Quantum

Fractals
From Heisenberg's Uncertainty
to Barnsley's Fractality

8992_9789814569866_tp.indd 1 17/6/14 9:08 am

May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

 Quantum
Fractals
From Heisenberg's Uncertainty
to Barnsley's Fractality

Arkadiusz Jadczyk
Quantum Future Group Inc., USA

World Scientific
NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TA I P E I • CHENNAI

8992_9789814569866_tp.indd 2 17/6/14 9:08 am

 Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

Library of Congress Cataloging-in-Publication Data

Jadczyk, Arkadiusz.
Quantum fractals : from Heisenberg’s uncertainty to Barnsley’s fractality / Arkadiusz Jadczyk,
Quantum Future Group Inc, USA.
pages cm
Includes bibliographical references and index.
ISBN 978-9814569866 (hardcover : alk. paper)
1. Fractals. 2. Mathematical physics. 3. Quantum theory. I. Title.
QC20.7.F73J33 2014
530.1201'514742--dc23
2014013078

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

Copyright © 2014 by World Scientific Publishing Co. Pte. Ltd.

All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy
is not required from the publisher.

In-house Editor: Ng Kah Fee

Printed in Singapore

KahFee - Quantum Fractals.indd 1 17/6/2014 12:13:33 PM

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. v/0

To Laura
for her support of the creative principle

v
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. vii/0

Preface

This book provides a handy introduction to quantum fractals — a new

kind of fractals arising in quantum-like jump random processes involving
non-commuting operations. It describes the basic theoretical concepts, al-
gorithms and also touches upon philosophical questions of the foundations
of quantum theory.

An overview The science of fractals is young and growing fast. Quantum

fractals are even younger and are still crawling on all fours. But the time
seems to be ripe for them to get up and look around. As we hope it will
become clear from this book — various possible applications abound.
Roughly: quantum fractals are patterns generated by iterated function
systems, with place dependent probabilities, of Möbius transformations on
spheres or on more general projective spaces. In quantum physics quantum
fractals can be interpreted as traces of quantum jumps during simultaneous
monitoring of several non-commuting observables. These quantum jumps
accompany events with information exchanges between the quantum system
and the classical information processing devices.
While mathematically completely clear such a concept brings an almost
revolutionary novelty into quantum physics. Until now it has usually been
assumed that simultaneous “measurements” of non-commuting observables
makes no sense, and that it cannot lead to any useful predictions. In this
book we challenge the standard position by proposing that such experi-
ments may lead to organized chaotic behavior that can be experimentally
verified. The phenomenon is general enough to be present in applications of
the quantum formalism beyond physics and beyond quantum computing,
for instance in quantum games, quantum psychology etc. Here possible
deviations from linearity are also touched upon.

vii
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. viii/0

viii Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Another area where quantum fractals may appear is relativistic light

aberration, though there the particular form of place dependent probabili-
ties that are derived from linear event enhanced quantum theory (EEQT)
is not, at the present moment, justified.

About the Book This book combines a number of different topics such
as: fractals generation and analysis, elements of geometry, linear and multi-
linear algebra and group theory, special relativity, quantum measurements
and, in particular, Heisenberg’s uncertainty relations and their interpreta-
tion, as well as some elements of random processes.
Since it is rather unusual to find a single person that would be inter-
ested in all of these areas, this book has been organized in such a way that
the reader should be able to extract from it the information that is of a
particular interest in her/his research. Nevertheless the primary idea of
the book is to bring together a diversity of ideas and, in this way, encour-
age cooperation and stimulate mutual interest between various branches of
quantum physics and fractal research for the benefit of all.
For this reason the book has not been organized in strict linear or-
der. To facilitate the process of extracting the information of interest there
are repetitions: the same concept may appear in the book several times,
though in a somewhat different context and stressing different aspects. The
reader that would like to know more about a given concept can always find
additional information by perusing the index.
For those who wish to start with looking first at examples: they can
start with “the impossible quantum fractal” — Sec. 2.5 — and then check
examples of hyperbolic quantum fractals in Sec. 3.1.
Those who are simply looking for algorithms and examples of the code
that were used for generating these examples may like to start directly with
Sec. 3.4.
On the other hand readers interested in the foundations of quantum
theory can start with Chap. 4 or one of its sections.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. ix/0

Contents

Preface vii

1. Introduction 1

2. What are Quantum Fractals? 11

2.1 Cantor set . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Cantor set through “Chaos Game” . . . . . . . . 13
2.2 Iterated function systems . . . . . . . . . . . . . . . . . . 15
2.2.1 Definition of IFS . . . . . . . . . . . . . . . . . . . 18
2.2.2 Frobenius-Perron operator . . . . . . . . . . . . . 22
2.3 Cantor set through matrix eigenvector . . . . . . . . . . . 23
2.4 Quantum iterated function systems . . . . . . . . . . . . . 25
2.5 Example: The “impossible” quantum fractal . . . . . . . . 29
2.5.1 24 symmetries — the octahedral group . . . . . . 30
2.5.2 Construction of the 24-elements SQIFS . . . . . . 31
2.5.3 Open problems . . . . . . . . . . . . . . . . . . . . 37
2.6 Action on the plane . . . . . . . . . . . . . . . . . . . . . 41
2.7 Lorentz group, SL(2, C), and relativistic aberration . . . . 43
2.7.1 The Lorentz group . . . . . . . . . . . . . . . . . . 43
2.7.2 Action of the Lorentz group on the sphere . . . . 45
2.7.3 The group SL(2, C) . . . . . . . . . . . . . . . . . 48
2.7.4 Action of SL(2, C) on the two-sphere S 2 . . . . . 50
2.7.5 Projection operators representations of the Bloch
sphere S 2 . . . . . . . . . . . . . . . . . . . . . . . 55
2.7.6 Visualization of quantum spin states and state
vectors . . . . . . . . . . . . . . . . . . . . . . . . 60

ix
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. x/0

x Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

2.7.7 Action on orthogonal projections and Möbius

transformations . . . . . . . . . . . . . . . . . . . 68
2.7.8 Exponential map in SL(2, C) . . . . . . . . . . . . 70
2.7.9 Two different eigenvalues . . . . . . . . . . . . . . 75
2.7.10 Classification of Möbius transformations . . . . . 76
2.7.11 Area transformation law . . . . . . . . . . . . . . 78
2.7.12 Relativistic aberration . . . . . . . . . . . . . . . 85
2.7.13 Example: Special subgroup of parabolic
transformations . . . . . . . . . . . . . . . . . . . 94
2.7.14 Pythagorean triples and quadruples . . . . . . . . 96

3. Examples 109
3.1 Hyperbolic quantum fractals . . . . . . . . . . . . . . . . 109
3.1.1 The circle . . . . . . . . . . . . . . . . . . . . . . . 110
3.1.2 Platonic quantum fractals for a qubit . . . . . . . 120
3.2 Controlling chaotic behavior and fractal dimension . . . . 167
3.3 Quantum fractals on n-spheres . . . . . . . . . . . . . . . 171
3.3.1 Clifford algebras . . . . . . . . . . . . . . . . . . . 173
3.3.2 Stereographic projection . . . . . . . . . . . . . . 194
3.3.3 Conformal maps and Frobenius-Perron operator . 195
3.4 Algorithms for generating hyperbolic quantum fractals . . 199
3.4.1 Chaos game on n-sphere . . . . . . . . . . . . . . 202
3.4.2 Approximation to the invariant measure . . . . . 208

4. Foundational Questions 213

4.1 Stochastic nature of quantum measurement processes . . . 213
4.2 Are there quantum jumps? . . . . . . . . . . . . . . . . . 227
4.3 Bohmian mechanics . . . . . . . . . . . . . . . . . . . . . 233
4.4 Event Enhanced Quantum Theory . . . . . . . . . . . . . 240
4.4.1 Piecewise deterministic process . . . . . . . . . . . 242
4.4.2 Algorithm for the piecewise deterministic
process (PDP) . . . . . . . . . . . . . . . . . . . . 244
4.4.3 Association of the semigroup with PDP . . . . . . 244
4.4.4 Central classical observables . . . . . . . . . . . . 245
4.4.5 Quantum Events Theory — Duality . . . . . . . . 249
4.4.6 Completely positive maps . . . . . . . . . . . . . . 251
4.4.7 Dynamical semigroups on an algebra with a
center . . . . . . . . . . . . . . . . . . . . . . . . . 254
June 16, 2014 8:52 World Scientific Book - 9in x 6in QuantumFractals3 pg. xi/0

Contents xi

4.4.8 Liouville equation for states . . . . . . . . . . . . 256

4.4.9 Ensemble and individual descriptions . . . . . . . 256
4.5 Ghirardi-Rimini-Weber spontaneous localization . . . . . 259
4.5.1 The coupling . . . . . . . . . . . . . . . . . . . . . 262
4.6 Heisenberg’s uncertainty principle and quantum fractals . 264
4.6.1 Simple examples . . . . . . . . . . . . . . . . . . . 268
4.6.2 A single detector . . . . . . . . . . . . . . . . . . 269
4.6.3 Measurement of non-commuting observables . . . 270
4.6.4 The simplest toy model — space and momentum
are each only two-points . . . . . . . . . . . . . . 272
4.7 Are quantum fractals real? . . . . . . . . . . . . . . . . . 279

Appendix A Mathematical Concepts 297

A.1 Metric spaces . . . . . . . . . . . . . . . . . . . . . . . . . 297
A.1.1 Compact metric spaces . . . . . . . . . . . . . . . 300
A.1.2 Locally compact metric spaces . . . . . . . . . . . 300
A.2 Normed spaces . . . . . . . . . . . . . . . . . . . . . . . . 301
A.2.1 Banach spaces . . . . . . . . . . . . . . . . . . . . 301
A.2.2 The space C(X, Y ) . . . . . . . . . . . . . . . . . 302
A.3 Measure and integral . . . . . . . . . . . . . . . . . . . . . 302
A.3.1 Borel sets . . . . . . . . . . . . . . . . . . . . . . . 303
A.3.2 Measure . . . . . . . . . . . . . . . . . . . . . . . 303
A.3.3 Integral . . . . . . . . . . . . . . . . . . . . . . . . 304
A.3.4 Lp spaces . . . . . . . . . . . . . . . . . . . . . . . 306
A.4 Markov, Frobenius-Perron and Koopman operators . . . . 308

Appendix B Minkowski Space Generalization of Euler-Rodrigues

Formula 311
B.1 Alternative derivation via SL(2, C) . . . . . . . . . . . . . 314

Bibliography 317

Index 329
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 1/1

Chapter 1

Introduction

This book brings together two concepts. The first is over a hundred years
old — the “quantum”, while the second, “fractals”, is newer, achieving
popularity after the pioneering work of Benoit Mandelbrot. Both areas of
research are expanding dramatically day by day. It is somewhat amaz-
ing that quantum theory, in spite of its age, is still a boiling mystery as
we see in some quotes from recent publications addressed to non-expert
readers:

“Heisenberg uncertainty principle stressed in new test”:

Pioneering experiments have cast doubt on a founding idea
of the branch of physics called quantum mechanics. The
Heisenberg uncertainty principle is in part an embodiment
of the idea that in the quantum world, the mere act of
measuring can affect the result. But the idea had never
been put to the test, and a team writing in Physical Review
Letters says “weak measurements” prove the rule was never
quite right. BBC News, September 7, 2012 [Palmer (2012)]

“Will we ever understand quantum theory?”: Quantum

mechanics must be one of the most successful theories in
science. (....) Yet the weird thing is that no one actually
understands quantum theory. BBC Future, January 25,
2013 [Ball (2013b)]

“Quantum Weirdness? It’s All in Your Mind”: ...

Quantum mechanics is an incredibly successful theory but
one full of strange paradoxes. ... Scientific American, June
10, 2013 [von Baeyer (2013)]

1
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 2/1

2 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

“Proof mooted for quantum uncertainty: Study confirms

principle’s limits on measurement accuracy”: Encapsu-
lating the strangeness of quantum mechanics is a single
mathematical expression. According to every undergradu-
ate physics textbook, the uncertainty principle states that
it is impossible to simultaneously know the exact position
and momentum of a subatomic particle — the more pre-
cisely one knows the particle’s position at a given moment,
the less precisely one can know the value of its momen-
tum. But the original version of the principle, put forward
by physicist Werner Heisenberg in 1927, couches quan-
tum indeterminism in a different way — as a fundamental
limit to how well a detector can measure quantum proper-
ties. Heisenberg offered no direct proof for this version of
his principle, and expressed his ideas only informally and
intuitively’, says physicist Jos Uffink of the University of
Minnesota in Minneapolis. Nature, June 27, 2013 [Cowen
(2013)]

It’s “an inertia” of thinking, says David Deutsch, a British physicist at

the University of Oxford [Deutsch (2003)] in his paper on “Physics, Phi-
losophy and Quantum Technology”, suggesting the need of a more general
“Quantum Constructor Theory.” The main issues here are “quantum infor-
mation”, “quantum computing”, and “quantum cryptography”. But we do
not yet understand quantum theory, its limitations, its full potential, and
its connections to other areas of science.
There is no doubt about the successful applications of quantum theory
in technology of today and of tomorrow. Yet if we want to make the full use
of this theory, we need to cure it from its diseases and apparent paradoxes
that follow from its vagueness in some areas. If the solution to these prob-
lems have not been found till now, something must have been overlooked.
Probably something related to “quantum measurement”. Chapter 4 of this
book discusses those foundational questions which also happened to be re-
lated to the processes relevant for creation of quantum fractals.
One such important problem is Heisenberg’s uncertainty and its inter-
pretations and implications. There are precise mathematical formulas, and
there are their interpretations. These interpretations do not follow from
the so successful mathematical machinery — they are imposed on top of it
with questionable philosophical underpinnings.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 3/1

Introduction 3

My research related to quantum fractals targets that question: can we

expand the formalism so as to describe the processes not yet described and
predict what has not been predicted so far? Can we do it in such a way that
we will still be in agreement with the numerous successful applications?
Quantum fractals result from a new kind of measurement processes. I
became interested in quantum fractals because they are considered “im-
possible” and/or irrelevant. Yet they exist within a well defined theory
that describes the so far vague concept of “measurement”. Indeed, after
sufficient amount of research we find that their “impossibility” and/or irrel-
evance is the result of inertia and prejudices. If Quantum Theory still poses
unresolved problems, it can only be because of deeply ingrained prejudices
— nothing else.
String theories and other “theories of everything” will not help a least
in this respect. They do not even try to address the foundational problems
of quantum theory. As a result fundamental unresolved issues become even
more obscured.
There are many different interpretations of quantum theory: the stan-
dard, orthodox, Copenhagen interpretation, Bohmian, CSL (continuous
spontaneous localization), GRW (Ghirardi-Rimini-Weber), transactional,
Bayesian, ... The poll conducted among the participants of the conference
“Quantum Physics and the Nature of Reality” held in July 2011 in Aus-
tria, and organized by Anton Zeilinger, revealed that opinions of the leading
experts evenly split among different interpretations and their implications
[Schlosshauer et al. (2013)]. But none of these different interpretations
really predict dramatic new effects.
Some physicists say “do not worry about interpretation”, just “shut up
and calculate”. Louisiana State University physicist Jonathan P. Dowling,
specializing in quantum computing, describes this attitude in detail using
himself as an example:
When it comes to interpretations of quantum theory, I
am a pantheist. (...) Pantheism is not without its pitfalls,
but unlike Pascal’s original wager, you are hedging more
of your bets. I’m happy to revel in the Copenhagen inter-
pretation on Monday, Wednesday, and Friday; exploit the
Many-Worlds interpretation on Tuesday, Thursday, and
Saturday; and on Sunday turn in desperation to Bohm the-
ory. Unlike Pascal’s wager, however, I have nothing to lose
by this strategy and everything to gain. Often a perplexing
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 4/1

4 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

quantum problem becomes completely clear to me when I

switch from Copenhagen to Many Worlds. (Nothing is ever
cleared up when I resort to Bohm theory but then I usu-
ally sleep in on Sundays.) I don’t believe any of them!
or perhaps I believe all of them. The physicist Edwin T.
Jaynes once said, So long as I can use and teach a physical
theory, I don’t have to believe it. Well played! [Dowling,
2013b, p. 34–35]
Yet interpretation does matter: it suggests to us the directions of our re-
search. It shapes questions that we ask. It tells us what to not bother
about while designing new experiments. We are convinced that new effects
will come very soon from overcoming the tabus imposed by our current in-
terpretation of Heisenberg’s uncertainty relations. Discussing these issues
in detail would take a whole new book, however we discuss the main points
specifically in Sec. 4.6.
One may wonder why I devote space to those issues that this book is
really not about? The book is mainly about fractals — of a special kind.
But I am first of all a physicist. Quantum Fractals emerged as a byproduct
of my research. Here I am concentrating upon the fractal aspects, related
methodology questions, structure and properties of these fractals. I hope
that this book will encourage quantum physicists, busy with their quantum
cryptographical qubits, to look into something having not only applied
appeal but also aesthetical values: quantum fractals that are hidden in
their qubits. I also hope that the fractal industry will expand owing to the
new perspective stemming from quantum theoretical applications.
The mathematical formalism developed here should have applications
in those areas of research that stem from quantum concepts applied within
quantum theory itself but also beyond. Since, here and there, while dis-
cussing quantum fractals I have to use precise mathematical terms that are
rather specialized, the book has an Appendix where a concise but precise
definition of the most important mathematical terms is given.
New questions will reveal new fields of research and cast new light on
old areas. With Quantum Fractals come challenges that will open new
paths. Can we find these quantum fractals in nature today? If so, where?
This question is addressed, within the framework of Quantum Theory, in
Sec. 4.7. But my guess is that most promising is the search for quantum
fractals in the cosmos and in the study of light.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 5/1

Introduction 5

Quantum Fractals have a history and, as I am convinced you will see

from this book, also a bright future. Originally they emerged from my
interest in the foundational problems of the still mysterious quantum theory,
as well as my work on filling up its most disturbing blind spots — “the
measurement problem”.
John Stewart Bell (1928–1990), the distinguished CERN quantum
physicist, gave one of his papers the provocative title: “Against Measure-
ment” [Bell (1990)]. There he suggested banning the word “measurement”
from physicist vocabulary altogether! Nevertheless quantum theory is a
theory of measurements — as it is often repeated. But what are these
“measurements”? What kind of processes are involved there?

Fig. 1.1 Spikes and bursts, in a way similar to quantum jumps, interrupt the continuous
accumulation of the electric potential in neuronal activity. URL: http://en.wikipedia.
org/wiki/Neural\%20oscillation.

It is while trying to find answers to these questions, answers that would

satisfy me, I came up with an idea of a “piecewise deterministic process”
that produced quantum jumps as discrete events taking place in time.
Continuous time evolution of potentialities encoded in the quantum wave
function is, in this approach, interrupted by occasional discontinuities —
“quantum jumps” — according to a precisely defined piecewise determinis-
tic random process. The resulting process produces a picture that is similar
to that observed in neuronal activity in the brain (see Fig. 1.1). Similar
spikes can be seen, for instance, in sunspot activity and financial markets
— which is intriguing enough in itself.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 6/1

6 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

From the mathematical description of the process it became evident to

me that measuring simultaneously quantum complementary quantities is
possible — in a sense that is contrary to the prevailing interpretations of
Heisenberg’s uncertainty relations. Mathematics not only allows us to do
what the standard quantum theory never dare to do, it also predicts the
patterns — though chaotic and fractal-like. That may be the reason why
there were no attempts to see these patterns experimentally, as there was
no theory indicating what to look for.
To look for something we must first have a theory, or at least an idea,
even if only rough and approximate, perhaps even a false one, of what to
look for. Now that we have such a theory, with the piecewise-deterministic
process paradigm, with quantum fractals, perhaps experiments will follow?
As of today, quantum fractals are just images and algorithms. They are
pretty. But in the future, I am sure, we will find them all around us. And
not only in the highly specialized realm of quantum microphysics, but also
in macrocosmic light patterns, and also in the extensions of the quantum
formalism that regulate social and psychic phenomena, processes in the
brain, thought formation, etc.
In the above mentioned 2011 poll [Schlosshauer et al. (2013)] (see also
[Ball (2013a)]) on the foundational issues of quantum theory we find

“More than two-thirds believed that there is no funda-

mental limit to quantum theory — that it should be possi-
ble for objects, no matter how big, to be prepared in quan-
tum superpositions like Schrödinger’s cat. So the era where
quantum theory was associated only with the atomic realm
appears finally over.”

Certainly Quantum Fractals do have their Quantum Future, but let me

return to their history. Originally quantum fractals came up as patterns
formed as the results of millions of consecutive quantum jumps of a spin
1/2 quantum state during a continuous monitoring of mutually incompati-
ble spin directions. More generally they can be also formed during continu-
ous monitoring of position and momentum, when the wave functions jump
wildly. Millions of wave function reductions1 would give us a fractal-like
pattern on the phase space — something that is impossible according to
the standard quantum mechanics (except, perhaps, for Wigner’s function,
1 In some experiments demonstrating “Quantum Zeno Effect” [Streed et al. (2006);

Balzer et al. (2002)], thousands of jumps have been repeatedly registered.

June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 7/1

Introduction 7

which hides the resulting chaos behind “negative probabilities”). This pro-
gram has not yet been implemented, but I see no reason why it cannot be.
Wave functions, or quantum state vectors are strange objects. Their
ontological status is debatable. Can we really “see”, in some kind of an
experiment, the patterns that they form? What are these wave functions
anyway?
According to some interpretations wave functions (or “pure quantum
states”) are purely subjective, they represent our knowledge. As Henry
Stapp puts it in his book “Mind, Matter and Quantum Mechanics” [Stapp
(1993)]:

“The Copenhagen interpretation is often criticized on

the grounds that it is subjective, i.e., that it deals with the
observer’s knowledge of things, rather than those things
themselves. This charge arises mainly from Heisenberg’s
frequent use of the words ‘knowledge’ and ‘observer’. Since
quantum theory is fundamentally a procedure by which sci-
entists make predictions, it is completely appropriate that
it refer to the knowledge of the observer. For human ob-
servers play a vital role in setting up experiments and in
noting their results.”

A pure objective interpretation of quantum wave functions can hardly be

spotted. Yet a subjective (or Bayesian) interpretation is not the only one
possible. Wave functions can be shadows of some deeper, perhaps hy-
perdimensional, reality, encoding space-time-energy information about the
objects, information based on the data that have been irreversibly regis-
tered and stored. Shadows can be seen — even if they are only shadows.
Shadows can have shadows of their own. Shadows can be registered and
analysed through the set of discrete data. Once we have a theory that
tells us how to decode the data, we can reconstruct the shadows. Once we
have a theory that tells us how to reconstruct from shadows of certain level
shadows of a deeper level, we can reconstruct these deeper levels as well.
In this book we will touch upon all these problems and describe first
steps in these directions.
But the mathematical formalism and algorithms of quantum fractals
apply not only to quantum state vectors but also to relativistic light aber-
ration. Therefore they can be, in principle, observed, in the sky — as there
is the formal relation between the quantum spin state Bloch sphere and the
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 8/1

8 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

heavenly sky of Special Relativity Theory.2 My favorite example here is

somewhat out of touch with reality, but it serves the purpose of activating
our imagination. To set the scene let me quote a passage from “Beyond
Startrek Physics from Alien Invasions to the End of Time” by Lawrence M.
Krauss [Krauss, 1998, p. 21]:

The traditional notion has been that UFOs don’t be-

have like rockets or planes (this is, after all, what makes
them UFOs). Strange lights that flit unlikely distances back
and forth across the sky, like the dazzling display in Steven
Spielberg’s Close Encounters of the Third Kind, are typ-
ical. More recently, in one of the early episodes of The
X-Files, the ardent UFO hunter and FBI agent Fox Mul-
der finally gets to see some real UFOs in a secret air force
installation somewhere in the Southwest (could it be Area
51?), and these vehicles do just what UFOs are supposed
to do — namely, everything our own aircraft can’t. Mulder
and his colleague Dana Scully are astounded by a series of
bright disks moving at incredible speeds through the skies
above the remote base, turning at 90-degree angles on a
dime.

Now imagine a monkey flying such an UFO disk. The monkey operates
a joystick with 24 positions — each position defines an instant maneuver
that changes the direction and rotates the disk. The monkey marks on the
hemispherical transparent dome the position of a fixed star — say Alpha
Centauri. After each such maneuver the new mark appears on the dome
— the change of the visible star position is the result of the relativistic
aberration. The monkey operates the joystick randomly and is not getting
tired. After about a hundred thousand of such random operations a fractal
pattern starts to appear on the screen.
Today the above scene, although it is supported by rigorous mathemat-
ics, is fit only for a Sci-Fi movie. The author of “Beyond Star Trek” rightly
argues that such maneuvers are impossible according to the conventional
science and technology. But human beings seem to fear no challenges, and
tomorrow our science and technology can be ready for experiments resulting
in essentially the same effect.
2 For a somewhat different perspective on quantum fractals see [Slomczyński et al.

(2000); Wojcik et al. (2000); Lozinski et al. (2003)].

June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 9/1

Introduction 9

Fig. 1.2 Parabolic Quantum Fractal. The ocean around is a symbolic (“artistic”) rep-
resentation of the classical world.

Perhaps the most amazing of all quantum fractals is the one I have
discovered after starting the work on this book — the parabolic quan-
tum fractal stemming from the most simple formula z → z + 4 and its
24 octahedral variations — you see on the cover of this book. In Fig. 1.2
we have exterior view of this object.3 Its shape reminds us of the Buck-
minster Fuller geodesic dome — we may call it the Quantum Dome. It has
huge circular “windows”. I do not have a mathematical formula for these
windows, though certainly there is one — yet to be discovered. When stere-
ographically projected from the sphere on the plane — the windows that
have different sizes on the sphere acquire the same size on the plane. The
Quantum Dome fractal is also “an impossible one” — according to a math-
ematical theorem proved by Andrew Vince [Vince (2013)] (based on the
definition of an attractor proposed by Barnsley and Vince [Barnsley and
Vince (2011)]) we can’t get a “true fractal” from parabolic transformations
such as z → z + 4.4 And yet here it is, and its existence asks for a better
understanding of what should be called a fractal and what not.

3 You can see it rotating on YouTube channel:

http://www.youtube.com/watch?v=YNzf1o4GAAY
4 Of course I tried also other parabolic transformations such as z + 1 and z + 2, but

z + 4 produces the most pleasing pattern.

June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 10/1

10 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

To speculate even further5 the Quantum Dome (or similar fractals) may
also have engineering applications. Can such a design have some special
optimizing material properties? Can it be used as a “fractal antenna”
of some kind, the way fractal antennas are being used in sophisticated
engineering cell phones. Can it be used for space-propulsion of some sort?
I am having these thoughts while looking at the intriguing and unexpected
shapes and design patterns.
There is also a mysterious connection to Pythagorean triples and
quadruples (see Sec. 2.7.14), discrete space-time structures etc. All of these
deserve further research. This book may serve, I hope, as a jumping-off
platform in this respect.

5 To quote from Irving John Good, a British brilliant mathematician, who worked as

a cryptologist with Alan Turing: “It is often better to be stimulating and wrong than
boring and right.” [Good, 1962, p. 1]
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 11/2

Chapter 2

What are Quantum Fractals?

Before discussing quantum fractals let us start with the classical example
of a fractal — the Cantor set. This simple example allows us to become
accustomed with important concepts that will be needed later on.

2.1 Cantor set

One of the simplest, and probably most important fractals is called “the
Cantor set”. George Cantor, a maverick German mathematician (1845–
1918) was the first one who dared to introduce a rigorous concept of in-
finities into mathematics. His life was not an easy one as he had to fight
against strong prejudices both from his colleagues mathematicians as well
as from theologians. At that time the concept of “infinity” was reserved for
God. The great French mathematician of that time, Henri Poincaré was
accusing Cantor of introducing a “great disease” into mathematics, while
Cantor’s famous German colleague, Leopold Kronecker, considered him to
be a scientific charlatan, a renegade and a corrupter of youth. The Church
accused Cantor of propagating pantheism. Perhaps partly as the result of
all this Cantor died after being driven mad [Dauben (1990)]. However his
infinities have found their way into mathematics rather quickly, and today
his calculus of infinities (so called “cardinal numbers”) is being taught to
undergraduate students.1
The Cantor set, that this section is about, is a fractal which is one-
dimensional, therefore, visually, rather dull, unless you are a mathematician
that can appreciate its abstract properties. The algorithm for creating the
Cantor set is extremely simple:
1 The same is not however true about Cantor’s philosophy. As a philosopher Cantor

was a dualist — he was dividing reality into a material part and an “aetherial” part.

11
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 12/2

12 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

• Take the unit interval. Remove the middle third.

• From each of the remaining thirds remove their middle thirds.
• And so on.

For the true Cantor set this construction never stops. The Cantor set is
the limiting set of the algorithm. In practice we stop after few steps and
enjoy the result. We get the idea.
Visually it is better to present the Cantor set as a two-dimensional strip
— similar to the strips we see when looking at the spectra of chemical
elements — cf. Fig. 2.1.

Fig. 2.1 Cantor set.

Saturn’s rings have also certain features of the Cantor set [Avron and
Simon (1981); Heck and Perdang (1991)].

Fig. 2.2 Fractal analysis of Saturn rings — [Li and Ostoja-Starzewski (2012)].

A recent detailed analysis [Li and Ostoja-Starzewski (2012)] (cf.

Fig. 2.2) estimates the fractal dimension of Saturn’s rings in the range
1.63–1.77, while the mathematical Cantor set has fractal dimension equal
to log(2)/ log(3) ≈ 0.63. The difference of 1.0 is due to the fact that Saturn’s
ring are two-dimensional, while the Cantor set is one-dimensional. If we
would take a one-dimensional cross-section of Saturn’s ring, we would get
estimates 0.63–0.77, thus not far from the fractal dimension of the Cantor
set.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 13/2

What are Quantum Fractals? 13

A nice and illuminating animated demonstration of the Cantor set, with

zooming and panning capability, can be found on Wolfram’s Demonstra-
tions Project site
http://demonstrations.wolfram.com/CantorSet/

2.1.1 Cantor set through “Chaos Game”

The algorithm for generating Cantor set via “remove the middle third ad
infinitum” is simple, but it does not suit our purpose here for the follow-
ing reason: quantum fractals — the main subject of this book — relate to
quantum theory, they have something to do with physics, not just mathe-
matics.
Remove the “middle set of what”? — a physicist will ask. In the case of
quantum fractals “the set” is “the set of quantum states”. We can’t remove
a part of the set of quantum states easily. We do not have such scissors!
On the other hand quantum states can change with time. Usually they
evolve continuously in time, but, once in a while, they also abruptly “jump”.
We can exercise some control over their time evolution, the continuous one,
and we can influence the jumping process as well.
For instance: an external homogeneous magnetic field causes the spin
to continuously precess around the magnetic field direction. This is called
Larmor precession. If we want to concentrate only on jumps, since it is
these jumps that are responsible for quantum fractals, we need to freeze or
“subtract” the continuous part. A spin direction measuring device causes
quantum spin state to jump in a discontinuous way.
In a similar way the Cantor set can be also described through a long
sequence of jumps — just like those quantum jumps that accompany quan-
tum measurements, and this is the method we will discuss now. This is
closer to physics. The corresponding algorithm is often called “the Chaos
Game”.
The Chaos Game involves a lottery — much like in all hazard games.
The cards are being shuffled (or the dice are being shaken) well enough so
that the outcome is approximately random. For the construction of the
Cantor set having just one fair coin is enough. The two sides of the coin,
the head and the tail, cause two different transformations of real numbers
in the interval (0, 1). Let us denote these transformations T1 and T2 . They
are defined as follows:
1
T1 : x → x, (2.1)
3
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 14/2

14 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

1
T2 : x → (x + 2). (2.2)
3

The transformation T1 transforms 0 into 0, and 1 into 13 , while transforma-

tion T2 transforms 0 into 23 and 1 into 1. T1 squeezes [0, 1] into [0, 1/3], T2
into [2/3, 1]. Then the Cantor Chaos Game goes as follows:

• In order to get a good approximation of the Cantor set start with

a randomly selected x0 between 0 and 1.
• Then throw a coin.
• If head, then apply T1 to x0 , otherwise apply T2 .
• Thus, after the first choice, you obtain x1 .
• Then throw the coin again, check whether what comes is the head
or the tail, and apply again either T1 or T2 to x1 , to get x2 .
• Throw the coin again to obtain x3 . Do it one million and ten thou-
sands times.
• You will get one million and ten thousand real numbers x1 ,
x2 , ..., x1010000 , all in the interval (0, 1).
• Skip the first ten thousand (to get well onto the “attractor”).
• Put the remaining million of real numbers, all between 0 and 1,
into a list L. Divide the interval [0, 1] into N = 1000 subintervals of
equal length. Count the number of elements in L that fall into each
of the subintervals. You will get 1000 numbers n1 , ..., n1000 that add
to the total of one million. Plot these numbers as black bars over
the corresponding subintervals — you will get a histogram. If you
repeat the whole experiment again, you will get a similar plot.

Of course no one is going to throw a coin a million of times. Instead we

use a computer and pseudo-random numbers. There are different random
number generators, differing in their quality (cf. e.g. [Dutang and Wuertz
(2009)]), but for our purpose it does not really matter which one we use.
The random number generator running twice usually produces the same
sequence of numbers. Unless we change its “seed”.
It is good to normalize our numbers for plotting. That is instead of
plotting n1 , ..., n100 we plot the frequencies n1 /s, ..., n100 /s, where s =
n1 + ... + n100 is the total number of points in our list.
If we repeat this experiment several times, with different random num-
bers, each run will produce “almost the same plot”. We will get something
similar to what is plotted in Fig. 2.3. It is an approximation to what is
called “the Cantor measure”.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 15/2

What are Quantum Fractals? 15

N = 1000
0.01

0.008

0.006

0.004

0.002

0.0
0 1/9 2/9 3/9 4/9 5/9 6/9 7/9 8/9 1

Fig. 2.3 Cantor measure through Chaos Game.

In our Chaos Game we can select only those point that fall in the interval
(0, 1/9) and plot only their relative frequencies. This way we “zoom” into
our picture. Yet, as you can see in Fig. 2.4 the pattern is the same. The
small differences are easily explained by the numerical approximations and
by random fluctuations.
What we see on these pictures is, in fact, a “Cantor measure”. The
Cantor set itself is the “support” of this measure — the set on the horizontal
axis over which we have black bars of a nonzero height.
The Cantor set is one of the simplest examples of Iterated Function
Systems (IFS).

2.2 Iterated function systems

There is an important class of fractals that come from applications of ran-

domly selected transformations from a finite system of transformations of
a given set. One example is the Cantor set, discussed above, where the
set is the one-dimensional interval [0, 1]. Another example is probably even
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 16/2

16 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

N = 1000
0.012

0.01

0.008

0.006

0.004

0.002

0.0
0 1/81 2/81 3/81 4/81 5/81 6/81 7/81 8/81 1/9

Fig. 2.4 Cantor measure through the Chaos Game — zoom into the subinterval [0, 1/9].

more famous — the so called ‘Sierpinski triangle’. While in the case of the
Cantor set we had a pair of transformations acting on a one-dimensional
interval, here we have three transformations (or “maps”) acting on the unit
square [0, 1] × [0, 1] in the two-dimensional plane R2 .
The maps are, in the case of Sierpinski triangle, coded in the following
three 3 × 3 real matrices L1 , L2 , L3 :
⎛ ⎞
0.5 0 0
L1 = ⎝ 0 0.5 0⎠ ,
0 0 1
⎛ ⎞
0.5 0 0.5
L2 = ⎝ 0 0.5 0 ⎠ , (2.3)
0 0 1
⎛ ⎞
0.5 0 0.25
L3 = ⎝ 0 0.5 0.5 ⎠
0 0 1
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 17/2

What are Quantum Fractals? 17

The first two rows and columns of these matrices define a uniform x, y
contraction by the factor 0.5, — the same contraction for each of the three
maps. The third column describes translation — different for each of the
three transformations.
We act with each matrix on the column vector (x, y, 1), to get (x , y  , 1),
using the standard matrix-vector multiplication. Then we act on the re-
sult with another randomly selected matrix. Repeating this procedure say,
100,000 times and plotting the points, we obtain the fractal image shown
in Fig. 2.5.

Fig. 2.5 Sierpinski triangle.

It is important to notice that the three operations L1 , L2 , L3 do not com-

mute. We can estimate the non-commutativity by calculating the matrix
norms of the “commutators”, as they are usually defined in group theory,
Li Lj Li −1 Lj −1 . For i = j we obtain the values between 1.13 and 1.15.
Another well know example of a fractal resulting from an IFS, this time
in three dimension, is the Menger-Sierpinski sponge. It is generated by 20
transformations. All of them have the same uniform contraction factor 1/3,
only translations differ. The twenty translations are given by the vertical
columns of the following matrix
⎛ ⎞
0 0 0 0 0 0 0 0 13 13 13 31 23 32 23 23 32 23 32 23
⎜ ⎟
⎜0 0 0 1 1 2 2 2 0 0 2 2 0 0 0 1 1 2 2 2 ⎟ (2.4)
⎝ 3 3 3 3 3 3 3 3 3 3 3 3 ⎠
1 2 2 1 2 2 2 1 2 2 1 2
0 3 3 0 3 0 3 3 0 3 0 3 0 3 3 0 3 0 3 3
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 18/2

18 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.6 1 mln points for Menger-Sierpinski sponge.

The procedure is the same as for the Sierpinski triangle. However the
fractal is living now in 3D, therefore its visualization becomes a problem.
If we try to plot 1 mln points, we get a 3D image which, when projected on
the plane, is impossible to interpret — Fig. 2.6. To get an idea what kind
of a fractal we are dealing with it is necessary, in this case, to take the view
of just a thin slice along one of the six faces of the cube, as in Fig. 2.7. Yet
even then it is difficult to imagine what kind of a 3D object we are dealing
with. The solution, in this case, consists of using a different method for
creating the same 3D fractal, namely start with a solid cube and iteratively
remove the appropriate sub-cubes, as in Fig. 2.8.

2.2.1 Definition of IFS

The Cantor set, Sierpinski triangle, Menger-Sierpinski sponge, like many
other fractals, live within a bounded region of an Euclidean space. But
fractals can also live on curved surfaces, for instance on the surface of the
sphere, as it is the case of typical quantum fractals.
Therefore it is convenient to give a definition of an iterated function
system that is general enough to include most of the cases that have been
studied. The proper arena for such a definition is a “complete metric space”
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 19/2

What are Quantum Fractals? 19

Fig. 2.7 Top view Menger-Sierpinski sponge IFS.

(for a mathematical definition see Appendix Sec. A.1)

Quantum fractals that we will consider are defined by surjective (i.e.
“onto”) maps and they live on spheres which are bounded with respect to
their natural metrics. These maps are never everywhere contractive.
We will now review, for the Reader’s convenience, the main definitions
and properties (without proofs) used in the discussion of iterated function
systems. Let us start with the most general settings.
Definition 2.1 (Iterated function system). Let (X, Σ) be a measur-
able space (cf. Appendix Sec. A.3), and let W = {w1 , ..., wn } be a finite set
of measurable transformations of X. Then W is called an iterated function
system on X. If, in addition, there is given a system p = {pi , i = 1, ..., n}
of measurable functions
pi : X → [0, 1], i = 1, .., n (2.5)
satisfying

n
pi (x) = 1 a.e. (2.6)
i=1

then (W, p) is called an iterated function system, or IFS, with place-

dependent probabilities.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 20/2

20 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.8 3D solid view of Menger-Sierpinski sponge.

If (X, d) is a complete metric space, and W = {w1 , ..., wn } is a finite set of

contractions (c.f Appendix (Sec. A.4)) on X, then the system W is called
a (contractive) hyperbolic iterated function system.

Thus, for instance, the two maps Eq. (2.2) defining the Cantor set form
a hyperbolic IFS. Originally hyperbolic iterated function systems (for
comprehensive reviews see e.g. Refs. [Barnsely (1988); Peitgen et al.
(1992)]) consisted of a collection of affine maps acting on the plane R2 .
An affine map is a map of the form x = Ax + b:

x
1
x1 a11 a12 x1 b1
=φ = + . (2.7)
x
2
x2 a21 a22 x2 b2
It is only the matrix A that determines the contraction ratio. The trans-
lation by vector b does not affect this ratio. The contraction ratio k can
be computed as the square root of the highest eigenvalue of the symmetric
matrix AT A. Explicitly:

k = α + β + (α − β)2 + γ 2 , (2.8)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 21/2

What are Quantum Fractals? 21

where

α = (a211 + a221 )/2, β = (a211 + a222 )/2, γ = a11 a12 + a21 a22 . (2.9)

Each hyperbolic IFS has an attractor set. In the case of the Cantor IFS
the attractor is the Cantor set. A general definition of the attractor for a
hyperbolic IFS requires a little bit of a preparation.

Definition 2.2. Let W = {w1 , ..., wn } be a collection of maps defined on a

set X. Then (by the abuse of notation) W defines the map W : A → W(A)
on the set of all subsets of X defined as follows:

W(A) = w1 (A) ∪ ... ∪ wn (A). (2.10)

That is W(A) is the union of all images wi (A) of A under the maps wi . We
then have the following theorem:

Theorem 2.1. If W is a hyperbolic IFS on a complete metric space (X, d),

then there is a unique compact subset A of X with the property:

W(A) = A. (2.11)

The set A is then called the attractor of W.

Usually, in order to visualize the attractor, we start with the whole space
A0 = X, and construct iteratively the sequence Ak+1 = W(Ak ). Then,
as k → ∞, the sets Ak approximate the attractor better and better. In
order to define “better and better” the concept of the Hausdorff distance
is needed.

Definition 2.3. Let (X, d) be a metric space, and let H(X) be the col-
lection of all nonempty closed, bounded subsets of X. For any nonempty
set A ⊂ X we denote by A the union of all open balls with radius  and
centers in A.
For A, B ∈ H(X) we define

dH (A, B) = inf{ : A ⊂ B and B ⊂ A }. (2.12)

Then dH is called the Hausdorff distance on H(X).

Theorem 2.2. The Hausdorff distance dH is a metric on H(X). If (X, d)

is complete (resp. totally bounded, resp. compact) metric space, then
(H(X), dH ) is also complete (resp. totally bounded, resp. compact).
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 22/2

22 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

The sketch of the proof can be found in [Munkres (2000)]. For a hyperbolic
IFS we have:
A = lim W k (X), (2.13)
k→∞

where the limit is taken with respect to the Hausdorff distance dH

Definition 2.4. If W is an IFS, W = {w1 , ..., wn }, and P = {p1 , ..., pn }

is a sequence of positive numbers pi < 1, i = 1, .., n, p1 + ...pn = 1, then
(W, P ) is an IFS with (constant) probabilities.

2.2.2 Frobenius-Perron operator

Given an IFS with constant probabilities, we can define action of this system
not only on subsets of X, but also on measures on X.

Definition 2.5. Let (W, P ) be an IFS with constant probabilities. For

every measure μ on X, and for every subset A ⊂ X let F μ be the set
function defined by the formula:

n
(F μ)(A) = pi μ(wi−1 (A)). (2.14)
i=1

The operator F is called the Frobenius-Perron operator associated with

(W, P ).

Remark 2.1. Notice that transformations wi in Eq. (2.14) do not have to

be invertible. wi−1 (A) there denotes the counterimage of A:
wi−1 (A) = {x ∈ A : wi (x) ∈ A}.
On the other hand wi must be invertible in Eq. (2.17) below, where densities
are involved instead of measures.

Of course the standard condition of measurability must be satisfied in order

for the above formula to make sense. In application such conditions are
usually automatically satisfied.

Remark 2.2. The Frobenius-Perron operator defined above may be con-

sidered as a particular case of the Frobenius-Perron operator for one trans-
formation, by using an appropriate construction on product spaces.

For a general IFS with place-dependent probabilities we may first define

the Koopman operator P ∗ in analogy to Eq. (A.38) on p. 310:
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 23/2

What are Quantum Fractals? 23


n
(P ∗ f )(x) = pi (x)f (wi (x)). (2.15)
i=1

It is dual to the Frobenius-Perron operator F defined in Eq. (2.14):

 
∗
(P f )(x) dμ(x) = f (x) dF μ(x). (2.16)
X X

If μ is a fixed measure on X, and if the transformations wi are invertible,

then P acts on densities of measures (with respect to μ) according to the
formula that is a slight generalization of the formula Eq. (A.36) in the
Appendix, p. 309.
 
m
 −1   −1  dμ wi−1 (x)
(P f )(x) = pi wi (x) f wi (x) . (2.17)
i=1
dμ(x)

2.3 Cantor set through matrix eigenvector

We have defined the Cantor set through an iterated function system con-
sisting of two transformations that are selected with equal probabilities.
These transformations, let us call them here w1 , w2 :
1
w1 (x) = x, (2.18)
3
1 2
w2 (x) = x+ . (2.19)
3 3
act on points of the interval [0, 1]. Whenever we have a transformation
acting on points, it induces transformation of functions, we denote it by
T ∗:

(T ∗ f )(x) = f (T (x)). (2.20)

When we have two transformations, T1 and T2 , that are selected with prob-
abilities p1 and p2 , the resulting function is also weighted with probabilities.
This way we arrive at what is called the Koopman operator T associated
with the IFS:

(T ∗ f )(x) = p1 f (T1 (x)) + p2 f (T2 (x)). (2.21)

Dual to the space of functions is the space of bounded measures. Given a

measure μ, we can associate with each function (that is ‘measurable’ etc.)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 24/2

24 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality


the number (μ, f ) = f dμ. We define the dual operator T∗ on measures
by the formula:

(T∗ μ, f ) = (μ, T ∗ f ). (2.22)

Then T∗ is called the Frobenius-Perron operator associated with the IFS.

We are looking for a probabilistic measure that is invariant with respect
to T∗ . In most cases studied in the literature one can prove that such a
measure exists and is unique. For the Cantor system we can calculate and
graphically represent an approximation for this invariant measure. To this
end we discretize the interval [0, 1] into, say, N = 1000 small intervals Δi =
[(i−1)∗Δ, i∗Δ], i = 1, ..., N, Δ = 1.0/N. In the space of functions we chose
an orthogonal basis ei (x) = χΔi (x), where χΔ denotes the characteristic
function of the set Δ ⊂ [0, 1]: χΔ (x) = 1 for x ∈ Δ, otherwise χΔ (x) = 0..
Then we approximate the operator T ∗ by a finite dimensional. Namely, we
want to decompose T ∗ ei , projected onto the subspace generated by ei , into
ej :

T ∗ ei = Tji∗ ej . (2.23)
j

In order to calculate the matrix coefficients Tji∗ , we take scalar products (in
L2 ) of the above formula with ek .

(ek , T ∗ ei ) = Tji∗ (ek , ej ). (2.24)
j

The functions ei are orthogonal and (ek , ej ) = Δδk,j , where δkj is the
Kronecker delta. This way we get the formula:

∗ 1 1 x x+2
Tki = (ek , T ∗ ei ) = χΔi ( ) + χΔi ( ) dx, (2.25)
2Δ 2Δ Δk 3 3
where the factor 2 in the denominator comes from the fact that each of the
two transformations is selected with the probability 1/2.
Now, χΔ ( x3 ) = χ3Δ (x), and χΔ ( x+2
3 ) = χ3Δ−2 (x). Therefore we obtain
the following formula:
1  
∗
Tki = |3Δi ∩ Δk | + |(3Δi − 2) ∩ Δk | , (2.26)
2Δ
where we denote by |.| the length of the corresponding interval.

Notice that owing to the fact that i Δi is the whole interval [0, 1], we
 ∗ ∗
get i Tki = 1. Thus the sum of elements in every row of the matrix Tki
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 25/2

What are Quantum Fractals? 25

is one, therefore it has an eigenvector belonging to the eigenvalue 1 (the

vector with all its components equal 1). But a matrix and a transposed
matrix have the same eigenvalues. The Frobenius-Perron operator is dual
to T ∗ , therefore it is represented by the transposed matrix, let us call it T :
Tik = Tki . It’s eigenvector to the eigenvalue one is exactly the eigenvector
we are looking for — our approximation to the invariant measure.
Given two intervals (a1 , b1 ) and (c1 , d1 ) we have the following formula
for the length of their intersection:

|(a1 , b1 ) ∩ (a2 , b2 )| = max (0, min(b1 , b2 ) − max(a1 , a2 )) . (2.27)

Applying to our case we obtain:

1
Tik = max(0, min(3i, k) − max(3i − 3, k − 1))
2
1
+ max(0, min(3i − 2N, k) − max(3i − 3 − 2N, k − 1)). (2.28)
2

The matrix T has a simple band structure, especially regular when N =

3k, where k is an integer. Figure 2.9 is a graphical representation of this
structure for k = 18. We have there either 0 or 12 . The problem of finding the
invariant eigenvector of such a matrix can be solved exactly. For N = 1000
the solution can be found numerically — it is represented on Fig. 2.10.

2.4 Quantum iterated function systems

Quantum iterated function systems which result in quantum fractals consist

of special kind of transformations of the unit sphere in three-dimensional
Euclidean space — so called Möbius transformations. Möbius transforma-
tions are special transformations of the sphere — in particular they map
circles into circles and preserve angles between circles, while the centers and
radii of the circles in general change. The simplest way to encode them is
via 2 × 2 complex matrices of determinant one. An example of the action
of a Möbius transformation is shown in Fig. 2.11.
Quantum Iterated Function System (QIFS) is a special kind of an IFS
consisting of a finite number of Möbius transformations acting on the two-
dimensional sphere. Standard QIFS (SQIFS) are additionally endowed with
a specific formula defining the place-dependent probabilities. Each Möbius
transformation w is determined by a two-by-two complex matrix A of
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 26/2

26 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

1 20 40 54
1 1

20 20

40 40

54 54
1 20 40 54

Fig. 2.9 Band structure of the Frobenius-Perron matrix for the Cantor set for N = 54.

determinant one, with w determining A uniquely up to sign. An exam-

ple of how this happens is discussed in Sec. 2.5, while a general theory is
described in detail in Sec. 2.7.
Let {wα , (α = 1, ..., n)}, be a finite set of Möbius transformations acting
on the sphere, and let Aα , det Aα = 1, be a set of 2×2 matrices determining
the wα -s. The condition that needs to be satisfied in order for {wα } to define
n
a QIFS is that the sum α=1 A∗α Aα must be proportional to the identity
matrix I:

n
A∗α Aα = λI, λ > 0. (Balancing Condition) (2.29)
α=1
n ∗
In general the sum Λ = α=1 Aα Aα is just a positive definite matrix.
As such it has two positive eigenvalues λ1 , λ2 . The balancing condition is
equivalent to the fact that both eigenvalues are equal.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 27/2

What are Quantum Fractals? 27

0.010
0.008
0.006
0.004
0.002
0.000

0 1/9 2/9 1/3 4/9 5/9 2/3 7/9 8/9 1

Fig. 2.10 Frobenius-Perron matrix for the Cantor set for N = 1000.

The requirement that this sum is proportional to the identity matrix is

a strong requirement. It is called “balancing condition” for the following
reason: In the context of quantum measurement theory each Aα determines
a quantum jump — a discontinuous change of state of the quantum system
that is being continuously monitored. Between jumps the quantum state
evolves according to a modified, in general non-unitary Schrödinger equa-
tion. The balancing condition assures then that the standard Schrödinger
evolution of quantum states between jumps is unmodified, that is that be-
tween jumps we have a unitary evolution of the standard quantum theory
as it is described in all textbooks. In other words, if in Nature there would
be only quantum jumps, the Chaos Game producing a trajectory of a QIFS
would describe one possible history of an individual quantum system (for
instance of the whole Universe).
An SQIFS must additionally satisfy the condition that the place de-
pendent probabilities are given by a particular formula — they are not
arbitrary, but they are completely determined by the transformations them-
selves. If we denote by μ the standard rotation invariant measure on the
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 28/2

28 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.11 Action of a Möbius transformation on the sphere. Circles are mapped into
circles, angles between circles are preserved. This particular Möbius transformation is
encoded in the matrix ( 10 11 ).

sphere, and if we denote by J(x) the change of the infinitesimal surface

area at x, x2 = 1, resulting from the application of a transformation w,
then the formula for the place dependent probabilities of SQIFS reads:

pα (x) = cJα (x)−1/2 , (2.30)

where c is a positive constant assuring the condition that


n
pi (x) = 1. (2.31)
α=1

That such a constant exists is guaranteed by the balancing condition. The

place dependent probabilities pα (x) can be computed in two ways. The
first way is in terms of the matrices Aα , the second way in terms of Lorentz
matrices L(α)μν determined by Aα . The sphere can be interpreted as the
Bloch sphere — the sphere of spin 1/2 directions. If ξ is a unit vector in
the two-dimensional complex space C2 representing the spin direction x,
the probabilities pα (x) are given by the formula

||Aα ξ||2 ||Aα ξ||2

pα (x) = n = , (2.32)
β=1 ||Aβ ||
2 λ

where λ is given by Eq. (2.29). In terms of the Lorentz matrices L(α)

determined by Aα , as in Proposition 2.2, p. 53, the formula for probabilities
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 29/2

What are Quantum Fractals? 29

takes the form2

L(α)4i xi + L(α)44 L(α)4i xi + L(α)44
pα (x) =  4 i 4 = . (2.33)
β (L(β) i x + L(β) 4 ) λ

Remark 2.3. Each point x0 on the sphere can be represented by the point
(x, 1) on the light cone x2 − (x4 )2 = 0, in Minkowski space, with its time
coordinate x4 normalized to x4 = 1. The probabilities given by Eq. (2.33)
are thus proportional to the time component of the transformed point on
the light cone. Within the framework of Lorentz transformations such a
condition may seem to be counterintuitive. But, as it is explained further in
this book, such a formula for place-dependent probabilities is required by a
quantum measurement theory that is able to reproduce the standard Born’s
interpretation of the quantum-mechanical wave function. Any departure
from this formula, and one such departure is discussed in Sec. 4.7, would
lead to a version of a nonlinear quantum mechanics with all its attractions
and all its dangers.

In the next section we will see one example of a quantum fractal “in
action”.

2.5 Example: The “impossible” quantum fractal

Probably the simplest quantum fractal is the one that is similar to the
Cantor set, except that it lives on the unit circle instead of the unit interval.
We will discuss it later, in Sec. 3.1.1. But in order to get a real taste of
what quantum fractals are about, it is useful to start with something that is
more illustrative, even though more complicated. This way we will acquaint
ourselves with the whole machinery that will be described in detail in the
following sections.
Our first quantum fractal will be created by an iterated function sys-
tem consisting of 24 transformations, and it will live on a two-dimensional
sphere. It will satisfy the Balancing Condition Eq. (2.29), and its place
dependent probabilities will given by Eq. (2.32).
The two-dimensional unit sphere can be projected stereographically on
the (complex) plane. We will first describe our construction on the plane,
then realize it on the sphere, and then stereographically project from the
sphere onto the plane.
2 In Eq. (2.33), and in the rest of this book, we use Einstein’s convention of a summation

over a repeating index — cf. Remark 2.4 p. 44.

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 30/2

30 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

We will do it in this order for the following reason: the plane is infinite;
iterations of Möbius transformations acting on the plane occasionally pro-
duce very large numbers (they may even produce infinities!), and it may
lead to unnecessary computational inaccuracies. These numerical inaccu-
racies and instabilities can be avoided if iterations are taking place on the
sphere — which constitutes the natural arena on which quantum fractals
live.

2.5.1 24 symmetries — the octahedral group

Our QIFS will consists of 24 transformations. Why 24? Why not a smaller
number? This is because we want this particular fractal to be symmetric.
More precisely: we want it to have the symmetry group of the cube. The
symmetry group of the cube has 24 elements. Indeed, the cube has 6 faces.
When one of these six faces is facing up, there are 4 faces possible that
can face us — when we are looking from the side. Therefore 6 × 4 = 24
symmetries. We will represent this group by 24 two-by-two complex unitary
matrices, all of them having determinant 1, as follows:

1, 0 1 1, −i 1 1, i 0, i
U1 = , U2 = √ , U3 = √ , U4 = ,
0, 1 2 −i, 1 2 i, 1 i, 0

1 0, 1 − i 1 1 − i, 1 − i
U5 = √ , U6 = ,
2 −1 − i, 0 2 −1 − i, 1 + i

1 1 − i, −1 + i 1 1 − i, 0
U7 = , U8 = √ ,
2 1 + i, 1 + i 2 0, 1 + i

1 1 + i, −1 + i 1 1, −1
U9 = , U10 = √ ,
2 1 + i, 1 − i 2 1, 1

1 1 + i, 1 − i 1 1, 1
U11 = , U12 = √ ,
2 −1 − i, 1 − i 2 −1, 1

1 i, −i 1 1 − i, 1 + i 1 i, i
U13 = √ , U14 = , U15 = √ ,
2 −i, −i 2 −1 + i, 1 + i 2 i, −i

1 1 − i, −1 − i 1 1 + i, 0
U16 = , U17 = √ ,
2 1 − i, 1 + i 2 0, 1 − i

1 1 + i, 1 + i 1 1 + i, −1 − i
U18 = , U19 = ,
2 −1 + i, 1 − i 2 1 − i, 1 − i
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 31/2

What are Quantum Fractals? 31

1 0, 1 + i 0, 1
U20 = √ , U21 = ,
2 −1 + i, 0 −1, 0

1 i, 1 1 i, −1 i, 0
U22 = √ , U23 = √ , U24 = .
2 −1, −i 2 1, −i 0, −i
These 24 matrices can be obtained in the following way (cf. [Springer
(1977); Stekolshchik (2008)]): we start with a set G0 of three matrices
G0 = {g1 , g2 , g3 } defined by
 0 0i 1 −1 −1
g1 = , g2 = , g3 = √ , (2.34)
0 −1 i0 2 − 

where  = eπi/4 . We generate the group of 48 elements from this set, for
instance by first taking their products, G1 = G0 · G0 , then G2 = G1 · G1 ,
then G3 = G2 · G2 . This way we obtain 48 different elements that form
the binary octahedral group. The group elements occur in pairs ±g. We
remove one element from each such pair. Our 24 matrices Uα were obtained
in such a way.
All these 24 matrices are unitary of determinant one, they belong to
the group SU (2). SU (2) — the special unitary group in two dimensions,
is the double covering group of the group SO(3) — the (proper, i.e. of
determinant +1) rotation group in three dimensions.3
The matrices Uα form a projective representation of the symmetry group
of the cube in the following sense: for every pair α, β = 1, .., 24, there exists
a unique γ such that Uα Uβ = ±Uγ . We could use the whole group of 48
elements, but this is not needed for our purpose because matrices Uα and
−Uα define the same Möbius transformations of the sphere. The extra
24 elements would differ only by ±1 from Uα , and this sign would not
contribute to the construction of the IFS.

2.5.2 Construction of the 24-elements SQIFS

Let us construct our IFS. We start with the matrix A1 defined as follows:
14
A1 = . (2.35)
01
We have chosen this particular matrix for several reasons. My first reason
is that, as you will see, it produces a nice quantum fractal with several
interesting properties. The second reason is that, as mentioned in Chap. 1,
3 SU (2) is also the group of unit quaternions.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 32/2

32 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

according to the theorem proved by Andrew Vince [Vince (2013)], it can

not produce a fractal — because it has trace equal 2, therefore it defines
a parabolic Möbius transformation (cf. Sec. 2.7.10).4 All our 24 matrices,
being unitarily equivalent to A1 will be also parabolic. Therefore our exam-
ple creates a certain puzzle — that is why we used the word “impossible” in
the title of this section. There is something in it that is asking for further
research.
The matrix A1 is of determinant 1, it belongs to the group SL(2, C)
— the special linear group in two complex dimensions. SL(2, C) is the
double covering group of the restricted Lorentz group of Special Relativ-
ity Theory. This fact will be of relevance when we will (alternatively, as
opposed to quantum spin measurements) interpret our transformations via
the relativistic light aberration in Sec. 2.7.12.
Using the SU (2) symmetry matrices Uα we construct now 24 matrices
Aα by 3D rotating the matrix A1 :
Aα = Uα A1 Uα ∗ , α = 1, ..., 24. (2.36)
Explicitly:
 1, 4     
1+2i, 2 1−2i, 2
A1 = 0, 1 , A2 = 2, 1−2i , A3 = 2, 1+2i ,
 1, 0     
1, 0 3, −2i
A4 = 4, 1 , A5 = −4i, 1 , A6 = −2i, −1 ,
     
−1, −2i 1, −4i 1−2i, 2i
A7 = −2i, 3 , A8 = 0, 1 , A9 = −2i, 1+2i ,
 −1, 2     
1+2i, 2i 3, 2
A10 = −2, 3 , A11 = −2i, 1−2i , A12 = −2, −1 , (2.37)
    3, −2 
−2 
A13 = −1, 2, 3 , A14 = 1−2i, −2i
2i, 1+2i , A15 = 2, −1 ,
     
−2i
A16 = 1+2i, 2i, 1−2i , A17 = 1, 4i
0, 1 , A18 = 3, 2i
2i, −1 ,
     1, 0 
A19 = −1, 2i
2i, 3 , A20 = 1, 0
4i, 1 , A21 = −4, 1 ,
     1, −4 
−2 1−2i, −2
A22 = 1+2i,−2, 1−2i , A23 = −2, 1+2i , A24 = 0, 1 .

Our notation is consistent owing to the fact that U1 is the identity matrix.
Now all matrices Aα are all in SL(2, C). They all have trace 2 — they define
parabolic transformations.
It may be of interest to notice that the matrices Aα are constructed
in such a way that (Aα )−1 = A25−α . This property is not at all evident
4 From the Abelian group of parabolic transformations discussed in more detail in

Sec. 2.7.13.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 33/2

What are Quantum Fractals? 33

from the construction, but it can be easily verified by a direct matrix mul-
tiplication. Therefore, our set of 24 matrices is closed with respect to the
operation of taking of the inverse.
This fact5 is interesting, but it is not really important. What is im-
portant is that:

24
A∗α Aα = 216 I. (2.38)
α=1

Therefore the balancing condition Eq. (2.29) is satisfied.

We can now calculate probabilities given by Eq. (2.32) explicitly. To this
end we parametrize quantum spin states, norm 1 vectors in C2 , (excluding
the one corresponding to ξ = ( 10 )) by complex numbers z = x + iy:

1 x + iy
ξ= , (2.39)
1 + x2 + y 2 1

to obtain:
1 qα (x, y)
pα (x, y) = + , (2.40)
24 54(1 + x2 + y 2 )
where qα (x, y) are given by the following sequence of 24 quadratic polyno-
mials in x, y:
   
2 1 + x − x2 − y 2 , 2(x − 2y), 2(x + 2y), 2 x2 + x + y 2 − 1 ,
 
2 x2 + y 2 + y − 1 , x2 + y 2 − 4y − 1, −x2 − y 2 + 4y + 1,
 
2 1 − x2 − y 2 − y , 2y − 4x, −x2 − 4x − y 2 + 1,
2(2x + y), x2 + 4x + y 2 − 1, −x2 + 4x − y 2 + 1,
 
4x − 2y, x2 − 4x + y 2 − 1, −2(2x + y), 2 1 − x2 − y 2 + y ,
 
x2 + y 2 + 4y − 1, −x2 − y 2 − 4y + 1, 2 x2 + y 2 − y − 1 ,
   
2 x2 − x + y 2 − 1 , 4y − 2x, −2(x + 2y), −2 x2 + x + y 2 − 1 .
24 24
It can be verified that α=1 qα (x, y) = 0, therefore α=1 pα (x, y) = 1.
The minima and maxima of the probability functions pα (x, y) form a
regular pattern at the points whose coordinates are rational functions of the
Golden Ratio ϕ — cf. Fig. 2.12. A typical behavior of one such probability
function is shown in Fig. 2.13.
5 As well as the fact that the entries of our matrices are Gaussian integers — complex

numbers with integer real and imaginary coefficients. Therefore our quantum IFS can
be used as a generator of Pythagorean quadruples — cf. Sec. 2.7.14.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 34/2

34 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

2 (0, ϕ)

1
(−ϕ, 0) (ϕ, 0)
0

−1

−2 (0, −ϕ)

−3

−4

−5
−5 −4 −3 −2 −1 0 1 2 3 4 5

Fig. 2.12 Maxima and minima √ of the 24 probability functions on the complex plane. ϕ
is the Golden Ratio ϕ = (1 + 5)/2 = 1.618...

For several reasons it is more convenient to define QIFS acting on the

sphere and consisting of 24 particular transformations L(α), α = 1, ..., 24,
from the restricted Lorentz group. They can be calculated from the formula
in Proposition 2.2, on p. 53.

1 0 −4 4 1 −4 0 4 1 4 0 4
0 1 0 0 4 −7 0 8 −4 −7 0 −8
L(1) = 4 0 −7 8 , L(2) = 0 0 1 0
, L(3) = 0 0 1 0
,
4 0 −8 9 4 −8 0 9 4 8 0 9

1 0 4 4 1 0 0 0 1 0 0 0
0 1 0 0 0 1 4 4 0 −7 4 8
L(4) = −4 0 −7 −8 , L(5) = 0 −4 −7 −8 , L(6) = 0 −4 1 4 ,
4 0 8 9 0 4 8 9 0 −8 4 9

1 0 0 0 1 0 0 0 −7 4 0 8
0 −7 4 −8 0 1 4 −4 −4 1 0 4
L(7) = 0 −4 1 −4 , L(8) = 0 −4 −7 8 , L(9) = 0 0 1 0
,
0 8 −4 9 0 −4 −8 9 −8 4 0 9

−7 0 −4 8 −7 −4 0 −8 −7 0 −4 −8
0 1 0 0 4 1 0 4 0 1 0 0
L(10) = 4 0 1 −4 , L(11) = 0 0 1 0
, L(12) = 4 0 1 4
,
−8 0 −4 9 8 4 0 9 8 0 4 9

−7 0 4 −8 −7 4 0 −8 −7 0 4 8
0 1 0 0 −4 1 0 −4 0 1 0 0
L(13) = −4 0 1 −4 , L(14) = 0 0 1 0
, L(15) = −4 0 1 4 ,
8 0 −4 9 8 −4 0 9 −8 0 4 9
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 35/2

What are Quantum Fractals? 35

Fig. 2.13 Graph of the probability function p(x, y, 9) near its min-max points.

−7 −4 0 8 1 0 0 0 1 0 0 0
4 1 0 −4 0 1 −4 4 0 −7 −4 −8
L(16) = 0 0 1 0
, L(17) = 0 4 −7 8 , L(18) = 0 4 1 4
,
−8 −4 0 9 0 4 −8 9 0 8 4 9

1 0 0 0 1 0 0 0 1 0 −4 −4
0 −7 −4 8 0 1 −4 −4 0 1 0 0
L(19) = 0 4 1 −4 , L(20) = 0 4 −7 −8 , L(21) = 4 0 −7 −8 ,
0 −8 −4 9 0 −4 8 9 −4 0 8 9

1 −4 0 −4 1 4 0 −4 1 0 4 −4
4 −7 0 −8 −4 −7 0 8 0 1 0 0
L(22) = 0 0 1 0
, L(23) = 0 0 1 0
, L(24) = −4 0 −7 8 .
−4 8 0 9 −4 −8 0 9 −4 0 −8 9

Each of the matrices L(α) has exactly one fixed point on the sphere. For
instance L(1) has one fixed point — the north pole of the sphere. Fig. 2.14
shows the action of L(1) on the sphere. Other transformations are all
similar (related by a 3D rotation) to L(1).
The probabilities pα , in terms of the coordinates x, y, z of the points on
the unit sphere x2 + y 2 + z 2 = 1, can now be calculated from Eq. (2.33).
They are of the form:
1 qα (x, y, z)
pα (x, y, z) = + , (2.41)
24 54
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 36/2

36 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.14 Action of L(1) on the sphere. The shading shows the Jacobian: the level of
the compression of a given infinitesimal area. L(1) maps the highly distorted region on
the left to the corresponding image of the distorted region on the right.

where qα are given by:

q1 = (x − 2 ∗ z), q2 = (x − 2 ∗ y), q3 = (x + 2 ∗ y),
q4 = (x + 2 ∗ z), q5 = (y + 2 ∗ z), q6 = (−2 ∗ y + z),
q7 = (2 ∗ y − z), q8 = (−y − 2 ∗ z), q9 = (−2 ∗ x + y),
q10 = (−2 ∗ x − z), q11 = (2 ∗ x + y), q12 = (2 ∗ x + z),
q13 = (2 ∗ x − z), q14 = (2 ∗ x − y), q15 = (−2 ∗ x + z), (2.42)
q16 = (−2 ∗ x − y), q17 = (y − 2 ∗ z), q18 = (2 ∗ y + z),
q19 = (−2 ∗ y − z), q20 = (−y + 2 ∗ z), q21 = (−x + 2 ∗ z),
q22 = (−x + 2 ∗ y), q23 = (−x − 2 ∗ y), q24 = (−x − 2 ∗ z).
These probabilities tend to increase the contrast of the resulting “Chaos
Game” image. In Fig. 2.15 we can see the resulting pattern of 1,000,000,000
iterations projected vertically, along the z-axis from the sphere, but with
uniform probabilities, each equal to 1/24. The dark areas correspond to
those regions on the sphere where practically no hits have been registered.
Figure 2.16 shows the stereographic projection of the lower hemisphere.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 37/2

What are Quantum Fractals? 37

Fig. 2.15 Parabolic quantum fractal made of QIFS of 24 Möbius transformations with
dodecahedral symmetry. 1,000,000,000 points on the sphere projected vertically on the
plane. The square [−1, 1] × [−1, 1] has been divided into 1200 × 1200 square cells and the
hits has been counted for each of the cells. The color bar displays log10 (n + 1), where n
is the number of hits in a given cell.

The pattern has an evident translational symmetry with the period 4.0. It
can be seen from Fig. 2.17 and Fig. 2.18, that the correspondence between
the regions in both figures, one on the sphere and one in the plane, is not
that evident: the circles 1, 2, 3, 4 after stereographic projection become of
the same size as circles A, B, C, D, the reason being that they are closer to
the projection point — the north pole of the sphere.

2.5.3 Open problems

2.5.3.1 Is it a fractal?
The first question that needs to be asked is whether the resulting pattern
is really a fractal, and if so, in which sense? There are circular regions that
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 38/2

38 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.16 Stereographic projection.

are being avoided. But do we have an attractor here that is not the whole
plane?

2.5.3.2 What are these circles?

The large circles on the complex plane seem to have the same radius. Well,
they should have the same radius because our 24 transformations contain
translations by 4. But what is this radius, if there such a thing as a sharp
radius there?
√ From the picture, by measurement, the radius seems to be
close to 3. √If this is the case, then the next smaller circles perhaps would
have radius 3/5, and center at 2/5 + 2i/5, since this is what √ the unitary
transformation U3 does when acting on the circle with radius 3 and center
at 2+2i. But can one find an explicit expression for all these circles, big and
small? Figure 2.19 shows the result of applying all sequences of 1, 2, 3, 4, 5
transformations (words of at most five letters from our list of 24) to the
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 39/2

What are Quantum Fractals? 39

Fig. 2.17 Vertical projection.

point 2 + 2i. This results in 5063328 points (coinciding points are counted
as one) shown in the picture. Visually — see Fig. 2.19 — there is a strong
similarity with a part of the result of the Chaos Game. Additionally, two
exact circles have been superimposed in Fig. 2.19 in order to check how
exact is the above guess. While the larger circle fits pretty well, it is not
quite the same with the small one.

2.5.3.3 Can we find a formula for the frequencies of pairs?

The first two questions are essentially independent of the definition of place
dependent probabilities. But suppose we use the formulas Eqs. (2.41) and
(2.42). Then while the frequency of selecting any particular transformation
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 40/2

40 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.18 Stereographic projection.

in a long sequence of our “Chaos Game” is the same and equal to 1/24,
the frequencies of consecutive pairs are not the same, cf. Fig. 2.20, where
the lighter squares correspond to higher frequencies. It can be seen from
this picture that repeating the same transformation twice (the diagonal)
happens rarely. This is probably due to the fact that each of the transfor-
mations contains a unitary rotation along an axis that is orthogonal to the
contraction-expansion direction. But can this particular pattern be some-
how understood by examining the set of 24 matrices and nothing else? To
test the hypothesis Fig. 2.21 shows the density plot of the distance between
Lorentz matrices L(i) and L(j) measured by the function

d(i, j) = ||(L(i) − L(j))t (L(i) − L(j))||2 . (2.43)

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 41/2

What are Quantum Fractals? 41

Fig. 2.19
√ Five letter words applied to 2 + 2i with superimposed
√ two circles: one of
radius 3 and center at 2 + 2i, the other one with radius 3/5 and center 2/5 + 2i/5.
Plot range x, y from −14 to 14.

Figure 2.21 is, of course, perfectly symmetric, which is not the case with
Fig. 2.20, but otherwise the similarity is rather striking. How to understand
this phenomenon? In principle the statistics of pairs can be deduced from
the formula (12) in [Jadczyk (2006)], but this is yet to be done.

2.6 Action on the plane

a b
Every nonsingular 2 × 2 complex matrix A = c d acts on the complex
plane by linear fractional transformation:
az + b
A : z → A · z = . (2.44)
cz + d
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 42/2

42 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

20

15

10

5 10 15 20

Fig. 2.20 Frequencies of pairs (i, j), where the transformation Ai is followed by Aj
using place dependent probabilities Eqs. (2.41) and (2.42). Darker squares correspond
to smaller frequency. The counting was performed on a sample 100,000 iterations long.

This action can be derived form the natural action on complex vectors using
the projective method as follows:
ab z1 az1 + bz2
= . (2.45)
cd z2 cz1 + dz2
Then, with z1 = z, z2 = 1 we obtain:
az+b
ab z cz+d
= (cz + d) . (2.46)
cd 1 1
The action in Eq. (2.44) is singular for z = −d/c. This point is mapped into
infinity. That is why it is convenient to add the point ∞ to the complex
plane, to compactify it, to make it into the Riemann sphere. But this is not
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 43/2

What are Quantum Fractals? 43

20

15

10

5 10 15 20

Fig. 2.21 Distance d(i, j) between Lorentz matrices L(i) and L(j).

the optimal solution. A much better solution is to act on the sphere from
the very beginning.
Another disadvantage of the complex plane representation is that ma-
trices A and cA, 0 = c ∈ C, define the same transformation. We have a
redundancy in our description. It is much better to get rid of these two
problems from the very beginning. In the next section we will show how
this can be done by replacing 2 × 2 complex matrices by 4 × 4 real ones.

2.7 Lorentz group, SL(2, C), and relativistic aberration

2.7.1 The Lorentz group

In Special Relativity space-time is modeled by the Minkowski space M. We
have four coordinates xμ , μ = 1, 2, 3, 4, or (x, x4 ), where x = (x1 , x2 , x3 )
stands for space, while x4 stands for the time coordinate.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 44/2

44 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

The time coordinate is usually denoted by x0 rather than by x4 .

We will interchangeably use either x0 or x4 .
In some computing environments numbering 1, .., 4 is usually better
than 0, ..., 3 — since arrays may like to start with index = 1.

So, essentially, while as a vector space, M is R4 , the geometry of M is

defined by its fundamental invariant: the quadratic form q(x)

q(x) = (x1 )2 + (x2 )2 + (x3 )2 − (x4 )2 (2.47)

= −(x ) + (x ) + (x ) + (x ) .
0 2 1 2 2 2 3 2

The form q(x) can be also written in a matrix form as

q(x) = xT η x, (2.48)

where, if x denotes a column vector, then xT = (x1 , x2 , x3 , x4 ),

⎛ ⎞
1 0 0 0
⎜0 1 0 0 ⎟
η=⎜ ⎝0 0 1 0 ⎠ .
⎟ (2.49)
0 0 0 −1

We say that the metric tensor η has signature (1, 1, 1, −1) or (3, 1). More
popular among physicists is the opposite convention: one plus, three mi-
nuses. This other choice is particularly convenient when we do not want
to interfere with the sign of energy, the quantity dual to “time”. On the
first pages of Ref. [Misner et al., 1973, Table of sign conventions] one can
find a useful table presenting conventions that are used in most popular
textbooks.

Remark 2.4. Throughout this book we will be using Einstein’s convention:

summation over repeated indices, whenever one is an upper index and one
is a lower index. Lower and upper indices are, as a rule, related by the
metric matrix ηαβ , η = η −1 , η −1 = {η αβ }. Thus, for instance,

xα = ηαβ xβ , xα = ηαβ xβ ,
q(x) = ηαβ xα xβ = ηαβ xα xβ . (2.50)

Roman indices i, j, k, ... as a rule, are space indices running from 1 to 3,

while Greek indices μ, ν, σ, ... run from 0 to 3, or from 1 to 4, depending on
the context.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 45/2

What are Quantum Fractals? 45

If M is considered as an affine space, then the symmetry group of M

is the Poincaré group, or Inhomogeneous Lorentz group. It contains trans-
lation subgroup, and it contains the subgroup of transformations that pre-
serve the origin x = 0 — the Lorentz group, denoted O(3, 1) (which is the
same as O(1, 3)). O(3, 1) is the set of all real 4 × 4 matrices L that preserve
the quadratic form q. Thus a matrix L is in O(3, 1) if and only if q(Lx) = x
for all x in M. This is equivalent to the matrix equation:
LT ηL = η. (2.51)
From Eq. (2.51) it follows that the product of two Lorentz matrices is
again a Lorentz matrix. On the other hand, by taking the determinant
of both sides of Eq. (2.51), and using the fact that det(LT ) = det(L) and
det(η) = −1, we find that det(L)2 = 1, therefore det(L) = ±1. Thus, in
particular, all Lorentz matrices are invertible. Evidently, if a matrix L
preserves the quadratic form q, so does its inverse. Therefore O(3, 1) is a
group of matrices.

2.7.2 Action of the Lorentz group on the sphere

Lorentz transformations preserve the form q(x), therefore they preserve, in
particular, the double cone — the light cone — defined by the equation
q(x) = 0, i.e.
(x4 )2 = (x1 )2 + (x2 )2 + (x3 )2 . (2.52)
Since Lorentz transformations are linear, they map generator lines of the
cone onto generator lines. These generator lines can be faithfully repre-
sented by the two-sphere that results from the intersection of the cone with
the plane x4 = 1 — as shown in Fig. 2.22, where we omitted the x2 coor-
dinate.
This cone can be split into three parts: the origin (or apex) x = 0, the
future cone defined by the formula

x4 = (x1 )2 + (x2 )2 + (x3 )2 > 0, (2.53)
and the past cone

x4 = − (x1 )2 + (x2 )2 + (x3 )2 < 0. (2.54)
Every Lorentz transformation leaves the origin x = 0 invariant. We will be
interested only in those Lorentz transformations that map the future cone
into itself. The Lorentz matrices implementing such transformations form
a subgroup of the Lorentz group; it is denoted by O+ (3, 1) and often called
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 46/2

46 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.22 Intersection of the future light cone with the plane x4 = 1, including two
generator lines.

the orthochronous Lorentz group. The subgroup of O+ (3, 1) consisting of

matrices L ∈ O+ (3, 1) with det(L) = +1 is called the restricted Lorentz
group and denoted SO + (3, 1).

Remark 2.5. In applications to quantum fractals, in principle, an iterated

function system may contain transformations from the whole group O(3, 1),
thus space or time inversions are allowed. Yet in all our examples we will
have space and time orientation preserved — all our transformations will
be from the restricted Lorentz group SO+ (3, 1).

Knowing that our group acts on the sphere, the next thing is to get a
parametric representation of this action. To this end let us consider the
intersection of the future light cone with the hyperplane x4 = 1. This
intersection is the two-sphere described by the equations
(x1 )2 + (x2 )2 + (x3 )2 = 1, x4 = 1. (2.55)
For a point x belonging to this intersection, define
ni = xi , i = 1, 2, 3, n = (n1 , n2 , n3 ), n4 = 1. (2.56)
Then, owing to the equation 0 = q(n) = n − (n ) = n − 1, the vector
2 4 2 2

n is a unit vector.6
6 We are using here the letter n instead of x in order to stress the fact that our vector

is normalized.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 47/2

What are Quantum Fractals? 47

Let L be a matrix from the restricted Lorentz group SO+ (3, 1). The
vector Ln has coordinates:
(Ln)i = Lij nj + Li4 n4 , (2.57)
(Ln)4 = L4j nj + L44 n4 . (2.58)
Now, L maps the future light cone onto itself. Therefore (cf. Eq. (2.53)):

 3

(Ln) =  ((Ln)i )2 > 0.
4
(2.59)
i=1

Define
(Ln)i
(L · n)i = . (2.60)
(Ln)4
Or, explicitly:
3
j=1 Lij nj + Li4
(L · n) = 3
i
. (2.61)
j=1 L4j nj + L44
Now, from Eq. (2.60), if n2 = 1, then also L · n has square of one. Thus
the restricted Lorentz group SO+ (3, 1) acts, through n → L · n, on the
two-sphere.

Remark 2.6. In computer implementations of quantum fractals we will

use the formula Eq. (2.61) repeatedly. However it is better to implement it
in a different way. Let us denote by ñ = {ñi } the numerator:

3
ñi = Lij nj + Li4 . (2.62)
j=1

Then the denominator of Eq. (2.61) is nothing else than the norm of the
vector ñ. Taking the quotient in Eq. (2.61) is equivalent to the normalization
of the vector. Yet, computationally, there will be always small numerical
differences between ||ñ|| and the denominator in Eq. (2.61) calculated using
the coefficients of the matrix L. On certain occasions these differences tend
to accumulate in iterations. Therefore it is numerically more stable to
implement the transformations Eq. (2.61) as
ñ
L·n= , (2.63)
||ñ||
that is
ñi
(L · n)i =  . (2.64)
3 j 2
j=1 (ñ )
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 48/2

48 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

2.7.3 The group SL(2, C)

The restricted Lorentz group SO+ (3, 1) is path connected — any two differ-
ent elements of the group can be connected by a continuous path within the
group. But SO+ (3, 1) is not simply connected. There are closed loops within
the group that cannot be continuously deformed to a point. Responsible
for this fact is the rotation subgroup SO(3). The simply connected group
that is the double cover of SO+ (3, 1) is its spin group realized as SL(2, C)
— the group of 2 × 2 complex matrices of determinant one.
In this part of the book we will be working in a two-dimensional complex
vector space C2 . A typical element of this space is a column vector ξ = ( uv ),
where u, v ∈ C — the complex numbers. The Hermitian scalar product and
the norm in C2 are defined by the formulas

(ξ, ξ  ) = ūu + v̄v  , (2.65)

||ξ||2 = (ξ, ξ) = ūu + v̄v = |u|2 + |v|2 . (2.66)

Linear transformations on vectors

 of C2 are represented by 2 × 2 complex
matrices, typically A = ac db . The determinant map det : A → det(A)
associates with each matrix A the complex number det(A) = ad − bc. De-
noting by I the identity matrix I = ( 10 01 ) , we have det(I) = 1. Since
determinant of a product is a product of determinants, invertible matri-
ces (which are matrices A with det(A) = 0), form a group with respect
to matrix multiplication. This group is denoted GL(2, C) (sometimes also
written as GL2 (C) or M at(2, C)) — the General Linear Group. Matrices
with determinant 1 form a subgroup  SL(2, C) of GL(2, C) — the Spe-
cial Linear Group. Thus for A = ac db to be in SL(2, C) one must have
ad − bc = 1.
 
Remark 2.7. Notice that if A = ac db is in SL(2, C), then
−1
ab d −b
= , (2.67)
cd −c a

which can be easily verified by multiplication and by taking into account

the determinant condition ad − bc = 1.

Example 2.1. The matrix U (φ) given by

eiφ 0
U = U (φ) = , φ∈R (2.68)
0 e−iφ
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 49/2

What are Quantum Fractals? 49

is not only in SL(2, C) but also in SU (2) — the group of unitary ma-
trices (i.e. U ∗ U = I) of determinant 1. We will see that it can be inter-
preted as describing a rotation in the three-dimensional space R3 around the
z-axis by the angle 2φ. The group SU (2) is the double covering group of
the proper orthogonal group SO+ (3).

Example 2.2. The matrix T given by

eα 0
T = T (α) = , α∈R (2.69)
0 e−α
is in SL(2, C). Moreover, it is Hermitian and positive (i.e. (ξ, T ξ) ≥ 0 for
all ξ ∈ C2 ). We will see that it can be interpreted as describing “Lorentz
boost” in the direction of the negative z-axis with velocity v = tanh 2α.

Both U (φ) and T (α) are one-parameter subgroups of SL(2, C). We have
U (φ)U (φ ) = U (φ + φ ) and T (α)T (α ) = T (α + α ).

Example 2.3. The following two-parameter family of matrices B(z),

z ∈ C:
1z
B(z) = (2.70)
01
forms a two-parameter commutative (Abelian) subgroup of SL(2, C):
B(z)B(z  ) = B(z + z  ). (2.71)
These matrices (except of the identity matrix I = B(0)) are neither unitary
nor Hermitian. But each of them can be uniquely decomposed (via the
polar decomposition) into a product of a unitary matrix (rotation) and a
positive matrix (Lorentz boost). Unitary matrices in SL(2, C) can be easily
parametrized — as it is shown in Proposition 2.1, Eq. (2.72) below.
 
Proposition 2.1. Let A = ac db be in SL(2, C). Then A is unitary, i.e.
A∗ A = I if and only if c = −b̄, d = ā and |a|2 + |b|2 = 1. In other words:
U is in SU (2) if and only if U is of the form
a b
U= , |a|2 + |b|2 = 1. (2.72)
−b̄ ā
In particular the whole group SU (2) is homeomorphic to the sphere S 3 .
 
Proof. With A = ac db , writing down the unitarity conditions A∗ A = I,
we get:
(1) |a|2 + |b|2 = 1,
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 50/2

50 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

(2) ac̄ + bd¯ = 0,

(3) |c|2 + |d|2 = 1.
If a = 0, then, from (1), |b| = 1, thus, from (2), d = 0, and therefore,
from the determinant condition ad − bc = 1, it follows that c = −b̄. Let us
suppose that a = 0, then, multiplying the conjugated equation (2) by a,
and multiplying the determinant equation by −b̄ we get:
|a|2 c + ab̄d = 0,
|b|2 c − ab̄d = −b̄.
Adding these two together and using (2) we get c = −b̄. Then the de-
terminant equation ad − bc = 1 becomes ad + |b|2 = 1. Comparing this
equation with (1) we deduce that |a|2 = ad, therefore d = ā. Let us write
a = x1 +ix2 , b = x3 +ix4 , then (2) translates to (x1 )2 +(x2 )2 +(x3 )2 +(x4 )2 =
1, which is the equation of the 3-sphere S 3 in R4 . 

2.7.4 Action of SL(2, C) on the two-sphere S 2

The group SL(2, C) of 2 × 2 matrices with complex coefficients and deter-
minant 1 acts on the two-sphere, and this action can be described in three
different but equivalent ways:
(1) We have the 2 : 1 covering homomorphism A → Λ(A) from SL(2, C)
onto the restricted Lorentz group SO+ (3, 1) defined explicitly in
Eq. (2.86).
(2) SL(2, C) acts on one-dimensional complex subspaces of C2 represented
by orthogonal projection operators P acting on C2 :
A : P → P  , AP A∗ = λP  , λ > 0. (2.73)
(3) SL(2, C) acts via linear fractional transformations on the complex
plane:
ab az + b
= A : z → , (2.74)
cd cz + d
but the complex plane z = x+iy is identified through the stereographic
projection with the two-sphere X, Y, Z, X 2 + Y 2 + Z 2 = 1 (minus its
north pole Z = 1):
X Y
(x, y) = , , z = x + iy ∈ C. (2.75)
1−Z 1−Z
Here, exceptionally, we are using capital letters X, Y, Z for denoting the
coordinates in the three-dimensional Euclidean space.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 51/2

What are Quantum Fractals? 51

2.7.4.1 Action via group homomorphism SL(2, C) → SO+ (3, 1)

We will describe now the 2:1 covering homomorphism SL(2, C) →
SO+ (3, 1).

Pauli matrices. The first step, that is widely used in relativistic physics,
is to realize the four-dimensional Minkowski space M as the space of all
Hermitian 2 × 2 matrices. To this end we introduce four Pauli matrices
sμ , (μ = 0, 1, 2, 3), with coefficients (sμ )ab , as follows7 :
10 01 0 i 1 0
s0 = , s1 = , s2 = , s3 = . (2.76)
01 10 −i 0 0 −1
The matrices sμ have the (easily verified) property that
Tr (sμ sν ) = 2δμν , (2.77)
where δμν is the “Kronecker delta”. Moreover, we have the following com-
pleteness formula (that can also be verified by direct calculations):

3
(sμ )ab (sμ )cd = 2δad δbc , (a, b, c, d = 1, 2). (2.78)
μ=0

These four matrices are Hermitian: sμ ∗ = sμ . The matrices s0 , s1 , s3 are

real symmetric, while s2 is imaginary antisymmetric. They form a basis
in the real four-dimensional vector space of all Hermitian 2 × 2 matrices.
Thus every Hermitian matrix X can be uniquely decomposed as a linear
combination of sμ with real coefficients:
X = xμ sμ
= x0 s0 + x1 s1 + x2 s2 + x3 s3
x0 + x3 x1 + ix2
= . (2.79)
x1 − ix2 x0 − x3
Minkowski space realized by 2 ×2 Hermitian matrices: With X as
in Eq. (2.79), we get
det(X) = (x0 )2 − (x1 ) − (x2 )2 − (x3 )2 , (2.80)
which is (up to a sign, and up to renaming x0 = x4 ) the quadratic form
in R4 defining the Minkowski geometry. The coordinate x0 is the time
coordinate, while xi , (i = 1, 2, 3), are space coordinates of the space-time
7 Often s is defined with the opposite sign. With our choice the three matrices s , i =
2 i
1, 2, 3 satisfy the rule si sj = δij s0 − iijk sk , where δij is the Kronecker δ and ijk is the
totally antisymmetric Levi-Civita symbol.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 52/2

52 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

point (event). Notice that, owing to the formula Eq. (2.77), the coordinates
xμ can be read from the matrix X:
1
xμ = Tr (Xσμ ). (2.81)
2
In particular, what follows also from the explicit form of X in Eq. (2.79),
1
x0 = Tr (X). (2.82)
2
Action of SL(2, C): If X is a Hermitian matrix, X = X ∗ , and if A is
an arbitrary matrix, then X  = AXA∗ is also Hermitian, as we have
∗
X  = (AXA∗ )∗ = A∗∗ XA∗ = AXA∗ = X  . (2.83)
If A is a matrix in SL(2, C), then det(A) = 1, therefore
det(X  ) = det(AXA∗ ) = | det(A)|2 det(X) = det(X), (2.84)
or
(x )2 − (x ) − (x )2 − (x )2 = (x0 )2 − (x1 )2 − (x2 )2 − (x3 )2 .
0 1 2 3
(2.85)

From SL(2, C) to Lorentz matrices: If A is a matrix in SL(2, C),

then Asμ A∗ are Hermitian matrices and therefore can be decomposed into
a linear combination of sμ with real coefficients L(A)μν .
Asμ A∗ = sν L(A)ν μ . (2.86)

Remark 2.8. Although L(A)ν μ are (real) numbers and sμ are (complex,
but Hermitian) matrices, it is convenient to write the formula Eq. (2.86)
with numbers on the right hand side. It helps keeping track of the left
and right group actions. With the notation as above we have L(AA ) =
L(A)L(A ), where the matrix multiplication is the standard one:
(LL )μν = Lμρ Lρν . (2.87)

By taking the trace of both sides of Eq. (2.86) multiplied by sλ , and using
Eq. (2.77) we obtain the following important formula8 :
1
L(A)μν =
Tr (sμ Asν A ). (2.88)
2
It follows now from Eq. (2.78) that
Tr (L(A)) = |Tr (A)|2 . (2.89)
8A similar formula can be found in [Carmeli and Malin, 2000, p. 56, (3.84a)], however
with an error. Carmeli and Malin have there (incorrectly) Λαβ instead of the correct
Λαβ . On the other hand, in Ref. [Carmeli, 1977, p. 36] the formula is correct.
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 53/2

What are Quantum Fractals? 53

Remark 2.9. In the formulas below we are using index 4 instead of the
index 0 for the “time coordinate”. We do so because these formulas are
being used in the computer implementation of quantum fractals. In some
computer languages such as Fortran, Mathematica, Matlab, some Pascal
versions, the “native” indexing of arrays starts with the index 1 rather than
from 0. One can always go to the base index 0 representation by renaming
the index 4 → 0.

Proposition 2.2. For an SL(2, C) matrix

ab
A= , ad − bc = 1 (2.90)
cd
the 16 entries of the real 4 × 4 matrix L = L(A) are given explicitly by the
following formulas:

L11 = (bc̄ + ad¯), L12 = ((bc̄ − ad¯), L13 = (ac̄ − bd¯),

L14 = (ac̄ + bd¯), L21 = (bc̄ + ad¯), L22 = (ad¯ − bc̄),
L23 = (ac̄ − bd¯), L24 = (ac̄ + bd¯), L31 = (ab̄ − cd¯),
L32 = (cd¯ − ab̄), L33 = (aā − bb̄ − cc̄ + dd¯)/2, (2.91)
L34 = (aā + bb̄ − cc̄ − dd¯)/2), L41 = (ab̄ + cd¯),
L42 = −(ab̄ + cd¯), L43 = (aā − bb̄ + cc̄ − dd¯)/2),
L44 = (aā + bb̄ + cc̄ + dd¯)/2.

(z) and (z) denote respectively the real and the imaginary part of a
complex number z.

Proof. The proof follows by direct calculations based on Eq. (2.88).9 

One can show that the map A → L(A) is a group homomorphism from
SL(2, C) onto the connected component of the identity SO+ (3, 1) of the
Lorentz group — cf. [Carmeli and Malin, 2000, Ch. 3.4.2].
Let X be a Hermitian matrix representing a point (“event”) with coor-
dinates xμ in Minkowski’s space. Then X  = AXA∗ is also Hermitian and
can be written as X  = xμ sμ . We then have

xμ sμ = Axμ sμ A∗ = xμ Asμ A∗ = xμ sν L(A)ν μ , (2.92)

9 These formulas agree, up to the order of terms, with those given in Ref. [Carmeli and

Malin, 2000, p. 62] and [Naber (2012)].

June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 54/2

54 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

and, from the fact that the matrices sμ form a basis in the space of Hermi-
tian matrices, after renaming the labelling indices, we obtain:
xν = L(A)ν μ xμ . (2.93)
Since, owing to Eq. (2.84), Eq. (2.85), det(X  ) = det(X), the Minkowski
quadratic form has the same value on xμ and on xμ . Therefore L is a
Lorentz transformation.
Notice that, owing to Eq. (2.91), the component L(A)44 of the matrix
L(A) is necessarily positive. In fact, since L(A)44 = Tr (A∗ A)/2 ≥ 1,
L(A)44 ≥ 1. (2.94)

Preservation of the future light cone. We will now show that L(A)
preserves the future light cone. Let x, with coordinates xμ , be a point on
the future light cone. Then det(X) = 0 and x0 = Tr (X) > 0; the matrix X
has two eigenvalues. The determinant is the product of these eigenvalues,
therefore one of these eigenvalues must be zero, the other one must be
positive. X is a positive definite matrix.
Recall that a matrix A is called positive (or positive definite) if and only
if u∗ Au > 0 for any nonzero vector u. Or, written differently,
(ξ, Xξ) > 0 for all 0 = ξ ∈ C2 . (2.95)
A matrix is positive if and only if it is Hermitian and all its eigenvalues are
positive. If A and B are positive matrices, then tA + (1 − t)B is positive
for 0 ≤ t ≤ 1. It is easy to see that for 2 × 2 matrices the property of
positivity is equivalent to the property of both, determinant (the product
of eigenvalues) and the trace (the sum of eigenvalues) of the matrix, being
positive.
But then (ξ, X  ξ) = (A∗ ξ, XA∗ ξ) = (ξ  , Xξ  ), where ξ  = A∗ ξ, is
also positive. Therefore X  is positive definite as well. Since det(X  ) =
det(X) = 0, it follows that the trace of X  is positive, therefore x > 0.
0

Topology of SO+ (3, 1). The image of SL(2, C) under the group homo-
morphism SL(2, C) → SO+ (3, 1) not only preserves the time direction, but
also does not contain space inversions, that is that all the matrices L(A)
have determinant +1. This property follows from the fact that SL(2, C),
as a topological space, is path connected, which, in turn, follows from the
polar decomposition theorem (cf. e.g. [Axler (1997)]). Namely, every non-
singular matrix A can √ be uniquely decomposed as A = T U, where T is a
positive matrix (T = AA∗ ), and U is a unitary matrix (U U ∗ = I). If
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 55/2

What are Quantum Fractals? 55

A is in SL(2, C), then det(T ) = 1, therefore det(U ) = 1, that is, U is in

SU (2). The group SU (2) (isomorphic to the group of unit quaternions)
has the topology of the 3-sphere S 3 (cf. Proposition 2.1), and S 3 is path
connected. The set of positive matrices is a convex cone, and we can al-
ways normalize a path there to get determinant 1. SO+ (3, 1) is an image
of a path connected space, therefore it is also path connected, thus the
determinant of every matrix L from SO+ (3, 1) must be = +1.
The action of SL(2, C) on the two-sphere via the group homomorphism
A → L(A), SL(2, C) → SO+ (3, 1) and Eq. (2.64) is now clear: we define
A · n = L(A) · n, n2 = 1, n ∈ S 2 , A ∈ SL(2, C). (2.96)

2.7.5 Projection operators representations of the Bloch

sphere S 2
Quantum fractals emerged from quantum theory. In quantum theory the
fundamental mathematical objects are state vectors and operators. The
simplest quantum system is a pure (thus no space degrees from freedom)
spin 1/2. Its states are represented by vectors in a 2-dimensional Hilbert
space realized as C2 . Operators operate on states, and linear operators in
C2 are represented by complex 2 × 2 matrices.

2.7.5.1 The Bloch sphere

Of particular interest are Hermitian matrices A = A , where A∗ is defined
by the relation (ξ, Aξ  ) = (A∗ ξ, ξ  ) for all vectors ξ, ξ  (cf. Eq. (2.66)). They
represent, what arecalled  in quantum theory,“observables”. Such a matrix
is of the form A = ac db , with a = ā, d = d, ¯ c = b̄. Real linear combination
of Hermitian matrices are again Hermitian. The space of Hermitian matri-
ces is, in our case, a four-dimensional real vector space. A convenient basis
for this vector space consist of the four matrices sμ defined by Eq. (2.76).
In quantum theory matrices si , (i = 1, 2, 3), (often with the difference of
the sign of s2 ), describe “spin 1/2 directions”.
As already noticed in Eq. (2.79), every Hermitian matrix X decomposes
uniquely into a linear combination:

3
x0 + x3 x1 + ix2
X= xμ sμ = , (2.97)
x1 − ix2 x0 − x3
μ=0

with real coefficients xμ . The coefficients xμ can be read back from the
matrix X using the simple formulas in Eq. (2.81), Eq. (2.82).
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 56/2

56 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Among Hermitian matrices there are special ones that are idempotents.
They have the property that X 2 = X. The matrices X = 0, and X = I
do have this property, but they are considered as being trivial idempotents.
What are the nontrivial ones? As we will show, they are orthogonal projec-
tion operators on one-dimensional subspaces of C2 .
As a rule, we will use the letter P to denote them. Using Eq. (2.97) and
writing down the equation X 2 = X we obtain that any nontrivial Hermitian
idempotent P is necessarily of the form
1
Pn = (I + n · s), (2.98)
2
where n1 , n2 , n3 are real,
n · s = n1 s 1 + n 2 s 2 + n 3 s 3 , n2 = n21 + n22 + n23 = 1. (2.99)
Conversely, every such Pn is an idempotent. From Eq. (2.81) we can then
calculate ni in terms of Pn :
1
ni = Tr ((2Pn − I)si ) = Tr (Pn si ), i = 1, 2, 3, (2.100)
2
where we have used the fact that Tr (si ) = 0 for i = 1, 2, 3.
Therefore Hermitian idempotents are uniquely parametrized by vectors
n on the unit sphere S 2 — according to Eq. (2.98).

Example 2.4. This elementary example is standard in the textbooks of

quantum mechanics. Take n = (0, 0, 1) (the north pole of the sphere), and
find explicitly the subspace of C2 onto which the operator Pn is projecting
onto.
We have
1 20 10
P(0,0,1) = = , (2.101)
2 00 00
therefore
z1 z1
P(0,0,1) = , (2.102)
z2 0
thus P(0,0,1) projects onto the subspace spanned by the vector ξ0
1
ξ0 = (2.103)
0
that will be excluded from the complex plane representation.

In quantum mechanics the sphere n2 = 1, whose points are associated

with orthogonal projection operators on one-dimensional subspaces of C2 ,
is called the Bloch sphere, named after the physicist Felix Bloch.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 57/2

What are Quantum Fractals? 57

2.7.5.2 The complex plane representation

In this section we will examine the complex plane representation of spin
1/2 states in more detail. As a byproduct we will derive the stereographic
projection formulas.
What is the relation between other points n = (n1 , n2 , n3 ) of the unit
sphere n21 + n22 + n23 = 1 and the complex numbers z of the complex plane
representation?
It is easier to start with the complex plane representation, so let z =
x + iy, x, y ∈ R be a complex number.
 We  associate with z the vector
ξ = ( z1 ) . Now, the norm of ξ is 1 + |z|2 = 1 + x2 + y 2 , therefore the
normalized vector (it represents the same quantum spin state as ξ) is
⎛ x+iy ⎞
√
ξ ⎠.
=⎝
1+x2 +y 2
ζ= (2.104)
||ξ|| √ 12 2
1+x +y

The general formula for the orthogonal projection operator onto the sub-
space spanned by a normalized vector ζ reads:
ζ1  ¯ ¯ 
Pζ = ζ ⊗ ζ † = ζ1 ζ2 , (2.105)
ζ2
or, in action on a vector χ

Pζ (χ) = (ζ, χ)ζ, (2.106)

where (ζ, χ) is the standard Hermitian scalar product in C2 . Therefore Pζ

is represented by the matrix:
ζ1 ζ¯1 ζ1 ζ¯2
Pζ = . (2.107)
ζ2 ζ¯1 ζ2 ζ¯2

We can now use Eq. (2.104) in order to express now Pζ in term of complex
plane coordinates (x, y):

Pζ = ζ ⊗ ζ † (2.108)
1 z  
== z̄ 1
1 + |z|2 1
1 |z|2 z
=
1 + |z|2 z̄ 1
1 1 + x2 + y 2 x + iy
=
1 + x2 + y 2 x − iy 1
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 58/2

58 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

In the following we will use the formula:

1 |z|2 z
Pz = . (2.109)
1 + |z|2 z̄ 1
Notice that Tr (Pz ) = 1, as it should be, since the trace of an orthogonal
projection is the dimension of the corresponding subspace. The determinant
of Pz is zero, since Pz is singular (non-invertible).

Stereographic projection. We can use now equations Eq. (2.100) and

Eq. (2.98) to calculate n(z) with coordinates ni in terms of z = x + iy. The
result is:
2x
n1 = , (2.110)
1 + x2 + y 2
2y
n2 = , (2.111)
1 + x2 + y 2
−1 + x2 + y 2
n3 = . (2.112)
1 + x2 + y 2
These formulas describe, in fact, the inverse stereographic projection. We
can calculate x, y in terms of n1 , n2 , n3 subjected to the unit length vector
constraint n2 = 1. We obtain the stereographic projection formulas:
n1
x= , (2.113)
1 − n3
n2
y= . (2.114)
1 − n3
The projection is from the north pole onto the plane cutting the sphere at
the equator. The north pole of the sphere itself, n1 = n2 = 0, n3 = 1, is
excluded from the projection.
The upper hemisphere is mapped onto the interior of the unit disk, the
lower hemisphere is mapped onto its exterior. Our sphere is nothing but
the Riemann sphere — the compactification of the complex plane C by
adding one point set — {∞}. Schematically (cutting off the y-dimension)
this is depicted in Fig. 2.23.
The above can be extended to a general relation between the complex
vector coordinates and real sphere coordinates. Let ξ be a complex vector:
z1
ξ= , z 1 , z2 ∈ C (2.115)
z2
with
z1 = x1 + iy1 , z2 = x2 + iy2 , x1 , y1 , x2 , y2 ∈ R. (2.116)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 59/2

What are Quantum Fractals? 59

2x
n1 = 1+|z|2
{∞} = (0, 0, 1) 2y
n2 = 1+|z|2
−1+|z|2
n3 = 1+|z|2
n = (n1 , n2 , n3 )

z=0 z = x + iy
n1
x = 1−n 3
n2
y = 1−n3

Fig. 2.23 Stereographic projection relating complex plane and projection operator rep-
resentations.

Assuming, temporarily, that z2 = 0, we define the proportional vector ξ  :

z1
z
ξ = z2
= , (2.117)
1 1
where
1
z= (x1 x2 + y1 y2 + i(x2 y1 − x1 y2 )). (2.118)
x22
+ y22
We can apply now the formulas for stereographic projection given in
Eq. (2.112) with the following result
2(x1 x2 + y1 y2 )
n1 = 2 , (2.119)
x1 + x22 + y12 + y22
2(x2 y1 − x1 y2 )
n2 = , (2.120)
x21 + x22 + y12 + y22
x21 − x22 + y12 − y22
n3 = . (2.121)
x21 + x22 + y12 + y22
The above formulas are now valid even without the restriction z2 = 0,
(provided that ξ is not the zero vector with both components equal to
zero) and can be written as
2(z1z¯2 )
n1 = , (2.122)
|z1 |2 + |z2 |2
2(z1z¯2 )
n2 = , (2.123)
|z1 |2 + |z2 |2
|z1 |2 − |z2 |2
n3 = . (2.124)
|z1 |2 + |z2 |2
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 60/2

60 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

It is now evident that this sphere representation (n is a unit vector in

R3 defining the point on the sphere) of the spin 1/2 state vectors maps
proportional complex vectors into the same point on the sphere. Indeed,
the expressions for the coordinates of n are insensitive to scalings ξ → λξ,
λ > 0, and also insensitive to phase transformations ξ → eiφ ξ, φ ∈ R.

2.7.6 Visualization of quantum spin states and state

vectors
According to the standard postulates of quantum theory, and also according
to its present usage, the phase of the vector in the Hilbert space C2 of spin
1/2 is “unobservable”. Even if unobservable, that does not mean that it is
unimportant. There is the phenomenon of “quantum interference”, in fact
one of the most important concepts in quantum theory, where the relative
phase of two state vectors play a decisive role. For this reason it is useful
to have a geometric interpretation not only of quantum states, but also of
their relation to state vectors. State vectors carry information about both:
quantum state and phase. We know that quantum states of spin 1/2 are
represented by points on the unit sphere — the Bloch sphere. But how to
visualize quantum state vectors, we will call them spin vectors, and their
relation to the points on the Bloch sphere? We will discuss this point, and
illustrate it in the present section.
A spin vector ξ is a unit vector in C2 :
z1
ξ= , |z1 |2 + |z2 |2 = 1. (2.125)
z2
Let us write the components of ξ in terms of real coordinates X, Y, Z, W
as follows:
z1 = X + iY,
z2 = Z + iW. (2.126)
We can associate with this vector in the two-dimensional complex space a
pair of vectors a, b in the Euclidean plane R2 :
X Z
a= ,b= , (2.127)
Y W
with
a2 + b2 = 1, (2.128)
where
 
a = |a| = X 2 + Y 2, b = |b| = Z 2 + W 2. (2.129)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 61/2

What are Quantum Fractals? 61

We can now express the coordinates nx , ny , nz of the corresponding point

(spin state) on the Bloch sphere using Eq. (2.124). But, in this section,
for the identification of spin states with points nx , ny , nz we will use in
Eq. (2.98) the standard Pauli matrices σ = (σx , σy , σz ) instead of the ma-
trices s1 , s2 , s3 . That is equivalent to changing the sign of ny .
With this convention the result reads:

nx = 2(XZ + Y W ) = 2 a · b,
ny = XW − Y Z = 2 a × b, (2.130)
nz = X + Y − Z − W = a − b = 2a − 1.
2 2 2 2 2 2 2

Now, a change of the phase

ξ → eiψ ξ (2.131)

of the spin vector ξ, (also known as a gauge transformation), is realized by

a simultaneous rotation of the two vectors a, b by the same angle ψ in R2 .
It is clear that nx , ny , nz do not change under such a transformation for the
reason that they are expressed in terms of rotational invariants of pairs of
vectors in R2 .
We can express each of the pair of two vectors a, b in terms of the polar
coordinates ψ, θ on the plane R2 as follows:

X = a cos(ψ), (2.132)
Y = a sin(ψ), (2.133)
Z = b cos(θ + ψ), (2.134)
W = b sin(θ + ψ). (2.135)

For the vector b we have introduced θ + ψ instead of just one independent

angle. This has the advantage that by changing ψ we are rotating both
vectors by the same angle, what corresponds to the gauge transformation
Eq. (2.131) — cf. Fig. 2.24.
For a better geometric interpretation and visualization we still have
to solve the constraint a2 + b2 = 1. A convenient way of doing it is by
introducing a new angle φ in the range [0, π], and to set
φ
a = cos , (2.136)
2
φ
b = sin . (2.137)
2
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 62/2

62 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

b
b

ψψ++θθ
θ a
a

ψ
ψ

a2a+
2
+bb22 =
= 11

Fig. 2.24 Spinor represented by a pair of two vectors a, b with a2 + b2 = 1.

In terms of the angles φ, θ, ψ we can now express X, Y, Z, W and nx , ny , nz

as follows:
φ
X = cos cos(ψ), (2.138)
2
φ
Y = cos sin(ψ), (2.139)
2
φ
Z = sin cos(θ + ψ), (2.140)
2
φ
W = sin sin(θ + ψ). (2.141)
2

nx = sin(φ) cos(θ),
ny = sin(φ) sin(θ), (2.142)
nz = cos(φ).
We see that the phase angle ψ disappears completely from the expressions
for the components of the vector n, while the angles φ and θ play the
role of the standard spherical angle coordinates on the sphere: φ is the
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 63/2

What are Quantum Fractals? 63

latitude measured from the north pole, and θ is the longitude measured
from the meridian of y = 0.
This way we have implemented a map from the three-dimensional sphere
S 3 described by the equation X 2 + Y 2 + Z 2 + W 2 = 1, and parametrized
by three angles φ, θ, ψ, onto two-dimensional sphere S 2 described by the
equation n2x + n2y + n2z = 1 and parametrized by the angles φ, θ. This map
S 3 → S 2 is known as Hopf fibration. The fibers of the Hopf fibration are
the circles parametrized by the ψ angle while φ and θ are constant.
We will now describe the physical meaning of the angle φ.
Let n be a spin direction, and let P (n) be the orthogonal projection on
the corresponding spin state:
1
P (n) = (I + σ · n) . (2.143)
2
According to the standard quantum theory the probability pξ (n) of regis-
tering spin in the direction n, when the spin state is described by a unit
vector ξ in C2 , is given by the formula:

pξ (n) = (ξ, P (n)ξ). (2.144)

Suppose now that n = (0, 0, 1), that is that we want to measure the prob-
ability of the spin direction pointing towards the north pole of the sphere.
In this case, using the notation of this section, Eq. (2.144) gives us

φ 1 + cos(φ)
pξ (n) = |z1 |2 = a2 = cos2 = . (2.145)
2 2
The foci of constant probability on the Bloch sphere are therefore parallels
of latitude. The north pole, φ = 0, gives us probability 1; the south pole,
φ = π — probability 0; while the equator, φ = π/2 — probability 1/2.
But how to visualize not just states, but spin vectors representing spin
states of a constant probability? To this end we will use the stereographic
projection from the three-sphere S 3 to the Euclidean space R3 . The formu-
las for the stereographic projection are similar to those in two dimension:
X
x= ,
1−W
Y
y= , (2.146)
1−W
Z
z= .
1−W
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 64/2

64 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Substituting Eq. (2.141) we get parametric equations of the two-

dimensional surfaces of constant probability:
 
cos φ2 cos(ψ)
x(θ, ψ) =   ,
1 − sin φ2 sin(θ + ψ)
 
cos φ2 sin(ψ)
y(θ, ψ) =   , (2.147)
1 − sin φ2 sin(θ + ψ)
 
sin φ2 cos(θ + ψ)
z(θ, ψ) =   ,
1 − sin φ2 sin(θ + ψ)

where φ is constant on each of the surfaces. An illustrative example of three

such isosurfaces is shown in Fig. 2.25. What is still missing in Fig. 2.25 are
the lines of the phase parameter ψ. Figure 2.26 shows also these lines. Each
of the circles in Fig. 2.26 represents just one point on the Bloch sphere,

Fig. 2.25 Tori of constant probability. Stereographic projection from the sphere S 3 .
Shown three isosurfaces for p = 0.853553, 0.5, 0.146447 corresponding respectively to
φ = π/4, π/2, 3π/4. The most inner torus (cut from the top and the bottom) is that of
φ = 3π/4.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 65/2

What are Quantum Fractals? 65

Fig. 2.26 Isosurfaces of constant probability with a family of phase lines (the “Villarceau
circles”) — the fibers of the Hopf fibration.

that is just one spin state.10 Our choice of the singular point W = 1 in
the stereographic projection formula is preferring the z-axis of the spin
direction. It is therefore instructive to consider also the measurement of,
say, the x-component of the spin. In this case the projection P (n) becomes:
1 1 11
P (n) = (I + σx ) = , (2.148)
2 2 11
which easily translates into the probability formula:
1
px (φ, θ, ψ) = (1 + sin(φ) cos(θ)) . (2.149)
2
Suppose now that we want to visualize the isosurface of px = 1/2. In this
case the isosurface equation is cos(θ) sin(φ) = 0, therefore the isosurface can
be build from two parts: one with θ = π/2, the other one with θ = 3π/2,
while φ ∈ [0, π] and ψ ∈ [0, 2π] are arbitrary. The resulting surface is a
deformed torus, closing at infinity, as can be seen in Fig. 2.28. Figure 2.29
shows a family of 16 such isosurfaces (which happen to be, in this case,
half-planes) of constant phase ψ. In EEQT (cf. Sec. 4.4) we use fuzzy
10 In a different context (related to Penrose twistors) essentially the same picture can be

found in, for instance, [Penrose and Rind, 1986, p. 62].

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 66/2

66 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.27 A family of Villarceau circles making one isosurface torus viewed from above.
The torus represents one parallel of latitude on the Bloch sphere. Each circle represents
one point on the Bloch sphere.

projections P (n, k) instead of sharp projection P (n):

1
P (n, k) = (I + kn · σ). (2.150)
2
The isosurfaces of constant probability are, in this case, the same as in
the case of sharp spin measurements. The only difference between the two
cases is that there are no states with probability zero — there is a lower
limit for the probability, decreasing to 0 when k approaches 1.
In all examples of quantum fractals in this book we pay no attention
to the phase. Quantum fractals are drawn on the Bloch sphere, or on its
stereographic projection. Iterated function systems are implemented by
Lorentz transformations. All information about the phase is lost. And
indeed in the standard quantum theory the global phase of the state vector
is irrelevant. In fact the fact that such a global phase is unobservable, and
thus unimportant, is at the basis of “gauge invariance” (or U (1) symmetry),
which leads, via Noether’s theorem, to “electric charge conservation”. The
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 67/2

What are Quantum Fractals? 67

Fig. 2.28 Two isosurfaces of constant probability: one for spin z-component (torus),
and one for spin x-component (deformed torus, closing at infinity).

conservation of the electric charge is considered to be one of the basic laws

of physics.
Yet, as noted in, for instance, [Lämmerzahl et al. (2005)], no unique
tests of charge conservation have been found and physicists have reasons
for speculating about the possible limits and deviations from the strict
charge conservation law, especially when one of the versions of theories of
gravity starts to play a role in quantum processes (cf. also [Dolgov et al.
(2002)]). Thus, at some point in the future, the so far essentially neglected
phase may become important in some physical effects.
In this book we are formulating the process of generation of quantum
fractals in several ways, one of them is in terms of complex 2 × 2 matrices
from the group SL(2, C). Two proportional matrices, A and eiψ A, give
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 68/2

68 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.29 A family of 16 isosurfaces (half-planes) of constant phase ψ.

the same Lorentz matrices and the same quantum fractals on the Bloch
sphere. But, in principle, quantum fractals can be also drawn in the space
of spin state vectors — that is on the three-dimensional S 3 — or, using the
stereographic projection, in the three-dimensional Euclidean space R3 . In
such a case the phases contained in the complex matrices become important.
We did not implement this idea for the same reason we did not show higher
dimensional quantum fractals discussed in Sec. 3.3 — mainly because it
would need an implementation of a 3D graphical representation of fractals,
which constitutes a technical challenge.

2.7.7 Action on orthogonal projections and Möbius

transformations
Let n ∈ R3 , n2 = 1, and let Pn be defined as in Eq. (2.98). Let A be an
SL(2, C) matrix. Then APn A∗ is a Hermitian matrix with one-dimensional
range. Therefore it is proportional to an orthogonal projection.

Proposition 2.3. Let n be a vector on the unit sphere S 2 ∈ R3 , and let

Pn be the corresponding projection as in Eq. (2.98). With A ∈ SL(2, C) we
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 69/2

What are Quantum Fractals? 69

have
APn A∗ = λPn , (2.151)
where n = A · n, as in Eq. (2.96) and
1
λ = λ(A, n) = . (2.152)
L(A)0i ni + L00
The function λ : SL(2, C) × S 2 → R+ satisfies the cocycle condition
λ(A A, n) = λ(A , A · n)λ(A, n). (2.153)

Proof. Let n be the vector in R4 with coordinates (1, n). Then, with
n0 = 1, (cf. Eq. (2.98))
1 1
Pn = (I + ns ) = nμ sμ . (2.154)
2 2
Therefore (cf. Eq. (2.86))
1 μ
APn A∗ = n Asμ A ∗
2
1
= nμ Lνμ sν . (2.155)
2
Since s0 = I, and taking into account Eq. (2.61),we get:
1 L(A)iμ nμ si
APn A∗ = I+
L(A)0μ nμ L(A)0μ nμ
1
= Pn (2.156)
L(A)0i ni + L00
where
n = A · n = L(A) · n. (2.157)
Therefore
1
λ(A, n) = . (2.158)
L(A)0i ni + L00
The cocycle identity follows now from
(A A)Pn (A A)∗ = λ(A A, n)PA A·n , (2.159)
on one hand, while on the other hand, we have
∗
(A A)Pn (A A)∗ = A APn A∗ A
∗
= A λ(A, n)PA·n A
= λ(A, n)λ(A , A · n)PA A·n .

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 70/2

70 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

We can also parametrize the sphere (minus one point) using the stereo-
graphic projection parametrization by expressing the unit vector n in terms
of z = x + iy as in Eq. (2.112), then substitute for L(A) the explicit for-
mulas in Proposition 2.2. But we can also use the formula Eq. (2.109) to
obtain:
 
Proposition 2.4. With A = ac db ∈ SL(2, C), the factor λ in Eq. (2.158)
can be also written as
|az + b|2 + |cz + d|2
λ(A, z) = . (2.160)
1 + |z|2
Proof. In order to prove the statement we take the trace of both sides of
Eq. (2.151) and notice that the trace of a projection on a one-dimensional
subspace is 1. Thus we get:
λ = Tr (APz A∗ ) = Tr (A∗ APz ). (2.161)
Now we have that
|a2 | + |c|2 āb + c̄d
A∗ A = ¯ |b|2 + |d|2 . (2.162)
b̄a + dc
Then the direct calculation gives:
1  2 
λ= ¯ + |b|2 + |d|2 .
(|a| + |c|2 )|z|2 + (āb + c̄d)z̄ + (b̄a + dc)z
1 + |z| 2
(2.163)
The statement in the Proposition then follows by a direct calculation of
|az + b|2 + |cz + d|2 , and by checking that the result coincides with the
expression in the parenthesis. 
The linear fractional transformations z → (az +b)/(cz +d) implemented
by SL(2, C) matrices A are called Möbius transformations.

2.7.8 Exponential map in SL(2, C)

This section can be thought of as a preparation for the classification of
Möbius transformations that will be given in Sec. 2.7.10.
SL(2, C) is a Lie group. Suppose we have a path A(t) in the group,
with A(0) = I. For each element
a(t) b(t)
A(t) = (2.164)
c(t) d(t)
we then have
1 = det(A(t)) = a(t)d(t) − b(t)c(t). (2.165)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 71/2

What are Quantum Fractals? 71

Let us denote by a dot ˙ the derivative with respect to t. Taking the

derivative of both sides of the equation det(A(t)) = a(t)d(t) − b(t)c(t) = 1
at t = 0, and taking into account the fact that A(0) = I, i.e. a(0) = d(0) =
˙
1, c(0) = b(0) = 0, we obtain ȧ(0) + d(0) = 0. Therefore the tangent vector
ȧ(0) ḃ(0)
(2.166)
˙
ċ(0) d(0)
to a path in SL(2, C) is a matrix of trace zero.
Traceless matrices form the Lie algebra sl(2, C) of SL(2, C). They are
generators of one-parameter subgroups. If X is a traceless matrix, Tr (X) =
0, then, owing to the general identity, valid for any matrix:
det(A) = exp(Tr (A)), (2.167)
we have
det(exp X) = 1. (2.168)
While every traceless matrix can be exponentiated to give a matrix in
SL(2, C), it does not necessarily follow that every matrix in SL(2, C) can
be obtained this way. In fact, as we will now see, it is “almost the case”,
but there is an exceptional class of SL(2, C) matrices.

2.7.8.1 Explicit form of the exponential of a traceless matrix

Let us start by looking at the form of eX for a traceless matrix X. The
exponential is given by the power series
X2 X3
exp(X) = I + X + + + .... (2.169)
2! 3!
Remark 2.10. Owing to the fact that our matrices are finite-dimensional,
the power series above is always absolutely convergent.
In fact, the sum of the power series for exp(X), X ∈ sl(2, C), can be com-
puted explicitly owing to the following lemma.
Lemma 2.1. If X ∈ sl(2, C) then X 2 = − det(X)I.
Proof. If X is traceless, it can be written as
p q
X= . (2.170)
r −p
Taking its square we obtain
p2 + qr 0
X2 = = (p2 + qr)I. (2.171)
0 p2 + qr
But p2 + qr = − det(X). Therefore we find that X 2 is proportional to the
identity matrix, the proportionality coefficient being − det(X). 
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 72/2

72 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Once we have that X 2 is proportional to I, we have X 3 being propor-

tional to X, X 4 again proportional to I, etc. The whole series for exp(X)
will thus be of the form
exp(X) = αI + βX, (2.172)
where α and β are complex numbers that we will now compute.
• If det(X) = 0, then only the first two terms survive
exp(X) = I + X. (2.173)
• If det(X) = 0, let us denote by ω one of the two square roots of
− det(X) = p2 + qr:

ω = − det(X). (2.174)
Then we have:
∞
 Xk
eX =
k!
k=0
∞ ∞
 X 2k+1
X 2k
= +
(2k)! (2k + 1)!
k=0 k=0
∞ 2k ∞
ω ω 2k
= I+ X
(2k)! (2k + 1)!
k=0 k=0
∞
 ω 2k
= cosh(ω) I + X.
(2k + 1)!
k=0

Since we are considering now the case of det(X) = 0, we can write the
series in the second term as sin(ω)/ω, and we obtain
sinh ω
eX = cosh(ω) I + X. (2.175)
ω
Notice that it does not matter which of the two square roots of det(X),
that differ by a sign, we use in this formula. Moreover, with understand-
ing that sin(0)/0 = 1, the last formula applies to the case of det(X) = 0
as well.

Remark 2.11. Notice that the argument ω of the sinh and cosh functions
is, in general, a complex number. If z = x + iy, where x, y are real, then
cosh z = cosh x cos y + i sinh x sin y,
sinh z = sinh x cos y + i cosh x sin y.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 73/2

What are Quantum Fractals? 73

2.7.8.2 The eigenvalue problem

Now, let us return to our original question, namely when a given matrix A
from SL(2, C) can be written as an exponential of a traceless matrix X?
To this end it proves to be useful to discuss first the eigenvalue problem of
A. So, suppose that A can be written as
A = exp(X) = αI + βX (2.176)
— cf. Eq. (2.172). Then, taking trace of both sides, and taking into account
the fact that Tr (X) = 0, Tr (I) = 2, we obtain α = Tr (A)/2. Therefore,
assuming that A = exp(X), we must have
Tr (A)
A− I = βX. (2.177)
2
It remains to find the coefficient β and our problem is solved. But can β
always be found?
In order to answer this question, and also in order to find β, whenever
it exists, let us study in detail the eigenvalue problem of the matrix A. We
are looking for nonzero vectors u in C2 with the property
Au = λu, λ ∈ C. (2.178)
The characteristic equation, the necessary condition for the eigenvalue
equation to have a solution, is
det(A − λI) = 0, (2.179)
that is,
a−λ b
0 = det = (a − λ)(d − λ) − bc. (2.180)
c d−λ
Taking into account the fact that 1 = det(A) = ad − bc, we can eliminate
bc from the characteristic equation to get the following simple form:
λ2 − (a + d)λ + 1 = 0. (2.181)
The discriminant Δ for our equation is
Δ = (a + d)2 − 4, (2.182)
and we have the solutions: √
a+d± Δ
λ± = . (2.183)
2
From the Vieta’s formulas we know that λ+ + λ− = a + d = Tr (A), while
λ+ λ− = 1. Denoting λ+ by λ we have
λ + 1/λ = Tr (A). (2.184)
Consider first the case when Δ = 0, i.e. Tr (A) = 2, or Tr (A) = −2. In this
case λ+ = λ− ; therefore, from λ+ λ− = 1, we have that λ = ±1.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 74/2

74 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

2.7.8.3 The case of Δ = 0, Tr (A) = 2

If Tr (A) = 2, then, from Eq. (2.183), λ = 1, since Tr (A − I) = Tr (A) −
Tr (I) = 2 − 2 = 0, we have that A − I is traceless and of determinant zero.
It follows from Lemma 2.1 that (A − I)2 = 0, and therefore exp(A − I) =
I + (A − I) + A−I
2! + ... = A. It follows that in this case A is the exponential
of the traceless nilpotent (i.e. of square zero) matrix X = A − I.
A = eX , X = A − I. (2.185)
As an example we can take
11
A= . (2.186)
01
Then, evidently, det(A) = 1 and Tr (A) = 2. In the next Proposition we
will show that every matrix with det(A) = 1 and Tr (A) = 2, if not the
identity matrix, is similar to the matrix above.

Proposition 2.5. Given a matrix A ∈ SL(2, C) with Tr (A) = 2, A = I,

there exists a matrix S ∈ SL(2, C) such that
11
A=S S −1 . (2.187)
01
Proof. With X = A − I as above let u be a nonzero eigenvector of A:
Au = u. Then Xu = 0. Let v be any vector linearly independent of u. Then
Xv = 0, otherwise X would vanish on two linearly independent vectors, u
and v; thus X = 0, i.e. A = I, the case we have excluded. Now, we claim
that Xv must be proportional to u. Indeed, we have that X(Xv) = X 2 v = 0
and if Xv is not proportional to u, X would then vanish on u and on Xv,
which would imply again X = 0. So we have Xv = cu, c = 0, and we can
always choose our v in such a way that c = 1. Let us write both vectors, u
and v as u = ( uu12 ) , v = ( vv12 ) , and let us define the matrix S as
u1 v1
S= . (2.188)
u2 v2
Since u and v are linearly independent, S is nonsingular. Then, since
Au = u, Av = (I + X)v = v + u, we obtain
u1 u1 + v1 11u1 v1 11
AS = = =S . (2.189)
u2 u2 + v2 01u2 v2 01

If necessary we can replace now S by S/ det(S), and we get the desired
result. 
June 16, 2014 8:52 World Scientific Book - 9in x 6in QuantumFractals3 pg. 75/2

What are Quantum Fractals? 75

It follows that the exceptional nontrivial SL(2, C) matrices with Tr (A) = 2

form up just one conjugacy class, with the matrix ( 10 11 ) as its typical rep-
resentative.

2.7.8.4 The case of Δ = 0, Tr (A) = −2

Consider now the second possibility, namely the case of Tr (A) = −2, when
both eigenvalues are −1. In this case A − Tr 2(A) = A + 1. There are two
cases there.
 −1 0 
If A = −I: The first case is that of A = −I, that
 is, A = 0 −1 is the
iπ 0
minus identity matrix. In this case X = 0 −iπ easily does the job:
iπ 0 eiπ 0 −1 0
exp = = . (2.190)
0 −iπ 0 e−iπ 0 −1

If A = −I: When A = −I, but Tr (A) = −2, then from A − Tr 2(A) I =

A + I = βX, we find that β = 0, and X = A+I
β . Therefore we should have
A+I
A=e β . (2.191)
Acting with both sides of this equation on the eigenvector u belonging to
the eigenvalue λ = −1 we deduce then that
−1+1
−1 = e β = e0 = 1, (2.192)
which is impossible. Therefore matrices A ∈ SL(2, C) with Tr (A) = −2
and both eigenvalues equal to −1 are not exponentials of traceless matrices.

2.7.9 Two different eigenvalues

Let us consider now the case of Tr 2 (A) = 4, when we have two different
eigenvalues λ, λ−1 , with λ = 1. Let u be the non-zero eigenvector Au = λu,
defined uniquely up to a constant multiplier. From A − Tr 2(A) I = βX
(Eq. (2.177)), we find that β = 0, and that, with α = β −1 , we have
1
X = α A − (λ + λ−1 ) , (2.193)
2
which entails the following identity valid on both eigenvectors
1
A = exp α (A − (λ + λ−1 ) . (2.194)
2
Acting with both sides of the last equation on u we obtain
1 −1
λ = exp α λ − (λ + λ−1 )
α
= e 2 (λ−λ ) . (2.195)
2
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 76/2

76 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

We can take now the logarithm of both sides to get

2λ log(λ)
α= . (2.196)
λ2 − 1
With this value of α we have the identity
1
A = eα(A− 2 Tr (A)I ) , (2.197)

providing the explicit form of X, verifiable by acting on eigenvectors of A.

We summarize our results in the following proposition.

Proposition 2.6. Let A be a matrix from SL(2, C), i.e. det(A) = 1. Then

i) If Tr (A) = ±2, then A has two different eigenvalues λ, λ−1 . With

2λ log(λ)
α= (2.198)
λ2 − 1
and
Tr (A)
X =A− I, (2.199)
2
we have Tr (X) = 0 and

A = eαX . (2.200)

ii) If Tr (A) = 2, then, with X = A − I, we have Tr (X) = 0, and

A = eX . (2.201)
 iπ 
iii) If Tr (A) = −2, and A = −I, then, with X = 0 −iπ 0
we have
X
Tr (X) = 0 and A = e .
iv) If Tr (A) = −2, and A = −I, then A is not an exponential of a traceless
matrix.

2.7.10 Classification of Möbius transformations

Möbius transformations are classified according to the value of the trace
of implementing matrices [Ford (1979); Ahlfors (1979)]. A matrix A in
SL(2, C) is called

i) loxodromic if Tr (A) ∈ [−2, 2],

ii) elliptic if Tr (A) ∈ (−2, 2),
iii) parabolic if Tr (A) = 2 or Tr (A) = −2.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 77/2

What are Quantum Fractals? 77

Classification of SL(2,C) matrices by their trace

5

3
loxodromic
2

1
)
-1

−2.0 2.0
( +

0
hyperbolic elliptic hyperbolic
-1

-2

-3
parabolic
-4

-5
-5 -4 -3 -2 -1 0 1 2 3 4 5
-1
( + )

Fig. 2.30 Classification of SL(2, C) matrices and associated Möbius transformations.

Loxodromic transformations with real trace are called hyperbolic. Hyper-

bolic transformations represent relativistic boosts. Elliptic transformations
represent three-dimensional rotations. Parabolic transformations are dis-
tinguished by the fact that they have just one fixed point on the Riemann
sphere. All other transformations (except of the identity) have exactly two
fixed points.
 
Proposition 2.7. Let A = ac db be in SL(2, C). If A is parabolic, then A
has exactly one fixed point on the sphere. If c = 0, then the fixed point is
given in the complex plane representation by the formula:

a−d
z0 = . (2.202)
2c
If c = 0, then A is necessarily of the form Eq. (2.70) and its fixed point is
∞ corresponding to the north pole n = (0, 0, 1) of the sphere. If A is not
parabolic, then A has exactly two fixed points. If c = 0, then the fixed points
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 78/2

78 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

are
a−d+ Tr 2 (A) − 4
z1 = , (2.203)
2c
a−d− Tr 2 (A) − 4
z2 = . (2.204)
2c
If c = 0, then one of the fixed points is
b
z1 = , (2.205)
d−a
the second one is
z2 = ∞. (2.206)

Proof. The fixed point equation is

az + b
= z, (2.207)
cz + d
or
cz 2 + (d − a)z − b = 0. (2.208)
• Suppose c = 0. Then the discriminant of the quadratic equation is
Δ = (a − d)2 + 4bc = a2 + d2 − 2ad + 4bc = (a + d)2 − 4ad + 4bc
= Tr 2 (A) − 4. (2.209)
If A is parabolic, that is if Tr 2 (A) = 4, we have Δ = 0 and there is just
one solution, as in Eq. (2.202). Otherwise we have two solutions as in
Eq. (2.205),Eq. (2.206).
• Suppose, on the other hand, that c = 0. If A is parabolic, then a +
d = 2 or a + d = −2. Since ad = 1, we have only two solutions:
a = d = 1, which gives the parabolic family Eq. (2.70) with ∞ as the
only fixed point, or a = d = −1, which gives essentially the same family
of transformations.


2.7.11 Area transformation law

Let us equip the sphere S 2 with the standard surface area element. In
spherical coordinates, if n is given by the formula
n1 = sin θ cos φ
n2 = sin θ sin φ (2.210)
3
n = cos θ
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 79/2

What are Quantum Fractals? 79

the infinitesimal area element dS is given by the formula

dS = sin2 θ dθ dφ. (2.211)

If we use the stereographic projection Eq. (2.112) onto the complex plane
z = x + iy, we get
sin θ cos φ
x=
1 − cos θ
sin θ sin φ
y= (2.212)
1 − cos θ
From this we can calculate the Jacobian determinant
 
 ∂x ∂x 
 ∂θ ∂φ  sin θ
dxdy =  ∂y ∂y  dθ dφ = dθ dφ. (2.213)
 ∂θ ∂φ  (1 − cos θ)2

It follows now from Eq. (2.211) that

dS = (1 − cos θ)2 dxdy. (2.214)

On the other hand

2 2
sin θ cos φ sin θ cos φ
1 + x2 + y 2 = 1 + +
1 − cos θ 1 − cos θ
sin2 θ
= 1+
(1 − cos θ)2
(1 − cos θ)2 + sin2 θ
= (2.215)
(1 − cos θ)2
2(1 − cos θ) 2
= = .
(1 − cos θ)2 1 − cos θ
Therefore from Eq. (2.211) we finally get
4
dS = dxdy. (2.216)
(1 + x2 + y 2 )2
In the following part of this section we will slightly change our assump-
tions. Namely, we will assume that A is an arbitrary nonsingular matrix —
A ∈ GL(2, C). The reason for this is the following: taking into account the
constraint det(A) = 1 in calculations can be tricky. It is more straightfor-
ward to deal with matrices without such constraints. The price for it will
be that our expressions will contain, here and there the term det(A). We
can then set it to 1 in our final expressions.
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 80/2

80 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Proposition 2.8.
 
(1) Let A = ac db be a GL(2, C) matrix. Let us denote
D = det(A) = ad − bc = 1.. (2.217)
Let A : z → z  = A · z = az+b
cz+d
be the corresponding linear fractional
transformation (Möbius transformation) of the complex plane. The
transformation induces the associated transformation of the sphere S 2
via stereographic projection. Let dS  be the transformed surface area
element of the sphere. Then
2
(1 + |z|2 )
dS  = D2 dS. (2.218)
|az + b|2 + |cz + d|2
In particular, if det(A) = 1, then
dS  1
= . (2.219)
dS λ(A, z)2
where λ(A, z) is given by Eq. (2.160).
(2) Denote by α, β, β̄, δ the matrix elements of the Hermitian and positive
matrix A∗ A:
αβ
A∗ A = . (2.220)
β̄ δ
Then
2
dS  1 + |z|2
= . (2.221)
dS α|z|2 + β̄z + β z̄ + δ
Therefore dS  /dS depends only on the matrix A∗ A and it is invariant
under the replacements of the type A → U A, where U is in SU (2).

Proof.
(1) Let z → f (z) be the linear fractional transformation of the complex
plane determined by A
az + b
.
f (z) = (2.222)
cz + d
Then the complex Jacobian JC of the transformation is
a(cz + d) − c(az + b) ad − bc D
JC = f  (z) = 2
= 2
= . (2.223)
(cz + d) (cz + d) (cz + d)2
The complex Jacobian can be also considered as a real Jacobian JR ,
where we express the derivative with respect to the complex variable z
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 81/2

What are Quantum Fractals? 81

in terms of partial derivatives with respect to x and y. There is a general

relation between real and complex Jacobians (cf. [Cross (2008)]):
|JR | = |JC |2 . (2.224)
Therefore the area transformation law in terms of x, y variables is (cf.
also [Ford, 1979, p. 24, Thm 16]):
D2
dx dy  = dxdy. (2.225)
|cz + d|4
Now, using Eq. (2.216), we have
4 4D2
dS  = dx
dy 
= dxdy
(1 + |z  |2 )2 (1 + |z  |2 )2 |cz + d|4
D2 (1 + |z|2 )2
= dS
(1 + |z  |2 )2 |cz + d|4
D2 (1 + |z|2 )2
=   2 dS
 az+b 2
1 +  cz+d  |cz + d| 4

D2 (1 + |z|2 )2
= dS (2.226)
(|cz + d|2 + |az + b|2 )2
The result follows now by comparing with Eq. (2.160).
(2) The second statement follows by noticing that
ā c̄ ab |a|2 + |c|2 āb + c̄d
A∗ A = = . (2.227)
b̄ d¯ cd ab̄ + cd¯ |b|2 + |d|2
Therefore
α = |a|2 + |c|2 ,
β = āb + c̄d, (2.228)
¯
β̄ = ab̄ + cd,
δ = |b|2 + |d|2 ,
while
|az + b|2 + |cz + d|2 =
(2.229)
¯ + (āb + c̄d)z̄ + |b|2 + |d|2 .
(|a|2 + |c|2 )|z|2 + (ab̄ + cd)z


Corollary 2.1. Given A ∈ SL(2, C), and with the notation as in proposi-
tion 2.8 let

w(z) ≡ dS  /dS. (2.230)
Then w(z) ≡ 1 if and only if A is unitary, i.e. if and only if A ∈ SU (2).
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 82/2

82 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Proof. We have

D(1 + |z|2 )
w(z) = . (2.231)
α|z|2 + β̄z + β z̄ + δ
If A is unitary, then A∗ A = I, therefore α = δ = 1, β = 0. It follows that
w(z) ≡ 1. Conversely, assume that w(z) ≡ 1. In fact it is enough to assume
that w(z) = 1 at three points, namely z = 0, z = 1, z = −1. Therefore
w(z) = 1 is equivalent to

1 + |z|2 = α|z|2 + β̄z + β z̄ + δ. (2.232)

With z = 0 we instantly get δ = 1. Thus

|z|2 = α|z|2 + β̄z + β z̄. (2.233)

Then from det(A∗ A) = | det(A)|2 = 1, we have that α = 1 + β 2 , therefore

|β|2 |z|2 + β̄z + β z̄ = 0. (2.234)

Putting now z = 1 and z = −1 into this equation we get |β|2 = −|β|2 ,

therefore β = 0 and thus α = 1. It follows that A is unitary. 

Proposition 2.9. With A ∈ GL(2,  C) let w(z) be as in Eq. (2.230),

Eq. (2.231), and let A∗ A = αγ βδ , then

(1) If A is not unitary, then the function w(z) has maximum and minimum
at the points z1 , z2 which are fixed points of the matrix A∗ A. If det(A) =
1, then the extremal values w+ , w− of w(z) are equal to the eigenvalues
of A∗ A.
(2) Assuming det(A) = 1 and w1 ≤ w ≤ w2 , the loci of points satisfying
the equation

w(z) = w, (2.235)

are circles of centers at

wβ
z0 = , (2.236)
1 − wα
and radii

w(α + δ) − w2 − 1
R(w) = . (2.237)
|1 − wα|
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 83/2

What are Quantum Fractals? 83

Proof. 1) Consider the function w(z):

D(1 + |z|2 )
w(z) = . (2.238)
α|z|2 + β̄z + β z̄ + δ
The denominator α|z|2 + β̄z + β z̄ + δ has the minimum at z = −β/α and its
value is (αδ − |β|2 )/α = D2 /α. Therefore w(z) ≤ α(1 + |z|2 ). Thus w(z) is
finite for all z ∈ C. In order to find the points of the complex plane where
the function attains its maximum and minimum values, we write down the
condition ∂w(z)/∂ z̄ = 0 that is necessary for the extremum. We obtain
this way the following equation:
β̄z 2 + (δ − α)z − β = 0. (2.239)
This equation is exactly the same as that for fixed points of A∗ A:
αz + β
= z. (2.240)
β̄z + δ
If β = 0, the solution is z1 = 0. The value of w(z) at this point is D/δ.
If, additionally, D = 1, we have that αδ = 1, therefore the extremal value
of w(z) at this point is 1/δ = α — one of the eigenvalues of the diagonal
matrix A∗ A. The other extremum is at z = ∞, in which case w(z) = δ —
the other eigenvalue. 
Now, suppose β = 0. Since w(z) = dS  /dS, and since unitary transfor-
mations preserve the areas, we can always diagonalize A∗ A by a unitary
transformation without affecting the extremal values of w(z). Therefore we
can reduce the case to that of β = 0. It follows that, also in the case of
β = 0, the extremal values of w(z) are the eigenvalues of A∗ A, that is, they
are solutions of the equation det(A∗ A − wI) = 0, or, explicitly
w2 − (α + δ)w + 1 = 0. (2.241)
Denoting
τ = Tr (A∗ A) = α + δ (2.242)
we obtain two solutions
√
τ2 − 4
τ+
w+ = , (2.243)
2
√
τ − τ2 − 4
w− = . (2.244)
2
These correspond to the two solutions of Eq. (2.239):
√
α − δ − τ2 − 4
z+ = , (2.245)
2β̄
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 84/2

84 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

√
α − δ + τ2 − 4
z− = . (2.246)
2β̄
The reason for such correspondence will become clear from the second part
of the proof.
2) The equation w(z) = w can be written as
wβ̄ wβ D − wδ
|z|2 − z− z̄ + = 0. (2.247)
D − wα D − wα D − wα
This is the same as the equation of a circle of radius R and center z0 :
|z − z0 |2 = R2 , (2.248)
or
z 2 − z¯0 z − z0 z̄ + |z0 |2 − R2 = 0, (2.249)
if we identify:
wβ
z0 = . (2.250)
D − wα
D − wδ
R2 = |z0 |2 − . (2.251)
D − wα
Putting now w = w+ and w = w− in Eq. (2.250) we get, after some
calculation and taking into account that D = 1, δ = τ − α, the formulas
Eq. (2.246). Let us first analyze the case when R = 0. These should be
the extremal points of w(z). And indeed, we get exactly the Eq. (2.250).
Otherwise, assuming D = 1, we get the formula for the radius:

1 − wδ
R(w) = |z0 |2 −
1 − wα

w(α + δ) − w2 − 1
= . (2.252)
|1 − wα|
Notice that the expression under the square root is always nonnegative.
This follows from the fact that R(w) vanishes only at the extremal points
and it is positive at w = τ /2. 

Example 2.5. Let us take, as an example,

√ √
1 2+1 2−1
A= √  √ . (2.253)
2 2−1 5 2+1
Then A is in SL(2, C) and
1 11
A∗ A = √ . (2.254)
2 13
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 85/2

What are Quantum Fractals? 85

The matrix A∗ A has then the eigenvalues w− = 0.414214, w+ = 2.41421

and the associated Möbius transformation has fixed points at z− =
0.414214 + 0i, z+ = −2.41421 + 0i. The contours of w(z)2 are plotted in
Fig. 2.31. Notice that w−2
≈ 0.2, w+
2
≈ 5.2. If we replace A by U A, with
∗
any U in SU 2, then A A will be the same, therefore w− , w+ , z− , z+ will
also be the same.

Area Ratio Contour Lines

5 1.9 1.3

3
3.6
2

1
4.7
5.3
0 5.8 0.2
4.1
-1

-2 0.7

-3

-4 3.0

-5 2.4
-5 -4 -3 -2 -1 0 1 2 3 4 5

Fig. 2.31 Contour plot of w(z)2 for A∗ A from Example 2.5.

2.7.12 Relativistic aberration

Mathematically Minkowski space is modeled by an affine space M based
on a vector space V endowed with a flat metric with signature (3, 1) (or
(1, 3)). Inertial frames correspond to selections of an origin O in M and of
an orthonormal basis in V. Assuming that an origin and an orthonormal
basis have been selected, Minkowski’s space-time can be identified with R4 ,
and elements of the matrix group SO+ (3, 1) can be interpreted as changing
the inertial frame to another frame, say associated with a moving space-
ship, under the assumption that such a change preserves the origin of the
coordinate system.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 86/2

86 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Lorentz transformations can be decomposed into “relativistic boosts”

and “space rotations”. Such a decomposition depends on the inertial frame.
In the following we will assume that an inertial has been fixed. If
we change this frame, all our decompositions will transform by a similarity
transformation within SO(3, 1).
It is algebraically simpler to work with the decompositions into boosts
and rotations within SL(2, C) rather than within SO + (3, 1). Therefore we
will follow this path. Let us start with recalling the polar decomposition
theorem (cf. e.g. [Axler, 1997, p. 153]) in its simplified version valid for
nonsingular matrices:

Proposition 2.10. Let A be a nonsingular complex matrix, A ∈ GL(2, C).

Then A can be uniquely represented as a product
A = U T, (2.255)
∗ ∗
where U is a unitary matrix U U = U U = I, and T is a Hermitian and
positive matrix. T is defined as the positive square root of A∗ A:
T = (A∗ A) 2 ,
1
(2.256)
while U is then defined as
U = AT −1 = A(A∗ A)− 2 .
1
(2.257)
If A is in SL(2, C) then T is also in SL(2, C) and U is in SU (2).
Remark 2.12. The decomposition above can also be written in a different
order, as
A = T U , (2.258)
where
T  = U T U ∗ = (AA∗ ) 2 , U  = U,
1
(2.259)
but we will mainly use the form as in Eq. (2.10).

The positive factor T represents a Lorentz boost, while U describes space

rotation of (x, y, z) axes of the reference frame, following after the boost.
In special relativity a boost is characterized by a direction n and a
velocity change v, v ∈ R, in this direction (assuming the speed of light
c = 1.) We will choose n as the direction pointing backward.
Rotation is characterized by an axis direction n and an angle ϕ of anti-
clockwise rotation around this axis. It should be noticed that changing
the signs of n and v is not affecting the boost itself. Therefore, in order
to preserve the uniqueness of the description we will assume that, for a
nontrivial boost, v should be always positive.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 87/2

What are Quantum Fractals? 87

The first thing to do is to be to decode these parameters from the

matrices T and U.

2.7.12.1 Relativistic boosts

We will start with boosts by establishing the standard form of a positive
matrix from SL(2, C). It is important here to notice that every Hermi-
tian matrix X is similar to a diagonal matrix D by a unitary matrix V :
V XV ∗ = D. Then Tr (X) = Tr (D), and det(X) = det(D). If X is a posi-
tive matrix from SL(2, C), then also D is positive and in SL(2, C), therefore
D is of the form
x0
D= , x > 0. (2.260)
0 x1
Then Tr (X) = Tr (D) = x + 1/x. The function f (x) = x + 1/x, x > 0 has
its minimum equal 2 at x = 1 — cf. Fig. 2.32. For x = 1, D (therefore
also X) becomes the identity matrix. For nontrivial matrices the trace is
always greater than 2 — cf. Fig. 2.32.

Fig. 2.32 Trace of a diagonal positive SL(2, C) matrix.

Another thing to notice about positive SL(2, C) matrices is that there is

a simple relation between Tr (X) and Tr (X 2 ). It is enough to establish this
relation for diagonal matrices because V that diagonalizes X, diagonalizes
also X 2 , and because trace is unitarily invariant. Denoting by τ the trace
of X and by s the trace of X 2 we have:
√
τ = s + 2, s = τ 2 − 2. (2.261)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 88/2

88 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Indeed, for a diagonal matrix D = diag (x, 1/x) we have τ = Tr (D) =

x + 1/x, and s = Tr (D 2 ) = x2 + 1/x2 . The relations in Eq. (2.261) follow
now by a straightforward verification.

Proposition 2.11. Let X be a nontrivial (i.e. different from the identity

matrix) positive matrix in SL(2, C). Then X can be uniquely represented
in the form
1
X= √ (I + k n · s), (2.262)
1 − k2
where 0 < k < 1, n = (n1 , n2 , n3 ) is a unit vector, n2 = 1, and s =
(s1 , s2 , s3 ) are Pauli’s spin matrices — cf. Eq. (2.76).

Proof. Let us write X = xμ sμ . Then, owing to the fact that X is positive,

Tr (X) = 2x0 > 2, we have that x0 > 2. From 1 = det(X) = (x0 )2 − x2 it
follows that
x
n=  (2.263)
(x )2 − 1
0

is a unit vector. Let us define k by the formula:


(x0 )2 − 1
k= . (2.264)
x0
Then 0 < k < 1 and
1
x0 = √ . (2.265)
1 − k2
The rest is a simple algebra verification. 

Remark 2.13.  With the assumptions  as in the Proposition above,

 if we
write X = ac db , then a = (1 + kn3)/ (1 − k 2 ), d = (1 − kn3 )/ (1 − k 2 ),
therefore, from 0 < k < 1 and −1 ≤ n3 ≤ 1, it follows that a > 0 and
d > 0.

We will show now that the vector n has the physical meaning of the direc-
tion from which the transformed reference frame is running away, while k
is a simple function of the velocity v of the boost described by X.
 
Proposition 2.12. Let X = ac db , a, d > 0, c = b̄ be a non-trivial (i.e.
X = I) positive matrix from SL(2, C), and let us write X 2 as
αβ
X2 = . (2.266)
β̄ δ
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 89/2

What are Quantum Fractals? 89

Let

s = Tr (X 2 ) = α + δ. (2.267)

Then X describes a boost with the velocity v given by

s2 − 4
v= , (2.268)
s
in the direction opposite to the unit vector n given by
2(β)
n1 = √ ,
s2 − 4
2(β)
n2 = √ , (2.269)
s2 − 4
α−δ
n3 = √ .
s2 − 4
Proof. Let us consider a Lorentz transformation between two inertial
systems xμ and x :
μ

x = Lμν xν , (μ = 0..3).
μ
(2.270)

The origin O with coordinates x0 , 0, 0, 0 has the transformed coordinates

x = L00 x0
0
(2.271)
x = Li0 x0 , (i = 1, 2, 3)
i
(2.272)

The velocity vector vi of O with respect to the moving system is then

x
i
Li0
vi = = , (2.273)
x 0 L00
and its square is
(L10 )2 + L20 )2 + L13 )2
v2 = . (2.274)
(L00 )2
Now, owing to the equation Lt ηL = η we have

(L10 )2 + (L20 )2 + (L13 )2 = (L00 )2 − 1. (2.275)

Therefore
(L00 )2 − 1
v2 = . (2.276)
(L00 )2
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 90/2

90 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Using Eq. (2.91) in proposition 2.2 and Eq. (2.227) we get

s
L00 = , (2.277)
2
and therefore
s2 − 4
v2 = (2.278)
s2
Since we decided to have v positive, we obtain
√
s2 − 4
v= , (2.279)
s
2
s= √ . (2.280)
1 − v2
Consider now the component v1 :
L10
v1 = . (2.281)
L00
From Eq. (2.91) we get
¯ = a(c̄) + d(b) = (āb + c̄d) = (β).
L10 = (ac̄ + bd) (2.282)

Therefore
2(β)
v1 = (2.283)
s
and
v1 2(β)
n1 = =√ . (2.284)
v s2 − 4
Similarly11 : v 2 = L20 /L00 , and

L20 = (ac̄ + bd)

¯ = (β), (2.285)

|a|2 + |b|2 − |c|2 − |d|2 a2 − d2 α−δ

L30 = = = . (2.286)
2 2 2

It is important to notice that the vector n points from the moving system
to the original system. So it has the direction opposite to the direction of
the boost.
11 Owing to our choice of notation a confusion may arise, since v 2 can denote either

the second component of the velocity vector, or the square of its norm. The intended
meaning should, however, always be clear from the context.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 91/2

What are Quantum Fractals? 91

2.7.12.2 Unitary 3D rotations

Let now U be a matrix from SU (2). Every such matrix has a standard
form.

Proposition 2.13. Every matrix U ∈ SU (2), U = ±I,

a b
U= , |a|2 + |b|2 = 1, (2.287)
−b̄ ā
has a unique representation in the form
ϕ ϕ n3 n1 + in2
U = cos I + i sin , (2.288)
2 2 n − in
1 2
−n3
where 0 < ϕ < 2π, and n = (n1 , n2 , n3 ) is a unit vector. We have
Tr (U )
ϕ = 2 arccos , (2.289)
2
(b)
n1 = , (2.290)
sin ϕ2
(b)
n2 = − , (2.291)
sin ϕ2
(a)
n3 = . (2.292)
sin ϕ2
Proof. We first notice that U is necessarily of the form given in
Eq. (2.72):
a b
U= , |a|2 + |b|2 = 1. (2.293)
−b̄ ā
It follows that Tr (U ) = a + ā is real and |Tr (U )| ≤ 2. Moreover, Tr (U ) =
±2 implies U = ±I. Therefore, with our assumptions, −1 < Tr (U )/2 < 1.
We can define then unambiguously
Tr (U )
ϕ = 2 arccos , (2.294)
2
with 0 < ϕ < 2π. With ϕ being determined we can now define n1 , n2 , n3
leading to Eq. (2.288).
Every unitary matrix can be diagonalized by an SU (2) similarity trans-
formation. Therefore let us restrict ourselves to a diagonal matrix corre-
sponding to n1 = n2 = 0, n3 = 1:
iϕ
e 2 0
U= −ϕ , (2.295)
0 e 2
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 92/2

92 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

iϕ
with a = e 2 , b = 0. From Eq. (2.91) we get the corresponding Lorentz
transformation:
x = x1 cos ϕ − x2 sin ϕ,
1
(2.296)
2 1 2
x = x sin ϕ + x cos ϕ, (2.297)
3 3
x =x , (2.298)
x = x0 .
0
(2.299)
This is an anticlockwise rotation around the axis n = (0, 0, 1) by the angle
ϕ. 

2.7.12.3 Aberration formula

Let us consider the special case of a diagonal positive SL(2, C) matrix
p0
X= , p > 0, p = 1. (2.300)
0 p1

The corresponding Lorentz transformation Lμν , (μ, ν = 0, ..., 3), calculated

from Eq. (2.91) reads:
⎛   ⎞
1 2 1 p4 −1
2 p + p2 0 0
⎜ 2p2
⎟
⎜ 0 1 0 0 ⎟
⎜ ⎟
⎜ ⎟ (2.301)
⎜ 0 0 1 0 ⎟
⎝  ⎠
p4 −1
2p2 0 0 12 p2 + p12
2
We have, using Eq. (2.280), Tr (X 2 ) = p2 + 1/p2 = √1−v , from which it
√ 2
√
follows that p = 1 − v 2 /(1 − v). Therefore (p − 1)/2p = v/ 1 − v 2 ,
2 4 2

and L takes the form

⎛ 1 v
⎞
√ 0 0 √1−v
1−v 2 2
⎜ ⎟
⎜ 0 1 0 0 ⎟
⎜ ⎟. (2.302)
⎜ 0
⎝ 0 1 0 ⎟ ⎠
√ v 0 0 √1−v 1
1−v 2 2

This is a relativistic boost in the negative direction of the third axis, that is
in the direction of the vector (0, 0, −1). The corresponding transformation
of the sphere (cf. Eq. (2.61)) reads now
√
1 1 − v2 1
n = n , (2.303)
1 + vn3
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 93/2

What are Quantum Fractals? 93

√
2 1 − v2 2
n = n , (2.304)
1 + vn3
n3 + v
n =
3
. (2.305)
1 + vn3
Now, the cosine of the angle θ between the vector n and the boost direction
(0, 0, −1) is −n3 , while the cosine of the angle θ between the transformed
3
vector and the boost direction is −n = √n1+vn
3 +v
3
. Therefore we obtain the
well known formula for the relativistic aberration:
cos θ − v
cos(θ ) = . (2.306)
1 − v cos θ
As an example, the result of the aberration corresponding to the boost
velocity v = 0.992 are shown in Fig. 2.33. The denominator in Eq. (2.306)
is responsible for the relativistic correction of the pre-relativistic, Galilei-
Newton, aberration formula based on simple adding of velocities. As the
result all incoming rays directions of the stars (except the one exactly at the
south pole) move towards the boost direction. In their book “Relativistic
Celestial Mechanics of the Solar System” S. Kopeikin and M. Efroimsky
put it this way [Kopeikin and Efroimsky, 2012, p. 150]:
“One astronomical consequence of this is that if all stars in
the sky were distributed uniformly over the celestial sphere, the
observer moving with ultra-relativistic speed would see the stars
displaced toward the point on the sky in the direction of its motion.
In the limit of V → c, the entire stellar sky of the observer would
shrink to a single bright point embracing all stars in the sky. It is
really impossible to use stars for navigating the spaceship moving
with such an ultra-relativistic speed!

Fig. 2.33 Relativistic aberration. Boost with velocity v = 0.992 towards the north pole.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 94/2

94 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

2.7.13 Example: Special subgroup of parabolic

transformations
Matrices
1z
A(z) = , z = x + iy = r (cos(ψ) + i sin(ψ)) ∈ C, (2.307)
01
form a special Abelian subgroup of parabolic transformations. We have
A(z)A(z  ) = A(z + z  ), A(z)−1 = A(−z). (2.308)
The action of A(z) on C (cf. Eq. (2.44)) is a pure complex translation:
A(z) · z  = z  + z. (2.309)
All A(z) share the same unique fixed point in C — the infinity. The corre-
sponding Lorentz matrix (cf. Eq. (2.91) on p. 53) Lμν , (μ, ν = 1, 2, 3, 4) is
easily computed:
⎛ ⎞
1 0 −x x
⎜0 1 −y ⎟
⎜ y ⎟
L(x, y) = ⎜ 
⎜x y 1 −x2 − y 2 + 2  1
 2 2
 ⎟.
⎟ (2.310)
⎝ 2 2 x +y ⎠
 2   2 
2 −x − y
1 2 1 2
x y 2 x +y +2

Proposition 2.14. Each matrix A(z) has the following explicit polar de-
composition A(z) = U (z)T (z), with U, T ∈ SL(2, C), where U is unitary
and T is positive:
1 2 z
U (z) =  , (2.311)
4 + |z|2 −z̄ 2

1 2 z
T (z) =  . (2.312)
4+ |z|2 z̄ 2 + |z|2

Proof. We first calculate A(z)∗ A(z):

1 z
A(z)∗ A(z) = . (2.313)
z̄ 1 + |z|2
The matrix T (z) in Eq. (2.312) has positive trace τ (z)
4 + |z|2 
τ (z) =  = 4 + z 2 > 0, (2.314)
4 + |z|2
and determinant det(T (z)) = 1:
2(2 + |z|2 ) − |z|2
det(T (z)) = = 1. (2.315)
4 + |z|2
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 95/2

What are Quantum Fractals? 95

Therefore T (z) is positive. Moreover, we easily verify that T (z)2 =

A(z)∗ A(z), therefore T (z) = (A(z)∗ A(z))1/2 .
Since T (z) is in SL(2, C) we can use Eq. (2.67) to get its inverse:

1 2 + |z|2 −z
T (z)−1 =  . (2.316)
4 + |z|2 −z̄ 2

We can now easily compute U (z) = A(z)T (z)−1 with the following result:
1 2 z
U (z) =  . (2.317)
4 + |z|2 −z̄ 2

Now U (z) ∈ SU (2), T (z) ∈ SL(2, C) is positive, and A(z) = U (z)T (z). 

Let us decode the transformations corresponding to T (z) and U (z) in

terms of the boost direction, velocity, rotation direction, and rotation angle.
The velocity v(z) of the boost can be read from Eq. (2.268). We have

s(z) = Tr (T (z)2 ) = Tr (A(z)∗ A(z)) = 2 + |z|2 , (2.318)

  √
s(z)2 − 4 |z| 4 + |z|2 r 4 + r2
v(z) = = = . (2.319)
s(z) 2 + |z|2 2 + r2
The boost direction nb can be calculated from Eq. (2.270) on page 89:
2(z) 2 cos ψ
n1b (z) =  = √ , (2.320)
|z| 4 + |z|2 4 + r2
2(z) 2 sin ψ
n2b (z) =  = √ , (2.321)
|z| 4 + |z|2 4 + r2
−|z| −r
n3b (z) =  =√ . (2.322)
4+ |z|2 4 + r2
The Lorentz transformation matrix Lμν , (μ, ν = 1, 2, 3, 4), corresponding to
T (z), z = x + iy, can be calculated from Eq. (2.91):
⎛ ⎞
4+3x2 +y 2 x|z|2
2xy
− 4+|z| x
⎜ ⎟
4+|z| 2 4+|z| 2 2

⎜ 2xy ⎟
⎜ 4+x2 +3y 2 y|z|2
− 4+|z|2 y ⎟
⎜ 4+|z|2 4+|z|2 ⎟
⎜ ⎟ (2.323)
⎜ x|z|2 y|z|2 8+x4 +2y 2 +y 4 +2x2 (1+y 2 ) |z|2 ⎟
⎜ − 4+|z|2 − 4+|z|2 − ⎟
⎝ ⎠
4+|z| 2 2
|z|2
1 + |z|2
2
x y − 2
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 96/2

96 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

By using Eq. (2.61) we find that it transforms the north pole unit vector
2 (4x, 4y, 4 − |z| ). The cosine of the angle α between the
1 2
(0, 0, 1) into 4+|z|
north pole and the transformed direction is then given by:
4 − |z|2
cos α = , (2.324)
4 + |z|2
We can now find the angle of rotation of the unitary transformation follow-
ing the boost. We use for this purpose Eq. (2.289):
2
ϕ(z) = 2 arccos  . (2.325)
4 + |z|2
Using the formula cos ϕ = 2 cos2 ϕ2 − 1 we obtain
4 − |z|2
cos ϕ(z) = . (2.326)
4 + |z|2
The axis nr of rotation is computed using Eq. (2.292).
It is seen that nr is a unit vector whose components are proportional
to (sin(ψ), − cos(ψ), 0), therefore:
n1r = sin(ψ), (2.327)
n2r = − cos(ψ), (2.328)
n3r = 0. (2.329)
3
We notice that the rotation axis nr is in the equatorial plane x = 0 and
that it is perpendicular to the boost direction nb . The boost T (z) moves
the north pole by the angle ϕ, the rotation U (z) moves it back through
rotating by the same angle with respect to the axis that is orthogonal to
both: the north pole and its image under T (z). That is why the north pole
is invariant under the composition A(z) = U (z)T (z).
As an example Fig. 2.34 shows the deformation of the spherical coor-
dinate system by the boost alone, and by the complete transformation for
the case of A(1).

2.7.14 Pythagorean triples and quadruples

Consider again the two-parameter
  Abelian group of parabolic SL(2, C)
transformations A(z) = 10 x+iy
1 and the corresponding family of Lorentz
matrices Eq. (2.310). For x = 2, y = 0, we get the following matrix L(2, 0):
⎛ ⎞
1 0 −2 2
⎜0 1 0 0⎟
L(2, 0) = ⎜⎝2 0 −1 2⎠ ,
⎟ (2.330)
2 0 −2 3
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 97/2

What are Quantum Fractals? 97

Fig. 2.34 On the left: boost T (1) extracted from the parabolic matrix A(1). The black
dot indicates boost’s direction. On the right: The unitary rotation U (1) with respect to
the y-axis returns the north pole to its original position.

together with its inverse:

⎛ ⎞
1 0 2 −2
⎜0 1 0 0⎟
L(−2, 0) = ⎜
⎝−2
⎟. (2.331)
0 −1 2 ⎠
−2 0 −2 3
Both are matrices with integer coefficients, therefore they transform
vectors in Minkowski space with integer coefficients into other integer co-
efficient vectors. Being Lorentz matrices they transform vectors on the
future light cone into other vectors on the future light cone. Moreover, the
second component is left untransformed by these matrices — the trans-
formation affects only the coordinates x1 , x3 , x4 . Notice that the matrices
A(2, 0) and A(−2, 0) are, in fact, in SL(2, Z) ⊂ SL(2, R) ⊂ SL(2, C). The
group SL(2, R) is the covering group of the restricted Lorentz group in 2+1
space-time dimensions. The integer component vectors in (x, z, t) on the
future light cone there form Pythagorean triples x2 + z 2 = t2 .

2.7.14.1 Pythagorean triples and Hall matrices

A Pythagorean triple is a set (a, b, c) of positive integers such that a2 + b2 =
c2 . Such a triple is called primitive if a, b, c do not have a common divisor.
The smallest Pythagorean triple is the well known (3, 4, 5).
Barning [Barning (1963)] (cf. also A. Hall in [Hall (1970)]) generated
the whole “Pythagorean genealogical tree” starting with “the mother triple”
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 98/2

98 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

(3, 4, 5), and then applying successively one of the following three SL(2, R)
matrices — see Fig. 2.35.
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 −2 2 1 2 2 −1 2 2
A1 = ⎝2 −1 2⎠ , A2 = ⎝2 1 2⎠ , A3 = ⎝−2 1 2⎠ . (2.332)
2 −2 3 2 2 3 −2 2 3

2.7.14.2 Pythagorean quadruples

Pythagorean quadruples consists of four positive integers (x, y, z, t) with
x2 +y 2 +z 2 = t2 . In other words: Pythagorean quadruples (with respect to a
given reference frame) are those events on the future light cone whose space-
time coordinates are integer-valued. In number theory (cf. e.g. [Sierpiński
(1988)]) some easy properties of Pythagorean quadruples are derived as
follows.
Proposition 2.15. If (x, y, z, t) is a primitive Pythagorean quadruple, then
two of the three numbers x, y, z are even, while the third one is odd. t is
always odd.
Proof. We start with recalling some elementary properties of numbers.
While they are evident for mathematicians, it may not be so for people
outside this community.
Every even number can be written as 2k, k being an integer. Therefore
sums, products (therefore squares) of even numbers are even numbers.
Every odd number can be written as 2k − 1. Number 1 is the first odd
number, where k = 1. Thus products of odd numbers are odd numbers,
therefore also their squares.
If square of some number is even (resp. odd), then the number itself
is even (resp. odd). Product of an even number and an odd number is an
even number. Given two consecutive integers k, k + 1 their product k(k + 1)
is always even.
What else can we say about squared odd number beyond the fact that
it is odd? Assume n is odd, and let n = 2k − 1. Then
n2 = (2k − 1)2 = 4k 2 − 4k + 1. (2.333)
We can write it as
n2 = 4k(k − 1) + 1. (2.334)
Here we have a product of two consecutive integers k and k − 1, which is
even. Therefore
n2 = 8p + 1. (2.335)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 99/2

What are Quantum Fractals? 99

(63, 16, 65)

(45, 28, 53) (133, 156, 205)
(85, 132, 157)
(273, 136, 305)
(15, 8, 17) (55, 48, 73) (403, 396, 565)
(115, 252, 277)
(209, 120, 241)
(7, 24, 25) (275, 252, 373)
(51, 140, 149)
(165, 52, 173)
(45, 28, 53) (319, 360, 481)
(175, 288, 337)
(459, 220, 509)
(3, 4, 5) (21, 20, 29) (55, 48, 73) (697, 696, 985)
(217, 456, 505)
(299, 180, 349)
(7, 24, 25) (377, 336, 505)
(57, 176, 185)
(117, 44, 125)
(45, 28, 53) (207, 224, 305)
(95, 168, 193)
(187, 84, 205)
(5, 12, 13) (55, 48, 73) (297, 304, 425)
(105, 208, 233)
(91, 60, 109)
(7, 24, 25) (105, 88, 137)
(9, 40, 41)

Fig. 2.35 Pythagorean genealogical tree based on three matrices A1 , A2 , A3 and the
starting triple (3, 4, 5).

It follows that the remainder from the division of the square of any odd
number by 8 is always 1. Therefore also the remainder from the division of
the square of any odd number by 4 is also 1.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 100/2

100 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

What can be said about the remainder from the division by 4 of any
even number? If n = 2p, then n2 = 4p2 , therefore the remainder from the
division by 4 of any even number is always 0.
Let us return to the proof of the proposition. First of all we exclude the
possibility of all three numbers x, y, z being even. Indeed, if x, y, z are even,
then x2 , y 2 , z 2 are even, therefore t2 is also even, thus t must be even as well.
It would follow then x, y, z, t are all divisible by 2 — a contradiction with
the assumption that the Pythagorean quadruple (x, y, z, t) is primitive.
Next we exclude the possibility that (x, y, z) are all odd. Since, if we
assume it being the case, then

x2 = 8p + 1,
y 2 = 8q + 1,
z 2 = 8r + 1,

therefore

x2 + y 2 + z 2 = 8(p + q + r) + 3 = t2 . (2.336)

It would follow that the remainder from the division of t2 by 8 is 3. That

means t would have to be odd. But the remainder from the division by 8
of an odd number is 1 — a contradiction. Therefore x, y, z can not be all
odd.
Can two of them be odd, the third one being even? Let us suppose, for
instance, that x, y are odd, while z is even. Then

x2 = 4k + 1,
y 2 = 4p + 1,
z 2 = 4q,

therefore

t2 = 4(k + p + q) + 2,

which implies that t2 is divisible by 4 with the remainder 2. But if x, y

are odd, and z is odd, then t2 is even, thus t is even, therefore t2 should
be divisible by 4 without any remainder. Therefore the assumption that
x, y are odd while z is even leads to a contradiction. The only remaining
possibility is that of two of the three numbers x, y, z being even, while the
third one is odd. 
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 101/2

What are Quantum Fractals? 101

2.7.14.3 The 6 generators

The pair of matrices L(2, 0), L(−2, 0) distinguishes a fixed point — the
north pole — and a direction in space — the north-south axis on the
heavenly sphere. In order to treat all three directions the same way, we
introduce the 3-element group of transpositions of the axes. These are all
orthogonal matrices that do not affect the fourth coordinate. Explicitly,
let L(i), (i = 1, 2, 3) be the following set of Lorentz matrices in one row
simplified matrix notation:
L(1) = [1, 0, 0, 0; 0, 1, 0, 0; 0, 0, 1, 0; 0, 0, 0, 1]
L(2) = [0, 1, 0, 0; 0, 0, 1, 0; 1, 0, 0, 0; 0, 0, 0, 1]
L(3) = [0, 0, 1, 0; 1, 0, 0, 0; 0, 1, 0, 0; 0, 0, 0, 1]
Using these transformations we rotate, h → L(i)hL(i)−1, the two original
h(1) = L(2, 0), h(2) = L(−2, 0) matrices to obtain 6 Lorentz matrices h(i),
with integer coefficients
−1 −2 0 2 1 0 −2 2 1 0 0 0
2 1 0 −2 0 1 0 0 0 −1 −2 2
h(1) = 0 0 1 0
, h(2) = 2 0 −1 2 , h(3) = 0 2 1 −2 ,
−2 −2 0 3 2 0 −2 3 0 −2 −2 3
1 0 0 0 1 0 2 −2 −1 2 0 2
0 −1 2 2 0 1 0 0 −2 1 0 2
h(4) = 0 −2 1 2 , h(5) = −2 0 −1 2 , h(6) = 0 0 1 0
.
0 −2 2 3 −2 0 −2 3 −2 2 0 3
(2.337)
Our matrices h(i), if we consider only those acting on (1,3,4) coordinates,
have a similar structure to Hall’s matrices Eq. (2.332) [Hall (1970)], except
for the fact that all our matrices are of determinant +1.

Remark 2.14. The set {h(i) : i = 1, ..., 6} contains, with each matrix h(i),
its inverse matrix h(j), where we have ordered our six matrices in such a
way that h(j) = h(i)−1 for i + j = 7.

Definition 2.6. A quadruple a, b, c, d of integers, not all zero, d >= 1, with

the property
a2 + b2 + c2 = d2 (2.338)
is called a generalized Pythagorean quadruple (GPQ). If a, b, c, d are co-
prime, the quadruple is called primitive.

For a, b, c we allow all integers here, including negative and two zeroes.
Therefore the standard Pythagorean quadruples and also triples (when one
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 102/2

102 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

of the four numbers is zero) belong to this set. We introduce the following
generating set of primitive degenerate quadruples:
q1 = (1, 0, 0, 1)
q2 = (0, 1, 0, 1)
q3 = (0, 0, 1, 1)

Theorem 2.3. The set of 6 matrices h generates, from the generating set
(q1 , q2 , q3 ), all primitive generalized Pythagorean quadruples.

Proof. We first show that acting with the matrices h(i), (i = 1, ..., 6) on
a primitive GPQ produces again a primitive GPQ. Owing to the symmetry
it is enough to consider h(1). Other cases will differ only by permutation
and sign changes that bear no influence on the final result.
We have
h(1) : (a, b, c, d) → (−a − 2b + 2d, 2a + b − 2d, c, −2a − 2b + 3d). (2.339)
Suppose the resulting GPQ has a common divisor k, that is: −a− 2b + 2d =
km, 2a + b − 2d = kn, c = kp, −2a − 2b + 3d = kq, all integers. These
four equations for a, b, c, d solve to a = −k(m − 2n − 2q), b = k(−2m +
n + 2q), c = kp, d = k(−2m + 2n + 3q). Thus k would be then a common
divisor for a, b, c, d. Therefore the matrices h generate only primitive GPQs.
We will prove now that the matrices h, starting with the three generating
quadruples, generate all primitive GPQs.
We consider first the case of d = 1. Let (a, b, c, d) be a primitive GPQ.
If d = 1, then, since a2 + b2 + c2 = 1, and a, b, c are integers, two of a, b, c
must be zero, the third one must be ±1. We have
h(5)q1 = (−1, 0, 0, 1), (2.340)
h(1)q2 = (0, −1, 0, 1), (2.341)
h(3)q3 = (0, 0, −1, 1). (2.342)
It follows that among all six GPQs with d = 1 three are the generators and
the other three are obtained from them using h(5), h(1), h(3). Therefore it
will be enough to prove that starting with any GPQ with d > 1, we can
always reach one of the three elements of the generating set q1 , q2 , q3 . If this
is the case, we can take the complementary inverse elements (cf. Remark
2.14) to reverse the path.
To prove that it will be enough to show that, unless two numbers are
zero and the third is greater or equal to 1 (what implies d = 1), we can
always find a transformation in our set that strictly decreases the value
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 103/2

What are Quantum Fractals? 103

of d. Since our transformations preserve nonnegativity of d, we will then

finally end with one of the already considered six GPQs with d = 1.
To prove this let us suppose that, to the contrary, (a, b, c, d) is a primitive
GPQ such that for all h(i), the resulting quadruple (a , b , c , d ) has d ≥ d.
For any Lorentz transformation L we have
d = Λ41 a + Λ42 b + Λ43 c + Λ44 d. (2.343)
Denote w(i)μ = h(i)4μ . Since h(i)44 = 3 for all i, the inequality d ≥ d
translates to
w(i)1 a + w(i)2 b + w(i)3 c + 3d ≥ 1,
or
w(i)1 a + w(i)2 b + w(i)3 c + 2d ≥ 0. (2.344)
Consider first the case √ Eq. (2.344), we get d ≥ a+b.
√ of i = 1. For i = 1, form
Since d > 0 and d = a2 + b2 + c2 , we have a2 + b2 + c2 ≥ a+b, therefore
a2 + b2 + c2 ≥ (a + b)2 , or c2 ≥ 2ab. On the other hand application of the
inverse matrix, h(6) = h(1)−1 , leads to d ≥ a − b, which, by the same
reasoning, implies c2 ≥ −2ab. Therefore c2 ≥ 2|ab|. Doing the same for
the pairs (h(2), h(5)) and (h(3), h(4)) we obtain the following system of
inequalities:
a2 ≥ 2|bc|, (2.345)
b ≥ 2|ac|,
2
(2.346)
c ≥ 2|ab|.
2
(2.347)
By multiplying the LHS’s and RHS’s, we get
a2 b2 c2 ≥ 8a2 b2 c2 . (2.348)
With a, b, c being integers Eq. (2.348) is possible only if one of them is
zero. Suppose a = 0. Then, from a2 ≥ 2|bc|, we deduce that b or c must
be zero. Suppose b = 0. Then, from a2 + b2 + c2 = d2 , we get |c| = d. If
d > 1, then the quadruple (0, 0, ±d, d) is not a primitive one, contrary to
our assumption.
The same reasoning applies if b = 0 or d = 0. 
Generation of Pythagorean quadruples using our six generators is not very
effective as there are repetitions. For instance, although h(4) and h(5) do
not commute:
⎛ ⎞ ⎛ ⎞
1 0 −2 −2 1 0 2 −2
⎜ 0 −1 2 2 ⎟ ⎜ −8 −1 −6 10 ⎟
h(5)h(4) = ⎜ ⎟ ⎜
⎝−2 −2 3 4 ⎠ , h(4)h(5) = ⎝ −6 −2 −5 8 ⎠
⎟ (2.349)
−2 −2 4 5 −10 −2 −8 13
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 104/2

104 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

we have
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 1 −1
⎜0⎟ ⎜0⎟ ⎜ 2 ⎟
h(2)h(3) ⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0⎠ = h(3)h(2) ⎝0⎠ = ⎝ 2 ⎠ . (2.350)
1 1 3
Also the results of the IFS based on the six generators h(i) are not visually
pleasing — see Fig. 2.36.

Fig. 2.36 IFS based on six generators h(i).

2.7.14.4 Results of Robert Spira

For the sake of convenience we recall here the results obtained by Robert
Spira in his 1962 paper “The Diophantine equation x2 + y 2 + z 2 = m2 ”
[Spira (1962)]. However, we will change his original notation as follows: his
(x, y, z, m) we translate to ours (y, x, z, t), his (t, u, v, w) we translate into
ours (q, n, m, p). With these changes his theorems 1 and 2 read as follows:
Theorem 2.4 (Spira). Let (x, y, z, t) be a primitive Pythagorean quadru-
ple with x, y, z, t > 0, x, y even, z odd. Then there exist integers m, n, p, q
such that
x = 2(nq + mp)
y = 2(np − mq),
z = m 2 + n 2 − p2 − q 2 (2.351)
2 2 2 2
t = m +n +p +q
The parameters n, m, p, q can be chosen to satisfy the following conditions
(where gcd denotes “greatest common divisor”)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 105/2

What are Quantum Fractals? 105

i) n, p ≥ 1, m, q ≥ 0, m + q ≥ 1,
ii) mn > mq, m2 + n2 > p2 + q 2 ,
iii) m + n + p + q ≡ 1 (mod 2),
iv) gcd(m2 + n2 , p2 + t2 , nq + mp) = 1,
v) If q = 0 then n ≤ m, if m = 0 then p ≤ q.
With the conditions above satisfied, the representation of (x, y, z, t) in terms
of (m, n, p, q) is unique.

Remark 2.15. The representation Eq. (2.352) remains valid if y = 0, in

which case we have a Pythagorean triple. One should put then n = 0, q = 0.
See e.g [Sierpiński, 1988, p. 38].

Interpretation of Spira’s representation in terms of two-

component spinors Let C2 be the two-dimensional complex vector
space (endowed with the standard Hermitian scalar product). Its ele-
ments are pairs of complex numbers, written as ( ab ) , sometimes calledalso 
“qubits”. Given a vector v = ( ab ) , its complex conjugate vector is v̄ = āb̄ ,
while its hermitian conjugate v† = ( ā, b̄ ) . We can take the product of v
and v † to obtain the hermitian 2 × 2 matrix denoted, using Dirac’s bra and
ket notation, as |v >< v|:
a   aā ab̄
|v v| = ā, b̄ = . (2.352)
b bā bb̄
The matrix |v v| is automatically hermitian, and it projects onto the sub-
space spanned by v and is always singular. Its determinant is automatically
zero.
Hermitian 2 × 2 matrices have simple interpretation in terms of coor-
dinates (x, y, z, t) of Minkowski space given by the following well known
one-to-one correspondence:
1 t + z x + iy
(x, y, z, t) → . (2.353)
2 x − iy t − z
The factor 12 is convenient, so that the coordinate t comes as the trace of
the associated matrix. Its determinant is (t2 − x2 − y 2 − z 2 )/4. Therefore
the matrix is singular if and only if the point (x, y, z, t) is on the light cone.
Let us take a spinor v written explicitly in terms of real numbers
m, n, p, q, a = m + in, b = p + iq. Then the associated matrix |v v| reads
as follows:
m2 + n2 mp + nq + i(np − mq)
|v v| = . (2.354)
mp + nq − i(np − mq) p2 + q 2
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 106/2

106 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

In terms of associated Minkowski coordinates (x, y, z, t) we obtain:

x = 2(mp + nq),
y = 2(np − mq),
z = m2 + n2 − p2 − q 2 ,
t = m2 + n2 + p2 + q 2 ,

that is exactly the same formulas as in Spiras’s parametrization Eq. (2.352),

provided m, n, p, q are integers. Assume now that the vector v is of the form

1+i m + in
v+ = . (2.355)
2 p + iq

Then we obtain:

x = = m2 + n2 − p2 − q 2 ,
y = 2(mp + nq),
z = 2np − 2mq,
t = m2 + n 2 + p2 + q 2 .

Similarly, for

1−i m + in
v− = . (2.356)
2 p + iq

In this get we get

x = = m2 + n2 − p2 − q 2 ,
y = 2(mp + nq),
z = 2np − 2mq,
t = m2 + n 2 + p2 + q 2 .

Primitive Pythagorean quadruples. James D. Harper [Harper (2012)]

gave the following characterization of all primitive Pythagorean quadruples:

Proposition 2.16. Let m, n, p be positive integers satisfying the following

constraints:

(1) m2 + p2 > n2 ,
(2) gcd(m, n, p) = 1,
(3) If n is odd, then m + p is even.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 107/2

What are Quantum Fractals? 107

Then, with g = gcd(m2 + p2 , n2 ), the quadruple a = 2mn/g, b = 2np/g,

c = (m2 + p2 − n2 )/g, d = (m2 + n2 + p2 )/g is a primitive Pythagorean
quadruple with a and b even and c and d odd. Conversely, every primitive
Pythagorean quadruple with a and b even and c odd can be obtained in this
way.

Figure 2.37 and Fig. 2.38 show primitive Pythagorean quadruples obtained
via the above prescription for m, n, p < 30, symmetrized with respect to
the odd element of a, b, c.

Fig. 2.37 Primitive Pythagorean quadruples (symmetrized) generated by Harper’s al-

gorithm for m, n, p < 30. Upper hemisphere.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 108/2

108 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 2.38 Primitive Pythagorean quadruples generated by Harper’s algorithm for

m, n, p < 30. Symmetrized stereographic projection of the upper hemisphere.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 109/3

Chapter 3

Examples

This chapter contains a number of examples of quantum fractals. All of

them are symmetric, even if the symmetry requirement is not necessary.
What is necessary is the balancing condition Eq. (2.29). Symmetry, however
easily guarantees the balancing condition and, moreover, leads to visually
pleasant images.
The methods we use in these examples are those discussed in detail in
Chap. 2. However, at the cost of some repetitions, we will make an attempt
to make all examples below to a large degree self-contained.
In this book we do not provide graphic examples of quantum frac-
tals living on three-dimensional sphere S 3 , though the formulas needed
for generating higher dimensional hyperbolic quantum fractals are derived
in Sec. 3.3 and are given explicitly at the end of that section. Several ex-
amples based on six four-dimensional Platonic solids can be found in Ref.
[Jadczyk (2007)]. The sphere S 3 , embedded in 4D, can be stereographically
projected into 3D. However graphical rendering of 3D fractals is not an easy
task, and we leave it for the future.

3.1 Hyperbolic quantum fractals

All fractals in this chapter are based on hyperbolic Möbius transforma-

tions (cf. Sec. 2.7.10). Notice that there may be some confusion about
the term “hyperbolic”. General iterated function systems are often called
“hyperbolic” when they are based on contractive maps (cf. Sec. 2.2.1). Hy-
perbolic Möbius transformations, on the other hand, are never everywhere
contractive. They have their contraction and expansion regions.
We will start with one of the simplest quantum fractal — on the circle.
It is an analogy to the Cantor set. One may even give it the name The
Quantum Cantor Set .
109
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 110/3

110 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

3.1.1 The circle

If we cut the sphere with the plane y = 0, we get a circle. The subgroup of
the Lorentz group consisting of those transformations that do not affect the
y-coordinate act on this circle. We will define now four such transformations
that we will use for constructing an iterated function system on the circle.1
Let k ∈ (0, 1) be a real parameter. For a given k we define four 2 × 2
matrices A(i), (i = 1, .., 4), as follows:
1 1k
A(1) = √ , (3.1)
1−k 2 k1
1 1+k 0
A(2) = √ , (3.2)
1 − k2 0 1−k
1 1 −k
A(3) = √ , (3.3)
1 − k2 −k 1
1 1−k 0
A(4) = √ . (3.4)
1 − k2 0 1+k
Notice that we have A(3) = A(1)−1 , A(4) = A(2)−1 . These matrices are in
SL(2, R) ⊂ SL(2, C), the group of real 2 × 2 matrices with determinant 1,
and, when they act on C via fractional linear transformations, they leave
the real axis invariant.
In effect they act on the one-point compactification of R — the circle.
Alternatively, they define Lorentz transformations that do not affect the
y-component. The corresponding 3 × 3 Lorentz matrices, we will denote
them M (1), ..., M (4), acting on coordinates (x, z, t), are easily derived from
the explicit formulas Eq. (2.91). Denoting
1 + k2 2k
a= , b= (3.5)
1−k 2 1 − k2
we obtain:
⎛ ⎞ ⎛ ⎞
a 0 b 1 0 0
M (1) = ⎝0 1 0⎠ , M (2) = ⎝0 a b ⎠ ,
b 0 a 0 b a
⎛ ⎞ ⎛ ⎞ (3.6)
a 0 −b 1 0 0
M (3) = ⎝ 0 1 0⎠ , M (4) = ⎝0 a −b ⎠ .
−b 0 a 0 −b a
1 The simplest nontrivial quantum fractal on the circle needs three transformations.

Two transformations, when the balancing condition is imposed, imply that these trans-
formations commute — therefore we would get just two points as the attractor. Although
three transformations would be enough, we choose four because the unit square has in-
teger coordinates.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 111/3

Examples 111

The matrices M (i) are in the connected component of the identity of the
group O(2, 1), and they preserve the light cone x2 + z 2 − t2 = 0, as well as
the sign of the t-coordinate. Therefore they define the projective action on
the unit circle — the intersection of the light cone with the plane t = 1 —
with the following fractional linear transformation:
M (i)11 x + M (i)12 z + M (i)13 M (i)21 x + M (i)22 z + M (i)23
fi (x, z) = , .
M (i)31 x + M (i)32 z + M (i)33 M (i)31 x + M (i)32 z + M (i)33
(3.7)
2 2
If x + z = 1, then the point fi (x, z) has also this property. Thus fi maps
the unit circle into itself. This way we obtain an iterated function system
on the unit circle.

3.1.1.1 Contracting and expanding regions

Each of the four transformations has two fixed points on the circle. For
i = 1, .., 4 they are listed below, first the attracting point, then the repelling
point:

(1) (1, 0), (−1, 0)

(2) (0, 1), (0, −1)
(3) (−1, 0), (1, 0)
(4) (0, −1), (0, 1)

These four points are the vertices of the regular square inscribed into the
unit circle. In order to find the contracting and expanding regions for each
of our transformations, let us parametrize the circle by the angle:
x = cos φ, z = sin φ, (3.8)
and write the four transformations explicitly in terms of the angle φ. Using
the defining formulas we easily obtain the following transformations (by an
abuse of notation we use the same symbols fi for our four maps expressed
in terms of the angle φ):
   
f1 (φ) = atan2 k 2 cos(φ) + 2k + cos(φ), 1 − k 2 sin(φ) , (3.9)
  
f2 (φ) = atan2 1 − k cos(φ), k sin(φ) + 2k + sin(φ) ,
2 2
(3.10)
 2    
f3 (φ) = atan2 k + 1 cos(φ) − 2k, 1 − k sin(φ) ,
2
(3.11)
   2  
f4 (φ) = atan2 1 − k cos(φ), k + 1 sin(φ) − 2k .
2
(3.12)

Remark 3.1. One may notice that the arguments of the function atan2
above are not normalized — they are not on the unit circle. That is because
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 112/3

112 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

we have skipped the denominators (they are always positive) present in

Eq. (3.7). The function atan2, returning the argument of a complex number
z = x + iy, is insensitive to the scalings z → cz, c > 0.

For the derivatives we find:

1 − k2
f1 (φ) = , (3.13)
k 2 + 2k cos(φ) + 1
1 − k2
f2 (φ) = , (3.14)
k2 + 2k sin(φ) + 1
1 − k2
f3 (φ) = , (3.15)
k 2 − 2k cos(φ) + 1
1 − k2
f4 (φ) = . (3.16)
k 2 − 2k sin(φ) + 1
It is seen that the contracting regions grow from (−π/2, π/2) for
k → 0+, to (−π, π) for k → 1− around each of the four attracting points.
Therefore each of the four attracting points is always contained in the con-
tracting region of its two nearest neighbours — cf. Fig. 3.1, Fig. 3.2.

3.1.1.2 Place dependent probabilities

We still need to define (place-dependent) probabilities for our transforma-
tions. In our particular case there is a way of choosing these probabilities

Jacobians fi (φ) for k = 0.7.

6
i = 1
5 i = 2
i = 3
4 i = 6
fi (φ)

0
-π -π/2 0 π/2 π
φ

Fig. 3.1 Jacobians of the four transformations fi , for k = 0.7. Contracting regions are
those where the value of the Jacobian is smaller than 1.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 113/3

Examples 113

Jacobians f3 (φ) for k = 0.5, 0.7, 0.9

20
18 k=0.5
k=0.7
16
k=0.9
14
12
f3 (φ)

10
8
6
4
2
0
-π -π/2 0 π/2 π
φ

Fig. 3.2 Jacobians of the transformation f3 , for k = 0.5, 0.7, 0.9. Contracting regions
are those where the value of the Jacobian is smaller than 1.

in a natural way. First, we may easily verify that the denominators in

Eq. (3.7) are always positive and that their sum, for i = 1, ..., 4 is constant,
independent of x, z, and is equal to 4a. This suggests taking the denomi-
nators of Eq. (3.7) and dividing them by the constant 4a, that leads, after
simple calculation, to the following explicit expressions for the probabilities:
p(1; x, z) = 0.25 + cx, (3.17)
p(2; x, z) = 0.25 + cz, (3.18)
p(3; x, z) = 0.25 − cx, (3.19)
p(4; x, z) = 0.25 − cz, (3.20)
where c = 0.5k/(1 + k 2 ).
For k = 0.5, 0.7, 0.9, the values of c are 0.2, 0.234899, 0.248619 resp.
Notice that x and z vary between −1 and 1 (with the constraint x2 +z 2 = 1.)
Remark 3.2. It can be seen that these probabilities are inversely propor-
tional to the Jacobians f  (φ) of the corresponding transformations. This
may come as a surprise, since, based on a heuristic reasoning and practice,
it is usually assumed that probabilities for IFS should be chosen as directly
proportional to the Jacobians [Barnsely, 1988, p. 87]. However, even if
it is not relevant to our particular subject here, these formulas for prob-
abilities are automaticaly generated when similar constructions are being
made in simulations of quantum jumps within the ordinary linear quantum
mechanics — cf. Sec. 2.4, Sec. 4.7.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 114/3

114 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Remark 3.3. Selecting such or another set of probabilities does not, as a

rule, affect much the shape of the obtained image. In computer simulations
of the parabolic system of transformations discussed in Sec. 2.5 we used the
recipe above, as well as the uniform distribution of probabilities among the
transformations — the results were, visually, almost the same.

3.1.1.3 Images
The images below were generated each using 100,000,000 points, skipping
the first 100,000. The points on the circle were collected into 1200 beans
in the φ interval (0, 2π). We have run the simulations for three different
values of the parameter k: k = 0.9, 0.7, 0.5. While for k = 0.9 the generated
image (Fig. 3.3) resembles similar images of the Cantor set, the picture in
Fig. 3.5, representing the invariant measure for k = 0.5, does not appear
to be representing a fractal set at all. The invariant measure seems to be
continuous with respect to the Lebesgue measure on the circle. Apparently

IFS log10 density for: k = 0.9

log10 of bin counts

Fig. 3.3 IFS density for k = 0.9.

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 115/3

Examples 115

IFS log10 density for: k = 0.7

log10 of bin counts

Fig. 3.4 IFS density for k = 0.7.

there are no regions on the circle that are not being visited. The case of
k = 0.7 (Fig. 3.4) is in the middle between the two other cases. Question:
Do we know if we are getting a fractal set for k = 0.7? Are we sure that
we are getting fractal set for k = 0.9? Or, perhaps, there is some special
value of k — the onset of fractality?

3.1.1.4 Frobenius-Perron operator

In this section we discuss the Frobenius-Perron operator for our IFS on the
circle. We will assume that our IFS has a unique invariant measure, and
that this measure is continuous (nonsingular) with respect to the natural
measure λ (arc length) on the circle (though this last assumption is not
necessary). Our analysis here is similar to that in Sec. 2.3, where we dis-
cussed the Frobenius-Perron operator for the classical Cantor set — except
that now the formulas are much more complicated.
To each IFS there is an associated Frobenius-Perron operator (cf.
Sec. A.4) acting on measures (signed measures) on the underlying set, in
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 116/3

116 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

IFS log10 density for: k = 0.5

log10 of bin counts

Fig. 3.5 IFS density for k = 0.5.

our case — on the circle. If there is a unique invariant measure, then it is an

eigenvector of the Frobenius-Perron operator belonging to the eigenvalue
+1. The support of this measure is the attractor set for the system. We
will apply the standard formula for this operator acting on functions on the
circle.
We are interested here in positive densities. Suppose we have transfor-
mations x → wκ (φ), κ = 1, ..., K, and place-dependent probabilities pκ (φ),
where φ is a point on the circle.
Then the Frobenius-Perron P operator acting on measure densities f is
given by

K
  dλ(wκ−1 (x))
(P f )(φ) = pκ wκ−1 (φ) fκ (wκ−1 (φ)). (3.21)
κ=1
dλ(φ)

In order to construct a finite-dimensional approximation to P we divide

the circle into n intervals Ji of length δ, and chose a finite dimensional sub-
space Ln in the space of densities, spanned by the characteristic functions
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 117/3

Examples 117

ei (x) = χJi (x), i = 1, ..., n. We take

Ji = [ai , bi ], ai = iδ − δ/2, bi = iδ + δ/2, i = 0, ..., n − 1. (3.22)

Remark 3.4. There may be other choices of the basis, for instance using
continuous functions, that are better adapted to our particular problem,
and giving a better finite-dimensional approximation.

The finite dimensional approximation Pn for P is obtained essentially by

computing the matrix Mij of scalar products Mij = (ei , P ej ). For the
particular case of our IFS we adapt the formula given by Góra and Boyarsky
[Boyarsky and Gora, 1997, Sec. 4], which in our case reads as follows
  
K
λ wκ−1 (Ji ) ∩ Jj ) 1
Mi,j = pκ (x)dx . (3.23)
κ=1
δ δ Jj
The matrix Mij , by its very construction, is a left-stochastic matrix:

n
Mij = 1, j = 1, ..., k, (3.24)
i=1

therefore λ = 1 is one of its eigenvalues. We need to find the corresponding

eigenvector. This eigenvector will be then a discrete finite approximation
to the invariant density.

3.1.1.5 Details of the calculations

There are two technical problems involved in calculating the matrix M. The
first problem is that of calculating the mean values of pκ in the intervals Jj .
In our case of slowly varying functions p(φ) we can guess that there should
be no essential difference between calculating these numbers exactly and
replacing the mean values by the values of p in the middles of the intervals.
In fact we shall compare the resulting eigenvectors in these two cases: with
the integral calculated exactly, and the second case being that of assuming
the uniform probabilities pκ (x) = 1/K, that is just replacing the second
factor in Eq. (3.23) by 1/K.
 The second problem is that of finding an explicit working formula for
λ wκ−1 (Ji ∩ Jj ) .

Calculation of integrals. We set x = cos(φ), z = sin(φ), and consider

the case of κ = 1 in Eq. (3.17). Then

p(1; φ) = 0.25δ + c(sin(iδ + δ/2) − sin(iδ − δ/2)).
Ji
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 118/3

118 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

In order to avoid the subtractions of close numbers we use the formula

φ+ψ φ−ψ
sin(φ) − sin(ψ) = 2 cos( ) sin( ).
2 2
This way we obtain the following formulas:

k δ
p(1; φ) = 0.25δ + 2
cos(iδ) sin( ), (3.25)
Ji 1 + k 2

k δ
p(2; φ) = 0.25δ + sin(iδ) sin( ), (3.26)
Ji 1 + k2 2

k δ
p(3; φ) = 0.25δ − 2
cos(iδ) sin( ), (3.27)
Ji 1+k 2

k δ
p(2; φ) = 0.25δ − 2
sin(iδ) sin( ). (3.28)
Ji 1 + k 2
 −1 
Calculation of λ wκ (Ji ) ∩ Jj ) . This part needs special care due to
the fact that we are on a circle, therefore modulo 2π periodicity needs to
be taken into account.
The two intervals can overlap in several ways, therefore several cases
and subcases must be considered. We do not have a general algorithm
for this problem, but in the case of the hyperbolic transformation f3 , with
Ji = (A, B), w3 (Jj ) = (A1 , B1 ), our algorithm (using Fortran code) goes
as follows:

! CASE 1
IF ( (A < B) .AND. (A1 < B1) ) THEN
TM = MM0(A,A1,B,B1)

! CASE 2
ELSE IF ( ( A > B ) .AND. ( A1 < B1 ) ) THEN
TM = MM1(A,A1,B,B1) !Done

!CASE 3
ELSE IF ( ( A < B ) .AND. ( A1 >= B1) ) THEN
TM = MM2(A,A1,B,B1) !Done
!CASE 4
ELSE IF ( (A1 < Pi) ) THEN
IF ( B1 > 0 ) THEN
TM = TWOPI - A + MIN( B,B1 )
ELSE
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 119/3

Examples 119

TM = MIN (B,A1) - B1
END IF
ELSE
TM = TWOPI - MAX( A, A1 ) + MIN( B,B1 )
END IF
MM = TM/DELTA
END FUNCTION MM

REAL*8 FUNCTION MM0(A,A1,B,B1)

REAL*8 A,A1,B,B1

IF ( ( A > B1 ) .OR. ( A1 > B) ) THEN

MM0 = 0D0
ELSE
MM0 = MIN(B,B1) - MAX(A,A1)
ENDIF
END FUNCTION MM0

REAL*8 FUNCTION MM1(A,A1,B,B1)

REAL*8 A,A1,B,B1
IF ( ( A > B1 ) .AND. (A1 > B ) ) THEN
MM1 = 0
ELSE IF ( (A > B1 ) .AND. (A1 <= B) ) THEN
MM1 = MIN( B,B1 ) - A1
ELSE IF ( (A <= B1) .AND. ( A1 > B ) ) THEN
MM1 = B1 - MAX( A,A1 )
ELSE IF ( ( A <= B1 ) .AND. ( A1 <= B ) ) THEN
MM1 = B1 - A + B - A1
END IF
END FUNCTION MM1

REAL*8 FUNCTION MM2(A,A1,B,B1)

REAL*8 A,A1,B,B1
IF ( ( A > B1 ) .AND. (A1> B ) ) THEN
MM2 = 0
ELSE IF ( (A > B1 ) .AND. (A1 <= B) ) THEN
MM2 = B - MAX (A,A1)
ELSE IF ( (A <= B1) .AND. ( A1 > B ) ) THEN
MM2 = MIN( B1,B ) - A
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 120/3

120 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

ELSE IF ( ( A <= B1 ) .AND. ( A1 <= B ) ) THEN

MM2 = B1 - A + B - A1
END IF
END FUNCTION MM2

REAL*8 FUNCTION MM3(A,A1,B,B1)

REAL*8 A,A1,B,B1
IF ( A1 > PI ) THEN
IF ( B1 > 0 ) THEN
MM3 = TWOPI - A + MIN(B,B1)
ELSE
MM3 = MIN(B,B1) - B1
END IF
ELSE
MM3 = TWOPI - MAX( A, A1 ) + MIN( B, B1 )
ENDIF
END FUNCTION MM3

Other transformations can be reduced to this one owing to the symmetry

of the problem: each of the remaining three transformations is a rotation
by π/2, π, 3π/2 of f3 .

Eigenvalues of M. We calculated the eigenvalues of M for k =

0.5, 0.7 , 0.9, each in two cases: uniform probabilities = 0.25, and place
dependent probabilities Eq. (3.17),..., Eq. (3.20). Figures 3.6–3.8 show the
plots of eigenvalues (real or complex in conjugate pairs) for each case.

The eigenvector for the eigenvalue 1. The fixed point of the

Frobenius-Perron matrix represents an approximation of the invariant mea-
sure for the IFS. We normalized this vector to the probability density and,
again, compared the two cases: of place dependent and uniform probabili-
ties resp. See Fig. 3.9, Fig. 3.10, Fig. 3.11. It is of interest to notice that
uniform probabilities give less concentrated probability distribution.

3.1.2 Platonic quantum fractals for a qubit

The next five quantum fractals are based on Platonic solids. Each vertex
serves as an attraction point. Thus a solid with v vertices will define a
SQIFS (Standard Quantum Iterated Function System) based on v Möbius
transformations — pure boost corresponding to the same velocity.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 121/3

Examples 121

Fig. 3.6 Eigenvalues of the Frobenius-Perron matrix for k = 0.5. Lower graphics con-
cerns the case of uniform probabilities.

Fig. 3.7 Eigenvalues of the Frobenius-Perron matrix for k = 0.7. Lower graphics con-
cerns the case of uniform probabilities.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 122/3

122 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.8 Eigenvalues of the Frobenius-Perron matrix for k = 0.9. Lower graphics con-
cerns the case of uniform probabilities.

3.1.2.1 Platonic solids

Among regular polyhedra the most regular ones are Platonic solids. All
faces of a platonic solid are congruent polytopes, and the number of faces
that meet at one vertex is the same for each vertex. There are five Platonic
solids — they are listed in Table 3.1, where v is the number of vertices; e —
the number of edges; f — the number of faces; n — the number of vertices
in each polygon; m — the number of polygons meeting at each vertex.

Table 3.1 The Five Platonic solids.

v e f n m
Tetrahedron 4 6 4 3 3
Octahedron 6 12 8 3 4
Hexahedron (Cube) 8 12 6 4 3
Icosahedron 12 30 20 3 5
Dodecahedron 20 30 12 5 3

v is the number of vertices; e — the number of edges; f — the

number of faces; n — the number of vertices in each polygon; m
— the number of polygons meeting at each vertex.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 123/3

Examples 123

Fig. 3.9 Frobenius-Perron fixed point vector k = 0.5.

Below is the list of vertices of the five Platonic solids. We have cho-
sen the directions in such a way that in each case the north pole, with
coordinates (0, 0, 1), of the sphere is occupied by one of the vertices.

(1) tetrahedron: 4 vertices n[i], i = 1, . . . , 4

{{0, 0, 1.}, {a[17], 0, -a[3]}, {-a[6], a[12], -a[3]}, {-a[6], -a[12], -a[3]}}
(2) octahedron: 6 vertices n[i], i = 1, . . . , 6
{{0, 0, 1.}, {1., 0, 0}, {0, 1., 0},
{-1., 0, 0}, {0, -1., 0}, {0, 0, -1.}}
(3) cube: 8 vertices n[i], i = 1, . . . , 8
{{0, 0, 1.}, {a[17], 0, a[3]}, {-a[6], a[12], a[3]}, {-a[6], -a[12], a[3]},
{a[6], a[12], -a[3]}, {a[6], -a[12], -a[3]}, {-a[17], 0, -a[3]}, {0, 0, -1.}}
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 124/3

124 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.10 Frobenius-Perron fixed point vector k = 0.7.

(4) icosahedron: 12 vertices n[i], i = 1, . . . , 12

{{0, 0, 1.}, {0.a[15], 0, a[5]}, {a[2], a[13], a[5]}, {-a[10], a[7], a[5]},
{-a[10], -a[7], a[5]}, {a[2], -a[13], a[5]}, {a[10], a[7], -a[5]},
{a[10], -a[7], -a[5]}, {-a[2], a[13], -a[5]}, {-a[15], 0, -a[5]},
{-a[2], -a[13], -a[5]}, {0, 0, -1.}}
(5) dodecahedron: 20 vertices n[i], i = 1, . . . , 20
{{0, 0, 1.}, {a[9], 0, a[11]}, {-a[3], a[8], a[11]}, {-a[3], -a[8], a[11]},
{a[11], a[8], a[3]}, {a[11], -a[8], a[3]}, {-a[14], a[4], a[3]},
{a[1], a[16], a[3]}, {a[1], -a[16], a[3]}, {-a[14], -a[4], a[3]},
{a[14], a[4], -a[3]}, {a[14], -a[4], -a[3]}, {-a[11], a[8], -a[3]},
{-a[1], a[16], -a[3]}, {-a[1], -a[16], -a[3]}, {-a[11], -a[8], -a[3]},
{a[3], a[8], -a[11]}, {a[3], -a[8], -a[11]}, {-a[9], 0, -a[11]},
{0, 0, -1.}}
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 125/3

Examples 125

Fig. 3.11 Frobenius-Perron fixed point vector k = 0.9.

where the array of real numbers a[i], i = 1, . . . , 17, is given in the following
table.
√ √ √
3− 5 5− 5 1 5−1
√
a[1] = 6 a[2] = 10 a[3] = 3 a[4] = 2 3
√ √
a[5] = √1 a[6] = 2
a[7] = 5− 5
a[8] = √1
5 3 10 3
√ √
2 5+ 5 5 2
a[9] = 3
a[10] = 10
a[11] = 3
a[12] = 3
√ √ √
5+ 5 3+ 5 √2 3+ 5
a[13] = 10 a[14] = 6 a[15] = 5
a[16] = 6
√
2 2
a[17] = 3
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 126/3

126 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

For each of the five solids the center of mass is at the origin:
N
n[i] = 0. (3.29)
i=1
Therefore, for each of the solids, and a real parameter 0 < k < 1, we can
construct a hyperbolic QIFS using the formula Eq. (2.262) on p. 88. For a
given unit length vector n it is convenient to introduce vector k of length
k:
k = k n. (3.30)
Then
1
Ak = √ (I + k · s), (i = 1, ..., N ), 0 < k < 1. (3.31)
1 − k2
Notice that
k12 + k22 + k32 = k 2 . (3.32)
The parameter k tells us how close the given matrix is to a projection.
For k = 0 we have the identity transformation, for k close to 1 we have a
transformation that approximates the orthogonal projection on the one di-
mensional subspace spanned
 by the vector k. Matrices Ak of this form have
real trace equal to 2/ (1 − k 2 ). This number is always contained outside of
the interval [0, 2]. Therefore they define hyperbolic Möbius transformations.
By using Eq. (2.91), we can calculate the Lorentz matrix corresponding
to Ak . The result is given by:
⎛ ⎞
(1 − k 2 ) + 2k12 2k1 k2 2k1 k3 2k1
1 ⎜ ⎜ 2k1 k2 (1 − k 2 ) + 2k22 2k2 k3 2k2 ⎟
⎟
Lk = ⎝
1−k 2 2k1 k3 2k2 k3 (1 − k ) + 2k
2 2
3 2k3 ⎠
2k1 2k2 2k31 + k2
(3.33)
The corresponding transformation wk of the sphere is now simply calculated
(1 − k 2 )x + 2(1 + k · x) k
x = wk (x) = . (3.34)
1 + k2 + 2 k · x
The inverse transformation is obtained from the above one by replacing k
with −k:
(1 − k2 )x − 2(1 − k · x) k
wk −1 (x) = . (3.35)
1 + k2 − 2 k · x
Remark 3.5. In numerical calculations it is better to express Eq. (3.34)
in a different, though equivalent form:
(1 − k 2 )x + 2(1 + k · x) k
x = wk (x) = . (3.36)
||(1 − k 2 )x + 2(1 + k · x) k||
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 127/3

Examples 127

The point is that the denominator of Eq. (3.34) contains the term
2 k · x, which can be negative. Numerically it is more stable to replace
the denominator by the norm of the numerators, especially when the for-
mula is being iterated many times. This way we are making sure that we
are not departing from the unit sphere.

Using Eq. (2.268), we find the velocity of the boost described by the trans-
formation Eq. (3.34):
2k
v(k) = . (3.37)
1 + k2
The increase of v(k) with k becomes very slow as k → 1:

Fig. 3.12 Dependence of the boost velocity v(k) on k for positive SL(2, C) matrices of
the form Ak = √ 1 2 (I + k · s).
1−k

3.1.2.2 Approximations to the invariant measure

We recall here the action of the Frobenius-Perron operator on densities of
measures on the sphere. The general formula (cf. Eq. (2.17)) reads:
 
m
 −1   −1  dμ wi−1 (x)
(P f )(x) = pi wi (x) f wi (x) . (3.38)
dμ(x)
i=1

For the measure μ in this formula we choose the natural rotation invariant
measure on the sphere, while for the transformations wi we choose the trans-
formations implemented by the SL(2, C) matrices of the form Eq. (3.31).
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 128/3

128 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

The area transformation corresponding to such a matrix can be then cal-

culated from Eq. (2.218), where we need to put:

a = (1 + k3 )/ 1 − k 2 ,

b = (1 − k3 )/ 1 − k 2 ,

c = (k1 + ik2 )/ 1 − k 2 ,

d = (1 − k3 )/ 1 − k 2 ,
z = (x + iy)/(1 − z).

The vector x is the unit vector on the sphere. It is then a question of a

straightforward computation to find out the result:
2
dS  1 − k2
Jk (x) = = . (3.39)
dS 1 + k 2 + 2k · x
In order to calculate the Frobenius-Perron operator we also need the for-
mula for the place dependent probabilities. Here we have essentially two
choices:

(1) Uniform distribution of constant probabilities:

pi (x) = 1/n, (3.40)

(2) SQIFS distribution defined in Eq. (2.30)

1 + k 2 + 2ki · x
pi (x) = cJki (x)−1/2 = c . (3.41)
1 − k2
n n
Since i=1 ki = 0, the condition i=1 pi (x) = 1 gives us the value of
the constant c:
1 − k2
c= , (3.42)
n(1 + k 2 )
therefore
1 + k 2 + 2ki · x
pi (x) = cJki (x)−1/2 = . (3.43)
n(1 + k 2 )
 
We need to calculate pi wi−1 (x) . It is a matter of a straightforward
calculation to find out that substituting x in 1 + k 2 + 2k · x for w−1 (x)
from Eq. (3.35) gives (1 − k2 )2 /(1 + k 2 − 2k · x). Therefore we get:
  1 − k2 1
pi wi−1 (x) = · . (3.44)
n(1 + k 2 ) 1 + k 2 − 2ki · x
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 129/3

Examples 129

On the other hand, substituting −k for k in Eq. (3.43), we get

 
dμ wi −1 (x) 1 − k2
2
= . (3.45)
dμ(x) 1 + k 2 − 2k · x
Therefore we obtain the following result:

Proposition 3.1. Given a system of hyperbolic transformations defined

n
by n matrices Aki as in Eq. (3.31), with i=1 ki = 0, the corresponding
Frobenius-Perron operator P acting on densities f (x) with respect to the
rotation invariant measure on the sphere is given by the formula:
(1) In the case of constant probabilities
(1 − k 2 )2 1
(P f )(x) = · . (3.46)
n (1 + k − 2k · x)2
2

(2) In the case of place dependent probabilities according to the SQIFS

method:
(1 − k 2 )4 1
(P f )(x) = · . (3.47)
n(1 + k 2 ) (1 + k 2 − 2k · x)3
The following question naturally arises in this context: does a unique
invariant measure exist for the above Frobenius-Perron operator?
In Ref. [Stenflo (2002)] Stenflo states the following classical results,
attributed to Barnsley et al [Barnsley et al. (1988)].
Theorem 3.1 (Barnsley, Stenflo). Let {(X, d), pi (x), wi (x), i ∈ S =
{1, 2, . . . , N }} be an IFS with place-dependent probabilities, with all wi be-
ing Lipshitz continuous, and all pi being Dini-continuous, and bounded away
from 0. Suppose

N
d (wi (x), wi (y))
g(w, p) = sup pi (x) log < 0. (3.48)
x=y i=1 d(x, y)
Then the generated Markov chain has a unique invariant probability mea-
sure.

The log-average contraction condition Eq. (3.48) is somewhat more gen-

eral than the average contraction condition

.  N
d (wi (x), wi (y))
M ax(w, p) = sup pi (x) < 1. (3.49)
x=y i=1 d(x, y)
1−2
In our case wi (x) and pi (x) are analytic, with pi (r) ≥ N (1+2 )
.
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 130/3

130 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

We were unable to show analytically that the condition g(w) < 0 is

satisfied for the family of Platonic quantum fractals analyzed here. Never-
theless it was possible to estimate the value of g for each of the five fractals,
and for a whole range of the parameter k. Except for the case of the tetra-
hedron the differences between uniform probabilities and place-dependent
probabilities were very small.
Therefore, except for the tetrahedron, we give in Fig. 3.13 the behavior
of g(w, k) for uniform probabilities only.

Log-average function g(k), 0<k<1

0.0

-0.5

-1.0

-1.5
g(k)

-2.0 Tetrahedron
Octahedron
-2.5
Cube
-3.0 Icosahedron
-3.5 Dodecahedron
Tetrahedron SQIFS
-4.0
0.0 0.2 0.4 0.6 0.8 1.0
k
Fig. 3.13 Numerical estimates of the log-average function g(k) for all five Platonic
quantum fractals. Uniform probabilities are used, except for the Tetrahedron, where
the results for both uniform and place dependent probabilities according to the SQIFS
formula are plotted.

3.1.2.3 Geometric interpretation of transformations and

probabilities
We will give now a simple geometric interpretation of the transformations
Eq. (3.34) of the point on the sphere by the Lorentz matrices Lk — recall
that Lk describe a pure relativistic boost in the directions of k with velocity
v(k) = 2k/(1 + k 2 ).2
2 The idea for such a simple interpretation came from a file that I have downloaded some

time ago from some site on the internet. The file has no title and no author, Google
search by keywords returns nothing. Nevertheless the input from the anonymous author
is acknowledged! The interpretation for probabilities is my own addition.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 131/3

Examples 131

p(x)k

x x+k

Fig. 3.14 Geometric interpretation of transformations and probability function by a

pure boost transformation.

Let x be a unit vector, x2 = 1, and let k be a vector of a positive length

less than 1 — cf. Fig. 3.14. We want to find the formula for the reflection
of x with respect to the line defined by the vector x + k. The solution can
be found explicitly by solving the following pair of equation for an unknown
vector x :

x · x = 1, x · (x + k) = x · (x + k). (3.50)

These are quadratic equations, therefore we have two solutions: one is

x, the second solution is:
(1 − k2 )x + 2(1 + kx) k
x = . (3.51)
1 + k 2 − 2k · x
Therefore we have obtained the geometric interpretation of the transfor-
mation formula Eq. (3.34). But what about the formula for probability
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 132/3

132 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

function p(x) given by Eq. (3.41)? The values of probabilities in the iter-
ated function system depend on the other transformations; so here, when
we are discussing only one transformation, we can only talk about the de-
pendence of the probability p(x) on x — up to a scale. We will show that
the function p(x) of Eq. (3.41) with c = 1, which is the inverse square root
of the Jacobian, also has a simple geometric interpretation.
First consider the vector Lk (x) defined by


3
Lk (x)i = (Lk )i j xj + (Lk )i 0
i=1
 i
= (1 − k 2 )x + 2(1 + k · x) k . (3.52)

This is the same as x — up to the normalization. A straightforward

calculation shows that its norm is proportional to the probability function:

||Lk (x)|| = 1 + k 2 + 2 k · x. (3.53)

This fact can be handy in computer implementation of quantum fractals:

we can use the value of this norm in two ways: for the normalization of the
vector Lk (x) and also for calculation of the place dependent probabilities.
But there is also another way of obtaining a geometric interpretation of
the probability function. We can rewrite Eq. (3.51) as
x+k
x = + k, (3.54)
λ
where λ is given by

1 + k2 + 2 k · x
λ= . (3.55)
1 − k2
It is exactly our probability function. The geometric interpretation follows
at once: from

λx = x + k + λk (3.56)

we deduce that λ is the unique factor for which the sum of x + k and λk is
on the same line as the vector x . This geometric interpretations is depicted
in Fig. 3.14.
We will now give explicit formulas for the vertices and probabilities of
QIFS based on the five Platonic solids, as well as images resulting from the
Chaos Game for different values of the parameter k.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 133/3

Examples 133

3.1.2.4 Quantum Tetrahedron

The tetrahedron has four vertices. Therefore our IFS consists of four func-
tions. It appears that it is not enough for creating a nice looking and
sophisticated fractal pattern.

Vertices
⎛ ⎞
√
0 0 1.
⎜ 2 2
⎜ 3 0 − 13 ⎟
⎟
⎜ √ ⎟
⎜ ⎟ (3.57)
⎜ − 32 2
− 13 ⎟
⎝ √ 3 ⎠
− 32 − 2
3 −3 1

Probabilities
prob(1) = 1/4 + 0.5 k z/(1 + k 2 )
prob(2) = 1/4 + k (0.471405 x − 0.166667 z)/(1 + k 2 )
prob(3) = 1/4 + k (−0.235702 x + 0.408248 y − 0.166667 z)/(1 + k 2 )
prob(4) = 1/4 + k (−0.235702 x − 0.408248 y − 0.166667 z/(1 + k 2 )
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 134/3

134 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.15 Quantum tetrahedron for k = 0.2, 109 iterations. With this value of k we see
mainly chaos.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 135/3

Examples 135

Fig. 3.16 Quantum tetrahedron for k = 0.3, 109 iterations. Somewhat unexpected
pattern appears on the sphere.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 136/3

136 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.17 Quantum tetrahedron for k = 0.4, 109 iterations. The pattern gets more
intricate.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 137/3

Examples 137

Fig. 3.18 Quantum tetrahedron for k = 0.4, 109 iterations. The southern hemisphere.
where the three other vertices are located.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 138/3

138 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.19 Quantum tetrahedron for k = 0.5, 109 iterations. The pattern becomes more
intricate.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 139/3

Examples 139

Fig. 3.20 Quantum tetrahedron for k = 0.6, 109 iterations. Vertices becomes more and
more attractive.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 140/3

140 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.21 Quantum tetrahedron for k = 0.7, 109 iterations. Vertices become more and
more attractive.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 141/3

Examples 141

Fig. 3.22 Quantum tetrahedron for k = 0.8, 109 iterations. Vertices are strongly at-
tractive.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 142/3

142 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

3.1.2.5 Quantum Octahedron

Vertices
⎛ ⎞
1 0 0
⎜ 0 1 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 1 ⎟
⎜ ⎟ (3.58)
⎜ 0 0 −1 ⎟
⎜ ⎟
⎝ 0 −1 0 ⎠
−1 0 0
Probabilities
prob(1) = 1/6 + k x/(3 + 3 k 2 )
prob(2) = 1/6 + k y/(3 + 3 k 2 )
prob(3) = 1/6 + k z/(3 + 3 k2 )
prob(4) = 1/6 − k z/(3 + 3 k2 )
prob(5) = 1/6 − k y/(3 + 3 k 2 )
prob(6) = 1/6 − k x/(3 + 3 k 2 )
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 143/3

Examples 143

Fig. 3.23 Quantum octahedron for k = 0.3, 109 iterations. Mainly chaos.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 144/3

144 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.24 Quantum octahedron for k = 0.4, 109 iterations. First circles become visible.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 145/3

Examples 145

Fig. 3.25 Quantum octahedron for k = 0.5, 109 iterations. More circles.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 146/3

146 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.26 Quantum octahedron for k = 0.6, 109 iterations. Black (small probability of
hits) circularly shaped regions appear.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 147/3

Examples 147

Fig. 3.27 Quantum octahedron for k = 0.7, 109 iterations. Black circular regions grow.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 148/3

148 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.28 Quantum octahedron for k = 0.8, 109 iterations. Almost all hits are concen-
trated near the vertices.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 149/3

Examples 149

3.1.2.6 Quantum Cube

Vertices
⎛ ⎞
0 0 1.
⎜ √ ⎟
⎜ 2 2
0 1⎟
⎜ 3 3⎟
⎜ √ ⎟
⎜ − 2 2 1 ⎟
⎜ 3 3 3 ⎟
⎜ √ ⎟
⎜ − 2 − 23 1 ⎟
⎜ 3 ⎟
⎜ √
3
⎟ (3.59)
⎜ ⎟
⎜ 2 2
− 13 ⎟
⎜ √
3 3 ⎟
⎜ ⎟
⎜ 2
− 23 − 13 ⎟
⎜ 3
√  ⎟
⎜−1 2 2 − 13 ⎟
⎝ 3 0 ⎠
0 0 −1.
Probabilities
prob(1) = 1/8 + 0.25 k z/(1 + k 2 )
prob(2) = 1/8 + k (0.235702 x + 0.0833333 z)/(1 + k2 )
prob(3) = 1/8 + k (−0.117851 x + 0.204124 y + 0.0833333 z)/(1 + k2 )
prob(4) = 1/8 + k (−0.117851 x − 0.204124 y + 0.0833333 z)/(1 + k 2 )
prob(5) = 1/8 + k (0.117851 x + 0.204124 y − 0.0833333 z)/(1 + k 2 )
prob(6) = 1/8 + k (0.117851 x − 0.204124 y − 0.0833333 z)/(1 + k 2 )
prob(7) = 1/8 + k (−0.235702 x − 0.0833333 z)/(1 + k 2 )
prob(8) = 1/8 − 0.25 k z/(1 + k 2 )
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 150/3

150 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.29 Quantum cube for k = 0.5, 109 iterations. First features appear.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 151/3

Examples 151

Fig. 3.30 Quantum cube for k = 0.6, 109 iterations. Circular regions appear.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 152/3

152 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.31 Quantum cube for k = 0.7, 109 iterations. Pattern becomes more intricate.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 153/3

Examples 153

Fig. 3.32 Quantum cube for k = 0.8, 109 iterations. Big circular, overlapping empty
regions.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 154/3

154 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

3.1.2.7 Quantum Icosahedron

Vertices
⎛ ⎞
0 0 1.
⎜ ⎟
⎜ √2 0 √1 ⎟
⎜ 5 5 ⎟
⎜ 1  √   √  ⎟
⎜ 5 − 5 1
5+ 5 √1 ⎟
⎜ 10 10 5 ⎟
⎜  √ √  ⎟
⎜ 1 −5 − 5 
√1 ⎟
10 5 −
1
⎜ 10 5 ⎟
5
⎜  √   √  ⎟
⎜ 1 ⎟
⎜ 10 −5 − 5 − 101
5− 5 √1 ⎟
5 ⎟
⎜ √  √ 
⎜ 1   ⎟
⎜ 10 5 − 5 − 101
5+ 5 √1 ⎟
⎜ 5 ⎟ (3.60)
⎜ 1  √   √  ⎟
⎜
⎜ 10 5 + 5 10 5 −
1
5 − √15 ⎟
⎟
⎜ ⎟
⎜ 1 5 + √5 − 101
 √
5− 5
 1 ⎟
− 5⎟
√
⎜ 10
⎜  √   √  ⎟
⎜ 1 ⎟
⎜ 10 −5 + 5 1
10 5 + 5 − √15 ⎟
⎜ ⎟
⎜
⎜ − √25 0 − √15 ⎟
⎟
⎜  ⎟
⎜ 1 −5 + √5 − 1
 √ 
− √5 ⎟
1
⎝ 10 10 5 + 5 ⎠
0 0 −1.
Probabilities
prob(1) = 1/12 + 0.166667 k z/(1 + k 2 )
prob(2) = 1/12 + k (0.149071 x + 0.0745356 z)/(1 + k 2 )
prob(3) = 1/12 + k (0.0460655 x + 0.141775 y + 0.0745356 z)/(1 + k2 )
prob(4) = 1/12 + k (−0.120601 x + 0.0876219 y + 0.0745356 z)/(1 + k2 )
prob(5) = 1/12 + k (−0.120601 x − 0.0876219 y + 0.0745356 z)/(1 + k2 )
prob(6) = 1/12 + k (0.0460655 x − 0.141775 y + 0.0745356 z)/(1 + k2 )
prob(7) = 1/12 + k (0.120601 x + 0.0876219 y − 0.0745356 z)/(1 + k2 )
prob(8) = 1/12 + k (0.120601 x − 0.0876219 y − 0.0745356 z)/(1 + k2 )
prob(9) = 1/12 + k (−0.0460655 x + 0.141775 y − 0.0745356 z)/(1 + k2 )
prob(10) = 1/12 + k (−0.149071 x − 0.0745356 z)/(1 + k 2 )
prob(11) = 1/12 + k (−0.0460655 x − 0.141775 y − 0.0745356 z)/(1 + k 2 )
prob(12) = 1/12 − 0.166667 k z/(1 + k 2 )
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 155/3

Examples 155

Fig. 3.33 Quantum icosahedron for k = 0.6, 109 iterations. Five fold symmetry becomes
identifiable.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 156/3

156 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.34 Quantum icosahedron for k = 0.7, 109 iterations. Patterns become more
apparent.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 157/3

Examples 157

Fig. 3.35 Quantum icosahedron for k = 0.8, 109 iterations. Circular regions appear.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 158/3

158 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.36 Quantum icosahedron for k = 0.85, 109 iterations. Five-fold snowflake pat-
terns predominate.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 159/3

Examples 159

3.1.2.8 Quantum Dodecahedron

Vertices
⎛ ⎞
0 0 1.
⎜ √ ⎟
⎜ 2
0 5 ⎟
⎜ 3
√
3 ⎟
⎜ ⎟
⎜ −3 1 √1 5 ⎟
⎜ 3 √
3 ⎟
⎜ ⎟
⎜ − 13 − √13 5
⎟
⎜ √
3 ⎟
⎜ ⎟
⎜ 5 √1 1
⎟
⎜ √
3 3 3 ⎟
⎜ 5
− 1 1 ⎟
⎜ √ ⎟
⎜  3 √  √3 3
⎟
⎜ 1 −3 − 5 −1+√ 5 1 ⎟
⎜6 ⎟
⎜  √  
2 3
√ 
3
⎟
⎜ 1 1 ⎟
⎜ 6 3− 5 1
3+ 5 3 ⎟
⎜ √ 
6
√  ⎟
⎜ 1  1 ⎟
⎜ 6 3− 5 − 1
3 + 5 ⎟
⎜  √ 
6
√
3 ⎟
⎜1 1 ⎟
⎜ 6 −3 − 5 − −1+ √ 5 ⎟
⎜  √  2 √3 3 ⎟
(3.61)
⎜ 1 −1+ 1 ⎟
⎜ 6 3+ 5 √ 5 −3 ⎟
⎜  √  2 3√ ⎟
⎜ 1 ⎟
⎜ 6 3+ 5 − −1+ √ 5 − 31 ⎟
⎜ √ 2 3 ⎟
⎜ − 35 √1 − 31 ⎟
⎜ ⎟
⎜  √  
3
√  ⎟
⎜1 ⎟
⎜ 6 −3 + 5 1
3+ 5 − 31 ⎟
⎜ 6 ⎟
⎜1 √   √  ⎟
⎜ −3 + 5 − 1 3 + 5 − 31 ⎟
⎜6 6 ⎟
⎜ √ ⎟
⎜ − 5
− √13 − 31 ⎟
⎜ 3
√ ⎟
⎜ ⎟
⎜ 1 √1 − 35 ⎟
⎜ 3 3 √ ⎟
⎜ ⎟
⎜ 1
− √1 − 35 ⎟
⎜ 3 3 √ ⎟
⎜
⎝ − 23 0 − 35 ⎟
⎠
0 0 −1.

Probabilities

prob(1) = 1/20 + 0.1 k z/(1 + k 2 )

prob(2) = 1/20 + k(0.0666667 x + 0.0745356 z)/(1 + k2 )
prob(3) = 1/20 + k(−0.0333333 x + 0.057735 y + 0.0745356 z)/(1 + k 2 )
prob(4) = 1/20 + k(−0.0333333 x − 0.057735 y + 0.0745356 z)/(1 + k 2 )
prob(5) = 1/20 + k(0.0745356 x + 0.057735 y + 0.0333333 z)/(1 + k 2 )
prob(6) = 1/20 + k(0.0745356 x − 0.057735 y + 0.0333333 z)/(1 + k 2 )
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 160/3

160 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

prob(7) = 1/20 + k(−0.0872678 x + 0.0356822 y + 0.0333333 z)/(1 + k 2 )

prob(8) = 1/20 + k(0.0127322 x + 0.0934172 y + 0.0333333 z)/(1 + k 2 )
prob(9) = 1/20 + k(0.0127322 x − 0.0934172 y + 0.0333333 z)/(1 + k2 )
prob(10) = 1/20 + k(−0.0872678 x − 0.0356822 y + 0.0333333 z)/(1 + k2 )
prob(11) = 1/20 + k(0.0872678 x + 0.0356822 y − 0.0333333 z)/(1 + k 2 )
prob(12) = 1/20 + k(0.0872678 x − 0.0356822 y − 0.0333333 z)/(1 + k2 )
prob(13) = 1/20 + k(−0.0745356 x + 0.057735 y − 0.0333333 k z)/(1 + k 2 )
prob(14) = 1/20 + k(−0.0127322 x + 0.0934172 y − 0.0333333 z)/(1 + k 2 )
prob(15) = 1/20 + k(−0.0127322 x − 0.0934172 y − 0.0333333 z)/(1 + k 2 )
prob(16) = 1/20 + k(−0.0745356 x − 0.057735 y − 0.0333333 z)/(1 + k 2 )
prob(17) = 1/20 + k(0.0333333 x + 0.057735 y − 0.0745356 z)/(1 + k 2 )
prob(18) = 1/20 − k(0.0333333 x − 0.057735 y − 0.0745356 z)/(1 + k 2 )
prob(19) = 1/20 + k(−0.0666667 x − 0.0745356 z)/(1 + k 2 )
prob(20) = 1/20 − 0.1kz/(1 + k2 )
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 161/3

Examples 161

Fig. 3.37 Quantum dodecahedron for k = 0.65, 109 iterations. Place independent
probabilities.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 162/3

162 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.38 Quantum dodecahedron for k = 0.65, 109 iterations. Place dependent prob-
abilities. More details can be distinguished.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 163/3

Examples 163

Fig. 3.39 Quantum dodecahedron for k = 0.7, 109 iterations.

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 164/3

164 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.40 Quantum dodecahedron for k = 0.75, 109 iterations.

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 165/3

Examples 165

Fig. 3.41 Quantum dodecahedron for k = 0.85, 109 iterations.

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 166/3

166 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.42 Quantum dodecahedron for k = 0.87, 109 iterations.

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 167/3

Examples 167

3.2 Controlling chaotic behavior and fractal dimension

Quantum fractals live on projective spaces. In general a projective space

is a rather complex concept, but here we are interested mainly in the two-
dimensional sphere. In physics the points on the sphere can be interpreted
in several ways: as pure quantum states of a spin 1/2 quantum system, as
pure quantum states of a two-level atom, or as light directions plotted on
the heavenly sphere.
Any finite system of 2 × 2 nonsingular complex matrices can be used
to define an iterated function system of Möbius transformations on the
sphere. Given a nonsingular complex matrix A, it decomposes uniquely,
by the polar decomposition theorem (cf. Proposition 2.10, p. 86) into a
product A = U T, where U is unitary and T is positive.
Unitary matrices describe rotations, and rotations preserve distances
on the sphere. It is the positive part T that is responsible for contractions
and expansions of regions of the sphere. Contractions and expansions lead
to self-similarity and to fractal-like patterns. Therefore let us limit our
discussion to iterated function systems on the sphere that are generated by
families of positive matrices.
Every positive 2 × 2 complex matrix, if not proportional to the identity
matrix, has two mutually orthogonal eigenvectors. These two vectors are
represented by two opposite points on the sphere that represent complex
vectors in C2 modulo constant factor (see Eq. (2.124)). Therefore we have
two fixed points, one of them is attracting; the other one is repelling.
Given a point n, n2 = 1, on the sphere, there is a unique (up to a scale)
one-parameter family of positive matrices A(k), 0 < k < 1, with n as the
attraction point. These are the matrices that are being used in most of our
examples of quantum fractals.
Instead of working with complex matrices acting on complex vectors or
complex plane, it is often more convenient to work directly with the real
sphere in R3 , and with real transformations. While working in R3 , the one
parameter family of transformations wk,n defined by n (sometimes we write
kn = k) can be then written as (see Eq. (3.34))

r = wk,n (r) = (1 − k 2 )r + 2k(1 + 2k(n · r))n, (3.62)

that is followed by the normalization:

r
r → . (3.63)
||r ||
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 168/3

168 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

As an example in this section we will take the Quantum Octahedron with

its six vertices given by vectors
ni = ((1, 0, 0), (0, 1, 0), (0, 0, 1), (−1, 0, 0), (0, −1, 0), (0, 0, −1)). (3.64)
The fractal patterns defined by the IFS described in Sec. 3.1.2.5 depend
now on the value of the parameter k, 0 < k < 1. We are interested in the
question how the estimated fractal dimension depends on the value of the
parameter.
Here we encounter several problems. First of all, with our prescrip-
tion Eq. (2.30) for place dependent probabilities, it is expected that the
resulting estimate will be different than if the probabilities are constant
and uniformly distributed among the six Möbius transformations.
Second, we are working on the sphere, while the known algorithms for
estimating fractal dimensions are adapted to the flat space. Dividing the
sphere into equal area cells is not a simple task. We could try Lambert’s
equal area projection from the half-sphere onto the disk, but such a projec-
tion, at the boundaries, will distort the shape of the pattern, and that can
influence the estimated dimensions. Stereographic projection, on the other
hand, while preserving shapes, distorts areas.
Here, in order to estimate the fractal dimension of the generated pat-
tern, we will take a middle way: we will project vertically the upper hemi-
sphere on the plane, but we will take, for our calculation, only a small
neighborhood of the north pole. While the whole hemisphere is projected
on the disk x2 + y 2 = 1, for our estimations of the dimension of the re-
sulting fractal measure we take only the pattern projected on the square
(−0.2 < x < 0.2) × (−0.2 < y < 0.2). Within such a small square around
the north pole the distortion of the area is of the order of a few percent.
We can afford such a distortion, because the existing algorithms for fractal
dimension estimates differ from one another by that much or more.
From all existing freely available packages we have chosen the one that is
widely used, FD3.3 FD3 is based on the algorithm by Liebovitch and Toth
[Liebovitch and Toth (1989)]; which, in its turn, is an implementation of
the earlier ideas of Grassberger and Procaccia [Grassberger and Procaccia
(1983)]. The algorithm estimates three kinds of fractal dimensions: capacity
dimension, information dimension, and correlation dimension.

3 At the time of this writing FD3 package is available for dos, unix and mac platforms

from ftp://ftp.cs.csustan.edu/pub/fd3/.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 169/3

Examples 169

Capacity (or “box”) dimension is insensitive to the frequency with which

different cells are occupied. While it gives us some information about the
pattern of the supporting fractal set, it tells us nothing about the properties
of the fractal measure over the support set.
Concerning the information dimension, the documentation that is pro-
vided together with the FD3 package summarizes it as follows: “It measures
average surprise in learning which cell a point is in.”
Correlation dimension, on the other hand, gives us information about
distribution of pairs of points in cells with different distances. In the FD3
algorithm the counting of pairs is replaced by an essentially equivalent
estimate: Rényi generalized dimension with q = 2 (cf. e.g. [Schroeder,
1991, p. 220], [Harte, 2001, Ch. 8]).
The algorithm for producing the data set that we fed into the FD3
program is given below. We use the following formulas:
yi = (1 − k2 )x + 2k(1 + α x · ni )ni , (3.65)
pi (x) = ||yi ||. (3.66)
Begin
Select k in the range 0 < k < 1.
Select random unit vector x = (x(1), x(2), x(3))
Repeat n iter times
Begin
Calculate yi using the formula Eq. (3.65)
Calculate pi using Eq. (3.66)
Normalize pi : pi → pi /(sum(pj )), i, j = 1, ..6
Select i with probability pi
Set new x to be yi /||yi ||
If |x(1)| < 0.2 and |x(2)| < 0.2 and x(3) > 0 then write x(1), x(2)
to a file
End
End
In the estimation of fractal dimensions we normally run a certain num-
ber of iterations (for instance 100,000) just to get onto the attractor, there-
fore without saving the data, and then run the main loop until we get a
required number of points (for instance 1,000,000 points).
The data that have been obtained in this way were then fed into FD3
estimator. The results, for k varying between 0.3 and 0.9 are shown in
Fig. 3.43. The overall curves are rather regular, with somewhat strange
behavior around k = 0.1. The reason for these “hiccups” are unknown.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 170/3

170 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

k
Fig. 3.43 Fractal dimensions for Quantum Octahedron estimated by FD3 package. De-
pendence of the dimensions on the “fuzziness” parameter α.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 171/3

Examples 171

Estimation of the fractal dimension with another method leads, in gen-

eral, to deviations from the numbers outputted by FD3. For instance, for
k = 0.81 the algorithms used in FD3 produce outputs that are different
from those one would get by manual slope finding techniques — this is
demonstrated in Fig. 3.44.

3.3 Quantum fractals on n-spheres

Möbius transformations that are at the basis of quantum fractals can be also
defined in higher dimensions. Therefore quantum fractals exist in higher di-
mensions as well. In principle it would be possible to provide several simple
formulas — and we can start computer simulations of higher dimensional
quantum fractals. Some of these formulas have exactly the same form as
in the case of the two-sphere. Some others, however, as we will see, are not
so evident.
Generalizing certain concepts is usually not unique. Often it happens
that one can generalize a given result this way or another way. Different
generalizations may lead to different theories, each one with its own math-
ematical apparatus. For instance Euclidean geometry can be generalized
to an affine geometry (based on affine connection), Riemannian geometry
(based on metric tensor), Finsler geometry (based on Finslerian distance),
symmetric space geometry (based on group action) etc. These different
generalizations may partly overlap, but each one has its own path and
direction.
For quantum fractals two paths seems to be natural. Path one is to
follow the standard mathematical formalism of complex Hilbert spaces.
Quantum fractals generalized along this path would live on complex pro-
jective spaces. But then we lose sight of the growing number of applications
of quantum mechanical ideas beyond physics (cf. [Khrennikov (2008)], and
the corresponding discussion in Sec. 4.1). In physics, for some reasons that
are not completely clear, we are always using complex spaces.4 But in more
general, axiomatized approaches to quantum theory that start with, for
instance, quantum logic, other models are possible, in particular models in
real Hilbert spaces, without complex structure (cf. e.g. [Blank et al., 2008,
Notes to Chapter 13], and references therein).
4 David Hestenes, for instance, cf. [Hestenes (1985)], attributes the appearance of the

imaginary unit i in quantum theory to space-time signature (3, 1) and its Clifford algebra
structure.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 172/3

172 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

k = 0.81, Db = 0.78732
20

18

16

14

12
log2 (N ())

10

2
y = 0.82x + 0.53
0
0 5 10 15 20 25 30
log2 (1/)

Fig. 3.44 Estimation of the capacity dimension for k = 0.81. Demonstrates the differ-
ence between the standard least square fit based on a “visual” selection of the linear part
of the log2 of counts versus log2 of the cell sizes, and the one given by FD3 method of
averaging.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 173/3

Examples 173

While every complex Hilbert space carries in a natural way the structure
of a real space, the converse is not true. For quantum jumps and for
quantum fractals complex structure is not needed. It is needed for the
continuous part of the Schrödinger evolution in quantum physics, but not
for the discontinuous transitions. For that reason in our generalization of
quantum fractals we will follow the path pointed out by the formalism of
Clifford algebras — as it seems to be most natural mathematically (even
though it needs a more advanced algebraic apparatus), while also offering
a new insight into possible new applications of the quantum mechanical
formalism.
We will explain now how the mathematical formalism of Clifford alge-
bras enters the scene here.
First of all the simplest generalization of the two-sphere is the n-sphere
S — the one-point compactification of Rn . Like in the case of the two-
n

sphere, the group that acts on S n by angle-preserving (conformal) transfor-

mations is SO(n + 1, 1). Its two-fold covering is the Spin group Spin(n, 1)
that is realized within the Clifford algebra C(n + 1, 1). It is within this
Clifford algebra that each point of S n can be realized as an idempotent —
much like in the two-dimensional case.
Such an approach, although it requires the use of Clifford algebra tech-
niques, is straightforward and leads to simple generalization of the formulas
needed for realization of quantum fractals.
We will start with a short technical introduction of the algebraic tools
needed for that purpose.

3.3.1 Clifford algebras

3.3.1.1 Notation
We denote by R the field of real numbers, and by R∗ the multiplicative
group R \ {0}. Let V be an n-dimensional real vector space endowed with
a non-degenerate quadratic form Q of signature (r, s), r + s = n. That
is, V admits an orthonormal basis ei , with Q(e1 ) = . . . = Q(er ) = 1,
Q(er+1 ) = . . . = Q(en ) = −1.
The scalar product of two vectors v, w ∈ V can be expressed in terms
of the quadratic form Q by the polarization identity:
1
(v, w) = (Q(v + w) − Q(v − w)) , (3.67)
4
so that we have:
||v||2 = (v, v) = Q(v). (3.68)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 174/3

174 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

When an orthonormal basis is selected and fixed, (V, Q) can be identified

with the n = r + s-dimensional space Rr,s of real vectors v with the scalar
product
(v, w) = v 1 w1 + ... + v r wr − v r+1 wr+1 − ... − v r+s wr+s . (3.69)

Definition 3.1. The Clifford algebra C(V, Q) is the free associative algebra
with unit 1, generated by V and such that
v 2 = Q(v)1 for all v ∈ V. (3.70)
The Clifford algebra for R r,s
is often denoted as Cr,s , (sometimes C(r, s)).

Notice that, according to the definition, the vector space V itself becomes
a part of the algebra: V ⊂ C(V, Q). It follows then from the polarization
identity Eq. (3.67) that, within the algebra,
1 
(v, w) = (v + w)2 − (v − w)2
4
1 2 
= (v + w2 + vw + wv) − (v2 + w2 − vw − wv)
4
1
= (vw + wvw). (3.71)
2
Therefore in C(V, Q) we have
vw + wv = 2(v, w). (3.72)
In particular orthogonal vectors anticommute within the algebra.
Let C = C(V, Q) be the Clifford algebra of (V, Q). The even and the odd
parts of C are denoted as C + and C − respectively. We shall consider R and
V as vector subspaces of C, so that v2 = Q(v) ∈ C, v ∈ V ⊂ C.
The definition of the Clifford algebra contains the word ‘free’. Freeness
means in this case that there are no other relations within the algebra except
of those implied by the defining relations Eq. (3.70). To give an example:
consider the real line R with the standard quadratic form Q(x) = x2 , x ∈ R.
The Clifford algebra C(R) should have in this case one generator, denoted
by x, with the property x2 = 1. In the free algebra this generator should be
different from 1. Yet, if we set x = 1, then the relations Eq. (3.70) are also
satisfied. The free algebra C(V, Q) can be characterized by its “universality
property” as follows:

Theorem 3.2. Let A be a real, associative, algebra with unit 1A , and let
φ : V → A be a linear mapping with the property φ(v)2 = Q(v)1A for all
v ∈ V. Then there exists a unique algebra homomorphism Φ : C(V, Q) → A
that extends φ, that is such that Φ(v) = φ(v) for v ∈ V.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 175/3

Examples 175

Remark 3.6. Some authors adopt a different sign convention. They write
the defining equation as v2 = −Q(v)1. Therefore one has to be careful
comparing different texts. The translation between the two conventions is
easy: Cr,s in one convention corresponds to Cs,r in the other one.

It follows from this universal property that the Clifford algebra C(V, Q) is
essentially unique, i.e. unique up to an isomorphism.

Theorem 3.3. As a (real) vector space the Clifford algebra Cr,s has the
dimension 2r+s . If ei , i = 1, ..., n = r + s, is an orthonormal basis in V,
with e2i = ±1, then 2n elements ei11 ei22 ...einn , ij = 0 or 1, form a basis for
the Clifford algebra C(V, Q).

Proof. See [Deheuvels, 1981, Thm. VIII.2.A, p. 293].5 

In practical applications it is often useful to have the Clifford algebra “at
hand”, that is to have a particular realization of the properties that are
given in a non-constructive way in the definition. One way to define the
Clifford algebra semi-constructively is by using the infinite-dimensional ten-
sor algebra of V and divide it by the (also infinite-dimensional) ideal gen-
erated by the defining relations. Yet using infinite-dimensional spaces in
finite-dimensional applications seems to be like an overkill. Although math-
ematically correct it does not help in applications. Therefore a better and
often used way is to describe the Clifford as a matrix algebra. This is al-
ways possible and rather simple, provided that we are using matrices with
entries that are either real or complex, or quaternions, depending on the
signature (r, s).

3.3.1.2 Examples of Clifford algebras

• The algebra of complex numbers C. As a real algebra it is 2 = 21 -
dimensional. Its defining relation is i2 = −1. Therefore C is the Clifford
algebra C0,1 .
• The algebra of quaternions. Its real dimension is 4 = 22 . Its defining
relations are i2 = j 2 = k2 = −1, ij = k, jk = i, ki = j. Because of the
dimensionality it may be only the Clifford algebra of a two-dimensional
space. The relation i2 = j 2 = −1 suggest that it is the Clifford algebra
C0,2 . And indeed this is the case.
• The algebra R[2] of real 2 × 2 matrices is four-dimensional. Therefore
it can be a Clifford algebra of a two-dimensional real vector space. We
5 More details in Theorem 3.7.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 176/3

176 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

can introduce the following two matrices:

1 0 01
e1 = e2 = . (3.73)
0 −1 10

Then e1 and e2 anticommute and

0 1
e1 e 2 = . (3.74)
−1 0

The matrices e1 , e2 , e1 e2 , together with the identity matrix 1, span the

whole algebra. We have e21 = e22 = 1. Therefore R[2] can be identified
with the Clifford algebra C2,0 .

On the other hand we can choose:

1 0 0 1
e1 = , e2 = . (3.75)
0 −1 −1 0

By a similar reasoning as above R[2] can be then identified with the

Clifford algebra C1,1 . On this example we see that Clifford algebras of
two different scalar product spaces can be isomorphic.
• The algebra C[2] of complex 2 × 2 matrices has real dimension 8 = 23 .
Therefore it may be the Clifford algebra of a three-dimensional real
vector space. Let s1 , s2 , s3 be the standard Pauli matrices:

01 0 −i 1 0
s1 = , s2 = , s3 = . (3.76)
10 i 0 0 −1
Then the eight matrices 1, i, s1 , s2 , s3 , is1 , is2 , is3 span the whole alge-
bra. As generators we can choose s1 , s2 , s3 which leads to the algebra
C3,0 . On the other hand we can choose is1 , is2 , is3 , which leads to the
algebra C1,2 . One could think that we can also choose s1 , s2 , is3 , but
this is not a good choice as s1 s2 = is3 , so s1 , s2 , is3 are not algebraically
independent.

Table 3.2 gives the table of low-dimensional Clifford algebras Cr,s . R[n],
C[n], H[n] denote the (real) algebras of n × n matrices with entries that
are real numbers, complex numbers, and quaternions respectively. R2 [n]
denotes the algebra of real 2n × 2n block matrices with two n × n matrices
on the diagonal, zeros otherwise. Thus, for instance, R2 [1] denotes the
algebra of 2 × 2 matrices of the form ( a0 0b ) , where a, b ∈ R. Similarly for
C2 [n] and H2 [n].
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 177/3

Examples 177

Table 3.2 Clifford algebras of low-dimensional spaces.

0 1 2 3 4 5 6 7 8
0 R R[1] R[2] C[2] H[2] H2 [2] H[4] C[8] R[16]
1 C R[2] R2 [2] R[4] C[4] H[4] H2 [4] H[8] C[16]
2 H C[2] R[4] R2 [4] R[8] C[8] H[8] H2 [8] H[16]
3 H2 H[2] C[4] R[8] R2 [8] R[16] C[16] H[16] H2 [16]
4 H[2] H2 [2] H[4] C[8] R[16] R2 [16] R[32] C[32] H[32]
5 C[4] H[4] H2 [4] H[8] C[16] R[32] R2 [32] R[64] C[64]
6 R[8] C[8] H[8] H2 [8] H[16] C[32] R[64] R2 [64] R[128]
7 R2 [8] R[16] C[16] H[16] H2 [16] H[32] C[64] R[128] R2 [128]
8 R[16] R2 [16] R[32] C[32] H[32] H2 [32] H[64] C[128] R[256]
The algebras Cr,s . The index r increases vertically, s increases horizontally.

3.3.1.3 Even and odd parts, principal automorphism and

anti-automorphism, conjugation
The subspace spanned by the products of an even number of vectors of
V forms a subalgebra C + (V, Q). If ei is an orthonormal basis of V, then
C + (V, Q) is spanned by the products ei11 ei22 ...einn with an even number of ij
being equal to 1. The complementary subspace C − (V, Q) is spanned by the
products of an odd number of vectors ei . The linear subspace C − (V, Q) is
not an algebra as the product of two elements of C − (V, Q) is in C + (V, Q).
Both C + (V, Q) and C − (V, Q) are of dimension 2n−1 — half of the dimension
of the whole Clifford algebra.

Theorem 3.4. Let C be the Clifford algebra of (V, Q).

• There exists a unique algebra automorphism π of C such that

π(v) = −v, v ∈ V.
• There exists a unique algebra anti-automorphism τ, denoted also as
τ (a) = aτ , such that v τ = v, v ∈ V.
• Their composition ν is the unique anti-automorphism satisfying
ν(v) = −v for all v ∈ V.

Proof. See [Deheuvels, 1981, Ch. VIII.2]. 

The automorphism π is called the principal automorphism of C. It is the

identity map on C +, while it multiplies every element of C − by −1. The
anti-automorphism τ is called the principal anti-automorphism of C. Their
composition ν = π ◦ τ = τ ◦ π is called the conjugation of C.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 178/3

178 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

3.3.1.4 Examples
(1) Consider the algebra R[2] as C2,0 with generators e1 , e2 as in Eq. (3.73).
Denote by C the matrix e1 e2 as in Eq. (3.74). Then C 2 = −1, C −1 =
−C, and Ce1 C −1 = −e1 , Ce2 C −1 = −e2 . Therefore the map A →
CAC −1 is an automorphism of C2,0 = R[2] that maps v ∈ V into −v.
Thus, in C2,0 realized as R[2], we have that π(A) = CAC −1 .
On the other hand the matrix transpose operation A → AT is an
automorphism of R[2]. Moreover, the matrices e1 , e2 are symmetric,
therefore eT1 = e1 , eT2 = e2 . It follows that τ (A) = AT , and ν(A) =
CAT C −1 .
(2) Now let us consider R[2] as C1,1 with generators e1 , e2 as in Eq. (3.75).
Let B = e1 , C = e1 e2 . Then π(A) = CAC −1 , τ (A) = BAT B −1 , ν(A) =
e2 AT e−12 .
(3) Consider the algebra C[2] as a realization of C3,0 with generators
e1 = s1 , e2 = s2 , e3 = s3 as in Eq. (3.76). The matrices e1 , e2 , e3 are
Hermitian, and the Hermitian conjugation is an anti-automorphism in
C[2]. Therefore in C3,0 realized as C[2] we have τ (A) = A∗ . Now, let
C = s2 . Then C 2 = 1, C̄ = −C, C T = −C. It follows that A → C ĀC −1
is an automorphism of C[2] that acts as multiplication by −1 on all
three generators. It follows that π(A) = C ĀC −1 . For the conjugation
in this Clifford algebra we then get ν(A) = CAT C −1 .
(4) On the other hand consider C[2] as C1,2 with generators e1 = s1 , e2 =
is2 , e3 = is3 . This time π(A) = C ĀC −1 , with C = e3 , τ (A) = e1 A∗ e−1
1 ,
T
ν(A) = s2 A s2 .

3.3.1.5 The norm function

We will denote by Δ the norm function Δ : C −→ C, defined by

Δ(a) = aν a. (3.77)

While Δ(a) is, in general, an element of the algebra C, for some a ∈ C it

may happen that Δ(a) is, in fact, a real number Δ(a) ∈ R ⊂ C. In fact, in
all four examples of Clifford algebras in 3.3.1.4 Δ(A) = det(A)1.

Theorem 3.5. If Δ(a), Δ(b) ∈ R, then

(i) Δ(ab) = Δ(a)Δ(b),

(ii) Δ(π(a)) = Δ(τ (a)) = Δ(aν ) = Δ(a),
(iii) For all λ ∈ R, Δ(λa) = λ2 Δ(a).
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 179/3

Examples 179

(iv) Moreover, if Δ(a) ∈ R∗ , then a is invertible, and a−1 = (1/Δ(a))aν .

(v) In particular, if Δ(a) = aν a ∈ R, then also aaν = Δ(a).

Proof. See [Gilbert and Murray, 1991, 5.14–5.16]. 

3.3.1.6 Spin group

From theorem 3.5 it follows that the set of all elements a of C(V, Q) for
which Δ(a) = 1 is a group. In applications it often happens that Clifford
algebra elements are implementing operations on the vector space V.

Definition 3.2. We denote by Spin+ (V, Q) the group6 :

Spin+ (V, Q) = {g ∈ C + (V, Q) : Δ(g) = 1, gV g −1 = V }. (3.78)

Theorem 3.6.

(i) Every element g ∈ Spin+ (V, Q) is a product of an even number of

positive unit vectors (i.e. vectors v ∈ V such that Q(u) = +1) and an
even number of negative unit vectors (i.e. v ∈ V such that Q(u) = −1).
(ii) The map s : Spin+ (V, Q) → SO+ (V, Q), s(g) : v → gvg −1 is a two-
fold covering homomorphism from Spin+(V, Q) onto SO+ (V, Q), the
connected group of “proper rotations”, that is orthogonal transforma-
tions of (V, Q) of determinant one, which preserve the orientation of
maximal negative subspaces of V.

Proof. See [Deheuvels (1981)]. 

3.3.1.7 Vector space isomorphism between the Clifford and the

Grassmann algebra
As a vector space, Clifford algebra C = C(V, Q) is naturally graded and

isomorphic to the exterior algebra V of V. In particular we have the
following result:

Theorem 3.7. Let ei , i = 1, 2, . . . , n be an orthonormal basis for V, and

let eI : I = (i1 , i2 , . . . , ip ), 1 ≤ i1 < i2 < . . . < ip ≤ n be defined as the
Clifford products eI = ei1 ei1 . . . eip , with eI = 1 for I = ∅. Then the set
6 The group Spin+ is denoted simply as Spin in [Deheuvels (1981)], [Anglès, 2007,

2.4.2], and as Spin0 in [Gilbert and Murray (1991)]. The case of (r, s) = (1, 1) is special,
as in this case the group Spin+ has two disconnected components — cf. [Deheuvels,
1981, p. 369]
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 180/3

180 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

{eI } of 2n vectors in C is a linear basis of C, the subspaces Cp generated by

eI , I = (i1 , . . . ip ) are independent of the choice of the orthonormal basis
ei , and C is the direct sum of vector subspaces Cp :
 n
C= Cp (3.79)
k=0
Moreover, for each p = 0, . . . , n the skew-symmetric map αp from V × V ×
. . . × V (p times) to C given by:
1 
αp (x1 , x2 , . . . , xp ) = (−1)s xs1 xs2 . . . xsp ,
p! s

determines an isomorphism of the vector subspace p V of the exterior
 p
algebra V onto Cp that sends ei1 ∧ . . . ∧eip ∈ V to ei1 . . . eip ∈ Cp ⊂ C.

Proof. See [Deheuvels, 1981, Theoreme VIII.10]. 

Example Consider the algebra C3,0 realized as C[2], with generators ei =

si , i = 1, 2, 3. The subspace C0 is just the real line R. The subspace C3 is
spanned by the matrix e1 e2 e3 = i1. The subspace C1 is spanned by three
Hermitian traceless matrices e1 , e2 , e3 . Therefore it consists of Hermitian
traceless matrices. It is three-dimensional. The subspace C2 is spanned
by the products e1 e2 = is3 , e2 e3 = is1 , e3 e1 = is2 . All these matrices are
anti-Hermitian and traceless. Therefore C2 is also three-dimensional and
consists of anti-Hermitian traceless matrices.

3.3.1.8 Paravectors

Definition 3.3. Of particular interest for us will be the (n+1)-dimensional

subspace C0 ⊕ C1 ⊂ C of paravectors. We will denote this subspace by V 1
and endow it with the quadratic form Q1 defined by
Q1 (x0 , v) = (x0 )2 − v 2 , x0 ∈ R, v ∈ V. (3.80)
1
If Q is of signature (r, s), then Q has signature (s + 1, r).

3.3.1.9 Example
In the case of C3,0 realized as C[2] the subspace of paravectors consists of
all Hermitian matrices. Every Hermitian matrix can be decomposed into a
sum of (real) multiples of the identity matrix, and the traceless part. Thus
we can write:
x0 + x3 x1 − ix2
A = x0 · 1 + x1 s1 + x2 s2 + x3 s3 = (3.81)
x1 + ix2 x0 − x3 .
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 181/3

Examples 181

This is the most general paravector. We can calculate then Δ(A) =

det(A) = (x0 )2 − (x1 )2 − (x2 )2 − (x3 )2 , which is nothing but the standard
“length squared” of a vector in Minkowski space-time.

3.3.1.10 The trace

We denote by Φ the linear functional on C assigning to each element a ∈ C
its scalar part Φ(a) = a0 ∈ C0 in the decomposition Eq. (3.79). Then the
following proposition holds:

Proposition 3.2. The functional Φ has the following properties:

(i) Φ(1) = 1,
(ii) Φ(aτ ) = Φ(a), ∀a ∈ C,
(iii) Φ(ab) = Φ(ba), ∀a, b ∈ C,
df
(iv) (a, b) = Φ(aτ b) is a nondegenerate, symmetric, bilinear form on
C, that is positive definite if the original quadratic form on V
is positive definite. We have Φ(a) = (1, a) = (a, 1), ∀a ∈ C.
(v) (ab, c) = (b, aτ c) = (a, cbτ ), ∀a, b, c ∈ C.

Proof. (i) and (ii) follow immediately from the definition. In order to
prove (iii) notice that if {ei }, i = 1, . . . , n, is an orthonormal basis in
V, {eI }, I = {i1 < . . . < ip } is the corresponding basis in C, and a =
 I

I aI e , 
b = I bI eI are the decompositions of a and b in the basis eI , then
Φ(ab) = I aI bI Φ(eI eI ) = Φ(ba). From the very definition of the scalar
product (a, b) it follows that (a, b) = Φ(aτ b) = Φ((aτ b)τ ) = Φ(bτ a) = (b, a).
Moreover, we have (eI , eJ ) = 0 if I = J, and also (eI , eI ) = ei1 2 . . . eip 2 =
(−1)s(I) , where s(I) is the number of negative norm square vectors in I. In
particular eI is orthonormal with respect to the scalar product in C, and so
(iv) holds. We have (ab, c) = Φ((ab)τ c) = Φ(bτ aτ c) = Φ(aτ cbτ ) = (a, cbτ ),
which establishes (v). 

Remark 3.7. : It is easy to see that Φ(a) = (1/2n)tr(L(a)), where L(a) is

the left multiplication by a acting on C: L(a)b = ab, and the trace is taken
over C, see e.g. [Weil, 1980, p. 601], for a general discussion. Because of
this property Φ will be called a trace.

We will call an element a ∈ C positive, which we will write a ≥ 0, if a = aτ

and (v, av) ≥ 0 for all v ∈ C. Equivalently, a ≥ 0 if and only if a is of
the form a = bτ b, for some b ∈ C (cf. e.g. [Axler, 1997, 7.27]). If a is
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 182/3

182 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

positive and a = 0, we will write a > 0. If a ≥ 0, then, in particular,

Φ(a) = (1, a1) ≥ 0 and, if a > 0 then Φ(a) > 0.

3.3.1.11 Algebra isomorphism between C +(V 1 , Q1 )

and R(2, C(V, Q))
It is a well known fact (see e.g. Ref. [Gilbert and Murray (1991)], I.6.13)
that the algebras C +(V 1 , Q1 ) and C(V, Q) are isomorphic. What we need
here is a description of this isomorphism in more detail.

Notation: In what follows we will use the notation C = C(V, Q), and
C 1 = C(V 1 , Q1 ).

Let the map A : C × C → R(2, C) be defined by

 
a b
A(a, b) = : a, b ∈ C , (3.82)
π(b) π(a)
and let γ : V 1 → R(2, C) be the linear map given by
0 x0 + v 0 x0 + v
γ(x0 , v) = = . (3.83)
x −v 0
0 0
π(x + v) 0
Then γ is evidently the Clifford map, γ(x0 , v)2 = Q1 (x0 , v)I, and therefore
it extends to a unique algebra homomorphism, which we will denote by the
same symbol γ, from C 1 to R(2, C). We will define now several maps shown in
Fig. 3.45, and study their properties. The map pr11 : R(2, C) → C assigns
to each matrix in R(2, C) its top-left entry. For instance pr11 (A(a, b)) =
a. Im(A) is the set of all matrices of the form Eq. (3.82). We will not

C×C
A - Im(A)
QQA 
3
Ã= γ̃ −1 ◦A γ̃
? γ QQ
s ?
C1 - R(2, C)
6 QQ ψ
Q pr11
Q
s ?
ψ+ Q
Q
1+ - C
C
Fig. 3.45 Maps between the important algebras.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 183/3

Examples 183

distinguish between the maps A : C × C → R(2, R) and A : C × C → Im(A),

which differ only by the canonical inclusion Im(A) → R(2, C). But we will
distinguish between γ : C 1 → R(2, C) and γ̃ : C 1 → Im(A). The latter map
is an algebra isomorphism, therefore γ̃ −1 : Im(A) → C 1 is well defined. The
map ψ is defined as ψ = pr11 ◦ γ̃, and is an algebra homomorphism, and ψ +
+
is its restriction to C 1 . We will use the notation Ã(a, b) for γ̃ −1 (A(a, b)).

Theorem 3.8.
(i) Let us realize the Clifford algebra C(1, −1) as the matrix algebra R[2]
using the following basis
0 1 01 −1 0
f0 = , f1 = , f01 = , (3.84)
−1 0 10 0 1
so that we have
f02k = 12 , f02k+1 = f0 , f12k = (−1k )12 , f12k+1 = (−1)k f1 . (3.85)
Let {e0 ∈ R, ei ∈ V, i = 1, . . . , n + 1} be an orthonormal basis of V 1 .
Then, in terms of this basis the map γ : C 1 −→ M at(2, C) reads:
γ(1) = 12 ⊗ 1C
γ(e0 ei1 . . . ei2k ) = (−1)k f0 ⊗ ei1 . . . ei2k ,
γ(e0 ei1 . . . ei2k+1 ) = (−1)k f01 ⊗ ei1 . . . ei2k+1 ,
γ(ei1 . . . ei2k ) = (−1)k 12 ⊗ ei1 . . . ei2k ,
γ(ei1 . . . ei2k+1 ) = (−1)k f1 ⊗ ei1 . . . ei2k+1 .
−
(ii) ker(ψ) = C 1 , and ψ restricts to the algebra isomorphisms ψ + from
+
C 1 onto C. In terms of the basis we have:
⎫
ψ + (1C 1 ) = 1C , ⎬
ψ+ (ei1 . . . ei2k ) = (−1)k ei1 . . . ei2k , (3.86)
⎭
ψ+ (e0 ei1 . . . ei2k+1 ) = (−1)k ei1 . . . ei2k+1 .
(iii) With the notation as above, we have
A(a, b)A(a , b ) = A(a , b ), where
a = aa +bπ(b ), b = ab +bπ(a ).
(3.87)
The principal involution π and the principal anti-involution τ of C 1
can be expressed through their corresponding operations in C as
π(Ã(a, b)) = Ã(a, −b), (3.88)

τ (Ã(a, b)) = Ã(ν(a), τ (b)). (3.89)

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 184/3

184 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

+
The even subalgebra C 1 of C 1 can then be identified with the set of all
A(a, b), with b = 0, that is, using the map pr11 , with C.
(iv) Denoting by Φ1 (resp. Δ1 ), and Φ (resp. Δ) the trace (resp. norm
function) of C 1 and C respectively, we have

Φ1 = Φ ◦ ψ, (3.90)

Δ1 (ã) = Δ(ψ + (ã)), ∀ ã ∈ C 1 , (3.91)

(ã, b̃) = (π(ψ + (ã)), ψ + (b̃)), ∀ã, b̃ ∈ C 1 . (3.92)

(v) g̃ ∈ Spin(V 1 , Q1 ) if and only if g = ψ + (g̃) satisfies

a) Δ(g) = 1, and
b) gV 1 g τ = V 1 .

Proof. (i) and (ii) follow by a straightforward calculation.

(iii) By a straightforward matrix multiplication we get from Eq. (3.82) that

A(a, b)A(a , b ) = A(a , b ), where a = aa + bπ(b ), b = ab + bπ(a ).
(3.93)
It follows that the range (image) of the map A is an algebra and, because
it has the right dimension 2 × dim(V ), the Clifford map γ extends to the
isomorphism of C 1 onto Im(A). It is also clear that the even subalgebra
of C 1 is represented by the matrices A(a, 0), while the odd subspace is
represented by matrices A(0, b).
It follows from the very definition that π and τ defined by Eq. (3.88)
and Eq. (3.89) are involutions, and that π(ψ(w)) = ψ(−w), τ (γ(w)) = γ(w)
for w ∈ V 1 . Therefore we need to show that π, defined by Eq. (3.88), is an
automorphism, and that τ, defined by Eq. (3.89), is an anti-automorphism.
(Notice that although, by abuse of the notation, we denote by the same
symbol π the main automorphisms of C and C 1 , the meaning is always clear
from the context.)
1 0
Let C be the matrix7 : C = , then
0 −1

CA(a, b)C −1 = A(a, −b), (3.94)

therefore the formula Eq. (3.88) defines an involutive automorphism of C 1 ,

and, since it reverses the signs of vectors, it defines the principal involution
of C 1 .
7 Cf. also [Deheuvels, 1981, Ch. VIII.6, p. 310], where the matrix C is used to define

an anti-involution of the algebra C[2].

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 185/3

Examples 185

Proving that τ is an anti-automorphism of C 1 follows by a straightfor-

ward calculation using Eq. (3.82) and the properties of π and τ on C.
(iv) follows from (i)–(ii). Finally, (v) follows from (ii) and (iv) — (cf.
also Ref. [Gilbert and Murray, 1991, Theorem 6.12]). 

From now on we will assume that (V, Q) is an (n+1)-dimensional Euclidean

space, that is that Q has the signature (n + 1, 0).

3.3.1.12 Möbius transformations

Notation Let (V, Q) be an (n + 1)-dimensional Euclidean space, n > 0.
Vectors in V we denote by bold symbols: x, y, etc. The term “Euclidean”
means that the quadratic form Q is positive definite:

n+1
Q(x) = x2 = (xi )2 . (3.95)
i=1

The bold symbol n will be reserved for unit vectors, thus


Q(n) = n2 = i = 1n+1 (ni )2 = 1. (3.96)

We will denote by B n+1 the open unit ball

B n+1 = {x ∈ V : x2 < 1}, (3.97)

by B̄ n+1 its closure

B̄ n+1 = {x ∈ V : x2 ≤ 1}, (3.98)

and by S n its boundary, the unit sphere

S n = {n ∈ V : n2 = 1}. (3.99)

We will denote by C the Clifford algebra C(V, Q), by Spin+n+1 the group
+
Spin+(V, Q), and by Spin1n+1 the group ψ + (Spin+ (V1 , Q1 )), described
by the conditions a) and b) in Theorem 3.8, (v). We define the Clifford
group Γ(V) as

Γ(V) = {w1 . . . wk : wj ∈ V1 , Δ(wj ) = 0}. (3.100)

It is evident that this group is closed under π, τ, ν.

+
We will describe the action of Spin1n+1 on the unit sphere S n , and on
its interior B n+1 . As the main tool we will use the special class of elements
of CV , that are called transformers.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 186/3

186 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

3.3.1.13 Conformal spin geometry of n-spheres

The sphere S 2 is contained in R3 . Of course the rotation group of R3 acts
on S 2 preserving its geometry. But the Lorentz group acts also on the
sphere S 2 , this time by conformal transformations (Möbius transforma-
tions), preserving all angles, though not lengths. Angles are important in
optics. Now, the Lorentz group has its double covering group, the spin
group, that is isomorphic to SL(2, C). In SL(2, C) we can easily distin-
guish rotation subgroup SU (2), but SL(2, C) contains also Lorentz boosts
(special Lorentz transformations). These are represented by positive ma-
trices. In a strange way optics connects here to quantum mechanics of spin
1/2 particles. Rotations correspond to unitary transformations of quantum
mechanics, boosts correspond to simple operators implementing quantum
spin measurements. We want to generalize this correspondence replacing
S 2 by S n . In fact this is possible using the mathematics of Clifford algebra,
though it requires some effort. First step in this direction have been made
by Gilbert and Murray in their monograph [Gilbert and Murray (1991)].

Definitions We will be working with the (n + 1)-dimensional space de-

noted here simply as V. Its Clifford algebra will be denoted simply as C. As
an example one should think of V as R3 and of C as of the complex 2 × 2
matrices C[2].
C contains V, but it also contains real numbers R. Together they form
the space V1 of paravectors. V1 is (n + 2)-dimensional and equipped with
the quadratic form of signature (1, n + 1). One should think of V1 as of a
generalization of Minkowski space of Special Relativity. Vectors of V1 will
be transformed by some special elements of the Clifford algebra C.
The important role in these transformations will be played by the prin-
cipal anti-automorphism τ. In the case of C = C[2], τ is nothing but the
Hermitian conjugate of a matrix. We will restrict our attention to invert-
ible elements of the Clifford algebra. Let us now define several subsets of
C that will be used in the following discussion. After this definition we will
identify these subsets for the case of n = 2.
Definition 3.4. An element a of C is called a transformer if for every
element w ∈ V1 there exists another w ∈ V1 such that
aw = w π(a). (3.101)
The set T of all transformers is a multiplicative semigroup. Moreover we
have the following important result proven in [Gilbert and Murray, 1991,
5.24–5.29]:
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 187/3

Examples 187

Theorem 3.9. The set of all transformers T is closed under the principal
automorphism π. Moreover, for every a ∈ T , Δ(a) ∈ R, and if Δ(a) = 0,
then also aτ ∈ T . The set of all invertible transformers coincides with
Clifford group Γ(V).

Lemma 3.1. If a is an invertible transformer, then for every w ∈ V1 we

have
df
sa (w) = awaτ ∈ V1 . (3.102)

Proof. We first notice that τ (w) = w, ∀w ∈ V1 . Applying τ to both sides

of the defining equation Eq. (3.101) we get waτ = aν w . Multiplying by a
from the left, we get awaτ = aaν w . But since Δ(a) = Δ(π(a)) = aν a =
aaν ∈ R, we get awaτ = Δ(a)w ∈ V1 . 
Motivated by the above lemma we define the subsets M, M+ ∈ C as
follows:
M = {a ∈ C : aV 1 aτ ⊂ V1 }. (3.103)
M+ = {a ∈ M : a > 0, Φ(a) = 1} (3.104)

Definition 3.5. We define the following important subsets of T and of

M+ :
G = {a ∈ T : Δ(a) = 1}, (3.105)
GR = {a ∈ G : aa = 1}, τ
(3.106)
G+ = {a ∈ G : a ≥ 0}, (3.107)
M1+ = {a ∈ M+ : Δ(a) > 0}, (3.108)
M̄1+ = {a ∈ M+ : Δ(a) ≥ 0}, (3.109)
M0+ = {a ∈ M+ : Δ(a) = 0}. (3.110)

Notice that, by the Theorem 3.9, G is invariant under both π and τ. It

is sometimes denoted as Spin0 (V), and the map s (cf. Eq. (3.102)) is
a two-fold covering homomorphism: G → SO+ (V1 , Q1 ) — cf. [Gilbert
+
and Murray, 1991, 6.12]. Thus G is nothing but Spin1 . GR leaves the
subspace V ⊂ V1 invariant and s, when restricted to GR , is a two-fold
covering homomorphism of SO(V, Q).
The elements of G+ , that will be studied in our paper in some details,
will be called spin-boosts. M is a multiplicative semigroup, and G ⊂ M.
We will show that M0+ is naturally isomorphic to the unit sphere S n ,
while M1+ (respectively M̄1+ ) corresponds to the open unit ball B n+1
(respectively its closure B̄ n+1 ).
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 188/3

188 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Lemma 3.2. Let a ∈ V1 , a =  0, 1, Φ(a) = 1, Δ(a) ≥ 0. Then a > 0, and

a is of the form a = 1 + kn, 0 < k ≤ 1, n ∈ V, n2 = 1. If Δ(a) > 0, then
√
√ 1 1 − 1 − k2
a= √ (1 + n), where  = . (3.111)
1 + 2 k
√
If Δ(a) = 0, then k = 1, a = 1 + n, and a = √12 a.

Proof. Since a ∈ V1 , a = x0 + x, x0 ∈ R, x ∈ V. Since Φ(a) = x0 ,

and Δ(a) = (x0 )2 − x2 , it follows that x0 = 1, x2 ≤ 1. Let us write a as
a = 1 + kn, 0 < k ≤ 1, n2 = 1. Consider first the case of Δ(a) = 1 − k2 > 0,
√
df 2
i.e. k < 1. Let b = √1+ 1
2
(1 + n), where  = 1− k1−k . Then bτ = b,
and, by simple algebra, we get 0 <  < 1, bτ b = b2 = a. Thus a > 0.
But now b has the same form as a (up to a positive multiplicative factor),
√
therefore also b > 0. Then b = a follows from the uniqueness of a positive
square root of a positive element. If Δ(a) = 0, i.e. a = 1 + n, √then
√
a2 = (1 + n)2 = 1 + 2n + n2 = 2a, and therefore a > 0, and a = a/ 2.
The following proposition characterizes explicitly the sets M0+ , M1+ ,
M̄1+ .

Proposition 3.3. Let P : V ⊃ B̄ n+1 → V1 ⊂ C be the map

P (x) = 1 + x, x ∈ B̄ n+1 . (3.112)
Then P is a bijection P : B̄ n+1 → M̄+
1 , P (S ) = M0+ , and P (B
n n+1
)=
M1 .
+

Proof. If x = 0, then P (x) = 1, which is evidently in M̄+ 1 . Let us

therefore assume 0 < x2 ≤ 1. With a = P (x), we have a = aτ , Φ(a) = 1,
Δ(a) = 1 − x2 ≥ 0, therefore, by Lemma 3.2, a > 0. Moreover, by a simple
calculation, we find that if w = y 0 + y ∈ V 1 , then
awa = y 0 (1 + x2 ) + 2(x · y) + (1 − x2 )y + 2(y 0 + (x · y))x ∈ V 1 . (3.113)
Therefore a ∈ M̄+ 1 . To show that P is a surjection onto M̄1+ , let a be an
arbitrary element in M̄1+ . Then a2 = a(1 + 0)a must be in V1 . Let us
therefore write a2 = y 0 + y. Now y 0 = Φ(a2 ) > 0, and Δ(a2 ) = Δ(a)2 ≥ 0,
Therefore, we can write a2 = y 0 (1+kn), k ≤ 1. Then it follows from Lemma
3.2 that a2 has a square root in V 1 and, because of the uniqueness of the
square root, a itself must be in V 1 . But, since Φ(a) = 1, and Δ(a) ≥ 0,
it follows that a = 1 + x, x2 ≤ 1. This shows that P is a bijection. The
remaining statements follow from Δ(P (x)) = Δ(1 + x) = 1 − x2 . 
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 189/3

Examples 189

The following proposition and its corollary describe the set of spin-boosts
G+ , and the Iwasawa-type decomposition of G.

Proposition 3.4. m ∈ G+ if and only if m is of the form

1 + kn
m= √ , n ∈ Sn, 0 ≤ k < 1. (3.114)
1 − k2
An equivalent form is that of
η 
m = exp n , k = tanh(η/2), η > 0. (3.115)
2
Proof. The sufficient condition: With m, n, and k as in Eq. (3.114), it
follows from the Proposition 3.3 that 1 + kn > 0. On one hand Δ(m) = 1,
thus m ∈ G+ . On the other hand, since n2 = 1, is easy to calculate the
exponential in Eq. (3.115), the result being:
η 
exp n = cosh(η/2) + sinh(η/2)n. (3.116)
2
It is then easy to see that by setting k = tanh(η/2), η > 0, we recover
Eq. (3.114).
The necessary condition. We can assume that m = 1. Suppose m ∈ G+ ,
then Φ(m) > 0, and thus m/Φ(m) ∈ M+ 1 . It follows from the Lemma
3.2 that m is proportional to 1 + kn, 0 < k < 1, n2 = 1. √ Then, from
Δ(m) = 1 it follows that the proportionality coefficient is 1/ 1 − k 2 . 

Corollary 3.1. G = G+ GR . Every element g ∈ G has a unique decomposi-

tion into the product
g = mu, m ∈ G+ , u ∈ GR . (3.117)

Proof. Let g ∈ G. If g = 1, then there is nothing to prove, as we take

m = 1, u = 1. Let us therefore assume g = 1. Using the Polar Decomposition
Theorem (cf. e.g. [Axler, 1997, p. 153]), g can be written, in a unique way,
as g = mu, where m2 = gg τ > 0, and uuτ = uτ u = 1. We need to show
that m ∈ G+ and u ∈ GR . Now, since G is invariant under τ, it follows that
m2 = gg τ ∈ G+ . Therefore, by (i), m2 can be written as m2 = exp(ηn/2)
and, from the uniqueness of the square root, m = exp(ηn/4). Therefore
m ∈ G+ . It follows that u = m−1 g ∈ G, and so u ∈ GR . 

Remark 3.8. The decomposition given in Eq. (3.117) corresponds to the

well known decomposition of Lorentz transformations into “boosts” and
“space rotations.” The special case of n = 2, and SO+ (1, 3), though not at
the Clifford algebra level, is treated in details in Ref. [Moretti (2002)]
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 190/3

190 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Let us now describe the action of the group G on B̄ n+1 . We will need the
following lemma, which is the result of a simple, though somewhat lengthy,
calculation in the Clifford algebra C.
√
Lemma 3.3. If m = P (kn)/ 1 − k2 ∈ G+ , then for all x ∈ B̄ n+1 we have
P (kn)(1 + x)P (kn) = (1 + k 2 + 2k (n · x))(1 + x ), (3.118)

1 + α2 + 2α(n · x)
m(1 + x)m = (1 + x ), (3.119)
1 − k2
where
(1 − k 2 )x + 2k(1 + k (n · x))n
x = . (3.120)
1 + k 2 + 2k (n · x)
Proof. Straightforward calculation. 
Before stating the next theorem let us notice that if x2 ≤ 1, then P (x) =
1 + x > 0. If g ∈ G, then also gP (x)g τ > 0 and, therefore, Φ(gP (x)g τ ) > 0.
Since G ⊂ M, and since M is a multiplicative semigroup, it follows that
gP (x)g τ ∈ M, and therefore gP (x)g τ /Φ(gP (x)g τ ) ∈ M̄1+ .
We also recall the definition of a conformal transformation (see e.g. Ref.
[Goldberg, 1998, Ch. 3.7].

Definition 3.6. A diffeomorphism φ of a Riemannian manifold (M, G) is

called a conformal transformation if there is a function ρ > 0 on M such
that
(φ∗ G)αβ = ρ2 Gαβ .
If n = dim(M ) ≥ 3 then the group of conformal transformations of M
is a Lie group of dimension ≤ (n+1)(n+2)
2 , and for the spheres S n , that
are of particular interest in our paper, the upper limit is reached — cf.
e.g. [Kobayashi and Nomizu, 1996, Note 11, p. 309] and also references in
[Anglès, 2007, Ch. 2].

Remark 3.9. The case of n = 2 is exceptional, as in this case every com-

plex analytic transformation of the complex plane generates a conformal
transformation on the Riemann sphere. In this case it is better to deal
with the subgroup of all conformal transformations of S 2 , called “Möbius
transformations.” These are the transformations of S n that preserve cross-
ratios
d(u, x)d(v, y)
, u, v, x, y ∈ S n ,
d(u, v)d(x, y)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 191/3

Examples 191

where d is the natural distance on S n . More information about various

equivalent definitions and properties of Möbius transformations of S n and
of B n+1 can be found, for example, in Refs. [Ratcliffe, 1994, Ch. 4] and
[Anglès, 2007, Ch. 2].

Theorem 3.10.

(i) Let for each g ∈ G, let φg : B̄ n+1 → B̄ n+1 be defined by

sg (P (x))
φg (x) = P −1 . (3.121)
Φ(sg (P (x)))
Then g → φg is a homomorphism from G onto a group of transforma-
tions of B̄ n+1 . √
(ii) If m ∈ G+ is written as in Eq. (3.114): m = (1 + kn)/ 1 − k 2 , then
the Möbius transformation φm is explicitly given by the formula:
(1 − k2 )x + 2k(1 + k (n · x))n
φm (x) = , x ∈ B̄ n+1 . (3.122)
1 + k 2 + 2k (n · x)
(iii) When restricted to the unit sphere S n , φ is a two-fold covering homo-
morphism from G onto
 the group of Möbius transformations of S n .
(iv) For m = (1 + kn)/ (1 − k ) ∈ G+ , the map φm : S  x → x ∈ S n ,
2 n

given by Eq. (3.120), is conformal with the conformal factor

(1 − k2 )
ρ= . (3.123)
(1 + k 2 + 2k(n · x))
That is, if G = (Gαβ ) is the natural Riemannian metric on the unit
sphere then
(1 − k2 )2
(φ∗m G)αβ = Gαβ . (3.124)
(1 + k 2 + 2k(n · x))2
Thus φm does not, in general, preserve the canonical, SO(V )-
invariant, volume form dS of S n . Denoting by dS  the pullback 8
φ∗m (dS) of dS by φm , for every x ∈ S n we have:
n
dS  1 − k2
(x) = . (3.125)
dS 1 + α2 + 2α(n · x)
8 Let us recall that if φ : M −→ N is a C 1 map between differentiable manifolds M and

N, and if ω is a k-form on N, then its pullback φ∗ (ω) is the k-form on M defined by

φ∗ (ω)(ξ1 p , . . . , ξk p ) = ω(dφp (ξ1 p ), . . . , dφp (ξk p )) for all ξ1 p , . . . , ξk p ∈ Tp (M ), p ∈ M
where dφp : Tp M −→ Tφ(p) N is the derivative of φ at p. For a composition of maps we
have (φ ◦ ψ)∗ = ψ ∗ ◦ φ∗ — cf. e.g. [Dieudonne, 1972, Ch. XVI.20].
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 192/3

192 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

If the map Eq. (3.120) is applied to the ball B (n+1) (rather than to its
boundary S n ), and if dV denotes the standard Euclidean volume form
of V 1 , then
n+2
dV  1 − k2
= . (3.126)
dV 1 + α + 2α(n · x)
2

Remark 3.10. It is easy to see that our definition of conformal (Möbius)

transformations of S n is equivalent to one given by Pierre Anglès in Ref.
[Anglès, 2007, 2.4.1,2.4.2.1]. In particular M0+ can be identified with
P (Q1 − {0}) in the notation of Ref. [Anglès (2007)]. But we do not need
the stereographic projection that distinguishes the vector en+1 ∈ V.

Remark 3.11. The transformations φg : B̄ n+1 → B̄ n+1 , defined in

Eq. (3.121) are also called Poincaré extensions of those restricted to S n
— cf. [Ratcliffe, 1994, Ch. 4.4, 4.5].

Proof. (i) That φg is a group homomorphism follows directly from the

defining formula. In order to show that each φg maps S n onto S n , we
first notice that from Δ(P (x)) = 1 − x2 , it follows that x ∈ S n if and
only if Δ(P (x)) = 0. If Δ(P (x)) = 0, then, since Δ(g) = Δ(g τ ) = 1, also
Δ(gP (x)g τ ) = Δ(g)2 Δ(P (x)) = Δ(P (x)) = 0, thus φg (S n )  S n . In fact,
since g −1 = g ν ∈ G, we have that φg (S n ) = S n .
(ii) Follows from Eq. (3.119).
(iii) Let us show that φ restricted to S n has kernel Z2 . We first notice that
if g ∈ kerφ then g τ ∈ kerφ. Indeed, from the very definition of φ it follows
that g ∈ ker φ if and only if g(1 + n)g τ is proportional to 1 + n for all
n ∈ Sn:

g(1 + n)g τ = λ(1 + n).

By applying π to both sides of this equation, we get

π(g)(1 − n)g ν = λ(1 − n).

Now, multiplying by g τ from the left, and by g from the right, and taking
into account the fact that Δ(g) = g ν g = Δ(g τ ) = g τ π(g) = 1, we find
g τ (1 − n)g = (1/λ)(1 − n); and, since n ∈ S n is arbitrary, g τ ∈ kerφ. Now,
assuming that g ∈ kerφ, let g = mu be the decomposition of g into a spin-
boost m ∈ G+ and a rotation u ∈ GR . Then g τ = uτ m ∈ kerφ; and, since
the kernel of a group homomorphism is a group, we get m2 =√gg τ ∈ kerφ,
i.e. φm2 (x) = x, x ∈ S n . Let us write m2 as m2 = (1 + kn)/ 1 − k 2 and,
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 193/3

Examples 193

since we have assumed that dim(V ) ≥ 2, we can choose for x a unit vector
in V, orthogonal to n. Then, from Eq. (3.120) we get

(1 − k)2 x + 2kn
x = φm2 (x) = ,
1 + k2

which is possible only for k = 0, i.e. if m2 = 1. But then, from the

uniqueness of the square root, m = 1, and so g = u. Now, u(1 + x)uτ =
(1 + x) implies uxuτ = x, which extends, by simple scaling to all x ∈ V.
Since uτ π(u) = Δ(uτ ) = Δ(u) = 1, the last equation can be rewritten as
ux = xπ(u), and it follows from [Gilbert and Murray, 1991, Lemma 5.25]
that u ∈ R. Then, since Δ(u) = 1, we get u2 = 1, so that u = ±1. The
homomorphism φ is surjective, as its image is a connected Lie group of
conformal transformations of dimension equal to that of Spin+ (1, n + 1),
that is (n + 2)(n + 1)/2 — cf. Definition 3.6 and Remark 3.9.
(iv) Let us endow V with an orthonormal basis e1 , . . . , en+1 , and the cor-
responding coordinates x1 , . . . , xn+1 . Let G = (Gij = δij ) be the natu-
ral Riemannian metric in V. From Eq. (3.120) it is then easy to compute
k
∂xl
G∗ij = (φ∗m G)ij = ∂x
∂xi ∂xj kl
δ . The result is

4k 2 (x2 − 1) 2k
G∗ij = ρ2 δij + 2
ni nj − (ni xj + nj xi ) (3.127)
f f

where

1 − k2
f = 1 + k2 + 2k(n · x), ρ= . (3.128)
f

If v = (vi ) and w = (wi ) are vectors tangent to S n , so that (v · n) =

(w · n) = 0 then, when computing G∗ij vi wj , the two last terms vanish,
and we obtain G∗ij v i wj = ρ2 Gij v i wj , which proves Eq. (3.124). Equation
(3.125) follows immediately from Eq. (3.124). It is also easy to calculate the
determinant of the matrix G∗ . It has eigenvalue equal to ρ2 on the whole
(n − 1)-dimensional subspace orthogonal to n and x, while the product of
its two eigenvalues in the subspace spanned by n and x is equal to ρ4 . So
the determinant is ρ2(n+1) , and the square root of the determinant is ρn+2 ,
which proves Eq. (3.126).9 

9 The same way one gets Eq. (3.126) also for xn. An alternative method of proving

Eq. (3.125) and Eq. (3.126), using (n + 1)-dimensional polar coordinates can be found
in a previous version of this paper, available as an arxiv preprint [Jadczyk (2007)].
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 194/3

194 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

3.3.2 Stereographic projection

In order to get a better insight into the geometrical nature of our transfor-
mations, and also to understand why in Eq. (3.115), following Ref. [Anglès
(2007)], we have used η/2, rather than just η as the parameter of the ex-
ponential, it is instructive to discuss the action of our transformations on
the stereographic projection of the sphere S n . As before, we fix the vec-
tor n ∈ S n , and let sn be the stereographic projection from S n onto the
hyperplane through the origin of V, orthogonal to n, with the origin at n.
Explicitly, we have
x − (n · x)n
sn (x) = , x ∈ Sn. (3.129)
1 − (n · x)
Indeed, the vector sn (x) is on the straight line connecting n and x, and is
orthogonal to n, which with these two properties uniquely characterize the
stereographic projection. Let us recall now the action of φm on S n . From
the formula Eq. (3.120) we have:
(1 − k 2 )x + 2k(1 + k (n · x))n
x = . (3.130)
1 + k 2 + 2k (n · x)
Let us compare now sn (x ) with sn (x). By a straightforward calculation we
obtain:
2k + (1 + k 2 )n · x
(n · x ) = , (3.131)
1 + k 2 + 2k(n · x)
(1 − k)2 (1 − (n · x))
1 − (n · x ) = , (3.132)
1 + k 2 + 2k(n · x)
(1 − k 2 )(x − (n · x)n)
x − (n · x )n = , (3.133)
1 + k2 + 2k(n · x)
and therefore
x − (n · x )n (1 − k 2 )(x − (n · x)n)
sn (x ) = =
1 − (n · x )
 (1 − k)2 + (1 − k 2 )(n · x)
(3.134)
1 − k 2 x − (n · x)n 1+k
= = sn (x).
(1 − k)2 1 − (n · x) 1−k
Now, since k = tanh(η/2), we have
1+k cosh(η/2) + sin(η/2) 2 exp(η/2)
= = = exp(η),
1−k cosh(η/2) − sinh(η/2) 2 exp(−η/2)
and therefore
sn (x ) = eη sn (x), (3.135)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 195/3

Examples 195

so that the family of Möbius transformations gn (), when parametrized by

η = 2 arctanh(k), act as a one-parameter group of uniform dilations on the
stereographic projection sn (S n ) = Rn .

3.3.3 Conformal maps and Frobenius-Perron operator

Let us first recall the definition and the role of the Frobenius-Perron oper-
ator for an iterated function system with place dependent probabilities.
Let S be a set, let {wα : α = 1, 2, . . . , N } be a family of maps wα :
S −→ S, and let pα (x), α = 1, 2, . . . , N be nonnegative functions on S
N
satisfying α=1 pα (x) = 1, ∀x ∈ S.
The maps wα and the functions pα (x) define an iterated function sys-
tem with place dependent probabilities. Starting with an initial point x0
we select one of the transformations wα with the probability distribution
pα (x0 ). If wα1 is selected, we get the next point x1 = wα1 (x0 ), and we
repeat the process again, selecting the next transformation wα2 , according
to the probability distribution pα (x1 ). By iterating the process we produce
a random sequence of integers α0 , α1 , . . . and a random sequence of points
xl = wαl (xl−1 ) ∈ S, l = 1, 2, . . . . In interesting cases the sequence xl
accumulates on an “attractor set” which has fractal properties.
Instead of looking at the points of S we can take a dual look at the
functions on S. Let F (S) be the set of all real-valued functions on S. F (S)
is a vector space, and each transformation w : S → S induces a linear
transformation w : F (S) → F(S) defined by (w f )(x) = f (w(x)), x ∈
S, f ∈ F (S).
Given an iterated function system {wα , pα ( . )} on S one naturally as-
sociates with it the Frobenius-Perron operator T ∗ : F (S) → F(S) defined
by


N 
N
∗
(T f )(x) = pα (x)(wα∗ f )(x) = pα (x)f (wα (x)). (3.136)
α=1 α=1

There is a dual Frobenius-Perron operator T∗ , acting on measures on S.

Suppose S has a measurable structure, wα and pi ( . ) are measurable, and
let F (S) be the space of all bounded measurable functions on S.
Let M(S) be the space of all finite measures on S. Then T∗ : M(S) →
. 
M(S) is defined by duality: (T∗ μ, f ) = (μ, T ∗ f ), where (μ, f ) = f dμ.
 
Since T ∗ (1) = 1, where 1(x) = 1, ∀x ∈ S, we have that dT∗ μ = dμ and,
in particular, T∗ maps probabilistic measures into probabilistic measures.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 196/3

196 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

In many interesting cases the sequence of iterates (T∗ )r μ converges,

in some appropriate topology, to a limit μ∞ = limr→∞ (T∗ )r μ, that is
independent of the initial measure μ, and which is the unique fixed point
of T∗ . The support set of μ∞ is then the attractor set mentioned above.
Let μ0 be a fixed, normalized measure on S, and assume that the maps
wα−1 map sets of measure μ zero into sets of measure μ zero. Then, for any
finite r, the measure T r μ0 is continuous with respect to μ0 and therefore
can be written as
T r μ0 (r) = fr (r) μ0 (r). (3.137)
The sequence of functions fr (r) gives a convenient graphic represen-
tation of the limit invariant measure. In our case, as it follows from the
formula Eq. (3.137)), the maps wα are bijections, and the functions fr can
be computed explicitly via the following recurrence formula:
 
N
  dμ0 wα−1 (r)  
fr+1 (r) = pα wα−1 (r) fr wα−1 (r) . (3.138)
α=1
dμ0 (r)

3.3.3.1 Towards real linear quantum formalism

What is the main difference between quantum mechanics and classical me-
chanics? While there are many possible ways of answering this question,
here we choose the one that fits our context: in quantum mechanics we re-
frain from using “nonlinear observables”. Abandoning nonlinear functions
on state-space has its advantages — it leads to a simple mathematical
formalism of linear algebra. Yet it also has its price — we are losing in-
formation. Whether such a compromise is really necessary or not, only the
future development of physics will tell.
We have defined quantum fractals in such a way that linearity is taken
into account even in the framework of real Hilbert spaces. That this is the
case can be seen from the following.
Let S be the sphere S n , and let the maps w be of the form Eq. (3.120),
determined by vectors kn ∈ B (n+1) . Then, given k, 0 < k < 1, and N unit
vectors nα ∈ S n , we have N maps
(1 − k 2 )x + 2k(1 + k(nα · x))nα
wα (x) = , (3.139)
1 + k 2 + 2k(nα · x)
as in Eq. (3.120). In such a case we have an additional structure in the set
S and in the maps wα , namely the one stemming from the Clifford algebra
realization.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 197/3

Examples 197

First of all to each x ∈ S we have associated the positive element P (x) =

(1 + x), and then we have a special class of functions on S, namely the
functions of the form:

fb (x) = (P (x), a), b ∈ C, x ∈ S. (3.140)

We denote by L the vector space of these functions. Notice that func-

tions in L separate the points x ∈ S. Indeed, for x, y ∈ S we have
fy (x) = x·y/2, thus our statement reduces to: for any two different vectors
x1 , x2 one can always find another vector y such that x1 · y = x2 · y, which
is evident.
Notice that the space L is (n + 2)-dimensional, as it is clear that
fb (x) = 0, ∀ b ∈ Cp ⊂ C, p > 1.

Proposition 3.5. With the notation as in the beginning of this section, let
0 < k < 1, nα ∈ S n , α = 1, 2, . . . N, and wα as in Eq. (3.139). Suppose
that

1)

N
nα = 0, (3.141)
α=1

2)
1 + k 2 + 2k(nα · x)
pα (x) = , (3.142)
Z(k)
where

N
Z(k) = (1 + k 2 + 2k(nα · x)) = N (1 + k 2 ),
α=1

then the Frobenius-Perron operator T ∗ of the iterated function system

{(wα , pα )} maps the space L into itself: T ∗ : fb → fV (b) , where

1 N
V (b) = P (knα ) b P (knα ). (3.143)
N (1 + k 2 ) α=1

Proof. From Eq. (3.119) it follows that if α nα = 0, then
. 
Z= 1 + k 2 + 2k(nα · x) = N (1 + k 2 )
α

is a constant, independent of x.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 198/3

198 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

From the very definition of the Frobenius-Perron operator, as well as

from Eq. (3.140), Eq. (3.118) it follows then that
 
(T ∗ fb )(x) = pα (x)fb (wα (x)) = pα (x)Φ(b P (wα (x)))
α α
 1 − k2
= pα (x)Φ b P (knα )P (x)P (knα )
α
(1 + k 2 + 2k(nα · x))
 (1 − k 2 )
= pα (x) Φ (P (knα )bP (knα )P (x))
α
1 + k2 + 2k(nα · x)
1 
= Φ (P (knα )bP (knα )P (x)) = fV (a) (x). (3.144)
Z(k) α

The Frobenius-Perron operator T acts on measures, while its dual T ∗ acts
on functions on S. Every probabilistic measure μ on S determines an algebra
element P (μ) defined by:
⎛ ⎞
  
P (μ) = P (x) dμ(x) = 1 + x dμ(x) = P ⎝ x dμ(x)⎠ , (3.145)
S S S

so that automatically Φ(P (μ)) = 1. P (μ)/2 is an idempotent if and only if

μ is concentrated at just one point on the boundary S n .
In general there are infinitely many measures μ giving rise to the same
algebra element P (μ). The process of integration on one hand leads to
simplification (linearization) but, on the other hand, it also leads to the
loss of information.

Corollary 3.2. Under the assumptions 1) and 2) of Proposition 3.5, if μ1

and μ2 are two probabilistic measures on S such that P (μ1 ) = P (μ2 ) = P,
then P (T ∗ μ1 ) = P (T ∗ μ2 ) = V (P ), where V (P ) is given by the formula
Eq. (3.143), with b replaced by P.

Proof. Because functions fb , b ∈ C separate the elements of C, it is

enough to show that fb (P (T ∗ μ)) = fb (V (P (μ))) for all b ∈ C. Now, from the
very definition of the functions fb , fb (x) = Φ(bP (x)), and from the linearity
. 
of the trace functional Φ, it follows that (fb , μ) = fb (x)dμ(x) = Φ(bP (μ)),
and so fb (V (P (μ)) = Φ(bV (P (μ))) = Φ(V (b)P (μ)) = fV (b) (P (μ)) =
fb (P (T ∗ μ)). 

Corollary 3.3. Under the assumptions 1) and 2) of Proposition 3.5,

the Frobenius-Perron operator recurrence formula Eq. (3.138) is explicitly
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 199/3

Examples 199

given by
 
(1 − k2 )n+2  fr wα−1 (x)
N
fr+1 (x) = , (3.146)
N (1 + k2 ) α=1 (1 + k 2 − 2k(nα · x))n+1
where
(1 − k2 )x − 2k(1 − α(nα · x))nα
wα−1 (x) = . (3.147)
1 + k2 − 2k(nα · x)

Proof. The proof follows from the definitions, using the following, easily
verifiable, algebraic property: if
g(x) = 1 + k2 + 2k(nα · x),
and if
(1 − k 2 )x − 2k(1 − k(nα · x))nα
wα−1 (x) = , n2α = 1,
1 + k 2 − 2k(nα · x)
then
(1 − k 2 )2
g(wα−1 (x)) = .
1 + k 2 − 2k(nα · x) 

3.4 Algorithms for generating hyperbolic quantum fractals

There are two methods of generating the data set for graphical representa-
tion of hyperbolic quantum fractals: one is through the Chaos Game, the
other one is through the Frobenius-Perron operator. We give both of them
here, adapted to a general case of n-sphere. For the standard two-sphere
n = 2. In both cases it is possible to render the data points either on the
sphere or on its stereographic projection. In the algorithms below x · y
stands for the standard Euclidean scalar product of two vectors, while ||x||
stands for the standard Euclidean norm of a vector.
In addition to the pseudocodes, Fig. 3.46 shows the output of the For-
tran code listed in Sec. 3.4.1.1, while Fig. 3.47 shows the output of the
Fortran code listed in Sec. 3.4.2.1. Both codes should run under Linux
gfortran. For the graphics part we used DISLIN package available from
http://www.dislin.de.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 200/3

200 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 3.46 Quantum octahedron generated by the Fortran code listed in Sec. 3.4.1.1.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 201/3

Examples 201

Fig. 3.47 Fourth level iteration of the Frobenius-Perron operator for Quantum Do-
decahedron (with stereographic projection) generated by the Fortran code listed in
Sec. 3.4.2.1.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 202/3

202 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

3.4.1 Chaos game on n-sphere

Algorithm 3.1. Hyperbolic QIFS through the Chaos Game

Require: N vectors q[j], (j = 1, 2, ..., N ) in Rn+1 with 0 < ||q[j]|| < 1,


that add to zero: j q[j] = 0.
Require: integer niter — number of iterations
Require: (random) starting unit vector x in Rn+1
1: s ← 0  Precompute array k2 and constant s
2: for j ← 1, N do
3: k2[j] ← ||q[j]||2
4: s ← s + 1 + k2[j]
5: end for
6: function prob(j,x)  place dependent probabilities return
(1 + k2[j] + 2 ∗ (q[j] · x))/s
7: end function
8: function jump(j,x)
9: y ← (1 − k2[j]) ∗ x + 2 ∗ (1 + (q[j] · x)) ∗ q[j]
10: return y/||y||
11: end function
12: for i ← 1, niter do  Main loop start
13: r ← random real in (0, 1)
14: p←0
15: for j ← 1, N do
16: p[j] ← prob(j, x)
17: p ← p + p[j]
18: if p > r then
19: exit
20: end if
21: end for
22: x ← jump(j, x)
23: write x to a file, or increase pixel density corresponding to x
24: end for  Main loop end
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 203/3

Examples 203

3.4.1.1 Fortran code

PROGRAM sphere
USE DISLIN
IMPLICIT none
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! This part contains parameters that need to be set
! (or commented and uncommented) before the run.
! It can be all done in single precision, but with
! fractals, where we have possibly millions of
! iterations, double precision seems to be more
! appropriate.
!!!!!!!!!!!!!!!!
REAL (KIND = 8), PARAMETER :: k = 0.57D0
! The weight.
! Here the same for all vertices because of the
! symmetry of the regular polyhedra
INTEGER ( KIND = 4 ), PARAMETER :: nvert = 6
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
INTEGER ( KIND = 8 ), PARAMETER :: NIT = 10**9
! 1000 millions
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! If we want to plot the image
! (vertical view of the hemisphere), we set the
! resolution. We do not need it if we want just
! to collect the points
! RESOLUTION:
INTEGER ( KIND = 4 ), PARAMETER :: RES = 1000
! Declarations of variables and setting the
! length of the arrays
REAL*8 K2(NVERT)
REAL*8 V(3,NVERT)
REAL*8 Q(3,NVERT) ! Vertices multiplied by k
REAL*8 X(3) ! Current point
REAL*4 S
! If we want to store all generated points on the
! sphere, we need these arrays for x,y,z
! coordinates. Of course if we want to have 100
! millions of them, we may have memory problems
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 204/3

204 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

! REAL*4 XS(NIT,3)
! On the other hand we can choose a given
! resolution RES and divide the unit square on the
! equatorial plane into a grid of RESxRES small
! squares. Then to each square we add a counter.
! Whenever XS,YS,ZS vertically projected falls
! into a given square we add 1 to the counter of
! this square. Then we process XS,YS,ZS to
! get new ones. So we do not have to store all the
! points on the sphere, we store only the counters.
! One counter for the density of hits, another
! counter for logarithms of these hits. Logarithmic
! counters are often used in the fractal community.
! They allow us to see more details.
INTEGER DENSITY(RES,RES)
REAL*4 DENSITYL(RES,RES)
! Dummy variables used in calculations
INTEGER*4 M,N
REAL*4 MAXL
! Then we need random numbers:
REAL*4 RR ! For random number between 0 and 1
INTEGER R ! Integer random in 1,...,NVERT
INTEGER*8 I,J !(to count iterations etc.)
REAL*4 P !To be used for place dependent
! random choice of an active vertex
V = RESHAPE((/0.D0, 0.D0, 1.0D0,&
1.D0, 0.D0, 0.D0,&
0.D0, 1.D0, 0.D0, &
-1.D0, 0.D0, 0.D0,&
0.D0, -1.D0, 0.D0,&
0.D0, 0.D0, -1.0D0/),(/3,NVERT/))
Q = k*V
s = 0
DO M=1,NVERT
k2(M) = sc(Q(:,M),Q(:,M))
s = s + 1.0D0 + k2(M)
ENDDO
! We choose the starting point on the sphere at
! random. It really does not matter which one we
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 205/3

Examples 205

! choose. Can be the North Pole as well:

RR = RAND(0)
!CALL RANDOM(RR)
! Initialize random x-component.
! The name of the function RANDOM[]
! may depend on the compiler
! It should give a real number in (0,1)
X(1) = 2*RR - 1 ! Move from (0,1) to (-1,1)
RR = RAND(0) ! Initialize random y-component
X(2) = 2 * RR - 1 ! Move from (0,1) to (-1,1)
RR = RAND(0) ! Initialize random z-component
X(3) = 2 * RR - 1 ! Move from (0,1) to (-1,1)
X=X/N2(X) ! Normalize
! Here the main loop of the program starts.
! We are using four functions: sc (scalar product),
! N2 (norm), PROB (place dependent probabilities),
! JUMP (jumps on the sphere).
! They are defined at the end of the program
DO I=1,NIT
IF (modulo(I,CEILING(NIT/100.0)) == 0) THEN
! To inform about the progress
WRITE(*,’(I10)’) I
END IF
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! The following two lines are for evenly
! distributed probabilities
!RR = RAND(0)
!R = CEILING(REAL(NVERT)*RR)
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! For place dependent probabilities use the
! following 9 lines:
RR = RAND(0)
p = 0
DO M=1,NVERT
p = p + PROB(M,X)
IF (RR < p) THEN
R = M
EXIT
END IF
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 206/3

206 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

END DO
X = JUMP(R,X)
! The line below should be commented out
! if we do not need to store the points
! on the sphere.
! XS(I,:) = X
! Here we classify the point according to the
! little square into which it falls. Since
! this is vertical projection, only x,z
! coordinates count.
IF (X(3)>0D0) THEN
! IF (X(3)<0D0) THEN !For the southern hemisphere
M = MAX(CEILING((REAL(X(1))+1.0)&
&*REAL(RES)/2.0),1)
N = MAX(CEILING((REAL(X(2))+1.0)*&
&REAL(RES)/2.0),1)
DENSITY(M,N)=DENSITY(M,N)+1
END IF
END DO
! Calculate log of the density (+1) and its
! maximum value:
MAXL = 0
DO I=1,RES
DO J=1,RES
DENSITYL(I,J)=LOG10(REAL(DENSITY(I,J)+1.0))
IF (DENSITYL(I,J)>MAXL) THEN
MAXL = DENSITYL(I,J)
END IF
END DO
END DO
! Normalize to values in [0,1]:
DENSITYL(1:RES,1:RES) = DENSITYL(1:RES,1:RES)/MAXL
! Graphics part:
CALL FILMOD ("VERSION")
CALL winsiz(1000,1000)
CALL PAGE(1440,1440)
CALL sclmod("full")
CALL METAFL(’CONS’)
!CALL SCRMOD (’REVERSE’)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 207/3

Examples 207

CALL DISINI()
CALL SETGRF ("NONE", "NONE", "NONE", "NONE")
CALL SETVLT (’SPEC’)
!CALL SETVLT (’GREYR’)
CALL AUTRES(RES,RES)
CALL AXSPOS(25,1410)
CALL AX3LEN(1390,1390,1000)
CALL NOBAR
CALL GRAF3(-1.,1.,-1.,0.5,-1.,1.,-1.,0.5,0.,1.0,0.,&
&1.0)
CALL CRVMAT(DENSITYL,RES,RES,1,2)
CALL RBMP(’image.bmp’)
CALL NOLINE(’X’)
CALL NOLINE(’Y’)
CALL DISFIN()
STOP

CONTAINS
! Scalar product and norm in double precision
! In single precision we can also use standard
! Fortran 90 functions. Or we can use some other
! library.
REAL*8 FUNCTION SC(V1,V2) ! Scalar product
REAL*8 V1(3),V2(3)
SC = V1(1)*V2(1) + V1(2)*V2(2) + V1(3)*V2(3)
END FUNCTION SC
! Norm in double precision
REAL*8 FUNCTION N2(W)
REAL*8 W(3)
N2 = DSQRT( W(1)*W(1) + W(2)*W(2) + W(3)*W(3) )
END FUNCTION N2
! Place dependent probabilities
REAL*4 FUNCTION PROB(J,X)
REAL*8 X(3)
INTEGER*4 J
PROB = (1.0 + 2.0*REAL( sc(X,Q(:,J)) ))/s
END FUNCTION PROB
! Jump to the next point on the sphere
FUNCTION JUMP(M,X)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 208/3

208 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

REAL*8 X(3),Y(3)
REAL*8, DIMENSION(3) :: JUMP
INTEGER*4 M
Y = (1.0D0 - k2(M))*X + &
2.0D0*(1.0D0 + sc(Q(:,M),X))*Q(:,M)
JUMP= Y(1:3)/N2(Y)
END FUNCTION JUMP
END PROGRAM sphere

3.4.2 Approximation to the invariant measure

Algorithm 3.2. Frobenius-Perron operator density

Require: N unit vectors n[j], (j = 1, 2, ..., N ) in Rn+1 that add to zero
Require: real number k between 0 and 1
Require: n-dimensional grid of ngrid points x[i], (i = 1, ..., ngrid) on the
sphere
Require: integer level the recursion level
for j ← 1, N do
q[j] ← k ∗ x[j]
end for
f ac ← (1 − k 2 )n+2 /(N ∗ (1 + k2 ))
function f(l,x)
if l == 0 then
result ← 1
else
result ← 0
for j ← 1, N do
dot ← (x · q[j])
y ← (1 − k2 ) ∗ x − 2 ∗ (1 − dot) ∗ q[j]
x ← y/||y||
result ← result + F (l − 1, x)/(1 + k 2 − 2 ∗ dot)n+1
end for
end if
return f ac ∗ result
end function
for i ← 1, ngrid do
plot F (level, x[i]) at the point corresponding to x[i]
end for
June 16, 2014 8:52 World Scientific Book - 9in x 6in QuantumFractals3 pg. 209/3

Examples 209

3.4.2.1 Frobenius-Perron Fortran code

PROGRAM DODECAFP
USE DISLIN
IMPLICIT none
REAL (KIND = 8), PARAMETER :: k = 0.6
INTEGER ( KIND = 4 ), PARAMETER :: nn = 20
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
INTEGER ( KIND = 4 ), PARAMETER :: level = 2 ! Recursion level
! Size of the square on the complex plane\index{complex!plane}
REAL (KIND = 8), PARAMETER :: xymin = -5.0D0
REAL (KIND = 8), PARAMETER :: xymax = 5.0D0
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
REAL*8 Q(3,nn) ! Vertices multiplied by k
INTEGER ( KIND = 4 ), PARAMETER :: RES = 1024 ! Resolution
REAL*8 k1,k3,k4,fac,delta,delta2,xysize,zi,zr,z2
REAL*8 V0(3,nn)
REAL*8 r(3)
REAL*8 pic(res,res)
REAL*4 maxm
REAL*4 picl(res,res)
real clock_start, clock_finish
INTEGER ip,iq

call cpu_time (clock_start)

xysize = xymax-xymin
delta = xysize/real(res)
delta2 = delta/2.0D0
!!!!!!!!!!!
V0 = RESHAPE((/0D0, 0D0, 1.D0, &
0.D0, -0.666667D0, 0.745356D0, &
0.57735D0, 0.333333D0, 0.745356D0, &
-0.57735D0, 0.333333D0, 0.745356D0, &
0.57735D0, -0.745356D0, 0.333333D0, &
-0.57735D0, -0.745356D0, 0.333333D0, &
0.356822D0, 0.872678D0, 0.333333D0, &
0.934172D0, -0.127322D0, 0.333333D0, &
-0.934172D0, -0.127322D0, 0.333333D0, &
-0.356822D0, 0.872678D0, 0.333333D0, &
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 210/3

210 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

0.356822D0, -0.872678D0, -0.333333D0, &

-0.356822D0, -0.872678D0, -0.333333D0, &
0.57735D0, 0.745356D0, -0.333333D0, &
0.934172D0, 0.127322D0, -0.333333D0, &
-0.934172D0, 0.127322D0, -0.333333D0, &
-0.57735D0, 0.745356D0, -0.333333D0, &
0.57735D0, -0.333333D0, -0.745356D0, &
-0.57735D0, -0.333333D0, -0.745356D0, &
0.D0, 0.666667D0, -0.745356D0, &
0D0, 0D0, -1.D0/),(/3,nn/))
!!!!!!!!!!!!!!!!!!!!!!
! Calculating constants
k1 = 1.0D0 - k*k
k3 = 1.0D0 + k*k
k4 = 1.0D0/( nn*k3 )
fac = k1**4/(nn*k3) !nonuniform
!fac = k1**2/nn !uniform
Q = k*V0 ! Vertices multiplied by k
do ip = 1,res
if (modulo(ip,100)==0) then
write(*,*) ip
end if
zr = xymin + ( ip*delta - delta2)
do iq = 1,res
zi = xymin + ( iq*delta - delta2)
z2 = zi*zi + zr*zr
r(1) = 2*zr/(1.0D0 + z2)
r(2) = 2*zi/(1.0D0 + z2)
r(3) = (-1.0D0 + z2)/(1.0D0 + z2)
pic(ip,iq) = fp(level,r)
end do
end do
maxm = 0.0
do ip = 1,res
do iq = 1,res
picl(ip,iq) = log10(real(pic(ip,iq))+1.0)
if ( picl(ip,iq)>maxm ) then
maxm = picl(ip,iq)
end if
June 16, 2014 8:52 World Scientific Book - 9in x 6in QuantumFractals3 pg. 211/3

Examples 211

end do
end do
call cpu_time(clock_finish)
write (*,’(A)’) "--------------------------------------"
write (*,’(A)’) " "
write (*,*) "CPU Time = ",(clock_finish - clock_start), &
" seconds"
picl = picl/maxm
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
CALL METAFL(’CONS’)
CALL PAGE(1300,1300)
CALL winsiz(1040,1150)
CALL sclmod("full")
CALL DISINI()
CALL SETVLT (’GREY’)
CALL HWFONT()
CALL INTAX()
CALL AUTRES(RES,RES)
CALL AXSPOS(150,1160)
CALL AX3LEN(1100,1100,1100)
CALL NOBAR
CALL LABDIG(1,’XYZ’)
CALL LABELS(’FLOAT’, ’XYZ’)
CALL GRAF3(real(xymin),real(xymax),real(xymin),&
real(xysize)/10.0,real(xymin),&
real(xymax),real(xymin),real(xysize)/10.0,0.,1.0,0.,1.0)
CALL CRVMAT(picl,RES,RES,1,2)
CALL RBMP(’dodecahedron_fp_06_3.bmp’)
CALL DISFIN()
STOP

CONTAINS

recursive function fp(n,v) result(b)

implicit none
integer i,j,n
real*8 v(3),w(3),vq(3),b,den,den3,dot
if (n==0) then
b = 1.0D0
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 212/3

212 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

else
b = 0.D0
do i=1,nn
do j=1,3
vq(j) = q(j,i);
end do
dot = vq(1)*v(1)+vq(2)*v(2)+vq(3)*v(3)
den = 1.0D0/(k3-2.0D0*dot)
den3 = den**3 ! nonuniform
! den3 = den**2 !uniform
w = den * (k1 * v - 2.0D0 * (1.0D0 - dot) * vq)
b = b + fp(n-1,w)*den3
end do
b = fac * b
end if
end function fp
END PROGRAM DODECAFP
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 213/4

Chapter 4

Foundational Questions

4.1 Stochastic nature of quantum measurement processes

The following are the main features of quantum theory:

• Unpredictability of the observed individual events (such as time in the

decay of radioactive elements, a position of a dot on the screen in the
two-slit experiment)
• Predictability of statistical averages in a long series of repeated experi-
ments
• Quantization of measured values of certain quantities characterizing the
system (projection of the spin of a particle in a specified direction, en-
ergies of photons emitted by an excited atom)
• Some kind of non-locality — apparent “action at a distance” (as in the
celebrated Einstein-Podolsky-Rosen experiment [Einstein et al. (1935)])

The most prominent feature, as of today, is the first one: unpredictability.

Sometimes, when we are looking for “true random” number generators, we
tend to consider the “quantum noise” as the only thing that deserves this
name. For instance, the exact time of radioactive decay of an unstable
atom seems to be completely unpredictable. We can say, with statistical
certainty, what will be the expected time for the decay of half of the radioac-
tive material, but single decay events seem to be completely random. Yet
questions arise: are they “truly random?”, and what does “truly random”
mean?
The belief that quantum phenomena are “truly random” seems to be
widespread. The “true randomness” of quantum event generating processes
is related to the “security” of quantum key distribution” (QKD). So, for
instance, in a letter to Nature entitled “Random Numbers Certified by Bell’s

213
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 214/4

214 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Theorem” [Pironio et al. (2010)], eleven distinguished international experts

in cryptography state:

“... we show that the non-local correlations of entangled quantum

particles can be used to certify the presence of genuine randomness.
... This strong form of randomness generation is impossible clas-
sically and possible in quantum systems only if certified by a Bell
inequality violation ...”

Other experts [Hänggi et al. (2010)] are somewhat more cautious:

“... The practical relevance is that the resulting security is device-

independent: We could even use devices manufactured by the ad-
versary to do key agreement. The theoretical relevance is that the
resulting protocol is secure if either relativity or quantum theory is
correct. This is in the spirit of modern cryptography’s quest to min-
imize assumptions on which security rests. ...”

In another paper [Scarani and Kurtsiefer (2009)] we can find a bit more of
an explanation:

II. QUANTUM SIGNALS AS INCORRUPTIBLE COURIERS

Even before unconditional security was technically proved,
“security based on the laws of physics” became the selling slogan of
QKD. It’s catchy, and it can be understood correctly — but it may
also be understood wrongly and has often been explicitly spelled out
as “security based only on the laws of physics”. Of course, a pause
of reflection shows that the statement cannot possibly be as strong
as that. For instance, the laws of physics do not prevent someone
from reading the outcomes of a detector; however, if the adversary
has access to that information, security is clearly compromised! But
many people were just carried away by the power of the slogan —
fair enough, this does not happen only with QKD.
On the wings of enthusiasm, some promoters of QKD also man-
aged to convey the impression that they were presenting the solution
for (almost) every task of secret communication. This may have im-
pressed some sponsors. However, the main result was to alienate
a great part of the community of experts in classical cryptography,
who, unfamiliar with quantum physics though they may be, could
not fail to spot the overstatement. ...
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 215/4

Foundational Questions 215

But we know that every theoretical scheme in physics has a limited do-
main of application. Beyond these domain the underlying assumptions,
and therefore also the model, are no longer valid. Why would quantum
theory be different?
Nevertheless, writing about the importance of problems of randomness
in physics three British educators, Jon Ogborn, Simon Collins and Mick
Brown wrote [Ogborn et al. (2003b)] (see also continuation in [Ogborn
et al. (2003a)]):

... Finally, random behaviour matters because it is at the root

of quantum behaviour, which is to say, at the root of our deepest
understanding of how things are. Quantum calculations predict, not
what will happen, but the probability of events. This is the reason,
for example, why radioactive decays come at random. It is also the
reason why photons arrive at random. Yet in physics teaching, these
principles are usually at best merely stated, not demonstrated. So a
last and fundamental reason why understanding randomness matters
is to be able to demonstrate its presence in real phenomena such as
radioactive decay.
... data give some experimental reason to believe that the counts
from a radioactive source — a true quantum phenomenon — do really
arrive randomly.

In another paper devoted to the education of physicists [Ogborn (2011)]

the distinguished British educator Jon Ogborn writes:

“Perhaps the most important first experience of quantum be-

havior is to listen to a Geiger counter detecting gamma photons:
click....click. click........click. The key point is that the gamma pho-
tons arrive at random.”

So, what is it? Do we know for sure? Do we just “believe”? The internet
site “random.org” advertises as follows

RANDOM.ORG is a true random number service that generates

randomness via atmospheric noise.

In an article “What’s this fuss about true randomness?” the authors

write [Random.org (2013)]:

Quantum Events or Chaotic Systems?

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 216/4

216 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

... One characteristic that builders of TRNGs [True Random

Number Generators] sometimes discuss is whether the physical phe-
nomenon used is a quantum phenomenon or a phenomenon with
chaotic behaviour. There is some disagreement about whether quan-
tum phenomena are better or not, and oddly enough it all comes
down to our beliefs about how the universe works. ...
We have two different problems here. The first one is: What exactly is
“true randomness”? Do we have a precise definition of this term? The
second one: “Assuming that we know what true randomness is, do we have
a sufficient reason to rely on the assumption that quantum phenomena are
truly random?”.
Let us consider the first question: what does it mean that something,
some pattern or some series of physical events, is truly random? One can
look at it from several different perspectives. One of these perspectives
goes to the very foundations of mathematics: the concept of infinity and to
Gödel’s incompleteness theorem. The interested reader may like to read the
excellent monograph by Anastasios A. Tsonis “Randomnicity: Rules and
Randomness in the Realm of the Infinite” [Tsonis (2008)]. Another useful
reference is “Randomness and Complexity” by Christian S. Calude [Calude
(2007)]. The general conclusion of all this work is that: “randomness” is
a tricky subject. We can relate it to “complexity” — roughly the length
of the shortest program on a given PC that produces the given sequence of
bits. But complexity is always defined in relative terms, for instance to a
given Turing machine. If we change our machine, what was considered to
be complex on the first one, can prove to be simple on another. Therefore
we need to define the complexity of the machine itself. For this we will
need a meta-machine. Such a sequence will never end, unless we arbitrarily
fix some standards.
The second perspective can be formulated as follows: Is the calculus of
probability adequate for modeling phenomena that are supposedly random?
In quantum theory, even if it does not start with an explicit probabilistic
formulation — which comes only later, through interpretative postulates
— at some point we start using probabilistic arguments. That is why
quantum theory is considered to be intrinsically probabilistic. But does
it tell us something about the real nature of the quantum world? In his
monograph “Decision Systems and Nonstochastic Randomness” [Ivanenko
(2010)] V. I. Ivanenko makes these important remarks:
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 217/4

Foundational Questions 217

In probability theory there is a notion of an elementary (indivis-

ible) event, in contrast to an event that consists of a set of several
elementary events. In real life, we are never able to state with abso-
lute assurance that an event is elementary. So, if the introduction of
food is an event for the researcher, for the dog it may be the intro-
duction of food of a certain quality; if for one statistician the event
is that all plates are broken by one sister, for another it may be that
they are broken by the youngest one; and so on.
At first glance, the situation seems to be similar to scenarios from
other fields of human endeavor: we never know anything for certain,
and so we use approximate knowledge. Nevertheless, here there is a
principal difference. Imagine that we have finally found the true first
principles and thus have succeeded in describing, following Laplace,
all the coordinates and velocities of all the particles of the universe.
We may try to consider these points of the phase space as elementary
events and thus describe all our experience in these terms.
Immediately, we see that the value of our experience will be zero,
since none of these elementary events will be ever repeated. This
means that even if we were able to describe and memorize our ex-
perience exactly, we would be able to use it only after a certain
coarse-graining to some other events (or notions).

Then he concludes as follows:

(1) In any statistical research, or even more generally in any conscious

activity, in any reasoning, we deal with events none of which is elemen-
tary.
(2) Events that we consider predetermine the conclusions that we obtain.
(3) The reasons leading to the consideration of this or that event are linked,
in each particular case, with what is called common sense. However,
common sense differs from person to person. Moreover, even when we
consider a professional activity of an outstanding specialist in his field
(William Feller), even in this case common sense does not guarantee
the uniqueness of conclusions.

Randomness is always relative. It is relative to the model, relative to

the state of knowledge (or information), relative to the assumptions that
have been made, either explicitly or implicitly. Quantum physics is not an
exception here. At the present moment we may consider radioactive decay
events as elementary events taking zero time, in some other model we may
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 218/4

218 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

deal with these events as processes of infinite complexity, without any sharp
beginning or end. These comments concern the theory. But what about
experiments? What can we learn from experiments?
As we have mentioned before, the process of radioactive decay is one of
the most typical examples of physical processes that are considered to be
of “truly random” character. Mathematical modeling of this process starts
with counting the data which are individual “events”. An extract from a
typical history of such counts may look as in Fig. 4.1.

Counting events
9

5
Events

1
t1 t 2 t3 t4 t5 t6 t7 t8 t9
0
0 1 2 3 4 5 6 7 8
Time

Fig. 4.1 Counting typical decay events.

In theory it is usually assumed (what else could be simpler?) that all

nuclei in a sample of a radioactive material are independent of each other.
Thus events occurring in nonoverlapping intervals are independent as well.
It is assumed that the process is stationary, that is that the distribution
of the number of events in any time interval depends only on the length of
the interval (as long as we ignore the fact that our sample is finite). It is
also assumed that the probability of two events occurring simultaneously
is negligible, and that the probability of exactly one event to occur in a
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 219/4

Foundational Questions 219

sufficiently small interval dt is, in the first approximation, proportional to

dt:
P [N (dt) = 1] = λdt + o(dt). (4.1)
All these assumptions lead to the Poisson distribution, which can be for-
mulated as follows (see Fig. 4.2): It is seen that for large λ the Poisson
distribution has the shape approaching that of a Gaussian distribution.

e−λ λn
Poisson distribution Pλ [n] = n!

λ = 1.0 λ = 5.0
Probability

n n

λ = 10.0 λ = 15.0
Probability

n n

Fig. 4.2 Poisson distribution for λ = 1, 5, 10, 15. Probability of n counts in a unit time
interval.

Denoting by X(t) the random variable describing the number of events

in a positive time interval of length t we have:
(λt)n
P [X = n] = e−λt , n = 0, 1, 2, ... (4.2)
n!
On the other hand the probability density fλ,n of registering n counts in
a time interval (t, t + dt) is given by the so called Erlang (also known as
Gamma) distribution:
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 220/4

220 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

e−λt (λt)n−1
fλ,n (t) = λ . (4.3)
(n − 1)!
Plots of the Erlang distribution for λ = 1, 5, 10, 15 are shown in Fig. 4.3.

−λt (λt)n−1
Erlang (Gamma) distribution fλ,n [t] = λ e (n−1)!

n=1 λ = 1.0

n=2
n=3
n=4
n=5
n=6

Fig. 4.3 Probability density of n counts for λ = 2.

From the above we easily obtain that the expected number of events in
the time interval t is given by:
∞
 (λt)n
ne−λt = λt. (4.4)
n=0
n!

We also have:
∞
 (λt)n
n2 e−λt = λt + λ2 t2 . (4.5)
n=0
n!

For the variance of the Poisson distribution we thus obtain:

E[(X(t) − λt)2 ] = E[X 2 (t)] − (λt)2 = λt. (4.6)

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 221/4

Foundational Questions 221

Therefore the standard deviation σ is given by

√
σ = λt. (4.7)

A series of experiments with 2000 counts may result in a graph like the one
shown in Fig. 4.4 that was produced by a computer simulation for λ = 0.5.
Five series of experiments with 2000 counts (λ = 0.5) show fluctuations
as in Fig. 4.5.
For a Poisson counting process with the rate λ we expect λΔt counts
in each interval of length Δt. But a real experiment will show fluctuations
around this expectation value. Computer simulation of 1 mln counts for λ =
0.5, that are binned in successive intervals of length Δt = 1000 produces
the graph, shown in Fig. 4.6, of fluctuations around the expected value that
has been normalized to 1.
But in the real world of experiments the situation is not that simple. For
a long time it was taken for granted that nuclear decays, governed somehow
by quantum processes, are truly “spontaneous” and “uncontrollable”. Re-
cently this belief has been seriously questioned. So, for instance, in recent

Observation of 2000 successive Poisson events

1000

800
Events observed

600

400

200

0
0 500 1000 1500 2000
Time

Fig. 4.4 Poisson process, λ = 0.5. Number of counts versus time. 2000 counts.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 222/4

222 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Observation of 2000 successive Poisson events

2000

1500
Events observed

1000

500

0
0 500 1000 1500 2000 2500 3000 3500 4000 4500
Time

Fig. 4.5 Poisson process, λ = 0.5. Number of counts versus time. Five runs, each of
2000 counts.

Fluctuations
1.2
Variation of the count rate

1.15
1.1
1.05
1.0
0.95
0.9
0.85
0.8
0 500 1000 1500 2000
Time

Fig. 4.6 Poisson process. Fluctuations of the normalized decay rate for λ = 0.5. Com-
puter simulation with 1 mln events.

years it was discovered that radioactive decay rates show periodicities and
variations of the cosmic origin.
Probably the first scientific paper documenting unexplained annual
changes in the rates of radioactive β-decay processes was published in 2001
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 223/4

Foundational Questions 223

by E. D. Falkenberg in the journal “Apeiron” under the title: “Radioac-

tive Decay Caused by Neutrinos?” [Falkenberg (2001)].1 A year later the
methodology used by Falkenberg for evaluating the data was strongly crit-
icised by G. W. Bruhn [Bruhn (2002)] (see also Falkenberg’s reply [Falken-
berg (2002)]). The reality of the effect was not clear.
In July 2010 the American team of researchers from Purdue and Stan-
ford Universities and also Edwards AFB analyzed the nuclear decay data
collected over the years from Brookhaven National Laboratory, and came
to the conclusion that [Fischbach et al. (2010)]:

“This result appears to confirm our proposal that the Rieger

periodicity is due to an r-mode oscillation, and to indicate that
such an oscillation occurs in the solar core, influencing the solar
neutrino flux and thereby influencing certain nuclear decay-rates.”

Two month earlier, in May 2010, Russian physicist A. G. Parkhomov from

The Institute of Time Nature Explorations (part of Lomonosov’s Moscow
State University) published an internal report entitled “Periodic Changes
in Beta Decay Rates” [Parkhomov (2010)]. He concluded that:

“So, the results of the analysis of the Cassini spacecraft power

output seem unlikely to refute the idea about a possible correlation
of β radioactivity decay rates with the distance between the Sun
and Earth, because this effect is not present in α decays. Other
matter, that idea about connection of changes of radioactivity of
1-year period with Sun neutrino fluence oscillations [...] looks ex-
tremely doubtful because of exclusive weakness of interaction such
neutrino with substance. On the other hand, the presence of this
effect for β decays and its lack for α radioactivity hints at a pos-
sible involvement of neutrino in this phenomenon (neutrino are an
essential ingredient in β processes, but do not take part in α de-
cays). It is conceivable that these periodic variations are related
to streams of “relic” neutrino [...]. A hypothesis about a probable
role of relic neutrinos offered also to explain the vague effects which
have been observed at a measurement of neutrino mass [...]. The
substantiation of these ideas requires special consideration.”

The American team was using rather sophisticated statistical tools for an-
alyzing the data for periodicities. The two teams, American and Russian
1 Thanks are due to A. G. Parkhomov for bringing this reference to my attention.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 224/4

224 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

joined together and reanalyzed the decay data collected by the Lomonosov
State University. They produced a joint paper [Sturrock et al. (2013)] in
which they concluded:

To sum up: It is clear that different nuclides behave differently.

We have found no evidence to date that any alpha decays are vari-
able. It appears that some — but not all — beta decay rates are
variable. The mechanism is presently unknown.

Parkhomov, in his paper [Parkhomov (2010)] shows the graph of periodic-

ities in decay rates — see Fig. 4.7. There are two kinds of “irregularities”
that are being observed in “true randomness” of quantum generated nuclear
decay events. The first one is just the change of the rate of the process. To
explain this kind of deviation from simple probabilistic laws we do not have
to revise our notions about randomness of quantum events. It is enough
to take into account celestial mechanics, cosmic influences, and our theo-
ries concerned with regulators of nuclear processes (neutrinos etc.). But,
in his 2006 paper about “Three types of variations of the rate of different

Fig. 4.7 Count rates for 60 Co and 90 Sr–90 Y β sources measured by GM counters,
adjusted for a decrease in source activity with half-lives of 5.27 and 28.6 years, and count
rate for 239 P u α source measured by the silicon detector. Of course the fluctuations of
counts in the real data graphs are much smaller than those shown in Fig. 4.6. This is
due to the fact that the counters analyzed by Parkhomov counted 10–40 mln events per
day!
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 225/4

Foundational Questions 225

processes” [Parkhomov (2006)] Parkhomov notes that the very level of ran-
domness of certain random processes can also change. In the conclusions
he writes (translated from Russian):

We are used to thinking that the only way to influence the

course of the process is through changing its rate or intensity. That
is what is being exploited in contemporary technologies. (...) But
there is also another kind of variability of the course of processes,
that shows up as a change of organization of the elements of a
system. Such a change can arise independently of the changes in the
energy profiles. Perhaps we are dealing here with some unknown
properties of information, and also with consciousness. Perhaps
research in this direction will help us in overcoming the crisis of
contemporary natural science and will open the way towards a new
understanding of the Universe in which we live.

As an example of the type of variability that A. G. Parkhomov has in mind

here we can look at Fig. 4.8 — adapted from his 2009 paper “Regulated
chaos” [Parkhomov (2009)]. We find there references to other experiments
along this line. Unfortunately we do not have yet a developed methodol-
ogy for dealing with these kind of phenomena. Yet they may indicate the
need to look at other models, that go beyond the standard quantum theory

Regulated Chaos
13.8
13.6
13.4
13.2
imp/s

13.0
12.8
12.6
12.4
12.2
14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0 18.5 19.0
Time

Fig. 4.8 An example of the variation of the fluctuations width of the data registered
by a Geiger counter. The source is a radioactive sample of 60 Co with the mean data
count of 13.5 impulses per second, with standard deviation of 0.3 imp/s. During the
time interval of about 15 minutes an external “device” has been turned on. During this
time the rate of the counts did not change, but the fluctuations around the mean value
became significantly smaller.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 226/4

226 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

and allow us to study the possible means of creating “windows of order” in

otherwise chaotic and/or random phenomena. For example, Andrei Khren-
nikov in his paper “Quantum-like Probabilistic Models Outside Physics”
[Khrennikov (2008)] writes:
This approach gives the possibility to apply the mathematical
quantum formalism to probabilities induced in any domain of sci-
ence. In our model quantum randomness appears not as irreducible
randomness (as it is commonly accepted in conventional quantum
mechanics, e.g. by von Neumann and Dirac), but as a consequence
of obtaining incomplete information about a system.
The very concept of “quantum probability” may also need a fresh look such
as, for instance, in [Etter and Noyes (1998)], where the authors suggest
that we may need to “extend probability theory by allowing cases to count
negatively ...”. Whether such a radical step is to be taken or not is yet
to be seen. What can be noticed, however, is the fact that more often
than not “quantum events” have something to do with time, and the role
played by time in quantum theory is not entirely clear. There is no “time
operator” in the standard quantum theory, and there are several different
candidates for “time of arrival”. For example EEQT provides one such
candidate [Blanchard and Jadczyk (1966, 1997)]. There is also Kijowski’s
“classical time of arrival” [Kijowski (1974, 1999, 2005)], and Olkhovsky-
Recami “non-Hermitian approach” [Olkhovsky (2009); Recami et al. (2010,
2013)]. Different approaches are discussed in [Muga et al. (2002b, 2009)].
But even within the standard formulation of quantum theory there is a
lot of space for playing with various probabilistic models by simply relaxing
some of the assumptions that are usually made because of their simplicity.
In decay processes, for instance, there are more and more doubts concerning
the Poisson character of the decay (see e.g. [Aston (2012)]). The Poisson
process is a particular case in the family of general renewal processes that
may better fit the logical scheme of generation of quantum events. In a
general renewal process the interarrival times are assumed to be indepen-
dent and identically distributed random variables, but the increments do
not have to be necessarily stationary. Thus in the long time limit we get
N (t)
limt→∞ t = λ, but there are possible fluctuations around this limiting
value for finite times.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 227/4

Foundational Questions 227

4.2 Are there quantum jumps?

John Bell [Bell (2004)]unhappy with the discrepancies between the math-
ematical formalism and applications of quantum theory popularized two
different “enhancements” of the standard quantum theory: “Bohmian me-
chanics” and Ghirardi-Rimini-Weber (GRW) “spontaneous localization”.
As two representatives of the “standard” formulation he selected the
popular textbook “Quantum Mechanics” by K. Gottfried [Gottfried (1966)]
and the review paper “Ten theorems about quantum mechanical measure-
ments” by N. G. van Kampen [van Kampen (1988)]. Bell began his article
“Are there quantum jumps?” by attributing (without giving a reference) to
Erwin Schrödinger the often quoted sentence:

“If we have to go on with these damned quantum jumps, then

I’m sorry that I ever got involved.”

The title of Bell’s paper is the same as Schrödinger’s 1952 article “Are there
quantum jumps?” [Schrödinger (1952)]. Here is what Schrödinger wrote in
another 1952 paper, “What is Matter?” [Schrödinger (1953)]:

Fifty years ago science seemed on the road to a clear-cut answer

to the ancient question which is the title of this article. It looked as
if matter would be reduced at last to its ultimate building blocks
to certain submicroscopic but nevertheless tangible and measur-
able particles. But it proved to be less simple than that. Today
a physicist no longer can distinguish significantly between matter
and something else.
....
Thus, the subject of this article is in fact the total picture of space-
time reality as envisaged by physics. We have to admit that our
conception of material reality today is more wavering and uncer-
tain than it has been for a long time.
....
Physics stands at a grave crisis of ideas. In the face of this crisis,
many maintain that no objective picture of reality is possible.
...
The theory of quantum jumps is becoming more and more unac-
ceptable, at least to me personally, as the years go on. Its aban-
donment has, however, far-reaching consequences.
...
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 228/4

228 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

In spite of everything, we cannot completely banish the concepts

of quantum jump and individual corpuscle from the vocabulary of
physics.
...
If you finally ask me: Well, what are these corpuscles, really? I
ought to confess honestly that I am almost as little prepared to
answer that as to tell where Sancho Panza’s second donkey came
from.
...
The conservation of charge and mass in the large must be consid-
ered as a statistical effect, based on the law of large numbers.

The question of quantum jumps is so important for the content of the

present book, and Schrödinger’s observations are still so relevant, that I
will quote here in extenso from the extremely important 1955 paper “The
Philosophy of Experiment” [Schrödinger (1955)], and will comment upon
them. In “The Philosophy of Experiment” E. Schrödinger wrote

“Quantum mechanics claims that it deals ultimately and di-

rectly with nothing but actual observations, since they are the only
real thing, the only source of information, which is only about them.
The theory of measurement is carefully phrased so as to make it
epistemologically unassailable. There is no question ever of what is
or is not at a given instant, only of what we should find if we made
this or that measurement; and the theory is only about the func-
tional connection between some group of such findings and some
other group. But what is all this epistemological fuss for, if we have
not to do with actual, real findings ‘in the flesh’, only with imag-
ined findings? And worse still, is not the whole epistemology of the
scheme exploded, if there are any measurements at all, valuable
sources of information, that do not fall under the scheme?”

This is evidently the problem that later on troubled John Bell. Quantum
theory is not easy to learn. One has to practice it. This practice is often
far away from the theoretical principles that one learns from “good books”.
And so Bell noticed that “conventional formulations of quantum theory,
and of quantum field theory in particular, are unprofessionally vague and
ambiguous.” “Professional theoretical physicists ought to be able to do bet-
ter.” Indeed, when learning an advanced course of quantum mechanics, we
learn about Hilbert spaces, operators, ‘observables’, eigenvalues, eigenvec-
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 229/4

Foundational Questions 229

tors, scalar products, Fourier transforms, etc. Yet we are not told which of
these things has “real existence”. We can only guess that atoms are real,
space is real, time is real, but all these real things are just the arena, they
are not a part of quantum dynamics. In experiments we measure time of
events. But is ‘time of event’ a quantum-mechanical observable? Is the
age of the universe an eigenvalue of some quantum mechanical Hermitian
operator? What about the distance between two points? What about the
number of dimensions of our space? What about circumference to radius
ratio of a circle?

The new science (q.m.) arrogates the right to bully our whole
philosophical outlook. It is pretended that refined measurements
which lend themselves to easy discussions by the quantum me-
chanical formalism could actually be made. (I am alluding to the
gamma-ray-microscope, to the location of the electron in a ‘given’
hydrogen atom, and the sort). Actual measurements on single indi-
vidual systems are never discussed in this fundamental way, because
the theory is not fit for it. This in itself is no blame. What is objec-
tionable is the philosophical presumption, which claims reality for
anything the quantum theorist chooses to imagine as measurable,
while he closes his eyes to the fact that few, if any, actual measuring
devices are amenable to discussion under his scheme [Schrödinger,
1955, p. 565].

Today the situation has changed. Several of so called ‘Gedankexperimente’

have been actually performed, perhaps not exactly, but, as a rule, address-
ing the essential issue. More and more experiments are being performed
with single quantum objects. Superconductors, for example, behave like
such objects, and we can make experiments with SQUIDs — testing the-
ories that are able to describe the process of a continuous monitoring of
single quantum systems. We can hold single atoms in magnetic traps long
enough to observe transitions happening with one atom over time. This
dramatic increase of experimental capabilities in recent time has acceler-
ated the development of fundamental and phenomenological theories that
expand the predictive power of the standard scheme of quantum theory.

One can certainly make a case for the view that the sum to-
tal of all observations which have been and ever will be made is
after all the only reality, the only thing that physical science is
concerned with. This view is not self-evident, but it is worth dis-
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 230/4

230 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

cussing. However to maintain the same about all observations that

some school of theoreticians fancies, while in actual fact such ob-
servations are not made and differ in bulk from those that have
been made and on which physical science is based, such a view is
not founded on reason and cannot pretend to passing for serious
philosophy [Schrödinger, 1955, p. 565].

One such issue is, for instance, the tunneling time. According to quan-
tum mechanics, as we are being taught in almost every course, a quantum
particle, say, electron, can tunnel through an extended barrier even if its
energy is, according to the classical physics, insufficient for overcoming the
obstacle. The same phenomenon is observed with photons. But then the
question arises: “How long it takes to tunnel?” Textbook quantum mechan-
ics does not have “time observable”, but tunneling time can be measured.
The problem of tunneling time and arrival time have been discussed in
many papers and causes a lot of controversies. In 1972, ten years after
receiving the Nobel Prize ‘for his contributions to the theory of the atomic
nucleus and the elementary particles, particularly through the discovery
and application of fundamental symmetry principles’, E. P. Wigner ad-
dressed the issue of time in his paper “On the Time-Energy Uncertainty
Relation” [Wigner (1972)]. However his results were inconclusive and his
paper contains errors (cf. [Blanchard and Jadczyk (1997)]). In 1974 J. Ki-
jowski addressed the question of ‘time of arrival’ in his paper “On the time
operator in quantum mechanics and the Heisenberg uncertainty relation
for energy and time” [Kijowski (1974)]. He derived a natural solution for
a self-adjoint “time of arrival” operator for a free Schrödinger particle. In
1994 B. Mielnik re-analyzed the “waiting screen problem” [Mielnik (1994)]
to conclude that “the problem is still open”. In 1999 Kijowski replied to
critics of his solution pointing out that he is not using “unconventional
quantum mechanics” as has been suggested by those authors who did not
study his first paper carefully. In his 2005 paper “ ‘Time Operator’: the
Challenge Persists” [Mielnik and Torres-Vega (2005)] Mielnik and Torres-
Vega notice that “Quite obviously, the waiting detectors form a new class
of measurements, which have rather little to do with traditional Dirac–v.
Neumann observables”. Kijowski replies [Kijowski (2005)]: “Unfortunately,
at the moment there is no measurement theory, which could replace this
(naive and very unsatisfactory!) [conventional] picture.”
Concerning the time needed for a quantum signal to traverse a barrier
there is also an ongoing controversy. For instance in the two volumes of
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 231/4

Foundational Questions 231

“Time in Quantum Mechanics” [Muga et al. (2002b, 2009)] several differ-

ent approaches are being discussed by various authors. In a recent paper
concerning ‘the speed of quantum tunneling’, entitled “The Superluminal
Tunneling Story” Horst Aichman and Günter Nimtz describe this contro-
versy in these words:

“The predicted and measured zero tunneling time inside a bar-

rier was taken as a fantastic nonsense. (...) The antagonists were
scared that a superluminal signal velocity would violate Einstein
causality and time machines would become possible. (...) Several
theoretical studies concluded that tunneling takes place inside a
barrier in zero time. (...) Other studies and text-books denied
fiercely such results up to now ...”

These are just a few examples illustrating the point made above by
Schrödinger, who continues as follows:

“In using such plain language I hate to give offence to those of

my friends who adhere to this kind of view (without realising that
it is of this kind). But I wish to make it clear, that I shoulder now
and ever after the full responsibility for my refractoriness. I am
moving against the stream. But the tide will change.

Schrödinger was right, the wind has already started to change.

There is a habit in some quarters to answer objections of the

kind raised here by saying that they are a matter of philosophical
taste and not relevant to any question physics is really concerned
with. This attitude is an instance of the fact that scientists are
inclined to take their own outlook for the natural way of looking at
things, while the outlooks of others, in as much as they differ from
theirs, are adulterated by preconceived and unwarranted philosoph-
ical tenets, which unprejudiced science must avoid. The ingenious
new-comer to quantum mechanics asks many inconvenient ques-
tions from which, in the considered opinion of the adepts, he must
be weaned. He asks for instance whether the state-transitions in
the atom that accompany the emission of a light quanta are instan-
taneous or whether they take time and pass through intermediate
states. He is told that this question is meaningless and cannot be
answered. Meaning is only attached to the value we find for the
energy if we measure it, this can (by axiom) only be either the
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 232/4

232 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

value of the initial state or that of the final state, the probabil-
ity of finding the latter rather than the former increases with time
continuously in a way that the theory foretells.

The answer to the question of how long it takes to change the quantum state
depends on the model accepted theory of quantum measurement. Today
we are more prepared to address such questions than we were fifty years
ago. Of course there are always textbook answers, but these textbooks are
not addressing the problem directly for the very reason that it is impossible
to describe state-transition using Schrödinger’s continuous evolution alone.
Without approximations the standard quantum mechanics gives us a not
very satisfactory answer: “At best infinite time”.

Another example: our bright disciple may find out for himself,
that according to his theoretical instructions nothing prevents the
velocity of a particle being measured by the time-honoured method
which is practiced on the race-course and by the police (to trace
offenders against the speed-limits), viz. by recording the time taken
by the particle to cover a known distance; and he is perturbed in
noticing that nothing is in the way of carrying the accuracy of
this measurement far beyond the limit imposed by the Uncertainty
Principle. The answer he gets from the initiates is, that this is
indeed so, but causes no worry, since the conflicting data refer to a
bygone moment and cannot be used for predicting the future.

This is an important issue and it is addressed in Sec. 4.6.1.

These examples could be multiplied. The answers are intrigu-

ing; they appear to be unassailable, for they seem to rest on the
simple and safe principle that sound and sober reality, for the pur-
poses of science, coincides with what is (or might be) observed. But
actually this is not the whole story. We are also supposed to admit
that the extent of what is, or might be, observed coincides exactly
with what quantum mechanics is pleased to call observable. I have
endeavoured to adumbrate here that it does not. And my point
is that this is not an irrelevant issue of philosophical taste; it will
compel us to recast the conceptual scheme of quantum mechanics.”

Today several attempts of such a recasting are being actively developed,

as, for instance, ‘Bohmian mechanics’ and GRW spontaneous localization
theory.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 233/4

Foundational Questions 233

4.3 Bohmian mechanics

The idea of “Bohmian mechanics” has been around for so. What is nowa-
days called ‘Bohmian mechanics’ was created and described in detail by
Louis de Broglie. Chapter 15 of his 1930 book “An Introduction to the
Study of Wave Mechanics” [De Broglie (1930)] contains a complete de-
scription of this “mechanics”, even more general than it is done nowadays,
because de Broglie writes his equations for a general, geometrically inho-
mogeneous medium rather than just for a homogenous empty space. Why
then is it called “Bohmian mechanics”? One of the reasons may be the
fact that de Broglie himself considered this particular formalism and ac-
companying “causal interpretation” as too primitive to be true. He was
looking for a more satisfactory solution, which he called “The Double So-
lution”, a solution that would include nonlinearity. This is what he wrote
in the Preface to his 1960 book “Non-linear Wave Mechanics. A Causal
Interpretation”[De Broglie (1960)]:

“Such was the idea that had taken shape in my mind, and its
curious subtlety astonishes me to this day. I called it the “theory of
the Double Solution,” and it was that idea which translated my real
thinking in all its complexity. But, in order to facilitate explaining
it, I had sometimes given it a simplified form, much less profound,
to my way of thinking, which I had named ‘the pilot-wave theory’
in which the particle, assumed given a priori, was considered to be
piloted by the continuous wave.”
In 1952 David Bohm [Bohm (1952)] took de Broglie’s idea, the idea that de
Broglie himself considered as “simplified” and “not profound enough”, and
developed it here and there. Other physicists, for some reasons, decided to
call the whole idea “Bohmian mechanics”. Bohm, writes in his 1952 paper
[Bohm (1952)]:

“(...) After this article was completed, the author’s attention

was called to similar proposals for an alternative interpretation of
the quantum theory made by de Broglie [De Broglie (1930)] in 1926,
but later given up by him partly as a result of certain criticisms
made by Pauli and partly because of additional objections raised
by de Broglie himself.”
One can only wonder how a physicist at Princeton university could not be
aware of a book published in English by Louis de Broglie, one of founders
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 234/4

234 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

of Quantum Theory, twenty years before! In his review of the theory [Bell
(1982)] John Bell makes this simple comment:
“Moreover, the essential idea was one that had been advanced
already by de Broglie in 1927, in his pilot wave picture.
But why then had Born not told me of this pilot wave?”
Apparently, for some obscure reasons, de Broglie’s name, as well as his
ideas, were not worth attention among other western quantum physicists.
The idea described by de Broglie is indeed a simple one. We have a
wave function Ψ(t, x) satisfying the standard Schrödinger equation (suppose
without magnetic field and just for one particle):
∂Ψ 2 2
i =− ∇ ψ + V (x)Ψ. (4.8)
∂t 2m
At a given time t consider those points x in which Ψ(t, x) = 0. In a generic
case it will be “almost the whole space” — exceptional points must be
excluded, the formalism fails at those exceptional points, they need a special
treatment. But at those points where Ψ is non-zero, we can decompose Ψ
into its modulus and phase. De Broglie [De Broglie (1930)] writes it as
2πiφ
Ψ = ae h , (4.9)
while Bohm [Bohm (1952)] as
Ψ = R exp(iS/). (4.10)
Notice that formally
S
=  log Ψ.

Then, as John Bell terms it in [Bell (1980)]:
... the particle rides along on the wave at some position x(t)
with velocity
1 ∂
ẋ(t) =  log Ψ(t, r)|r=x . (4.11)
m ∂r
This equation has the property that a probability distribution for
x at time t
d3 x|Ψ(t, x)|2 (4.12)
evolves into a distribution
d3 x|Ψ(t , x)|2 (4.13)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 235/4

Foundational Questions 235

at time t . It is assumed that the particles are delivered initially by

the source, and then the familiar probability distribution of wave
mechanics holds automatically at later times. Notice that the only
use of probability here is, as in classical mechanics, to take into
account uncertainty in initial conditions.
De Broglie wrote Eq. (4.11) in a slightly different way:
 ∂φ
q̇i = − μik . (4.14)
∂qk
k
The difference of sign arises from the fact that de Broglie was using a dif-
ferent sign convention for writing the Schrödinger equation. In de Broglie’s
“simplified version”, which later became known as “Bohmian mechanics”
we have to solve the standard Schrödinger equation first, then calculate par-
ticle trajectories. It is implied that one of these trajectories is “real”, but
we are uncertain which one — therefore only probabilistic predictions are
possible. This simple idea was then renamed “Bohmian mechanics” and
propagated while distorting the facts. In his review article “Beyond the
Quantum” [Valentini (2009)] Antony Valentini, then at Imperial College,
wrote:
“[At the fifth Solvay conference in 1927] De Broglie’s pilot-wave
theory has been particularly neglected, and its high profile at the
conference severely downplayed. According to Max Jammer’s clas-
sic historical study The Philosophy of Quantum Mechanics, at the
conference, de Broglie’s theory ‘was hardly discussed at all’ and
‘the only serious reaction came from Pauli’, a view that is typi-
cal of standard historical accounts throughout the 20th century.
And yet, the published proceedings show that de Broglie’s the-
ory was in fact discussed extensively: at the end of de Broglie’s
talk, there are nine pages of discussion about his theory; while of
the 42 pages of general discussion (which took place at the end of
the conference), 15 pages include discussion of de Broglie’s theory.
And there were serious reactions and comments from Born, Bril-
louin, Einstein, Kramers, Lorentz, Schrödinger and others, as well
as from Pauli.”
Mike Towler, of Cavendish Laboratory, in his 2009 lecture course at Cam-
bridge University [Towler (2009)], calls de Broglie’s pilot wave idea “Abused
and ignored throughout its history but currently undergoing a major resur-
gence”, and adds to the above the following observations:
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 236/4

236 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

“The theory published in the Proceedings is absolutely pilot-

wave dynamics as we know it today, which is why it is usually
called, er.., ‘Bohmian mechanics’.
....
Bohm continued downplaying de Broglie’s contribution until his
death, see e.g. the following rather naughty extract (from Bohm
and Hiley’s 1993 textbook [Bohm and Hiley (1993)]). Given the
existence of a clear question of priority (which Bohm would lose
under any serious analysis) one would expect him to have paid more
attention to finding out exactly what it was that de Broglie had
done. However, this passage does express the common viewpoint:
The idea of a ‘pilot wave’ that guides the movement of the elec-
tron was first suggested by de Broglie in 1927, but only in con-
nection with the one-body system. De Broglie presented this
idea at the 1927 Solvay Congress where it was strongly criticised
by Pauli. His most important criticism was that, in a two-body
scattering process, the model could not be applied coherently. In
consequence de Broglie abandoned his suggestion. The idea
of a pilot wave was proposed again in 1952 by Bohm in which an
interpretation for the many-body system was given. This
latter made it possible to answer Pauli’s criticism.
The bold remarks are incorrect or misleading. Bohm’s character
was such that he was simply not interested in historical questions
of priority.”
In fact there is a subtle difference between the original version of the pilot
wave theory by de Broglie and that of Bohm. Let us write the Schrödinger
equation (for one particle, in absence of magnetic field, with potential en-
ergy time-independent, and with Planck constant  = 1) as
∂Ψ(t, x) 1 2
i = ∇ Ψ(t, x) + V (x)Ψ(t, x). (4.15)
∂t 2m
Assuming Ψ(x, t) = 0, let us write Ψ(t, x) = R(t, x) exp(iS(t, x), with
R(t, x) > 0, S(t, x). In de Broglie’s theory particle trajectories are obtained
by solving the first order differential equation:
dxi 1 ∂S(t, x)
= . (4.16)
dt m ∂xi
The initial position of the particle is supposed to be ‘random’ (whatever it
means), with the probability distribution
P (0, x) = |Ψ(0, x)|2 . (4.17)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 237/4

Foundational Questions 237

Bohm however reformulated de Broglie’s particle trajectories (de Broglie

himself did not use the ‘particles’ here, he was talking about “probability
elements”) as a second order differential equation:
d2 xi ∂V (x) + Q(t, x)
m 2
=− , (4.18)
dt ∂xi
where the “quantum potential Q” is defined by Ψ:
1 ∇2 R(t, x)
Q(t, x) = . (4.19)
2m |R(t, x)|
Then, to make his trajectories identical with de Broglie’s trajectories, Bohm
introduced an additional postulate, namely that the initial particle mo-
menta are constrained by the equation
∂S(0, x)
pi (0) = . (4.20)
∂xi
The postulate Eq. (4.17) was given the name “quantum equilibrium” by
Dürr, Goldstein and Zanghi [Dürr et al. (1992)]. The term “equilibrium”
was probably selected in such a way that the reader will think about it as
something ‘natural and almost self-evident’. And yet it created controver-
sies and attempts to justify this postulate, one way or another. First of all
it was found that Bohmian trajectories are numerically unstable, and their
computation leads to an exponentially accumulation of errors. Therefore
special numerical procedures are discussed that could partially cure these
problems [Nerukh and Frederick (2000)]. Valentini and Westman, in their
paper entitled “Dynamical origin of quantum probabilities” [Valentini and
Westman (2005)] found that the numerical simulations seemed to indicate
that starting with an arbitrary initial distribution the computed trajecto-
ries seem to evolve the “quantum equilibrium”. On the other hand Colin
and Valentini point out that while de Broglie’s constraint of the quantum
equilibrium may be stable, the second constraint Eq. (4.20), introduced by
Bohm in his reformulation of de Broglie’s theory, is unstable. It is interest-
ing to notice that while de Broglie, Bohm, and “Bohmian mechanicians”
talk about probability distribution of the ‘particle positions’, that, perhaps,
somehow originated at the Big Bang, they are not discussing the probabil-
ity distribution of the pilot waves Ψ. P si is not fluctuating, but x s are.
But why not Ψ?
Dürr and Teufel, in their monograph on “Bohmian Mechanics” [Dürr
and Teufel (2009)] develop another justification of “quantum equilibrium”
formula Eq. (4.17) — they introduced the concept of “typicality”. It
is amazing that they quote John Bell’s paper “Against measurement”
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 238/4

238 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

[Bell (1990)], yet ignore completely his criticism concerning the use of un-
defined concepts such as “macroscopic”. The term “macroscopic” appears
more than 100 times in [Dürr and Teufel (2009)] and is being used in their
arguments supposedly supporting ‘typicality of quantum equilibrium’.
Today quantum trajectories are being used in quantum chemistry. Ev-
idently they are of help in visualization of certain quantum phenomena.
On the other hand quantum opticians seem to be rather critical concerning
the improper use of these trajectories. In 1991 Scully, Englert and Walther
published a paper entitled “Surrealistic Bohm trajectories” [Englert et al.
(1992)]. In 1998 Marlan O. Scully published another paper, this time with
the title “Do Bohm Trajectories Always Provide a Trustworthy Physical
Picture of Particle Motion?” [Scully (1998)], with his short answer in the
abstract:

Abstract
“No. When particle detectors are included particles do not form
trajectories as we would expect from a classical type model.”

He also wrote:

“(...) But the aspect of the talk which precipitated the most
spirited discussion was the contention of Englert, Scully, Süssman
and Walther [Englert et al. (1992)] (ESSW) that the Bohm tra-
jectories can be shown to be ‘surrealistic’ by using a micromaser
which way or Welcher Weg (WW) detector. That this conclu-
sion should precipitate debate is understandable since Bohmian
mechanics (BM) has undeniable charm. It provides an interesting
perspective into quantum phenomena in that the Bohmian trajec-
tories seem to provide more information and insight than quantum
mechanics (QM) typically allows, see Fig. 1. Furthermore, it is
widely held that BM always agrees with QM; so it would not, one
would think, lead us into error.”

Indeed, there is some charm in these trajectories. For a typical two-slit

quantum interference experiment we get trajectories that seem to indicate
the paths of some ‘real particles’ as in Fig. 4.9. But, as Scully notices
[Scully (1998)],

“Unfortunately nature does not seem to allow us to view BM

as just another version of QM. The crux of the matter is this: In
analyzing several kinds of interferometers containing WW detectors
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 239/4

Foundational Questions 239

we find that the observed track of the particle is macroscopically

at variance with its Bohm trajectory”.

Basil Hiley, David’s Bohm collaborator and the coauthor of “Undivided

Universe’ [Bohm and Hiley (1993)] tried to solve the situation that so much
“upset the Bohmians at the University of Munich”. He did it by explaining
that

“To make this clear it was necessary to recall that the Bohm
trajectories were none other than the streamlines of probability.”

Thus Hiley, discussing the ontological status of “Bohmian trajectories” re-

turns to the original de Broglie’s terminology. De Broglie, in his original
presentation, was talking about motion of “probability elements”. Hiley is
not acknowledging this fact. On the other hand Hiley writes:

“All Bohm does is to assume that individual particles stick to

the flux lines”.

The picture resulting from these debates may indeed look somewhat ‘sur-
realistic’.

Fig. 4.9 ‘Bohmian trajectories’ for the double slit configuration. Instead solving of
Schrödinger equation for this very problem (which can be done only numerically) a
number of free Gaussian moving packets, with origins at selected points inside the slits,
are superposed.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 240/4

240 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

4.4 Event Enhanced Quantum Theory

It is a common belief of quantum physicists and of other enthusiasts of

quantum theory that “everything is quantum”. And if it not yet, “it should
be quantized”. Why? Because, they say, everything is build of atoms, and
atoms belong to the quantum world and obey quantum laws. That is what
they say.
But let us examine this logic. While it is true that quantum theory ap-
plies to atoms, and it does with an amazing success, it does not follow that
‘everything about atoms” must be quantum. Protons, electrons, they have
their masses and electric charges. Is the unit of the electric charge fluctuat-
ing? Is it random? Indeterminate? Did we ever notice “complementarity”
between the electric charge and some other quantity? The answer is “no”.
It seems to be constant.
We are not able, today, to explain this constancy. It seems that we have
a conservation law here, a law that can be related to some deeper symmetry,
“gauge symmetry” as we call it, yet invoking this symmetry does not help
us in understanding the deeper nature, or a “mechanism” that works behind
the scenes. We are not able to “compute” the electric charge starting from
some first principles. We are also not able to calculate from such principles
the value of the dimensionless “fine structure constant” α that is related to
charge e, speed of light c, and Planck constant . Quantum theory is of no
help here. And yet these are distinct properties of the quantum world.
Quantum waves are known for their interference effects. The two-slit
experiment, with electrons emitted one after another, as in Tonomura’s
and earlier experiments, is an example here. We are now able to get similar
effects with huge molecules consisting of hundred of atoms, yet we do not
see such effects with tables and chairs and cats. These effects occur in very
special experimental arrangements.
Quantum effects would be impossible without first setting up proper
conditions that are described in terms of classical concepts. Quantum the-
ory without classical logic, classical concepts, classical language, is mean-
ingless. We do have macroscopic quantum effects, like in supersensitive
quantum interference devices (SQUIDs) used for measuring of ultra-weak
magnetic fields. Yet in these quantum devices, together with superconduct-
ing elements obeying quantum laws, we have also classical electric circuits,
without which SQUID’s would be useless (see Fig. 4.11).
With time we will probably discover more and more macroscopic quan-
tum phenomena, perhaps even on a cosmic scale. Yet they will occur within
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 241/4

Foundational Questions 241

the classical framework. We, human beings, are partly quantum and partly
classical. Why should we deny this observation?
In a sense quantum theory “explains” why it is so difficult to have
quantum phenomena on a macroscopic scale. But is it a real explanation
when quantum theory itself remains unexplained? As Richard Feynman
succinctly puts it: “Nobody understands quantum theory”.
Usually quantum physicists blame the “environment” for the fact that
quantum phenomena are not a part of our everyday experience. But what
exactly is this “environment”? Splitting the universe into “a system under
observation” and “environment” is subjective. Attempts to make it objec-
tive have failed, and they must fail, because the very concept of a hard
“splitting” is classical. Why not accept this fact from the very beginning?
Talking about the system and its environment may be useful, may be even
good FAPP (“for all practical purposes”), but it does not belong to “fun-
damental physics”. It has nothing to do with “Laws of Nature”, it has
everything to do with what is convenient. It is certainly not in the spirit
of, as the French philosopher Bernard d’Espagnat terms it, “objectizing
physics”. Therefore why not try the “Columbus solution”?
Christopher Columbus, when challenged with the problem of how to
make an egg stand vertically, the problem that others could not solve despite
their efforts, simply broke the shell from one end — the simple and bold
idea that did not occur to others, but which was still within the unspoken
rules of the game. Therefore, in simple terms: Not all is quantum.
While the future is uncertain and may need quantum description, the
past is rather well set and can be described in classical terms. Even if the
past can be partly erased, nevertheless it belongs to the classical world.
Facts and events are classical, and their formal description should be based
on classical concepts. Possibilities (or “propensities”) belong to the quan-
tum world.
The past is evidently coupled to the future. Past events can influence
future possibilities and probabilities. Probabilities, when they actualize,
create events that form the past. EEQT, the Event Enhanced Quantum
Theory, is a mathematical model that describes such a coupling through
equations and algorithms. Equations describe the continuous time evolu-
tion of statistical averages. Algorithms describe creation of histories which
then can be statically averaged over.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 242/4

242 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

4.4.1 Piecewise deterministic process

In his two monographs [Davis (1984, 1993)], dealing with the subject of
stochastic control and optimization, M. H. A. Davis, having in mind mainly
queuing and insurance models, described a special class of piecewise deter-
ministic processes that was later found to fit almost perfectly the needs of
quantum measurement theory. For the quantum model we sometimes have
to extend the original Davis’ framework, and to work with jumps between
continuously parameterized states, and not just between a denumerable
family of manifolds. Yet for the present purpose we restrict ourselves to
the discrete case, and leave the problem of a rigorous formulation of its
evident extension to continuous families aside.
Let α be an index running over a finite or countable set J. In ap-
plications to quantum-classical couplings, α will number the states of the
classical system. For instance, suppose that we have just one detector with
two possible states, 0 and 1, then α will take only two values. For n such
detectors α will take 2n values. If, on the other case, the classical sys-
tem is a mechanical system with phase space parametrized by continuous
parameters p, q, then α will have values in R2 .
In general, for each α we may have a different Hilbert space. This takes
into account the possibility of phase transitions. Such a possibility, till
now, was never analyzed in any detail. In the future it may prove to be
the most important aspect of the formalism. The classical phase space may
then be replaced by the space of representations (or of equivalence classes
of representations) of some operator algebra.
We will be interested in functions defined on the product of the spaces
of classical and quantum states. While the index α runs over the space of
the classical system, the quantum system, for a given α, is represented by
a vector ξ in a complex Hilbert space. In some applications ξ may happen
to be an “improper vector” (with a formalism of Gelfand triples or “rigid
Hilbert spaces”), let us leave these possible generalizations aside.
Therefore, let us consider functions, with possibly complex values
f (ξ, α), where for each α the variable ξ is continuous and runs through
some set Mα . This set can, in particular, coincide with the unit ball in
some Hilbert space L2 (E) (modulo the phase). We will be dealing with
irreversible evolutions, since the random event creation mechanism is irre-
versible. In the simplest case of homogeneity in time, we expect the time
evolution will be described by a semigroup of transformations. That means
our translations will go from the past into the future, and we should not
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 243/4

Foundational Questions 243

expect that they will necessarily have an inverse. Even if they have an
inverse, the inverse does not have to be defined for all functions under con-
sideration. Therefore suppose we have a semigroup of transformations σt
acting, by linear transformations, on the space of such functions. We can
define then the infinitesimal generator of the semigroup by

1
(Df )(ξ, α) = lim (σt f − σ0 f )(ξ, α), (4.21)
t↓0 t

where the exact meaning of the convergence remains to be precisely defined,

depending on the case in hand.
Let us assume that the infinitesimal generator D is an integro-differential
operator of the following form:

(Df ) (ξ, α) = (Zα f )(ξ, α)

  (4.22)
+λ(ξ, α) β Mβ
Q(ξ, α; dξ  , β) (f (ξ  , β) − f (ξ, α)) ,

where Zα are vector fields that generate one-parameter flows φα on Mα ,

λ(ξ, α) are non-negative functions, while Q(ξ, α; dξ  , β) are (non-negative)
transition measures — thus satisfying


Q(ξ, α; dξ  , β) = 1, (4.23)
β Mβ

and also

Q(ξ, α; dξ  , α) = 0, (4.24)
{ξ}

for all α and ξ ∈ Mα .

The last condition means that we exclude from our description jumps
that are not accompanied by classical events. We notice that by the very
definition we have

Zα (ξ) = dφα (ξ, t)/dt |t=0 . (4.25)

Then, as it is shown in Refs. [Davis (1984, 1993)], one can associate with
this generator D a piecewise-deterministic stationary Markov process that
is described as follows.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 244/4

244 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Mα0

Mα1
φα0 (ξ0 , t1 )
ξ0
Mα2
ξ2 φα1 (ξ2 , t3 )

φα2 (ξ1 , t2 )
ξ1

Fig. 4.10 Trajectory of a piecewise deterministic process. Continuous evolution on

manifolds is interspersed with jumps. In general the manifolds Mα0 , Mα1 , Mα2 , ... can
be different, even of different dimensions. But in most applications they can be identified
either in a natural way, or in an “observer dependent way”. Therefore in most cases we
may restrict ourselves to the case when Mα0 = Mα1 = Mα2 = ... = M.

4.4.2 Algorithm for the piecewise deterministic process

(PDP)

Suppose the process starts at some point (ξ0 , α0 ). Then ξ evolves con-
tinuously along the vector field Zα , ξt = φα (ξ0 , t), while α0 remains con-
stant until a jump occurs at a certain random time t1 . The time of this
jump is governed by a (inhomogeneous) Poisson process with rate func-
tion λ(t) = λ(ξt , α0 ). When jump occurs at t = t1 , then (ξt1 , α0 ) jumps
to (ξ  , α) with probability density Q(ξt1 , α0 ; dξ  , α) and the process starts
again.

Remark 4.1. Notice that the probability that the jump will occur between
t and t + dt, provided it did not occur yet, is equal to
 t+dt
1 − exp − λ(s)ds ≈ λ(t)dt. (4.26)
t

This justifies calling λ the rate function.

4.4.3 Association of the semigroup with PDP

Association of the random process with the semigroup σt is canonical and
can be described as follows: first one goes from σt that acts on functions
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 245/4

Foundational Questions 245

f (ξ, α) to its dual σ t acting on measures. Formally:

 
σt (f )μ dξdα = f σ t (μ)dξdα. (4.27)

Then, choosing the Dirac measure δξ0 ,α0 concentrated at (ξ0 , α0 ) as the
initial point μ0 in the space of measures, we apply to it σ t to get μt = σt (μ0 ).
The resulting measure μt is then characterized by the fact that dμt (ξ, α)
is equal to the probability density that the process starting at t = 0 from
(ξ0 , α0 ) will end, at time t, at the point (ξ, α).
A detailed and precise description of the above correspondence should
include specification of the involved measure structures and domains of
definition. These are given in detail in Refs. [Davis (1984, 1993)]. Here
let us only notice the following important relation: let K(t; ξ, α; dξ  , β) be
the transition function for the process. That means K(t; ξ, α; dξ  , β) is the
probability that the process starting at time t = 0 at (ξ, α) will reach, after
time t the volume element dξ  on the manifold Mβ . Then the semigroup σt
is given by the formula

(σt f )(ξ, α) = K(t; ξ, α; dξ  , β)f (ξ  , β). (4.28)
β

Of course, if the index α is continuous, then the sum in the above formula
should be replaced by an integral.

4.4.4 Central classical observables

In EEQT we assume, as it is usually done in quantum theory, that the
important object is a -algebra of operators A. In quantum theory it is
usually assumed that all algebras and all Hilbert spaces are over the field of
complex numbers C. But the reasons why it must be so are not completely
clear. As the abstract formalism of quantum theory is finding, with time,
more and more applications outside physics (‘quantum games’, ‘quantum
psychology’, etc) may well happen in these domains; we will be dealing with
algebras and spaces over different number fields, and even not necessarily
fields, in the mathematical sense.
On the other hand there are extensions of the quantum formalism that
are more general, for instance based on ‘quantum logic’. At the present
moment the formalism of EEQT should work both for real and for complex
fields, as long as the starting point is some -algebra of linear operators (or
matrices).
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 246/4

246 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

The star operation in the algebra is either the hermitian (or, in case
of indefinite scalar product, pseudo-hermitian) conjugation or, in real case,
the operation of taking the transpose.
For historical reasons A is called an ‘algebra of observables’, even if
only normal operators, that is, those which commute with their adjoints,
are believed to be directly related to observable physical quantities.
In EEQT the elements of A, even if they can represent ‘physical quanti-
ties’, can neither be observed nor do they represent, as it is assumed within
the standard interpretation, ‘observational procedures’ — except in a limit
that is rather unrealistic. In EEQT the operators in A do exactly what
they are supposed to do: they operate on states to produce new states that
result from quantum events. They implement quantum jumps that ac-
company any event and any process of information transfer related to the
quantum system.
It should be noted that in EEQT we do not import any a priori proba-
bilistic interpretation of the standard quantum theory. All interpretation is
being derived from the Piecewise Deterministic Process (PDP) introduced
above. Interpretation of eigenvectors, eigenvalues, mean values of observ-
ables, etc. should be derived from the dynamics of EEQT. A major part
of the standard wisdom about eigenvalues and eigenvectors can, in fact,
be justified within EEQT, and so it can be used as a heuristic tool for
constructing mathematical models of ‘real world’ situations.
The algebra A is usually assumed to be a C or a von Neumann algebra,
but EEQT can work also in spaces with indefinite scalar product or within
a Clifford algebra framework.

Definition 4.1. An algebra A, with unit 1, over R or C, is called a normed

algebra if A, as a vector space, is endowed with a norm such that

||AB|| ≤ ||A|| · ||B|| (4.29)

for all A, B ∈ A, and ||1|| = 1.

If, as normed space, A is a Banach space (i.e. it is a complete metric
space, that means all Cauchy sequences are automatically convergent), then
A is called a Banach algebra. In a normed algebra multiplication is contin-
uous. Every finite dimensional matrix algebra is automatically a Banach
algebra.
Let A be a real algebra with unit (notice that any complex algebra can
be considered as a real one). A mapping  : A → A of A onto itself is
called an involution provided the following properties hold:
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 247/4

Foundational Questions 247

i) (A ) = A.
ii) (A + B) = A + B .
iii) (AB) = B A .
iv) (αA) = αA , α ∈ R,
v) 1 = 1.
If A is over C, we require that
iv’) (αA) = ᾱA , α ∈ R,
where α → ᾱ is the complex conjugation. A Banach algebra with involution
is called a Banach  algebra if
vi) ||A || = ||A||.
In a Banach  algebra the involution is automatically continuous. A Banach
 algebra is called a C algebra if, additionally,
vii) ||A A|| = ||A||2 .
If, additionally, a C algebra A can be realized as an algebra of operators
on some (real or complex) Hilbert space, and if A is closed in the weak
operator topology, then A is called a von Neumann algebra.

Thus every von Neumann algebra is a C algebra, every C algebra is

a Banach  algebra, every Banach  algebra is an normed algebra with
involution. The physical meaning of involution is not clear. Perhaps it is
somehow related to time reversal.
A generic algebra A will have a non-trivial center Z — the set of all
A ∈ A which commute with all the elements of A.

Definition 4.2. The center Z of an algebra A consists of all those elements

C ∈ A that commute with all elements of the algebra:
CA = AC, ∀A ∈ A. (4.30)
The center of an algebra with unit 1 is an algebra with the same unit. In
fact the center contains all elements α1, where α runs over all scalars. The
center is always a commutative algebra. If the center of the algebra consists
only of scalar multiples of its unit, the algebra is called simple.

Since the center Z is Abelian — it represents the classical subsystem.

Von Neumann algebras with trivial center (i.e. center consisting of oper-
ators that are complex multiples of the identity) are also called factors.
Physicists insisting on the idea that there are no genuine classical degrees
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 248/4

248 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

of freedom are, in fact, insisting on the idea that only factors should be
used for an algebraic description of quantum systems. While it is true that
every algebra can be decomposed, essentially uniquely, into a direct sum (or
integral) of factors, restricting to factors alone is like restricting to prime
numbers alone.
While it is true that any integer can be decomposed into a product
of prime numbers, insisting on the idea that only prime numbers should
be used would be simply silly. Atoms build molecules. There would be no
life without molecules. Similarly factors build more complex non-factors.
According to our definition below, there would be no ‘events’ without non-
factors! Thus there would be no data (recording a datum is an event) that
could be used in experiments.
Every Abelian (i.e. commutative) algebra has only one-dimensional ir-
reducible representations. These are called characters, and the set of all
characters of Z is called the spectrum of Z. The above is not a very precise
definition — some assumptions are missing. In order to get an idea about
this important subject (Abelian algebras, after all, represent the classical
world) let us give an illustrative example.

Example 4.1. consider the algebra C(K) of all functions (real or complex
valued) on a set K. To be precise, we may assume that K is a compact
Hausdorff space, the functions are continuous, and the norm on C(K) is
defined as the sup norm:
||f || = sup |f (x)|. (4.31)
x∈K

Then C(K) is a C algebra. Suppose now that K is the unit disk in the
complex plane:
K = {z ∈ C : |z| ≤ 1}.
The set D(K) of all continuous complex valued functions on K that are
analytic inside K is a Banach subalgebra of C(K). If z is a point in K, then
the map σz : D(K) → C defined by
σz (f ) = f (z) (4.32)
is a continuous algebra homomorphism (a character) from D(K) to C.
Moreover, every character on D(K) is of such a form. In other words:
given the algebra of functions D(K), with all its topology and algebra
structure, we can rediscover the space K. It is this idea that motivates
non-commutative geometry. Sometimes the ordinary concept of “space”
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 249/4

Foundational Questions 249

may become too restrictive. For instance the space may have some fractal
properties, may cease to have a well defined dimension or even topology.
The hope is that the algebra of space can be somehow handled better than
space itself. Even if the algebra happens to be noncommutative, in which
case it can not be simply interpreted as a space of functions on some un-
derlying set.

By quite general representation theorems, each Abelian algebra is nat-

urally isomorphic to an algebra of functions over its spectrum (continuous,
measurable etc., depending on the type of the algebra). For simplicity let
us assume that the spectrum of Z is discrete-countable, or even finite. With
proper care we could consider more general cases — as for instance in the
SQUID-tank model (cf. Fig. 4.11, p. 250), where the spectrum of Z is
a symplectic manifold — the phase space of a radio-frequency oscillator
(cf. [Blanchard and Jadczyk (1994); Olkiewicz (1997)], and also [Olkiewicz
(1999); Blanchard and Olkiewicz (1999a,b)] for other examples of working
EEQT models with a continuous spectrum of Z).
Heuristically the points of the spectrum of Z are the ‘pointer positions’
— that is, states of the classical subsystem — we will denote the spectrum
of Z by the letter C.

Definition 4.3. Discrete changes of states of C are called events.

When the set of classical states is discrete, then any change of the state is
discrete. But, for instance, in models with a continuous spectrum (as, for
instance, when C is a phase space {q, p}) we will have a continuous evolution
of the state of C that is interrupted by events, for instance jumps in the
momentum p (instantaneous boosts) in C. It is important to observe at this
place, that in such models we may have discontinuity in the momentum
(or velocity) without discontinuity in space. Thus the classical trajectory
will suddenly change the direction, but will not jump to another point in
space. Somehow such jumps are (philosophically) more acceptable than
jumps from one point in space to another.

4.4.5 Quantum Events Theory — Duality

EEQT starts therefore with the realization that any formal description
of Reality must have a dual, partly classical and partly quantum nature.
Those who deny this, contradict themselves since the very act of denial is
a classical event!
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 250/4

250 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Amplifier
detector electronics
μ
Josephson Junction

coupling loop
Φ̂, L LT CT

SrTiO3

Quantum SQUID Classical RF Tank Circuit SQUID

(a) (b)

Fig. 4.11 SQUID — a model for quantum-classical coupling. (a) A schematic view of
the quantum-classical coupling as implemented in a SQUID (Superconducting Quantum
Interference Device) magnetometer. The flux in the macroscopic (of the size of about
1000 nm) superconductive ring is governed by the quantum laws. The coupled radio-
frequency circuit is governed by the classical laws. (b) Using substrate resonators, the
performance of a SQUID can be greatly enhanced. A standard SrTiO3 substrate with
dimensions 10 × 10 × 1 mm3 serves as a tank circuit (resonator), a YBCO thin film
SQUID washer structure is patterned on it. On the resonator substrate, a small RF
washer SQUID with a step-edge junction is positioned in flip-chip geometry, thus forming
a magnetometer sensor. At 77 K, the field sensitivity of this SQUID magnetometer
reaches 24 fT/root(Hz) in the white noise range. (Courtesy of www.jsquid.com)

Indeed, as stressed already by Niels Bohr, the sentences that physicists

write, the conclusions they come to, are all classical in nature. In [Bell
(2004)] John Bell wrote:

“But we cannot include the whole world in the wavy

part. For the wave of the world is no more like the world
we know than the extended wave of the single electron is
like the tiny flash on the screen. We must always exclude
part of the world from the wavy ‘system’, to be described
in a ‘classical’ ‘particulate’ way, as involving definite events
rather than just wavy possibilities.”

The fact of communicating anything through some channel, in finite time,

is an ‘event’ — and as such, it is classical. It happens. However, there
are no events in standard quantum theory, they do not belong to quan-
tum dynamics, and the standard quantum theory does not provide us with
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 251/4

Foundational Questions 251

any understanding of why, how, and when they happen. That is why the
standard theory is incomplete.
In 1986 John Bell, envisioning a possibility of creating a new, more
complete theory wrote [Bell (1987b)]:

“And surely in fundamental theory this merging [of classical and quan-
tum] should be described not just by vague words but by precise math-
ematics? This mathematics would allow electrons to enjoy the cloudi-
ness of waves, while allowing tables and chairs, and ourselves, and
black marks on photographs, to be rather definitely in one place rather
than another, and to be described in ‘classical’ terms. The necessary
technical theoretical development involves introducing what is called
‘nonlinearity’, and perhaps what is called ‘stochasticity’, into the basic
‘Schrödinger equation’.”

EEQT is a step in this direction, a step involving a minimal nonlinearity,

non-unitarity, and stochasticity (randomness). The new mathematics of
EEQT, based on piecewise deterministic processes, enables us not only to
understand how events come into being, but also to understand how and
why the simultaneous measurement of non-commuting observables leads to
chaotic dynamics that could not have been anticipated by the founders of
quantum theory.

4.4.6 Completely positive maps

Historically, EEQT started with an attempt at describing time evolution
of a system with a non-trivial center, in the simplest case with the total
system algebra being a tensor product A = Aq ⊗ Acl , and Z ≈ Acl , where
there would be a dynamical coupling and mutual exchange of information
between the quantum and the classical degrees of freedom. In a more
general case we would have an algebra A that has a non-trivial center, but
the algebra is not isomorphic to the tensor product of some simple “pure
quantum algebra” and an Abelian “pure classical algebra”. In such a more
general case the “classical part” of the whole system can be precisely defined
(the center of the total algebra), but there is no such thing as its “quantum
part”!
In algebraic quantum theory an important role is being played by alge-
bra automorphisms. Usually they are associated with “symmetries of the
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 252/4

252 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

system”, including “time translation symmetry” related to the dynamics of

the system. Now, algebra automorphisms preserve the center of any alge-
bra, therefore it is clear that automorphisms could not be used as tool for
dynamical coupling between classical and quantum degrees of freedom.
In my private discussions with one of the founders of the algebraic
approach to quantum theory, professor at the University of Hamburg,
Rudolph Haag, in the eighties, he repeatedly expressed his personal doubts
about the physical significance of the algebraic product in the “algebra of
observables”. Even if the algebraic product AB is useful in setting up the
canonical commutation relations of quantum theory, the product of observ-
ables, if they do not commute, is not itself an observable and, therefore,
need not be necessarily preserved by the time evolution when irreversible
recording is taking place.
What does seem to have a clear physical meaning is, however, positiv-
ity in the algebra, therefore the simplest generalization of the automorphic
evolution takes us to semigroups of positive maps.
Positivity itself is not, however, a stable property. Adding spurious
degrees of freedom which do not participate in the dynamics can destroy
positivity. The more stable condition is called ‘complete positivity’. It is
defined as follows:

Definition 4.4. Let A, B be C -algebras. A linear map

φ:A→B

is Hermitian if

φ(A ) = φ(A) .

It is positive if and only if A ≥ 0, A ∈ A implies φ(A) ≥ 0.

Because Hermitian elements of a C -algebra are differences of two pos-
itive ones — each positive map is automatically Hermitian.
Let now Mn denote the n by n matrix algebra, and let

Mn (A) = Mn ⊗ A

be the algebra of n × n matrices with entries from A. Then Mn (A) carries a

natural structure of a C -algebra. With respect to this structure a matrix
A = (Aij ) from Mn (A) is positive if and only if it is a sum of matrices of
the form

(Aij ) = (Ai Aj ), Ai ∈ A.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 253/4

Foundational Questions 253

If A is an algebra of operators on a Hilbert space H, then Mn (A) can

be considered as acting on
.
Hn = H ⊗ Cn = ⊕ni=1 H.

Positivity of A = (Aij ) is then equivalent to

(Ψ, AΨ) ≥ 0 , Ψ ∈ Hn , (4.33)

or, equivalently, to

(Ψi , Aij Ψj ) ≥ 0 for all Ψ1 , . . . , Ψn ∈ H. (4.34)
i,j

Definition 4.5. A positive map φ is said to be completely positive or,

briefly, CP if and only if φ ⊗ idn : A ⊗ Mn → B ⊗ Mn defined by
(φ ⊗ idn )(A ⊗ M ) = φ(A) ⊗ M, M ∈ Mn , is positive for all n = 2, 3, . . . .

When written explicitly, complete positivity is equivalent to


n
Bi φ(Ai Aj )Bj ≥ 0 (4.35)
i,j=1

for every A1 , . . . , An ∈ A and B1 , . . . , Bn ∈ B. In particular every homo-

morphism of C algebras is completely positive.

One can show that if either A or B is Abelian, then positivity implies

complete positivity. Another important example: if A is a C algebra of
operators on a Hilbert space H, and if V ∈ B(H), then φ(A) = V AV is a
CP map φ : A → φ(A).
In quantum dynamics of open systems the unitary time evolution de-
scribed by the Schrödinger equation is usually replaced by a semigroup of
completely positive maps (also known as a ‘dynamical semigroup’) [Alicki
and Lendi (1987); Alicki (2002)]. Usually such semigroups are being studied
on the von Neumann algebra A of all bounded linear operators A = L(H)
on a separable Hilbert space H.
In the algebraic framework [Emch (1972)] we learn that more general
von Neumann algebras can also appear in physical applications, in particu-
lar, as discussed above, algebras with a non-trivial center Z = A ∩ A , where
A is the commutant of A. The nontrivial central elements lead to supers-
election sectors (cf. [Landsman (1991)], and references therein), and, due
to their commutativity with all observables, they represent the ‘classical
observables’ of the theory.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 254/4

254 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Applying open system dynamics to an algebra with a non-trivial center

brings in new possibilities, with an interesting new result that there is a one-
to-one correspondence between a class of completely positive semigroups
and piecewise deterministic random processes (PDP — cf. [Davis (1993)])
on the space of pure states of the algebra. In some cases, the associated
piecewise deterministic process can be interpreted as a nonlinear iterated
function system on a complex projective space of rays in the Hilbert space
H, with a fractal attractor, and with a range of Lyapunov’s exponents
depending on a particular value of the coupling constant in the semigroup
generator [Blanchard et al. (2001)]. Pure jump processes, on the other
hand, lead to quantum fractals — the main subject of this book.
Let us now assume that the algebra A of observables is a von Neumann
algebra. The points of the spectrum of its center Z represent (pure) states
of the Abelian subalgebra — we may call them superselection sectors. Let
us denote these states α = 1, . . . , m. The algebra A is then of the form
A = ⊕m α=1 Aα , where Aα are factors (that is, they have a trivial center).
We are interested in the simplest case, where Aα = L(Hα ), where Hα
is a Hilbert space of dimension (possibly infinite) nα . Thus every element
A ∈ A is represented by a family {Aα } of operators Aα ∈ L(Hα ), or as a
.
block diagonal matrix operator A = diag(A1 , . . . , Am ) on H = ⊕m α=1 Hα .
Every normal state ρ of A is represented by a density matrix on H ,
that is, by a family {ρα } of positive, trace-class operators on Hα , with
m m
α=1 Tr (ρα ) = 1, and ρ(A) = α=1 Tr (ρα Aα ).

4.4.7 Dynamical semigroups on an algebra with a center

The most general form of a generator of a completely positive semigroup is
then given by the formula of Christensen and Evans [Christensen and Evans
(1978)], which generalizes the classical results of Gorini, Kossakowski and
Sudarshan [Gorini et al. (1976)] and of Lindblad [Lindblad (1976)] to the
case of an arbitrary C -algebra. It is worthwhile to cite, after Lindblad, his
original motivation:

The dynamics of a finite closed quantum system is convention-

ally represented by a one-parameter group of unitary transforma-
tions in Hilbert space. This formalism makes it difficult to describe
irreversible processes like the decay of unstable particles, approach
to thermodynamic equilibrium and measurement processes. [. . .]
It seems that the only possibility of introducing an irreversible
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 255/4

Foundational Questions 255

behavior in a finite system is to avoid the unitary time develop-

ment altogether by considering non-Hamiltonian systems.

Theorem 4.1 (Christensen-Evans). Let σt = exp(Lt) be a norm-

continuous semigroup of CP maps of a C -algebra of operators A ⊂ L(H).
Then there exists a CP map φ of A into the ultraweak closure Ā and an
operator K ∈ Ā such that the generator L is of the form:

L(A) = φ(A) + K A + AK . (4.36)

The set of all CP maps φ : A → A is convex. Of particular interest to us

are generators L for which φ is extremal. Arveson [Arveson (1969)], using
the celebrated Stinespring theorem [Stinespring (1955)], proved that this is
the case if and only if φ is of the form

φ(A) = V π(A)V , (4.37)

where π is an irreducible representation of A on a Hilbert space K, and V :

H → K is a bounded operator (it must be, however, such that V π(A)V ⊂
A). Then φ(I) = V V. In the following we will assume that all nα < ∞,
then Ā = A, so that K = {Kα } ∈ A.
We will always assume that σt (I) = I or, equivalently, that L(I) = 0. It
is convenient to introduce Hα = i(Kα −Kα )/2 ∈ L(Hα ), then from L(I) = 0
we get Kα + Kα = −φ(I)α , and so Kα = −iHα − φ(1)α /2. Therefore we
have

L(A)α = i [Hα , Aα ] + φ(A)α − {φ(1)α , Aα }/2, (4.38)

where { , } denotes the anticommutator.

Using the Arveson result it is easy to see that, in our case, φ is a non-
zero extremal CP map if and only if V is if of the form V = {Vαβ }, where
only one matrix entry Vα0 β0 : Hβ0 → Hα0 is non-zero. Taking for φ a sum
of maps of such a type we end up with a generator L of the form:
 1
L(A)α = i[Hα , Aα ] + gβα Aβ gβα − {Λα , Aα }, (4.39)
2
β

where gαβ ∈ L(Hβ , Hα ) and


Λα = gβα gβα ∈ L(Hα ). (4.40)
β
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 256/4

256 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

4.4.8 Liouville equation for states

Taking into account the duality between observables and states, given by
m
the valuation < ρ, A > = Tr (ρA) = α=1 Tr (ρα Aα ), the evolution equa-
tion for the semigroup Ȧ = L(A) can be rewritten in terms of states:
 1
ρ̇α = −i[Hα , ρα ] + gαβ ρβ gαβ − {Λα , ρα }. (4.41)
2
β

Notice that the total trace is automatically conserved:

d 
Tr (ρ) = Tr (ρ̇α ) = 0.
dt α

In problems that are explicitly time-dependent, as it is in most cases where

there is an explicit intervention of the ‘experimenter’, who sets up the
characteristics of the measuring device according to the needs of the exper-
iment, the maps φ and K, and thus the operators Hα and gαβ will depend
on time, and they will generate a family σt of CP maps, which will not have
the semigroup property.

4.4.9 Ensemble and individual descriptions

There are two levels of descriptions in EEQT: the ensemble level and the
individual level. At the ensemble level the description is through a deter-
ministic, smooth Liouville evolution of statistical states.
At the individual level the description is through a piecewise determin-
istic process on the space of pure states, where a continuous, nonunitary,
evolution is interrupted by discontinuous catastrophic events.
One goes from the individual to the ensemble description by averaging
over many sample paths. The averaging process smoothes out discontinu-
ities and nonlinearities.
The jump probabilities in the process are computed from the formula:
gβα (t)ψ2
pα→β (ψ, t) = . (4.42)
< ψ, Λα (t)ψ >
It has been shown in [Jadczyk et al. (1996)] that when the diago-
nal terms gαα all vanish (cf. Eq. (4.24)), then there is a one-to-one cor-
respondence between the solutions of the Liouville equation Eq. (4.41),
and PDP processes on the space of pure states of the algebra A, where
the process realizing the solution of Eq. (4.41) with the initial pure state
ρ = (0, . . . , |ψα0 >< ψα0 |, . . . , 0) is described as follows:
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 257/4

Foundational Questions 257

PDP Process: Given an input t0 , α0 , and ψ0 ∈ Hα0 , with ψ0  = 1, it

produces on output t1 , α1 and ψ1 ∈ Hα1 , with ψ1  = 1.
1) Choose uniform random number r ∈ [0, 1].
2) Propagate ψ0 in Hα0 forward in time by solving:
1
ψ̇(t) = −iHα0 (t) − Λα0 (t) ψ(t) (4.43)
2
with initial condition ψ(t0 ) = ψ0 until t = t1 , where t1 is defined by 2
ψ(t1 )2 = r. (4.44)

3) Choose a uniform random number r ∈ [0, 1].
4) Run through the classical states α = 1, 2, . . . , m until you reach α = α1
for which

α1
pα0 →α (ψ(t1 ), t1 ) ≥ r . (4.45)
α=1

5) Set
gα1 α0 (t1 )ψ(t1 )
ψ1 = . (4.46)
gα1 α0 (t1 )ψ(t1 )
Time evolution of an individual system is described by repeated appli-
cation of the above algorithm, using its output as the input for each next
step. If we want to study time evolution in a given interval [tin , tf in ], then
we apply the algorithm by starting with t0 = tin , repeating it until we reach
t = tf in somewhere in the middle of the propagation in step 2). Then we
normalize the resulting state.
According to the theory developed in [Davis (1993)] the jump pro-
cess is an inhomogeneous Poisson process with intensity function λα (t) =
(ψ(t), Λα (t)ψ(t)). One way to simulate such a process is to move forward
in time by small time intervals Δt, and make independent decisions for
jumping with probability λα (t)Δt. This leads to the probability p of the
first jump to occur in the time interval (t0 , t) given by
 t
p = 1 − exp(− λα (s)ds). (4.47)
t0
By using the identity
 t
f˙(s)
log f (t) − log f (t0 ) = ds,
t0 f (s)
2 Note that, as can be seen from the equation Eq. (4.47), the norm of ψ(t) is a mono-
tonically decreasing function of t.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 258/4

258 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

.
with f (s) = ||ψα (t)||2 , it is easy to see that
p = 1 − ψα (t)2 , (4.48)
which may sometimes to simplify computer simulations — as we did in the
step 2) above. This observation also throws some new light upon those
approaches to the quantum mechanical description of particle decays that
were based on non-unitary evolution.
By repeating the above event generating algorithm many times, always
starting with the same state at the same initial time t0 , and ending it at
the same final time t, we will arrive at different final states with different
probabilities.
Let α0 , ψα0 , t0 be the initial state, and let μ(α0 , ψα0 , t0 ; α, ψα , t) be the
probability density of arriving at the state (α, ψα ) at time t. We may
associate with this probability distribution a family of density matrices:

ρα (t) = μ(α0 , ψα0 , t0 ; α, ψα , t)|ψα >< ψα |dψα , (4.49)

so that α Tr ρα (t) = 1. This association is many to one. We lose in this
way information. Nevertheless, as shown in [Jadczyk et al. (1996)], the
following theorem holds:

Theorem 4.2. The family ρα (t) satisfies the Liouville linear differential
equation Eq. (4.41). Conversely, the PDP process with values in the pure
states α, ψα described above is the unique one leading to Eq. (4.41).

The Liouville equation Eq. (4.41) describes the time evolution of the sta-
tistical states of the total system. This is the standard, linear, Master
Equation of statistical quantum physics, an equation that describes infi-
nite statistical ensembles, not individual systems. Although the theorem
quoted above tells us that the event generating algorithm follows essentially
uniquely from the Liouville equation, we believe that it is the PDP process
rather than the statistical description that will lead to future generaliza-
tions and extensions of the applicability of quantum theory.3
For instance, in the above formalism it is assumed that the operators gαβ
are linear. But they do not have to be. The operators gαβ represent cou-
plings between the quantum system and a classical ‘detector pointer’, and
jumps represent ‘events’ i.e. changes of the pointer state. The formalism
has been, in particular, applied to the calculation of arrival times [Blanchard
3 Individual description gives us a deeper insight into the real mechanism, and also is

closer to reality, where some experiments can be repeated only a few times, or even only
once, as it is with the Universe between Big Bang and Big Crunch.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 259/4

Foundational Questions 259

and Jadczyk (1966)] and tunneling times for quantum particles tunneling
through a potential barrier [Palao et al. (1997); Muga et al. (2002a)], to the
calculation of relativistic time of arrival [Ruschhaupt (2002a,b)], and also
for studying classical interventions in quantum systems [Peres (2000)].

4.5 Ghirardi-Rimini-Weber spontaneous localization

EEQT may be considered as a logical extension of the theories devel-

oped in the years 1970–1990 known under the names “spontaneous local-
ization”, “dynamical reduction” and, finally, “GRW”, or the “Ghirardi-
Rimini-Weber” model. The reasons for such “improvements” of the stan-
dard quantum theory was clearly stated by John Bell in his paper “Are
there quantum jumps” published in the proceedings of the conference held
from March 31 to April 3, 1987, at Imperial College London to celebrate the
centenary of the birth of Erwin Schrödinger [Kilmister (1987)]. Speaking
about Erwin Schrödinger, John Bell wrote:
At an early stage, he had tried to replace ‘particles’ by
wavepackets (Schrödinger, 1926). But wavepackets diffuse. And
the paper of 1952 ends, rather lamely, with the admission that
Schrödinger does not see how, for the present, to account for par-
ticle tracks in track chambers ... nor, more generally, for the defi-
niteness, the particularity, of the world of experience, as compared
with the indefiniteness, the waviness, of the wavefunction. It is the
problem that he had had (Schrödinger, 1935a) with his cat. He
thought that she could not be both dead and alive. But the wave-
function showed no such commitment, superposing the possibilities.
Either the wavefunction, as given by the Schrödinger equation, is
not everything, or it is not right.
GRW theory was built on the second assumption: that the wave function
given by the Schrödinger equation is not right. In EEQT, on the other hand,
it is clearly realized from the very beginning that the wave function given
by the Schrödinger equation is neither everything nor right. But, before
going into a few of the GRW theory, let us ponder on the next paragraph
written by John Bell in the same paper:
Of these two possibilities, that the wavefunction is not every-
thing, or not right, the first is developed especially in the de Broglie-
Bohm ‘pilot wave’ picture. Absurdly, such theories are known as
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 260/4

260 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

‘hidden variable’ theories. Absurdly, for there it is not in the wave-

function that one finds an image of the visible world, and the results
of experiments, but in the complementary ‘hidden’(!) variables. Of
course the extra variables are not confined to the visible ‘macro-
scopic’ scale. For no sharp definition of such a scale could be made.
The ‘microscopic’ aspect of the complementary variable is indeed
hidden from us. But to admit things not visible to the gross crea-
tures that we are is, in my opinion, to show a decent humility,
and not just a lamentable addiction to metaphysics. In any case,
the most hidden of all variables, in the pilot wave picture, is the
wavefunction, which manifests itself to us only by its influence on
the complementary variables. If, with Schrödinger, we reject extra
variables, then we must allow that his equation is not always right.
I do not know that he contemplated this conclusion, but it seems
to me inescapable.

In fact not only in the pilot wave picture, but in the whole quantum theory
as it is being taught, the wave function itself is “the hidden variable”. It
lives in a world that is not ours, in some infinite dimensional Hilbert space,
or in some abstract “space of quantum states” or “complex probability
amplitudes”.
Ghirardi, Rimini and Weber (as well as almost all other physicists work-
ing in this direction) did not pay any attention to this fact. Later on
Roderich Tumulka [Tumulka (2006)] (cf. also [Allori et al. (2012); Covan
and Tumulka (2013)]) understood that something is rotten in the Kingdom
of Dynamical Reduction Models and proposed, what he has called “the flash
ontology”. Yet he stopped half-way admitting that there must be some-
thing that is “real” — he called it “flashes”, but he did not dare to come
to the logical conclusion that these “flashes” are nothing else but “events”
that can have their own dynamics, dynamics that can be formulated in the
everyday language of “classical physics” and, this way, bring new insights
into the whole structure of theoretical physics and its relation to the world
as we perceive it and try to describe.
Although this will be out of synch with the true historical development,
we will describe the GRW theory as a particular case of EEQT. This has
the advantage that what is hidden within the old GRW or the newer “flash”
ontology, will be made clear and explicit.
Another advantage is that the non-uniqueness of the jump process of the
GRW theory (that was sometimes noted and criticised) is gone — within
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 261/4

Foundational Questions 261

EEQT the algorithm for the piecewise deterministic evolution of the wave
function uniquely follows from the Master Equation for the density matrix
of the total system.
In order to make the derivation of GRW from EEQT as simple as pos-
sible let us restrict ourselves to the case when space is discrete and even
finite. The formulas derived under these assumptions can be easily gener-
alize to the case of a continuous space, though the rigorous mathematics,
here and there, is more complicated in the continuous case. The continuous
case is discussed in more detail in [Jadczyk (1995b)], though even there not
to a satisfactory end. For instance “continuous tensor products” of Hilbert
spaces are mentioned in [Jadczyk (1995b)], but the work in this direction is
still to be done. We will touch this problem again at the end of this section.
Consider a quantum particle and a finite number N of detectors at
positions ri ∈ Rn , i = 1, .., N. Each detector is described by a function
gi (x). It should be noticed at this point that although in our derivation of
the GRW model we will assume that the functions gi are Gaussian, in fact,
that they all have the same shape
gi (r) = g(r − ri ), (4.50)
there is no need for the functions gi to be positive, or even real, and they
do not need to be (square) integrable. For instance, we could choose, in a
compact version of the GRW model (where space is the sphere S 3 instead
of non-compact R3 ), the function g to be a function of Airy type (cf. eg.
[Fernandez and Varadarajan (2008)]). What we need is that squares of the
absolute values of the functions gi form a“partition of unity”, that is:

gi (r)∗ gi (r) = λ, (4.51)
where λ is a positive constant, thus it does not depend on r.
The constant λ will be later on interpreted as the “reduction” rate. In
the GRW model it is assumed that λ is a universal physical constant, and
that its value is about:
λ ≈ 10−16 s−1 . (4.52)
We have now a quantum system (one “quantum particle”), and a classical
system — N detectors. We assume that each detector can be in one of the
two possible states: 0 or 1. We will furthermore assume that originally all
detectors are in the state 0. Then, with time, some of them one by one, can
flip their state from 0 to 1 or from 1 to 0. Such a change will constitute an
“event” — the “detection of the particle”.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 262/4

262 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

The actual state of the classical system (of N detectors) can be therefore
identified with a sequence consisting of N zeros or ones. We have therefore
2N possible states. Let us denote by S the set of all these states:

S = 2N . (4.53)

We can associate a Hilbert space with classical system. It will be

the space of the complex sequences cα , α = 1, .., 2N . Statistical states of
the classical system are then sequences of nonnegative numbers pα , with
2N
α=1 = 1. Pure states of the classical system can be identified with the
sequences pα where only one of the numbers is different from zero (and thus
equal 1).
The “total Hilbert space” of our hybrid system is the tensor product
of Hilbert spaces of the classical and the quantum system, that can be
identified with the direct sum of 2N copies of L2 (RN ). A general state of
the total system is therefore a measure on the classical state space S with
values in density matrices of the quantum system. In other words, it is a
family of positive operators ρα such that


N
2
Tr ρα = 1. (4.54)
α=1

4.5.1 The coupling

Now comes the coupling, the coupling between the classical and the quan-
tum system. We will use the general method of EEQT — to describe the
coupling we need

(1) A Hamiltonian operator Hα for every state α of the classical system.

(2) A coupling operator gαβ for every pair of states of the classical system.

Since at the end we are going to assume that detectors are uniformly dis-
tributed all over the space, we assume that all Hα are identical: Hα = H
for all α ∈ S. The Hamiltonian H will be responsible for the evolution
of the wave packets between detecting events. The events themselves will
consist of flipping the state of one of the detectors from 0 to 1 or from 1
to 0.
There is a small difference here between the GRW type model and the
particle track modeling scheme. If we want to model the particle track (in a
nuclear emulsion or in a cloud chamber), we assume that the detectors can
only change their state from, say, 0 to 1. After such a change the detector
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 263/4

Foundational Questions 263

is not active anymore. In the continuous limit such a change in the GRW
modeling scheme is of no importance, because just one point is of measure
zero in the continuum. But for a finite number of detectors the two schemes
would show a difference.
According to the general scheme of EEQT we need a family gαβ of op-
erators that implement quantum jumps when the classical system changes
its state from β to α. So, we take:

gαβ = ĝ(r − ri ), (4.55)

if α differs from β just in one place — the ith place. Otherwise we set gαβ
to zero. Here ĝ is the multiplication operator:

(gαβ ψ)(r) = g(r − ri )ψ(r). (4.56)

We then have
 
N
∗
Λα = gαβ gαβ = gi2 = λ, α = 1, ..., 2N . (4.57)
β i=1

With these simplifications the GRW algorithm is reproduced by the

EEQT algorithm — if we restrict our attention to the quantum states of
the particle and do not pay attention to the detectors and their states. But
EEQT suggests also a possible way of deriving the whole process from a
coupling between a quantum system and a classically described medium
with a nontrivial internal dynamics, for instance with classically described
geometrical (gravitational) field. The SQUID-tank model (cf. Fig. 4.11(a))
can serve here as an example of such a coupling.
Of course the whole idea behind the GRW model was an explana-
tion of the difference in jump rates between individual quantum particles
and quantum systems consisting of many particles (macroscopic systems).
The EEQT algorithm works also for many particle systems (cf. [Jadczyk
(1995a)]), with the same effect.

Remark 4.2. We have seen that for N detectors we have 2N states of the
classical system. When we go to the continuum limit, we will have to deal
with a classical system of an enormous set of states — which will lead,
mathematically, to the advanced theory of continuous tensor products (cf.
e.g. [Napiorkowski (1971); Arveson (2003)] and references therein), with
possible different phases of the medium and phase transitions. Such effects
may be important for cosmology.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 264/4

264 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

4.6 Heisenberg’s uncertainty principle and quantum fractals

This section contains the discussion of our main hypothesis: simultaneous

measurement of non-commuting quantum observables is not only possible,
but it may lead to experimentally verifiable chaotic behavior, with distin-
guished organization patterns (cf. [Jadczyk (2005)]). The problem is very
serious, and has been mostly treated in a rather superficial way in text-
books. Recently, however, research papers appear, where the subject has
been resurrected and where various authors begin looking afresh both into
the theory and experiments.
From the very beginning quantum mechanics has been formulated in
rather abstract mathematical terms: operators, commutators, eigenvalues,
eigenvectors, etc. For the most part, the accompanying physical interpre-
tations were discovered as surprises rather than due to any deeper under-
standing of what all this new theory was about.
Much of the axiomatization of quantum theory originated in the works
of John von Neumann, culminating in his classic monograph “Mathematical
Foundations of Quantum Theory” [von Neumann (1932)]. But physics is
not always as simple as mathematicians would like it to be. Even if the
criteria of mathematical elegance and simplicity are often useful in sorting
out candidates for possible formal descriptions of reality, Nature herself has
proven to have a sense of elegance that quite often goes deeper than what
we would naively expect.
The unfortunate result of the lack of deeper understanding of the physi-
cal foundations of quantum theory (as exemplified by the famous discussions
between Einstein and Bohr, with Einstein exclaiming: “God does not play
dice”, and Bohr responding: “Einstein, stop telling God what to do”) was
that the theory has been axiomatized, including the concept of “measure-
ment” that has been introduced there as an “evident” concept that does
not need any explanation whatsoever.
In this way for many years only a few brave physicists dared to note
that the emperor has no clothes, and to say it aloud. As we stressed in
Sec. 4.2 John Bell [Bell (1989, 1990)] deplored the misleading use of the
term “measurement” in quantum theory:

“Why did such serious people take so seriously axioms which

now seem so arbitrary? I suspect that they were misled by the
pernicious misuse of the word ‘measurement’ in contemporary
theory.”
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 265/4

Foundational Questions 265

So wrote CERN’s theoretical physicist John Bell in [Bell (2004)]. He opted

for banning this word (measurement) altogether from our quantum vocab-
ulary, together with other vague terms such as ‘macroscopic’, ‘microscopic’,
and ‘observable’. Today he would probably add to his list two other terms
of similarly dubious validity: ‘environment’, and ‘environmentally induced
decoherence’.
John Bell also suggested that we ought to replace the term ‘measure-
ment’ with that of ‘experiment’ [Bell (2004)], and also not to even speak
of ‘observables’ (the things that seem to call for an ‘observer’) but to in-
troduce, instead, the concept of ‘beables’ [Bell (1987a)] — the things that
objectively ‘happen-to-be (or not-to-be)’, independent of whether there is
some ‘observer’, even if only in the future [Wheeler (1984)], or not. In
his scrupulous critical analysis of the quantum measurement problem [Bell
(1990)], “Against Measurement,” John Bell indicates that to make sense of
the usual ‘mumbo jumbo’ one must assume either that
i) in addition to the wave function ψ of a system one must also have
variables describing the classical configuration of the apparatus or
ii) one must abrogate the Schrödinger evolution during measurement, re-
placing it by some sort of collapse dynamics.
The theory of quantum events, EEQT — the ‘Event Enhanced Quan-
tum Theory’, outlined in Sec. 4.4, combines (i) and (ii): there are additional
classical variables, commonly referred to as ‘superselection rules’, and be-
cause of the coupling between these variables and the quantum degrees of
freedom, the evolution is not exactly the unitary Schrödinger evolution,
and it leads to collapses, in particular in measurement-like situations.
It is to be noted that Bell criticized both (i) and (ii), because both as-
cribe a special fundamental role to ‘measurement’, which seems implausible
and makes vagueness unavoidable. EEQT takes his valid criticism into ac-
count. In EEQT we make a distinction between a measurement and an
experiment. Both have a definite meaning within EEQT.
According to the general philosophy of EEQT, our universe, one that
we perceive and are trying to describe and understand, can be considered
as being ‘an experiment’ — performed by Nature herself. This is in total
agreement with Bell; it is also in agreement with the philosophy of John
Wheeler, as outlined in [Wheeler (1984, 1990)].
John A. Wheeler stressed repeatedly [Wheeler (1984)]: “No elementary
quantum phenomenon is a phenomenon until it is a registered (‘observed,’
‘indelibly recorded’) phenomenon.” But, he did not give a definition of
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 266/4

266 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

‘being recorded’ (though he stressed that human ‘observers’ are neither

primary nor even necessary means by which quantum potentials become
‘real’) — and we now understand why: Because such a definition could not
have been given within the orthodox quantum theory. It is given in EEQT.
Historically, physicists arrived at the quantum formalism by a formal
process known as ‘quantization’. Bohr’s quantization, Sommerfeld’s quan-
tization, geometric quantization, deformation quantization ... Today there
is a multitude of formal quantization procedures, each leading to the end
result that classical quantities are being formally replaced by linear opera-
tors that, in general, do not commute. The same components of position
and momentum do not commute. Different components of spin do not com-
mute. In each case the quantum commutation relations involve Planck’s
constant on the right hand side.
It is usually considered that it is not possible to measure simultaneously
several non-commuting observables. One usually quotes in this respect the
celebrated Heisenberg’s uncertainty relations. One must notice that, in his
classic monograph [von Neumann (1932)], John von Neumann was very
careful in this respect, and he stressed explicitly that formal mathematical
relations in no way indicate impossibility of a simultaneous and precise
measurements of, say, position and momentum. He relied completely, in
his account of the ‘physical interpretation’ of uncertainty relations, on the
‘thought experiments’ of Bohr and Heisenberg.
Various textbook authors treat the subject in different ways. A rea-
sonable and modern account of the problem is presented [Grabowski and
Ingarden (1989)], where the authors describe the standard derivation of
Robertson’s inequality Eq. (4.58), and then add the following commentary:

“It follows from the Heisenberg’s uncertainty principle, and

from the Theorem VII.1, that momentum and position are not
commensurable, that is there is no generalized observable A such
that

A(Δ × R1 ) = E Q (Δ),
A(R1 × Δ) = E P (Δ),

for Δ ⊂ B(R1 ). However, that does not mean that quantum me-
chanics excludes the possibility of a simultaneous measurement of
P and Q. In experimental technique we are dealing with a simulta-
neous measurement of the momentum and position. For instance,
we observe a particle in a Wilson chamber. From the observation
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 267/4

Foundational Questions 267

of a particle track we determine its momentum and position. For

a charged particle we deduce its momentum by placing the Wilson
chamber in a magnetic field, and by measuring the curvature of
the track. Even in a situation when we are only measuring the
momentum of the particle, we have some knowledge of its position,
for instance that the particle is within the volume of the measur-
ing apparatus. The point is that in those situations we are not
talking about the simultaneous measurement in the exact sense
(description by spectral measures), but only about an approximate
measurement, with a given uncertainty — such as a measurement
described in example 6, section 12.1. The advantage of the formal-
ism of generalized observables [i.e. using positive operators rather
than idempotents ] is a possibility of a mathematical description of
such a situation.”
In EEQT indeed we are using positive operators and projections, but
that is not important for the very modeling of the simultaneous measure-
ment of non-commuting observables. In EEQT fuzziness leads to chaotic
dynamics resulting from non-commeasurability, and may sometimes result
in self-similarity and fractal patterns as exemplified in the present book.
Masanao Ozawa, in a series of papers [Ozawa (2001, 2002, 2003,
2004b,a)], reviewed the status of theories of state reduction and joint mea-
surement of non-commuting observables. For any pair of observables A and
B we have the following relation [Robertson (1929)]:
1
Δρ AΔρ B ≥ |[A, B]ρ |, (4.58)
2
where · · · ρ stands for the mean value in the given state ρ, Δρ A and
Δρ B are the standard deviations of A and B, defined by Δρ X = (X 2 ρ −
X2ρ )1/2 for X = A, B, and the square bracket stands for the commutator,
i.e., [A, B] = AB − BA. In particular, for two conjugate observables Q and
P , which satisfy the canonical commutation relation
[Q, P ] = i, (4.59)
we obtain Kennard’s inequality [Kennard (1927)]

Δρ QΔρ P ≥ . (4.60)
2
In [Ozawa (2001)] Ozawa concludes that
“(...) the prevailing Heisenberg’s lower bound for the noise-
disturbance product is valid for measurements with independent
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 268/4

268 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

intervention, but can be circumvented by a measurement with de-

pendent intervention. An experimental confirmation of the viola-
tion of Heisenberg’s lower bound is proposed for a measurement of
optical quadrature with currently available techniques in quantum
optics.”

In another paper of this series [Ozawa (2004b)] Ozawa wrote

“Robertson’s and Kennard’s relations are naturally interpreted

as the limitation of state preparations or the limitation of the ideal
independent measurements on identically prepared systems [Bal-
lentine (1970); Peres (1993)]. Moreover, the standard deviation, a
notion dependent on the state of the system but independent of the
apparatus, cannot be identified with the imprecision of the appa-
ratus such as the resolution power of the γ ray microscope. Thus,
it is still missing to correctly describe the unavoidable imprecisions
inherent to joint measurements of non-commuting observables. ”

Although our criticism of the standard treatment of the measurement

process and of the interpretation of the uncertainty relations goes much
deeper, we do agree with the above conclusions.

4.6.1 Simple examples

Physicists have long experience with constructing Hamiltonians Hα de-
scribing the action of external force fields and different known interactions
between particles. But how do we construct the transition operators gαβ
entering the EEQT evolution equation Eq. (4.4.8)?
As has been noted by many authors, any ‘measurement’ can be, in prin-
ciple, reduced to a position measurement. Once we know how to measure
the ‘pointer position’, it is argued, it is enough to set up an interaction be-
tween the apparatus and the system, both considered as quantum systems,
and, when the measurement is ‘done’, read the pointer position.
While we do not think that life is that simple, there is certainly some
truth in the above, and therefore let us start with a simple model of position
measurement. The position variable can be analyzed in terms of yes-no
observations as to whether a given region of space is occupied or not. Thus
our first example will deal with a simple particle detector. Then we will
consider simultaneous monitoring of several non-commuting observables as
it is modeled within EEQT.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 269/4

Foundational Questions 269

4.6.2 A single detector

A detector is a two-state device. It is often assumed that a detector destroys
the particle, but, as a typical track in a cloud chamber shows, this need not
be the case.
While there are several ways of building a model of a detector, let us
concentrate on the simplest one, even if not quite real. We would like to
think of a detector as a two-state device, with two meta-stable states, which
we denote 0 and 1. The detector has the ability to jump from one state to
another when detecting a signal.
For simplicity we will assume zero relaxation time, which means that
after detecting a signal, the detector is instantly ready to detect another
signal.
Heuristically a particle passing close to the detector can trigger its ‘flip’
from 0 to 1, or from 1 to 0. We will consider only the simplest case, when
the detection capability depends only on the particle location, and not on
its energy or other characteristics.4
Let us consider a detector of particle presence at a location a in space (of
n dimensions). Our detector has a certain range of detection and a certain
efficiency. In our simple model, like it was in the GRW case Eq. (4.50), we
encode these detector characteristics in a Gaussian function:
n/2
1 x2
g(x) = κ1/2 √ exp(− ), (4.61)
σ π 2σ 2

where κ is the detector sensitivity constant, σ is a width parameter, and n

stands for the number of space dimensions.
If the detector is moving in space along some trajectory a(t), and if the
detector characteristics κ and σ are constant in time and space, then we
put: gt (x) = g(x − a(t)).
Let us suppose that the detector is in one of its two states at time t = t0 ,
and that initially the particle wave function is ψ0 (x). Then, according to
the algorithm described in Sec. 4.4.9, the probability p of detection in the
infinitesimal time interval (t0 , t0 + Δt) is given by

p ≈ gt20 (x)|ψ0 (x)|2 dxΔt. (4.62)

4 Adding a relaxation time, even with an assigned probability distribution, as well as

modeling detectors with sensitivity dependent not only on particle’s location but also on
its energy, or momentum, or spin, is not a problem within EEQT.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 270/4

270 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

In the sharp detection limit of σ → 0, when gt2 (x) → κ δ(x − a(t)), we get
p ≈ κ|ψ0 (a(t0 ))|2 Δt.
Thus, when Δt << 1/κ, we approximately recover the usual Born in-
terpretation, with the evident and necessary correction that the probability
of detection is proportional to the length of exposure time of the detector.

4.6.3 Measurement of non-commuting observables

EEQT enhances the predictive power of the standard quantum theory, and
it does it in a rather simple way. Once enhanced, it predicts new facts and
solves old mysteries.
The model that we have outlined above has several important advan-
tages. One such advantage is of a practical nature: for example in [Blan-
chard and Jadczyk (1994)] it is shown how to generate pointer readings in
a tank radio-circuit coupled to a SQUID (cf. Fig. 4.11(b)).
In [Jadczyk (1995b,a)] the algorithm generating detection events of an
arbitrary geometrical configuration of particle position detectors, as for in-
stance in a Wilson chamber, has been derived. As a particular case, in a
continuous homogeneous limit we reproduced GRW spontaneous localiza-
tion model (cf. Sec. 4.5, also [Jadczyk (1995a)] and references therein).
Many other examples come from quantum optics, since the Quantum
Monte Carlo model used there is a special case of our approach, namely
when events are not fed-back into the system and thus do not really matter.
Another advantage of EEQT is of a conceptual nature: in EEQT we
need only one postulate: that events can be observed . All the rest can
and should be derived from this postulate. All probabilistic interpretation,
everything that we have learned, or postulated, about eigenvalues, eigen-
vectors, transition probabilities, etc. can be derived from the formalism of
EEQT.
Thus in [Blanchard and Jadczyk (1993)] we have shown that the prob-
ability distribution of the eigenvalues of Hermitian observables can be de-
rived from the simplest, measurement-like, coupling. Moreover, in [Jadczyk
(1995c)] it was shown that EEQT can also give definite predictions for
non-standard measurements, which are of particular interest here, namely
those involving non-commuting observables, and that is so because in our
scheme the contributions gαβ from different, non-commuting, devices add
rather than multiply. In this respect, because the measurement process
is dynamical in our approach, it is like adding non-commuting terms in a
Hamiltonian — nobody has difficulty with adding a position function Aμ (x)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 271/4

Foundational Questions 271

to the momentum pμ in a Hamiltonian. They act simultaneously and they

act together.
Before we discuss the model, and the resulting chaotic behavior and
strange attractor on the quantum state space, let us first discuss in more de-
tail the very question of the simultaneous measurability of non-commuting
observables.
As we have already mentioned, this subject has become quite contro-
versial since the early formulation of Heisenberg’s uncertainty relations.
Mathematically these relations are precise and leave no doubt about their
validity, but the question of how to interpret them physically and philo-
sophically has become a subject of hot discussions.
Various views have been expressed on this subject. Clearly there are dif-
ferent opinions. For instance it is notoriously claimed that within the stan-
dard quantum theory “the simultaneous measurement of non-commuting
observables is impossible”. Such a statement evidently contradict the other
already quoted sentence, from the textbook by Ingarden and Grabowski,
where the authors wrote:

“(...) However, that does not mean that quantum mechanics

excludes the possibility of a simultaneous measurement of P and
Q. In experimental technique we are dealing with a simultaneous
measurement of the momentum and position.”

It seems that even experts sometimes do not distinguish state preparation

procedure from measurements. To quote from Popper’s ‘Unended Quest’
[Popper (1993)]:

“The Heisenberg formula do not refer to measurements; which

implies that the whole current ‘quantum theory of measurement’ is
packed with misinterpretations. Measurements which according to
the usual interpretation of the Heisenberg formula are ‘forbidden’
are according to my results not only allowed, but actually required
for testing these very formula.”

Hilary Putnam came to a similar conclusion [Putnam (1981)]:

“Recently I have observed that it follows from just the quantum

mechanical criterion for measurement itself that the ‘minority view’
is right to at least the following extent: simultaneous measurements
of incompatible observables can be made. That such measurement
cannot have ‘predictive value’ is true ...”
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 272/4

272 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

These words, written more than twenty years ago, suggested that one should
expect chaotic behavior, and that this chaos and its characteristics ought
to be studied, both theoretically and experimentally. Yet, for some reason,
either no one noticed, or no one was interested in looking into the problem
quantitatively.
Of course the main theoretical obstacle was the unsolved quantum-
mechanical measurement problem. A good, critical, discussion of the issues
involved here, can be found in [Landsman (1995)]. In the abstract of his
paper Landsman states:

“We attempt to clarify the main conceptual issues in approaches

to ‘objectification’ or ‘measurement’ in quantum mechanics which
are based on superselection rules. Such approaches venture to de-
rive the emergence of classical ‘reality’ relative to a class of ob-
servers; those believing that the classical world exists intrinsically
and absolutely are advised against reading this paper.”

Even if Landsman is guilty of using the undefined, magical, word “environ-

ment”, as, for instance, in

“The prototype approach (Hepp) where superselection sectors

are assumed in the state space of the apparatus is shown to be
untenable. Instead, one should couple system and apparatus to an
environment, and postulate superselection rules for the latter.”

he does a pretty good job in taking apart different approaches and in ana-
lyzing their weaknesses.

4.6.4 The simplest toy model — space and momentum are

each only two-points
In this example we will describe the simplest possible toy model of a si-
multaneous measurement of several non-commuting observables. The most
celebrated textbook example is, of course, the canonical pair of position
and momentum “observables”.
Although in principle easy, technically it is difficult to simulate measure-
ments of the position and momentum on a computer, because the Hilbert
space is infinite-dimensional. It is also somewhat difficult to discuss the
problem analytically, due to the fact that position and momentum have
continuous spectra. Let us therefore choose here the maximally simplified
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 273/4

Foundational Questions 273

model — technically almost trivial, and yet demonstrating well the main
idea.
The simplest, nontrivial “space” has just two points, we will denote
them ‘-1’ and ‘1’. The “translation group” which operates on these two
points will have two elements: the identity element and the “flip” that
exchanges these two points. We realize this simple “imprimitivity system”5
in a two-dimensional Hilbert space H = C2 .
With Pauli’s matrices s1 , s2 , s3 defined in Eq. (2.76), p. 51 we represent
the ‘position operator’ by s3 , and the ‘momentum operator’ by s1 . Note
that in our case s1 represents the unitary ‘flip’, while (I + s1 )/2 represents
the ‘momentum’.
For symmetry we will use four detectors, two for detecting the position
eigenvalues q = −1 and q = +1, and two for detecting the momentum
eigenvalues p = 0 and p = +1.
Formally, this is a particular case of a more general situation, when we
model a monitoring of several non-commuting spin projections, therefore
let us discuss this more general situation. As with the simple choice above,
with four detectors, we may consider a simple, highly symmetric geometric
pattern, so that the fractal and self-similarity effect are easily recognizable.
Our quantum system will be therefore a single spin 1/2, with no spatial
degrees of freedom.
In order to construct a model within the framework of EEQT we need
to specify the classical system, its states, and the operations implementing
transitions between the states. Thus we will have a family of n detec-
tors, each of them can be excited independently of the others, so that the
probability of two detectors being excited at the same time is zero.
Therefore a state of the classical system will be a sequence of n numbers,
each number being 0 or 1. There are 2n of such states, and a possible change
of state consists of adding 1 mod 2 at i-th place.
For instance, if we have three detectors, a possible transition between
states can be α = (1, 0, 1) −→ β = (0, 0, 1) — a flip of the first detector.
Only one detector can flip at a time.
As for the spin system, we will identify the Hilbert spaces Hα ≡ H ≡
C2 , α = 1, . . . , 2n , corresponding to different states of the classical subsys-
tem. In this way the total algebra A will be represented as a tensor product
A = Aq ⊗ Ac of its quantum and classical parts.
5 An imprimitivity system consists of a spectral measure, and covariantly acting on this

spectral measure, a unitary representation of a group of transformations — cf. [Mackey

(1978); Varadarajan (1985)]
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 274/4

274 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Because the quantum system is a two-state system, the quantum algebra

Aq = L(H) can be identified with the algebra of 2 × 2 complex matrices.
Pure states of the spin system are uniquely represented by points of the
complex projective space P1 (C), which is isomorphic to the sphere S 2 or,
equivalently, by one-dimensional projections of the form 12 (I + n ·s), where
n is a unit vector in R3 — pointing in the direction of the spin.
To be specific, let us consider a simple and symmetric configuration of
detectors, when the measuring apparatus consists of six yes-no polarizers
corresponding to n = 6 spin directions ni , i = 1, ..., n, arranged at the
vertices of a regular octahedron along the directions ni , i = 1, . . . , n:
{{{0, 0, 1}, {1, 0, 0}, {0, 1, 0}, {−1, 0, 0}, {0, −1, 0}, {0, 0, −1}}.
Notice that the six vectors sum up to zero

n
ni = 0. (4.63)
i=1

We may assume that our spin evolves according to the Hamiltonian H =

2 s3 , ω ≥ 0. The coupling between the spin system and the detectors is
ω

specified by choosing six operators ai , which correspond to the six vectors

ni
1
ai = (I +  ni · s), (4.64)
2
where  ∈ (0, 1). These operators correspond to the events in the detectors:
whenever the i-th detector changes its state, and irrespective of the actual
state of other detectors, the quantum state makes a jump implemented by
the operator ai . Thus the ai -s play the role of operators gαβ :
. √
gαβ = κ ai (4.65)
whenever the states α and β differ just at the i-th place, otherwise gαβ = 0.
The coupling constant κ is introduced here for dimensional reasons.
Note that for  = 1 the ai are projection operators. For  < 1 the equation
Eq. (4.64) implies that a projection valued measure corresponding to a
sharp measurement has been replaced by a fuzzy positive operator valued
measure. Because of this, as a result of a jump, not all of the old state is
forgotten. The new states depend, to some degree, on the old state.
Here EEQT differs in an essential way from the naive von Neumann’s
projection postulate of quantum theory. The parameter  becomes impor-
tant. If  = 1 — the case where ai is a projection operator — the new state,
after the jump, is always the same, it does not matter what was the state
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 275/4

Foundational Questions 275

before the jump. There is no memory of the previous state, no ‘learning’

is possible, no ‘lesson’ is taken. This kind of a ‘projection postulate’ was
rightly criticized in the physical literature as being in contradiction to real
world events, contradicting, for instance, the experiments when we take
photographs of elementary particles tracks.
But when  is just close to the value 1, but smaller than 1, the con-
tradiction disappears. This has been demonstrated in the cloud chamber
model [Jadczyk (1995b,a)], where particles leave tracks, in real time, much
like in real life, and that happens because the multiplication operator by
a Gaussian function Eq. (4.61) does not kill the information about the
momentum content of the original wave function. Figure 4.12 shows one
track from the computer simulation of the particle track formation using
the EEQT model. The whole simulation can be found on the YouTube
channel https://www.youtube.com/watch?v=ndIopfN_zzk.

Fig. 4.12 One frame from the computer simulation of particle track formation according
to the EEQT model. Part of the wave function gets reflected from one of the detectors
and starts moving back.

Let us denote by P (n, ) the fuzzy projections

1
P (n, ) = (I +  n · s). (4.66)
2
Notice that P (n, ) have properties similar to those of Gaussian functions,
namely
1 + 2 2
P (n, )2 = P (n, ). (4.67)
2 1 + 2
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 276/4

276 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

We will describe now a sample path of the process. Let us first dis-
cuss the algebraic operation that is associated with each quantum jump.
Suppose before the jump the state of the quantum system is described by
a projection operator P (r), r being a unit vector (spin direction) on the
sphere. That is, suppose, before one the detectors flip, the spin “has” the
direction r. Now, suppose that the detector P (n, ) flips, and the spin right
after the flip has some other direction, r . What is the relation between r
and r ?
We know the answer from, for instance, Eq. (3.120):
(1 − 2 )r + 2(1 + (n · r))n
r = , (4.68)
1 + 2 + 2(n · r)
where (n · r) denotes the scalar product,
n · r = n1 r1 + n2 r2 + n3 r3 . (4.69)
According to EEQT the probabilities pi are computed from the formula
Eq. (4.42) which, in our case, translates to
1 + 2 + 2(ni · r)
pi = . (4.70)
N (1 + 2 )
N N
Notice that, owing to the fact that k=1 n[k] = 0, we have i=1 pi = 1, as
it should be. Assume that at time t = 0 the quantum system is in the state
r(0) ∈ S 2 (we identify here the space of pure states of the quantum system
with a two-dimensional sphere S 2 with radius 1). Under time evolution it
evolves to the state r(t) which is given by the rotation of r(0) with respect
to the z-axis. Then, at time t1 a jump occurs. The time rate of jumps is
governed by a homogeneous Poisson process with rate constant κ. When
jumping r(t) moves to
(1 − 2 )r(t) + 2(1 + r(t) · ni )ni
ri =
1 + 2 + 2r(t) · ni
with probability
1 + 2 + 2r(t) · ni
pi (r(t)) = ,
4(1 + 2 )
and the process starts again.
The resulting fractal pattern in the case of six detectors placed at the
vertices of the regular octahedron, with no external magnetic field, are
shown in Fig. 3.46. If there is an external magnetic field, the state will
rotate between jumps by Poisson distributed random angle. As the result
the fractal pattern will become fuzzy. The resulting fuzziness depends then
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 277/4

Foundational Questions 277

on the rotation speed and on the value of the constant κ that determines
the length of the random intervals between jump. If we set rotation speed
ω = 1, and κ = 100, we obtain a fuzzy version of Fig. 3.46 obtained from
the modified algorithm of Sec. 3.1, namely Fig. 4.13.

Fig. 4.13 Fuzzy (because of Larmor precession) quantum octahedron.

Removing two vertices of the octahedron we get four points that repre-
sent our ‘position-momentum’ simultaneous measurement toy model. Since
the four remaining vertices are in a plane (we choose the plane x = 0), which
intersects the sphere along a great circle, it is clear that the attractor will
be on this circle, and that the fractal pattern will be, in this case, one-
dimensional. In Fig. 4.14 we show the path to the attractor, starting with
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 278/4

278 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Fig. 4.14 Quantum Square. 200,000 jumps on the sphere,  = 0.01.

a randomly chosen initial point (left upper corner). The resolution con-
stant chosen , had to be very small,  = 0.0045, since otherwise the state
reaches the attractor set on the circle in just few steps. With a much higher
resolution ( = 0.7) the Cantor set-like fractal structure on the circle can
be seen. Figure 4.15 shows one million jumps, first the whole picture, and
then ×1000 zoom into the fractal attractor set.

Fig. 4.15 Quantum Square. First 200 steps of the random walk. The square is located
on the plane perpendicular to the viewing plane. One of the four vertices is in the center
in front.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 279/4

Foundational Questions 279

4.7 Are quantum fractals real?

Quantum fractals are generated by a special kind of Iterated Functions Sys-

tem. Thus we have a finite set of transformations (or “functions”), and a
specific rule that assigns a place-dependent probability for selecting each
of the transformations of the system. In quantum fractals the transforma-
tions are always Möbius transformations that act on a sphere (usually the
two-dimensional sphere, but they can as well act on n-dimensional spheres
for any n). Möbius transformations are conformal transformations: they
preserve infinitesimal angles, but not necessarily distances of the Euclidean
geometry. In physics they occur within the theory of relativistic light aber-
ration, but they also occur within the quantum measurement theory —
they implement quantum jumps of spin 1/2 states. Mathematics in both
cases is the same, only the physical interpretation changes. In the first
case we have incoming light directions — they are points on the visible
heavenly sphere. In the second case they are direction of the spin — they
are represented on the invisible sphere of spin 1/2 states — the so called
“Bloch sphere”. Of course no one doubts the reality of light. Therefore
those quantum fractals that are made of light direction have a good chance
to be considered as real — to realize them is just a question of technology
(we need quickly changing directions, close to the light speed).
Are quantum fractals real? First: what is real? In order to answer
the question whether something is real or not we need to define reality.
Probably everybody agrees that tables and chairs are real. First of all
they are heavy — we can feel their weight. Second, if we kick them —
they kick back. We can feel the pain. Real is something that when we
kick it, it kicks back. The term “kickability” was used by the renowned
British philosopher of science Karl Raimund Popper in the postscript to
his famous treatise “The Logic of Scientific Discovery”. Popper was trying
to avoid philosophical battles about the precise definition of reality. He
thus introduced the term “kickability” attributing it to the physicist Alfred
Lande [Popper (1982)]:

However, I do not intend to argue about words, including the

word ‘real’, though by and large I regard as excellent Lande’s sug-
gestion that we call physically real what is ‘kickable’ (and able to
kick back if kicked).

Yet there are various degrees of being real. For instance it is better when
the strength and the frequency of kicking back depends on how we kick.
Otherwise reality can be quite dull.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 280/4

280 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

So, are quantum fractals real? As far as we know, we do not have a direct
access to the reality of quantum waves. We can model them mathematically,
but they seem to be living outside of space and time. Nevertheless we can
try to kick them, and see if they kick back. We can try to kick them one
way or another, and see if the response, if any, changes according to our
predictions. If it does, we can learn how to associate our mathematical
models with experimental arrangements. However, since we are in the
quantum domain, we need statistics.
Quantum fractals are real if we can detect them. But detecting is not
enough. A higher degree of reality can be acquired by objects that we can
produce, manipulate, and examine the results of these manipulations. So,
are quantum fractals real, and if so, to what extent?
Quantum fractals are patterns on the state space. These patterns are
made of pure states. Let us take a simple example: the classical die — see
Fig. 4.16. The die has six faces. The sums of the numbers of dots on the
opposite sides of a standard die are all equal to seven. The quantum spin
1/2 is similar, except that instead of six faces it has infinitely many faces.
It is a sphere rather than a cube! Each point on the sphere represents
one face — the spin direction. One face of the classical die is marked
with a number of dots. Each face of the quantum spin die is marked with
three real numbers nx , ny , nz — the three coordinates of a unit vector n
— the direction of the spin. Its opposite face is marked with the vector
n. A pure state of the standard die is one of the six numbers 1, 2, ..., 6,

Fig. 4.16 Classical dice as a model of a simple quantum system.

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 281/4

Foundational Questions 281

n
−n

n
−n

Fig. 4.17 Quantum spin die. Four pure spin states (“faces”) are drawn here. Face −n
is opposite to n, face −n is opposite to n . For simplicity the circle is drawn instead of
the sphere.

as in Fig. 4.16. A pure state of the quantum spin die is the unit vector
(direction in three dimensional space) n, as in Fig. 4.17. When we shake
the die in our hands and then toss and let it roll across the table, the
result is unpredictable. If the die is not biased, each face can come with
equal probability 1/6 We start with a pure state, say 6 at the top, and
end with a mixed state: each of the six numbers with probability 1/6. In
the case of the classical die this final state is independent of the starting
state. Let us try to make an analogy, this time with the quantum spin
die. Suppose that the initial spin direction n0 is fixed by an appropriate
preparation procedure and that after that there are no forces (thus no
magnetic field) that would force this direction to precess or to change the
initial direction. After that we do not have to “shake” and toss. Just
“looking” at it “shakes” the quantum spin die. Using quantum physics
terminology we “perform spin direction measurement”. Suppose we do not
know the initial spin state. Then a typical elementary spin measurement
answers the following alternative: is the spin direction oriented along a
given direction n? Or is it oriented in the opposite direction −n? According
to the simplest and standard quantum mechanical interpretation, if the
answer was “yes” to the question “is the spin direction oriented along a
given direction n?” right after the measurement the spin is indeed aligned
with n direction. All that assuming that we can measure the spin direction
sharply. In reality there are no sharp measurements. Every measurement in
the laboratory is to some extend smeared up, fuzzy. Therefore the simplest
algorithm needs to be adjusted in order to allow for fuzzy determinations
of the spin direction. To this end we introduce the fuzzy spin direction
operators P (n, k), 0 < k < 1. For k → 1 we obtain the standard sharp
spin measurement operator, while for k → 0 no information about the
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 282/4

282 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

direction of the spin is obtained at all. It is this generalization of sharp

measurements that leads to the phenomenon of quantum fractals. Suppose
the spin is originally directed along the unit vector r, then the probability
that a given (fuzzy) spin n detector is activated is given by the formula:
p(r) = const (1 + k 2 + 2k cos(n, r)). (4.71)
When the detector n is activated, the quantum spin state jumps from r to
r given by:
(1 − k2 )r + 2k(1 + k(n · r))n
r = . (4.72)
1 + k 2 + 2k(n · r)
When k = 0, that is when no information at all about the spin direction
is being obtained, we have r = r, thus no quantum jumps. On the other
hand, when k = 1, the limiting case of the sharp measurement, we get
r = n — the result does not depend on the initial spin state! But in the
middle we have both: place dependent probabilities, and place-dependent
results! With several spin detectors acting at the same time we obtain an
iterated function system, we obtain fractal-like patterns of jump positions.
But is it real? Or is it just a mathematical phantom that will never be
observed in reality?
What follows now is, essentially, science fiction. We do not have yet the
technology that would enable us to perform the experiments we are going
to describe below, even though some preliminary steps in this direction can
be taken today [Wu (2011)]. The first step in this direction consists at
using several spin direction detectors at the same time. Why it is not being
done? Probably because the standard quantum theory is not able to tell
us what to look at in this case. We expect chaos, we would not even call
it a “measurement”. Yet, we think, this will pass and one day in a not so
distant future a new path will open and will result in a whole bunch of new
and interesting data. What we propose is the experimental setup depicted
in Fig. 4.18. The first thing we should notice are eight external detectors,
numbered from 0 to 7. They are symmetrically positioned and perform
fuzzy measurements of spin directions in the x-z plane (the plane of the
picture). Quantum spin states or represented by the points on the circle. All
eight detectors are identically constructed. In my model they are relatively
sharp. Each of them has the same “sharpness coefficient” k = 0.98. The
probability curve depends on the coefficient k as shown in Fig. 4.19. What
should we expect when the eight detectors are simultaneously turned on and
keep registering the counts while the quantum spin system is performing
quantum jumps? Theoretically the jumps result in a fractal-like pattern on
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 283/4

Foundational Questions 283

9
3 1

4 10 8 0

5 7
11

Fig. 4.18 Are quantum fractals real? Proposed experiment: Eight weak detectors acting
as monitors. Four strong detectors interacting with the quantum spin system.

(1+k2 +2k cos(φ))

p(φ) = (2∗(1+k2 )∗π)

Fig. 4.19 Normalized probability function for different values of k.

May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 284/4

284 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

IFS density for t = 0, β = 1.0

log 10 of bin counts

t = 0, β = 0.7
log 10 of bin counts

Angle

Fig. 4.20 Circle fractal measure for eight monitors only.

the circle — the result of the computer simulation for this particular case
are shown in Fig. 4.20. In fact, as we will describe it in the sequel, we will
be playing with some variations of the linear (in cos(φ) = n · r) formula by
deforming it to
pβ (r) = const (1 + k 2 + 2k cos(n, r))β . (4.73)
The resulting fractal measure for unorthodox β = 0.7 is almost indistin-
guishable from the orthodox one, β = 1.0 — see Fig. 4.20. For the moment
we are interested in the case where only the eight monitoring detectors are
turned on. This corresponds to the value t = 0 of the parameter t repre-
senting the relative sensitivity of the main four (inner) detectors and the
monitoring eight (external) detectors. In the next step we will turn the four
inner detectors on and we will let them be t = 1000 times more sensitive
than the monitors. The fact that the fractal measures for β = 1.0 and
β = 0.7 show essentially the same pattern can be understood by comparing
the corresponding probability functions — see Fig. 4.21.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 285/4

Foundational Questions 285

 β
p(φ) = 1 + k2 + 2k cos(φ)
4.0

3.5

3.0

2.5

2.0
p

1.5

1.0 β = 1
β = 0.75
0.5 β = 0.5
β = 0.25
0.0
0 π/4 π/2 3π/4 π 5π/4 3π/2 7π/4 2π
φ

Fig. 4.21 Probability function.

When only the detectors 1–7 are on, the detectors count in a random
way — according to the EEQT algorithm. That is all that we can see
— we can only register detector’s counts, we can’t see the fractal that is
made of jumping quantum spin states. Owing to the symmetry of the
arrangement we expect that, on average, each of the detectors will register
the same number of counts, therefore no useful information can be obtained
by these simple counts. As can be seen in Fig. 4.22 simple counts give slight
variations (within two standard deviations) from the average — as can be
expected from the symmetry of the detectors geometrical configuration.
Instead of just analyzing the counting statistics of each counter separately,
we could also analyze the statistics of consecutive pairs (i, j). We will do it
in the next step, but only after we turn on the four primary detectors, so
that we can compare the results with the primary detectors 8, ..., 11 turned
off and on.
When the four primary spin direction detectors 0, ..., 3 are turned on,
they also start causing the spin state to jump. In the simulation they are
strongly coupled to the spin, so that, on average, they act t = 1000 more
often than the monitoring weakly coupled detectors. The theoretical fractal
pattern of spin states on the circle changes — see Fig. 4.23. We can see
now all four strong peaks. Notice that the vertical axis is in a logarithmic
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 286/4

286 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

t = 0, β = 1.0
2σ

−σ

−2σ

t = 0, β = 0.7
2σ

−σ

−2σ

Fig. 4.22 Monitors counting statistics with the primary system disconnected.

scale. The main question concerning our issue of reality of quantum fractals
becomes: will the monitoring detectors be able to “feel” the action of the
primary detectors? It is the primary detectors that are mainly responsible
for the fractal pattern. Can these strongly coupled detectors be detected
by the system of weakly coupled monitors interacting with the spin sys-
tem? Let us first look at the statistics of pairs of consecutive counts of
the monitoring system — Fig. 4.24. When all twelve detectors are turned
on, then the events registered by monitoring detectors are interspersed by
(on average) 1000 events registered by the primary counters. Neverthe-
less, once in a while, there will be two events registered one immediately
after another by the weakly coupled counters. The analysis of such pairs
is shown in Fig. 4.24. We can see that the very presence of strongly cou-
pled detectors disturbs the original pattern of pairs. That is, even with the
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 287/4

Foundational Questions 287

IFS density for: t = 1000, β = 1.0

log 10 of bin counts

t = 1000, β = 0.7
log 10 of bin counts

Angle

Fig. 4.23 Circle fractal measure for the joint system: four primary and eight secondary
detectors.

standard quantum mechanical probability formula (β = 1.0) we can deduce

that “something” is disturbing the monitoring process. But nothing can be
deduced about the geometry of the four primary detectors.
Why is it so that the orthodox linear quantum mechanics, even when
enhanced by the EEQT algorithms that provide the mechanism of gen-
eration of times series of events for an individual quantum system, even
then so little can be known about what happens to the system from simply
monitoring the coupled classical detectors?
The first observation is that the linear probability law (β = 1.0) of quan-
tum mechanics is such that the effective statistical state of the quantum
system is described by a density matrix. The fractal pattern on the circle
(in general, on the Bloch sphere) of pure states is symmetric. Therefore
the resulting density matrix is located at the center of the circle — it is
trivial, it is proportional to the identity matrix, it carries no information
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 288/4

288 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

t = 0, β = 1.0

σ = 1240.2

t = 1000, β = 1.0
2σ

−σ

−2σ

i×j

Fig. 4.24 Statistics of consecutive pairs for β = 1.0.

about the quantum system at all! But this is too crude an observation. We
may want to understand the details — how does it happen? When both
systems, the primary and the monitoring one, are turned on, the quantum
state starts jumping and most of the time it will be close to the position of
one of the four primary detectors. Therefore, we may expect, that the four
monitoring detectors located at the same positions as the primary ones will
count more often than those positioned in between the primary ones. Yet
this is not what happens, and that is because of the particular linear prob-
ability law. Let μ(x) denote the density of the fractal measure depicted
in Fig. 4.20. Owing to the symmetry of the pattern the moments of this
probability distribution are zero:

x μ(x) = 0,
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 289/4

Foundational Questions 289

that is:
 2π  2π
μ(φ) cos(φ) = μ(φ) sin(φ) = 0. (4.74)
0 0

The probability response functions of different monitoring counters differ

only by the term n · x, where n is the position vector of the counter. There-
fore the differences in the probabilities of responding will be given by the
formula:
 2π
Δpi = cos(φi + φ)μ(φ)dφ (4.75)
0
 2π
= cos(φi ) cos(φ)μ(φ)dφ (4.76)
0
 2π
− sin(φi ) sin(φ)μ(φ)dφ
0
= 0. (4.77)

Thus we should not expect any difference in the counting rates of even
and odd monitoring counters. The change of the distribution of pairs re-
flects only the disturbance being present, but tells us nothing about the
geometrical configuration of the disturbing primary detectors.
But is quantum theory necessarily linear? Must it be so? And here
we are discussing not just the orthodox quantum mechanics as formulated
axiomatically by a mathematician, John von Neumann. Here we go a bit
beyond the standard quantum mechanical formalism assuming that the
EEQ algorithm is more fundamental than the standard statistical descrip-
tion. Why is it so that the probability formula must be linear (or, in a
formulation with state vectors, bi-linear)? Can it be changed? We know
that while this particular formula reproduces the standard Master Equation
for the density matrix, in other areas where iterated function systems are
studied, as for instance in fractal compression of images, a different formula
is used. If we just change the probability formula in the algorithm, the alge-
braic framework will still be preserved: the linear Schrödinger’s equation for
the evolution of wave packets between jumps will be unchanged; quantum
mechanics will still be mostly linear; only the additional quantum mechan-
ical postulate concerning the probability interpretation would change. We
will probably lose the nice “probability current”, but why should we pay
so much attention to a current that does not really flow? So, here, let
us see what happens if we just deform the linear formula by allowing the
exponent β to oscillate below the standard quantum mechanical value of
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 290/4

290 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

 β
p(φ) = 1 + k 2 + 2k cos(φ)

Fig. 4.25 Linear and nonlinear probability sums for even and odd counters.

β = 1.0. The probability curves for β = 1.0 and β = 0.7 are shown in
Fig. 4.25. Running the simulation with β = 0.7 we get the statistics of
the monitoring counters shown in Fig. 4.26 Evidently the counting statis-
tics shows now that the four monitoring counters placed at the positions
of primary detectors register significantly more counts (by seven standard
deviations) than the other four placed at the intermediate positions. We
can also see a somewhat better organization of pairs, but the message from
these statistics is not that obvious — see Fig. 4.27. To sum up: While it is
not clear whether within the standard quantum mechanical formalism the
monitoring detectors can detect the geometric configuration of the primary
detectors (though they can detect the disturbance), it is possible to read
such information if the standard quantum mechanical formula for proba-
bilities is slightly deformed. Such a possibility would, of course, change
our estimates of security codes for quantum cryptography. The decisive
role will be played, of course, by experiments. But experiments in this
direction should be performed without prejudices. There may be surprises
there, similar to those found in nuclear decay rate fluctuations [Parkhomov
(2009)]. Checking quantum probability rules in direct measurements may
not be easy. As stated in [Sinha et al. (2010)]:
A double slit experiment could be used to test Born’s rule, but
then one would have to measure the non-zero double slit interfer-
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 291/4

Foundational Questions 291

ence term and compare it with the theoretical prediction. This

would be sensitive to experimental parameters such as slit dimen-
sions, wavelength of incident photons and distance between detec-
tor and slits; each with its attendant error.

Quantum theory, although created some 100 years ago, only now begins
to bear real fruits. We are entering the age of quantum technological rev-
olution, the results of which we can’t imagine or predict. According to
some experts the Universe is a Quantum Computer [Lloyd (2013b)]. It is
not exactly clear what is meant by this statement since it is accompanied
by another statement: it processes quantum information. The concept of

t = 1000, β = 1.0
2σ

−σ

−2σ

t = 1000, β = 0.7

7σ

2σ
σ
−σ
−2σ

−7σ

Fig. 4.26 Monitors counting statistics with the primary system connected.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 292/4

292 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

t = 0, β = 0.7

σ = 1240.2

t = 1000, β = 0.7

2σ
σ
−σ
−2σ

i×j

Fig. 4.27 Statistics of consecutive pairs for β = 0.7.

information is, also, not very clear in itself. Imagine there are no sentient
beings in the Universe — how shall we define “information”? There are
many different mathematical kinds of information, usually they depend on
the concept of “probability”, but is there such a thing as an “objective
probability”? And if there is, does it belong to the quantum universe or
just to “its interpretation”? These questions are usually not confronted
by the enthusiasts of quantum computing. This is understandable because
quantum computing indeed opens the window to new and extremely pow-
erful technologies, therefore it creates new jobs and new perspectives. It
certainly leads us into a whole new world, beyond our imagination. The
limits for quantum computations, as assessed today [Lloyd (2000)] may eas-
ily prove to be unrealistic, one or another way, since “It seems to be as hard
for classical computers to simulate quantum weirdness as it is for human
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 293/4

Foundational Questions 293

beings to comprehend it.” [Lloyd (2013a)] Very few authors, taken over by
the quantum excitement, pause to think about the limitations of quantum
theory itself. And even if they do, they do so only on odd days of the week,
forgetting it completely on even days.
Jonathan P. Dowling of Quantum Science & Technology Group and
Hearne Institute of Theoretical Physics at Louisiana State University pre-
dicts that some day, in a not so distant future, perhaps, there will arise a
quantum mind [Dowling (2013b)].

The appropriately programmed and sufficiently powerful quan-

tum computer, with the right inputs and outputs, has a mind in
exactly the same sense human beings have minds, but it will have
a mind that, unlike me, also thinks in Hilbert space and therefore
super-exponentially transcends the human mind. ... Suppose the
quantum mind gives us and our classical AI colleagues the quantum
Turing test, and we fail, and it then decides that we are a threat
(or more likely a waste of resources) and kills all of us off.6

While there is certainly such a possibility, the same author, when in a more
relaxed mode, admits that there are some bizarre things about quantum
theory and relativity that cause him to contemplate the idea that [Dowling
(2013a)]

... there is some ur-theory, likely a phenomenological one, which

unifies non-relativistic quantum theory and non-quantum relativ-
ity theory. ... some intermediate unified theory between quantum
gravity and what we have now and that this theory in certain limits
produces non-relativistic quantum theory and non-quantum rela-
tivity theory.

It is clear from the above that quantum theory, as it is conceived today,

is not assumed to be applicable beyond the domains where it has been so
successful thus far. Where are the limits of present day linear, Hilbert space
based quantum algorithms? Are they down in the Planck scale phenomena?
Or are they in the large scale cosmology? Or, perhaps, in the intermediate
scale complexity? Perhaps they are far away from us, or, perhaps, they
6 If quantum AI has so much space in the infinity of dimensions of the Hilbert space,

why would they care at all about matter restricted to just three dimensions of space
and, perhaps, only one dimension of time? Unless it is exactly “time” that they can’t,
in spite of their super-exponential capabilities, fit into their Hilbert space?
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 294/4

294 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

are right under our noses? George Francis Rayner Ellis, Emeritus Distin-
guished Professor of Complex Systems in the Department of Mathematics
and Applied Mathematics at the University of Cape Town in South Africa,
in his paper on the limits of quantum theory recognizes the difficulty and
admits [Ellis (2012)]:
Unitary quantum physics is fundamental in that it applies to
everything at a foundational level in the hierarchy of complexity,
except when state vector reduction takes place in consequence of a
process that is yet to be determined.
So, here we have it: there is still the mysterious state vector reduc-
tion. Louisiana State University physicist and NASA Space Act awardee
Jonathan P. Dowling, quoted above, laughs loudly at Roger Penrose with
his ideas of gravity induced wave collapses [Dowling, 2013b, p. 417]:
What is Penrose’s beef with the classical computer? From read-
ing The Emperor’s New Mind, it is difficult to tell as that book is all
over the map. The reader is introduced to a wildly disparate collec-
tion of topics such as Newtonian mechanics, quantum mechanics,
cosmology, and quantum gravity before Penrose attacks the strong
AI hypothesis in the last couple of chapters. No physicist in his
or her right mind would think quantum gravity has anything to do
with human consciousness.
And yet he comes close, though by a different way, to essentially the same
conclusion as Penrose: gravity may be important for understanding some of
the puzzling features of quantum mechanics. The idea is not new. The first
rough formulation belongs probably to a Hungarian physicist F. Karoly-
hazy who suggested that uncertainties of the gravitational field may cause
reduction of wave functions for sufficiently massive bodies. In a 1966 paper,
[Karolyhazy (1966)], he sketched just such a general idea without even using
explicitly the notion of quantum jumps. Twenty years later, F. Karolyhazy,
A. Frenkel and B. Lukacs [Karolyhazy et al. (1986)] suggested a stochas-
tic influence of the gravitational field, based on the ideas by Ph. Pearle
[Pearle (1984)]. Diosi and Lukacs [Diosi and Lukács (1987); Diosi and
Lukacs (1989)] then suggested the need to create a unified theory of New-
tonian Quantum Mechanics and Gravity, while Holba and Lukacs [Holba
and Lukacs (1991)] pointed out the elusive phenomenon of the anomalous
Brownian motion as a possible result of gravity related stochasticity in
quantum wave packet dynamics.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 295/4

Foundational Questions 295

Yet there are no convincing reasons why it is gravity and not something
else that participates in the processes of transformation of the quantum
potentiality into classical actuality. Usually it is stated that gravity is a
universal interaction, and quantum wave packet reduction processes seem
to be also a universal phenomenon. Yet such thinking may also distract us
from looking for other reasons. For instance, if the experiments described
by A. G. Parkhomov [Parkhomov (2009)] are to be confirmed in different
experimental environments, then it is, perhaps, another long range weak
interaction, not yet known, that can be responsible for the control and
regulation of quantum randomness and stochasticity. If so, then, while the
standard linear quantum predictive schemes may be perfectly justified for
making predictions about averages, they may be inadequate for other goals
involving one-shot (or several-shots) processes.
Essential nonlinearity of the involved processes may be necessary for
their understanding. Nonlinearity in quantum phenomena is usually dis-
missed by invoking its incompatibility with Einstein’s causality [Gisin
(1989); Polchinski (1991); Czachor (1991)]. These arguments, as is usual
with no-go theories, can be circumvented (for instance, as in [Kent (2005)]).
Nonlinearity in quantum theory would have serious consequences for
quantum cryptography, and also for quantum computation [Abrams and
Lloyd (1998)]. Yet nonlinear deformations of quantum mechanics can have
more than just one form. Usually, when discussing nonlinearity, only a
mild form of Weinberg’s type nonlinearity [Weinberg (1989a,b)] is being in-
voked — nonlinear evolution of pure quantum states (nonlinear Schrödinger
equation) or density matrices (nonlinear Liouville equation).
But there exist other types of nonlinearities. For instance we may keep
the Schrödinger equation linear, but admit some nonlinear “observables”.
This is similar to the type of nonlinearity we have introduced in the section
“Are quantum fractals real?”, where the Schrödinger equation is linear,
but the transition probabilities in jump processes are of the form |(x, y)|2β ,
with β not necessarily being equal to one. This is similar to what we see
in Nature when studying 1/f noise: there are natural phenomena of self-
organized criticality with power spectrum of the type 1/f α , with α near one,
but α = 1 [Bak (1996); Jensen (1988); Buchanan (2000); Milotti (2002)].
Whether such a tweaking of the standard probabilistic interpretation of
quantum theory is of any use can be answered only by experiments, but
before making any experiment at least an outline of a theory should be
available.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 296/

296 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Quantum computing is about the coupling of quantum to classical com-

puters. EEQT provides the framework for a mathematical description of
a one-run behavior of such a system. The border between classical and
quantum can be put at a convenient and reasonable place — depending on
the needs. Defining borders is always phenomenological. Every particular
physical theory has its limits of application and these limits are never sharp.
It is the user that decides where to put the limit.
One of the most important conclusions from the EEQT formalism is
that the simultaneous measurement of several non-commuting observables
does not necessarily lead to obtaining a useless set of data. Data at the
output can be chaotic, but they can have an internal structure that car-
ries information. So far the main source of chaos and disorder has been
ascribed to (also phenomenological) environment and decoherence. That
is too rough. Simultaneous measurement of non-commuting observables
may well lead to fractal-like patterns and to self-organization [Nicolis and
Prigogine (1989)]. Nature measures itself, and it does not have human
prejudices concerning the interpretation of Heisenberg’s uncertainty rela-
tions! The future of quantum computing may benefit from such, thus far
neglected, processes.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 297/A

Appendix A

Mathematical Concepts

A.1 Metric spaces

Definition A.6. A metric space (X, d) consists of a set X on which is

defined a distance function which assigns to each pair of points of X a
distance between them, and which satisfies the following four axioms:
(1) d(x, y) ≥ 0 for all points x and y of X;
(2) d(x, y) = d(y, x) for all points x and y of X;
(3) d(x, z) ≤ d(x, y) + d(y, z) for all points x, y and z of X;
(4) d(x, y) = 0 if and only if the points x and y coincide.

The spaces Rn and Cn with the standard Euclidean distance function de-
fined as
1/2

n
d(x, y) = |x − y |
i i 2
(A.1)
i=1

are metric spaces.

Definition A.7. Let (X, d) and (X  , d ) be two metric spaces. A map f :

X → X  is called an isometry if d (f (x), f (y)) = d(x, y) for all x, y, ∈ X.

An isometry is necessarily 1 − 1.

Definition A.8. A sequence {xn ∈ X} is said to be a Cauchy sequence if

for every real  > 0 there is an integer N such that d(xn , xm ) < , provided
that n, m ≥ N .

Definition A.9. A sequence of x1 , x2 , ... in a metric space (X, d) is said

to converge to a point x0 ∈ X if for each  > 0 there is an integer N > 0

297
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 298/A

298 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

such that

d(xn , x0 ) <  whenever n ≥ N

We then write limn→∞ xn = x0 , or xn → x0 . If (X1 , d1 ) and X2 , d2 ) are two

metric spaces, then a map φ : X1 → X2 is said to be continuous if xn → x0
implies f (xn ) → x0 . f is called uniformly continuous if for every  > 0
there exists δ > 0 such that d(x, y) < δ implies d(f (x), f (y)) < .

The metric d itself is a continuous map from X × X to R equipped with its

standard distance metric.

Definition A.10. A subset Y of a metric space (X, d) is said to be closed

if every convergent sequence {xn } of points in Y converges to a point that
is also in Y. If Y is a subset of X, then the smallest closed set containing
Y is called the closure of Y and denoted Y .

Definition A.11. A metric space (X, d) is called complete if every

Cauchy sequence in X converges to some point in X.

The spaces Rn and Cn equipped with their standard distances are complete
metric spaces. Every closed subset of a complete metric space is a complete
metric space. In particular the cube and the sphere in R3 are complete
metric spaces.

Definition A.12. A subset Y of a metric space X is called dense in X

if every point x ∈ X is a limit of a convergent sequence of points from Y.
Equivalently: if X = Y .

Theorem A.3. Given a metric space (X, d) there exists a complete metric
˜ and an isometry φ from X onto a dense subspace φ(X) of X̃.
space (X̃, d)
˜ is called the completion of (X, d).
The space (X̃, d)

There is a canonical construction of X̃ in terms of equivalence classes of

Cauchy’s sequences on X. Usually, we identify X with φ(X) and consider
X as a dense subset of X̃.

Theorem A.4. Every uniformly continuous map f : X → Y form X to a

complete metric space Y extends to a unique uniformly continuous function
on the completion X̃ of X. In particular every uniformly continuous func-
tion f : X → R extends to a unique uniformly continuous function defined
on the completion X̃ of X.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 299/A

Mathematical Concepts 299

Let (X, d) be a metric space, let x be a point of X, and let r be a positive

real number. We define the open ball of center x and radius r to be
the set Br (x) defined by:
Br (x) = {y ∈ X : d(x, y) < r}. (A.2)

Definition A.13. A subset A of a metric space (X, d) is called open if

with each point x ∈ A it contains some open ball Br (x).

Definition A.14. Given a nonempty set A ⊂ X, the -neighborhood A

of A (or -expansion of A) is defined as the union of all open balls of
radius , with centers in A:
"
A = B (x) = {y ∈ X : there exists x ∈ A such that d(x, y) <  }.
x∈A
(A.3)

Definition A.15. A subset A of X is called totally bounded if it can be

covered by a finite collection of open balls Bri (xi ) with centers xi in A.

Definition A.16. Let (X, d) be a metric space, and let φ be a map X → X.

Then
i) φ is called a weak contraction if for all x, y in X, x = y, we have
d(φ(x), φ(y)) < d(x, y).
ii) φ satisfies Lipschitz condition with Lipschitz constant k > 0 if for
all x, y in X,
d(φ(x), φ(y)) < k d(x, y).
iii) φ is a contraction (or a shrinking map) on X if it satisfies the
Lipschitz condition with a Lipschitz constant k that is < 1. That is, if
there exists a positive constant 0k < 1 such that
d(φ(x), φ(y)) < k d(x, y). (A.4)

It is important to observe that whether a given map φ : x → X is a

contraction depends on the choice of metric on X. In applications to iterated
function systems we are usually given the underlying set X and a number of
maps on X, but not a metric. The choice of a metric depends on us. Usually
we choose some kind of a “natural metric”, for instance the Euclidean
distance. But there may be other choices of metrics. Some of them may
suit our purpose better than others. Usually we restrict our choices to
metrics which are “equivalent”.
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 300/A

300 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Definition A.17. We say that two metrics d and d on X are equivalent

on X if a convergent sequence in (X, d) is a convergent in (X, d ), and vice
versa.

Given any metric d on X, the formula

d(x, y)
d (x, y) = (A.5)
1 + d(x, y)
defines another metric that is equivalent to d. The metric d defined in
Eq. (A.5) is bounded. Indeed, it has values in the bounded interval [0, 1].
Example: Consider X = R and the map φ(x) = 12 x. Evidently φ is a
contraction with respect to the Euclidean metric d(x, y) = |x − y|. But it is
|x−y]
not a contraction with respect to the bounded metric d (x, y) = 1+|x−y . In
fact, with respect to a bounded metric no surjective map is a contraction.
This fact is easy to see. Suppose that φ is a contraction from X onto X,
with Lipschitz constant k < 1. Then the inverse map ψ = φ−1 is well
defined, and is an expansion with constant 1/k. Therefore we obtain
d(ψ n (x), ψ n (y)) > (1/k)d(x, y). If X has at least two different points, then
this inequality contradicts the boundness of d.

A.1.1 Compact metric spaces

Let (X, d) be a metric space. X is called compact if X is complete and
totally bounded. A subset K of a metric space X is compact if K endowed
with the metric d of X is compact.

There are other, equivalent definitions of a compact set and compact space,
but this one is probably the one that is most easy to visualize and check in
practical applications. Notice that every closed subset of a compact space
is compact. In Rn a set is compact if and only if it is closed and bounded.

Theorem A.5. Let X, Y be two metric spaces, with X compact. If f :

X → Y is continuous, then the image f (X) of X in Y is compact. If
f : X → R is continuous, X being compacts, then f attains its maximum
and minimum on X.

A.1.2 Locally compact metric spaces

Definition A.18. A metric space (X, d) is said to be locally compact if

every point x of X is contained in open ball Br (x) whose closure is compact.
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 301/A

Mathematical Concepts 301

Closed or open subsets of locally compact spaces are locally compact. Rn

with its natural metric is locally compact.

Theorem A.6. If X, Y are metric spaces and if Y is compact then every

continuous mapping from X to Y is uniformly continuous.

A.2 Normed spaces

Definition A.19. Let V be a vector space over K, where K = R or K = C.

A real-valued function on V
|| · || : x → ||x||,
is called a norm on V if the following conditions hold for all x, y in V and
all scalars α ∈ K
i) ||αx|| = |α| ||x||
ii) ||x + y|| ≤ ||x|| + ||y||
(1) ||x|| ≥ 0, and ||x|| = 0 if and only if x = 0
A vector space with a norm is called a normed space.

If V is a normed vector space, then d defined by

d(x, y) = ||x − y|| (A.6)
is a metric on V.

Theorem A.7. If X, Y are normed spaces and if f : X → Y is a linear

map, them the following assertions are equivalent
i) f is continuous
ii) f is uniformly continuous
iii) f is bounded in the following g sense: ||f || defined as
||f || = sup {||f (x|| : x ∈ X, ||x|| = 1} (A.7)
is finite. The space L(X, Y ) of all continuous linear maps f : X → Y
equipped with the function f → ||f || is a normed vector space.

A.2.1 Banach spaces

Definition A.20. A normed space that, equipped with the metric

Eq. (A.6), is complete is called a Banach space.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 302/A

302 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Every finite-dimensional normed space is a Banach space.

Theorem A.8. Every normed space can be completed to a Banach space.

Every continuous linear mapping between two normed spaces extends, with-
out changing its norm, to a unique continuous linear mapping between their
completions.

A.2.2 The space C(X, Y )

Definition A.21. Let (X, dX ), (Y, dY ) be two metric spaces. We denote

by C(X, Y ) the space of all continuous mappings from X to Y.

Theorem A.9. Let (X, dX ) and (Y, dY ) be two metric spaces, with X being
compact. Then the following formula defines metric on the space C(X, Y ):

d(f, g) = sup{dY (f (x), g(x) : x ∈ X}. (A.8)

If, additionally, Y is complete, then C(X, Y ) is complete. If X is compact

and Y is a normed space then C(X, Y ) is a vector space and

||f || = max{||f (x)|| : x ∈ X} (A.9)

is a norm on C(X, Y ). Additionally, if Y is a Banach space, the C(X, Y )

is a Banach space.

Let V be a normed space. The space C(V, R) of all continuous linear

functionals on V is called the dual of V and is denoted V ∗ . If V is a Banach
space, its dual X ∗ is also a Banach space. V can be considered as being
isometrically embedded in its second dual V ∗∗ according to the definition:

v(f ) = f (v). (A.10)

A.3 Measure and integral

Definition A.22. Let X be a set. A collection Σ of subsets of X is a

σ-algebra if it has the following properties:

(1) ∅ ∈ Σ and X ∈ Σ
(2) If A is in Σ, then the complement Ac = X \ A of A is in Σ.
(3) A countable union of sets from Σ is in Σ.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 303/A

Mathematical Concepts 303

It follows that also a countable intersection of sets in Σ is in Σ. If Σ is a

σ-algebra on X, then the pair (X, Σ) is called a measurable space. If
Σ is understood, then X itself is called a measurable space. The elements
of Σ are called measurable sets. If (X, Σ) and (X  , Σ ) are measurable
spaces and f is a map f : X → X  , then f is called measurable if the
counter-image f −1 (A) of every measurable subset A of X is measurable in
X  . If Σ and Σ are understood, we just say that f is measurable.

The collection 2X of all subsets of X is a σ-algebra. If A is any nonempty

collection of subsets of a set X, then there exists the smallest σ-algebra
containing A. It is denoted s(A).

A.3.1 Borel sets

Definition A.23. Let (X, d) be a metric space. The Borel σ-algebra

B(X) is the smallest σ-algebra in X that contains all open subsets of X.
Elements of B(X) are called Borel sets.

In other words: B(X) is the intersection of all σ-algebras that contain all
open sets. This collection is non-empty since 2X is its member. All open
sets and all closed sets are Borel sets. Every continuous mapping from one
metric space to another is Borel-measurable.

A.3.2 Measure

Definition A.24. A measure on a measurable space (X, Σ) is a function

μ : Σ → [0, ∞] such that
i) μ(∅) = 0
ii) IfAi is a countable collection of pairwise disjoint measurable sets then
∞
" ∞

μ = μ(Ai ). (A.11)
i=1 i=1

If μ(X) < ∞ the measure is called finite If μ(X) = 1, then μ is called a

probability measure. If μ(A) = 0, the A is called set of measure zero.
If some property holds everywhere except on a set of measure zero, we say
that it holds almost everywhere or, in short, a.e.

In Eq. (A.11) it is assumed that for any real number a we have a + ∞ =

∞ and ∞ + ∞ = ∞. The condition in Eq. (A.11) is known as sigma-
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 304/A

304 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

additivity. Every measure is, in particular, finally additive, since we can

always add empty sets to a finite collection of sets. A signed measure
is a real-valued (or complex-valued) (i.e. assuming only finite values) set
function on Σ that is countably additive.

Theorem A.10. If μ is a measure on (X, Σ), then:

i) If A, B ∈ Σ and A ⊂ B then μ(A) ≤ μ(B)

ii) If Ai ∈ Σ and A1 ⊂ A2 ⊂ ... then
∞
"
μ( Ai ) = sup μ(Ai ) = lim μ(Ai ). (A.12)
i i→∞
i=1

Definition A.25. Let (X, Σ) be a measurable space and let x0 be a point

in X. We define δx0 to be the measure on (X, Σ) defined by:

δx0 (A) = 1 if x0 ∈ A, and = 0 otherwise. (A.13)

δx0 is called the Dirac measure concentrated at x0 .

Definition A.26. Let (X, Σ, μ) be a measure spaces. A measurable trans-

formation φ : X → X is called non-singular if, for all A ∈ Σ, μ(A) = 0
implies μ(φ−1 (A)) = 0.

Definition A.27. A measure space (X, Σ, μ) is called s-finite if X is a

countable union of sets with finite measure.

Rn and Cn , endowed with their natural Borel structures and the standard
Lebesgue measure are s-finite.

Definition A.28. Let (X, Σ, μ) be a measure space, and let ν be an arbi-

trary set function on Σ. We write

ν  μ,

if, for every A ∈ Σ with μ(A) = 0, we have ν(A) = 0. If μ and ν are two
measures on Σ, and both ν  μ and μ  ν hold, we say that μ and ν are
equivalent and write μ ∼= ν.

A.3.3 Integral

Definition A.29. A real-valued function f on X is called a simple func-

tion if there is a finite collection of mutually disjoint measurable sets
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 305/A

Mathematical Concepts 305

Ei ⊆ X, i = 1, ..., n and real numbers c1 , ..., cn , such that f has a con-

stant value ci on each Ei and f (x) = 0 outside of the union of Ei . In other
words:
n
f= χEi , (A.14)
i=1

where χA denotes the characteristic function of a set A:

χA (x) = 1 if x ∈ A, = 0 otherwise. (A.15)

Notice that if S : X → X is a transformation of X, then

χA (S(x)) = χS −1 (A) (x). (A.16)
Simple functions are measurable. In fact we have the following theorem:

Theorem A.11. Every extended (that is possibly assuming infinite val-

ues) real-valued measurable function f is a limit of a sequence fk of simple
functions:
For every x ∈ X, lim fk (x) = f (x). (A.17)
k→∞

If f is non-negative, then each fk may be taken non-negative and the se-

quence fk may be assumed increasing. If f is bounded, then the sequence
fk may be made to converge to f uniformly: for every  > 0 there exists
N > 0 such that |fk (x) − f (x)| <  for all x ∈ X, and for all k > N. A real
valued function f
n
Definition A.30. Let (X, Σ, μ) be a measure space. If f = i=1 ci χEi is
a simple function, with Ei measurable and mutually
 disjoint, and if A is a
measurable subset of X, then the integral A f dμ is defined as
 n
f dμ = ci μ(Ei ∩ A). (A.18)
A i=1

If f : X → [0, ∞] is measurable, then we define

 
f dμ = sup gdμ, (A.19)
A A
where the supremum is taken over all simple nonnegative functions g such
that g ≤ f. Sometimes we use equivalent variants of this notation the
notation:
  
f dμ = f (x)μ(dx) = f (x)dμ(x). (A.20)
A A A
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 306/A

306 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

A.3.4 Lp spaces

Definition A.31. Let (X, Σ, μ) be a measure space. We define L1 (μ) to

be the set of all complex measurable functions f on x for which

|f |dμ < ∞. (A.21)
X
1

If f is in L (μ), the integral A f dμ is defined by first splitting f into real
and imaginary parts, then splitting each part into positive and negative part
and using the formula for the integral of measurable nonnegative functions.
More generally, for any real number 1 ≤ p < ∞ and for any complex
measurable function f on X we define
 1/p
||f ||p = |f |p dμ , (A.22)
X

and the space L (μ) as consisting of all functions f for which ||f ||p is finite.
p

Theorem A.12. For every 1 ≤ p < ∞, Lp (μ), is a vector space. ||f ||p is
norm on Lp (μ). Lp (μ) equipped with its norm is a Banach space.

Definition A.32. A measurable function f is called essentially bounded

if f is bounded outside of some set of measure zero, that is if there exists
a positive, finite constant a such that the set {x : |f (x)| < a} is of measure
zero. The infimum of the set of all such a is called the essential supremum
of |f | and denoted ess sup |f |. The set of all essentially bounded functions
on (X, Σ, μ) is denoted L∞ (μ).

Theorem A.13. The function || · ||∞ : L∞ (μ) → R defined by

||f ||∞ = ess sup |f | (A.23)
is a norm on L∞ (μ). L∞ (μ) equipped with this essential norm is a Banach
space.

Definition A.33. If p and q satisfy 1 ≤ p, q ≤ ∞ and

1 1
+ = 1, (A.24)
p q
we say that p and q are conjugate or dual. It is understood that 1/∞ = 0
and 1/0 = ∞.

Notice being dual is a symmetric relation. p = 1/2 is dual to itself, p = ∞

is dual to q = 0.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 307/A

Mathematical Concepts 307

Theorem A.14. Hölder inequality Assume 1 < p < ∞ and q < q < ∞
are dual to each other.If f ∈ Lp (μ), g ∈ Lq (μ) then f g ∈ L1 (μ) and
||f g||1 ≤ ||f ||p ||g||q . (A.25)
p q ∞
The spaces L (μ) and L (μ) are dual to each ether. The space L (μ) is
dual to L1 (μ).

However the dual to L∞ (μ) is, in general, larger than L1 (μ).

Definition A.34. Let (X, d) be a compact metric space. A linear func-

tional α : C(X, R) → R is called positive if for every f ∈ C(X, BR) f ≥ 0
implies α(f ) ≥ 0.

Theorem A.15. Every positive linear functional α on C(X) is bounded.

In fact, we have
|α(f )| ≥ α(1) ||f ||∞ . (A.26)

Definition A.35. Let (X, d) be a compact metric space. A continuous

positive linear functional μ on C(x) is called a Radon measure. For
f ∈ C(X) we use the notation:

μ(f ) = f dμ. (A.27)
X

Usually integration is defined in a different way, via Borel measures. On

compact metric spaces the definition of the integral can be simplified as
above. The two definitions are equivalent.

Definition A.36. Let (X, d) be a metric space. The support supp (f )

of a function f : X → R is the closure of the set on which the function is
nonzero:
supp (f ) = {x ∈ X : ¯F (x) = 0}. (A.28)

Theorem A.16. Let μ be a Radon measure on a compact metric space X.

Then the function U → μ(U ) defined on open subsets of X by the formula
μ(A) = sup{α(f ) : 0 ≤ f ≤ 1, f ∈ C(X), supp (f ) ⊆ U, (A.29)
has a unique extension to a Borel measure on X.

Definition A.37. Let μ be  a Radon measure on a compact metric space

X. Then the function f → |f |dμ from C(X) to R is a norm on C(X). We
denote this norm as ||f ||1 :

||f ||1 = |f |dμ, f ∈ C(X). (A.30)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 308/A

308 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

The completion of the space C(X) with respect to ||·||1 is denoted L1 (X, μ).
When μ is understood, we write simply L1 (X).

Theorem A.17. The function f → f dμ extends to a unique continuous
function on L1 (X). It is written the same way, as f → f dμ.

Theorem A.18. Radon-Nikodym theorem Let (X, Σ, μ) be a sigma-finite

measure space, and let ν be a real (or complex) finite measure that is ab-
solutely continuous with respect to μ. Then there exists a unique function
f ∈ L1 (μ) such that

For all A ∈ Σ ν(A) = f dμ. (A.31)
A

Definition A.38. The function f defined in Eq. (A.31) is called the

Radon-Nikodym derivative of ν with respect to μ and denoted as ν/dμ.

Theorem A.19. If ν and μ are s-finite measures such that ν  μ, and if

f is a finite-valued measurable function for which X f dν is defined, then

dν
intX f dν = f dμ. (A.32)
X dμ

A.4 Markov, Frobenius-Perron and Koopman operators

Definition A.39. Let (X, Σ, μ) be a measure space. A linear operator P :

L1 (μ) → L1 (μ) is called a Markov operator if for every f ∈ L1 (μ), f ≥ 0

i) P f ≥ 0
i) ||P f ||1 = ||f ||1 .

The inequalities are understood in the a.e. sense.

Definition A.40. A function f ∈ L1 (μ) is called a density if f ≥ 0 and

||f ||1 = 1. If f is a density, then μf defined by the formula

μf (A) = f dμ, A ∈ Σ, (A.33)
A

is a probabilistic measure. We say that f is a density or Radon-Nikodym

derivative of μf with respect to μ. If f is a density and P f = f, then f is
said to be stationary with respect to P.
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 309/A

Mathematical Concepts 309

Every Markov operator transforms densities into densities.

Theorem A.20. If P is a Markov operator, then P is a continuous linear

operator on the Banach space L1 (μ) with norm P ≤ 1.

Definition A.41. Let (X, Σ) be a measurable space, and let S : X → X

be a measurable transformation. The operator P defined on measures on
(XΣ) by the formula
(P μ)(A) = μ(S −1 (A)), A∈Σ (A.34)
transforms finite measures into finite measures and probabilistic measures
into probabilistic measures. It is called Frobenius-Perron operator as-
sociated to S.

Frobenius-Perron operator is a Markov operator. If x0 is a point in X and

δx0 is the Dirac measure concentrated at x0 , then
P n δx0 = δS n (x0 ) .
In other words the knowledge of P allows us to recover the whole trajectory
of iterates of S acing on an arbitrary point x0 ∈ X.
When not only a measurable space is given, but also a measure μ on X is
fixed, and if the transformation S is non-singular, then we can define the
Frobenius-Perron operator acting on densities.

Definition A.42. Let (X, Σ, μ) be a measure space and let S be a measur-

able and non-singular transformation S :→ X. The operator P : L1 (μ) →
L1 (μ) defined by the formula
  
P f dμ = f dμ = χA (S(X)f (x)dμ(x), (A.35)
A S −1 (A) X

for all A ∈ Σ and all f ∈ L1 (μ), is called the Frobenius-Perron operator

on densities associated to S..

Theorem A.21. [Lasota and Mackey, 1985, p. 42] Let (X, Σ, μ) be a mea-
sure space and let S : X → X be a measurable invertible non-singular
transformation. Let P be the Frobenius-Perron operator on densities asso-
ciated with S. Denote by J −1 the Radon-Nikodym derivative
dμ(S −1 (x))
J −1 (x) = . (A.36)
dμ(x)
Then for every integrable essentially bounded function f ∈ L1 (μ) ∩ L∞ (μ)
we have
(P f )(x) = f (S −1 (x))J −1 (x). (A.37)
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 310/A

310 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Definition A.43. Let (X, Σ, μ) be a measure space, let S be a measurable

non-singular transformation S :→ X. The Koopman operator associ-
ated to S is the operator P ∗ on L∞ (μ) defined as
(P ∗ f )(x) = f (S(x)), f ∈ L∞ (μ). (A.38)

Theorem A.22. The Koopman operator P ∗ is the adjoint of the Frobenius-

Perron operator P : For every f ∈ L1 (μ), g ∈ L∞ (μ) we have
(P f, g) = (f, P ∗ g), (A.39)
that is,
 
(P f )(x)g(x)dμ(x) = f (x)(P ∗ g)(x)dμ(x). (A.40)
X X
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 311/B

Appendix B

Minkowski Space Generalization of

Euler-Rodrigues Formula

Let F be a generator of a one-parameter subgroup of the Lorentz group.

We write F in the following general form that resembles the form of the
electromagnetic field mixed tensor expressed F μ ν in terms of the electric
and magnetic field vectors e and b:
⎛ ⎞
0 b3 −b2 e1
−b3 0
F =⎝ ⎠.
b1 e2
b2 −b1 0 e3
(B.1)
e1 e2 e3 0

The “dual” matrix, denoted F̃ is obtained from F by “dual rotation”, that

is by replacing e → b, b → −e:
⎛ ⎞
0 −e3 e2 b1
0 −e1
F̃ = ⎝ −e3 ⎠
e b2
2 e1 0 b3
(B.2)
b1 b2 b3 0

We also introduce real numbers u and v defined as

1 1 1
u = Tr F F̃ = e · b, v = Tr F 2 = (e2 − b2 ). (B.3)
4 4 2
The characteristic polynomials for F and F̃ can now be expressed in terms
of u and v

det(F − λI) = λ4 − 2vλ2 − u2 , det(F̃ − λI) = λ4 + 2vλ2 − u2 . (B.4)

Let σ and θ be defined as

 
σ= u2 + v 2 + v, θ = sgn(u) u2 + v 2 − v. (B.5)

It is clear that σ is nonnegative and θ has the sign of u, where the sgn is
defined to be right continuous, that is sgn(0) = 1. The eigenvalues of F
are the zeros of the characteristic polynomial, they are ±σ and ±iθ. The

311
May 29, 2014 10:6 World Scientific Book - 9in x 6in QuantumFractals3 pg. 312/B

312 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

eigenvalues of F̃ are ±θ and ±iσ. The following identities follow directly

from the definitions:
σ2 − θ2
v= , u = σθ. (B.6)
2
Let us define T as follows:
F 2 + F̃ 2 = T. (B.7)
Then the following lemma holds:

Lemma B.1. The matrices F, F̃ , T commute. Moreover, the following

identities hold:
F F̃ = F̃ F = σθ I (B.8)
F 2 − F̃ 2 = (σ 2 − θ2 ) I (B.9)
F 3 = (σ 2 − θ2 )F + σθF̃ , (B.10)
T + (σ − θ )I
2 2
F2 = , (B.11)
2
T − (σ 2 − θ2 )I
F̃ 2 = , (B.12)
2
F T = (σ 2 − θ2 )F + 2σθF̃ , (B.13)
F̃ T = 2σθF − (σ 2 − θ2 )F̃ , (B.14)
2 2 2 2
T = (σ + θ ) I. (B.15)

Proof. The proof of equations (B.8-B.10) follows by the direct verifica-

tion, with matrices, using the fact that σθ = e · b and σ2 − θ2 = e2 − b2 .
Eq. (B.11) (resp. (B.12)) follows by adding (resp. subtracting) Eq. (B.7)
and Eq. (B.9). Eq. (B.13) results in a smilar way. In order to show
Eq. (B.14) we first multiply Eq. (B.9) by F̃ and use Eq. (B.8). Finally,
Eq. (B.15) can be derived from Eq. (B.13) multiplied by F , and using
(B.8) and (B.9). 

Proposition B.1 (Generalized Euler-Rodrigues formula). Assume

that σ2 + θ2 > 0. Then the following general formula holds:
cosh(tσ) + cos(tθ) σ sinh(tσ) + θ sin(tθ)
exp(F t) = I+ F
2 σ2 + θ2
θ sinh(tσ) − σ sin(tθ) cosh(tσ) − cos(tθ)
+ F̃ + T. (B.16)
σ2 + θ2 2(σ 2 + θ2 )
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 313/B

Minkowski Space Generalization of Euler-Rodrigues Formula 313

Equivalently, using F 2 instead of T

θ2 cosh(tσ) + σ 2 cos(tθ) σ sinh(tσ) + θ sin(tθ)
exp(F t) = I+ F
σ 2 + θ2 σ2 + θ2
θ sinh(tσ) − σ sin(tθ) cosh(tσ) − cos(tθ) 2
+ 2 2
F̃ + F . (B.17)
σ +θ σ2 + θ2
If σ = θ = 0, then
t2 t2
exp(tF ) = I + tF + T = I + tF + F 2 . (B.18)
4 2

Proof. In the proof we use the following theorem about generators of

one–parameter matrix groups: If γ(t) is a one-parameter group of matri-
ces, then γ(t) = exp(Xt), where X = γ  (0). 7 We consider first the case of
at least one of the numbers σ, θ being nonzero, so that σ2 + θ2 > 0.
Let L(t) denote the right hand side of the formula Eq. (B.16). It is immedi-
ate that L(0) = I, L (0) = F. Our aim to show that L(t) is a one-parameter
matrix group, that is that
L(t + s) = L(t)L(s). (B.19)
The proof is somewhat tedious but straightforward. We write L(t + s) and
expand the functions sin(x + y), cos(x + y), sinh(x + y), cosh(x + y) in terms
of products of functions of the corresponding arguments x, y. This way we
get a long expression with coefficients in front of I, F, F̃ , T.
On the other hand we multiply L(t)L(s) and obtain coefficients in
front of the products of I, F, F̃ , T. All of these products can be reduced to
I, F, F̃ , T using Lemma B.1. Comparing the coefficients in front I, F, F̃ , T
establishes the result.
Suppose now that σ = θ = 0. Then, from Lemma B.1, we have that
F T = T 2 = 0, F 2 = T /2 (B.20)
The group property of L(t) given by Eq. (B.19) follows then by the following
observation:
t2 s2
L(t)L(s) = (I + tF + T )(I + sF + T )
4 4
s2 ts t2 1
= I + sF + T + tF + T + T = I + (s + t)F + (s + t)2 T.
4 2 4 4
7 The proof of this classical theorem can be found, for instance, in ‘An introduction to

matrix groups and their applications‘ by Andrew Baker, Springer 2002, Theorem 2.17,
also available online, the same title and author, Theorem 2.5: http://www.maths.gla.
ac.uk/~ajb/dvi-ps/lie-bern.pdf.
June 12, 2014 16:56 World Scientific Book - 9in x 6in QuantumFractals3 pg. 314/B

314 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

On the other hand L(0) = I, L (0) = F, which completes the proof.

Eq. (B.17) follows from Eq. (B.16) and Eq. (B.11). 

Remark B.1. Since we are dealing with commuting matrices, the prob-
lem reduces to a simple commutative symbolic algebra. Using computer
software able to do commutative symbolic operations can therefore save
time.

B.1 Alternative derivation via SL(2, C)

With the four hermitian matrices8 σμ = σ μ ,

 0 i 0 0 
σ1 = ( 01 10 ) , σ2 = −i 10
0 , σ3 = 0 −1 , σ4 = ( 0 1 ) , (B.21)
the group homomorphism A → Λ(A) from the group of unimodular matri-
ces SL(2, C) onto the connected component of the identity SO(3, 1)0 of the
homogenous Lorentz group is given by
1
Tr (Aσ μ A† σν ), (μ, ν = 1, ..., 4).
Λ(A)μ ν = (B.22)
2
The completeness relations for σ matrices

4
σ μAB σμCD = 2δD
A B
δC , (A, B, C, D = 1, 2) (B.23)
μ=1

entail
Tr (Λ(A)) = |Tr (A)|2 . (B.24)
Taking the derivative of Eq. (B.22) we arrive at the linear relation (isomor-
phism) between infinitesimal generators f (traceless 2×2 complex matrices)
from the Lie algebra SL(2, C) to the Lie algebra elements F in SO(3, 1):
1
F μν = Tr (f σ μ σν + σ μ f † σν ). (B.25)
2
With f defined by
1
3
def
f = (ei + ibi)σi (B.26)
2
i=i
def
we arrive at F given by (B.1), while f˜ = −if gives F̃ . The characteristic
polynomial for f is: det(f − λ I) = λ2 − 12 (v + iu), with two roots ±ω.
8 By abuse of notation σ μ constitute exactly the same set matrices. Their
μ and σ
components are σμAB and σμAB , (μ = 1, ..., 4), (A, B = 1, 2).
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 315/B

Minkowski Space Generalization of Euler-Rodrigues Formula 315

There is a simple relation between ω and σ, θ: ω = 12 (σ + iθ). Every 2 × 2

complex matrix X determines a vector in the complex Minkowski space
with complex coordinates xμ = Tr (σ μ X)/2. There are two scalar products
in this space: (x, y) = xT Jy and {x, y} = x† Jy. The first one is bilinear,
while the second one is hermitian. Both are SO(3, 1) invariant. X is
†
hermitian if and only if xμ are real, moreover  Tr (X
 Y )/2 = {x, y} and
0 1
Tr (X T X)/2 = det(X) = (x, y), where  = −1 0 . If ξ± are eigenvectors
†
of f belonging to eigenvalues ±ω = 0, and if X± = ξ± ⊗ ξ± , then x± are
real isotropic (i.e. (x, y) = {x, y} = 0) eigenvectors of F corresponding to
† †
real eigenvalues ±2(ω). Vectors y± corresponding to ξ+ ⊗ ξ− and ξ− ⊗ ξ+
†
are hermitian space-like (we have {y, y} = 2(||ξ+ ||2 ||ξ− ||2 − |ξ+ ξ− |2 ) >
0), bilinear isotropic (i.e. (y± , y± ) = 0), and J-orthogonal to x± , resp.
They are eigenvectors of F corresponding to imaginary eigenvalues ±2(ω).
Since f 2 = (v + iu)I/2 = ω 2 , exp(tf ) is easily computed
sinh ωt
etf = cosh(ωt) I + f, (B.27)
ω
where it does not matter which of the two possible signs of ω is chosen. If
ω = 0, then f has just one eigenvector ξ, vector x, corresponding to ξ is
real isotropic, and F annihilates 2-dimensional plane in the 3-dimensional
hyperplane orthogonal to x. Moreover, when ω = 0, which happens if and
only if f 2 = 0, we get instantly
etf = I + tf, (B.28)
which can be also obtained by taking the limit of ω → 0 in Eq. (B.27). We
can now expand the functions cos(tω), sinh(tω) of the complex argument
tω = t(σ + iθ) and use Eq. (B.22) to recover, by straightforward though
somewhat lengthy calculations, the results of Proposition B.1.
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 317/4

Bibliography

Abrams, D. and Lloyd, S. (1998). Nonlinear quantum mechanics implies

polynomial-time solution for np-complete and #p problems, Physical Re-
view letters 81, 18, pp. 3992–3995.
Ahlfors, L. V. (1979). Complex Analysis (McGraw-Hill, Inc.).
Alicki, R. (2002). Invitation to quantum dynamical semigroups, in
P. Grabaczewski and R. Olkiewicz (eds.), Dynamics of Dissipation, Vol.
LNP 597 (Springer), pp. 239–264.
Alicki, R. and Lendi, K. (1987). Quantum Dynamical Semigroups and Applica-
tions (Springer).
Allori, V., Goldstein, S., Tumulka, R. and Zanghi, N. (2012). Predictions and
primitive ontology in quantum foundations: A study of examples, URL
http://arxiv.org/abs/1206.0019.
Anglès, P. (2007). Conformal Groups in Geometry and Spin Structures
(Birkhauser, Progress in Mathematical Physics, Vol. 50).
Arveson, W. (2003). Noncommutative Dynamics and E-Semigroups (Springer).
Arveson, W. B. (1969). Subalgebras of C  –algebras, Acta Math. 123, pp. 141–224.
Aston, P. J. (2012). Is radioactive decay really exponential? Europhysics Letters
97, 5, p. 52001.
Avron, J. E. and Simon, B. (1981). Almost periodic Hill’s equation and the rings
of Saturn, Physical Review Letters 46, 7, pp. 1166–1168.
Axler, S. (1997). Linear Algebra Done Right (Springer).
Bak, P. (1996). How Nature Works: The science of self-organized criticality
(Springer).
Ball, P. (2013a). Experts still split about what quantum theory means. Poll reveals
diverse views about foundational questions in physics. Nature News, URL
http://www.nature.com/news/experts-still-split-about-what-
quantum-theory-means-1.12198.
Ball, P. (2013b). Will we ever understand quantum theory? BBC Future, URL
http://www.bbc.com/future/story/20130124-will-we-ever-get-
quantum-theory.
Ballentine, L. E. (1970). The statistical interpretation of quantum mechanics,
Rev. Mod. Phys. 42, pp. 358–381.

317
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 318/4

318 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Balzer, C., Hannemann, T., Reiß, D., Wunderlich, C., Neuhauser, W. and
Toschek, P. E. (2002). A relaxationless demonstration of the quantum zeno
paradox on an individual atom, Optics Communications 211, pp. 235–241.
Barning, F. J. M. (1963). Over Pythagorese en bijna-Pythagorese driehoeken en
een generatieproces met behulp van unimodulaire matrices, Master’s thesis,
Stischting Mathematisch Centrum Amsterdam, URL http://oai.cwi.nl/
oai/asset/7151/7151A.pdf.
Barnsley, M. F. (1988). Fractals Everywhere (Academic Press).
Barnsley, M. F., Demko, S. G., Elton, J. H. and Geronimo, J. S. (1988). Invariant
measures for Markov processes arising from iterated function systems with
place-dependent probabilities, Ann. Inst. H. Poincaré (B) 24, 3, pp. 367–
394.
Barnsley, M. F. and Vince, A. (2011). The chaos game on a general iterated
function system, Ergodic Theory and Dynamical Systems 31, 04, pp. 1073–
1079.
Bell, J. (1980). de Broglie-Bohm, delayed-choice double-slit experiment, and den-
sity matrix, Int. J. Quant. Chem. 14, pp. 155–159.
Bell, J. (1982). On the impossible pilot wave, Found. Phys. 12, pp. 989–999.
Bell, J. (1987a). Beables for quantum theory, in Speakable and Unspeakable in
Quantum Mechanics (Cambridge University Press), pp. 173–180.
Bell, J. (1987b). Six possible worlds of quantum mechanics, in Speakable and Un-
speakable in Quantum Mechanics (Cambridge University Press), pp. 181–
195.
Bell, J. (1989). Towards an exact quantum mechanics, in Themes in Contempo-
rary Physics II. Essays in honor of Julian Schwinger’s 70th birthday (World
Scientific), pp. 1–26.
Bell, J. (1990). Against measurement, Physics World August, pp. 33–40.
Bell, J. S. (2004). Speakable and Unspeakable in Quantum Mechanics, 2nd edn.
(Cambridge University Press).
Blanchard, P. and Jadczyk, A. (1966). Time of events in quantum theory, Helv.
Phys. Acta 69, pp. 613–635.
Blanchard, P. and Jadczyk, A. (1993). On the interaction between classical and
quantum systems, Phys. Lett. A 175, pp. 157–164.
Blanchard, P. and Jadczyk, A. (1994). How and when quantum phenomena be-
come real, in Z. Haba et al. (eds.), Stochasticity and Quantum Chaos, Vol.
Proc. Third Max Born Symp., Stochasticity and Quantum Chaos (Kluver),
pp. 13–31.
Blanchard, P. and Jadczyk, A. (1997). Time of arrival in event enhanced quantum
theory, in H. D. Doebner, P. Nattermann, and W. Scherer (eds.), GROUP
21: Proceedings of the XXI International Colloquium on Group Theoretical
Methods in Physics, Vol. 1 (World Scientific), pp. 314–320.
Blanchard, P., Jadczyk, A. and Olkiewicz, R. (2001). Completely mixing quantum
open systems and quantum fractals, Physica D: Nonlinear Phenomena 148,
3–4, pp. 227–241.
Blanchard, P. and Olkiewicz, R. (1999a). Interacting quantum and classical con-
tinuous systems I. The piecewise deterministic dynamics, J. Stat. Phys. 94,
p. 913.
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 319/4

Bibliography 319

Blanchard, P. and Olkiewicz, R. (1999b). Interacting quantum and classical con-

tinuous systems II. Asymptotic behavior of the quantum system, J. Stat.
Phys. 94, p. 933.
Blank, J., Exner, P. and Havlı́ček, M. (2008). Hilbert Space Operators in Quantum
Physics, 2nd edn. (Springer, New York).
Bohm, D. (1952). A suggested interpretation of the quantum theory in terms of
‘hidden’ variables I, Phys. Rev. 85, 2, pp. 166–179.
Bohm, D. and Hiley, B. (1993). The Undivided Universe (Routledge).
Boyarsky, A. and Gora, P. (1997). Laws of Chaos (Birkhauser).
Bruhn, G. W. (2002). Does radioactivity correlate with the annual orbit of earth
around sun? Apeiron 9, 2, pp. 28–40.
Buchanan, M. (2000). Ubiquity (Phoenix).
Calude, C. S. (2007). Randomness and Complexity (World Scientific).
Carmeli, M. (1977). Group Theory and Special Relativity (McGraw-Hill).
Carmeli, M. and Malin, S. (2000). Theory of Spinors: An introduction. (World
Scientific).
Christensen, E. and Evans, D. (1978). Cohomology of operator algebras and quan-
tum dynamical semigroups, London. Math. Soc. 20, pp. 358–368.
Covan, C. W. and Tumulka, R. (2013). Epistemology of wave function collapse
in quantum physics, URL http://arxiv.org/abs/1307.0827.
Cowen, R. (2013). Proof mooted for quantum uncertainty, Nature 498, 7455,
pp. 419–420, doi:10.1038/498419a, URL http://dx.doi.org/10.1038/
498419a.
Cross, D. J. (2008). On the relation between real and complex Jacobian
determinants, URL http://www.haverford.edu/physics-astro/dcross/
academics/papers/jacobian.pdf.
Czachor, M. (1991). Mobility and non-separability, Foundations of Physics Letters
4, 4, pp. 351–361.
Dauben, J. W. (1990). Georg Cantor: His mathematics and philosophy of the inifi-
nite (Princeton University Press), ISBN 0691024472, URL http://press.
princeton.edu/titles/4740.html.
Davis, M. H. (1993). Markov models and optimization (Chapman and Hall).
Davis, M. H. A. (1984). Lectures on Stochastic Control and Nonlinear Filtering
(Springer).
De Broglie, L. (1930). An Introduction to the Study of Wave Mechanics (Methuen
& CO. LTD.).
De Broglie, L. (1960). Non-linear Wave Mechanics. A causal interpretation
(Elsevier).
Deheuvels, R. (1981). Formes Quadratiques et Groupes Classiques (Presses Uni-
versitaires de France).
Deutsch, D. (2003). Physics, philosophy, and quantum technology, in J. H.
Shapiro and O. Hirota (eds.), Proceedings of the Sixth International
Conference on Quantum Communication, Measurement and Computing
(Rinton Press, Princeton, NJ.), pp. 419–426.
Dieudonné, J.(1972). Treatise on Analysis, III (Academic Press).
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 320/4

320 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Diosi, L. and Lukács, B. (1987). In favor of a newtonian quantum gravity, Annalen

der Physik 44, 7, pp. 488–492.
Diosi, L. and Lukács, B. (1989). On the minimum uncertainty of space-time
geodesics, Physics Letters A 142, 6,7, pp. 331–334.
Dolgov, A. D., Maeda, H. and Torii, T. (2002). One more mechanism of electric
charge non-conservation, arxiv.org, URL http://arxiv.org/abs/hep-ph/
0210267.
Dowling, J. P. (2013a). On the curious consistency of non-relativistic
quantum theory with non-quantum relativity theory, Quantum Pun-
dit Blog entry, URL http://quantumpundit.blogspot.fr/2013/09/
on-curious-consistency-of-non.html.
Dowling, J. P. (2013b). Schrödinger’s Killer App: Race to build the world’s first
quantum computer (CRC Press).
Dürr, D., Sheldon, G. and Zanghi, N. (1992). Quantum equilibrium and the
origin of absolute uncertainty, J. Sta. Phys. 67, pp. 843–907, URL http:
//arxiv.org/pdf/quant-ph/0308039.
Dürr, D. and Teufel, S. (2009). Bohmian Mechanics (Springer).
Dutang, C. and Wuertz, D. (2009). A note on random number generation, on-
line, cran.r-project.org, URL http://cran.r-project.org/web/packages/
randtoolbox/vignettes/fullpres.pdf.
Einstein, A., Podolsky, B. and Rosen, N. (1935). Can quantum–mechanical de-
scription of physical reareal be considered complete? Physical Review, pp.
777–780.
Ellis, G. F. R. (2012). On the limits of quantum theory: Contextuality and the
quantumclassical cut, Annals of Physics 327, pp. 1890–1932.
Emch, G. G. (1972). Algebraic methods in Statistical Mechanics and Quantum
Field Theory (Wiley - Interscience).
Englert, B.-G., Scully, M. O., Süssman, G. and Walther, H. (1992). Surrealistic
Bohm trajectories, Z. Naturforsch. 47a, pp. 1175–1186.
Etter, T. and Noyes, H. P. (1998). Process, system, causality, and quantum me-
chanics, a psychoanalysis of animal faith, URL http://arxiv.org/abs/
quant-ph/9808011.
Falkenberg, E. D. (2001). Radioactive decay caused by neutrinos? Apeiron 8, 2,
pp. 32–45.
Falkenberg, E. D. (2002). Reply to does radioactivity correlate with the annual
orbit of earth around sun? by G. W. Bruhn, Apeiron 9, 2, pp. 41–42.
Fernandez, R. N. and Varadarajan, V. S. (2008). Matrix airy functions for com-
pact lie groups, Int. J. Math. 20, pp. 945–977, URL http://ncatlab.org/
nlab/show/Airy+function.
Fischbach, E., Jenkins, J. H., Buncher, J. B. and Gruenwald, J. T. (2010). Evi-
dence for solar influences on nuclear decay rates, in V. A. Kostelecký (ed.),
Proceedings of the Fifth Meeting on CPT and Lorentz Symmetry (World
Scientific), pp. 168–173.
Ford, L. R. (1979). Automorphic Functions (McGraw-Hill, Inc.).
Gilbert, J. E. and Murray, M. (1991). Clifford Algebras and Dirac Operators in
Harmonic Analysis (Cambridge University Press).
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 321/4

Bibliography 321

Gisin, N. (1989). Stochastic quantum dynamics and relativity, Helvetica Physica

Acta 62, pp. 363–371.
Goldberg, S. I. (1998). Curvature and Homology (Dover).
Good, I. J. (ed.) (1962). The Scientist Speculates: An anthology of partly-baked
ideas (Basic Books).
Gorini, V., Kossakowski, A. and Sudarshan, E. C. G. (1976). Completely positive
dynamical semigroups of n-level systems, J. Math. Phys. 17, pp. 821–825.
Gottfried, K. (1966). Quantum Mechanics (Plenum).
Grabowski, M. and Ingarden, R. S. (1989). Mechanika Kwantowa (PWN,
Warszawa), in Polish.
Grassberger, P. and Procaccia, I. (1983). Measuring the strangeness of strange
attractors, Physica D 9, pp. 189–208.
Hall, A. (1970). Genealogy of pythagorean triads, Mathematical Gazette 390, pp.
377–379.
Hänggi, E., Renner, R. and Wolf, S. (2010). Efficient device-independent quantum
key distribution, in H. Gilbert (ed.), Advances in Cryptology EUROCRYPT
2010 (Springer), pp. 216–234.
Harper, J. D. (2012). Pythagorean quadruples, Online preprint, URL http://
www.cwu.edu/~ harperj/PythagoreanQuadruples.pdf.
Harte, D. (2001). Multifractals: Theory and applications (Chapman & Hall/
CRC).
Heck, A. and Perdang, J. M. (eds.) (1991). Applying Fractals in Astronomy
(Springer).
Hestenes, D. (1985). Clifford algebra and the interpretation of quantum mechan-
ics, in J. S. R. Chisholm and A. Common (eds.), Clifford Algebras and
their Applications in Mathematical Physics, (NATO ASI series. Series C,
Mathematical afnd physical sciences; vol. 183) (Reidel), pp. 321–346.
Holba, A. and Lukacs, B. (1991). Is the anomalous brownian motion seen in
emulsions? Acta Physica Hungarica 70, 1-2, pp. 121–139.
Ivanenko, V. I. (2010). Decision Systems and Nonstochastic Randomness
(Springer).
Jadczyk, A. (1995a). On quantum jumps, events and spontaneous localization
models, Found. Phys. 25, pp. 743–762.
Jadczyk, A. (1995b). Particle tracks, events and quantum theory, Progr. Theor.
Phys. 93, pp. 631–646.
Jadczyk, A. (1995c). Topics in quantum dynamics, in R. Coquereaux et al. (eds.),
Infinite Dimensional Geometry, Noncommutative Geometry, Operator Al-
gebras and Fundamental Interactions, Vol. Proc. First Caribb. School of
Math. and Theor. Phys., Saint–Francois–Guadeloupe 1993 (World Scien-
tific), pp. 58–93.
Jadczyk, A. (2005). Simultaneous measurement of non-commuting observables
and quantum fractals on complex projective spaces, Chinese J. of Physics
43, 2, pp. 301–327.
Jadczyk, A. (2006). Some comment on the formal structure of spontaneous lo-
calization models, in A. Bassi, D. Dürr, T. Weber, and N. Zanghi (eds.),
Quantum Mechanics: Are there quantum jumps? and on the present status
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 322/4

322 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

of quantum mechanics, AIP Conference Proceedings, Vol. 844 (American

Institute of Physics), pp. 192–199.
Jadczyk, A. (2007). Quantum fractals on n-spheres. Clifford algebra approach,
Adv. Appl. Clifford Alg. 17, pp. 201–240.
Jadczyk, A., Kondrat, G. and Olkiewicz, R. (1996). On uniqueness of the jump
process in quantum measurement theory, J. Phys. A 30, pp. 1–18.
Jensen, H. J. (1988). Self-Organized Criticality (Cambridge University Press).
Karolyhazy, F. (1966). Gravitation and quantum mechanics of macroscopic ob-
jects, Il Nuovo Cimento A 42, 2, pp. 390–402.
Karolyhazy, F., Frenkel, A. and Lukács, B. (1986). On the possible role of gravity
in the reduction of the wave function, in R. Penrose and C. J. Isham (eds.),
Quantum Concepts in Space and Time (Clanderon Press), pp. 109–128.
Kennard, E. H. (1927). Zur quantenmechanik einfacher bewegungstypen, Z. Phys.
44, pp. 326–352.
Kent, A. (2005). Nonlinearity without superluminality, Physical Review A 72,
pp. 012108-1–4.
Khrennikov, A. (2008). Quantum-like probabilistic models outside physics, in
I. Licata and A. Sakaji (eds.), Physics of Emergence and Organization
(World Scientific), pp. 135–163.
Kijowski, J. (1974). On the time operator in quantum mechanics and the Heisen-
berg uncertainty relation for energy and time, Rep. Math. Phys. 6, pp.
361–386.
Kijowski, J. (1999). Comment on the “arival time” in quantum mechanics, Phys.
Rev. A 59, pp. 897–899.
Kijowski, J. (2005). Comment on ‘time operator’: The challenge persists, Conc.
Phys. II, pp. 99–102.
Kilmister, C. W. (ed.) (1987). Schrödinger — Centenary celebration of a polymath
(Cambridge University Press).
Kobayashi, S. and Nomizu, K. (1996). Foundations of Differential Geometry (Wi-
ley).
Kopeikin, S. and Efroimsky, M. (2012). Relativistic Celestial Mechanics of the
Solar System (Wiley-VCH).
Krauss, L. M. (1998). Beyond Star Trek: From alien invasions to the end of time
(BasicBooks).
Lämmerzahl, C., Macı́as, A. and Müller, H. (2005). Lorentz invariance violation
and charge (non)conservation: A general theoretical frame for extensions
of the maxwell equations, Phys. Rev. D 71, 2, pp. 025007–025022.
Landsman, N. P. (1991). Algebraic theory of superselection sectors and the mea-
surement problem in quantum mechanics, J. Mod. Phys. A 6, pp. 5349–
5371, requested via RG.
Landsman, N. P. (1995). Observation and superselection in quantum mechanics,
Stud. Hist. Phil. Mod. Phys. 26, pp. 45–73.
Lasota, A. and Mackey, M. C. (1985). Probabilistic Properies of Deterministic
Systems (Cambridge University Press).
Li, J. and Ostoja-Starzewski, M. (2012). Saturn’s rings are fractal, arxiv.org,
URL http://arxiv.org/abs/1207.0155.
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 323/4

Bibliography 323

Liebovitch, L. S. and Toth, T. (1989). A fast algorithm to determine fractal

dimensions by box counting, Physics Letters A 141, 8,9, pp. 386–390.
Lindblad, G. (1976). On the generators of quantum mechanical semigroups,
Comm. Math. Phys. 48, pp. 119–130.
Lloyd, S. (2000). Ultimate physical limits to computation, Nature 406, pp. 1047–
1054.
Lloyd, S. (2013a). Uncomputability and physical law, URL http://arxiv.org/
abs/1312.4456.
Lloyd, S. (2013b). The universe as quantum computer, URL http://arxiv.org/
abs/1312.4455.
Lozinski, A., Zyczkowski, K. and Slomczyński, W. (2003). Quantum iterated
function systems, Phys. Rev. E 68, 046110, http://arXiv.org/abs/quant-
ph/0210029.
Mackey, G. W. (1978). Unitary Group Representations in Physics, Probability,
and Number Theory (Benjamin).
Mielnik, B. (1994). The screen problem, Foundations of Physics 24, 8, pp. 1113–
1129.
Mielnik, B. and Torres-Vega, G. (2005). “Time operator”: the challenge persists,
Conc. Phys. II, pp. 81–97.
Milotti, E. (2002). 1/f noise: A pedagogical review, URL http://arxiv.org/
abs/physics/0204033.
Misner, C. W., Thorne, K. S. and Wheeler, J. A. (1973). Gravitation (W. H.
Freeman and Company).
Moretti, V. (2002). The interplay of the polar decomposition theorem and the
Lorentz group, URL http://arxiv.org/abs/math-ph/0211047.
Muga, J. G., Egusquiza, I. L., Damborenea, J. A. and Delgado, F. (2002a).
Bounds and enhancements for the negative scattering-time delays, Phys.
Rev. A 66, p. 042115.
Muga, J. G., Mayato, R. and Egusquiza, I. (eds.) (2002b). Time in Quantum
Mechanics, Vol 1 (Springer).
Muga, J. G., Ruschaupt, A. and del Campo, A. (eds.) (2009). Time in Quantum
Mechanics, Vol 2 (Springer).
Munkres, J. R. (2000). Topology (Prentice Hall).
Naber, G. L. (2012). The Geometry of Minkowski Spacetime (Springer).
Napiorkowski, K. (1971). Continuous tensor products of hilbert spaces and prod-
uct operators, Studia Mathematica XXXIX, pp. 307–327.
Nerukh, D. and Frederick, J. H. (2000). Multidimensional quantum dynamics
with trajectories: A novel numerical implementation of Bohmian mechan-
ics, Chem. Phys. Lett. 332, pp. 145–553.
Nicolis, G. and Prigogine, I. (1989). Exploring Complexity: An introduction
(W. H. Freeman & Company).
Ogborn, J. (2011). A first introduction to quantum behavior, URL http://www.
if.ufrj.br/~pef/aulas_seminarios/notas_de_aula/carlos_2011_1/
ensinoMQ/Ogborn_QuantumBehavior.pdf.
Ogborn, J., Collins, S. and Brown, M. (2003a). Randomness at the root of things
2: Poisson sequences, Physics Education 38, pp. 398–405.
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 324/4

324 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Ogborn, J., Collins, S. and Brown, M. (2003b). Randomness at the root of things
1: Random walks, Physics Education 38, pp. 391–397.
Olkhovsky, V. S. (2009). Time as a quantum observable, canonically conjugated
to energy, and foundations of self-consistent time analysis of quantum pro-
cesses, Advances in Mathematical Physics 2009, p. 83.
Olkiewicz, R. (1997). Some mathematical problems related to classical-quantum
interactions, Rev. Math. Phys. 9, p. 719.
Olkiewicz, R. (1999). Dynamical semigroups for interacting quantum and classical
systems, J. Math. Phys. 40, p. 1300.
Ozawa, M. (2001). Controlling quantum state reduction, Phys. Lett. A 282, p.
336.
Ozawa, M. (2002). Position measuring interactions and the Heisenberg uncer-
tainty principle, Phys. Lett. A 299, p. 1.
Ozawa, M. (2003). Universally valid reformulation of the Heisenberg uncertainty
principle on noise and disturbance in measurement, Phys. Rev. A 67,
p. 042105.
Ozawa, M. (2004a). Uncertainty principle for conservative measurement and com-
puting, URL http://arxiv.org/abs/quant-ph/0310071v1.
Ozawa, M. (2004b). Uncertainty relations for joint measurements of noncommut-
ing observables, Phys. Lett. A 320, pp. 367–374.
Palao, J. P., Muga, J. G. and Jadczyk, A. (1997). Barrier traversal times using a
phenomenological track formation model, Phys. Lett. A 233, pp. 227–232.
Palmer, J. (2012). Heisenberg uncertainty principle stressed in new test,
BBC News, Science & Environment, URL http://www.bbc.co.uk/news/
science-environment-19489385.
Parkhomov, A. G. (2006). Tri tipa izmenchivosti hoda razlichnyh processov, URL
http://www.chronos.msu.ru/RREPORTS/
parhomov_tritipaizmenchivosti.pdf.
Parkhomov, A. G. (2009). Upravljaemyj haos, in Materialy konferencii “Tor-
sionnye polja i informacionnye vzaimodejstvija - 2009, pp. 259–265, URL
http://www.second-physics.ru/sochi2009/pdf/p259-265.pdf.
Parkhomov, A. G. (2010). Periodic changes in beta decay rates, URL http:
//www.chronos.msu.ru/EREPORTS/PeriodicEng.pdf.
Pearle, P. (1984). Experimental tests of dynamical state-vector reduction, Phys-
ical Review D 29, 2, pp. 235–240.
Peitgen, H. O., Hartmut, J. and Saupe, D. (1992). Chaos and Fractals. New
frontiers of science (Springer).
Penrose, R. and Rindler, W. (1986). Spinors and Space-Time, Vol. 2 (Cambridge).
Peres, A. (1993). Quantum Theory: Concepts and methods (Kluwer).
Peres, A. (2000). Classical interventions in quantum systems, Phys. Rev. A 61,
p. 022117.
Pironio, S., Acin, A., Massar, R., Giroday, A. B. D. L., Matsukevich, D. N.,
Maunz, P., Olmschenk, S., Hayes, D., Luo, L., Manning, T. A. and Monroe,
C. (2010). Random numbers certified by Bell’s theorem, Nature 464, pp.
1021–1024.
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 325/4

Bibliography 325

Polchinski, J. (1991). Weinberg’s nonlinear quantum mechanics and the Einstein-

Podolsky-Rosen paradox, Physical Review Letters 66, 4, pp. 397–400.
Popper, K. (1982). Quantum Theory and the Schism of Physics (Rowman and
Littlefield).
Popper, K. (1993). Unended Quest. An intellectual autobiography (Routledge,
London).
Putnam, H. (1981). Quantum mechanics and the observer, Erkenntnis 16 2, pp.
193–219.
Random.org (2013). What’s this fuss about true randomness? URL http://www.
random.org/randomness/.
Ratcliffe, J. G. (1994). Foundations of Hyperbolic Manifolds (Springer).
Recami, E., Olkhovsky, V. S. and Maydanyuk, S. P. (2010). On non-self-adjoint
operators for observables in quantum mechanics and quantum field theory,
International Journal of Modern Physics A 25, 09, p. 175.
Recami, E., Zamboni-Rached, M. and Ignazio, L. (2013). On a time-space opera-
tor (and other non-selfadjoint operators) for observables in QM and QFT,
URL http://arxiv.org/abs/1305.3591.
Robertson, H. P. (1929). The uncertainty principle, Phys. Rev. 34, pp. 163–164.
Ruschhaupt, A. (2002a). A relativistic extension of event-enhanced quantum the-
ory, J. Phys. A: Math. Gen. 35, pp. 9227–9243.
Ruschhaupt, A. (2002b). Relativistic time-of-arrival and traversal time, J. Phys.
A: Math. Gen. 35, pp. 10429–10443.
Scarani, V. and Kurtsiefer, C. (2009). The black paper of quantum cryptography:
Real implementation problems, URL http://arxiv.org/abs/0906.4547.
Schlosshauer, M., Kofler, J. and Zeilinger, A. (2013). A snapshot of foundational
attitudes toward quantum mechanics, preprint, URL http://arxiv.org/
abs/1301.1069.
Schrödinger, E. (1952). Are there quantum jumps? The British Journal of the
Philosophy of Science 3, 10, pp. 478–502.
Schrödinger, E. (1955). The philosophy of experiment, Il Nuovo Cimento 1, pp. 5–
15.
Schrödinger, W. (1953). What is matter? Scientific American 189, pp. 52–57.
Schroeder, M. (1991). Fractals, Chaos, Power Laws (W. H).
Scully, M. O. (1998). Do Bohm trajectories always provide a trustworthy physical
picture of particle motion? Physica Scripta T76, pp. 41–46.
Sierpiński, W. (1988). Elementary Theory of Numbers (North–Holland, PWN).
Sinha, U., Couteau, C., Jennewein, T., Laflamme, R. and Weihs, G. (2010).
Ruling out multi-order interference in quantum mechanics, Science 329,
pp. 418–4121.
Slomczyński, W., Kwapień, J. and Życzkowski, K. (2000). Entropy computing
via integration over fractal measures, CHAOS 10, 180–188.
Spira, R. (1962). The diophantine equation x2 + y 2 + z 2 = m2 , The American
Mathematical Monthly 69, 5, pp. 360–365.
Springer, Y. A. (1977). Invariant Theory, Lect. Notes in Math. 585 (Springer).
Stapp, H. P. (1993). Mind, Matter and Quantum Mechanics (Springer).
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 326/4

326 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

Stekolshchik, R. (2008). Notes on Coxeter Transformations and the McKay Cor-

respondence (Springer).
Stenflo, O. (2002). Uniqueness of invariant measures for place-dependent random
iteration functions, IMA Vol. Math. Appl 132, pp. 13–32.
Stinespring, W. F. (1955). Positive functions on C ∗ -algebras, Proc. Amer. Math.
Soc. 6, pp. 211–216.
Streed, E. W., Mun, J., Boyd, M., Campbell, G. K., Medley, P., Ketterle, W. and
Pritchard, D. E. (2006). Continuous and pulsed quantum zeno effect, Phys.
Rev. Lett. 97, pp. 260402–260406, doi:10.1103/PhysRevLett.97.260402,
URL http://link.aps.org/doi/10.1103/PhysRevLett.97.260402.
Sturrock, P. A., Bertello, L., Fischbach, E., Javorsek II, D., Jenkins, J. H., Koso-
vichev, A. and Parkhomov, A. G. (2013). An analysis of apparent r-mode
oscillations in solar activity, the solar diameter, the solar neutrino flux, and
nuclear decay rates, with implications concerning the sun’s internal struc-
ture and rotation, and neutrino processes, Astroparticle Physics 42, pp. 62–
69, doi:http://dx.doi.org/10.1016/j.astropartphys.2012.11.011, URL http:
//www.sciencedirect.com/science/article/pii/S0927650512002150.
Towler, M. (2009). The return of pilot waves, URL http://www.tcm.phy.cam.
ac.uk/~mdt26/pilot_waves.html.
Tsonis, A. A. (2008). Randomnicity: Rules and randomness in the realm of the
infinite (Imperial College Press and World Scientific).
Tumulka, R. (2006). On spontaneous wave function collapse and quantum field
theory, Proc. Roy. Soc. Lond. A 462, pp. 1897–1908.
Valentini, A. (2009). Beyond the quantum, Physics World 22, 11, pp. 32–37.
Valentini, A. and Westman, H. (2005). Dynamical origin of quantum probabilities,
Proc. Roy. Soc. A 461, pp. 253–272.
van Kampen, N. G. (1988). Ten theorems about quantum mechanical measure-
ments, Physica A 153, pp. 97–113.
Varadarajan, V. S. (1985). Geometry of Quantum Theory (Springer).
Vince, A. (2013). Möbius iterated function systems, Trans. Amer. Math. Soc.
365, 1, pp. 491–509.
von Baeyer, H. C. (2013). Quantum weirdness? It’s all in your mind, Scientific
American 308, 6, pp. 46–51, doi:10.1038/scientificamerican0613-46, URL
http://dx.doi.org/10.1038/scientificamerican0613-46.
von Neumann, J. (1932). Mathematische Grundlagen der Quantenmechanik
(Springer).
Weil, A. (1980). Algebras with involutions and the classical groups, in Collected
papers, Vol. II (Springer), pp. 601–623.
Weinberg, S. (1989a). Precision tests of quantum mechanics, Physical Review
Letters 62, 5, pp. 485–488.
Weinberg, S. (1989b). Testing quantum mechanics, Annals of Physics 194, pp.
336–386.
Wheeler, J. A. (1984). Delayed-choice experiments and Bohr’s elementary quan-
tum phenomenon, in Proc. Int. Symp. Found. of Quantum Mechanics
(Physical Society of Japan), pp. 140–152.
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 327/4

Bibliography 327

Wheeler, J. A. (1990). It from bit, in N. B. Gakkai (ed.), Proceedings of 3rd

International Symposium on Foundations of Quantum Mechanics (Physical
Society of Japan), pp. 354–368.
Wigner, E. P. (1972). On the time-energy uncertainty relation, in A. Salam and
E. P. Wigner (eds.), Aspects of Quantum Theory (Cambridge University
Press), pp. 237–247.
Wojcik, D., Bialynicki-Birula, I. and Slomczyński, W. (2000). Time evolution of
quantum fractals, Phys. Rev. Lett. 85, 5022–5025.
Wu, L.-A. (2011). Nuclear spin polarization and manipulation by repeated mea-
surements, J. Phys. A: Math. Theor. 44, pp. 325302–325311.
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 329/4

Index

aberration, see light, aberration Grassmann, 179

affine connection, 171 homomorphism, 174
affine geometry, 171 Lie, 71
affine maps, 20 linear, 196
affine space, 45, 85 matrix, 175, 252
AI (Artificial Intelligence), 293 multilinear, viii
quantum, 293 normed, 246, 247
strong, hypothesis, 294 of observables, 246, 252, 254
Aichman, H., 231 of quaternions, see quaternions
algebra of space, 249
Banach, 246, 247 operator, 242, 247, 253, 255
Banach  algebra, 247 pure classical, 251
C0,2 , 175 pure quantum, 251, 274
C[2], 178, 184 R2 [1], 176
C2,0 , 176, 178 R[2], 175, 178, 183
C(R), 174 R2 [n], 176
Rr,s , 174, 175 real, associative, 174
C(V, Q), 174, 175, 179, 182 σ-, 303
C + (V, Q), 177 simple, 247
Clifford, 171, 173–193 , 245, 248
C1,−1 , 183 tensor, 175
C1,1 , 176 total system, of, 251, 273
C1,2 , 176 von Neumann, 246, 247, 253, 254
C3,0 , 176, 178, 180 algebra automorphisms, 252
Cr,s , table of, 176 algebraic description of quantum
C + (V 1 , Q1 ), 182 systems, 248
C[2], 176 algebraic framework, 253, 289
commutative, 248, 249 algebraic product of observables, 252
C  , 246, 247, 252–254 algorithm, 6, 13, 118, 241, 257, 258,
exterior, 179, 180 263, 269, 277, 281, 285, 289
free, 174 FD3, 168, 169, 171
free, associative, 174 for Cantor set, 11–13

329
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 330/4

330 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

for fractal dimension, 168 Big Crunch, 258

Fortran code, 118 Bloch sphere, 7, 28, 55–56, 60, 61, 63,
generating detection events, 270 64, 66, 68, 279, 287
hyperbolic quantum fractals, made of circles, 66
199–202 Bloch, F., 56
Liebovitch and Toth, 168 Bohm theory, 3, 4
piecewise deterministic process, Bohm, D., 233, 234, 236, 237, 239
244, 261 Bohmian mechanics, 3, 227, 232–239
quantum, 293 Bohmian trajectories, 237–239
Alien Invasions, 8 double slit, 239
alpha decays, 224 Bohmians, 239
Area 51, 8 Bohr, N., 250, 264, 266
area transformation, 128 boost
area transformation law, 78–85 direction, 49, 89, 93, 95, 96, 97
area, distortion of, 168 Lorentz, 49, 77, 86–90, 92, 95–97,
arrival time, see time, arrival of 130, 131, 186, 189, 249
Arveson, W., 255 spin, 187, 189, 192
atomic nucleus, 230 velocity, 49, 88, 89, 93, 95, 120, 127
atomic realm, 6 Borel measure, 307
atoms, 229, 240, 248 Borel sets, 303
attracting point, 111, 112 Borel structure, 304
attractor, 9, 14, 21, 38, 110, 116, 169, Born’s interpretation, 29, 270
195, 196, 254, 271, 277, 278 Born’s rule, 290
average contraction condition, 129 Born, M., 234, 235
averaging, 256 box dimension, see dimension,
fractal, capacity
balancing condition, 26–29, 33, 109, Boyarsky, A., 117
110 brain processes, 6
Banach space, 246, 301–302, 306, 309 Brillouin, L., 235
Banach subalgebra, 248 Brown, M., 215
band structure, 25 Brownian motion, anomalous, 294
Barning, F. J. M., 97 Bruhn, G. W., 223
Barnsley, M., 9, 20, 129 Buckminster Fuller geodesic dome, 9
basis
in the space of Hermitian matrices, Calude, C. S., 216
55 canonical commutation relations, 252,
Bayesian quantum theory, 3 267
Bayesian quantum theory Cantor
interpretation, 7 IFS, 21, 24
beables, 265 measure, 14, 15, 16, 24
Bell inequality violation, 214 chaos game, 15
Bell’s theorem, 214 set, 11–16, 18, 20, 21, 26, 23–27,
Bell, J. S., 5, 227, 228, 234, 237, 250, 29, 109, 114, 115, 278
251, 259, 264, 265 lines, 12
beta decay rates, 224 quantum, 109
Big Bang, 237, 258 Cantor, G., 11
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 331/4

Index 331

capacity dimension, see dimension, classical subsystem, 249

fractal, capacity Clifford algebra, see algebra, Clifford
cardinal numbers, 11 closed bounded subsets, 21
Carmeli, M., 52 closed group, 185
Cassini, 223 closed loops, 48
Cauchy sequence, 246, 297, 298 closed quantum system, 254
causality, see Einstein’s causality closed set, 33, 247, 303
CCR, see canonical commutation closed subset, 298, 300, 301
relations closure, 185, 187, 255, 298, 300, 307
celestial mechanics, 224 cloud chamber, 262, 269, 275
cell phones, 10 coarse-graining, 217
cells, 169, 172 cocycle condition, 69
center of algebra, 247, 252 coin, 13, 14
chaos, 7, 134, 143, 272, 282, 296 collapses, 265
regulated, 225 Collins, S., 215
chaos game, 13–15, 15, 16, 27, 36, Columbus, C., 241
39, 40, 132, 199, 202 commutative algebra, 247
chaotic behavior, vii, 167–171, 216, commutator, 17
264, 271, 272 compact set, 21
chaotic data, 296 compactification, 58
chaotic dynamics, 251, 267 complete metric space, 18, 20, 21,
chaotic patterns, 6 246, 298, 300, 302
chaotic phenomena, 226 complete normed space, 301, 302
chaotic systems, 215 complete positivity, 252, 253
characteristic equation, 73 completely positive maps, 251–254
characteristic function, 24, 116, 305 completely positive semigroup, 254
charge completeness formula, 51
conservation, 66, 67 completion, 298, 302, 308
non-conservation, 67 complex
charged particle, 267 algebra, 246
Christensen, E., 254 analytic transformation, 190
circle, 25, 37–39, 41, 63, 64, 66, 82, coefficients, 50
84, 109–120 conjugate vector, 105
mapped, 28 conjugation, 247
circle center, 38 dimensions, 32
circle radius, 38 field, 245
circular regions, 37, 157 finite measure, 308
circular windows, 9 Jacobian, 80, 81
classical computers, 292–294, 296 matrix, 25, 41, 43, 48, 52, 55, 67,
classical degrees of freedom, 248, 251 68, 86, 167, 175, 176, 186,
classical events, see events, classical 274
classical logic, 240 multiples of the identity, 247
classical mechanics, 196, 235, 295 plane, 29, 34, 38, 41, 42, 50, 79,
classical observables, 253 80, 83, 167, 190, 248
classical physics, 230, 260 plane representation, 43, 56–60, 77
classical state space, 262 probability amplitudes, 260
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 332/4

332 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

projective space, 171, 254, 274 index, 245

sequences, 262 limit, 263
space, 28, 60, 171 linear functional, 302
structure, 171, 173 linear mapping, 301, 302
subspaces, 50 linear operator, 309
translation, 94 Lipshitz, 129
unitary matrices, 30 mapping, 301–303
valued functions, 248 measure, 114, 115, 196
valued functions, algebra of, 248 monitoring, 6, 27, 229
valued set function, 304 multiplication, 246
vector space, 48, 105 parameters, 242
vectors, 42, 167 parametrization, 242
complex numbers, 57 path, 48
complex plane coordinates, 57 precession, 13
complex systems, 294 semigroup, 255
complexity, 216, 218, 293, 294 space, 261
compression, 36 spectrum, 249
computational inaccuracies, 30 tensor product, 263
computed trajectories, 237 variable, 242
computer implementation, 47, 53, 132 wave, 233
computer simulations, 114, 171, 221, continuous part, 13
222, 258, 275, 284 continuous spontaneous localization,
particle track, 275 3
conformal contracting region, 112
factor, 191 contraction, 20, 299, 300
maps, 195–199 factor, 17
spin geometry, 186–193 map, 19, 109, 299
transformation, 173, 186, 190–193, ratio, 20
279 uniform, 17
congruent polytopes, 122 weak, 299
conjugate pairs, 120 contraction and expansion regions,
connected component of the identity, 109, 111–112, 167
53, 111 contraction-expansion direction, 40
conscious activity, 217 contractive everywhere
consciousness, 225 transformation, 109
human, 294 contractive system, 20
constant phase, 65 Copenhagen interpretation, 3, 4, 7
constant probability, 63, 64 correlation dimension, see dimension,
continuous fractal, correlation
algebra homomorphism, 248 cosmic influences, 224
Dini, 129 cosmic origin, 222
evolution, 5, 27, 173, 232, 241, 244, cosmic scale, 240
249, 256 cosmology, 263, 293, 294
evolution with jumps, 244 cosmos, 4
functions, 117, 248, 249 countable
increase of probability, 232 collection, 303
June 16, 2014 9:20 World Scientific Book - 9in x 6in QuantumFractals3 pg. 333/4

Index 333

intersection, 303 capacity, 169

set, 242 correlation, 169
spectrum, 249 information, 169
union, 302, 304 quantum, 170
counting statistics, 285, 286 quantum octahedron, 172
covering homomorphism, 50, 51, 187 information, 168, 169
cryptography, 214 Diophantine equation, 104
CSL, see continuous spontaneous Diosi, L., 294
localization Dirac measure, 245
Cube, 18, 30, 31, 122, 123, 298 Dirac, P. A. M., 226
Quantum, 150, 151, 152, 153, discontinuous jumps, 13
149–154 discontinuous transitions, 173
curved surfaces, 18 discrete changes of states, 249
discrete data, 7
d’Espagnat, B., 241 discrete events, 5
Davis, M. H. A., 242 discrete finite approximation, 117
de Broglie, L., 233–237, 239 discrete set of classical states, 249
decay events, see events, decay discrete space, 261
decay processes, 226 discrete space-time structure, 10
decay rate, 222 discrete spectrum, 249
decoherence, 265, 296 discriminant, 73, 78
dense subset, 298 Dodecahedron, 122, 123
dense subspace, 298 Quantum, 166, 159–167, 201, 209
densities, 22, 23 double slit interference, 291
density matrix, 254, 258, 261, 287, double solution, 233
289 Dowling, J. P., 3, 293
density plot, 40 Dürr, D., 237
detector pointer, 258 dynamical
determinant, 25, 26, 30–32, 45, 48–50, coupling, 251, 252
54, 55, 58, 74, 94, 101, 105, 110, origin of quantum probabilities,
179, 193 237
determinant condition, 48, 50 reduction, 259, 260
determinant equation, 50 semigroup, 253–255
determinant map, 48 dynamical process, 270
determinant of a product, 48
Deutsch, D., 2 EEQT, see event enhanced quantum
dice, 13 theory
die eigenvalue, 25, 26, 54, 75–76, 82, 83,
classical, 280, 281 85, 116, 117, 120, 193, 228, 229,
quantum spin, 280, 281 246, 264, 270, 315
spin states, 281 complex, 120
standard, 280 equation, 73
dimension highest, 20
capacity, 168, 169 momentum, 273
correlation, 168 of the Frobenius-Perron matrix,
fractal, 12, 167–171 121, 122
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 334/4

334 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

position, 273 philosophy, of, 265

positive, 26, 54 track formation, 275
problem, 73 events, vii, 5, 98, 216–218, 222,
product of, 54 248–251, 258, 260, 262, 270, 275,
sum of, 54 286
eigenvector, 23, 25, 75, 117, 120, 167, catastrophic, 256
229, 246, 264, 270, 315 classical, 241, 243
invariant, 25 creation, 241
non-zero, 74, 75 decay, 213, 217, 221, 224
of the Frobenius-Perron operator, count rates, 224
116 counting, 218
Einstein’s causality, 231, 295 elementary, 217
Einstein’s convention, 29, 44 in detectors, 274
Einstein, A., 213, 235, 264 individual, 213, 218
electric charge, 240 number, of, 219, 220
electric potential, 5 past, 241
electron, 229 probability, of, 215, 218
elementary particles, 230 quantum, 215, 224, 226, 246,
Ellis, G. F. R., 294 249–251, 265
End of Time, 8 time series, of, 287
Englert, B. G., 238 time, of, 229
ensemble level, 256 excited atom, 213
entangled quantum particles, 214 experiment, 6–8, 218
environment, 241, 265
epistemology, 228 faces, 30
equal area cells, 168 factors, 247, 248, 254
equator, 63 fair coin, 13
equivalent metrics, 300 Falkenberg, E. D., 223
Erlang distribution, 219, 220, 276 FD3, see algorithm, FD3
Euclidean Feynman, R., 241
distance, 297, 299 fifth Solvay conference, 235
geometry, 171, 279 financial markets, 5
metric, 300 fine structure constant, 240
norm, 199 finite measures, 195
plane, 60 Finslerian geometry, 171
scalar product, 199 fixed point, 35, 77, 78, 82, 83, 85, 94,
space, 18, 25, 50, 63, 68, 185 101, 111, 120, 167, 196
volume, 192 fixed point equation, 78
Euler-Rodrigues formula, 311, 312 fixed point vector, 123–125
Evans, D., 254 flash ontology, 260
event creation mechanism, 242 flashes, 260
event enhanced quantum theory, viii, fluctuations, 221, 226
65, 226, 240–263, 265–270, flux lines, 239
273–276, 285, 296 foci of constant probability, 63
algorithm, 263, 287, 289 Fortran, 53, 199, 203–212
formalism, 296 Fourier transforms, 229
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 335/4

Index 335

fractal, 1, 4, 9, 11, 15, 17, 18, 30, 32, fundamental physics, 241
285 fuzzy projection, see projection, fuzzy
3D, 18, 68, 109
analysis of Saturn’s rings, 12 γ ray microscope, 229, 268
antenna, 9, 10 gauge
attractor, 254, 278 invariance, 66
behavior transformation, 61
control of, 167–171 gauge symmetry, 240
community, 203 Gaussian distribution, 219
compression, 289 Gaussian function, 261, 269, 275
dimension, see dimension, fractal, Gaussian integers, 33
168, 169 Gaussian packet, 239
effect, 273 Gedankexperimente, 229
image, 17 Geiger counter, 215, 225
industry, 4 Gelfand triples, 242
measure, 168, 169, 284, 287, 288 generalized Pythagorean quadruples,
circle, 284 102, 103
pattern, 6, 8, 37, 133, 267, 276, geometric interpretation, 60, 61
277, 285–287 Ghirardi-Rimini-Weber model, 3, 227,
properties, 195, 249 232, 259–263, 269, 270
quantum, 2–6, 8, 11, 13, 18, 19, 25, Gilbert, J. E., 186
29–31, 46, 47, 53, 55, 66–68, God, 11
109, 171, 173, 196, 254, 279, Gödel’s incompleteness theorem, 216
280, 282, 286 Golden Ratio, 33, 34
algorithms, 7 Goldstein, S., 237
higher dimensional, 68, 171 Good, I. J., 9
hyperbolic, 109–167 Góra, P., 117
parabolic, see parabolic Gorini, V., 254
quantum fractal GPQ, see generalized Pythagorean
set, 115 quadruple
structure, 278 Grabowski, M., 271
true, 9 Grassberger, P., 168
fractal-like patterns, 167, 282, 296 gravitational field, 263, 294
fractional linear transformation, 110, gravity, 67, 294, 295
111 quantum, 293, 294
Frenkel, A., 294 greatest common divisor, 104
frequencies of pairs, 39–41, 42 group
Frobenius-Perron operator, 22–25, O(2, 1), 111
115–117, 120, 121, 122, 125, SL(2, C), 32, 48–50, 67, 70, 110,
127–129, 195–199, 201, 208, 209, 314
308–310 action on S 2 , 50–55
band structure, 26 SL(2, R), 97
eigenvector, 27 SO(3, 1), 45
fixed point, 121, 123, 124 SO+ (3, 1), 85
function, measurable, see measurable, SU (2), 31, 32, 49, 55, 80, 81, 85,
function 86, 91, 95, 186
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 336/4

336 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

parametrization, 49 Hexahedron, see Cube

Spin, 179 hidden variables, 260
Spin(n, 1), 173 Hilbert space, 60, 228, 242, 245,
Abelian, 32, 96 253–255, 262, 273, 293
binary octahedral, 31 L2 (E), 242
Clifford, 185 2-dimensional, 55
covering, 32, 97 complex, 171, 173, 242, 247
Lie, 70, 190, 193 continuous tensor product, of, 261
Lorentz, 43–45, 110, 186 infinite dimensional, 260, 272, 293
action on sphere, 45–47 linear, 293
restricted, 32, 34 real, 171, 196, 247
octahedral, 30–31 rigid, 242
Poincaré, 45 separable, 253
theory, 17 tensor product, of, 262
group action, 52 total, of a system, 262
group homomorphism, 53–55 two-dimensional, 273
GRW, see Ghirardi-Rimini-Weber Hiley, B. J., 236, 239
model histogram, 14
history of an individual quantum
Haag, R., 252 system, 27
half-planes, 65 Holba, A., 294
Hall’s matrices, 101 Hopf fibration, 63, 65
Hall, A., 97 human beings, 8, 241, 293
Hamiltonian, 262, 268, 270, 271, 274 human endeavor, 217
Harper’s algorithm, 107, 108 human mind, 293
Harper, J. D., 106 human observers, 7, 266
Hausdorff human prejudices, 296
distance, 21, 22 hydrogen atom, 229
topological space, 248 hyperbolic iterated function system,
Hausdorff dimension, see dimension, 20
fractal, capacity hyperbolic transformation, 118, 129
hazard games, 13 hyperdimensional reality, 7
heavenly sky, 7 hyperplane, 46, 194
Heisenberg
formula, 271 Icosahedron, 122, 123
Heisenberg uncertainty Quantum, 154–159
interpretation, viii, 268, 296 circular regions, 157
principle, 1, 2, 232, 264, 266 five fold symmetry, 155
relation patterns, 156
time-energy, 230 snowflake patterns, 158
relations, viii, 4, 6, 266, 268, 271 idempotent, 56, 173, 267
Heisenberg, W., 2, 7, 266 identity matrix, 26, 27, 32, 48
Hermitian idempotent, 56 IFS, see iterated function system
Hermitian matrix, 51–56, 68 imaginary part, 53
Hermitian scalar product, 48, 57 imagination, 8
hermitian:matrix, 314 incomplete information, 226
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 337/4

Index 337

indefinite scalar product, 246 jumps, 13

indefiniteness, 259
independent and identically Karolyhazy, F., 294
distributed random variables, 226 Kennard’s inequality, 267
indeterminism, 2 Khrennikov, A., 226
individual level, 256 kickability, 279
individual quantum system, 287 kicking back, 279
individual systems, 258 Kijowski, J., 226, 230
infinitesimal generator, 243 knowledge, 7, 217
infinity, 11, 30, 42, 65 Koopman operator, 22, 23, 308–310
information, 7, 217, 292 Kossakowski, A., 254
information dimension, see Krauss, L. M., 8
dimension, fractal, information Kronecker δ, 24, 51
information transfer, 246 Kronecker, L., 11
Ingarden, R. S., 271
integro-differential operator, 243 Lambert’s projection, see projection,
interarrival times, 226 equal area
interferometers, 238 Lande, A., 279
interior of the unit disk, 58 Larmor precession, 13, 277
interval, 12, 16, 23, 24 law of large numbers, 228
invariant density, 117 laws of nature, 241
invariant measure, 25, 114–116, 120, laws of physics, 214
127–130, 196 Lebesgue measure, 114, 304
inverse matrix, 33, 48 left-stochastic matrix, 117
invertible transformation, 22, 23 light, 4
irreducible representations, 248 aberration, viii, 7, 8, 32, 85–87,
irreversibility, 242, 252 92–93, 279
irreversible processes, 254 relativistic correction, 93
isosurface, 64, 65, 65, 66, 67 cone, 29, 45, 46, 105, 111
tori, 64 future, 45–47, 54, 97, 98
iterated function system, 15–25, 29, generator lines, 45
31, 46, 66, 104, 113, 115, 115, 117, past, 45
120, 129, 132, 133, 167, 168, 195, quanta, 231
197, 254, 279 speed, 86, 240, 279
density, 114, 116 limiting set, 12
hyperbolic, 20–22 Lindblad, G., 254
six generators, 104 linear fractional transformation, 41,
iterations, 36, 47 50, 70, 80
Ivanenko, V. I., 216 Liouville equation, 256, 258, 261, 289
Iwasawa-type decomposition, 189 nonlinear, 295
Lipschitz constant, 299
Jacobian, 79, 113, 132 log-average contraction condition, 129
transformation dependence, 112 log-average function, 130
velocity dependence, 113 longitude, 63
Jaynes, E. T., 4 Lorentz
jumping process, 13 matrix, 68, 94
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 338/4

338 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

transformation, 29, 45, 54, 66, 89, structure, 195

92, 103, 110 transformation, 19, 304, 309, 310
Lorentz boost, see boost, Lorentz measure, 22–24
Lorentz group, see group, Lorentz, 53 measure densities, 116
orthochronous, 46 measurement, 3, 5, 265
restricted, 46–48, 50, 314 measurement accuracy, 2
Lorentz matrices, 28, 40, 45, 52, 96, measurement process, 254, 270
101, 110, 126, 130 measurement theory, 230
distance, 43 measuring device, 13, 229
lottery, 13 Menger-Sierpinski sponge, 17, 18, 20
lower hemisphere, 36, 58 slice, 18
Lukács, B., 294 top view, 19
Lyapunov’s exponents, 254 meridian, 63
metaphysics, 260
macrocosmic light patterns, 6 metric, 19, 21, 301, 302
macroscopic quantum effects, 240 flat, 85
macroscopic quantum phenomena, matrix, 44
240 Riemannian, see Riemannian
macroscopic systems, 263 metric
magnetic field, 13, 234, 236, 240, 267, space, 21, 297–300, 303, 307
276, 281, 311 compact, 300, 307
magnetic trap, 229 complete, see complete metric
magnetometer, 250 space
Malin, S., 52 locally compact, 300–301
Mandelbrot, B., 1 tensor, 44, 171
manifold, 191, 242, 244, 245 micromaser, 238
many-worlds interpretation, 3, 4 Mielnik, B., 230
Markov chain, 129 mind, 7, 293
Markov operator, 308–310 Minkowski
Markov process, 243 coordinates, 106
Master Equation, see Liouville geometry, 51
equation quadratic form, 54
Mathematica, 53 space, 29, 43, 51–53, 85, 97, 105,
Matlab, 53 181, 186, 311
matrix Minkowski space
positive, 54 symmetry group, 45
positivity, 54 Möbius transformation, vii, 25, 26,
matrix multiplication, 33, 48 28, 30, 31, 37, 68–70, 77, 80, 85,
matrix norm, 17 120, 185–186, 190–192, 195, 279
matrix-vector multiplication, 17 classification, 76–78
measurability, 22 elliptic, 76, 77
measurable hyperbolic, 77, 109, 126, 167, 168,
function, 19, 23, 195, 249, 305, 306, 171
308 loxodromic, 76, 77
set, 303–305 parabolic, 9, 32, 37, 38, 40, 76–78,
space, 19, 302–304, 309 94–96, 97, 114
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 339/4

Index 339

modulo 2π periodicity, 118 nuclear decay-rates, 223

molecules, 240, 248 fluctuations, 225
monkey, 8 periodicities, 224
multiplicative semigroup, 186, 187, nuclear emulsion, 262
190 number theory, 98
Murray, A. M., 186 numerator, 47
numerical inaccuracies and
Nature, 27, 265 instabilities, 30
neuronal activity, 5
pikes and bursts, 5 O(2, 1), see group, O(2, 1)
neutrino, 223 objective probability, 292
neutrino fluence oscillations, 223 objects, 7
neutrino mass, 223 observable, 55, 232, 265
Newtonian mechanics, 294 observational procedures, 246
Nimtz G., 231 observer, 7
Noether’s theorem, 66 observer’s knowledge, 7
non-commeasurability, 267 Octahedron, 122, 123, 274, 276, 277
non-commutativity, 17 Quantum, 143, 144, 145, 146,
non-commuting 147, 148, 142–149, 168,
observables 170, 200, 277
simultaneous measurement,
Ogborn, J., 215
of, vii, 251, 266–268,
one-dimensional subspaces of C2 ., 56
270–273, 296
one-parameter flows, 243
operations, vii
one-parameter group of unitary
non-Hamiltonian system, 255
transformations, 254
non-linear
one-parameter subgroups, 49, 71
wave mechanics, 233
one-point compactification, 110, 173
non-local correlations, 214
onset of fractality, 115
non-locality, 213
ontological status, 7
non-singular matrix, 54
non-trivial center, 247, 251, 253, 254 open ball, 21, 299, 300
non-unitary evolution, 27, 258 open subset, 299, 301, 303, 307
nonlinear open system dynamics, 254
deformations of QM, 295 open systems, 253
evolution, 295 open unit ball, 185, 187
functions, 196 operators
IFS, 254 in quantum theory, 55
observables, 196, 295 optics, 186
probability, 290 quantum, 268, 270
quantum mechanics, 29 orthogonal axis, 40, 96
nonlinearity, 233, 251, 256, 295 orthogonal basis, 24
norm, 48 orthogonal eigenvectors, 167
normalized vector, 57 orthogonal functions, 24
north pole, 35, 37, 56, 58, 63 orthogonal group, 49
nuclear decay events, see events, orthogonal hyperplane, 194
decay orthogonal matrices, 101
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 340/4

340 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

orthogonal projection, see projection, place dependent probabilities, 132

orthogonal Planck
orthogonal straight line, 194 constant, 236, 240, 266
orthogonal subspace, 193 scale, 293
orthogonal transformations, 179 Platonic solids, 109, 120, 132
orthogonal vector, 174, 193 Podolsky, B., 213
orthonormal basis, 85, 173–175, 177, Poincaré group, see group, Poincaré
179–181, 183, 193 Poincaré, H., 11
Ozawa, M., 267, 268 pointer position, 249, 268
pointer state, 258
pantheism, 3, 11 Poisson
parabolic matrices, 32 distribution, 219, 220
parabolic quantum fractal, 8, 9, 37, counts, 220
38, 40 density, 219
paradoxes, 2 process, 221, 222, 226
parallels of latitude, 63 counts, 221
parametric equations, 64 homogeneous, 276
Parkhomov, A. G., 223–225, 290, 295 inhomogeneous, 244, 257
particle decays, 258 polar coordinates, 61, 193
particle detectors, 238 polar decomposition, 94
particle momentum, 2, 6, 249 polar decomposition theorem, 86,
particle position, 2, 6 167, 189
particle track, 262, 275 polarization identity, 173, 174
particle trajectories, 235, 236 positive definite matrix, 26, 54
Pascal’s wager, 3 positive matrix, 54, 167
path connected space, 54, 55 positive operator valued measure, 274
Pauli matrices, 61, 176 positive operators, 267
Pauli, W., 233, 235, 236 positive square root, 188
PDP, see piecewise deterministic positivity, 252
process possibilities, 241, 250, 259
Pearle, Ph., 294 potentialities, 5
Penrose, R., 65, 294 predictions, 7
periodic variations, 223 prime numbers, 248
periodicities, 222, 223 principal anti-automorphism, 177, 186
phase, 61, 66, 67 principal automorphism, 177, 187
phase angle, 62 principal involution, 184
phase isosurface, 68 probabilistic measure, 24, 195, 198
phase parameter, 64 probabilistic models, 226
phase space, 6, 242, 249 probabilities, 23, 28, 33, 35, 36, 129,
phase transitions, 242, 263 226
photons, 213, 215, 230, 291 actualization, of, 241
piecewise deterministic evolution, 261 constant, 22, 128, 129, 168
piecewise deterministic process, 5, 6, cube, 149
242–246, 251, 254, 256–258 dodecahedron, 159
trajectory, 244 equal, 23
pilot wave, 233–237, 259, 260 evenly distributed, 203–208
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 341/4

Index 341

formula, for, 28 projection valued measure, 274

icosahedron, 154 projective method, 42
in EEQT, 258, 276 projective representation, 31
interpretation, of, 130–132 projective space, 167
jump, 256 propensities, 241
negative, 7 proportional complex vectors, 60
octahedron, 142 pseudo-random numbers, 14
of responding, 289 pseudocodes, 199
place dependent, vii, viii, 19, 22, psychic phenomena, 6
25, 27–29, 39, 42, 112–114, pure jump processes, 254
120, 121, 122, 128–130, pure quantum states, 7, 167
161, 162, 168, 195, pure state, 256, 258, 280, 281
202–208, 282 Pythagorean
quantum, 237 quadruples, 10, 33, 98–104, 107,
quantum, deformed, 290 108
tetrahedron, 133 primitive, 106–107, 108
time, 29 triples, 10, 97–98, 99, 101, 105
transition, 270 Pythagorean genealogical tree, 97, 99
transition, in jump processes, 295
uniform, 36, 117, 120, 121, 130 QIFS, see quantum iterated function
probability current, 289 system
probability density, 120, 244, 245 QKD, see quantum key distribution
probability distribution, 120, 195, quadratic form, 44, 45, 173, 174, 186
234–237 quadratic polynomials, 33
probability elements, 237, 239 quantum
probability formula, 65 chemistry, 238
probability function, 33, 34, 35, 132, complementarity, 240
283, 285 complementary quantities, 6
probability of the spin direction, 63 complementary variables, 260
probability theory, 217, 226 computer, 291
Procaccia, I., 168 computing, vii, 2, 3, 292, 295, 296
product spaces, 22 constructor theory, 2
projection cryptography, 2, 214, 290
equal area, 168 dynamics, 229
fuzzy, 66, 275 equilibrium, 237, 238
operator, 56 field theory, 228
orthogonal, 50, 56–58, 63, 68–70, fractals, 167
126, 274, 276 future, 6
postulate, 274, 275 games, vii, 245
sharp, 66 information, 2, 291
stereographic, 29, 36, 37, 38, 40, interference, 60, 238
50, 57–60, 63, 64, 65, 66, jumps, vii, 5, 6, 13, 27, 113, 173,
68, 70, 79, 80, 108, 126, 227–232, 242–244, 246, 249,
168, 192, 194–195, 199, 201 258, 259, 263, 276, 279, 282,
complex plane, 59 294
vertical, 39, 203–208 key distribution, 213
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 342/4

342 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

logic, 245 quantum-like probabilistic models,

measurement problem, 5, 265 226
measurement theory, 29, 271 quaternions, 31, 55, 175, 176
measurements, 2, 13, 232 qubit, 4, 105, 120
microphysics, 6
mind, 293 radio frequency oscillator, 249
noise, 213 radioactivity, 213, 215, 217, 218, 223
opticians, 238 decay rates, 223
optics, see optics, quantum variations of, 222
phase, 60, 66, 68 radius, 38
phenomena, 216, 238 Radon measure, 307
physicists, 4 Radon-Nikodym derivative, 308, 309
physics, vii, 3, 214, 217, 258, 281 Radon-Nikodym theorem, 308
probability, 226 random fluctuations, 15
processes, 221 random number generator, 14
properties, 2 random numbers, 14
psychology, vii, 245 random walk, 278
randomness, 214–217, 224, 251
signal, 230
irreducible, 226
state, 27, 60
nonstochastic, 216
state vector, 7, 60
quantum, 226, 295
superpositions, 6
true, 213–216, 224
system, 27
rational functions, 33
technology, 2
real coordinates, 60
theory
real part, 53
interpretations, 3
real sphere coordinates, 58
limits, 6 reality, 11
theory textbooks, 6 redundancy, 43
theory, algebraic, 251, 252 relative frequencies, 15
theory, axiomatization, 264 relative phase, 60
theory, mathematical foundations relativistic aberration, see light,
of, 264 aberration
theory, orthodox, 266 relativistic boost, see boost, Lorentz
theory, physical foundations of, 264 relativistic physics, 51
theory, probabilistic interpretation relic neutrino, 223
of, 295 renewal processes, 226
uncertainty, 2 repelling point, 111
weirdness, 1 restricted Lorentz group, 97
world, 1, 240, 241 Rieger periodicity, 223
Zeno effect, 6 Riemann sphere, 42, 58, 77, 190
Quantum Cube, see Cube, Quantum Riemannian geometry, 171
quantum dome fractal, 9 Riemannian manifold, 190
quantum events, see events, quantum Riemannian metric, 191, 193
quantum iterated function system, Robertson’s inequality, 266
25–30, 34, 37, 126, 132, 202 Rodrigues formula, see
quantum-classical couplings, 242 Euler-Rodrigues formula
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 343/4

Index 343

Rosen, N., 213 slice, 18

rotation, 49 snowflake patterns, 158
rotation group, 31 SO(3, 1), see group, SO(3, 1)
rotation invariant measure, 27, 129 SO+ (3, 1), see group, SO+ (3, 1)
rotation subgroup, 48 social phenomena, 6
rotational invariants, 61 solar neutrino, 223
Solar System, 93
Saturn’s rings, 12 space inversions, 54
scalar product, 24, 173 space of bounded measures, 23
scalings, 60 space of functions, 23, 24
Schrödinger space propulsion, 10
equation, 27, 234–236, 239, 251, space rotation, 86, 189
253, 259, 260, 289, 295 space-time, 43
nonlinear, 295 space-time signature, 171
evolution, 27, 173, 232, 265 space-time-energy information, 7
Schrödinger’s cat, 6, 259 special relativity, viii, 7, 32, 43, 86,
Schrödinger, E., 227, 228, 231, 235, 186, 214, 293
259, 260 spectra of chemical elements, 12
screen, 250 spectral measure, 267
Scully, M. O., 238 speed of light, see light, speed
secret communication, 214 sphere
seed, 14 S 3 , 49, 63, 68
self-organization, 296 spherical coordinate system, 96
self-organized criticality, 295 spherical coordinates, 78
self-similarity, 167, 267, 273 spin, 13, 55, 213, 266, 269, 273, 274,
semigroup of transformation, 242, 243 280, 281, 285
semigroups of positive maps, 252 1/2 quantum system, 167
set function, 22 component, 65
set of quantum states, 13 detector, 282
set, measurable, see measurable, set direction, 13, 28, 55, 63, 65, 274,
shadows, 7 276, 279–282
Sierpinski sponge, see detectors, 282, 285
Menger-Sierpinski sponge mutually incompatible, 6
Sierpinski triangle, 16, 17, 18 operators, 281
sigma-additivity, 304 evolution, 274
σ-algebra, 303 group, 48, 173, 179
sign conventions, 44 matrices, 88, 314
sign of energy, 44 measurement, 32, 65, 66, 281
signed measures, 115 operator, 281
similarity transformation, 91 projection, 273
simultaneous measurement, 264, 267 quantum state, 6
single atoms, 229 state, 7, 13, 33, 57, 60–68, 279,
single quantum systems, 229 281, 282, 285
sky, 7 state vector, 60
SL(2, C), see group, SL(2, C) system, 273, 274, 282, 283, 286
SL(2, R), see group, SL(2, R) vector, 60
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 344/4

344 Quantum Fractals: From Heisenberg’s Uncertainty to Barnsley’s Fractality

spin boost, see boost, spin Tetrahedron, 122, 123, 130

spinor, 62, 105 Quantum, 134, 135, 136, 137,
Spira, R., 104–106 138, 139, 140, 141,
SQIFS, see standard quantum 133–141
iterated function system Teufel, S., 237
square cells, 37 theoretical physics, 260, 293
SQUID, superconducting quantum theories of everything, 3
interference device, 229, 240, 249, theory of measurement, 228
250, 263, 270 thermodynamic equilibrium, 254
standard quantum iterated function thought formation, 6
system, 25, 27, 28, 31–40, 120, time, 5, 13, 44, 226, 230, 293
128–130 coordinate, 29, 43, 44, 51, 53
Stapp, H., 1, 7 dependent, 256
Startrek Physics, 8 dimension, of, 293
state vectors, 55, 60 direction, 54
state-transitions in atom, 231 evolution, 13, 241, 242, 252, 253,
statistical certainty, 213 257, 258, 276
statistical ensembles, 258 exposure, 270
statistics of consecutive pairs, 41, homogeneity, 242
288, 292 in quantum mechanics, 231
stereographic projection, see inversion, 46
projection, stereographic machine, 231
Stinespring theorem, 255 nature explorations, of, 223
stochastic control, 242 observable, 230
stochasticity, 251, 294 of arrival, 226, 230, 258
quantum, 295 relativistic, 259
string theories, 3 of decay, 213
SU (2), see group, SU (2) of jump, 244
sub-cubes, 18 operator, 226, 230
subatomic particle, 2 orientation, 46
subintervals, 14 random, 244
Sudarshan, E. C. G., 254 rate of jumps, 276
Süssman, G., 238 reality of, 229
sunspot activity, 5 relaxation, 269
superconductors, 229 reversal, 247
superluminal tunneling, 231 series, 287
superselection sectors, 253, 254 translation symmetry, 252
surface area, 28, 78, 80 Tonomura, A., 240
surjection, 188 topology, 54, 196, 248, 249
surjective maps, 19 torus, 64, 64, 65, 66
symmetric matrix, 20 totally antisymmetric, 51
symmetry group of the cube, 30 totally bounded set, 21
symmetry matrix, 32 trace, 52, 54, 70, 76, 198
symplectic manifold, 249 trace-class operators, 254
traceless matrix, 71–73, 75, 76
tables and chairs and cats, 240 traceless nilpotent, 74
June 13, 2014 9:0 World Scientific Book - 9in x 6in QuantumFractals3 pg. 345/4

Index 345

transformation, measurable, see unstable atom, 213

measurable, transformation ur-theory, 293
transformer, 187
transition measures, 243 Valentini, A., 235, 237
transition operators, 268 van Kampen, N. G., 227
translation, 17, 20, 38 vector
translation subgroup, 45 unobservable, 60
transposed matrix, 25 vertical projection, see projection,
trivial center, 247 vertical
true random number generators, 213, Vieta’s formulas, 73
216 Villarceau circles, 65, 66
true randomness, 213 Vince, A., 9, 32
Tsonis, A. A., 216 visualization, 18, 60–68
tunneling, 231, 259 von Neumann, J., 226
speed of, 231
time, 230, 231, 259 waiting detectors, 230
Turing machine, 216 Walter, H., 238
Turing, A., 9 wave
twistors, 65 packet, 259, 262, 289, 294
two-level atom, 167 reduction, 295
two-parameter commutative quantum, 240, 280
subgroup, 49 wave function, 5, 7, 29, 234, 259, 260,
two-slit experiment, 213, 240 265, 269, 275
collapse, 294
Uffink, J., 2 evolution, of, 261
UFO disk, 8 interpretation, 7
UFOs, 8 jumps, 6
ultra-relativistic speed, 93 reduction, 6, 294
unit circle, 29, 111 wave mechanics, 233, 235
unit interval, 29 waviness, 259
unit sphere, 56 weak experiments, 1
unit square, 16 Wheeler, J. A., 265
unit vector, 28, 46, 63 Wigner’s function, 6
unitarily equivalent matrices, 32 Wigner, E. P., 230
unitarity conditions, 49 Wilson chamber, 270
unitary group, 31 windows of order, 226
unitary matrix, 49, 54 Wolfram’s Demonstrations Project,
unitary quantum physics, 294 13
unitary rotation, 40
unitary transformation, 38 X-files, 8
universal property, 175
universality property, 174 YouTube, 9, 275
universe, 27, 216, 217, 225, 229, 258,
265, 291 Zanghi, N., 237
quantum, 292 Zeilinger, A., 3
unpredictability, 213