕ x 2 fag. The only object that has such a
property is a itself, so a has at least one property distinguishing it from any other
object. Leibniz law applies and, thus, there cannot be indistinguishable but non-
identical objects.
Indiscernible entities can be accommodated in a set theory via equivalence
relations. The elements of an equivalent class can be taken as representing the
same object, but this is clearly a mathematical trick and does not work as part of a
philosophically well-motivated proposal. A trick similar to this is used in orthodox
quantum mechanics when symmetric and antisymmetric wave functions are
chosen to stand for certain quantum systems: Functions are selected that do not
alter the probabilities when particle labels are exchanged. (This trick was called
“Weyl’s strategy” because it was used by Hermann Weyl; see French and Krause
2006: section 6.5.1).
As a result, within standard mathematics, there are no absolutely indiscernible
objects as quantum objects are said to be in certain situations. Thus, if we
use standard mathematics in our preferred formulation of quantum mechanics
(the same point applies to quantum field theories), from the simple fact that there
are two quantum objects, it results from the mathematics alone that the objects are
different (they are not identical), and by Leibniz law, there is at least one property
that one of them has and the other does not. However, if the objects in question are
indiscernible, such as two bosons in the same state, which property would that be?
The assumption of the existence of such a property amounts to the introduction of
hidden variables – even in those formulations of quantum mechanics that do not
accept them. But the fact that there is such a property follows from Leibniz law
(which, as noted, is part of the package formed by classical mathematics, which
includes a corresponding logic, and the standard theory of identity). Thus, in any
situation, given two quantum objects, there is a difference between them. Such a
difference cannot be given by a substratum (a haecceity), because the existence of
such a substratum is ruled out in quantum theories (see Teller 1998). The differ-
ence can be expressed in terms of a bundle theory of properties, which leads to the
conclusion that there is a property that only one of the quantum objects in question
have, but not the other. The problem is that, according to quantum theories,
assuming their usual interpretations, this is not a viable possibility. Otherwise,
quantum objects would be discernible. In the end, what is needed is a framework
that does not preclude the possibility of indiscernible but potentially distinct
systems of entities – a framework that makes room for nonindividuals.
An appropriate, philosophically well-motivated, strategy would then be to leave
standard set theories behind and adopt a set theory in which identity is not taken to
hold in general, namely, a quasi-set theory. This is a mathematical framework
which can be used as a metamathematics for quantum theories (see French and
Krause 2006, Domenech, Holik, and Krause 2008, Krause and Arenhart 2016). In
this theory, collections (called quasi-sets) can be formed by absolutely indiscern-
ible elements without thereby becoming identical. As a result, Leibniz law is
violated for some objects (although it remains valid for another kind of objects,
called classical). These collections of indiscernible entities can have a cardinal,
called its quasi-cardinal, even if they do not have an ordinal. The theory provides a
framework to examine collections of objects without ordering them, without
identifying or individuating them. And differently from classical set theories, the
theory offers a framework in which nonindividuals can be formulated and
thoroughly studied without the incoherence found in the use of classical set
theories for the formulation of the foundations of quantum mechanics (for details,
see French and Krause 2006, Krause and Arenhart 2016). We conclude this chapter
by noting the significance of foundational studies of physics, of which quasi-set
theories provide a clear case.
10.6 Conclusion
In this chapter, the metaphysical underpinnings of the idea that quantum entities
are nonindividuals have been examined. Schrödinger’s claim that identity does not
make sense for quantum entities was interpreted, and the connections between this
claim and some issues related to continuous trajectories in quantum theory were
investigated. The resulting metaphysics of nonindividuals assumes a tight connec-
tion between identity and individuality, so that if individuality has to go, so does
identity. Given that identity is a logical relation, which is part of classical math-
ematics, and since almost every system of logic has a version with identity, it is
important to provide an account of what it is like for a system of logic to have no
identity. A few more words on the importance of developing such formalisms and
their relation with physics are, thus, in order.
First, consider some reasons to look for alternative mathematics (and logic) for
quantum mechanics. Leaving aside historical considerations (an up-to-date analy-
sis, which also considers some historical facts, can be found in Maudlin 2018), the
motivation has to do with the foundations of physics. One could argue that physics
works fine with standard (Leibnizian) mathematics (and logic), as it can be seen by
considering any book on quantum mechanics. In particular, the argument goes,
questions about the foundations of physics could be regarded as “mere philosoph-
ical” problems that, on their own, contribute nothing to the clear understanding of
physics. That this view is untenable becomes clear by considering some of
the papers in Cao (1999) and the significant insights that a careful reflection on
the foundations of physics provides (for the sake of brevity, we will not revisit the
various arguments here).
It could be argued that something similar happens with current physical theor-
ies. Current physics works fine with two incompatible theories, namely, the
standard model of particle physics and general relativity. The former provides
the best way developed so far to account for the physics of the small, while the
latter offers the best physics of the big, as it were. One or the other is applied
depending on the subject matter under study. However, these two theories are
logically incompatible with one another, for gravity has not been quantized.
Should the situation be left at that, with everyone being encouraged to accept that
physics has reached its final limit, and no unification is ultimately necessary?
Of course not. Novelties have always emerged when foundational issues have
been pursued. In mathematics, this is undeniably the case, as the development of
logic and set theory clearly illustrates. There is no reason to think that quantum
mechanics would be any different. Indeed, the relevance of string theories, loop
gravity, and any other attempt to find a more fundamental theory, in particular, the
quantization of gravity, cannot be appreciated without acknowledging the signifi-
cance of foundational research. In fact, without the latter, it would be difficult to
make sense of why physicists systematically pursue such enterprises.
To look for more suitable mathematical bases for a coherent metaphysical
conception of quantum objects as nonindividuals is reasonable and even necessary
(to prevent inconsistencies). Arguably, no one seems to know, and perhaps no one
will ever know, what quantum entities ultimately are. All one has are one’s
theories. Even the concept of particle changes from theory to theory (see Falken-
burg 2007: chapter 6). Foundational research provides some perspective and
insight to pursue the search for understanding that is integral to the attempt of
making sense of these issues as well as their significance.
This brings the second topic to be addressed here: ontology. Physicists, in
general, have a broad and intuitive idea of what ontology is, but some of them
do not find it relevant to their work in physics. Ontology, it was noted, is tradition-
ally occupied with what there is (in the world) – with existential questions, such as:
Are there winged horses? Are there electrons? Are there transcendental numbers?
Metaphysics is more general and includes ontology as a proper part. For instance,
Democritus’ claim that the world is composed by atoms (indivisibles) is a meta-
physical view. It concerns the basic structures of the world. Even in classical logic
one finds metaphysical assumptions. For example, classical propositional logic
assumes a metaphysical, semantic principle to the effect that the truth of a complex
formula depends on the truth of its component formulas, usually referred to as
Frege’s Principle of Compositionality (Szabó 2017).
Physics is not different, and it also has its share of metaphysical claims. One of
them, crucial to the entire discussion examined in this work, is that quantum
objects – whether they are particles in orthodox quantum mechanics or field
excitations in quantum field theories – are ultimately nonindividuals. This claim,
of course, does not force us to assume that nonindividuals exist. Situations can be
presented in a conditional form: If there are things like quantum entities, then they
can be interpreted as nonindividuals. This process of interpretation, itself an
integral part of foundational research, provides a possible approach to the under-
standing of the nature of such entities.
Just to be clear, we are not asserting that quantum objects are nonindividuals. It
is unclear how this claim could be established. Rather, the goal is to develop the
view as coherently as possible and indicate how it helps one to understand
quantum entities. Bohmian mechanics (BM) works with a “classical” metaphysics

involving “classical” individuals, each one having their own identity. But Boh-
mians should also be careful and acknowledge that their hypothesis is just a
hypothesis, a metaphysical view among several others. The physics works fine
with BM, at least at the level of its nonrelativistic counterpart (we make no
commitment regarding relativistic BM). Clearly, Bohmian mechanics provides
an additional example of the significance of foundational research.
To conclude, placed in classical metaphysics and in standard underlying math-
ematics (based on ZFC), quantum objects cannot be completely indiscernible. The
resulting theory can provide predictions of quantum phenomena, such as in the
two-slit experiment (see Holland 2010), but these are just predictions that are able
to make the physics work. The problem, however, is the logic that is being invoked
is inconsistent with the indiscernibility of the phenomena in question. Quantum
statistics, Gibbs paradox, and many other quantum phenomena presuppose abso-
lute indiscernibility. As Wiczek and Devine have said, “in the microworld, we
need uniformity of a strong kind: complete indistinguishability” (Wilczek and
Devine 1987: 135). So, from a logical point of view, predictions are not enough:
One needs a proper foundational mathematical framework. The commitment that
classical logic and standard mathematics have to Leibniz law questions their
adequacy to accommodate, in a proper way, truly indistinguishable things.
A different path is then called for.
References
Arenhart, J. R. B. (2017). “The received view on quantum non-individuality: Formal and
metaphysical analysis,” Synthese, 194: 1323–1347.
Bueno, O. (2014). “Why identity is fundamental,” American Philosophical Quarterly, 51:
325–332.
Bueno, O. (2019). “Weyl, identity, indistinguishability, realism,” in A. Cordero (ed.),
Philosophers Look at Quantum Mechanics. Dordrecht: Springer.
Cantor, G. (1915/1955). Contributions to the Founding of the Theory of Transfinite
Numbers. New York: Dover.
Cao, T. (ed.) (1999). Conceptual Foundations of Quantum Field Theories. Cambridge:
Church, A. (1956). Introduction to Mathematical Logic. Princeton: Princeton University
Press.
da Costa, N. C. A. and Bueno, O. (2009). “Non-reflexive logics,” Revista Brasileira de
Filosofia, 232: 181–196.
da Costa, N. C. A., Krause, D., and Bueno, O. (2007). “Paraconsistent logics and para-
consistency,” pp. 791–911 in D. Jacquette (ed.), Philosophy of Logic. Amsterdam:
North-Holland.
Domenech, G., Holik, F., and Krause, D. (2008). “Q-Spaces and the foundations of
quantum mechanics,” Foundations of Physics, 38: 969–994.
Falkenburg, B. (2007). Particle Metaphysics: A Critical Account of Subatomic Reality.

Dordrecht: Springer.
Forster, T. (2014). “Quine’s new foundations,” in E. N. Zalta (ed.), The Stanford Encyclo-
pedia of Philosophy (Fall 2014 Edition), https://plato.stanford.edu/archives/fall2014/
entries/quine-nf/
Frege, G. (1960). “On sense and reference,” pp. 56–78 in P. Geach and M. Black (eds.),
Translations from the Philosophical Writings of Gottlob Frege. Oxford: Basil
Blackwell.
French, S. and Krause, D. (2006). Identity in Physics: A Historical, Philosophical, and
Formal Analysis. Oxford: Oxford University Press.
Heisenberg, W. (1998). “The nature of elementary particles,” pp. 211–222 in E. Castellani
(ed.), Interpreting Bodies: Classical and Quantum Objects in Modern Physics.
Holland, P. R. (2010). The Quantum Theory of Motion. Cambridge: Cambridge University
Press.
Krause, D. (2019). “Does Newtonian space provide identity for quantum systems?,”
Foundations of Science, https://link.springer.com/article/10.1007/s10699-018-9561-3
Krause, D. and Arenhart, J. R. B. (2016). “Presenting non-reflexive quantum mechanics:
Formalism and metaphysics,” Cadernos de História e Filosofia da Ciência, 4: 59–91.
Krause, D. and Arenhart, J. R. B. (2018). “Quantum non-individuality: Background
concepts and possibilities,” pp. 281–306 in S. Wuppuluri and F. A. Doria. (eds.),
The Map and the Territory: Exploring the Foundations of Science, Thought and
Reality. Dordrecht: Springer.
Lowe, E. J. (2003). “Individuation,” pp. 75–95 in M. J. Loux and D. W. Zimmerman
(eds.), The Oxford Handbook of Metaphysics. Oxford: Oxford University Press.
Lowe, E. J. (2016). “Non-individuals,” pp. 49–60 in A. Guay and T. Pradeu, T. (eds.),
Individuals Across the Sciences. New York: Oxford University Press.
Maitland Wright, J. D. (1973). “All operators on a Hilbert space are bounded,” Bulletin of
the American Mathematical Society, 79: 1247–1251.
Maudlin, T. (2018). “The labyrinth of quantum theory,” https://arxiv.org/abs/1802.01834
Muller, F. A. and Saunders, S. (2008). “Discerning fermions,” British Journal for the
Nozick, R. (1981). Philosophical Explanations. Cambridge, MA: Harvard University Press.
Schrödinger, E. (1996). Nature and the Greeks and Science and Humanism (Foreword by
Roger Penrose). Cambridge: Cambridge University Press.
Schrödinger, E. (1998). “What is an elementary particle?,” pp. 197–210 in E. Castellani
Styer, D. F., Balkin, M. S., Becker, K. M., Burns, M. R., Dudley, C. E., Forth, S. T., . . .
Wotherspoon, T. D. (2002). “Nine formulations of quantum mechanics,” American
Journal of Physics, 70: 288–297.
Szabó, Z. G. (2018). “Compositionality,” in E. N. Zalta (ed.), The Stanford Encyclopedia
of Philosophy (Summer 2017 Edition), https://plato.stanford.edu/archives/sum2017/
entries/compositionality/
Teller, P. (1998). “Quantum mechanics and haecceities,” pp. 114–141 in E. Castellani
Weinberg, S. (1992). Dreams of a Final Theory. New York: Vintage Books.
Weyl, H. (1950). The Theory of Groups and Quantum Mechanics. New York: Dover.
Wilczek, F. and Devine, B. (1987). Longing for Harmonies: Themes and Variations from
Modern Physics. London: Penguin Books.
11
From Quantum to Classical Physics:
The Role of Distinguishability
ruth kastner
11.1 Introduction
This chapter consists of three main sections. In Section 11.2, I review Nick
Huggett’s finding regarding the nonrelevance of permutable labeling for the
classical/quantum divide (Huggett 1999). In Section 11.3, I review the derivations
of the classical and quantum statistics and argue that a form of separability is a key
feature of the quantum-to-classical transition. In Section 11.4, I consider the
question: What allows separability to serve as a form of distinguishability in the
classical limit? First, let us review some basic considerations regarding issues of
individuality and distinguishability.
Steven French (2015) has noted that the concept of individuality is primarily a
metaphysical issue, while that of distinguishability is primarily an epistemological
issue. Nevertheless, distinguishability does have bearing on ontological questions
such as:
What is an individual?
Are there any true individuals?
Does Leibniz’ Principle of Identity of Indiscernibles apply to nature?
However, I will not enter here into the metaphysical debate concerning questions
such as: What is an individual? and Are quantum systems individuals? Rather,
I will focus on the issue of distinguishability regarding the quantum-classical
divide and attempt to identify some ontological features that may underlie the
form of distinguishability obtaining in that context.
11.2 Huggett’s Finding on Haecceitism and Classical Objects

Davis Lewis (1986) introduced the term haecceitism, which denotes a form of
strong individuality: An individual’s identity is taken as a primitive “this-ness”
205
206 Ruth Kastner
which transcends all its qualitative features. This concept can be identified with the
term ‘transcendent individuality’ (TI) as discussed in French and Redhead (1988).
Anti-haecceitism consists in saying that an individual’s identity is constituted by its
qualitative features and nothing more. Although the precise definition of haecceit-
ism is still a matter under discussion, for our purposes we can think of it as the
capacity of an entity to carry a label or “name,” where that label is not contingent
on any of its qualitative features. Thus, what makes a person named Fred “Fred the
individual” is his primitive this-ness, not the color of his eyes or hair or his height,
weight, etc.
Now let us consider this notion as applied to some typically classical objects,
such as a pair of coins that are assumed to be completely identical and can be either
“heads” or “tails.” Give them name-labels, say ‘Fred’ and ‘Joe’ their assumed
haecceitism is represented by their name-labels. In this context, haecceitism
implies that if we consider the case in which Fred and Joe are in different states
(one of them being “heads” and the other “tails”), then interchanging Fred and Joe
constitutes two different possible configurations. If we include the cases in which
Fred and Joe are both “heads” or both “tails,” we have four possible states of the
coins, as seen in Figure 11.1.
In contrast, for a pair of hypothetical “boson coins,” the usual story is that there
is no such thing as Fred or Joe no nameable identity that transcends the
qualitative properties of the quantum coins. So the possible configurations are just
three in number (Figure 11.2).
It should be noted that French and Redhead (1988) dissent from this usual
identification of individuality with the capacity to carry a label such that permuta-
tion of the labels establishes a different state of the total system. They argue that a
form of individuality can still be retained for quantum systems if one argues that
certain states are not accessible to the total system. For purposes of this discussion,
Figure 11.1 States of two classical coins.

From Quantum to Classical Physics: The Role of Distinguishability 207
Figure 11.2 States of two hypothetical “boson” coins.
we work with Huggett’s formulation, but note that his interpretation of the
metaphysical bearing of the labels is not obligatory.
Huggett (1999) notes that there are two ways of describing the state-space of a
composite system such as the set of two coins. We can either use a phase space
(Γ-space) description, which specifies which component system is in which state,
or we can use a distribution space (Z-space) description, which just specifies how
many systems are in each state. The Γ-space description assumes that each
component can be meaningfully labeled and/or distinguished from the others, so
it supports haecceitism in that respect. In contrast, since Z-space specifies only the
occupancy number of each state, without identifying any particular system with
any particular state, it does not support haecceitism in the same way. Since it is
typically supposed that the key distinction between classical and quantum objects
is the ability of the former to carry a label, one would think that the two kinds of
descriptions phase space and distribution space would lead to different kinds
of statistics; i.e., classical and quantum statistics, respectively.
However, Huggett shows that if we assume that classical objects are impene-
trable, i.e., that no more than one such object can never occupy a given spacetime
point, then it turns out that the Γ- and Z-space descriptions give the same empirical
predictions. Thus, we cannot use any experimental data to decide between them.
This means that there is no empirical support for the idea that classical and
quantum objects differ fundamentally in their metaphysical nature as individuals.
The basic argument goes like this: In terms of the coin analogy, we have to pretend
that there are no other qualitative differences between the coins and forbid the two
coins, Fred and Joe (they can keep their labels), from occupying the same state. Of
course, real coins would not fulfill this criterion. For the more realistic case of classical
gas molecules, the operative condition is that no two molecules can ever occupy the
same individual phase space state, since they can never be at the same spacetime point.
208 Ruth Kastner
Figure 11.3 States of two classical coins if we forbid them from occupying the
same state.
In the case of the idealized coins, if we forbid them from occupying the same
state, there are now only two available composite Γ-states for Fred and Joe the
ones in which they are in different “heads” or “tails” states (Figure 11.3).
Additionally, since both of these correspond to the distribution “one coin in each
state,” the frequency of this distribution is 2/2 = 1. Meanwhile, the frequency of
this distribution in terms of the Z-space representation, which ignores the phase
space configurations, is just 1/1 = 1. We see that, for the idealized classical coins,
the frequency of occurrence of the distribution is exactly the same in either
representation, so there is no empirical difference between the two spaces they
both predict the same probabilities. Huggett shows that this holds in general, for an
arbitrary number of systems and states (i.e., in which the frequency of a given
distribution may differ from unity, in contrast to the trivial example shown earlier).
Thus, it turns out that there is no empirical support for the Γ-space description
over the Z-space description for classical systems, if they are correctly character-
ized as impenetrable, and thus no empirical support for haecceitism as applying to
classical objects—if one identifies that as a criterion for haecceitism, as does
Huggett. Based on the dissent of French and Redhead (1988), from the criterion
described earlier for transcendental individuality, these authors could, of course,
still argue that both quantum and classical systems possess metaphysical individu-
ality. What is off the table, in view of Huggett’s argument, is the idea there is any
empirical support for a fundamental difference between quantum and classical
systems regarding their status as individuals. While somewhat bewildering for our
intuitions about the difference between classical and quantum objects, we actually
need this result. Why? Because, in keeping with the correspondence principle, the
classical (Maxwell-Boltzmann) distribution must (and does) emerge as a limit from
quantum statistics (either Bose-Einstein or Fermi-Dirac). That is, the quantum

distributions transition smoothly into the classical distribution. This calls into
question the idea that classical objects have any sort of “digital” or on/off
distinguishability or individuality feature or features that differ from quantum
objects.
Thus, the challenge facing us is that the transition from the quantum domain to the
classical domain seems continuous, not discrete and essential (as in a change of
intrinsic character or essence). This is another puzzling feature of the micro/macro
divide. In the next section, we review the derivations of each kind of distribution and
consider what clues we might find therein to better understand the ontology under-
lying the transition between the quantum and classical statistics.
11.3 Classical versus Quantum Statistics

Let us begin by simply listing the three major distributions: the classical Maxwell-
Boltzmann, Bose-Einstein for bosons, and Fermi-Dirac for fermions, respectively:
ðMBÞ eβεi
ni ¼ N P βε (11.1a)
e j
j
ðBEÞ 1
ni ¼ (11.1b)
eβðεi μÞ 1
ðFDÞ 1
ni ¼ (11.1c)
eβðεi μÞ þ1
In the quantum distributions Eq. (11.1b) and Eq. (11.1c), the chemical potential μ
(related to the number of degrees of freedom N) necessarily enters for systems with
a fixed N. This will turn out to be significant, as we shall see.
Now, recall that classical distributions can only be wavelike or particle-like. The
particle-like classical distribution (applying to systems such as ideal gases) is just
the Maxwell-Boltzmann distribution Eq. (11.1a). Meanwhile, the classical wave
distribution is what was applied to blackbody radiation prior to the advent of
quantum theory, resulting in the Rayleigh-Jeans distribution and the “ultraviolet
catastrophe”:
2ε2 kT
PðεÞRJ ¼ (11.2)
ðhcÞ2
In view of “wave-particle duality,” it is well known that the quantum distri-
butions interpolate between these two extremes, as follows. Consider the
210 Ruth Kastner
correct quantum distribution for electromagnetic blackbody radiation (the

“Planck Distribution”):
2ε3 1
PðεÞ ¼ 2 eε=kT
(11.3)
ðhcÞ 1
For energies ε small compared to kT, this becomes

2ε3 1 2ε3 kT 2ε2 kT
Pðε kT Þ ¼ ¼ , (11.4)
ðhcÞ2 ð1 þ ε=kT þ Þ 1 ðhcÞ2 ε ðhcÞ2
i.e., it yields the Rayleigh-Jeans law Eq. (11.2).
On the other hand, for ε large compared to kT, the exponential in the denomin-
ator of Eq. (11.3) swamps the unity term and we get
1
Pðε kT Þ / (11.5)
eε=kT
which is the Maxwell-Boltzmann distribution, reflecting particle-like (or at least
discrete) behavior on the part of the radiation. (To better reveal the basic form of
the distribution in this limit, we neglect the factors corresponding to the density of
states for blackbody radiation.)
Thus, we see that the quantum statistics interpolate between wavelike and
particle-like behavior. This is a key aspect of quantum systems as opposed to
classical systems; the latter can be unambiguously categorized as either waves or
particles. In contrast, the quantum statistics must cover both situations in the same
distribution indicating that they describe entities that are (somehow) both wave
and particle.
If we look at the assumptions that go into deriving the various distributions, we
can get some additional clues as to the key differences between the classical and
the quantum situation. For example, the classical Maxwell-Boltzmann distribution
for an ideal gas (e.g., a system of N molecules) is obtained from a partition function
assumed to be the direct product of the N individual molecular partition functions.
This cannot be done for the quantum statistics for fixed N. We will now consider
those issues in detail.
First, let us recall the general procedure for obtaining a distribution describ-
ing the mean number of systems nr in a given energy state εr . This procedure
holds regardless of the type of system considered (whether quantum or
classical).
• The number of degrees of freedom having energy εr is denoted nr .

• Thus, the possible energy states ER of the whole gas (having N particles) are:
X
E R ¼ n1 ε1 þ n2 ε2 þ ¼ nε
r r r
(11.6)
and
X
N¼ n
r r
(11.7)
At this point, it may already be noted that Eq. (11.6) represents a distribution over
the possible energy states, and in that sense is the “Z-space” representation. Since this is
a general derivation (leading to both classical and quantum statistics), it is clear that the
Z-space representation is applicable for both cases, reinforcing Huggett’s observation.
Now, for the case in which the total system of N degrees of freedom is taken as
capable of exchanging energy with its environment at temperature T (the “canon-
ical ensemble”), the probability that the total system is in the state R is given by:
PR ¼ C eβER (11.8)
where β ¼ 1=kT.
The constant of proportionality C is 1/Z, where Z is the total system partition
function:
X
Z¼ j
eβEj (11.9)
So the probability that the gas is in state R is:

eβER
PR ¼ (11.10)
Z
From this we can find the average number of degrees of freedom in energy state εr :
P P
X nr eβER 1 1 X ∂ β nr εr 1 1 ∂Z
nr ¼ nr PR ¼ R
¼ e r ¼ (11.11)
R
Z β Z R
∂εr β Z ∂εr
So that in compact form,

1 ∂ ln Z
nr ¼ (11.12)
β ∂εr
Again, this is a general result for any partition function Z.
Now, for a single degree of freedom with possible energy states εi , the partition
function ζ (i.e., the weighted sum over the possible energy states) is given by:
X
ζ ¼ j
eβεj (11.13)
So, analogously with Eq. (11.10), the probability that a single system is in state εi is
eβεi eβεi
Pðεi Þ ¼ P βε ¼ (11.14)
je
j ζ
212 Ruth Kastner
We make note of this because for a classical gas of N degrees of freedom, one finds
the average number simply by taking Z(N) for the entire gas as the product of the
individual partition functions:
Z ðN Þ ¼ ζ N (11.15)
So that, using Eq. (11.12), the distribution for ni becomes:
1 ∂ ln Z ðN Þ 1 ∂ ln ζ eβεi
ni ¼ ¼ N ¼ N P βε (11.16)
β ∂εi β ∂εi je
j
which is just the Maxwell-Boltzmann distribution Eq. (11.1a).

However, one cannot use the expression Eq. (11.15) for quantum systems
that have a constrained number of degrees of freedom N and this is of
crucial significance. Instead, one must incorporate the restriction to N by way
of the chemical potential μ, which acts as a Lagrange multiplier. This dictates
that we are working with the “grand canonical ensemble,” which allows N to
^
vary. The corresponding grand canonical partition function Z is obtained as
follows:
^ X
Z¼ exp ½βER exp ½βμN
R
X
¼ exp ½βðn1 ε1 þ n2 ε2 þ n3 ε3 þ . . .Þ exp ½βμN ¼
n1 , n2 , n3 , ...
X
¼ exp ½βðn1 ε1 þ n2 ε2 þ n3 ε3 þ . . .Þ exp ½βμðn1 þ n2 þ n3 þ . . .Þ ¼
n1 , n2 , n3 , ...
X
¼ eβn1 ðε1 μÞ eβn2 ðε2 μÞ eβn3 ðε3 μÞ . . .
n1 , n2 , n3 , ...
! ! !
X
∞ X
∞ X
∞
βn1 ðε1 μÞ βn2 ðε2 μÞ βn3 ðε3 μÞ
¼ e e e ... (11.17)
n1 ¼0 n2 ¼0 n3 ¼0
P
This is just a product of infinite sums of the form ∞n¼0 xn ¼ 1=ð1 xÞ, jxj < 1;
and given that μ < εr , we therefore have

^ 1 1 1
Z¼ ... (11.18)
1 eβðε1 μÞ 1 eβðε2 μÞ 1 eβðε3 μÞ
Taking logs of both sides, we get the more useful form:
^ X
∞
ln Z ¼ ln 1 eβðεr μÞ (11.19)
r¼0
Then we can use Eq. (11.12) to get the distribution for average occupation
number ns :
" #
^
1 ∂ ln Z 1 ∂ X∞
ns ¼ ¼ N ln 1 eβðεr μÞ
β ∂εs β ∂εi r¼0
eβðεs μÞ 1
¼ β ð ε μÞ
¼ βðε μÞ (11.20)
1e s e s 1
which is the Bose-Einstein distribution.
The first thing to notice here (besides the fact that we could not use Eq. (11.15)
to obtain this quantum distribution) is that the total number of degrees of freedom,
N, seems to have “disappeared.” It got “dissolved” into infinite sums over all the
possible values of the ns ðεs Þ. Thus, ironically, N has to become a variable in order
to be able to “fix” N for a gas of quantum systems. We recover N as the sum over
the average occupation numbers ns :
X
NN ¼ ns (11.21)
s
The situation is similar for fermions, except that they obey the Pauli Exclusion
principle which limits state occupancy to zero or one. Without going through the
derivation here, we note that, given the restriction described earlier on occupancy,
the inability to express the partition function Z(N) as a direct product of N
individual degrees of freedom yields for the mean occupancy number:
1
nðsFDÞ ¼ (11.22)
eβðεs μÞ þ1
which is the Fermi-Dirac distribution.
Thus, for both bosons and fermions, the chemical potential μ is involved in a
crucial, nonseparable way. Its relation to N is fixed by (11.21), i.e.,
X 1
¼
N¼N (11.23)
s eβðεs μÞ 1
Does the chemical potential play any role in the classical case? Yes, but only
trivially, as a normalizing factor. In the “dilute” (low-occupancy) limit yielding the
classical case, the exponential factor involving μ approaches the particle number N
divided by the single-particle partition function, i.e.,
N N
eβμ ! ¼ P βε (11.24)
ζ e j
j
214 Ruth Kastner
So that the Maxwell-Boltzmann distribution can be expressed in terms of μ as:

ðMBÞ N βεi 1
ni ¼ e ¼ eβμ eβεi ¼ βðε μÞ (11.25)
ζ e i
In this form, it is easy to see that the classical case emerges from the quantum
distributions when εi μ for all i.
In summary, we make the following observations based on the derivations of the
respective distributions. For a classical system comprising N degrees of freedom,
we can simply assume that N is fixed, and use the “canonical ensemble” to obtain
the distribution. For that purpose, we can express the total canonical partition
function Z(N) as simply the product of the individual partition functions ζ i for the
component degrees of freedom.
However, for a quantum system with fixed N (and nonvanishing mass), we
cannot use the canonical ensemble; we must use the “grand canonical ensemble”
^
(i.e., partition function Z ) – representing a system in contact with both an energy
and particle reservoir. That is, we must in-principle allow N to vary. The physical
content of this procedure is as follows: The chemical potential μ is a Lagrange
multiplier, representing a constraint force that is present in the quantum case, even
if there is no contact with an external particle reservoir. Thus the natural physical
interpretation is that the quantum degrees of freedom are imposing a constraint
force on one another that is not present in the classical case.
Based on the previous discussion, what can we conclude about the classical/
quantum divide? We cannot treat a collection of N quantum objects as elements of
separable probability spaces, because in that case we do not obtain the quantum-
mechanical statistics. Nonseparability of the spaces confirms that we are dealing
with quantum coherence, with all its attendant features such as entanglement and
the requirement for symmetrization rendering labels superfluous (the latter usually
and reasonably understood as reflecting indistinguishability). Moreover, the
mutual constraint of quantum systems expressed by the chemical potential μ (even
at T = 0) reflects a peculiarly quantum sort of physical correlation or interaction not
present in the classical case, probably expressing the so-called ‘exchange forces’
associated with symmetrization (of course, this is a misnomer; there is no real
“force” operating here in the usual physical sense).
Thus, our finding is that there is no empirical support for any ‘digital’ on/off form
of metaphysical individuality at the classical/quantum border. Since the classical
statistics are straightforwardly obtainable as a limit from the quantum statistics, and
representable in terms of Z-space, we can confirm Huggett’s result that the capacity
of classical systems to carry labels that simply permute to form new system states
(i.e., new Γ-space configurations) is not reflected in the statistics. Yet clearly, the
classical limit brings with it some sort of new capacity for permutable labeling of the
component systems, in that the collective partition function can be obtained from
individual partition functions ζ i in-principle capable of carrying the permutable
label i. This indicates that in the classical limit, the component systems acquire a
form of distinguishability. In the next section, we investigate the nature of this
emergence, in the classical limit, of the capacity to carry a label.
11.4 Whence Quasi-Classical Distinguishability in the Dilute Limit?

The dilute limit, yielding classicality, is known to be obtained in the “small
wavelength limit,” through the use of the so-called thermal wavelength λth . The
usual ways of obtaining the λth condition can be criticized for conflating classical
and quantum quantities. For example, one typical method for deriving λth is by
treating it as a kind of quantum-mechanical position uncertainty Δx, corresponding
to the root-mean-square uncertainty ΔpRMS of the momentum of the component
degrees of freedom (at a given temperature T). That is, one starts with the expression:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ΔpRMS
¼ hp2 i hpi2 (11.26)
The average momentum hpi appearing in Eq. (11.26) is assumed to be zero

because of “random motion” thus, it is an average over N independent degrees
of freedom, not an expectation value for any quantum state.ΔpRMS is then taken as
equal to the square root of the average squared momentum p2 , which is obtained
from the equipartition theorem:
2
p ¼ 3mkT (11.27)
So, from Eq. (11.26) and Eq. (11.27), the quantity ΔpRMS is taken to be:
pffiffiffiffiffiffiffiffiffiffiffi
ΔpRMS ¼ 3mkT (11.28)
and this (despite the fact that it is not a real momentum uncertainty but rather a
root-mean-square error) is plugged into the uncertainty relation to obtain a corres-
ponding thermal wavelength λth :
h
λth ¼ pffiffiffiffiffiffiffiffiffiffiffi (11.29)
3mkT
Clearly, in this context, λth is assumed to be a kind of position uncertainty. It is then
demanded that this be much smaller than the average interparticle spacing d, where:
1=3
V
d¼ (11.30)
N
216 Ruth Kastner
Figure 11.4 One way of picturing the thermal wavelength condition for the
classical limit of a quantum distribution.
the idea being that this condition makes the gas ‘dilute’ (i.e., no particles ever
occupying same position x; and many positions unoccupied, see Figure 11.4.) So
the thermal wavelength condition for classicality becomes:
1=3
h V
pffiffiffiffiffiffiffiffiffiffiffi (11.31)
3mkT N
However, as alluded to earlier, the preceding derivation applies a classical, not

quantum, uncertainty to momentum in two distinct ways:
• Averaging over N systems assumed to be in random motion to get a vanishing

value for hpi
• Using the classical Equipartition Theorem to get a value for p2
The resulting momentum uncertainty is only an epistemic average over many
different phase space points, not an intrinsic quantum uncertainty arising from a
single state having an intrinsic “coarse graining,” represented by a finite-sized
element of phase space. Therefore, it is arguably not the correct quantity for
applicability of the uncertainty relation between position and momentum. More-
over, the derivation of the Equipartition Theorem, which is used to obtain that
root-mean-square “momentum uncertainty,” presupposes classical Maxwell-
Boltzmann statistics!
In addition, the derivation conflates a wavelength with a position uncertainty,
which is problematic: If a system has a well-defined wavelength λ, then it has
infinite position uncertainty. Thus, λth is really being interpreted as the spread of a
“wave packet,” despite the fact that the states available to the systems making up
the gas are not necessarily described by wave packets (i.e., in the quantum limit,
they are plane waves).
Of course, it is well known that classical behavior emerges in the small-
wavelength limit, so rather than try to pretend that a wavelength is a position
uncertainty, one can simply work with the de Broglie wavelength of the average
momentum (still using the Equipartition Theorem) to obtain the same condition
Eq. (11.31). But in this case, one cannot explain the classical behavior in this limit
by saying that the gas is dilute, as pictured in Figure 11.4, because it retains a
nonlocal character arising from the presumed exact wavelengths of its degrees of
freedom. So the question is: Why does condition Eq. (11.31) seem to work so well
as a criterion for the classical limit?
It turns out that if we reexpress Eq. (11.31) in terms of thermal energy kT, we
find the condition (neglecting numerical constants of order unity):
2=
h2 N 3
kT (11.32)
m V
But the quantity on the right hand side is then recognized as the Fermi energy,
which is the chemical potential μ for fermions at T = 0:
2=
h2 N 3
EF ¼ ¼ μðT ¼ 0Þ (11.33)
m V
And in fact, this condition kT E F is also the well-known condition for the
classical limit of the Fermi-Dirac distribution. So what happens in this limit that
could justify a classical description? Once again, the chemical potential μðT Þ has
much to teach us.
Specifically, μ is crucially related to the Helmholtz free energy F, defined as
F ¼ U TS (11.34)
The chemical potential μ is the change in F when adding a degree of freedom (at a
given T and V), i.e.,

ΔF
μ¼ (11.35)
ΔN T , V
Now, μ is of large magnitude and negative in the classical limit of large T, small
N/V, and small λth , as can be seen from the well-known relation (see, e.g., Kelly
2002 for details):
!
V
μ ! kT ln (11.36)
clas Nλ3th
218 Ruth Kastner
If we can understand the physical significance of the negativity (and large

magnitude) of μ in the classical limit as expressed by Eq. (11.36), we may hope
to gain insight into the ontology of the quantum/classical transition. A large
negative value of μ ¼ ΔF=ΔN means that F decreases significantly when a degree
of freedom is added to the system. Since F ¼ U TS, increasing the entropy S is
the only way to decrease F. This indicates that the addition of a degree of freedom
in the classical limit increases the entropy far more than it increases the internal
energy U. The higher the temperature T, the larger this decrease, and the more
negative μ becomes. Thus, a large negative μ corresponds to a large entropy
increase of the whole system whenever a degree of freedom is added, accompanied
by a negligible increase in internal energy. The same basic situation applies to
bosons, although in that case μ can never be positive, and with increasing T, it
becomes more and more negative while more of the higher energy states εs become
populated (and the system becomes more dilute in terms of state occupancy).
We can therefore summarize as follows, guided by the clues provided by the
behavior of the chemical potential. In the classical (dilute) limit, adding a degree of
freedom results in a statistically independent increase in the overall state-space,
increasing the entropy with comparatively small increase in U. In contrast, in the
quantum domain, we have two cases: (i) for fermions, adding a degree of freedom
increases U more than it increases TS; and (ii) for bosons, the entropy term (TS)
is always larger in magnitude than U, but the increase in U is nonnegligible
compared to the increase in the magnitude of TS. The physical origin of the
relatively small increase in TS when adding a degree of freedom in the quantum
limit is the following: The new degree of freedom has to find an energy level
contingent on the preexisting energy level structure, which reduces the availability
of states that would have been available in the classical case. Thus the entropy
increase (which is a measure of the increase in the number of available states) is
much smaller than what obtains in the classical case. Once again, the quantum
degrees of freedom “know about each other” and evidently have some form of
interaction (quantified by the chemical potential), even at T = 0 when there are no
thermal interactions at all. They are not independent and separable.
11.5 Conclusions
By examining the derivations of the quantum and classical distributions, we have
found that separability of the individual probability spaces fails in the quantum
domain. In contrast, in the classical limit, the probability spaces of the component
degrees of freedom are fully separable. In addition, by examining the role of the
chemical potential μ, we find a clear manifestation of the highly nonclassical
constraints that quantum degrees of freedom impose on one another via the so-
called exchange forces corresponding to the need for symmetrization. We also

confirm, via the behavior of μ, that in the classical limit, adding a degree of
freedom gives rise to new energy states for the whole system of N degrees of
freedom, independently of the state occupancies of the preexisting N 1 degrees
of freedom increasing entropy with minimal increase in internal energy.
Thus, classical systems (those obeying Maxwell-Boltzmann statistics) have a
form of separability and independence not applying to quantum systems. This
separability amounts to distinguishability, since one could in-principle apply labels
to the N individual state-spaces making up the collective state-space (as in a
Γ-space representation).
However, it is notable that impenetrability does not come into the picture in any
fundamental way, because we need only consider energy states (not position) in
order to obtain Maxwell-Boltzmann statistics. Nevertheless, what about our intu-
ition (also reflected in the usual derivation of the thermal wavelength criterion as
illustrated in Figure 11.4) that classical objects do not overlap in spacetime and are
fundamentally independent from one another? Einstein addressed this classical
notion of separability in terms of his “being thus” concept:
An essential aspect of this arrangement of things in physics is that they lay claim, at a certain
time, to an existence independent of one another, provided that these objects ‘are situated in
different parts of space.’ Unless one makes this kind of assumption about the independence
of the existence (the ‘being-thus’) of objects which are far apart from one another in space
which stems in the first place from everyday thinking physical thinking in the familiar
sense would not be possible. It is also hard to see any way of formulating and testing the laws
of physics unless one makes a clear distinction of this kind.
(Einstein 1948/1971: 170)
Of course, when he made this statement, Einstein was resisting the quantum
nonlocality and/or nonseparability that was evident in the context of the famous
EPR experiment (Einstein, Podolsky, Rosen 1935). It has since become clear that it
is indeed possible, and necessary, to formulate and test the laws of physics without
relying on this sort of classical picture at all levels.
We can trace the emergence of Einstein’s “being thus” in the classical limit by
noting that the latter obtains for high thermal energies kT Eq. (11.32). What can be
deduced from that depends on one’s interpretation of the quantum formalism. In a
unitary-only account, high thermal energies enable decoherence arguments to
proceed (Joos and Zeh 1985), although that account has been criticized on the
basis of entanglement relativity and circularity (e.g., Fields 2010, Dugić and
Jeknić-Dugić 2012, Kastner 2014, 2016a).
Another interpretive approach is to take the projection postulate of von
Neumann as a real physical process (i.e., a “collapse” interpretation). One
such approach is actually a different theory from quantum mechanics: The
220 Ruth Kastner
Ghirardi-Rimini-Weber (GRW; Ghirardi, Rimini, and Weber 1985) mechanism

requires an ad hoc modification to the Schrödinger evolution. In contrast, a
collapse interpretation that does not change the basic quantum theory is the
(Relativistic) Transactional Interpretation (RTI). The original TI, as proposed in
Cramer (1986), was limited to the nonrelativistic domain and took emitters and
absorbers as primitive. The extension of TI to the relativistic domain (RTI) by the
present author has allowed a quantitative definition of emitters and absorbers from
underlying principles (Kastner 2012: chapter 6, Kastner 2016a), and full refutation
of the consistency challenge raised by Maudlin (1996) (the refutation is presented
in Kastner 2016b). The RTI takes the advanced states as playing a physical role in
measurement by breaking the linearity of the evolution and giving rise to the von
Neumann “measurement transition” (Kastner 2012: chapter 3, 2016a). High ther-
mal energies kT give rise to frequent inelastic scatterings among the degrees of
freedom of the gas and thermal photons. According to RTI, inelastic scatterings
correspond to collapses, which serve to localize the component degrees of freedom
giving them effective separate and distinct spacetime trajectories, conferring
independence, and thus restoring Einstein’s notion of being thus. Another advan-
tage of this approach is to provide a physical grounding for the Second Law of
Thermodynamics at the micro-level (see Kastner 2017).
It should also be noted that, under the RTI model with a real nonunitary
transition defining “measurement,” interference truly does disappear upon
measurement, in contrast to the usual assumption of unitary-only evolution.
This resolves the issue alluded to (for example) in French and Redhead (1988),
who say:
But, of course, ontologically speaking, ‘interference’ is never strictly absent. That, after all,
is what constitutes the ‘problem of measurement’ in QM, so the involvement of every
electron with the state of every other electron in the universe, although negligible for
practical purposes, remains an ontological commitment of QM, under the interpretation
where the particles are treated as individuals.
(French and Redhead 1988: 245)
In contrast, under RTI, the previously discussed form of global interference does
vanish upon the nonunitary transition in which a transaction is actualized. Since
transactions are very frequent in the conditions defining the classical limit, this can
be seen as directly supporting the independence, or being thus, of systems in the
classical limit.
Whichever interpretation one adopts, in the domain of high thermal energies kT,
one gets at least effective determinacy of position over time, and thus a unique
spacetime trajectory for each degree of freedom. Such a trajectory confers the capacity
for a unique label and therefore supports distinguishability of the degree of freedom to
which it corresponds. This does not amount to a haecceitistic label, because it is
conferred based on qualitative features of the degrees of freedom (i.e., their trajector-
ies). Nevertheless, as noted by French and Redhead, one may still regard all systems
(classical and quantum) as haecceitistic under a suitable interpretation of individuality.
This chapter takes no position on that metaphysical issue.
Acknowledgments
The author is grateful for valuable correspondence from Jeffrey Bub and Steven
French.
References
Cramer J. G. (1986). “The transactional interpretation of quantum mechanics,”’ Reviews of
Modern Physics, 58: 647–688.
Dugić, M. and Jeknić-Dugić, J. (2012). “Parallel decoherence in composite quantum
systems,” Pramana, 79: 199–209.
Einstein, A. (1948/1971). “Quantum mechanics and reality,” pp. 168–173 in M. Born
(trans.). The Born-Einstein Letters. London: Walker and Co.
Einstein, A., Podolsky, B., and Rosen, N. (1935). “Can quantum mechanical description of
reality be considered complete?”, Physical Review, 47: 777–780.
Fields, C. (2010). “Quantum Darwinism requires an extra-theoretical assumption of
encoding redundancy,” International Journal of Theoretical Physics, 49: 2523–2527.
French, S. (2015). “Identity and individuality in quantum theory,” in E. N. Zalta (ed.), The
Stanford Encyclopedia of Philosophy (Fall 2015 Edition), https://plato.stanford.edu/
archives/fall2015/entries/qt-idind/
French, S. and Redhead, M. (1988). “Quantum physics and the identity of indiscernibles,”
The British Journal for the Philosophy of Science, 39: 233–246.
Huggett, N. (1999). “Atomic metaphysics,” The Journal of Philosophy, 96: 5–24.
Joos, J. and Zeh, D. H. (1985). “The emergence of classical properties through interaction
with the environment,” Zeitschrift für Physik B, 59: 223–243.
Kastner, R. E. (2012). The Transactional Interpretation of Quantum Mechanics: The
Reality of Possibility. Cambridge: Cambridge University Press.
Kastner, R. E. (2014). “Einselection of pointer observables: The new H-theorem?”, Studies
in History and Philosophy of Modern Physics, 48: 56–58.
Kastner, R. E. (2016a). “The Transactional Interpretation and its evolution into the 21st
century: An overview,” Philosophy Compass, 11: 923–932. Preprint version: https://
arxiv.org/abs/1608.00660.
Kastner, R. E. (2016b). “The Relativistic Transactional Interpretation: Immune to the
Maudlin challenge,” https://arxiv.org/abs/1610.04609
Kastner, R. E. (2017). “On quantum non-unitarity as a basis for the second law of
thermodynamics,” Entropy, 19: 106. Preprint version: https://arxiv.org/abs/
1612.08734.
Kelly, J. (2002). “Semiclassical statistical mechanics,” (lecture notes), www.physics.umd
.edu/courses/Phys603/kelly/Notes/Semiclassical.pdf
Lewis, D. (1986). On the Plurality of Worlds. Oxford: Blackwell.
Maudlin, T. (1996). Quantum Nonlocality and Relativity. Oxford: Blackwell.
12
Individuality and the Account of Nonlocality:
The Case for the Particle Ontology in Quantum Physics
michael esfeld
12.1 The Measurement Problem

Quantum mechanics is a highly successful theory as far as the prediction, confirm-
ation, and application of measurement outcome statistics is concerned. The central
tool for these predictions is the Born Rule, according to which, in brief, the squared
modulus of the wave function of a quantum system indicates the probability to find
a particle at a certain location if a measurement is made. Consequently, the
measurement outcomes show a jψ j2 distribution. Furthermore, all the information
that we obtain in experiments is knowledge about positions, as Bell (2004: 166)
stressed: the measurement outcomes are recorded in macroscopic positions, such
as the positions of dots on a screen, pointer positions, etc. and finally brain
configurations, and they provide information about where the investigated objects
are. This insight holds whatever observables one defines and measures in terms of
operators. Thus, for instance, all the information that the outcome of a spin
measurement of an electron by means of a Stern-Gerlach experiment provides is
information about the wave packet in which the electron is located, etc.
However, an algorithm to calculate measurement outcome statistics is not a
physical theory. Physics is about nature, physis in ancient Greek. Consequently, a
physical theory has to (i) spell out an ontology of what there is in nature according
to the theory, (ii) provide a dynamics for the elements of the ontology and (iii)
deduce measurement outcome statistics from the ontology and dynamics by
treating measurement interactions within the ontology and dynamics; in order to
do so, the ontology and dynamics have to be linked with an appropriate probability
measure. Thus, the question is: What is the law that describes the individual
processes that occur in nature (dynamics) and what are the entities that make up
these individual processes (ontology)?
Consider as an illustration the double-slit experiment in quantum mechanics:
One can do this experiment with individual particles, say prepare a source that
222
Individuality and the Account of Nonlocality 223
emits one particle every morning at 8 a.m. so that one gets an outcome recorded in
the form of exactly one dot on a screen every morning. The question then is what
happens between the emission of the particle from the source and the measurement
record on the screen. Is there a particle that goes through one of the two slits?
A wave that goes through both slits and that contracts afterwards to be recorded as
a dot on a screen? Or something else? The constraint on the ontology and dynamics
that are to answer this question is that they have to account for the characteristic
distribution of the dots on the screen that shows up if one runs this experiment
many times. In other words, the ontology and the dynamics have to explain the
measurement outcome distribution.
It is not possible to infer the law that describes the individual processes that
occur in nature from the rule to calculate the measurement outcome statistics. The
characteristic pattern of the measurement outcome distribution in the double-slit
experiment by no means reveals what happens between the source and the screen.
Trying to make such inferences runs into the famous measurement problem of
quantum physics. The measurement outcomes show a jψ j2 distribution, the law for
the evolution of the wave function ψ is the Schrödinger equation, but the Schrö-
dinger evolution does, in general, not lead to measurement outcomes. More
precisely, the by now standard formulation of the measurement problem is the
one of Maudlin (1995: 7):
1.A The wave-function of a system is complete, i.e., the wave-function specifies

(directly or indirectly) all of the physical properties of a system.
2.A The wave-function always evolves in accord with a linear dynamical equation
(e.g., the Schrödinger equation).
3.A Measurements of, e.g., the spin of an electron always (or at least usually) have
determinate outcomes, i.e., at the end of the measurement the measuring device is
either in a state which indicates spin up (and not down) or spin down (and not up).
Any two of these propositions are consistent with one another, but the conjunction
of all three of them is inconsistent. This can be easily illustrated by Schrödinger’s
cat paradox (Schrödinger 1935: 812): If the entire system is completely described
by the wave function, and if the wave function always evolves according to the
Schrödinger equation, then, due to the linearity of this wave equation, superpos-
itions and entangled states will, in general, be preserved. Consequently, a meas-
urement of the cat will, in general, not have a determinate outcome: At the end of
the measurement, the cat will not be in the state of either being alive or being dead.
Hence, the measurement problem is not just a philosophical problem of the
interpretation of a given formalism. It concerns the very formulation of a consistent
quantum theory. Even if one takes (1.A) and (2.A) to define the core formalism of
224 Michael Esfeld
quantum mechanics and abandons (3.A), one has to put forward a formulation of
quantum physics that establishes a link with at least the appearance of determinate
measurement outcomes. If one retains (3.A), one has to develop a formulation of a
quantum theory that goes beyond a theory in which only a wave function and a
linear dynamical equation for the evolution of the wave function figure. Accord-
ingly, the solution space for the formulation of a consistent quantum theory can be
divided into many-worlds theories, rejecting (3.A); collapse theories, rejecting
(2.A); and additional variable theories, rejecting (1.A).
However, research in the last decade has made clear that we do not face three
equally distinct possibilities to solve the measurement problem, but just two: The
main dividing line is between endorsing (3.A) and rejecting it. If one endorses
(3.A), the consequence is not that one has to abandon either (1.A) or (2.A), but that
one has to amend both (1.A) and (2.A). Determinate measurement outcomes as
described in (3.A) are outcomes occurring in ordinary physical space, that is, in
three-dimensional space or four-dimensional spacetime. Hence, endorsing (3.A)
entails being committed to the existence of a determinate configuration of matter in
physical space that constitutes measurement outcomes (such as a live cat or an
apparatus configuration that indicates spin up, etc.). If one does so, one cannot stop
at amending (2.A). The central issue then is not whether or not a collapse term for
the wave function has to be added to the Schrödinger equation, because even with
the addition of such a term, this equation still is an equation for the evolution of the
wave function, by contrast to an equation for the evolution of a configuration of
matter in physical space. Consequently, over and above the Schrödinger equation –
however amended – a law or rule is called for that establishes an explicit link
between the wave function and the configuration of matter in physical space. By
the same token, (1.A) has to be changed in such a way that reference is made to the
configuration of matter in physical space and not just the quantum state as encoded
in the wave function (see Allori et al. 2008). This fact underlines the point made
earlier: We need both a dynamics, filling in proposition (2.A), and an ontology,
filling in proposition (1.A), that specifies the entities to which the dynamics
refers and whose evolution it describes. The way in which these two interplay
then has to account for the measurement outcomes and their distribution, filling in
proposition (3.A).
This point can be further illustrated by a second formulation of the measurement
problem that Maudlin (1995: 11) provides:
1.B The wave function of a system is complete, i.e., the wave function specifies
(directly or indirectly) all of the physical properties of a system.
2.B The wave function always evolves in accord with a deterministic dynamical
equation (e.g., the Schrödinger equation).
3.B Measurement situations which are described by identical initial wave functions
sometimes have different outcomes, and the probability of each possible outcome is
given (at least approximately) by the Born Rule.
Again, any two of these propositions are consistent with one another, but the
conjunction of all three of them is inconsistent. Again, the issue is what the law is
and what the physical entities are to which the law refers such that, if one takes
(3.B) for granted, configurations of matter in physical space that constitute definite
measurement outcomes are accounted for.
All mathematical formulations of nonrelativistic quantum mechanics work with a
formalism in terms of a definite, finite number of point particles and a wave function
that is attributed to these particles, with the basic law for the evolution of the wave
function being the Schrödinger equation. The wave function is defined on configur-
ation space, thereby taking for granted that the particles have a position in three-
dimensional space: For N particles, the configuration space has 3N dimensions so
that each point of configuration space represents a possible configuration of the N
particles in three-dimensional space. This fact speaks also in quantum physics
against regarding configuration space as the physical space, because its dimension
depends on a definite number of particles admitted in three-dimensional space.
Even if one pursues an ontology of configuration space being the physical space
in quantum physics, one can add further stuff than the wave function to configur-
ation space, such as, e.g., the position of a world-particle in configuration space, or
take the wave function to collapse occasionally in configuration space in order to
solve the measurement problem (see Albert 2015: chapters 6–8). However, the
problem remains how to connect what there is in configuration space and its
evolution with our experience of three-dimensional physical objects, their relative
positions, and motions. That experience is the main reason to retain (3.A and 3.B).
This chapter is situated in the framework that envisages abandoning (3.A and 3.B)
only as a last resort, that is, only in case it turned out that the consequences of the options
that endorse (3.A and 3.B) were even more unpalatable than the ones of rejecting (3.A
and 3.B). It seeks to make a contribution to assessing the options that are available in this
framework, namely, the option that starts from amending the Schrödinger dynamics by
admitting a dynamics of the collapse of the wave function – (not 2.A and 2.B); see next
section – and the option that starts from admitting particles, although the information
about their positions is not contained in the wave function – (not 1.A and 1.B); Section
12.3. The chapter closes with a few remarks on quantum field theory (Section 12.4).
12.2 The Collapse Solution and Its Ontology

Against this background, let us take (3.A and 3.B) for granted. On the one hand,
the fact that all formulations of nonrelativistic quantum mechanics work with a
226 Michael Esfeld
formalism in terms of a definite, finite number of point particles and a wave

function that is attributed to these particles suggests to propose a particle ontology
for quantum physics. Following this suggestion, the basic ontology – that is, the
objects in nature to which the formalism refers – is the same in classical and
quantum mechanics. By contrast, the dynamics that is postulated for these objects
is radically different: There is no wave function in classical mechanics. This view
is supported by the fact that, as mentioned in the previous section, all recorded
measurement outcomes consist in definite positions of macroscopic, discrete
objects that provide information about where microscopic, discrete objects (i.e.,
particles) are located. Furthermore, the measurement instruments are composed of
particles that, hence, are located where these instruments are.
On the other hand, the dynamics as given by the wave function evolving
according to the Schrödinger equation does not describe the evolution of particle
positions: It does not provide for determinate trajectories of individual particles.
Even if one starts with an initial condition of precise information about particle
positions, the Schrödinger equation will, in general, describe these particles as
evolving into a superposition of different trajectories. Consequently, the Schrö-
dinger evolution does not establish an intertemporal identity of these objects: It
fails to distinguish them. Moreover, the Heisenberg uncertainty relations put a limit
on the epistemic accessibility of particle positions: Operators for position and
momentum cannot both be measured with arbitrary accuracy on a quantum system.
These facts motivate going for another quantum ontology than the one of particles.
In brief, if there are no precise particle positions, it makes no sense to maintain a
particle ontology. An object that does not have a precise position is not a particle,
but something else. The foremost candidate for that something else is a wave, since
the Schrödinger equation is a wave equation.
There is a proposal for a quantum ontology that takes the wave equation to
describe a wave in three-dimensional physical space, namely, a continuous matter
density field (see Ghirardi, Grassi, and Benatti 1995). Determinate measurement
outcomes – as well as the formation of discrete objects in general – are accounted
for in terms of a spontaneous contraction of the matter density field at certain
points or regions of space. This spontaneous contraction is represented in terms of
the collapse of the wave function. Consequently, Ghirardi et al. (1995) use a
modified Schrödinger dynamics that breaks the linearity and the determinism of
the Schrödinger equation by including the collapse of the wave function under
certain circumstances – rejection of (2.A and 2.B).
On the quantum dynamics proposed by Ghirardi, Rimini, and Weber (GRW; see
Ghirardi, Rimini, and Weber 1986), the wave function undergoes spontaneous
jumps in configuration space at random times, distributed according to the Poisson
distribution with rate N λ, with N being the particle number and λ being the mean
collapse rate. Between two successive jumps, the wave function Ψt evolves
according to the usual Schrödinger equation. At the time of a jump the
kth component of the wave function Ψt undergoes an instantaneous collapse
according to
1=2
Lxxk Ψt ðx1 ; . . . ; xk ; . . . ; xN Þ
Ψt ðx1 ; . . . ; xk ; . . . ; xN Þ↦ 1= (12.1)
x 2
L
xk Ψt
where the localization operator Lxxk is given as a multiplication operator of the form
1 2
e2σ2 ðxk xÞ
1
Lxxk ≔ (12.2)
ð2πσ2 Þ3=2
and x, the center of the collapse, is a random position distributed according to the

x 1=2 2

probability density pðxÞ ¼ Lxk Ψt
. This modified Schrödinger evolution
captures in a mathematically precise way what the collapse postulate in the
textbooks introduces by a fiat, namely the collapse of the wave function so that
it can represent localized objects in physical space, including in particular meas-
urement outcomes. GRW thereby introduces two additional parameters, the mean
rate λ as well as the width σ of the localization operator. An accepted value of λ is
of the order of 1015s1. This value implies that the spontaneous localization
process for a single particle occurs only at astronomical time scales of the order
of 1015s, while for a macroscopic system of N~1023 particles, the collapse happens
so fast that possible superpositions are resolved long before they would be experi-
mentally observable. Moreover, the value of σ can be regarded as localization
width; an accepted value is of the order of 107m.
A further law then is needed to link this modified Schrödinger equation with a
wave ontology in the guise of an ontology of a continuous matter density field
mt ðxÞ in space:
XN ð
mt ðxÞ ¼ mk dx1 . . . dxN δ3 ðx xk Þ jΨt ðx1 ; . . . ; xN Þj2 (12.3)
k¼1
This field mt ðxÞ is to be understood as the density of matter in three-dimensional

physical space at time t (see Allori et al. 2008: section 3.1). The thus defined theory
of a GRW collapse dynamics describing the evolution of a matter density field in
physical space is known as GRWm.
Consequently, although GRWm is formulated in terms of particle numbers,
there are no particles in the ontology. More generally speaking, there is no plurality
228 Michael Esfeld
of fundamental physical systems. There is just one object in the universe, namely a
matter density field that stretches out throughout space and that has varying
degrees of density at different points of space, with these degrees changing in
time. Hence, there are no individual systems in nature according to this theory so
that the issue of identity and distinguishability of individual quantum systems does
not arise. That notwithstanding, this theory accounts for measurement outcomes
that appear as individual particle outcomes in terms of a spontaneous contraction of
the matter density field at certain locations. Position thus is distinguished in the
form of degrees of matter density at points of space. It is the only fundamental
physical property, as Allori et al. (2014) point out:
Moreover, the matter that we postulate in GRWm and whose density is given by the m
function does not ipso facto have any such properties as mass or charge; it can only assume
various levels of density.
(Allori et al. 2014: 331–332)
This ontology is not committed to a dualism of an absolute space and time on

the one hand and a matter field that fills that space and that develops in time on the
other hand. It can also be conceived as a relationalism about space and time,
namely, a field relationalism (see Rovelli 1997 for a field relationalism in the
context of general relativity theory): There is just the matter density field as an
autonomous entity (substance), with an internal differentiation into various degrees
of density, and change of these degrees. The geometry of a three-dimensional,
Euclidean space then is a means to represent that differentiation, and time is a
means to represent that change. In any case, on this view, matter is a continuous
stuff, known as gunk, and it is a primitive stuff or bare substratum that, moreover,
admits different degrees of density as a primitive matter of fact. There is nothing
that accounts for the difference in degrees of density of matter in different regions
of space as expressed by the m function in the formalism.
Making the collapse postulate of textbook quantum mechanics precise by
amending the Schrödinger equation with the two new parameters λ and σ, indicat-
ing the mean rate of spontaneous collapse and the width of the localization
operator, paradoxically has the consequence that the GRW formalism cannot
reproduce the predictions that textbook quantum mechanics achieves by applying
the Born Rule in all situations. Rather than being a decisive drawback, this,
however, opens up the way for testing collapse theories like GRW against theories
that solve the measurement problem without the collapse postulate (see Curceanu
et al. 2016 for such experiments).
On a more fundamental level, however, it is in dispute whether the ontology of a
continuous matter density field that develops according to the GRW equation is
sufficient to solve the measurement problem. The reason is the so-called problem
of the tails of the wave function. This problem arises from the fact that the GRW
formalism mathematically implements the collapse postulate by multiplying the
wave function with a Gaussian, such that the collapsed wave function, although
being sharply peaked in a small region of configuration space, does not actually
vanish outside that region; it has tails spreading to infinity. On this basis, one can
object that GRWm does not achieve its aim, namely to describe measurement
outcomes in the form of macrophysical objects having a definite position (see e.g.,
Maudlin 2010: 135–138). However, one can also make a case for the view that this
mathematical fact does not prevent GRWm from accounting for definite measure-
ment outcomes in physical space (see e.g., Wallace 2014, Egg and Esfeld 2015:
section 3).
The main drawback of GRWm, arguably, is its account of quantum nonlocality,
which occurs when the wave function collapses all over space. Consider a simple
example, namely the thought experiment of one particle in a box that Einstein
presented at the Solvay conference in 1927 (the following presentation is based on
de Broglie’s version of the thought experiment in de Broglie 1964: 28–29, and on
Norsen 2005). The box is split in two halves that are sent in opposite directions,
say from Brussels to Paris and Tokyo. When the half-box arriving in Tokyo is
opened and found to be empty, there is on all accounts of quantum mechanics that
acknowledge that measurements have outcomes a fact that the particle is in the
half-box in Paris.
On GRWm, the particle is a matter density field that stretches over the whole
box and that is split in two halves of equal density when the box is split, these
matter densities travelling in opposite directions. Upon interaction with a meas-
urement device, one of these matter densities (the one in Tokyo in the example
given earlier) vanishes, while the matter density in the other half-box (the one in
Paris) increases so that the whole matter is concentrated in one of the half-boxes.
One might be tempted to say that some matter travels from Tokyo to Paris;
however, because it is impossible to assign any finite velocity to this travel, the
use of the term ‘travel’ is inappropriate. For lack of a better expression, let us say
that some matter is delocated from Tokyo to Paris (this expression was proposed
by Matthias Egg; see Egg and Esfeld 2014: 193). For even if the spontaneous
localization of the wave function is conceived as a continuous process, as in
Ghirardi, Pearle, and Rimini (1990), the time it takes for the matter density to
disappear in one place and to reappear in another place does not depend on the
distance between the two places. This delocation of matter, which is not a travel
with any finite velocity, is a quite mysterious process that the GRWm ontology
asks us to countenance.
Apart from the matter density ontology, there is another ontology available
for the GRW collapse formalism. This ontology goes back to Bell (2004:
230 Michael Esfeld
chapter 22, originally published 1987): Whenever there is a spontaneous

localization of the wave function in configuration space, that development of
the wave function in configuration space represents an event occurring at a
point in physical space. These point-events are known as flashes; the term
‘flash’ was coined by Tumulka (2006: 826). According to the GRW flash theory
(GRWf), the flashes are all there is in physical space. Macroscopic objects are,
in the terms of Bell (2004: 205), galaxies of such flashes. Consequently, the
temporal development of the wave function in configuration space does not
represent the distribution of matter in physical space. It represents the objective
probabilities for the occurrence of further flashes, given an initial configuration
of flashes. Hence, space is not filled with persisting objects such as particles or
fields. There only is a sparse distribution of single events. These events are
individual and distinguishable, because they are absolutely discernible by their
position in space; but there is no intertemporal identity of anything, because
these events are ephemeral.
GRWf is committed to absolute space and time (or spacetime) as the substance
within which the flashes occur. There can also be times at which there are no
flashes at all. The flashes, again, are bare particulars. There is no informative
answer to the question of what distinguishes an empty spacetime point from a point
at which a flash occurs: it is a primitive matter of fact that there are flashes at some
points of spacetime. The flashes are only characterized by their spacetime location.
They come into existence at some points of spacetime out of nothing and they
disappear into nothing.
The most serious drawback of GRWf is that this theory covers only the
spontaneous appearance and disappearance of flashes, but offers no account of
interactions, given that there are no persisting objects at all. The idea that motivates
the GRW collapse dynamics is that a macroscopic object such as a measurement
device consists of a great number of particles so that the entanglement of the wave
function of the apparatus with the one of the measured quantum objects will be
immediately reduced due to the spontaneous localization of the wave function of
the apparatus. However, even if one supposes that a macroscopic object such as a
measurement apparatus can be conceived as a galaxy of flashes (however see the
reservations of Maudlin 2011: 257–258), there is on GRWf nothing with which the
apparatus could interact: There is no particle that enters it, no matter density, and in
general, no field that gets in touch with it either (even if one conceives the wave
function as a field, it is a field in configuration space and not a field in physical
space). There is only one flash (standing for what is usually supposed to be a
quantum object) in its past light cone, but there is nothing left of that flash with
which the apparatus could interact. In brief, there simply are no objects that could
interact in GRWf.
12.3 The Particle Solution and Its Dynamics

One may regard the flashes as particles that are deprived of their trajectories, so
that there are only disconnected particle localizations left, being represented by the
collapses of the wave function in GRWf. Nonetheless, of course, particles without
trajectories and only disconnected point-like localization events is no longer a
particle ontology. Let us therefore now consider a particle ontology for quantum
mechanics. The quantum theory in that sense is the one going back to de Broglie
(1928) and Bohm (1952). Its dominant contemporary version is known as Boh-
mian mechanics (BM; see Dürr, Goldstein, and Zanghì 2013). BM is based on the
following four axioms:
1. Particle configuration: There always is a configuration of N permanent point

particles in the universe that are characterized only by their positions x1 , . . . , xN
in three-dimensional, physical space at any time t.
2. Guiding equation: A wave function is attributed to the particle configuration,
being the central dynamical parameter for its evolution. The wave function has
the task to determine a velocity field along which the particles move, given their
positions. It accomplishes this task by figuring in the law of motion of the
particles, which is known as the guiding equation:
dxk ℏ rk ψ
¼ Im ðx1 ; . . . ; xN Þ (12.4)
dt mk ψ
This equation yields the evolution of the kth particle at a time t as depending on,
via the wave function, the position of all the other particles at that time.
3. Schrödinger equation: The wave function always evolves according to the
usual Schrödinger equation.
4. Typicality measure: On the basis of the universal wave function Ψ, a typicality
measure can be defined in terms of the jΨj2 -density. Given that typicality
measure, it can then be shown that for almost all initial conditions, the distribu-
tion of particle configurations in an ensemble of subsystems of the universe that
admit of a wave function ψ of their own (known as effective wave function) is a
jψ j2 -distribution. A universe in which this distribution of the particles in
subconfigurations obtains is considered to be in quantum equilibrium.
Assuming that the actual universe is a typical Bohmian universe in that it is in

quantum equilibrium, one can hence deduce the Born Rule for the calculation of
measurement outcome statistics on subsystems of the universe in BM (instead of
simply stipulating that rule). In a nutshell, the axiom of jΨj2 providing a typicality
measure with Ψ being the universal wave function justifies applying the jψ j2 -rule
for the calculation of the probabilities of measurement outcomes on particular
232 Michael Esfeld
systems within the universe, with ψ being the effective wave function of the
particular systems in question (see Dürr et al. 2013: chapter 2).
Axiom (1) defines the ontology of the theory. The universe is a configuration of
point particles that, consequently, always have a precise position relative to one
another: They stand in determinate distance relations to each other. Indeed, BM
does not require the commitment to an absolute space in which the particles are
embedded and an absolute time in which their configuration evolves. The geom-
etry and the time with its metric can be conceived as a mere means to represent the
particle configuration and its change, that is, the change in the relative distances of
the particles (see Esfeld and Deckert 2017: chapter 3.2 for the philosophical
argument, and Vassallo and Ip 2016 for a relationalist formulation of BM).
Consequently, the particles are individuals that are absolutely discernible by their
position in a configuration. They have an identity in time that is provided by the
continuous trajectory that their motion, i.e., the change in their relative positions,
traces out. In the framework of relationalism about space and time, one can employ
the distance to individuate the particles, so that the particles in BM are not bare
particulars: The distance relations make them the objects that they are and account
for their numerical plurality (see Esfeld and Deckert 2017: chapter 2.1).
This ontology contains the core of the solution to the measurement problem that
BM provides: There are always point particles with definite positions, and these
particles compose the macroscopic objects. Hence, Schrödinger’s cat is always
either alive or dead, a radioactive atom is always either decayed or not decayed, an
electron in the double-slit experiment with both slides open always goes either
through the upper or the lower slit, etc. That is to say: There are no superpositions
of anything in nature. Superpositions concern only the wave function and its
dynamics according to the Schrödinger equation – Axiom (3) – but not the matter –
the objects – that exist in the world, although, of course, the superpositions in the
wave function and its dynamics can be relevant for the explanation of trajectories
of the matter in physical space.
Axiom (2) then provides the particle dynamics. As it is evident from the guiding
equation, the evolution of the position of any particle depends, strictly speaking, on
the position of all the other particles in the universe via the wave function. This is
the manner in which BM implements quantum nonlocality: It is correlated particle
motion, with the correlation being established by the wave function and being
independent of the distance of the particles. However, it is only correlated motion.
By contrast to the collapse dynamics with the wave function representing an
ontology of a wave in physical space in the guise of a matter density field
(GRWm), there is never a delocation of matter in physical space. There are always
only particles, moving on continuous trajectories and thus with a finite velocity in
physical space (in the relativistic setting, a velocity that is not greater than the
velocity of light), with their motions being correlated with one another as a
primitive matter of fact. The particle ontology and this dynamics provide a
considerable advantage of the Bohmian account of measurement results over the
collapse one: The particles are always there in space, instead of spontaneously
localizing upon the collapse of the wave function – either from nowhere, as the
flashes do in GRWf, or being delocated all over space, as the matter density does in
GRWm.
The ontology of BM – known as the primitive ontology, that is, the referent of
the formalism – is given by the particles and their relative distances as well as the
change of these distances (i.e., the particle motion). The wave function is defined
by its dynamical role of providing the velocity field along which the particles
move. The wave function is set out on configuration space. It can be conceived as a
wave or a field; but, then, it is a wave or field on configuration space by contrast to
an entity in physical space over and above the particles: The wave function does
not have values at the points of physical space. It is therefore misleading to
consider BM as an ontology of a dualism of particles and a wave and to imagine
the wave function as a pilot wave that guides or pilots the particles in physical
space. These are metaphorical ways of speaking, since the wave function cannot be
or represent a wave in physical space.
The wave function is nomological in the sense that it is introduced and defined
through its dynamical role for the particle motion. Consequent upon its being
nomological in that sense, all the stances in the metaphysics of laws of nature are
applicable to the wave function. In particular, in recent years, a Bohmian
Humeanism has been developed according to which the universal wave function
is fixed by or supervenes on the history of the particle positions, being a variable
that figures in the Humean best system (see Miller 2014, Esfeld 2014, Callender
2015, Bhogal and Perry 2017). This stance is applicable to all the theories that
introduce the wave function through its dynamical role for the evolution of the
configuration of matter in physical space, including GRWm and GRWf (see
Dowker and Herbauts 2005 for a precise physical model based on GRWf).
This Humean stance makes clear that one can be a scientific realist without
subscribing to an ontological commitment to the wave function and without falling
into instrumentalism about the wave function: The role of the wave function is, in the
first place, a dynamical one through its position in the law of motion – Axiom (2) –
with its role for the calculation of measurement outcome statistics deriving from that
nomological role – Axiom (4). But the laws, including the dynamical parameters
that figure in them, can be the axioms of the system that achieves the best balance
between simplicity and information in representing the particle motion, as on
Humeanism, instead of being entities that exist in the physical world over and above
the configuration of matter (see Esfeld and Deckert 2017: chapter 2.3).
234 Michael Esfeld
The particle positions are an additional parameter in BM in the sense that the
wave function and its evolution according to the Schrödinger equation do not
contain the information about the exact particle positions and their evolution. All
there is to the Bohmian particles are their relative positions, that is, the distances
among them. Although the parameter of particle mass figures in the guiding
equation – Axiom (4) – mass cannot be considered as an intrinsic property of the
particles in BM. It is not situated where the particles are, but rather in the
superposed wave packets. The same holds for the charge (see e.g., Brown,
Dewdney, and Horton 1995, Brown, Elby, and Weingard 1996, and references
therein; see also, most recently, Pylkkänen, Hiley, and Pättiniemi 2015 and Esfeld
et al. 2017).
The particle positions are the only additional parameter. Theorems like the one
of Kochen and Specker (1967) prove that it is not possible to take the operators or
observables of quantum mechanics to have definite values independently of
measurement contexts, on pain of violating the predictions of quantum mechanics
for measurement outcome statistics. Nonetheless, these theorems leave the possi-
bility open to admit one additional parameter of the quantum objects that has a
definite value, without the precise information about that value figuring in the wave
function. The natural choice for the additional parameter then is position, since all
measurement outcomes consist in macroscopically recorded positions of discrete
objects. In making this choice, one therefore lays the ground for solving the
measurement problem. By contrast, pursuing a strategy that accords definite values
to different parameters over time – as done in so-called modal interpretations apart
from BM – falls victim to the measurement problem, as has been proven by
Maudlin (1995: 13–14).
It is often maintained that, apart from position, all the other operators or
observables are treated as contextual properties in BM, in the sense that they
acquire a definite value, signifying that they are realized as properties of quantum
systems, only in the context of measurements. But this is wrong-headed. The
operators or observables defined on a Hilbert space are never properties of
anything. Suggesting that measurements somehow bring into existence properties
that are contextual, in the sense that they do not exist independently of measure-
ment situations, is a confused manner of talking. The operators or observables
defined on a Hilbert space are instruments that provide information about how the
quantum objects behave in certain situations, and that finally comes down to
information about how their positions develop. Thus, there is no property of spin
that quantum objects possess, and there is no contextual property of a definite
value of spin in a certain direction that quantum objects acquire in the context of
measurement situations. What these measurements do, essentially, is to provide
information about particle positions. For instance, the measurement result “spin up”
of the measurement of an electron provides the information that the particle is

situated in the upwards wave packet, and not in the downwards one, etc. (see Bell
2004: chapter 4, originally published 1971, and Norsen 2014). A similar remark
applies to the matter density field in GRWm and the flashes in GRWf; again, they
only have position, and the operators or observables provide information about
their position (i.e., the position of flashes or the matter density at points of space).
Consequently, although if one formulates the relationship of the operators or
observables defined on a Hilbert space in first order logic, the result is that this
relationship does not fulfill the laws of classical logic; but there is no reason to
assume that the world does not conform to the laws of classical logic, on pain of
confusing operators with properties of objects in the world. BM in particular shows
why there is no problem with classical logic in quantum mechanics. More precisely
and more generally speaking, when it comes to the ontology of what there is in the
world and the dynamics of these entities, one solves the measurement problem by
sticking to classical logic, and one would remain trapped by this problem if one
were to abandon classical logic in the formulation of an ontology of quantum
physics.
The upshot of these considerations is this one: In classical physics, one can be
liberal about the properties of objects. That is to say, one can take the parameters
that figure in the dynamical equations of classical physical theories to designate
properties of the objects that these theories admit, such as the particles. Quantum
physics teaches us that such a liberal ontological attitude is misplaced: The
parameters that one employs to describe the dynamics of the physical objects –
including, in particular, their behavior in measurement situations – cannot without
further reflection be attributed as properties to the objects. This holds even for the
classical parameters of mass and charge, which are, in quantum physics, situated at
the level of the wave function and, hence, cannot be considered as intrinsic
properties of the objects, although their value remains constant.
Quantum physics thereby teaches us that it is mandatory to draw the following
distinction: On the one hand, there is the basic or primitive ontology of the theory,
namely the referents of the formalism, which are supposed to be simply there in
nature, such as particles, which are then only characterized by what is necessary
and minimally sufficient for something to be a particle, namely relative positions
(the same holds for the flashes in GRWf and the positions of matter density values
in GRWm). On the other hand, there is the dynamical structure of the theory. All
and only the parameters that are introduced in terms of the functional or causal role
that they exercise for the evolution of the elements of the basic or primitive
ontology belong to the dynamical structure. They are nomological in the sense
that they are there to perform a certain role for the evolution of the referents of the
theory through their place in the laws of that evolution (but, of course, they are not
236 Michael Esfeld
themselves laws – not even the universal wave function in BM is a law, because the
theory admits of models with different universal wave functions). It then depends on
the stance that one takes with respect to laws whether or not one accords the
dynamical structure a place in the ontology over and above the basic or primitive
ontology or takes it merely to be a means of representation (as on Humeanism).
Not only the operators in quantum mechanics, but also the classical parameters
of mass and charge belong to the dynamical structure. Mass and charge are also
introduced in classical mechanics through their functional role for the particle
motion (see Mach 1919: 241 on mass in Newtonian mechanics). Consequently,
also in classical mechanics, the distinction is available between, on the one hand,
the primitive ontology of the theory in the guise of particle positions and particle
motion and, on the other hand, the dynamical structure in terms of mass and
charge, forces and fields, and energy and potentials introduced through their causal
role for particle motion. Thus, also in classical mechanics, one can suspend or
refuse an ontological commitment to the dynamical structure, being committed
only to the particle positions and their change (see Hall 2009: section 5.2 and
Esfeld and Deckert 2017: chapter 2.3). In short, quantum physics simply makes
evident that it is mandatory to draw a distinction that was already there from the
very beginning.
Although the particle positions make up the ontology in BM, there are limits
to their accessibility. These limits are given in Axiom (4), which implies that
BM cannot make more precise predictions about measurement outcomes on
subsystems of the universe than those generated by using the Born Rule. The
link between the dynamical laws of BM and the jΨj2 -density on the level of the
universal wave function as typicality measure is at least as tight as the link
between the dynamical laws of classical mechanics in the Hamiltonian formu-
lation and the Lebesgue measure (see Goldstein and Struyve 2007). Indeed, in
BM, the quantum probabilities have the same status as the probabilities in
classical statistical mechanics: They enter into the theory as the answer to the
question of what evolution of a given system we can typically expect in
situations in which the evolution of the system is highly sensitive to slight
variations of its initial conditions, and we do not know the exact initial condi-
tions. In such situations, the deterministic laws of motion cannot be employed
to generate deterministic predictions. Nonetheless, the probabilities are object-
ive: They capture patterns in the evolution of the objects in the universe that
show up when one considers many situations of the same type, such as many
coin tosses in classical physics or running the double-slit experiment with many
particles in quantum physics. To put it in a nutshell, the Bohmian universe is
like a classical universe in which not the motion of the planets, but the coin toss
is the standard situation.
However, one may wonder whether, if one lived in a classical universe of coin
tosses, one would endorse the Hamiltonian laws as providing the dynamics of that
universe; Hamiltonian mechanics could then well be a minority position like BM
today. The reason for endorsing Hamiltonian mechanics in that classical case
would be the same as the one for endorsing BM in the quantum case: We need
not only statistical predictions, but also a dynamics that describes the individual
processes occurring in nature, on pain of falling into a measurement problem in the
guise of the inability to account for the occurrence of determinate measurement
outcomes. That is to say: In a classical universe of coin tosses, the Hamiltonian
laws would be useless for predictions, as the calculation of Bohmian particle
trajectories is useless for predictions. But both are indispensable for physics as a
theory of nature. This is not only philosophical ontology, it is the business of
physics to provide dynamical laws that apply to the individual processes in nature.
Nonetheless, in contrast to the relationship between classical statistical
mechanics and classical mechanics, there is a principled limit to the accessibility
of initial conditions of physical systems in the quantum case. That limit becomes
evident, for instance, in the famous Heisenberg uncertainty relations. It is trivial
that measurement is an interaction, so that the measurement changes the measured
system, and can thus not simply reveal the position and velocity that it had
independently of the measurement interaction. Some limit to the accessibility of
physical systems may follow from the triviality that any measurement is an
interaction. Thus, it is well known in classical physics that no observer within
the universe could obtain the data that Laplace’s demon would need for its
predictions. However, the quantum case is not simply an illustration of that
triviality, since there is a precise principled limit of the epistemic accessibility of
quantum systems (as illustrated, for instance, by the Heisenberg uncertainty
relations). In BM, this principled limit follows from applying the theory as defined
earlier to measurement interactions (see Dürr et al. 2013: chapter 2).
In the last resort, of course, it is the particle motion in the world that makes
possible stable particle correlations such that one particle configuration (say a
measurement device or a brain) records the position and traces the motion of other
particles and particle configurations, and it is also the particle motion in the world
that puts a limit on such correlations. The laws of BM, including the typicality
measure and the assumption that the actual universe is a typical Bohmian universe,
bring out these facts about the actual particle motion. Again, the wave function in
configuration space represents that particle motion; it is not the wave function that
puts the limit on the epistemic accessibility of the particle positions, although we
understand that limit by representing the particle motion through the wave func-
tion. Instead of taking this limit to be a drawback and hoping for a physical theory
like classical mechanics in which there are paradigmatic cases of deterministic
238 Michael Esfeld
laws of motion enabling deterministic predictions (e.g., the motion of the planets),
it is fortunate and by no means trivial that there are such stable particle correlations
in the universe at all so that we can represent actual particle positions and motions
and make reliable predictions.
As these considerations make clear, posing a limit to the accessibility of the
objects in physical space is by no means a feature that is peculiar to BM. Such a
limit applies not only to the particles in BM, but, for instance, also to the flash
distribution in GRWf and the matter density field in GRWm (see Cowan and
Tumulka 2016). In any case of a quantum ontology of objects in physical space, if
these objects were fully accessible, we could employ the thus gained information
to exploit quantum nonlocality for superluminal signaling. This limited accessibil-
ity of the particle configuration, the flash distribution or the matter density field
confirms that if one endorses proposition (3.A and 3.B) of the measurement
problem, i.e., determinate measurement outcomes whose statistical distributions
are given by the Born Rule, the situation is not that one has to reject either
proposition (1.A and 1.B) or proposition (2.A and 2.B). One has in this case to
modify both proposition (1.A and 1.B) and proposition (2.A and 2.B). If one starts
from admitting particle positions that are not revealed by the wave function –
rejection of proposition (1. A and 1.B) – one can retain the Schrödinger dynamics
for the wave function – proposition (2.A and 2.B) – but then this is not the
complete dynamics: The central dynamical law is the law of the evolution of the
additional variables, namely, the guiding equation that tells us how the particle
positions evolve in physical space. If one starts from amending the Schrödinger
equation by collapse parameters – rejection of proposition (2.A and 2.B) – one can
retain the wave function and its dynamics as describing the evolution of the objects
in physical space – flashes, matter density field; proposition (1.A and 1.B) – but
that distribution then nevertheless is “hidden” in the sense that it follows from the
theory that it is not fully accessible.
At least three conclusions can be drawn from this situation:
1. Any theory that admits definite measurement outcomes distinguishes position –
be it particle positions, positions of flashes, or values of matter density at points
of space. All the other observables are accounted for on this basis.
2. If one is not prepared to accept a principled limit to the epistemic accessibility
of the objects in physical space, one remains trapped by the measurement
problem, because one then does not have a dynamics at one’s disposal that
accounts for determinate measurement outcomes.
3. The solution space to the measurement problem reduces to this one: Either one
abandons determinate measurement outcomes in physical space, in which case
one can retain the propositions of the wave function being complete and its
evolving always according to the Schrödinger equation; one then has to come
up with an Everett-style account of why it appears to us as if there were
determinate measurement outcomes in physical space. Or, one retains determin-
ate measurement outcomes in physical space, and the account of these meas-
urement outcomes then commits one to endorsing a distribution of objects in
physical space whose evolution cannot be given by the Schrödinger equation
and is not fully accessible.
BM can be seen as the answer to the following question: What is the minimal
deviation from classical mechanics that is necessary in order to obtain quantum
mechanics? BM shows that the physical ontology can remain the same – point
particles moving on continuous trajectories – and that the status of probabilities can
remain unchanged. What has to change is the dynamics, that is, a wave function
parameter has to be introduced with the wave function binding the evolution of the
particle positions together independently of their distance in physical space.
That notwithstanding, this conceptualization of quantum nonlocality is like
Newtonian gravitation, in that there is never any matter instantaneously delocated
in space. As in Newtonian gravitation, the distribution of the particle positions,
velocities, and masses all over space at any time t fixes the acceleration of the
particles at that t, so in BM the distribution of the particle positions and the
universal wave function at any time t fix the velocity of the particles at that t. Of
course, Newtonian gravitation concerns all particles indiscriminately and depends
on the square of their distance, whereas quantum nonlocality is de facto highly
selective, i.e., concerns de facto only specific particles, and is independent of their
distance. Nonetheless, in both cases, nonlocality means that there are correlations
in the particle motion without these correlations being mediated by the instantan-
eous transport of anything all over space. There is no reason to change more.
Doing so only leads to unpalatable consequences beyond the quantum nonlocality
with which one has to come to terms anyway.
12.4 Permanent Particles in Quantum Field Theory

The same conclusions apply to quantum field theory (QFT). Let us briefly point out
why (for details see Esfeld and Deckert 2017: chapter 4.2). The measurement
problem hits QFT in the same way as quantum mechanics (see Barrett 2014).
Again, we have a highly successful formalism to calculate measurement outcome
statistics at our disposal. However, the measurement problem as formulated by
Maudlin (1995) arises as soon as it comes to accommodating measurement
outcomes in physical space, for this formalism does as such not include a dynamics
that describes the individual processes in nature that lead to determinate
240 Michael Esfeld
measurement outcomes. In particular, only the formalism to calculate measurement

outcome statistics is Lorentz-invariant (i.e., it is irrelevant for it in which temporal
order space-like separated measurement outcomes occur). But we do not have a
relativistic, Lorentz-invariant dynamics of the individual processes at our disposal
that lead to determinate measurement outcomes in physical space (despite what
may look like claims to the contrary in the context of GRWf and GRWm; see
Tumulka 2006 and Bedingham et al. 2014 for these claims; see Barrett 2014 and
Esfeld and Gisin 2014 for pointing out their limits).
Despite its name, QFT is not an ontology of fields. The fields in the formalism
are operator valued fields, by contrast to fields that have determinate values at the
points of physical space. In the standard model of particle physics, fields are there
to model the interactions (i.e., the electromagnetic, the weak, and the strong
interaction, without gravitation). If one endorses an ontology of fields in physical
space, then the problem is, like in GRWm for quantum mechanics, to formulate a
credible dynamics of the contraction of fields so that they can constitute determin-
ate measurement outcomes and, in general, discrete macroscopic objects.
However, a particle ontology also faces obstacles in QFT, namely, new obs-
tacles that do not arise in quantum mechanics: In QFT, measurement records not
only fail to keep track of particle trajectories but, moreover, they fail to keep track
of a fixed number of particles. Also in QFT, as in any other area of physics, the
experimental evidence recorded by the measurement devices is particle evidence.
However this evidence includes what appears to be particle creation and annihila-
tion events, so that there seems to be no fixed number of particles that persist.
Nonetheless, these experimental facts as such do not entitle any inferences for
ontology. More precisely, as it is a non sequitur to take particle trajectories to be
ruled out in quantum mechanics due to the Heisenberg uncertainty relations, so it is
a non sequitur to take permanent particles moving on definite trajectories according
to a deterministic law to be ruled out in QFT due to the statistics of particle creation
and annihilation phenomena. In both cases, the experimental evidence leaves open
whether the particle trajectories do not exist or are simply not accessible in
measurements, and whether the particles come into being and are annihilated, or
it is simply not possible to keep track of them in the experiments. These issues
have to be settled by the theory. The standard for assessing the theory is the
solution to the measurement problem.
It is possible to pursue a Bohmian solution to the measurement problem in QFT
along the same lines as in quantum mechanics. As BM has no ambition to improve
on the statistical predictions of measurement outcomes in quantum mechanics, but
deduces these predictions from the axiom that the universe is in quantum equilib-
rium, so there is no ambition that a Bohmian solution to the measurement problem
in QFT can resolve the mathematical difficulties that QFT currently faces. That is
to say: Bohmian QFT has to rely on cut-offs, as does the standard model of particle
physics when it comes to dynamical laws of interactions (by contrast to scattering
theory). Given appropriate cut-offs, one can formulate a Bohmian theory for QFT
in the same way as for quantum mechanics: On what is known as Bohmian Dirac
sea QFT, the ontology is one of a very large, but finite and fixed number of
permanent point particles that move on continuous trajectories as given by a
deterministic dynamical law (guiding equation) by means of the universal wave
function.
More precisely, one can define a ground state for these particles that is a state of
equilibrium. This state is one of a homogeneous particle motion, in the sense that
the particle interactions cancel each other out. Consequently, the particle motion is
not accessible. This state corresponds to what is known as the vacuum state.
However, on this view, it is not at all a vacuum, but a sea full of particles (known
as the Dirac sea) in which the particles are not accessible. What is accessible, and
what is effectively modeled by the Fock space formalism of calculating measure-
ment outcome statistics, are the excitations of this ground state that show up in
what appears to be particle creation and annihilation events. Again, by defining a
typicality measure on the level of the universal wave function, one can derive the
predictions of measurement outcome statistics in the guise of, in this case, statistics
of excitation events from the ground state. Thus, again, the quantum probabilities
are due to a – principled – limit to the accessibility of the particle motion (see Colin
and Struyve 2007, Esfeld and Deckert 2017: chapter 4.2).
When pursuing a solution to the measurement problem in terms of an ontology
of particles, it is worthwhile to go down all the Bohmian way also in QFT. What is
known as Bell-type Bohmian QFT (Bell 2004: chapter 19, and further elaborated
on in Dürr et al. 2005) goes only half the way down: On this theory, the particles
come into and go out of existence with statistical jumps between sectors of
different particle numbers in the dynamics. However, this proposal amounts to
elevating what is known as quasi-particles that are dependent on the contingent
choice of a reference frame to the status of particles in the ontology. Furthermore, it
is committed to absolute space as the substance in which the particles come into
and go out of existence. The Bell-type (quasi-) particles are much like the GRWf
flashes, apart from the fact that they can persist for a limited time, instead of being
ephemeral. By contrast, on the Bohmian Dirac sea theory, the particles are
permanent so that they can be conceived of as being individuated by the distances
among them in any given configuration and as having an identity in change
through the continuous trajectories that their motion trace out. Consequently, there
is no need for a commitment to a surplus structure in the guise of absolute space
and time in the ontology; there is no need for a medium in which the particles exist
(viz. come into and go out of existence). Probabilities then come in through linking
242 Michael Esfeld
the deterministic dynamics with a typicality measure. Filling negative energy states
with particles is no problem on this ontology, because the only property of the
particles is their position; energy is not a property of anything, but a variable in the
formalism to track particle motion.
In sum, on the Bohmian Dirac sea ontology, the account of measurement
outcomes is of the same type as in Bohmian quantum mechanics, with the
difference that there are many more particles in the sea than one would expect in
an ontology of particle positions that are only given by the distances among the
particles (corresponding to empty space in the representation in terms of a space in
which the particles are embedded). The account again has two stages: The ontol-
ogy of particles – in this case, the excitations of particles against the background of
the particle motion in the Dirac sea – accounts for the presence of the measured
quantum objects as well as the one of the macroscopic systems; the latter are
constituted by these particle excitations, with the particle dynamics that yields
these excitations explaining their stability. The measurement outcome statistics
then are accounted for in terms of the limited accessibility of the quantum particles
by means of defining a typicality measure from which one then deduces the
formalism to calculate these statistics.
In any case, the objects that one poses in an ontology of quantum physics are
theoretical entities. They are admitted to explain the phenomena as given by the
measurement outcome statistics. That is why the solution to the measurement
problem is the standard for assessing these proposals. In any case, if one admits
quantum objects in physical space beyond the wave function, there is a limit to
their accessibility; the wave function has, in this case, an exclusively dynamical
status, namely, yielding the dynamics for these objects. The Bohmian solution to
the measurement problem provides the least deviation from the ontology of
classical mechanics that is necessary to accommodate quantum physics, both in
the case of quantum mechanics and in the case of quantum field theory. There is no
cogent reason to go beyond that minimum.
References
Albert, D. Z. (2015). After Physics. Cambridge, MA: Harvard University Press.
Allori, V., Goldstein, S., Tumulka, R., and Zanghì, N. (2014). “Predictions and primitive
ontology in quantum foundations: A study of examples,” The British Journal for the
Barrett, J. A. (2014). “Entanglement and disentanglement in relativistic quantum mechan-
ics,” Studies in History and Philosophy of Modern Physics, 48: 168–174.
Bedingham, D., Dürr, D., Ghirardi, G. C., Goldstein, S., Tumulka, R., and Zanghì,
N. (2014). “Matter density and relativistic models of wave function collapse,” Journal
Bell, J. S. (2004). Speakable and Unspeakable in Quantum Mechanics, 2nd edition.
Bhogal, H. and Perry, Z. R. (2017). “What the Humean should say about entanglement,”
Noûs, 51: 74–94.
variables,” Physical Review, 85: 166–179, 180–193.
Brown, H. R., Dewdney, C., and Horton, G. (1995). “Bohm particles and their detection in
the light of neutron interferometry,” Foundations of Physics, 25: 329–347.
Brown, H. R., Elby, A., and Weingard, R. (1996). “Cause and effect in the pilot-wave
interpretation of quantum mechanics,” pp. 309–319 in J. T. Cushing, A. Fine, and
S. Goldstein (eds.), Boston Studies in the Philosophy of Science: Bohmian Mechanics
and Quantum Theory: An Appraisal, Vol. 184. Dordrecht: Springer.
Callender, C. (2015). “One world, one beable,” Synthese, 192: 3153–3177.
Colin, S. and Struyve, W. (2007). “A Dirac sea pilot-wave model for quantum field
theory,” Journal of Physics A, 40: 7309–7341.
Cowan, C. W. and Tumulka, R. (2016). “Epistemology of wave function collapse in
quantum physics,” The British Journal for the Philosophy of Science, 67: 405–434.
Curceanu, C., Bartalucci, S., Bassi, A., Bazzi, M., Bertolucci, S., Berucci, C., . . . Zmeskal,
J. (2016). “Spontaneously emitted x-rays: An experimental signature of the dynamical
reduction models,” Foundations of Physics, 46: 263–268.
de Broglie, L. (1928). “La nouvelle dynamique des quanta,” pp. 105–132 in Electrons et
photons. Rapports et discussions du cinquième Conseil de Physique tenu à Bruxelles
du 24 au 29 octobre 1927 sous les auspices de l’Institut International de Physique
Solvay. Paris: Gauthier-Villars. English translation: (2009), pp. 341–371 in
G. Bacciagaluppi and A. Valentini (eds.), Quantum Theory at the Crossroads.
Reconsidering the 1927 Solvay Conference. Cambridge: Cambridge University Press.
de Broglie, L. (1964). The Current Interpretation of Wave Mechanics. A Critical Study.
Amsterdam: Elsevier.
Dowker, F. and Herbauts, I. (2005). “The status of the wave function in dynamical collapse
models,” Foundations of Physics Letters, 18: 499–518.
Dürr, D., Goldstein, S., Tumulka, R., and Zanghì, N. (2005). “Bell-type quantum field
theories,” Journal of Physics A: Mathematical and General, 38: R1–R43.
Dürr, D., Goldstein, S., and Zanghì, N. (2013). Quantum Physics without Quantum
Philosophy. Berlin: Springer.
Egg, M. and Esfeld, M. (2014). “Non-local common cause explanations for EPR,”
European Journal for Philosophy of Science, 4: 181–196.
Egg, M. and Esfeld, M. (2015). “Primitive ontology and quantum state in the GRW matter
density theory,” Synthese, 192: 3229–3245.
Esfeld, M. (2014). “Quantum Humeanism, or: Physicalism without properties,” The
Philosophical Quarterly, 64: 453–470.
Esfeld, M. and Deckert, D.-A. (2017). A Minimalist Ontology of the Natural World. New
York: Routledge.
Esfeld, M. and Gisin, N. (2014). “The GRW flash theory: A relativistic quantum ontology
of matter in space-time?”, Philosophy of Science, 81: 248–264.
Esfeld, M., Lazarovici, D., Lam, V., and Hubert, M. (2017). “The physics and
metaphysics of primitive stuff,” The British Journal for the Philosophy of Science,
68: 133–161.
244 Michael Esfeld
Ghirardi, G. C., Grassi, R., and Benatti, F. (1995). “Describing the macroscopic world:
Closing the circle within the dynamical reduction program,” Foundations of Physics,
25: 5–38.
Ghirardi, G. C., Pearle, P., and Rimini, A. (1990). “Markov processes in Hilbert space and
continuous spontaneous localization of systems of identical particles,” Physical
Review A, 42: 78–89.
Goldstein, S. and Struyve, W. (2007). “On the uniqueness of quantum equilibrium in
Bohmian mechanics,” Journal of Statistical Physics, 128: 1197–1209.
Hall, N. (2009). “Humean reductionism about laws of nature,” unpublished manuscript.
http://philpapers.org/rec/halhra.
Mach, E. (1919). The Science of Mechanics: A Critical and Historical Account of Its
Development, 4th edition. T. J. McCormack (trans.). Chicago: Open Court.
Maudlin, T. (1995). “Three measurement problems,” Topoi, 14: 7–15.
Maudlin, T. (2010). “Can the world be only wave-function?” pp. 121–143 in S. Saunders,
J. Barrett, A. Kent, and D. Wallace (eds.), Many Worlds? Everett, Quantum Theory,
and Reality. Oxford: Oxford University Press.
Maudlin, T. (2011). Quantum Non-Locality and Relativity, 3rd edition. Chichester: Wiley-
Blackwell.
Miller, E. (2014). “Quantum entanglement, Bohmian mechanics, and Humean superve-
nience,” Australasian Journal of Philosophy, 92: 567–583.
Norsen, T. (2005). “Einstein’s boxes,” American Journal of Physics, 73: 164–176.
Norsen, T. (2014). “The pilot-wave perspective on spin,” American Journal of Physics, 82:
337–348.
Pylkkänen, P., Hiley, B. J., and Pättiniemi, I. (2015). “Bohm’s approach and individu-
ality,” pp. 226–246 in A. Guay and T. Pradeu (eds.), Individuals Across the Sciences.
Rovelli, C. (1997). “Halfway through the woods: Contemporary research on space and
time,” pp. 180–223 in J. Earman and J. Norton (eds.), The Cosmos of Science.
Pittsburgh: University of Pittsburgh Press.
Schrödinger, E. (1935). “Die gegenwärtige Situation in der Quantenmechanik,” Naturwis-
senschaften, 23: 807–812.
Tumulka, R. (2006). “A relativistic version of the Ghirardi-Rimini-Weber model,” Journal
Vassallo, A. and Ip, P. H. (2016). “On the conceptual issues surrounding the notion of
relational Bohmian dynamics,” Foundations of Physics, 46: 943–972.
Wallace, D. (2014). “Life and death in the tails of the GRW wave function,”
arXiv:1407.4746 [quant-ph].
13
Beyond Loophole-Free Experiments:
A Search for Nonergodicity
alejandro a. hnilo
13.1 Introduction
Quantum mechanics (QM) has been controversial since its very inception. In the
climactic point of a famous debate with Niels Bohr, Albert Einstein argued that the
correlations between measurements performed on distant entangled particles dem-
onstrated that the description of physical reality provided by QM was incomplete
(note the subtle difference: There was no objection on the completeness of the
theory, it was the description of physical reality that was claimed to be incom-
plete). Bohr answered that the idea of a “physical reality” independent of an
observer was meaningless. In my opinion, Einstein’s argument was impeccable
and Bohr’s reply was dangerously close to Saint Bellarmino’s refutation to
Galileo’s observation of mountains in the moon. Yet, quite incomprehensibly to
me, physicists’ public opinion gave the reason to Bohr. Using a football analogy, it
was a beautiful Einstein’s goal disallowed by off-side (the most arcane of foot-
ball’s rules, see Figure 13.1). The history of QM would have been very different if
Bohr had replied in 1935: “Wow Albert, you have a good point. I don’t know. But
let’s use this new theory. There is a lot of exciting work to do. Perhaps, by
applying the theory to new problems, this issue of completeness will become
clearer.”
Bohr’s actual answer or point of view (the so-called Copenhagen interpretation
of QM) opens the door to paradoxes that even have negative social consequences.
I quote E. T. Jaynes’ opinion (Jaynes 1980):
Defenders of the (quantum) theory say that this notion (“real physical situation”) is
philosophically naïve, a throwback to outmoded ways of thinking, and that recognition
of this constitutes deep new wisdom about the nature of human knowledge. I say that it
constitutes a violent irrationality, that somewhere in this theory the distinction between
reality and our knowledge of reality has become lost, and the result has more the character
of medieval necromancy than science.
(Jaynes 1980: 42)
245
246 Alejandro A. Hnilo
Figure 13.1 Einstein’s goal in the “EPR paradox” argument was disallowed by the
referee of public opinion. By the way, Bohr did play as goalkeeper (and a very
good one) in his youth.
In fact, there is an increase of what can be called “quantum mysticism” in social

media and even in politics, supporting a magical conception of the world derived
from the claimed influence of the consciousness in the results of observations. Of
course, many serious scientists who defend with intellectual honesty the Copen-
hagen interpretation dislike this unexpected, and undesired, outcome. And of
course, social consequences have nothing to do with the truth of an idea in physics.
But the ontological debate is not the only controversial feature of QM. Other,
more technical issues are: (i) the failure of the “principle of correspondence” for
chaotic (classical) trajectories, (ii) the collapse of quantum field theory if a (even
infinitesimal) deviation from the principle of superposition is introduced (Gisin
1990), and (iii) the well-known “measurement problem,” i.e., QM is confessedly
unable to describe what happens in a single measurement. Note that these issues
are related, more or less directly, to nonlinear effects.
Before going on, I would like to emphasize the importance I give to the issue
(ii). In my opinion, the essential features of a good theory must be robust against a
small change in any of its parameters. Where it is not, the resulting “structural
instability” of the theory is a clue that some new physics may be lurking there. My
Beyond Loophole-Free Experiments: A Search for Nonergodicity 247
favorite example is extracted from a commentary in Goldstein’s superb textbook

(Goldstein 1950) and refers to Hamilton’s principle: If Hamilton had speculated
that the variation of the Lagrangian time integral was not exactly zero, but a small
quantity h, he would have “discovered” QM a century earlier. Goldstein sensibly
argues that Hamilton had no experimental reason, in the first half of the ninteenth
century, to suppose that variation to take any value but zero. However, Hamilton
might have noted that assuming that value to be infinitesimally nonzero led to a
theory completely different from what he knew. In other words, he would have
found that classical mechanics was structurally unstable at the point (in parameters’
space) h = 0. If he had used, as a general principle, that a satisfactory theory must
be structurally stable, he would have revealed classical mechanics as the limit of a
more general theory (QM) that “embeds” it. Following the same reasoning,
I believe the structural instability of quantum field theory in the point μ = 0 (where
μ is some general nonlinear parameter) to be a clue that new physics are hidden
there. But, as Goldstein wisely remarked, there is no experimental evidence (yet) to
step into that more general, nonlinear theory. Essentially, this chapter deals with
one of the possible ways to find that evidence.
An appealing solution to the previously mentioned controversial features is
interpreting QM as a statistical theory. According to this interpretation, it is not
physical reality that depends on the information available to the observer, but the
predictions of the QM theory. This dependence is an expected feature of any
statistical, probability-based, incomplete-knowledge description. It follows that
there must exist a more complete and still unknown description or theory, from
which QM is able to provide only the statistically averaged results. Hence, QM
would not provide a complete description of physical reality, in agreement with
Einstein’s opinion. The statistical interpretation has been developed in recent times
by Ballentine (1998). Yet, the Copenhagen interpretation of QM is widely
accepted.
There is a sound reason for this acceptance nowadays (although not in 1935).
An important step forward in the debate was the derivation by J. S. Bell of
measurable boundaries (Bell’s inequalities) that any classical theory must obey
and that are violated by some QM predictions (see e.g., Clauser and Shimony
1978). This result moved a purely theoretical discussion into the realm of experi-
mental physics. Bell’s experiments measure the correlations between the results of
observations performed in two remote entangled particles. If a reality independent
of an observer exists and there are no instantaneous interactions at a distance (these
two assumptions together often receive the name of local realism, or LR), then the
correlations cannot be higher than a certain number, which depends on the
particular inequality used. The mere derivation of the inequalities demonstrates
that QM (as we know it, at least) is incompatible with LR. Early experiments
measured a violation of the inequalities, confirming QM predictions and hence

refuting the validity of LR in nature. This is a most relevant result, for LR is
assumed not only in everyday life, but also in all the scientific practice (except if
QM is involved, of course). Those experiments disproved or, at least, reduced
much of the appealingness of the statistical interpretation. For some time, experi-
mental imperfections (generally called logical loopholes) left some room for
hoping that the inequalities were not violated in nature after all. In recent years,
a series of experiments with improved techniques practically closed all the loop-
holes (for a critical review, see Hnilo 2017a). Yet the consequences of abandoning
LR in nature are so serious and deep, that it is sensible to double-check if some
hypothesis, additional to LR, has slipped inadvertently into the reasoning.
In fact, a detailed analysis on how Bell’s inequalities are measured (rather than
derived) shows that there is at least one such additional hypothesis involved, which
I call “ergodicity” (as shorthand) in what follows. In order to not to interrupt the
flow of this introduction, that analysis is reviewed in the next section. The issue is
also discussed in detail in Hnilo (2013, 2014, 2017b); see also Khrennikov (2017).
In few words: It is commonly believed that the following logical relationship
holds:
Locality + Realism ) Bell’s inequalities are valid.
The analysis of how measured numbers are inserted into Bell’s inequalities shows
that the actual logical relationship is:
Locality + Realism + “ergodicity” ) Bell’s inequalities are valid.
That naturally leads to speculating that it is “ergodicity,” and not LR, that has been
disproved by the recent loophole-free experiments. It is convenient to recall here
that the ergodic hypothesis means that the average of the dynamical variables (of
the system being considered) calculated over the phase space, which is called
ensemble average, is equal to the average obtained over the actual evolution of the
system, or time average. The interest of the ergodic hypothesis is that ensemble
averages are far easier to calculate than time averages. The former can be deduced
from the system’s symmetry or from laws of conservation. The latter requires the
complete solution of the equations of motion of the system. But the ergodic
hypothesis is not always valid. The Fermi-Pasta-Ulam system (a series of coupled
nonlinear oscillators) is the best known example of a nonergodic system. The idea
that ergodicity may be involved in the QM vs. LR controversy is not new. It was
indicated long ago by V. Buonomano (1978). It has probably passed mostly
unnoticed because he linked the nonergodic possibility (which is a general argu-
ment) to specific mechanisms, or loopholes, which always have a conspiratorial
flavor. The issue has been updated by Khrennikov (2017 and references therein).
I stress that I use “ergodicity” here as a shorthand for naming a set of hypotheses
that retrieve the validity of the usual form of the Bell’s inequalities. In this set,
some hypotheses are weaker and others stronger than strictly speaking ergodicity.
The origin and meaning of the hypotheses in the set have been discussed in detail
in Hnilo (2013). In that paper, an example of a LR model relevant to the Bell’s
experiment, which is able to violate ergodicity (in its broadest sense), is also
presented. The main conclusion is that nonergodicity (in its broadest sense) is a
condition necessary, but not sufficient, to violate Bell’s inequalities without violat-
ing LR. Be aware that that model was devised as an example of how ergodicity can
be reasonably violated, but that it does not survive (it was not intended to survive)
the performed loophole-free Bell’s experiments.
The pertinent question now is: How to determine whether ergodicity is violated
or not in a Bell’s experiment? Pragmatically speaking, randomness ) ergodicity,
hence nonergodicity ) nonrandomness (I leave aside the precise definition of
“randomness,” which is a difficult issue). In consequence, the study of the devi-
ations from randomness in a time series of measurements in a Bell’s experiment
may reveal that it is ergodicity (instead of LR) that has been refuted in the recent
loophole-free experiments. It is worth noting here that a Bell’s experiment is
equivalent to the “quantum link” of a quantum key distribution (QKD) setup.
Therefore, such deviation from randomness may imply a QKD vulnerability of a
fundamental origin. This possibility has practical consequences.
In summary, the key to test this way to save the validity of LR in nature is the
search for evidence of nonergodic dynamics in Bell’s experiments. This evidence
cannot be found by measuring average magnitudes, as it was done in almost all
Bell’s experiments performed until now, but by analyzing the time evolution of
these magnitudes. This requires time-resolved acquisition of data, a procedure also
known as “stamping” or “tagging” the time values when the entangled particles are
detected. If the analysis of the time series obtained in this way showed the
existence of nonergodic dynamics, QM could be then interpreted as the steady-
state approximation of a more general theory, still unknown. Recall that ergodicity
) steady-state and that the states of QM belong to a Hilbert space, where the
ergodic hypothesis is valid. In any case, such revelation would entail rewriting
the basis of microscopic physics. It would also give the reason to Einstein in the
famous debate with Bohr.
In the next section, the reason why the ergodic hypothesis is necessary to insert
measured data into the Bell’s inequalities is reviewed. A brief analysis of the case
of an experiment using Greenberger-Horne-Zeilinger (GHZ) states is presented. In
Section 13.3, the basic idea of the experiment to detect nonergodic behavior is
presented, together with the discussion of some existing antecedents. In Section
13.4, I indulge myself in some free considerations.
13.2 An Unnoticed Hypothesis

It is convenient to review the derivation of Bell’s inequalities. Here it is done for
the Clauser-Horne one, which includes the Eberhardt’s inequality. The discussion
on the necessity of the ergodic hypothesis for the other experimentally relevant
inequality, Clauser-Horne-Shimony and Holt (CHSH), is similar and can be found
in Hnilo (2017b).
13.2.1 Review of the Derivation of the Clauser-Horne Inequality (CH)

Consider the usual experiment with photons entangled in polarization, sketched in
Figure 13.2. Assume that the probability to detect a photon after an analyzer
A oriented at an angle α is PA(α,λ), where λ is an arbitrary “hidden” variable.
The strength of the Bell’s inequalities is that λ can be anything: real or complex,
a vector, a tensor, etc. The only necessary assumption is that the integrals over
the λ-space exist and are “well behaved.” The observable probability of detection
is then:
ð
PA ðαÞ ¼ dλ:ρðλÞ:PA ðα; λÞ (13.1)
Ðwhere ρ(λ) is a normalized probability distribution in the λ-space (ρ(λ) 0,

dλ.ρ(λ) = 1). These assumptions are in compliance with realism. Consider now
two photons carrying the same value of λ. The probability that both photons are
detected after analyzers A and B set at angles {α,β} is, by definition, PAB(α,β,λ).
Locality implies that PAB(α,β,λ) = PA(α,λ) PB(β,λ), so that the probability to
observe a double detection is:
ð
PAB ðα; βÞ ¼ dλ:ρðλÞ:PA ðα; λÞ:PB ðβ; λÞ: (13.2)
α β
L
A B
(+) (+)
(–) (–)
Figure 13.2 Scheme of a typical Bell’s experiment (actually, Einstein-Podolsky-

Rosen-Bohm setup). The source S emits two photons entangled in polarization
towards two remote stations A and B, where analyzers are oriented at angles α and
β. Detectors after the analyzers count single photons. Relevant measured numbers
are the rate of “singles” (detections at each station) and “coincidences” (detections
simultaneous at both stations). The efficiency is defined as the ratio between
coincidences and singles.
Locality also implies that {α,β,λ} are statistically independent variables: PA(α,λ) =
PA(α)PA(λ), and that ρ(λ) is independent of {α,β}. The set of all these properties
often receive the specific name of measurement independence. To enforce it in the
practice, outstanding experiments have been performed. I point out the ones by
Giustina et al. (2015), Shalm et al. (2015), and Weihs et al. (1998). In these
experiments, the angle setting {α,β} is randomly changed in a time shorter than
L/c. In what follows, measurement independence is taken for granted.
Given {x,y 0, X x’, Y y’} the following equality holds: 1 xy – xy’ +
x’y + x’y’ – Xy – Yx’ 0. Choosing x = PA(α,λ), x’ = PA(α’,λ), y = PB(β,λ), y’ =
PB(β’,λ) and X = Y = 1, where {α,β,α’,β’} are different analyzers’ orientations,
Ð and
after integration over the space of the hidden variables applying dλ.ρ(λ) and
Eq. (13.2), we get:

1 PAB ðα; βÞ PAB α; β’ þ PAB ðα’ ; βÞ þ PAB α’ ; β’ PB ðβÞ PA ðα’ Þ
J0
(13.3)
which is the Clauser-Horne (CH) inequality. The QM predictions violate it. That is,
for the entangled state jφ+i = (1/√2){jxa,xbi + jya,ybi}, PAB(α,β) = ½.cos2(α,β), and
choosing {α,β,α’,β’} = {0, π/8, π/4, 3π/8}, we get: J = 0.427 – 0.073 + 0.427 +
0.427 – ½ – ½ = 0.208, violating the right-hand side (r.h.s.) of the inequality. It is
concluded that QM is incompatible with at least one of the assumptions (i.e.,
locality and/or realism).
13.2.2 The Necessity of a Hypothesis Additional to Local Realism

Note that all real measurements are made in time. The expression for any observ-
able probability is (e.g., for PA):
ð θþΔt
PA ðαÞ ¼ ð1=ΔtÞ dt ρðtÞPA ðα; tÞ (13.4)
θ
This equation represents the result of the following real process: Set A = α during
the time interval [θ, θ + Δt], sum up the number of photons detected after the
analyzer A, and obtain PA(α) as the ratio of detected over incident photons. But the
integrals in Eq. (13.1) and Eq. (13.4) are different: The former is an average over
the possible states of the hidden variables; the latter is an average over time. They
are not necessarily equal. Assuming they are equal means assuming the ergodic
hypothesis valid. In all Bell’s experiments to date, this assumption has been
(implicitly) made to insert measured numbers into derived inequalities. If this
assumption is not made, the insertion of measured data into the derived Bell’s
inequalities implies nothing about the validity of LR in nature; regardless, the

inequalities computed in this way are violated, or not.
The usual logical value of the inequality can be formally retrieved by integrating
over the total measuring time (in the same way than the integral over λ in the
previous subsection). But, the result turns out to include time integrals for setting
angles, say {α,β}, calculated over times when the actual setting angles were
different, say {α’,β}. That is, we have to deal with the results of counterfactual
measurements (see Hnilo 2013, 2014, 2017b for details). In the LR framework,
Bell’s inequalities are derived and counterfactual measurements are assumed to
have definite outcomes (what is called counterfactual definiteness). Therefore, no
hypothesis additional to LR is needed to legitimately deal with counterfactuals.
Yet, a “possible world” must still be defined, in addition to counterfactual definite-
ness, to ensure logical consistency and to assign numerical values to the counter-
factual terms (see d’Espagnat 1984). Depending on the “possible world” chosen,
the values taken by the counterfactual terms are different, and the Bell’s
inequalities have a different form. In other words, assuming counterfactual defin-
iteness is not enough. There still remains the problem of assigning numerical
values to the counterfactual terms.
There is no mystery in this situation, simply lack of information. Let us examine
an example from everyday life: If when I go to the cafeteria I have 30% probability
of finding my friend Alice there, then what is the probability for me to find Alice in
the cafeteria when I don’t go there? If the question is strictly considered, there is no
answer. If it is assumed that Alice and the cafeteria have a well defined existence
even when I do not go there (roughly speaking, if counterfactual definiteness is
assumed), the question does have an answer, say, q. But the only available infor-
mation at this point is that the probability that I find Alice is 30%, so that the
numerical value of q is definite, but unknown. More information is needed about
what happens in the cafeteria when I don’t go there (i.e., a “possible world” must
be defined) to assign a numerical value to q. That is, if an ergodic “possible world”
is defined, then the counterfactual terms take values such that the usual meaning of
Bell’s inequalities is retrieved. Yet this definition means a hypothesis additional to
LR. Several conceivable “possible worlds” are explored in Hnilo (2013), leading to
different inequalities. Some of them are not violated by experiments and not even
by QM.
In summary, a “possible world” must be defined (it may be ergodic or not) to
assign numerical values to the counterfactual terms. That definition unavoidably
means one assumption additional to LR, thus weakening the consequences of the
observed violation of the Bell’s inequalities. Note that this weakening does not
arise from an experimental imperfection, as it is in the case of the loopholes. The
setup is assumed ideally perfect. The weakening arises only from the fact that real
measurements are performed during time and that it is impossible to measure with
two different angle settings simultaneously. It is impossible to travel in time to
measure again, at the same value of time, with a different angle setting.
13.2.3 The Case of Greenberger-Horne-Zeilinger States

Professor Lev Vaidman put forward the interesting question of whether the ergodic
hypothesis is also necessary in the case of tests using GHZ states. Here I show that
the answer is “yes.” Note that this subsection is independent from the rest of the
text, and it can be skipped by the reader with no further consequences.
The question arises from the widespread belief that GHZ states can provide a
“single-shot” disproval of LR, hence making the time integral in Eq. (13.4) irrele-
vant. Therefore, let first see why this belief is erroneous. Consider, e.g., the GHZ
state of three photons:
pffiffiffi
j ϕð3Þ i ¼ 1= 2 fjx1 ; x2 ; x3 i þ ijy1 ; y2 ; y3 ig (13.5)
where x,y are the planes of linear polarization and 1, 2, 3 label the entangled
photons. The Pauli operator σl, which acts: σljxi = jyi, σljyi = jxi, represents
a polarization analyzer that fully transmits a photon linearly polarized at 45o of the
x,y axes. The Pauli operator σr, which acts: σrjxi = ijyi, σrjyi = ijxi, does the
same with circularly polarized photons (rightwards: transmitted, leftwards:
reflected). These operators are applied to each particle of the state and are chosen
in a random and independent way in three remote stations. The configuration of the
whole setup can be lll (the three photons find linear polarization analyzers), llr
(photons 1 and 2 find linear analyzers, the third one a circular one), etc. with equal
probability. There are eight possible configurations. A transmitted photon means a
result +1, a reflected one 1, in each station. The result of a complete measurement
is the product of the results obtained at the three stations. That is, if the photons are
transmitted at the first two stations and reflected at the third one, the result of the
measurement is: (+1) (+1) (1) = 1. The state jϕ(3)i is an eigenstate with
eigenvalue +1 for the configurations rll, lrl, and llr, and with eigenvalue 1 for the
configuration rrr. For the other four possible configurations (the ones having an
even number of r) jϕ(3)i is not an eigenstate, and the result of a measurement may
be 1 or +1 with equal probability. In a long run, they average to zero. Let call the
four elements of the set {rll, lrl, llr, rrr} “words,” and the other four possible
configurations “strings.”
A simple hidden variables theory can be constructed in the form of a set of 2 3
matrices. The matrix determines the result of detection for each photon of the trio,
depending what type of analyzer it finds. For example, the following matrix yields
Table 13.1: Example of a matrix that yields the result (+1) if the configuration is lrl or llr,
and (1) if it is rrr, all in agreement with QM predictions; but, it yields (1) for rll,
contrarily to QM predictions.
Analyzer found Result in station 1 Result in station 2 Result in station 3

l + +
r + +
the result (+1) if the configuration is lrl or llr and (1) if it is rrr, all in agreement
with QM predictions. However, it yields (1) for rll, contrarily to QM prediction
(see Table 13.1).
There are 26 = 64 of these matrices. Each matrix reproduces the QM predictions
for three of the four “words” and for all the “strings” (in the average). Let’s call
“bad word” the configuration for which a given matrix cannot reproduce the QM
prediction (in the example stated earlier, the “bad word” is rll). There are eight
matrices sharing the same “bad word,” and each configuration is the “bad word” of
eight matrices. The probability that a (randomly chosen) configuration is the “bad
word” of the (randomly chosen) matrix carried by the entangled trio is hence:
8/64 = 1/8. This is the probability of the matrix theory to not reproduce the QM
predictions. In other words, the probability of the matrix model to reproduce the
QM predictions for a single trio is 7/8. In order to disprove LR with a reliability
>99%, 35 trios (all of them showing results coincident with the QM predictions)
must be detected in an ideal setup. There are actually 70 trios, because half of the
configurations are “strings,” in a setup where the bases are randomly and independ-
ently chosen in each station. There is, in consequence, no single-shot disproval of
LR with GHZ states, but a statistical one, as in the usual two-particles Bell’s case.
What can be single-shot disproved is QM. It suffices to observe (+1) for rrr, or
(1) for any of the other four “words,” to get a result that refutes QM. Of course,
this is for an ideal setup. In a real setup the imperfections must be taken into
account, and the conditions to discriminate QM from LR become much like in the
two-particles case. The issue is discussed in detail in Hnilo (1994) for GHZ states
with arbitrary numbers of particles. The main conclusion of that paper is, in short,
that experiments with GHZ states do not provide any significant advantage (for the
tests of LR) over the two-particles Bell states, even if the difficulty in the
preparation of the states is not taken into account.
Now that we have seen that a single-shot disproval of LR is not provided by
GHZ states, let go back to the main question. Mermin (1990) has demonstrated that
classical theories are limited by the inequality:
ð

Fn ¼ Im dλ:ρðλÞ:Π El j þ iEr j 2n=2 ðn evenÞ, or 2ðn-1Þ=2 ðn oddÞ: (13.6)
where n is the number of particles in the GHZ state, λ is the hidden variable, the
product goes from j = 1 to n, and Elj (Nl+ Nl)/(Nl+ + Nl), where Nl+ (Nl) is
the number of detections that produced a result “+1” (1) in the j-station when the
setting was l (analogously for r). In the case of jϕ(3)i Mermin’s inequality takes the
form:
ð

F3 ¼ dλ:ρðλÞ: El 1 :El 2 :Er 3 þ El 1 :Er 2 :El 3 þ Er 1 :El 2 :El 3 Er 1 :Er 2 :Er 3 2
(13.7)
Note that the settings are randomly and independently chosen, so that in any
experiment (even in an ideal one) the “string” configurations (say, lrr) also appear
in the set of measured data. However, they average to zero, so that they are
dropped in Eq. (13.7). The QM prediction in the case of jϕ(3)i is F3 = 4, which
violates the inequality. In the case of the matrix model, the integral over λ is
replaced by a sum over the 64 matrices. The matrix model saturates the inequality
(F3 = 2), because all the matrices fail to reproduce at least one of the “words” (e.g.,
“bad word” of each matrix).
The definition of Fn involves an integral over the hidden variables, so that it is
clear that the ergodic hypothesis is necessary in an experiment using GHZ states,
too. To be specific, what is actually measured for n = 3 is:
ð ð ð
Fmeas: ¼ dt:ρðtÞ El :El :Er þ dt:ρðtÞ El :Er :El þ dt:ρðtÞ Er 1 :El 2 :El 3
1 2 3 1 2 3
ð
dt:ρðtÞ Er 1 :Er 2 :Er 3 (13.8)
where each integral spans over a different period of time (regardless of whether
they are continuous or divided in many randomly chosen, separate small intervals).
If we make λ = t and integrate over the whole measuring time, we are faced again
with the problem of assigning values to the counterfactual integrals, as before.
Therefore, F3 6¼ Fmeas., unless the ergodic hypothesis (or something like it) is
assumed. The involved inequality is different (Mermin’s instead of Bell’s), but the
situation is the same, as in the case of the experiment with two entangled particles.
13.3 The Search for Nonergodicity

As it was shown, ergodicity must be assumed to validly insert measured values
into the mathematical expression of Bell’s inequalities. Taking for granted that the
performed loophole-free experiments have demonstrated the violation of Bell’s
inequalities, then locality, realism, or ergodicity, must be invalid in nature. It is my
intention to save LR, so let us see how to demonstrate that it is ergodicity that is
violated in the experiments. The direct way cannot be taken, for it is impossible to
measure an average over the space of the hidden variables (to check whether it is
equal to the time average or not). Therefore, one must follow an indirect approach.
As was discussed in the introduction, the key may be deviations from randomness
in time series of data obtained in a Bell’s experiment. This means measuring not
only the average values of magnitudes (say, J, SCHSH, or concurrence), as in almost
all experiments performed to date, but also the time evolution of these or other
magnitudes.
13.3.1 Experiments to Detect Nonergodicity

In a chaotic system, the dynamical variables are linked through nonlinear equations
in such a way that the evolution is very complex and apparently random. Yet, the
dynamics are described by few degrees of freedom. This is a difference with “true”
random processes, which can be thought of as having a very high (eventually,
infinite) number of degrees of freedom. Nonlinear analysis (Abarbanel 1983)
allows, under favorable circumstances, measuring the number of dimensions of
the phase space where the system evolves (this is called the dimension of embed-
ding dE), and hence, discriminating chaos from true randomness. In the case of
interest here, it may allow detecting the existence of the hypothesized nonergodic
dynamics, as opposed to the fundamental randomness assumed by QM.
However, revealing chaotic dynamics from observations is not an easy task. In
principle, one should see a complex time behavior involving quasi-periodicities
and long transients. In order to reconstruct the (hypothesized) underlying object in
phase space, one must record a time series with a density of data capable of
detecting oscillations at some basic frequency f. The value of f in the case of
interest here is unknown. Nevertheless, assuming that locality holds (recall that the
purpose here is to save LR), it is intuitive to expect f c/L, where L is the physical
distance between the two stations and c is the speed of light (see Figure 13.2). The
capacity of recording oscillations at a frequency as high as c/L is hence the main
condition to reveal nonergodic behavior. No experiment performed until now has
fulfilled it. In order to get an idea of what that condition means, let us use
Shannon’s criterion of two samples in the period of interest. A sample of, e.g.,
the value of the probability of coincidence P++(α,β) with the necessary resolution
requires a minimum of 19 coincidences before the analyzers (0.052 1/19 is the
largest difference between the QM prediction for P++(α,β) and the limit imposed by
Bell’s inequalities for a maximally entangled state). Therefore, at least 38 coinci-
dences detected in a period L/c are needed. Let see now how far performed
experiments are from this figure. The highest reported rate obtained in a laboratory
environment is 3 105 s1 coincidences (see Kurtsiefer, Oberparleiter, and
Weinfurter 2001). This number cannot be much improved nowadays, for the
fastest currently available single-photon detectors (avalanche photodiodes) cannot
be used reliably if the rate approaches 106 s1 because of the high number of
secondary (false) counts. In consequence, to detect 38 pairs with 3105 s1 coinci-
dences in a time L/c, one must have L > 38 km. Bell’s setups with L = 13 km (Peng
et al. 2005) and even 144 km (Scheidl et al. 2010) have been performed, but the
achieved coincidence rate was much lower than necessary: 50 and 8 s–1 (scaled
values are 2 10–3 c/L and 4 10–3 c/L). The recent loophole-free experiment in
Vienna (Giustina et al. 2015) reached 200 s–1 for L = 58 m, or 4 10–5 c/L; the one
in Boulder (Shalm et al. 2015) 5 s–1 for L = 185 m, or 3 10–6 c/L. The rate of
detected pairs should be thus increased by several orders of magnitude to enter the
range where the basic oscillations are expected to be detectable. A direct search for
nonergodic dynamics seems beyond the current technical capacity.
Yet, the task is reachable under a “stroboscopic” approximation. That is, by
supposing that the system decays to a “ground state” in a time unknown, but finite
τdecay after the source of entangled states is turned off. The hypothesized picture is
then as follows: Once the source of entangled states is turned on, the nonclassical
correlation between measurements in the remote stations starts to evolve in a
nonergodic way. After the source is turned off, the correlation decays with time
τdecay (unknown, but finite). We have studied a general form that the dynamics
may take in this picture in Hnilo (2012) and Hnilo and Agüero (2015) if locality is
imposed. Noteworthy, oscillations with period 4L/c are predicted for a broad
region in parameters’ space.
If the hypothesized picture is correct, a stroboscopic reconstruction of the
system’s evolution is possible by using a pulsed source of entangled states. The
time between pulses is adjusted longer than τdecay. The pulse duration is sliced in
periods shorter than L/2c, and the number of photons detected in each time slice is
recorded. Due to technical reasons (see Agüero, Hnilo, and Kovalsky 2014), less
than one photon per pulse must be recorded in the average but, after millions of
pulses are detected, the time slices are gradually “filled” with data and the evolu-
tion of the correlation during the pulse duration can be reconstructed with arbitrary
precision. The value of τdecay is unknown, but the pulse repetition rate can be
lowered as much as necessary, at the only cost of increasing the total duration of
the experimental run. In summary, the stroboscopic approach allows the search for
nonergodic dynamics with accessible means. There is, however, a risk of failure: If
the system did not decay to the same ground state after each pump pulse, then the
initial condition before each new pulse would not always be the same, jamming the
reconstruction. This is an unavoidable risk in any stroboscopic observation.
Anyway, even an imperfect reconstruction of the dynamics may provide a valuable
antecedent to consider the realization (or not) of the “complete,” and much more
Pump laser
(pulsed)
Time stamper A Time stamper B
Trigger C+ C– C– C+ Trigger
Frequency down
conversion crystals
SPCM SPCM
Single-mode fiber Single-mode fiber

SPCM HWP HWP SPCM
“Bat-ears” “Bat-ears”
Photodiode
Photodiode L
Figure 13.3 Sketch of the proposed experiment. SPCM: single-photon counting

modules (avalanche photodiodes). The “bat ears” allow the compensation of
birefringence in the optical fibers. HWP (“half wave plate”) adjust the angle
setting before each analyzer. Samples from the pump laser are used to trigger
(synchronize) the time stampers. The time values of detection of each single
photon and of the trigger signals are saved for further analysis.
difficult, L > 38 km experiment. Note that the “complete” experiment would be, in
any case, a simpler setup than, e.g., a laser interferometer gravitational-wave
observatory (LIGO) or a supercollider.
The specific experiment proposed (Figure 13.3) uses a pulsed laser to pump the
nonlinear crystals to generate entangled (in polarization) states of photons. Both
the pump repetition rate and the pulse length are adjustable. This is to explore the
unknown value of τdecay and of the time of evolution of the dynamics. The
entangled photons are inserted into single-mode optical fibers and transmitted to
remote stations. The time-stamped files allow the calculation of the variables of
interest (say, concurrence, efficiency, etc.) after the experimental run has ended.
The distance between the stations is adjustable. Varying the value of L allows
discarding artifacts in the case some dynamics are actually observed, for L is
supposed to define the timescale of the problem. It is relevant to mention here that
the effect of polarization mode dispersion limits the use of optical fibers to about
L 1 km. Photons are detected at the stations with avalanche photodiodes, and
detections’ time values stored in time-stamped files with a resolution of 1 or 2 ns.
A better resolution is meaningless, because of the intrinsic time jitter of the
photodiodes. A sample of the pump pulse is sent to each station to synchronize
the time-stamping devices, to set a uniform and stable starting point for the
stroboscopic reconstruction, and to avoid the drift of the clocks that apparently
occurred in some previous experiments (see the next section).
Entanglement
Efficiency
Pump pulse
Time
Figure 13.4 Sketch of a possible result of the proposed experiment. Here, it is

supposed that the efficiency increases monotonically (as observed in Hnilo and
Agüero 2015). It is also assumed that the entanglement, which is now measured
with sufficient time resolution, displays a complex behavior. Be aware that these
results would be a stroboscopic reconstruction obtained from the record of several
millions of pump pulses separated, from each other, by a time longer than τdecay.
The basic oscillation would have a frequency f < c/L. The nonlinear analysis of
this time series may reveal the presence of a compact object in phase space, i.e., a
deviation from randomness, and hence, the existence of nonergodic dynamics.
The stroboscopically obtained time series (Figure 13.4) are then analyzed,
looking for the existence of a low-dimension object in phase space.
13.3.2 Some Antecedents

An early search of nonergodic dynamics in a Bell’s experiment was completed by
our group (Hnilo, Peuriot, and Santiago 2002) using the raw data of the Innsbruck
experiment (Weihs et al. 1998). This experiment had implemented time stamping
for reasons different from the test of ergodicity. The focus of our study was to find
time series with a measurable value of dE. The theorem of embedding ensures that
dE can be measured using the time series of any dynamical scalar variable. In
practice, favorable experimental conditions and choosing the appropriate observed
variable are necessary. Our group developed skills and techniques to deal with
these problems for the experimental study of the chaotic dynamics of Kerr-lens
mode locked femtosecond lasers (Kovalsky and Hnilo 2004) and the formation of
optical rogue waves in lasers with a saturable absorber (Bonazzola et al. 2015).
Our search in the data of the Innsbruck experiment involved dozens of files of
raw time-stamped data, generously provided by professor Gregor Weihs, and
several possible observables. The hope to find evidence of nonergodic dynamics
was dim, for the detection rate in the Innsbruck experiment was too low to detect
oscillations at c/L (the scaled coincidence rate was 2 10–3 c/L << 38 c/L). The
result of our search was that no definite value of dE was reliably measured in any of
the data files, except one. The exception was the longest run in real time (6 minutes,
rather than 10 to 30 seconds of most files), and the result was dE = 10 with four
positive Lyapunov exponents, which meant that the series was hyperchaotic. Based
on the reconstructed attractor, we were able to “predict” the future outcomes in the
series, with satisfactory precision, up to an average of five to six pairs, that roughly
corresponded to the inverse of the largest positive Lyapunov exponent. As there
were four possible settings in the series, about 20 bits of the key were predictable.
The importance of this finding for the security of QKD was remarked. When we
found this result, the Innsbruck experiment had been dismantled, so that the cause
of the detected dynamics is impossible to know for sure. It is believed to be a drift
between the clocks in each station. This belief is supported by the fact that a time-
stamped experiment performed by our group some years later (Agüero et al. 2009),
with an even longer session of data recording (>30 min) but a single clock,
produced no measurable value of dE. Regardless whether the cause of the chaotic
dynamics was instrumental or fundamental, the nonlinear analysis approach was
able to reveal it in one file of the Innsbruck experiment. This result proves the
capacity and power of the approach.
Another antecedent is the stroboscopic reconstruction of the evolution of
entanglement achieved by our group (Agüero, Hnilo, and Kovalsky 2012). The
experiment’s aim was not to test ergodicity, but to close the time-coincidence
loophole. No evolution of entanglement was observed during the pulse duration.
Entanglement was constant during the pulse, and was born and dead “instantan-
eously” (strictly speaking, in a time shorter than the time resolution of the device,
12.5 ns). There was only an increase of statistical errors at the pulse’s edges,
naturally caused by scarcer data. Nevertheless, the time resolution of photon
detectors and time-stamping devices was insufficient to detect f for the small value
of L used which was, to make things even worse, fixed. After the setup was
dismantled, we realized there was a linear increase of efficiency with time during
the pulse duration. This result is consistent with the predictions of a simple LR
hidden variables theory, but we consider it in no way conclusive (see Hnilo and
Agüero 2015). A repetition of the experiment is in order to discard possible
artifacts.
Finally, in my opinion, the proposed stroboscopic experiment is worth doing for
the following reasons:
1) It is new. Instead of measuring Bell’s inequalities over and over again with
different random number generators (as in the recent “Big Bell Test,” or by
using light from remote stars) and/or improved detectors, this proposal involves
analyzing the time evolution of the system, not only the average values. I do not
mean those experiments are worthless. On the contrary, they are formidable
technical achievements – useful and meaningful. However, you cannot keep

doing the same thing and expect different results.
2) It is important. I find it difficult to imagine an experiment with consequences
deeper and broader than one whose scope is to reveal one of the limits to the
validity of QM. That the limits exist, that QM is only an approximate, incom-
plete description of physical reality should be evident to everybody. The fate of
all human knowledge is to be incomplete and provisory. Claiming the opposite
is not only methodologically wrong (recall Popper) but, to my taste, of an
unbearable arrogance.
3) It is at hand, and some encouraging antecedents exist. Besides, even if the
experiment were not fully successful (say, because a failure of the stroboscopic
assumption or a value of τdecay too long, such that it is impossible to be handled
in practice), it may still provide clues useful to consider (or to discard) the
“complete” version of the test with L > 38 km.
13.4 Some Personal Views and Conclusions

As famously stated by Schrödinger, “entanglement is the characteristic trait of
QM.” All performed quantum optical experiments can be explained by semiclas-
sical theories, except for the ones involving spatially spread entangled states (see
Scully and Zubairy 1997). The key to describing these experiments in QM is
interference of waves, yet not in real space, but in an abstract space. Now I would
like to share with the reader some imprecise (yet hopefully funny) thoughts.
Interference is, in itself, a formidable phenomenon. Although not often noted, it
is alien to our intuitive way of thinking. It sets us apart from the ancients’
philosophical realm. Parmenides stated that “something” cannot arise from “noth-
ing” nor to vanish into “nothing,” then nihil novum sub sole. The eternity of
Democritus’ atoms follows. But, in the interference phenomenon, two (or more)
“somethings” can become “nothing,” and vice versa. The natural way to describe
interference is by using arrows. Here I use the term “arrow” instead of “vector”
because the latter is a specific mathematical entity. A vector is an element of
(obviously) a vector space, where the principle of superposition holds, i.e., the
linear combination of vectors is also a vector. Arrow, instead, is a more general
entity; e.g., the solutions of nonlinear equations can sum up to zero (they can
interfere) but their linear combination is not necessarily a solution of the equation
(so, they do not form a vector space). Vectors rotating in time are solutions of
linear equations of physics, like Maxwell’s equations, and are conveniently
described with complex numbers. In this way we get a complex Hilbert (vector)
space, the realm of QM, where ergodicity is valid.
Now I would like to remark that arrows and vectors are really strange things. If
one thinks in the familiar terms of sets, the properties common to different sets are
found by their intersection. The same for arrows is found by projecting one arrow
into the other. Orthogonal arrows have “nothing” in common. The intersection of
sets is associative and commutative. The projection of arrows is not. In a popular
children’s game, one has to find a person in a set by asking whether the person is a
man or a woman, if he or she is blond or not, wears glasses or not, etc. The result is
the same regardless of the order in which the questions are posed. If played with
arrows, however, the game may have a different result depending the order of the
questions. Some centuries-old mysteries are impossible to grasp by thinking in
terms of sets; e.g., Christian Trinity – three different persons, but only one God.
Considered in terms of sets, it has no solution other than faith. In terms of arrows,
God can be thought as the sum of three orthogonal (i.e., completely different)
entities: Son, Father, and Holy Spirit. This picture also explains the Jesuits’
commandments to study nature (to get close to the Father), to do charitable actions
(to get close to the Son), and to exercise introspection (to get close to the Holy
Spirit). Failing to complete any of these three commandments means to fall short
of reaching God by a distance 1/√3 (in God’s space, assumed to be Euclidean).
The QM description of Bell’s experiments requires interference of waves lying in
remote positions in real space. It is interference in an abstract, nonlocal space. This is
not only anti-intuitive (interference in real space has already led us far from intuition,
so that this is not too serious) but, much more important in my opinion, structurally
unstable. For nonlocal interference and nonlinearity (even if infinitesimally small)
lead to the possibility of transmission of information at infinite speed and quantum
field theory collapses. Some way to include nonlinear terms without leading to faster-
than-light signaling (or something even stranger, see Polchinski 1991 and references
therein) would be most welcome, for nonlinearities exist everywhere. They exist at
the large scale of the universe; the equations of general relativity are nonlinear. They
also exist at our scale – we can describe the evolution of the surrounding world with
linear equations only as an approximation. Yet, QM claims nonlinearities to do not
exist at the microscopic scale, not even infinitesimal ones. I prefer thinking that
nonlinearities also exist at the microscopic scale, but that they cannot lead to
instantaneous transmission of information, because the strong correlations character-
istic of entanglement appear only after a time L/c has elapsed. This is an effect that is
not predicted by QM. In other words, I believe entanglement to be an average
property of the setup’s symmetries and, for spatially spread systems, it requires a
time longer than L/c to grow. In shorter times, transient deviations from the high
correlations predicted by QM should be observed. I find this the most economical
way to save quantum field theory, the statistical interpretation of QM, and LR.
QM claims that only probabilities can be predicted, that information is physical,
and that an independent physical reality does not exist. This has to me the smell of an
approximation used beyond its range of validity. Consider the alternative: QM is a

statistical approximation, a theory to be used when the knowledge we have on the
system is incomplete. Therefore it is QM theory which can only predict probabilities.
Information does not change physical reality, but affects QM theory predictions (for
a change in the available information changes the predictions of a probabilistic
theory, as it is well known). An independent physical reality does exist; QM is an
incomplete (statistical) description of this physical reality. This is a very good and
useful description, beyond any doubt, but, as Einstein stated, it is an incomplete one.
And, as one of the LIGO’s leaders recently said: Don’t bet against Einstein.
Historically, the laws of evolution of particles (whose properties obey the logic of
sets) were developed before their statistical approximation. Classical statistical mech-
anics describes the collective, average behavior of a very large number of particles
(say, a mesoscopic volume of ideal gas), where Newton’s equations are impossible
(and unnecessary) to solve. Extending its concepts and results to the evolution of two
particles leads to paradoxes. However, everyone knows that they are only apparent
paradoxes. They are the mere consequence of having used the theory outside the
range of validity of the statistical approximation. Historically, we have entered the
realm of arrows through the opposite door (for good, practical reasons, no doubt). We
have got the statistical mechanics of vectors (i.e., QM) before developing the
“mechanics of single arrows” (strictly speaking, the mechanics of entities whose
properties obey, in the statistical approximation, the logic of vectors). This missing
(hopefully LR) theory, where the principle of superposition does not necessarily hold
and from which QM is the statistical, steady-state, ergodic approximation, is the
phantom whose vanishing tracks the proposed experiments are intended to find.
Acknowledgments
I would like to offer many thanks to the participants on the workshop Identity,
indistinguishability and non-locality in quantum physics (Buenos Aires, June
2017), especially to Federico Holik, Nino Zanghì, Olimpia Lombardi, Lev Vaid-
man, and Sebastian Fortin, for so many exciting and fruitful discussions on the
subject of this contribution. This research was partially supported by the grant
PIP11–077 of CONICET, Argentina.
References
Abarbanel, H. (1983) “The analysis of observed chaotic data in physical systems,” Reviews
of Modern Physics, 65: 1331–1392.
Agüero, M., Hnilo, A., Kovalsky, M., and Larotonda, M. (2009). “Time stamping in EPRB
experiments: Application on the test of non-ergodic theories,” European Physical
Journal D, 55: 705–709.
Agüero, M., Hnilo, A., and Kovalsky, M. (2012). “Time resolved measurement of the
Bell’s inequalities and the coincidence-loophole,” Physical Review A, 86: 052121.
Agüero, M., Hnilo, A., and Kovalsky, M. (2014). “Measuring the entanglement of photons
produced by a nanosecond pulsed source,” Journal of the Optical Society of America B,
31: 3088–3096.
Ballentine, L. E. (1998). Quantum Mechanics. A Modern Development. Singapore: World
Scientific Publishing.
Bonazzola, C., Hnilo, A., Kovalsky, M., and Tredicce, J. (2015). “Features of the extreme
events observed in an all-solid-state laser with saturable absorber,” Physical Review A,
92: 053816.
Buonomano, V. (1978). “A limitation on Bell’s inequality,” Annales de l’Institut Henri
Poincaré, 29A: 379–394.
Clauser, J. and Shimony, A. (1978). “Bell’s theorem: Experimental tests and implications,”
Reports on Progress in Physics, 41: 1881–1927.
d’Espagnat, B. (1984). “Nonseparability and the tentative descriptions of reality,” Physics
Reports, 110: 201–264.
Gisin, N. (1990). “Weinberg’s non-linear quantum mechanics and superluminal communi-
cations,” Physics Letters A, 143: 1–2.
Giustina, M., Versteegh, M. A. M., Wengerowsky, S., Handsteiner, J., Hochrainer, A.,
Phelan, K., . . . Zeilinger, A. (2015). “A significant loophole-free test of Bell’s
theorem with entangled photons,” Physical Review Letters, 115: 250401.
Goldstein, H. (1950). Classical Mechanics. Reading, MA: Addison-Wesley.
Hnilo, A. (1994). “On testing objective local theories by using GHZ states,” Foundations
of Physics, 24: 139–162.
Hnilo, A. (2012). “Observable consequences of a hypothetical transient deviation from
Quantum Mechanics,” arXiv/quant-ph/1212.5722.
Hnilo, A. (2013). “Time weakens the Bell’s inequalities,” arXiv/quant-ph/1306.1383v2.
Hnilo, A. (2014). “On the meaning of an additional hypothesis in the Bell’s inequalities,”
arXiv/quant-ph/1402.6177.
Hnilo, A. (2017a). “Consequences of recent loophole-free experiments on a relaxation of
measurement independence,” Physical Review A, 95: 022102.
Hnilo, A. (2017b). “Using measured values in Bell’s inequalities entails at least one
hypothesis additional to Local Realism,” Entropy, 19: 80.
Hnilo, A. and Agüero, M. (2015). “Simple experiment to test a hypothetical transient
deviation from Quantum Mechanics,” arXiv/abs/1507.01766.
Hnilo, A., Peuriot, A., and Santiago, G. (2002). “Local realistic models tested by the EPRB
experiment with random variable analyzers,” Foundations of Physics Letters, 15:
359–371.
Jaynes, E. T. (1980). “Quantum beats,” pp. 37–43 in A. Barut (ed.), Foundations of
Radiation Theory and Quantum Electrodynamics. New York: Plenum Press.
Khrennikov, A. (2017). “Buonomano against Bell: Nonergodicity or nonlocality?”, Inter-
national Journal of Quantum Information, 8: 1740010.
Kovalsky, M. and Hnilo, A. (2004). “Different routes to chaos in the Ti: Sapphire laser,”
Physical Review A, 70: 043813.
Kurtsiefer, C., Oberparleiter, M., and Weinfurter, H. (2001). “High efficiency entangled
photon pair collection in type II parametric fluorescence,” Physical Review A, 64:
023802.
Mermin, D. (1990). “Extreme quantum entanglement in a superposition of macroscopic-
ally distinct states,” Physical Review, 65: 1838–1840.
Peng, C., Yang, T., Bao, X., Zhang, J., Jin, X., Feng, F., . . . Pan, J. W. (2005). “Experi-
mental free-space distribution of entangled photon pairs over 13km: towards satellite-
based global quantum communication,” Physical Review Letters, 94: 150501.
Polchinski, J. (1991). “Weinberg’s nonlinear quantum mechanics and the Einstein-
Podolsky-Rosen paradox,” Physical Review Letters, 66: 397–400.
Scheidl, T., Ursin, R., Kofler, J., Ramelow, S., Ma, X. -S., Herbst, T., . . . Zeilinger, A.
(2010). “Violation of local realism with freedom of choice,” Proceedings of the
National Academy of Sciences of the United States of America, 107: 19708–19713.
Scully, O. M. and Zubairy, M. S. (1997). Quantum Optics. Cambridge: Cambridge
University Press.
Shalm, L. Meyer-Scott, E., Christensen, B., Bierhorst, P., Wayne, M., Stevens, M., . . .
Nam, S. W. (2015). “A strong loophole-free test of local realism,” Physical Review
Letters, 115: 250402.
Weihs, G., Jennewein, T., Simon, C., Weinfurter, H., and Zeilinger, A. (1998). “Violation
of Bell’s inequality under strict Einstein locality conditions,” Physical Review Letters,
81: 5039–5043.
Part IV
Symmetries and Structure in Quantum Mechanics
14
Spacetime Symmetries in Quantum Mechanics
cristian lópez and olimpia lombardi
14.1 Introduction
In the last decades, the philosophy of physics has begun to pay attention to the
meaning and the role of symmetries, an issue that has, however, had a great
relevance in physics since, at least, the middle of the twentieth century. Notwith-
standing this fact, this increasing interest in symmetries has not yet been trans-
ferred to the field of the interpretation of quantum mechanics. Although it is
usually accepted that the Galilean group is the group of invariance of the theory,
discussions about interpretations of quantum mechanics, with very few exceptions,
have not taken into account symmetry considerations. But the invariance of a
theory under a group does not guarantee the invariance of its interpretations, as
they usually add interpretive assumptions to the formal structure of the theory.
Symmetry considerations should thus be seriously taken into account in the field of
the interpretation of quantum mechanics.
For this reason, in this chapter we shall focus on the spacetime symmetries of
quantum mechanics. After briefly introducing certain terminological clarifications,
we shall focus on two aspects of spacetime transformations. First, we shall
consider the behavior of nonrelativistic quantum mechanics under the Galilean
group, aiming at assessing its Galilean invariance in relation to interpretive con-
cerns. Second, we shall analyze the widely-accepted view about the invariance of
the Schrödinger equation under time reversal, in order to unveil some implicit
assumptions underlying such a claim.
14.2 The Concepts of Invariance and Covariance

The meaning of the term ‘symmetry’ is rooted in ordinary language: Symmetry is a
geometrical property of a body whose parts are equal in a certain sense. In
mathematics, the term acquires a precise meaning in terms of invariance – an
269
270 Cristian López and Olimpia Lombardi
object is symmetric with respect to a certain transformation when it is invariant

under such transformation; that is, it remains unchanged under its application. The
concept of group, originally proposed by Galois in the first half of the nineteenth
century, comes to supplement the notion of symmetry – a group clusters different
transformations into a specific structure.
Despite its mathematical precision, in physics, the concept of symmetry has
given rise to some disagreements on the meaning and the scope of the concept of
invariance and of the closely-related concept of covariance. Commonly, the
property of invariance only applies to mathematical objects and, derivatively, to
the physical items to which they refer, and the property of covariance is reserved
for equations and, derivatively, for the physical laws they express. However, some
authors claim that the difference between invariance and covariance not only
makes sense but is also relevant when applied to laws (Ohanian and Ruffini
1994, Suppes 2000, Brading and Castellani 2007). Hans Ohanian and Remo
Ruffini (1994), for instance, claim that an equation is said to be covariant when
its form is left unaltered under a certain transformation, and it is said to be invariant
when it is covariant and its content, that is, its absolute objects (constants and
nondynamical quantities) are also left unchanged by the transformation. Although
inspired in this idea, we will not follow it in every detail. Here we will consider that
an equation is invariant under a certain transformation when it does not change
under the application of such transformation, and it is covariant under that trans-
formation when its form is left unchanged by it (Suppes 2000). From this perspec-
tive, the invariance of a law does not imply the invariance of the objects contained
in its representing equation.
Once one accepts that the concept of invariance makes sense in its application to
laws, the conceptual implications both of the invariance of the law and of the
involved objects under a particular group of transformations deserve to be con-
sidered. Moreover, when a law is covariant under a transformation and all the
objects it contains are also invariant under the same transformation, the law is
invariant under the transformation as well. Nevertheless, this is not the only way
for a law to be invariant – if a law is covariant under a certain transformation, it can
turn out to be invariant under the transformation even in the case that some of the
objects it contains are not invariant under the same transformation (we will come
back to this point in the next section, when discussing the invariance of the
Schrödinger equation).
On the basis of these conceptual clarifications, some formal definitions can now
be introduced.
Def. 1 Let us consider a set A of objects ai 2 A, and a group G of transformations
gα 2 G, where the gα : A ! A act upon the ai as ai ! a~i . An object ai 2 A is
Spacetime Symmetries in Quantum Mechanics 271
invariant under the transformation gα if, for that transformation, a~i ¼ ai . In turn, the
object ai 2 A is invariant under the group G if it is invariant under all the trans-
formations gα 2 G.
In physics, the objects to which transformations apply are usually those represent-
ing states s, observables O, and differential operators D, and each transformation
acts upon them in a particular way. In turn, those objects are combined in equations
representing the laws of a theory. Then,

Def. 2 Let L be a law represented by an equation E s; Oi ; Dj ¼ 0, where s
represents a state, the Oi represent observables, and the Dj represent differential
operators, and let G be a group of transformations gα 2 G acting upon the objects
~ ~
involved in the equation as s !~s , Oi ! O i , and Dj ! D j . L is covariant under the
~ ~
transformation
gα if E ~s ; O i ; D j ¼ 0, and L is invariant under the transformation gα
if E ~s ; Oi ; Dj ¼ 0. Moreover, L is covariant invariant under the group G if it is
covariant invariant under all the transformations gα 2 G.
On this basis, it is usually said that a certain group is the symmetry group of a theory:
Def. 3 A group G of transformations is said to be the symmetry group of a theory if
the laws of the theory are covariant under the group G.
This means that the laws preserve their validity even when the transformations of
the group are applied to the involved objects.
Some authors prefer to talk about symmetry instead of covariance. This is the case
of John Earman (2004), who defines symmetry in the language of model theory:
Def. 4 Let M be the set of the models of a certain mathematical structure, and let
ML M be the subset of the models satisfying the law L. A symmetry of the law
L is a map S : M ! M that preserves ML , that is, for any m 2 ML , SðmÞ 2 ML .

When L is represented by a differential equation E s; Oi ; Dj ¼ 0, each model
m 2 ML is represented by a solution s ¼ F ðOi ; s0 Þ of the equation, corresponding
to a possible evolution of the system.
Then, the covariance of L under a transform-
~ ~
ation g – that is, the fact that E ~s ; O i ;D j ¼ 0 – implies that if s ¼ F ðOi ; s0 Þ is a
solution of the equation, ~s ¼ F~ O ~ i ; s0 is also a solution and, as a consequence, it
represents a model SðmÞ 2 ML . This means that the definition of covariance given
by Def. 2 and the definition of symmetry given by Def. 4 are equivalent.
In turn, the covariance of a dynamical law – represented by a differential
equation – does not imply the invariance of the possible evolutions – represented
by the solutions
of the equation.
In fact, the covariance of the law L, represented
by
the equation E s; Oi ; Dj ¼ 0, implies that s ¼ F ðOi ; s0 Þ and ~s ¼ F~ O ~ i ; s0 are both
solutions of the equation, but does not imply that s ¼ ~s . In the model-theory
language, the symmetry of L does not imply that SðmÞ ¼ m. By contrast,

invariance

is a stronger property of the law: The invariance of L means that E ~s ; Oi ; Dj ¼ 0; in
this case s ¼ ~s ¼ F ðOi ; s0 Þ or, in the model language, SðmÞ ¼ m.
Def. 5 Let M be the set of the models of a certain mathematical structure, and let
ML M be the subset of the models satisfying the law L. Let a transformation be a
map S : M ! M that preserves ML . The law L is invariant under the transform-
ation S if, for any m 2 ML , SðmÞ ¼ m.
The general definitions just described can be applied to the Schrödinger equation
so as to explicitly state the conditions of covariance and invariance for quantum
mechanics. Here we will focus on the evolution equation of the theory, leaving
aside the collapse postulate, since it is an interpretive postulate in orthodox
quantum mechanics. Given a transformation g acting as jφi ! jφ ~
~ i, O ! O,
~ ~
d=dt ! d=dt, and i ! i (considering i as the shorthand for the operator i I), by
making ℏ ¼ 1 the Schrödinger equation is covariant under g when
d~ jφ
~i
¼ ~i H
~ jφ
~i (14.1)
dt
and it is invariant under g when
~i
djφ
~ i:
¼ i H jφ (14.2)
dt
14.3 Quantum Mechanics and the Galilean Group

14.3.1 The Galilean Group
As time is represented by the variable t 2 R and position is represented by the
variable r ¼ ðx; y; zÞ 2 R3 , the Galilean group G ¼ fT α g, with α ¼ 1 to 10, is a
group of continuous spacetime transformations T α : R3 R ! R3 R such that
0
• t !t ¼tþτ (time-displacement)
0
• r!r ¼rþρ (space-displacement)
0
• r ! r ¼ Rθ r (space-rotation)
0
• r ! r ¼ r þ ut (velocity-boost)

where τ 2 R is a real number representing a time interval, ρ ¼ ρx , ρy , ρz 2 R3 is
a triple of real numbers representing a space interval, Rθ 2 M33 is a 3 3 matrix

representing a space rotation by an angle θ, and u ¼ ux ; uy ; uz 2 R3 is a triple of
real numbers representing a constant velocity.
For the Galilean group, G is a Lie group, the Galilean transformations T α
can be represented by unitary operators U α over the Hilbert space, with the
exponential parametrization U α ¼ eiK α sα , where sα is a continuous parameter

and K α is a hermitian operator independent of sα , the generator of the trans-
formation T α . Then, G is defined by 10 group generators K α : one time-
displacement K τ , three space-displacements K ρi , three space-rotations K θi ,
and three velocity-boosts K ui , with i ¼ x, y, z. Therefore, by taking ℏ ¼ 1 as
usual, the Galilean group is defined by the commutation relations between its
generators:
h i h i
ðaÞ K ρi ; K ρj ¼ 0 ðf Þ K ui ; K ρj ¼ iδij M

ðbÞ K ui ; K uj ¼ 0 ðgÞ K ρi ; K τ ¼ 0

ðcÞ K θi ; K θj ¼ iεijk K θj ðhÞ ½K θi ; K τ ¼ 0 (14.3)
h i
ðdÞ K θi ; K ρj ¼ iεijk K ρk ðiÞ ½K ui ; K τ ¼ iK ρi

ðeÞ K θi ; K uj ¼ iεijk K uk
where εijk is the Levi-Civita tensor. Strictly speaking, in the case of quantum
mechanics the symmetry group is the group corresponding to the central extension
of the Galilean algebra, obtained as a semi-direct product between the Galilean
algebra and the algebra generated by a central charge, which in this case is the
mass operator M ¼ mI, where I is the identity operator and m is the mass. The
mass operator as a central charge is a consequence of the projective representation
of the Galilean group (see Bose 1995, Weinberg 1995). However, in order to
simplify the presentation, we will use the expression “Galilean group” from now
on to refer to the corresponding central extension.
In a closed, constant-energy system free from external fields, the generators K α
are given by the basic magnitudes of the theory: the energy H ¼ ℏK τ , the three
momentum components Pi ¼ ℏK ρi , the three angular momentum components
J i ¼ ℏK θi , and the three boost components Gi ¼ ℏK ui . Then, in this case the
commutation relations turn out to be

ðaÞ Pi ; Pj ¼ 0 ðf Þ Gi ; Pj ¼ iδij M

ðbÞ Gi ; Gj ¼ 0 ðgÞ ½Pi ; H ¼ 0

ðcÞ J i ; J j ¼ iεijk J k ðhÞ ½J i ; H ¼ 0 (14.4)

ðdÞ J i ; Pj ¼ iεijk Pk ðiÞ ½Gi ; H ¼ iPi

ðeÞ J i ; Gj ¼ iεijk Gk
The rest of the physical magnitudes can be. defined in terms of these basic ones. For
instance, the three position components are Qi ¼ Gi =m, the three orbital angular
momentum components are Li ¼ εijk Qj Pk , and the three spin components are
Si ¼ J i Li . In the Hilbert formulation of quantum mechanics, each Galilean

transformation gα 2 G acts upon states and upon observables as
~ i ¼ U sα jφi ¼ eiK α sα jφi
jφi ! jφ (14.5)
~ ¼ U sα O U 1 ¼ eiK α sα O eiK α sα
O!O (14.6)
sα
The invariance of an observable O under a Galilean transformation T α amounts to

the commutation between O and the corresponding generator K α :
~ ¼ eiK α sα OeiK α sα ¼ O , ½O; K α ¼ 0
O (14.7)
It is worth bearing in mind that there are operators that are invariant under all
the transformations of the group, and thereby, commute with all the generators of
the group – the Casimir operators of a group. In the case of the Galilean group, the
2 the internal energy W ¼ H P =2M, the square of total spin
2
Casimir operators are
S ¼ J M G P , and the mass M, which are multiples of the identity in any
2 1
irreducible representation.
14.3.2 The Covariance of the Schrödinger Equation

Given the Schrödinger equation, let us begin by (i) premultiplying its two members
by U ¼ eiKs , (ii) adding and subtracting ðdU=dtÞjφi to its first member, and (iii)
using the property U 1 U ¼ I:
djφi dU dU 1
U þ jφi U U jφi ¼ UiU 1 UHU 1 U jφi (14.8)
dt dt dt
Then, by recalling the transformations of states and observables of Eq. (14.5) and
Eq. (14.6), we obtain
~ i dU 1
d jφ
~ i ¼ ~i H
U jφ ~ jφ
~i (14.9)
dt dt
This shows that covariance obtains when the time-derivative operator transforms
as
d d~ D d dU 1 d~ jφ
~i
! ¼ ¼ U ) ¼ ~i H
~ jφ
~i (14.10)
dt dt Dt dt dt dt
~
This means that the transformed differential operator d=dt is a covariant time-
derivative D=Dt, which makes the Schrödinger equation to be Galilean-covariant
in the sense of Eq. (14.1).
In a closed, constant-energy system free from external fields, H is time-
independent and the Pi and the J i are constants of motion (see Eq. (14.4g) and
Eq. (14.4h)). Then, for time-translations, space-translations and space-rotations,

dU=dt ¼ deiKs =dt ¼ 0, where K and s stand for H and τ, Pi and ρi , and J i and θi ,
respectively. As a consequence, the time-derivative is invariant under time-
displacements, space-displacements, and space-rotations (see Eq. (14.10)):
~
d=dt ! d=dt ¼ d=dt. But for boost-transformations this is not the case: The
covariance of the Schrödinger equation implies the transformation of the differen-
tial operator as d=dt ! D=Dt. This means that covariance under boosts amounts to
a sort of “nonhomogeneity” of time, which requires the covariant adjustment of the
time-derivative. This conclusion should not be surprising since, when the system is
described in a reference frame F~ at uniform motion corresponding to a velocity ux
with respect to the original frame F, the boost-transformed state depends on a
generator that is a linear function of time:
Gx ¼ mQx ¼ mðQx0 þ V x t Þ ¼ mQx0 þ Px t (14.11)

~ where the state is jφ
Then, if the Schrödinger equation is to be valid in F, ~ i, the
transformed time-derivative has to be adjusted to compensate the time-depending
transformation of the state.
14.3.3 The Invariance of the Schrödinger Equation

As we have seen in the previous section, in a closed, constant-energy system free from
external fields, H is time-independent and the Pi and the J i are constants of motion.
Then, for time-translations, space-translations, and space-rotations, it follows that
dU=dt ¼ deiKs =dt ¼ 0 and d=dt ! d=dt ~ ¼ d=dt. Moreover, for those transform-
~ ~
ations, i ¼ i follows trivially, and H ¼ H because (see Eq. (14.7)) (i) ½H; H ¼ 0, (ii)
½Pi ; H ¼ 0 (Eq. (14.4g)), and (iii) ½J i ; H ¼ 0 (Eq. (14.4h)). When these results apply
to Eq. (14.9), it is easy to see that the Schrödinger equation is invariant under time-
displacements, space-displacements, and space-rotations in the sense of Eq. (14.2).
The case of boost-transformations is different from the previous cases because,
although ~i ¼ i still holds, the Hamiltonian is not boost-invariant even when the
system is free from external fields (the same happens in classical mechanics, see
Butterfield 2007: 6). In fact, under a boost-transformation corresponding to a
velocity ux , H changes as (see Eq. (14.4i): ½Gx ; H ¼ iPx 6¼ 0)
~ ¼ eiGx ux HeiGx ux 6¼ H
H (14.12)
Since Gx is not time-independent, dU=dt ¼ deiGx ux =dt 6¼ 0, and Eq. (14.9) yields

~i
d jφ ~ deiGx ux iGx ux
¼ i H þ i e ~i
jφ (14.13)
dt dt
In order to know the value of the bracket in the right-hand side (r.h.s.) of Eq.
(14.13), the two terms in the bracket must be computed. When the task is
performed, it can be proved that the terms added to H in H~ cancel out with those
coming from the term containing the time-derivative (see Lombardi, Castagnino,
and Ardenghi 2010: appendices). Therefore, Eq. (14.2) is again obtained and the
invariance of the Schrödinger equation is proved to hold also for boost-
transformations.
The case of boost-transformations illustrates a claim previously mentioned in
Section 14.2: Even though a law is invariant under a transformation when it is
covariant and all the involved objects are invariant, this is not the only way to
obtain invariance. When the quantum system is free from external fields, the
Schrödinger equation is invariant under boost-transformations, in spite of the fact
that the Hamiltonian and the differential operator d=dt are not boost-invariant
objects.
14.3.4 Galilean Group and External Fields

As explained in the previous subsection, when there are no external fields acting on
the system, the Hamiltonian is invariant under time-displacements, space-displace-
ments, and space-rotations, but not under boost-transformations. Despite this fact,
the Schrödinger equation is completely invariant under the Galilean group, and
this conceptually means that the state vector jφi does not “see” the effect of
the transformations – the evolutions of jφi and jφ ~ i are identical. In other words,
the time-behavior of the system is independent of the reference frame used for the
description.
When the system is under the action of external fields, the fields modify the
evolution of the system. But, in nonrelativistic quantum mechanics, fields are not
quantized: They do not play the role of quantum systems that interact with other
systems. For this reason, the effect of the fields on a system must be included in its
Hamiltonian, because it is the only observable involved in the time-evolution law.
It can be proved that the most general form of the Hamiltonian in the presence of
external fields is (see, e.g., Ballentine 1998)
ðP AðQÞÞ2
H¼ þ V ð QÞ (14.14)
2M
where AðQÞ is a vector potential and V ðQÞ is a scalar potential. The covariance of
the Schrödinger equation, as expressed in Eq. (14.9), fixes the way in which the
potentials AðQÞ and V ðQÞ must transform under the Galilean group. The electro-
magnetic field may be derived from a vector potential and a scalar potential; thus, a
fully Galilean-covariant quantum theory of the Schrödinger field interacting with
an external electromagnetic field is possible. However, the electric and magnetic

fields that transform as required to preserve Galilean covariance, although related
to the scalar and vector potentials in the usual way, are ruled by one of two sets of
electromagnetic “field” equations. Those equations can be considered the nonre-
lativistic limits of Maxwell’s equations in cases where either (i) magnetic effects
predominate over electric ones (“magnetic limit”: cjBj >> jE j), or (ii) electric
effects predominate over magnetic ones (“electric limit”: cjBj << jE j) (see Brown
and Holland 1999, Colussi and Wickramasekara 2008). Nevertheless, AðQÞ and
V ðQÞ should not necessarily be identified with the electromagnetic potentials,
because they are arbitrary functions that need not satisfy Maxwell’s equations;
for example, the Newtonian gravitational potential can also be included in the
scalar V ðQÞ (Ballentine 1998).
At this point, a relevant issue must be stressed. Space-displacements and space-
rotations are purely geometric operations of displacing and rotating the system
self-congruently to another place and to another direction, respectively. Analo-
gously, time-displacements are purely geometric operations of displacing the
system self-congruently to another time, and they may agree or not with dynamical
evolutions. The commutation of two transformation generators means that the
corresponding geometric operations can be performed in either order with the
same result; for instance, the commutation ½K ρi , K ρj ¼ 0 (see Eq. (14.3a)) means
that the order in which space-displacements in different directions are performed
does not modify the result. In particular, the validity of the Galilean group implies
that time-displacements commute both with space-displacements and with space-
rotations (see Eq. (14.3g) and Eq. (14.3h)). When there are no external fields acting
on the system, this feature is given by the commutation relations involving the
Hamiltonian, the three momentum components Pi , and the three angular momen-
tum components J i , ½Pi ; H ¼ 0 and ½J i ; H ¼ 0 (Eq. (14.4g) and Eq. (14.4h)),
because here the Hamiltonian is the time-displacement generator. But in the
presence of external fields, since the action of the fields is incorporated into the
Hamiltonian of the system, the Hamiltonian is no longer the generator of time-
displacements: It only retains its role as the generator of the dynamical evolution
(see Laue 1996, Ballentine 1998). For this reason, the commutation with the
momentum components and with the angular momentum components gets broken:
½Pi ; H 6¼ 0 and ½J i ; H 6¼ 0. However, to the extent that the covariance of the
Schrödinger equation is retained, the commutation of time-displacements with
both space-displacements and space-rotations still holds, and is still represented

by K ρi ; K τ ¼ 0 and ½K θi ; K τ ¼ 0, respectively (see Eq. (14.3g) and Eq. (14.3h)),
where the momentum components are still the generators of space-displacements,
Pi ¼ ℏK ρi , and the angular momentum components are still the generators of
space-rotations, J i ¼ ℏK θi , but the Hamiltonian is no longer the generator of time-

displacements, H 6¼ ℏK τ (we will come back to this point in the next section, about
time-reversal invariance).
14.3.5 The Relevance to Interpretation

In principle, there are two possible interpretations of a transformation: active and
passive. Under the active interpretation, the transformation corresponds to a
change from a system to the transformed system; for instance, a system translated
in space with respect to the original one. Under the passive interpretation, the
transformation consists in a change of the viewpoint – the reference frame – from
which the system is described; for instance, the space-translation of the observer
that describes the system. In the case of continuous spacetime transformations,
both active and passive interpretation are equally allowed; but such a situation is
not so clear in the case of discrete transformations. In general, it is accepted that
only the active interpretation makes sense in the case of discrete transformations
(Sklar 1974: 363). Nevertheless, no matter which interpretation is adopted, the
covariance of the fundamental law of a theory under its continuous symmetry
group implies that the law still holds when the transformations are applied. In the
active interpretation language, the original and the transformed systems are equiva-
lent; in the passive interpretation language, the original and the transformed
reference frames are equivalent.
As is typically accepted, the Galilean group is the symmetry group of continuous
spacetime transformations of classical and quantum mechanics. In the language of
the passive interpretation, the covariance of the dynamical laws amounts to the
equivalence among inertial reference frames (time-translated, space-translated,
space-rotated, or uniformly moving with respect to each other). In other words,
Galilean transformations do not introduce any modification in the physical situation,
but only express a change in the perspective from which the system is described.
These remarks are related to the fact that certain quantities are physically
irrelevant in the light of theory’s symmetries. For instance, the space-translation
symmetry of a dynamical law means that the specific place where the system is
located in space is irrelevant to its evolution governed by such law: “A global
symmetry reflects the irrelevance of absolute values of a certain quantity: only
relative values are relevant” (see Brading and Castellani 2007: 1360). In classical
mechanics, for example, space-translation invariance implies that absolute position
is irrelevant to the system’s behavior – the equations of motion do not depend on
absolute positions, only relative positions matter. The physical irrelevance of
certain quantities is strongly linked with the issue of objectivity.
The intuition about a strong link between invariance and objectivity is rooted in
a natural idea: What is objective should not depend on the particular perspective
used for the description. When this intuition is translated into the group-theoretical
language, it can be said that what is objective according to a theory is what is
invariant under the symmetry group of the theory. This idea appeared in the
domain of formal sciences in Felix Klein’s “Erlangen Program” of 1872, with
the attempt to characterize all known geometries by their invariants (see Kramer
1970). This idea passed to physics with the advent of relativity, regarding the
ontological status of space and time (Minkowski 1923). The claim that objectivity
means invariance becomes a main thesis of Hemann Weyl’s book Symmetry
(1952). Max Born also very clearly expressed his conviction about the strong link
between invariance and objectivity: “I think the idea of invariance is the clue to a
rational concept of reality” (Born 1953: 144). In recent times, the idea has strongly
reappeared in several works. For instance, in her deep analysis of quantum field
theory, Sunny Auyang (1995) makes her general concept of “object” to be founded
on its invariance under transformations among all representations. The assumption
of invariance as the root of objectivity is also the central theme of Robert Nozick’s
book Invariances: The Structure of the Objective World (2001). In the same vein,
David Baker (2010) has argued that symmetries are a guide to finding out which
quantities represent fundamental natural properties in a physical theory.
If the ontological meaning of symmetries is accepted, it is easy to see that
symmetries must play an active role in the understanding of a physical theory. In
the particular case of quantum mechanics, the consideration of its Galilean
covariance cannot be overlooked in the discussions about interpretation.
As it is well known, the Kochen-Specker theorem (Kochen and Specker 1967)
establishes a barrier to any realist classical-like interpretation of quantum mechan-
ics: It proves the impossibility of ascribing definite values to all the physical
quantities (observables) of a quantum system simultaneously, while preserving the
functional relations between commuting observables. This result is a manifestation
of the contextuality of quantum mechanics – the ascription of definite values to the
observables of a quantum system is always contextual. As a consequence of the
Kochen-Specker theorem, any realist interpretation of quantum mechanics is com-
mitted to selecting a subset of definite-valued observables from the set of all the
observables of the system (or a preferred basis from all the formally equivalent bases
of the Hilbert space). The observables of that subset will be those that acquire
definite values without violating quantum contextuality. It is at this point that the
symmetry group of the theory becomes a leading character. As noticed by Harvey
Brown, Mauricio Suárez, and Guido Bacciagaluppi (1998), any interpretation that
selects the set of the definite-valued observables of a quantum system in a given state
is committed to considering how that set is transformed under the Galilean group.
However, now the link between invariance and objectivity comes into play. The
study of the role of symmetries is particularly pressing in the case of realist
interpretations of quantum mechanics, which conceive a definite-valued observ-
able as a physical magnitude that objectively acquires an actual definite value
among all its possible values: The fact that a certain observable acquires a definite
value should be an objective fact that should not depend on the descriptive
perspective. Therefore, the set of the definite-valued observables of a system
picked out by the interpretation should be left invariant by the Galilean
transformations. From a realist viewpoint, it would be unacceptable that such a
set changed as result of a mere change in the perspective from which the system is
described.
In his article “Aspects of objectivity in quantum mechanics,” Harvey Brown
(1999) explicitly tackles the problem in discussing the objectivity of “sharp
values.” In particular, he focuses on interpretations that specify state-dependent
rules for assigning sharp values to some of the self-adjoint operators representing
quantum magnitudes, such as the interpretations whose value-assignment rules
coincide with the eigenstate-eigenvalue link, or the modal interpretations that make
the set of definite-valued observables to depend on the instantaneous state of the
system. Brown clearly explains the difference between the classical and the
quantum case. In classical mechanics, Galilean noninvariant magnitudes modify
their values with the change of reference frame; for this reason, if their objectivity
is to be retained, they must be regarded not as intrinsic properties but as relational
properties. For instance, the values of classical position and momentum can be
conceived as relational properties that link the system and the reference frame. In
quantum mechanics, by contrast, the relational nature acquires a further degree;
whereas, in the classical case the sharp value of a magnitude depends on the
reference frame, in the quantum case the very sharpness of an observable’s value
must be relational in order to preserve its objectivity. For instance, the fact that the
position of a system has a sharp (definite) value in a certain reference frame, and,
as a consequence, that the system can be conceived as a localized particle, is itself
relational. In a different reference frame, the system may have an unsharp value of
position and, then, may behave as a delocalized particle. Brown also correctly
stresses that it is not just boosts that produce this kind of situation; passive spatial
translations can cause that some sharp-valued observables to become unsharp. On
the basis of this analysis, he concludes that
If, in the hope of providing an ontological interpretation of quantum mechanics, we
introduce state-dependent rules for assigning sharp values to magnitudes associated with
a specific quantum system, we should recognise that the objective status of such sharp
values is relational, not absolute.
(Brown 1999: 66–67)
In the same article, Brown considers an interpretation in which the rule of

definite-value ascription is not state-dependent; he analyzes the problem of covar-
iance in the de Broglie-Bohm pilot-wave interpretation of quantum mechanics, but
he discards it because, although no privileged frame is picked out by the hidden
dynamics of the corpuscles, the forces acting upon the corpuscles and generated by
the guiding wave are Aristotelian, not Newtonian, because they produce velocities,
not accelerations. This seems to lead him to interpretations that introduce state-
dependent rules of definite-value ascription as the only alternative. This maneuver
is opposed to that which led Jeffrey Bub (1997) to advocate for Bohmian mechan-
ics, conceived as a modal interpretation whose rule of definite-value ascription
picks out the position observable: The difficulties of the original modal
interpretations to deal with nonideal measurements due to their state-dependent
rules turns Bohmian mechanics into a natural alternative. But what both Bub and
Brown seem to overlook is that there are other interpretive strategies beyond
Bohmian mechanics and traditional modal interpretations that make the definite-
valued observables to depend on the state of the system. One of them is even more
natural than the Bohmian proposal when the aim is to preserve the objectivity of
definite-valuedness (or of sharpness, in Brown’s terms) in the light of Galilean
symmetry. In fact, the natural way to reach this goal, without making the objective
status of definite-valuedness relational, is to appeal to the Casimir operators of the
Galilean group: If the interpretation has to select a Galilean-invariant set of
definite-valued observables, such a set must depend on those Casimir operators,
insofar as they are invariant under all the transformations of the Galilean group. An
interpretation that has adopted this interpretive strategy is the modal-Hamiltonian
interpretation in its Galilean invariant version (Ardenghi, Castagnino, and Lom-
bardi 2009, Lombardi, Castagnino, and Ardenghi 2010), which has been success-
fully applied to many well-known physical situations and has proved to be
effective for solving the measurement problem, both in its ideal and its nonideal
versions (Lombardi and Castagnino 2008).
Considering that the Casimir operators of the Galilean group represent the
definite-valued observables of a quantum system has the advantage of being very
general. When the system is free from external fields and the Galilean group is
defined by the commutation relations Eq. (14.4), the Casimir operators correspond
to the observables mass M, squared-spin S2 , and internal energy W. Yet the
Casimir operators of the Galilean group can always be defined, even when Eq.
(14.4) does not hold. The group must thus be defined in the completely general
way as expressed by Eq. (14.3). Therefore, when there are external fields applied to
the system, and such fields do not break the covariance of quantum mechanics
under the Galilean group, the strategy of defining the definite-valued observables
in terms of the Casimir operators remains valid, whatever they represent.
Furthermore, the strategy admits a further generalization in its application to

relativistic quantum theories, such as relativistic quantum mechanics and quantum
field theory: The interpretive postulate of endowing the Casimir operators with
definite-valuedness and objectivity is retained, but the relevant group structure is
replaced by changing the Galilean group by the Poincaré group.
14.4 Quantum Mechanics and Time Reversal

14.4.1 A General Notion of Time Reversal
What was said in the previous section seems not to straightforwardly apply to the
question about time-reversal symmetry. To begin with, time reversal is a trans-
formation that does not belong to the Galilean group. Furthermore, while all
Galilean transformations are continuous, time reversal is a discrete transformation
that, at first glance, simply performs the transformation t ! t. Notwithstanding
these facts, time reversal encloses even more subtle features as it is somehow
related to the nature of time and its unavoidable differences with respect to space;
whereas one can move freely in all directions of space, it seems that one “moves”
in just one direction of time, from past to future and never the other way around.
From early twentieth century, the very notion of time reversal was quite relevant
not only for many physicists working on the foundations of physics, but also for
many philosophers aiming at getting a grasp of the nature of time.
Unfortunately, time is not the kind of thing one can experiment on: Unlike
electrons, pendulums or electromagnetic fields, one cannot directly find out time’s
properties by running an experiment. However, physicists and philosophers have
managed this setback by devising a formal way to dig into time’s properties – the
notions of time reversal and time-reversal invariance have been keystones for the
famous problem of the arrow of time in physics (problem lying on the borderline
between the physics and the metaphysics of time). In fact, the arrow of time has
largely been introduced in terms of time-reversal symmetry: If physical laws
somehow fail to be time-reversal invariant, then one might come to the conclusion
that time is headed according to such a physical law. To put it differently, time-
reversal symmetry is supposed to shed light on the structure of time according to a
given theory. Quoting Jill North:
If the fundamental laws cannot be formulated without reference to a particular kind of
structure, then this structure must exist in order to support the laws – “support” in the sense
that the laws could not be formulated without making reference to that structure.
(North 2009: 203)
The idea is that, by knowing how dynamical equations (standing for physical laws)
behave under time reversal, one can learn about the nature of time according to a
theory. This kind of principle is well-seated in the literature (see also Earman 1974,
Sklar 1974, Arntzenius 1997), and it is the key element that links time’s properties,
physical laws, and the time-reversal transformation.
However, here nothing has yet been said about the time-reversal transformation
in itself, other than that it minimally performs the transformation t ! t. As
typically noted in the literature, the topic is somewhat tricky as there is no shared
understanding of what time reversal is exactly supposed to do nor of what proper-
ties it should instantiate (see Savitt 1996 for a careful analysis of varied time-
reversal operators; see Peterson 2015 for an updated approach). Furthermore, time
reversal’s properties seem to change from theory to theory, to the extent that
dynamics also changes. This remark already assumes a strong premise – that the
structure of time, in particular, regarding its time-reversal symmetry property, is
closely tied up to the theory’s dynamics.
14.4.2 Time Reversal in Quantum Mechanics

Independently of the considerations discussed previously, the community of
physicists has reached a wide consensus about the appropriate time-reversal
transformation for standard quantum mechanics. The traditional procedure starts
out by arguing that time reversal can no longer be considered as in Hamiltonian
classical mechanics. As it is well known, the time-reversal operator in classical
contexts is typically defined as the operator that changes the sign of the variable t.
But in quantum mechanics the rationale seems to be different; in fact, Ballentine
(1998) warns us:
One might suppose that time reversal would be closely analogous to space inversion, with
the operation t ! t replacing x ! x. In fact, this simple analogy proves to be
misleading at almost every step.
(Ballentine 1998: 377)
Quantum mechanics textbooks rarely offer a thorough justification for such a claim
and commonly go on by formally introducing the “proper” way to reverse time in
quantum mechanics. In some cases, the only justification is based on a classical
analogy: The transformation t ! t does not lead to the transformation of
momentum as P ! P, which is expected because this is the way in which
momentum transforms under time reversal in classical mechanics. But mere
analogy does not seem to be a sufficiently good argument; for this reason, Bryan
Roberts (2017) has very recently brought up an updated and purely quantum-
mechanic-based reasoning for defending the standard procedure. Here we will not
analyze in detail those arguments; rather, we will consider the problem in the light
of the symmetries of the Schrödinger equation related to the reversal of time.
Let us use θ to call a generic time-reversal operator, which performs at least the
transformation t ! t but whose precise form is not defined yet. As explained in
Section 14.2, this operator acts as jφi ! jφ~ i, O ! O, ~
~ d=dt ! d=dt, and i ! ~i; in
~ i ¼ θjφi, O
particular, jφ ~
~ ¼ θO θ , d=dt ¼ θ d=dt θ , and ~i ¼ θ i θ . The Schrö-
1 1 1
dinger equation is covariant/invariant under θ when Eq. (14.2)/(14.3) holds,

respectively. So, let us begin by premultiplying the two members of the Schrö-
dinger equation by θ and using the property θ1 θ ¼ I:

d 1
θ θ θjφi ¼ θiθ1 θHθ1 θjφi (14.15)
dt
As long as θ is not a function of t, it is easy to prove that the Schrödinger equation
is covariant under the application of θ:
d~ jφ
~i
¼ ~i H
~ jφ
~i (14.16)
dt
Now, in order to know if the Schrödinger equation is also invariant under the
application of θ, it is necessary to define the precise form of θ to see how it acts
upon d=dt, i, and H. As in the case of the Galilean group, the situation of a closed,
constant-energy system will be considered.
Case (i): If θ ¼ T only performs the transformation t ! t, then

d~ d d ~i ¼ TiT 1 ¼ i ~ ¼ THT 1 ¼ H
¼T T 1 ¼ H (14.17)
dt dt dt
Introducing these equations into Eq. (14.16) leads to the conclusion that the
Schrödinger equation is not T-invariant, since
~i
d jφ
~i
¼ iH jφ (14.18)
dt
Case (ii): If θ ¼ T ∗ performs the transformation t ! t and the complex

conjugation i ! i, then

d~ ∗ d d ~i ¼ T ∗ iT ∗1 ¼ i ~ ¼ T ∗ HT ∗1 ¼ H
¼T T ∗1 ¼ H
dt dt dt
(14.19)
Introducing these equations into Eq. (14.16) leads to the conclusion the Schrödinger
equation is T ∗ -invariant, since
~i
djφ
~i
¼ i H jφ (14.20)
dt
Summing up, independently of any interpretation, the results given by Eq. (14.18)
and Eq. (14.20) show that the Schrödinger equation is not invariant under the unitary
operator T, and it is invariant under the antiunitary operator T ∗ . The conceptual question
is which of the two operators, T or T ∗ , represents the operation of time reversal.
14.4.3 Time-Reversal Invariance: Between Petitio Principii

and a Priori Truth
A common answer to that conceptual question is saying that the unitary operator T is
unacceptable as the time-reversal operator because it breaks the requirement that the
energy of the system must be bounded from below. In Jun John Sakurai’s words:
Consider an energy eigenket jni with energy eigenvalue E n . The corresponding time-
reversed state would be Θjni [where Θ stands for our T], and we would have, because of
(4.4.27) [HΘ ¼ ΘH]
HΘjni ¼ ΘH jni ¼ ðE n ÞΘjni (4.4.28)

This equation says that Θjni is an eigenket of the Hamiltonian with energy eigenvalues
En . But this is nonsensical even in the very elementary case of a free particle. We know
that the energy spectrum of the free particle is positive semidefinite – from 0 to +∞. There
is no state lower than a particle at rest (momentum eigenstate with momentum eigenvalue
zero); the energy spectrum ranging from ∞ to 0 would be completely unacceptable.
(Sakurai 1994: 272–273)
Why does the unitary operator T not meet that requirement? It is not unusual to read
that the reason is that the unitary operator T transforms the Hamiltonian as
THT 1 ¼ H. This sounds very strange because, by performing only the transform-
ation t ! t, the operator T should leave the time-independent Hamiltonian invari-
ant. So, now the question is: Why does the Hamiltonian transform as H ! H?
Although not always explicit, a typical answer is that offered by Stephen Gasiorowicz:
we find that [the equation of motion] can be invariant only if
THT 1 ¼ H
This, however, is an unacceptable condition, because time reversal cannot change the spectrum
of H, which consists of positive energies only. If T is taken to be anti-unitary [our T ∗ ], the
* operator changes the i to i [in the equation of motion] and the trouble does not occur.
(Gasiorowicz 1966: 27; italics added)
In a similar vein, Robert Sachs clearly explains that

we require that the [time-reversal] transformations leave the equations of motion invariant
when all forces or interactions vanish.
(Sachs 1987: 7; italics added)
In other words, under time reversal, the Hamiltonian should transform as H ! H

in order to preserve the time-reversal invariance of the Schrödinger equation; for
this reason, the unitary T is unacceptable as time-reversal operator, and the right
one is the antiunitary T ∗ .
At this point, one seems to be caught in the dilemma between petitio principii
and a priori truth. If our original question was whether the Schrödinger equation is
time-reversal invariant, an argument that selects the right operator describing time
reversal by taking the time-reversal invariance of the equation as one of its
premises clearly begs the question. However, some authors do not fall in petitio
principii by considering that certain symmetries of the physical laws have an a
priori status:
A symmetry can be a priori, i.e., the physical law is built in such a way that it respects that
particular symmetry by construction. This is exemplified by spacetime symmetries,
because spacetime is the theater in which the physical law acts . . . and must therefore
respect the rules of the theater.
(Dürr and Teufel 2009: 43–44)
From this perspective, the invariance under the Galilean group must be built into
the Schrödinger equation due to the homogeneity and the isotropy of space and the
homogeneity of time. This view may sound reasonable to the extent that those are
features of space and time that we, in a certain sense, can experience. But, why
should we impose time-reversal invariance? We have no experience of the isotropy
of time since we cannot travel backwards in time. Despite this, time-reversal
invariance must be introduced as a postulate:
One should ask why we are so keen to have this feature in the fundamental laws when we
experience the contrary. Indeed, we typically experience thermodynamic changes which
are irreversible, i.e., which are not time reversible. The simple answer is that our platonic
idea (or mathematical idea) of time and space is that they are without preferred direction,
and that the “directed” experience we have is to be explained from the underlying time
symmetric law.
(Dürr and Teufel 2009: 47)
Challenging the most widely-held position about time reversal in the field
of quantum mechanics, a few authors have raised their voices against it by
appealing to philosophical reasons: The antiunitary operator T ∗ would fail to
offer a conceptually sound and clear-cut representation of time reversal. On
the one hand, a far-reaching tradition, which tracks back to the work of Giulio
Racah (1937) and Satosi Watanabe (1955), pleads for a unitary time-reversal
operator in quantum theories. Oliver Costa de Beauregard (1980) has argued
for such a view by claiming that a unitary time-reversal operator that merely
reverses the direction of time by flipping the sign of the variable t goes more
naturally along with relativistic contexts and is more naturally akin to the
Feynmann zig-zag philosophy. On the other hand, philosophers such as Craig
Callender (2000) and David Albert (2000) have claimed that the Schrödinger
equation should actually be considered as nontime-reversal invariant, since
it is not invariant under T, which more fairly represents what one means
by “reversing the direction of time.” Without any further ado, Jill North
claims:
What is a time-reversal transformation? Just a flipping of the direction of time! That is all
there is to a transformation that changes how things are with respect to time: change the
direction of time itself.
(North 2009: 212)
For these philosophers, if the time-reversal invariance of a theory will tell us

something about the structure of time, time reversal should only reverse the
direction of time without extra additions. In particular, if the question at issue is
the problem of the arrow or time and the time-reversal invariance of the theory is
considered relevant to this problem, imposing the time-reversal invariance of the
Schrödinger equation as a requirement that the theory must satisfy is a circular
strategy.
14.4.4 Wigner’s Definition

The line of argumentation sketched in the previous section seems to be not
completely convincing for adopting the antiunitary operator T ∗ as the adequate
representation of time reversal. Yet a thorough argument can be introduced by
appealing to the authority of Eugene Wigner, who defines time reversal as a
transformation such that:
The following four operations, carried out in succession on an arbitrary state, will result in
the system returning to its original state. The first operation is time inversion, the second
time displacement by t, the third again time inversion, and the last on again time
displacement by t.
(Wigner 1931/1959: 326)
In other words,
time reversal displacement by Δt time reversal displacement by Δt = identity
This requirement is commonly interpreted in formal terms as follows:
U Δt θ1 U Δt θ s ¼ s (14.21)
where s is the arbitrary state and U Δt is the evolution operator for Δt. This
requirement is precisely Sakurai’s starting point in the argument that led him to
the conclusion stated previously. Under the assumption that the evolution oper-
ators form a group (an assumption that is not always satisfied, see Bohm and
Gadella [1989], where the time evolution is represented by a semigroup), a U Δt
exists such that U Δt U Δt ¼ I. In this case, Eq. (14.21) becomes
θ1 U Δt θ s ¼ U Δt s (14.22)

In quantum mechanics, the argument continues, U Δt ¼ eiHΔt :
θ1 eiHΔt θ jφi ¼ eiHΔt jφi (14.23)

Therefore,
θ1 ðiH Þθ ¼ iH (14.24)

So, when θ is the unitary operator T, this leads to
H ¼ T H T 1 (14.25)
which is unacceptable because it leads to values of energy that are unbounded from
below. Since Wigner (1931/1959) also proved that any symmetry transformation is
represented by a unitary or an antiunitary operator, the argument concludes that the
right time-reversal operator is the antiunitary operator T ∗ .
This argument in favor of T ∗ is certainly much better than the previous one,
which takes the time-reversal invariance of the Schrödinger equation as one of the
premises. Nevertheless, as we will see, this second argument imposes the time-
reversal invariance of the dynamical equation beforehand as well, even if in a more
subtle way.
Let us begin by noticing that, when Eq. (14.21) is used to formalize Wigner’s
requirement, the “displacement by Δt” is represented by the time evolution of the
system by Δt according to the dynamical law of the theory – the Schrödinger
equation, here expressed as s ¼ U Δt s0 . However, as stressed in Section 14.3.4,
when considering the Galilean group, time displacement is not time evolution.
Time evolution is ruled by the dynamical law of the theory, in this case quantum
mechanics: According to the Schrödinger equation, the Hamiltonian is the gener-
ator of the dynamical evolution. By contrast, spacetime transformations are
purely geometric operations of displacing or rotating the system self-congruently.
In particular, time-displacement is a purely geometric operation that displaces the
system self-congruently to another time, and may agree or not with time evolu-
tion. In fact, as Hans Laue (1996) and Leslie Ballentine (1998) stress, in a generic
case, the Hamiltonian is not the generator of time displacements and only retains
its role as the generator of the dynamical evolution. This clearly shows that
time displacement and time evolution are different concepts. Hence, insofar as
Wigner’s definition involves time displacement and not time evolution, it must be
formally expressed as (recall how Galilean transformations act upon states,
Eq. (14.5))
U τ θ1 U τ θ s ¼ eiK τ Δt θ1 eiK τ Δt θ s ¼ s (14.26)
where K τ is the generator of time-displacement (see Eq. (14.3)). Now, since U τ

exists such that U τ U τ ¼ I, an analogue of Eq. (14.23) can be obtained:
θ1 eiK τ Δt θ jφi ¼ eiK τ Δt jφi (14.27)
where the difference in sign between Eq. (14.27) and Eq. (14.22) is due to the
inverse relation between transformations on function space and transformations on
coordinates (see Ballentine 1998: 67). Eq. (14.27) expresses what time reversal
means: It should be a transformation such that
time reversal displacement by Δt time reversal = displacement by Δt
More explicitly, take an arbitrary state, time-reverse it, time-displace it by Δt in a

given time-displacement direction, and time-reverse it again; these three operations
must be equivalent to time-displace the original state the same time interval Δt in
the opposite time-displacement direction. Eq. (14.27) represents formally this
condition, but no conclusion about how the Hamiltonian is transformed by time
reversal can be drawn from it.
Even if accepting the conceptual difference between time displacement and time
evolution in Wigner’s definition of time reversal, somebody might retort by saying
that there are cases in which time displacement amounts to time evolution: as
discussed in Section 14.3, in those cases the generator K τ of time displacement is
equal to the generator H of time evolution. If K τ is replaced by H in Eq. (14.27),
when θ is the unitary operator T the relation H ¼ T H T 1 of Eq. (14.25) obtains
again, and this is sufficient to discard the unitary operator T as the adequate
representation of time reversal.
Although the argument just discussed seems to be conclusive, when con-
sidered in detail, the implicit assumptions come to light. In fact, the argument
equates time displacement and time evolution, both in the case of Δt and in the
case of Δt. We have good empirical reasons to accept that, in certain cases, the
time displacement toward the future by Δt is equivalent to the time evolution
given by U Δt ¼ eiHΔt . But we do not know how the system would evolve in
time toward the past, as we have no experience at all of such an evolution; this is
the specific feature that makes time so different than space. If we impose that,
when time displacement is time evolution toward the future, this is the case
toward the past too, then we are introducing the time-reversal invariance of the
dynamical law by hand.
In other words, the question about the time-reversal invariance of a law is
precisely the question of whether the time displacement of the system toward the
past is also ruled by the dynamical law, that is, whether it is also a time evolution. If
the answer is positive, the law is time-reversal invariant, if the answer is negative, the
law is not time-reversal invariant. Therefore, supposing from the very beginning that
any time displacement toward the past is a dynamical evolution amounts to putting
the cart before the horse.
But, then, what do Eq. (14.21) and Eq. (14.22) mean? Actually, those equations
express the conditions that define what can be called motion reversal:
motion reversal evolution by Δt motion reversal evolution by Δt = identity
motion reversal evolution by Δt motion reversal = evolution by Δt
To put it precisely, the motion-reversal operator is the operator that reverses the
direction of a lawful motion of the system so as to obtain another lawful motion.
Then, the argument that, starting by Eq. (14.21), concludes with discarding the
unitary operator T is a proof of the fact that the antiunitary operator T ∗ is the right
motion-reversal operator for quantum mechanics.
Even though the difference between motion reversal and time reversal has not
been sufficiently stressed, it is acknowledged by some authors. For example,
Ballentine clearly states:
In the first place, the term “time reversal” is misleading, and the operation that is the
subject of this section would be more accurately described as motion reversal. We shall
continue to use the traditional but less accurate expression “time reversal”, because it is so
firmly entrenched.
(Ballentine 1998: 377; italics in original)
Sakurai also emphasizes the point just at the beginning of the section devoted to
time reversal:
In this section we study another discrete symmetry operator, called time reversal. This is a
difficult topic for the novice, partly because the term time reversal is a misnomer; it reminds
us of science fiction. Actually what we do in this section can be more appropriately
characterized by the term reversal of motion. Indeed, that is the terminology used by
E. Wigner, who formulated time reversal in a very fundamental paper written in 1932.
(Sakurai 1994: 266; bold and italics in original)
Summing up, it is quite clear that the antiunitary operator T ∗ is the motion-reversal
operator in quantum mechanics. But the initial question still remains: which is the
right quantum time-reversal operator?
14.5 Conclusions
In this chapter we have focused on the spacetime symmetries of quantum mechan-
ics under the assumption that exploring the meaning of those symmetries is
relevant to the interpretation of the theory.
In the first part, we have considered the behavior of nonrelativistic quantum
mechanics under the Galilean group. We have shown that the Schrödinger equa-
tion is always covariant under the Galilean group, but its Galilean invariance can
only be guaranteed when it is applied to a closed system free from external fields.
We have also discussed the relevance of symmetries to interpretation; in particular,
any realist interpretation that intends to select a Galilean-invariant set of definite-
valued observables should make that set to depend on the Casimir operators of the
Galilean group, since they are invariant under all the transformations of the group.
In future works, these conclusions can be extended in two senses. On the one hand,
they can be transferred to quantum field theory by changing the symmetry group
accordingly: The definite-valued observables of a system in quantum field theory
would be those represented by the Casimir operators of the Poincaré group. Since
the mass operator M and the squared-spin operator S2 are the only Casimir
operators of the Poincaré group, they would always represent definite-valued
observables, a view that stands in agreement with a usual physical assumption in
quantum field theory. On the other hand, if invariance is a mark of objectivity,
there is no reason to focus only on spacetime global symmetries. Internal or gauge
symmetries should also be considered as relevant in the definition of objectivity
and, as a consequence, in the identification of the definite-valued observables of
the system.
In the second part of the chapter, we have carefully disentangled the different
notions involved in the issue of the time-reversal invariance of the Schrödinger
equation. We have assessed the usual claim about the matter, according to which
the Schrödinger equation is time-reversal invariant and the quantum time-reversal
operator is antiunitary. We have argued that the antiunitary operator is actually a
motion-reversal operator and that the question about the right time-reversal oper-
ator in quantum mechanics is still an open question. Those who think that time is
ontologically independent of and prior to the processes in it will stress the
difference between time reversal and motion reversal and, consequently, may tend
to prefer a time-reversal operator that only flips the direction of time. Others, by
contrast, may claim that the very concept of time as independent of motion has no
meaning. From this relationalist-like view, distinguishing between time reversal
and motion reversal as different operations makes no sense and, as a consequence,
the right time-reversal operator is necessarily a motion-reversal operator. This
shows that the question about the time-reversal invariance of quantum mechanics
involves deep issues about the very nature of time. But the further development of
this aspect of the problem will be the subject of future work.
Acknowledgments
We want to thank the participants of the workshop Identity, indistinguishability
and non-locality in quantum physics (Buenos Aires, June 2017) for their contribu-
tion to a philosophically exciting and fruitful time. This work was made possible
through the support of Grant 57919 from the John Templeton Foundation and
Grant PICT-2014–2812 from the National Agency of Scientific and Technological
Promotion of Argentina.
References
Albert, D. (2000). Time and Chance. Cambridge, MA: Harvard University Press.
Ardenghi, J. S., Castagnino, M., and Lombardi, O. (2009). “Quantum mechanics: Modal
interpretation and Galilean transformations,” Foundations of Physics, 39: 1023–1045.
Arntzenius, F. (1997). “Mirrors and the direction of time,” Philosophy of Science, 64: 213–222.
Auyang, S. Y. (1995). How Is Quantum Field Theory Possible?. Oxford: Oxford Univer-
sity Press.
Baker, D. (2010). “Symmetry and the metaphysics of physics,” Philosophy Compass, 5:
1157–1166.
Ballentine, L. (1998). Quantum Mechanics: A Modern Development. Singapore: World
Scientific.
Bohm, A. and Gadella, M. (1989). Dirac Kets, Gamow Vectors and Gel’fand Triplets.
New York: Springer.
Born, M. (1953). “Physical reality,” Philosophical Quarterly, 3: 139–149.
Bose, S. K. (1995). “The Galilean group in 2+1 space-times and its central extension,”
Communications in Mathematical Physics, 169: 385–395.
Brading, K. and Castellani, E. (2007). “Symmetries and invariances in classical physics,”
pp. 1331–1367 in J. Butterfield and J. Earman (eds.), Philosophy of Physics: Part B.
Amsterdam: Elsevier.
Brown, H. (1999). “Aspects of objectivity in quantum mechanics,” pp. 45–70 in
J. Butterfield and C. Pagonis (eds.), From Physics to Philosophy. Cambridge: Cam-
bridge University Press.
Brown, H. and Holland P. (1999). “The Galilean covariance of quantum mechanics in the
case of external fields,” American Journal of Physics, 67: 204–214.
Brown, H., Suárez, M., and Bacciagaluppi, G. (1998). “Are ‘sharp values’ of observables
always objective elements of reality?”, pp. 289–306 in D. Dieks and P. E. Vermaas.
The Modal Interpretation of Quantum Mechanics. Dordrecht: Kluwer Academic
Publishers.
Butterfield, J. (2007). “On symplectic reduction in classical mechanics,” pp. 1–131 in
J. Butterfield and J. Earman (eds.), Philosophy of Physics: Part A. Amsterdam: Elsevier.
Callender, C. (2000). “Is time ‘handed’ in a quantum world?”, Proceedings of the
Aristotelian Society, 100: 247–269.
Colussi, V. and Wickramasekara, S. (2008). “Galilean and U(1)-gauge symmetry of the
Schrödinger field,” Annals of Physics, 323: 3020–3036.
Costa de Beauregard, O. (1980). “CPT invariance and interpretation of quantum mechan-

ics,” Foundations of Physics, 10: 513–530.
Dürr, D. and Teufel, S. (2009). Bohmian Mechanics: The Physics and Mathematics of
Quantum Theory. Dordrecht: Springer.
Earman, J. (1974). “An attempt to add a little direction to ‘The Problem of the Direction of
Time’”, Philosophy of Science, 41: 15–47.
Earman, J. (2004). “Laws, symmetry, and symmetry breaking: Invariance, conservation
principles, and objectivity,” Philosophy of Science, 71: 1227–1241.
Gasiorowicz, S. (1966). Elementary Particle Physics. New York: John Wiley and Sons.
Kramer, E. (1970). The Nature and Growth of Modern Mathematics. Princeton: Princeton
University Press.
Laue, H. (1996). “Space and time translations commute, don’t they?”, American Journal of
Physics, 64: 1203–1205.
tum mechanics,” Studies in History and Philosophy of Modern Physics, 39: 380–443
Minkowski, H. (1923). “Space and time,” pp. 75–91 in W. Perrett and G. B. Jeffrey (eds.),
The Principle of Relativity: A Collection of Original Memoirs on the Special and
General Theory of Relativity. New York: Dover.
North, J. (2009). “Two views on time reversal,” Philosophy of Science, 75: 201–223.
Nozick, R. (2001). Invariances: The Structure of the Objective World. Harvard: Harvard
University Press.
Ohanian, H. and Ruffini, R. (1994). Gravitation and Spacetime, 2nd edition. London:
W. W. Norton and Company.
Peterson, D. (2015). “Prospect for a new account of time reversal,” Studies in History and
Philosophy of Modern Physics, 49: 42–56.
Racah, G. (1937). “Sulla simmetria tra particelle e antiparticelle,” Il Nuovo Cimento, 14:
322–326.
Roberts, B. (2017). “Three myths about time reversal invariance,” Philosophy of Science,
84: 315–334.
Sachs, R. (1987). The Physics of Time Reversal. London: University Chicago Press.
Sakurai, J. J. (1994). Modern Quantum Mechanics. Revised Edition. Reading, MA:
Addison-Wesley.
Savitt, S. (1996). “The direction of time,” The British Journal for the Philosophy of
Science, 47: 347–370.
Sklar, L. (1974). Space, Time and Spacetime. Berkeley-Los Angeles: University of
California Press.
Suppes, P. (2000). “Invariance, symmetry and meaning,” Foundations of Physics, 30:
1569–1585.
Watanabe, S. (1955). “Symmetry of physical laws. Part 3: Prediction and retrodiction,”
Reviews of Modern Physics, 27: 179–186.
Weinberg, S. (1995). The Quantum Theory of Fields, Volume I: Foundations. Cambridge:
Weyl, H. (1952). Symmetry. Princeton: Princeton University Press.
Wigner, E. P. (1931/1959). Group Theory and Its Application to the Quantum Mechanics
of Atomic Spectra. New York: Academic Press.
15
Symmetry, Structure, and Emergent Subsystems
nathan harshman
15.1 Introduction
One of the most evocative results in the whole history of mathematical physics is
that there are exactly five polyhedra with perfect symmetry, i.e., all faces, edges,
and vertices are congruent. Starting from the intuitive and reasonable definitions
and axioms of Euclidean geometry and applying the constraint of symmetry, these
five structures are logically inevitable. Book 13, the climax of Euclid’s Elements,
constructs these solids and proves they exhaust the possibilities. While the beauty
of symmetric geometry provokes attention, I believe it is as much their “fiveness”
that inspires. Plato, Kepler, and others sought five-fold explanations for the
physical structures of the world (see Wilczek 2015). Those chimerical hopes
notwithstanding, Platonic solids do appear throughout science and technology
for natural and practical reasons: They are fundamental structures.
Starting in the nineteenth century, combining generalized notions of geometry
and symmetry yielded a bounty of new structures. In turn, these structures pro-
vided a framework for interpreting, classifying, and generating mathematical
models of physical reality. Glossing over the technical details (upon which hang
critical distinctions and academic careers), the same structural system that classifies
Platonic solids and their generalizations in any dimension also classifies Lie
groups, dynamical catastrophes, symmetric manifolds, random matrices, topo-
logical insulators, gauge quantum field theories, and so on. The imposition of
symmetry on vector spaces is surprisingly rigid, often giving finite or countable
possibilities. And then, more often than not, it seems we find these possibilities
manifest in our models of systems and dynamics. This coincidence of logically
inevitable mathematical structures with elements of physical reality remains as
seductive now as it did to Plato and Kepler. Some theoretical physicists spend their
career chasing these beautifully symmetric models and enumerating their qualities,
hoping one day the model will play an important role in the next “paradigm shift.”
294
Symmetry, Structure, and Emergent Subsystems 295
My proposal is more modest. Symmetry is not the answer to every question, and
the universe may have contingent features that no model can predict or explain.
However, symmetries seem to exist in reality and certainly exist in many effective
and productive models of reality. Within a framework like quantum mechanics, the
presence of symmetry entails the existence of mathematical structures that are
privileged by their relation to the symmetry. The focus of this chapter is the
particular structures called irreducible representations of symmetry groups. Similar
to how a Hamiltonian gives a spectrum in quantum mechanics, the irreducible
representations (irreps) of a group of symmetry transformations form a kind of
basis for possible manifestations of a symmetry on vector spaces. Like the Platonic
solids, these irreps are the denumerable “atoms” of the Hilbert spaces of quantum
mechanical models with symmetry.
For example, any quantum model with special unitary, or SU(2) symmetry has a
description in terms of units that are often called spins, even when they have no
rotational origin. These spin units are Hilbert subspaces that carry irreps of SU(2),
and the total Hilbert space is reducible into products and sums of these “atoms”
of spin.
Similarly, quantum field theory is built by associating free particles to irreps of
the Poincaré group of symmetry transformations of Minkowski spacetime (see, for
example, Streater and Wightman 1964, Haag 1992, Weinberg 1995). As Eugene
Wigner (1939) discovered, the irreps of the Poincaré group are conveniently
labeled by three invariants: mass, internal energy, and spin. Irreps of the Poincaré
group are equivalent to relativistic particles for most technical and interpretational
purposes. A similar construction with the Galilean group and nonrelativistic
particles is discussed later.
Let us recall that, technically, relativistic particles correspond to irreps of the
universal covering group of the Poincaré group. This is necessary to account for
the fact that, in quantum mechanics, we are typically interested in projective
representations, i.e., representations up to a phase. Projective representations are
more general than precise representations, and they preserve the interpretation of
pure states as rays in the Hilbert space. This technicality is mostly swept under the
rug for the rest of this chapter.
These examples make it clear that using irreps to analyze symmetry in quantum
mechanics is an old story, full of many bold successes and productive technical
details. Why rehash it here? The purpose of this contributed chapter is to explore
the connections between the mathematical units of symmetry embodied by irreps,
arguably the “most inevitable” symmetric structures of quantum mechanics, and
the conceptual units of reality that form the basis for interpretation of quantum
theories. Since irreps are symmetric structures that have the appealing properties of
being denumerable, they hold the same appeal as Platonic “fiveness.” Their logical
296 Nathan Harshman
inevitability (given the standard formulation of quantum mechanics) makes them

the natural vocabulary for asking and answering questions about the fundamental
nature of quantum reality, whether a more epistemic or ontic interpretation is
advanced. At the very least, any interpretation of quantum mechanics must provide
some justification for the “unreasonable effectiveness” of these mathematical
structures as conceptual units. At the same time, the conceptual overreach perpet-
rated by natural philosophers enamored by the Platonic solids should serve as a
cautionary tale.
With this motivation, Section 15.2 gives the local definitions of quantum
mechanical models and symmetries. Section 15.3 and Section 15.4 subject the
reader to a technical presentation about “Hilbert space arithmetic”. The Hilbert
space can be built “bottom-up” out of products and sums of irrep spaces, or
reduced “top-down” into products and sums of subspaces that carry irreps.
I argue that Hilbert space arithmetic, and the identification of how symmetry
manifests itself in that arithmetic, provide a lens for understanding quantum
concepts like energy levels, entanglement, locality (and its generalization, specifi-
city), distinguishability, and emergence.
But the view through that lens is not yet clear, so I also offer to the skeptical
reader an application of symmetrized Hilbert space arithmetic. In Section 15.5,
I investigate whether the mathematical structures of irreps have interpretative
power when considered as conceptual units of reality within a model of quantum
few-body systems in one dimension. This choice is partially due to the charming
coincidence that such systems can carry realizations of the same symmetries as the
Platonic solids and their generalizations. More meaningfully, quantum few-body
systems in one dimension are at the knife’s edge in terms of dynamical regimes for
closed quantum systems. Depending on the symmetry of the Hamiltonian, such
systems can have dynamics that are regular, integrable, and solvable or irregular,
chaotic, and ill-conditioned. I argue that irreps form conceptual units to interpret
this rich physics of few-body systems.
The stakes of this analysis have been raised by recent experiments with ultracold
atoms in effectively one-dimensional optical traps (see Serwane et al. 2011, Zürn
et al. 2012, 2013, Wenz et al. 2013). These offer the possibility of implementing
controllable few-body models, and they provide a relevant milieu for evaluating
claims of interpretative utility of irreps as conceptual units. Because irreps are
derived from symmetries of the Hamiltonian of a few-body model, they can be
used to find useful observables to describe the collective degrees of freedom of
few-body systems (Harshman 2016a, 2016b). Further, the connections between
reducibility into irreps and solvability, while not completely clear, can still be
productive in the quest to identify new solvable models in few-body physics (see
Harshman 2017a, Andersen, Harshman, and Zinner 2017 for recent examples).
Finally, the connections between solvability and controllability provide opportun-

ities for technological applications with ultracold atoms. Controllability is a key
requirement for the relativity of entanglement pioneered by Zanardi (see Section
15.4). As a practical concern for developing quantum information processing
devices, it motivates the search for emergent subsystems (Zanardi and Lloyd
2004).
15.2 Quantum Mechanical Models and Symmetry

Symmetries are often invoked as grand, unifying principles at the foundations of
quantum mechanics. For example, the standard model of particle physics is
presented as though it emerges as a logical consequence of combining the sym-
metries of Minkowski spacetime with gauge symmetries of internal flavors. Such a
presentation is an exaggeration, as the numerous unexplained free parameters in
the standard model attest. Here, I want to present symmetries in the narrower
scope – symmetries as defined within the context of a specific quantum
mechanical model.
15.2.1 Model Definition and Scope

Following Haag (1992), I define a quantum mechanical model as a set of observ-
ables A that are represented by Hermitian operators acting on a Hilbert space H .
This rather airless definition of quantum mechanical model makes minimal onto-
logical claims beyond the standard framework of quantum mechanics: Pure states
are unit vectors in H , measurements are expectation values of observables, Born
Rule, and all that. The origin of the observables and their representation within the
model could be established by scientific utility alone, such as making predictions,
or by other epistemic concerns. Using this definition of model, I am consciously
attempting to restrict my ontological commitment (as much as possible) to a
structure of relatively defined mathematical objects while still doing “normal”
quantum mechanics. To those who have ontological commitments in their inter-
pretations of quantum mechanics, I believe that this framework for models is
general enough to encompass those theories, and so I ask indulgence in this
exercise in structuralism.
The simplest nontrivial example of a model is given by the algebra of Pauli
matrices acting on H C2 . Any two-level system can be described in this
framework, and although there is an analogy to a spin-1/2 system, the model
makes no ontological commitment to any fundamental particle or other unit of
reality. The model could describe an isolated discrete degree of freedom, like the
isospin of a nucleon, or could emerge as an effective theory, like the lowest two
298 Nathan Harshman
energy levels of a more complicated system. It could also be derived from the
smallest nontrivial projective representation of an underlying three-dimensional
rotation group, i.e., a true spin. In any case, the model as a mathematical structure
can be studied without reference to any physical embodiment, ontological com-
mitment, or larger theoretical framework.
Nonrelativistic few-body models often do start with an ontological commitment,
such as N particles on an underlying d-dimensional
Euclidean space X Rd . The
Hilbert space is realized as H L R , Lebesgue square-integrable functions
2 Nd
on the configuration space X N constructed from the N-fold Cartesian product of

single-particle spaces. The corresponding set of observables is generated by the
Heisenberg algebra of 2Nd canonical position and momentum operators which are
represented as variable multiplication and differentiation operators on L2 RNd .
There are certainly technical demons (like unbounded operators) hidden in this
model, but for now let us assume they can be exorcised.
The subsequent analysis presumes the existence of a model, whether it has been
defined in the abstract or derived from underlying particles or other units of reality.
The goal is then to use symmetry to identify structures within the model that aid
interpretation of empirical phenomena without making further commitments. Note
that in both of the examples presented earlier, the set of observables was actually
an algebra of observables, closed under addition and multiplication. That will not
always be the case for models I consider. For models with finite-dimensional
Hilbert spaces the distinction is minor, but it is important for infinite-dimensional
spaces.
In order to make progress, I restrict my attention to models with two additional
properties. First, I assume that within the set of observables there is a particular
observable H 2 A called the Hamiltonian. The Hamiltonian generates time trans-
lations by the unitary operator U ðt Þ ¼ exp ðiHt=ℏÞ defined on all of H . This
assumption thereby excludes models that describe systems with time-dependent
Hamiltonians or open dynamics. These are obviously cases of physical interest, but
the complications they introduce require attention to detail beyond the scope of this
chapter.
Second, I assume that the set of observables is represented on a Hilbert space
that is separable in the topological sense. A separable Hilbert space has a countable
orthonormal basis (see Streater and Wightman 1964). Equivalently, it implies the
existence of a decomposition of the Hilbert space H of the system into a direct sum
of a countable (but still possibly infinite) set of one-dimensional subspaces H i :
H ¼ ⊕i H i (15.1)
Restricting to models with separable Hilbert spaces is a convenience that (to my
knowledge) is not much of a restriction for models of quantum mechanical systems
with a finite number of particles. However, there are many-body systems and
quantum field theories where this kind of separability cannot be assumed (again,
see Streater and Wightman 1964).
15.2.2 Model Symmetries

Following Wigner (1959), a symmetry of a model is a group of unitary (and
possibly antiunitary) operators on H . Such operators preserve the magnitude of
inner products and therefore of probabilities. For finite dimensional Hilbert spaces
H CD , the symmetry group of the model must be a subgroup of the unitary
matrices U ðDÞ. For infinite dimensional separable Hilbert spaces, the unitary group
U ð∞Þ can be defined by the inductive limit of U ðDÞ (Olshanski 2003) and any
subgroup of U ð∞Þ could be a symmetry group. But most of these subgroups do not
have any physical interpretation; they are purely formal. So how do we distinguish
and classify useful or meaningful symmetry groups?
Within the context of a formal model, the empirical content of a symmetry is
inferred from its relation to the set of observables, specifically, how the unitary
group of symmetry operators transform the Hilbert-space representations of the
observables. For example, the symmetry identifies invariant operators of the
model, or more generally, can be used to classify operators based on their
transformation properties (Lombardi and Fortin 2015). A familiar example is
classifying operators by their tensor rank under a representation of a matrix group,
e.g., scalar, vector, pseudovector, and so on for the rotation group.
However, starting from a formal model with Hilbert space and algebra and then
identifying meaningful symmetries is not always how physical analysis proceeds.
The logic is sometimes reversed and the model is constructed from the symmetry
of an underlying space or spacetime. The group of unitary operators on the Hilbert
space is a representation of the universal covering group of some spacetime
symmetry. The set of observables is the enveloping algebra built from the gener-
ators of the symmetry representation. Three relevant examples are: (1) spin models
from rotational symmetry; (2) free nonrelativistic particle models from Galilean
symmetry (Lévy-Leblond 1967); and (3) free relativistic particle models from
Poincaré symmetry (Wigner 1939). In each of these cases, the Hilbert space of a
single spin or a single particle corresponds to an irreducible representation space of
the spacetime symmetry group. As mentioned before, technically “particles”
correspond to irreps of the universal covering groups of these spacetime symmet-
ries. This distinction allows for projective representations. Further, in order to get
massive representations, an observable corresponding to mass must be added as a
central element to the Galilean algebra. Bottom-up approaches to quantum
300 Nathan Harshman
mechanics often take these basic models as starting units and build more compli-
cated models out of their irreps spaces and algebras of observables.
In practice, identifying the symmetries of a nonrelativistic few-body model
takes place using a hybrid of formal, top-down and reductionist, bottom-up
approaches. Symmetries in physical space are important, but so are symmetries
in the spaces of configuration space and phase space. Although these auxiliary
spaces can be derived from the bottom-up approach, starting with free particle
models, the symmetries of these derived spaces may not be easily reducible into
free-particle symmetries and their corresponding irreps can describe collective or
emergent degrees of freedom. Additionally, few-body models can have symmet-
ries that are defined by unitary operators on the Hilbert space itself without
reference to a symmetry of an underlying space. For example, these kind of
symmetries are present when there are accidental degeneracies (Harshman 2017a).
15.2.3 Kinematic and Dynamic Symmetries

The symmetry groups with which I am primarily concerned are those represented
by unitary operators that leave the Hamiltonian invariant. Call these kinematic
symmetries of the model. These unitary operators transform stationary states of the
Hamiltonian, i.e., energy eigenstates, into other energy eigenstates with the same
energy. Sometimes instead of considering a single Hamiltonian, it is profitable to
consider the kinematic symmetry of a family of Hamiltonians that (for example)
depends on a parameter or parameters but exist within the same model. This can be
a tricky business, because sometimes as the parameter is varied the Hamiltonian
loses the property of self-adjointness on the original Hilbert space of the model.
A notable example is when parameter variation makes configuration space effect-
ively disconnected (Harshman 2017a).
As opposed to the kinematic symmetry group of a Hamiltonian, a dynamic
symmetry group does not commute with the Hamiltonian as a whole (although
some elements may). This definition encompasses a wide class of possibilities. In
their simplest manifestation, dynamic symmetries provide algebraic relationships
between the group and the Hamiltonian that map energy eigenstates into other
energy eigenstates (like ladder operators) and induce algebraic relationships among
expectation values of noncommuting observables (Wybourne 1974). Poincaré and
Galilean transformations are also dynamic symmetries in this sense; boosts change
the energy of a state in an algebraic way. Dynamic symmetries can also describe
maps among Hamiltonians within a parametrized family, like a scale transform-
ations (Jackiw 1972). They can even serve as maps between models and Hamilto-
nians that on the surface seem radically different, like supersymmetric partner
Hamiltonians in quantum mechanics (see Cooper, Khare, and Sukhatme 1995).
15.3 Hilbert Space Arithmetic: Direct Sum Decompositions

The definition of model provided here seemingly relies on a the Hilbert space as an
essential feature, and this section and the next are going to outline how Hilbert
spaces can be decomposed into direct sums and factored into tensor products using
observables and symmetries as the starting point. Topological separability of the
Hilbert space is assumed throughout this vector space arithmetic, and the goal is to
see how much interpretive power this kind of analysis can provide. Carving models
into subspaces by decomposing into direct sum and into submodels by factoring into
tensor products has a long history in quantum mechanics for practical reasons of
mathematical analysis. I argue that it also forms units of conceptual analysis that are
called different names in different contexts but are all manifestations of the same
underlying structure. The tentative claim is that these structures are as valid as
conceptual units of reality as are the more tangible concepts like “particles.”
However, before proceeding, I must declare that I am not a Hilbert space
fetishist. The specific topology of the Hilbert space is at once too loose and too
restrictive for some purposes. For example, the domains on which the set of
observables are bounded and self-adjoint may not be the entire Hilbert space,
and their eigenspaces may not be contained in the Hilbert space. This technical
aspect can be rigorously handled using Gel’fand triplets and rigged Hilbert spaces
(Bohm and Gadella 1989, Bohm and Harshman 1998), and so for the rest of this
section when I say “Hilbert space,” assume that I am talking about decomposing
and factoring some kind topological vector space with suffciently nice properties.
15.3.1 Hamiltonian-Induced Direct-Sum Decomposition

Observables can be used to partition the Hilbert space into subspaces. Each sub-
space is associated to an eigenvalue of the operator, and on that subspace that
operator acts like a multiple of the identity. The most familiar example is the
Hamiltonian H. Assume for simplicity that the spectrum of the Hamiltonian σðH Þ
is discrete, as it would be for most systems with finite extent. Then the Hilbert
space for the system is decomposed on the spectrum of H as
H ¼ ⊕ HE (15.2)
E2σðH Þ
On the surface, this looks like Eq. (15.1), but here, instead of a decomposition into
one-dimensional spaces, each energy eigenspace H E is realized by a finite-
dimensional space CdðEÞ with dimension equal to the degeneracy d ðE Þ of the
energy E.
If the Hamiltonian under consideration is time-independent and describes a
closed system, then the system has time-translation symmetry and the
302 Nathan Harshman
decomposition Eq. (15.2) has another interpretation in terms of irreps. Time

translation symmetry is an abelian symmetry, i.e., a system translated in time by
intervals t then t 0 is the same as if translated by t 0 then t. Abelian symmetries have
one-dimensional irreps, and the unitary operators are just phases. Different irreps
of time translation are distinguished by a scale-setting parameter (called the
energy) that determines how fast the phase advances:
0 0 0
U ðt 0 Þ U ðtÞ H E ¼ eiEt =ℏ eiEt =ℏ H E ¼ eiEðt þtÞ=ℏ H E ¼ U ðt 0 þ t Þ H E (15.3)
When H E has more than one dimension, it is therefore a sum of multiple time-
translation irrep spaces. This implies the existence of at least one other operator
that commutes with H and can be used to diagonalize the subspace H E . Formally,
one can always construct a single operator on the Hilbert space that commutes with
H and whose eigenvalues uniquely distinguish every vector in every degenerate
space. Such an operator can be chosen as block diagonal, one block for each H E
consisting of a d ðE Þ-dimensional diagonal matrix with, for example, the numbers
1 through dðEÞ on the diagonal. Call this operator D. Then H can be reduced to a
direct sum of one-dimensional subspaces on the joint spectrum of H and D. In this
sense, H and D are a complete set of commuting observables. However, the
observable D is defined by its construction as an operator on H and has no
fundamental origin as, for example, an observable defined by a measuring appar-
atus or the generator of a symmetry transformation. It is an example of a formal
mathematical structure without physical interpretation. One goal, when analyzing a
system, is to find operators that perform the same diagonalizing function as D, but
have some other physical meaning, perhaps from kinematic symmetries.
For a particular Hamiltonian, the decomposition of H into energy subspaces H E
is fixed. However, consider a family of Hamiltonians from the same model that can
be characterized by a variable parameter g. The subspaces H E may coalesce or split
as g is varied. The energy eigenspaces are called “levels,” and I argue they function
as interpretational units that are treated as “real” objects. One speaks of level
“dynamics,” for example, levels “shifting,” “splitting,” “diverging,” etc., but what
is doing these actions is quite abstract – a subspace built from irreps in a model
with a family of Hamiltonians.
15.3.2 Observable-Induced Direct-Sum Decomposition

Any observable Λ with a discrete spectrum λ 2 σðΛÞ can serve as the origin of a
decomposition like Eq. (15.2), not just the Hamiltonian:
H ¼ ⊕ Hλ (15.4)
λ2σðΛÞ
As before, the dimensionality of H λ is the degeneracy of the eigenvalue λ. To

diagonalize this degeneracy, one can add additional observables and arrive at a set
of k commuting observables Λ ¼ fΛ1 ; . . . ; Λk g with discrete eigenvalues denoted
λ ¼ fλ1 ; . . . ; λk g 2 σðΛÞ. Of course the Hamiltonian could be one of these Λi .
Note that σðΛÞ is the joint spectrum of all k observables. For many models, the
joint spectrum cannot be decomposed into the product of spectra. When it can (see
Harshman 2016c for examples of separable three-body Hamiltonians), then the
decomposition can be further reduced into independent sums
H ¼ ⊕ ... ⊕ Hλ (15.5)
λ1 2σðΛ1 Þ λk 2σðΛk Þ
As discussed in (Harshman 2016c), this kind of spectral separability distinguishes

Hamiltonians with “silver” and “gold” separability from “bronze” separability (this
is a different notion of separability than topological separability; see more discus-
sion in Subsection 15.4.4).
15.3.3 Symmetry-Induced Direct-Sum Decomposition

The two previous subsections used observables to decompose the Hilbert space
into a direct sum of observable eigenspaces. Groups of symmetry transformations
can be used for the same purpose, but now the subspaces are not necessarily
eigenspaces. Instead, the subspaces are irreducible representation spaces, also
known as modules, and carry a linear representation of the transformation group.
A representation that is irreducible means that, within the corresponding module,
there are no invariant subspaces.
For specificity (and to avoid details that are important but technical) consider a
group G that has only finite-dimensional representations. Discrete finite groups,
compact Lie groups, and their combinations have this property. Denote the label
for an irrep by a Greek letter in parentheses like ðμÞ and the corresponding irrep
space with dimension dðμÞ by V ðμÞ CdðμÞ . Then the Hilbert space can be broken
into sectors labeled by ðμÞ:
H ¼ ⊕ H ðμÞ (15.6a)
ðμÞ
Each sector H ðμÞ is a tower of irreps spaces

ðμÞ
H ðμÞ ¼ ⊕ V i (15.6b)
i
This may seem all a bit abstract, so here are a few examples. Perhaps the
simplest and most familiar is the case of parity. Parity is realized by a finite group
of two elements Z 2 . This abelian group has two irreps denoted + for even states
304 Nathan Harshman
under parity and for odd states. So the Hilbert space can be divided into sectors
of even and odd states H ¼ H þ ⊕ H . Because the group Z 2 has only one-
dimensional irreps, nothing more can be said except that H þ and H are built
out of one-dimensional subspaces that are invariant under parity. If parity is a
kinematic symmetry of a model, then all energy eigenstates are in one of those two
sectors and the expectation value of parity is a dynamical invariant. If parity is a
kinematic symmetry of a family of Hamiltonians, then varying the parameters of
the family mixes states within a sector, but not across sectors, i.e., changing the
control parameter does not change the parity of a state.
Another familiar example is rotational symmetry in three dimensions. The
eigenvalue of the operator representing angular momentum squared ℏ2 sðs þ 1Þ
can be used to characterize irreps and irrep spaces. The spin s comes in two infinite
series: nonnegative integers and nonnegative half-integers. A startling fact, called a
superselection rule, is that a decomposition into rotation group irreps only consist
of one of those two types, either integer or half-integer irreps. There can never be a
superposition of states with integer and half-integer total angular momentum. That
means in any quantum model, the Hilbert space only has sectors of integer or half-
integer irrep spaces.
A final example is the symmetry of particle permutations for N identical
particles, realized by the symmetric group SN . Here the irrep decomposition
provides a conceptual unit for analyzing the meaning of identical particles. Irreps
of SN are labeled by positive integer partitions of N. For example, there are four
partitions of N = 4: [4], [31], [22][22], [211][212], and [1111][14]. That
means the Hilbert space for a model with four identical particles can be decom-
posed into five sectors:
H ¼ H ½4 ⊕H ½31 ⊕H ½22 ⊕H ½212 ⊕H ½14 (15.7)
The sectors have transformation properties dependent on their corresponding irrep.

The irrep labeled [4] is the trivial one-dimensional representation in which all
permutations are represented by multiplication by +1. So this sector H ½4 is
appropriate for representing bosons; in fact, it is the full scope of possibilities for
four identical bosons. The irrep labeled [14] is the one-dimensional totally anti-
symmetric representations where odd permutations are represented by 1 and even
permutations by +1. The corresponding sector H ½14 is therefore where the fer-
mions live. The three other sectors become necessary when considering particles
with parastatistics (an exotic generalization between fermions and bosons), or more
prosaically, when considering particles with spin and spatial degrees of freedom,
but fixing or tracing over either the spin or spatial degrees.
A sector like H ½4 or H ½14 is not a monolithic space; it is a tower of irrep spaces.
Alternatively, it can be decomposed into energy subspaces, parity subspaces, etc.
The mathematical structure of decomposition into irreps provides not just a

technical tool for solving problems with identical particles, but also a key concep-
tual unit for a minimal interpretation of what identical particles mean. This is
discussed in more detail later, in the specific discussion of the few-body model.
As noted earlier, the Hamiltonian itself is the generator for time translation
symmetry, so the decomposition into energy subspaces Eq. (15.2) is also an example
of the irrep decomposition Eq. (15.6). Kinematic symmetries add more structure.
Then each energy eigenspace is a sum of irrep spaces of the kinematic symmetry:
ðμÞ
H E ¼ ⊕ V E, i (15.8)
ðμÞi
A familiar example is the reduction of hydrogen energy levels into subspaces with
fixed orbital angular momentum, doubled by the presence of spin. In principle, the
direct sum in Eq. (15.8) extends over all irreps of the kinematic group and there can
be multiple copies of the same irrep. However, when there are multiple irreps with
multiplicity, that usually signifies the presence of additional kinematic symmetries.
In the hydrogen atom example, there is the larger kinematic symmetry group SO(4)
that explains why subspaces with different orbital angular momentum have the
same energy.
Define the maximal kinematic symmetry group GH of the Hamiltonian H as the
group such that every H E corresponds to a single irrep ðμÞ of GH . When this group
can be found, energy levels are irreps of the maximal kinematic symmetry group,
and this is a powerful tool for the analysis of the model. It allows the physics of
degeneracy to be handled in a systematic, algebraic fashion because the symmetry
group provides all invariant observables necessary to diagonalize degeneracies.
Other observables in the model can be characterized by their transformation
properties under the group, simplifying calculations of expectation values, transi-
tion rates, and perturbation theory. Further, if H is part of a family of Hamiltonians,
then how GH changes with varying parameters determines how the energy levels
(irreps) split and merge and how invariants are broken and reformed.
To close this section on decomposition, consider a dynamic symmetry group G
with irrep labeled by ðμÞ representations. Since the symmetry group does not
commute with the Hamiltonian, there is no necessary relationship between irreps
of G and H E . However, one possibility is that each irrep of G is decomposable in a
sum of energy eigenspaces, i.e., the reversal of Eq. (15.8):
ðμÞ
Vi ¼ ⊕ HE (15.9)
ðμÞ
E2σi ðH Þ
ðμÞ
where σi ðH Þ is a purely symbolic shorthand for the spectrum of energies E
ðμÞ
corresponding to the irrep V i and depends on the Hamiltonian H and the
306 Nathan Harshman
symmetry group G in a model-specific way. In this case, G is called a spectrum-

generating group for the Hamiltonian H (see Wybourne 1974).
15.4 Hilbert Space Arithmetic: Tensor Product Factorizations

The other method of partitioning a model is through factoring the model Hilbert
space into a tensor product structure. Any finite-dimensional Hilbert space
H Cd can be factorized into a k-fold tensor product
H ¼ H 1 ⊗H 2 ⊗ . . . ⊗H k Cd1 ⊗Cd2 ⊗ . . . ⊗Cdk (15.10)

as long as d 1 d2 . . . d k is a factorization of the positive integer d. Such a
factorization is not unique; any unitary matrix in U ðd Þ that cannot be factorized
into U ðd 1 Þ U ðd 2 Þ . . . U ðd k Þ defines another factorization with the same
structure but different subspaces H i . The situation with a separable but infinite-
dimensional Hilbert space is even wilder, where at least formally, Hilbert sub-
spaces of any finite dimension can be factored off willy-nilly.
However, as with the generic decomposition Eq. (15.1), such factorizations do
not necessarily have any physical meaning, even within the limited ontology of a
formal model. It is the allowed set of observables that distinguishes which factor-
izations have functional, conceptual, or interpretational value. In this section,
several methods are presented in which a factorization of a model Hilbert space
has an operational meaning in terms of observables.
15.4.1 Models of Independent, Distinguishable Subsystems

The most transparent case is when a model describes a system that is composed of
denumerable, independent, distinguishable systems. This is the bottom-up
approach to building a model. Consider each of these systems as a submodel with
a Hilbert space H i and set of observables Ai . The total Hilbert space is the tensor
product of k subspaces H i
k
H ¼ ⊗ Hi (15.11)
i¼1
Pure states in H can be classified as to whether they are separable or not separable
with respect to this factorization. In this context, separable is used in the algebraic
sense that a separable pure state jψi can be written at the tensor product of states
jψi i 2 H i as
jψi ¼ jψ1 i⊗jψ2 i⊗ . . . ⊗jψk i (15.12)
An entangled pure state is not separable and does not admit a factorization like
Eq. (15.12). To be clear, this is separable in a totally different sense than the
topological notion of a separable Hilbert space admitting a decomposition into

denumerable one-dimensional subspaces Eq. (15.1), and it is also not equivalent to
the separability of a differential equation discussed later.
The set of observables denoted A⊕ is constructed using the Kronecker sum of
the subsets of observables Ai . The Kronecker sum of two operators in different
algebras A 2 A1 and B 2 A2 is
A⊕B ¼ A⊗I2 þ I1 ⊗B (15.13)
where I1 is the identity operator in H i . This definition of the Kronecker sum can
be generalized to more factors. (Unfortunately, note that the same symbol is
typically used for the direct sum of vector spaces and the Kronecker sum of
operators.) For observables in A⊕ , expectation values of measurements factor
into a sum of expectation values in each subsystem, and this holds for both
separable and entangled pure states. I argue this can be taken as the definition of
a model of independent systems. Note that A⊕ is not an algebra of observables,
even if each Ai were algebras. The set A⊕ is closed under sums, but not under
products.
Technically, a distinction is necessary between the submodel observables Ai
and the representation of those observables in A⊕ . The representation of Ai in A⊕
is found by taking the Kronecker product of elements of Ai with the identity
elements in all other subspaces H i . In other words, the element A 2 A1 is mapped
into the element A⊗I2 ⊗ . . . ⊗Ik in the set A⊕ . It is in this sense that we can say
that the subsets of operators in the model commute with each other, another
indication of their independence in this model.
Within each submodel there is a Hamiltonian H i 2 Ai . Considered as operators
of the total space H , the sub-Hamiltonian H 1 is realized by H 1 ⊗I2 ⊗ . . . ⊗Ik and so
on for all k sub-Hamiltonians. The total Hamiltonian
k
H ¼ ⊕ Hi (15.14)
i¼1
describes a system composed of subsystems that are noninteracting. When H is

exponentiated exp ðiHt=ℏÞ to generate time translations, it factors into a product
of unitary operators on each subspace H i
U ðt Þ ¼ exp ðiH 1 t=ℏÞ⊗ exp ðiH 2 t=ℏÞ⊗ . . . ⊗ exp ðiH k t=ℏÞ (15.15)
Because time translation factors into operators that do not mix subspaces, entangle-
ment is time-invariant. The only entanglement in the system is present in the initial
state and remains unchanged. This definition can be lifted to describe the tensor
product structure in Eq. (15.11) – it is a dynamically invariant tensor product
structure (see Harshman and Wickramasekara 2007a).
308 Nathan Harshman
One physical interpretation of such a model is that the subsystems are on

isolated patches of space. For example, in this case each Hilbert subspace H i
might be a spin (i.e., an irrep of the rotation group). Even if these subsystems are
internally identical, they can be distinguished by their location. In this sense, the
factorization of the tensor product corresponds to the intuitive notion of locality,
the observables in A⊕ are local observables, and the unitary operator is a local
unitary. However, the subsystems could just as well be in the same location, but
noninteracting and distinguishable. In this case, one could still use the term locality
to refer to operators in A⊕ that are local with respect to the tensor product of the
structure. In fact, following Zanardi (see later), I have used this generalization of
the term locality in numerous talks, including at the workshop that inspired this
chapter, and received angry rebukes. A significant portion of the audience always
seems to prefer that local retain its original meaning in terms of space (or space-
time for relativistic systems). In response to the persistent headwind I have finally
conceded and tentatively propose the terms specific and specificity to replace local
and locality in this context.
15.4.2 Models of Interacting Subsystems

A standard assumption of quantum mechanics is that the model for an interacting
system can be built from the models of the subsystems. The construction of the
Hilbert space by tensor product is the same as before, and therewith follow the
same notions of specificity and entanglement among subsystems. The difference is
now that the set of observables is extended to include operators not in A⊕ . In the
most extreme case, the total set of observables could be the algebra of observables
is constructed as the tensor product of the subsets Ai as
k
A ¼ ⊗ Ai (15.16)
i¼1
More generally, the set A includes at least some observables that are not specific to
the tensor product structure induced by the subsystems.
For an interacting system, the Hamiltonian must be nonspecific, i.e., an operator
that cannot be constructed by Kronecker products of sub-Hamiltonians, like Eq.
(15.14). As a result, time evolution no longer factors into a specific unitary
operator like Eq. (15.15). Entanglement of a state evolving in time is typically
no longer a dynamical invariant with respect to the tensor product structure Eq.
(15.11). However, there may be other observables that are dynamical invariants.
There may even be other tensor product structures besides the original construction
that are dynamically invariant. The question becomes: Can one exploit these
observables to find an alternate factorization and an alternate notion of specificity?
15.4.3 Zanardi’s Theorem and Virtual Subsystems

For systems with finite-dimensional Hilbert spaces, one step towards answering
this question is addressed by Zanardi’s theorem (Zanardi 2001, Zanardi, Lidar, and
Lloyd 2004; see also Harshman and Ranade 2011). It provides criteria for whether
a partition of the observables leads to a tensor product structure and a notion of
specificity.
A version of the theorem can be stated as follows:
Given a state space Φ 2 H Cd and a collection of subalgebras fA1 ; A2 ; . . .g of the total
algebra of observables A acting on H , the subalgebras induce a tensor product structure
H ¼ ⊗i H i if they satisfy the following criteria:
• Subsystem independence: The subalgebras commute ½A1 ; A2 ¼ 0 for all i, j.
• Completeness: The subalgebras generate the total algebra of observables A ¼ ⊗i Ai .
• Specific (née local) accessibility: Each subalgebra corresponds to a set of controllable
observables.
In this statement, the first two requirements on the subalgebras of observables are
mathematical in nature, and they could be assessed from within the model as true
or false for any particular partition of the observables. However, the third require-
ment is a physical criterion about empirical accessibility of measurement and
control. There could be partitions of the observables that satisfy the first two, but
are inadmissible based on a physical limitation of reality or some other constraint
from outside the model.
An extension of Zanardi’s theorem, called the “tailored observables theorem,”
demonstrates the flexibility provided by the first two requirements in constructing
subalgebras that factor the Hilbert space into “virtual” subsystems. For a finite-
dimensional Hilbert space H Cd , one can construct subalgebras of observables
that induce a tensor product structure from a finite basis of operators, such that any
known pure state can have any entanglement that is possible for any factorization of d
(Harshman and Ranade 2011). The proof relies on the unitary equivalence of Hilbert
spaces with the same dimension, and it is constructive in the sense that a procedure is
given to construct the generators for the subalgebras in a finite number of steps
(depending on the factorization of d). A consequence of this theorem is that, for any
pure state, observables can be found that will detect as much or as little entanglement
as is possible in a Hilbert space with dimension d. The only hitch is that entanglement
is completely relative only when the control of the system is absolute, and therefore the
third criterion of Zanardi’s theorem is unrestrictive. For example, in a system of linear
quantum optics (i.e., using only mirrors, phase shifters, and beam splitters) any finite-
dimensional unitary operator can be implemented (Reck et al. 1994). Combined with
Mach-Zehnder interferometers, this system has enough control to extract any
observer-relevant entanglement from any pure state.
310 Nathan Harshman
15.4.4 Top-Down Approach to Emergent Tensor Product Structures

As stated and proved, Zanardi’s theorem and the tailored observables theorem hold
for finite-dimensional Hilbert spaces and algebras of observables. It should have
generalizations to separable Hilbert spaces and more general sets of observables,
but I am unaware of any work in that direction.
Nonetheless, there are other cases of greater scope where partitions of observ-
ables provided by extra-model considerations lead to novel tensor product
structures. Decoherence can select preferred tensor structures and thereby notions
of subsystems (Jeknić-Dugić, Arsenijević, and Dugić 2013). This approach to
building the classical-quantum correspondence using decoherence and emergent
tensor product structures has been referred to as the “top-down approach” (Fortin
and Lombardi 2016).
Another category of top-down virtual subsystems is provided by yet another
notion of separability – separation of variables and separation of integration
constants. The Hamiltonian can be represented as the Schrödinger operator for
models with an underlying D-dimensional space or configuration space. This kind
of separability describes the existence of an orthogonal coordinate system that
totally separates the Schrödinger equation into D one-dimensional differential
equations (or partially separates it into d < D differential equations). There are
different levels of separability depending on how the separated differential equa-
tions depend on the separation constants.
The “gold standard” is when each differential equation only depends on a single
separation constant λi (Harshman 2016c). Then each differential operator Λi defines a
Zanardi-like subsystem and the whole Hilbert space can be decomposed as
H ¼ ⊕ . . . ⊕ H λ1 ⊗ . . . ⊗H λD (15.17)
σ ðλ 1 Þ σðλD Þ
where σðλi Þ is the spectrum of the differential operator Λi . This is a more robust
separability than “silver” separability described in Eq. (15.5), where only the
spectrum was separable, not the Hilbert space. For “bronze” separability, the
spectrum of each differential operator depends on the values of other separation
constants, and so spectral separability is lost. Only for “gold” separability does
differential separability correspond to a tensor product structure and therewith to
an algebraic notion of separability.
A final method for identifying top-down tensor product structure is requiring
that the tensor product structure be invariant with respect to a symmetry group of
the model. A general theory of when this is possible has not been developed, but
two examples from nonrelativistic physics illustrate the idea.
Consider a model whose Hilbert space is an irrep space of the Galilean group
and whose algebra of observables is the Galilean algebra extended by a central
mass observable, i.e., a model of a single nonrelativistic particle (Lévy-Leblond

1967). Elements of this irrep space are wave functions in three-dimensional space
with an internal spin degree of freedom. The Hilbert space has a natural
factorization

H ¼ H space ⊗H spin L2 R3 ⊗C2sþ1 (15.18)
Because it is an irrep of the Galilean group, no vector in this space is invariant

under all transformations. However, this factorization of spatial and spin degrees of
freedom is invariant. The unitary operator that represents any Galilean transform-
ation factors into a product of a specific unitary operator on each subspace (Harsh-
man and Wickramasekara 2007b). In contrast, irrep spaces of Poincaré
transformations, corresponding to relativistic particles, do not have an invariant
factorization between spatial and spin degrees of freedom, although they may
under a subgroup of transformations (Harshman 2005).
Another example is a model built from two Galilean irreps H ¼ H 1 ⊗H 0 2 and a
Hamiltonian with interactions that is not specific to that factorization, but for which
the center-of-mass degrees of freedom still separate (in the sense of a differential
equation). Then there is an alternate tensor product structure
H ¼ H com ⊗H rel (15.19)
between the center-of-mass and relative degrees of freedom. This tensor product
structure is symmetry-invariant with respect to Galilean transformation, and fur-
ther, is it a dynamically-invariant tensor product structure. The unitary operators
for a general Galilean transformation and for time-translation with the interacting
Hamiltonian both factor into specific unitaries (Harshman and Wickramasekara
2007a). As a result, entanglement between center-of-mass and relative degrees of
freedom is invariant under transformation of coordinate systems and invariant in
time. This kind of entanglement between center-of-mass and relative degrees of
freedom is not peculiar; it is present even in typical initial states of a scattering
experiment where the particles are not originally entangled with respect to the
interparticle tensor product structure H ¼ H 1 ⊗H 2 .
15.5 Context: The Simplest Quantum Few-Body Problem

The purpose of this section is to apply the techniques of Hilbert space decom-
position and factorization to a specific quantum mechanical model describing a few
interacting particles in one dimension. I consider distinguishable and indistinguish-
able particles without internal degrees of freedom like spin (or if they have spin,
the spin is fixed and not dynamical). It is arguably the simplest quantum model
that exhibits the full range of complexity of quantum dynamical systems.
312 Nathan Harshman
Understanding the model requires assessing the interplay of interaction,

indistinguishability, identity, integrability, solvability, and entanglement.
A consistent, coherent synthesis of these issues provides a challenge for any
interpretation of quantum mechanics in terms of units of reality. I claim that
combining the bottom-up and top-down structural perspectives of model symmet-
ries and Hilbert space arithmetic responds to this challenge with surprising depth.
15.5.1 Single-Particle Hamiltonian

A first step in building the interacting few-body model is describing the single-
particle model. The Hamiltonian for a single particle in one dimension experi-
encing a static externally generated potential:
1 ∂2
H1 ¼ þ V ðxÞ (15.20)
2m ∂x2
Here is the first place the restriction to one-dimensional systems pays dividends. First,
one-dimensional systems are always integrable. An integrable system has as many
algebraically independent, globally defined constants of the motion as the number of
degrees of freedom. (The classical definition of integrability is usually formulated in
term of operators generating flows on phase space. In a classical one-dimensional
system, the constraint provided by this integral of motion reduces the two-dimensional
phase space to a one-dimensional manifold, i.e., the trajectory of the particle. There is
some ambiguity in the quantum case. See Caux and Mossel 2011 for a review of the
difficulties). For a one-dimensional system, the Hamiltonian itself is the single
conserved integral of motion necessary for integrability. This does not mean that the
system is necessarily solvable, in the sense that the eigenvalues and eigenstates of the
Hamiltonian can be expressed in closed-form analytic expressions. However, for
moderately well-behaved trapping potentials (Harshman 2017a), the Sturm-Liouville
theory guarantees that the energy spectrum of Eq. (15.20) is a denumerable tower of
singly-degenerate states bounded from below. The wave function with lowest energy
ε0 has no nodes, and each successive state with energy εn has n nodes.
Again assuming a reasonable trapping potential, the Hilbert space of the one-particle
system H 1 is the space of Lebesgue-square-integrable functions L2 ðRÞ on the real line.
This space carries an irrep of the one-dimensional Galilean group, although the
trapping potential breaks that symmetry. Each energy eigenstate jεn i spans a one-
dimensional subspace H 1εn 2 H 1 , leading to the decomposition of H 1 into a direct sum
of energy eigenspaces (or equivalently, time translation symmetry irrep spaces):
∞
H 1 ¼ ⊕ H 1εn (15.21)
n¼0
Note that in two and more dimensions, integrability is not guaranteed without
more knowledge of the potential and its symmetries. Although states can still be
labeled by a spectrum of energies and a decomposition in energy subspaces like
Eq. (15.21) is still possible, the subspaces are not necessarily one-dimensional and
much less can be inferred about the properties of the wave functions.
15.5.2 N Noninteracting, Distinguishable Identical Particles

The next step in constructing this minimal model is to combine N noninteracting
particles. The Hilbert space for the system is constructed by the tensor product of
single-particle Hilbert spaces
N
H N ¼ ⊗ H 1i (15.22)
i¼1
One realization of this Hilbert

space is as Lebesgue square-integrable functions on
configuration space L R . The Hamiltonian can be written as a sum of differen-
2 N
tial operators acting on functions of N one-dimensional coordinates xi :

XN
1 ∂2
H0 ¼
N
þ V ðxi Þ (15.23)
i¼1
2m ∂x2i
The noninteracting dynamical model with Hamiltonian Eq. (15.23) is separable.

Each degree of freedom xi is independent and integrable; each single-particle
Hamiltonian is an integral of the motion. For distinguishable particles without
interactions, the N coordinates xi remain dynamically uncoupled. Though occupy-
ing the same physical space, the particles might as well be scattered throughout the
galaxy as far as the dynamics are concerned. The Hamiltonian H N0 is specific to the
tensor product structure Eq. (15.22), and any entanglement among the particles is a
dynamical invariant.
The kinematic symmetry group of H N0 includes the finite group of particle
permutations isomorphic to SN . The transformations in this group are represented
by operations on H N , but they can also be realized as orthogonal transformations,
i.e., reflections and rotations, on the configuration space RN (Harshman 2016a).
This realization of particle permutations by geometrical point transformations
connects back to the Platonic solids of the introduction. For example, the realiza-
tion of S3 in configuration space R3 is the point group of a triangle, the realization
of S4 in R4 is the point group of a tetrahedron, and so on for N-dimensional
symmetries of (N-1)-simplices. If the single-particle system has parity symmetry
(i.e., reflection symmetry about some point) then, in addition, a sequence of cubic-
type symmetries appears (Harshman 2016b).
314 Nathan Harshman
An energy eigenstate basis for the total system is formed by all tensor products
of single-particle basis vectors like jni jεn1 i⊗jεn2 i⊗ . . . ⊗jεnN i. The energy En
of a basis state is the sum of the single-particle energies. Most energies are no
longer singly degenerate, but they are still denumerable and provide a decom-
position of H N into a tower of energy eigenspaces
H N ¼ ⊕ H En (15.24)
En
where the direct sum is over all possible energies constructed as sums of N single-
particle energies εn . Note that each space H En has a complete basis that is
unentangled with respect to the tensor product structure Eq. (15.22).
The dimension of H En is determined by the number of ways the set of single-
particle energies that sum to E n can be permuted. For example, for three particles
the spaces H En can have one, three, or six dimensions. Irreps of S3 either have one
or two dimensions and that signals the presence of additional kinematic symmet-
ries beyond S3 (Leyvraz et al. 1997, Fernández 2013).
In fact, the decomposition Eq. (15.24) is a reduction into the irrep spaces of the
kinematic symmetry group SN o T t , there T t is the time translation group of a
single-particle system and o is the wreath product (Harshman 2016b). There could
be an even larger kinematic symmetry group of the same form incorporating
additional single-particle symmetries, like parity.
Note that I have not made the claim that all the spaces H En correspond to
different energies. That depends one whether each energy can be uniquely associ-
ated to a set of N single-particle energies. If several energy sums coincide, then
there must be an even larger kinematic symmetry group. I call this an emergent
kinematic symmetry, because it cannot be generated by single-particle symmetries
and particle permutations. One example is when the trapping potential is a har-
monic trap and then the maximal kinematic symmetry is U ðN Þ and can be realized
as symmetry transformations on phase space (Baker 1956, Louck 1965). For this
system, the degeneracies of the energy eigenspaces grow like a factorial in the
energy but can be reduced into spaces like H En . Another example of an emergent
kinematic symmetry is when the trapping potential is an infinite square well. Then
there are “pythagorean degeneracies” that do not appear to have a description as a
group of transformations realized on configuration space or phase space (Shaw
1974).
15.5.3 N Noninteracting Indistinguishable Particles

If the particles are indistinguishable fermions or bosons, then the model of N
noninteracting particles described earlier contains states and observables that
cannot physically be realized or measured. There are several approaches to refining

the model to account for indistinguishability, some more “bottom-up” and others
more “top-down.”
One traditional bottom-up approach starts with the overcomplete Hilbert space
N
H and uses the symmetry group of particle exchanges to decompose it into
sectors. Each sector is labeled by an irrep of the symmetric group SN , and each
sector is a tower of irrep spaces like Eq. (15.6). In the example given earlier with
four particles Eq. (15.7), there were five kinds of irreps. More generally, there are
as many irreps as there are positive integer partitions of N. The number PðN Þ of
inequivalent irreps for a given N, or equivalently, the number of partitions of N, is a
combinatoric problem that does not have an algebraic expression. However, one of
these is always the totally symmetric irrep [N] which is one-dimensional and on
which every particle permutation is represented
by multiplication by +1. Another
of these is the totally symmetric irrep 1N which is also one-dimensional, but now
every odd particle permutation is represented by multiplication by 1 and the even
permutations by +1.
To summarize, the Hilbert space H N can always be decomposed as
H N ¼ H N½N ⊕H N1N ⊕H Neverything else (15.25)

½
The space H N½N contains all the allowed bosonic states, and the space H N1N contains
½
the allowed fermionic states. Further, because particle permutations are a kinematic
symmetry of H N0 , all of the energy eigenspaces H En can be similarly decomposed.
For one-dimensional systems, there is a single bosonic state in each H En , but there is
only a fermionic state in H En when all the single-particle states are different.
The sectors H N½N and H N1N do not inherit the tensor product structure Eq. (15.23)
½
from the distinguishable particle construction. In fact for fermions, there are no
pure states that are separable with respect to the tensor product structure Eq. (15.23).
That sounds exciting on the surface, but because of indistinguishability, there are
also no observables that can detect the implied interparticle entanglement
correlations of pure states (Benatti, Floreanini, and Titimbo 2014). One method to
define entanglement in bosonic and fermionic systems is to find several complete
commuting subalgebras of observables, a la Zanardi, and use them to partition the
symmetrized spaces H N½N and H N1N .
½
A distinct, somewhat more “top-down” approach to indistinguishable particles
builds the model on the symmetrized few-body configuration space (Leinaas and
Myrheim 1977). In the case of N one-dimensional particles, one takes the quotient
of the configuration space RN with the symmetric group RN =SN . This means that
0 0 0 0
points in configuration space x ¼ fx1 ; x2 ; . . . ; xN g and x ¼ x1 ; x2 ; . . . ; xN , which
316 Nathan Harshman
are equivalent up to a permutation of coordinates, are identified as the same point.

In this way, indistinguishability becomes a topological notion. The topologically
trivial manifold RN is warped into a topologically nontrivial orbifold RN =SN . For
three-dimensional systems in Euclidean space, this approach leads to an equivalent
formulation containing bosons and fermions. However, for two-dimensional
systems, or other systems with topologically nontrivial one-particle spaces, this
topological approach to identical particles can yield new physics, most famously
anyons in two dimensions (Wilczek 1982).
15.5.4 Contact Pairwise Interactions

The final piece of the model adds contact (or zero-range) interactions between each
pair hi; ji:
XN X
1 ∂2
H ¼N
þ V ð xi Þ þ g δ xi xj (15.26)
i¼1
2m ∂xi2
hi;ji
The contact interaction is given functional form by the one-dimensional delta-

function, weighted by the interaction parameter g. The contact interaction is
appropriate for modeling physical scenarios where the range of the pairwise
interaction is much shorter than any other length scale in the problem, e.g., the
length scale determined by the trap and from the de Broglie wavelengths. The fact
that the contact interaction is characterized by a single parameter makes it particu-
larly amenable to analysis, as described later.
This model has been well studied for over 50 years, going back at least as far as
analyses for trapped bosonic particles in the g ! ∞ limit by Girardeau (1960), free
bosonic particles by Lieb and Liniger (1963), and free fermions by Yang (1967).
For reviews, including experimental implementations, see Cazalilla et al. (2011),
and Guan, Batchelor, and Lee (2013).
The Hamiltonian H N is no longer specific to the single-particle tensor product
structure. Interactions break the separability and integrability of H N0 and engender
correlations among the distinguishable or indistinguishable particles. However
there are two important limiting cases.
• When g ! ∞, the interactions are called “hard-core interactions.” In one dimen-

sion, there is no way for particles to move past each other and so distinguishable
particles would remain in a fixed order. In this limit, the system is no longer
separable, but integrability reemerges for any trapping potential V ðxÞ. As
Girardeau showed, the wave functions can be expressed as algebraic combin-
ations of the noninteracting wave functions restricted to specific orderings (see
Girardeau 1960, Harshman 2017b). However, in this limit H N is not continuous

and self-adjoint on all L2 RN , which creates some difficulties for level dynam-
ics (Sen 1999).
• When the trapping potential is homogenous but finite in extent, i.e., the infinite
square well, then the system is integrable for any value of g using a method
called the Bethe ansatz (Batchelor et al. 2005, Oelkers et al. 2006). There are N
integrals of motion that are symmetrized polynomials in the single-particle
momenta.
Both of these limiting cases can be understood as examples when the Yang-Baxter
equation holds and there is diffractionless scattering (Sutherland 2004).
One more special case is when the external trap is quadratic in position. The
noninteracting system H N0 is equivalent to an isotropic harmonic oscillator in N
dimensions and, as mentioned earlier, has kinematic symmetry group U ðN Þ. That
system is maximally superintegrable, meaning there are 2N 1 integrals of
motion, and exactly solvable, meaning that the energy is an algebraic function of
the quantum numbers, and all excited states are products of the ground state with
polynomials (Post, Tsujimoto, and Vinet 2012). At finite interaction strength, most
of this additional analytical tractability is lost, but there is one extra integral of the
motion corresponding to the separable center-of-mass degree of freedom (Harsh-
man 2014). This separability survives symmetrization of indistinguishable
particles and, therefore, entanglement between center-of-mass and relative degrees
of freedom remains a dynamical invariant.
For general traps and arbitrary g, the model Eq. (15.26) is not integrable, nor is it
solvable except numerically. Then two questions become: How far from integra-
bility and deep into chaos and is it? How difficult is it to achieve convergent
numerical solutions? The second question has been investigated exhaustively, by
this author and many others, because of the relevance to current experiments
(Serwane et al. 2011, Zürn et al. 2012, 2013, Wenz et al. 2013; for a partial list,
see the references of Harshman 2016a). However, I would claim a productive
metatheory of when particular approximation methods work well has not arrived.
One way to answer the first question about chaos is by comparing the spectrum
to the Wigner-Dyson distribution of eigenvalues of a random matrix (Gutzwiller
1990). According to the Bohigas-Giannoni-Schmit conjecture, this is a universal
feature of systems with quantum chaos. Work on closely related systems suggests
that chaos is present in these systems (Bohigas, Giannoni, and Schmit 1984), but
the world currently waits for a more detailed analysis, especially one that situates
the model in the hierarchy of chaos from ergodic, mixing, Kolmogorov, and
Bernoulli (Ullmo 2016, Gomez, Losada, and Lombardi 2017).
In summary, depending on the trap shape and interaction strength, the model
Eq. (15.26) can manifest the full range of possible dynamic behaviors, from (super)
318 Nathan Harshman
integrability to (conjectured) hard chaos. For integrable cases, there are observ-
ables privileged by dynamical conservation laws that fully characterize the system,
i.e., a complete set of commuting observables. The specific nature of the integrals
of motion depends on the trap and interaction strength. For the noninteracting case,
the integrals of motion are single-particle observables. In integrable interacting
cases like the hard-core limit or Bethe-ansatz solvable cases, the conserved quan-
tities are collective observables built from symmetrized polynomials of single-
particle observables. In contrast, for chaotic cases, the unique conserved observ-
able is the energy itself, and the spectrum matches with a relevant form of
randomness. A goal of this avenue of research is to see if the approach to chaos
can be understood as the dissolution of structures based on irreps of
symmetry group.
15.6 Conclusion: A Few Comments on Symmetry, Structures, and Solvability

The previous section introduced in situ certain technical terms of art, like integra-
bility and solvability. The reader should not be misled by my breezy usage of the
terms to infer that there is universal acceptance among the community of math-
ematical physicists on how these terms should be applied. For example, Liouvillian
integrability is naturally defined in classical dynamics on phase spaces, but there is
debate on how this works for quantum systems with a mixture of continuous and
discrete degrees of freedom (Caux and Mossel 2011). And there is not consensus
on the relationship between Bethe-ansatz integrability relevant to the model dis-
cussed earlier and Liouvillian integrability, and I know of only a few cases where
they coexist (e.g., Harshman et al. 2017). Even more egregious is the dissent and
dissemblance around the term solvable. Some physicists throw around the term
exact solution when they have actually found a somewhere-convergent, asymptot-
ically approximate numerical solution to the limiting case of a mathematically ill-
conditioned dynamical model.
I argue that solvability should be considered as a property of a dynamical model,
and it should be considered as a continuum. At the most solvable extreme are models
where one can push analysis deep into the realm of pure algebra. Examples include
superintegrable systems and exactly solvable systems (which, in fact, are conjectured
to be the same thing). Moving down the spectrum we have systems whose solutions
are formulated in exact analytic relations, but those relations may require (for example)
solving transcendental equations for the spectrum or other model parameters.
My argument is that this same continuum, from more solvable to less solvable,
coincides roughly with two other features: (1) the amount of symmetry in the
system, as measured by the size or complexity of the symmetry group and (2) the
tractability of Hilbert space arithmetic, meaning the variety of inequivalent ways
the system can be decomposed or factorized into subspaces. The structural unit that
unifies these two features is the irreducible representation.
Irreps appear in many guises – as invariant subspaces in direct sums and tensor
products, as the building blocks of towers for describing identical particles that
generate unusable entanglement and frustrate algebraic separability, as the concept
of energy levels that vary across families of Hamiltonians in the same model, and
more. No matter the underlying ontological commitment of an interpretation, any
formulation of quantum mechanics must account for the prevalence and utility of
these structures. Unlike the Platonic solids, these are not metaphors. They are
mathematical building blocks of quantum mechanics and will remain so even when
new or reformed ontologies emerge.
The most egregious oversight of this chapter is that I have not discussed how
symmetry groups partition the set of observables into irreps. This is more technic-
ally challenging that the Hilbert space arithmetic I have presented, but I think the
potential rewards are a deeper understanding of the connections among the observ-
ables, separability (in all three senses), and integrability.
At this stage, the investigation is still incomplete, but I argue that one notion
emerges – the importance of solvability. Solvability is a concept lying in that
awkward place of being a technical term with multiple overlapping and connecting
definitions in different contexts. One of the unifying themes across these contexts
is that solvable systems play a central role in the interpretation of physical
phenomena. Solvable models in mechanics, like coupled harmonic oscillators
and hydrogenic atoms, are not just touchstones for mathematical analysis. They
are ubiquitous as direct and approximate models in nature, and they provide the
cognitive framework for understanding other physical systems. Is it a coincidence
that solvable models are so useful? Is it just attention bias, i.e., we pay more
attention to things we understand more thoroughly? Or, is there something more
deeply “real” about solvable systems, either in an epistemic or ontic sense?
Acknowledgments
I would like to thank Olimpia Lombardi and the other organizers of the workshop
Identity, indistinguishability and non-locality in quantum physics (Buenos Aires,
June 2017) for assembling such a stimulating group of physicists and philosophers.
References
Andersen, M. E. S., Harshman, N. L., and Zinner, N. T. (2017). “Hybrid model of
separable, zero-range, few-body interactions in one-dimensional harmonic traps,”
320 Nathan Harshman
Baker, G. A. (1956). “Degeneracy of the n-dimensional, isotropic, harmonic oscillator,”

Physical Review, 103: 1119–1120.
Batchelor, M. T., Guan, X.-W., Oelkers, N., and Lee, C. (2005). “The 1d interacting Bose
gas in a hard wall box,” Journal of Physics A: Mathematical and General, 38:
7787–7806.
Benatti, F., Floreanini, R., and Titimbo, K. (2014). “Entanglement of identical particles,”
Open Systems & Information Dynamics, 21: 1440003.
Bohigas, O., Giannoni, M. J., and Schmit, C. (1984). “Characterization of chaotic quantum
spectra and universality of level fluctuation laws,” Physical Review Letters, 52: 1–4.
Bohm, A. and Gadella, M. (1989). Dirac Kets, Gamow Vectors and Gel’fand Triplets:
Lecture Notes in Physics No. 348. Berlin-Heidelberg: Springer-Verlag.
Bohm, A. and Harshman, N. L. (1998). “Quantum theory in the rigged Hilbert space
irreversibility from causality,” pp. 179–237 in A. Bohm, H.-D. Doebner, and
P. Kielanowski (eds.), Irreversibility and Causality Semigroups and Rigged Hilbert
Spaces: Lecture Notes in Physics. Berlin-Heidelberg: Springer.
Caux, J.-S. and Mossel, J. (2011). “Remarks on the notion of quantum integrability,”
Journal of Statistical Mechanics, 2011: P02023.
Cazalilla, M. A., Citro, R., Giamarchi, T., Orignac, E., and Rigol, M. (2011). “One-
dimensional bosons: From condensed matter systems to ultracold gases,” Reviews
of Modern Physics, 83: 1405–1466.
Cooper, F., Khare, A., and Sukhatme, U. (1995). “Supersymmetry and quantum mechan-
ics,” Physics Reports, 251: 267–385.
Fernández, F. M. (2013). “On the symmetry of the quantum-mechanical particle in a cubic
box,” arXiv:1310.5136.
Fortin, S. and Lombardi, O. (2016). “A top-down view of the classical limit of quantum
mechanics,” pp. 435–468 in R. E. Kastner, J. Jeknić-Dugić, and G. Jaroszkiewicz
(eds.), Quantum Structural Studies. Singapore: World Scientific.
Girardeau, M. (1960). “Relationship between systems of impenetrable bosons and fer-
mions in one dimension,” Journal of Mathematical Physics, 1: 516–523.
Gómez, I., Losada, M., and Lombardi, O. (2017). “About the concept of quantum chaos,”
Entropy, 19: 205.
Guan, X.-W., Batchelor, M. T., and Lee, C. (2013). “Fermi gases in one dimension: From
Bethe ansatz to experiments,” Reviews of Modern Physics, 85: 1633–1691.
Gutzwiller, M. C. (1990). Chaos in Classical and Quantum Mechanics: Interdisciplinary
Applied Mathematics No. 1. New York: Springer-Verlag.
Haag, R. (1992). Local Quantum Physics: Fields, Particles, Algebras: Texts and Mono-
graphs in Physics. Berlin-New York: Springer-Verlag.
Harshman, N. L. (2005). “Basis states for relativistic dynamically entangled particles,”
Harshman, N. L. (2014). “Spectroscopy for a few atoms harmonically trapped in one
dimension,” Physical Review A, 89: 033633.
Harshman, N. L. (2016a) “One-dimensional traps, two-body interactions, few-body sym-
metries: I. One, two, and three particles,” Few-Body Systems, 57: 11–43.
Harshman, N. L. (2016b) “One-dimensional traps, two-body interactions, few-body sym-
metries: II. N particles,” Few-Body Systems, 57: 45–69.
Harshman, N. L. (2016c). “Symmetry and natural quantum structures for three-particles in
one-dimension,” pp. 373–400 in R. E. Kastner, J. Jeknić-Dugić, and G. Jaroszkiewicz
(eds.), Quantum Structural Studies. Singapore: World Scientific.
Harshman, N. L. (2017a). “Infinite barriers and symmetries for a few trapped particles in
one dimension,” Physical Review A, 95: 053616.
Harshman, N. L. (2017b). “Identical wells, symmetry breaking, and the near-unitary limit,”
Few-Body Systems, 58: 41.
Harshman, N. L., Olshanii, M., Dehkharghani, A. S., Volosniev, A. G., Jackson, S. G., and
Zinner, N. T. (2017). “Integrable families of hard-core particles with unequal masses
in a one-dimensional harmonic trap,” Physical Review X, 7: 041001.
Harshman, N. L. and Ranade, K. S. (2011). “Observables can be tailored to change the
entanglement of any pure state,” Physical Review A, 84: 012303.
Harshman, N. L. and Wickramasekara, S. (2007a). “Galilean and dynamical invariance of
entanglement in particle scattering,” Physical Review Letters, 98: 080406–4.
Harshman, N. L. and Wickramasekara, S. (2007b). “Tensor product structures, entangle-
ment, and particle scattering,” Open Systems & Information Dynamics, 14: 341–351.
Jackiw, R. (1972). “Introducing scale symmetry,” Physics Today, 25: 23–27.
Jeknić-Dugić, J., Arsenijević, M., and Dugić, M. (2013). Quantum Structures: A View of
the Quantum World. Saarbrücken: Lambert Academic Publishing.
Leinaas, J. M. and Myrheim, J. (1977). “On the theory of identical particles,” Nuovo
Cimento B, 37: 1–23.
Lévy-Leblond, J.-M. (1967). “Nonrelativistic particles and wave equations,” Communi-
cations in Mathematical Physics, 6: 286–311.
Leyvraz, F., Frank, A., Lemus, R., and Andrés, M. V. (1997). “Accidental degeneracy in a
simple quantum system: A new symmetry group for a particle in an impenetrable
square-well potential,” American Journal of Physics, 65: 1087–1094.
Lieb, E. H. and Liniger, W. (1963). “Exact analysis of an interacting bose gas. I. The
general solution and the ground state,” Physical Review, 130: 1605–1616.
Lombardi, O. and Fortin, S. (2015). “The role of symmetry in the interpretation of quantum
mechanics,” Electronic Journal of Theoretical Physics, 12: 255–275.
Louck, J. D. (1965). “Group theory of harmonic oscillators in n-dimensional space,”
Journal of Mathematical Physics, 6: 1786–1804.
Oelkers, N., Batchelor, M. T., Bortz, M., and Guan, X.-W. (2006). “Bethe ansatz study of
one-dimensional Bose and Fermi gases with periodic and hard wall boundary condi-
tions,” Journal of Physics A: Mathematical and General, 39: 1073–1098.
Olshanski, G. (2003). “The problem of harmonic analysis on the in_nite-dimensional
unitary group,” Journal of Functional Analysis, 205: 464–524.
Post, S., Tsujimoto, S., and Vinet, L. (2012). “Families of superintegrable Hamiltonians
constructed from exceptional polynomials,” Journal of Physics A: Mathematical and
Theoretical, 45: 405202.
Reck, M., Zeilinger, A., Bernstein, H. J., and Bertani, P. (1994). “Experimental realization
of any discrete unitary operator,” Physical Review Letters, 73: 58–61.
Sen, D. (1999). “Perturbation theory for singular potentials in quantum mechanics,”
International Journal of Modern Physics A, 14: 1789–1807.
Serwane, F., Zürn, G., Lompe, T., Ottenstein, T. B., Wenz, A. N., and Jochim, S. (2011).
“Deterministic preparation of a tunable few-fermion system,” Science, 332: 336–338.
Shaw, G. B. (1974). “Degeneracy in the particle-in-a-box problem,” Journal of Physics A:
Mathematica, Nuclear and General, 7: 1537–1543.
Streater, R. F. and Wightman, A. S. (1964). PCT, Spin and Statistics, and All That: The
Mathematical Physics Monograph Series. New York: W.A. Benjamin.
Sutherland, B. (2004). Beautiful Models: 70 Years of Exactly Solved Quantum Many-Body
Problems. River Edge, NJ: World Scientific.
Ullmo, D. (2016). “Bohigas-Giannoni-Schmit conjecture,” Scholarpedia, 11: 31721.
Weinberg, S. (1995). The Quantum Theory of Fields, 1st edition. Cambridge: Cambridge
University Press.
322 Nathan Harshman
Wenz, A. N., Zürn, G., Murmann, S., Brouzos, I., Lompe, T., and Jochim, S. (2013).
“From few to many: Observing the formation of a Fermi sea one atom at a time,”
Science, 342: 457–460.
Wigner, E. P. (1939). “On unitary representations of the inhomogeneous Lorentz group,”
Annals of Mathematics, 40: 149–204.
Wigner, E. P. (1959). Group Theory and its Application to the Quantum Mechanics of
Atomic Spectra. New York: Academic Press.
Wilczek, F. (1982). “Quantum mechanics of fractional-spin particles,” Physical Review
Letters, 49: 957–959.
Wilczek, F. (2015). A Beautiful Question: Finding Nature’s Deep Design. New York:
Penguin Press.
Wybourne, B. G. (1974). Classical Groups for Physicists. New York: Wiley.
Yang, C. N. (1967). “Some exact results for the many-body problem in one dimension with
repulsive delta-function interaction,” Physical Review Letters, 19: 1312–1315.
Zanardi, P. (2001). “Virtual quantum subsystems,” Physical Review Letters, 87: 077901.
Zanardi, P., Lidar, D. A., and Lloyd, S. (2004). “Quantum tensor product structures are
observable induced,” Physical Review Letters, 92: 060402.
Zanardi, P. and Lloyd, S. (2004). “Universal control of quantum subspaces and subsys-
tems,” Physical Review A, 69: 022313.
Zürn, G., Serwane, F., Lompe, T., Wenz, A. N., Ries, M. G., Bohn, J. E., and Jochim, S.
(2012). “Fermionization of two distinguishable fermions,” Physical Review Letters,
108: 075303.
Zürn, G., Wenz, A. N., Murmann, S., Bergschneider, A., Lompe, T., and Jochim, S.
(2013). “Pairing in few-fermion systems with attractive interactions,” Physical
Review Letters, 111: 175302.
16
Majorization, across the (Quantum) Universe
guido bellomo and gustavo m. bosyk
16.1 Introduction
In how many ways can one represent a given quantum mixed state as a mixture of
pure states? Why (and in which sense) are separable states more disordered
globally than locally? Is it possible to transform a given pure state into another
by means of local operations and classical communication? How should an
adequate formulation of the uncertainty principle be? All these questions, as
dissimilar as they may seem, share one element in common: They can be answered
by appealing to the notion of majorization partial order. Majorization is nowadays
a well-established and powerful mathematical tool with many and different appli-
cations in several disciplines, such as economics, biology, and physics, among
others. Indeed, the seminal idea of this concept had already been glimpsed by
Lorenz (1905) while studying the inequality of wealth distribution and developing
the representation of the (nowadays called) Lorenz curves. Moreover, the famous
Gini coefficient (Gini 1912), widely accepted as a legitimate quantifier of income
distribution inequality, is merely a ratio between graphical areas defined by a
Lorenz curve. Other key contributions to the subject were those by Muirhead
(1903), Dalton (1920), Schur (1923), and Hardy, Littlewood, and Pólya (1929).
The name “majorization,” though, appears first in the prominent book by Hardy,
Littlewood, and Pólya (1934).
At present, it is clear that anyone who is interested in the field would find it
appropriate to begin by the celebrated book by Marshall and Olkin (1979), whose
second edition was coauthored by Arnold (2010). Recently, Arnold (2007) pub-
lished an article entitled “Majorization: Here, there and everywhere,” in which he
presents a sampling of diverse areas in which majorization has been found useful
in the last few years, such as geometry, probability, statistical mechanics, and
graph theory. However, those contributions do not explore its quantum theoretical
implications, which are much more extensively covered, for example, in Nielsen’s
323
324 Guido Bellomo and Gustavo M. Bosyk
lecture notes (Nielsen 2002, see also Nielsen and Vidal 2001). In this chapter, we
attempt to make a brief review of the subject and then to highlight the most
important results of this research line in the quantum realm, in order to offer a
kind of quantum counterpart of Arnold’s work.
A natural question for our current work regards the roots beneath the wide
applicability of majorization in quantum mechanics. Some authors argue that the
connection arises as a result of two important theorems that link majorization to
unitary matrices, Horn’s lemma, and Uhlmann’s theorem, together with the
ubiquity of unitary matrices in quantum mechanics (Nielsen 2002: 5). Here, we
present and discuss a variety of situations to show that the spread applicability of
majorization in the quantum realm emerges as a consequence of deep connections
among majorization, partially ordered probability vectors, unitary matrices, and
the probabilistic structure of quantum mechanics. To this end, we review basic
aspects of majorization, focusing on its connections with some quantum
information problems. In particular, we organize our study here into three
different facets. The first one consists in the role played by majorization as a
disorder signature on quantum states, how quantum operations affect this, and the
connection with quantum entropies. The second one is the study of entanglement
transformations of bipartite pure states, applying local operations, and classical
communication. Third, the problem of how to formulate the uncertainty principle
is posed, and different proposals of majorization uncertainty relations are
reviewed.
In the following section, in order to make the ensuing study more self-
contained, we present some elementary definitions and mathematical properties
related to majorization theory.
16.2 Majorization Theory: Definitions and Mathematical Properties

Let us begin with some basic definitions and notation. Hereafter, we consider the
set of d-dimensional probability vectors whose components are sorted in nonde-
creasing order, namely
( )
X
d
Δd ¼ ½x1 ; . . . ; xd : xi xiþ1 0 and xi ¼ 1 : (16.1)
i¼1
We take for granted that any vector that appears below lives in Δd , that is, has its
elements sorted in a nonincreasing manner. In order to define the concept of
majorization between probability vectors, let x and y be two members of Δd . It is
said that x is majorized by y, denoted as x ≺ y, if and only if the entire set of d 1
inequalities
Majorization, across the (Quantum) Universe 325
X
n X
n
xi yi for all n 2 f1; . . . ; d 1g (16.2)
i¼1 i¼1
P P
is satisfied. (Notice that the condition di¼1 xi ¼ di¼1 yi is trivially satisfied; for
that reason we discard this condition from the definition of majorization). The
intuitive idea is that a probability distribution majorizes another if the former is
more concentrated than the latter. In this sense, majorization provides a quantifica-
tion of the notion of disorder for given probability vectors. To fix ideas, let us
observe that any probability vector x 2 Δd trivially satisfies the relation

1 1
. . . ≺ x ≺½1 0 . . . 0 (16.3)
d d
where the left-hand side (l.h.s.) corresponds to the most uncertain case (uniform
probability vector), whereas the right-hand side (r.h.s.) represents complete
certainty.
It is essential to note that majorization defines a partial order over Δd , meaning
that there exist x, y 2 Δd , such that neither x ≺ y nor y ≺ x. That is the case, for
example, for x ¼ ½0:6; 0:2; 0:2 and y ¼ ½0:5; 0:4; 0:1.
There is a way to easily visualize whether, given two probability vectors, there
is a majorization relation between them. This is done through the notion of Lorenz
curve (Lorenz 1905). A Lorenz curve of a probability
vector x 2 Δd consists of the
Pn
linear segments joining the points n; i¼1 xi for n 2 f0; . . . ; dg. Therefore, a
Lorenz curve is always concave and has extreme points ð0; 0Þ and ð1; 0Þ. In this
way, there is a majorization relation between x, y 2 Δd , if and only if their
corresponding Lorenz curves do not intersect each other, except in the extreme
points. In Figure 16.1, we illustrate these situations.
Figure 16.1 Lorenz curve for four-dimensional probability vectors x ¼

½1; 0; 0; 0 (line with circles), z1 ¼ ½0:6; 0:15; 0:15; 0:1 (line with triangles),
z2 ¼ ½0:5; 0:25; 0:2; 0:05 (line with diamonds), and y ¼ ½0:25; 0:25; 0:25; 0:25
(line with squares). Notice that all the Lorenz curves are concave and lie between
the corresponding extremes: ones of x and y. Futhermore, notice that there is an
intersection between the Lorenz curves of z1 and z2 , showing that there is not a
majorization relation.
Figure 16.2 Set of vectors majorized by a fixed vector y 2 Δ3 , fx : x≺yg. The set
is given by the points inside the convex hull of the orbit of y under the group of
permutation matrices.
Remarkably, the partial order defined by majorization can be posed in several

equivalent ways. We are particularly interested in the following two:
and only if there exists a doubly stochastic matrix B, i.e., Bij 0 for all
• x ≺ y if P P
i, j and di¼1 Bij ¼ 1 ¼ dj¼1 Bij , such that x ¼ By
Pd Pd
• x ≺ y if and only if i¼1 ϕðxi Þ i¼1 ϕðyi Þ for every concave function ϕ
The first equivalent characterization was originally discussed by Schur (1923).
With the aid of Birkhoff’s theorem (Birkhoff 1946), which states that the class of
d d doubly stochastic matrices coincides with the convex hull of the set of d d
permutation matrices, this condition turns into a more geometrical one:
• x ≺ y if and only if x lies in the convex hull of the orbit of y under the group of
permutation matrices (see Figure 16.2)
The second equivalent condition, linking majorization with concave functions,
relates further to the notions of Schur-concavity and entropy. Functions that
preserve the majorization order form a large class, which was formerly studied
by Schur (1923). In his honor, we say that a function Φ : Δd ↦ R is Schur-concave
if it (anti)preserves the majorization relation, that is,
if x ≺ y ) ΦðxÞ ΦðyÞ for all x, y 2 Δd : (16.4)
For example, the very general family of ðh; ϕÞ-entropies satisfies the Schur-
concavity (Bosyk, Zozor, Holik, et al. 2016). They are defined as
!
Xd
H ðh;ϕÞ ðxÞ ¼ h ϕðxi Þ , (16.5)
i¼1
where the entropic functionals h : R ↦ R and ϕ : ½0; 1 ↦ R are such that either: (i)
h is increasing and ϕ is concave, or (ii) h is decreasing and ϕ is convex, together
with the conventions ϕð0Þ ¼ 0 and hðϕð1ÞÞ ¼ 0. The ðh; ϕÞ-entropies can be
classified according to whether they satisfy the additivity relation
H ðh;ϕÞ ðx⊗yÞ ¼ H ðh;ϕÞ ðxÞ þ H ðh;ϕÞ ðyÞ (16.6)
or not. Particular instances of Eq. (16.6) that are additive are the well-known
Shannon (1948), Rényi (1961) and Burg (1967) entropies, given respectively by
X
d
H 1 ðxÞ ¼ xi log xi , (16.7)
i¼1
sign α Xd
H α ðxÞ ¼ log xαi , for all α 2 R∖f0; 1g, (16.8)
1α i¼1
1X d
H Burg ðxÞ ¼ log xi : (16.9)
d i¼1
On the other hand, a paradigmatic example of a nonadditive entropy is that by

Tsallis (1988),
!
1 Xd
q
H q ðxÞ ¼ 1 xi , for all q 2 R∖f1g, (16.10)
1q i¼1
for which the additivity rule is given by

H q ðx⊗yÞ ¼ H q ðxÞ þ H q ðyÞ þ ð1 qÞH q ðxÞH q ðyÞ: (16.11)
We have already pointed out that, from the viewpoint of order theory, majoriza-
tion gives a partial order among probability vectors belonging to Δd . This means
that this binary relation fulfils, for all x, y, z 2 Δd , the following three properties:
• reflexivity: x ≺ x,
• symmetry: if x ≺ y and y ≺ x, then x ¼ y,
• transitivity: if x ≺ z and z ≺ y, then x ≺ y.
Remarkably enough, Cicalese and Vaccaro (2002) have shown that majorization
over Δd defines an even more complex structure: a lattice. This means that there
always exist the infimum (join), x∧y, and the supremum (meet), x∨y. By defin-
ition, x∧y means that x∧y ≺ x, x∧y ≺ y, and z ≺ x∧y for all z such that z ≺ x and
z ≺ y. In a similar way, x∨y means that x ≺ x∨y, y ≺ x∨y, and x∨y ≺ z0 for all z0
such that x ≺ z0 and y ≺ z0 (see Figure 16.3). The algorithms to obtain both the
infimum and the supremum between arbitrary vectors in terms of its elements have
also been given by Cicalese and Vaccaro (2002).
Furthermore, by appealing to the subadditivity and supermodularity of Shannon
entropy, the authors introduced a proper distance on the lattice (Cicalese, Gargano,
and Vaccaro 2013): given x, y 2 Δd , the distance D : Δd Δd ↦ R is defined as
Figure 16.3 Hasse diagram for probability vectors x and y, its infimum (join) x∧y,
and its supremum (meet) x∨y.
Dðx; yÞ ¼ H 1 ðxÞ þ H 1 ðyÞ 2H 1 ðx∨yÞ: (16.12)

It can be shown that this definition provides a proper distance that satisfies, for
x, y, z 2 Δd :
• nonnegativity: Dðx; yÞ 0, with Dðx; yÞ ¼ 0 if and only if x ¼ y;

• symmetry: Dðx; yÞ ¼ Dðy; xÞ;
• triangle inequality: Dðx; yÞ þ Dðy; zÞ Dðx; zÞ;
• compatibility with the majorization lattice: if x ≺ y ≺ z, then Dðx; yÞþ
Dðy; zÞ ¼ Dðx; zÞ.
Before surfing across the quantum universe, we should introduce a couple of
important results involving majorization and matrices, because our study of finite
dimensional quantum systems ultimately relies on the density matrix representa-
tion of their states. The first theorem is due to Schur (1923), and establishes that
given a selfadjoint matrix H 2 CNN with diagonal d ¼ ½d 1 ; d2 . . . d N and eigen-
values λ ¼ ½λ1 ; λ2 . . . λN , then it is always the case that d ≺ λ. On the other hand,
Horn’s theorem (Horn 1954) tells us that for any two vectors d ¼ ½d1 ; d 2 . . . dN
and λ ¼ ½λ1 ; λ2 . . . λN such that d ≺ λ, there exists a selfadjoint matrix H 2 CNN
with diagonal d and eigenvalues λ.
16.3 Majorization in the Quantum Universe

16.3.1 Quantum Basics and Our Roadmap
We are now in position to translate some of the previous discussion into the
quantum setting. First, we should review the concepts of quantum states and
quantum maps. Now, instead of considering the set Δd of real d-dimensional
Figure 16.4 Illustration of the convex set of quantum states. The pure states are
the extreme points of the set (the border). Notice that a given mixed state can be
written in infinitely many ways as convex combinations of pure states. In the
example in this figure, ρ can be expressed by mixing either ρ1 and ρ2 , or ρ3 and ρ4 .
vectors, we are going to work over a set of finite dimensional (positive semidefinite
and trace-class) density matrices or operators over a Hilbert space H ffi Cd ,

D ¼ ρ 2 Cdd : ρ 0 and Trρ ¼ 1 (16.13)
These matrices are also selfadjoint, ρ† ¼ ρ. The set D is convex, meaning that
given ρ1 , ρ2 2 D, then any mixture of the form p ρ1 þ ð1 pÞρ2 2 D, when
p 2 ½0; 1. The extreme states of D are the pure states that satisfy
ρ2 ¼ ρ ¼j ψihψ j, with jψ i 2 H . States which are not pure are called “mixed.”
Notice that there exist infinite ways to decompose a mixed state in terms of convex
combinations of pure states, but they are not arbitrary (we will go more deeply into
this when we discuss the Schrödinger theorem; see Figure 16.4).
Moreover, any state ρ has a spectral decomposition
X
d
ρ¼ λi jei ihei j, (16.14)
i¼1
P
with λi 0, i λi ¼ 1, and ei jej ¼ δij .
Since we plan to later talk about quantum correlations, we should remember
how to describe composite systems. The states of multipartite systems act over
the tensor product of the individual Hilbert spaces. For example, for a bipartite
A [ B system, where H A and H B are the respective individual Hilbert spaces, the
joint state acts over H AB ¼ H A ⊗H B . If ρAB is the state of the bipartite system,
we can obtain the (reduced) states of each subsystem by performing the partial
traces: ρAðBÞ ¼ Tr BðAÞ ρAB . When the joint state is pure, ρAB ¼j ψ AB i ψ AB j, we
Acan

use the Schmidt technic to decompose the same in terms of local basis, ji i
and fjiB ig, as

AB Xn
pffiffiffiffiffi
B X

ψ ¼ ψ i i i , with ψ i 0, ψ i ¼ 1 and n ¼ min d A ; dB :
i¼1 i
(16.15)
The Schmidt decomposition offers a natural viewpoint to study the correlations of

bipartite systems. Indeed, if only one Schmidt
coefficient

A is different

B to zero then

the state is a product (or noncorrelated) state: ψ AB

¼ ψ ⊗ ψ . In other words,
the system is in a pure separable state. When more than one Schmidt coefficient is
positive, the system is in an entangled state. The extreme case is that of a
maximally entangled
pffiffiffiffiffiffiffiffi state, characterized by the equality of the Schmidt coeffi-
cients, ψ i ¼ 1=n for all i. Hereafter, let σ ðψ Þ denote the probability

vector
formed by the squared Schmidt coefficients of a bipartite pure state
ψ AB i.
Mixed joint states demand a more complex hierarchy of correlations. A mixed
state ρAB is a product state whenever it is expressible as a product of individual
states as ρA ⊗ρB . In turn, a mixed separable state can be written as a convex
combination of product states,
X X
ρAB ¼ pi ρAi ⊗ρBi , with pi 0 and pi ¼ 1: (16.16)
i i
Mixed product states are particular cases of separable states. Any state that is
nonseparable is called “entangled.”
We are also going to consider linear, completely positive, trace-preserving maps
from the set of density operators into itself ℰ : D ↦ D, also called “quantum
operations.” All these operations have a Kraus representation of the form
X X
ℰðρÞ ¼ Pi ρP†i with P†i Pi ¼ I (16.17)
i i
in terms of its Kraus operators Pi and I the corresponding identity operator.

Quantum operations include a wide variety of possible quantum dynamical
maps. For example, if we restrict ourselves to the set of bistochastic maps,
P † †
for which i Pi Pi ¼ I, we find the unitary evolutions, ℰðρÞ ¼ UρU with
† †
UU ¼ U U ¼ I, as well as the projective measurements,
X
d
ℰðρÞ ¼ pi Pi with pi ¼ Tr ðρPi Þ and Pi Pj ¼ δij , (16.18)
i¼1
where fPi ¼ jiihij; jii 2 H g is a set of orthogonal rank-one projectors.

Finally, we want to introduce some quantum analogues for the classical entro-
pies. The von Neumann entropy (see, e.g., Wehrl 1978),
SðρÞ ¼ Trρ log ρ, (16.19)

can be viewed as the quantum version of the classical Shannon entropy, by
replacing the sum operation with a trace. We recall that for a selfadjoint operator,
P
over H , X ¼ i xi j iX ihiX j, with j iX i being its eigenvectors in H and xi being the
P
corresponding eigenvalues, one has f ðX Þ ¼ i f ðxi Þ j iX ihiX j, and the trace oper-
ation is the sum of the corresponding eigenvalues.
So far, we have identified two privileged descriptions of the quantum states in
terms of probability vectors. First, for an arbitrary d-dimensional mixed state, we
have the spectral decomposition with its associated probability vector of eigen-
values, λðρÞ 2 Δd . Second, for any d-dimensional bipartite pure state j ψ AB i, we
have its Schmidt decomposition with the corresponding coefficients, which after
being squared also give a probability vector, σ ðψ Þ 2 Δn , with n ¼ min dA ; d B .
For a given pure state, both descriptions are connected, because the squared
Schmidt coefficients coincide with the eigenvalues of the
reduced states, that is,
AðBÞ AB
σ ðψ Þ ¼ λðψ Þ ¼ λðψ Þ, where ψ

A B
¼ Tr BðAÞ j ψ i ψ . Each of these probabil-
AB
ity vectors entails complementary information about the quantum state. Addition-
ally, there is still a third probability vector that one can obtain from the state ρ, if
one considers a preferred observable quantity X, namely, the one defined by the
components of the expectation value in this way. Let X be an observable with
P
discrete and nondegenerate spectrum, that is, X ¼ di¼1 xi jiX ihiX j. The correspond-
ing probability vector, pðX; ρÞ, has i component pi ðX; ρÞ ¼ Tr ðρjiX ihiX jÞ. In what
follows, we are going to present some paradigmatic problems in quantum mech-
anics in which the notion of majorization, together with one of those descriptions,
gives the right starting point toward their solutions (see Figure 16.5).
16.3.2 Schrödinger’s and Ulhmann’s Theorems

Let us start by attacking the question mentioned in the first line of this chapter,
namely: In which ways can one represent a given quantum mixed state as a mixture
of pure states? It seems that the same was answered for the first time by
P
Schrödinger (1936), who proved that a given state ρ ¼ di¼1 λi jei ihei j can be
PM P
written in the form ρ ¼ i¼1 pi jψ i ihψ i j, with pi 0 and i pi ¼ 1, if and only
if there exists a unitary matrix U 2 CMM such that
1 X d pffiffiffiffi
jψ i i ¼ pffiffiffiffi U ik λk jek i: (16.20)
pi k¼1
It follows that the proposed expression is possible whenever the vector p, with
components pi , satisfies the relation p ¼ BλðρÞ, where B is some doubly stochastic
matrix. Hence, recalling Schur’s theorem, we have (Nielsen 2000)
X
M
9fpi ; jψ i ig : ρ ¼ pi jψ i ihψ i j , p ≺ λðρÞ: (16.21)
i¼1
This last equation gives a conclusive requisite that any ensemble of pure states
must fulfill to represent a given mixed state.
Figure 16.5 Our roadmap. Given a quantum state, ρ, there are several associated
probability vectors, which involve different informational aspects of the same.
Each probability vector allows complementary descriptions that are related to
different quantum information problems, which are discussed throughout this
chapter in the sections indicated between brackets.
From the example just shown, we can also have some insight on how to
naturally define a notion of majorization between quantum states by appealing to
its spectral decomposition. Given ρ, σ over H , we say that ρ is majorized by σ as
follows:
ρ ≺ σ , λðρÞ ≺ λðσ Þ: (16.22)
Now, a reasonable question regards the connection between two given states that
satisfy a majorization relation from the viewpoint of quantum operations. In other
words, whether there is any operation linking ρ and σ which satisfies ρ ≺ σ. What
Uhlmann proved is that ρ ≺ σ if and only if the majorized state can be obtained
from the latter by means of convex combinations of unitary maps (Uhlmann
1970)
X X
ρ¼ pi U i σU †i , with U i U †i ¼ I ¼ U †i U i , pi 0 and pi ¼ 1: (16.23)
i i
That is, ρ lies in the convex hull of the unitary orbit to which σ belongs. It is
interesting to observe that, contrary to the classical case, convex combinations of
unitary operations give a more general class of bistochastic channels, also known
as random external fields (Alicki and Lendi 1987). Moreover, one can also prove
that there exists a completely positive quantum bistochastic map transforming

every possible initial state σ into ρ if and only if ρ ≺ σ (Chefles 2002):
bistochastic
σ ! ρ , ρ ≺ σ: (16.24)
In other words, majorization between quantum states offers an equivalent classifi-
cation to that of the action of quantum bistochastic maps, in the same way as
majorization between probability vectors can be rephrased as transformations by
doubly stochastic matrices. Finally, notice that any map of the form Eq. (16.23) is a
unital map, that is, a map with the identity map as a fixed point.
16.3.3 Quantum Entropies and Operations

We have already seen that there is a large class of functions that preserve the
majorization relation between two probability vectors, commonly referred to as
entropies. The same can be settled in the quantum case by appealing to the
quantum analogues of the classical entropies. Indeed, the quantum ðh; ϕÞ-entropies
are defined as (Bosyk, Zozor, Holik, et al. 2016)
Sðh;ϕÞ ðρÞ ¼ hðTrϕðρÞÞ, (16.25)
where the entropic functions h and ϕ satisfy the same requirements as that in the
classical case. Notice that the quantum entropy of a given state ρ coincides with the
classical one for the probability vector formed by the eigenvalues of ρ, that is,
Sðh;ϕÞ ðρÞ ¼ H ðh;ϕÞ ðλðρÞÞ. In this way, we find, in the quantum realm, the same close
connection between majorization, bistochastic maps, and entropies.
Several properties of Eq. (16.25) have been studied by Bosyk, Zozor, Holik,
et al. (2016). In addition to the Schur-concavity of Eq. (16.25), two specific links to
majorization are the following. On the one hand, as a consequence of Schur-
concavity and Schrödinger theorem, one has that the quantum ðh; ϕÞ-entropy of
P
an arbitrary statistical mixture of pure states ρ ¼ M i¼1 pi jψ i ihψ i j, is upper-
bounded by the classical ðh; ϕÞ-entropy of the probability vector formed by the
mixture weights, that is,
Sðh;ϕÞ ðρÞ H ðh;ϕÞ ðpÞ: (16.26)
On the other hand, when dealing with quantum systems, it is of interest to estimate
the effect of a given quantum operation on them. In particular, one may guess that
a measurement can only perturb the state and, thus, that the entropy will increase.
This is also true for more general quantum operations. Indeed, as a consequence of
the Schur-concavity and a result by Chefles (2002), one has that any bistochastic
map ℰ has a nondecreasing effect on the entropy, that is,
Sðh;ϕÞ ðℰðρÞÞ Sðh;ϕÞ ðρÞ for all ρ 2 D, (16.27)
where the equality holds if and only if the map is unitary ℰðρÞ ¼ UρU † : Notice that
in the case of a nonbistochastic map, the entropy can decrease. For instance, let ρ
be the density operator of an arbitrary mixed qubit system, with nonvanishing
quantum ðh; ϕÞ-entropy, and ε a completely positive and trace-preserving (but not
unital) map characterized by Kraus operators P1 ¼ j0ih0 j and P2 ¼ j0ih1 j. Then,
the system, after the action of this map, is on the pure state ℰðρÞ ¼ j0ih0 j; thus, it
has zero quantum ðh; ϕÞ-entropy.
Projective measurements are particular cases of bistochastic maps (see Eq. (16.18)).
Hence, for a given quantum system, the difference of quantum entropies between the
postmeasurement and the premeasurement states works as a signature of the disturb-
ance of the state of a system due to the measurement. In particular, consideration of
local disturbances gives place to a type of quantum correlations measures and a way to
characterize them (see, e.g., Luo 2008, Horodecki et al. 2005). As the main ingredient
needed to guarantee the validity of these measures is the property expressed in
Eq. (16.27), these measures can be defined in terms of generalized quantum entropies
(see, e.g., Rossignoli, Canosa, and Ciliberti 2010, Bosyk, Bellomo, Zozor, et al. 2016).
16.3.4 Entanglement, Separability and Entropy

There is another important application of majorization theory to the problem of
determining the separable or entangled character of quantum states. The result
depends, not surprisingly, on the relation between the global and local spectral
properties of the states.
Let us consider a bipartite scenario: A quantum system in the state ρAB with
marginal states ρA ¼ Tr B ρAB and ρB ¼ Tr A ρAB . A remarkable feature of quantum
systems is that there are states that describe situations such that the local disorder
(in terms of the entropy of its reduced states) is greater than the global disorder.
And that feature implies nonseparability of the state. That is,

Sðh;ϕÞ ρA > Sðh;ϕÞ ρAB or Sðh;ϕÞ ρB > Sðh;ϕÞ ρAB ) ρAB is entangled:
(16.28)
The most radical example of these bounds is surely given by a Bell state, in which
case we have minimal (null) global entropy and maximal local entropies. Classic-
ally, this is never the case: Given two classical random variables X and Y, with
joint probability distribution pXY , the Shannon entropy always obeys the relation
H 1 ðX Þ H 1 ðX; Y Þ and H 1 ðY Þ H 1 ðX; Y Þ. From Eq. (16.27), we see that an
analogous result holds in the quantum realm only for separable states (Horodecki,
Horodecki, and Horodecki 1996).
Relying on the strong relationship between disorder, entropies, and majoriza-

tion, it is natural to expect some connection between separability and the local and
global spectrums. Indeed, Nielsen and Kempe (2001) proved that
ρAB separable ) ρAB ≺ ρA and ρAB ≺ ρB : (16.29)

Notice that violation of the r.h.s of Eq. (16.29) is a sufficient condition to have
entanglement. However, we remark that the spectral properties do not determine
separability in general. Indeed, there exist pairs of states such that they are globally
and locally isospectral, but one of them is separable and the other is not (Bengtsson
and Życzkowski 2017: 491).
16.3.5 Pure States Interconversion

Evoking the third introductory question that we asked at the very beginning of this
chapter, which referred to the possibility of transforming a given pure state into
another by means of local operations and classical communication, we now
propose to study entanglement transformations by using local operations and
classical communication (LOCC). So far, we have always considered majorization
between quantum states (namely, between the corresponding spectra), but in this
case, we have to recall majorization relations among Schmidt coefficients.
So, the question is whether an initial bipartite pure state j ψ AB i 2 H AB can be
transformed into another bipartite state j ϕAB i 2 H AB (the target), by using LOCC.
Here, by LOCC one refers to product (noncorrelated) operations, acting over H A
and H B independently, assisted by two-way classical communication. That is, one
supposes a channel that allows communicating the results of a given local oper-
ation to the other part. The problem has been originally addressed by Nielsen
(1999), who identified the necessary and sufficient condition that enables this
entanglement transformation process. An auxiliary result due to Lo and Popescu
(1999), used by Nielsen to prove his theorem, regards the possibility to simulate
the two-way classical communication by a unidirectional classical channel and
local generalized measurements. Interestingly enough, the conditions elucidated by
Nielsen, under which entangled states can be achieved by this process, can be
established in terms of a majorization relation. More precisely, if σ ðψ Þ and σ ðϕÞ are
the probability vectors formed

by the
corresponding
squared Schmidt coefficients,
the LOCC transformation
ψ AB !
ϕAB is possible if and only if σ ðψ Þ ≺ σ ðϕÞ:

AB LOCC
AB

ψ !
ϕ , σ ðψ Þ ≺ σ ðϕÞ: (16.30)
In other words, since those squared coefficients coincide with the reduced states’
eigenvalues

AB LOCC
AB

ψ !
ϕ , Tr A
ψ AB ψ AB
≺ Tr A
ϕAB ϕAB
: (16.31)
We stress that the majorization relationship constitutes the necessary and sufficient
condition under which this transformation is allowed, without any reference to the
corresponding Schmidt bases.
As already mentioned, majorization gives a partial order over Δd and, as such,
Nielsen’s result does not hold in general, in the sense that there exists a pair of
states that neither of them majorizes each other: jψ i ↮ j ϕi. For instance, it is easy
to check that j ψi ↮ j ϕi by LOCC when the squared Schmidt coefficients are
σ ðψ Þ ¼ ½0:60; 0:15; 0:15; 0:10 and σ ðϕÞ ¼ ½0:50; 0:25; 0:20; 0:05, because
σ ðψ Þ ⊀ σ ðϕÞ and σ ðϕÞ ⊀ σ ðψ Þ. With this in mind, the celebrated result discussed
earlier, due to Nielsen, has subsequently been extended to the case of nondetermi-
nistic LOCC transformations by Vidal (1999). In that case, one looks for the
maximal probability of success. On the other hand, it is also possible to provide
further insight if one considers deterministic transformations but with approximate
target states. Vidal, Jonathan, and Nielsen (2000) have solved this problem by
invoking a criterion of maximal fidelity. The same problem has been tackled from
a different perspective, exploiting the lattice structure of majorization and showing
that both proposals are linked via a majorization relation (Bosyk, Sergioli, Freytes,
et al. 2017). The latter seems to be the first attempt to exploit the lattice character of
the majorization partial order in a quantum information context, beyond its well-
known partial-order properties.
Another generalization, proposed by Jonathan and Plenio (1999), consists in the
extension of the set of initial and final states by appealing to deterministic
entangled-assisted LOCC, that is, considering a shared catalytic entangled state
between both parts. In this protocol, we have a new partial-order relation that it is
called “trumping majorization” and reads as follows: Given x, y, z 2 Δd , it is said
that x is trumping majorized by y (and denoted by x ≺ T y) if and only if there exists
a catalytic r such that x ⊗ z ≺ y ⊗ z (see Daftuar and Klimesh 2001 for some
mathematical properties of trumping, and Müller and Pastena 2016 for an exten-
sion of this concept related to Shannon entropy). Although it is an open question
whether trumping majorization can be endowed with a lattice structure in the
general case (Harremoës 2004), it has been recently shown that the structure holds
for the minimal nontrivial case, namely the case of four-dimensional vectors and
two-dimensional catalysts (Bosyk, Freytes, Bellomo, et al. 2018).
It is notable that all these questions can be enclosed under the problem
of convertibility of one kind of physical resource into another. Lately, this
resource theoretic approach has been extensively applied to attack a bunch of
quantum information-related topics such as, for instance, nonlocal correlations
(see, e.g., Barrett et al. 2005, de Vicente 2014), quantum coherence and asymmetry
(see, e.g., Ahmadi, Jennings, and Rudolph 2013, Piani et al. 2016), quantum
thermodynamics (see, e.g., Brandao et al. 2013, Gour et al. 2015) and super-
selection rules (see, e.g., Gour and Spekkens 2008). Remarkably enough, this
formalism has recently been applied out of the quantum domain, for instance, to
the study of polarization-coherence properties of classical electromagnetic fields
(Bosyk, Bellomo, and Luis 2018a, 2018b) as well as to the study of measures of
statistical complexity (Rudnicki et al. 2016).
16.3.6 Uncertainty Relations

The uncertainty principle is, without any doubt, another of the fundamental
characteristics of quantum mechanics. Its relation to majorization theory is the
end of our trip across the quantum universe.
Heisenberg (1927), in his seminal article, whose 90th anniversary was cele-
brated in 2017, appealed to a heuristic formulation in order to quantify a funda-
mental operational limitation imposed by the quantum laws, namely: the
impossibility of preparing states that give well-defined values for complementary
observables (such as the position and momentum of a particle).
The best-known formulation of this principle is the Robertson (1929) uncer-
tainty relation
1
V ðX; ρÞV ðY; ρÞ

h½X; Y iρ
, (16.32)
4

where V ðO; ρÞ ¼ O2 ρ hOi2ρ andh Oiρ ¼ Tr ðρOÞ denote the variance and
expectation value of an observable O ¼ X, Y given the state preparation ρ, respect-
ively. This preparation uncertainty relation thus describes a tradeoff between the
variance of two incompatible observables for the same quantum state, but separ-
ately obtained in different experiments. However, several authors criticized that
formulation because it does not capture the essence of the uncertainty principle in
general. The main shortcomings of Eq. (16.32) appear when the observables have
discrete spectrum. On the one hand, the use of variance as degree of uncertainty of
a given observable with discrete spectrum loses its operational meaning, because a
mere relabeling of the observable outcomes (without changing its probability of
occurrence) can give a variation of the uncertainty. Indeed, the only justification
for the use of the variance given by Robertson is because it is “in accordance to
statistical usage” (Robertson 1929). It seems that he had in mind Gaussian
distributions, where the variance is enough to completely describe it. On the other
hand, the measure of incompatibility of the observables is given by the mean value
of the commutator. For observables with discrete spectrum, the commutator is
another observable for which there always exists a quantum state such that its mean
value vanishes. For example, this happens when the quantum state is an eigenstate
of one the observables. In other words, the r.h.s of Eq. (16.32) is state-dependent
(that is, it is not universal) and does not fully characterize the incompatibility of the
observables.
For those reasons, several alternative uncertainty relations have been proposed
in order to overcome these issues. Among them, geometric (Landau and Pollak
1961, Bosyk et al. 2014), entropic (Deutsch 1983, Maassen and Uffink 1988,
Zozor, Bosyk, and Portesi 2014) and majorization (Partovi 2011, Friedland,
Gheorghiu, and Gour 2013, Puchała, Rudnicki, and Życzkowski 2013) uncertainty
relations have appeared as the most prominent ones. In general, an uncertainty
relation is an inequality of the form
U ðA; B; ρÞ ℬðA; BÞ, (16.33)

where U ðA; B; ρÞ measures the degree of uncertainty in the observables’ outcomes
when the quantum system is prepared in the state ρ, and ℬðA; BÞ is a measure of the
observables’ incompatibility, which is state-independent and strictly greater than
zero except when the observables share at least one eigenstate.
Here, we are interested in a majorization-based formulation of the uncertainty
principle. For simplicity, let us consider observables with discrete and nondegene-
P Pd
rate spectrum, that is, X ¼ di¼1 xi jiX ihiX j and i¼1 yi jiY ihiY j. A majorization-
based uncertainty relation has the form
pðX; ρÞ⊗pðY; ρÞ ≺ ωðX; Y Þ, (16.34)

where pi ðO; ρÞ ¼ Tr ðρjiO ihiO jÞ is the i-th component of the d-dimensional prob-
ability vector pðO; ρÞ and ωðX; Y Þ is a d 2 -probability vector that measures the
incompatibility between the observables. Clearly, if the observables do not share a
common eigenstate, then ωðX; Y Þ 6¼ ½1; 0; . . . ; 0, giving a nontrivial and universal
(i.e., state-independent) bound on how the product distributions must be. The
explicit expression of the optimal bound ωðX; Y Þ is very difficult to calculate in
general, because it involves a hard optimization problem. It can be shown that a
weaker uncertainty relation is given as follows,
pðX; ρÞ⊗pðY; ρÞ ≺ ωðc; c0 Þ ¼ ½ω1 ðcÞ; ω2 ðc0 Þ ω1 ðcÞ; 1 ω2 ðc0 Þ; 0; . . . ; 0,

(16.35)
1þc 2
where ω1 ðcÞ ¼ and c denotes the maximum overlap between the eigenbasis
2
of the observables, that is,

1
c ¼ max jhiX jiY ij 2 pffiffiffiffi ; 1 , (16.36)
fiX ;iY g N
2
0 1 þ c0
and ω1 ðc Þ ¼ with
2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2
c0 ¼ max jhiX jiY ij2 þ
i0X ji0Y
, (16.37)
where the maximum is taken over the all indexes iX ¼ i0X and iY 6¼ i0Y , and over the
all indexes iX 6¼ i0X and iY ¼ i0Y .
Finally, let us observe that from a majorization uncertainty relation one can always
obtain the corresponding entropic version by using Schur-concave additive entropies.
For instance, the Rényi entropic uncertainty relation obtained from Eq. (16.34) is
1 α α
H α ðpðX;ρÞÞ þH α ðpðY;ρÞÞ log ω1 ðcÞα þ ðω2 ðc0 Þ ω1 ðcÞÞ þ ð1 ω2 ðc0 ÞÞ :
1 α
(16.38)
It can be shown that for Shannon entropy (α ¼ 1) this entropic uncertainty relation
is stronger than the one derived by Deutsch (1983)
1þc
H ðpðX; ρÞÞ þ H ðpðY; ρÞÞ 2 log : (16.39)
2
Therefore, majorization-based uncertainty relations not only give adequate formu-
lations of the uncertainty principle, but also allow stronger entropic-based expres-
sions to be obtained.

We began our tour with the aim of explaining how a variety of quantum problems
ultimately depend on possible hierarchizations of quantum states based on the notion
of majorization. Our task has been accomplished, after heterogeneous discussions
about classification of quantum mixtures, entropies, and bistochastic operations,
correlations, entanglement, conversion by LOCC, and uncertainty relations. As we
anticipate, this journey has been by no means comprehensive. However we hope the
paradigmatic examples that we have discussed throughout this chapter could draw
attention to the ubiquity of majorization as a natural way to compare quantum states
and as a powerful tool to study quantum information problems.
Acknowledgments
We are extremely grateful to the organizers of the workshop Identity, indistinguish-
ability and non-locality in quantum physics (Buenos Aires, June 2017). This work
was partially supported by CONICET and UNLP and Grant 57919 from the John
Templeton Foundation.
References
Ahmadi, M., Jennings, D., and Rudolph, T. (2013). “The Wigner–Araki–Yanase theorem
and the quantum resource theory of asymmetry,” New Journal of Physics, 15:
013057.
Alicki, R. and Lendi, K. (1987). Quantum Dynamical Semigroups and Applications.
Berlin: Springer-Verlag.
Arnold, B. (2007). “Majorization: Here, there and everywhere,” Statistical Science, 22:
407–413.
Barrett, J., Linden, N., Massar, S., Pironio, S., Popescu, S., and Roberts, D. (2005).
“Nonlocal correlations as an information-theoretic resource,” Physical Review A,
71: 022101.
Bengtsson, I. and Życzkowski, K. (2017). Geometry of Quantum States: An Introduction to
Quantum Entanglement. Cambridge: Cambridge University Press.
Birkhoff, G. (1946). “Tres observaciones sobre el álgebra lineal,” Universidad Nacional de
Tucumán Revista, Serie A, 5: 147–151.
Bosyk, G. M., Bellomo, G., and Luis, A. (2018a). “A resource-theoretic approach to
vectorial coherence,” Optics Letters, 43: 1463–1466.
Bosyk, G. M., Bellomo, G., and Luis, A. (2018b). “Polarization monotones of two-
dimensional and three-dimensional random electromagnetic fields,” Physical Review
A, 97: 023804.
Bosyk, G. M., Bellomo, G., Zozor, S., Portesi, M., and Lamberti, P. W. (2016). “Unified
entropic measures of quantum correlations induced by local measurements,” Physica
A: Statistical Mechanics and its Applications, 462: 930–939.
Bosyk, G. M., Freytes, H., Bellomo, G., and Sergioli, G. (2018). “The lattice of
trumping majorization for 4D probability vectors and 2D catalysts.” Scientific
Reports, 8: 3671.
Bosyk, G. M., Sergioli, G., Freytes, H., Holik, F., and Bellomo, G. (2017). “Approximate
transformations of bipartite pure-state entanglement from the majorization lattice,”
Physica A: Statistical Mechanics and its Applications, 473: 403–411.
Bosyk, G. M., Tristán, M. O., Lamberti, P. W., and Portesi, M. (2014). “Geometric
formulation of the uncertainty principle,” Physical Review A, 89: 034101.
Bosyk, G. M., Zozor, S., Holik, F., Portesi, M., and Lamberti, P. W. (2016). “A family of
generalized quantum entropies: Definition and properties,” Quantum Information
Processing, 15: 3393–3420.
Brandao, F. G., Horodecki, M., Oppenheim, J., Renes, J. M., and Spekkens, R. W. (2013).
“Resource theory of quantum states out of thermal equilibrium,” Physical Review
Letters, 111: 250404.
Burg, J. P. (1967). “Maximum entropy spectral analysis,” Proceedings 37th Annual
Meeting of the Society of Exploration Geophysicists. Oklahoma City, OK.
Chefles, A. (2002). “Quantum operations, state transformations and probabilities,” Phys-
ical Review A, 65: 052314.
Cicalese, F., Gargano, L., and Vaccaro, U. (2013). “Information theoretic measures of
distances and their econometric applications,” pp. 409–413 in Information Theory
Proceedings (ISIT), 2013 IEEE International Symposium. IEEE.
Cicalese, F. and Vaccaro, U. (2002). “Supermodularity and subadditivity properties of the
entropy on the majorization lattice,” IEEE Transactions on Information Theory, 48:
933–938.
Daftuar, S. and Klimesh, M. (2001). “Mathematical structure of entanglement catalysis,”
Dalton, H. (1920). “The measurement of the inequality of incomes,” The Economic

Journal, 30: 348–361.
Deutsch, D. (1983). “Uncertainty in quantum measurements,” Physical Review Letters, 50:
631–633.
de Vicente, J. I. (2014). “On nonlocality as a resource theory and nonlocality measures,”
Journal of Physics A: Mathematical and Theoretical, 47: 424017.
Friedland, S., Gheorghiu, V., and Gour, G. (2013). “Universal uncertainty relations,”
Physical Review Letters, 111: 230401.
Gini, C. (1912). “Variabilità e mutabilità,” reprinted in Memorie di Metodologica Statis-
tica. Rome: Libreria Eredi Virgilio Veschi.
Gour, G., Müller, M. P., Narasimhachar, V., Spekkens, R. W., and Halpern, N. Y. (2015).
“The resource theory of informational nonequilibrium in thermodynamics,” Physics
Reports, 583: 1–58.
Gour, G. and Spekkens, R. W. (2008). “The resource theory of quantum reference frames:
Manipulations and monotones,” New Journal of Physics, 10: 033023.
Hardy, G. H., Littlewood, J. E., and Pólya, G. (1929). “Some simple inequalities satisfied
by convex functions,” Messenger of Mathematics, 58: 145–152.
Hardy, G. H., Littlewood, J. E., and Pólya, G. (1934). Inequalities, 1st edition. (2nd edition
1952). London and New York: Cambridge University Press.
Harremoës, P. (2004). “A new look on majorization,” pp. 1422–1425 in Proceedings
ISITA. Parma, Italy.
Heisenberg, W. (1927). “Über den anschaulichen Inhalt der quantentheoretischen Kine-
matik und Mechanik,” Zeitschrift für Physik, 43: 172–198.
Horn, A. (1954). “Doubly stochastic matrices and the diagonal of a rotation matrix,”
American Journal of Mathematics, 76: 620–630.
Horodecki, R., Horodecki, P., and Horodecki, M. (1996). “Quantum α-entropy inequal-
ities: Independent condition for local realism?”, Physics Letters A, 210: 377–381.
Horodecki, M., Horodecki, P., Horodecki, R., Oppenheim, J., Sen, A., Sen, U., and Synak-
Radtke, B. (2005). “Local versus nonlocal information in quantum-information
theory: Formalism and phenomena,” Physical Review A, 71: 062307.
Jonathan, D. and Plenio, M. B. (1999). “Entanglement-assisted local manipulation of pure
quantum states,” Physical Review Letters, 83: 3566.
Landau, H. J. and Pollak, H. O. (1961). “Prolate spheroidal wave functions, Fourier
analysis and uncertainty–II,” Bell System Technical Journal, 40: 65–84.
Lo, H. K. and Popescu, S. (1999). “Classical communication cost of entanglement
manipulation: Is entanglement an interconvertible resource?”, Physical Review
Letters, 83: 1459–1462.
Lorenz, M. O. (1905). “Methods of measuring the concentration of wealth,” Publications
of the American Statistical Association, 9: 209–219.
Luo, S. (2008). “Using measurement-induced disturbance to characterize correlations as
classical or quantum,” Physical Review A, 77: 022301.
Maassen, H. and Uffink, J. (1988). “Generalized entropic uncertainty relations,” Physical
Review Letters, 60: 1103–1106.
Marshall, A. W. and Olkin, I. (1979). Inequalities: Theory of Majorization and Its
Applications. New York: Springer Science & Business Media.
Marshall, A. W., Olkin, I., and Arnold, B. (2010). Inequalities: Theory of Majorization and
Its Applications. New York: Academic Press.
Muirhead, R. F. (1903). “Some methods applicable to identities and inequalities of
symmetric algebraic functions of n letters,” Proceedings of the Edinburgh Mathemat-
ical Society, 21: 144–157.
Müller, M. P. and Pastena, M. (2016). “A generalization of majorization that characterizes

Shannon entropy,” IEEE Transactions on Information Theory, 62: 1711–1720.
Nielsen, M. A. (1999). “Conditions for a class of entanglement transformations,” Physical
Review Letters, 83: 436–439.
Nielsen, M. A. (2000). “Probability distributions consistent with a mixed state,” Physical
Review A, 62: 052308.
Nielsen, M. A. (2002). An Introduction to Majorization and Its Applications to Quantum
Mechanics. Lecture Notes, Department of Physics. Brisbane: University of
Queensland.
Nielsen, M. A. and Kempe, J. (2001). “Separable states are more disordered globally than
locally,” Physical Review Letters, 86: 5184–5187.
Nielsen, M. A. and Vidal, G. (2001). “Majorization and the interconversion of bipartite
states,” Quantum Information & Computation, 1: 76–93.
Partovi, M. H. (2011). “Majorization formulation of uncertainty in quantum mechanics,”
Piani, M., Cianciaruso, M., Bromley, T. R., Napoli, C., Johnston, N., and Adesso, G.
(2016). “Robustness of asymmetry and coherence of quantum states,” Physical
Review A, 93: 042107.
Puchała, Z., Rudnicki, Ł., and Życzkowski, K. (2013). “Majorization entropic uncertainty
relations,” Journal of Physics A: Mathematical and Theoretical, 46: 272002.
Rényi, A. (1961). “On measures of information and entropy,” pp. 547–561 in Proceedings
of the Fourth Berkeley Symposium on Mathematics, Statistics and Probability.
Berkeley.
Robertson, H. P. (1929). “The uncertainty principle,” Physical Review, 34: 163–164.
Rossignoli, R., Canosa, N., and Ciliberti, L. (2010). “Generalized entropic measures of
quantum correlations,” Physical Review A, 82: 052342.
Rudnicki, Ł., Toranzo, I. V., Sánchez-Moreno, P., and Dehesa, J. S. (2016). “Monotone
measures of statistical complexity,” Physics Letters A, 380: 377–380.
Schrödinger, E. (1936). “Probability relations between separated systems,” Mathematical
Proceedings of the Cambridge Philosophical Society, 32: 446–452.
Schur, I. (1923). “Über eine Klasse von Mittelbildungen mit Anwendungen auf die
Determinantentheorie,” Sitzungsberichte der Berliner Mathematischen Gesellschaft,
22: 9–20. English translation: (1973). pp. 416–427 in A. Brauer and H. Rohrbach
(eds.), Issai Schur Collected Works. Berlin: Springer-Verlag.
Shannon, C. (1948). “The mathematical theory of communication,” Bell System Technical
Journal, 27: 379–423.
Tsallis, C. (1988). “Possible generalization of Boltzmann-Gibbs statistics,” Journal of
Statistical Physics, 52: 479–487.
Uhlmann, A. (1970). “On the Shannon entropy and related functionals on convex sets,”
Reports on Mathematical Physics, 1: 147–159.
Vidal, G. (1999). “Entanglement of pure states for a single copy,” Physical Review Letters,
83: 1046–1049.
Vidal, G., Jonathan, D., and Nielsen, M. A. (2000). “Approximate transformations and
robust manipulation of bipartite pure-state entanglement,” Physical Review A, 62:
012304.
Wehrl, A. (1978). “General properties of entropy,” Reviews of Modern Physics, 50:
221–260.
Zozor, S., Bosyk, G. M., and Portesi, M. (2014). “General entropy-like uncertainty
relations in finite dimensions,” Journal of Physics A: Mathematical and Theoretical,
47: 495302.
Part V
The Relationship between the Quantum Ontology and
the Classical World
17
A Closed-System Approach to Decoherence
sebastian fortin and olimpia lombardi
17.1 Introduction
Decoherence is a process that leads to spontaneous suppression of quantum
interference. The orthodox explanation of the phenomenon is given by the environ-
ment-induced-decoherence approach (see, e.g., Zurek 1982, 1993, 2003),
according to which decoherence is a process resulting from the interaction of an
open quantum system and its environment. By studying different physical models,
it was proved that, when the environment has a huge number of degrees of freedom
and for certain interactions, the reduced state of the open system rapidly diagona-
lizes in a well-defined preferred basis.
The environment-induced approach has been extensively applied to many areas
of physics, such as atomic physics, quantum optics, and condensed matter, and has
acquired a great importance in quantum computation, where the loss of coherence
represents a major difficulty for the implementation of the information processing
hardware that takes advantage of superpositions. In the field of the foundations of
physics, this approach has been conceived as the key ingredient to explain
the emergence of classicality from the quantum world, because the preferred
basis identifies the candidates for classical states (see, e.g., Elby 1994, Healey
1995, Paz and Zurek 2002). It has been also considered a relevant element in
different interpretations or approaches to quantum mechanics (for a survey, see
Bacciagaluppi 2016).
The wide success of the environment-induced approach to decoherence over-
shadowed any conceptual difficulty: Only a few works were devoted to analyze the
assumptions and limitations of the orthodox approach. In resonance with this fact,
the different approaches to decoherence that have arisen to face those difficulties
were not taken into account with the care that they deserve. In this chapter we will
show that there is a different perspective to understand decoherence – a closed-
system approach – which not only solves or dissolves the problems of the orthodox
345
346 Sebastian Fortin and Olimpia Lombardi
approach, but also is in agreement with a top-down view of quantum mechanics

that offers a new perspective about the traditional interpretive problems.
With this purpose, the chapter is organized as follows. In Section 17.2, we will
begin by contrasting a bottom-up view versus a top-down view of quantum
mechanics. In Section 17.3, the decoherence resulting from the interaction with
the environment will be explained from a closed-system perspective. This will
allow us to introduce, in Section 17.4, a general top-down, closed-system approach
to decoherence, in the context of which environment-induced decoherence is a
particular case. The chapter closes with some final remarks.
17.2 Bottom-Up View versus Top-Down View of Quantum Mechanics

The idea that nature consists of tiny elemental entities is deeply entrenched in our
way of conceiving reality. It finds its roots in ancient Greece with atomism, and
reappears in the early Modern Age with the corpuscularist philosophy of Robert
Boyle, which influenced many contemporary thinkers, including Newton. Since
those days, it has taken different forms in chemistry, as in Dalton’s atomic theory,
and in physics, from the kinetic theory of gases to the standard model of particle
physics. An epistemological strategy becomes natural in the light of this onto-
logical picture: In order to understand nature, it is necessary to decompose it into
simple systems. The knowledge about the whole is obtained by first studying the
simple systems and then combining them through their interactions. Of course,
there are cases in which this analytical strategy leads to descriptions that cannot be
solved by formal means. This is the case of the three-body problem in classical
mechanics. Nevertheless, even if there is no general closed-form solution for the
equations describing the many-body system, nobody doubts that the behavior of
the whole system is determined by the components and its interactions; precisely
for this reason, those problems are commonly solved by numerical methods.
With the advent of quantum mechanics, this ontological picture went into crisis.
The phenomenon of entanglement, which is not a traditional physical interaction,
is responsible of correlations that cannot be understood in classical terms. There-
fore, in quantum mechanics, the assumption that the best knowledge of the whole
is obtained by studying the simple systems and their interactions breaks down:
Here the state of the composite system is not uniquely determined by the states of
the component subsystems. Nevertheless, in spite of this well-known fact, it is
usual to begin with quantum systems, represented by Hilbert spaces, which
become subsystems when they constitute a composite system. The implicit
assumption is the atomistic assumption that there are certain elemental “particles”
out of which everything is composed. This assumption has even been made
explicit by the atomic modal interpretation of quantum mechanics, according to
A Closed-System Approach to Decoherence 347
which there is, in nature, a fixed set of mutually disjoint atomic quantum systems
that constitute the building blocks of all the other quantum systems (Bacciagaluppi
and Dickson 1999). Good candidates for elemental systems are those represented
by the irreducible representations of the symmetry group of the theory.
From this viewpoint, when quantum systems interact, their states may become
entangled: “By the interaction the two representatives [the quantum states] have
become entangled” (Schrödinger 1935: 555, when he coined the term ‘entangled’).
In this case, it is said that the composite system is an entangled state, because it
cannot be obtained as the tensor product of the components’ states. Entanglement
is, therefore, responsible for the correlations between the values of the observables
of the two subsystems.
This bottom-up ontological view leads us to first consider two particles, say, a
proton p and an electron e, represented by the Hilbert spaces H p and H e and in
states ψp 2 H p and ψe 2 H e , respectively. Then, the state ψ 2 H p ⊗H e of the
hydrogen atom as a composite system is said to be entangled when ψ 6¼ ψp ⊗ψe for
any pair of states ψp and ψe . This suggests that “entangled” is a property that
applies or not to the state of a composite system. However, the hydrogen atom can
also be represented as constituted by two different subsystems, the center-of-mass
system ψc 2 H c and the relative system ψr 2 H r , such that the state of the
hydrogen atom ψ 2 H c ⊗H r can be obtained as ψ ¼ ψc ⊗ψr : Now the state of
the composite system is not entangled. Although conceiving the hydrogen atom as
being composed of a proton and an electron seems more natural, there are group
reasons that may lead to considering that the decomposition in a center-of-mass
system and a relative system is more fundamental (see Ardenghi, Castagnino, and
Lombardi 2009). This means that it cannot be said that a state of a composite
system is entangled or not without first deciding which decomposition of the
system will be considered.
John Earman stresses this fact by saying:
[A] state may be entangled with respect to one decomposition but not another; hence,
unless there is some principled way to choose a decomposition, entanglement is a radically
ambiguous notion.
(Earman 2015: 303)
As a consequence, it is necessary to single out the “correct” decomposition, and

two positions can be distinguished (Earman 2015: 324–327). For the realist, there
are certain subsystems that are ontologically “real” systems, whereas others are
merely fictional. For the pragmatist, by contrast, the legitimate criterion for decom-
position is empirical accessibility.
Although in certain passages of his article Earman talks about relativity, the
stronger idea is that of the “rampant ambiguity” of the notion of entanglement
(2015: 324, 325, 327). A notion is ambiguous if it has more than one meaning; so,
in science and in philosophy ambiguity must be avoided. Therefore, if the notion of
entanglement is ambiguous, the need for a clear-cut decision about how to split the
composite system into subsystems seems completely reasonable. Nevertheless, a
different view is possible: The notion of entanglement is not ambiguous; it is
relative to the decomposition. The difference between ambiguity and relativity is
not irrelevant at all. Whereas the first is a conceptual problem to be solved, the
second is a common feature of physical concepts. In fact, the concept of velocity is
not ambiguous because it is relative to a reference frame. In the same sense,
entanglement is a notion that acquires a precise meaning when relativized to a
certain partition of the composite system and, as a consequence, no absolute
criterion to select the right decomposition is needed.
The relative conception of entanglement invites us to reverse the general
approach to quantum mechanics – from the traditional, classically inspired
bottom-up view, to a top-down view that endows the composite system with
ontological priority. From this perspective, even if two systems exist independ-
ently before interaction, after the interaction their existence is only derivative, they
become components of the composite system on a par with other subsystems
resulting from any different decomposition. This view finds a significant affinity
with the so called quantum structure studies, which deal with the different ways in
which a quantum system can be decomposed into subsystems according to differ-
ent tensor product structures (Harshman and Wickramasekara 2007a,b, Jeknić-
Dugić, Arsenijević, and Dugić 2013, Arsenijević, Jeknić-Dugić, and Dugić 2016,
Harshman 2016).
But the top-down view can be generalized a step further. Up to this point, the
relation between “top” and “down” was described in terms of decomposing the
composite system into its subsystems: The result of decomposition are subsystems,
represented by Hilbert spaces; the tensor product of the Hilbert spaces of the
subsystems is the Hilbert space of the composite system. But the top-down
relationship can also be conceptualized in terms of algebras of observables, in
resonance with the algebraic approach to quantum mechanics (Haag 1992). The
whole system, represented by its algebra of observables, can be partitioned into
different parts, identified by the subalgebras, even when these subalgebras do not
correspond to subsystems represented by Hilbert spaces. This perspective, released
from the subsystem-dependent view anchored in tensor product structures, was
proposed by Howard Barnum and colleagues (2003) as the basis for a
generalization of the notion of entanglement to partitions of algebras. This gener-
alized notion becomes the usual notion of entanglement when the partition of the
algebra of the whole system defines a decomposition of the system into subsystems
(Barnum et al. 2004, Viola et al. 2005; Viola and Barnum 2010). A further
characterization of pure entangled states can be given by appealing to the notion of

restriction to a subalgebra, which is a natural algebraic generalization of the partial
trace operation (Balachandran et al. 2013a,b). As a consequence, entanglement is
not a relationship between systems or states, but between algebras of observables
(Harshman and Ranade 2011).
At present, this subsystem-independent view has been formally studied with
great detail in many works and is still in development. However, the point that we
want to stress here is that this view suggests a top-down closed-system ontological
picture, according to which the whole closed system is the only autonomous entity:
The subentities represented by subalgebras of the whole algebra of observables are
only partial perspectives of the closed system without autonomous existence. In the
following section we will show that the phenomenon of decoherence can be
explained from this top-down closed-system view, which, in turn, leads to a
generalized approach to decoherence.
17.3 Environment-Induced Decoherence from a Closed-System Perspective

17.3.1 What Are the Systems That Decohere?
The environment-induced-decoherence program quickly became a new orthodoxy in
the physicists’ community (Bub 1997). Despite this, the program is still threatened by
a serious conceptual problem, which is precisely derived from its open system.
According to the orthodox view, the first step is to split the universe into the
degrees of freedom that are of direct interest for the observer, “the system of
interest,” and the remaining degrees of freedom that are usually referred to as “the
environment.” In many models, distinguishing between the system of interest and its
environment seems to be a simple matter. This is the case in many typical applica-
tions of the decoherence formalism to spin-bath models – devoted to study the
behavior of a particle immersed in a large “bath” of many particles (see, e.g., Zurek
1982). But the environment can also be internal, such as phonons or other inside
excitations. This is typically the case when the formalism is applied to cosmology:
The universe is split into some degrees of freedom representing the system, and the
remaining degrees of freedom that are supposed to be nonaccessible and, therefore,
play the role of the environment (see, e.g., Calzetta, Hu, and Mazzitelli 2001). The
possibility of internal environments leads to the need for a general criterion to
distinguish between the system and its environment. The problem is that the envir-
onment-induced-decoherence program does not provide such a criterion. Wojciech
Zurek recognized this shortcoming of his proposal early on:
one issue which has been often taken for granted is looming big, as a foundation of the
whole decoherence program. It is the question of what are the “systems” which play such a
crucial role in all the discussions of the emergent classicality. This issue was raised earlier,
but the progress to date has been slow at best. Moreover, replacing “systems” with, say,
“coarse grainings” does not seem to help at all, we have at least tangible evidence of the
objectivity of the existence of systems, while coarse-grainings are completely “in the eye
of the observer.”
(Zurek 2000: 338; see also Zurek 1998).
It is quite clear that the problem can be removed from a top-down closed-system
perspective as that delineated in the previous section.
In order to explain decoherence from a closed-system perspective, let us begin
by recalling the definition of the concept of reduced state, because the environ-
ment-induced-decoherence program decides to study the time behavior of the
reduced state of the system of interest. The reduced state ρr1 of a system S1 ,
subsystem of a system S, is defined as the density operator by means of which
the expectation values of all the observables of S belonging exclusively to S1 can
be computed. As Maximilian Schlosshauer emphasizes, strictly speaking, a
reduced density operator is only a “calculational tool” for computing expectation
values (Schlosshauer 2007: 48). This means that the description of decoherence in
terms of the reduced state of the open system is conceptually equivalent to the
description in terms of the expectation values of the observables of the open
system but viewed from the perspective of the whole closed system. This is the
path we will follow here.
17.3.2 The Perspective of the Closed System

Let us consider a closed system U partitioned as U ¼ S [ E, where S is the open
system of interest and E is the environment. Let us call OU the space of observ-
ables of U, and OS and OE the spaces of observables of S and E, respectively; then
OU ¼ OS ⊗OE . If ρU is the state of U, the reduced state of S can be computed by
means of the partial trace as ρS ¼ Tr E ρU . The environment-induced-decoherence
formalism proves that, in many physically relevant models with environments of
many degrees of freedom, the nondiagonal terms of the reduced state ρS ðtÞ rapidly
tend to vanish after an extremely short decoherence time t D :
tt D
ρS ðt Þ ! ρdS ðtÞ (17.1)
where ρdS ðt Þ is diagonal in the preferred basis of OS . The evolution just described
expresses the following evolution in the expectation values of the observables
OS 2 OS of the open system S:
ttD
hOS iρS ðtÞ ! hOS iρd ðtÞ (17.2)
S
But, by definition, ρS is the density operator by means of which the expectation

values of all the observables OS 2 OS in the state ρS can be computed, that is:
8ðOUS ¼ OS ⊗I E Þ 2 OU hOUS iρU ¼ hOS iρS (17.3)
where I E 2 OE is the identity of the space of observables of the environment E.

Then, it is clear that even when the task is to describe only S, its reduced state is not
indispensable. The physically relevant information about that subsystem can also
be obtained by studying the state ρU of the whole closed system U and its relevant
observables OUS ¼ OS ⊗I E . This means that there is no difference between describ-
ing the open system S by means of its reduced state ρS and describing it from a
closed-system perspective by means of the expectation values of the relevant
observables OUS of the closed composite system U in the state ρU . Therefore,
the evolution of Eq. (17.3) can be expressed from the viewpoint of the closed
system U as:
tt D
hOUS iρU ðtÞ ! hOUS iρd ðtÞ (17.4)
U
where ρdU ðtÞ is not completely diagonal, but is diagonal in the preferred basis
of OS .
17.3.3 The Emergence of Classicality

The emergence of classicality through decoherence can be explained strictly in
terms of expectation values. The general idea is that the expectation value of an
observable O when the system is in the certain state ρ can be expressed as:
X X
hOiρ ¼ Oii ρii þ Oij ρij (17.5)
i i6¼j
where the ρii and the Oii are the diagonal components, and the ρij and the Oij are the
nondiagonal components of ρ and O, respectively, in a certain basis. The second
sum of Eq. (17.5) represents the specifically quantum interference terms of the
expectation value. If those terms vanished, the expectation value would adopt the
structure of a classical expectation value, where the Oii might be interpreted as
possible values, and the ρii might play the role of probabilities, since they are
positive numbers that are less than or equal to one and sum to one.
In the light of this idea, the process of decoherence described by the evolution of
Eq. (17.2) leads to a classical-like expectation value, since ρdS ðtÞ is diagonal in the
preferred basis of OS :
tt D X
hOS iρS ðtÞ ! hOS iρd ðtÞ ¼ OSii ρdSii ðtÞ (17.6)
S
i
where the ρSii and the OSii are the diagonal components of ρS and OS , respectively,
in the preferred basis.
However, the same move cannot be applied to the evolution as expressed in
Eq. (17.4), because ρdU ðt Þ is not completely diagonal: It is diagonal only in the
components corresponding to the preferred basis of OS . Nevertheless, decoherence
can be described from the closed-system perspective analogously to Eq. (17.6) if a
coarse-grained state ρG ðt Þ of the closed system U is defined as the operator such
that:
8ðOUS ¼ OS ⊗I E Þ 2 OU hOUS iρU ðtÞ ¼ hOUS iρG ðtÞ (17.7)
The density operator ρG represents a coarse-grained state because it can be

obtained as ρG ¼ Π ρU ¼ Π Π ρU . The projector Π performs the following
operation:
Π ρU ¼ ðTr E ρU Þ⊗~δ E ¼ ρS ⊗δ~E (17.8)

where δ~E 2 OE is a normalized identity operator with coefficients
~δ Eαβ ¼ δαβ =P δγγ (see Fortin and Lombardi 2014). Now, the process of decoher-
γ
ence can be expressed as
ttD
hOUS iρU ðtÞ ! hOUS iρd ðtÞ (17.9)
G
where ρdG ðtÞ remains completely diagonal for all times t t D . Now it can be said
that the expectation value also acquires a classical form from the closed-system
perspective since:
tt D X
hOUS iρU ðtÞ ! hOUS iρd ðtÞ ¼ OUSii ρdGii ðt Þ (17.10)
G
i
where the ρdGii and the OUSii are the diagonal components of ρdG and OUS ,
respectively, in the basis of decoherence. It is quite clear that ρG , although
operating onto OU , is not the quantum state of U: It is a coarse-grained state
of the closed system that disregards certain information of its quantum state.
However, ρG supplies the same information about the open system S as the
reduced state ρS , but now from the viewpoint of the composite system S. In fact,
if the degrees of freedom of the environment are traced off, the reduced state ρS is
obtained:
Tr E ρG ¼ ρS (17.11)
Therefore, the reduced density operator ρS can also be conceived of as a kind of
coarse-grained state of U, which disregards certain degrees of freedom considered
as irrelevant.
17.3.4 The Applications of the Closed-System Approach

The closed-system approach was presented from different perspectives, from the
more conceptual (Castagnino, Laura, and Lombardi 2007, Lombardi, Fortin, and
Castagnino 2012), to the more technical (Castagnino and Lombardi 2005, Castag-
nino and Fortin 2011, Fortin, Lombardi, and Castagnino 2014). It was also applied
to a generalization of the spin-bath model (Castagnino, Fortin, and Lombardi
2010): A generalized spin-bath model of m þ n spin-1/2 particles, where the m
particles interact with each other and the n particles also interact with each other,
but the particles of the m group do not interact with those of the n group. The study
of the model shows that there are definite conditions under which all the particles
decohere, but neither the system composed of the m group nor the system
composed of the n group decoheres.
Once decoherence is understood from this new perspective, the defining-system
problem, that is, the problem that there is no criterion to distinguish between the
system and the environment, disappears. In fact, the same closed system can be
decomposed in many different ways. Since there is no privileged or “essential”
decomposition, there is no need for an unequivocal criterion to decide where to
place the cut between “the” system and “the” environment. If all the ways of
selecting the system of interest are equally legitimate, decoherence is relative to the
decomposition of the whole system (Lombardi et al. 2012, see also Lychkovskiy
2013). In other words, Zurek’s “looming big” problem is not a real threat to the
environment-induced-decoherence approach: The supposed challenge dissolves
once it is understood that decoherence is not a yes-or-no process but a relative
phenomenon.
17.4 The Top-Down Approach to Decoherence

17.4.1 The Formalism
In the previous section, the closed-system approach to decoherence was still
discussed in terms of the possibility of different tensor product structures: Deco-
herence is relative to the particular decomposition of the composite system into
subsystems. In this section, the generalization will be taken a step further from the
algebraic viewpoint, by admitting that a closed system may be partitioned into
parts that do not constitute subsystems.
The starting point of the algebraic approach to quantum mechanics (Haag 1992;
see also Bratteli and Robinson 1987) is the algebra of observables AðOÞ, which is
the algebra spanned by a certain set O of observables O represented by self-adjoint
operators mapping a suitable Hilbert space H onto itself. When the algebra AðOÞ
identifies a quantum system, the quantum state ω of the system is a prescription of
the expectation values of the observables, and it is formalized as an expectation

value functional from the observables to the unit interval, ω : AðOÞ ! ½0; 1.
A quantum state is said to be normal when there is an associate density operator
ρω (with ρω 0 and Tr ρω ¼ 1) acting on the same Hilbert space H and such that
ωðOÞ ¼ Tr ðOρω Þ. The expectation value ωðOÞ gives the expected value if one
measures the observable O when the system is in the state ω, and the equation
ωðOÞ ¼ Tr ðOρω Þ is essentially the Born Rule extended to mixed states.
The algebraic notions just stated are sufficient to formulate a top-down approach
to decoherence that is independent of the tensor product structures of the Hilbert
spaces. Let us consider a closed system U identified by its algebra of observables
AðOU Þ, and its state, represented by the density operator ρU . Now U is not
decomposed into subsystems, but a certain set of relevant observables OR is
selected. It is interesting to notice that this move agrees with the approaches of
the first period in the historical development in the general program of decoherence
(see Fortin et al. 2014), when the aim was to understand how classical macroscopic
properties emerge from the quantum microscopic evolution of a closed system. In
this first period, the approach to equilibrium of quantum systems was studied from
the behavior of certain observables that supposedly should behave classically
because they are accessible from the macroscopic viewpoint, e.g., “gross observ-
ables” (van Kampen 1954), “macroscopic observables of the apparatus” (Daneri,
Loinger, and Prosperi 1962). In the present case, no restriction is imposed on the
selection of the relevant observables: Any set of observables can be selected. In
any case, the algebra of the relevant observables, subalgebra of AðOU Þ, will be
considered: AðOR Þ AðOU Þ.
Once the relevant observables are selected, the second step consists in comput-
ing the expectation values of the observables of the relevant algebra AðOR Þ:
8OR 2 AðOR Þ hOR iρU ðtÞ (17.12)
Then, a coarse-grained state ρG ðt Þ is defined, such that:

8OR 2 AðOR Þ hOR iρU ðtÞ ¼ hOR iρG ðtÞ (17.13)
Now, the nonunitary evolution (governed by a master equation) of this expectation

value is computed. Decoherence occurs when, after an extremely short decoher-
ence time t D , the expectation acquires a particular form:
tt D
hOR iρU ðtÞ ¼ hOR iρG ðtÞ ! hOR iρd ðtÞ (17.14)
G
where ρdG ðtÞ remains diagonal in the preferred basis for all times t t D . This
means that, although the off-diagonal terms of ρU ðt Þ never vanish through its
unitary evolution, it might be said that the system decoheres relatively to the
observational point of view given by any observable belonging to the algebra of

the relevant observables AðOR Þ.
17.4.2 Classically-Behaving Observables

Let us recall that decoherence has been considered the essential element to explain
the emergence of classicality from the quantum world. But if decoherence is a
relative phenomenon, classicality also seems to be relative – the fact that a system
behaves classically or not cannot depend on the way in which the observer decides
to split the original closed system into relevant and irrelevant observables. This
situation also challenges the orthodox open-system approach: In certain situations
the fact that classicality emerges in an open system or not depends on what
composite system that open subsystem is embedded in. More precisely, given
two partitions of a closed system U, U ¼ S1 [ E1 and U ¼ S2 [ E 2 , it may be the
case that S1 and S2 decohere and behave classically, but S1 [ S2 does not decohere
and, so, classicality does not emerge in it (see the model in Castagnino et al. 2010).
This is a difficulty if one considers that the classical world is objective, independ-
ent of any observer’s decision. Recall Zurek’s rejection of any solution of the
defining-system problem that relies on “the eye of the observer” (Zurek
2000: 338).
Despite what it seems, the top-down view of decoherence based on the algebraic
approach is not affected by this difficulty. Given the closed system U, saying that it
decoheres from the perspective of the relevant observables OR 2 AðOR Þ amounts
to saying that, after a very short decoherence time, the interference terms of the
expectation values of those observables tend to vanish with the unitary time-
evolution of the state ρU of U. But the vanishing of the interference terms of the
expectation values of an observable is not a relative fact that depends on the
observer: What depends on the observer is the selection of the relevant observables
with the purpose to see whether the closed system decoheres relative to it or not.
When this fact is understood, it turns out to be clear that all the observables of
the closed system U can be considered one by one, their trivial algebras can be
defined, and the decoherence of the system U relative to each one of those algebras
can be studied. As a result, one is in a position to know the set of all the
observables of U that behave classically after a certain time, with neither ambiguity
nor relativity.
Another difficulty of the orthodox approach that is not usually stressed is that
certain systems have a classical behavior with respect to certain observables and a
quantum behavior with respect to others. For instance, a transistor behaves classic-
ally with respect to its center of mass when it falls off the table, but it also has the
quantum behavior characteristic of its specific use. When decoherence is conceived
as a phenomenon that occurs or not to a quantum system, these common situations

cannot be accounted for. By contrast, the top-down approach that relies on the
subalgebras of observables can easily explain how a single system may combine
classical and quantum behaviors of its different observables.
In summary, according to the explanation of the emergence of the classical
world given by the top-down algebraic approach just proposed, strictly speaking,
classicality is not a property of systems: Thinking of systems that become classical
in their whole leads to the difficulties mentioned earlier. The difficulties can be
overcome once it is recognized that classicality is a property of observables. The
emergent classical world is, then, the world described by the observables that
behave classically with respect to their expectation values.

In this chapter we have proposed a closed-system approach to decoherence which,
at first sight, seems to be a rival of the orthodox open-system approach. However,
as we have argued, our proposal is compatible with the environment-induced-
decoherence view, but generalizes it by including the treatment of situations that
could not be studied with that orthodox view.
As already explained, this closed-system approach is in resonance with a top-
down view of quantum mechanics, usually based on the algebraic formalism,
which is gaining ground in the physics community. It is also interesting to notice
that understanding decoherence from the viewpoint of a closed system represented
by its algebra of observables stands in close agreement with the modal-
Hamiltonian interpretation of quantum mechanics (Lombardi and Castagnino
2008, Ardenghi et al. 2009, Lombardi, Castagnino and Ardenghi 2010; see also
Chapter 2) also developed in our research group. This interpretation, also based on
the algebraic approach, makes the rule that selects the definite-valued observables
to depend on the Hamiltonian of the closed system. Moreover, the definition of the
system in terms of its algebra of observables leads to an ontological picture where
quantum systems are bundles of properties without individuality (da Costa, Lom-
bardi, and Lastiri 2013, da Costa and Lombardi 2014, Lombardi and Dieks 2016).
In summary, the general view that endows closed systems with ontological priority
has different but converging manifestations, in the light of which it deserves to be
further developed.
Acknowledgments
We are grateful to the participants of the workshop Identity, indistinguishability
and non-locality in quantum physics (Buenos Aires, June 2017) for their
interesting comments. This work was made possible through the support of Grant
57919 from the John Templeton Foundation and Grant PICT-2014–2812 from the
National Agency of Scientific and Technological Promotion of Argentina.
References
Ardenghi, J. S., Castagnino, M., and Lombardi, O. (2009). “Quantum mechanics:
Modal interpretation and Galilean transformations,” Foundations of Physics, 39:
1023–1045.
Arsenijević, M., Jeknić-Dugić, J., and Dugić, M. (2016). “A top-down versus a bottom-up
hidden-variables description of the Stern-Gerlach experiment,” pp. 469–484 in R. E.
Kastner, J. Jeknić-Dugić, and G. Jaroszkiewicz (eds.), Quantum Structural Studies:
Classical Emergence from the Quantum Level. Singapore: World Scientific.
Bacciagaluppi, G. (2016). “The role of decoherence in quantum mechanics,” in E. N. Zalta
(ed.), The Stanford Encyclopedia of Philosophy (Fall 2016 Edition). https://plato
.stanford.edu/archives/fall2016/entries/qm-decoherence/
Bacciagaluppi, G. and Dickson, M. (1999). “Dynamics for modal interpretations,” Foun-
dations of Physics, 29: 1165–1201.
Balachandran, A. P., Govindarajan, T. R., de Queiroz, A. R., and Reyes-Lega, A. F.
(2013a). “Algebraic approach to entanglement and entropy,” Physical Review A,
88: 022301.
Balachandran, A. P., Govindarajan, T. R., de Queiroz, A. R., and Reyes-Lega, A. F.
(2013b). “Entanglement and particle identity: A unifying approach,” Physical Review
Letters, 110: 080503
Barnum, H., Knill, E., Ortiz, G., Somma, R., and Viola, L. (2003). “Generalizations of
entanglement based on coherent states and convex sets,” Physical Review A, 68:
032308.
Barnum, H., Knill, E., Ortiz, G., Somma, R., and Viola, L. (2004). “A subsystem-inde-
pendent generalization of entanglement,” Physical Review Letters, 92: 107902.
Bratteli, O., and Robinson, D. W. (1987). Operator algebras and quantum statistical
mechanics 1, 2nd edition. New York: Springer.
Calzetta, E., Hu, B. L., and Mazzitelli, F. (2001). “Coarse-grained effective action and
renormalization group theory in semiclassical gravity and cosmology,” Physics
Reports, 352: 459–520.
Castagnino, M. and Fortin, S. (2011). “New bases for a general definition for the moving
preferred basis,” Modern Physics Letters A, 26: 2365–2373.
Castagnino, M., Fortin, S., and Lombardi, O. (2010). “Suppression of decoherence in a
generalization of the spin-bath model,” Journal of Physics A: Mathematical and
Theoretical, 43: 065304.
Castagnino, M., Laura, R., and Lombardi, O. (2007). “A general conceptual framework for
decoherence in closed and open systems,” Philosophy of Science, 74: 968–980.
Castagnino, M. and Lombardi, O. (2005). “Decoherence time in self-induced decoher-
ence,” Physical Review A, 72: 012102.
da Costa, N. and Lombardi, O. (2014). “Quantum mechanics: Ontology without individ-
uals,” Foundations of Physics, 44: 1246–1257.
da Costa, N., Lombardi, O., and Lastiri, M. (2013). “A modal ontology of properties for
quantum mechanics,” Synthese, 190: 3671–3693.
Daneri, A., Loinger, A., and Prosperi, G. (1962). “Quantum theory of measurement and
ergodicity conditions,” Nuclear Physics, 33: 297–319.
Earman, J. (2015). “Some puzzles and unresolved issues about quantum entanglement,”
Erkenntnis, 80: 303–337.
Elby, A. (1994). “The ‘decoherence’ approach to the measurement problem in quantum
mechanics,” Proceedings of the 1994 Biennial Meeting of the Philosophy of Science
Association, 1: 355–365.
What do reduced states refer to?”, Foundations of Physics, 44: 426–446.
Fortin, S., Lombardi, O., and Castagnino, M. (2014). “Decoherence: A closed-system
approach,” Brazilian Journal of Physics, 44: 138–153.
Haag, R. (1992). Local Quantum Physics: Fields, Particles, Algebras. Berlin: Springer.
Harshman, N. (2016). “Symmetry and natural quantum structures for three-particles in
one-dimension,” pp. 373–400 in R. E. Kastner, J. Jeknić-Dugić, and G. Jaroszkiewicz
(eds.), Quantum Structural Studies: Classical Emergence from the Quantum Level.
Singapore: World Scientific.
Harshman, N. and Ranade, K. (2011). “Observables can be tailored to change the entangle-
ment of any pure state,” Physical Review A, 84: 012303.
Harshman, N. and Wickramasekara, S. (2007a). “Galilean and dynamical invariance of
entanglement in particle scattering,” Physical Review Letters, 98: 080406.
Harshman, N. and Wickramasekara, S. (2007b). “Tensor product structures, entanglement,
and particle scattering,” Open Systems & Information Dynamics, 14: 341–351.
Healey, R. (1995). “Dissipating the quantum measurement problem,” Topoi, 14: 55–65.
Jeknić-Dugić, J., Arsenijević, M., and Dugić, M. (2013). Quantum Structures: A View of
the Quantum World. Saarbrücken: Lambert Academic Publishing.
Lombardi, O. and Dieks, D. (2016). “Particles in a quantum ontology of properties,”
pp. 123–143 in T. Bigaj and C. Wüthrich (eds.), Metaphysics in Contemporary
Physics. Leiden: Brill-Rodopi.
Lombardi, O., Fortin, S., and Castagnino, M. (2012). “The problem of identifying the
system and the environment in the phenomenon of decoherence,” pp. 161–174 in
H. W. de Regt, S. Hartmann, and S. Okasha (eds.), Philosophical Issues in the
Sciences Vol. 3. Berlin: Springer.
Lychkovskiy, O. (2013). “Dependence of decoherence-assisted classicality on the way a
system is partitioned into subsystems,” Physical Review A, 87: 022112.
Paz, J. P. and Zurek, W. H. (2002). “Environment-induced decoherence and the transition
from quantum to classical,” pp. 77–148, in D. Heiss (ed.), Fundamentals of Quantum
Information: Quantum Computation, Communication, Decoherence and All That.
Heidelberg-Berlin: Springer.
Schlosshauer, M. (2007). Decoherence and the Quantum-to-Classical Transition. Berlin:
Springer.
Schrödinger, E. (1935). “Discussion of probability relations between separated systems,”
Proceedings of the Cambridge Philosophical Society, 31: 555–563.
van Kampen, N. G. (1954). “Quantum statistics of irreversible processes,” Physica, 20:
603–622.
Viola, L. and Barnum, H. (2010). “Entanglement and subsystems, entanglement beyond

subsystems, and all that,” pp. 16–43 in A. Bokulich and G. Jaeger (eds.), Philosophy
of Quantum Information and Entanglement. Cambridge: Cambridge University Press.
Viola, L., Barnum, H., Knill, E., Ortiz, G., and Somma, R. (2005). “Entanglement beyond
subsystems,” Contemporary Mathematics, 381: 117–130.
Zurek, W. H. (1982). “Environment-induced superselection rules,” Physical Review D, 26:
1862–1880.
Zurek, W. H. (1993). “Preferred states, predictability, classicality and the environment-
induced decoherence,” Progress of Theoretical Physics, 89: 281–312.
Zurek, W. H. (1998). “Decoherence, einselection, and the existential interpretation,”
Philosophical Transactions of the Royal Society A, 356: 1793–1820.
Zurek, W. H. (2000). “Decoherence and einselection,” pp. 309–342 in P. Blanchard,
D. Giulini, E. Joos, C. Kiefer, and I.-O. Stamatescu (eds.), Decoherence: Theoretical,
Experimental, and Conceptual Problems. Berlin-Heidelberg: Springer-Verlag.
Zurek, W. H. (2003). “Decoherence, einselection, and the quantum origins of the clas-
sical,” Reviews of Modern Physics, 75: 715–776.
18
A Logical Approach to the
Quantum-to-Classical Transition
sebastian fortin, manuel gadella, federico holik, and
marcelo losada
18.1 Introduction
The description of the classical limit of a quantum system is one of the most
important issues in the foundations of quantum mechanics (see Cohen 1989). This
problem has been formulated in different ways and explained by appealing to
different interpretations (see Schlosshauer 2007). The attempts to explain the
classical limit go back to the correspondence principle, proposed by Niels Bohr.
This principle establishes a connection between quantum observables and their
classical counterparts when Planck’s constant is small enough in comparison with
relevant quantities of the quantum system. In particular, this happens in the limit of
large quantum numbers.
Nowadays, the most important approach to describe the classical limit is based
on the decoherence process (see Schlosshauer 2007). The general idea of this
approach is to explain the disappearance of the interference terms of quantum
states by appealing to the decoherence process induced by the environment. In this
way, the coherence needed for most typical quantum phenomena is lost, and the
classical features appear instead.
As is well known, the set of observables associated with a quantum system
forms a noncommutative algebra. This differs from the classical description of
physical systems, in which observables are represented by functions over a phase
space, which form a commutative algebra. This difference between quantum and
classical systems has a correlate in terms of the elementary properties of physical
systems. The elementary properties of quantum systems (also known as “yes-no
tests” or elementary experiments) are represented by orthogonal projectors acting
on a Hilbert space. These projectors form a non-Boolean lattice (more specifically,
a complete, atomic, atomistic, orthomodular lattice, satisfying the covering law,
see Kalmbach 1983). Instead, the elementary properties of a classical system are
the measurable subsets of the phase space, which form a Boolean lattice.
360
A Logical Approach to the Quantum-to-Classical Transition 361
The decoherence approach to the quantum-to-classical transition is based on the

Schrödinger picture, in which states evolve over time, while observables and
physical properties are taken to be constants. As a result, the structure of quantum
properties remains the same for all times: The quantum logic associated with the
system does not change (Bub 1997). Therefore, in this approach it is not explained
how the structure of quantum properties becomes classical. However, as it was
remarked in Fortin and Vanni (2014), a reasonable condition for the existence of a
classical limit is that the lattice of elementary properties becomes Boolean, or
equivalently, that the algebra of observables becomes commutative (Fortin and
Vanni 2014, Fortin, Holik, and Vanni 2016, Losada, Fortin, and Holik 2018).
In this chapter, we present a logical approach to the classical limit, which
describes how the logical structure of the elementary properties of a quantum
system becomes classical when the classical limit is reached. In order to describe
the evolution of logical structure, we consider the Heisenberg picture. According to
this picture, observables and physical properties evolve in time, while states remain
constant. In this way, we can consider the algebra of observables and the lattice of
elementary properties as dynamical objects, depending on time or other relevant
parameters, such as action, temperature, particle number, or energy.
As we will show later, this offers an interesting perspective for studying
different physical processes. In particular, we discuss the possibility of connecting
the approach of dynamical algebras developed in recent papers (Fortin and Vanni
2014, Fortin et al. 2016, Losada et al. 2018) with the description of the classical
limit based on deformation of algebras. We also discuss the case of quantum
statistical mechanics, where intermediate logics are interpreted as phase transitions.
The chapter is organized as follows. In Section 18.2, we review the problem of
the classical limit as it was traditionally considered in the literature. In Section
18.3, we briefly summarize the logical structure of the elementary properties of
classical and quantum physical systems, and we discuss the main difference
between both logical structures. In Section 18.4, we introduce the logical approach
to the classical limit, and we illustrate this approach with four different examples.
Finally, in Section 18.5 we draw our conclusions.
18.2 Different Approaches to the Classical Limit

One of the first explanations was proposed by Niels Bohr, who appealed to the
correspondence principle. This principle establishes a connection between quantum
observables and their classical counterparts by asserting that, if the ratio between the
action of the system and Planck’s constant is large enough, the classical limit should
be recovered. This implies that the quantum-to-classical transition should be
attained in the limit of large quantum numbers, such as large orbits, large energies,
362 Sebastian Fortin, Manuel Gadella, Federico Holik, and Marcelo Losada
or a large number of particles. A result that goes in line with the correspondence
principle is the Ehrenfest theorem. This approach is still important today, in particu-
lar for studying quantum phenomena in the semiclassical level.
Paul Dirac proposed another explanation of the classical limit, appealing to the
destructive interference among all the possible paths of the physical system (Dirac
1933). In this way, he showed that the classical action path has the dominant
contribution. This idea was subsequently elaborated by Richard Feynman (1942)
in his thesis, opening the door to the celebrated path-integral formulation of
quantum mechanics.
All these approaches presented problems, which where extensively discussed in
the literature. In particular, it is important to remark that Bohr himself did not
considered the classical limit as an explanation of the emergence of classical
reality. Quite on the contrary, Bohr believed that the classical realm exists inde-
pendently of quantum theory and cannot be derived from it. As is well known, the
discussion about the classical limit is subtle and problematic, and there is no real
agreement on a solution.
Nowadays, the most important approach for describing the classical limit is
based on the environment-induced decoherence. In this approach, it is considered
that the quantum-to-classical transition is the result of the loss of coherence of the
system due to the interaction with its environment (Schlosshauer 2007). Many
physicists considered this proposal as the correct explanation of the classical limit
(and also of the measurement process); however, some objections were raised,
because the decoherence process would not explain how the logical structure of the
elementary properties becomes a classical logic.
Another important approach to the study of the classical limit is based on
algebras’ deformation (see Landsman 1993). In this formalism, quantum commu-
tators (or equivalently, Moyal brackets) reduce to Poisson brackets, deforming the
algebra involved.
In what follows, we present an alternative approach to describe the classical
limit. This is a logical approach, based on the evolution of the quantum observ-
ables, and it allows describing the quantum-to-classical transition of the logical
structure of the quantum systems. In the next section, we review some basic
features about the lattice of the elementary properties of classical and quantum
systems, which are relevant to our logical approach of the classical limit.
18.3 Logical Structure of Quantum Mechanics

In classical and quantum mechanics, the physical properties of a system are
endowed with a lattice structure. These structures are different in the classical
and quantum case, and they determine the logical structure of the physical system.
In classical mechanics, physical systems are represented by the phase space, and
their properties are represented by subsets of the phase space (see Kalmbach 1983).
In quantum mechanics, the physical systems are represented by Hilbert spaces and
the properties are represented by closed vector subspaces or by their corresponding
orthogonal projectors (von Neumann 1932; for a recent discussion about the logic
approach to quantum mechanics, see Domenech, Holik, and Massri 2010, Holik,
Massri, and Ciancaglini 2012, Holik, Massri, Plastino, and Zuberman 2013, Holik,
Plastino, and Sáenz 2014, Holik and Plastino 2015; for applications to quantum
histories see Omnès 1994, Losada, Vanni, and Laura 2013, 2016, Griffiths 2014,
Losada and Laura 2014a, b). In both cases, the set of all the properties of a system
has an orthocomplemented lattice structure. This implies that there is an order
relation () such that for each pair of properties there are an infimum (∧) and a
supremum (∨), and each property p has a complement p⊥ with adequate proper-
ties. All orthocomplemented lattices satisfy certain inequalities, called distributive
inequalities (see Kalmbach 1983):
a∧ðb∨cÞ ða∧bÞ∨ða∧cÞ
(18.1)
a∨ðb∧cÞ ða∨bÞ∧ða∨cÞ
When the equalities hold, the lattice is distributive. An orthocomplemented and
distributive lattice is called a Boolean lattice. The distributive property is an
essential feature that differentiates classical and quantum lattices of properties.
In the classical case, the properties of the system are represented by the subsets
of its phase space. The partial-order relation is given by the inclusion () of sets.
The infimum and the supremum are the intersection (\) and the union ([) of sets,
respectively, and the complement of a property p is the complement of sets pc. The
set of classical properties is not only an orthocomplemented lattice, but also a
distributive one, i.e., classical properties satisfy the distributive equalities. There-
fore, the logical structure of a classical system is Boolean. This structure is usually
called classical logic.
The quantum case is very different. The properties are represented by closed-
vector subspaces (or by their corresponding orthogonal projectors; von Neumann
1932). Thus, the logical structure of quantum systems is given by the algebraic
structure of closed subspaces. The set of all quantum properties is also an
orthocomplemented lattice, and, as in the classical case, the partial order relation
is given by the inclusion of subspaces, and the infimum is given by the intersection
of subspaces. However, the supremum and the complement of properties are
different from the classical ones. The supremum is given by the sum of subspaces
and the complement of a property is its orthogonal subspace. The resulting lattice
is nondistributive (see Kalmbach 1983), and therefore, it is not Boolean. This
structure is called quantum logic (Birkhoff and von Neumann 1936).
The distributive inequalities are the main difference between classical and
quantum logic. In the classical lattice, all properties satisfy the distributive equal-
ities, but in the quantum lattice, only distributive inequalities hold, in general.
However, for some subsets of quantum properties the equalities hold. When a
subset of properties satisfies the distributive equalities, they are called compatible
properties. It can be proved that a sufficient and necessary condition for a set of
properties to be compatible is that the projectors associated with the properties
commute. Moreover, it can be shown that properties associated with different
observables are compatible if the observables commute. If, on the contrary, two
observables do not commute, some of the properties associated with them are not
compatible. Therefore, by extension, commuting observables are called compatible
observables.
The differences between classical and quantum logic are of fundamental import-
ance for the classical limit problem. If a quantum system undergoes a physical
process such that its behavior becomes classical, then its logical structure of
properties should undergo a transition from quantum logic to classical logic, i.e.,
its lattice structure should become distributive. However, the description of the
classical limit of a quantum system usually focuses on the state of the system.
The mathematical description of this process does not explain how the logical
structure changes on time. Therefore, in these approaches it is not possible to
describe how the structure of quantum properties becomes classical. In order to
give an adequate approach to the classical limit, we need a description in which
observables and physical properties evolve over time, changing the logical struc-
ture of the system.
18.4 Logical Quantum-to-Classical Transition

A complete description of the quantum-to-classical transition should explain how
the logical structure of the system changes from a quantum logic to a classical
logic. In order to adequately describe this transition, we consider a quantum system
with a general
time evolution and a time-dependent set of relevant observables
^
OðtÞ ¼ Oi ðtÞgiϵI (where I is a set of indexes).
Each set of relevant observables OðtÞ generates an algebra of observables V ðtÞ,
and each algebra has associated an orthocomplemented lattice of properties L V ðtÞ .
We assume that in the initial set of observables Oð0Þ there are some incompatible
observables, and therefore its corresponding algebra of observables V ð0Þ is non-
commutative and the associated lattice of properties L V ð0Þ is nondistributive.
For a quantum system with a decoherence time tD , the quantum-to-classical
transition is characterized by a process that transforms noncommutative observ-
ables into commutative ones
h i h i
Oî ð0Þ; O^j ð0Þ 6¼ 0 ⟶ Oî ðt D Þ; O^j ðt D Þ ¼ 0, 8i, j: (18.2)
After time tD , the algebra of observables V ðtÞ becomes commutative, and the
corresponding orthomodular lattice L V ðtÞ becomes nondistributive. The logical
classical limit is expressed by the fact that, while L V ð0Þ is a nondistributive lattice,
L V ðtD Þ is a Boolean one. In this way, we obtain an adequate description of the
logical evolution of a quantum system.
In what follows, we discuss the dynamics of the quantum algebra of observables
and the logic structure of properties in some physical models.
18.4.1 Quantum Operations

A quantum operation is a linear and completely positive map from the set of
density operators into itself (Nielsen and Chuang 2000). For each time t, we
consider a quantum operation ℰt , which maps the initial state of the system ^ρ 0 to
the state at time t, i.e.,
ℰt ð^ρ 0 Þ ¼ ^ρ ðt Þ: (18.3)
If we use the sum representation, we can express the quantum operation ℰt as
follows (Nielsen and Chuang 2000):
X
ℰt ð^ρ 0 Þ ¼ ^ ðtÞ^ρ 0 E
E b †μ ðt Þ, (18.4)
μ μ
^ μ ðt Þ are the Kraus operators of ℰt .

where E
We define the Heisenberg representation of ℰt as the ~ t that evolves the
ℰ
operator

~ ^ ^
observables from an initial time up to time t, i.e., ℰ t O ¼ O ðt Þ. The operator ℰ ~t
must preserve the mean values of the observables for all times,
X
^ ¼ Tr
Tr ^ρ ðt ÞO ^ μ ðtÞ^ρ 0 E
E ^ ¼
b †μ ðt ÞO
μ
X
¼ Tr ^ρ 0 μ E Ê
b †μ ðt ÞO ^ ðt Þ :
^ μ ðt Þ ¼ Tr ^ρ 0 O (18.5)
Hence, we can express the operator ℰ ~ t as follows:

X
~t O
ℰ ^ ¼O ^ ðt Þ ¼ Ê
b † ðt ÞO
E ^ μ ðt Þ: (18.6)
μ μ

Since O^ is self-adjoint, then ℰ ~t O ~t
^ is also a self-adjoint operator. Therefore, ℰ
maps observables to observables.
Once we have defined the temporal evolution of quantum operations, we can
describe the logical classical limit of a quantum system as it was explained before.
We illustrate the logical approach with a simple example: the amplitude damping
channel.
The amplitude damping channel is useful for describing the energy dissipation
due to the environment effects. It is relevant for quantum information processing,
because it is an adequate model for quantum noise. In particular, this model can be
applied to the decay of an excited state of a two-level atom due to spontaneous
emission of photons. If the atom is in the ground state, no photon is emitted, and
the atom continues in the same state. But, if the atom is in the excited state, after an
interval of time τ, there is a probability p that the state has decayed to the ground
state and a photon has been emitted (see Nielsen and Chuang 2000).
The quantum operation of the amplitude damping channel can be expressed as
follows:
^ 0 ^ρ 0 E
ℰτ ð^ρ 0 Þ ¼ E ^ 1 ^ρ 0 E
^ †0 þ E ^ †1 , (18.7)
where the Kraus operators are
! pffiffiffi !
1 0 0 p
^0 ¼
E ^1 ¼
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi , E : (18.8)
0 ð1 pÞ 0 0
The quantum map ℰ ~ τ , acting on the space of observables is given by

associated

~τ O
ℰ ^ ¼E b 0O
† Ê
^0 þ E
b 1O
† Ê^ 1 . In matrix form, we have
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !
O00 ð1 pÞO01
~τ O
ℰ ^ ¼ pffiffiffiffiffiffiffiffiffiffiffi : (18.9)
1 pO10 pO00 þ ð1 pÞO11

Applying the amplitude damping channel n times, we obtain the map ℰ~nτ O ^ ,
which has the form
0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffinffi 1
O 00 ð1 pÞ O01
ℰ~nτ O
^ ¼ @ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A: (18.10)
n n n
ð1 pÞ O10 ð1 pÞ O11 þ O00 1 ð1 pÞ
If n ⟶ ∞, all the observables become proportional to the identity. This implies that
the algebra of observables becomes trivially commutative, and its corresponding
lattice of properties becomes a classical logic.
18.4.2 Rigged Hilbert Space

One of the most investigated fields in quantum foundations is the quantization
problem, which consists in obtaining quantum observables from their classical
counterpart. Much less considered has been the problem of dequantization: the
transition from quantum to classical observables. It was shown (Castagnino and
Gadella 2006) that dequantization may require two steps, one is a type of deco-
herence and the other is the notion of macroscopicity, which is implemented by the
limit ℏ ! 0. In the present section, we intend to give a brief account of another
notion of dequantization in the presence of unstable quantum systems or
resonances.
In a previous paper (Fortin et al. 2016), we argued that an essential characteristic
of the quantum-to-classical transition should be the transition from a noncommu-
tative algebra of observables to a commutative one, when t ⟶ þ ∞. This can be
rigorously formulated for unstable quantum systems, provided we considered the
linear space spanned by the resonance state vectors, also called Gamow vectors.
Resonances are usually characterized as poles of some analytic continuations of
a reduced resolvent or a scattering matrix. Both formulations are not always
equivalent – one may construct models for which the poles in one of these two
formulations are not poles in the other. In the energy representation, these poles
appear in complex conjugate pairs and have the form E R iΓ=2, where E R is the
resonance energy and Γ is related to the inverse of the mean lifetime (see Bohm
1993). Notice that Γ must be always positive.
From the observational point of view, single resonances show an exponential
decay, provided that the time intervals are not too short and not very large either
(Fonda, Ghirardi, and Rimini 1978). However, these deviations are very difficult to
observe. Therefore, most experiments with resonances show exponential decays
for practically all values of time (Fischer, Gutierrez-Medina, and Raizen 2001,
Rothe, Hintschich, and Monkman 2006).
Now, pure stable states have a mathematical representation in terms of vector
states. The difference between a stable state and a resonance state is just that
the value of the parameter Γ is equal to zero for stable states. Then, one is
tempted to introduce a definition of resonance states in such a way that, if the
resonance poles are ER iΓ=2, we have either H jψ D i ¼ ðE R iΓ=2Þjψ D i or
H jψ G i ¼ ðE R þ iΓ=2Þjψ G i (Nakanishi 1958). Here H ¼ H 0 þ V is the total
Hamiltonian which produces the resonance phenomenon. Note that, in the first
case, formal time evolution gives eitH jψ D i ¼ eiER t etΓ=2 jψ D i, which is an expo-
nential decay for t ⟶ þ ∞. On the other hand, a similar formal time evolution
gives eitH jψ G i ¼ eiER t etΓ=2 jψ G i, which decays exponentially as t ⟶ ∞. Vector
states jψ D i and jψ G i are known as the decaying and growing Gamow vectors,
respectively.
As a matter of fact, both vectors jψ D i and jψ G i (where the D stands for decay
and the G stands for growing) are equally suitable for a vector state for the
considered resonance. Nevertheless, the choice jψ D i seems more natural as the
time flows in the positive direction. The point is that both are time reversal of each
other and represent the same physical phenomenon. Note that both vector states
describe the part of the resonance that behaves exponentially with time. Deviations
add a background term (Bohm and Gadella 1989), but here we can consider it as
negligible.
The previous considerations have an important mathematical flaw, however.
Gamow vectors are eigenvectors of the total Hamiltonian H with complex eigen-
values. This is not compatible with the assumption that H is self-adjoint. However,
this property is essential if we want the Gamow vectors to have an exponential
behavior with time. There are two possible remedies for this problem:
18.4.2.1 Non-Hermitian Hamiltonian

This is the approach known as dilation analytic potentials (Balslev and Combes
1971). It gives normalizable Gamow vectors belonging to the Hilbert space on
which the total Hamiltonian H is defined as a self-adjoint operator. However, these
Gamow vectors depend on a nonphysical parameter, precisely the parameter that
provides the dilation, which is arbitrary at some extent (Reed and Simon 1978).
There are other possibilities for using non-Hermitian Hamiltonians (see for
instance Eleuch and Rotter 2017).
18.4.2.2 Rigged Hilbert Space

The second possibility is the extension of the Hilbert space to a rigged Hilbert space
(RHS). A RHS is a triplet of three spaces Φ H Φ , with the following proper-
ties: (i) H is an infinite dimensional separable Hilbert space (a Hilbert space is
separable if any orthonormal basis is countable); (ii) Φ is a dense subspace (a
subspace of H is dense if any neighborhood of any vector in H contains vectors
in Φ) with a topology such that the Φ has less convergent sequences than what it
would have with the topology inherited from H ; and (iii) Φ is the vector space of all
continuous linear mappings from Φ to the space C of complex numbers. RHS also
serves for a rigorous presentation of the Dirac formulation of quantum mechanics
(Roberts 1966, Antoine 1969, Melsheimer 1974, Bohm 1978, Gadella and Gómez
2002, 2003), and it has some other applications concerning group representations
and special functions (Celeghini, Gadella, and del Olmo 2016, 2017, 2018).
Then, if H is the Hilbert space on which the total Hamiltonian H ¼ H 0 þ V
acts, we may construct two RHS Φ H Φ , with the property that
E
E
ψ j 2 Φ þ and
ψ j
G
2 Φ , where the index j stands for the number of reson-
ances in the system with resonance complex energies ERj iΓj =2. Decaying and
growing Gamow vectors have the desired time behavior, a fact that can be
rigorously proven (Bohm and Gadella 1989, Civitarese and Gadella 2004).
A nonrelativistic quantum system may have infinitely many resonances. This
means that only a finite number of resonances may be considered. We recall that
resonances are determined by the poles of a complex analytic function which are
always isolated points in the complex plane. For large values of ER , the energies go
to the relativistic regime, so that we have to discard this possibility. But then,
resonances with large imaginary part are not observable, because their mean
lifetimes are extremely small. This means that only a finite number of resonances
may be considered within the nonrelativistic regime for a given unstable quantum
system.
In addition, if we only focus our attention on the resonance behavior, we may
consider the space spanned by the Gamow vectors. For decaying (growing)
Gamow vectors, this is a finite dimensional subspace of Φ þ Φ
. Let us assume
∗
that our system has N resonances with zj ≔E Rj þ iΓj =2 and zj being its complex
conjugate. We may consider the 2N dimensional space H G spanned by all Gamow
vectors n

o

ψ D ;
ψ G ;
ψ D ;
ψ G ; : . . . ;
ψ D ;
ψ G :::: (18.11)
1 1 2 2 N N
We define on H G a pseudometric that on the vectors of the basis (2) is

ψD
ψ D ¼ ψ G
ψ G ¼ 0, ψD
G ¼ ψ G
ψ D ¼ δij ,
i j i j i ψj i j (18.12)
where δij is the Kronecker delta. We extend this pseudometric to the whole of H G
by linearity.
We may write the restriction of the total Hamiltonian H to H G as (Losada,
Fortin, Gadella, and Holik 2018)
XN

XN

∗
G
H¼ z
ψ
j¼1 j i
D
ψj
þ
G
z
ψ i
j¼1 j
ψDj
: (18.13)
Note that H in Eq. (18.13) is formally Hermitian. Using the pseudometric Eq.
(18.12), we find that
XN

XN n

n
D

G
Hn ¼ j¼1
z j
ψ i j
þ
ψG j¼1
z∗
j
ψ i ψDj
: (18.14)
This suggests a possible choice for the time evolution operator on H G as

XN n

o
itzj
D G
itz∗

U ðt Þ≔eitH ¼ j¼1
e
ψ i ψ j
þ e j
ψ G
i ψD
j
: (18.15)
The identity I on H G in this representation is given by

XN n

o
I≔ j¼1

ψ D
j ψGj
þ
ψ j
G
ψD
j
: (18.16)

D
G
G
Using the pseudometric, we obtain that I
ψ D

j Þ ¼ ψ j Þ and I ψ j Þ ¼ ψ j Þ,
j ¼ 1, . . . , N, so that this is indeed the identity. With this identity, one possible
choice of the inverse of U ðtÞ is
XN n

o

G
itz∗

U 1 ðtÞ ¼ U ðtÞ ¼ j¼1

eitzj
ψ D
i ψ j
þ e j
ψ G
i j
:
ψD (18.17)
Then, we obtain U ðt ÞU 1 ðt Þ ¼ I as we would expect.

The time evolution for any observable O ¼ Oð0Þ on H G should be defined as
Oðt Þ ≔ U ðt ÞOU ðt Þ: (18.18)
This Oðt Þ is well defined for all values of time t. However, with definitions Eq.
(18.15) and Eq. (18.16), Oðt Þ diverges as t ⟶ ∞. This result is not satisfactory.
This is the reason why we have chosen instead, as evolution operator on H G :
XN n

o
itzj
D G
itz∗

U ðt Þ ≔ j¼1
e
ψ i ψ j
þ e
ψ G
j
i ψD
j
, (18.19)
which is Hermitian. In this case, we have that U ðtÞU † ðtÞ ¼ etΓ I. Nevertheless, we
should keep the definition of O at time t as Oðt Þ≔U † ðt ÞOU ðtÞ. In this case, we
have the following relationship for the commutator of two observables at time t:
XN n

o
2tΓj
D G

G
½O1 ðt Þ; O2 ðt Þ ¼ j¼1
e αj ðt Þ
ψ j ψ j
þ β j ð t Þ
ψ j ψD
j
, (18.20)
where αj ðt Þ and βj ðtÞ, i ¼ 1, 2, . . . , N, are constants for which the dependence on t

is just a phase of the form e2itERj . Since all Γ j > 0, one concludes that, in the limit
t ! þ∞, the commutator Eq. (18.20) vanishes.
In conclusion, for quantum decaying systems and with a correct choice of the
form of our operators, commutators vanish for long values of time.
18.4.3 Decoherence and Irreversible Processes

Many attempts have been made to recover the laws of classical mechanics through
some classical limit. The more relevant approaches include the quantum decoher-
ence process, which is responsible for the disappearance of the interference terms
of quantum states, which are inadmissible for a classical description. In addition,
decoherence provides a rule for choosing the candidates for classical states.
As it is indicated in Castagnino, Fortin, Laura, and Lombardi (2008), three
periods can be identified in the development of the general program of deco-
herence (see also Omnès 2005). In the first period, the arrival to the equilibrium
of irreversible systems was studied. During this period, authors such as van
Kampen, van Hove, Daneri, et al. developed a formalism that was not success-
ful for explaining the decoherence phenomenon, but it established the basis
for its future development. The main problem of this period was that the
decoherence times that were found were too long in comparison with the
experimental ones.
In the second period, the decoherence in open systems was studied. The main
characters of this period were Zeh (1970, 1973) and Zurek (1982, 1991). The
decoherence process is described as an interaction process between an open
quantum system and its environment. This process, called environment-induced
decoherence (EID), determines a privileged basis (usually called pointer basis or
moving decoherence basis), which defines the observables that acquire classical
features. Nowadays, this is the orthodox position on the subject (Bub 1997). The
decoherence times in this period were much smaller, solving the problem of the
first period.
In the third period, the arrival to equilibrium of closed systems was studied (Casati
and Chirikov 1995a, b, Ford and O’Connel 2001, Frasca 2003, Casati and Prosen
2005, Gambini, Porto, and Pulin 2007, Gambini and Pulin 2007, 2010). Within this
period, a new approach to the decoherence was presented by Castagnino et al.
According to this approach, the decoherence process can occur in closed systems,
and it depends on the choice of some observables with some particular physical
relevance (for example, the van Hove observables). This process, called self-induced
decoherence (SID), also determines which is the privileged basis, called the final
decoherence basis, that defines which observables acquire classical features.
In some works (Castagnino and Lombardi 2004, Castagnino et al. 2008, Cas-
tagnino and Fortin 2013), the common characteristics of the different approaches
to decoherence were summarized, and a general framework for decoherence was
proposed. According to the general framework, decoherence is just a particular
case of the general problem of irreversibility in quantum mechanics. Since the
quantum state follows a unitary evolution, it cannot reach a final equilibrium state
when time goes to infinity. Therefore, another element must be considered in such
a way that a nonunitary evolution is obtained. The way to introduce this nonunitary
evolution has to include the splitting of the whole space of observables O into a
relevant subspace OR O and an irrelevant subspace. Once the essential role
played by the selection of the relevant observables is clearly understood, the
phenomenon of decoherence can be explained in four general steps (reproduced
from Castagnino and Fortin 2013):
• First step:
The space of relevant observables OR is defined. For example, in the EID
approach the relevant observables are OR ¼ OS ⊗I E , where OS is an arbitrary
observable of the system S, and I E is the unit operator of the environment E.
SID-relevant observables were defined in Castagnino and Fortin (2013).
• Second step:
The expectation value hOR iρðtÞ , for any OR 2 OR , is obtained. This step can be
formulated in two different but equivalent ways:
i. hOR iρðtÞ is obtained as the expectation value of OR in the unitarily evolving

state ρðt Þ (this way is typical of SID) and its evolution is studied.
ii. A coarse-grained state ρR ðt Þ is defined as
hOR iρðtÞ ¼ hOR iρR ðtÞ , (18.21)
for any OR 2 O and its nonunitary evolution, governed by a master equation, is

obtained (this way is typical of EID).
• Third step:
It is proved that hOR iρðtÞ ¼ hOR iρR ðtÞ reaches a final equilibrium value hOR iρ∗ :
lim t!∞ hOR iρðtÞ ¼ lim t!∞ hOR iρR ðtÞ ¼ hOR iρ∗ , 8OR 2 OR : (18.22)
This also means that the coarse-grained state ρR ðt Þ evolves, with a nonunitary
evolution, toward a final equilibrium state:
lim t!∞ hOR iρR ðtÞ ¼ hOR iρ∗ , 8OR 2 OR (18.23)
• Fourth step: n o
The moving preferred basis j jð~t Þi is defined. This basis is the eigenbasis of a
state ρP ðtÞ such that
lim t!∞ hOR iðρR ðtÞρP ðtÞÞ ¼ 0, 8OR 2 OR : (18.24)
The characteristic time for this limit is the decoherence time tD .

The approaches to decoherence all have one thing in common: They need to
introduce a nonunitary evolution. From a general point of view, it is possible to
approximate the evolution of the system through an effective non-Hermitian
Hamiltonian H eff . It can be proved that the evolution of the mean value is given
by (Castagnino and Fortin 2012)
X
hOR iρðtÞ ffi hOR iρ∗ þ C eγi t ,
i i
(18.25)
where γ1
i are the characteristic times of the system, which are associated with the
complex eigenvalues of the effective Hamiltonian. Then, it is easy to see that the
commutator between two relevant observables is (Fortin and Vanni 2014)

OR ; O0R ρðtÞ ⟶0: (18.26)
This means that, when t ⟶ þ ∞, the expectation value of the commutator between
OR and O0R becomes zero. Therefore, the Heisenberg uncertainty relation becomes
undetectable from the experimental viewpoint.
18.4.4 Quantum Statistics and the Classical Limit

In this section, we show that the dynamics of logics can be related to other
parameters, different from time. In quantum statistics, the mean number of par-
ticles occupying a quantum state is given by the formula
1
ns ¼ (18.27)
exp ðα þ βϵ s Þ 1
in which the “+” sign corresponds to Fermi-Dirac statistics and the “” to Bose-
Einstein. The parameter α is related to the particle number according to the
condition
X X 1
s

ns ¼ s exp ðα þ βϵ Þ 1
¼N: (18.28)
s
being N the total particle number. The partition function reads

X
ln ðZ Þ ¼ αN s
ln ½1 exp ðα βϵ s Þ : (18.29)
When the concentration of the gas is made sufficiently low, quantum effects
should be important. This limit corresponds to small N. Equivalently, we should
have ns 1 (or exp ðα þ βϵ s Þ 1).
If we now assume that the particle number is fixed, and we increase the
temperature (this is equivalent to β⟶0), we obtain that the most important terms
are those satisfying βϵ s α. Under these conditions, we obtain that
exp ðα þ βϵ s Þ 1. Or equivalently, that ns 1. This is the condition for the
classical limit. In other words, the condition under which quantum effects are
negligible. In this limit, and for both cases, Fermi-Dirac and Bose-Einstein, we
obtain
ns ¼ exp ðα βϵ s Þ: (18.30)
This constraint reduces to
X
s
exp ðα βϵ s Þ ¼ N; (18.31)
then, we can express

exp ðα βϵ s Þ
ns ¼ N P : (18.32)
s exp ðβϵ s Þ
Thus, at sufficiently low density or high temperature, we obtain the Maxwell–

Boltzmann distribution, which is a signature of classicality. But the fact that we can
attain the classical limit by adjusting the temperature suggests that time is not the
only parameter that allows us to observe a logic transition. The algebraic aspects of
this transition will be discussed in a future work, but we can advance some
points here.
First, some interpretations of quantum mechanics suggest that, when the clas-
sical limit is obtained, an irreversible process should be observed. Under this
perspective, this can be related to the mathematical formalism of Gamow vectors.
Second, the approach of dynamical logics can be useful to interpret the quant-
ization deformation formalism under a new light. Indeed, some authors (see, e.g.,
Landsman 1993) have proposed to study the classical limit and the quantization of
a given theory by appealing to the formalism of deformation quantization. In this
approach, one starts with a classical (commutative) algebra of observables A0 ,
endowed with a pointwise product ∙ and a Poisson bracket f;g. Then, a family of
algebras Ah is introduced, indexed with a parameter h 0. The parameter h is
intended to represent a dimensionless combination of some characteristic param-
eters associated with the system and Planck’s constant. An associative product ?h
is introduced in the indexed algebras, and it is required that (see Landsman 1993
for details)
i
lim h!0 ½ f ; g h ¼ f f ; gg (18.33)
h
and
1
lim h!0 ½ f ; g hþ ¼ f g: (18.34)
2
The examples shown in this section suggest that the parameters involved in the
classical limit process could be time, temperature, particle number, or others. Thus,
our dynamical logics approach could be connected in a natural way with the
formalism of deformation of algebras. We will discuss this possibility elsewhere.
18.5 Conclusions
In this chapter, we have presented a logical approach for the description of the
quantum-to-classical transition of physical systems. This approach consists in
describing the system as a collection of observables that evolve over time,
according to the Heisenberg picture, but with a nonunitary evolution.
In turn, the algebra of observables determines a lattice of elementary physical
properties with a logical structure. In the classical case, the properties have a
classical logic structure, and in the quantum case, they have a quantum logic
structure. The time evolution of the algebra induces a time evolution of the lattice
of properties. Therefore, in this approach, the classical limit is attained when the
final structure of properties becomes a classical logic, or equivalently, when the

resulting algebra of observables becomes commutative.
We have shown some examples in which this logical transition occurs, among
them, quantum channels, unstable physical processes and models of self-
induced decoherence. We have also shown that our formulation has a natural
application in quantum statistical mechanics, where the temperature parameter
or the particle number can play the role of the time in reaching the classical
limit. In other words, the classical limit of quantum statistical systems indicates
that time is not the only parameter that may show a transition from quantum to
classical logic. Furthermore, we have connected our approach with the formal-
ism of quantization deformation. In future works, we will develop these ideas in
more detail.
Acknowledgments
We are grateful to the participants of the workshop Identity, indistinguishability
and non-locality in quantum physics (Buenos Aires, June 2017) for their useful
comments. This work was made possible through the support of Grant 57919 from
the John Templeton Foundation and Grant PICT-2014–2812 from the National
Agency of Scientific and Technological Promotion of Argentina.
References
Antoine, J. P. (1969). “Dirac formalism and symmetry problems in quantum mechanics I:
General Dirac formalism,” Journal of Mathematical Physics, 10: 53–69.
Balslev, E. and Combes, J. M. (1971). “Spectral properties of many body Schrödinger
operators with dilation analytic intercations,” Communications in Mathematical
Physics, 22: 280–294.
Birkhoff, G. and von Neumann, J. (1936). “The logic of quantum mechanics,” Annals of
Mathematics, 37: 823–843.
Bohm, A. (1978). The Rigged Hilbert Space and Quantum Mechanics, Springer Lecture
Notes in Physics. New York: Springer.
Bohm, A. (1993). Quantum Mechanics: Foundations and Applications. Berlin and New
York: Springer.
Bohm, A. and Gadella, M. (1989). Dirac Kets, Gamow Vectors, and Gel’fand Triplets: The
Rigged Hilbert Space Formulation of Quantum Mechanics. Berlin: Springer.
Casati, G. and Chirikov, B. (1995a). “Comment on ‘Decoherence, chaos, and the second
Law’,” Physical Review Letters, 75: 350.
Casati, G. and Chirikov, B. (1995b). “Quantum chaos: Unexpected complexity,” Physical
Review D, 86: 220–237.
Casati, G., and Prosen, T. (2005). “Quantum chaos and the double-slit experiment,”
Castagnino, M. and Fortin, S. (2012). “Non-Hermitian Hamiltonians in decoherence and

equilibrium theory,” Journal of Physics A, 45: 444009.
Castagnino, M. and Fortin, S. (2013). “Formal features of a general theoretical framework
for decoherence in open and closed systems,” International Journal of Theoretical
Physics, 52: 1379–1398.
Castagnino, M., Fortin, S., Laura, R., and Lombardi, O. (2008). “A general theoretical
framework for decoherence in open and closed systems,” Classical and Quantum
Gravity, 25: 154002.
Castagnino, M. and Gadella, M. (2006). “The problem of the classical limit of quantum
mechanics and the role of self-induced decoherence,” Foundations of Physics, 36:
920–952.
Castagnino, M. and Lombardi, O. (2004). “Self-induced decoherence: A new approach,”
Celeghini, E., Gadella, M., and del Olmo, M. A. (2016). “Applications of rigged Hilbert
spaces in quantum mechanics and signal processing,” Journal of Mathematical
Physics, 57: 072105.
Celeghini, E., Gadella, M., and del Olmo, M. A. (2017). “Lie algebra representations and
rigged Hilbert spaces: The SO(2) case,” Acta Polytechnica, 57: 379–384.
Celeghini, E., Gadella, M., and del Olmo, M. A. (2018). “Spherical harmonics and rigged
Hilbert spaces,” Journal of Mathematical Physics, 59: 053502.
Civitarese, O. and Gadella, M. (2004). “Physical and mathematical aspects of Gamow
states,” Physics Reports, 396: 41–113.
Cohen, D. (1989). An Introduction to Hilbert Space and Quantum Logic. Berlin: Springer-
Verlag.
Dirac, P. A. M. (1933). “The Lagrangian in quantum mechanics,” Physikalische Zeitschrift
der Sowjetunion, 3: 64–72.
Domenech, G., Holik, F., and Massri, C. (2010). “A quantum logical and geometrical
approach to the study of improper mixtures,” Journal of Mathematical Physics, 51:
052108.
Eleuch, H. and Rotter, I. (2017). “Resonances in open quantum systems,” Physical Review
A, 98: 0221117.
Feynman, R. P. (1942). The Principle of Least Action in Quantum Mechanics. Princeton:
Princeton University. Reproduced in R. P. Feynman and L. M. Brown (eds.). (2005).
Feynman’s Thesis: a New Approach to Quantum Theory. Singapore: World
Scientific.
Fischer, M. C., Gutierrez-Medina, B., and Raizen, M. G. (2001). “Observation of the
quantum Zeno and anti-Zeno effects in an unstable system,” Physical Review Letters,
87: 40402.
Fonda, L., Ghirardi, G. C., and Rimini, A. (1978). “Decay theory of unstable quantum
systems,” Reports on Progress in Physics, 41: 587–631.
Ford, G. W. and O’Connel, R. F. (2001). “Decoherence without dissipation,” Physics
Letters A, 286: 87–90.
Fortin, S., Holik, F., and Vanni, L. (2016). “Non-unitary evolution of quantum logics,”
Springer Proceedings in Physics, 184: 219–234.
Fortin, S. and Vanni, L. (2014). “Quantum decoherence: A logical perspective,” Founda-
tions of Physics, 44: 1258–1268.
Frasca, M. (2003). “General theorems on decoherence in the thermodynamic limit,”
Physics Letters A, 308: 135–139.
Gadella, M. and Gómez, F. (2002). “A unified mathematical formalism for the Dirac
formulation of quantum mechanics,” Foundations of Physics, 32: 815–869.
Gadella, M. and Gómez, F. (2003). “On the mathematical basis of the Dirac formulation of
quantum mechanics,” Foundations of Physics, 42: 2225–2254.
Gambini, R., Porto, R. A., and Pulin, J. (2007). “Fundamental decoherence from
quantum gravity: A pedagogical review,” General Relativity and Gravitation, 39:
1143–1156.
Gambini, R. and Pulin, J. (2007). “Relational physics with real rods and clocks and the
measurement problem of quantum mechanics,” Foundations of Physics, 37:
1074–1092.
Gambini, R. and Pulin, J. (2010). “Modern space-time and undecidability,” pp. 149–161 in
V. Petkov (ed.), Minkowski Spacetime: A Hundred Years Later. Fundamental Theor-
ies of Physics 165. Heidelberg: Springer.
Griffiths, R. B. (2014). “The new quantum logic,” Foundations of Physics, 44: 610–640.
Holik, F., Massri, C., and Ciancaglini, N. (2012). “Convex quantum logic,” International
Journal of Theoretical Physics, 51: 1600–1620.
Holik, F., Massri, C., Plastino, A., and Zuberman, L. (2013). “On the lattice structure of
probability spaces in quantum mechanics,” International Journal of Theoretical
Physics, 52: 1836–1876.
Holik, F. and Plastino, A. (2015). “Quantum mechanics: A new turn in probability theory,”
pp. 399–414 in Z. Ezziane (ed.), Contemporary Research in Quantum Systems. New
York: Nova Publishers.
Holik, F., Plastino, A., and Sáenz, M. (2014). “A discussion on the origin of quantum
probabilities,” Annals of Physics, 340: 293–310.
Kalmbach, G. (1983). Orthomodular Lattices. San Diego: Academic Press.
Landsman, N. P. (1993). “Deformation of algebras of observables and the classical limit of
quantum mechanics,” Reviews in Mathematical Physics, 5: 775–806.
Losada, M., Fortin, F., Gadella, M., and Holik, F. (2018). “Dynamical algebras in quantum
unstable systems,” International Journal of Modern Physics A, 33: 1850109.
Losada, M., Fortin, S., and Holik, F. (2018). “Classical limit and quantum logic,” Inter-
national Journal of Theoretical Physics, 57: 465–475.
Losada, M. and Laura, R. (2014a). “Quantum histories without contrary inferences,”
Annals of Physics, 351: 418–425.
Losada, M. and Laura, R. (2014b). “Generalized contexts and consistent histories in
quantum mechanics,” Annals of Physics, 344: 263–274.
Losada, M., Vanni, L., and Laura, R. (2013). “Probabilities for time–dependent properties
in classical and quantum mechanics,” Physical Review A, 87: 052128.
Losada, M., Vanni, L., and Laura, R. (2016). “The measurement process in the generalized
contexts formalism for quantum histories,” International Journal of Theoretical
Physics, 55: 817–824.
Melsheimer, O. (1974). “Rigged Hilbert space formalism as an extended mathematical
formalism for quantum systems. 1. General theory,” Journal of Mathematical Phys-
ics, 15: 902–916.
Nakanishi, N. (1958). “A theory of clothed unstable particles,” Progress of Theoretical
Physics, 19: 607–621.
Nielsen, M. and Chuang, I. (2000). Quantum Computation and Quantum Information.
Omnès, R. (1994). The Interpretation of Quantum Mechanics. Princeton: Princeton Uni-
versity Press.
Omnès, R. (2005). “Results and problems in decoherence theory,” Brazilian Journal of
Physics, 35: 207–210.
Reed, M. and Simon, B. (1978). Analysis of Operators. New York: Academic.
Roberts, J. E. (1966). “Rigged Hilbert spaces in quantum mechanics,” Communications in

Mathematical Physics, 3: 98–119.
Rothe, C., Hintschich, S. L., and Monkman, A. P. (2006). “Violation of the exponential-
decay law at long times,” Physical Review Letters, 96: 163601.
Springer.
von Neumann, J. (1932). Mathematische Grundlagen der Quantenmechanik. Heidelberg:
University Press.
Zeh, H. D. (1970). “On the interpretation of measurement in quantum theory,” Founda-
tions of Physics, 1: 69–76.
Zeh, H. D. (1973). “Toward a quantum theory of observation,” Foundations of Physics, 3:
109–116.
Zurek, W. (1982). “Environment-induced superselection rules,” Physical Review D, 26:
1862–1880.
Zurek, W. (1991). “Decoherence and the transition from quantum to classical,” Physics
Today, 44: 36–44.
19
Quantum Mechanics and Molecular Structure:
The Case of Optical Isomers
juan camilo martı́nez gonzález, jesús jaimes arriaga, and
sebastian fortin
19.1 Introduction
Since its birth, quantum mechanics has enjoyed high prestige thanks to its success
in the explanation and prediction of phenomena at the atomic and molecular scales.
Indeed, this theory began by explaining the emission lines observed in the hydro-
gen atom and, after a few years after its first formulation, it could explain the
energy spectrum of simple molecules. This type of success quickly leads scientists
to suppose that all chemistry can be explained by physics. The famous claim by
Paul Dirac is an example of such an assumption:
The underlying physical laws necessary for the mathematical theory of a large part of
physics and the whole of chemistry are thus completely known, and the difficulty is only
that the exact application of these equations leads to equations much too complicated to be
soluble. It therefore becomes desirable that approximate practical methods of applying
quantum mechanics should be developed, which can lead to an explanation of the main
features of complex atomic systems without too much computation.”
(Dirac 1929: 714)
However, as time went by, it turned out to be clear that the attempt to explain
chemistry from physics leads to complications that allow us to question Dirac’s
claim. One of these problems is to explain molecular structure from quantum
mechanics. There are several ways to approach this problem, but in this work we
will do it by means of a particular case: optical isomers and the Hund paradox.
When young Kant meditated upon the distinction between his right hand and
his left hand, he could not foresee that the problem of incongruent counterparts
would be reborn in the twentieth century under a new form. The so-called Hund
paradox points to the difficulty of giving a quantum explanation to chirality, that
is, to the difference between the members of a pair of optical isomers or
enantiomers. The question about whether the quantum formalism can account
for chirality concerns philosophy of science for (at least) three reasons. First, it
379
380 Juan Camilo Martínez González, Jesús Jaimes Arriaga, and Sebastian Fortin
introduces an interesting case for the debate about the relation between physics
(quantum mechanics) and chemistry (molecular chemistry), which has been the
focus of many philosophical works in recent years. Second, and related to the
previous point, the analysis of the paradox can enrich the discussion about
whether quantum mechanics can provide an explanation of molecular structure.
Third, since some approaches attribute the origins of the paradox to a focus on
isolated molecules, the solution is believed to be found in considering molecules
in interaction; these views pose a relevant question to the ontology of chemistry:
Is chirality an intrinsic property of a molecule? These three problematic points
make the resolution of the Hund paradox an issue of the utmost importance for
the philosophy of science.
On this basis, in this chapter we will analyze the problem of optical isomerism
by proceeding in the following steps. In Section 19.2, the Hund paradox will be
presented in formal terms. Section 19.3 will be devoted to showing the relevance
of the paradox to the relation between physics and chemistry, to the explanation of
molecular structure, and to the ontology of chemistry. In Section 19.4 the paradox
will be conceptualized as a case of quantum measurement, stressing that decoher-
ence does not offer a way out for this problem. Finally, in Section 19.5 we will
argue for the need of adopting a clear interpretation of quantum mechanics; in
particular, we will claim that the modal-Hamiltonian interpretation, which con-
ceives measurement as a breaking symmetry process, supplies the tools required to
solve the Hund paradox.
19.2 The Hund Paradox

As it is well known, a chemical formula such as H2O indicates the elements in a
compound and their relative proportions, but it does not offer information about the
geometric structure of the molecule. Molecules with the same chemical formula
but differing in the spatial disposition of their atoms are called isomers. The class
of isomers includes the subclass of optical isomers or enantiomers: the members of
a pair of enantiomers are mirror images of each other; the property that distin-
guishes them is called chirality. The peculiarity of enantiomers of a same com-
pound is that they share almost all their chemical and physical properties: They
differ in how they rotate the plane of polarization of plane-polarized light.
Depending on the direction of the rotation, dextro-rotation or levo-rotation, optical
isomers are called D or L.
The problem of the enantiomers was first formulated by Friedrich Hund (1927),
a pioneer in the development of quantum chemistry. From a structural point of
view, the two members of a pair of enantiomers have the same bonds, i.e., the
“distance between atoms” is the same. Since the quantum Coulombic Hamiltonian
Quantum Mechanics and Molecular Structure: The Case of Optical Isomers 381
depends only on the internuclear distances, the Hamiltonian is exactly the same for
the two members of the pair. Consequently, quantum mechanics provides the same
description for two chemical species that can effectively be differentiated in
practice by their optical activity (Harris and Stodolsky 1981, Wolley 1982, Berlin,
Burin, and Goldanskii 1996, Quack and Stohner 2005, Schlosshauer 2007).
In the quantum domain, the parity operator P ^ is associated with spatial reflec-
tion: if jDi and jLi are the states of isomers D and L, respectively, P ^ transforms jDi
into jLi and vice versa: P ^ jLi ¼ jDi, P^ jDi ¼ jLi. Let us suppose that the molecule
consists of A atomic nuclei and N electrons. Then, the Coulombic Hamiltonian of
the complete molecule reads
!
XA
P 2 X A
Z Z X i
P 2 XA
Z X N
1
^ ¼
H
g
þ e2
g h
þ
i
e2
g
þ e2
g
2mg g<h
2mg N
2me g
^r ig i<j
^r ij
(19.1)
where P^ g , Z g , and mg are the momentum operator, the atomic number, and the
mass of the nucleus g, respectively, with g = 1, 2, . . ., A; e and me are the charge
and the mass of the electrons, respectively; P ^ i is the momentum operator of the
electron i, with i = 1, 2, . . ., N; ^r ij is the operator “distance” between the electron i
and the electron j, and ^r ig is the operator “distance” between the electron i and the
nucleus g. Since the Coulombic interaction only depends on the distance between
the interacting particles, it is symmetric under spatial reflection; therefore, the
Hamiltonian commutes with the parity operator P: ^

^ H
P; ^ ¼0 (19.2)
This means that the eigenstates j ωn i of the Hamiltonian have definite parity. In
particular, the ground state j ω0 i is invariant under space reflection:
^ j ω0 i ¼ j ω0 i. As a consequence, j ω0 i cannot be a chiral state j Di or j Li, but
P
it is a superposition of them:
1
j ω0 i ¼ pffiffiffi ðjDi þ jLiÞ (19.3)
2
The question is, then, why chiral molecules are never found in this superposition
state. The states obtained in the laboratory are j Di and j Li, which are not
eigenstates of the Hamiltonian and do not correspond to the ground state. So,
why do certain chiral molecules display an optical activity that is stable over time,
associated with a well-defined chiral state? Why do chiral molecules have a
definite chirality? (Berlin et al. 1996). The Hund paradox points to the core of
certain traditional problems of the philosophy of chemistry. Let us consider them
briefly.
19.3 Intertheoretic Relationships and Molecular Structure

The relationship between chemistry and physics is one of the hottest topics in the
philosophy of chemistry. In this context, the links between theories coming from
the two disciplines have been explored in great detail from different perspectives.
However, despite this effort, there is no agreement yet with respect to the best
model of intertheoretic relationships to describe those links. Although the idea of
reducing different disciplines to physics is much older, the success of the applica-
tions of quantum mechanics to chemical systems turned reduction into a regulative
idea in the accounts of the relationship between chemistry and physics. Following
Dirac’s famous dictum (1929), the idea that chemistry can be reduced to quantum
mechanics pervaded both the communities of physicists and chemists. The neces-
sary approximations for such a reduction led to the constitution of quantum
chemistry as a new area of scientific research (Gavroglu and Simões 2012).
However, the approximate methods frequently distort the principles of quantum
mechanics in such a way that the interpretation of the intertheoretical links as
reductive is seriously disputed (Woolley 1978, 1982; Primas 1983). In fact, the
strategies that make possible the description of chemical phenomena in quantum
terms, such as the Born-Oppenheimer approximation or the models of valence
bond and molecular orbital, do not strictly satisfy the conditions of Nagelian
reduction – not only do they establish loose and noncontinuous connections
between chemistry and physics (Lombardi 2014), but they also introduce assump-
tions that stand in conflict with quantum mechanics itself (Lombardi and Castag-
nino 2010).
The case of enantiomers would provide a new insight in the discussion about the
reduction of chemistry, because it involves a difficulty that does not depend on
approximations. In this case the challenge is more fundamental, because it stands
beyond the Born-Oppenheimer approximation. In fact, even if we cannot write
down the exact Hamiltonian due to its complexity, we know that it only depends
on the distance of the component particles and, therefore, it cannot account for the
difference between the members of a pair of enantiomers.
19.3.1 The Concept of Molecular Structure

Molecular structure, given by the spatial arrangement of the nuclei in a molecule, is
a main character of molecular chemistry: It is “the central dogma of molecular
science” (Woolley 1978: 1074), as it plays a key role in the explanation of
reactivity. However, the concept of molecular structure finds no comfortable place
in the theoretical framework of quantum mechanics, inasmuch as it appeals to the
classical notion of individual nuclei in fixed positions. This problem can be viewed
as a particular manifestation of the general problem of the intertheoretical links

between molecular chemistry and quantum mechanics. Following the traditional
reductionist perspective, some authors consider that the difficulties are only due to
our partial knowledge of molecular systems in the theoretical framework of
quantum mechanics (Sutcliffe and Woolley 2011, 2012).
From the opposite perspective, Robin Hendry (2004, 2008, 2010), who rejects
the epistemological reduction of chemistry to physics, claims that the problem of
molecular structure must be addressed within the ontological domain: The inter-
esting philosophical question is how the entities and processes studied by different
disciplines are related to each other. In particular, the author considers that the
links between quantum mechanics and molecular chemistry, embodied in the
problem of molecular structure, must be conceived nonreductively, in terms of
emergence.
Optical isomerism introduces a new perspective to the discussion. In fact, the
difference between two enantiomers lies in their structure. Again, in this case no
arguments concerning how to interpret approximations in quantum mechanics are
involved – the Hund paradox arises in terms of the exact Hamiltonian of the
molecules. Due to this particularity, a clarification of the paradox would certainly
enrich the debate about whether quantum mechanics can provide an explanation of
molecular structure.
19.3.2 The Ontology of Chemistry

Although not as extensively treated as the relationship between chemistry and
physics, a central topic in the philosophy of chemistry is that related to the
ontology of chemistry, that is, to the object of study of the discipline (see Hendry
2018). The central question in this field of inquiry concerns what items inhabit the
realm of chemistry and to which ontological categories they belong. For instance,
certain works analyze the nature of the chemical bond (Vemulapalli 2008, Hendry
2012), or the reference of the concept of chemical orbital (Labarca and Lombardi
2010, Llored 2010), or even the ontological status of electronegativity (Leach
2013). The common ground in those discussions is the effort to identify and to
categorize chemical properties. Does the term ‘chemical bond’ have an ontic
reference? Is the shape of the orbitals a real property or a mere theoretical tool?
Is electronegativity an intrinsic property of atoms or a loosely defined property of
elements?
From a traditional perspective, chirality was usually conceived as an intrinsic
property of molecules, defined by the molecular structure. However, more recently,
some approaches consider that the Hund paradox is the consequence of focusing
on isolated molecules. From this viewpoint the solution must be based on studying
the interactions in which the molecule is involved. Therefore, chirality would no

longer be a property of a particular molecule, but it would rather turn out to express
the result of a relation linking the molecules with its environment. This view, that
modifies the traditional ontological picture regarding chirality, began to prevail
with the development of the theory of decoherence.
19.3.3 The Reductionist Program: Quantum Mechanics Reloaded

Hinne Hettema (2009, 2012) claims that molecular structure can be understood in
term of reduction, in light of certain recent developments in quantum chemistry,
such as the quantum theory of atoms in molecules (QTAiM; Bader 1994).
The QTAiM was introduced by Richard Bader in the nineties, and it is based on
the assumption that the distribution of the electron density associated with atoms and
molecules represents the physical manifestation of matter in space (Bader 2010b).
Therefore, the topology of the electron density encodes and reflects the notions of
atom, bond, structure, and structural stability (Bader 2011). In this way, Bader argues
that a theory based on electron density allows linking the language of chemistry with
that of physics. According to Bader, atoms are bounded by zero-flux surfaces in the
gradient vector field, which is a result of the dominant morphology of the electron
density distribution, which leads to the natural partition of the molecular space into
mononuclear regions associated with the atoms. The zero-flux surface is not crossed
by any trajectory of the gradient vector field. This fact is interpreted in principle as
follows: no electron crosses such a surface in such a way that the electron density of
each atom in the molecule remains unchanged over time.
Likewise, it is possible to define molecular structure by mean of the critical points
at which the gradient vector cancels. A local maximum is associated with nuclear
positions and a saddle point is linked to a chemical bond (which is called bond path
in the theory). Moreover, with the help of these critical points, the classical view of
molecular structure with “balls and sticks” is recovered through molecular graphs,
which are constructed by employing “balls” to represent nuclear positions and
“sticks” to represent bond paths (Bader 2010a). Under this context, QTAiM claims
to offer a reductionist scheme for molecular structure. In the author’s words:
The reductionist approach afforded by QTAiM offers a clear solution to the myriad of
personal views and models of bonding. As has been amply demonstrated by appeal to
physics, the presence of a bond path linking a pair of atoms fulfills the sufficient and
necessary conditions that the atoms are bonded to one another. This definition, which
necessarily applies to quantum mechanical densities, transcends all bonding schemes and
categories and provides a unified physical understanding of atomic interactions. One
assumes such unification to be a primary goal of any physical theory.
(Bader 2011: 20)
QTAiM was very successful to describe many chemical molecules. However, it

is doubtful that it offers a complete reduction of chemistry to quantum mechanics:
– Bader discards the wave function and adopts the density of electrons as a
fundamental entity.
– QTAiM considers that the maximum of electron density are the positions of
the atoms. Why? Is this a new postulate of quantum mechanics?
– The theory finds zero flux surfaces and claims that these surfaces enclose
atoms. However, it does not consider the holism of quantum systems and the
peculiar problem of individuality in quantum mechanics.
Bader himself presents his theory as an extension of quantum mechanics for
chemical systems. Then, he has to modify quantum mechanics to obtain his results.
The idea of modifying quantum mechanics just to explain the results of chemistry
is not acceptable in a reductionist program with ontological pretensions. In fact, the
most accepted proposal to describe molecular structure (in general, and isomers in
particular) from quantum mechanics is based on quantum decoherence (Scerri
2011), a phenomenon that can be accounted for in standard quantum-
mechanical terms.
19.4 Decoherence, Enantiomers, and Quantum Measurement

As explained in Section 19.2, the ground state of the molecule is not one of the
chiral states j Di or j Li, but a superposition of them (see Eq. (19.3)). Then why do
we always find the molecule either in the state j Di or in the state j Li? It is not hard
to see that the question is the same as that underlying the quantum measurement
problem: Following Schrödinger’s famous example, if the cat is in a superposition
of “alive” and “dead,” we have to explain why we always see the cat dead or alive.
In technical terms, the problem is to explain why we measure definite values of an
observable when the system is in state of superposition of the eigenstates of that
observable. In the particular case of chirality, the problem is to account for the fact
that, although the molecule is in a superposition of the chiral states, it always
manifests a definite chirality.
The orthodox answer to the measurement problem is the collapse hypothesis (or
von Neumann’s projection postulate), according to which, when we measure the
system, the state collapses to one of the states of the superposition. Then, if the
result of a single measurement is, say, dextro-rotation, then the system is actually
in the state j Di:
1
j ω0 i ¼ pffiffiffi ðjDi þ jLiÞ ! j Di (19.4)
2
If many measurements are performed on identical systems with the same initial
conditions, it is possible to define an ensemble, whose state is represented by a
density operator:
1
^ρ collapsed ¼ ðjDihDj þ jLihLjÞ (19.5)
2
This state is interpreted as stating that there is a probability 0.5 of finding the
system in the state j Di and a probability 0.5 of finding the system in the statej Li:
the state is a mixture of an equal number of definite chiral states. The collapse
hypothesis is very successful in reproducing the experimental results, but it has no
explanatory power, to the extent that it is an ad hoc hypothesis specifically
designed to account for the quantum measurement problem. Moreover, collapse
is a nonunitary process that breaks the Schrödinger evolution; however, the
hypothesis does not explain why or when the process happens. For this reason,
during the last decades, quantum measurements have been approached from
different perspectives; one of them is that given by the theory of decoherence.
According to the orthodox approach – the so-called environment-induced deco-
herence (Zurek 1981, 1993, 2003) – decoherence is a phenomenon resulting from
the interaction between an open quantum system and its environment. Let us
consider a closed system U with two subsystems: the open system S in the initial
state ^ρ S , and the environment E in the initial state ^ρ E . Then, the initial state of
the total system is ^ρ U ¼ ^ρ S ⊗^ρ E . This state evolves in a unitary way according to
the Schrödinger equation. But, the theory of decoherence studies the behavior
of the reduced state of the open system, ^ρ reduced ¼ Tr E ð^ρ U Þ, obtained by applying
the partial trace on the state of the whole closed system; the partial trace is an
operation that removes the degrees of freedom of the environment from ^ρ U . As
a consequence, the reduced state of the open system is no longer governed
by the Schrödinger equation, but is instead ruled by a master equation: ^ρ reduced
may evolve in a nonunitary way. Moreover, when the number of degrees
of freedom of the environment is very high, the reduced state may become
diagonal and mimic the ^ρ collapsed , obtained by means of the collapse hypothesis
(see Eq. (19.5)).
In his “Editorial 37” in Foundations of Chemistry, Eric Scerri (2011) explicitly
relates the problem of optical isomerism to the quantum measurement problem.
According to Scerri, the Hund paradox would be dissolved if the interaction of the
molecule with its environment were taken into account:
The study of decoherence has shown that it is not just observations that serve to collapse
the superpositions in the quantum mechanics. The collapse can also be brought about by
molecules interacting with their environment.
(Scerri 2011: 4; see Scerri 2013 for a similar claim)
The idea is that the enantiomer molecule is in interaction with the environment (air,
particles, other molecules, etc.). If the initial states of the molecule and the
pffiffiffi are j ω0 i and j ε0 i, respectively, the initial state of the whole system
environment
is 1= 2 ðjDi þ jLiÞ ⊗ j ε0 i. The interaction between the molecule and its envir-
onment define the evolution of the total system, which, in some cases, produces a
correlation between the possible states of the system and the environment:
1 1 1
pffiffiffi ðjDi þ jLiÞ ⊗ jε0 i ! pffiffiffi j Di ⊗ jεD i þ pffiffiffi j Li ⊗ jεL i (19.6)
2 2 2
Decoherence occurs when, as the result of the evolution, the states of the environ-
ment become rapidly orthogonal: hεL jεL i ! 0. As a consequence, after an
extremely short decoherence time, the reduced state of the molecule acquires the
same structure as that of the mixed state after collapse (see Eq. (19.5)):
1
^ρ decohered ¼ ðjDihDj þ jLihLjÞ (19.7)
2
As in the case of quantum measurement, this state is interpreted as stating that the
molecule is in one of the states j Li or j Di, and that probabilities measure our
ignorance about which state it is. In this way, the theory of decoherence would
solve the problem underlying the Hund paradox.
Although there was a time when, as stressed by Anthony Leggett (1987) and
Jeffrey Bub (1997), decoherence was considered the “new orthodoxy” in the
physics community to explain quantum measurements, at present it is quite clear
that decoherence does not solve the measurement problem. In fact, collapse is the
change of the state of the system, from a superposition to a definite state; on this
basis, ^ρ decohered can be interpreted as a legitimate mixture. On the contrary, in the
case of decoherence, the state of the whole system never collapses, but always
evolves according to the Schrödinger equation – the superposition never vanishes
through the unitary evolution. Therefore, it cannot be supposed that what is
observed at the end of the decoherence process is one of two definite events: either
that associated with j Li or that associated with j Di (see Adler 2003). Bub (1997)
even claims that the assumption of a definite event at the end of the process is not
only unjustified, but also contradicts the eigenstate-eigenvalue link. These conclu-
sions about decoherence can also be drawn from the traditional distinction between
a proper mixture – the mixed state of a closed system – and an improper mixture –
the reduced state of an open system. As Bernard d’Espagnat (1966, 1976) repeat-
edly stressed, improper mixtures cannot be interpreted in terms of ignorance (for
additional arguments, see Fortin and Lombardi 2014).
Summing up, at present some authors still consider that decoherence, by itself,
solves many conceptual problems in quantum physics (e.g., Crull 2015).
Nevertheless, in the community of the philosophy of physics it is well known that,

although decoherence is a powerful tool to deal with conceptual problems, it does
not allow us to dispense with interpreting the formalism (Vassallo and Esfeld
2015). In the next section we will follow precisely an interpretive path to deal with
the Hund paradox.
19.5 Symmetry Breaking, Enantiomers, and the

Modal-Hamiltonian Interpretation
As it is well-known, the contextuality of quantum mechanics, derived from the
Kochen-Specker theorem, implies that all the observables of a quantum system
cannot acquire definite actual values simultaneously. Therefore, any realist inter-
pretation of quantum mechanics is forced to select a preferred context, that is, the
set of the definite-valued observables of the system. The modal-Hamiltonian
interpretation (MHI; Lombardi and Castagnino 2008, Lombardi, Castagnino, and
Ardenghi 2010) is a realist, noncollapse interpretation that places the Hamiltonian
of the system in the center of the stage. According to the modal-Hamiltonian
actualization rule, the observables that acquire actual definite values are the
Hamiltonian H ^ and all the observables that commute with H ^ and have, at least,
the same symmetries (that is, that do not break the symmetries of H). ^
The justification for selecting the Hamiltonian as the preferred observable
ultimately lies in the physical relevance of the MHI and in its ability to solve
interpretive difficulties. In fact, the MHI actualization rule can be applied to several
well-known physical situations, leading to results consistent with empirical evi-
dence (Lombardi and Castagnino 2008: section 5), and to the account of quantum
measurements both in the ideal an in the nonideal cases (Lombardi and Castagnino
2008: section 6, Lombardi, Fortin, and López 2015). In the present discussion
about the Hund paradox, the relevant point is that the MHI describes measurement
as a symmetry-breaking process – measurement breaks the symmetry of the
Hamiltonian and then turns an otherwise nonactualized observable into an actually
definite-valued observable, which thus becomes empirically accessible.
As a simple example, let us consider the case of a free particle. In this case the
Hamiltonian is symmetric under space-displacements in all space directions: All
the directions are equivalent with respect to the linear motion of the particle. The
three components of the momentum, Px, Py, Pz, are the generators of this sym-
metry, but they cannot acquire definite values simultaneously because they do not
commute with each other. According to the MHI, none of these three observables
belongs to the preferred context: They are not definite-valued observables because
they have less symmetries than H, ^ and this means that the actualization of any of
them would distinguish a direction of space in a completely arbitrary way. Let us
now suppose that we want to measure one of those observables, say, the compon-
ent Py in direction y. For this purpose, we have to place a wall normal to the
direction y, in such a way that the new Hamiltonian is the original one plus a term
that represents the asymmetric potential barrier. It is precisely this term that breaks
the symmetry of the original Hamiltonian and renders the observable Py actual and
definite-valued and, as a consequence, accessible to measurement. But the point to
stress here is that now the system is no longer the free particle; it is a new system,
whose Hamiltonian is not symmetric with respect to displacements in direction y.
In light of these interpretive ideas, the Hund paradox can now be rephrased in
MHI’s language. As stressed in Section 19.2, the exact Hamiltonian H ^ of the
enantiomeric molecule (see Eq. (19.1)) is symmetric under spatial reflection – it
commutes with the parity operator P ^ (see Eq. (19.2)). Now, let us consider the
^ whose eigenstates are j Di and j Li. The eigenvalues d and l
observable chirality C,
^
of C represent the properties dextro-rotation or levo-rotation,
respectively. It is
^ ^ ^ ^
easy to see that C does not commute with P: P; C 6¼ 0. In this case, as in the
^ would determine
example of the free particle, the actualization of the observable C,
the chirality of the molecule in a completely arbitrary way: It would introduce in
the molecule an asymmetry not contained in its Hamiltonian. As a consequence,
from the MHI viewpoint, C ^ has no actual value: Chirality is indefinite in the
isolated molecule.
If the observable C^ is to be measured, the parity symmetry of the molecule has
to be broken. For this purpose, the molecule must interact with another system M,
which plays the role of the apparatus, in such a way that the Hamiltonian H ^ T of the
new composite system is no longer parity invariant. For instance, this is obtained
when
H ^ þH
^T ¼ H ^ M, (19.8)
where the Hamiltonian ^ M of the new system breaks the original parity
H

invariance: H ^ 6¼ 0 ) H
^ M; P ^ T; P
^ 6¼ 0. A good candidate for H ^ M is the Hamil-
tonian usually introduced in quantum chemistry to describe the interaction between
molecules and polarized light (see Shao and Hänggi 1997), which is a function of
the electric field E and the magnetic field B ^
of the polarized light. Additionally, C
must commute with the total Hamiltonian H ^ T in order to obtain a stable reading of
chirality. Under these conditions, according to the MHI the observable C ^ acquires
a definite actual value: We measure dextro-rotation or levo-rotation, but now the
system is no longer the isolated molecule, but the molecule in interaction with the
polarized light. In a certain sense, this answer to the Hund paradox agrees with
the view according to which the solution must be looked for in the interaction of
the molecule with its surroundings: Chirality is not an intrinsic property of the
molecule, but of the composite system molecule plus polarized light. However, our
view does not appeal to decoherence, but rather to an interpretation of quantum

mechanics that explicitly accounts for measurement from the perspective of the
symmetries of the system.
19.6 Final Remarks

In this chapter we have argued that the problem of enantiomers cannot be solved
by appealing to decoherence, but it requires a precise interpretation of quantum
mechanics capable of dealing with quantum measurement. In particular, we have
shown that the MHI provides us with the adequate tools, since it conceptualizes
measurement as a symmetry breaking process. It is important to stress that,
nevertheless, this result does not supply an indisputable answer to the problem
of molecular structure. In fact, the proposed interpretive approach only accounts
for the different behavior of the members of a pair of enantiomers in their
interaction with polarized light, but it does not take a stand about molecular
structure understood as a spatial geometric property. This is an issue that deserves
a further discussion, even in the context of the present solution of the Hund
paradox.
Acknowledgments
This work was made possible by the support of Grant 57919 from the John
Templeton Foundation and Grant PICT-2014–2812 from the National Agency of
Scientific and Technological Promotion of Argentina.
References
Adler, S. (2003). “Why decoherence has not solved the measurement problem: A response
to P. W. Anderson,” Studies in History and Philosophy of Modern Physics, 34:
135–142.
Bader, R. F. W. (1994). Atoms in Molecules: A Quantum Theory. Oxford: Oxford
University Press.
Bader, R. F. W. (2010a). “The density in density functional theory,” Journal of Molecular
Structure: TEOCHEM, 943: 2–18.
Bader, R. F. W. (2010b). “Definition of molecular structure: By choice or by appeal to
observation,” Journal Physical Chemistry A, 114: 7431–7444.
Bader, R. F. W. (2011). “On the non-existence of parallel universes in chemistry,”
Foundations of Chemistry, 13: 11–37.
Berlin, Y., Burin, A., and Goldanskii, V. (1996). “The Hund paradox and stabilization of
molecular chiral states,” Zeitschrift für Physik D, 37: 333–339.
Crull, E. (2015). “Less interpretation and more decoherence in quantum gravity and
inflationary cosmology,” Foundations of Physics, 45: 1019–1045.
d’Espagnat, B. (1966). “An elementary note about mixtures,” pp. 185–191 in A. de-Shalit,
H. Feshbach, and L. van Hove (eds.), Preludes in Theoretical Physics. Amsterdam:
North-Holland.
d’Espagnat, B. (1976). Conceptual Foundations of Quantum Mechanics. Reading, MA:
Benjamin.
Dirac, P. A. M. (1929). “Quantum mechanics of many-electron systems,” Proceedings of
the Royal Society of London A, 123: 714–733.
What do reduced states refer to?”, Foundations of Physics, 44: 426–446.
Gavroglu, K. and Simões, A. (2012). Neither Physics nor Chemistry: A History of
Quantum Chemistry. Cambridge, MA: MIT Press.
Harris, R. and Stodolsky, L. (1981). “Time dependence of optical activity,” The Journal of
Chemical Physics, 74: 2145–2155.
Hendry, R. (2004). “The physicists, the chemists, and the pragmatics of explanation,”
Hendry, R. (2008). “Two conceptions of the chemical bond,” Philosophy of Science, 75:
909–920.
Hendry, R. (2010). “Ontological reduction and molecular structure,” Studies in History
Hendry, R. (2012). “The chemical bond,” pp. 293–308 in A. Woody, R. Hendry, and
P. Needham (eds.), Handbook of the Philosophy of Science Volume 6: Philosophy of
Chemistry. Amsterdam: Elsevier.
Hendry, R. (2018). The Metaphysics of Chemistry. Oxford: Oxford University Press.
Hettema, H. (2009). “Explanation and theory foundation in quantum chemistry,” Founda-
tions of Chemistry, 11: 145–174.
Hettema, H. (2012). Reducing Chemistry to Physics. Limits, Models, Consequences.
Groningen, Netherlands: University of Groningen.
Hund, F. (1927). “Zur Deutung der Molekelspektren. III,” Zeitschrift für Physik, 43:
805–826.
Labarca, M. and Lombardi, O. (2010). “Why orbitals do not exist?”, Foundations of
Chemistry, 12: 149–157.
Leach, M. (2013). “Concerning electronegativity as a basic elemental property and why the
periodic table is usually represented in its medium form,” Foundations of Chemistry,
15: 13–29.
Leggett, A. (1987). “Reflections on the quantum measurement paradox,” pp. 85–104 in
B. Hiley and F. Peat (eds.), Quantum Implications. London: Routledge and Kegan
Paul.
Llored, J. -P. (2010). “Mereology and quantum chemistry: The approximation of molecular
orbital,” Foundations of Chemistry, 12: 203–221.
Lombardi, O. (2014). “Linking chemistry with physics: Arguments and counterargu-
ments,” Foundations of Chemistry, 16: 181–192.
Lombardi, O. and Castagnino, M. (2010). “Matters are not so clear on the physical side,”
Foundations of Chemistry, 12: 159–166.
Lombardi, O., Fortin, S., and López, L. (2015). “Measurement, interpretation and infor-
mation,” Entropy, 17: 7310–7330.
Primas, H. (1983). Chemistry, Quantum Mechanics and Reductionism. Berlin: Springer.

Quack, M. and Stohner, J. (2005). “Parity violation in chiral molecules,” International
Journal for Chemistry, 59: 530–538.
Scerri, E. (2011). “Editorial 37,” Foundations of Chemistry, 13: 1–7.
Scerri, E. (2013). “Philosophy of chemistry: Where has it been and where is it going,”
pp. 208–225 in J.-P. Llored (ed.), The Philosophy of Chemistry: Practices, Method-
ologies, and Concepts. Newcastle: Cambridge Scholars Publishing.
Springer.
Shao, J. and Hänggi, P. (1997). “Control of molecular chirality,” The Journal of Chemical
Physics, 107: 9935–9941.
Sutcliffe, B. and Woolley, G. (2011). “A comment on Editorial 37,” Foundations of
Chemistry, 13: 93–95.
Sutcliffe, B. and Woolley, G. (2012). “Atoms and molecules in classical chemistry and
quantum mechanics,” pp. 387–426 in A. Woody, R. Hendry, and P. Needham (eds.),
Handbook of the Philosophy of Science Volume 6: Philosophy of Chemistry. Amster-
dam: Elsevier.
Vassallo, A. and Esfeld, M. (2015). “On the importance of interpretation in quantum
physics: A reply to Elise Crull,” Foundations of Physics, 45: 1533–1536.
Vemulapalli, K. (2008). “Theories of the chemical bond and its true nature,” Foundations
of Chemistry, 10: 167–176.
Wolley, G. (1978). “Must a molecule have a shape?”, Journal of the American Chemical
Society, 100: 1073–1078.
Wolley, G. (1982). “Natural optical activity and the molecular hypothesis,” Structure and
Bonding, 52: 1–35.
Zurek, W. (1981). “Pointer basis of quantum apparatus: Into what mixture does the wave
packet collapse?”, Physical Review D, 24: 1516–1525.
Zurek, W. (1993). “Preferred states, predictability, classicality and the environment-
induced decoherence,” Progress of Theoretical Physics, 89: 281–312.
Zurek, W. (2003). “Decoherence, einselection, and the quantum origins of the classical,”
Reviews of Modern Physics, 75: 715–776.
Index
Action at a distance, 45, 96, 99, 127, 172, 176, Bohmian

178–179 mechanics, 71
Aharanov, Yakir, 129 theory, 78, 90–91, 99, 241
Aharonov-Bohm effect, 47, 96 Bohr, Niels, 52–53, 71, 73, 75–76, 83, 93, 121–123,
Albert, David, 13, 37, 95, 166, 287 130, 133, 135, 154–160, 245, 249, 360–361
Algebra Born Rule, 65, 97, 104, 107–108, 111–112, 170, 174,
Galilean, 273, 299, 310 297, 354
Heisenberg, 298 Born, Max, 83, 279
non-commutative, 74, 155, 367 Born-Oppenheimer approximation, 38, 382
of observables, 153, 298, 307–310, 348, 353–354, Born-Vaidman rule, 101–103
356, 361, 364, 366, 374 Bosyk, Gustavo, 4, 323, 333
von Neumann, 22 Boyle, Robert, 346
Algebraic approach to quantum mechanics, 34, 43, 47, Brown, Harvey, 39, 126–127, 279–281
348, 353, 355–356 Brun, Todd, 21
Allori, Valia, 15, 166, 178, 228 Bub, Jeffrey, 33, 60, 62–63, 136, 281, 387
Araújo, Mateus, 59 Bueno, Otávio, 3, 185
Arenhart, Jonas, 3, 185 Buonomano, Vincent, 248
Arnold, Barry, 323 Butterfield, Jeremy, 23
Arrow of time, 282
Auyang, Sunny, 279 Callender, Craig, 136, 287
Cantor, Georg, 199
Bacciagaluppi, Guido, 39, 279 Carnap, Rudolf, 135, 154, 157–159
Bader, Richard, 384–385 Carroll, Sean, 103, 122, 134–135, 142–143
Baker, David, 279 Cartwright, Nancy, 123
Ballentine, Leslie, 247, 283, 288, 290 Casimir operators, 40–42, 274, 281, 291
Barnum, Howard, 348 Cassirer, Ernst, 155–158
Bauer, Edmond, 52–57 Cauchy surfaces, 16–17, 19–22, 26–27
Beables, 9–12, 14, 22–25, 131, 172 Causation, 169, 175
Bedingham, Daniel, 25, 28 Chakravartty, Anjan, 150–151
Bell, John, 74, 77, 79–80, 86, 126, 133, 165, 167, Chaos, 256, 317–318
171–172, 175, 181, 222, 229, 247 Chen, Eddy K., 158
Bell’s Chirality, 379–381, 383–385, 389
experiment, 247, 249, 251, 262 Church, Alonzo, 198
inequalities, 247–250, 252, 255–256, 260 Clarke, Samuel, 179
theorem, 75, 124, 127, 129 Classical
Bellomo, Guido, 4, 323 configuration space, 164, 166, 170–171
Benatti, Fabio, 13, 15, 25 limit, 42, 68, 82, 205, 214, 216–220, 360–362,
Big Bell Test, 260 364–365, 370, 373–375
Bohm, David, 77, 81, 83, 124, 126, 129, 133, 135, logic, 72, 188, 196, 199, 202–203, 235, 362–364,
159, 161, 231 366, 374–375
393
394 Index
Classical (cont.) Earman, John, 271, 347
mechanics, 71, 76–77, 81–82, 89, 93, 108, 130, 191, Egg, Matthias, 229
226, 236–237, 239, 242, 275, 278, 280, 283, 346, Eigenstate-eigenvalue link, 9, 13–15, 23, 153, 280,
363 387
physics, 11, 67, 73, 76, 81, 94, 96, 137–139, 158, Einstein, Albert, 45, 77, 90, 121, 123, 150, 164, 168,
177, 205, 235–237 176–178, 219–220, 229, 245–247, 249, 263
statistical mechanics, 89, 236–237 Electron density, 384–385
statistics, 209, 214 Ensemble
universe, 237 canonical, 211, 214
variables, 76–77, 83 grand canonical, 212, 214
Clifton, Rob, 14 Entanglement
Collapse as a relationship between algebras of observables,
hypothesis, 385–386 349
of the wave function, 87–88, 97, 225–227, 233 generalization of, 348
relativistic, 9, 16, 20, 22, 24, 26 relativity of, 348
theories, 9–10, 13, 15–16, 20, 22, 25, 95, 100, 104, Entropy
125, 165–167, 173–174, 224, 228 Burg, 327
Collier, John, 179 quantum, 333
Contextuality, 32, 44, 75–76, 279, 388 Rényi, 327
Correlations, 36–37, 45, 53, 63, 94, 168, 175, Shannon, 327, 330, 334, 336, 339
237–239, 245, 247, 262, 315–316, 329, 334, 336, Tsallis, 327
339, 346–347 von Neumann, 330
Costa de Beauregard, Oliver, 286 Ergodicity, 245, 248–249, 255–256, 259–261
Cramer, John, 220 Esfeld, Michael, 3, 136, 168, 222
Cushing, James, 124, 126, 128 Euclid, 294
Everett, Hugh, 54, 126, 133, 135, 161, 239
d’Espagnat, Bernard, 387 Experiment
Dalton, John, 346 Aspect, 127
de Broglie, Louis, 80, 124, 126, 229, 231 double-slit, 114–115
Decision theory, 103 EPR, 125, 219
Decoherence loophole-free, 245, 248–249, 255, 257
environment-induced, 345–346 noninteracting, 44
relativity of, 353 Stern–Gerlach, 38, 84, 86, 125, 170, 222
self-induced, 375
top-down, closed-system approach to, 346 Faye, Jan, 154
Democritus, 261 Feynman, Richard, 72, 77, 180, 362
Deutsch, David, 103, 121, 129 Finkelstein, Jerry, 21
Dieks, Dennis, 2, 51 Fleming, Gordon, 23
Diosi, Lajos, 95 Folse, Henry, 121
Dirac, Paul, 124, 133, 135, 157, 362, 379, 382 Forman, Paul, 124, 126
Discernibility, 187, 192 Fortin, Sebastian, 4, 345, 360, 379
Distinguishability, 205, 209, 215, 219–220, 228, Frauchiger, Daniela, 58–60, 63, 65
296 French, Steven, 193–194, 196, 206, 208, 220
Distribution Frigg, Roman, 145
Bose-Einstein, 186, 209, 213, 373 Fuchs, Christopher, 136
equilibrium, 83, 85
Fermi-Dirac, 186, 209, 213, 217, 373 Gadella, Manuel, 4, 360
flash, 238 Galilean
Maxwell–Boltzmann, 208–210, 212, 216, 219, 373 covariance, 277, 279
Planck, 210 invariance, 291
Poisson, 226 transformations, 272, 278, 280, 282, 289, 300
Rayleigh-Jeans, 209 Galileo Galilei, 71, 245
Dürr, Detlef, 166 Galois, Évariste, 270
Dynamics Gamow vectors, 367–369, 374
collapse, 21, 167, 227, 230, 232 Gasiorowicz, Stephen, 285
deterministic, 242 Gell-Mann, Murray, 107
Markovian, 16, 18 Generalized contexts, 107, 112
nonlinear, 16 Ghirardi, GianCarlo, 9, 13, 15, 22–23, 25, 95, 226
Index 395
Girardeau, Marvin, 316 Clauser-Horne-Shimony-Holt (CHSH), 250
Gisin, Nicolas, 16, 18 Eberhardt’s, 250
Gleason, Andrew, 75 Mermin’s, 255
Goethe, Johann, 155 Instrumentalism, 122, 129, 233
Goldstein, Sheldon, 15, 136, 166, 247 Interpretation of quantum mechanics
Grassi, Renata, 9, 13, 15, 22–23, 25 atomic modal, 346
Griffiths, Robert, 55, 107 Bohr’s, 85
Group consistent-histories, 55
Galilean, 34, 39–42, 269, 272–274, 276–279, Copenhagen, 72, 77, 83, 121–122, 133, 245,
281–282, 284, 286, 288, 291, 295, 310, 247
312 de Broglie-Bohm pilot-wave, 122, 281
Lie, 272 Everett, 54, 122, 134, 173–174
maximal kinematic, 305, 314 many worlds (MWI), 59, 68, 98–99, 101–104, 126,
Poincaré, 42, 282, 291, 295 165–167, 224
Schrödinger, 35 modal, 32–33, 36, 54, 57, 234,
GRW 280–281
flash theory (GRWf ), 230, 233, 235, 238, 240–241 modal-Casimir, 41
matter density theory (GRWm), 227, 229, 232–233, modal-Hamiltonian (MHI), 33–34, 39–40, 46, 54,
235, 238, 240 281, 356, 388
theory, 9, 53, 161, 165, 173, 220, 226–230 perspectival, 59
Guiding equation, 78–80, 82–84, 86, 231–232, 234, QBist (quantum Bayesian), 141, 160
238, 241 relational, 55, 57
statistical, 247–248, 262
Haag, Rudolf, 297 transactional, 220
Hacking, Ian, 123 Invariance
Haecceity, 193–197, 200 dynamical, 304, 308, 313, 317
Halvorson, Hans, 3, 133, 145 Galilei, 41, 269
Hamilton, William, 247 Parity, 389
Hamiltonian Poincaré, 42
effective, 372 time-reversal, 278, 282, 286–288,
non-Hermitian, 368, 372 290–291
Harshman, Nathan, 4, 294 Irreducible representations (irreps), 40, 295,
Hartle, James, 107 347
Healey, Richard, 136, 140–141, 159–160 Ismael, Jenann, 175
Hegel, Georg W. Friedrich, 155
Heisenberg, Werner, 57, 133, 135, 157, 185, 337 Jaimes Arriaga, Jesús, 4, 379
Hendry, Robin, 383 Jaynes, Edwin, 245
Hettema, Hinne, 384
Hidden variables, 75, 78, 87, 123–124, 165–166, 171, Kant, Immanuel, 124, 155, 379
174, 195, 200, 251, 253, 255, 260 Kastner, Ruth, 3, 205
Hnilo, Alejandro, 4, 245 Kelvin, Lord, 93–94, 96
Holik, Federico, 4, 360 Kent, Adrian, 17
Howard, Don, 164, 168 Kepler, Johannes, 294
Huggett, Nick, 205, 207–208, 211, 214 Klein, Felix, 279
Humeanism, 169, 195, 233, 236 Kochen, Simon, 75, 234
Hund paradox, 379, 381, 383, 386–390 Krause, Décio, 3, 185, 193–194, 196
Hund, Friedrich, 380
Ladyman, James, 3, 121, 179
Improper mixture, 56, 387 Lakatos, Imre, 126, 130
Indiscernibility, 187–188, 197–198, 203 Landau, Lev, 76
Indistinguishability, 44–46, 198, 203, 214, 312, Lange, Marc, 173
315 Laplace, Pierre-Simon, 93, 135
Individuality Laplace’s demon, 237
non-, 186, 188, 191, 193–196 Lattice
transcendental, 193, 208 Boolean, 108, 360, 363
Inequality distributive, 108, 111–112, 363, 365
Clauser-Horne (CH), 251 orthocomplemented, 108, 111–113, 116,
Clauser-Horne inequality (CH), 250 363–364
396 Index
Laue, Hans, 288 Noncommutativity, 73
Laura, Roberto, 2, 107 Nonlocality, 45, 74, 80, 90, 222, 229, 232, 238–239
Leegwater, Gijs, 65 Nonseparability, 214
Leggett, Anthony, 387 North, Jill, 282, 287
Leibniz law (principle), 198, 200, 203, 205 Nozick, Robert, 279
Leibniz, Wilhelm, 179
Lewis, David, 100 Ohanian, Hans, 270
Lewis, Davis, 205 Olkin, Ingram, 323
Lewis, Peter, 167, 176 Omnès, Roland, 107
Lieb, Elliott, 316 Ontology
Lifshitz, Evgeny, 76 distributional, 9–10
Liniger, Werner, 316 of properties, 43, 46
Lo, Hoi-Kwong, 335 particle, 159, 222, 226, 231, 233, 240
Loewer, Barry, 13, 37, 169 primitive, 15, 91, 166–167, 178, 233, 235–236
Lombardi, Olimpia, 2, 4, 32, 269, 345 Optical isomerism, 38, 380, 386
London, Fritz, 52–57
López, Cristian, 4, 269 Parmenides, 261
Lorenz, Max, 323 Particles
Losada, Marcelo, 2, 4, 107, 360 Bohmian, 97–98, 234
Bohmian Dirac sea of, 241
Mach-Zehnder interferometer, 116, 309 Dirac sea of, 241–242
Majorization, 323 identical, 44–45, 78, 304–305, 313, 316, 319
Marshall, Albert, 323 indistinguishable, 311, 314–317
Martínez González, Juan Camilo, 4, 379 individual, 167, 222, 226
Mass density, 13, 15, 22, 25 numerical difference of, 191–192
Matter density point, 186, 225–226, 231–232, 239, 241
field, 226–229, 232, 235, 238 Pauli, Wolfgang, 122
ontology, 25, 167, 229 Pearle, Philip, 9, 14, 22, 24, 28, 95
Maudlin, Tim, 21, 95, 121, 130, 220, 223–224, 230, Peirce, Charles Sanders, 124
234, 239 Penrose, Roger, 95, 133
Maxwell equations, 94, 128, 156, 158, 261, 277 Peres, Asher, 136
Measurement Perspectivalism, 51, 57–59, 61, 64, 67–68
as a symmetry breaking process, 33, 39, 390 Picture/representation
determinate outcomes of, 223 Heisenberg, 90, 108, 110–113, 127, 361, 365, 374
frequency, 36 interaction, 17
reliable, 37 Schrödinger, 111
single, 36–37, 246, 385 Tomonaga–Schwinger, 17, 20
von Neumann model of, 36–38, 55 Planck constant, 79, 360–361, 374
Mermin, David, 21, 180, 254 Plato, 294
Mill, John Stuart, 123 Popescu, Sandu, 335
Minkowski spacetime, 10, 16, 28, 64, 66 Popper, Karl, 126, 261
Molecular Positive-operator-valued measures, 86
chemistry, 38, 380, 382 Principle
structure, 379–380, 382–385, 390 of compositionality, 202
Møller, Christian, 155 of correspondence, 246
Monton, Bradley, 14 of identity of indiscernibles, 191
Motion reversal, 290–291 of impenetrability, 191
Mott, Nevill Francis, 124 of individuality, 45, 189, 191–192
Muller, Fred, 192 of local causality, 172
Musgrave, Alan, 126 of metaphysical continuity, 10–11, 14
Myrvold, Wayne, 2, 9, 65, 67, 168, 173, 176 of superposition, 246, 261, 263
of uncertainty, 78, 323–324, 337–339
Newton laws, 94 Pauli exclusion, 213
Newton, Isaac, 71, 178–179, 346 Projection postulate, 161, 219, 385
Ney, Alyssa, 3, 136, 164–165 Properties
Nguyen, James, 145 atomic, 108–109, 111–112
Nielsen, Michael, 323, 335 bundles of, 43–44, 46, 356
No-go theorems, 43, 65–66, 123 case-, 43, 45
Index 397
contrary, 107, 113, 116 Schrödinger, Erwin, 72, 80, 83, 186–191, 193–196,
definite, 65 201, 261, 331
elementary, 360–362 Schrödinger’s cat, 72, 88, 223, 232
natural physical, 148–149 Schur, Issai, 323, 326, 328
possible, 44, 46 Schur-concavity, 326, 333
relational, 57–58, 280 Sebens, Charles, 103
type-, 43, 45 Self-induced decoherence, 371
Semi-group, 19
Quantum Shimony, Abner, 13, 171
chemistry, 128, 380, 382, 384, 389 Signalling, 16, 18–19
computation, 72, 129, 345 Spacetime
information, 72, 86, 297, 324, 332, 336, 339, Galilean, 16, 127
366 globally hyperbolic, 16
noise, 366 relativistic, 10, 16–17, 20
potential, 81–82 Specker, Ernst, 75, 234
statistics, 45, 78, 205, 207, 209–211, 214, 373 Spontaneous localization, 9, 227, 229–230
Quantum field theory, 11, 21, 42, 47, 73, 78, 90, 128, Spurrett, Don, 179
225, 239, 242, 279, 282, 291, 295 Standard model, 11, 128, 201, 240–241, 297, 346
Quantum Theory of Atoms in Molecules (QTAiM), State
384–385 actual, 54, 137
Quasi-set theory, 200 Bell, 254, 334
Quine, Willard V. O., 157 entangled, 16, 18, 45, 51, 53–54, 62, 89, 99, 170, 187,
223, 251, 256–257, 261, 330, 335–336, 347, 349
Racah, Giulio, 286 extrinsic, 20–22
Randomness, 78, 83–84, 87, 100, 124, 249, 256, 259, GHZ, 253–255
318 perspectival, 56
Realism potential, 137
about the quantum state, 135, 159 reduced, 12, 20–22, 331, 334–335, 345, 350–352,
entity, 123, 129 386–387
local, 247, 251 relational, 56
metaphysical, 129 Suárez, Mauricio, 39, 279
ontic structural, 168 Sudbery, Anthony, 59
scientific, 72, 121–122, 124–125, 127–130, 143, Superselection rules, 304, 337
149 Symmetrization, 46, 214, 219, 317
spacetime state, 151, 168 Symmetry
Redhead, Michael, 206, 208, 220 abelian, 302
Reichenbach, Hans, 145 dynamic, 300, 305
Relativity Galilean, 79, 281, 299
general, 150, 185, 201, 228 gauge, 42
special, 64, 66, 96, 176 group of, 40, 295, 299
Renner, Renato, 58–60, 63, 65 kinematic, 300, 304–305, 313–315, 317
Rigged Hilbert space, 301, 366, 368 of particle permutations, 304
Rimini, Alberto, 226 of the Hamiltonian, 35, 38, 296, 388
Roberts, Bryan, 283 permutation, 103, 188, 192
Ross, Don, 179 Poincaré, 299
Rovelli, Carlo, 55, 57, 135, 141, 160 rotational, 35, 299, 304
Ruffini, Remo, 270 space-time, 299
SU(2), 295
Sachs, Robert, 285
Sakurai, Jun, 285, 287, 290 Tails problem, 13
Saunders, Simon, 126, 134, 136, 142, 192 Tappenden, Paul, 103
Scerri, Eric, 386 Tarski, Alfred, 44
Schaffer, Jonathan, 175 Teller, Paul, 168
Schlosshauer, Maximilian, 350 Tensor product structure, 34, 306–311, 313–316, 348,
Schmidt decomposition, 330–331 353
Schrödinger equation Theorem
covariance of the, 274–277 Birkhoff’s, 326
invariance of the, 41, 269–270, 275–276 Ehrenfest, 362
398 Index
Theorem (cont.) Wave function
equipartition, 215–217 as a field on configuration space, 136
Horn’s, 328 as a mere calculational device, 140
Kochen-Specker, 41, 45, 139, 279, 388 as a multi-field on physical space, 136
Schrödinger, 329, 333 as an abstract mathematical object, 143
Schur’s, 331 conditional, 85, 88
Uhlmann’s, 324 realism, 136, 151, 158
Zanardi’s, 309 representational status of the, 157
Timpson, Christopher, 142–143, 149, 151–153, 158, 168 Weber, Tullio, 226
Tumulka, Roderich, 25, 166, 230 Weihs, Gregor, 259
Typicality measure, 231, 236–237, 241–242 Weyl, Hermann, 187–188, 191–192, 200, 279
Wheeler, John, 121
Uncertainty relations, 52, 123, 130, 215–216, 226, Wigner, Eugene, 287–290, 295, 299
237, 240, 324, 337–339, 372 Wigner’s friend, 55, 57–58, 65, 67
Urelemente, 199 Wilson, Jessica, 175
Wiseman, Howard, 171–172
Vaidman, Lev, 2, 93, 135, 253
van Fraassen, Bas, 32, 129, 149, 151 Yang, Chen Ning, 316
Vanni, Leonardo, 2, 107
von Helmholtz, Hermann, 155 Zanardi, Paolo, 297, 308, 310, 315
von Neumann, John, 52–54, 56, 75, 95, 122 Zanghí, Nino, 2, 71, 136, 166
Zeeman effect, 38
Wallace, David, 103, 134, 136, 142–143, 148–149, Zeh, Heinz-Dieter, 371
151–153, 158, 161, 168 Zeilinger, Anton, 121
Watanabe, Satosi, 286 Zurek, Wojciech, 349, 353, 355, 371

Fortin, Sebastian - Holik, Federico - Lombardi, Olimpia - López, Cristian - Quantum Worlds - Perspectives On The Ontology of Quantum Mechanics-Cambridge University Press (2019)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fortin, Sebastian - Holik, Federico - Lombardi, Olimpia - López, Cristian - Quantum Worlds - Perspectives On The Ontology of Quantum Mechanics-Cambridge University Press (2019)

Uploaded by

Copyright:

Available Formats

QUANTUM WORLDS

olimpia lombardi is a principal researcher at the National Scientiﬁc and

Cambridge University Press is part of the University of Cambridge.

List of Contributors page vii

2 The Modal-Hamiltonian Interpretation: Measurement,

3 Quantum Mechanics and Perspectivalism 51

4 Quantum Physics Grounded on Bohmian Mechanics 71

5 Ontology of the Wave Function and the Many-Worlds Interpretation 93

6 Generalized Contexts for Quantum Histories 107

Part II Realism, Wave Function, and Primitive Ontology 119

8 To Be a Realist about Quantum Theory 133

9 Locality and Wave Function Realism 164

Part III Individuality, Distinguishability, and Locality 183

11 From Quantum to Classical Physics: The Role of Distinguishability 205

12 Individuality and the Account of Nonlocality: The Case for the

13 Beyond Loophole-Free Experiments: A Search for Nonergodicity 245

Part IV Symmetries and Structure in Quantum Mechanics 267

15 Symmetry, Structure, and Emergent Subsystems 294

16 Majorization, across the (Quantum) Universe 323

Part V The Relationship between the Quantum Ontology and

18 A Logical Approach to the Quantum-to-Classical Transition 360

19 Quantum Mechanics and Molecular Structure:

Jesús Jaimes Arriaga

Juan Camilo Martínez González

Developing a research group in philosophy of physics is not an easy task in

complexity. If each different perspective tells us a different story about the

Sebastian Fortin address the difﬁculty of giving a quantum explanation of chirality,

an account, dynamical quantities such as charge or mass within a speciﬁed region

1.2 The Case for Distributional Ontology

structure of a world in which there is no gravitation and in which the standard

1.2.2 The Requirement of Local Beables

1.2.3 Ontology for Ideal Collapse Theories

1.2.4 Distributional Ontology

On this view, the value of every dynamical variable is distributional. A collapse

1.2.5 Primitive Ontology as an Alternative?

1.3 A Schema for Relativistic Collapse Theories

1.3.2 Collapse Theories in Relativistic Spacetime

The relevant condition is the following:

As Kent (2005) has argued, although violations of linearity permit signalling,

difference between states on different Cauchy surfaces is due to the stochastic

in a set Γ. We expect our theories to specify, given two surfaces σ, σ0 , with σ σ0 ,

with probabilities for which state is realized given by

The linearity condition entails that

Consider two Cauchy surfaces σ, σ0 , with σ σ0 , that coincide everywhere

condition for this is that evolution operators corresponding to spacelike separated

1.4 Beables for Relativistic Collapse Theories

Figure 1.2 Bounded spacelike region α, that is common to Cauchy surfaces

1.4.2 Compatibility of Extrinsic States

1.4.3 Local Beables for Collapse Theories

This means that, for x within α, T μν ðxÞ will be regarded as representing an

criterion for property attribution is satisﬁed. If Bob chooses to do a spin experiment

As mentioned previously, Ghirardi et al. (1995) have argued that a theory on

We consider a ﬁnite set of Cauchy surfaces fσ1 ; σ2 ; . . . ; σn g, all containing an open

spacelike separated from α. Therefore, K γ commutes with all self-adjoint elements

Proof. This is immediate from Lemma 1.

Bassi, A. and Ghirardi, G. C. (2003). “Dynamical reduction models,” Physics Reports,

Shimony, A. (1991). “Desiderata for a modiﬁed quantum mechanics,” pp. 49–59 in A.

interpretations (Harvey Brown, cited in Bacciagaluppi and Hemmo 1996). This

2.2 The Modal-Hamiltonian Interpretation

Composite systems postulate: A quantum system S : ðO; H Þ, with initial state

The justiﬁcation for selecting the Hamiltonian as the preferred observable

2.3 The Modal-Hamiltonian View of Quantum Measurement

• Single measurement: It is a single process, in which the reading of the pointer is