s3.amazonaws.com...acknowledgments this bookhas had a somewhatunusual history. originally,...
TRANSCRIPT
RIGOGIN* \342\226\240'VJ.'tfo/T.PT'TrTT^^ $$
The End of
CERTAINTYTime,Chaos,and the
New Laws ofNature
ILYA PRIGOGINEIn collaborationwith IsabelleStengers
THE FREE PRESS
New York London Toronto Sydney Singapore
THE FREE PRESSA Division ofSimon & Schuster Inc.
1230Avenue of the Americas
New York, NY 10020
Copyright \302\251 1996by Editions Odile JacobEnglish language translation copyright \302\251 1997by Editions Odile JacobAll rights reserved,
including the right of reproductionin whole or in part in any form.
First Free PressEdition 1997Published by arrangement with Editions Odile Jacob, Paris, France
THE FREE PRESSand colophon are trademarks
of Simon &Schuster Inc.
Designed by Carla Bolte
Manufactured in the United StatesofAmerica
10 987654321Library ofCongressCataloging-in-Publication Data
Prigogine, I. (Ilya)
[La fin des certitudes, English]
The end of certainty; time, chaos, and the new laws ofnature/Uya
Prigogine; in collaboration with Isabelle Stengers.p. cm.
Includes bibliographical references and index.
1.Science\342\200\224Philosophy, 2. Spaceand time. 3. Chaotic behavior
in systems, 4.Natural history. I. Stengers, Isabelle.
II. Tide.Q175.P75131997530.11\342\200\224dc21 97-3001
CIPISBN 0-684-83705-6
Original French edition entitled La Fin desCertitudes, by Ilya Prigogine in
collaboration with Isabelle Stengers,published by Editions Odile Jacob
CONTENTS
Acknowledgments vii
Author's Note ix
Introduction:A NewRationality? 1
1.Epicurus'Dilemma 9
2.Only an Illusion? 57
3.FromProbability to Irreversibility 73
4.TheLaws ofChaos 89
5.BeyondNewton'sLaws 107
6.A Unified Formulation ofQuantum Theory 129
7. OurDialoguewith Nature 153
8.DoesTimePrecedeExistence? 163
9.A Narrow Path 183
Notes 191Glossary 201Index 207
ACKNOWLEDGMENTS
Thisbookhas had a somewhatunusual history.
Originally,IsabelleStengersand I had intendedto translate
our bookEntre le Temps et I'Etemite(BetweenTimeand
Eternity) into English.1We had already prepared several
versions,oneofwhichappearedin German,and anotherin Russian.2But at the sametime,we weremakingimportant progressin the mathematical formulation ofourapproach. As a result, we abandoned our translation of the
originalbookand decidedto write a new version,whichhas recently appeared in French.3IsabelleStengershasaskednot to bedesignatedas a coauthorofthis newpresentation, but only as my collaborator.Although I felt
obligedto respecther wishes, I would like to stressthat
without her, this bookwould never have beenwritten.Iam most grateful for herassistance.
This textis the result ofdecadesofwork by both theBrusselsand Austin groups.While the physicalideas had
beenclearfor a considerablelengthoftime,their precisemathematical formulation has emergedonly during the
past few years.4I expressmy gratitude to the young,enthusiastic coworkerswho have beeninstrumental in
definingthe new approachto the nature oftime that forms the
basisofthis book,especiallyIoannisAntoniou (Brussels),
Vll
viii Acknowledgments
DeanDriebe(Austin), HiroshiHasegawa(Austin), TomioPetrosky(Austin), and ShuichiTasaki (Kyoto).I wouldalsoliketo mentionmy oldgroup in Brussels,who laid thefoundations that made further progresspossible.My thanks
to Radu Balescu,Michel de Haan, FrancoiseHenin,ClaudeGeorge,Alkis Grecos,and Fernand Mayne.Unfortunately, PierreResiboisand LeonRosenfeldare nolongerwith us.
Thework presentedin this bookcouldnot have beenaccomplishedwithout the support ofa numberoforganizations. I especiallywant to thank the BelgianCom-munaute Francaise,the Belgianfederal government,theInternationalSolvay Institutes(Brussels),the UnitedStates
Departmentof Energy, the EuropeanUnion,and theWelch Foundation (Texas).
English is not my native language,and I am muchobligedto Dr. E. C. GeorgeSudarshan and Dr. DeanDriebe,both from the University ofTexasat Austin, andDavid Lortimer(London),who have read the text with
great care.I also thank my Frenchpublisher,OdileJacob,who encouragedmeto write this new book,and StephenMorrow,my editorin the UnitedStates,as well asJudythSchaubhutSmith, for their help in preparingthe Englishedition.
I believethat we are at an important turning point in
the history ofscience.We have cometo the end of theroad paved by Galileoand Newton,whichpresentedus
with an imageofa time-reversible,deterministicuniverse.We now seethe erosionofdeterminismand the
emergenceofa new formulation ofthe laws ofphysics.
Ilya Prigogine
AUTHOR'SNOTE
I have triedto make this booka readable,self-containedaccountaccessibleto generalreaders.However,
especiallyin chapters5 and 6,I decidedto go into more
technical detail becausethe findings I have presenteddeviate
significantly from traditional views.In spiteofthe fact that
this volume is the result ofdecadesofwork, manyquestions still await answers.But taking into considerationthefinite life span ofeachofus, the fruits ofmy labors areshownsuch as they are today. I invite readersnot on a visit
to an archaeologicalmuseum,but rather on an adventurein sciencein the making.
IX
Introduction
A NEW RATIONALITY?
Earlierthis century in The Open Universe:An
Argument/or Indeterminism, Karl Popperwrote,\"Commonsenseinclines,on the onehand, to assertthat every event is
causedby someprecedingevents, so that every event canbe explainedor predicted....On the other hand, ...commonsenseattributes to mature and sane humanpersons ...the ability to choosefreely betweenalternative
possibilitiesofacting.\"1This \"dilemma ofdeterminism,\"as William Jamescalledit, is closelyrelatedto the meaningof time.2Is the future given, or is it under perpetualconstruction?A profound dilemma for all of mankind,as time is the fundamental dimensionof our existence.It was the incorporationof time into the conceptualschemeof Galilean physics that marked the originsofmodernscience.
This triumph ofhuman thought is also at the rootofthe main problemaddressedby this book:the denial ofwhat has beencalledthe arrow of time. As is well known,Albert Einstein often asserted,\"Timeis an illusion.\"In-
1
2 The End of Certainty
deedtime, as describedby the basiclaws ofphysics,from
classicalNewtoniandynamics to relativity and quantumphysics,doesnot includeany distinctionbetweenpast andfuture. Even today, for many physicistsit is a matter offaith
that as far as the fundamental descriptionofnature isconcerned, thereis no arrow oftime.
Yet everywhere\342\200\224in chemistry, geology,cosmology,biology, and the human sciences\342\200\224past and future playdifferent roles.Howcan the arrow oftime emergefrom what
physicsdescribesas a time-symmetricalworld?This is thetime paradox, oneofthe central concernsofthis book.
Thetime paradoxwas identifiedonly in the secondhalfof the nineteenthcentury after the ViennesephysicistLudwigBoltzmanntriedto emulatewhat CharlesDarwinhad donein biologyin an effort to formulate an
evolutionary approach to physics.The laws ofNewtonian physicshad longsincebeenacceptedas expressingthe idealofobjective knowledge.As they impliedthe equivalencebetween past and future, any attempt to confera fundamental
meaningon the arrow oftime was resistedas a threat tothis ideal.Isaac Newton'slaws were consideredfinal in
theirdomain ofapplication,somewhat the way quantummechanicsis now consideredto be final by manyphysicists. How then can we introduceunidirectionaltimewithout destroying these amazing achievementsof thehuman mind?
SinceBoltzmann,the arrow oftime has beenrelegatedto the realm ofphenomenology.We, as imperfecthuman
observers,are responsiblefor the difference betweenpastand future through the approximationswe introducein
ourdescriptionofnature.This is still the prevailingscientific wisdom.Certainexpertslament that we stand beforean unsolvable mystery for whichsciencecan provide no
Introduction 3
answer.We believethat this is no longerthe casebecauseof two recentdevelopments:the spectaculargrowth ofnonequilibriumphysicsand the dynamicsofunstable
systems, beginningwith the idea ofchaos.Over the past several decades,a new sciencehas been
born,the physics of nonequilibrium processes,and has ledto
conceptssuch as self-organizationand dissipative structures,
whichare widely used today in a largespectrumofdisciplines, includingcosmology,chemistry, and biology, aswell as ecologyand the socialsciences.The physicsofnonequilibriumprocessesdescribesthe effectsofunidirectional time and gives fresh meaningto the termirreversibility. In the past, the arrow of time appeared in
physicsonly through simpleprocessessuch as diffusion orviscosity, whichcouldbe understoodwithout anyextension of the usual time-reversibledynamics. This is nolongerthe case.We now know that irreversibility leadstoa host of novel phenomena,such as vortex formation,chemicaloscillations,and laserlight, all illustrating theessential constructive roleofthe arrow oftime.Irreversibilitycan no longerbe identifiedwith a mereappearancethat
would disappearif we had perfect knowledge.Instead, it
leads to coherence,to effects that encompassbillionsandbillionsofparticles.Figuratively speaking,matter at
equilibrium, with no arrow oftime,is \"blind,\" but with the
arrowoftime,it beginsto \"see.\"Without this newcoherence due to irreversible,nonequilibriumprocesses,life onearth would beimpossibleto envision.Theclaim that thearrow oftime is \"only phenomenological,\"or subjective,is thereforeabsurd.We are actually the childrenof thearrow oftime,ofevolution, not its progenitors.
Thesecondcrucialdevelopmentin revising the conceptof time was the formulation of the physicsof unstable
4 The End of Certainty
systems.Classicalscienceemphasizedorderand stability;
now, in contrast, we seefluctuations, instability, multiplechoices,and limited predictabilityat all levels ofobservation. Ideassuchas chaoshave becomequite popular,influencing our thinking in practically all fieldsofscience,from
cosmologyto economics.As we shall demonstrate,we cannow extendclassicaland quantum physicsto includeinstability
and chaos.We are then able to obtain a formulationofthe laws ofnature appropriatefor the descriptionofourevolving universe, a descriptionthat containsthe arrow oftime,sincepast and future no longerplay symmetricalroles.In the classicalview\342\200\224and herewe includequantummechanicsand relativity\342\200\224laws of nature expresscertitudes. When appropriate initial conditionsare given, wecan predictwith certainty the future, or \"retrodict\" the
past.Onceinstability is included,this is no longerthe case,and the meaningof the laws ofnature changesradically,for they now expresspossibilitiesorprobabilities.Herewe
go against oneofthe basictraditions ofWestern thought,the beliefin certainty. As statedby GerdGigerenzeret al.in The Empire of Chance,\"Despitethe upheavals in sciencein the over two millennia separatingAristotle from theParis ofClaudeBernard,they sharedat least oneattitudeoffaith: Sciencewas about causes,not chance.Kant even
promoteduniversal causal determinismto the status ofa
necessaryconditionofall scientificknowledge.\"3Therewere, however, dissenting voices.The great
physicistJamesClerkMaxwell spoke ofa \"new kind ofknowledge\"that would overcomethe prejudiceofdeterminism.4 But, on the whole,the prevailing opinionwasthat probabilitieswerestates ofmind rather than states ofthe world.This is so even today in spite of the fact that
quantum mechanicshas includedstatisticalconceptsin the
Introduction 5
coreof physics.But the fundamental objectofquantum mechanics,the wave function, satisfiesa deterministic,time-reversibleequation.To introduceprobability and
irreversibility, the orthodoxformulation ofquantummechanics requiresan observer.
Throughhis measurements,the observerwould bringirreversibility to a time-symmetricuniverse.Again, as in
the time paradox,we would be responsiblein somesensefor the evolutionary patterns ofthe universe.This roleofthe observer,whichgave quantum mechanicsits subjectiveflavor, was the main reasonthat preventedEinstein from
endorsingquantum mechanics,and it has sinceledto
unending controversies.Theroleofthe observerwas a necessaryconceptin the
introductionof irreversibility, or the flow of time,into
quantum theory. But once it is shown that instabilitybreakstime symmetry, the observeris no longeressential.In solving the time paradox,we also solve the quantumparadox,and obtain a new, realisticformulation ofquantum theory.This doesnot meana return to classicaldeterministic orthodoxy; on the contrary, we go beyond thecertitudesassociatedwith the traditional laws ofquantumtheory and emphasizethe fundamental roleofprobabilities. In both classicaland quantum physics,the basiclawsnow expresspossibilities.We neednot only laws, but alsoevents that bringan elementofradicalnovelty to the
description ofnature. This novelty leadsus to the \"new kindofknowledge\"anticipatedby Maxwell.ForAbraham DeMoivre, one of the founders of the classicaltheory ofprobabilities,chancecan neitherbe defined norunderstood.5 As we shall illustrate, we are now able to includeprobabilities in the formulation of the basic laws ofphysics.Oncethis is done,Newtoniandeterminismfails;
6 The End of Certainty
the future is no longerdeterminedby the present,and the
symmetry betweenpast and future is broken.Thisconfronts us with the most difficult questionsofall: What arethe roots oftime? Didtime start with the \"big bang\"?Ordoestime preexistour universe?
Thesequestionsplaceus at the very frontiers ofspaceand time.A detailedexplanationofthe cosmologicalimplications ofour positionwould requirea specialmonograph. Briefly stated,however, we believethat the bigbangwas an event associatedwith an instability within the
mediumthat producedour universe.It markedthe start ofour universe but not the start oftime.Although ouruniverse has an age,the mediumthat producedour universehas none.Timehas no beginning,and probablyno end.
But herewe enterthe world ofspeculation.The main
purposeofthis bookis to presentthe formulation ofthelaws ofnature within the rangeoflow energies.Thisis thedomain ofmacroscopicphysics,chemistry, and biology.Itis the domain in which human existenceactually takes
place.The problemsoftime and determinismhave remained
at the coreofWestern thought sincethe pre-Socratics.How can we conceiveofhuman creativity or ethics in a
deterministicworld?This questionreflects a profoundcontradictionin
Western humanistic tradition, which emphasizesthe
importance ofknowledgeand objectivity, as well as individual
responsibilityand freedomof choiceas impliedby the
ideal ofdemocracy.Popperand many otherphilosophershave pointed out that we are faced with an unsolvable
problemas long as nature is describedsolelyby a
deterministic science.6Consideringourselvesas distinct from
the natural worldwould imply a dualismthat is difficult for
Introduction 7
the modernmind to accept.Our aim in this work is toshow that we can now overcomethis obstacle.If\"the
passion ofthe westernworld is to reunitewith the groundofits being,\"as RichardTarnas has written, perhapsit is nottooboldto say that we are closingin on the objectofourpassion.7
Mankind is at a turning point, the beginningofa new
rationality in whichscienceis no longeridentified with
certitudeand probabilitywith ignorance.We agreecompletely with Yvor Leclercwhenhe writes, \"In the presentcenturywe are suffering from the separationofscienceand
philosophywhich followedupon the triumph ofNewtonian physicsin the eighteenthcentury8JacobBronowski
beautifully expressedthe same thought in this way: \"Theunderstandingofhuman nature and ofthe humancondition within nature is oneof the central themesofsci-
\"9ence.At the endofthis century, it is often askedwhat the
future of sciencemay be.Forsome,such as StephenW.
Hawking in his BriefHistory of Time, we are closeto the
end,the momentwhenwe shallbeableto read the \"mind
ofGod.\"10In contrast, we believe that we are actually at
the beginningofa new scientificera.We are observingthebirth ofa sciencethat is no longerlimited to idealizedand
simplifiedsituations but reflects the complexityofthe real
world, a sciencethat views us and our creativity as part ofa fundamental trend presentat all levels ofnature.
Chapter1
EPICURUS'DILEMMA
I
Isthe universe ruledby deterministiclaws? What is the
nature oftime? Thesequestionswereformulated by the
pre-Socraticsat the very start ofWestern rationality. After
morethan twenty-five hundredyears,they are still with us.However, recentdevelopmentsin physicsandmathematics associatedwith chaosand instability have openedupdifferent avenues ofinvestigation. We are beginningto seethese problems,which deal with the very position ofmankind in nature, in a new light,and can now avoid the
contradictionsofthe past.TheGreekphilosopherEpicuruswas the first to address
a fundamental dilemma.As a follower ofDemocritus,hebelievedthat the world is made of atoms and the void.Moreover,heconcluded,atoms fall through the void at thesame speedand on parallel paths. How then could theycollide?Howcouldnovelty associatedwith combinationsofatoms ever appear?ForEpicurus,the problemsofsci-
9
10 The End of Certainty
ence,the intelligibility ofnature, and human destinycouldnot be separated.What couldbe the meaningofhumanfreedomin a deterministicworld ofatoms? As Epicuruswrote to Meneceus,\"Ourwill is autonomousand
independent and to it we can attribute praise or disapproval.Thus,in orderto keepour freedom,it would have beenbetterto remain attached to the beliefin godsrather than
beingslavesto the fate ofthe physicists:Theformergivesus the hopeofwinningthe benevolenceofdeitiesthroughpromiseand sacrifices;the latter, on the contrary, bringswith it an inviolable necessity.\"1How contemporary this
quotationsounds!Again and again, the greatestthinkersin
Western tradition, suchas ImmanuelKant, Alfred NorthWhitehead, and Martin Heidegger,felt that they had tomake a tragic choicebetweenan alienating scienceor anantiscientific philosophy.They attempted to find somecompromise,but noneprovedto besatisfactory.
Epicurus thought that he had found a solutionto this
dilemma,whichhe termedthe clinamen. As expressedbyLucretius,\"While the first bodiesare beingcarrieddownwards by their own weight in straight lines through the
void, at times quite uncertain and at uncertain places,they
deviate slightly from their course,just enoughto bedefinedas
having changeddirection.\"2But no mechanismwas given forthis clinamen.No wonderthat it has always beenconsidered a foreign, arbitrary element.
But do we needthis novelty at all?ForHeraclitus,asunderstood by Popper,\"Truth lies in having grasped theessentialbecoming ofnature, i.e.,having representedit as
implicitly infinite, as a processin itself.\"3 Parmenidestookthe oppositeview. In his celebratedpoemon the uniquereality ofexistence,he wrote,\"Norwas it ever, norwill it
be,sincenow it is, all together.\"4
Epicurus'Dilemma 11
It is amusing that the Epicurus clinamenhas appearedrepeatedlyin the scienceofourcentury. Inhis classicpaperon the emissionofphotonsassociatedwith the transitionsbetweenatomicstates (1916),Einsteinexplicitlyexpressedhis confidencein scientificdeterminism,although heassumed that theseemissionsare ruledby chance.
Greekphilosophy was unable to solve this dilemma.Plato linked truth with being,that is,with the unchangingreality beyond becoming.Yet he was consciousof the
paradoxicalcharacterofthis positionbecauseit woulddebase both life and thought.In The Sophist,he concludedthat we needboth beingand becoming.5
This duality has plaguedWestern thought ever since.As
observedby the FrenchphilosopherJeanWahl, the historyofWestern philosophyis, on the whole,an unhappy one,characterizedby perpetualoscillationsbetweenthe worldas an automaton and a theologyin whichGodgovernstheuniverse.6Bothare forms ofdeterminism.
This debate tooka turn in the eighteenthcentury with
the discoveryofthe \"laws ofnature.\" Theforemost
examplewas Newton'slaw relating forceand acceleration,
whichwas both deterministicand, moreimportant, time
reversible.Oncewe know the initial conditions,we cancalculate all subsequentstates aswellas the precedingones.Moreover,past and future play the same rolebecauseNewton'slaw is invariant with respectto the time
inversion \302\243 \342\200\224> -t.This leads to nightmares such as the demonimaginedby Pierre-Simonde Laplace,capableofobserving
the currentstate ofthe universe and predictingitsevolution.7
As is wellknown, Newton'slaw has beensupersededinthe twentieth century by quantum mechanicsand
relativity.Still, the basiccharacteristicsofhis laws\342\200\224determinism
12 The End of Certainty
and time symmetry\342\200\224have survived. It is true that
quantum mechanicsno longerdealswith trajectoriesbut with
wave functions (seeSectionIV ofthis chapterandChapter 6),but it is important to note that the basicequationofquantum mechanics,Schrodinger'sequation,is onceagaindeterministicand time reversible.
By way ofsuch equations,laws ofnature lead tocertitudes. Onceinitial conditionsare given, everything isdetermined. Nature is an automaton, whichwe can control,at least in principle.Novelty, choice,and spontaneousaction are real only from our human point ofview.
Many historiansbelievethat an essentialrolein this
vision ofnature was playedby the Christian Godasconceived in the seventeenth century as an omnipotentlegislator.Theologyand scienceagreed.As Gottfried vonLeibnizwrote,\"In the least ofsubstances,eyesas piercingas thoseofGodcouldread the wholecourseofthings in
the universe, quaesint, quaefuerint, quaemoxfutura trahantur\"
(thosewhichare, whichhave been,and whichshall be in
the future).8 The discoveryofnature'sdeterministiclawswas thus bringinghuman knowledgecloserto the divine,
atemporalpoint ofview.Theconceptofa passivenature subjectto deterministic
and time-reversiblelaws is quite specificto the Westernworld.In Chinaand Japan, nature means \"what is byitself.\" In his excellentbookScienceand Societyin East and
West, JosephNeedhamtells us of the irony with whichChinesemenoflettersgreetedthe Jesuits'announcementof the triumphs ofmodernscience.9Forthem, the ideathat nature is governedby simple,knowablelaws seemedto be a perfect exampleof anthropocentricfoolishness.Accordingto Chinesetradition, nature isspontaneoushar-
Epicurus'Dilemma 13
mony; speakingabout \"laws ofnature\" would thus subjectnature to someexternalauthority.
In a messageto the great Indian poet,Rabindranath
Tagore,Einsteinwrote:
If the moon, in the act of completingits eternal path round
the earth, were gifted with self-consciousness,it would feel
thoroughly convincedthat it would travel its path on its own,in accordancewith a resolutiontaken onceand for all.
Sowould a Being,endowedwith higher insight and more
perfect intelligence, watching man and his doings, smile
about this illusion of his that he was acting accordingto his
own freewill.
This is my belief,although I know well that it is not fully
demonstrable.If one thinks out to the very last consequencewhat one exactlyknows and understands, there would hardly
be any human beingwho could be imperviousto this view,
providedhis self-lovedid not rub up against it. Man defendshimself from being regarded as an impotent object in the
courseof the Universe.But shouldthe lawfulness of
happenings,such as unveils itselfmore and more clearlyin inorganic
nature, ceaseto function in the activities in our brain?10
To Einstein, this appearedto be the only positioncompatible with the achievementsofscience.But this
conclusion is as difficult to acceptnow as it was to Epicurus.Timeis our basic existentialdimension.Sincethe nineteenthcentury, philosophyhas becomemoreand moretime
centered, aswe seein the work ofGeorgWilhelm Hegel,Edmund Husserl,William James,HenriBergson,Martin
Heidegger,and Alfred North Whitehead.ForphysicistssuchasEinstein,the problemhas beensolved.Forphiloso-
14 The End of Certainty
phers,it remains the central question ofontology,at the
very basisofthe meaningofhuman existence.In The Open Universe:An Argument for Indeterminism,
Popper wrote, \"I regard Laplaciandeterminism\342\200\224confirmed
as it may seemto beby the prima facie deterministictheories of physics,and by their marvelous success\342\200\224as themost solid and seriousobstacleto our understandingand
justifying the nature ofhuman freedom,creativity, and
responsibility.\" ForPopper,\"Thereality oftime and changeis the cruxofrealism.\"11
In his short essay,\"ThePossibleand the Real,\"Bergsonargued, \"What is the role of time? .. .Timepreventseverything from beinggiven at once....Is it not thevehicle ofcreativity and choice?Is not the existenceoftime
the proofofindeterminismin nature?\"12Forboth Popperand Bergson,we need\"indeterminism.\"But how do we
gobeyonddeterminism?This difficulty is wellanalyzed in
an essayby William Jamesentitled \"TheDilemmaofDeterminism.\"13 In accordwith well-definedmechanisms,determinismis \"mathematizable,\" as shownby the laws ofnature formulated by Newton,Schrodinger,and Einstein.Incontrast, deviations from determinismseemto introduceanthropomorphicconceptssuchas chanceoraccident.
The conflict betweenthe time-reversibleview ofphysics and time-centeredphilosophyhas ledto an openclash.What is the purposeofscienceif it cannotincorporatesomeofthe basicaspectsofhuman experience?The anti-
scientificattitude of Heideggeris well known.AlreadyFriedrichNietzschehad concludedthat thereare no facts,
only interpretations.As statedby JohnR.Searle,postmodern philosophy,with its idea ofdeconstruction,challengesWestern traditions regardingthe nature oftruth, objectiv-
Epicurus'Dilemma 15
ity, and reality.14In addition, the roleof evolution, ofevents, in our descriptionofnature is steadily increasing.How then can we maintain a time-reversibleview ofphysics?
In October1994,thereappeareda specialissueofScientific
American devoted to \"life in the universe.\"15At all
levels\342\200\224cosmology, geology,biology,and humansociety\342\200\224
we seea processofevolution in regardto instabilitiesandfluctuations. We thereforecannotavoid the question:Howare theseevolutionary patterns rootedin the fundamentallaws ofphysics?Only onearticle,written by the eminentphysicistStevenWeinberg,is relevant to this problem.Hewrites, \"As muchas we would like to take a unified viewofnature, we keepencounteringa stubborn duality in the
roleofintelligentlife in the universe, as both subjectandstudent. ...On the one hand, there is the Schrodingerequation,whichdescribesin a perfectly deterministic wayhow the wave function ofany systemchangeswith time.Then, quite separate,there is a set ofprinciplesthat tellshow to use the wave function to calculatethe probabilitiesofvarious possibleoutcomeswhen someonemakes themeasurement.\"16
Doesthis suggest that through our measurements,weourselvesare at the originofcosmicevolution? Weinbergspeaksof a stubborn duality, a point ofview found in
many recentpublicationssuch as StephenW Hawking'sBriefHistory of Time.17HereHawking advocatesa purelygeometricalinterpretation of cosmology.In short, timewouldbean accidentofspace.But heunderstandsthat this
interpretationis not enough.We needan arrow oftime todealwith intelligentlife.Therefore,alongwith many othercosmologists,Hawking introducesthe so-calledanthropic
16 The End of Certainty
principle.Nevertheless,this principleis as arbitrary as was
Epicurus'clinamen.Hawking gives no indicationofhowthe anthropic principlecouldever emergefrom a static
geometricaluniverse.As mentionedearlier,Einstein attempted to maintain
the unity ofnature, includingmankind, at the costofreducing us to mereautomata. This was also the view ofBaruchSpinoza.But therewasanotherapproachsuggestedby ReneDescartes,also in the seventeenthcentury, whichinvolved the conceptofdualism:on onesideis matter, res
externa, as describedby geometry,and on the other, the
mind, associatedwith res cogitans.iSIn this way, Descartesdescribedthe striking difference betweenthe behaviorofsimplephysicalsystemssuchas a frictionlesspendulumand
the functioning of the human brain.Curiously, the
anthropic principlebrings us back to Cartesiandualism.In The Emperor'sNew Mind, RogerPenrosestates,\"It is
ourpresentlackofunderstandingofthe fundamental lawsofphysicsthat prevents us from comingto grips with the
conceptof 'mind'in physicalorlogicalterms.\"19We
believe that Penroseis right:We needa new formulation ofthe fundamental laws ofphysics.Theevolutionary aspectsofnature have to beexpressedin termsofthe basiclaws ofphysics.Only in this way can we give a satisfactory answerto Epicurus'dilemma.Thereasons for indeterminism,for
temporalasymmetry, must berootedin dynamics.Formulations that do not containthesefeatures are incomplete,exactly as would be formulations ofphysics that ignoregravitation orelectricity.
Probability plays an essentialrolein most sciences,from
economicsto genetics.Still, the idea that probabilityis
merelya state ofmind has survived. We now have to go a
step farther and show how probabilityentersthe funda-
Epicurus'Dilemma 17
mental laws ofphysics,whetherclassicalor quantum. A
new formulation ofthe laws ofnature is now possible.Inthis way, we obtain a moreacceptabledescriptionin whichthereis roomfor both the laws ofnature and novelty and
creativity.At the beginningof this chapter, we mentionedthe
pre-Socratics.In fact, we owe to the ancientGreekstwoidealsthat have sinceshapedhuman history. The first is the
intelligibility ofnature, or in Whitehead'swords, \"the
attempt to frame a coherent,logical,necessarysystemofgeneralideasin terms ofwhichevery elementofourexperiencecan be interpreted.\"20The secondis the idea ofdemocracybased on the assumptionofhuman freedom,creativity, and responsibility.As longas scienceledto the
descriptionof nature as an automaton, thesetwo idealswerecontradictory.It is this contradictionthat we are
beginning to overcome.
IIIn SectionI,we emphasizedthat the problemsoftime and
determinismform the dividing linebetweenscienceand
philosophy,oralternatively,betweenC.P.Snow'stwo
cultures.21 But scienceis far from beinga monolithicbloc.Infact, the nineteenthcentury left us a doubleheritage:the
laws ofnature, such as Newton'slaw, whichdescribesatime-reversibleuniverse, and an evolutionary descriptionassociatedwith entropy.
Entropy is an essentialpart ofthermodynamics,thescience that dealsspecificallywith irreversible,time-orientedprocesses.Everyoneis to someextentfamiliar with theseprocesses.Think about radioactive decay, orabout
viscosity,whichslowsthe motionofa fluid. In contrast to time-
18 The End of Certainty
reversibleprocesses,such as the motionof a frictionless
pendulum,wherefuture and past play the same role(wecan interchangefuture, that is, +t, with past, \342\200\224t),
irreversible processeshave a directionin time.A radioactivesubstancepreparedin the pastwill disappearin the future.Becauseofviscosity, the liquid flow slowsover time.
The primordialroleofthe directionoftime is evidentin the processeswe study at the macroscopiclevel, such aschemicalreactionsor transport processes.We start with
chemicalcompoundsthat may react.As time goeson, theyreachequilibriumand the reactionstops.Similarly, if westart with an inhomogeneousstate, diffusion will tend to
homogenizethe system.Solarradiation is the result ofirreversible nuclearprocesses.No descriptionof the ecos-pherewould bepossiblewithout taking into accounttheinnumerableirreversibleprocessesthat determineweatherand climate.Nature involves both time-reversible and time-
irreversible processes,but it is fair to say that irreversible
processesare the rule and reversibleprocessestheexception. Reversibleprocessescorrespondto idealizations:We
have to ignorefriction to make the pendulummove re-versiblySuchidealizationsare problematicbecausethereis
no absolutevoid in nature. As previouslymentioned,time-reversibleprocessesare describedby equationsofmotion,whichare invariant with respectto time inversion,as is thecase in Newton's equation in classicalmechanicsorSchrodinger'sequationin quantum mechanics.Forirreversible processes,however, we needa descriptionthat
breakstime symmetry.The distinction between reversible and irreversible
processeswas introducedthrough the conceptofentropyassociatedwith the so-calledsecondlaw ofthermodynamics. Entropy had already beendefined by RudolfJulius
Epicurus'Dilemma 19
Clausiusin 1865(in Greek,entropysimply means
\"evolution\.22") Accordingto this law, irreversibleprocessesproduce entropy. In contrast, reversibleprocessesleave the
entropyconstant.We shall comeback repeatedlyto this secondlaw. For
now, let us recallClausius'scelebratedformulation: \"Theenergyofthe universe is constant.Theentropyoftheuniverse is increasing.\"This increasein entropy is due to theirreversibleprocessesthat take placein the universe.Clausius's statement was the first formulation ofan
evolutionaryview ofthe universe basedon the existenceofthese
processes.Arthur Stanley Eddington calledentropy the\"arrow of time.\"23 Nevertheless,accordingto thefundamental laws of physics, there should be no irreversible
processes.We thereforesee that we have inheritedtwo
conflictingviews ofnature from the nineteenthcentury:the time-reversibleview based on the laws ofdynamicsand the evolutionary view based on entropy. How cantheseconflictingviews bereconciled?After so many years,this problemis still with us.
For the Viennesephysicist Ludwig Boltzmann,thenineteenthcentury was the century ofCharlesDarwin,the man who defined life as the result ofa never-endingprocessofevolution and thus placedbecomingat the
center ofour understandingofnature. Still, for most
physicists, Boltzmannis now associatedwith a conclusionquiteoppositeto that ofDarwin; he is creditedwith havingshown that irreversibility is only an illusion.It wasBoltzmann's tragedy to have attempted in physicswhat Darwinhad accomplishedin biology\342\200\224only to reachan impasse.
At first glance,the similaritiesbetweenthe approachesofthesetwo giants ofthe nineteenthcentury are striking.Darwin showedthat if we start with the study ofpopula-
20 The End of Certainty
tions,and not individuals, we can understandhowindividual variability, subject to selectionpressure,producesadrift. Correspondingly,Boltzmannarguedthat we cannotunderstand the secondlaw of thermodynamics,and the
spontaneous increasein entropy it predicts,by startingwith individual dynamical trajectories;we must begininstead with largepopulationsofparticles.The increaseinentropy would be the globaldrift resultingfrom the
numerous collisionsbetweentheseparticles.In 1872,Boltzmannpublishedhis famous H-theorem,
whichincludedthe H-function,a microscopicanalogueofentropy24 This theoremtakes into accountthe effects ofcollisionsthat modify the velocitiesofparticlesat eachinstant. It showsthat collisionsbringthe distributionofvelocities ofthe populationofparticlescloserto equilibrium(the so-calledMaxwell-Boltzmanndistribution). As the
population approachesequilibrium,Boltzmann'sH-function decreasesand reachesits minimum value at
equilibrium;this minimum value means that collisionsno longer
modify the distributionofvelocities.ForBoltzmann,the
particlecollisionsare thus the mechanismthat leads the
systemto equilibrium.BothBoltzmannand Darwin replacedthe study of
\"individuals\" with the study ofpopulations,and showedthat
slight variations (the variability of individuals, ormicroscopic collisions)taking placeover a longperiodoftime
can generateevolution at a collectivelevel.(Inlater
chapters, we shall comeback to the roleofpopulations.)Exactly as biologicalevolution cannotbedefinedat the level
ofindividuals, the flow of time is also a globalproperty(seeChapters5 and 6).But whileDarwinattemptedto
explain the appearanceofnew species,Boltzmanndescribedan evolution toward equilibriumand uniformity. Signifi-
Epicurus'Dilemma 21
cantly, thesetwo theorieshave had very different fortunes.Darwin's theory of evolution, whichwas to triumph in
spite of fiercecontroversies,remains the basis for ourunderstandingoflife. On the otherhand, Boltzmann's
interpretation ofirreversibility succumbedto its critics,andhe was gradually forcedto retreat.He couldnot excludethe possibilityof \"antithermodynamical\" evolutions,as aresult of which entropy would diminish and inhomo-geneities,insteadofbeingleveled,would increasespontaneously.
Thesituation confronting Boltzmann was indeeddramatic. Hewas convincedthat in orderto understandnature we have to includeevolutionary features and that
irreversibility, as definedby the secondlaw ofthermodynamics, was a decisivestepin this direction.But hewas alsoheirto the grand tradition ofdynamics,and realizedthat it
stoodin the way of his attempt to give a microscopicmeaningto the arrow oftime.
Fromtoday'svantage point,Boltzmann'sneedto choosebetweenhis convictionthat physicshad to understand
becoming, and his loyalty to its traditional role,seemsparticularly poignant.The fact that his attempt would end in
failure now seemsself-evident.Every student learns that
a trajectory is time reversible,and thus allowsnodistinction betweenfuture and past. As HenriPoincarenoted,explainingirreversibility in terms of trajectoriesthat are
time-reversibleprocesses,however numerous,appears to
bea purelylogicalerror.25Supposethat we invert the signofthe velocity ofall molecules.Thesystemwould then gointo its own \"past.\" Even if entropywas increasingbeforevelocity inversion,it would now decrease.This wasJosephLoschmidt'svelocity-reversal paradox,whichwas thereason why Boltzmanncouldnot excludeantithermodynam-
22 The End of Certainty
ical behavior.When faced with severecriticism,hereplaced his microscopicinterpretation of the secondlawwith a probabilistic interpretationbaseduponourlackofinformation.
In a complexsystemformedby hugenumbersofmolecules (onthe orderof1023,orAvogadro's number),suchas a gas or liquid, it is obviousthat we are unable to
computethe behaviorofeachmolecule.Forthis reason,Boltz-
mann introducedthe assumptionthat all microscopicstatesofsucha systemhave the sameaprioriprobability.Thedifference would beassociatedwith the macroscopicstate, asdescribedby temperature,pressure,and otherparameters.Boltzmann defined the probabilityof eachmacroscopicstate by calculating the numberofmicroscopicstates that
give riseto it.Boltzmannwould have us imagine,for instance,a
volume dividedinto two equalcompartmentsthat
communicate with eachother.This volume containsa largenumberofmolecules,whichwe shall call N. Although we areunable to follow the path of each individual molecule,through measuringa macroscopicquantity, such as the
pressure in each compartment,we can determinethenumberof moleculesit contains.We can also preparea starting point,or \"initial state\" as it is generallyreferredto by physicists,whereoneof the two compartmentsisnearly empty.What can we expectto observe?Over the
courseoftime,moleculeswill populate the emptycompartment. Indeed,the greatmajority ofall possiblemicroscopic
states correspondsto a macroscopicsituation whereeachcompartmentcontains the same numberofmolecules. Thesestates correspondto equilibrium,or to
pressures that would beequal in the two compartments.Oncethis state has beenachieved,the moleculeswill continueto
Epicurus'Dilemma 23
passfrom onecompartmentto the other,but on average,the numberofmoleculesgoingto the right and left will beequal.Apart from slight transitory fluctuations, thenumber ofmoleculesin the two compartmentswill remainconstant over time,and equilibriumwill be preserved.However, there is a basic weaknessin this argument.Aspontaneouslong-termdeviation from equilibriumis not
impossible,even if it is, as Boltzmann concluded,\"improbable.\"
Boltzmann'sprobability-basedinterpretationmakesthe
macroscopiccharacterofour observationsresponsibleforthe irreversibility we observe.If we couldfollow theindividual motionof the molecules,we would seea time-reversiblesystemin whicheachmoleculefollows the lawsofNewtonianphysics.Becausewe can only describethenumberofmoleculesin eachcompartment,we concludethat the system evolves toward equilibrium.Accordingto this interpretation,irreversibility is not a basic law ofnature; it is merely a consequenceof the approximate,macroscopiccharacterofour observations.
Ernst Zermeloadded anothercriticismofBoltzmann's
argument to Loschmidt's reversal paradox26 in quotingPoincare'srecurrencetheorem,which shows that if wewereto wait longenough,we couldobservethespontaneous return ofa dynamical systemto a state ascloseto theinitial state as we might wish.As the physicistRomanSmoluchowskiconcluded,\"If we continuedourobservation for an immeasurablylong time,all processeswould
appear to be reversible.\"27This appliesdirectly toBoltzmann's two-compartmentmodel.After a sufficiently longtime,the initially empty compartmentwill again becomeempty. Irreversibility correspondsonly to an appearancethat is devoidofany fundamental significance.
24 TheEnd of Certainty
Let us now return to the situation discussedin SectionI.Throughourown approximations,we wouldberesponsible for the evolutionary characterof the universe. Inorderto makesuch an argument plausible,the first step in
assuring that irreversibility will be the result ofourapproximations is to view the consequencesof the secondlaw as trivial and self-evident.In his recentbook,The
Quark and the Jaguar, Murray Gell-Mannwrites,
The explanation [of irreversibility] is that there are more
ways for nails or pennies to be mixed up than sorted.Therearemoreways for peanut butter and jelly to contaminateeachother's containersthan there are to remain completelypure.And there are more ways for oxygenand nitrogen gasmolecules to be mixed up man segregated.To the extent that
chance is operating, it is likely that a closedsystem that has
someorder will move toward disorder,which offers so many
more possibilities.Howare thosepossibilitiesto be counted?An entire closedsystem, exactlydescribed,can existin a
variety of states,often calledmicrostates.In quantum
mechanics,theseare understoodto bepossiblequantum states of the
system. Thesemicrostatesare groupedinto categories(sometimes calledmacrostates)accordingto the various propertiesthat are beingdistinguished by coarsegraining. The microstatesin a given macrostateare then treated as equivalent, so that
only their number matters. . . .
Entropy and information are very closelyrelated.In fact,
entropy can be regardedas a measureof ignorance.When it
is known only that a system is in a given macrostate,the
entropy of the macrostatemeasuresthe degreeof ignorancethe
microstatesystem is in by counting the number of bits of
additional information neededto specifyit, with all the
microstates in the macrostatetreatedas equally probable.28
Epicurus'Dilemma 25
Similarargumentscan be found in most booksdealingwith the arrow oftime.We believe that theseargumentsare untenable.They imply that it is our own ignorance,our coarsegraining,that leads to the secondlaw. Forawell-informedobserver,such as the demonimaginedby
Laplace, the world would appear as perfectly timereversible. We wouldbethe father oftime,ofevolution, andnot its children.Irreversibility subsists,whatever theprecision ofourexperiments.This meansthat attributing thesepropertiesto incompleteinformation can hardly beconsidered. It is interestingto note that Max Planckhad
already opposedthe idea of incompleteinformation todescribethe secondlaw. In his Treatiseon Thermodynamics
he wrote,
It would be absurd to assumethat the validity of the secondlaw dependsin any way on the skill of the physicist or
chemist in observingor experimenting.The gist of the
second law has nothing to do with experiment; the law asserts
briefly that there exists in nature a quantity which always changes
in the same way in all natural processes.The propositionstatedin
this generalform may be correct or incorrect;but whicheverit may be, it will remain so, irrespectiveof whether thinking
and measuring beingsexist on the earth or not, and whether
or not, assuming they do exist, they are able to measurethe
details of physical or chemicalprocessesmore accuratelyby
one,two, or a hundred decimalplacesthan we can.The
limitation of the law, if any, must lie in the sameprovinceas its
essentialidea, in the observedNature, and not in the
Observer. That man's experienceis calledupon in the deductionof the law is of no consequence;for that is, in fact, our only
way of arriving at a knowledgeof natural law.29
26 TheEnd of Certainty
However,Planck'sviews wereto remain isolated.As wehave indicated,most scientistsconsideredthe secondlawthe result ofapproximations,or the intrusionofsubjectiveviews into the exactworldofphysics.In a celebratedstatement, Max Bornassertedthat \"irreversibility is the effectof the introductionof ignoranceinto the basic laws ofphysics.\"30
Our own point ofview is that the laws ofphysics,asformulated in the traditional way, describean idealized,stableworld that is quite different from the unstable,evolving
world in whichwe live.Themain reasonto discardthebanalization ofirreversibility is that we can no longerassociate the arrow oftime only with an increasein disorder.Recent developments in nonequilibriumphysics and
chemistrypoint in the oppositedirection.They show
unambiguously that the arrow of time is a sourceof order.
This is already clearin simpleexperimentssuchas thermal
diffusion, which has beenknown since the nineteenthcentury. Let us considera boxcontainingtwo components(such as hydrogen and nitrogen)where we heat oneboundary and coolthe other(seeFigure1.1).Thesystemevolves to a steady state in whichonecomponentisenriched in the hot part and the other in the coldpart. Theentropyproducedby the irreversibleheat flow leadsto an
orderingprocess,whichwould be impossibleif taken
independently from the heat flow. Irreversibility leads toboth orderand disorder.
The constructive roleof irreversibility is even morestriking in far-from-equilibriumsituations where non-equilibriumleads to new forms ofcoherence.(We shall
comeback to nonequilibriumphysicsin Chapter2.)We
have now learnedthat it is preciselythrough irreversible
processesassociatedwith the arrow of time that nature
achievesits most delicateand complexstructures.Life is
Epicurus'Dilemma 27
Figure1.1ThermalDiffusionAs a result of the differencein temperature betweenthe two
containers, the black moleculeshave a higher concentrationin the left
compartment. Thiscorrespondsto thermal diffusion.
T\\ T2
O\342\200\242
\342\200\242
Oo o
\342\200\242 oo # oo o
possibleonly in a nonequilibriumuniverse.Nonequilib-rium leadsto conceptssuch as self-organizationand dissi-
pative structures,whichwill be describedin moredetailin Chapter2.In From Being to Becoming, we had alreadyformulated the following conclusionsbasedon theremarkable developmentsin nonequilibriumphysicsand
chemistryover the past several decades:
\342\200\242 Irreversible processes(associatedwith the arrow oftime)are as real as reversibleprocessesdescribedby thefundamental laws ofphysics;they do not correspondtoapproximationsadded to the basiclaws.
\342\200\242 Irreversibleprocessesplay a fundamental constructive
rolein nature.31
What impact do theseconceptshave on currentthinkingabout dynamical systems?Boltzmann was well aware that
nothinganalogousto irreversibility exists in classicaldynamics; he therefore concludedthat irreversibility can bederivedonly from assumptionsabout the initial conditions
O
o\342\200\242
28 The End of Certainty
in the early stagesofour universe.We can keepour usual
formulations of dynamics,but we need to supplementthemwith appropriateinitial conditions.In this view, the
originaluniverse was highly organized,and therefore in an
improbablestate\342\200\224a suggestionstill acceptedin a numberofrecentbooks.32Theinitial conditionsprevailing in ouruniverse leadto interestingand largely unsolvedproblems(seeChapter8),but we believethat Boltzmann'sargument is nolongerdefensible.Whatever the past, thereexistat presenttwo typesofprocesses:time-reversibleprocesses,wherethe
applicationofexistingdynamicshas provedto besuccessful
(i.e.,the motionofthe moonin classicalmechanics,orthe
hydrogen atom in quantum mechanics),and irreversible
processeslike heat conditions,where the asymmetrybetween past and future is obvious.Our objectiveis to devisea new formulation ofphysicsthat explains,independentlyofany cosmologicalconsiderations,the difference betweenthesebehaviors.This can indeedbe achievedfor unstableand thermodynamic systems.We can overcomewhat lookedlike an apparent contradictionbetweenthe time-reversiblelaws of dynamics and the evolutionary view of naturebasedon entropy. But let us not getaheadofourselves.
Nearly two hundred years ago,Joseph-LouisLagrangedescribedanalytical mechanicsbasedon Newton'slaws asa branch ofmathematics.33 In the Frenchscientificliterature, one often speaks of \"rational mechanics.\"In this
sense,Newton'slaws would define the laws ofreasonand
representa truth ofabsolutegenerality. Sincethe birth ofquantum mechanicsand relativity, we know that this is not
the case.Thetemptation is now strong to ascribea similar
status ofabsolute truth to quantum theory. In The Quarkand the Jaguar, Gell-Mannasserts,\"Quantummechanicsis
Epicurus'Dilemma 29
not itselfa theory; rather it is the framework into whichall
contemporaryphysicaltheory must fit.\"34 Is this really so?As statedby my late friend LeonRosenfeld,\"Every theoryis basedonphysicalconceptsexpressedthroughmathematical idealizations.They are introducedto give an adequaterepresentationof the physical phenomena.No physical
conceptis sufficiently defined without the knowledge of its domain
ofvalidity\"35
It is this \"domainofvalidity\" requiredfor the basic
conceptsofphysics,such as trajectoriesin classicalmechanics
orwave functions in quantum theory, that we are
beginningto delineate.Theselimits are associatedwith
instabilityand chaos,whichwe shall briefly introducein the next
section.Oncewe includetheseconcepts,we cometo a
new formulation of the laws of nature, one that is nolongerbuilt on certitudes,as is the casefor
deterministic laws, but rather on possibilities.Moreover,in this
probabilisticformulation, time symmetry is destroyed.Theevolutionary characterofthe universe has to bereflectedwithin the contextof the fundamental laws of physics.Rememberthe idealofthe intelligibility ofnature asformulated by Whitehead (seeSectionI): Every elementofour experiencehas to beincludedin a coherentsystemofgeneralideas.Basedon this rewritingofthe lawsofnature,we can now completethe work pioneeredby Boltzmannmorethan a century ago.
It is interestingthat great mathematicians, suchas EmileBorel,alsounderstoodthe needto overcomedeterminism.Borelnotedthat considerationsofisolatedsystems,suchas
moon-earth,are always idealizations,and that determinismmay fail whenwe leave this reductionistview.36That is
indeed what our own researchshows.
30 TheEnd of Certainty
III
Everyoneis to someextentfamiliar with the differencebetween stableand unstable systems.Considera pendulum,for example.Supposethat it is originally at equilibrium,wherethe potentialenergyis at a minimum.If a smallperturbation is followedby a return to equilibrium(seeFigure1.2),this systemrepresentsa stableequilibrium.In contrast,if we put a pencilon its head, the smallestperturbationwill causeit to fall to the left orright,giving us a modelofunstable equilibrium.
Thereis a basicdistinctionbetweenstableand unstablemotions.In short, stable dynamical systemsare those in
whichslight changesin the initial conditionsproducecorrespondingly slight effects.But for a largeclassofdynamical systems,small perturbationsin the initial conditionsare
amplified over the courseoftime.Chaoticsystemsare an
extremeexampleofunstable motionbecausetrajectoriesidentified by distinct initial conditions,no matter howclose,diverge exponentiallyover time.This is known as
\"sensitivity to initial conditions.\"A classicillustration ofamplification through chaosis the \"butterfly effect\"; by
Figure1.2Stableand UnstableEquilibrium
kpot kpot
Stable
Unstable
Epicurus'Dilemma 31
just flapping its wings, a butterfly in Amazonia may affectthe weather in the UnitedStates.We shall seeexamplesofchaoticsystemslater on in Chapters3 and 4.
Thetermdeterministic chaoshas also enteredthediscussion ofchaoticsystems.Indeed,the equationsofmotionremain deterministic,as is the casein Newtoniandynamics,even if a particular outcomeappearsto be random.Thediscoveryofthe important roleofinstability has led to arevival ofclassicaldynamics,previously considereda closedsubject.In fact, until recentlyit was thought that all systemsdescribedby Newton'slaws are alike.Ofcourse,everyoneknew that the trajectory of a falling stone was easiertosolve than a \"three-bodyproblem,\"suchas the oneinvolving
the sun, Earth, and Jupiter.But this was consideredtobemerelya questionofcomputation.It was only at the endof the nineteenthcentury that Poincareshowed that this
was not the case.Theproblemsare fundamentally different
dependingon whetherornot a dynamical systemis stable.We have mentionedchaoticsystems,but thereare other
typesofinstability to beconsidered.Let us first describeinqualitative terms in what senseinstability leads to an
extension ofthe laws ofdynamics.In classicaldynamics,the
initial state is determinedby the positionsq and velocitiesv (ormomentump).*Oncetheseare known, we can
determine the trajectory by using Newton'slaws (or anyother equivalent formulation ofdynamics).We can then
representthe dynamical state by a point q0,p0 in a spaceformedby the coordinatesand momenta.This is known asthe phasespace(Figure1.3).Insteadofexamininga singlesystem,we can also study a collectionofsystems\342\200\224an \"en-
*For the purpose of simplification, we have used a single letter even when we
are discussing a system formed by many particles.
32 TheEnd of Certainty
Figure1.3Trajectory in PhaseSpaceThe dynamical state is represented by a point in the phase spaceq, p.The time evolution is describedby a trajectory starting at the initial
point q0, p0.
Po'lo
semble,\"as it has beencalledsincethe pioneeringwork ofAlbert EinsteinandJosiahWillard Gibbsat the beginningofthis century.
At this point, it would be helpful to reproducepart ofGibbs'sfamous prefaceto his Elementary PrinciplesinStatistical Mechanics:
We may imagine a great number of systems of the same
nature, but differing in the configurations and velocitieswhich
they have at a given instant, and differing not merely infini-
tesimally, but it may be so as to embrace every conceivablecombination of configurations and velocities.And here we
may set the problem, not to follow a particular system
through its successionof configurations, but to determine
how the whole number of systems will be distributed amongthe various conceivableconfigurations and velocitiesat any
requiredtime, when the distribution has beengiven for someone time. . . .
Epicurus' Dilemma 33
The laws of thermodynamics,as empirically determined,
expressthe approximate and probable behaviorof systems of a
great number of particles,or, more precisely,they expressthe
laws of mechanicsfor such systems as they appear to beingswho have not the finenessof perception to appreciatequantities of the order of magnitude of those 'which relate to
single particles,and who cannot repeat their experimentsoften
enough to obtain any but the most probableresults,37
Gibbs introducedpopulation dynamicsinto physicsby
using an ensembleapproach.An ensembleis representedby a cloudofpoints in phasespace(seeFigure1.4).Thecloudis describedby a function p(q,p,t),whichhas a
simple physicalinterpretation:the probability offinding at time
t, a point in the small regionofphase spacearound the
point q,p. A trajectory correspondsto a specialcasein
whichp is vanishing everywhereexceptat the point q0,p0-This situation is describedby a specialform ofp.Functions that have the property ofvanishing everywhereexcept at a singlepoint are calledDiracdelta functions d(x).The function d(x \342\200\224
xQ) is vanishing for all points x 5* xQ.Therefore,for a singletrajectoryat time zero,the
distribution function p takesthe form p= 8(q-q0)${p\342\200\224
p0)-*We
shall comeback to the propertiesofdelta functions later.
*When we take x= x0, the function 8(x-xQ) diverges to infinity. The8-func-
tion therefore has \"abnormal\" properties as compared to a continuous function
such as xor sin x. It is called a generalized junction or distribution (not to beconfused with probability distribution p). Generalized functions are used in
conjunction with test functions <f(x), which are continuous functions (i.e.,J dx(p(x)B(x- x())
=<p(x0)). Also note that at time t we have for a free particle
moving at the speedm the probability p = 8(p-p0)B(q-
qQ-
m ), as the
momentum remains constant and the coordinate varies linearly with time.
34 The End of Certainty
Figure1.4Ensemblesin PhaseSpaceGibbs'sensembleis representedby a cloudof particlesdiffering
according to their initial conditions.The shapeof the cloudchangesover time.
As was clearly statedby Gibbs, however,the ensembleapproachwas merelya convenientcomputationaltoolforhim whenexactinitial conditionswerenot available.In his
opinion,probabilitiesexpressignorance,or lack ofinformation. Moreover,it has always beenacceptedthat fromthe dynamical point ofview, individual trajectoriesand
probabilitydistributionspresent equivalent problems.We
can start with individual trajectoriesand then derive theevolution of probability functions, and vice versa. Theprobabilitypcorrespondssimplyto a superpositionoftrajectories, and leadsto no new properties.Thetwo levelsofdescription,the individual level (correspondingto singletrajectories)and the statistical level (correspondingtoensembles),would beequivalent.
Is this always the case?Forsimplestablesystems,wherewe donot expectany irreversibility, this is indeedtrue.Gibbs and Einstein were right. The individual point ofview (in terms oftrajectories)and the statistical point ofview (in terms ofprobabilities)are then equivalent.This
Epicurus'Dilemma 35
canbeeasilyverified, and we shall comeback to this pointin Chapter5.However, is this also true for unstablesystems? Howis it that all theoriesdealingwith irreversible
processeson the molecularlevel, such as Boltzmann'skinetictheory, involve probabilitiesand not trajectories?Is this again becauseof our approximations,our coarsegraining? Howcan we then explainthe successofkinetictheory, the quantitative predictionsofmany propertiesofdilutegases,suchas thermal conductivity and diffusion, all
ofwhichhave beenverified by experimentation?Poincarewas so impressedby the successofkinetic
theorythat hewrote,\"Perhapsthe kinetictheory ofgaseswill
serveas a model.. . Physical laws will then take on a
completely new form; they will take on a statistical character.\"3*
Thesewereindeedpropheticwords.In an extraordinarilydaring move, Boltzmannintroducedprobabilityas an
empirical tool.Now, morethan onehundred years later, weare beginningto understand how probabilisticconceptsemergewhenwe go from dynamicsto thermodynamics.Instability destroysthe equivalencebetweenthe individual
and statistical levels ofdescription.Probabilitiesthen
acquire an intrinsical dynamical meaning.This knowledgehas ledto a new kind ofphysics,the physicsofpopulations, whichis the basicsubjectofthis book.
By way ofexplanation,let us considera simplifiedexample ofchaos.Supposethat we have two typesofmotiondenotedas + or \342\200\224 (i.e.,motion\"up\" or \"down\") within
the phase spaceillustrated in Figure1.4.This leadsus tothe two typesofsituations representedby Figures1.5and1.6.In Figure1.5,thereare two different regionsin phasespace,onecorrespondingto the motion
\342\200\224,
the otherto themotion+.If we discardthe regioncloseto the boundary,each\342\200\224 is surroundedby \342\200\224,
and each+ by +.This corre-
36 The End of Certainty
Figure1.5StableDynamicalSystemThe motions denotedas + or-lie in distinct regionsofphasespace.
spondsto a stablesystem.Slightchangesin the initial
conditions do not alter the result.In Figure1.6,instead,each+ is surroundedby \342\200\224,
andvice versa. Theslightestchangein initial conditionsisamplified, and the systemis thereforeunstable.A primary re-
Figure1.6UnstableDynamicalSystemEach motion + is surrounded by
-and viceversa.
Epicurus'Dilemma 37
suit ofthis instability is that trajectoriesnow becomeidealizations. We can no longerprepare a singletrajectory, asthis would imply infinite precision.Forstablesystems,this
is without significance,but for unstable systems,with their
sensitivity to initial conditions,we can only prepareprobability distributions,includingvarious typesofmotion.
Is this difficulty merelya practicalone?Yes,if weconsider that trajectorieshave now becomeuncomputable.But there is more:Probabilitydistribution permits us to
incorporatewithin the framework of the dynamical
description the complexmicrostructureofthe phasespace.Itthereforecontainsadditional information that is lackingat
the level ofindividual trajectories.As we shall seein
Chapter 4, this has fundamental consequences.At the level ofdistributionfunctions p,we obtain a new dynamicaldescription that permits us to predict the future evolution ofthe ensemble,includingcharacteristictime scales.This is
impossible at the level of individual trajectories.Theequivalencebetweenthe individual and statistical levels is
indeedbroken.We obtain new solutionsfor the
probabilitydistributionp that are irreducible becausethey do not
apply to singletrajectories.The laws ofchaoshave to beformulated at the statisticallevel.That is what we meant in
the precedingsectionwhenwe spokeabout ageneralization ofdynamicsthat cannotbeexpressedin terms oftrajectories. This leads to a situation that has never beenencounteredin the past.Theinitial conditionis no longera point in the phasespacebut someregiondescribedby pat the initial time (= zero.We thus have a nonlocal
description. Thereare still trajectories,but they are the outcomeofa stochastic,probabilisticprocess.No matter how
precisely matchedour initial conditionsare, we obtaindifferent trajectoriesfrom them.Moreover,as we shall see,time
38 The End of Certainty
symmetry is broken,as past and future play different rolesin the statistical formulation. Ofcourse,for stable systems,we revert to the usual descriptionin terms ofdeterministic trajectories.
Why has it taken so longto arrive at a generalizationofthe laws ofnature that includesirreversibility and
probability? One of the reasons is ideological\342\200\224the desiretoachieve a quasi-divinepoint ofview in our descriptionofnature. But therehas also beena technical,mathematical
probleminvolved. Our work is basedon recentprogressinfunctional analysis,a field ofmathematics that has cometothe forefront only in recentdecades.As we shall see,ourformulation requiresan extendedfunctional space.Thisnew field ofmathematics,which uses generalizedfunctions orfractals, as BenoitMandelbrotcalledthem,is now
playing a criticalrolein the understandingofthe laws ofnature.39We needa \"divine\" point ofview to retain theidea of determinism.But no human measurements,notheoreticalpredictions,can give us initial conditionswith
infinite precision.It is interestingto contemplatewhat becomesof the
Laplacedemonin the world ofdeterministicchaos.Hecan no longerpredictthe future unlessheknowsthe initial
conditionswith infinite precision.Only then can hecontinue to use a trajectory description.But there is an evenmorepowerful instability that leads to the destructionoftrajectories,whatever the precisionof the initial description.This form ofinstability is offundamental importancebecause it appliesto both classicaland quantum mechanics.
Our story actually beginsat the end ofthe nineteenth
century with the work ofJules-HenriPoincare.Accordingto Poincare,a dynamical systemis characterizedin termsofthe kineticenergyofits particlesplus the potentialen-
Epicurus'Dilemma 39
ergy dueto their interaction.40A simpleexamplewouldbefree, noninteractingparticles,wherethere is no potentialenergy,and the calculationoftrajectoriesis trivial. Suchsystemsare by definition integrable.Poincarethen askedthe question:Are all systemsintegrable?Can we choosesuitable variables to eliminatepotential energy?Byshowing
that this was generallyimpossible,he proved that
dynamical systemswerelargely nonintegrable.
It is worthwhile to pause for a momentand reflect onPoincares conclusions.Supposehe had proved that all
dynamical systemsare integrable.This would mean that all
dynamical motionsare isomorphicto free noninteractingparticles.Therewould beno placefor the arrow oftime,for self-organization,or life itself. Integrablesystemsdescribea static, deterministicworld.Poincarenot onlydemonstratednonintegrability, but alsoidentifiedthe
reason for it: the existenceofresonancesbetween the degreesoffreedom. As we shall seein greaterdetail in Chapter5,thereis
a frequency that correspondsto eachmodeofmotion.Thesimplestexampleofthis is the harmonicoscillator,inwhicha particleand central point are given.Theparticleis
heldby a forceproportionalto its distancefrom that point.If we displacethe particlefrom the center,it will oscillatewith a well-definedfrequency.It is through thesefrequencies that we arrive at the notionof resonance,which is
crucialto Poincare'stheorem.We are all moreorlessfamiliar with the conceptof
resonance. When we forcea spring to deviate from its
equilibrium position,it vibrates with a characteristicfrequency.Now let us subject this spring to an externalforcewith a
frequency that can bevaried.When the two frequencies,that of the spring and that of the externalforce,have a
simplenumericalratio (that is, whenoneofthe frequen-
40 The End of Certainty
ciesis either equal to the other, or two, three,four . . .
times larger),the amplitudeof the motionofthe springincreasesdramatically. The same phenomenon occurswhenwe play a noteon a musicalinstrument.We heartheharmonics.Resonance\"couples\"sounds.
Nowconsiderthe caseofa systemcharacterizedby two
frequencies.By definition,wheneverthe sum n10)1+n2<CQ2
= 0,wheren, and \302\253, are nonvanishing integers,we have9h \342\200\224 _ Hiresonance.This means that \302\2532
\342\200\224
\302\273i;the ratio ofthe
frequencies is then a rational number.As Poincarehas shown,in dynamicsresonanceslead to terms with \"dangerous\"denominatorssuch as
\342\200\236,,,
.\342\200\236 ,\342\200\236
Whenever thereare res-11 \"2 2\"
onances(i.e.,points in phase spacewheren1a>1 + n2a>2=
zero),theseterms diverge.As a result,we encounterobstacles wheneverwe try to calculatetrajectories.
This is the origin of Poincare'snonintegrability. The\"problemof small denominators\"was already known by
eighteenth-centuryastronomers,but Poincare'stheoremshowed that this difficulty, whichhe called the \"generalproblemofdynamics,\" is sharedby the great majority ofdynamical systems.For a considerablelengthoftime,however, the importanceofPoincare'sfindings was overlooked.
Max Bornwrote, \"It would indeedbe remarkable if
Nature had fortified herselfagainst further advances in
knowledgebehindthe analytical difficulties ofthe many-
bodyproblem.\"41It was hard to believethat a technicaldifficulty (divergencesdue to resonances)could alter the
conceptualstructureofdynamics.We now seethis
problem in a different way. Forus,Poincare'sdivergencesare an
opportunity.Indeed,we can go beyond his negativestatement and show that nonintegrabilitypaves the way, as doeschaos,for a new statistical formulation ofthe laws ofdy-
Epicurus'Dilemma 41
namics.It tooksixty years after Poincare,through the workofAndrei N.Kolmogorov,continuedby Vladimir Igore-vich Arnold andJiirgenKurt Moser(the so-calledKAM
theory), for nonintegrabilityto be understoodnot as the
frustrating manifestation ofsomeresistanceofnature againstthe advances ofknowledge,to paraphraseBorn,but as anewstarting point for dynamics.42
The KAM theory deals with the influenceofresonances on trajectories.ThefrequenciesCO dependin
general on the values ofdynamic variables suchas coordinatesand momenta.They thereforetake on different values at
different points in the phasespace.Theresult is that somepointswill becharacterizedby resonances,and otherswill
not.Again, for chaosthis leads to an extraordinarycomplexity in the phasespace.Accordingto the KAM theory,we observetwo typesoftrajectories:\"nice\" deterministictrajectoriesand \"random\" trajectoriesassociatedwith
resonances, which wander erratically through regionsofphasespace.
Another important result ofthis theory is that whenweincreasethe value ofenergy,we increasethe regionswhererandomness prevails. Forsomecriticalvalue of energy,chaosappears:over time we observe the exponentialdivergence of neighboringtrajectories.Furthermore,for
fully developedchaos,the cloudofpoints generatedbya trajectory leads to diffusion. But diffusion is associatedwith the approachto uniformity in ourfuture. It is an
irreversible processthat createsentropy (seeSectionI).Although we started with classicaldynamics,we can nowobservethe breakingoftime symmetryHow this is
possible is the main problemwe have to solve in ordertoovercome the timeparadox.
42 The End of Certainty
Poincareresonancesplay a fundamental rolein physics.Emissionor absorption of light is due to resonances,asis the approachto equilibrium in a systemof interactingparticles.Interactingfields again lead to resonances.It is
difficult to identify an important problemin classicalorquantum physicswhereresonancesdo not play a significantpart.But how can we overcomethe divergencesassociatedwith resonances?Heresomeessentialprogresshas beenmade.As in SectionIII,we have to distinguishtheindividual level (trajectories)from the statisticallevel (ensembles,as
describedby the probabilitydistributionp).At theindividual level we have divergences,but thesecan be solvedat
a statistical level (seeChapters 5 and 6), whereresonances producea couplingofevents looselyanalogousto the
couplingofsoundsby resonance.This leadsto new, non-Newtonianterms that are incompatible with a trajectory
description and instead requirea statistical,probabilisticdescription.This is not astonishing.Resonancesare not localevents,inasmuch as they do not occurat a given point or instant.
They imply a nonlocaldescriptionand therefore cannot beincludedin the trajectory descriptionassociatedwith
Newtonian dynamics.As we shall see,they lead to diffusive
motion. When we start at a point PQin phasespace,we can nolongerpredictwith certainty its positionPx after a time
period X. In short, the initial point PQleadsto many possiblepoints\342\200\224Pv P2, P3\342\200\224with
well-definedprobabilities.In Figure1.7,eachpoint in the domainD has a nonva-
nishingorwell-definedtransition probabilityofappearingat time X. This situation is similar to that ofthe \"random
walk,\" or \"Brownian motion.\"In the simplestcase,this
conditionmay beillustrated by a particleon a
one-dimensional lattice that makesa one-steptransition at regular time
intervals (seeFigure1.8).
Epicurus'Dilemma 43
Figure1.7Diffusive MotionAfter a time t, the system may producea result at any point, suchas Pj, P2, P3, in the domain D.
At every step, the probabilityis i that the particlewill
go to the left and 2 that it will go to the right.At every
step, the future is uncertain.Fromthe very beginning,itis impossible to speak of trajectories.Mathematically,Brownianmotionis describedby diffusion-type equations(the so-calledFokker-Planckequations).Sincediffusion is
Figure1.8A RandomWalk
Brownian motion on a one-dimensionallattice.At every step,the
probability is\\
that the particle will go to the left and\\
that it will
go to the right.
1 12 2
44 TheEnd of Certainty
time oriented,if we start with a cloudofpoints, all ofwhichare situatedat the sameorigin,as time goeson thecloudwill disperse.Someparticleswill be found fartherfrom the origin,otherscloser.It is quite remarkablethat,
starting with classicaldynamics,resonanceslead preciselyto diffusive terms,which is to say, resonancesintroduceuncertainty even within the framework of classicalmechanics, and so break time symmetry.
Forintegrablesystems,when thesediffusive
contributions are absent,we comeback to the trajectorydescription, but in generalthe laws of dynamics have to beformulated at the level of probabilitydistributions.Thebasicquestion is therefore:In whichsituations can we
expect the diffusive terms to beobservable?When this is so,probabilitybecomesa basicpropertyofnature. Thisquestion, whichinvolves defining the limits ofthe validity ofNewtoniandynamics(or the validity ofquantum theory,whichwe shall considerin the next section),is nothingshort of revolutionary. For centuries,trajectorieshavebeenconsideredthe basic, primitive objectof classical
physics.In contrast,we now considerthemoflimited
validity for resonant systems.We shall return repeatedlytothis question in Chapter5, and to a parallel question for
quantum mechanicsin Chapter6.Forthe moment,however, let us presentsomeprovisional answers.Fortransient
interactions(a beamofparticlescollideswith an obstacleand escapes),diffusive terms are negligible.But forpersistent interactions(a steady flow ofparticlesfalls onto the
obstacle),they becomedominant.In computersimulations, as in the real world, we can reproducebothsituations and thereforetest our predictions.Theresultsshow
unambiguouslythe appearanceofdiffusive terms for per-
Epicurus'Dilemma 45
sistent interactions,and thereforethe breakdown of the
Newtonian,as wellas the orthodox,quantum mechanicaldescriptions.In both thesecases,we obtain \"irreducible\"
probabilisticdescriptions,as in deterministicchaos.But thereis yet anothersituation that is even more
remarkable. Macroscopicsystemsare generally defined in
terms ofthe thermodynamic limit, accordingto whichboththe numberN of particles and the volume V becomelarge.We shall study this limit in Chapters 5 and 6.In theobservationofphenomenaassociatedwith this limit, thenew propertiesofmatter becomeobvious.
As longas we considermerelya few particles,wecannot say if they form a liquidorgas.Statesofmatter as wellas phase transitions are ultimately definedby the
thermodynamiclimit.The existenceofphase transitions shows
that we have to becareful whenwe adopta reductionistattitude. Phase transitions correspondto emergingproperties. They are meaningful only at the level ofpopulations,and not ofsingleparticles.This contentionis somewhat
analogousto that whichis based on Poincareresonances.Persistentinteractionsmeanthat we cannottake a part ofthe systemand considerit in isolation.It is at this globallevel, at the level ofpopulations, that the symmetrybetween past and future is broken,and sciencecan recognizethe flow oftime.This solvesa long-standingpuzzle.It is
indeedin macroscopicphysicsthat irreversibility and
probabilityare the most conspicuous.
Thermodynamics applies to non-integrablesystems.This means that we cannot solve the dynamical problemin terms of trajectories,but we can solve it in terms ofprobabilities.Therefore,as is the casefor deterministicchaos,the new statisticalformulation ofclassicalmechanics
46 The End of Certainty
leadsto an extensionofthe mathematical framework.Tosomeextent,this isreminiscentofgeneralrelativity. As
Einstein showed,we have to move from Euclideangeometryto Riemanniangeometryto includegravitation. Infunctional calculus,a special roleis played by the so-calledHilbertspace,which extends Euclideangeometry tosituations involving an infinite number of dimensions(the \"function space\.")Traditionally, quantum mechanicsand statistical mechanicshave utilized Hilbertspace.Toobtain our new formulation, which is valid for unstable
systemsand the thermodynamic limit, we have to movefrom Hilbertspaceto moregeneralfunctional spaces.Thisobservation will be explainedin detail in Chapters 4
through 6.Sincethe beginningof this century, we have become
used to the ideathat classicalmechanicshas to beextendedwhen we considermicroscopicobjects,such as atoms orelementaryparticles,or when we deal with astrophysicaldimensions.Surprisingly,instability alsorequiresan
extension ofclassicalmechanics.Thesituation in quantummechanics, to whichwe now turn, is quite similar.Instabilitydriven by resonancesplays a fundamental rolein changingthe formulation ofquantum theory.
IVIn quantum mechanics,we encountera rather strangesituation. As is well known, this theory has beenremarkablysuccessfulin all its predictions.Still,morethan sixty yearsafter its formulation, discussionsabout its meaningand
scopeare as heatedas ever.This is uniquein the history ofscience.43In spiteofall its successes,most physicistsshare
Epicurus'Dilemma 47
somefeelingof uneasiness.RichardFeynman onceremarked that nobody really \"understands\"quantum theory.
Here,the basicquantity is the wave function*\302\245,
whichplays somewhat the roleofthe trajectory in classicalmechanics. Indeed,the fundamental equationof quantumtheory, the Schrodingerequation,describesthe timeevolution ofthe wave function.It transforms the wavefunction ^(tf), as given at the initial time
tQ, into the wavefunction
*\302\245(t)at time t, exactlyas trajectoriesin classical
mechanicslead from onephasepoint to another.Like Newton'sequation,Schrodinger'sequationis
deterministic and time reversible. Again, as in classical
dynamics,thereappearsa gap betweenthe dynamicaldescription of quantum mechanicsand the evolutionary
descriptionassociatedwith entropy.Thephysicalinterpretation ofthe wave function *\302\245 is that ofa probability
amplitude. This implies that the square | 4*12 = W* (4* hasboth a real and imaginary part; 4** is the complexconjugate
of4*) is a probability, whichwe shall again denoteby
p.Thereare moregeneralforms of probabilitycorresponding to ensemblesobtained by the superimposingofvarious wave functions.Theseare calledmixtures,as
opposed to pure casesthat obtain from a singlewave
function.
Thebasicassumptionofquantum theory is that everydynamical problemcan be solvedat the level ofprobability amplitudes exactly as every dynamical problemin
classical mechanicswas traditionally associatedwith trajectorydynamics.But strangely, in orderto attribute well-definedpropertiesto matter, we have to gobeyondprobabilityamplitudes; we needprobabilitiesthemselves.To understandthis difficulty, let us considera simple example.Suppose
48 The End of Certainty
that energycan take on two values, Ex and Er Thecorresponding wave function is ux or uT Nowconsiderthelinear superimposition*\302\245
= ciui + c2u2.Thewave functionthen \"participates\"at both levels.Thesystemis neitherat
level 1nor level 2,but rather in a kind of intermediatestate.Let us now measurethe energy associatedwith 4*.
Accordingto quantum mechanics,we then find eitherExorE2 with probabilitiesgiven by the squaresofthe
probability amplitudes I q I
2 and I c212.Initially we startedwith a singlewave function 4*, but
we still end up with a mixtureoftwo wave functions, ux
and wr This is often calledthe \"reduction,\"or \"collapse,\"ofthe wave function.We needto move from potentialitiesdescribedby the wave function *\302\245 to actualities that we canmeasure.In the traditional languageofquantum theory,we move from a pure state (the wave function) to an
ensemble, ormixture.But how is this possible?As mentionedearlier,Schrodinger'sequationtransforms a wave functioninto anotherwave function, and not into an ensemble.This has often beencalledthe quantum paradox.It has beensuggestedthat the transition from potentiality to actualityis dueto ourown measurements.This is the point ofview
expressedby StevenWeinbergin SectionI ofthis chapterand in a considerablenumberoftextbooks.It is the same
type ofexplanationas waspresentedfor the time paradoxin classicalmechanics.In that caseas well, it is difficult tounderstand how a human action, such as observation,couldbe made responsiblefor the transition frompotentialities to actualities.Would the evolution ofthe universe
be different in the absenceofhumankind? In his
Introduction to The New Physics:ASynthesis, Paul C.W Davieswrites:
Epicurus'Dilemma 49
At rockbottom, quantum mechanicsprovidesa highly
successful procedurefor predictingthe results of observationsof
microsystems,but when we ask what actually happenswhenan observation takes place, we get nonsense!Attempts to
breakout of this paradoxrange from the bizarre,such as the
many universes interpretation of Hugh Everett, to the
mystical ideasofJohn von Neumann and EugeneWigner, whoinvoke the observer'sconsciousness.After half a century of
argument, the quantum observationdebate remains as lively
as ever.The problemsof the physics of the very small and the
very largeareformidable,but it may be that this frontier\342\200\224the
interfaceof mind and matter\342\200\224will turn out to be the most
challenginglegacyof the NewPhysics.44
This \"interfacebetweenmind and matter\" is also at thecoreofthe time paradox.Ifthe arrow oftime existedonlybecauseour human consciousnessinterferedwith a worldotherwiseruledby time-symmetricallaws, the veryacquisition ofknowledgewould becomeparadoxical,sinceany
measure already implies an irreversible process.If we wish tolearn anything at all about a time-reversibleobject,wecannotavoid the irreversibleprocessesinvolved in
measurement, whetherat the level ofan apparatus or ofourown sensorymechanisms.Thus,in classicalphysics,whenwe askhow we can understand\"observation\"in terms offundamental time-reversiblelaws, we get \"nonsense,\"asDaviesputs it.In classicalphysics,this intrusion ofirreversibility was perceivedas a minor problem.The greatsuccessofclassicaldynamicsleft no doubt about its
objective character.The situation is quite different in quantumtheory.Herethe needto includemeasurementin ourfundamental descriptionofnature is explicitlyassertedin the
50 The End of Certainty
very structure of the theory. It thereforeseemsthat wehave an irreducibleduality: on the onehand, thetime-reversible Schrodingerequation,and on the other,the
collapse ofthe wave function.This dualistic nature ofquantum mechanicswas
repeatedly emphasizedby the great physicistWolfgang Pauli.Ina letterto Markus Fierz in 1947,he wrote, \"Somethingonly really happenswhen an observationis made,and in
conjunctionwith that. ..entropy necessarilyincreases.Betweenobservations,nothingat all happens.45Still, the
paper on which we write ages and becomesyellow,whetherornot we observeit.
Howcan this paradoxbesolved?Therehave beenmany
proposalsput forth in addition to the extremepositionsmentionedby Davies, includingNielsBohr's\"Copenhagen interpretation.\"*Bohrconcludedthat themeasurement apparatus has to be treated classically.It is as if we,who belongto the macroworld,needan intermediary tocommunicatewith the microworld,just as in somereligions we needa priestorshaman to communicatewith theother world.
But this hardly solvesthe problem,as the Copenhageninterpretationdoesnot lead to any prescriptionofwhatshould characterizethe physicalsystemswe may use as ameasurementdevice.Bohravoids the basicquestion:What
kind ofdynamical processesare responsiblefor the collapseof the wave function? LeonRosenfeld,Bohr'sclosestcoworker,was quite consciousof the limitations of the
Copenhageninterpretation.He consideredit only a first
step, the nextbeingto give a dynamical interpretationof
*Wehighly recommend Rae, Quantum Physics, and A. Shimony, \"Conceptual
Foundations of Quantum Mechanics,\" in Davies' New Physics.
Epicurus'Dilemma 51
the roleofthe apparatus.His convictionledto a numberofpublicationsin commonwith our own researchgroup,whichanticipatedourpresentapproach.46
Otherphysicistshave proposedidentifying the
measuringinstrument with some\"macroscopic\"device.In their
minds, the conceptofsuch a deviceis associatedwith
approximations. Forpracticalreasons,we wouldbeunabletomeasurethe quantum propertiesofthe apparatus.Furthermore, it has often beensuggestedthat we shouldconsiderthe apparatus as an \"open\"quantum systemconnectedtothe entireworld.47Contingentperturbationsandfluctuations stemmingfrom the environment would beresponsible for our ability to perform measurements.But what ismeant by \"environment\"?Who makesthe distinctionbetween an objectand its environment? This distinctionis
only a modified version of the von Neumannproposal,whichstates that through ouractionsand observations,it iswe who producethe collapseofthe wave function.
Theneedto eliminatethe subjectiveelementassociatedwith the observer has beenstressedby John Bellin his
excellentbook,Speakableand Unspeakablein QuantumMechanics.^ It is also an important considerationin the recentwork of Murray Gell-Mannand JamesB.Hartle,who
arguethat the appeal to an observerbecomeseven moreobscurein connectionwith cosmology.49Who measuresthe universe? This is not the placefor a detaileddiscussionoftheir approach;nevertheless,a briefdescriptionoftheir
latest findings would seemto be in order.Gell-Mannand othersintroducea coarse-grained
description ofthe quantum mechanicalhistoriesoftheuniverse that transforms the structureofquantum mechanics,leadingfrom a theory ofprobabilityamplitudesto a theoryofprobabilitiesproper.As an example,let us again con-
52 TheEnd of Certainty
siderthe wave function 4* =cxux + c2u2 obtainedby the su-
perimpositionofthe wave functions ux and uT Ifwe thentake the square (for purposesof simplification,we may
suppose*\302\245 is real) we have *\302\2452=
clxulx + ^2^2+ ^c\\c2u\\ut
Let us now presume that we can ignore the doubleproductcalledthe \"interferenceterm.\" All the mystery ofquantum theory then disappears.The probability *\302\2452 is
\"simply\" the sum ofprobabilities.Thereis no longeranyneedto speakofthe transition from potentiality to
actuality,and we can work directlywith probabilities.But how is
this possible?Interferencetermsplay a centralrolein many
applications ofquantum theory.Still, suppressingtheinterference termis preciselywhat Gell-Mannand his
colleagues propose.Why then,in somesituations,do we needexact,fine-grainedquantum descriptions,includinginterference, and in others,coarse-grainedones suppressinginterferences?Again, who actually doesthe coarsegraining?
Is it in any way reasonableto discussthe solution offundamental problemsin terms ofapproximations?How is
this consistentwith Gell-Mann'sown statement, alreadyquoted in SectionII, that quantum mechanicsis theframework into whichall theory must fit?
Still othersin the field hopeto solve the quantummechanical puzzle by reintroducingthe Epicurus clinamenin a modernform.Indeed,GiancarloGhirardi,EmanueleRimini,and Tullio Weber supposethat at sometime,forsomeunknown reason,a spontaneouscollapseofthe wave
function occurs.50Herethe conceptofchanceentersthe
discussion,but without any deeperjustification as a deus exmachina. Why doesthis new clinamenapply to somesituations and not to others?
What is especiallyunsatisfactory about all theseattemptsto elucidatethe conceptualfoundations ofquantum the-
Epicurus'Dilemma 53
ory is that they makeno new predictionsthat can actuallybetested.
Our own conclusioncoincideswith that ofmany otherspecialistssuchas AbnerShimonyin the UnitedStatesand
Bernardd'Espagnatin France.51Accordingto them,radical innovations have to bemade that wouldpreserveall theachievementsof quantum mechanics,but eliminatethedifficulties related to the theory'sdualistic structure.Notethat the measurementproblemis not isolated.As
emphasized by LeonRosenfeld,measurementis associatedwith
irreversibility.But in quantum mechanics,thereis noplacefor irreversibleprocesses,whetherornot they are involvedwith measurement.The difficulty of introducingirreversibility into quantum theory was already establisheddecadesago (in the contextofergodictheory) by von
Neumann, Pauli, and Fierz.52As in classicalmechanics,theytriedto solve the problemby coarsegraining,but their
attempts remainedunsuccessful.Thismay bethe reasonthat
von Neumanneventually adopted a dual formulation: the
Schrodingerequation on oneside,and the collapseofthewave function on the other.53But this is hardly satisfactoryas longas the collapseis not describedin dynamical terms.This is preciselywhat our own theory achieves.Thecentral roleis again playedby instability. However,deterministic chaosguidedby exponentiallydivergingtrajectoriesis
not applicablehere.In quantum mechanics,thereare notrajectories.Therefore,we have to considerinstability in
terms ofPoincareresonances.We can incorporatePoincareresonancesinto a statistical
descriptionand derive diffusive terms that lieoutside the
rangeofquantum mechanicsin terms ofwave functions.Thedescriptionis onceagain basedon the level ofprobability p (also calledthe density matrix in quantum me-
54 The End of Certainty
chanics;seeChapter6) and no longeron wave functions.
ThroughPoincareresonances,we achieve the transitionfrom probabilityamplitudesto probabilityproperwithout
drawingon any nondynamicalassumptions.As in classicaldynamics,the basicquestionis,When are
these diffusive terms observable?What are the limits
of traditional quantum theory? The answeris similar tothat for classicaldynamics(seeSectionIII).In short, it is in
persistent interactionsthat the diffusive termsbecomedominant (seeChapter7).As in classicalmechanics,this
prediction has beenverified by numericalsimulations.Onlyby goingbeyond a reductionistdescriptioncan we give arealisticinterpretationofquantum theory.Thereis nocollapse ofthe wave function, as the dynamical laws are nowat the level ofp,the densitymatrix, and not ofwavefunctions *\302\245. Moreover,the observerno longerplays anyspecial role.Themeasurementdevicehas to presenta brokentime symmetry. Forthesesystems,thereis a privilegeddirection oftime,exactlyas thereis a privilegeddirectionoftime in ourperceptionofnature. It is this common arrow oftime that is the necessaryconditionofourcommunicationwith the physicalworld; it is the basisofourcommunication with our fellow human beings.
Thus,instability plays a centralrolein bothclassicaland
quantum mechanics,and as such,obligesus to extendthe
scopeofboth disciplines.In so doing,we have to leave thefield of simple integrablesystems.The possibilityof aunified formulation ofquantum theory is particularly
exciting becausethis problemhas beenso hotly debatedover
the pastdecades,but the needfor an extensionofclassical
theory is even moreunexpected.We recognizethat this
meansa break with a rational tradition that harksbackto
the very foundationsofWestern scienceas conceivedby
Epicurus'Dilemma 55
Galileoand Newton.But it is no merecoincidencethat
the applicationofrecentmathematical methodstounstable systemsleadspreciselyto the extensionsdefinedin this
book.They allow us to includea descriptionofthe
evolutionary characteristicsofour universe basedon aprobabilistic descriptionofnature.In a recentarticle,I.Bernard Cohenspokeof the probabilistic revolution as arevolution in applications.Hewrote,\"Even if the decades18QD\342\200\2241930 do not showa singlerevolution in the domainofprobability,they provideevidenceofa probabilizingrevolution, that is, of a true revolution of fantastic
consequences attendant on the introductionofprobabilityandstatistics into areas that have undergonerevolutionarychangesas a result.\"54 This \"probabilizingrevolution\" isstill goingon.
V
We now cometo the closeofthis chapter.We beganwith
Epicurusand Lucretius,and their invention ofthe clina-men to permitthe appearanceofnovelty. After twenty-five hundred years,we can at last give a precisephysicalmeaningto this concept,which originatesininstabilities identifiedby the moderntheory ofdynamical systems.If the world wereformed by stabledynamical systems,itwould be radically different from the one we observearound us.It would bea static, predictableworld, but wewould not behereto make the predictions.In our world,we discoverfluctuations, bifurcations, and instabilitiesat all
levels.Stablesystemsleadingto certitudescorrespondonlyto idealizations,or approximations.Curiously,this insightwas anticipatedby Poincare.In discussingthe laws ofthermodynamics he wrote,
56 The End of Certainty
Theselaws canhave only one significance,which is that there
is a property common to all possibilities;but in the
deterministic hypothesis there is only a singlepossibility, and the
laws no longer have any meaning.In the indeterministic
hypothesis, on the other hand, they would have meaning, evenif they were taken in an absolutesense;they would appearas
a limitation imposedupon freedom.But thesewords remind
me that I am digressingand am on the point of leaving the
domains of mathematics and physics.55
Today we are not afraid ofthe \"indeterministichypothesis.\" It is the natural outcomeof the moderntheory ofinstability and chaos.Oncewe have an arrow oftime,weunderstand immediately the two main characteristicsofnature:its unity and its diversity: unity, becausethe arrowoftime is commonto all parts ofthe universe (your future
is my future; the future ofthe sun is the future ofany otherstar); diversity, as in the roomwhereI write,becausethereis air, a mixtureofgasesthat has moreorlessreachedthermal equilibriumand is in a state ofmoleculardisorder,andthereare the beautiful flowers arrangedby my wife, whichare objectsfar from equilibrium,highly organizedthanks
to temporal,irreversible,nonequilibriumprocesses.Noformulation ofthe laws ofnature that doesnot take intoaccountthis constructiveroleoftime can ever besatisfactory.
Chapter2
ONLY AN ILLUSION?
I
Theresultspresentedin this bookhave matured slowly.
It is now morethan fifty years sinceI publishedmy
first paper on nonequilibriumthermodynamics,in whichI pointedout the constructive roleof irreversibility.1 To
my knowledge,this was also the first paperthat dealt with
self-organizationas associatedwith distancefromequilibrium. After so many years,I often wonderwhy I wasfascinated with the problemoftime,and why it tookso verylongto establishits relationshipwith dynamics.While this
is not the placeto discussthe history ofthermodynamicsand statistical mechanicsover the past half century, I dowant to explainmy own motivations, and indicatesomeofthe main difficulties I encounteredalongthe way.
I have always consideredscienceto be a dialoguewith
nature. As in a real dialogue,the answersare often
unexpected\342\200\224andsometimesastonishing.
As an adolescent,I was enchantedwith archaeology,
57
58 The End of Certainty
philosophy,and especiallymusic.My motherused to saythat I couldread musicbeforereadingbooks.When I
entered the university, I spent muchmoretime at the pianothan in lecturehalls. In all the subjects I enjoyed, time
playedan essentialrole,whetherin the gradual emergenceofcivilizations, the ethical problemsassociatedwith human
freedom,or the temporalorganization ofsoundsin music.Thencamethe threat ofwar. It seemedmoreappropriateto undertake a careerin hard sciences,and so I startedto study physicsand chemistry at the FreeUniversity ofBrussels.
I often questionedmy teachersabout the meaningoftime,but their answerswereconflicting.Forthe
philosophers, this was the most difficult problemofall, closelyrelated to ethicsand the very nature ofhuman existence.Thephysicistsfound my question somewhatnaive, as theanswerhad already beengiven by Newton,and later
improved uponby Einstein.As a consequence,I felt bothastonished and frustrated. In science,time was consideredameregeometricalparameter.In 1796,more than onehundred years beforeAlbert Einstein and HermannMinkowski,Joseph-LouisLagrangehad calleddynamicsa\"four-dimensionalgeometry\"2Einstein went on to say,
\"Time[as associatedwith irreversibility] is an illusion.\"With my own background,thesestatementswereimpossible for meto accept.Nevertheless,the tradition ofspatial-izedtime is still very muchalive today, as witnessedby the
work of scientistssuch as StephenW Hawking.3In his
Brief History of Time, Hawking introduces\"imaginarytime\" to eliminateany distinctionbetweenspaceand time,a conceptwe shall examinein greaterdepth in Chapter8.
I am certainly not the first to have felt that the spatial-ization of time is incompatiblewith both the evolving
Only an Illusion? 59
universe, which we observe around us, and our ownhuman experience.This was the starting point for theFrenchphilosopherHenriBergson,for whom\"time isinvention ornothingat all.\"4 In Chapter1,I mentionedoneofBergson'slater articles,\"ThePossibleand the Real,\"
published on the occasionof his NobelPrizein 1930,wherehe expressedhis feelingthat human existenceconsists of \"the continual creationofunpredictablenovelty,\"
concludingthat time provesthat thereis indetermination in
nature.5 Theuniverse around us is only oneofa numberofpossibleworlds.Bergsonwould have beenquite amazedto read HenriPoincare'squotation at the endofChapterl.6Curiously, though, their conclusionspointed in thesame direction.I also quoted Alfred North Whiteheadfrom Processand Reality, for whomthe ultimate goalwas toreconcilepermanenceand change,to conceiveofexistence as a process.Accordingto him, classicalscience,whichoriginatedin the seventeenthcentury, was an
example ofmisplacedconcretenessunableto expresscreativity
as the basic property of nature, \"wherebythe actualworld has its characterof temporal passageto novelty.\"
Whiteheads conceptionofthe actual worldwasobviouslyincompatiblewith any deterministicdescription.7
I couldgoon by quotingMartin Heideggerand others,includingArthur Stanley Eddington,who wrote,\"In any
attempt to bridgethe domainsofexperiencebelongingtothe spiritual and physicalsidesofournature, time occupiesthe keyposition.\"8But insteadofbuildingthis bridge,timehas remaineda controversial issuefrom the pre-Socraticstothe present day. As mentionedfor classicalscience,the
problemoftime had beensolvedby Newtonand Einstein,but for mostphilosophers,this solutionwas incomplete.In
their opinion,we had to turn to metaphysics.
60 The End of Certainty
My personalconvictionwas quite different.
Abandoningscienceappearedto be tooheavy a priceto pay.After
all, sciencehad ledto a uniqueand fruitful dialoguebetween mankindand nature.Perhapsclassicalsciencecouldindeedlimit timeto a geometricalparameter becauseitwas dealingonly with simpleproblems.Therewas noneedto extendthe conceptof time when we dealt with africtionless pendulum, for instance.But once scienceencounteredcomplexsystems,it would have to modify its
approachto time.An examplethat often cameto mind wasassociatedwith architecture.Thereis not muchdifferencebetweenan Iranian brickfrom the fifth century beforeChrist and a neogothicbrickfrom the nineteenthcentury, but the results\342\200\224the palacesof Persepolisand the
neogothicchurches\342\200\224are in striking contrast.Timewouldthen be an \"emerging\"property.But what couldbe theroots oftime? I becameconvincedthat macroscopicirreversibility was the manifestation of the randomness ofprobabilisticprocesseson a microscopicscale.What thenwas the originofthis randomness?
With thesepreoccupations,it was only natural that I
turn to thermodynamics, especiallybecausein Brusselsthere was already an establishedschoolin the subjectfounded by TheophileDeDonder(1870-1957).
IIIn Chapter1,we mentionedthe classicalformulation ofthe secondlaw ofthermodynamicsattributed to Clausius.This law is basedon an inequality:Theentropy, S, ofan
isolatedsystemincreasesmonotonicallyuntil it reachesitsmaximum value at thermodynamic equilibrium.We
thereforehave dS > 0 for the changein entropy over the
Only an Illusion? 61
courseoftime.Howcan we extendthis statement tosystems that are not isolated,but whichexchangeenergyandmatter with the outsideworld?We must then distinguishtwo terms in the entropy change,dS:the first, d S, is thetransfer ofentropyacrossthe boundariesofthe system;the
second,rf.S, is the entropyproducedwithin the system.As
a result,we have dS= d S + d.S.We can now expressthesecondlaw by stating that whatever the boundaryconditions, the entropyproductiond.Sis positive,that is, d.S>0.Irreversible processesare creating entropy. DeDonderwenteven farther: Heexpressedthe productionofentropyperunit time P = -j^ in terms of the rates ofvariousirreversible processes(chemicalreactionrates, diffusion, etc.)and thermodynamic forces.In fact, he consideredonlychemicalreactions,but further generalizationwas easy.9
DeDonderhimselfdid not go very far alongthis road.He was concernedmainly with equilibriumand the
neighborhoodofequilibrium.Limitedas it was, his work
representedan important step in the formulation ofnon-equilibriumthermodynamics,even if it seemedto leadnowherefor a considerablelengthoftime.I still rememberthe hostility with whichDeDonder'swork was met.Forthe vast majority ofscientists,thermodynamicshad to belimited strictly to equilibrium.
That was the opinionofJ.Willard Gibbs, as well as ofGilbertN.Lewis,the mostrenownedthermodynamicistofhis day. Forthem, irreversibility associatedwith
unidirectional time was anathema. Lewiswent so far as to write,\"We shall see that nearly everywhere the physicisthas
purgedfrom his sciencethe useofone-way time ...aliento the idealsofphysics.\"10
I myself experiencedthis type of hostility in 1946,when I organizedthe first Conferenceon Statistical
62 The End of Certainty
Mechanicsand Thermodynamics under the auspicesofthe InternationalUnion for Pure and Applied Physics(IUPAP).Thesemeetingshave sincebeenheldon a
regular basisand continueto attract largecrowds,but at that
time we were a small group of approximatelythirty to
forty people.After I had presentedmy own lectureonirreversible thermodynamics,the greatestexpertin the fieldofthermodynamicsmadethe following comment:\"I amastonishedthat this young man is so interestedin nonequi-librium physics.Irreversibleprocessesare transient.Whynot wait and study equilibriumas everyoneelsedoes?\"Iwas so amazedat this responsethat I did not have thepresence ofmind to answer:\"But we are all transient.Is it notnatural to be interestedin our commonhuman
condition?\"
Throughoutmy entirelife I have encounteredhostilityto the conceptofunidirectionaltime.It is still the
prevailingview that thermodynamicsas a disciplineshould
remain limited to equilibrium.In Chapter1,I mentionedthe attempts to banalize the secondlaw that are so mucha
part ofthe credoofa numberoffamous physicists.Icontinue to beastonishedby this attitude. Everywherearoundus we seethe emergenceofstructuresthat bearwitnesstothe \"creativity ofnature,\" to useWhitehead'sterm.I have
always felt that this creativity had to beconnectedin someway to the distancefrom equilibrium,and was thus the
result ofirreversibleprocesses.Compare,for example,a crystal and a town.A crystal is
an equilibriumstructure that can bemaintainedin a
vacuum, but if we isolatedthe town, it would diebecauseitsstructure dependson its function.Functionand structureare inseparablein that the latter expressesthe interactionsofthe town with its environment.
Only an Illusion? 63
In Erwin Schrodingers beautiful bookWhat Is Life?hediscussesthe metabolismofa living body in terms ofentropy productionand entropy flow. Ifan organismis in a
steady state, its entropy remains constant over time,andthereforedS= 0.As a result,the entropyproductiond.Siscompensatedby the entropy flow d S + d S = 0,or d S =d.S< 0.Life, concludesSchrodinger,feedson a \"negativeentropy flow\"11The moreimportant point, however, isthat life is associatedwith entropy productionand
therefore with irreversibleprocesses.But how can structure, as in living systemsor towns,
emergein nonequilibriumconditions?Hereagain, as in
dynamics,the problemofstability playsan essentialrole.At
thermodynamic equilibrium,entropy has a maximumvalue whenthe systemis isolated.Fora systemmaintainedat temperatureT, we have a similar situation.We thenintroduce \"freeenergy,\" F=E \342\200\224 TS,a linearcombinationofenergyE and entropyS.As shown in all texts onthermodynamics, free energy,F, is at its minimum at equilibrium(seeFigure2.1).Consequently,perturbations orfluctuations have no effect becausethey are followedby a return
to equilibrium.Thesituation is not unlike that ofthe
stable pendulumconsideredin Chapter1,SectionIII.What happens in a steadystate correspondingto non-
equilibrium?We saw such an examplein a discussionofthermal diffusion in Chapter1,SectionII.Isa
nonequilibrium steady state truly stable? In near-equilibriumsituations (known as \"linear\" nonequilibriumthermodynamics), the answeris yes.As shown in 1945,the steadystate correspondsto a minimum ofentropyproductionperunit time P= -^.12At equilibriumP= 0,entropyproduction vanishes,while in the linear regimearoundequilibrium, P is minimum (seeFigure2.2).13
64 The End of Certainty
Figure2.1Minimum ofFFreeenergy is minimum at equilibrium (X
=X\342\200\236
Again fluctuations dieout.But therealready appearsaremarkablenew characteristic:A nonequilibriumsystemmay evolve spontaneouslyto a state ofincreasedcomplexity.The orderingwe observe is the outcomeof irreversible
processes,and couldnot beachievedat equilibrium.Thisis clearin the exampleofthermal difFusion mentionedin
Chapter1,wherethe temperaturegradient leadsto apartial separationofthe compounds.Many other caseshavesincebeenstudied in which complexity has consistentlybeenassociatedwith irreversibility. Theseresultsbecamethe guidelinesfor our future research.
But can we extrapolatethe resultsoffar-from-equilib-rium situations from thoseat near-equilibrium?Mycolleague Paul GlansdorfF and I investigated this problemfor
many years,14and arrived at a surprisingconclusion:Contrary to what happensat equilibrium,ornearequilibrium,systemsfar from equilibriumdo not conformto anyminimum principlethat is valid for functions offree energyorentropyproduction.As a consequence,thereis noguarantee that fluctuations are damped.We can only achieve
Only an Illusion? 65
Figure2.2Minimum ofPEntropy production P =
d\\S/dt is minimum in a steady state (X=
A.J().
a formulation of sufficient conditions for stability, whichwe call the \"generalevolution criterion.\"This requiresspecifyingthe mechanismofirreversibleprocesses.Near-equilibriumlaws ofnature are universal, but whenthey arefar from equilibrium,they becomemechanismdependent.We thereforebeginto perceivethe originofthe variety in
nature we observearound us.Matter acquiresnewproperties whenfar from equilibriumin that fluctuations andinstabilities are now the norm. Matter becomesmore\"active.\" Although thereis at presentan enormousliterature surroundingthis subject,15for the momentwe shall
consideronly a simpleexample.Supposethat we have achemicalreaction{A} =^{X} -^ {F}in which{A}is a
set ofinitial products,{X}a set ofintermediateones,and
{F}a set offinal ones.At equilibrium,we have a detailedbalancewherethereare as many transitions from {^4}to
{X}as from {X}to {^4},with the sameapplyingto {X}and {F}.The ratio of initial to final products {^4}/{F}takeson a well-definedvalue correspondingto maximumentropy if the systemis isolated.Now consideran open
66 The End of Certainty
system,such as a chemicalreactor.By controllingthe flowofmatter, we may fix the values ofboth the initial andfinal products {A}and {F}.We progressivelyincreasetheratio {A}/{F},starting from its equilibriumvalue. What
will happen to the intermediateproducts {X}when wemove away from equilibrium?
Chemicalreactionsare generallydescribedby nonlinearequations.Thereare many solutionsfor the intermediateconcentrations{X}for given values of{A}and {F},but
only onecorrespondsto thermodynamicequilibriumandmaximumentropy.This solution,whichwe call the
\"thermodynamic branch,\" may be extendedto the domain ofnonequilibrium.The unexpectedresult is that this branch
generallybecomesunstable at somecriticaldistancefrom
equilibrium(seeFigure2.3).Thepoint wherethis occursis known as the bifurcation point.
Beyond the bifurcation point,a set ofnew phenomenaarises;we may have oscillatingchemicalreactions,non-equilibriumspatial structures,orchemicalwaves.We have
given the name dissipative structures to thesespatiotemporalorganizations.Thermodynamicsleads us to theformulation of two conditionsfor the occurrenceofdissipativestructuresin chemistry:(1)far-from-equilibriumsituationsdefinedby a criticaldistance;and (2) catalytic steps,suchas
the productionof the intermediatecompoundY from
compoundX together with the productionofX from Y.
It is interestingto note that theseconditionsare satisfiedin all living systems:Nucleotidescodefor proteins,whichin turn codefor nucleotides.
We wereextremelyfortunate in that soonafter we had
predictedthesevarious possibilities,the experimentalresults of the Belousov-Zhabotinskireaction\342\200\224a spectacularexampleofchemicaloscillations\342\200\224became widelyknown.16I rememberouramazement whenwe sawthe reactingso-
Only an Illusion? 67
Figure2.3ThermodynamicBranchThe two steady-state solutions th and d are functions of the ratio
A/F. At the bifurcation point, the thennodynamic branch th
becomesunstable, and another branch d becomesstable.
X .i**'
Equilibrium
YA/F
Point ofbifurcation
lution becomeblue,and then red,and then blue again.Today, many other oscillatoryreactionsare known,17but
the Belousov-Zhabotinskireactionremains historicbecause it proved that matter far from equilibriumacquiresnew properties.Billionsofmoleculesbecomesimultaneously blue, and then red. This entails the appearanceoflong-rangecorrelationsin far-from-equilibriumconditions that are absent in a state ofequilibrium.Again, wecan say that matter at equilibriumis \"blind,\" but far from
equilibriumit beginsto \"see.\"We have observed that at
nearequilibrium,dissipationassociatedwith entropyproduction is at a minimum.Far from equilibrium,it is justthe opposite.New processesset in and increasetheproduction ofentropy.
Therehas beensteadyprogressin far-from-equilibriumchemistry.In recentyears, nonequilibriumspatial struc-
68 The End of Certainty
tures have beenobserved.18Thesewerefirst predictedbyAlan MathisonTuring in the contextofmorphogenesis.19
When we push the systemfarther into nonequilibrium,new bifurcations typical of chaoticbehavior may arise.Neighboringtrajectoriesdivergeexponentiallyas in
deterministic chaosakin to the dynamical systemsweconsidered in Chapter1,SectionIII.
In short, distancefrom equilibriumbecomesan essential
parameterin describingnature much like temperature in
equilibriumthermodynamics.When we lower the
temperature, we observe a successionof phase transitions
through various states ofmatter. But in nonequilibriumphysics,the variety ofbehaviorsis muchgreater.We haveconsideredchemistryfor the purposesof this discussion,but similar processesassociatedwith nonequilibriumdissi-
pative structureshave beenstudied in many other fields,includinghydrodynamics,optics,and liquid crystals.
Let us now lookmorecloselyat the critical effect offluctuations. As we have seen,near-equilibriumfluctuations are harmless,but far from equilibrium,they play acentralrole.Not only dowe needirreversibility, but wealso have to abandon the deterministicdescriptionassociated with dynamics.Thesystem\"chooses\"oneofthepossible branchesavailable when far from equilibrium.But
nothingin the macroscopicequationsjustifies thepreference for any onesolution.This introducesan irreducibleprobabilisticelement.Oneof the simplestbifurcations isthe so-called\"pitchfork bifurcation\" representedin Figure2.4,whereX = 0correspondsto equilibrium.
The thermodynamic branch is stable from X =0 toX = X.BeyondX, it becomesunstable,and a symmetricalpair ofnew stable solutionsemerges.It is the fluctuationsthat decidewhichbranch will be selected.If we wereto
suppressfluctuations, the systemwould maintain itself in
Only an Illusion? 69
Figure2.4Pitchfork BifurcationConcentrationX is a function of the parameter X, which measuresthe distance from equilibrium. At the bifurcation point, the
thermodynamic branch becomesunstable, and the two new solutions
bx and b2 emerge.
X
Thermodynamic branch (\\
\\
Stable ,
~\342\200\224
\342\226\240
b2
Multiple solutions X
an unstablestate.Attempts have beenmade to decreasethefluctuations so that we can subject the unstableregionto
experiment;nevertheless,sooneror later, fluctuations ofinternal or externalorigintake over and bringthe systemto oneofthe branchesb1 or b2.
Bifurcations are a sourceofsymmetry breaking.In fact,the solutionsof the equationbeyond X generallyhave a
lowersymmetry than the thermodynamicbranch.20Bifurcations are the manifestation ofan intrinsicdifferentiationbetweenparts of the systemitself and the systemand its
environment.Oncea dissipative structure is formed,the
homogeneityoftime (as in oscillatorychemicalreactions)orspace(as in nonequilibriumTuring structures),orboth,is broken.
In general,we have a successionofbifurcations asrepresented schematicallyin Figure2.5.The temporaldescription of such systems involves both deterministicprocesses(betweenbifurcations) and probabilisticprocesses(in the choiceofthe branches).Thereis also a historical
70 TheEnd of Certainty
dimensioninvolved. If we observe that the system is in
state d2, that means that it has gonethrough the statesbx
and q (seeFigure2.5).Oncewe have dissipative structures,we can speak of
self-organization.Even if we know the initial values and
boundaryconstraints,thereare still many states availabletothe systemamongwhichit \"chooses\"as a result offluctuations. Suchconclusionsare ofinterestbeyond the realmsofphysicsand chemistry.Indeed,bifurcations can beconsidered the sourceof diversification and innovation.21Theseconceptsare now appliedto a wide group ofproblems in biology,sociology,and economicsat
interdisciplinarycentersthroughoutthe world.In Western Europe
alone,therehave beenmorethan fifty centersfornonlinear processesfounded over the past ten years.
Freudwrote that the history ofscienceis the history ofalienation.Copernicusshowed that the earth is not at the
centerof the planetary system,Darwin that we are onespeciesofanimal amongmany others,and Freud that our
Figure2.5SuccessiveBifurcations with IncreasingDistancefrom
Equilibrium
X
Only an Illusion? 71
rational activity is only part of the unconscious.We can
now invert this perspective:We seethat human creativityand innovation can beunderstoodas the amplification oflaws ofnature already present in physicsorchemistry.
IllThe resultspresentedthus far show that the attempts totrivialize thermodynamics mentionedin Chapter 1 are
necessarilydoomedto failure. The arrow oftime plays an
essentialrolein the formation of structures in both the
physicalsciencesand biology.But we are only at the
beginning ofourquest.Thereis still a gap betweenthe most
complexstructureswe can producein nonequilibriumsituations in chemistryand the complexitywe find in
biology.This is not only a problemfor pure science.In a recent
report to the European Communities,Christof Karl
Biebracher,GregoireNicolis,and PeterSchusterwrote,
The maintenanceof organization in nature is not\342\200\224and
cannot be\342\200\224achieved by central management; order can only bemaintained by self-organization. Self-organizing systems
allow adaptation to the prevailing environment, i.e.,they
react to changesin the environment with a thermodynamic
responsewhich makes the systems extraordinarily flexibleand
robust against perturbations from outside conditions. We
want to point out the superiority of self-organizingsystems
over conventionalhuman technologywhich carefully avoids
complexity and hierarchically managesnearly all technical
processes.For instance,in synthetic chemistry, different
reaction stepsare usually carefully separatedfrom each other and
contributions from the diffusion of the reactants are avoided
by stirring reactors.An entirely new technologywill have to
be developedto tap the high guidanceand regulation poten-
72 The End of Certainty
tial of self-organizingsystems for technicalprocesses.The
superiority of self-organizingsystems is illustrated by biologicalsystems where complexproductscan be formed with
unsurpassed accuracy,efficiencyand speed!\"22
The results of nonequilibriumthermodynamics arecloseto the views expressedby Bergsonand Whitehead.Nature is indeedrelated to the creationofunpredictablenovelty, wherethe possibleis richerthan the real.Ouruniverse has followeda path involving a successionofbifurcations. While other universes may have followed otherpaths,we are fortunate that ourshas ledto life, culture,andthe arts.
The dreamofmy youth was to contributeto theunification ofscienceand philosophyby resolvingthe enigmaof time.*Nonequilibriumphysicsshows that this is
entirely possible.The resultsdescribedin this chaptergavemethe impetusto explorethe conceptoftime on the
microscopic level.I have emphasized the roleoffluctuations\342\200\224but what is their origin?How can we reconciletheir behavior with the deterministicdescriptionbased
upon the traditional formulation of the laws of nature?Were we to doso,we would losethe distinctionbetweennear-and far-from-equilibriumprocesses.Moreover,wewould becalling into questionsuch unique and marvelousconstructionsofthe human mind as classicaland quantummechanics.
I must confessthat thesethoughts ledto many sleeplessnights.Without the supportofmy colleaguesand students,I would most certainly have given up.
*I expressed this dream in three short essays written for a student journal as
early as 1937!
Chapter3
FROM PROBABILITYTOIRREVERSIBILITY
I
Aswe saw in Chapter2,irreversibleprocessesdescribefundamental features ofnature leadingto nonequi-
librium dissipativestructures.Suchprocesseswould not bepossible in a world ruled by the time-reversiblelaws ofclassicaland quantum mechanics.Dissipativestructures
require an arrow oftime.Furthermore,there is no hopeofexplainingthe appearanceofsuch structuresthroughapproximations that would be introducedby theselaws.
I have always beenconvincedthat an understandingofthe dynamical origin of dissipative structures, and moregenerallyof complexity, is one of the most fascinatingconceptualproblemsofcontemporaryscience.As alreadystated in Chapter1,for unstable systemswe have toformulate the laws of dynamicsat the statistical level.This
73
74 The End of Certainty
changesourdescriptionofnature in a radicalway In sucha formulation, the basic objectsofphysicsare no longertrajectoriesor wave functions; they are probabilities.We
have thus cometo the end of the \"probabilisticrevolution\" that could already be found in areas other than
physicsby the eighteenthcentury However,when facedwith the implicationsofthis radicalconclusion,I hesitatedfor sometime,reachingfor lessextremesolutions.In From
Being to Becoming, I wrote,\"In quantum mechanics,thereare observationswhosenumericalvalue cannotbedetermined simultaneously, i.e.,coordinatesand momentum.(This is the essenceofHeisenberg'suncertainty relationsand Bohr'scomplementarityprinciple.)Herewe also havea complementarity\342\200\224one betweendynamical and thermo-dynamical descriptions.\"1This would have beena muchless extremeapproach to the conceptualproblemassociated with irreversibility.
In retrospect,I regretthis statement in my earlierbook.If there is morethan a single description,who wouldchoosethe right one?The existenceofthe arrow oftimeis not a matter ofconvenience.It is a fact imposedbyobservation. However,it is only in recentyears that the resultswe obtainedby studyingthe dynamicsofunstable systemsforcedus to reformulate dynamicsat the statistical level,and to concludethat this formulation leadsto an extensionofclassicaland quantum mechanics.In this chapter, I
describe someofthe stepsinvolved.Forapproximatelyonehundred years,we have known
that even simpleprobabilisticprocessesare time oriented.In Chapter1,we mentionedthe \"randomwalk.\" Another
exampleis the \"urn model\"proposedby Paul and TatianaEhrenfest(seeFigure3.1).2
ConsiderN objects(such as balls) distributedbetween
From Probability to Irreversibility 75
Figure3.1TheEhrenfestUrnModelN balls are distributed between two urns, A and B.At time n,there are k balls in A and N-k balls in B.At regular time
intervals, a ball is removedat random from one urn and placedinto
the other.
pottery
N-k
k-1 \\\"^ ^*V N-k+1
or /^ ^\\ or
fe+1 / \\iV-fe-l.
two urns A and B.At regular time intervals (for example,every second)a ball is chosenat random and moved fromoneurn to the other.Supposethat at time n, thereare k
balls in A, and thereforeN \342\200\224 k in B.At time n + 1therecan be eitherk \342\200\224 1or k + 1balls in A. Theseare well-defined transition probabilities.But let us go on with the
game.We expectthat as a result ofthe exchangeofballs,we shall reach a situation where there will beapproximately ~2
balls in eachurn.However,fluctuations willcontinue. We might even end up in the situation at time n
wherethereare again k balls in urn A. It is at the level ofprobability distribution that we seean irreversibleapproachto
76 TheEnd of Certainty
equilibrium.Whatever the starting point, it can beshownthat the probabilitypn(k) offinding k balls in oneurn aftern moves as n \342\200\224>
\302\260o tends to the binomial distributionN\\ ... M
fe!(N-fe)! .This expressionhas a maximumvalue ofk = ^,but also takesinto accountfluctuations in distribution.Inthe Boltzmannmodel,the maximumentropycorrespondspreciselyto the binomialdistribution.
The Ehrenfest modelis an exampleof a \"Markov
process\"(or\"Markovchain\,")namedafter the greatRussian mathematician, Andrei Markov, who was the first todescribesuch processes.Oncewe have a probabilisticdescription, it is often possibleto derive irreversibility. Buthow do we relate theseprobabilisticprocessesto dynamics?That is the fundamental problem.
We have seenthat a basicstep in this directionwastaken
by the fathers ofstatisticalphysics,or the physicsofpopulations. Maxwell,Boltzmann,Gibbs, and Einstein all
emphasized the roleofensemblesdescribedby a probabilitydistributionp.An important questionthen is, What is theform of this distribution function onceequilibriumis
reached?Let qv . . . ,qs be the coordinatesand pl, \342\226\240 \342\226\240 \342\226\240,psthe momentaof the particles forming this system. In
Chapter1,the phasespacewas definedby the coordinatesand momenta.We also introducedthe probabilitydistributions p(q,p, t) (seeChapter1,SectionIII).We shall nowuse the singleletterq for all coordinatesand p for all
momenta. Equilibrium is reachedwhenp becomestime
independent. In every textbook,it is shown that this occurswhenp dependsonly on the total energy.As mentionedin
Chapter1,SectionIII,the total energy is the sum ofthekineticenergy (dueto the motionofthe particles)and the
potential energy (due to interactions).When expressedinterms ofq and p, this energy,whichis calledthe Hamilton-
ian H(p,q),remainsconstant over time.This is the princi-
From Probability to Irreversibility 77
pie ofconservationofenergy,the first principleofthermodynamics. It is therefore natural that at equilibrium,p is
a function ofthe Hamiltonian H.An important exceptioncaseis that of ensemblesin
whichall systemshave the same energy E. Thedistribution function then vanishes throughout the phase space,save on the surface H (p, q) = E, where the distributionfunction is constant.This is calledthe \"microcanonicalensemble.\" Gibbsshowedthat suchensemblesdo indeedsatisfy
the laws of equilibriumthermodynamics.He alsoconsideredotherensemblessuch as the \"canonicalensemble,\"
in whichall systemsinteract with a reservoirat
temperature T. This leads to a distribution function that
dependsexponentiallyon the Hamiltonian,p now beingproportionalto exp (\342\200\224 ^f), whereT is the temperatureofthe reservoirand k the Boltzmannconstant,whichmakesthe exponentdimensionless.
Oncethe equilibriumdistributionis given, we cancalculate all thermodynamic equilibriumpropertiessuch as
pressure,specificheat, etc.We can even gobeyondmacroscopic thermodynamicsbecausewe are able to includefluctuations. It is generallyacceptedthat in the vast field ofequilibriumstatistical thermodynamics,thereare noconceptual difficulties left, only computational ones,whichcan be solvedlargely through numericalsimulations.Theapplicationofensembletheory to equilibriumsituationshas undoubtedlybeenquite successful.Notethat the
dynamical interpretationofequilibriumthermodynamicsbyGibbs is in terms ofensembles,and not in terms oftrajectories. It is this approachthat we have to extendin ordertoincludeirreversibility.
This is quite natural, as thereis no time orderingat thelevel oftrajectories(orwave functions) becausefuture and
past play the same rolein accordancewith classicaland
78 TheEnd of Certainty
quantum physics.However,what happens at the level ofstatistical description,in terms ofdistribution functions?Let us lookat a glassofwater.In this glass,thereis a hugenumberofmolecules,a quantity on the orderof 1023.Fromthe dynamical point ofview, this is a nonintegrablePoincaresystem,as defined in Chapter1,sincethereareinteractionsbetweenthe moleculesthat we cannoteliminate. We may visualize theseinteractionsas leadingtocollisions betweenthe molecules(the term\"collision\"will bedefined moreprecisely in Chapter5), and describethewater containingthemin terms ofthe statistical ensemblep.Is the water aging? Certainlynot, if we considertheindividual water molecules,whichare stableover geologicaltime.Still,thereis a natural time orderin this systemfromthe point ofview ofthe statistical description.Aging is a
property ofpopulations,exactlyas it is in the Darwinian
theory ofbiologicalevolution.It is the statisticaldistribution that approachesthe equilibriumdistribution,such asthe canonicaldistribution defined above.To describethis
approachto equilibrium,we needthe idea ofcorrelation.
Considera probabilitydistributionp(xpx2),
dependingon two variables xp x2.If xt and x2 are independent,
we have the factorization p(xp x2)=p1(x1)p2(x2).Theprobabilityp(xpx2) is then the productoftwoprobabilities. In contrast,if p(xpx2) cannot befactorized,x; and x2are correlated.Now let us return to the moleculesin the
glassofwater.Thecollisionsbetweenthesemoleculeshavetwo effects: They make the velocity distribution moresymmetrical, and they producecorrelations(seeFigure3.2).But two correlatedparticleswill eventually collidewith a third one.Binary correlationsare then transformedinto ternary ones,and so on (seeFigure3.3).
We now have a flow ofcorrelationsthat are orderedin
From Probability to Irreversibility 79
Figure3.2Collisionsand CorrelationsThe collisionof two particles createsa correlationbetween them
(representedby a wavy line).
oo
o
o oBeforecollision After collision
time.A valuable and provocative analogy to this flowwould be human communication.When two peoplemeet,they converse,and consequendymodify their
thinkingto someextent.Thesemodificationsare brought to
subsequentmeetings,and modifiedfurther. The word forthis phenomenonis dissemination. Thereis a flow ofcommunication in society,just as thereis a flow ofcorrelationsin matter. Of course,we may also conceiveof inverse
processesthat make the velocity distributionlesssymmetrical by destroyingcorrelations(seeFigure3.4).
We thereforeneedan elementthat will validate the
processesthat make the velocity distribution moresymmetrical over the courseof time.As we shall see,this is
preciselythe roleofPoincareresonances.We now beginto
geta glimpseofa statistical descriptionthat includesirreversibility. This descriptionwill be a dynamics ojcorrelations
leadingto the equilibriumdistribution.Theexistenceofa flow ofcorrelationsorderedin time,
asrepresentedin Figure3.3,has beenverified by computersimulations.3We can also reproduceprocessessuchas those
80 The End of Certainty
Figure3.3FlowofCorrelationsSuccessivecollisionsleadto binary, ternary, ...correlations.
o o^oo
o -o
\342\200\224\342\226\272 o o -o
\342\200\224\342\226\272
oo
representedin Figure3.4through time inversion,wherewe invert the velocity ofthe particles.But we can achievethis inverted flow ofcorrelationsonly for briefperiodsoftime and for a limitednumberofparticles,after whichwe
again have a directedflow of correlationsinvolving an
ever-increasingnumberofparticlesleadingthe systemto
equilibrium.
Figure3.4DestructionofCorrelationsIn (a) the particles (representedby black points) interact with the
obstacle(representedby the circle).Initially all particles have the
same velocity. The collisionvaries the velocitiesand createscorrelations betweenthe particles and the obstacle.In (b) we representthe oppositeprocess.We considerthe effect of a velocityinversion; as a result of the inverted collision,correlationswith the
obstacle aredestroyed,and the initial velocity is recovered.
o \342\200\224\342\226\272 Ct^*-^ \"*^~>D \342\200\224\342\226\272 o
(a) (b)
From Probability to Irreversibility 81
Theseresults,whichgive meaningto irreversibility at
the statistical level, wereobtained nearly thirty years ago.4At that time,however, certainbasic questions still
remained unanswered:Howcan irreversibility appearat thestatistical level ofdescription,and not whenwe describedynamicsin terms oftrajectories?Is this due to ourapproximations? Moreover,is the successionofcorrelationsthat we observe, for examplein computerexperiments,perhaps the result of the limitations of computertime?
Obviously,a shorterprogramis requiredto prepareuncor-related particles that producecorrelationsthroughcollisions than to prepareensemblesthat couldlead to inverse
processesin whichcorrelationsare destroyed.But why start at all with probabilitydistributions?Such
distributionsdescribethe behaviorofbundlesoftrajectories,
or ensembles.Dowe use ensemblesbecauseofour\"ignorance,\"or is there,as argued in Chapter1,a deeperreasoninvolved? Forunstablesystems,ensemblesindeeddisplay new propertiesas comparedwith individualtrajectories. This is what we shall now demonstratewith several
simpleexamples.
IIIn this section,we shall be concernedwith deterministicchaos,as well as an especiallysimple type ofchaos,both
correspondingto chaotic maps. Contrary to what occursinordinary dynamics,time in maps acts only at discreteintervals, as is the casein the Ehrenfesturn modelwestudied in SectionI.Mapsthereforerepresenta simplifiedformofdynamicsthat makesit easy for us to comparetheindividual level ofdescription(the trajectories)with thestatistical description.We shall considertwo maps; the first
82 The End of Certainty
charts simpleperiodicbehavior, the seconddeterministicchaos.
In the first instance,we shall considerthe \"equationsofmotion\" x , , =x + ?,modulo1,whichmeansthat we are
n + 1 n *\342\226\240''
dealing only with numbers between0 and 1.After two1 3
shifts, we are back to the initial point (i.e.,xo = 4, xt =4, x23 2 5 1=4+4=4 =4).This situation is representedin Figure3.5.Insteadofconsideringindividual points locatedby
trajectories, it is worth examiningensemblesdescribedby the
probabilitydistributionp(x).A trajectorycorrespondsto a
specificsetofensembleswherethe coordinatex takeson a
well-definedvalue xn, and the distribution function p is
then reducedto a singlepoint.As mentionedin Chapter1,SectionIII,this can bewritten as Pn(x) =8(x\342\200\224 x ). (Deltais a symbolfor a function that vanishes for all values ofx
Figure3.5PeriodicMapThere is a simple geometrical construction that moves from the
initial point P0to the next point Px accordingto the map xn + j \342\200\224>
x\342\200\236
+ l /2.We go from P0to P',then to P\" on the bisector,and fromthere to Px.Obviously, if we start with Plt wecomeback to P0.
X\342\200\236
From Probability to Irreversibility 83
exceptx= x .)By usingdistributionfunction p, the
mappingcan be expressedas a relation betweenp +l(x) and
Pn(x).We may then write Pn + l(x)= U p (x). Formally,p + j
is obtained through the operatorU, known as thePerron-Frobeniusoperator,acting onPn{x).5At this point,although its explicitform is not important to us, it is
interesting to note that no new element(in addition to the
equationofmotion)entersinto the constructionof U.
Obviously,the ensembledescriptionmust allow the
trajectory descriptionas a specialcase;we thereforehave 8(x\342\200\224
xn + i) = U8(x\342\200\224 xj.This is simplya way ofrewriting the
equationofmotion,as xn becomesxn + after one shift.
Themain questionis,however, 75 this the only solution, or arethere new solutionsfor the evolution ofensembles,as describedby
the Perron-Frobeniusoperator, which cannot beexpressedin terms
of trajectories?In ourexampleofa periodicmap, the answeris no.Thereis not any difference betweenthe behaviorofindividual trajectoriesand ensemblesfor stablesystems.It is
this equivalencebetween the individual point of view
(correspondingto trajectoriesor wave functions) and thestatistical point ofview (correspondingto ensembles)that
is brokenfor unstable dynamical systems.Thesimplestexampleofa chaoticmap is the Bernoulli
map. Herewe doublethe value ofa numberbetween0and1every second.Theequationofmotionis now xn+l
= 2x(modulo1).This map is representedin Figure3.6.Theequationofmotionis again deterministicin that onceweknow x , the numberx ,,is determined.Herewe have anrv \302\253+l
exampleofdeterministicchaos,so calledbecauseif wefollow a trajectory through numericalsimulations,we seethat it becomeserratic.As the coordinatex is multipliedbytwo at eachstep,the distancebetweentwo trajectorieswill
84 The End of Certainty
Figure3.6BernoulliMapIn this exampleof deterministic chaos,we start from point P0and
go to point Pu as the value of x doubles(modulo 1).
be (2\") = exp(n log2), again modulo1.In terms ofcontinuous time t, this can bewritten asexp(tX), with A = log2,whereX is calledthe Lyapunov exponent.This showsthat trajectoriesdivergeexponentially,and it is this
divergencethat is the signature ofdeterministicchaos.Ifwe
wait longenough,any arbitrarily selectedpoint between0and 1will eventually be approachedby the trajectory (seeFigure3.7).Herewe have a dynamical processleadingto
randomness.In the past, this apparent flow in thedeterministic universe was repeatedlyinvestigated by greatmathematicians such as LeopoldKronecker(1884) and
HermannWeyl (1916).Accordingto Jan von Plato, similar
resultshad beenobtainedas early as medieval times,so this
is certainly not a new problem.6What is new, however, is
the statistical formulation of the Bernoullimap, whichlinks randomnessto operatortheory.
From Probability to Irreversibility 85
Figure3.7NumericalSimulationsofTrajectoriesfor theBernoulliMapThe initial conditions are slightly different for each simulation.
This differenceis amplified as time goes on. (These numericalsimulations are the work of DeanDriebe).
1
0.8
0.6
0.4
0.2
10 20 30 40
86 TheEnd of Certainty
Figure3.8Simulationofpn(x) for the BernoulliMapNumerical simulation of the evolution of the probabilitydistribution. In contrast with the trajectory description,the probabilities
rapidly reachthe asymptotic uniform distribution. (Thesenumerical simulations are the work ofDeanDriebe.)
0 0.2 0.4 0.6 0.8 1 X 0 0.2 0.4 0.6 O.i 1 X
PW
2 -
1.5-1
0.5-
t = 2
-J l_
0 0.2 0.4 0.6 O.i 1 X
PW
21.5-
1 -
0.5-_J l l l_
0 0.2 0.4 0.6 O.i 1 X
PW
21.5
1
0.5
( = 4
_i l l i_
0 0.2 0.4 0.6 0.8 1 X
From Probability to Irreversibility 87
We now turn to the statistical descriptionin terms ofthe Perron-Frobeniusoperator.In Figure3.8,we seetheeffect ofthe operatorU on the distributionfunction.Thedifferencefrom the trajectory descriptionis strikingbecause the distributionfunction p (x) leadsrapidly to a
constant. We may thereforeconcludethat theremust bea basicdifference betweenthe descriptionin terms oftrajectorieson the onehand and in terms ofensembleson the other.In short, instability at the level oftrajectoriesleadsto
stability at the level ofstatistical descriptions.How is this possible?The Perron-Frobeniusoperator
still admits a trajectory description8(x\342\200\224 x +1)=U8(x\342\200\224
x ), but the unexpectedfeature is that it also allows newsolutions that are applicableonly to statistical ensembles,and not to individual trajectories.The equivalencebetween the individual point ofview and the statistical
description is broken.This remarkablefact leads to a new chapter in
mathematics and theoreticalphysics.7Although the problemofchaoscannotbe solvedat the level ofindividual
trajectories,it can besolvedat the level ofensembles.We can now
speakofthe laws ofchaos.sAs we shall seein Chapter4, we
may even predict the speedat whichthe distributionpapproaches equilibrium(which for the Bernoullimap is a
constant),and establishthe relationshipbetweenthis speedand the Lyapunov exponent.
How can we understand the difference betweenindividual descriptionand statisticaldescription?We shall
analyzethis situation in moredetail in Chapter4, wherewe
shall seethat thesenew solutionsrequiresmoothnessin thedistribution functions.This is the reasonthat suchsolutions are not applicableto individual trajectories.A
trajectory representedby 8(x\342\200\224 x) is not a smoothfunction; it is
88 TheEnd of Certainty
different from zeroonly for d vanishes if xdiffers
at all from x .n
The descriptionin terms of distribution functions is
thereforericherthan that derivedfrom individualtrajectories. This agreeswith the conclusionswe arrived at in
Chapter1,SectionIII.Trajectoriesare merelyspecialsolutions ofthe Perron-Frobeniusequation for unstable maps.This also appliesto systemswith Poincareresonances(seeChapters5 and 6).Thetime-orientedflow ofcorrelationsis an essentialelementin the new solutionsfor the
probability distribution,while no time-orientedprocessesexistat the level ofindividual trajectories.
This break in the equivalencebetweenthe individualand the statisticaldescriptionis the fundamental inspirationofour approach.In the next chapter, we shall discussin
greaterdetail the new solutionsthat arise in chaoticmapsat the statisticallevel.
Thesituation whichwe now find ourselvesin is
reminiscent ofthe onewe encounteredin thermodynamics(Chapter 2). The very successof equilibriumthermodynamicshas retarded the discovery ofnew propertiesofmatter in
nonequilibriumsituations wheredissipative structures and
self-organization appear.In parallel, the successofclassical
trajectory theory and quantum mechanicshas hamperedthe
extensionofdynamicsto the statisticallevel in which
irreversibility can beincorporatedinto the basicdescriptionofnature.
Chapter4
THE LAWS OF CHAOS
I
Inthe precedingchapter, we formulated the principal
factor that makesit possiblefor us to extendclassicaland
quantum mechanicsfor unstable dynamical systems:the
breaking of the equivalencebetweenthe individual
description (in terms of trajectories)and the statistical
description (in termsofensembles).We now wish to analyzethis inequivalencemorecloselyfor simple chaoticmapsand illustrate how this observationrelatesto recentdevelopments in mathematics.1Let us first return to theBernoullimap, whichwe have already introducedas an
exampleofdeterministicchaos.We seefrom the equationofmotionxn+x
= 2xn (mod1)that we may calculatexn for arbitrary n oncewe know theinitial conditionxQ.However,an essentialelementofrandomness still appearsto bepresent.An arbitrary numberxbetween0and 1can berepresentedin a binary digital sys-
89
90 TheEnd of Certainty
tern:x =\"f + ~a + IT \342\200\242 \342\200\242 \342\200\242
>wherew. = 0or 1(we are
using the negative indicesu_v u_2 to introducethe bakertransformation, whichwe shall study in SectionIII.Eachnumberx is thus representedby a seriesofdigits.We can
easily verify that the Bernoullimap leadsto the shift w' =u . (for instance,u' ,= u ,) as it moves the numbers u. tothe left. Becausethe value ofeachdigit in the seriesw_pu _2,...is independentof the others,the result ofeachsuccessiveshift is as randomas flipping a coin.This systemis calleda \"Bernoullishift,\" in memory ofthe pioneeringwork in gamesofchancedoneby the great eighteenth-century mathematician, JakobBernoulli.Herewe can alsoobserve a sensitivity to initial conditions:Two numbers
differing only slightly (for example,by w_40, whichmeanslessthan 2~39)will differ by i after 40 steps.As we have
already explained,this sensitivity correspondsto a positiveLyapunov exponentwhosevalue is log2 as x doublesat
eachstep (seeChapter3,SectionII).From the outset,the Bernoullimap introducesan arrow
oftime that can only point in onedirection.If, insteadofxn+1
=2xn (mod1),we considerthe map xn+l
=hxn, we
find a single-pointattractor at x=0.Thetime symmetry is
brokenat the level of the equationofmotion,which is
thus not invertible. This is in contrast to the dynamicalsystems describedby Newton,whoseequations ofmotionare invariant with respectto time inversion.
Themost important point to keepin mind at this
juncture is that trajectoriesare inadequate.They are incapableofdescribingthe time evolution ofchaoticsystemseven if
they are governedby deterministicequationsofmotion.As Pierre-MauriceDuhemstatedas early as 1906,the
notion of trajectory is an adequate modeofrepresentationonly if the trajectoryremainsmoreor lessthe samewhen
The Laws of Chaos 91
we slightly modify the initial conditions.2Thedescriptionofchaoticsystemsin terms oftrajectorieslackspreciselythis robustness.This is the very meaningofsensitivity toinitial conditions:Two trajectoriestaking off from pointsas closetogetheras we can imaginewill divergeexponentially
over the courseoftime.On the contrary, there is no difficulty in describing
chaoticsystemsat the statistical level.It is thereforeat this
level that we have to formulate the lawsofchaos.InChapter 3,we introducedthe Perron-FrobeniusoperatorU,which transforms the probability distribution p (x) into
P\342\200\236+i(x), leadingus to concludethat thereexist newsolutions that are not applicableto individual trajectories.It is
thesenovel solutionsthat we wish to identify in this
chapter. Thestudy ofthe Perron-Frobeniusoperator,whichis
a rapidly growingfield, is ofspecialinterestherebecausechaoticmapsare perhapsthe simplestsystemsthat displayirreversibleprocesses.
Boltzmannapplieshis ideas to gasescontaininganimmense numberofparticles(onthe orderof1023).Here,onthe otherhand, we are dealingwith only a fewindependent variables (onefor the Bernoullimap and two for thebaker map, whichwe shall considershortly).Onceagain,we shall have to rejectthe contentionthat irreversibilityexistsonly becauseour measurementsare limited to
approximations. But first let us identify the new classofsolutions associatedwith the statistical description.
IIHow do we solve a dynamical problemat the statisticallevel?First we needto determinethe distributionfunction
p(x)so that we can observethe recurrencerelation p +1(x)
92 The End of Certainty
= U p (x).The distribution function Pn+l(x) after (n + 1)mapsis obtainedby the actionofthe operatorU on p (x),which is the distribution function after n maps.We shallmeetthe same type ofproblemin classicaland quantummechanics.Forreasonsthat we shall explainin Chapter6,operatorformalism was first introducedin quantumtheory, and then extendedto otherfields ofphysics,most
notably statistical mechanics.An operatoris simplya prescriptionfor how to act on a
given function; as such,it may involve multiplication,differentiation, orany othermathematical operation.Inorderto definethe operator,we must also specifyits domain.Onwhat types of functions doesthe operatoract? Are theycontinuousorbounded?Dothey have othercharacteristicsas well?Thesepropertiesdefine the function space.
In general,an operatorU actingon a function f(x)transforms it into a different function.(Forinstance,if U is
a derivative operator , then Ux2=2x).However,thereare
specialfunctions, known as the eigenfunctions ofthe
operator,which remain invariant when we apply U; they are
multiplied only by a numberknown as the eigenvalue. Inthe above example,e*x is an eigenfunctionto which the
eigenvalue k corresponds.A fundamental theoremin
operator analysisstates that we can expressan operatorin termsofits eigenfunctionsand eigenvalues,both ofwhichdepend on the function space.Of particular importanceisthe so-called\"Hilbertspace,\"whichhas beencarefullyexplored by theoreticalphysicistsworking in quantummechanics. It contains\"nicefunctions\" such as xorsin x,but
not the singular, generalizedfunctions that we shallneedin
orderto introduceirreversibility into the statistical
description. Every new theory in physicsalso requiresnew math-
The Laws of Chaos 93
ematical tools.Here,the basic novelty is our needto gobeyondHilbertspacefor unstable dynamical systems.
After theseinitial considerations,let us onceagainreturn to the Bernoullimap, wherewe can easily derive the
explicitform ofthe evolution operatorU, therebyobtaining
P\342\200\236+1(x) = Upn(x)=|[p\342\200\236(|)
+ Pjr-T1)]-This equationmeans that after
(\302\253
+ 1) iterations,the probabilitypK +1(x)at point x is determinedby the values ofp (x) at points 2
1-j- v
and\342\200\2242\342\200\224
\342\200\242As a consequenceofthe form ofU, if p is aconstantequalto a,pn + ^
is also equalto a,sinceUOL = 0C.
The uniform distribution p =OL, which correspondsto
equilibrium,is the distribution function reachedthroughiteration ofthe shift, for n \342\200\224> \302\260\302\260.
On the contrary, if p (x) =x,we obtain Pk +1(x)=4 + f.1 XIn otherwords, Ux= 4+2wherethe operatorU
transforms the function x into a different function, 4+2-Butwe can easily find the eigenfunctionsas defined above, in
which the operatorreproducesthe same function multi-
pliedby a constant.In the exampleU(x\342\200\224 2) = 2~(x\342\200\224 2)'tne
1 1
eigenfunctionis thereforex \342\200\224
2 and the eigenvalue 2 \342\200\242Ifwe
repeatthe Bernoullimap n times, we obtain U\"(x\342\200\224 2) =
1 1(2)n(x
\342\200\224 2),whichmoves toward 0for \302\253\342\200\224\302\273<*>. Thecontri-bution (x \342\200\224 2) to p(x) is thereforerapidly dampedat a raterelatedto the Lyapunov exponent.Thefunction x \342\200\224
2
belongs to a family ofpolynomialscalledthe Bernoulli
polynomials, denotedas B (x),whichare eigenfunctionsof Uwith eigenvalues(2)\", wheren is the degreeofthepolynomial.3 When p is written as a superpositionofBernoullipolynomials,the polynomialsofa higherdegreedisappearfirst becausetheir dampingfactor is greater.This is thereason that the distributionfunction moves rapidly toward aconstant.In the end,only BQ(x)= 1survives.
94 The End of Certainty
We now needto expressthe distributionfunction p andthe Perron-FrobeniusoperatorU in terms ofBernoullipolynomials.Beforewe describethe result, however, weshould once more emphasize the distinction between\"nice\" functions and \"singular\" functions (alsocalledgeneralized functions or distributions,which are not to beconfusedwith probabilitydistributions),as it playsa crucialrole.The simplestsingular function is the delta function
8(x).As we saw in Chapter1,SectionIII,8(x\342\200\224
x0) is zerofor all values where and infinite wherex = xn.We have already noted that singular functions have to beused in conjunctionwith nicefunctions.Forexample,iffix) is a nicecontinuousfunction, the integral J dxf(x)8(x
\342\200\224
xQ) =f(xQ)has a well-definedmeaning.In contrast, the
integral containinga productofsingular functions, suchas
/ dx8{x\342\200\224 x0)8(x\342\200\224
x0) =8(0)= \302\260\302\260, divergesand is thereforemeaningless.
Our basicmathematical problemis defining theoperator U in terms ofits eigenfunctionsand eigenvalues.Thisis calledthe spectral representationof the operatorU.Oncewe have this representation,we can use it to expressUp,that is, the effect ofthe Perron-Frobeniusoperatoronthe probability distributionp.Herewe find a quiteremarkable situation characteristicof deterministicchaos.We have already found a set ofeigenfunctions,B (x),the
Bernoullipolynomials,whichare nicefunctions, but thereis a secondset,B (x),whichis formedby singular functions
related to derivatives of the 5-function.4To obtain the
spectralrepresentationof U and thereforeU p, we needboth setsofeigenfunctions.As a result, the statisticalformulation for the Bernoullimap is applicableonly to niceprobability functions p and not to single trajectoriesthat correspondto singular distribution functions repre-
The Laws of Chaos 95
sentedby 5-functions.Thespectraldecompositionof Uwhenappliedto a 5-functioncontainsproductsofsingular functions that divergeand are meaningless.Theequivalence betweenthe individual description(in terms oftrajectoriesrepresentedby 5-functions)and the statistical
descriptionis broken.Forcontinuousdistributionp,however, we obtain consistentresultsthat gobeyond trajectorytheory. We can calculatethe rate ofapproachtoequilibrium and thereforeto an explicitdynamical formulation ofirreversibleprocessesthat take placein the Bernoullimap.This outcomeconfirmsthe qualitative discussioninChapter 1,SectionIII.Probability distribution takes intoaccount the complexmicrostructureofthe phasespace.Thedescriptionofdeterministicchaosin terms oftrajectoriescorrespondsto an overidealizationand is unableto expressthe approachto equilibrium.
Herewe already encountersomeofthe mostcriticalissues in modernmathematics.In fact, as we shall see in
Chapters5 and6,the determinationofeigenfunctionsand
eigenvaluesis the central problemofstatistical andquantum mechanics.The aim there,as well as for chaos,is to
expressan operator,such as U, in terms ofits
eigenfunctions and eigenvalues.When we succeedin doingso,weobtain the spectralrepresentationofthe operator.Inquantum mechanics,sucha representationhas beenachievedin
simplesituations in terms ofnicefunctions.We may thenuse Hilbertspace.The associationbetweenquantummechanics and operatorcalculusin Hilbertspaceis so closethat quantum mechanicsis often consideredan operatorcalculusin Hilbertspace.In Chapter6,we shall seethat
this is generallynot the case.Ultimately, to grasp the real world, we must leave
Hilbertspace.In the caseofchaoticmaps,we have to go
96 The End of Certainty
out of Hilbertspacebecausewe needboth the B (x),whichare nicefunctions, and the B (x),whichare singularfunctions.We can then speakofriggedHilbertspace,orGelfandspace.Inmoretechnicalterms,we obtain anirreducible spectral representationof the Perron-Frobeniusoperatoras it appliesexclusivelyto niceprobabilitydistributions, and not to individual trajectories.Thesefeaturesare fundamental inasmuchas they are typical ofunstable
dynamical systems.We shall find themagain in ourgeneralization ofclassicaldynamicsin Chapter5 and quantummechanicsin Chapter6.The physicalreasonsfor whichwe have to leave Hilbertspaceare related to the problemofpersistentinteractionsmentionedabove, whichrequiresa holistic,nonlocaldescription.It is only outside Hilbertspacethat the equivalencebetweenindividual andstatistical descriptionis irrevocably broken,and irreversibility is
incorporatedinto the laws ofnature.
IllThe Bernoullimap is not an invertible system.We
mentioned earlierthat an arrow of time already existsat thelevel ofequationsofmotion.As our main problemis todescribethe emergenceofirreversibility in invertibledynamical systems,we shall now considerthe baker map, orbaker transformation, which is a generalizationof theBernoullimap.Let us take a squarewhosesideshave length1.First we flatten the squareinto a rectanglewhoselengthis 2;then we cut it in half and build a new square.Ifweexaminethe lowerpart ofthe square,we seethat after oneiteration of this process(or mapping),it splits into two
bands (seeFigure4.1).Moreover,the transformation is re-
The Laws of Chaos 97
Figure4.1TheBakerTransformation
\342\200\224j ^_j % \342\200\224\342\200\224i
X X
(a) (b) (c)
versible:The inverse transformation, which first reshapesthe squareinto a rectanglewith lengthi and height2,returns eachpoint to its initial position.
For the Bernoullimap, the equations of motion are
very simple:At eachstep, the coordinates (x,y) become(2x,2) for 0< x < 2 and (2x-1,
\342\200\224y~)for 2 < x < 1.To
obtain the inverse baker transformation, we only have to
permutex and y.In the baker map, the two coordinatesplay different
roles.Thehorizontalcoordinatex is the dilatingcoordinate, which correspondsto the coordinatex in theBernoullimap as it is multipliedby 2 (mod1)at eachmapping.
Thearea ofthe square is preservedbecausewe alsohave a contractingcoordinatey; in the directionof thevertical coordinate,the points draw closertogetherwhilethe squareis beingflattened into a rectangle.Sincethedistance betweentwo points alongthe horizontalcoordinatex doubleswith eachtransformation, it will be multipliedby 2n after n transformations. Ifwe rewrite2\" as e\"10^2, asthe numbern oftransformations measurestime,the
Lyapunov exponentis log2,exactly as in the Bernoullimapconsideredin SectionII.Thereis also a secondLyapunov
98 The End of Certainty
exponentwith the negative value\342\200\224log
2,whichcorresponds to the contractingdirectiony.
The effect ofsuccessiveiterationsin the bakertransformation is worthy ofthe sameattention we gave to themin
the discussionofthe Bernoullimaps(seeFigure3.7).Herewe start with points localizedin a small portionof the
square (see Figure4.2),where we can clearly see the
stretching effect of the positive Lyapunov exponent.As
the coordinatesxand y are limitedto the interval 0\342\200\224 1,the
points are reinjected,leadingto their uniform distribution
throughoutthe square.By numericalsimulation,we arealso able to verify that if we start with the probabilityp (x,y), the distribution moves rapidly toward unifor-
Figure4.2NumericalSimulationofthe BakerTransformationThe maps are ordered accordingto the number of iterations,which representtime. (Thesenumerical simulations are the workofDeanDriebe.)
7
10
11 12
The Laws of Chaos 99
mity, as in the caseofthe Bernoullishift (seeFigure3.8).We can gain a great dealofinsight into the mechanism
ofthe bakertransformation by representingit as a Bernoullishift, as we did in SectionI.Herewe associatewith eachpoint (x,y) ofthe unit squarethe doubly infinite sequenceofnumbers {u}definedby the binary representation
0 oox = x 2^\302\253\342\200\236, y
= 2^ u\342\200\236>
n= -oo\342\200\236= i
whereeachu can take on the values 0or1.Eachpoint x,yis representedby the series...u_2, u_v uQ, uv u2 ... , in
which.. . u_\342\200\236 u_v uQ correspondsto the dilating
coordinate x and uv u2, ... to the contractingcoordinatey. Forinstance,the point x = i,y
= iwill berepresentedby aserieswith u_ j
= 1,u2= 1,with all otheru beingzero.By
inserting theseexpressionsinto the equationsofmotion,we obtain the shift u'= u .,whichis again the Bernoulli
n n\342\200\224Va
shift. We seethat the information containedin the initial
conditionsincludesthe entirepast and future history ofthe system(Figure4.3).
Successiveiterations ofthe baker transformation lead to
fragmentation ofthe shadedand unshadedareas,producing
an increasingnumberofdisconnectedregions.Notethat the digit uQ determineswhether the representativephasespacepoint is in the left half
(wQ=0) orthe right half
(\302\2530
= 1)ofthe unit square.Sincethe digits u , ...can bedeterminedby tossinga coin,the time iterates ofu , u' =
un-i> u\"\342\200\236
=u\342\200\236-7
W^ have the same random properties.This showsthat the processby whichthe point appearsin
the left or right half of the square can be consideredaBernoullishift.
The baker transformation also shares an importantproperty of all dynamical systems,known as recurrence.
100The End of Certainty
Figure4.3Iterationsofthe BakerTransformation
Starting with the partition 0 (calledthe generating partition), werepeatedlyapply the baker transformation. In moving toward the
future, we generatehorizontal bands. Similarly, by moving into
the past, we generatevertical bands.
10 12Past | Future
Generating partition
Considera point (x,y)for whichthe sequence{ujin the
binary digit representationis finite or infinite but periodic,and x and y are then rational numbers.Sinceall u areshifted in the sameway, every state ofthis kind will recycleidentically after a certain periodoftime.The sameholdstrue for mostotherstates.To illustrate this concept,we shall
considerthe binary representationof an irrational point(x,y),whichcontainsan infinity ofnontrivial,
nonrepeating digits.It can beshown that almost all irrationals contain
a finite sequenceofdigitsrepeatedan infinite numberoftimes.Thus, a given sequenceof2m digitsaroundposition0,whichdeterminesthe state ofthe systemto an errorof2~m, will reappearan infinite numberoftimes underthe
effect ofthe shift. Sincem can bemade as large as desired(although finite), almost every state will arbitrarily
approach any point, including,ofcourse,the initial position,an infinite numberoftimes.In otherwords, most ofthe
trajectorieswill traverse the entirephasespace.This is the
famous Poincarerecurrencetheorem,which, together
TheLaws of Chaos 101
with time reversibility, was long advancedas an essential
argument against the existenceof genuinely dissipativeprocesses.However,this view can no longerbesustained.
In summary, the baker transformation is invertible, time
reversible, deterministic, recurrent, and chaotic.Demonstratingthesepropertiesthrough this exampleis especiallyuseful,sincethesesamepropertiescharacterizemany real-worlddynamical systems.As we shall see,despite theseproperties,
chaosallows us to establishgenuineirreversibility by
settingup a descriptionat the statistical level.The dynamicsofconservative systemsinvolve laws of
motionand initial conditions.Herethe laws ofmotionare
simple,but the conceptofinitial data demandsa moredetailed analysis.The initial conditionsofa singletrajectory
correspondto an infinite set {u} (n = \342\200\224\302\260\302\260 to +
\302\260\302\260).
But in
the real world, we can only lookthrough a finite window.Thismeansthat we are ableto controlan arbitrary but
limited numberofdigitsu .Supposethat this window
correspondsto u_?u_2u.[u0
\342\226\240
uxii2uv all other digits beingunknown (thedot indicatesthe separationbetweenx and
y digits).TheBernoullishift u' = u _t impliesthat at thenext step, the previousseriesis replacedby u_4u_3u_2u_l
'u\342\200\236u.u2,
whichcontainsthe unknown digit u_4. Moreprecisely, owingto the existenceofa positiveLyapunovexponent, we needto know the initial positionofthe pointwith an accuracyofN+ n digitsin orderto beable to
determine its positionwith an accuracyofN digitsafter n
iterations.
As we saw in Chapter 1,the traditional meansofsolving
this problemwould be to introducea coarse-grainedprobability distribution, which is not defined by singlepoints, but rather by regions,as originally proposedbyPaul and Tatiana Ehrenfest.5However, two points on an
102The End of Certainty
expandingmanifold, even if not distinguishablebymeasurements ofa given finite precisionat time 0,will beseparated, and thus observable,over time.Traditional coarsegraining thereforecannotbeappliedto the dynamicalevolution. This is oneof the reasonsfor whichwe needamoresophisticatedmethod.
First, however, we should analyze in moredetail whatthe approachto equilibriummeans in terms ofthe bakertransformation.6 In spite ofthe fact that this
transformation is invertible, as are all dynamical systems,theevolutions for t \342\200\224> + \302\260\302\260 and t \342\200\224>
-\302\260\302\260 are different. For< \342\200\224> +
\302\260\302\260,
we move toward increasinglynarrow horizontal bands (seeFigure4.3).In contrast,for t \342\200\224>
\342\200\224
\302\260\302\260,
we move toward
increasingly narrow vertical bands.We see that for chaoticmaps, dynamics lead to two
typesofevolutions.We thus obtain two independentdescriptions, onecharacterizingthe approachto equilibriumin our future (for t \342\200\224> +
\302\260\302\260),
and the otherin ourpast (for t
\342\200\224>
\342\200\224
\302\260\302\260).
Such dynamical decompositionis possible forboth chaoticmaps and nonintegrableclassicalandquantum systems,as we shall seelater on.Fora simpledynamical system,whethera harmonicoscillatoror a two-bodysystem,suchdecompositiondoesnot exist;future and pastcannotbedistinguished.Which ofthe two descriptionsforchaoticmaps should we retain? We shall comeback
repeatedly to this question.Forthe moment,let us take intoaccountthe inherentuniversality that every irreversible
processhas in common.All arrowsoftime in nature havethe same orientation:They all produceentropy in the
samedirectionoftime,whichis by definition the future.
We thereforehave to retain the descriptioncorrespondingto equilibriumreachedin our future, that is, for t \342\200\224> + \302\260\302\260.
TheLaws of Chaos 103
In Chapter1,we mentionedthe time paradoxassociated with the bakermap:While the dynamicsdescribedbythis map are time reversible,irreversibleprocessesdo
appear at the statistical level.As in the Bernoullimap, we canintroducethe Perron-FrobeniusoperatorU defined by
Pn + \\{x,y)= U pn(x,y).But there is a fundamental
difference. A generaltheoremstates that for invertibledynamical systemsthereexistsa spectralrepresentation,definedonHilbertspace,whichinvolves only nicefunctions.7
Moreover, in this representationthere is no damping, as the
eigenvaluesare modulo1.Sucha representationalso existsfor the baker transformation, but it is not ofinterestto us
becauseit offers no new information regardingtrajectories. We simply comeback to 8(x\342\200\224
xn + j)<5(y\342\200\224
yn + t)=
US(x\342\200\224 x )S(y\342\200\224
y ), a solution that is equivalent to the
trajectory description.8Exactlyas we did for the Bernoullimap, we have to go
out ofHilbertspaceto obtain additionalinformation.Forspectralrepresentationsin generalizedspace,which have
recentlybeenobtained,the eigenvaluesare the same (i)mas for the Bernoullimap.9 Moreover,the eigenfunctionsare singular functions, such as the B (x) for the Bernoullimap. Again, theserepresentationsare irreduciblein that
they apply only to suitable test functions, obligingus tolimit ourselvesto continuousdistributionfunctions.Singletrajectoriesdescribedby singular 5-functionsare excluded.As is the casein the Bernoullimap, the equivalencebetween the individual descriptionand the statisticaldescription is broken.Only the statisticaldescriptionincludesthe
approachto equilibriumand thereforeirreversibility.Forthe bakermap, thereis oneimportant new element
involved, however, in comparisonto the Bernoullimap:
104 The End of Certainty
The Perron-Frobeniusequation can be applied to bothfuture and past (pn + 1= Upn and Pn_1=U_1pK; hereU\"1 is
the inverse of U).In the realm ofHilbertspacespectralrepresentations,this makesno difference becauseU\"\\
+n2 =U\"\\ U\"i, whatever the signofn1 and
\302\2532 (rememberingthat
the positive sign refers to the future, and the negative tothe past). Hilbertspacecan be describedas a dynamical
group. In contrast, for irreduciblespectralrepresentations,there is an essentialdifference betweenfuture and past.The eigenvaluesof U\" are expressedas (jm)\"
= e~n(-m log 2).This formula correspondsto damping in the future (n >0),and divergencein the past (n <0).Therenow existtwodifferent spectralrepresentations\342\200\224one for the future, andthe other for the past. Thesetwo time directions,whichare containedin the trajectory description(or Hilbertspace),are now disentangled.The dynamical group is
thereby brokeninto two semigroups.As previouslymentioned, in accordancewith our view that all irreversible
processesare orientedin the samedirection,we have toselect the semigroupin whichequilibriumis reachedin ourown future. Nature itself is describedby a semigroupthat
distinguishesbetweenpast and future. Thereis an arrow oftime.As a result,the traditional conflict betweendynamicsand thermodynamicsis eliminated.
In summary, as longaswe are consideringtrajectories,it
seemsparadoxicalto speakoflaws ofchaosbecausewe are
dealingwith the negative aspectsofchaos,such as the
exponential divergenceof trajectories,which lead to un-computability and apparent lawlessness.The situation
changesdrastically when we introducethe probabilisticdescription,which remains valid and computableat all
times.It is thereforeat the probabilisticlevel that the lawsofdynamicshave to be formulated for chaoticsystems.In
The Laws of Chaos 105
the simpleexamplesstudiedabove, irreversibility is linked
only to Lyapunov time,but ourresearchhas recentlybeenextendedto moregeneralmaps that includesuchirreversible phenomenasuch as diffusion and various othertransportprocesses.10
IVAs mentionedin Chapter1,the successofthe statistical
descriptionwhen applied to deterministicchaosstemsfrom the fact that it takes into accountthe complexmi-crostructureofphasespace.In eachfinite regionofphasespace,thereare exponentiallydiverging trajectories.Thevery definition of the Lyapunov exponentinvolves the
comparisonofneighboringtrajectories.It is remarkablethat irreversibility already emergesin simplesituations
involving only a few degreesoffreedom.This is,ofcourse,a blow to the anthropomorphicinterpretation ofirreversibility based on approximationsthat we ourselvesare
supposed to introduce.Unfortunately this interpretation,whichwas formulated after the defeat ofBoltzmann,continues to bepropagatedtoday.
It is true that thereis still a trajectory descriptionifinitial conditionsare known with infinite precision.But this
doesnot correspondto any realistic situation.Whenever we perform an experiment,whetherby computerorsomeothermeans,we are dealingwith situations in whichthe initial conditionsare given with a finite precisionand
lead, for chaoticsystems,to a breakingoftime symmetry.Similarly, we couldimagineinfinite velocities,andtherefore we would no longerneedrelativity theory, which is
based on the existenceofa maximumvelocity\342\200\224the
velocity oflight c in the vacuum\342\200\224but the assumptionofve-
106TheEnd of Certainty
locitiesgreaterthan ccorrespondsto no known observable
reality.Maps are idealizedmodelsthat cannotcapture time's
true continuity. As we now turn ourattention to morerealistic situations, ofspecialimportanceto us will benon-integrablePoincaresystems,wherethe breakbetweentheindividual description(trajectoriesorwave functions) andthe statistical descriptionis even morestriking.Forthesesystems, the Laplace demonis powerless,whether his
knowledgeofthe present is finite or infinite. Thefuture is
no morea given; it becomesa \"construction,\"to use an
expressionofthe FrenchpoetPaul Valery.
Chapter5
BEYOND NEWTON'S LAWS
I
Havinganalyzed maps that representsimplified models
in Chapter 4, we cometo the questionat the veryheart ofourquest:What is the roleofinstability andpersistent interactions in the framework ofclassicaland quantummechanics?Classicalmechanicsis the scienceupon whichour beliefin a deterministic,time-reversibledescriptionofnature is based.In respondingto this question,we must first
grapplewith Newton'slaws,the equationsthat havedominated theoreticalphysicsfor the past threecenturies.
Quantum mechanicslimits the validity ofclassicalmechanics when applied to atoms and elementaryparticles.Relativity shows that classicalmechanicsalso has to bemodifiedwhen dealingwith high energiesor cosmology.Whatever the situation, we may introduceeitheranindividual description(in termsoftrajectories,wave functions,orfields) ora statisticaldescription.Remarkably,at all
levels, instability and nonintegrabilitybreak the equivalence
107
108The End of Certainty
ofboth descriptions.Consequently,we have to revise theformulation ofthe laws ofphysicsin accordancewith the
open,evolving universe in whichmankindlives.As stated previously, our positionis that classical
mechanics is incompletebecauseit doesnot includeirreversible processesassociatedwith an increasein entropy.Toincludetheseprocessesin its formulation, we must
incorporate instability and nonintegrability.Integrablesystemsare the exception.Starting with the three-bodyproblem,most dynamical systemsare nonintegrable.Forintegrablesystems, the two modesof description\342\200\224the trajectorydescription,basedon Newton'slaws, and the statistical
description, basedon ensembles\342\200\224are equivalent.Fornonintegrable systems,this is not so.Even in classicaldynamics,then,we have to use the Gibbsianstatistical approach(seeChapter1,SectionIII).As we saw in Chapter3,SectionI,it is this approachthat leads to the dynamicalinterpretation ofequilibriumthermodynamics.It is thereforequitenatural that we alsohave to employ the statisticaldescription to includeirreversibleprocessesdriving systemsto
equilibrium.In this way we canincorporateirreversibilityinto dynamics.As a result, there appear non-Newtoniancontributionsthat can beconsistentlyincludedin
dynamics at the level of the statistical description.Moreover,thesenewcontributionsbreak timesymmetry. We
therefore obtain a probabilistic formulation of dynamics bymeans of which we can resolve the conflict betweentime-reversibledynamicsand the time-orientedview ofthermodynamics.
We are well aware that this step representsa radical
departure from the past.Trajectorieshave always beenconsidered primitive, fundamental toolsof the trade. This is
no longerthe case.We shall encountersituations where
Beyond Newton's Laws 109
trajectories\"collapse,\"to borrowa termfrom quantummechanics(seeSectionVII).
In hindsight, it is not surprising that we have had toabandon the trajectory description.As we saw in Chapter1,nonintegrability is due to resonances,which expressconditionsthat must be satisfied by frequencies.They arenot localevents that occurat given points in spaceand at a
given instant in time.As such,they introduceelementsthat
are quite foreign to the localtrajectory description.Instead, we needa statistical descriptionto formulatedynamics in situations wherewe expectirreversibleprocessesand thereforean increasein entropy.Suchsituations, after
all, are what we seein the world around us.Indeterminism,as conceivedby Whitehead, Bergson,
and Popper,now appearsin physics.This is no longerthe
result ofsomea priorimetaphysicalchoice,but rather the
needfor a statistical descriptionofunstable dynamicalsystems. Overthe past decades,many scientistshave proposedreformulations or extensionsofquantum theory.But the
fact that we now needto extendclassicalmechanicsaswellis quite unanticipated.Even moreunexpectedis the
realization that this revision ofclassicalmechanicscan guideus
in extendingquantum theory.
IIBeforewe begin our revision of Newton'slaws, let us
summarizethe fundamental conceptsofclassicalmechanics. Considerthe motionofa point ofmassm. With the
passageoftime,its trajectory is describedby its position,r(t), its velocity, v = dr/dt, and its acceleration,a = d2r/dt2.Newton'sbasic equationrelatesaccelerationa to forceFthrough the formula F = ma. This formula includesthe
110The End of Certainty
classicalprincipleof inertia, that is, where there is noforce,there is no acceleration,and the velocity remainsconstant.Newton'sequationremains invariant when weshift from oneobserver to anotherwho moves at aconstant velocity with respectto the first. This is known as theGalilean invariance, which has beenradically altered by
relativity, as we shall seein Chapter8.Herewe are dealingwith Newtonian,nonrelativistic physics.
We seethat time takes its placein Newton'sequationonly by meansofa secondderivative. Newton'stime,so to
speak, is reversible,and future and past assumethe samerole.Moreover,Newton'slaw is deterministic.
Now considera moregeneralsituation in whichasystem is formedby N particles.In three-dimensionalspace,we have the 3N coordinatesqv . . . ,q3N and the
corresponding velocitiesvy . . . ,v3N. Inmodernformulations ofdynamics,we usually defineboth the coordinatesandvelocities (orbetter,the momentapv . . . ,p3N, wherein
simplecasesp = mv) as independentvariables.As in Chapter1,
the state ofthe dynamical systemis then associatedwith
a point in phase space,and its motionwith a trajectoryin this space.The most important quantity in classical
dynamicsis the Hamiltonian H, which is defined as the
energyofthe systemexpressedin terms ofthe variables qAnd p. In general,Histhe sumofthe kineticenergyEkm(p)and the potential energy V(q), wherep and q signify theentireset ofindependentvariables.
Oncewe have obtainedthe HamiltonianH(p,q), we canderive the equationsofmotionthat determinetheevolution ofcoordinatesand momentaover the courseoftime.This procedureis familiar to all students ofmechanics.Such equations, as derived from the Hamiltonian,are
calledthe canonicalequations of motion.Contrary to
Beyond Newton's Laws 111
Newton'sequations,whichare ofthe secondorder(that
is, they containthe secondtime derivative), Hamiltonian
equationsare ofthe first order.Fora singlefree particle,H\342\200\224 j^, the momentump is constant over time,and thecoordinate varies linearly when time q =
q0 + ^t. Bydefinition, for integrablesystems, the Hamiltonian can beexpressedonly in terms ofmomenta(if necessary,after an
appropriatechangeofvariables).PoincarestudiedHamil-tonians in the form H=HQ{p)+ \"kV(q), whichis the sumofan integrablecontribution(the \"freeHamiltonian\" HQ)and a potential energy due to interactions (A, is a scalingfactor that will beused later on).Heshowedthat this classofHamiltoniansis generallynot integrable,whichis to saythat we cannoteliminateinteractionsand goback to
independent units. We already mentionedin Chapter1 that
nonintegrabilityis due to divergingdenominatorsassociated with Poincareresonances,as a result ofwhichwecannot solve the equationsofmotion(at least in powersofthe
couplingconstant A,).
In the following pages,we shall concernourselves
primarily with nonintegrablelargePoincaresystems(LPS).As
we have seen,Poincareresonancesare associatedwith
frequencies correspondingto various modesofmotion.Afrequency(0k dependson the wavelength k. (Usinglight asan example,ultraviolet has a higher frequency (0 and
shorterwavelength k than infrared light.)When weconsider nonintegrablesystemsin whichthe frequencyvaries
continuouslywith the wavelength, we arrive at the verydefinition ofLPS.This conditionis metwhenthe volumein whichthe systemis locatedis great enoughfor surfaceeffects to be ignored.This is why we call thesesystemslargePoincaresystems.
A simpleexampleofLPSwould be the interactionbe-
112The End of Certainty
tween an oscillatorwith frequency C0l coupledwith a
given field.In this centuryofradioand television,we haveall heard the termelectromagneticwaves. Theamplitudeofthesewaves is definedby a field describedby a function
<p(x,t) ofpositionand time.As was establishedat the
beginning ofthe century, a field can bethought ofas the
superposing of oscillations with frequenciesft)fewhose
wavelength k varies from the sizeofthe systemitselfto thedimensionsofelementaryparticles.In the oscillator-fieldinteractionthat we are considering,resonancesappeareachtime a field frequencyft)fe
is equalto the oscillatorfrequency C0l.When we try to solve the equationsofmotionofthe oscillatorin interactionwith the field, we encounterPoincareresonancesiw _ w\\, whichcorrespondtodivergences whenever
C0X= ft),. In otherwords,thesetermstend
toward infinity and thereforebecomemeaningless.As weshall see,we can eliminatethesedivergencesin ourstatistical description.
Poincareresonanceslead to a form ofchaos.Indeed,innumerable computersimulations have shown that theseresonanceselicitthe appearanceofrandomtrajectories,asis the casefor deterministicchaos.In this sense,there is acloseanalogy betweendeterministicchaosand Poincarenonintegrability.
IllAs in previouschapters,we shall considerthe probabilitydistributionp(q,p,t), whoseevolution over time can easilybederivedfrom the canonicalequationsofmotion.We are
now in the same situation as we were for chaoticmaps,wherewe replacedthe equationsofmotionwith statistical
descriptionsassociatedwith the Perron-Frobeniusoperator. In classicalmechanics,we also encounteran evolution
Beyond Newton's Laws 113
operatorknown as the Liouville operatorL, whichdetermines the evolution ofp through the equationi-\302\243
= Lp.Thetime changeofp is obtainedby acting onpwith the
operatorL. If the distribution function is timeindependent
-\302\243
= 0,then Lp = 0.This correspondstothermodynamic equilibrium.As we saw in Chapter3,SectionI,pthen dependson only the energy (or the Hamiltonian),whichis an invariant ofmotion.
The solution of dynamical problems at the statisticallevel requiresdeterminingthe spectralrepresentationofL,as was explainedin Chapter4 for chaoticsystems.We
thereforehave to defineits eigenfunctionsand eigenvalues.We have seenthat spectralrepresentationdependson thefunctions which,as used in the past (and still appropriatefor integrablesystems),are in Hilbertspace,the spaceof\"nice\" functions.Accordingto a fundamental textbooktheorem,operatorL has real eigenvalues/ in Hilbertspace.In this case,evolution over time provesto be a
superposition ofoscillatoryterms.In fact, the formalsolution ofthe Liouvilleequation is p(t) \342\200\224 exp(\342\200\224itL)p(0). Theoscillatoryterm
exp(\342\200\224itl)
= costl \342\200\224 i sin tl is associatedwith eigenvalue /,wherefuture and past play the samerole.In orderto includeirreversibility,we needcomplexeigenvalues such as / = CO
\342\200\224 iT , which lead to exponentialdamping e_)\"' for time evolution.This contributionprogressively diminishesin the future (t>0) but is increasedinthe past (t < 0),and thus time symmetry is broken.
However, obtaining complexeigenvaluesis possibleonly whenwe leave Hilbertspace.Our main objectiveisnow to understandfor whichphysicalreasonswe have todo so.This follows from the inescapablefact that thereare
persistent interactionsin the natural world.1When weconsider the roomin whichwe sit, the moleculesin the
atmosphere are constantly colliding.This is quite different
114TheEnd of Certainty
from transitory interactions,suchas a finite numberofmolecules in a vacuum. Themoleculesthen interact over afinite periodof time,and eventually may escapeinto
infinity. Thedistinctionbetweenpersistentand transitoryinteractionstakeson a crucialimportancein moving fromclassicaldynamicsto thermodynamics.Classicaldynamicsextracts a given numberofparticlesand considerstheirmotion in isolation;irreversibility occurswheninteractions never cease.In short, dynamicscorrespondsto areductionist point of view in the sensethat we considera finite number of moleculesin isolation.Irreversibility emergesfrom a moreholistic approach in whichweconsidersystemsdriven by a largenumberofparticlesas awhole.Inmakingthis distinctionmoreprecise,we shall
indicate why we needsingular distribution functions andmust thereforeleave Hilbertspace.
IVTransient interactionsmay bedescribedby localizeddistribution functions.To describepersistent interactions in a
largespacesuch as the atmosphere,we needdelocalizeddistribution functions. In defining more precisely thedistinctionbetweenlocalizedand delocalizeddistributionfunctions p, let us beginwith a simpleexample.In a one-dimensionalsystem,the coordinatex extendsfrom \342\200\224
\302\260\302\260 to+00.Localizeddistributionfunctions are concentratedon afinite sectionofthe line.A specialcaseis a singletrajectorythat is localizedat a given point and moves alongthe lineover the courseoftime.In contrast,delocalizeddistribution functions extendover the entire line.Thesetwo
classesoffunctions describevarious situations.As an
example, let us considerscattering. In the usual scatteringexperiments, we preparea beamofparticlesthat we shootat
Beyond Newton's Laws 115
an obstacle(the scattering\"center\.")We then have thethreestagesrepresentedin Figure5.1.
In this experiment,the beamfirst approachesthe
scattering center,then interacts with it, and is finally in freemotionagain.Theimportant point hereis that theinteraction processis transient.Fordelocalizeddistributions,onthe otherhand, the beamextendsover the entireaxis,and
scatteringneitherstarts nor stops.We then have what wecall persistentscattering.
Transient scatteringexperimentshave playedasignificant part in the history ofphysicsby allowingus to studythe interactionsbetweenelementaryparticlessuch asprotons and electrons.Still,in many situations\342\200\224particularly in
macroscopicsystemssuchas gasesorliquids\342\200\224we
havepersistent interactionsbecausecollisionsnever cease.In sum,transient interactionsare related to localizeddistributionfunctions, suchas trajectories,while persistentinteractionsare related to delocalizeddistributions,whichextendoverthe entiresystem.
Thermodynamicsystemsare characterizedby persistent
Figure5.1TheThreeStagesofScattering(a) The beam approachesthe scattering center, (b) The beamintersects the scattering center, (c)The beam is onceagain in free motion.
116The End of Certainty
interactions,and must thereforebe describedby delocal-izeddistributions.In defining thesesystems,we have toconsiderthe thermodynamic limit, wherethe numberofparticles JV and the volume V are increased,while theirratio,the concentrationN/l{remains constant.Although
formally we considerthe limits JV \342\200\224>
\302\260\302\260,
V \342\200\224>
\302\260\302\260,
thereare, ofcourse,no dynamical systems\342\200\224not even the universe\342\200\224
wherethe numberofparticlesis infinite. This limit simplymeansthat surface effectsdescribedby the terms of^orpcanbe ignored.The thermodynamic limit plays a centralrolein all macroscopicphysics.Without this concept,wecouldnot even define states ofmatter such as gases,liquids, or solids,or describethe phase transitions betweenthesestates ofmatter. We would also beunable to
distinguishbetweennear-equilibriumand far-from-equilibrium
situations, whichwerediscussedin Chapter2.We shall now illustrate why the introductionofdelocal-
ized distribution functions forces us to leave the classofnicefunctions and thereforeHilbertspace.In orderto do
so, we have to considerseveral elementarymathematicalnotions.In the first place,every student ofmathematicsis familiar with periodicfunctions such as sin (~^)- Thisfunction remainsinvariant whenwe add to the coordinatex the wavelength A,, as
2kx \342\226\240 2k(x + X)sin -y \342\200\224 sin \342\200\224
j^\342\200\224.
Otherperiodicfunctions are cos-=^,orthe morecomplexcombination
2nx,\342\200\224
2kx . \342\226\240 \342\226\240 2kxe *\342\226\240
= cos-y + i sin -yInsteadofthe wavelength A,, we often use the wave vectork =y.The exponentiale'kx is calleda planewave.
In the secondplace,the classicaltheory ofFourierseries
Beyond Newton's Laws 117
(orFourierintegrals) demonstratesthat a function ofthecoordinatex, whichwe shall callj(x),can beexpressedasa superpositionof periodicfunctions correspondingtowave vectorsk, or morespecifically,as a superpositionofplane waves e'kx. In this superposition,eachplane wave is
multipliedby an amplitude cp(fe), which is a function ofk.
Thisfunction <p(k) is knownas the Fouriertransform o{j(x).In short, we can go from a function J(x)ofcoordinatex
to a description<p(k) in wave vectorsk. Ofcourse,theinverse transformation is equally possible.It is also importantto note that there is a kind of duality betweenj{x)and
<p(k). KJ{x)extendsover a spatial interval Ax (and vanishes
outside),(p(fe) extendsover the spectral interval Ak ~^ .When the spatial interval Ax increases,the spectralintervalAk decreases,and vice versa.2
In Chapter1,SectionIIIand Chapter3,SectionII,wedefined the singular function 8(x).As we saw, 8(x)differs
from zeroonly at x=0.Thespatial interval Ax is thereforezero,and whenAk ~^r, the spectralinterval is infinite.
Inversely, delocalizedfunctions for whichAx \342\200\224>
\302\260\302\260 leads to
singular functions in k such as 8(fe).Thus,delocalizeddistribution functions are an essentialelementin describingpersistent interactions.At equilibrium,the distributionfunction p is a function ofthe HamiltonianH(seeChapter 3,SectionI).TheHamiltoniancontainsthe kineticenergy that is a function ofthe momentap and not ofthe
coordinates,and thus includesa delocalizedpart that has a
singular Fouriertransform. It is hardly astonishingthat
singular functions play a criticalrolein our dynamicaldescription. Indeed,it is our needfor thesefunctions that
forcesus to leave Hilbertspace.Equilibrium distributionsthat are functions ofthe Hamiltonianare already outsideHilbertspace.
118The End of Certainty
V
Let us now comparethe trajectory descriptionwith thestatistical descriptionin terms of the Liouville operator(seeSectionIII).Herewe are in for quite a surprisebecause the statistical descriptionintroducescompletelydifferent concepts.This is obviouseven in the simplestcasewherewe considerthe motionofa free particlealongaline.As we saw in SectionII,the coordinateq oftheparticle varies linearly over time,whilethe momentumpremains constant.On the contrary, the statistical descriptionis defined in terms ofthe wave vectors k, associatedwith
the Fouriertransform ofq, and the momentump.We areused to dealingwith wave vectorswhenwe studyacoustical or opticalproblems,but herewave vectorsappear in a
problemofdynamics.Thereasonis that for a free particle,the Liouville operatorL is simply a derivative operator,whereL = ^ gx.As we noted in Chapter4, SectionI,the
eigenfunctionsare then exponentialsexp (ikx) and the
eigenvalues .Theeigenfunctionexp (ikx) is a periodicfunction, orplanewave, sinceexp(ikx) =coskx + i sin kx.It extendsover the entirespace,in strikingcontrastwith a
trajectory localizedat a singlepoint.The solutionofthe
equationofmotionfor a free particleis obtainedin thestatistical descriptionthrough a superpositionof planewaves.Ofcourse,in this simpleexample,the twodescriptions are expectedto be equivalent. Usingthe theory ofFouriertransformation, we can reconstructthe trajectorystarting with planewaves (seeFigure5.2).Becausethe
trajectory is concentratedat onepoint,we have to superposeplane waves extendingover the entirelengthofthespectral interval (Ak \342\200\224>
\302\260\302\260).
As a result,for q = q0, the amplitudesofthe planewaves
Beyond Newton's Laws 119
Figure5.2SuperpositionofPlaneWaves
Trajectories resulting from the superpositionof plane wavesthrough constructive interference lead to a function characterizedby dramatic peaking around q
= 0.
M M
/^aS^>n<(a) (b)
increasethrough constructive interference,while for q ^ qQ,
they vanish through destructive interference.In integrablesystems,the wave vectork is constantover time.Bysuperposing
the plane waves, we can reconstructtrajectoriesat
any moment.But the important point to considerhereis
that the trajectory is no longera primitive concept,but
rather a derivedconceptas a constructofplanewaves. It is
thus conceivablethat resonancesmay threaten theconstructive interferencesleadingto a trajectory. This couldnot beconsideredas longas the trajectorywas treatedas a
primitive, irreducibleconcept.Given that a trajectory is
representedby a point in phasespace,we can seethat the
collapseoftrajectorieswould correspondto a situation in
whicha point decomposesover time into a multiplicity ofpoints, exactly as in the diffusion processwe analyzed in
Chapter1.Thesame initial conditionwould then lead toa multiplicity oftrajectories,as was also the casein thediffusion process.
120The End of Certainty
Theeigenvalues-\302\243
ofthe Liouvilleoperatorcorrespondto the frequenciesappearingin Poincareresonances.Theydependon both k and p, and not on the coordinates.Theuseofwave vectork is thereforea logicalstarting point for
discussingthe roleof these resonances.By using planewaves, we candescribenot only trajectories(whichcorrespond
to transient interactions),but alsodelocalizedsituations. As we have seen,this leads to singular functions in
the wave vectork. Let us now examinethe effect ofinteractions on the statistical descriptionby employingthe
language ofwave vectors.
VI
Supposethat the potential energy V in the Hamiltonianis
the sum ofbinary interactions.It then follows from well-establishedtheoremsthat interactionsbetweenparticlesjand n modify the two wave vectors k. and k , while theirsum is conserved,giving us the conservationlaw k. + k =k'.+ &',wherek\\ and k' are the wave vectorsafter inter-j n> J n
action.3We are able to describedynamical evolution within the
statistical formalism pictoriallyby consideringa successionofevents separatedby free motion.At eachevent, the wavevectors k and momentap are modified;between the
events, they remain constant.Let us now examinethenature oftheseevents in moredetail.
In Chapter3,SectionI,we introducedthe notionofcorrelations,whichwe shall now define with greaterprecision. Thedistributionfunction p(q,p,t) dependson bothcoordinatesand momenta.If we integratethis function
over the coordinates,we loseall information about the
position ofparticles,and thus correlations,in space.We
obtain a function p0(p, t), which offers information only
Beyond Newton's Laws 121
about momenta.Forthis reason,p0 is known as the vacuum
ofcorrelations.On the otherhand,by integrating over all
coordinates exceptthe coordinatesq.,q.ofparticlesi and j,we retain the information about possiblecorrelationsbetween particlesi andj.This function, p2, is calleda binarycorrelation.We can defineternary correlationsand beyondin a similar way. In the statisticaldescription,it is importantto replacethe coordinates,whichdependon thedistribution functions through theirFouriertransform, with wave
vectorsas they appearin the spectraldecompositionoftheLiouville operator.
We shall now take into accountthe law ofconservationofwave vectors,in whicheachevent is representedby a
point,with two entry lines,k., k , and two exitlines,k'.,k' , wherek. + k = k'.+ k' .Moreover,at eachpoint, then' j n j n rmomentap ofthe interactingparticlesare modified,and a
derivative operatorg- appears.Thesimplestevent ofthis
kind is illustrated in Figure5.3.We call the diagramin Figure5.3a propagationevent,
or propagationdiagram.This correspondsto a
modification ofthe binary correlationp2betweenparticlesj and n.
But we can also start from the vacuum ofcorrelationsp0,in which k. = k = 0, and producea binary correlationp,,, , with k.+k =0 to conservethe sumofthe wave vec-
kj kn j n
Figure5.3PropagationDiagramA dynamical event correspondingto the interaction of two
particles leadsfrom wave vectorski, kn to k',-, k'\342\200\236.
122TheEnd of Certainty
Figure5.4CreationFragmentA dynamical event transforms the vacuum of correlationsinto a
binary correlation/, \342\200\224/.
k'\342\200\236=l
tors (seeFigure5.4).We then have what is known as acreation ofcorrelationdiagram,orcreationfragment. We alsohave destructionfragments, as presentedin Figure5.5,which transform binary correlationsinto the vacuum ofcorrelations.4
We now beginto seedynamicsas a history ofcorrelations.
Figure5.6represents,for example,the emergenceof a
five-particlecorrelationstarting from the vacuum ofcorrelations. Events associatedwith interactionsproducecorrelations.
We can now introducethe effect ofPoincareresonancesinto the statistical descriptionof dynamics.Theseresonances coupledynamical processesexactlyas they couple
Figure5.5DestructionFragmentA dynamical event transforms the binary correlation/, -/ into the
vacuum of correlations.
kj= -l
Beyond Newton's Laws 123
Figure5.6EvolutionofCorrelationsThe four events at points 0X, 02,03, 04 transform the vacuum ofcorrelationsinto a five-particlecorrelation.
harmonicsin music.In our description,they couplecreation and destructionfragments (seeFigure5.7),whichleads to new dynamical processesthat start from a givenstate ofcorrelations(ofwhichthe vacuum ofcorrelationsis merely oneexample)and eventually return to exactlythe samestate.In Figure5.7,thesedynamical processesare
depictedas bubbles.While the state ofcorrelationsis
preserved, the distributionofmomentais changed(remembering
that eachvortex introducesa derivative operatorgr).Thesebubbles correspondto events that must be
consideredas a whole.They introducenon\342\200\224Newtonian ele-
124 The End of Certainty
Figure5.7BubbleDueto PoincareResonancesPoincareresonancescouplethe creationand destruction ofcorrelations, and leadto diffusion.
Destruction ofcorrelations Creation ofcorrelations
ments in that no analogueof such processesexists in
trajectory theory.Suchnew processeshave a dramaticeffect on dynamicsbecausethey break time symmetryIndeed, they lead to the type ofdiffusion that had alwaysbeenpostulated in phenomenologicaltheoriesofirreversible processes,includingBoltzmann'skineticequation.To mark the parallel with the phenomenologicaldescription, we have calledthe new elementscollision operators.They act on the distributionfunctions.*
*Wesaw in Chapter 1,Section III that Poincare resonances between frequencies
lead to divergences with small denominators. Herethe frequency of a particle ofmomentum p is kp/m, where fe is the wave vector (seeSection IV). For LPS, in
which fe is a continuous variable, we can avoid the divergences and express the
resonances in terms of 6-functions. This involves a branch of mathematics
associated with analytical continuation (seethe references in the chapter notes). For
a two-body process, the argument of the 6-function is fe/m(p{
\342\200\224
p2), leading to
contributions whenever the frequencies kpl/m and kp2/m are equal, and
otherwise vanishing. Thewave vector fe = 0 therefore plays an especially important
role wherein the argument of the 6-function vanishes, remembering that S(x)=\302\260\302\260 for x= 0 and S(x)= 0 for x * 0.A vanishing wave vector fe corresponds to an
infinite wavelength, and thus to a process that is delocalized in space.Hence,Poincare resonances cannot be included in the trajectory description.
Beyond Newton's Laws 125
Our approach includesthe usual kinetic theory, but
only as a specialcase.Traditionally this theory, asintroduced by Maxwell,was centeredaround the evolution ofthe velocity distribution,whereit appearedthat only a fewcollisionswould be sufficient to reestablishequilibriumifdisturbed at the initial time.Our approach, on the
contrary, takesinto accountthe progressivebuildup ofhigherand highercorrelationsinvolving moreand moreparticles.This processrequireslong time scales,in agreementwith
the numericalsimulations that have been available for
many years.5As a result,irreversibility leadsto longmemory
effects that profoundlyalter macroscopicphysics.6Many new resultsthat gobeyond the traditional kinetic
theory have already beenobtained.However,it is outsidethe scopeofthis bookto describethem.They will becovered in greaterdetail in a separatemonograph.7
Suffice it to say that we are beginningto understandwhat irreversibility really means.Let usconsiderthe simpleanalogy ofthe agingprocess.On our time scale,the atomsthat makeup ourbodiesare immortal.What is changingisthe relation between the atoms and molecules.In this
sense,agingis a property ofpopulations,and notindividuals. This is also true ofthe inanimate world.
VIILet us now return to our originalobjective,which is the
solutionofthe dynamical problemat the statistical level in
terms ofthe distribution function p.As was the casefordeterministicchaos,this solution involves the spectralrepresentation of the evolution operator,which in classical
dynamicsis the Liouvilleoperator.First we considerdelo-calizeddistributionfunctions associatedwith persistentinteractions that lead to singular functions (seeSectionsIII
126The End of Certainty
and IV).As a result,we have to leave Hilbertspace,whichis limited to localizednicefunctions.We then introducePoincareresonances,which,as we saw in SectionVI, leadto new dynamical processesconnectedwith diffusion.
Oncewe have includedthesetwo features, we obtain an
irreducible,complexspectral representation.Again,
complex meanstimesymmetry is broken,and irreducible meanswe cannotreturn to a trajectory description.The laws ofdynamicsnow take on new meaning.By incorporatingirreversibility they expressnot certitudesbut possibilities.Only if we relaxour conditionsand considerlocalizeddistribution functions associatedwith a finite numberofparticlescanwe recoverthe Newtoniantrajectorydescription. But in generaldiffusion processesdominate.
Thereare thereforemany situations in whichwe can
expectdeviations from Newtonianphysics,and whereourpredictionshave already beenverified by extensive
computer simulations.In SectionIV, we introducedthe
thermodynamic limit, wherethe numberofparticlesJV \342\200\224>
\302\260\302\260
and volume V \342\200\224>
\302\260\302\260,
while -p = the concentrationthat
remains constant.In this limit, interactionsgo on forever,and only the statistical descriptionapplies.It has beenshownby extensivenumericalsimulationsthat even if westart with a trajectory involving an ever-increasingnumberofparticles,diffusive processestake over, and the trajectory\"collapses\"becauseit is transformedover time into a delo-calizedsingular distributionfunction.8
Our new kinetictheory is ofgreat interestin describingdissipative processesfor all time scales,as observed in the
laboratory or the ecosphere.But this is only oneof its
many novel features.BecauseofPoincareresonances,the
dynamical processesdescribedin this sectionlead to long-rangecorrelations,even if the forcesbetweenthe particles
Beyond Newton's Laws 127
are short range.Theonly exceptionis the state ofequilibrium, wherethe range ofcorrelationsis determinedby that
ofthe forcesbetweenthe particles.Thisexplainsthe fact, asstated in Chapter2,that nonequilibriumallows for a newcoherence,which is clearly manifested by chemicaloscillations and hydrodynamic flows. We now recognizethat
equilibriumphysicsgave us a false imageofmatter. Onceagain, we are facedwith the fact that matter in equilibriumis \"blind,\" while in nonequilibriumit beginsto \"see.\"
In sum, we are now able to gobeyond Newtonianmechanics. Thevalidity ofthe trajectory descriptionused in
classicalmechanicsis severely limited.Thermodynamicsis
incompatiblewith trajectory description,as it requiresastatisticalapproachboth at equilibriumand out ofequilibrium. The fact that the vast majority ofthe dynamicalsystems correspondingto the phenomenathat surround usare LPS is the reasonwhy thermodynamicsis universallyvalid. Transient dynamical interactionssuchas scattering arenot representativeof the situations that we encounterinthe natural world, where interactionsare persistent.Thecollisionprocessesthat appear in our statistical descriptionas a result ofPoincareresonancesare essentialin that theybreak time symmetry and lead to evolutionary patterns in
accordancewith the thermodynamicdescription.The microscopicdepictionof nature associatedwith
thermodynamics has little to do with the comfortable
time-symmetricaldescriptionscientistshave traditionallytaken from Newtonian principles.Ours is a fluctuating,noisy, chaoticworldmoreakin to what the Greekatomists
imagined.In Chapter1,we describedEpicurus'dilemma.Theclinamenhe envisagedno longerbelongsto a
philosophical dream that is foreign to physics.It is the veryexpression ofdynamical instability.
128The End of Certainty
Of course,dynamical instability provides only thoseconditionsnecessaryto generateevolutionary patterns ofnature.Oncewe have achievedour statistical description,we can also formulate the additional factors we needin
orderto observethe emergenceofcomplexity\342\200\224of dissi-
pative structuresat the macroscopiclevel.We now begintounderstand the dynamical roots oforganization,thedynamics at the rootofcomplexitythat are essentialfor self-
organization and the emergenceoflife.
Chapter6
A UNIFIED FORMULATIONOF QUANTUM THEORY
I
Thereare fundamental differencesbetweenclassical
Newtonian dynamics and quantum theory But in
both casesthereexistan individual descriptionin terms oftrajectoriesorwave functions (seeChapter1,SectionIV)
and a statistical descriptionin terms ofprobabilitydistributions. As we have alreadyseen,Poincareresonancesappearin classicalas wellas quantum theory. We can therefore
anticipate that the resultsobtained in classicalmechanicswill
also apply to quantum theory.In fact, in both instanceswehave achieved a new statisticalformulation applicableto LPSoutsideHilbertspace.This descriptionincludes
time-symmetry breaking,and is irreducibleto the individual
description in termsofquantum wave functions.In spiteofquantum theory'sastonishingsuccess,
discussions about its conceptualfoundations have not abated.After seventy years,they are as lively as ever.
129
130TheEnd of Certainty
Forexample,in his recentbookShadowsof the Mind,
RogerPenrosedistinguishesbetween\"Z mysteries\"(forquantum puzzles) and \"X mysteries\"(forthe quantumparadox) in quantum behavior.1Furthermore,the roleofnonlocalityseemsintenselyproblematic.Given that
localityis a property associatedwith the Newtonian pointwise
trajectory description,it is not surprising that quantumtheory, whichincludesthe wave aspectofmatter, leadstoa form ofnonlocality.2
The \"collapse\"of the wave function, whichseemstorequirea dualistic formulation ofquantum theory,represents a further complication.On the onehand, we havethe basicSchrodingerequationfor wave functions, whichis time reversibleand deterministic,exactly as is Newton'sequation;on the other,we have the measurementprocessassociatedwith irreversibility and the collapseofthe wavefunction.This dualistic structure is the basisofJohn vonNeumann'sargument in his famous book,Mathematical
Foundations of Quantum Mechanics.3This situation is indeedbizarrebecausein addition to the basictime-reversible,deterministic Schrodingerequation,therewouldbea seconddynamical law associatedwith the collapse(or reduction)of the wave function.Until now, however, no one hasbeenable to describethe link betweenthesetwo laws ofquantum theory, nor has anyone succeededin giving arealistic interpretation of the reductionof the wavefunction. This is the quantum paradox.
Thequantum paradox,whichderivesfrom the dualisticstructure ofquantum theory, is closelyrelated to anotherproblem.Ourconclusionis that quantum theory is
incomplete.Like classicaltrajectory theory, it is timesymmetric,
and thereforecannotdescribeirreversibleprocessessuchasthe approachto thermodynamicequilibrium.This is par-
A Unified Formulation of Quantum Theory 131
ticularly curiousbecausequantum theory beganin 1900with Max Planck'ssuccessfuldescriptionofblackbodyradiation in equilibriumwith matter. Even today, in spiteofthe great advances madeby Albert Einsteinand Paul A. M.Dirac,we still have no exactquantum theory describingthe approachto equilibriumwhenradiation interacts with
matter. (As we shallsee,this is relatedto the fact that
quantum theory describesintegrablesystems.We shall comeback to this challengein SectionIV.) We needbothequilibrium and nonequilibriumphysicsto describethe worldaround us.An exampleofan equilibriumsituation is thefamous residualblackbody radiation at
3\302\260K,which
originated at a time closeto the big bang.A largepart ofmacroscopicphysicsalso dealswith equilibriumsystems,whetherthey are solids,liquids,or gases.Thereis thus a
gap between quantum theory and thermodynamics as
deepas that betweenclassicaltheory and thermodynamics.Remarkably, the same methodemployed in extendingclassicalmechanicsin Chapter5 also permits us to unify
quantum theory and thermodynamics.Indeed,ourapproach eliminatesthe dualistic structure ofquantummechanics, and thus eliminates the quantum paradox.We
arrive at a realistic interpretation ofquantum theorybecause the transition from wave functions to ensemblescannow be understoodas the result ofPoincareresonanceswithout the mysteriousintervention ofan \"observer\"orthe introductionofotheruncontrollableassumptions.Incontrast to otherattempts to extendquantum theory, asnotedin Chapter1,ourown approachmakeswell-definedpredictionsthat are testable.Thus far, they have beenconfirmed by every numericalsimulation performed.4
Our thinking constitutesa return to realism,but
emphatically not a return to determinism.On the contrary,
132TheEnd of Certainty
we move even farther away from the deterministicvisionofclassicalphysics.We agreewith Popperwhenhewrites,\"My own point ofview is that indeterminismis
compatible with realism,and that the acceptanceofthis fact allowsus to adopt a coherentobjectiveepistemologyof thewholeofquantum theory, and an objectivistinterpretationofprobability.\" We shall thereforeendeavorto bring intothe realm ofphysicswhat Poppercalledhis metaphysicaldream:\"It is likely that the worldwould bejust as indeter-ministicas it is even if therewereno observingsubjectsto
experimentwith it, and to interfere with it.\"5 Thus we will
show that the quantum theory ofunstabledynamicalsystems with persistentinteractionsleads,as in classicalsystems, to a descriptionthat is both statisticaland realistic.Inthis new formulation, the basic quantity is no longerthewave function correspondingto a probability amplitude,
but probability itself. As in classicalphysics, probabilityemergesfrom quantum mechanicsas a fundamental
concept.In this sense,we are on the eve ofthe triumph ofthe
\"probabilisticrevolution,\" which has beengoingon forcenturies.Probability is no longera state ofmind due toour ignorance,but the result ofthe laws ofnature.
IITheobservation that the interactionbetweenatoms and
light leads to well-definedabsorption and emissionfrequencies was the starting point for the formulation ofquantum mechanics.Theatom was describedby NielsH.D.Bohrin terms ofdiscreteenergy levels.In accordancewith experimentaldata (the Ritz-Rydbergprinciple),the
frequencyofspectrallinesis the difference between two energy
levels.Oncetheselevels are known,we can predictthe fre-
A Unified Formulation of Quantum Theory 133
quencyofspectrallines.Theproblemsofspectroscopycanbereducedto the calculationoflevels ofenergy.But howcan we reconcilethe existenceofwell-definedenergylevels, which decisivelyinfluencedthe history of quantumtheory, with the Hamiltonianconceptthat is so importantto classicaltheory?TheclassicalHamiltonianexpressesthe
energy ofa dynamical system in terms ofcoordinatesqand momentap, and thereforetakeson a continuousset ofvalues.It cannotlead to discreteenergylevels.Forthis
reason, the HamiltonianHis replacedin quantum theory bythe HamiltonianoperatorH .r
op
We have repeatedlyusedoperatorformalism (the Perron-Frobeniusoperatorwas introducedin Chapter4, and theLiouvilleoperatorin Chapter5), but it was in quantumtheory that operatorcalculus was first introducedinto
physics.In the situations studied in Chapters 4 and 5,weneededoperatorsto achieve the statistical description.Here,even the individual level ofdescriptioncorresponding
to wave functions requiresoperatorformalism.Thebasicproblemin quantum mechanicsis the
determination ofthe eigenfunctionsua and the eigenvaluesEaofthe HamiltonoperatorH(weshall omit the subscriptopwhereverpossible).TheeigenvaluesEa, whichareidentified with the observedvalues ofthe energy levels, form the
spectrumofH.We speakofa discretespectrum whensuccessive eigenvaluesare separatedby finite distances.If the
spacingbetweenlevelstends toward zero,we then speakofa continuous spectrum. Fora free particlein aone-dimensional boxwith a lengthofL, the spacingofthe energylevel is inversely proportionalto ]_}_ As a consequence,whenL \342\200\224>
\302\260\302\260,this spacingmoves toward zero,and we
obtain a continuousspectrum.By definition, the word\"large\" in largePoincaresystems(LPS)means precisely
134 The End of Certainty
that thesesystemshave a continuousspectrum.As in
classical theory, the Hamiltonianis herea function ofcoordinates and momenta.However,becausethe Hamiltonianis
now an operator,thesequantities, and thereforeall
dynamical variables, now have to be treatedalso as operators.For today'sphysicists,the transition from functions to
operatorsthat takesplacein quantum theory seemsperfectly natural. They now manipulate operatorswith theeasewith whichmost ofus manipulate natural numbers.Nonetheless,for classicalphysicistssuchas the great DutchscientistHendrikAntoon Lorentz,the introductionofoperators was barely acceptable,and even repulsive.In anycase,individuals such as Werner Heisenberg,Max Born,Pascual Jordan,Erwin Schrodinger,and Paul Dirac,who
daringly introducedoperatorformalism into physics,deserve our admiration.They drastically changedourdescription ofnature in defining the conceptualdifferencebetweena physicalquantity (representedby an operator)and the numericalvalues this physicalquantity may take on(the eigenvaluesofthe correspondingoperator).Thisradical changein outlookhas had far-reachingand profoundimplicationsfor ourconceptionofreality.
As an exampleofthe sophisticationofoperatorformalism,
considerthe commutationrelationsbetweentwooperators. Theseoperatorscommuteif the orderof their
applicationto a function is immaterial.They do notcommute if the orderoftheir applicationchangesthe result.Forinstance,multiplying a function j{x)by x and then
differentiating it with respectto x doesnot leadto the sameresult as first differentiating^^)and then multiplying it byx. This can easilybeverified.Operatorsthat do notcommute exhibit different eigenfunctions;if they docommute, they have commoneigenfunctions.
A Unified Formulation of Quantum Theory 135
Thefamous Heisenberguncertainty principlefollows fromthe fact that the coordinateand momentumoperators,asdefined in quantum theory, do not commute.In all
textbooks on quantum mechanics,it is shown that in the\"coordinate representation,\"the operatorcorrespondingto acoordinateq has eigenvaluesthat are the coordinatesofthe quantum object.The operatorqo may thereforebeidentifiedwith the classicalcoordinateq. In contrast, themomentumoperatorp is definedby the derivativeoperator 7 5:whichis a derivative in respectto q.Thetwooperators q and p thus do not commute,and have no
1op rop
'commoneigenfunctions.6In quantum mechanics,we mayuse various representations.In addition to the coordinaterepresentation,we have the momentumrepresentation,where the momentumoperatoris simplyp, and coordi-nates are representedby derivative operators7 5:.Whatever the representation,the two operatorsdo notcommute.
Thefact that q and p do not commutemeansthat we^op rop
cannotdefine states ofa quantum objectfor whichboththe coordinateand the momentumtake on well-definedvalues.This is the rootofHeisenberg'suncertaintyreaction, which forcesus to abandon the \"naive realism\" ofclassicalphysics.We are ableto measurethe momentumorthe coordinateofa given particle,but we cannot say that
this particle has well-definedvalues for both its
momentum and its coordinates.This conclusionwas reachedsixty
yearsagoby HeisenbergandBorn,amongothers.Even so,discussionsabout the meaningofuncertainty relationsstill
go on, and somescientistshave not yet given up the hopeofrestoringthe traditional deterministicrealismofclassical mechanics.7This was oneofthe reasonsfor Einstein'sdissatisfaction with quantum theory.We should note that
136The End of Certainty
Heisenberg'suncertainty principleis compatiblewith a
deterministictime-symmetricaldescriptionofnature (theSchrodingerequation).
What do we meanwhenwe say that a quantum systemis in a particular \"state\"? In classicalmechanics,the state is
a point in phasespace.Hereit is describedby a wavefunction whose evolution over time is expressedby the
Schrodingerequationih/2n9*F(f)/9f =Ho Wit).
This equationidentifiesthe time derivative ofthe wave
function *P with the actionof the Hamiltonian operatoron *F. It is not derived,but rather is assumedat the start,and can thus bevalidated only by experiment.It is thefundamental law ofnature in quantum theory*Notetheformal analogy with the Liouville equationin Chapter5,SectionIII,wherethe basicdifference is that L (theLiouville operator)acts on distributionfunctions p, while Hacts on wave functions.
We have already mentionedthat a wave function
corresponds to a probabilityamplitude.Theparallel that guidedErwin Schrodingerin formulating his equationwas that ofclassicaloptics.In contrast to the trajectory equations ofclassicalmechanics,the Schrodingerequationis a wave
equation.It is a partial differential equationbecauseinaddition to the time derivative, thereare alsoderivatives with
respectto coordinatesappearingin H (rememberingthat
in the coordinaterepresentation,the momentumoperatoris a derivative with respectto coordinates).But classicaland
quantum equationshave an essentialelementin common:They both correspondto a deterministicdescription.
There are various extensions of the Schrodinger equation and the relativistic
Dirac equation, but they are not necessary to this discussion.
A Unified Formulation of Quantum Theory 137
Once*P is known at somearbitrary time tQ, togetherwith
appropriateboundary conditions(suchas *P \342\200\224> 0 at
infinite distances),we may calculate *F for arbitrary times in
the future aswellas in the past.In this sense,we reinstitute
the deterministicview ofclassicalmechanics,but it now
appliesto wave functions, and not trajectories.As in the classicalequationsofmotion,the Schrodinger
equation is time reversible.When we replacet by \342\200\224t,
the
equationremainsvalid. We only have to replace*P with its
complexconjugate*F*. As a consequence,if we observethe transition of*P from *Ft to
XV2at time t2 wheret2
> tpwe can also observe a transition from V|/*2 to \\)/*1 It is
worth remindingourselvesofArthur Stanley Eddington'sremark at an early stagein quantum mechanicsto the
effect that quantum probabilitiesare \"obtainedby
introducingtwo symmetricalsystemsofwaves traveling in opposite
directionsof time.\"8 Indeed,as we have seen, the
Schrodingerequation is a wave equation describingthe
evolution ofprobabilityamplitudes.Ifwe now take the
complexconjugateofthe Schrodingerequation,that is, if
we replace/ by \342\200\224 i, *F by *P* (supposingthat Hg is real),and t by \342\200\224t,
we return to Schrodinger'sequation.As stated
by Eddington,*P* may thereforebe viewed as a wave
function propagatinginto the past.Furthermore,as
mentioned in Chapter 1, probability proper is obtained
through multiplying *P by its complexconjugate*P* (that
is, l^l2).Since*P* may be interpretedas *P evolvingbackwardin time,the definition ofprobabilityimpliesthe
meetingoftwo times,onestemmingfrom the past and theotherthe future. In quantum theory, probabilitiesare thus
time symmetric.We now seethat in spite of their fundamental
differences, both classicaland quantum mechanicscorrespond
138The End of Certainty
to laws ofnature that are deterministicand time reversible.No difference betweenpast and future appears in theseformulations.As we noted in Chapters 1and 2,this leadsto the time paradox.In quantum mechanics,it alsoleadstothe quantum paradox,due to the needto introducea du-alistic formulation of quantum theory. In both classical
theory and quantum theory, the Hamiltonianplays acentral role.In quantum theory, its eigenvaluesdeterminethe
energy levels, while,accordingto the Schrodingerequation, it also determinesthe time evolution of the wavefunction.
As in the precedingchapter, we shall concentrateonsystemsin whichthe HamiltonianH is the sum ofa freeHamiltonian H\342\200\236 and a termproducedby the interactionsA,V, wherebyH= HQ+ A,V. Thetime history ofsuchsystems can then bedescribedby transitions betweeneigen-states ofH0 inducedby theseinteractions.
As longas we remain in Hilbertspace,the eigenvaluesEa ofHare real (likethe Liouville operator,His also \"her-mitian,\" and hermitian operatorshave real eigenvaluesinHilbertspace).Theevolution ofthe wave function is a
superposition ofoscillatingterms such asexp(\342\200\224iE t). There
are, however, irreversibleprocessesin quantum mechanics,such as the quantum leapsin Bohr'stheory, whereexcitedatoms decaythrough the emissionofphotonsor unstable
particles(seeFigure6.1)or the decayofunstable particles.Cantheseprocessesbeincludedin Hilbertspacewithin
the framework of traditional quantum theory? Decayprocessesoccurin largesystems.Ifan excitedatom werekept in a cavity, the emittedelectronwould bounceback,and there would be no irreversibleprocess.As we have
seen,the time evolution ofthe wave function is described
A Unified Formulation of Quantum Theory 139
Figure6.1Decayofan ExcitedAtomThe atom \"falls\" from the excitedstate to the ground state with
the emissionof a photon.
r- Excited state
*-L Photon
* Ground state
by a superposition,or sum, ofoscillatoryterms.With thelimit oflargesystems,this sum becomesan integral,and
acquiresnewproperties.In the caseofthe decayofexcitedatoms as describedby Figure6.1,the probabilities l^l2decayalmost exponentiallyover time.Herethe word almost
is essential:As longaswe remain in Hilbertspace,therearedeviations from the exponentialfor both very brieftimes
(the sameorderas the frequencyofoscillationsoftheelectron around the nucleus~ 10~16seconds)and very longtimes (for example,ten to onehundred times the lifetimeofan excitedstate, whichis ~ 10-9).However,in spiteofa
great numberofexperimentalstudies,no deviations from
exponentialbehavior have yet beendetected.This is
indeed fortunate, becauseif they did exist,it would raiseserious questions about the entire theoreticalsystem ofparticlephysics.
Supposethat we preparea beamofunstable particles,letit decay, and later on preparea secondbeam.Imaginethe
strange situation of the two beams prepared at differenttimes having different decaylaws.We couldthen distin-
140 The End of Certainty
guish betweenthem just as we do betweenolderand
younger individuals!This fantasy would be a violation ofthe principleof indistinguishability for elementaryparticles, which has led to someof the greatest successesofquantum theory* Thepreciseexponentialbehaviorobserved thus far showsthe inadequacyofHilbertspacedescription. We shall comeback to decay processesin thenext section,but at this point we should note that such
processesoughtnot to beconfusedwith processesdrivingthe system to equilibrium.The decay processasrepresented in Fig.6.1only transfers the energyofthe atom tothe photons.
IllAs we have seen,the main issuein quantum mechanicsisthe solutionofthe eigenvalue for the Hamiltonian.Thereare only a few quantum systemsin whichthis problemhasbeensolvedexactly.In orderto do so,we generallyneedtousea perturbationalapproach.As mentioned,we start with
a Hamiltonianin the form H= H0+ Xl^whereHQcorresponds to a Hamiltonian operatorfor which we havesolvedthe eigenvalue (the \"free\" Hamiltonian)and Kis a
perturbation coupledwith HQ,through the so-calledcoupling constantX. We assumethat we know the solutionofthe eigenvalue Hnu '0)= E ^u '0),and that we wish to solve0 On n n 'the equationHu =E u .The standard procedure,whichis1 n n n r
Schrodinger'sperturbational method,is to expandboth
*These include the explanation of superfluidity and the quantum theory ofsolid state.
A Unified Formulation of Quantum Theory 141
the eigenvaluesand eigenfunctionsin terms ofpowersofthe couplingconstant X.
Theperturbationalapproachleadsto a recurrenceprocedure involving equationsfor eachorderin X. Thesolution oftheseequations implies the use of terms such as
1/(23(\302\260) \342\200\224 E(\302\260)),
whichbecomeill defined when the de-nominatorvanishes.This situation again correspondstoresonances,*and oncemorewe encounterthe divergenceproblemthat liesat the very centerofPoincares definitionofnonintegrablesystems.
However,thereis an essentialdifference here.We have
already introducedthe distinction betweendiscreteandcontinuousspectrums.In quantum mechanics,this
difference becomescrucial.In fact, when the spectrumis
discrete, it is generally possible to avoid the divergenceproblemthrough an appropriate choiceof theunperturbed Hamiltonian.^ Since all finite quantum systemshave a discretespectrum,we can then concludethat theyare integrable.
The situation changesdramatically when we turn to
largequantum systemsinvolving excitedatoms,scatteringsystems,and so on.In this case,the spectrumis
continuous,whichbrings us back to LPS.Theexampleofa
particle coupledwith a field, whichwe presentedin Chapter5,SectionV, also appliesto quantum systems.We then haveresonanceswheneverthe frequency00?associatedwith the
particleis equalto a frequencyC0fcassociatedwith the field.
*In quantum mechanics, to each energy E corresponds a frequency CO expressedas E = (/)/2lt)C0.
fin more technical terms, we first raise the degeneracy by an appropriatetransformation.
142 The End of Certainty
Theonly difference is that in quantum theory, frequenciesare associatedwith energies.The eigenvalueE corre-sponds to the frequency 2^00a,where h is Planck'sconstant.
Theexamplein Figure6.1,which correspondsto an
LPS,illustrates that we have resonanceeachtime the
energy difference betweenthe two levels is equal to the
energy ofthe photonthat is emitted.As in the caseofdeterministicchaosstudiedin Chapter
4, we can extendthe eigenvalue problemto singularfunctions outside Hilbertspace.The formal solution ofthe Schrodingerequationis ^(t) = U(r)*F(0),whereU(t) =e~'Ht; U(t) is the evolution operatorthat links the value ofthe wave function at time t to that at the initial time t =0.Both future and pastplay the samerole,sinceU(t:)U(t2)=
U(t1 + i2), whatever the sign of tt and tr This propertydefineswhat is calleda dynamical group.OutsideHilbertspace,the dynamical group splits into two semigroups.Thereare then two functions correspondingto the excitedatom:The first, <pv decays exponentially in the future
(<p^~e \"(/r), while the second,q>v decays in the past
(<pt~et/r)- Only oneofthesetwo semigroupsis realizedinnature.Inboth cases,thereis an exactexponentialdecay(incontrast to the approximateonedescribedin the precedingsection).This was the first such examplestudied, notablyby Arno Bohmand GeorgeSudarshan,who showed that
in orderto obtainexactexponentiallaws and avoid thedifficulties mentionedin SectionII,Hilbertspacemust beabandoned.9However,in theirapproach,the central
quantityremainsthe probabilityamplitude,and the basic
paradox of quantum mechanics(the collapseof the wave
function) is not solved.As already mentioned,the decayofexcitedatoms or unstable particlescorrespondsonly to a
A Unified Formulation of Quantum Theory 143
transfer ofenergy from onesystem(the excitedatom) tothe other (the photon).The approach to equilibriumrequires a fundamental modificationofquantum theory.As
in classicalmechanics,we have to go from the individual
description,associatedwith wave functions, to thestatistical description,associatedwith ensembles.
IVIn the transition from the individual to the statistical
description, quantum theory introducescertainspecificfeatures as comparedto classicalmechanics.There,as we sawin Chapter5, the statisticaldistributionfunction is a
function ofboth the coordinatesand momenta.A trajectorycorrespondsto the delta function (seeChapter1,SectionIII).In quantum mechanics,the quantum state, asassociated with a wave function, is describedby a continuousfunction ofthe independentvariables.We can eithertakethe coordinatesas independentvariables and consider^(q), or we can take the momentaand consider (p).Heisenberg'suncertainty principleprevents us from takingboth.The definition ofa quantum state thereforeinvolves
only half ofthe variables that are used in the definition ofthe classicalstate.
Thequantum state *P representsa probabilityamplitude
for whichthe correspondingprobabilityp is given by the
productofthe amplitudes*\302\245(q)and
x\302\245*(q'),and is therefore
a function oftwo setsofvariables, q and q'orp and p\\ We
can thus write p(q,q')orp(p,p'),wherethe first expressioncorrespondsto the coordinaterepresentation,and thesecond to the momentumrepresentation,whichwill beespecially
useful to us.In quantum mechanics,the probabilitypis often calledthe \"density matrix\" (matrices,as studied
144 The End of Certainty
in algebra,alsohave two indices).We can easily write the
equationofevolution for p becausethe equationfor *P
(the Schrodingerequation)is already known.Theevolution equationfor p is the quantum Liouville equation,whoseexplicitform is ih
(-\302\243)
= Hp \342\200\224 pH,which is the\"commutator\" ofp with H.This showsthat whenp is afunction ofH,we have an equilibriumsituation. Thendpldt = 0,as Hcommuteswith a function ofitself.
Now that we have consideredthe distributionfunction
p, whichcorrespondsto a singlewave function, we canalsoconsidersituations in whichp correspondsto a\"mixture\" ofvarious wave functions.In both cases,theLiouville equationremainsthe same.
For integrablesystems, the statistical formulationintroduces no new features. Supposethat we know the
eigenfunctions<pa(p) and the eigenvaluesEa ofH.TheeigenfunctionsofL are then the productsq>a(p)q>a(p')andthe eigenvaluesthe differencesEa
\342\200\224
E\342\200\236. Theproblemsinvolved in deriving the spectralrepresentationsofHand Lare thus equivalent.
TheeigenvaluesEa\342\200\224
E\342\200\236 ofL corresponddirectlyto the
frequenciesmeasuredin spectroscopy,wherethe timeevolution ofthe distribution function p is a superpositionofoscillatingterms e~'^Ea
~\302\243/3)(. Again, thereis no approachto
equilibrium.Moreover,for thosesituations in whichwecan derive the eigenvaluefor the Hamiltonian,eigenfunctions ofL, such as <Pa(p)<Pa(p),correspondto zeroeigenvalues of the Liouville operator,Ea \342\200\224 Ea = 0, and are
thereforeinvariants ofmotion.As a result the system is
integrable(as is a systemofnoninteractingparticles),andcannotreachequilibrium.This is a form ofthe quantum
paradox.We can now seeclearly why it is not sufficient to extend
wave functions beyond Hilbertspace.Indeed,as indicated
A Unified Formulation of Quantum Theory 145
in SectionIII,this leads to complexenergiesin the form
Ea = C0a\342\200\224 i y whereCOaisthe realpast and Ya the life span,
whichdescribethe decayofexcitedatoms orunstable
particles, but this still does not account for irreversible
processesassociatedwith the approach to equilibrium.Inspite ofthe complexelementin E , all diagonalelementsofp,whichare productssuchas q>a(p)(Pa(p'),wouldbeinvariants becausethe eigenvalue Ea \342\200\224 Ea again vanishes, andthe systemremainsintegrableand cannotapproachequilibrium.*
TheexperimentalbasisofBohr'stheory ofatoms andthe subsequentemergenceofquantum theory is basedonthe Ritz-Rydbergprinciple,accordingto whicheachfrequency v, as measuredin spectroscopy,is the differencebetween the two numbers Ea and
E\342\200\236,
whichrepresenttwo
quantum levels.This,however, can no longerbe true for
systemspresentingirreversibleprocessesthat lead thesystem to equilibrium.Quantum theory must thereforebefundamentally revised.
Historically,the roots ofmechanicsliein two branchesofphysics:the thermal equilibriumbetweenmatter andradiation that ledPlanckto introducehis famous constanth in 1900,and spectroscopy,which led from the Ritz-Rydbergprincipleto Bohr'satom, and finally, with Heisen-berg(1926),to quantum theory.However,the relationshipbetweenthese two domains has never beenelucidated.We seethat the Ritz-Rydbergprincipleis incompatiblewith the thermal approach to equilibriumdescribedbyPlanck'swork.Thus we needa new formulation makingthermal physicsand spectroscopycompatible.This can be
*Difficulties arise when Ea\342\200\224
E\342\200\236is replaced by Ea
\342\200\224
E*\342\200\236,where
E*\342\200\236is the
complex conjugate of E\342\200\236. Here, Ea-
E*a=-2fya^ 0, with no equilibrium state.
146 The End of Certainty
achieved at the level of probability distributions from
which we may derive observablefrequencies(includingtheir complexpart), but thesefrequenciesare no longerdifferencesin energy levels for the systemswe expecttoapproachequilibrium.We have to solve the quantum Liou-ville eigenvalue problemfor LPS in the contextofmoregeneralfunction spaces.As in classicalmechanics,this will
involve two basicingredients:delocalizeddistribution
functions, which lead to singularities, and Poincareresonances,which lead to new dynamical processes.As in classicaldynamics, there then appear new solutionsat the statisticallevel that cannot bereducedto the traditional wavefunction formalism ofquantum mechanics,and no longersatisfy
the Ritz-Rydbergprinciple.In this sense,we can truly
speakofa new formulation ofquantum theory.
V
With certainmodifications,we canfollow theprobabilistic formulation for classicalsystemsgiven in Chapter5.Theformal solutionofthe Liouvilleequationis i(dp/dt)=Lp,where in quantum theory Lp is the commutatorofthe Hamiltonianwith p (aswe have seen,Lp =Hp \342\200\224 pH).It can be written as eitherp(t) = e~iH'p(0)eHHtor p(t) =e~'Ltp(0).What is the differencebetweentheseequations?In the first formulation, it appearsthat we would have two
independent dynamic evolutions:oneassociatedwith e~tHt
and the otherwith e+,Ht, onemoving toward the \"future\"
and the other toward the \"past\" (as t is replacedby \342\200\224 t). If
this wereso,we couldexpectno time-symmetrybreaking,
and the statistical descriptionwould conservethe
timesymmetry of the Schrodingerequation.But this is
no longerthe casewhenwe includePoincareresonances,
A Unified Formulation of Quantum Theory 147
which couplethe two timeevolutions (e~'Ht and e+,Hf).
Thereis now only oneindependenttimeevolution (timehas \"onedimension\.")In orderto study time-symmetrybreaking,we have to begin with the expressionp(t) =e~\"Lp(0),whichdescribesa singletimesequencein the Li-ouville space.In otherwords,we have to orderdynamicalevents accordingto a singletime sequence.*We can thendescribeinteractions,as we did for classicalmechanics,asa successionofevents separatedby free motion.Inclassical mechanics,theseevents changethe values ofthe wavevectork and the momentap.In Chapter5,we introducedvarious events leadingto the creationand destructionofcorrelations,wherewe sawthat the decisivefactor was the
appearance,for LPS,ofnewevents (the bubblesin Figure5.7) that couplecreationand destruction.As such, they
radically changeclassicaldynamicsbecausethey introducediffusion, break determinism,and destroytimesymmetry.We can alsoidentify the sameevents in quantummechanics. To do so,we needto introducevariables that play thesameroleas the wave vectork in classicaltheory'sFourierrepresentation.In classicalmechanics,we start with a
statistical formulation in which the distribution functions
p(q,p) are expressedas functions ofthe coordinatesq andthe momentap. We then proceedto the Fouriertransformation pk{p)involving the wave vector k and themomenta.
In quantum mechanics,we can follow a similar
procedure.10 We start with the densitymatrix p(p,p')in the mo-
*If this is not done, we have to be very careful. Feynman's well-known
statement that an electron propagates toward the future and a position moves toward
the past refers to time as it appears in the Schrodinger equation before ordering
dynamical events according to a single time sequence.
148 The End of Certainty
mentumrepresentation,whichis a function oftwo setsofvariables, p and p'.We then introducenew variables, k \342\200\224
p \342\200\224 p'and P= (p + p*)/2;we can now write, as in classicalmechanics,pk(P)-It can then be shown that k plays thesamerolein quantum mechanicsas the wave vectordoesinclassicalmechanics.(Forexample,in interactions,the sumofthe wave vectorsis conserved,that is, k.+ k = k'.+k' .)' ' j n j n I
Again as in classicalmechanics,Poincareresonancesintroduce new dynamical events that couplethe creationanddestructionofcorrelations,and thereforedescribequantum diffusive processes.
The formulation of classicaland quantum theory forLPS is moreor lessparallel.A minordifference appearsin
the roleofthe momentumP. Foreachevent, as introducedin Chapter5, the momentaofthe interactingparticlesarealtered.In quantum mechanics,we use the two variables k
and P, wherethe variable P replacesthe classicalmomentum. As thesevariables interact, the modificationofPinvolves Planck'sconstant h. Forh \342\200\224> 0,however, we comeback to the classicalmomentump. But this difference hasno important effect on formal development,and we shall
not attempt to describeit in further detail.In the previouschapter,we introduceda fundamental
difference betweentransitory and persistent interactions.Persistent interactions are especiallysignificant becausethey appearin all situations wherethermodynamicscan beapplied.As in classicalmechanics,the distributionfunction
p correspondingto persistentinteractionsis describedby
singular functions ofthe variable k. In classicaldynamics,as wellas classicaland quantum mechanics,persistentscattering is typical of the situations describedby statistical
mechanicsand cosmology.For example,in the
atmosphere, particles collidecontinuously,are scattered,and
A Unified Formulation of Quantum Theory 149
then recollide.Persistentscatteringis describedby delocal-izeddistributionfunctions, whichare singular functions in
the wave vectorspace.As we saw in Chapter5, the latterforceus to go outsideHilbertspace.
By taking into accountdelocalizedsingular distributionfunctions and Poincareresonances,we obtain, as in
classical mechanics,complex,irreduciblespectralrepresentations for the Liouville operatorL. Again, as in classical
dynamics,irreversibility is associatedwith the appearanceofhigher- and higher-ordercorrelations.As in classicalmechanics,this leadsto new features in kinetictheory and
macroscopicphysics.Thebasicconclusionsofourformulation ofquantum mechanicsare as follows:
\342\200\242 Theeigenvaluesofthe Liouville operatorare no longerdifferencesbetween the eigenvaluesof the Hamilton-ian, which are obtained from the Schrodingerequation. Therefore,the Ritz-Rydbergprincipleis violated,whereby the systemsare no longerintegrableand the
approach to equilibriumis possible.\342\200\242 The quantum superpositionprincipleassociatedwith
the linearity ofthe Schrodingerequationis violated.\342\200\242 The eigenfunctionsof the Liouville operatorare not
expressedin terms of probability amplitudes or wave
functions, but rather in terms ofprobabilitiesproper.
Ourpredictionshave already beenverified in simplesituations wherewe can follow the collapseofwave functionsoutsideHilbertspace.11Moreover,they have ledto
interesting predictionsof the form ofspectrallines,and haveallowed us to accurately describethe approach toequilibrium. We regretthat we cannotgo into greaterdetailabout their specificapplications,but our objectivein this
150 The End of Certainty
bookis merely to provide a brieftour ofthe theoreticalbackground.
VIAt the Fifth Solvay Conferenceon Physicsthat tookplacein Brusselsin 1927,therewas an historicdebatebetweenEinsteinand Bohr.In the wordsofBohr:
To introduce the discussionon such points, I was askedat the
conferenceto give a report on the epistemologicalproblemsconfronting us in quantum physics and took the opportunity
to center upon the question of an appropriate terminologyand to stressthe viewpointof complementarity.The main
argument was that unambiguous communication of physicalevidencedemandsthat the experimentalarrangement as well
as the recordingof the observationsbe expressedin common
language, suitably refined by the vocabulary of classical
physics.12
But how can we describean apparatus in classicaltermsin a world dominatedby quantum laws? This is the weak
point in the so-calledCopenhageninterpretation.Nevertheless, there is an important elementof truth containedtherein.Measurementis a meansofcommunication.It is
becausewe are both \"actorsand spectators,\"to useBohr'swords, that we can learn somethingabout nature. Butcommunicationrequiresa commontime.Theexistenceofthis commontime is oneofthe basicconsequencesofourapproach.
Theapparatus that performsthe measurements,whethera physical constructor our own sensoryperception,must
follow the extendedlaws of dynamics, includingtime-
A Unified Formulation of Quantum Theory 151
symmetry breaking.Theredo exist integrabletime-reversible systems,but we cannot observethem in isolation.As emphasizedby Bohr,we needan apparatus that breakstime symmetry.LPSblur this distinctionin that they alreadybreak time symmetry and therefore, in a sense,measurethemselves.We do not have to describean apparatus in
classical terms.Commontime emergesat the quantum levelfor LPSassociatedwith thermodynamicsystems.
Thesubjectiveaspectofquantum theory, whichattributed an unreasonableroleto the observer,deeplytroubledEinstein.To our way of thinking, through his
measurements the observerno longerplays someextravagant rolein the evolution of nature\342\200\224at least no moreso than in
classicalphysics.We all transform information receivedfrom the outsideworld into actionson a human scale,but
we are far from beingthe demiurge,aspostulatedbyquantum physics,who would be responsiblefor the transition
from nature'spotentiality to actuality.In this sense,our approachrestoressanity. It eliminates
the anthropocentricfeatures implicitin the traditional
formulation of quantum theory. Perhaps this would havemade quantum theory moreacceptableto Einstein.
Chapter7
OUR DIALOGUEWITH NATURE
I
Scienceis a dialoguebetweenmankindand nature, the
resultsofwhichhave beenunpredictable.At the
beginning of the twentieth century, who would havedreamedofunstable particles,an expandinguniverse, self-
organization, and dissipative structures?But what makesthis dialoguepossible?A time-reversibleworld would alsobean unknowableworld.Knowledgepresupposesthat theworld affectsus and our instruments,that thereis an
interaction betweenthe knowerand the known, and that this
interactioncreatesa difference betweenpast and future.
Becomingis the sine qua non ofscience,and indeed,ofknowledgeitself.
The attempt to understandnature remainsoneof the
153
154 The End of Certainty
basicobjectivesofWestern thought.It should not,however, be identified with the idea ofcontrol.The masterwho believeshe understandshis slavesbecausethey obeyhis orderswould beblind.When we turn to physics,ourexpectationsare obviously very different, but hereas well,Vladimir Nabokov'sconvictionrings true:\"What can becontrolledis never completelyreal; what is real can neverbecompletelycontrolled.\"1Theclassicalidealofscience,aworldwithout time,memory,and history, recallsthetotalitarian nightmares describedby Aldous Huxley, MilanKundera,and GeorgeOrwell.
In our recentbook,Entre le Temps et I'Eternite, IsabelleStengersand I wrote:
Perhapswe need to start by emphasizing the almost
inconceivable characterof dynamic reversibility. The question oftime\342\200\224of what its flow preserves,createsand destroys\342\200\224has
always beenat the center of human concerns.Muchspeculation has calledthe ideaof novelty into questionand affirmed
the inexorablelinkage between causeand effect.Many forms
of mystical teachinghave deniedthe reality of this changingand uncertain world, and definedan idealexistence
permitting escapefrom life'safflictions. We know how important
the ideaof cyclicaltime was in antiquity. But, like the rhythm
of the seasonsor the generationsof man, this eternal return
to the point of origin is itselfmarked by the arrow of time.No speculation,no teachinghas everaffirmed an equivalencebetweenwhat is doneand what is undone:between a plant
that sprouts, flowers and dies,and a plant that resuscitates,
grows younger and returns to its original seed;between a
man who grows olderand learns, and one who becomesa
child, then an embryo,then a cell.2
Our Dialoguewith Nature 155
In Chapter1,we alludedto Epicurus'dilemmaand theatomisticapproachofthe ancients.Today,the situation has
changedsignificandy in the sensethat the morewe knowabout our universe, the moredifficult it becomestobelieve in determinism.We live in an evolutionary universewhoseroots,whichliein the fundamental laws ofphysics,we are now able to identify through the conceptofinstability
associatedwith deterministicchaosand nonintegra-bility. Chance,or probability, is no longera convenientway ofacceptingignorance,but rather part ofa new,extended rationality. As we have seen,for thesesystems,the
equivalenceis brokenbetweenthe individual description(trajectoriesand wave functions) and the statistical
description (in termsofensembles).At the statisticallevel, we can
incorporateinstability. The laws of nature, which nolongerdeal with certitudesbut possibilities,overrulethe
age-olddichotomy betweenbeingand becoming.Theydescribea world ofirregular, chaoticmotionsmoreakin
to the imageofthe ancientatomists than to the world ofregular Newtonianorbits.This disorderconstitutes the
very foundation ofthe macroscopicsystemsto whichwe
apply an evolutionary descriptionassociatedwith the
second law, the law ofincreasingentropy.We have considereddeterministicchaos,and we have
discussedthe roleofPoincareresonancesin both classicaland quantum mechanics.We have seenthat we needtwoconditionsto obtain our statistical formulations, whichgobeyond the usual onesfor classicaland quantummechanics: the existenceofPoincareresonances,which lead tonew diffusion-type processesthat can beincorporatedintothe statistical description,and extendedpersistentinteractions describedby delocalizeddistribution functions.
156 The End of Certainty
Theseconditionslead to a moregeneraldefinition ofchaos.As in the caseofdeterministicchaos,we thenobtain newsolutionsfor the statistical equations that cannotbeexpressedin terms oftrajectoriesorwave functions.Iftheseconditionsare not satisfied, we return to the usualformulations.This is the casein many simple examples,suchas two-body motion(for instance,the sun and earth),and typical scatteringexperiments,wherebeforeand after
scattering,the particlesare free.Theseexamples,however,
correspondto idealizations.Thesun and earth are part ofthe many-body planetary system;scattered particleswill
eventually meetother particles,and are thereforeneverfree.
It is only by isolatinga certainnumberofparticlesand
studying their dynamicsthat we obtain the traditionalformulations. Conversely,time-symmetrybreakingis a globalproperty encompassingHamiltoniandynamical systemsasa whole.In the chaoticmapsstudied in Chapters3 and 4,irreversibility occurseven in systemswith few degreesoffreedomdue to the simplificationsused to describethe
equationsofmotion.A remarkablefeature ofour approach is its application
to both classicaland quantum systems.All othertheoretical proposalsthat we are aware ofattempt to eliminatethe
quantum paradoxthrough an exclusivelyquantummechanism. On the contrary, in our view, the quantum paradoxis only oneaspectofthe time paradox.In the Copenhageninterpretation,the needto introducetwo different typesoftime evolution is engenderedby the measurementprocess.Accordingto Bohrhimself,\"Every atomicphenomenonisclosedin the sensethat its observationis basedona
recordingobtainedby means of suitable amplification devices
with irreversiblefunctions,such as permanentmarkson a
Our Dialoguewith Nature 157
photographicplate.\"3It was this measurementproblemthat ledto the needfor a collapseof the wave function,and forcedus to introducea secondtype of dynamicalevolution into quantum mechanics.It is thereforenot
surprising that the time paradoxand quantum paradoxare so
closelylinked.In solvingthe former,we alsosolve thelatter. As we have seenfor LPS,quantum dynamicscan onlybe describedat the statistical level.Moreover,to learnsomethingabout quantum processes,we again needanLPS actingas an apparatus.It is thus the secondlaw ofquantum time evolution,which includesirreversibility,that becomesthe generalone.
As statedby Alastair Rae,\"A pure quantum process(described by the Schrodingerequation) occursonly in oneor moreparametersthat have becomedetachedfrom therest ofthe universe, and perhapseven from space-timeitself, and leave no traceoftheir behavioron the rest oftheuniverse until a measurementinteractiontakes place.\"4Whatever the process,at somepoint irreversibility has tocomeinto the picture.An almost identicalstatement couldhave beenmade regardingclassicalmechanics!
It has often beensaid that in orderto make progressin
these difficult areas, we needthe inspiration of a truly
crazy idea.Heisenbergwas fond ofaskingwhat thedifference is betweenan abstract painter and a goodtheoreticalphysicist.Inhis opinion,an abstract painterneedsto bejustas originalas a goodtheoreticalphysicistneedsto beconservative.5 We have tried to follow Heisenberg'sadvice.Our lineofreasoningin this bookis certainly lessradicalthan mostotherattempts madein the past to solve the timeor quantum paradox.Perhapsour craziestidea is that
trajectories are not primary objects,but rather the result ofa
superpositionofplane waves. Poincareresonancesdestroy
158 TheEnd of Certainty
the coherenceofthesesuperpositions,and leadto anirreducible statisticaldescription.Oncethis is understood,the
generalizationto quantum mechanismsbecomeseasy.
IINumerousreferenceshave beenmadein this textto the
thermodynamic limit, which is defined by the limit N(numberofparticles)\342\200\224>
\302\260\302\260,
andvolume V \342\200\224>
\302\260\302\260,wherethe
concentrationN/Vremainsfinite. This limit simplymeansthat when the numberofparticlesN is sufficiendy large,termssuchas 1/N can beignored.This is true for the usual
thermodynamicsystemswhereN is typically on the orderof1023.However,thereare no systemsthat containaninfinite numberofparticles.
Theuniverse itself is highly heterogeneousand far from
equilibrium.This prevents systemsfrom reachinga state ofequilibrium.Forexample,the flow ofenergy that
originates in the irreversiblenuclearreactionswithin the sunmaintains our ecosystem far from equilibrium,and hasthus made it possiblefor life to developon earth.As wesaw in Chapter2,nonequilibriumleads to new collectiveeffects and to a newcoherence.It is interestingthat theseare precisely the consequencesof the dynamical theorypresentedin Chapters5 and 6.
Thereare two typesofeffects producedbynonequilibrium. If, as in the Benardinstability, we heat a liquid from
below, we producecollectiveflows ofmolecules.Whenwe stop the heatingprocess,the flows disintegrateand
return to the usual thermal motion.In chemistry, the
situation is different; irreversibility leads to the formation ofmoleculesthat cannotbe producedin near-equilibriumconditions.In this sense,irreversibility is inscribedin mat-
Our Dialoguewith Nature 159
ter.This is likely to be the originofself-replicatingbio-molecules.While we shall not pursue this question here,letus merelynote that moleculesofcomparablecomplexity
can indeedbe produced,at least through computersimulations,in nonequilibriumconditions.6In the nextchapter,whichdiscussescosmology,we arguethat matteritselfis the result ofirreversibleprocesses.
In nonrelativistic physics,whetherclassicalorquantum,time is universal, but the flow oftime as associatedwith
irreversible processesis not.It is to the fascinatingimplications ofthis distinctionthat we shall now turn.
IllLet us first considera chemicalmodel.Ifwe start at time
tQ
with two identicalsamplesofmixturesoftwo gases,suchascarbonmonoxide(CO)and oxygen(02),a chemicalreaction leadingto carbondioxide(C02)can becatalyzed bymetallic surfaces.In oneofthe samples,we introducesucha catalyst, and in the other,we do not.Ifwe comparethe
two samplesat a later time t, their compositionwill
therefore bequite different. Theentropyproducedin the
sample containingthe catalytic surface will bemuchgreaterasa result ofthe chemicalreaction.If we associatetheproduction ofentropyto the flow oftime,time itselfwill
appear to vary betweenthe two samples.This observationis
in agreementwith our dynamical description.Theflow oftime is rootedin Poincareresonancesthat dependon the
Hamiltonian,that is, on dynamics.Theintroductionofa
catalyst changesthe dynamics, and thereforealters the
microscopicdescription.In anotherexample,gravitationagain changesthe Hamiltonian,and thereforetheresonances. We then have a kind ofnonrelativistic analogueof
160The End of Certainty
the twin paradoxofrelativity, whichwe shallcomeback toin Chapter8.Forthe moment,supposethat we send twotwins (who are simply two LPS) into space,leaving theearth at
tQand comingbackat tx (seeFigure7.1).Before
their return,one twin goesthrough a gravitational field,and the otherdoesnot.Theentropyproduced(as a resultofPoincareresonances)will be different, and our twins
will comeback with different \"ages,\" leadingus to thebasicconclusionthat the flow oftime,even in aNewtonian universe, may have different effects accordingto the
processesconsidered.Our conclusionis in stark contrastwith the Newtonianview, whichwas basedon a universalflow oftime.But what can a flow oftime meanin a
description ofnature in whichpast and future play the samerole?It is irreversibility that leadsto a flow oftime.Timeevolution is no longerdescribedby groups wherepast and
Figure7.1Effect ofa Gravitational Fieldon the Flowof Time
h
Gravitational field /
<0
Our Dialoguewith Nature 161
future play the samerole,but rather by semigroups that
include the directionoftime.When we introducea timeassociatedwith the productionofentropy(seeChapter2),asthe sign ofentropy productionis positive, entropictime
always points in the samedirection.This is the casein thetwo examplesmentionedabove even thoughentropictimedoesnot keeppacewith clocktime.
We couldintroducean \"average\" entropictime for theentireuniverse, but this would not have a great deal ofmeaningbecauseof the heterogeneityof nature.Irreversible geologicalprocesseshave a time scaledistinctfrom
thoseofbiologicalprocesses.Even moreimportant, thereexistsa multiplicity ofevolutions,whichare particularlyevident in the field of biology.As stated by StephenJ.Gould,bacteriahave remainedbasicallythe samesincethePrecambrianera,while otherspecieshave evolved
dramatically,often over short time scales.7It would thereforebea
mistake to considera simple one-dimensionalevolution.Sometwo hundred million years ago, certain reptilesstarted to fly, while othersremainedon earth.At a later
stage,certainmammals returnedto the sea,while othersremainedon land.Similarly, certainapesevolved into hu-manoids,while others did not.
At the conclusionof this chapter, it is appropriate tociteGould'sdefinition ofthe historicalcharacteroflife:
Tounderstand the events and generalitiesof life'spathway, we
must go beyondprinciplesof evolutionary theory to a pale-ontologicalexamination of the contingent pattern of life's
history on our planet\342\200\224the singleactualizedversionamongmillions of plausible alternatives that happenednot to occur.Such a view of life'shistory is highly contrary both to con-
162The End of Certainty
ventional deterministicmodelsof Western scienceand to the
deepestsocialtraditions and psychologicalhopesof Western
cultures for a history culminating in humans as life'shighest
expressionand intended planetary steward.8
We are in a world of multiple fluctuations, someofwhichhave evolved, while othershave regressed.This is in
completeaccordwith the resultsoffar-from-equilibriumthermodynamicsobtainedin Chapter2.But we can now
go even farther. Thesefluctuations are the macroscopicmanifestations of fundamental propertiesof fluctuations
arising on the microscopiclevel of unstable dynamicalsystems.The difficulties emphasizedby Gould are nolongerpresent in our statistical formulation ofthe laws ofnature. Irreversibility, and thereforethe flow oftime,starts
at the dynamical level.It is amplified at the macroscopiclevel, then at the level of life, and finally at the level ofhuman activity. What drove these transitions from onelevel to the nextremainslargely unknown, but at leastwehave achieved a noncontradictory descriptionof naturerootedin dynamical instability. Thedescriptionsofnatureas presentedby biology and physicsnow begin to
converge.
Why doesa commonfuture exist at all? Why is thearrow oftime always pointedin the samedirection?Thiscan only mean that our universe forms a whole.It hasa commonorigin that already impliedtime-symmetrybreaking.Herewe encountercosmologicalproblems.Indealingwith them,we must embracegravity and enterthe
world ofEinstein'stheory ofrelativity.
Chapter8
DOESTIME PRECEDEEXISTENCE?
I
Severalyears ago, I delivereda physicscolloquiumat
LomonosoffUniversity in Moscow.Afterwards,Professor Ivanenko, oneofthe mostrespectedRussianphysicists,askedme to write a short inscriptionon a particular wall
wherethere werealready many sentimentsexpressedbyfamous scientistssuchas Diracand Bohr.I vaguely rememberthe sentencechosenby Dirac,which was somethinglike:\"Beauty and truth gotogetherin theoreticalphysics.\" After
somehesitation, I wrote:\"Timeprecedesexistence.\"Formany physicists,the acceptanceofthe bigbang
theory as the originofour universe means that time must
have a beginning,and perhapsan end.It seemsmorelikelyto methat the birth ofour universe was only oneevent in
the history of the entirecosmos,and that we therefore
163
164 The End of Certainty
have to ascribeto that so-called\"meta-universe\" a time
priorto the birth ofour own.We know that we are living in an expandinguniverse.
Thestandard model, whichdominatesthe field ofcosmology today, assertsthat if we wereto gobackwardin time,we would arrive at a singularity, a point that containsthe
totality ofthe energyand matter in the universe.However,the modeldoesnot enableus to describethis singularitybecausethe laws ofphysicscannot be applied to a pointcorrespondingto an infinite densityofmatter and energy.It is no wonderthat JohnArchibald Wheelerspeaksofthe
big bang as confronting us \"with the greatest crisis in
physics.\"1Canwe acceptthe bigbang as a real event, andhow is it possibleto reconcilethis event with laws ofnature that are time reversibleand deterministic?We comebackto the problemsofmeasurementand irreversibility,but now in the cosmologicalcontext.
Sincethe discoveryofthe bigbang,the scientific
community has reactedto the strangenature ofthis singularityby attempting to eliminatethe big bang entirely (seethe
steady-statetheory in SectionsI and III),orconsideringitas a kind of\"illusion\" arisingfrom the use ofan incorrectconceptoftime (seeHawking'simaginary time in SectionII),oreven viewing it as a sort ofmiracleakin to thebiblical descriptionin Genesis.
As we have already noted,it is impossibleto discuss
cosmology today without referringto the theory ofrelativity,
\"the most beautiful theory in physics,\" accordingto thecelebratedtextbookby Lev Davidovich Landau and
Evgeny Mikhailovich Lifschitz.2 In Newtonian physics,even when extendedby quantum theory, spaceand timeare given onceand for all. Moreover,there is a universaltime commonto all observers.In relativity, this is no
DoesTime PrecedeExistence? 165
longerthe case;spaceand time are now part ofthe picture.What consequencesdoesthis have for ourowninterpretation? In his recentbook,About Time, Paul C.W Daviescommentson the impactofrelativity, \"Thevery divisionoftime into past,presentand future seemsto bephysicallymeaningless.\"3He repeatsHermannMinkowski'sfamousstatement:\"Henceforthspaceby itself, and time by itself,are doomedto fade away into mereshadows.\"4
We have already alluded to Einstein'scelebratedassertion that \"for us convincedphysicists,the distinctionbetween past, present and future is an illusion, although a
persistentone.\"5At the end ofhis life, however, Einsteinseemsto have changedhis mind.In 1949,he was offered acollectionof essaysthat includeda contributionby the
great mathematician Kurt Godel,who had taken quiteseriously Einstein'sstatement that time as irreversibility was
only an illusion.When he presentedEinstein with a cos-mologicalmodelin whichit was possibleto return to one'sown past, Einstein was not enthusiastic.In his answer to
Godel,he wrote that he couldnot believethat he could\"telegraphback to his own past.\" Heeven added that this
impossibilityshouldleadphysiciststo reconsidertheproblem ofirreversibility.6That is preciselywhat we have
attempted to do.In any case,we wish to emphasizethat the revolution
brought about by relativity in no way affects our previousconclusions.Irreversibility, or the flow oftime,remainsas
\"real\" as in nonrelativistic physics.Perhapswe couldarguethat irreversibility playsan even greaterrolewhenwe go to
higherand higherenergies.It has beensuggested,mainly
by Hawking, that in the early universe, spaceand time losetheir distinction,and time becomesfully \"spatialized.\"Butno oneto ourknowledgehas deviseda mechanismfor this
166The End of Certainty
spatialization oftime,ora meansby whichspaceand timecouldemergefrom what is often describedas a \"foamymess.\"
Our positionis quite different from thosestated abovein that we considerthe bigbang an irreversibleprocessparexcellence.We suggest that there would have been anirreversiblephasetransition from a preuniversethat we callthe quantum vacuum. This irreversibility would result froman instability in the preuniverseinducedby the interactionsof gravitation and matter. Clearly we are at the edgeofpositiveknowledge,even dangerouslycloseto sciencefiction.
Nevertheless,we arguethat irreversibleprocessesassociated with dynamical processeshave probablyplayedadecisive rolein the birth ofouruniverse.From ourperspective,time is eternal.We have an age,ourcivilization has an age,our universe has an age, but time itself has neither a
beginningnor an end. This brings closertwo of thetraditional views of cosmology:the steady-statetheoryintroducedby HermannBondi,ThomasGold,and Fred
Hoyle, which may apply morepreciselyto the unstablemediumthat generatesour universe (the meta-orpreuniverse), and the standard bigbang approach.7
Again, speculativeelementscannot be avoided,but wefind it interestingthat views emphasizingthe roleoftimeand irreversibility can be formulated morepreciselythan
before,even thoughthe ultimate truth is still far beyondour reach.We agreeentirely with the Indian cosmologistJayant Vishnu Narlikar, who wrote, \"Astrophysicistsoftoday who hold the view that the 'ultimatecosmologicalproblem'has beenmoreor lesssolvedmay wellbe in for a
few surprisesbeforethis century is out.\"8
Does Time PrecedeExistence? 167
IIAs we proceedwith our investigation, let us considerEinstein's specialrelativity. This theory takesas its starting pointtwo inertial observersmoving at a constant velocity with
respectto oneanother.In prerelativistic, Galilean physics,it was acceptedthat the distancebetweenthe two
observers, / j2=
(x2\342\200\224xl)2+ (y2~yl)2 + (z2~zl)2,would remain
the same as the difference betweenthe two instants, (t2\342\200\224
tj)2. Spatial distance was defined in terms ofEuclideangeometry.This,however, ledto different values ofthe
velocity oflight c in the vacuum as measuredby the twoobservers. In accordancewith our experience,if we assumethat both observersmeasurethe samevalue ofthe velocityoflight, we must introduce(as did Lorentz,Poincare,and
Einstein)the spatiotemporalinterval, s22=
<?{tx\342\200\22412)2\342\200\224l22.
It is
this interval that is conservedwhenwe move from oneinertial observerto the other.In contrast to Euclideangeometry,
we now have the Minkowskispace-timeinterval. Thetransition from onecoordinatesystem,x,y, z, t, to another,x\\ y\\ z\\ t\\ is the famous Lorentztransformation that
combines spaceand time.At no point,however, is thedistinction betweenspaceand time lost; in the spatiotemporalinterval, the minus sign indicatesspacedimensions,andthe plussignindicatestime.
This situation is often illustrated by the spatiotemporaldiagram representedin Figure8.1.On oneaxis there is
time t, and on the othera singlegeometricalcoordinatex.In relativity, the velocity of light c in the vacuum is themaximumspeedat whichsignalscan be transmitted.We
can thereforedistinguishamongdifferent regionsin the
diagram.
168The End of Certainty
In this diagramthe observeris situatedat O.His future
is includedin the \"cone\"BOA, and his past in the coneA'O'B'.Theseconesare determinedby the velocity oflight c in that the velocitiesinside themare smallerthan c,and outside them are greater,and thereforeimpossibletorealize.In this diagram,the event C is simultaneouswith
O, while event DprecedesO.But this conclusionis purelyconventional becausea Lorentz transformation wouldrotate the axis t, x, in whichcaseDmight appearassimultaneous with O, and C posteriorto O. Simultaneity is
modified by the Lorentz transformation, but the coneoflight is not. The directionof time is thus invariant. Theproblemofascertainingwhetherornot the laws ofnatureare time symmetricremainsessentiallythe samein
relativityas in prerelativistic physics,but now this questionis
even morepertinent.At best, Oknows all the events that
occurredin his past, that is, in the coneA 'OB'.As
represented in Figure8.2,events starting in Cor D will reachhim only at later times, tx
and t2, even if they are associated
Figure8.1DistinctionBetweenFuture and Past in SpecialRelativity
/\342\226\240\342\226\240
r Absolute ^
C\302\273- x
x D
A' \"ua\"lult B'past
Does Time PrecedeExistence? 169
Figure8.2Events starting at C and D will reach the observerO at future
times fj and t2.
with signals traveling at the velocity oflight.As a result,Ocan collectonly limited data. In an amusing analogy with
deterministicchaosmadeby Baidyanath Misraand IoannisAntoniou, it is said that a relativistic observerhas only a
finite window on the outside world, and herealso a
deterministic descriptioncorrespondsto an overidealization.9This gives us yet anotherreasonto proceedto a statistical
description.Thereare, ofcourse,most interestingnew effects
introduced by relativity, such as the famous twin paradox,whereonetwin remainson earth at point x=0,while theother leaves in a spaceshipthat changesdirectionat
tQ (inthe coordinatesystem in which O is at rest), and comesback to earth at 2t(). The time interval, as measuredby the
moving twin, is greaterthan 2t0.This is Einstein'sremarkable time dilation prediction,whichhas beenverified by
170 The End of Certainty
usingunstable particles.The lifetime ofthesetwins
therefore dependson the path as predictedby relativity. InChapter7, we stated that the flow oftime dependson a
history ofevents, but Newtonian time is universal and
independent ofhistory. Now time itselfbecomeshistorydependent.
In his seminalbook,TheTheory ofSpace,Timeand
Gravitation,Vladimir A. Fockemphasizesthat we have to be
extremelycareful when discussingthe twin paradoxinasmuch as the effect ofaccelerationon the clockin the
moving spaceship is neglected.10He shows that when weconsidera moredetailedmodelin whichaccelerationis dueto a gravitational field describedby generalrelativity,
different results are obtained.The signoftime dilation can evenbe changed.Thesepredictionsofgeneralrelativity shouldleadto fascinatingnew experimentsto test their validity.
In his BriefHistory ofTime, Hawking introducesimaginary time,T = it, where all four dimensionsare \"spa-tialized\" in the Minkowski spatiotemporal interval.11Accordingto Hawking, real time may wellbe this
imaginary time,whereby the mathematical formula for theLorentz interval becomessymmetric.Hawking'sproposition doesindeedgo beyond relativity, but it is only onemoreattempt to negatethe reality of time in describingthe universe as a static, geometricalstructure,incontradiction to the rolethat the flow oftime plays at all levels ofobservation.
Let us now comeback to the cruxofourargument andconsiderthe effect ofrelativity on the systemsdescribedbyclassicalHamiltonian dynamics or quantum mechanics.Dirac,and otherswho cameafter him, showed how to
combinethe requirementsof specialrelativity with a
Hamiltoniandescription.12Relativity dictatesthat the laws
Does Time PrecedeExistence? 171
Figure8.3TheTwin ParadoxObserverO'is in motion in relation to observerO.
of physics remain the same for all inertial systems.InChapters5 and 6,we assumedimplicitly that the systemsas
a wholeare at rest.But accordingto relativity, a similar
description is valid whetheror not the systemas a wholeis
moving at uniform velocity with respectto someobserver.We have seenthat Poincareresonancesdestroythedynamical group in which past and future play the same role,wherebywe obtain semigroupsthat break time symmetry.Inprerelativistic physics,the groupsand semigroupsmaintain the distance
lf2invariant. In relativistic theory, we can
introduceas wellboth groupsand semigroupswhichleaveinvariant the Minkowskiinterval. Unfortunately, the proofis tootechnicalto begiven here.In any case,this
conclusion shows that the Minkowskispace-timeinterval is noway in contradictionto irreversibleprocesses.It is not truethat relativity impliesthe spatialization oftime.As stated
172 The End of Certainty
by Minkowski,spaceand time are no longerindependententities,but this doesnot precludethe existenceof anarrow oftime.
Sucha conclusioncouldbeanticipated.Iftime-symmetry breakingoccursin oneinertial frame, by the verydefinition ofrelativity, it has to appearin all inertial referencesystems.The theory ofirreversibleprocessesis thus quitesimilar (apart from certainformal changes)in both non-relativistic and relativistic systems.Thereis, however, onebasicdifference:Interactionsare no longerinstantaneous;rather, they propagateat the velocity oflight.Forchargedparticleswithin the framework ofquantum theory, for
example, interactionsare transmitted by photons.This leadsto additionalirreversibleprocessessuch as radiation
damping,whichresultsfrom the emissionofphotonsby
particles. In more generalterms,in relativistic physics weconsiderparticlesas associatedwith fields (the photonsarethe particlesassociatedwith the electromagneticfield),and
irreversibility resultsfrom the interactionofthesefields.Until now we have consideredthe Minkowskispace-
time interval as it correspondsto specialrelativity. In orderto completeour discussionofcosmology,we have toinclude gravitation, which first requiresa generalizationofthe space-timeinterval.
IllLet us first return to the questionofthe bigbang.As wementionedabove, by following our expandinguniversebackwardin time,we cometo a singularity in whichdensity, temperature,and curvature all becomeinfinite. Fromthe rate ofrecessionofthe galaxiesas observedtoday, we
Does Time PrecedeExistence? 173
can estimate that the birth of the universe occurredapproximately fifteen billionyears ago.This periodoftimethat separatesus from the bigbang is surprisinglyshort.To
expressit in years,we use the rotation of the earth as aclock.Fifteen billionrevolutions is indeeda small numberif we rememberthat in the hydrogenatom, the electronrotates some10,000billiontimespersecond!
Whatever the time scale,the existenceofa primordialevent at the originofour universe is certainly oneofthemost extraordinary suggestionssciencehas ever made.Physicsdealsonly with classesofphenomena,and the bigbangdoesnot seemto belongto any ofthese.At first view,it appearsto have no parallel elsewherein physics.
Many scientistshave beenwilling to explainthis
singularityin terms ofthe \"hand ofGod,\"or the triumph of
the biblicalstory ofcreation,whereby sciencewouldreconstruct the existenceofan act that transcendsphysicalrationality. Othershave triedto avoid what they seeas a
disquietingsituation.Oneremarkableattempt in this senseis the steady-stateuniverse proposedby Bondi,Gold,and
Hoyle.13This modelis basedon the perfect cosmologicalprinciple:Notonly is thereno privilegedplacein the
universe, but there is also no privilegedtime.Accordingtothis principle,every observer,in the past and in the future,is able to attribute to the universe the samevalues ofparameters suchas temperatureand matter density.Thesteady-state universe is characterizedby an exponentialexpansioncompensatedby a permanentcreationofmatter. Thesynchronization betweenexpansionand creationmaintains aconstant density ofmatter-energy,and thus leads to the
imageofan eternaluniverse in a state ofcontinouscreation. In spiteofits appeal,the steady-statemodelimplies
174 The End of Certainty
certainmajordifficulties. Inparticular, in orderto maintainthe steadystate, we needa fine-tuningbetweencosmolog-ical evolution (the expansionofthe universe) and
microscopicevents (the creationof matter).As long as no
mechanismfor this is proposed,the hypothesisofcompensation betweenexpansionand creationis highlyquestionable.
It was an experimentalresult that ledthe great majorityofcosmologiststo rejectthe steady-statemodelin favor ofthe bigbang,whichis now consideredthe standard model.This occurredin 1965,when Arno Penziasand RobertWilson identified the now-famous fossil radiation at 2.7K.14Theexistenceofsuch radiation had beenpredictedas
early as 1948by Ralph A. Alpher and RobertHerman,who reasonedthat if the universe was much hotter anddenserin the past than it is today, then it must have been\"opaque,\"with photonspossessingsufficient energy tointeract strongly with matter. It can be shown that at a
temperature of approximately3,000K, the equilibriumbetweenmatter and light is destroyed,and ouruniversebecomestransparent as radiation is \"detached\"from matter.Theonly subsequentchangein the propertiesofthephotons that form the thermal radiation is the changein their
wavelength, whichincreaseswith the sizeofthe universe.
Alpher and Hermanwerethus able to predict that if the
photonsindeedformeda blackbody radiation at 3,000Kat the time when their equilibriumwith matter was
destroyed (that is,some300,000years after the \"origin\,")the
temperatureofthis radiation shouldcorrespondtoday to a
temperatureofabout 3 K.This was a landmarkpredictionthat anticipatedoneofthe greatestexperimentalfindingsofthis century.15
Thestandard modelis very muchat the coreofpresent-
Does Time PrecedeExistence? 175
day cosmology,and scientistsgenerallyacceptthat it leadsto a correctdescriptionofthe universe starting onesecondafter the bigbang singularity. But the state ofthe universe
duringits first secondoflife still remainsan openquestion.Why is theresomethingrather than nothing? This
appears to be the ultimate questionbeyond the rangeofpositive knowledge.However,this questioncan beformulatedin physicalterms,and therebylinked to the problemofinstability and time.Onesuch formulation that has becomequite popular today definesthe birth ofour universe as a
free lunch. EdwardTryon presentedthis idea in 1973,but it
seemsto hark back to PascualJordan.In Tryon'sview, ouruniverse can bedescribedas having two forms ofenergy:onerelatedto attractive gravitational forces,whichis
negative, and the otherrelated to mass accordingto Einstein'scelebratedformula E= mc2, whichis positive.16
It is tempting to speculatethat the total energy of theuniverse couldbezero,as is the energyofan emptyuniverse. Thebigbangwould thus beassociatedwith
fluctuations in the vacuum conservingthe energy.This is truly an
appealingidea.Thegenerationofnonequilibriumstructures (such as Benard vortices or chemicaloscillations),whereenergyis conserved,alsocorrespondsto a free lunch,for the priceofnonequilibriumstructuresis entropy, andnot energy.In this context,can we specify the originofnegative gravitational energy and its transformation into
positive matter-energy?This is the question that we shall
now address.
IV
PerhapsEinstein'smostprofoundcontributionwas toassociate gravitation with the curvature ofspace-time.As we
176 The End of Certainty
have seenin specialrelativity, the Minkowskispace-timeinterval is ds2= c^di2 \342\200\224 dp. In generalrelativity, the space-time interval becomesds2=
~Lg vdx^dxy, where[i, v take onfour values: 0 (time),and 1,2,3 (space).The ten distinctfunctions obtained (given that g v
=gv ) characterize
space-time,or Riemanniangeometry.A simple examplethat illustrates Riemanniangeometry is a sphereconsidered as a curved two-dimensionalspace.
In the Newtonian view, space-timeis given onceandfor all, independentofthe matter it contains.Nowweunderstand, thanks to the Einsteinian revolution, that theconnectionbetweenspace-timeand matter is expressedbyEinstein's fundamental field equations,which relate two
objects:On the onehand, we have an expressionthat
describes the curvature ofspace-timein terms ofthe g andits derivatives with respectto spaceand time,and on theother an expressionthat defines the material content in
terms ofits matter-energycontentand pressure.Thismaterial contentis the sourceofthe curvature ofspace-time.Einsteinappliedhis equationsto the universe as a wholeas
early as 1917,and in so doing,set the courseofmoderncosmology.To achieve this application, he developedatimeless static modelin accordwith his philosophicalviews.BaruchSpinozawas Einstein'sfavorite philosopher,and we can recognizehis spirit in the choiceofthe model.
Then camea successionofsurprises.AlexanderFried-mann and Georges-HenriLemaitreproved that Einstein'suniverse was so unstable, the smallestfluctuation would
destroy it.17On the experimentalside, Edwin PowellHubbleand his colleaguesdiscoveredthe expansionofouruniverse.18Then in 1965camethe observationofresidualblack body radiation, which led to the present standard
cosmologicalmodel.
Does Time PrecedeExistence? 177
In orderto go from the basicequationsofgeneralrelativity to the field ofcosmology,we have to introducesimplifying assumptions.Thestandard modelassociatedwith
AlexanderFriedmann,Georges-HenriLemaitre,HowardRobertson,and Arthur Walker is founded on the cosmo-logicalprinciplethat the universe,whenviewedon a largescale,may beconsideredhomogeneousand isotropic.Themetricsthus take on the far simplerform ds2 = <?dt2 -R(t)2dP (the so-calledFriedmanninterval). Thisexpression differs from Minkowskispace-timein two respects:dp is a spatial elementthat correspondsto eithera zero-spacecurvature (as in the Minkowskispace)or to apositive ornegative curvature (as in a sphereorhyperboloid).R(t), which is usually calledthe radius of the universe,correspondsto the limit ofastronomicalobservationsat
time t. Einstein'sequations relate R(t) and the spacecurvature to the averagedensityand pressureofthe energy-matter. Einstein'scosmologicalevolution is alsoformulated as conservingentropy, and his equationsare
consequently time reversible.It is generallyacceptedthat the standard modelpermits
us to understand at least qualitatively what happenedtoouruniverse a fraction ofa secondafter its birth.This is an
extraordinaryachievement,but we are still left with the
questionofwhat occurredbefore.When we extrapolateback to the past, we cometo a point of infinite density.Canwe extrapolatebeyond this point? To give an idea ofthe rangeof values involved here,it is useful to definePlanck'sscales,which measurethe length, time, and
energy obtained by using three universal constants: h,
Planck'sconstant; G, the gravitational constant;and c, the
velocity of light. We then obtain Planck'slength, / =(G/i/c3)~ 1CT33cm,Planck'stime on the orderof 1CT44
178 The End of Certainty
seconds,and Planck'senergy,correspondingto a high
temperature on the orderof1032degrees.It is plausiblethat
thesescalesrelate to the very early universe characterizedby an extraordinarilybrieftime,a minusculegeometricalsize,and an enormousenergy.In this \"Planckera,\"quantum effects are likely to play an essentialrole.19We havenow arrived at the very limits of modern-day physics,wherewe are confrontedwith the fundamental problemofthe quantization ofgravity or,equivalently, ofspace-time.A generalsolutionis still far from our grasp,but we may at
least formulate a modelthat includesthe roleofPoincareresonancesand irreversibility at the very beginningofouruniverse.Let us now describesomeofthe stepsthat ledus
to this model.We have noted that the Friedmannspace-timeinterval
can be written (when we considerthe caseofEuclideanthree-dimensionalgeometry)as ds2=
Q.2(t)(dt2\342\200\224dl2), wheret is the conformal time. This is the Minkowskispace-timeinterval multipliedby the function Q,2,whichis calledthe
conformal factor. Suchconformalspace-timeintervals haveremarkablefeatures, includingtheir conservationof theconeof light, for which ds2 = 0.As Narlikar and othershave stated,they are the natural starting point for quantumcosmologybecausethey includethe Friedmannuniverse asa specialcase.20
Theconformalfactor as a function ofspace-timerelatesto a field in the same way as do other fieldssuch as the
electromagneticfield.(Rememberthat a field is a
dynamical system characterizedby a well-definedenergy and
thereforea Hamiltonian).As shown by RobertBrout andhis coworkers,this factor has a unique quality in that it
correspondsto a negative energy (that is, its energy is
unbounded from below),while the energyofany given mat-
Does Time PrecedeExistence? 179
ter field is positive.As a result, the gravitational fielddescribed by the conformalfactor may play the roleof areservoirofnegative energyfrom whichthe energytocreate matter is extracted.21
This is the theoreticalbasisofthe \"freelunch\" model,wherethe total energy (gravitational field plus matter) is
conserved,while the gravitational energy is transformedinto matter. Brout and his colleagueshave proposedamechanismfor this extractionofpositive energy.Inaddition to the conformalfield, they have introduceda matter
field, and demonstrated that Einstein'sequations lead toa cooperativeprocessinvolving the simultaneousappearance ofmatter and a curved space-timestarting from theMinkowskispace-time(containingzerogravitational andmassenergy).Theirmodelshows that such a cooperativeprocesscausesthe exponentialgrowth ofthe radius ofthe
universe over the courseoftime.(Thisis known as the deSitteruniverse.)
Theseconclusionsare intriguinginasmuchas theyindicate the possibilityof an irreversible processtransforming
gravitation into matter. They also focus our attention onthe preuniversestage, the Minkowski vacuum, which is
the starting point for irreversibletransformations. It is
important to note that this modeldoesnot describecreationexnihilo. Thequantum vacuum is already endorsedby the
universal constants,and it is assumedthat we can ascribetothemthe values they have today.
Thebirth ofouruniverse is no longerassociatedwith a
singularity, but rather with an instability that is analogousto a phase transition or bifurcation.However, this theorystill presents a numberofvexingproblems.Brout et al.have useda semiclassicalapproximationin whichthematter field is quantizedwhile the conformalfield is treated
180The End of Certainty
classically.This situation is highly unlikely in Planck'sera,wherequantum effects play an essentialrole.
Edgar Gunzigand PasqualeNardonehave asked whythis processdoesnot occuron a continuousbasis if the
quantum vacuum associatedwith a flat geometricalbackground is indeedunstable in the presenceofgravitationalinteractions.They have demonstratedthat in this semiclas-sical approximation,we needan initial fluctuation of a
cloudofheavy particlesofmass on the orderof50Planckmasses(~50.10~5^)in orderto start the process.22
Theseresultscan be incorporatedinto a macroscopicthermodynamic approach,where the universe has to betreated as an opensystem.Thus,we can observe matterand energy beingcreatedat the expenseofgravitationalenergy (seeFigure8.4).Thiscompelsus to makea numberof modifications to the first law of thermodynamics,wherethereis now a sourceofmatter-energyleadingto a
changein the definition of quantities suchas pressure.*Sinceentropy is specificallyassociatedwith matter, thetransformation ofspace-timeinto matter correspondsto a
dissipative, irreversibleprocessproducingentropy.Theinverse process,whichwould transform matter into space-time,is impossible.The birth ofour universe would thus
be the result ofa burst ofentropy.The interactionof the gravitational and matter fields
leadsto divergencesarisingfrom brieftimesand shortdistances that correspondin quantum theory to high valuesofenergy and momentum.Theseso-called\"ultraviolet\"
divergencesare the objectofa numberofinterestingin-
*The \"creation\" pressure is negative. Therefore, an often-quoted theorem of
Hawking and Penrose showing that the universe starts with a singularity and
involves positive pressure is not applicable.
Does Time PrecedeExistence? 181
Figure8.4Matter Is Createdat the Expenseofthe Gravitational FieldIn this simple model, the universe would have no stable ground state.
Creation of matter
Unstable ground state
Gravitation
vestigations that have led to a procedureknown as therenormalizationprogram,which has proved to be quitesuccessful.Still,certaindifficulties remain.Thereis a
striking analogy betweenfield theory and the thermodynamicsituation discussedin earlierchapters.Hereagain, we are
dealingwith persistent interactions that neitherstart norstop,and we thereforehave to go beyondHilbertspace.
Although this new field theory is still in the making, itsmain conclusionis reasonable:Theremay be no stable
groundstate at the cosmologicallevel, sincethe conformalfactor reacheslowerenergiesas it createsmatter. While this
lineofresearchcontinuesto bepursued,the two conceptsemphasized in this book,irreversibility and probability,clearly form an important part ofthis approach.Universes
appear at siteswhere the amplitudesof the gravitationaland matter fields have high values. The places and thetimeswherethis occurshave only a statistical meaning,as
they are associatedwith quantum fluctuations ofthe fields.This descriptionappliesnot only to our universe, but alsoto the meta-universe,the mediumin which individualuniverses are born.In ourview, hereagain we have an ex-
t
182The End of Certainty
ample ofPoincareresonancessimilar to that ofthe decayofan excitedatom.In this case,however,the decayprocesscreatesnot photonsbut universes!Even beforeour universe
was created,therewas an arrow of time,and this arrowwill go on forever.
Of course,thus far we have only a simplifiedmodel.Einstein'sdreamofa unified theory that would includeall
interactionsremainsalive today23Nonetheless,such a
theorywould have to take into accountthe time-oriented
characterof the universe as associatedwith its birth and
subsequentevolution.This can beachievedonly if certainfields (suchas gravitation) play different rolesfrom others(suchas matter).Inotherwords,unification is not enough.We needa moredialecticalview ofnature.
Questionsconcerningthe originsoftime will probablyalways be with us.But the idea that time has nobeginning\342\200\224that
indeedtime precedesthe existenceofouruniverse\342\200\224is becomingmoreand moreplausible.
Chapter9
A NARROW PATH
I
Ithas often beensuggestedthat irreversibility has a cos-
mologicaloriginassociatedwith the birth ofouruniverse. It is true that cosmologyis neededto explainwhythe arrow oftime is universal, but irreversibleprocessesdidnot ceasewith the creationofouruniverse;they still goontoday, on all levelsincludinggeologicaland biologicalevolution. Although the dissipative structures introducedin
Chapter2 are routinely observedin the laboratoryas wellas in large-scaleprocessesoccurringin the biosphere,irreversibility can be fully understoodonly in terms ofa
microscopic descriptionthat was traditionally identifiedwith
classicaland quantum mechanics.This requiresa newformulation ofthe laws ofnature that is no longerbasedoncertitudes,but rather possibilities.In acceptingthat thefuture is not determined,we cometo the endofcertainty. Isthis an admissionofdefeat for the human mind? On the
contrary, we believethat the oppositeis true.
183
184 TheEnd of Certainty
The Italian author Italo Calvinohas written a delightfulcollectionofstories,Cosmocomics,in whichindividuals
livingin a very early stageofour universe gathertogetherto
rememberthe terribletime whenthe universe was so smallthat their bodiesfilled it completely1What would havebeenthe history ofphysicsif Newtonhad beena memberofthis community? Hewould have observedthe birth and
decayofparticles,the mutual annihilation ofmatter andantimatter. Fromthe start, the universe would have
appeared as a thermodynamicsystemfar from equilibrium,with instabilitiesand bifurcations.
It is true that today we can isolatesimpledynamicalsystems and verify the laws ofclassicaland quantummechanics. Still, they correspondto idealizationsapplicable tostabledynamical systemswithin a universe that is a giantthermodynamic system far from equilibrium,where wefind fluctuations, instabilities,and evolutionary patterns at
all levels.On the otherhand, certainty has longbeenassociated with a denialoftime and creativity. It is interestingto considerthis conundrumin its historicalcontext.
IIHow can we reachcertainty? This is the question that
lies at the heart of the work ofReneDescartes.In his
thought-provokingbookCosmopolis,StephenToulmin
attempts to clarify the circumstancesthat ledDescartesonthis quest.2Hedescribesthe tragic situation ofthe
seventeenth century, a time ofpolitical instability and war
between Catholicsand Protestants in the nameofreligiousdogma.It was in the midst of this strife that Descartesbeganhis searchfor a different kind of certainty that all
humans, independentoftheirreligions,couldshare.This
A Narrow Path 185
ledhim to his famous cogito, the foundation ofhis
philosophy,as wellas his convictionthat sciencebasedon
mathematics was the only way to reachsuchcertainty. Descartes'views, which proved to be immenselysuccessful,influenced Leibniz'sconceptofthe laws ofnature discussedin
Chapter1.(Leibnizalso wanted to createa languagethat
would heal the divisionsamongreligionsand bringaboutthe endofreligiouswars.)Descartes'pursuit ofcertaintyfound its concreterealization in Newton'swork, whichhasremainedthe modelfor physicsfor threecenturies.
Toulmin'sanalysisreveals a remarkableparallel betweenthe historicalcircumstancessurroundingDescartes'questfor certainty and thoseofEinstein's.ForEinstein as well,sciencewas a meansofavoiding the turmoilofeverydayexistence.Hecomparedscientificactivity to the \"longingthat irresistiblypullsthe town-dwelleraway from his noisy,crampedquartersand toward the silenthigh mountains.\"3
Einstein'sview ofthe human conditionwas profoundlypessimistic.He had lived through a particularly tragicperiod in human history spanning the riseof fascism andanti-Semitismand two world wars.His vision ofphysicshas beendefinedas the ultimate triumph ofhuman reasonover a violent world, separatingobjectiveknowledgefromthe domainofthe uncertainand the subjective.
But is scienceas conceivedby Einstein\342\200\224an escapefromthe vagariesofhuman existence\342\200\224still the scienceoftoday?We cannotdesertthe pollutedtownsand citiesfor the highmountains.We have to participate in the building oftomorrow's society.In the wordsofPeterScott,\"Theworld,our world, tries ceaselesslyto extendthe frontiers oftheknowableand the valuable, to transcend the givennessofthings, to imaginea new and betterworld.\"4
Sciencebeganwith the Prometheanaffirmation ofthe
186The End of Certainty
powerof reason,but it seemedto end in alienation\342\200\224a
negationofeverything that gives meaningto human life.Our beliefis that our own age can be seenas oneof a
quest for a new type ofunity in our vision ofthe world,and that sciencemust play an important rolein definingthis new coherence.
As we mentionedin Chapter8,at the end ofhis life,Einsteinwas offered a collectionofessaysthat includedacontributionby the greatmathematician Kurt Godel.Inhis answer to Godel,he rejectedhis idea of a possibleequivalencebetweenpast and future. ForEinstein,nomatter how greatthe temptationofthe eternal,acceptingtheidea of traveling backin time was a denial of the realworld.Hecouldnot endorseGodel'sradicalinterpretationofhis very own views.3
As Carl Rubino has noted,Homer'sIliad revolvesaround the problemof time as Achilles embarks on asearchfor somethingpermanentand immutable:
The wisdomof the Iliad, a bitter lessonthat Achilles, its hero,learns too late, is that such perfection can be gained only at
the cost of one'shumanity: he must losehis life in order to
gain this new degreeof glory. For human men and women,for us, immutability, freedomfrom change,total security,
immunity from life'smaddening ups and downs,will comeonly
when we depart this life, by dying, or becominggods:the
gods,Horacetells us, are the only living beingswho leadsecure lives, free from anxiety and change.6
Homer'sOdysseyappearsas the dialecticalcounterpartto the Iliad.Odysseusis fortunate enoughto begiven the
choicebetween immortality, by remainingforever the
lover ofCalypso,and a return to humanity and ultimately
A Narrow Path 187
oldageand death.In the end,he choosestime over
eternity,human fate over the fate ofthe gods.
SinceHomer,time has beenthe central themeofliterature. We find a reactionquite similar to that ofEinsteinin
an essayby the great writerJorgeLuis Borgesentitled\"A New Refutation ofTime.\"After describingthedoctrines that make time an illusion,he concludes:\"And yet,and yet...denyingtemporalsuccession,denying the self,
denying the astronomicaluniverse, are apparentdesperations and secretconsolations... .Timeis the substance I
am madeof.Timeis a river whichsweepsmealong,but I
am the river; it is the tigerwhichdestroysme,but I am the
tiger;it is a fire whichconsumesme;but I am the fire. Theworld, unfortunately, is real; I,unfortunately, am Borges.\"7Timeand reality are irreduciblylinked.Denyingtimemayeitherbea consolationora triumph ofhuman reason.It is
always a negationofreality.The denialoftime was a temptation for both Einstein
the scientistand Borgesthe poet.Einsteinrepeatedlystatedthat he had learnedmorefrom Fyodor Dostoyevskythan
from any physicist.In a letterto Max Bornin 1924,hewrotethat if he wereforcedto abandonstrict causality,he\"wouldrather bea cobbler,oreven an employeein a
gaming house,than a physicist.\"8In orderto beofany value at
all, physicshad to satisfy his needto escapethe tragedy ofthe human condition.\"And yet, and yet,\" whenconfronted by Godelwith the extremeconsequencesof his
quest, the denialofthe very reality that physicistsendeavorto describe,Einsteinrecoiled.
We can certainly understandEinstein'srefusing chanceas the only answerto ourquestions.What we have triedtofollow is indeeda narrow path betweentwo conceptionsthat both lead to alienation:a world ruledby deterministic
188The End of Certainty
laws,whichleaves no placefor novelty, and a world ruledby a dice-playingGod,whereeverything is absurd,acausal,and incomprehensible.
We have attempted to make this booka journeyalongthe narrow path, and thereby illustrate the roleofhuman
creativity in science.Strangelyenough, this creativity is
often undervalued. We all realizethat if Shakespeare,Beethoven,orvan Goghhad diedsoonafter birth, no oneelsewould ever have achievedwhat they did.Is this alsotrue for scientists?Would someoneelsenot havediscovered the classicallaws of motionif there had beennoNewton?Didthe formulation ofthe secondlaw ofthermodynamics dependentirely on Clausius?Thereis sometruth in the contrast betweenartistic and scientific
creativity. Scienceis a collectiveenterprise.In orderto beacceptable,the solutionto a scientificproblemmust satisfy
exactingcriteriaand demands.Theseconstraints,however,do not eliminatecreativity.They provokeit.
Theformulation ofthe time paradox is itself an
extraordinary feat ofhuman creativity and imagination.Ifscience had beenrestrictedto empiricalfacts, how couldit
ever have dreamedof denying the arrow of time? Theelaborationof time-symmetrical laws was not achieved
merely by introducingarbitrary simplifications.Itcombined empiricalobservationswith the creationoftheoretical structures.This is why the resolutionof the time
paradoxcouldnot beaccomplishedby a simpleappealtocommonsenseor ad hoc modificationsofthe laws ofdynamics. It was not even a matter ofsimplyidentifying the
weaknessesin the classicaledifice.In orderto makefundamental progress,we neededto introducenew physicalconcepts,such as deterministicchaos and Poincareresonances, and new mathematical tools to turn theseweak-
A Narrow Path 189
nesses into strengths. In our dialoguewith nature, wetransform what first appearas obstaclesinto originalconceptual structuresprovidingfresh insights into the
relationshipbetweenthe knowerand the known.
What is now emergingis an \"intermediate\"descriptionthat liessomewherebetweenthe two alienating imagesofadeterministicworld and an arbitrary world ofpure chance.Physical laws lead to a new form of intelligibility as
expressed by irreducibleprobabilisticrepresentations.Whenassociatedwith instability, whetheron the microscopicormacroscopiclevel, the new laws ofnature deal with the-
possibilityof events, but do not reducetheseevents to
deductible,predictableconsequences.This delimitation ofwhat can and cannot bepredictedand controlledmay wellhave satisfiedEinstein'squestfor intelligibility.
As we follow alongthe narrowpath that avoids thedramatic alternatives ofblind laws and arbitrary events, wediscoverthat a largepart ofthe concreteworld around us
has until now \"slippedthrough the meshesofthe scientificnet,\" to use Alfred North Whitehead's expression.9We
face new horizonsat this privilegedmomentin the historyofscience,and it is our hopethat we have beenable tocommunicatethis convictionto our readers.
NOTES
Acknowledgments
1.I.Prigogineand I.Stengers,Entre le Temps et I'Eternite (Paris:Librairie Artheme Fayard, 1988(2nd ed.,Paris, Flammarion,1992).
2.I.Prigogineand I.Stengers,DasParadox der Zeit (Munich: R.Piper& Co.Verlag, 1993);I.Prigogineand I.Stengers,Time,Chaosand Quantum Theory (Moscow:Ed.Progress,1994).
3.I.Prigogine,LaFin desCertitudes (Paris:OdileJacob,1996).4. I.Prigogineand I.Stengers,Order Out of Chaos (New York:
Bantam Books,1984);I.Prigogine,From Being to Becoming (SanFrancisco:W. H.Freeman, 1980).
Introduction
1.K.R.Popper,The OpenUniverse: An Argument for Indeterminism
(Cambridge:Routledge,1982),p. xix.2.W.James, \"TheDilemma of Determinism,\" in TheWill to
Believe (NewYork: Dover,1956).3.G.Gigerenzer,Z.Swijtink, T. Porter, J.Daston,J.Beatty, and
L.Kriiger, TheEmpire of Chance (Cambridge:CambridgeUniversity Press,1989),p.xiii.
4. SeeL.Kriiger, J.Daston,and M.Heidelberger,eds.,TheProbabilistic Revolution (Cambridge,Mass.:MIT Press,1990),1:80.
5.Gigerenzeret al., Empire of Chance.6.Popper,OpenUniverse.
7.R. Tarnas, The Passion of the Western Mind (New York:
Harmony, 1991),p. 443.
191
192Notes
8.I.Leclerc,The Nature of Physical Existence (London:Allen and
Unwin; New York: Humanities Press,1972).9.J.Bronowski, A Senseof the Future (Cambridge,Mass.:MIT
Press,1978),p.ix.10.S.Hawking, A Brief History ofTime:From the Big Bang to Black
Holes(New York: Bantam Books,1988).
Chapter1.Epicurus' Dilemma
1.For Epicurus, seeJ.Barnes,ThePresocratic Philosophers (London:Routledge,1989).He probably had in mind the Stoics,whobelievedin a kind of determinism.
2.For Lucretius, seeTitus Lucretius Carus,DeNatura Rerum, ed.C.Bailey (Oxford:OxfordUniversity Press,1947).
3.K.R.Popper,TheOpenSociety and Its Enemies (Princeton,N.J.:PrincetonUniversity Press,1963).
4. For Parmenides,seeBarnes,Presocratic Philosophers.5.Plato, The Sophist (New York: Garland, 1979).6.J.Wahl, Traite deMetaphysique (Paris:Payot, 1968).7. P. S.Laplace,Oeuvres Completes de Laplace(Paris: Gauthier-
Vilars, 1967).8.G.von Leibniz, Discourse on Metaphysics and Other Essays, ed.
D.Garberand R.Ariew (Indianapolis: Hackett, 1991).9.J.Needham,Scienceand Society in East and West: The Grand
Titration (London:Allen and View, 1969).10.For the Einstein-Tagorecorrespondence,translated by A.
Robinson, seeK.Dutta and A. Robinson,Rabindranath Tagore(London: Bloomsbury, 1995).11.Popper,OpenUniverse, loc.cit.
12.H.Bergson,Oeuvres (Paris: PressesUniversitaires de France,1959),p.1331.
13.James,Dilemma ofDeterminism, loc.cit.14.J.Searle,\"Is There a Crisisin American Higher Education?\"
Bulletin of the American Academy ofArts and Sciences46, no. 4(January 1993):24.
15.Scientific American 271,no.4 (October1994).16.S.Weinberg, in ibid.,p. 44.17.Hawking, Brief History ofTime, loc.cit.18.R.Descartes,Meditations metaphysiques (Paris:J.Vrin, 1976).19.R.Penrose,TheEmperor's New Mind (Oxford:Oxford
University Press,1990),pp.4-5.
Notes 193
20.A. N.Whitehead, Process and Reality, ed.D.Griffin and D.Sherborne, correcteded.(New York: Macmillan, 1978).
21.C.P.Snow,TheTwo Cultures and the Scientific Revolution. TheTwo
Cultures and a SecondLook.(Cambridge:CambridgeUniversityPress,1964).
22.R. J. Clausius, Ann. Phys. 125(1865):353; Prigogine and
Stengers,OrderOut of Chaos,loc.cit.23.A. S.Eddington, The Nature of the Physical World (Ann Arbor:
University of Michigan Press,1958).24.SeePrigogine,From Being to Becoming.25.H.Poincare,\"La Mecaniqueet l'experience,\"in Revue deMeta-
physique et Morale 1 (1893):534\342\200\224537, and LeconsdeThermody-
namique, ed.J.Blondin (Paris:Herman, 1923).26.For Zermelo,seeS.Brush, Kinetic Theory (New York: Perga-
mon Press,1962),vol. 2.27.R. Smoluchowski, \"Vortrage iiber die kinetische Theorieder
Materieund Elektrizitat,\" 1914,quotedin H.Weyl, Philosophy
of Mathematics and Natural Science (Princeton,N.J.:PrincetonUniversity Press,1949).
28.M. Gell-Mann, The Quark and the Jaguar (London: Little,
Brown, 1994),pp.218-220.29.M. Planck, Treatise on Thermodynamics (New York: Dover,
1945).30.M.Born, The ClassicalMechanics ofAtoms (New York: Ungar,
1960);quoted in M.Tabor,Chaosand Integrability in Nonlinear
Dynamics (NewYork: Wiley, 1969).31.Prigogine,From Being to Becoming, p.213.32.SeeH.Price, Time'sArrow and Archimedes' Point: New Directions
for the Physics ofTime (Oxford:OxfordUniversity Press,1996).33.J.L.Lagrange, Theorie desfonctions analytiques (Paris:Imprimerie
de la Republique,1796).34.Gell-Mann,Quark and the Jaguar.35.L. Rosenfeld,\"Unphilosophical Considerationson Causality
in Physics,\" in SelectedPapers ofLion Rosenfeld, ed.R.S.Cohenand J.J.Stachel,Boston Studies in the Philosophy ofScience, vol.21(Dordrecht:Reidel,1979),pp. 666-690.
36.Borel,quoted in L. Kriiger, J.Daston,and M. Heidelberger,Probabilistic Revolution.
37'.J.W Gibbs,Elementary Principles in Statistical Mechanics (NewYork: Scribner's,1902).
38.H.Poincare,TheValue ofScience(New York: Dover,1958).
194 Notes
39.B.Mandelbrot, The Fractal Geometry of Nature (San Francisco:W.H. Freeman, 1983).
40.H.Poincare,New Methods of Celestial Mechanics, ed.D.GorofF(American Institute of Physics, 1993).
41.M.Born, quotedin M.Tabor,Chaosand Integrability in
Nonlinear Dynamics, p. 105.42.Tabor,Chaosand Integrability.
43.M.Jammer, The Philosophy of Quantum Mechanics (New York:
Wiley-Interscience,1974);A. I.M.Rae, Quantum Physics:Illusion or Reality? (Cambridge:CambridgeUniversity Press,1986).
44.P.Davies,TheNew Physics:A Synthesis (Cambridge:CambridgeUniversity Press,1989),p. 6.
45.Quotedby K.V Laurikainen, Beyond the Atom:The Philosophical
Thought ofWolfgang Pauli (Berlin:SpringerVerlag, 1988),p. 193.46.CI.George,I.Prigogine,and L.Rosenfeld,\"TheMacroscopic
Levelof Quantum Mechanics,\" Kong. Danske Viden. Selskab
Matematisk-fysiske Medd.38 (1972):1-44.47.See,e.g.,W. G. Unruh and W. H.Zurek, \"Reduction of a
Wavepacket in Quantum Brownian Motion,\" Phys. Rev. 40(1989):1070.
48.J. S. Bell, Speakable and Unspeakable in Quantum Mechanics
(Cambridge:CambridgeUniversity Press,1989).49.Gell-Mann,Quark and the Jaguar.50.G. C.Ghirardi, A. Rimini, and T Weber, Phys. Rev. D34
(1986):470.51.B.d'Espagnat, Conceptual Foundations of Quantum Theory,
Benjamin, California, 1976.52.SeeI.Farquhar, Ergodic Theory (London:Interscience
Publishers, 1964.)53.J.von Neumann, Mathematical Foundations of Quantum
Mechanics (Princeton,N.J.:PrincetonUniversity Press,1955).54.Cohen, Probabilistic Revolution.
55.H.Poincare,Scienceand Hypothesis (NewYork: SciencePress,1921).
Chapter2.Only an Illusion?
1.I.Prigogine,Bull. Acad. Roy. Belgique 31(1945):600.SeealsoEtude thermodynamique desphenomenes irreversibles (Liege:De-soer,1947).
Notes 195
2.Lagrange, Theorie desfonctions analytiques.
3.Hawking, Brief History ofTime.4. Bergson,L'Evolution creatrice, in Oeuvres, p. 784.5.Ibid.,p. 1344.6.Poincare,Scienceand Hypothesis.7.Whitehead, Process and Reality.
8.Eddington, Nature of the Physical World.
9.T.DeDonder and P.Van Rysselberghe,Affinity (MenloPark,Calif.:Stanford University Press,1967);I.Prigogine,Introduction to Thermodynamics ofIrreversible Processes, 3rd ed.(New York:
Wiley, 1967).10.G.N Lewis,Science 71(1930):570.11.E.Schrodinger,What Is Life?(Cambridge:Cambridge
University Press,1945).12.I.Prigogine,Bull. Acad. Roy. Belgique 3, (1945):600.13.L. Onsager,Phys. Rev. 37 (1931):405; 38 (1931):2265.The
proof of this theorem involves the celebratedOnsagerreciprocity relations.
14.P.Glandsdorff and I.Prigogine,Thermodynamic Theory ofStructure, Stability and Fluctuations (New York: Wiley-Interscience,1971).
15.G.Nicolisand I.Prigogine,Exploring Complexity (SanFrancisco: Freeman, 1989).
16.Ibid.17.For a review of oscillatory reactions,seeChemical Waves and
Patterns, ed. R. Kapral and K. Showalter (Newton, Mass:Kluwer, 1995).
18.For a review of nonequilibrium spatial structures, seeSpecialIssueof Physica A 213,nos.
1\342\200\2242, \"Inhomogeneous Phasesand
Pattern Formation,\" ed.J. Chanau and R. Lefever(North-Holland, 1995).
19.A. M.Turing, Phil. Trans. Roy. Soc.London, Ser.B, 237 (1952):37.
20.Nicolisand Prigogine,Self-Organization and Exploring
Complexity.
21.Nicolisand Prigogine,Exploring Complexity; Prigogine,From
Being to Becoming.22.C.K.Biebracher,G.Nicolis,and P.Schuster, Self-Organization
in the Physico-Chemical and Life Sciences, Report EUR 16546(European Commission,1995).
196Notes
Chapter3.From Probability to Irreversibility
1.Prigogine,From Being to Becoming.2.P.and T.Ehrenfest, Conceptual Foundations ofStatistical
Mechanics (Ithaca, N.Y.:CornellUniversity Press,1959).3.A. Bellemanns and J.Orban, Phys. Letters 24A (1967):620.4. I.Prigogine,Nonequilibrium Statistical Mechanics (New York:
Wiley, 1962);R.Balescu,Equilibrium and Non EquilibriumStatistical Mechanics (New York: Wiley, 1975);P.Resiboisand M.DeLeener,ClassicalKinetics ofFluids (New York: Wiley, 1977).
5.A. Lasotaand M.Mackey, Probabilistic Properties ofDeterministic
Systems (Cambridge:CambridgeUniversity Press,1985).6.Jan von Plato, Creating Modern Probability: Its Mathematics,
Physics, and Philosophy in Historical Perspective (Cambridge,Mass:CambridgeUniversity Press,1994).
7.D. Ruelle, Phys. Rev. Letters 56 (1986):405; Commun. Math
Phys. 125(1989):239;H.Hasegawaand W C.Saphir, Phys.
Rev. A 46 (1992):7401;H.Hasegawaand D.Driebe,Phys. Rev.
E 50 (1994):1781;P. Gaspard,J.of Physics A 25 (1992):L483;I.Antoniou and S.Tasaki,J.ofPhysics A: Math. Gen.26 (1993):73; PhysicaA 190(1992):303.
8.I.Prigogine,LesLoisdu Chaos(Paris:Flammarion, 1994),and
Le leggi del caos(Rome:Laterza, 1993).
Chapter4:TheLawsof Chaos
1.Hasegawaand Saphir, Phys. Rev. A 46 (1992):7401;Hasegawaand Driebe,Phys. Rev. E 50 (1994):1781;P.Colletand J.Eck-man, Iterated Maps on the Interval as Dynamical Systems (Boston:Birckhauser, 1980);R Shields,The Theory of Bernoulli Shifts
(Chicago:University of ChicagoPress,1973).2.P. Duhem, La theorie physique. Son objet. Sa structure (reprint,
Paris:Vrin, 1981),vol. 2.3.Hasegawaand Saphir, Phys. Rev. A 46 (1992):7401;Hasegawa
and Driebe,Phys. Rev. E 50 (1994):1781;Gaspard,Journal ofPhysics 25 (1992):L483;Antoniou and Tasaki, Journal ofPhysics
A: Math. Gen.26 (1993):73.4. Ibid.5.Mandelbrot, TheFractal Geometry ofNature; P.and T.Ehrenfest,
Conceptual Foundations ofStatistical Mechanics.
Notes 197
6.Nicolisand Prigogine,Exploring Complexity; Prigogine,From
Being to Becoming.1.See,e.g.,F. Riesz and B. Sz-Nagy, Functional Analysis (NewYork: Dover,1991).
8.Prigogine,From Being to Becoming; V. Arnold and A. Avez,
Ergodic Problems of Classical Mechanics (New York: Benjamin,1968).
9.Hasegawaand Saphir, Phys. Rev. A 46 (1992):7401;Hasegawaand Driebe,Phys. Rev. E 50 (1994):1781;Gaspard,Journal ofPhysics 25 (1992):L483;Antoniou and Tasaki, Journal ofPhysicsA: Math. Gen.26 (1993):73.
10.P.Gaspard,Physics Utters A 168(1992):13,and Chaos3 (1993):427;H.Hasegawaand D.Driebe,Physics Utters A 168(1992):18,and Phys. Rev. E 50 (1994):1781;H.Hasegawaand E.Luschei,\"Exact PowerSpectrum for a System of Intermittent
Chaos,\" Physics Utters A 186(1994):193.
Chapter5:Beyond Newton's Laws
1.T.Petrosky and I.Prigogine,\"Alternative Formulation ofClassical and Quantum Dynamics for Non-IntegrableSystems,\"
Physica A 175(1991);T. Petrosky and I.Prigogine,\"PoincareResonancesand the Limits of TrajectoryDynamics,\" PNAS90 (1993):9393;T.Petrosky and I.Prigogine,\"Poincare
Resonances and the Extension of ClassicalDynamics,\" Chaos,Soli-tons and Fractals 5 (1995).
2.Seeany text on Fourier series.3.Prigogine,Nonequilibrium Statistical Mechanics.
4. SeePetrosky and Prigogine,\"PoincareResonances.\"5.SeeS.G.Brush, KineticTheory (Oxford:Pergamon Press,1972),
vol. 3.6. SeeY Pomeauand P.Resibois,Physics Reports 19,2 (1975):63.7. T.Petrosky and I.Prigogine,\"New Methodsin Dynamics and
Statistical Physics\" (forthcoming).8.Prigogine,Nonequilibrium Statistical Mechanics; seealso the
citations in note 1 of this chapter.
198Notes
Chapter6:A Unified Formulation of Quantum Theory
1.R. Penrose,Shadows of the Mind (Oxford:OxfordUniversity
Press,1994),chap.5.2.P.Davies,TheNew Physics; Rae, Quantum Physics.3.J.C.von Neumann, Mathematical Foundations ofQuantum Theory.4. T. Petrosky and I.Prigogine, \"Quantum Chaos, Complex
Spectral Representationsand Time-Symmetry Breaking,\"
Chaos,Solitons and Fractals 4 (1994):311;TPetrosky and I.Prigogine, Physics Letters A 182(1993):5;T.Petrosky, I.Prigogine,and Z.Zhang (forthcoming).
5.K. R. Popper,Quantum Theory and the Schism in Physics (To-towa, N.J.:Rowman and Littlefield, 1982).
6.The standard text isby P.A. M.Dirac,ThePrinciples ofQuantumMechanics (Oxford:OxfordUniversity Press,1958).
7.M.Jammer, The Philosophy of Quantum Mechanics (NewYork:
John Wiley, 1974).8.A. Eddington, The Nature of the Physical World (Ann Arbor:
University of Michigan Press,1958).9.A. Bohm, Quantum Mechanics (Berlin: Springer, 1986);A.
Bohm and M. Gadella,Dirac Sets, Gamov Vectors and Gelfand
Triplets (Berlin:Springer, 1989);G.Sudarshan, Symmetry
Principlesat High Energies, ed.A. Perlmutter et al. (San Francisco:
Freeman, 1966);G.Sudarshan, C.B.Chiu, and V. Gorini,Physical Review D 18(1978):2914.
10.Petrosky and Prigogine,\"Quantum Chaos;\"T Petrosky and
Z.Zhang (forthcoming).11.Petrosky and Prigogine,\"Quantum Chaos\"and Physics Letters;
Petrosky, Prigogine,and Zhang (forthcoming).12.N. Bohr, \"The Solvay Meeting and the Developmentof
Quantum Physics,\" in La Theorie quantique des champs (NewYork: Interscience,1962).
Chapter7:Our Dialoguewith Nature
1.V. Nabokov, Look at the Harlequins (New York: McGraw-Hill,1974).
2.Prigogineand Stengers,Entre leTemps et I'Eternite.
3.N. Bohr, Atomic Physics and Human Knowledge (New York:
Wiley, 1958).
Notes 199
4. A. I.M.Rae, Quantum Physics.
5.W. Heisenberg,The Physical Principles of the Quantum Theory
(Chicago:University of ChicagoPress,1930).6.SeeNicolisand Prigogine,Exploring Complexity.7.S.J.Gould,Scientific American 271,no.4 (October1994):84.8.Ibid.
Chapter8:DoesTime PrecedeExistence?
1.J.Wheeler, quoted in H.Pagels,Perfect Symmetry (NewYork:
Bantam Books,1986),p.165.2.L.D.Landau and E.M.Lifschitz, The ClassicalTheory ofFields
(London:Pergamon Press,1959).3.P.Davies,About Time (London:Viking, 1995).4. H.Minkowski, The Principle ofRelativity: Original Papers
(Calcutta: University of Calcutta, 1920).5.A. Einstein, Correspondence Einstein-Michele Besso 1903\342\200\2241955
(Paris:Hermann, 1972).6.Albert Einstein: Philosopher-Scientist, ed.R A. Schlipp (Evanston,
III:Library of Living Philosophers,1949).7.H. Bondi, Cosmology (Cambridge:Cambridge University
Press,1960).8. SeeJ.V Narlikar and T. Padmanabhan, Gravity, Gauge Theory
and Quantum Cosmology (Dordrecht:Reidel,1986).9. I.Antoniou and B.Misra,Journal ofTheoretical Physics 31(1992):
119.10V Fock, TheTheory of Space,Time and Gravitation (New York:
Pergamon Press,1959).11.Hawking, Brief History ofTime.12.P.A. M.Dirac,Rev. Mod.Phys., 21(1949):392;D.J.Currie,T.
F.Jordan, and E.C.G. Sudarshan, Rev. Mod.Phys., 35 (1962):350;R.Balescuand T.Kotera,Physica 33 (1967):558;U BenYa'acov,Physica.
13.Bondi,Cosmology.14.Seethe excellentaccountby S.Weinberg, TheFirst Three
Minutes: A Modern View of the Origin of the Universe (New York:
BasicBooks,1977).15.SeeAlpher and Herman, in Nature 162(1948):774,and
Physical Review 75, no.7 (1949):1089.16.SeeE.P.Tryon, in Nature 266 (1973):396.
200 Notes
17.See, for a general account, S. Weinberg, Gravitation and
Cosmology: Principles and Applications of the General Theory ofRelativity (New York; Wiley, 1972).
18.Ibid.19.SeeJ.V. Narlikar and T.Padmanabham, Gravity.
20.Narlikar and Padmanabhan, Gravity.21.R.Brout, F. Englert, and E.Gunzig, Ann. Phys. 115(1978):78;
General Relativity and Gravitation 10(1979):1;R.Brout et al,Nuclear Physics B 170(1980):228;E. Gunzig and P.Nardone,Physics Letters B 188(1981):412,and also in Fundamentals ofCosmicPhysics 11(1987):311.
22.E. Gunzig, J.Geheniau,and I.Prigogine,Nature 330 (1987):621;I.Prigogne,J.Geheniau,E.Gunzig, and P.Nardone,Proc.Nat.Acad. Sci.USA 85 (1988):1428.
23.S.Weinberg, Dreams of a Final Theory (New York: PantheonBooks,1992).
Chapter9:A Narrow Path
1.I.Calvino, Cosmicomics,trans. W Weaver (NewYork: Harcourt,Brace& World, 1969).
2.S. Toulmin, Cosmopolis (Chicago:Chicago University Press,1990).
3.A. Einstein, Ideas and Opinions (New York: Crown, 1954),p. 225.
4. P. Scott, Knowledge, Culture and the Modern University, 75th
Jubileeof the Rijksuniversiteit (Groningen, Holland, 1984).5.Albert Einstein: Philosopher-Scientist.6.CarloRubino, unpublished.7.J.L.Borges,\"A New Refutation of Time,\" Labyrinths, Penguin
Modern Classics(Harmondsworth: Penguin Books,1970),p.269.
8.A. Einstein and M.Born, The Born-Einstein Letters (New York:
Walker, 1971),p. 82.9.A. N.Whitehead, Process and Reality.
GLOSSARY
anthropicprincipleThe ideathat the conditions of the universe
are explainedby the fact that we are hereto observethem.
bifurcation The branching of a solution into multiple solutions asa system parameter is varied.
big bangThe initial event of our universe, describedas an
explosive creationof matter and energy from a point.
chaosThe behavior of systems in which closetrajectoriesseparateexponentially in time.
clinamenThe idea,due to Epicurus, that an elementof chanceisneededto accountfor the deviation of material motion from rigid
predeterminedevolution.
coarsegrainingThe averaging of dynamics over finite regionsofphasespace.
collapseof the wave function The extradynamical elementneededin orthodoxquantum theory for the wave function,
representing potentialities, to yield an actual state.
degreesof freedom The number of independent variablesneededto specifythe configurational state of a system. A singleparticle in three-dimensionalspacehas three degreesof freedom.
determinismThe viewpoint that evolution isgovernedby a setofrules that, from any particular initial state, can generateone and
only one sequenceof future states.
201
202 Glossary
deterministicchaosChaoticbehavior arising from an entirelydeterministic evolution law.
Diracdelta function The mathematical object, introduced byDirac,which may be considereda function definedas infinity at
one point and zeroeverywhere else.
dissipative structureSpatiotemporal structures that appearin far-
from-equilibrium conditions, such as oscillating chemicalreactionsor regular spatial structures.
eigenstateA state that when actedon by a given operatoryieldsthe same state multiplied by a number.
eigenvalueThe number that an eigenstate is multiplied by after it
is actedupon by the correspondingoperator.
ensembleAn imagined collectionof identical systems with
different initial conditions.
entropy A function of the state of the system that increasesmo-notonically for isolatedsystems and reachesa maximum at
thermodynamic equilibrium.
fractal The term coined by BenoitMandelbrot for mathematical
objectsof noninteger dimension. For example,the length of the
irregular coastlineof a country increasesas the scaleusedto measureit decreases,and so the coastlinehas a dimension betweenone and
two.
Friedmannuniverse A cosmologicalmodelof an expandinguniverse basedon the assumption of homogeneity and isotropy of the
universe on large scales.
Gelfand spaceThe function spacecontaining both the
generalized functions and the well-behavedfunctions they act on.
generalizedfunction The classof mathematical objectsto which
the Diracdelta function belongs.A generalizedfunction is not aregular mathematical function but is defined by how it acts on regularfunctions.
Glossary 203
H-theoremBoltzmann's finding that a function (the H-function)involving the one-particledistribution function appearsunidirectional in time behavior under evolution of a dilute gas ofinteracting particles.
Hamiltonian The energy of a dynamical system expressedin
terms of its coordinatesand momenta.
Heisenberguncertainty principleThe product of the
accuracies by which the position and momentum of a quantum particlemay bedeterminedas limited by Planck'sconstant. Completeaccuracy of either the position or the momentum implies completeindeterminacy of the other one.
HilbertspaceThe spaceof functions for which the integral of the
square of the functions is well definedand finite. This is the
function spacethat was usedas the setting for orthodoxquantummechanics. It has been subsequently appliedto classicalmechanicsand
statistical mechanics.
KAM theory Describesthe dynamical behavior of classesof non-integrable systems. As the energy of a system is increased,chaoticbehavior becomesmore prevalent.
kinetic theory The study of the thermodynamic and transport
propertiesof fluid and gassystems in terms of interparticleinteractions.
largePoincaresystem (LPS)A nonintegrable system due toPoincareresonancestaken in the thermodynamic limit so that its
energy spectrum is continuous.
LaplacedemonThe entity imagined by Laplacethat would beable,given the exactinitial conditions, to calculate the precisefuture evolution of our universe.
Loschmidt'sreversal paradoxThe argument, raisedagainst the
conclusionsof Boltzmann, that sincethe equations of motion in
an interacting particle system are reversible,one can considerreversing all the velocitiesin a system so that any time-orientedfunctions of the state of the system would then behave in an
opposite manner.
204 Glossary
LyapunovexponentThe rate of exponential separationof nearby
trajectoriesin a chaoticsystem.
map A discrete-timedynamical process.Markov processA processwherein the future evolution of a state
dependsonly on the present state.For a continuous time system this
means that the processis localin time, that is, there are no memoryeffects.
Newtonian dynamicsThe rules of evolution that form the coreof classicalphysics and that, in pre-quantum eradeterminism, werebelievedto underlie all physical reality.
nonintegrablesystemAn interacting system that cannot betransformed to noninteracting parts. If such a transformation canbeperformed, the system is integrable and the equations of motion canbe trivially solved.
Perron\342\200\224Frobenius operatorThe time evolution operator for
probability distributions in discrete-timesystems (maps).
phasespaceThe abstract spaceof points in which the coordinatesare the positions and velocitiesof the particles in an evolvingsystem.
Planck era The universe just after the big bang characterizedbythe Planck scales,involving three fundamental constants of nature,
h, c,and G.
PoincarerecurrencetheoremThe finding that the state of aclosedsystem, as definedby the values of the positions and
velocities of all the particles,will recurarbitrarily closelyunder timeevolution of the system.
PoincareresonancesCouplingof degreesof freedomthat leadto
divergent expressionsdue to small denominators if there isresonance betweenthem. The resonancesmay prohibit the solution ofthe equations of motion.
probabilitydistributionfunction The function representing the
relative weights of the systems or initial conditions distributed in an
ensemble.
Glossary 205
resonanceThe constructive interferencethat appearswhen two
frequenciesin a system are rationally related.
Ritz\342\200\224Rydberg principleThe frequency of spectral linesrepresenting
the differencebetweentwo energy levels.
secondlaw ofthermodynamicsThe principle that the entropyof an isolatedsystem may only increaseor remain constant under
time evolution.
self-organizationThe choice between solutions appearing at
a bifurcation point, determinedby probabilistic laws. Far-from-
equilibrium self-organizationleadsto increasedcomplexity.
spectraldecompositionThe expressionof an operatorin termsof its eigenstatesand eigenvalues in a given function space.
steady-stateuniverse A cosmologicalmodelwherein the
expansion of the universe is compensatedby a continuous creationofmatter.
thermodynamiclimit Theprocedureof consideringthe numberN of particlesand the size17of a system becomingarbitrarily largewhile the concentration,c =N/ V, remains finite and constant.
thermodynamicsThe study of the macroscopicpropertiesof asystem and their relations without regard to the underlying dynamics.
Turing structuresPatterns in chemicalsystems arising from an
interplay of reactionand diffusion processes;thesearean exampleofdissipative structure.
INDEX
aging, 78, 125Alper, Ralph A., 174anthropic principle, 15-16Antoniou, Ioannis, 169approximation: dissipative
structures not explained by, 73;evolution as due to, 23,24-25;fundamental problems solved in
terms of, 52;irreversibility as
due to, 23,24,81,91,105architecture, 60Arnold, Vladimir Igorevich, 41arrow of time: all having same
orientation, 102,162;Bernoulli
maps introducing, 90,96;constructive role of, 3; for dealingwith intelligent life, 15;denial
of, 1-2;dissipative structures
requiring, 73;entropy as, 19;as
eternal, 182;in evolving
universe, 4; as fact imposed by
observation, 74;hostility to
conceptof, 61-62;in
macroscopic processes,18;in non-equilibrium physics, 3;nonintegrable systems requiredfor, 39;physics' denial of, 2; in
realistic interpretation ofquantum mechanics, 54;relegated to
phenomenology, 2, 3; as sourceoforder, 26;space-time asconsistent with, 172;in structure
formation, 71;subjective
interpretation of, 49atomism, 9-10,127atoms, Bohr's theory of, 132\342\200\22433,
145
bacteria, 161baker transformations (maps),
96-105,91;approach toequilibrium in terms of, 102;and
Bernoulli map, 90,91;Bernoulli
map compared with, 97-98,103-4;Bernoulli shift for
representing, 99;aschaotic and
deterministic, 101;eigenfunctionsand eigenvalues for, 103;equivalence of individual and statistical
description broken with, 103;as
invertible, 101,102;numerical
simulation of, 98;Perron-Frobenius operator with, 103;recurrence in, 99-101;spectral
representation in, 103;successiveiterations of, 98-99,100;time
paradox associated with, 103;astime reversible, 101,103
207
208 Index
becoming, 10Bell,John, 51Belousov-Zhabotinski reaction,
66-67Benard instability, 158,175Bergson, Henri, 13,14,59,72Bernoulli maps, 83-88,84;arrow
of time introduced in, 90,96;baker transformation asgeneralization of, 96;bakertransformation compared with, 97\342\200\22498,
103-4;as describing chaotic
systems, 89\342\200\22490; evolution operator,93;as not invertible, 96;simulation of distribution function for,
86;simulation of trajectoriesfor, 85
Bernoulli polynomials, 93-94Bernoulli shift, 90,99,101Biebracher, C.K.,71bifurcation point, 66,61,69bifurcations: in chaotic behavior,
68;our universe involving
successive, 72;pitchforkbifurcation, 68,69;as sourceofdiversification and innovation,70;as sourceof symmetry
breaking, 69;successive, 69\342\200\22470,
10big bang, 172-75;as beginning of
the universe, 6, 163-64;birth ofthe universe as a free lunch, 175,179;first secondafter, 175;instability associated with, 6; as
irreversible, 166;as occurringfifteen billion years ago, 173;residual black body radiation,
131,174,176binary correlations, 121,122,
122biology: arrow of time in structure
formation in, 71;Darwinian
evolution, 19,20,183;
multiplicityof evolutions in, 161-62;
self-replicating biomolecules,159.Seealso life
black body radiation, 131,145,174,176
Bohm, Arno, 142Bohr, Niels: atom describedin
terms of energy levels, 132,145;complementarity principle, 74,150;Copenhagen interpretationof quantum mechanics, 50,150,156-57;on quantum leaps, 138;on vocabulary for quantum
physics, 150Boltzmann, Ludwig: on ensembles,
76; evolutionary approach to
physics, 2, 19-21;H-functionand H-theorem, 20;on
irreversibility and dynamical
systems, 27;on irreversibility as
illusory, 2, 19,21;on secondlaw of thermodynamics as
probabilistic, 20,22;two-compartment model, 22-23,76, 91
Bondi, Hermann, 166,173Borel,Emile, 29Borges,JorgeLuis, 187Born, Max, 26,40,134,135,187Brief History ofTime (Hawking), 7,
15,170Bronowski, Jacob,7Brout, Robert, 178,179Brownian motion, 42,43,43butterfly effect, 30-31
Calvino, Italo, 184canonical ensemble, 77canonical equations of motion,
110-11,112causality, 4, 187certainty: coming to end of, 183;
denial of time and creativityassociated with, 184;Descartes'
Index 209
quest for, 184\342\200\22485; resonances
introducing uncertainty, 44chance.Seeprobabilitieschaos:baker transformation as
chaotic, 101;bifurcations
associated with, 68;chaotic systems,30-31;in classical physics, 4;conditions for general definition
of, 156;indeterminism as due
to, 56;and limits of physical
concepts,29;in nonequilibriumphysics, 3;probabilistic laws ofdynamics for, 104;problem of assolvable at ensemble level, 87;resonances'influence on, 41;resonances leading to, 112;simplified example of, 35-37;statistical nature of laws of, 37;time
symmetry broken in, 105;trajectories as inadequate for
describing, 90-91,105.Seealsodeterministic chaos; laws ofchaos
chaotic maps, 81-88;two types ofevolutions with, 102.Seealsobaker transformations; Bernoulli
mapschemical reactions: at equilibrium,
65-66;flow of time varying in,159;irreversibility in, 158-59,183;nonlinear equations for
describing, 66;oscillating
reactions, 66,127,175;synthetic
chemistry, 71;time
directionalityin, 18
chemistry, nonequilibrium, 26,27,67-68
classical physics: chaos included in,4; as deterministic, 136-37;extending to unstable systems, 89;extensions of, 46,109;fundamental concepts of, 109-10;Hamiltonian H in, 133;as
incomplete, 108;instability in, 4,54,107;irreversibility in, 49;laws of nature in, 4, 138,184;naive realism of, 135;predictability in, 4;probabilities in,
5; quantum mechanics limiting
validity of, 107;as reductionist,
114;relativity showing limits of,107;resonances introducinguncertainty into, 44;statistical
descriptions in, 108;time in, 59,60;time-reversible processesin,28.Seealso Newtonian physics;trajectories
Clausius, Rudolf Julius, 18-19clinamen, 10-11,52,55,127coarsegraining, 24,51,52,53,
101-2Cohen,I.Bernard, 55collapse of wave function. See
reduction (collapse)of wave
function
collision operators, 124collisions: in Boltzmann's H-
theorem, 20;and correlations,
78, 19,80;of molecules, 78; in
persistent interactions, 115communication: correlations and
human, 79;measurement as
means of, 150complementarity principle, 74,
150complexity: ofbiological and
chemical structures, 71;emergence of, 128;evolving in non-equilibrium systems, 64;irreversibility associated with, 64
Conferenceon Statistical
Mechanics and Thermodynamics, 61\342\200\22462
conformal factor, 178-79,181conformal time, 178conservation of energy, principle
of, 76-77
210Index
conservation of wave vectors, law
of, 121conservative systems, dynamics of,
101constructive interference, 119continuous spectrum, 133-34,141contracting coordinate, 97,99control, 154Copenhagen interpretation, 50,
150,156-57correlations: binary correlations,
121,122,122;and collision, 78,19,80;communication
compared with,. 79;creation of, 122,147;defined, 78, 120-21;destruction of, 79, 80,122,147;dynamics as a history of, 122;dynamics of, 79;evolution of,
123;flow of, 78-80,80,88;vacuum of correlations, 121,122,122,123,123
cosmologicalprinciple, 177
cosmology:anthropic principle,15\342\200\22416;
birth of universeassociated with instability, 179;birthof universe resulting from burst
of entropy, 180;cosmologicalprinciple, 177; in Einstein, 176,177;a meta-universe, 164,181;observer in, 51;originaluniverse as highly organized, 28;possible worlds, 59,72; standard
model, 164,174-75,177;succession ofbifurcations in our
universe, 72; the universe as
evolutionary, 4, 155;why isthere something rather than
nothing, 175.Seealso big bang;steady-state theory
coupling constant, 140creation fragments, 122,122,124creativity: as amplification of laws
of nature, 71;certainty and
denial of, 184;democracy as based
on, 17;and determinism, 6; anddistance from equilibrium, 62;of nature, 62;in science,188
crystals, 62
Darwin, Charles, 2, 19-21Davies, Paul C.W., 48-49,
165decay processes:in beams of
unstable particles, 139-40;of excited
atoms, 138-39,139;exponential decay, 139,142;radioactive
decay, 17,18;universes created
from, 182deconstruction, 14DeDonder,Theophile, 60,61delocalized distribution functions:
defined, 114;and going outsideHilbert space,116,117;forpersistent interactions, 114,115,117,125,148,155-56;
persistent scattering describedby, 149;in quantum physics, 146
delta functions, 33,94-95,117,124n,143
democracy, scienceas conflicting
with, 17Democritus, 9DeMoivre, Abraham, 5denominators, problem of small,
40density matrix, 47,53-54,143-44,
147-48Descartes,Rene,16,184-85description: as idealized in
traditional laws of physics, 26;intermediate description ofnature,
189;nonlocal, 37, 42,96.Seealso statistical level ofdescription; individual level ofdescription
de Sitter universe, 179
Index 211
destruction fragments, 122,122,124
destructive interference, 119determinism: baker transformation
as deterministic, 101;as basedon idealizations, 29;in classicaland quantum physics, 136\342\200\22437;
creativity and ethics and, 6;dilemma of, 1,6,14;divine
viewpoint required for, 38;as
marhematizable, 14;moving
away from, 131-32;Newton's
relation of acceleration andforceas deterministic, 110;in
pre-Socraticphilosophy, 9-10;in Western philosophy, 11
deterministic chaos:Bernoulli mapfor, 83\342\200\22488; in equations ofmotion, 31;exponential divergenceas signature of, 84;as
inapplicable in quantum mechanics, 53;individual and statistical
descriptions not equivalent in, 94\342\200\22495;
irreducible probabilisticdescriptions for, 45;Laplace demon in
world of, 38;and Poincare non-integrability, 112;statistical
description of, 105deterministic trajectories, 41diffusion: diffusive term in
persistent interactions, 44-45,54;diffusive terms in quantum
mechanics, 53\342\200\22454; entropyassociated with, 41;as irreversible,3, 105;resonances leading to,
42-44,43,126,155;thermal
diffusion, 26,21,35,64;trajectories leading to in chaos,41
dilating coordinate, 97,99dilemma of determinism, 1,6,
14\"Dilemma of Determinism\"
(James), 14
Dirac, Paul, 134,163Diracdelta functions, 33,94-95,
117,124n,143discrete spectrum, 133,141disorder, as constituting foundation
of microscopic systems, 155dissemination, 79dissipative structures: arrow of
time required by, 73; defined,66;emergenceof, 128;and
equilibrium, 67;homogeneityof spaceand time broken by, 69;and irreversibility, 73, 183;newkinetic theory for describing,126;in nonequilibrium physics,3, 27;and Poincare's recurrence
theorem, 101;self-organizationin, 70
distribution functions: additional
information provided by, 37;and Bernoulli maps, 86,87-88;as density matrices, 143-44;ensembles represented by, 33;at
equilibrium, 76-77,117;evolution over time, 112;integrating,120-21;Liouville operatordetermining evolution of, 113;localized, 114,115;microstructureof phase spaceaccounted for in,95;and Perron-Frobenius
operator, 87-88,91,94;in quantumstate representation, 143-44;smoothness in solutions of,87-88;in statistical descriptionof dynamical problems, 91\342\200\22496,
125-28;uniform distribution,93;written as Bernoulli
polynomials,93-94.Seealso delocalized
distribution functionsdistributions. Seegeneralized
functions
domain of validity,29
212Index
dualism: of Descartes,16;inquantum mechanics, 50,53,130,131;Weinberg on, 15
Duhem, Pierre-Maurice, 90dynamical decomposition, 102dynamical groups, 104,142,
171dynamical systems: dynamics as
history of correlations, 122;general problem of dynamics,40;instability in, 55,127-28;integrable systems, 39,44,54,108,131,144;and
irreversibility,27-28,126;as largely non-
integrable, 39;majority as
nonintegrable large Poincare
systems, 127;past and future in,102;phase spacerepresentation,31,32,110;Poincare on,
38-41;recurrence in, 99-101;returning to initial state in
immeasurable time, 23;solvingproblems at statistical level,91-96,113,125-28;stable,30-31,36,55;within athermodynamic system, 184.Seealso
nonintegrable systems; unstable
systems
Eddington, Arthur Stanley, 19,59,137
Ehrenfest, Paul and Tatiana, 74,81,101
eigenfunctions: with baker
transformation, 103;as central in
statistical and quantum mechanics,
95;defined, 92;of evolution
operator, 93;of Hamiltonian
operator, 133;of Liouville
operator, 113,118;of operators,134\342\200\22435;
in spectralrepresentation of an operator, 94-95;instatistical formulation ofquantum physics, 144
eigenvalues: with baker
transformation, 103;as central in
statistical and quantum mechanics,95;defined, 92;of evolution
operator, 93;of Hamiltonian
operator, 133,138,140-42;ofLiouville operator, 113,118,120,149;in spectral
representation of an operator, 94\342\200\22495;in
statistical formulation ofquantum physics, 144
Einstein, Albert: cosmology of,176,177; on ensembles,32,34,76;on freedom, 13;fundamental field equations, 176;generalrelativity, 46;on Godel'scosmo-logical model, 165,186,187;on
gravitation as curvature ofspace-time, 175\342\200\22476;
mass-
energy equation, 175;onquantum mechanics, 5, 135,151;on
science,185;on scientific
determinism, 11;special relativity,
167;and Spinoza, 176;on timeas an illusion, 1,58,165,187;time dilation prediction, 169;and unified theory, 182;on
unity ofnature, 16electromagnetic waves, 112elementary particles, 115,140Empire of Chance,The (Gigerenzer
et al.),4energy: chaos arising at critical
value of, 41;conservation of,76\342\200\22477; Einstein's mass-energy
equation, 175;free energy in
equilibrium conditions, 63,64;gravitational energy transformedinto matter, 174,179;total
energy, 76, 175energy levels:in Bohr description
of the atom, 132-33,145;eigenvalues determining,138
Index 213
ensembles:canonical ensemble,77;chaos problem solved at level of,87;defined, 31-32;distribution
function at equilibrium, 76-77;representation in phase space,33,34;trajectories comparedwith, 82,83,87; transition fromwave functions to, 131;forunstable systems, 81
Entre le Tempset VEternite (Pri-gogine and Stengers), 154
entropic time, 161entropy: as the arrow of time, 19;
birth of universe resulting fromburst of, 180;in Boltzmann
model, 76;diffusion associated
with, 41;entropic time, 161;flow of time correlated with,
159;and information, 24;and
irreversibility, 17, 18\342\200\22419;
irreversible processesas creating, 61;life as feeding on negative
entropy flow, 63;matter associated
with, 180;in nonrelativistic
twin paradox, 160;observation
increasing, 50;ordering comingfrom, 26;in secondlaw ofthermodynamics, 18-19,60
Epicurus: atomism of, 9-10;clina-men concept,10-11,52,55,127;on human freedom, 10
equations of motion: Bernoulli
maps for, 83-90;canonical
equations of motion, 110-11,112;as deterministic, 31;forfree particle, 118;Newton's law
relating forceand acceleration,
11,109-10;of oscillatorinteraction with field, 112;periodicmaps for, 82-83;time-reversible
processesdescribedby, 18equilibrium: approach to with
baker transformation, 102;black
body radiation in equilibrium
with matter, 131,145,174;inBoltzmann's two-compartmentmodel, 22\342\200\22423; calculating rate
of approach to, 95;and
correlation, 78, 80;creativity and
distance from, 62;and dissipation,67;distance from as parameterfor describing nature, 68;distribution function of ensembles at,
76-77;distribution functions at,
117;free energy and, 63,64;limiting thermodynamics to,61\342\200\22462;
matter acquiring new
properties when far from, 65,67;matter as blind in, 127;stable and unstable, 30,30-31;in
system of particles, 20;thermodynamic equilibrium, 60,63,66,77, 113,130;uniform
distribution corresponding to, 93;the
universe as far from, 158.Seealso nonequilibrium processes
equilibrium statistical
thermodynamics, 77
equilibrium thermodynamics, 88,108
Espagnat, Bernard, 53ethics: and determinism, 6; time
associated with, 58events, 5Everett, Hugh, 49evolution: approximation as
responsible for, 23,24-25;Boltzmann's evolutionary approach to
physics, 2, 19-21;ofcorrelations, 123;Darwinian, 19,20,183;described in terms of
probabilities, 55;of distribution
functions, 112;dynamical instabilityas condition of, 128;as multiplein biology, 161-62;in
reformulated laws of physics, 16;two
types with chaotic maps, 102;the universe as evolutionary, 155
214Index
evolution operator, 93,125-26,142
excitedatoms, decay of, 138-39,139,142-43
exponential decay, 139,142
Feynman, Richard, 47, 147nFierz, Markus, 50,53Fifth Solvay Conferenceon
Physics, 150first law of thermodynamics, 180flow of time: depending on history
of events, 170;as globalproperty, 20;gravitational fields
affecting, 159-60,160;introduced in quantum physics,5; in nonrelativistic physics, 159;progressing up levels oforganization, 162;resonances as sourceof, 159;as universal in
Newtonian physics, 160,164,170;as
varying, 159fluctuations: in big bang, 175,180;
in equilibrium conditions, 63;as
multiple, 162;in nonequilib-rium systems, 64,68-70;originof, 72; in urn model, 75,76
Fock, Vladimir A., 170Fokker-Planck equations, 43Fourier series,116-17Fourier transform, 117,118,147fractals, 38freedom: democracy as based on,
17;and determinism in Western
tradition, 6; Einstein's denial of,13;Epicurus on, 10;timeassociated with, 58
free energy, in equilibrium
conditions, 63,64free Hamiltonians, 111,138,140\"free lunch\" model, 175,179free particle, motion of, 111,118,
156
Freud, Sigmund, 70Friedmann, Alexander, 176,177Friedmann space-time interval,
177, 178From Being to Becoming (Prigogine),
27,74function, and structure, 62functional analysis, 38functional spaces,38,46,92future: in chaotic and simple
dynamical systems, 102,104;the
common future, 162;as aconstruction, 106;in formal
solution of Schrodinger equation,142;interaction betweenknower and known creating,153;in Newton's relation ofacceleration and force, 110;as not
determined, 183;as orientationof arrows of time in nature,
102;and past as asymmetrical in
irreversible processes,28;and
past meeting in probabilities,137;and past not distinguishedin physics, 2, 138;predicting in
classical science,4; in specialrelativity, 168
Galilean invariance, 110Galilean physics, 1,167gases, kinetic theory of.Seekinetic
theory of gasesGelfand space,96Gell-Mann, Murray, 24,28-29,
51,52general evolution criterion, 65generalized functions, 33n; Dirac
delta functions, 33,94-95,117,124n,143;in functional analysis,
38;and going outside Hilbert
space,117;nice functionscontrasted with, 94;as not included
in Hilbert space,92
Index 215
general problem of dynamics,40
general relativity: extension ofclassical mechanics in, 46;goingto cosmology from equations of,177;space-time interval in, 176
geological processes,time scaleof,161
Ghirardi, Giancarlo, 52Gibbs, Josiah Willard, 32-34,61,
76Gigerenzer, Gerd,4Glansdorff, Paul, 64God:and the big bang, 173;dice
playing by, 188;divine
viewpoint required for determinism,38;asgoverning the universe
deterministically, 11,12;scienceas reading the mind of, 7
Godel,Kurt, 165,186,187Gold,Thomas, 166,173Gould, Stephen J.,161gravitation: as curvature of space-
time, 175-76;quantization of,178
gravitational fields: matter as
created at expenseof, 179,180,181;time flow affected by,
159-60,160groups, dynamical, 104,142,171Gunzig, Edgar, 180
Hamiltonian H:in classical physics,133;defined, 110;distributionfunction as function of, 76-77,117;equations of motion
derived from, 110-11;flow oftime dependent on, 159;as her-mitian, 138;potential energy in
as sum of binary interactions,120;in quantum physics, 138;special relativity and descriptionby, 170-72
Hamiltonian operator H:eigenvalues of, 133,138,140-42;inquantum theory, 133;and
Schrodinger equation, 136;instatistical formulation ofquantum physics, 144
harmonic oscillators, 39,102harmonics, 40,123Hartle, JamesB.,51Hawking, Stephen W.: on an-
thropic principle, 15-16;on
future of science,7; on imaginary
time, 58,164,170;on spatializa-tion of time, 165;on universe
starting with a singularity, 180nHegel,GeorgWilhelm Friedrich,
13Heidegger, Martin, 10,13,14Heisenberg, Werner, 134,157Heisenberg uncertainty principle,
74, 135-36,143Heraclitus, 10Herman, Robert, 174hermitian operators, 138H-function of Boltzmann, 20Hilbert space:as dynamical group,
104;dynamical groups outside,142;eigenvalues of Hamiltonian
in, 138;eigenvalues of Liouville
operator in, 113;equivalence ofindividual and statistical
description breaking down outside, 96;extending wave functions
beyond, 144\342\200\22445;in functional
calculus, 46;generalized functions
not included in, 92;going
beyond, 93,95-96,114,116,117,126,181;quantum mechanics as
operator calculus in, 95;riggedHilbert space,96;and spectral
representation with baker
transformation, 103,104Homer, 186-87
216Index
Hoyle, Fred, 166,173H-theorem ofBoltzmann,
20Hubble, Edwin Powell, 176Husserl, Edmund, 13
Iliad (Homer), 186imaginary time, 58,164,170indeterminism: as
anthropomorphic,14;as compatible with
realism, 132;as due to instabilityand chaos,56;in fundamental
laws of physics, 16;statistical
description of unstable systemsrequiring, 109
indistinguishability of elementaryparticles, 140
individual level of description:breaking equivalence with
statistical description, 35,83,87,89,94-95,96,103,106,155;inclassical and quantum physics,129;for integrable systems, 108;operators required for, 133;statistical description in terms ofLiouville operators comparedwith, 118-19;statistical level as
equivalent to, 34-35,42,81,108;thermodynamics as
incompatible with, 127;transition tostatistical level in quantum
physics, 143-46;validity as
limited, 127inertia, 110infinite velocities, 105\342\200\2246
information: and entropy, 24;probabilities as expressing lack
of, 34;probability distributions
providing additional, 37;secondlaw of thermodynamics as dueto lack of, 25-26
initial conditions, sensitivity to, 30,37,90
instability: big bang associated
with, 6; birth of the universeassociated with, 179;and classical
dynamics' revival, 31;in classical
physics, 4, 54,107;classical
physics requiring extension for,46;and distance from
equilibrium, 66;in dynamical systems,55,127-28;equivalence ofindividual and statistical descriptionsdestroyed by, 35;indeterminismas due to, 56;in laws ofnature,4, 155,189;and limits ofphysical concepts,29;in quantum
physics, 4, 53-54,107;at
statistical level of description, 155;time linked to, 175;time
symmetry broken by, 5, 37-38;at
trajectory level leading to
stabilityat statistical level, 87.Seealso
unstable systems
integrable systems, 39,44,54,108,131,144
interactions: defined, 147;betweenknower and known, 153;in rel-ativistic systems, 172;statistical
description affected by, 120-25.Seealso persistent interactions;transient interactions
interference, constructive and
destructive, 119interference terms, 52inverse Fourier transform, 117irreversibility: all processes
oriented in same direction, 102,104\342\200\2245;
as appearance only in
immeasurably long time, 23;approximation as responsible for,
23,24,81,91,105;big bang as
irreversible, 166;and birth ofthe universe, 181,183;inchemical reactions, 158-59,183;constructive roleof, 3, 26,27, 57;
Index 217
and dissipative structures, 73,183;and dynamical systems,27-28,126;emergence of,96,105;and entropy, 17,18-19;entropy created by, 61;gravitation-energytransformation as irreversible, 179;as
illusory, 2, 19,21,165;in laws ofnature, 38,96;leading to longmemory effects, 125;lifeassociated with, 63;and Lyapunov
time, 105;in macroscopicphysics, 45;matter as result of,159;in measurement, 49,53;innature, 18;in nonequilibriumphysics, 3; novel phenomenafollowing from, 3; and the
observer, 5; order and disorder in,
26;past and future asasymmetrical in, 28;in persistentinteractions, 114;probabilities for
describing, 35;progressing uplevels of organization, 162;inquantum physics, 53,138;reality of, 3, 25,27, 165;andsecond law of thermodynamics, 21;statistical description for, 81,108,109;thermodynamics as
scienceof, 17;as transient, 62
James, William, 1,13,14Jordan, Pascual, 134,175
KAM theory, 41Kant, Immanuel, 4, 10kinetic theory ofgases:Boltz-
mann's two-compartmentmodel, 22-23,76,91;Poincare
on, 35;probability in, 35;as
special caseof new approach,125
Kolmogorov, Andrei N., 41Kronecker, Leopold,84
Lagrange, Joseph-Louis, 28,58Landau, Lev Davidovich, 164Laplace, Pierre-Simon de, 11,14,
25,38,106large Poincare systems. Seenonin-
tegrable large Poincare systemslaw of conservation of wave
vectors, 121laws of chaos, 89-106;possibility
of speaking of, 87, 104laws of nature: in classical physics,
4, 138,184;creativity asamplification of, 71;eighteenth-century laws as deterministicand time reversible, 11;fundamental law of quantum physics,136;idealized world describedby, 26,184;and instability, 4,155,189;irreversibility in, 38,96;Poincare on laws ofthermodynamics, 55-56;probabilitiesin, 5, 29,35,38,44,132,189;in quantum physics, 4, 138,184;within the range of low
energies,6; reformulating
fundamental laws ofphysics, 16-17,108;statistical formulation of, 162;time's constructive role for, 56;when far from equilibrium, 65
Leclerc,Yvor, 7Leibniz, Gottfried von, 12,185Lemaitre, Georges-Henri,176,177Lewis, Gilbert N., 61life: and dissipative structures, 66;
duality in accounting for, 15;dynamical instability requiredfor, 128;as feeding on negative
entropy flow, 63;historicalcharacter of, 161-62;irreversibilityassociated with, 3, 63;nonequilibrium processesrequired for, 3,26-27;nonintegrable systemsrequired for, 39
218Index
Lifichitz, Evgeny Mikhailovich,164
light, velocity of, 105-6,167-68linear nonequilibrium
thermodynamics,63
Liouville equation, 113,136,144,146
Liouville operator: defined, 113;eigenvalues of, 113,118,120,149;for free particle, 118;Hamiltonian operator comparedwith, 136;spectralrepresentation of, 113,125,149;statistical
description in terms of, 118-19literature, time as theme of, 187localized distribution functions:
defined, 114;for transient
interactions, 114,115Lorentz, Hendrik Antoon, 134Lorentz transformation, 167,168Loschmidt, Joseph, 21,23LPS.Seenonintegrable large Poin-
caresystemsLucretius, 10,55Lyapunov exponent: in baker
transformation, 97-98,101;in
Bernoulli map, 84,87,90,93;comparison of neighboring
trajectories in, 105Lyapunov time, 105
macroscopic systems, 6, 45,115,128,162
macrostates, 24Mandelbrot, Benoit, 38Markov process,76matter: black body radiation in
equilibrium with, 131,145,174;as blind in equilibrium,127;as blind without arrow oftime, 3; as created at expenseofgravitational fields, 179,180,181;entropy associated with,
180;mind and, 16,49;new
properties acquired when far
from equilibrium, 65,67;permanent creation in steady-statetheory, 173;phase transitions,45,116;probabilities required tounderstand properties of, 47-48;as result of irreversibility, 159;and space-time for Einstein,176;states of matter, 45,116;transformation of space-timeinto, 180
Maxwell, James Clerk:on
ensembles, 76;and kinetic theory, 125;on new kind of knowledge, 4, 5
Maxwell-Boltzmann distribution,20
measurement: as actualizing
potentiality, 48;in Copenhageninterpretation, 156-57;infundamental description ofnature, 49;irreversibility in, 49,53;as means of communication,150;as probabilistic, 15
measuring instruments, 51,54,150-51
Meneceus,10meta-universe, 164,181microcanonical ensemble, 77microstates, 24mind and matter, 16,49Minkowski, Hermann, 165Minkowski space-time interval,
167,171,172,176,177, 178Minkowski vacuum, 179Misra, B.,169mixtures of wave functions, 47,
48,144morphogenesis, 68Moser,Jiirgen Kurt, 41motion, equations of.See
equations of motion
Nabokov, Vladimir, 154Nardone, Pasquale, 180
Index 219
Narlikar, Jayant Vishnu, 166,178nature: as automaton, 12,17;
Chinese and Japanese view of,
12-13;creativity of, 62;dialectical view of required, 182;distance from equilibrium as
parameter in describing, 68;duality in, 15,16;intelligibility of,
17,29;intermediate descriptionof, 189;mankind's position in,9; measurement included in
fundamental description of, 49;microscopic depiction of, 127;nineteenth-century views on as
conflicting, 17,19;probabilityas property of, 44;reversible andirreversible processesin, 18;science as dialogue with, 57,60,153;as semigroup that
distinguishesfuture and past, 104;
unity and diversity of, 56;unpredictable novelty in, 72.Seealso laws ofnature
Needham, Joseph, 12New Physics, The:A Synthesis
(Davies),48\"New Refutation ofTime, A\"
(Borges),187Newtonian physics: as absolute, 2,
28;deviations from, 126;equations of motion invariant with
respect to time inversion, 90;flow of time as universal in, 160,164,170;going beyond, 127;law relating forceand
acceleration, 11,109-10;limits ofvalidity of, 44,107-8;quantum
physics compared with, 129;asrealization ofDescartes'questfor certainty, 185;spaceandtime as given onceand for all,
164,176;unidirectional timedenied in, 2.Seealso classical
physics
nicefunctions, 92,94,95,96Nicolis, Gregoire, 71Nietzsche, Friedrich, 14nonequilibrium chemistry, 26,27,
67-68nonequilibrium processes:
complexity evolving in, 64;effects
produced by, 158\342\200\22459;
fluctuations in nonequilibrium systems,64,68-70;generation of as free
lunch, 175;irreversibility asconstructive in, 26-27;non-equilibrium physics required to
describethe world, 131;structure in nonequilibriumconditions, 63;unidirectional time in,3
nonequilibrium thermodynamics:DeDonder'swork on, 61;linear
nonequilibriumthermodynamics, 63;and views of Bergsonand Whitehead, 72
nonintegrable large Poincare
systems (LPS):continuous spectrain, 133-34;defined, 111;and
deterministic chaos, 112;example of, 111-12;formulation in
quantum mechanics, 148;individual and statistical descriptionnot equivalent for, 106;majorityof dynamical systems as, 127;as
measuring themselves, 151;resonances in, 141-42
nonintegrable systems, 39-41;arrow of time requiring, 39;dynamical decomposition in, 102;glass ofwater as, 78;KAM
theory of, 41;resonances as reason
for, 39-40,109,111;as rule not
exception, 108;thermodynamiclimit corresponding to, 45.Seealso nonintegrable large Poincare
systemsnonlocal description, 37,42,96
220 Index
nonlocality, in quantum theory,130
observer: in cosmology, 51;entropy increasing with
observation, 50;indeterminism asindependent of, 132;macroscopic
character of observation,23;in quantum physics, 5,48-55,131,151;secondlaw ofthermodynamics based on
ignorance of, 25;in special relativity,
167,169Odyssey (Homer), 186Open Universe,The:An Argument for
Indeterminism (Popper), 1,14operator formalism, 92,133,134operators: collision operators, 124;
description requiring, 133;eigenfunctions of, 134\342\200\22435;
evolution operator, 93,125-26,142;hermitian operators, 138;introduction into physics, 134;operator formalism, 92,133,134;quantum physics as
operator calculus in Hilbert space,95.Seealso Hamiltonian operatorH; Liouville operator; Perron-Frobenius operator; spectralrepresentation of an operator
order:arrow of time as sourceof,
26;disorder as constitutingfoundation ofmicroscopicsystems, 155;entropy and, 26;self-organization maintaining,71-72.Seealso entropy; self-organization
oscillating chemical reactions, 66,127,175
oscillators, harmonic, 39,102
Parmenides, 10past: in chaotic and simple
dynamical systems, 102,104;complex
conjugate of wave functions
propagating into, 137;in formal
solution of Schrodingerequation, 142;and future as
asymmetrical in irreversible processes,28;and future meeting in
probabilities, 137;and future not
distinguished in physics, 2, 138;interaction between knower andknown creating, 153;inNewton's relation of acceleration and
force, 110;retrodicting in
classical science,4; in specialrelativity,
168Pauli, Wolfgang, 50,53pendulum, 30Penrose, Roger, 16,130,180nPenzias, Arno, 174periodicfunctions, 116,117periodicmaps, 82,82-83Perron-Frobenius equation, 88,
104Perron-Frobenius operator: in
baker transformations, 103;inBernoulli maps, 83,87-88,91,94,96
persistent interactions, 113-14;de-localized distribution functionsfor describing, 114,115,117,125,148,155-56;diffusive
terms in, 44\342\200\22445, 54;leavingHilbert spacedue to, 96,114;inmacroscopic systems, 115;persistent scattering, 148-49;inthermodynamic systems,115-16,148
perturbational approach to solvingfor eigenvalues, 140-42
phase space:defined, 31;distribution functions accounting for
microstructure of, 95;dynamicalstate represented in, 31,32,110;ensemblesrepresented in, 33,34,76; resonances'influence
Index 221
on, 41;state of a classical systemin, 136;statistical descriptionaccounting for microstructure of,105
phase transitions, 45,116phenomenology, arrow of time
relegated to, 2, 3philosophy: scienceas separated
from, 7, 14,72;as time
centered, 13;time for, 58;unhappy
history ofWestern, 11photons, 172,174physics: Boltzmann's evolutionary
approach to, 2, 19-21;domainof validity, 29;Galilean physics,1,167;idealized worlddescribed by, 26;operatorformalism in, 92;probabilities as basic
objectsof, 74; reformulatingfundamental laws of, 16-17,108;resonance playingfundamental role in, 42;statistical
mechanics, 46,92,95;time for, 58.Seealso classical physics;cosmology;
Newtonian physics;quantum physics; relativity;
thermodynamicsphysics of populations, 35pitchfork bifurcation, 68,
69Planck, Max, 131,145Planck era, 178Planck's constant, 142,145,148,
177Planck's energy, 178Planck's length, 177Planck's scales,177-78Planck's time, 177plane waves: defined, 116;
superposition of, 117,118-19,119,157-58;trajectory as construct
of, 119Plato, 11Plato, Jan von, 84
Poincare, Henri: on distinguishingbetween dynamical systems, 31;on dynamical systems, 38-41;on explaining irreversibility in
terms of trajectories, 21;on free
Hamiltonians, 111;on generalproblem of dynamics, 40;onkinetic theory of gases, 35;onlaws of thermodynamics, 55-56
Poincare resonances.Seeresonances
Poincare's recurrence theorem, 23,39,100-101
Popper, Karl, 1,6, 14,132population dynamics, 33populations: aging as property of,
78, 125;Boltzmann and Darwin
studying, 20;phase transitions
meaningful only at level of, 45;physics of, 35
possibilities, 5, 29\"Possible and the Real,The\"
(Bergson), 14,59possible worlds, 59,72postmodern philosophy, 14predictability: in classical science,
4, 11;with deterministic chaos,38;predictive successofquantum theory, 46;unpredictablenovelty in nature, 72
pre-Socratics,9-10,17principle of conservation of
energy, 76-77principle of indistinguishability of
elementary particles, 140probabilities: as basic objectsof
physics, 74;as basic property ofnature, 44;and birth of the
universe, 181;for diffusive motions,42,43;evolutionarycharacteristics describedin terms of, 55;as
expressing ignorance, 34;asextended form of rationality, 155;in fundamental laws of physics,
222 Index
probabilities (cont.)16-17;in laws of nature, 5, 29,35,38,44,132,189;inmacroscopic physics, 45;inmeasurement, 15;in nonequilibriumsystems, 68;past and future
meeting in, 137;probability
amplitudes giving way to, 54,132,149;the probabilizingrevolution, 55,74, 132;in quantum
physics, 5, 51-52,54,132,149;secondlaw of thermodynamicsas probabilistic, 20,22;subjective interpretation of, 4, 16;astime symmetric, 137;trajectories and, 34;transition
probabilities, 75; for unstable systems, 35.Seealso statistical level ofdescription
probability amplitudes: dynamical
problems solved in terms of, 47;giving way to probabilities, 54,132,149;in physical
interpretation of wave function, 47;quantum state representing, 143;Schrodinger equationdescribing, 137;wave function
corresponding to, 136probability distributions. See
distribution functions
problem of small denominators, 40propagation event (diagram), 121,
121
quantum leaps, of excitedatoms,138-39,139
quantum paradox, 48,130;andstatistical formulation ofquantum physics, 144;time paradoxand, 5, 48,138,156,157
quantum physics: as absolute,28-29;anthropomorphicfeatures of traditional, 151;basic
assumption of, 47;basicconclusions of, 149;basic problem of,133,140;chaos introduced in,4; classical mechanics limited by,
107;clinamen conceptintroduced into, 52;complementarity principle, 74, 150;Copenhagen interpretation, 50,150,156-57;as deterministic,136-37;discrete and continuous
spectrums in, 133,141;dualism
in, 50,53,130,131;eigenfunc-tions and eigenvalues in, 95,133;extending to unstable
systems, 89,131;Feynman on
nobody really understanding, 47;fundamental law of nature in,136;Hamiltonian in, 138;Heisenberg uncertainty
principle, 74, 135-36,143;as
incomplete, 130-31;instability in, 4,53-54,107;irreversibility in,
53,138;laws of nature in, 4,136,138,184;limits of validity
of, 44,54;meaning and scopedebated, 46-47,129-30;mi-crostates and macrostates, 24;Newtonian dynamics comparedwith, 129;nonlocality in, 130;observer's rolein, 5, 48-55,131,151;as operator calculus in
Hilbert space,95;operatorformalism in, 92,133,134;predictive successof, 46;probabilitiesin, 5, 51-52,54,132,149;realistic interpretation of, 54,131;reformulating, 46-55,129-51;resonances in, 53-54,146,148;state of a quantum system, 136,143;and thermodynamics, 131;time flow introduced in, 5;time-reversible processesin, 28;transition from individual to sta-
Index 223
tistical description in, 143\342\200\22446;
unidirectional time denied in, 2.Seealso quantum paradox; wave
functions
quantum vacuum, 175,179,180
radiation: black body, 131,145,174,176;solar, 18
radiation damping, 172radioactive decay, 17,18Rae,Alastair, 157random trajectories, 41random walk, 42,43,74realism: indeterminism as
compatible with, 132;naive realism ofclassical physics, 135;realistic
interpretation of quantum theory,54,131;realistic interpretationof reduction of wave function,130;time and change as crux of,14
recurrence:in baker
transformation, 99-101;Poincare'srecurrence theorem, 23,39,100-101
recurrence relation, 91-92reduction (collapse)of wave
function: with clinamen version ofquantum mechanics, 52;defined, 48;measurement problemleading to, 157;and quantum
paradox, 130;in realistic
interpretation of quantum
mechanics, 54relativity: classical mechanics' limits
shown by, 107;and infinite
velocities, 105;time for, 164-66;unidirectional time denied in, 2.Seealso general relativity; specialrelativity
resonances, 39-44;bubbles due to,123-24,124;in classical and
quantum physics, 129;constructive interference threatened by,
119;coupling creation anddestruction of correlations, 124;and diffusive motion, 42-44,43,126,155;dynamical groupsaffected by, 171;expressing in
terms of delta functions, 124n;flow of time rooted in, 159;fundamental role in physics, 42;leading to chaos, 112;leading toterms with dangerousdenominators, 40,124n;in LPS,141-42;and nonintegrability,
39-40,109,111;as nonlocal,42;in oscillator interaction with
field, 112;and perturbationalapproach, 141;in quantum
physics, 53-54,146,148;sounds
coupled by, 40;statistical
description affected by, 122-25;superpositions of plane wavesaffected by, 157-58;time
symmetry broken by, 41,44,146-47;trajectories influenced by, 41;intransition from wave functionsto ensembles, 131;and velocitydistribution over time, 79
retrodiction, 4reversible processes.Seetime-
reversible processesRiemannian geometry, 176rigged Hilbert space,96Rimini, Emanuele, 52Ritz-Rydberg principle, 132,145,
146,149Robertson, Howard, 177Rosenfeld, Leon,29,50,53Rubino, Carl, 186
scattering, 114\342\200\22415, 115;as not
representative of natural world,127;persistent scattering,148-49;typical experiments as
idealizations, 156
224 Index
scattering center, 115,115Schrodinger, Erwin, 63,134Schrodinger equation: complex
conjugate of, 137;asdeterministic and time reversible, 12,15,47, 137,146;formal solution of,142;as partial differential
equation, 136;probability amplitudesdescribedby, 137;and reductionof the wave function, 48,130;trajectories compared with, 47;wave function evolutiondescribed by, 15,47, 136
Schrodinger's perturbationalmethod, 140-41
Schuster, Peter, 71science:classical ideal of, 154;
creativity in, 188;democracy as
conflicting with, 17;Descarteson, 185;as dialogue with nature,57,60,153;Einstein on, 185;Freud on history of, 70;Hawking
on future of, 7; as not
monolithic, 17;philosophy as
separated from, 7, 14,72.Seealso biology; laws of nature;
physicsscientific laws. Seelaws of nature
Scott, Peter, 185Searle,JohnR.,14secondlaw of thermodynamics,
19;classical formulation of, 19,60;entropy associated with,
18-19,60;and irreversibility,
21;observer's ignorance as basis
of, 25-26;as probabilistic, 20,22;for systems that are not
isolated, 61self-organization: in dissipative
structures, 70;and distance from
equilibrium, 57; in nonequilib-rium physics, 3, 27;noninte-
grable systems required for, 39;
order maintained by, 71-72;technology compared with,
71-72;time orientation
required for, 128self-replicating biomolecules, 159semigroups, 104,142,171sensitivity to initial conditions, 30,
37,90Shimony, Abner, 53simultaneity, 168singular functions. Seegeneralized
functionssmall denominators, problem of,
40Smoluchowski, Roman, 23smoothness, statistical descriptions
requiring, 87\342\200\22488
Snow, C.P.,17solar radiation, 18space-time:arrow of time
consistent with, 172;conformalintervals in, 178;Friedmann
space-time interval, 177, 178;gravitation associated with
curvature of, 175-76;and matterfor Einstein, 176;Minkowski
space-time interval, 167,171,172,176,177,178;quantization
of, 178;Riemannian geometryas characterizing, 176;in specialrelativity, 167;transformed into
matter, 180spatialized time, 58-59,165-66,
171-72special relativity, 167-72;future
and past in, 168;and Hamilton-ian description, 170-72;spatialized time not implied by,
171-72;time dilation, 169,170;twin paradox, 169-70,111
spectral interval, 117spectral representation of an
operator: of evolution operator,
Index 225
125-26;of Liouville operator,113,125,149;ofPerron-Frobenius operator, 94\342\200\22495, 96,103,104
spectroscopy, 133,144,145spectrums, continuous and
discrete, 133,141Spinoza, Baruch, 16,176stability: instability at trajectory
level leading to at statistical level,87;stable dynamical systems,30-31,36,55;and structure in
nonequilibrium conditions, 63standard model (cosmology),164,
174-75,177state of a dynamical system, 31,32state ofa quantum system, 136,
143states of matter, 45,116statistical level of description:
breaking equivalence with
individual description, 35,83,87,89,94-95,96,103,106,155;for chaotic systems, 104;inclassical and quantum physics, 129;in classical dynamics, 108;fordeterministic chaos, 105;dynamical problems solved at,91-96,113,125-28;individual
level as equivalent to, 34\342\200\22435, 42,81,108;instability at, 155;interactions' effect on, 120-25;irreversibility given meaning at,81;for irreversible processes,81,108,109;of laws of nature, 162;of molecules in glass of water,78;operators required for, 133;replacing coordinates with wave
vectors, 121;resonances'effect
on, 122-25;in terms of Perron-Frobenius operator, 87-88;thermodynamics requiring, 127;trajectory description compared
with, 118-19;transition to in
quantum physics, 143-46;for
unstable systems, 109statistical mechanics, 46,92,95steady-state theory, 173-74;
difficulties with, 174;as unified with
big bang, 166Stengers, Isabelle, 154structure: arrow of time in
formation of, 71;and function, 62;innonequilibrium conditions, 63.Seealso dissipative structures
Sudarshan, George,142superimposition, 48,52superposition: of periodic
functions, 116;of plane waves,118-19,119,157-58;principleof, 149;of wave functions,138-39
symmetry breaking. Seetime
symmetry
synthetic chemistry, 71
Tagore, Rabindranath, 13Tarnas, Richard, 7
technology, self-organization
compared with, 71-72ternary correlations, 121test functions, 33n
thermal diffusion, 26,21,35,64thermodynamic branch: beyond
bifurcation point, 66,61;defined, 66;in pitchforkbifurcation, 68,69
thermodynamic equilibrium, 60,63,66,77, 113,130
thermodynamic limit, 45,116,126,158
thermodynamics: equilibriumstatistical thermodynamics, 77;equilibrium thermodynamics,88,108;first law of, 180;as
incompatible with trajectory
226 Index
thermodynamics (cont.)
description, 127;limiting to
equilibrium conditions, 61-62;nonequilibriumthermodynamics, 61,63,72;Poincare on laws
of, 55-56;and quantum theory,131;as scienceof the
irreversible, 17;traditional conflictwith dynamics eliminated, 104.Seealso entropy; secondlaw ofthermodynamics
thermodynamic systems, 115-16,148,158,184
three-body problem, 31,108time: as basic existential
dimension, 13;beginning of, 6,163-64;certainty and denial of,184;in classical science,59,60;communication requiringcommon, 150;constructive roleof,56;as crux of realism, 14;anddilemma of determinism, 1;as
emerging property, 60;entropictime, 161;as eternal, 166;ethicsassociated with, 58;freedomassociated with, 58;geologicaltime scale,161;as history
dependent, 170;as illusory, 1,58,165,187;imaginary time, 58,164,170;instability linked to,175;as literary theme, 187;inmaps, 81;of a meta-universe,164;in Newton's relation ofacceleration and force, 110;philosophers on, 58;philosophybecoming time centered, 13;physicists on, 58;as precedingexistence, 163,182;and realityas linked, 187;in relativity
theory, 164\342\200\22466; simultaneity, 168;spatialized time, 58-59,165-66,171-72;as universal in
Newtonian physics, 160,164,170.See
also arrow of time; evolution;flow of time; future; past; space-time; time paradox; time
symmetry
time dilation, 169,170time paradox: and baker
transformation, 103;defined, 2;formulation as creative act, 188;mind-matter interface at coreof,49;quantum paradox solved
along with, 5, 48,138,156,157time-reversible processes:as almost
inconceivable, 154;baker
transformation as, 101,103;inclassical and quantum mechanics, 28;classical physics as basis ofbeliefin, 107;and entropy, 18-19;irreversibility required for
studying, 49;irreversible processescompared with, 17-18;ofphysics versus time-centered
philosophy, 14;Schrodingerequation as, 137;time-reversibleworld as unknowable, 153;trajectory as, 21;wave function
satisfying time-reversible
equation, 5time symmetry: Bernoulli maps as
breaking, 90;breaking as global
property, 156;chaotic systemsbreaking, 105;complex spectralrepresentation as breaking, 126;dissipative structures breaking,69;instability breaking, 5,37-38;measuring deviceas
breaking, 54,150-51;non-Newtonian processesbreaking,108,124,129;probabilities as
time symmetric, 137;resonances
breaking, 41,44,146-47;semigroups breaking, 171
total energy, 76, 175Toulmin, Stephen, 184-85
Index 227
towns, 62trajectories: and Bernoulli maps,
83-88,85;collapse of, 109,119,126;as construct of planewaves, 119,157-58;and delta
functions, 33,143;ensembles
compared with, 82,83,87;as
idealizations, 37;as inadequatefor describing chaotic systems,90-91,105;laws of chaosparadoxical with, 104;limited
validity of, 44;in Lyapunov
exponent definition, 105;as not
smooth, 87\342\200\22488;as primitive,
108;and probability
distributions, 34;resonances' influence
on, 41;secondlaw ofthermodynamics and individual, 20;as
special solutions of Perron-Frobenius equation, 88;three-
body problem compared with,31;as time reversible, 21;wave
function compared with,
47
trajectory description. Seeindividual level of description
transient interactions: defined, 114;diffusive terms as negligible in,44;localized distributionfunctions for describing, 114,115;scattering as, 115,127;asunrepresentative of nature, 127
transition probabilities, 75
transport processes,18,105Tryon, Edward, 175Turing, Alan Mathison, 68,69twin paradox, 169-70,111;non-
relativistic analogue of, 160two-body motion, 102,156
\"ultraviolet\" divergences, 180-81uncertainty, resonances introducing
into classical mechanics, 44
unidirectional time. Seearrow oftime
unified theory, 182uniform distribution, 93universe, the. Seecosmologyunstable systems: contradiction
between reversible and irreversible
processesovercome in, 28;dynamics formulated at statistical
level for, 73-74;ensembles
deploying new properties for, 81;extending classical and quantummechanics to, 89;going beyondHilbert spacefor, 93,95-96,114;as probabilistic, 35,37;quantum theory of, 132;sensitivity
to initial conditions in, 37;statistical description of, 109;and unidirectional time, 2-3;unstable equilibrium, 30,30.Seealso instability
urn model, 74-76,75, 81
vacuum of correlations, 121,122,122,123,123
Valery, Paul, 106velocities, infinite, 105\342\200\2246
velocity inversion, 80velocity of light, 105-6,167-68velocity-reversal paradox, 21,23viscosity, 3, 17-18von Neumann, John, 49,51,53,
130
Wahl, Jean, 11Walker, Arthur, 177wave functions: as deterministic
and probabilistic, 15;deterministic, time-reversible equationsatisfied by, 5; evolution as
superposition of oscillating terms,138-39;extending beyondHilbert space,144-45;mixture
228 Index
wave functions (cont.)of, 47,48,144;operatorformalism required for descriptionof, 133;physical interpretationof, 47;and probability
amplitudes, 136;resonancesintroduced in terms of, 53-54;stateof quantum system describedby,
136,143;trajectory comparedwith, 47; transition to
ensembles,131.Seealso reduction
(collapse)of wave function;
Schrodinger equation;superposition
wave vectors: defined, 116;andFourier transform, 117;andinteractions' effect on statistical
description, 120-25;law ofconservation of, 121;quantummechanical counterpart for, 147,148;replacing coordinates
with, 121;in statistical
description of free particle motion,
118;vanishing wave vector,124n
Weber, Tullio, 52Weinberg, Steven, 15,48Weyl, Hermann, 84Wheeler, John Archibald, 164Whitehead, Alfred North:
compromise between scienceand
freedom, 10;on creativity ofnature, 62;on existence as
process,59;on intelligibility ofnature, 17,29;and nonequilib-rium thermodynamics, 72;philosophy as time centered, 13;on
slipping through the scientific
net, 189Wigner, Eugene, 49Wilson, Robert, 174
X mysteries, 130
Zermelo, Ernst, 23Z mysteries, 130
/p
ALSOBY ILYA PRIGOGINE
Order Out of Chaos
From Being to Becoming
Exploring Complexity
aaH.s
4/\"
'\302\253
^
oooat
_
J