PDF From random walks to random matrices - selected topics in modern theoretical physics 1st edition

Page 1


Zinn-Justin

Visit to download the full and correct content document: https://ebookmass.com/product/from-random-walks-to-random-matrices-selected-topi cs-in-modern-theoretical-physics-1st-edition-jean-zinn-justin/

More products digital (pdf, epub, mobi) instant download maybe you interests ...

Random in Death 1st Edition J. D. Robb

https://ebookmass.com/product/random-in-death-1st-edition-j-drobb/

Quantum Field Theory and Critical Phenomena 5th Edition

Jean Zinn-Justin

https://ebookmass.com/product/quantum-field-theory-and-criticalphenomena-5th-edition-jean-zinn-justin/

Stochastic processes and random matrices : lecture notes of the Les Houches Summer School First Edition. Edition Altland

https://ebookmass.com/product/stochastic-processes-and-randommatrices-lecture-notes-of-the-les-houches-summer-school-firstedition-edition-altland/

Schaum's Outline of Probability, Random Variables, and Random Processes, Fourth Edition Hwei P. Hsu

https://ebookmass.com/product/schaums-outline-of-probabilityrandom-variables-and-random-processes-fourth-edition-hwei-p-hsu/

Random Process Analysis With R Marco Bittelli

https://ebookmass.com/product/random-process-analysis-with-rmarco-bittelli/

Spatiotemporal Random Fields, Second Edition: Theory and Applications George Christakos

https://ebookmass.com/product/spatiotemporal-random-fieldssecond-edition-theory-and-applications-george-christakos/

Advances in Semiconductor Technologies: Selected Topics Beyond Conventional CMOS An Chen

https://ebookmass.com/product/advances-in-semiconductortechnologies-selected-topics-beyond-conventional-cmos-an-chen/

Special Functions

of

Fractional Calculus:

Applications to Diffusion and Random Search Processes Trifce Sandev

https://ebookmass.com/product/special-functions-of-fractionalcalculus-applications-to-diffusion-and-random-search-processestrifce-sandev/

Weather, macroweather, and the climate : our random yet predictable atmosphere Lovejoy

https://ebookmass.com/product/weather-macroweather-and-theclimate-our-random-yet-predictable-atmosphere-lovejoy/

FROMRANDOMWALKSTORANDOMMATRICES

FROMRANDOMWALKSTORANDOMMATRICES

SelectedTopicsinModernTheoreticalPhysics

IRFU/CEA,Paris-SaclayUniversity and FrenchAcademyofSciences

GreatClarendonStreet,Oxford,OX26DP, UnitedKingdom

OxfordUniversityPressisadepartmentoftheUniversityofOxford. ItfurtherstheUniversity’sobjectiveofexcellenceinresearch,scholarship, andeducationbypublishingworldwide.Oxfordisaregisteredtrademarkof OxfordUniversityPressintheUKandincertainothercountries

c JeanZinn-Justin2019

Themoralrightsoftheauthorhavebeenasserted FirstEditionpublishedin2019 Impression:1

Allrightsreserved.Nopartofthispublicationmaybereproduced,storedin aretrievalsystem,ortransmitted,inanyformorbyanymeans,withoutthe priorpermissioninwritingofOxfordUniversityPress,orasexpresslypermitted bylaw,bylicenceorundertermsagreedwiththeappropriatereprographics rightsorganization.Enquiriesconcerningreproductionoutsidethescopeofthe aboveshouldbesenttotheRightsDepartment,OxfordUniversityPress,atthe addressabove

Youmustnotcirculatethisworkinanyotherform andyoumustimposethissameconditiononanyacquirer

PublishedintheUnitedStatesofAmericabyOxfordUniversityPress 198MadisonAvenue,NewYork,NY10016,UnitedStatesofAmerica

BritishLibraryCataloguinginPublicationData Dataavailable

LibraryofCongressControlNumber:2019930266

ISBN978–0–19–878775–4 DOI:10.1093/oso/9780198787754.001.0001

Printedandboundby CPIGroup(UK)Ltd,Croydon,CR04YY

Preface

Thisworkgathersaselectionofseminarsandmini-coursesdelivered duringthelast decadeinanumberofcountries.Inaddition,fourarticlesfromScholarpediahave beenincluded,innewlyeditedform.

Asaconsequence,thevariouschapterstrytobeasself-containedaspossible, eventhoughthetopicshavebeenselectedtoinvolvesomecommonthemes,which giveageneralconsistencytothewholework.

Thisimpliessomeredundancy,whichwehavetriedtominimize.

Themainthemesaretherenormalizationgroup(RG),fixedpoints,universality andthecontinuumlimit,whichopenandconcludethework.Otherimportant andrelatedthemesarepathandfieldintegrals,thenotionofeffectivequantum orstatisticalfieldtheories,gaugetheories,whichisthemathematicalstructureat thebasisoftheinteractionsinfundamentalparticlephysics,includingquantizationproblemsandanomalies,stochasticdynamicalequationsandsummationof perturbativeseries.

Onpurpose,inthechaptersderivedfromseminars,wehavetriedtokeepthe technicallevelaslowasreasonablypossible,consideringthetopics

Somebasichistoricalbackgroundconcerningtheconstructionof quantumfield theory(QFT),theconstructionoftheStandardModelofinteractionsatthemicroscopicscaleandtheemergenceofRGisprovided.

Chapter1discussestheasymptoticpropertiesatlargetimeandspaceofthe familiarexampleoftherandomwalk.Theemphasishereison universality,the continuumlimit, Gaussiandistributionandscaling.Thesepropertiesarefirstderivedfromanexactsolutionandthenrecoveredby RG methods.Thismakesthe introductionofthewholeRGterminologypossible.

Chapter2islargelydescriptiveandintroducesthenotionoffunctional(pathand field)integrals,forbosonaswellasfermionsystems(basedonthe formalismof Grassmannintegration),astheyareusedinphysics.

Chapter3illustrates,usinganumberofexamples,theessentialroleoffunctional integralsinphysics:thepathintegralrepresentationofquantum mechanicsexplains whymanybasicequationsofclassicalphysicssatisfyavariationalprinciple,the relationbetweenQFTandthetheoryofcriticalphenomenainmacroscopicphase transitionsandsoon.

Chapter4describesafewimportantstepswhichhaveledfromthediscoveryof infinitiesinquantumelectrodynamicstotheconceptofrenormalizationandRG. RGtodayplaysanessentialroleinourunderstandingofthepropertiesofQFTand ofcontinuousmacroscopicphasetransitions.

InChapter5,wefirstrecalltheimportanceoftheconceptofscaledecouplingin physics.WethenemphasizethatQFTandthetheoryofcriticalphenomenahave providedtwoexampleswherethisconceptfails.

Todealwithsuchasituation,newtoolshavebeeninventedbasedon theconcept ofRG.

Chapter6describeshowtheperturbativeRGofQFThasmadeitpossible toderiveuniversalpropertiesofcontinuousmacroscopicphasetransitionsandto calculateuniversalquantitiesusingdimensionalcontinuationandWilson–Fisher’s ε =4 d expansion.Seriessummationmethodsarethenrequiredtodetermine precisevaluesofcriticalexponents.

Chapter7isdevotedtoadiscussionoftheRGflowwhentheeffective φ4 fieldtheorythatdescribestheuniversalpropertiesofcriticalphenomenadependsonseveral couplingconstants.Someexamplesillustratethenotionof emergentsymmetry

Chapter8ismoretechnical:wetrytoexplainthenotionofeffectivefieldtheory, illustratingthetopicwithexamples,andstressthatallQFTsappliedtoparticleor statisticalphysicsareeffective(i.e., notfundamental)lowenergyorlargedistance theories.

ManyRGresultsarederivedwithintheframeworkoftheperturbativeRG.However,thisRGistheasymptoticforminsomeneighbourhoodofaGaussianfixed pointofthemoregeneralandexactRG,asintroducedbyWilsonand Wegner.In Chapter9,wedescribethecorrespondingRGequationsandgivesomeindications abouttheirderivation.

Inthecaseofthespontaneousbreakingofacontinuoussymmetry,thecorrelation lengthdivergesatalltemperaturesintheorderedphase.Physicalsystemsthen displayuniversalpropertiesforalltemperaturesbelow Tc.Thisphenomenonis aconsequenceoftheexistenceofGoldstonemodes(masslessbosons,intheQFT terminology).InChapter10,weconsidertheexampleofthe O(N )symmetryand showthatanRGcanbeconstructedanduniversalpropertiesderivedatandnear dimension2intheframeworkofan ε = d 2expansion.

Chapter11isthefirstoffourchaptersthatdiscussvariousissuesconnectedwith theStandardModeloffundamentalinteractionsatthemicroscopicscale.

InChapter11,wediscussthenotionofgaugeinvariance,firstAbelianandthen non-Abelian,thebasicgeometricstructurethatgeneratesinteractions.

InChapter12,wedescribethemainstepsintheconstructionoftheelectroweak componentoftheStandardModel.TherecentdiscoveryofthelastpredictedparticleoftheStandardModel,theHiggsparticle(whichwassearchedforanumberof years),hasbeenanadditionalandmajorconfirmationofthevalidityofthemodel.

Chapter13isdevotedtosomeaspectsofquantumchromodynamics(QCD),the partoftheStandardModelresponsibleforstronginteractions. Asanintroduction, thegeometryofnon-Abeliangaugetheories,basedonparalleltransport,isrecalled. Thisgeometricpointofviewleadsnaturallytotheconstructionoflatticegauge theories,whichcanbestudiedbynumericalsimulations.

Chapter14describestheimportantBRSTsymmetryofquantizedgaugetheories. Theexplicitrealizationofthesymmetryisnotstableunderrenormalization.By contrast,somequadraticequationssatisfiedbythegeneratingfunctionalofvertex functions(the1PI(one-particle-irreducible)correlationfunctions),theZinn-Justin equation,isstableandisatthebasisofageneralproofoftherenormalizabilityof non-Abeliangaugetheories.

Chapter15isdevotedtothefollowingquestion:eventhoughQFTsareonly effectivelowenergyorlargedistancetheories,underwhichconditionsisitpossible todefinerenormalizedQFTsconsistentonallscales?Onthebasisof examples, ithasbeensuggestedthatanecessaryand,perhaps,sufficientconditionisthe existenceofultravioletfixedpoints,apropertycalled‘asymptoticsafety’.

Chapter16dealswiththeimportantproblemofquantizationwithsymmetries, thatis,howtoimplementsymmetriesoftheclassicalactioninthecorresponding quantumtheory.Theproposedsolutionsarebasedonregularizationmethods. Obstructionsencounteredinthecaseofchiraltheoriesareemphasized.

InChapter17,weexhibitvariousexplicitexamples,whereclassicalsymmetries cannotbetransferredtoquantumtheories.Theobstructionsarecharacterized byanomalies.Theexamplesinvolvechiralsymmetriescombinedwithcurrentsor gaugesymmetries.Inparticular,anomaliesleadtoobstructionsintheconstruction oftheoriesandhave,therefore,constrainedthestructureof theStandardModel. Otherapplicationsarealsodescribed.

Chapter18dealswithafewexampleswheretheclassicalactionhasaninfinite numberofdegenerateminimabut,inquantumtheory,thisdegeneracyislifted bybarrierpenetrationeffects.Technically,thiscorrespondstotheexistenceof instantons,solutionstoclassicalequationsinimaginarytime.Inthe caseofQCD, thisleadstothefamousstrongCPviolationproblem.

ComputersimulationsofcriticalstatisticalsystemsorQFTmodelsareperformed withsystemswheresizesarefinite.Intransfermatrixcalculations,allsizesbutone arealsofinite.Insystemswherethecorrelationlengthislarge,itisthusimportant tounderstandhowtheinfinitesizelimitisreached.Thisproblemisinvestigatedin Chapter19.

AfterthediscoveryofthepredictedBose–Einsteincondensation,whichisapropertyoffreebosons,aninterestingissuewastheeffectsofweakrepulsiveinteractions. InChapter20,itisshownthattheBose–Einsteincondensationisthenreplacedby asuperfluidphasetransition.Thesetheoreticalconsiderationsareillustratedbyan evaluationofthevariationofthetransitiontemperatureatweakcoupling.

Chapter21isdevotedtoastudyofQFTatfinitetemperature,atopicthat isrelevanttoheavyionhighenergycollisions.Aninterestingissueconcernsthe conditionsfordimensionalreduction,thatis,whencantheinitialfieldtheorybe replacedbyaneffectivefieldtheorywithouttimedimension?

InChapter22,westudystochasticdynamicalequations,generalizedLangevin equations,whichdescribeawiderangeofphenomenafromBrownian motionto criticaldynamicsinphasetransitions.Inthelattercase,dynamicRGequations canbederived.Dynamicscalingfollows,withascalingtimethatexhibits critical slowingdown.

InQFT,allperturbativeexpansionsaredivergentseries,inthemathematical sense.Thisleadstoadifficultywhentheexpansionparameterisnotsmall.In thecaseofBorelsummableseries,anumberoftechniqueshavebeendeveloped toderiveconvergentsequencesfromdivergentseries.Thisisthe centraltopicof Chapter23.

InanumberofquantummechanicsandQFTexamplesexhibitingdegenerate classicalminimarelatedbyquantumtunnelling,theperturbativeexpansionisnot Borelsummableandtheperturbationseriesdonotdefineuniquefunctions.An importantissueisthenwhatkindofadditionalinformationisrequired todetermine theexactexpandedfunctions.

WhiletheQFTexamplesarecomplicatedandtheirstudyisstillatthepreliminarystage,inquantummechanics,inthecaseofsomeanalyticpotentialsthathave degenerateminima(likethequarticdouble-wellpotential),theproblemhasbeen completelysolved.SomeexamplesaredescribedinChapter24.

Whilethestudyofthestatisticalpropertiesofrandommatricesof largesize hasalonghistorygoingbacktoWigner,inthemid-1980sitwasrealized that someensemblesofrandommatricesinthelargesizeandso-calleddoublescaling limitcouldbeusedastoymodelsfortwo-dimensionalquantumgravity coupledto conformalmatterandstringtheory,orasexamplesofcriticalstatisticalmodelson sometypeofrandomsurfaces.

Atremendousdevelopmentofrandommatrixtheoryfollowed,using increasinglysophisticatedmathematicalmethods,andnumberofmatrixmodelshave beensolvedexactly.However,thesomewhatparadoxicalsituationisthateither themodelscanbesolvedexactlyorlittlecanbesaidaboutthem.Therefore,since thesolvedmodelsexhibitcriticalpointsanduniversalproperties,it istemptingto useRGideastodetermineuniversalproperties,withoutsolvingmodelsexplicitly.

Somenon-trivialprogresshasbeenachievedalongtheselines,whicharereported inChapter25,butnosystematicmethodhasbeendiscoveredtogo beyondthe simplestapproximationschemes.

ThissituationagainillustratesthefactthatRGisafundamentalidea butnot amethod.Inorderforittobeefficient,itisnecessarytofindaproperwayto implementit.

Acknowledgements. ManythankstoElisabethFarrellforhercarefulproofreading ofthemanuscript.

Saclay,18April2019

Contents

1Therandomwalk:Universalityandcontinuumlimit .........1

1.1Randomwalkinvariantunderspaceanddiscretetimetranslations ..1

1.2Fourierrepresentation......................2

1.3Randomwalk:Asymptoticbehaviourfromadirectcalculation.... 3

1.4Correctionstocontinuumlimit..................5

1.5Randomwalk:Fixedpointsoftransformationsanduniversality.. ..6

1.6Localandglobalstabilityoffixedpoints..............8

1.7Brownianmotionandpathintegral................11

2Functionalintegration:Frompathtofieldintegrals ..........13

2.1Randomwalk,Brownianmotionandpathintegral.........14

2.2TheWienermeasureandstatisticalphysics............18

2.3Generalization.........................19

2.4Gaussianpathintegrals:Thequantumharmonicoscillator...... 23

2.5Pathintegrals:Perturbationtheory................24

2.6Pathintegral:Quantumtimeevolution..............26

2.7Barrierpenetrationinthesemi-classicallimit............27

2.8Pathintegrals:Afewgeneralizations...............28

2.9Pathintegralsforbosonsandfermions...............30

2.10Fieldintegrals:Newissues....................32

3Theessentialroleoffunctionalintegralsinmodernphysics ......35

3.1Classicalphysics:Themysteryofthevariationalprinciple..... .35

3.2Quantumevolution:FromHamiltoniantoLagrangianformalism...38

3.3Fromquantumevolutiontostatisticalphysics...........41

3.4Statisticalmodelsatcriticalityandquantumfieldtheory....... 43

3.5Barrierpenetration,vacuuminstability:Instantoncalculus... ...44

3.6LargeorderbehaviourandBorelsummability:Criticalexponents...45

3.7Quantizationofgaugetheories..................47

3.8Numericalsimulationsinquantumfieldtheory...........50

3.9Quantizationofthenon-linear σ-model..............51

3.10 N -componentfields:Large N techniques..............52

4Frominfinitiesinquantumelectrodynamicstothegeneral renormalizationgroup .......................53

4.1QFT,RG:Somemajorsteps...................54

4.2QEDandtheproblemofinfinities.................55

4.3Therenormalizationstrategy...................58

4.4Thenatureofdivergencesandthemeaningofrenormalization.. ..59

4.5QFTandRG..........................60

4.6Criticalphenomena:Otherinfinities................61

4.7Thefailureofscaledecoupling:TheRGidea............62

4.8Phasetransitions:ExactRGinthecontinuum...........63

4.9Effectivefieldtheory:Fromcriticalphenomenatoparticlephysics...66

5RenormalizationGroup:Fromageneralconcepttonumbers .....69

5.1Scaledecouplinginphysics:Abasicparadigm...........70

5.2Fundamentalmicroscopicinteractions...............71

5.3Macroscopicphasetransitions..................72

5.4Fixedpoints:TheQFTframework................75

5.5RG,correlationfunctionsandscalingrelations...........77

5.6Exponents:PracticalQFTcalculations..............78

5.7Resultsforthree-dimensionalcriticalexponents.......... .79

6Criticalphenomena:Thefieldtheoryapproach ............81

6.1UniversalityandRG......................82

6.2RGinthecontinuum:Abstractformulation............84

6.3Effectivefieldtheory......................85

6.4TheGaussianfieldtheory....................87

6.5GaussianfixedpointandGaussianrenormalization.........89

6.6Statisticalscalarfieldtheory:Perturbationtheory........ .90

6.7Dimensionalcontinuationandregularization............91

6.8PerturbativeRG........................92

6.9RGequations:Solutions.....................96

6.10Wilson–Fisher’sfixedpoint: ε-Expansion.............97

6.11Criticalexponentsas ε-expansions................99

6.12Three-dimensionalexponents:Summationofthe ε-expansion....100

7Stabilityofrenormalizationgroupfixedpointsanddecay ofcorrelations ..........................101

7.1Modelswithonlyonecorrelationlength.............101

7.2Cubicanisotropy,amodelwithtwocouplings..........103

7.3GeneralquarticHamiltonian:RGfunctions...........106

7.4Runningcouplingconstantsandgradientflows..........107

7.5Fixedpointstabilityandvalueofthepotential..........108

7.6Fixedpointstabilityandfielddimension.............110

8Quantumfieldtheory:Aneffectivetheory .............111

8.1Effectivelocalfieldtheory:Thescalarfield............112

8.2PerturbativeassumptionandGaussianrenormalization......113

8.3Fundamentalinteractionsatthemicroscopicscale.........120

8.4Fieldtheorywithalargemass:Anexplicittoymodel.......121

8.5Aneffectivefieldtheory:TheGross–Neveumodel.........124

8.6Non-linear σ-model:Anothereffectivefieldtheory.........130

9Thenon-perturbativerenormalizationgroup ............137

9.1IntuitiveRGformulation....................137

9.2Non-perturbativeRGequations.................139

9.3Partialfieldintegration:Someidentities.............142

9.4Partialfieldintegrationindifferentialform............143

10O(N)vectormodelintheorderedphase:Goldstonemodes .....145

10.1Classicallatticespinmodelandregularizednon-linear σ-model...146

10.2Perturbativeorlowtemperatureexpansion............149

10.3ZeromomentumorIRdivergences...............153

10.4Formalcontinuumlimit:Thenon-linear σ-model.........154

10.5Thecontinuumtheory:Regularization..............157

10.6Symmetryandrenormalization.................158

10.7Correlationfunctionsindimension d =2+ ε atoneloop......162

10.8RGequations.........................165

10.9ZerosoftheRG β-function:Fixedpoints............166

10.10Correlationfunctions:Scalingformbelow Tc ...........168

10.11Linearformulation......................170

10.12Twodimensions.......................174

11Gaugeinvarianceandgaugefixing ................177

11.1Gaugeinvariance:Afewhistoricalremarks............177

11.2Variationalprinciple,chargedparticleandgaugeinvariance... .178

11.3Gaugeinvariance:Achargedquantumparticle..........181

11.4Evolutionofachargedparticle:Pathintegralrepresentation ....185

11.5ClassicalelectromagnetismandMaxwell’sequations........185

11.6Gaugefixinginclassicalgaugetheories.............187

11.7QED............................188

11.8Non-Abeliangaugetheories..................190

11.9Quantizationofnon-Abeliangaugetheories:Gaugefixing..... 193

11.10GeneralRelativity......................194

12TheHiggsboson:Amajordiscoveryandaproblem ........195

12.1Perturbativequantumfieldtheory:Theconstruction...... .195

12.2Spontaneoussymmetrybreaking................196

12.3Non-Abeliangaugetheories..................198

12.4TheclassicalAbelianLandau–Ginzburg–Higgsmechanism.....199

12.5Abelianandnon-AbelianHiggsmechanism............200

12.6Non-Abeliangaugetheories:Quantizationandrenormalization. ..201

12.7Theself-coupledHiggsfield:AsimpleRGanalysis........202

12.8TheGross–Neveu–Yukawamodel:AHiggs–toptoymodel.....204

12.9GNYmodel:ThegeneralRGflowatoneloop..........207

12.10Thefinetuningissue.....................208

13Quantumchromodynamics:Anon-Abeliangaugetheory ......209

13.1Geometryofgaugetheories:Paralleltransport.......... 210

13.2Gaugeinvariantaction.....................212

13.3Hamiltonianformalism.Quantizationinthetemporalgauge....214

13.4Perturbationtheory,regularization...............219

13.5QCD:Renormalizationgroup..................220

13.6Anomalies:Generalremarks..................221

13.7QCD:Thesemi-classicalvacuumandinstantons.........222

13.8Latticegaugetheories:Generalities...............225

13.9Purelatticegaugetheory...................226

13.10Wilsonloopandtheconfinementproperty............229

13.11Fermionsonthelattice.Chiralsymmetry............235

14FromBRSTsymmetrytotheZinn-Justinequation .........237

14.1Non-Abeliangaugetheories:Classicalfieldtheory.........238

14.2Non-Abeliangaugetheories:Thequantizedaction........239

14.3BRSTsymmetryofthequantizedaction.............241

14.4TheZJequationandremormalization..............244

14.5TheZJequation:Afewgeneralproperties............247

14.6BRSTsymmetry:Thealgebraicorigin..............250

15Quantumfieldtheory:Asymptoticsafety

15.2Super-renormalizableeffectivefieldtheories:The(φ2)2 example..255 15.3Arenormalizablefieldtheory:The(φ2)2 theoryindimension4...258

15.4Thenon-linear σ-model....................260

15.5TheGross–Neveumodel....................262

15.6QCD............................263

15.7Generalinteractionsandsummary...............264

16Symmetries:Fromclassicaltoquantumfieldtheories ........265

16.1Symmetriesandregularization.................265

16.2Higherderivativesandmomentumcut-offregularization...... 268

16.3Regulatorfields........................271

16.4Abeliangaugetheory,thetheoreticalframeworkofQED.... ..273

16.5Non-Abeliangaugetheories..................276

16.6Dimensionalregularizationandchiralsymmetry.........278

16.7Latticeregularization.....................280

17Quantumanomalies:Afewphysicsapplications ..........283

17.1ElectromagneticdecayoftheneutralpionandAbeliananomaly. .284

17.2Atwo-dimensionalillustration:TheSchwingermodel.......291

17.3Abelianaxialcurrentandnon-Abeliangaugefields........298

17.4Non-Abeliananomalyandchiralgaugetheories..........301

17.5Weakandelectromagneticinteractions:Anomalycancellation.. .303

17.6Wess–Zuminoconsistencyconditions..............305

17.7Latticefermions:Ginsparg–Wilsonrelation............307

17.8Supersymmetricquantummechanicsanddomainwallfermions.. .313

18Periodicsemi-classicalvacuum,instantonsandanomalies ......319

18.1Theperiodiccosinepotential..................319

18.2Instantons,anomaliesand θ-vacua: CP N 1 models........322

18.3Non-Abeliangaugetheories:Instantonsandanomalies...... .328

18.4Thesemi-classicalvacuumandthestrongCPviolation......332

18.5Fermionsinaninstantonbackground:The U (1)problem.....332

19Fieldtheoryinafinitegeometry:Finitesizescaling ........335

19.1Periodicboundaryconditionsandtheproblemofthezeromode. .335

19.2Cylindricalgeometry:Two-dimensionalfieldtheory........338

19.3Effective(φ2)2 fieldtheoryatcriticalityinfinitegeometries....343

19.4Momentumquantizationinfinitegeometries...........346

19.5The(φ2)2 fieldtheoryinaperiodichypercube..........347

19.6The(φ2)2 fieldtheory:Cylindricalgeometry...........353

19.7Continuoussymmetries:Finitesizeeffectsatlowtemperature. ..358

20TheweaklyinteractingBosegasatthecriticaltemperature .....361 20.1Bosegas:Fieldintegralformulation...............361

20.2Independentbosons:Bose–Einsteincondensation........ .363

20.3TheweaklyinteractingBosegasandtheHeliumphasetransition. .364

20.4RGanduniversality......................365

20.5Theshiftofthecriticaltemperatureforweakinteraction... ...368

21Quantumfieldtheoryatfinitetemperature ............373

21.1FinitetemperatureQFT:Generalconsiderations.........374

21.2Scalarfieldtheory:Effectivetheoryforthezeromode...... .376

21.3The(φ2)2 1,d scalarQFT:Phasetransitions............379

21.4Temperatureeffects:Thetemperature-dependentmass.. ....380

21.5Phasestructureatfinitetemperatureatoneloop......... 381

21.6RGatfinitetemperature....................384

21.7Effectiveaction:Perturbativecalculation............386

21.8Effectiveaction: ϕ-Expansion.................388

21.9The(φ2)2 fieldtheoryatfinitetemperatureinthelarge N limit..389

21.10Thenon-linear σ-modelatfinitetemperatureforlarge N .....392

21.11TheGNmodelatfinitetemperatureforlarge N .........399

21.12Abeliangaugetheories:TheQEDexample............408 A21Appendix:One-loopcontributions ................415 A21.1Γand ζ functions.......................415 A21.2Theone-looptwo-pointcontributionat T =0..........416 A21.3Thethermalcorrectionsatoneloop...............416

22Fromrandomwalktocriticaldynamics ..............421 22.1Randomwalkwithgradientdrivingforce............422

22.2Anelementaryexample:Thelineardrivingforce.........423

22.3TheFokker–Planckformalism.................427

22.4Pathintegralrepresentation..................432

22.5ThedissipativeLangevinequation:Supersymmetricformulation ..435

22.6Criticaldynamics:TheLangevinequationinfieldtheory.....439

22.7Time-dependentcorrelationfunctionsanddynamicaction.... .442

22.8ThedissipativeLangevinequationandsupersymmetry......444

22.9RenormalizationofthedissipativeLangevinequation.......446

22.10DissipativeLangevinequation:RGequationsin4 ε dimensions..446

23Fieldtheory:Perturbativeexpansionandsummationmethods ....451

23.1Divergentseriesinquantumfieldtheory.............451

23.2Anexample:Theperturbative(φ2)2 fieldtheory.........453

23.3Renormalizedperturbationtheory:Callan–Symanzikequations ...455

23.4Summationmethodsandcriticalexponents...........457

23.5ODMsummationmethod...................461

23.6Application:Thesimpleintegral d =0..............465

23.7Thequarticanharmonicoscillator: d =1.............468

23.8 φ4 fieldtheoryin d =3dimensions...............469

24Hyper-asymptoticexpansionsandinstantons ............471

24.1DivergentseriesandBorelsummability.............472

24.2Perturbativeexpansionandpathintegral............474

24.3Thequarticanharmonicoscillator:ABorelsummableexample... 475

24.4Thedouble-wellpotential:GeneralizedBohr–Sommerfeld quantizationformulae.....................478

24.5Instantonsandmulti-instantons.................484

24.6PerturbativeandexactWKBexpansions.............488

24.7Otheranalyticpotentials:Afewexamples............490

25Renormalizationgroupapproachtomatrixmodels .........493

25.1One-Hermitianmatrixmodelsandrandomsurfaces:Asummary. .494

25.2Continuumanddoublescalinglimits..............495

25.3TheRGapproach.......................496

1Therandomwalk:Universalityandcontinuumlimit

The universalityofalargescalebehaviour and,correspondingly,theexistenceofa macroscopic continuumlimit,emergeascollectivepropertiesofsystemsinvolving a largenumberofrandomvariables whoseindividualdistributionissufficiently localized.

Theseproperties,aswellastheappearanceofanasymptotic Gaussiandistributionwhentherandomvariablesarestatisticallyindependent,areillustratedhere withthesimpleexampleofthe randomwalk withdiscretetimesteps.

Wefirstrecallhowtheasymptoticlargetime,largespacebehaviour canbederived andemphasizeits universalproperties

Wethentakea renormalizationgroup (RG)viewpoint.InspiredbyRGideas,we introducetransformations,actingonthetransitionprobability,whichdecreasethe numberoftimesteps[1].Weshowthat Gaussiandistributionsareattractivefixed points forthesetransformations. Thecontinuumasymptoticlimitwithuniversal scalingproperties isthenrecovered.

Thepropertiesofthe continuumlimit canthenbedescribedbya pathintegral.

1.1Randomwalkinvariantunderspaceanddiscretetimetranslations

Weconsiderastochasticprocess,arandomwalk,indiscretetimes, firstonthereal axisandthen,briefly,onthelatticeofpointswithintegercoordinates.

Therandomwalkisspecifiedby:

aninitialprobabilitydistribution P0(q) (q beingaposition)attime n =0, aprobabilitydensity ρ(q,q′) ≥ 0 forthetransitionfromthepoint q′ tothepoint q, whichweassume independentofthe(integer)time n. TheseconditionsdefineaMarkovchain,a Markovianprocess,inthesensethat thedisplacementattime n dependsonlyonthepositionattime n,butnotonthe positionsatpriortimes,homogeneousorstationary,thatis, invariantundertime translation,uptotheboundarycondition.

1.1.1Translationinvariantrandomwalkincontinuumspace

Probabilityconservationimplies

qρ(q,q ′)=1 . (1.1)

Theprobabilitydistribution Pn(q)forawalkertobeatpoint q attime n satisfies theevolutionequation Pn+1(q)= dq ′ ρ(q,q ′)Pn(q ′), dqP0(q)=1

Equation(1.1)thenimplies dqPn(q)=1.

Therandomwalk:Universalityandcontinuumlimit 1.2

Translationinvariance. Wehavealreadyassumed ρ independentof n and,thus, thetransitionprobabilityis invariantundertimetranslation.

Inaddition,wenowassumethatthetransitionprobabilityisalso invariantunder spacetranslations and,thus,

(q,q ′) ≡ ρ(q q ′)

Asaconsequence,theevolutionequationtakestheformofthe convolutionequation,

whichalsoappearsinthediscussionofthecentrallimittheoremofprobabilities.

Localrandomwalk. Weconsideronlytransitionfunctionspiecewisedifferentiable andwithboundedvariation.Wefurtherassumethatthetransition probability ρ(q) satisfiesaboundofexponentialform,

(q) ≤ M e A|q|,M,A> 0 ,

apropertyof exponentialdecay thatwecall locality.Qualitatively, largedisplacementshaveaverysmallprobability.

Generalizationto Rd Thegeneralizationtoatranslationinvariantwalkin Rd is simple.Inparticular,inthecaseofspacerotationsymmetry,themodificationof theevolutionequationisstraightforward.

1.2Fourierrepresentation

TheevolutionequationsimplifiesafterFouriertransformation.Wethusintroduce

whichisalsoa generatingfunctionofthemomentsofthedistribution Pn(q).

Therealityof Pn(q)andthenormalizationofthetotalprobabilityimply

Similarly,weintroduce

∗ n (k)= Pn( k), Pn(k =0)=1

(k)= dq

ikq ρ(q), whichisalsoa generatingfunctionofthemomentsofthedistribution ρ(q): qr = dqρ(q)qr .

Finally,theexponentialdecayconditionimpliesthat thefunction ρ(k) isanalytic inthestrip | Im k| <A and,thus,hasaconvergentexpansionat k =0.

Theevolutionequationthenbecomes Pn+1(k)=˜ ρ(k)Pn(k).

Toslightlysimplifytheanalysis,wetakeasaninitialdistribution P0(q)= δ(q), where δ isDirac’sdistribution(thewalkeratinitialtimeisat q =0withprobability 1).Withthischoiceofinitialconditions, ˜ P0(k)=1and,thus, Pn(k)=˜ ρn(k).

1.3

Therandomwalk:Universalityandcontinuumlimit 3

1.2.1Generatingfunctionofcumulants

Weintroducethegeneratingfunctionofthecumulantsof ρ(q),

)=ln˜

Then,

Theregularityof˜ ρ(k)andthecondition˜ ρ(0)=1implythat w(k)hasaregular expansionat k =0oftheform

where wr isthe rthcumulant,forexample,

1.3Randomwalk:Asymptoticbehaviourfromadirectcalculation

Withthehypothesessatisfiedby P0 and ρ,thedeterminationoftheasymptotic behaviourfor n →∞ followsfromargumentsidenticaltothoseleadingto the centrallimittheoremofprobabilities

For n →∞ and w1 =0, w(k)isdominatedbythefirsttermand,thus,

Therandomvariable q/n convergeswithprobability1towardsitsexpectationvalue w1 (themeanvelocity).

For n →∞ and w1 =0, w(k)isdominatedbythetermoforder k2 and,thus,

Thenitistherandomvariable q/√n thathasasitslimitingdistribution,aGaussian distributionwithwidth √w2

Therandomvariablethatcharacterizesthedeviationwithrespect tothemean trajectory,

and,thus, X =0,has,aslimitingdistribution,the universalGaussiandistribution Ln(X)= √nPn(nw1 + X√n) ∼ 1 √2πw2 e X2/2w2 ,

whichdependsonlyontheparameter w2

Theneglectedtermsareoftwotypes,multiplicativecorrectionsof order1/√n andadditivecorrectionsdecreasingexponentiallywith n

Theresultimpliesthat themeandeviationfromthemeantrajectoryincreasesas thesquarerootoftime,acharacteristicpropertyof Brownianmotion.

Therandomwalk:Universalityandcontinuumlimit 1.3

1.3.1Continuumtimelimit

TheasymptoticGaussiandistributionofthedeviation¯ q = q nw1 fromthemean trajectoryis

Bychangingthetimescaleandbyacontinuousinterpolation,onecan definea diffusionprocessorBrownianmotionincontinuoustime.

Let t and ε betworealpositivenumbersand n theintegerpartof t/ε:

Onethentakesthelimit ε → 0at t fixedand,thus, n →∞

Ifthetime t ismeasuredwithafiniteprecision∆t,assoonas∆t ≫ ε,timecan beconsideredasacontinuousvariableforwhatconcernsallexpectationvaluesof continuousfunctionsoftime.

Onethenperformsthechangeofdistancescale

SincetheGaussianfunctioniscontinuous,thelimitingdistributiontakestheform

(Thechangeofvariables q → x inducesachangeofnormalizationofthedistribution.)

Inthelimit n →∞ andinsuitablemacroscopicvariables,oneobtainsadiffusion processthatcanentirelybedescribedin continuumtimeandspace.Thelimiting distributionisasolutionofthediffusionorheatequation,

ThedistributionΠ(t,x)impliesa scalingproperty characteristicoftheBrownian motion.Themomentsofthedistributionsatisfy

.6)

Thevariable x/√t hastime-independentmoments.

Dimensions. Asthechange¯ q = x/√ε alsoindicates, onecanthusassignto position x ascalingdimension 1/2intimeunits(thisalsocorrespondsto assigning aHausdorffdimension 2 toaBrowniantrajectory inhigherdimensions).

Therandomwalk:Universalityandcontinuumlimit 5

1.4Correctionstocontinuumlimit

WenowstudyhowdeviationsfromthelimitingGaussiandistributiondecaywhen ε → 0.

Weexpressthedistributionof¯ q = q nw1 intermsof w(k)=ln˜ ρ(k),

wherewehaveintroduced w(

)= w(k)+ ikw1 .

Withourassumptions,theexpansionoftheregularfunction¯ w(k)inpowersof k takestheform

Aftertheintroductionofmacroscopicvariables,whichfortheFouriervariables correspondto k = κ√ε,onefinds

Oneobservesthat,when ε = t/n goestozero,toeachadditionalpowerof κ there correspondsanadditionalpowerof √ε. Inthecontinuumlimit,thedistributionbecomes

t,x)= 1 2π

Differentiatingwithrespecttothetime t,oneobtains

andin w(κ), κ canthenbereplacedbythedifferentialoperator i∂/∂x. OnethusfindsthatΠ(t,x)satisfiesthelineargeneralized‘partialdifferential equation’

Intheexpansion,eachadditionalderivativeimpliesanadditionalfactor √ε and, thus,the contributionsthatcontainmorederivativesdecreasefastertozero.

1.5Randomwalk:Fixedpointsoftransformationsanduniversality

Wenowderivethe universalproperties oftheasymptoticrandomwalk,thatis, theexistenceofa limitingGaussiandistribution independentoftheinitialdistributionwithits scalingproperty,byaquitedifferentmethodthat doesnotinvolve calculatingtheasymptoticdistributionexplicitly. Forsimplicity,weassumethattheinitialnumberoftimestepsisofthe form n = 2m.Theideathenisto recursivelycombinethetimestepstwobytwo,decreasing thenumberofstepsbyafactor2ateachiteration.Wethenlookforthe fixedpoints ofsuchatransformation.

Thismethodprovidesasimple applicationofRGideastothederivationofuniversalproperties.

Italsoallowsustointroducesomebasic RGterminology.

1.5.1Timescaletransformationandrenormalization

Ateachiteration,onereplaces ρ(q q′)by

rescalingthetimescalebyafactor 2.

Thetransformationofthedistribution ρ(q)isnon-linearbut,appliedtothefunction w(k)=ln˜ ρ(k),itbecomesalineartransformation,since

Thistransformationhasanimportantproperty:itisindependentof m or n.Inthe languageofdynamicalsystems,itsrepeatedapplicationgeneratesa stationary,or invariantundertimetranslation,Markoviandynamics

Largetimebehaviourandfixedpoints. Thelargetimebehaviourisobtainedby iteratingthetransformation,studying T m for m →∞. Alimitingdistributionnecessarilyisafixedpointofthetransformation. Itcorrespondstoafunction w∗(k)(thenotation ∗ isnotrelatedtocomplex conjugation)thatsatisfies

Fortheclassoffastdecreasingdistributions,thefunction w∗(k)hasanexpansion inpowersof k oftheform(w∗(0)=0)

Withthisassumption,thefixedpointequation(1 7)hasonlythetrivialsolution w∗(k) ≡ 0.

1.5

Therandomwalk:Universalityandcontinuumlimit 7

Tothetimerescalingmustbeassociatedarescaling(a renormalization,inquantumfieldtheoryterminology)oftherandomspacevariable q.

Randomvariable:renormalization. Non-trivialfixedpointscanbereachedifthe transformationiscombinedwitha renormalizationofthedistancescale, q → zq, with z> 0.Wethusconsiderthetransformation

[Tz w](k) ≡ 2w(k/z) .

Thetransformation Tz providesasimpleexampleofan RGtransformation,aconceptthatwedescribethoroughlyintheframeworkofphasetransitions.

Thefixedpointequationthenbecomes

[Tzw∗](k) ≡ 2w∗(k/z)= w∗(k),

whichdeterminesthepossiblevaluesof z andthecorrespondingfunctions w∗(k).

Dimensionoftherandomvariable. Comparingtherescalingoftimeandthe randomvariable q,onecanattachto q ascalingdimension dq,intimeunits,defined by dq =ln z/ ln2 (1 9)

1.5.2Fixedpoints:genericsituation: w1 =0

Intheexpansion(1 8)ofthefunction w∗(k),inthegenericsituation,thefirstterm w1 doesnotvanish.ExpandingtheRGequation,atorder k,onefinds

2w1/z = w1 and w1 =0 ⇒ z =2

Then,identifyingthetermsofhigherdegree,oneconcludes

21 ℓ wℓ = wℓ ⇒ wℓ =0for ℓ> 1

Therefore,afixedpointsolutionis w∗(k)= iw1k.

Thefixedpointsformaone-parameterfamily,buttheparameter w1 canalsobe absorbedintoanormalizationoftherandomvariable q

Since ρ∗(q)= 1 2π dk eikq iw1 k = δ(q w1),

fixedpointscorrespondtothecertaindistribution q = q = w1

Since spaceandtimearerescaledbythesamefactor2, q hasascalingdimension dq =ln z/ ln2=1intimeunits.

Consistently,thefixedpointcorrespondsto q(t)= w1t,the equationofthemean path.

1.5.3Centreddistribution

Foracentreddistribution, w1 =0andonehastoexpandtoorder k2.Onefinds theequation w2 =2w2/z2 .

Sincethevariance w2 isstrictlypositive,exceptforacertaindistribution,acase thatwenowexclude,theequationimplies z = √2. Again,thecoefficients wℓ vanishfor ℓ> 2andthefixedpointshavetheform

Therefore,onefindstheGaussiandistribution

Thedimension dq =ln z/ ln2= 1 2 isconsistentwiththescalingproperty x ∝ √t of Brownianmotion

Thetwoessentialasymptoticpropertiesoftherandomwalk, convergencetowards aGaussiandistribution,andscalingproperty,arethusreproducedbythisRG-type analysis.

1.6Localandglobalstabilityoffixedpoints

Foranon-linearRGtransformation,aglobalstabilityanalysisis,ingeneral,impossible.Onecanonlystudythelocalstabilityoffixedpoints.Here, sincethe transformationislinear,localandglobalstabilitiesareequivalent.

1.6.1GeneralanalysisandRGterminology

Setting w(k)= w∗(k)+ δw(k), onefinds, [Tz δw](k) ≡ 2δw(k/z)

Onethenlooksforthe eigenvectorsandeigenvalues ofthetransformation Tz:

[Tzδw](k)= τδw(k).

Totheeigenvalue τ,oneassociatesthe exponent

α =ln τ/ ln2 .

Theperturbation δw hasanexpansioninpowersof k oftheform,

δw(k)= ℓ=1 ( i)ℓ ℓ! δwℓkℓ .

1.6

Then,

Therandomwalk:Universalityandcontinuumlimit 9

[Tzδw](k)=2δw(k/z)=2 ℓ=1 ( ik)ℓ ℓ! z ℓδwℓ

Theexpressionshowsthatthefunctions kℓ with ℓ> 0arethe eigenvectors ofthe transformation Tz andthecorrespondingeigenvaluesare

τℓ =2z ℓ ⇒ αℓ =ln τℓ/ ln2=1 ℓ ln z/ ln2 .

For n =2m timesteps,after m iterations,thecomponent δwℓ ismultipliedby nαℓ since

Thebehaviour,for n →∞,ofthecomponentof δwℓ ontheeigenvector kℓ thus dependsonthesignoftheexponent αℓ forthevariousvaluesof ℓ

Definitions

α> 0:Theperturbationscorrespondtounstabledirections;acomponentonthe correspondingeigenvectordivergesfor m →∞.InRGterminology,aperturbation correspondingtoapositiveexponent α andwhichthusleadsawayfromthefixed point,iscalled relevant.

α =0:Suchperturbationsarecalled marginal.InthegeneralRGframework,a stabilitystudyrequiresgoingbeyondthelinearapproximation.

α< 0:Theperturbationscorrespondtostabledirectionsandarecalled irrelevant.

Thenotionofuniversality.Universality,intheRGformulation,isaconsequence ofthepropertythatalleigenvectors,butafinitenumber,areirrelevant.

Lineoffixedpoints. Quitegenerally,the existenceofaone-parameterfamilyof fixedpointsimplies theexistenceofaneigenvalue τ =1and,thus, anexponent α =0.Indeed,letusassumetheexistenceofaone-parameterfamily offixedpoints w∗(s),

w∗(s)= w∗(s) , where w∗(s)isadifferentiablefunctionoftheparameter s.Then,

1.6.2Fixedpointstability: w1 =0

Havingintroducedsomeelementsof RGterminology,wenowdiscusseigenvectors andeigenvaluesorcorrespondingexponents.

(i) ℓ =1 ⇒ τ1 =1, α1 =0.Ifoneaddsaterm δw proportionaltotheeigenvector k to w∗(k), δw(k)= iδw1k,then w1 → w1 + δw1 , whichcorrespondstoanewfixedpoint.Thischangecanalsobeinterpretedasa lineartransformationon k orontherandomvariable q

Sincetheexponent α1 vanishes,thecorrespondingeigen-perturbationis marginal

(ii) ℓ> 1 ⇒ τℓ =21 ℓ < 1, αℓ < 0.Theeigen-pertubationsare irrelevant since theycorrespondto negativeexponents.Thecomponentsof δw onsucheigenvectors convergetozerofor n or m →∞.

1.6.3Fixedpointstability: w1 =0

Wenowstudythestabilityofthefixedpointcorrespondingtothetransformation T√2.Onesets

(k)= w∗(k)+ δw(k),

andlooksfortheeigenvectorsandeigenvaluesofthetransformation

Theeigenvaluesare

Thecorrespondingexponentsare

Thevaluescanbeclassifiedas:

(i) ℓ =1 ⇒ τ1 = √2, α1 = 1 2 .Thiscorrespondstoanunstabledirection;a componentonsuchaeigenvectordivergesfor m →∞

IntheRGterminology,theperturbation,whichcorrespondstoapositiveexponent α andwhichthusleadsawayfromthefixedpoint,iscalled relevant

Here,aperturbationlinearin k violatesthecondition w1 =0.Oneisthen broughtbacktothestudyofthemorestablefixedpointswith w1 =0.

(ii) ℓ =2 ⇒ τ2 =1,α2 =0.Avanishingeigenvaluecharacterizesa marginal perturbation.Here,theperturbationonlymodifiesthevalueof w2 and,again,can beinterpretedasalineartransformationontherandomvariable.

(iii) ℓ> 2 ⇒ τℓ =21 ℓ/2 < 1 ,αℓ =1 ℓ/2 < 0.Finally,allperturbations ℓ> 2correspondtostabledirectionsinthesensethattheiramplitudes convergeto zerofor m →∞ andare irrelevant.

Redundantperturbations. Intheexamplesexaminedhere,themarginalperturbationscorrespondtosimplechangesinthenormalizationoftherandomvariables. Inmanyproblems,thisnormalizationplaysnorole.Onecanthenconsiderthat fixedpointscorrespondingtodifferentnormalizationsshouldnotbedistinguished. Fromthisviewpoint,inbothcasesonehasfoundreallyonlyonefixedpoint. Theperturbationcorrespondingtothevanishingeigenvalueisthen nolongercalled marginalbut redundant,inthesensethatitchangesonlyanarbitrarynormalization.

Otheruniversalityclasses. Othervaluesof z =21/µ,correspondformallytonew fixedpointsoftheform |k|µ,0 <µ< 2(µ> 2isexcludedbecausethecoefficient of k2 isstrictlypositive).

However,thesefixedpointsare nolongerregularfunctions of k.Theycorrespond todistributionsthathavenosecondmoment q2 andthusnovariance:theydecay onlyalgebraicallyforlargevaluesof q.IntheRGterminology,theycorrespondto different universalityclasses,distributionswithotherdecayproperties.

1.7 Therandomwalk:Universalityandcontinuumlimit 11

1.6.4Randomwalkonalatticeofpointswithintegercoordinates

Theanalysiscanbegeneralizedtoarandomwalkonthepointsofintegercoordinateslike Zd (thesimplestexamplebeingmotiononthelinebyonelatticespacing leftorrightwithprobability1/2).Then, w(k)isaperiodicfunctionofperiod2π foreachspacedirection.

However,ateachRGtransformation,theperiodismultipliedbyafactor z> 1. Thus,asymptotically,theperioddivergesand,atleastforcontinuousobservables, thediscretecharacteroftheinitiallatticedisappears.

Inthe d-dimensionallattice Zd,iftherandomwalkhas hypercubicsymmetry, ki →−ki ,ki

kj ,i,j =1,...,d, theleadingtermintheexpansionof w(k)for k smallis 1 2 w2k2,becauseitisthe onlyquadratichypercubicinvariant.Therefore,asymptotically,therandomwalkis Brownianmotionwithspacerotationsymmetry

Thelatticestructureisonlyapparentinthefirstirrelevantperturbationbecause thereexiststwoindependentcubicinvariantmonomialsofdegree4:

1.7Brownianmotionandpathintegral

Aniterationoftheevolutionequationofthetranslationinvariantrandomwalk

inthecaseofacertaininitialposition q = q0 =0,yields

Ifoneisinterestedonlyintheasymptoticpropertiesofthe distribution,which havebeenshowntobeindependentoftheinitialtransitionprobability,onecan derivethem,inthecontinuumlimit,startingdirectlyfrom Gaussiantransition probabilities oftheform

Theiteratedevolutionequationbecomes

with qn = q and

Therandomwalk:Universalityandcontinuumlimit 1.7

Wethenintroducemacroscopictimeandspacevariables(seeSection2.1),

τℓ = ℓε,τn = nε = t,xk = √εqk , with0 ≤ k ≤ n, andacontinuous,piecewiselinearpath x(τ)(Fig.1.1), x(τ)= √

Therelativepowersof ε betweenspaceandtimeagainreflectthescalingproperties oftherandomwalk.

Fig.1.1 A piecewiselinearpath contributingtothe time-discretized pathintegral.

Oneverifiesthat S canbewrittenas(withthenotation˙x(τ ) ≡ dx/dτ)

S x(τ) = 1 2w2 t 0 x(τ) 2dτ withtheboundaryconditions

x(0)=0 ,x(t)= √εq = x .

Moreover, Pn(q)= 1 (2πw2)1

Inthecontinuumlimit ε → 0, n →∞ with t fixed,theexpressionbecomesa representationofthedistributionofthecontinuumlimit,

Π(t,x) ∼ ε 1/2Pn(q), intheformofa pathintegral,whichwedenotesymbolicallyas

Π(t,x)= [dx(τ)] e−S(x(τ )) , where [dx(τ)]meanssumoverallcontinuouspathsthatstartfromtheoriginat time τ =0andreach x attime t.Thetrajectoriesthatcontributetothepath integralcorrespondto Brownianmotion,arandomwalkin continuumtimeand space.TherepresentationofBrownianmotionbypathintegrals,initiallyintroduced byWiener,isalsocalleda Wienerintegral [2](seeSection2.1).

2Functionalintegration:Frompathtofieldintegrals

Priortothesecondhalfofthetwentiethcentury,thetechnicaltoolsoftheoretical physicsweremainlydifferentialorpartialdifferentialequations.However,inthe twentiethcentury,systematicinvestigationsof largescalesystemswithquantumor statisticalfluctuations werestarted.Newtoolswerethenrequired.Amongthem, amajortoolhasbeen functionalintegration.

Inthisintroductorychapter,whichisaneditedversionofRef.[3],wedescribe variousformsoffunctionalintegralsastheyareusedindifferentbranchesoftheoreticalphysics.Arigorousstudyofthemathematicalpropertiesoffunctionalintegrals isstillanopensubtopicoffunctionalanalysisandwillnotbeconsideredhere.

Pathintegralsinvolveweightedsummationoverallpathssatisfyingsomeboundaryconditionsandcanbeunderstoodasextensionstoaninfinitenumberofintegrationvariablesofusualmultidimensionalintegrals.

Pathintegralsarepowerfultoolsforthestudyof quantummechanics.Indeed, inquantummechanics,physicalquantitiescanbeexpressedasaveragesoverall possiblepathsweightedbytheexponentialofatermproportional totheratioof theclassicalaction S associatedtoeachpath,dividedbythePlanck’sconstant [4].Thus,pathintegralsemphasizeveryexplicitlythecorrespondencebetween classicalandquantummechanicsandgiveanexplicitmeaningtothenotionof quantumfluctuations.Inparticular,inthesemi-classicallimit S/ →∞,the leadingcontributionstotheintegralcomefrompathsclosetoclassicalpaths,which arestationarypointsoftheaction:pathintegralsleadtoanintuitiveunderstanding andsimplecalculationsofphysicalquantitiesinthesemi-classicallimit.

Theformulationofquantummechanicsbasedonpathintegralsiswell adapted tosystemswithmanydegreesoffreedom,whereaformalismofSchr¨odingertypeis muchlessuseful.Therefore,itallowsaneasytransitionfromquantummechanicsto quantumfieldtheoryorstatisticalphysics.Inparticular,generalizedpathintegrals (knownasfieldintegrals)leadtoanunderstandingofthedeeprelationsbetween quantumfieldtheoryandthetheoryofcriticalphenomenaincontinuousphase transitions.

Wefirstdescribepathintegralsencounteredinthestudyof Brownianmotion [2] andofquantumstatisticalmechanics[5](also Euclideantimei.e., imaginarytime quantummechanics).Thismeansthatweconsiderthepathintegralrepresentation ofthematrixelementsofthequantumstatisticaloperator,ordensitymatrixat thermalequilibrium e βH , ˆ H beingthequantumHamiltonianand β theinverse temperature(measuredinaunitwheretheBoltzmannconstant kB is1).

Thisremarkablerepresentationofquantumstatisticalsystemsalsomakesitpossible,perhapsmoresurprisingly,toshowarelationshipbetweenclassicalandquantumstatisticalmechanics.Indeed,forawholeclassofquantumHamiltonians,the integrandintheEuclideantimepathintegraldefinesapositivemeasure. FromRandomWalkstoRandomMatrices.JeanZinn-Justin,OxfordUniversityPress(2019). c JeanZinn-Justin.DOI:10.1093/oso/9780198787754.001.0001

Wethendefinethe realtime (inrelativisticfieldtheoryMinkowskiantime)path integral[4],whichdescribesthetimeevolutionofquantumsystemsandcorresponds, fortime-translationinvariantsystems,totheevolutionoperator e it ˆ H/ (t being therealtime).

Finally,welistafewgeneralizations:pathintegralsintheHamiltonianformulation,pathintegralsintheholomorphicrepresentationrelevantforbosonsystems and,correspondingly,Grassmannpathintegralsforfermions[6].

Mostoftheapplicationsofthepathintegralideatophysicsinvolve, infact,integralsoverfields(seeChapter3).Inparticular,fieldintegralsareindispensablefor theconstructionofquantizedgaugeinvarianttheories(seeChapter13),whichare atthebasisoftheStandardModeloffundamentalinteractionsatthemicroscopic scale,aswellasfordescribing universalproperties ofphasetransitions(Chapter6). Theseapplicationsrelyonapragmaticapproach,focusingmoreondeveloping calculationtools(includinglargescalenumericalsimulations)thanonestablishing rigorousproperties.Indeed,eventhoughanumberofinterestingrigorousresults havebeenproved,theconstructionofsomerealisticquantumfield theories,like gaugetheoriesinfourspace-timedimensions,remainsaformidablemathematical challenge.

Pathintegrals:Theorigins. Thefirst pathintegral seemstohavebeendefined by Wiener [2],asatooltodescribethestatisticalpropertiesof Brownianmotion, inspiredbythefamousworkofEinstein.IfWiener’sworkisratherwellknown,a less-knownarticleof Wentzel [7]ofaboutthesameperiodintroduces,intheframeworkof quantumoptics,thenotionsof sumsoverpathsweightedbyaphasefactor, ofdestructiveinterferencebetweenpathsthatdonotsatisfyclassicalequationsof motion,andthe interpretationofthesumasatransitionprobabilityamplitude.

Later, Dirac [8]calculatedthematrixelementsofthequantumevolutionoperatorforinfinitesimaltimeintervals.Thegeneralizationoftheresult tofinitetime intervalswouldhaveledtoapathintegral.

Inphysics,themodernhistoryofpathintegralsreallybeginswiththearticles of Feynman [4],whoformulated quantumevolutionintermsofsumsoverasetof trajectoriesweightedby eiA/ ,where A isthevalueofthecorrespondingclassical action(timeintegraloftheLagrangian)and isPlanck’sconstant.

2.1Randomwalk,Brownianmotionandpathintegral

Asafirstexample,weconsiderarandomwalkonthereallinewithdiscretetimes n =0, 1, 2, (seeChapter1).Suchastochasticprocessisspecifiedbyaprobability distribution P0(q)fortheposition q atinitialtime n =0andadensity ρ describing theprobabilityoftransitionfromthepoint q′ tothepoint q.Weassumethat ρ is timeindependent,translationinvariantandsymmetric: ρ(q q′)= ρ(q′ q).

Theprobabilitydistribution Pn(q)attime n satisfiestherecursionrelationor masterequation

2.1 Functionalintegration:Frompathtofieldintegrals 15

Underrathergeneralconditions(seeChapter1),themostimportantbeingthat ρ(q)decreasesfastenoughfor |q| large(wecallthisa localMarkovprocess),one canprove(aconsequenceofthecentrallimittheoremofprobabilities)thatthe distribution Pn(q)convergesasymptoticallyforlargetimeandspacetowardsa Gaussiandistributionthat depends onlyonthesecondmomentofthetransition probability.

Therefore,ifoneisinterestedonlyinlargetimeanddistancepropertiesofthe randomwalk,onecanstartdirectlyfroman effective Gaussiantransitionprobability oftheform

where ξ> 0characterizesthewidthofthedistribution.

2.1.1Continuumlimitandpathintegral

IntheexampleofaGaussiantransitionprobability,itiseasytocalculate Pn(q) explicitlybysuccessiveGaussianintegrations.However,forourpurpose,itismore instructivetoimplementdirectlytherecursionrelation(2.1).Ifoneassumes,for example,thattheinitialdistributionisconcentratedatthepoint q = q0 (i.e., P0(q)= δ(q q0),where δ(q)isDirac’sgeneralizedfunction(adistribution)also knownas Diracfunction or δ-function),oneobtainsattime n theprobabilitydistribution

IntheGaussianexample(2.2),theexpressiontakestheform

where,defining q ≡ (q0,q1,...,qn)and q ≡ qn, S(q)= 1 2 n k=1 (qk qk 1)2 .

Wenowintroduceatimestep ε> 0,themacroscopictimeandspacevariables

k = t′ + kε,xk = √εqk , with0 ≤ k ≤ n, (suchthat τ0 = t′ , τn = t′ + nε ≡ t′′)andacontinuous,piecewiselinearpath(see Fig.1.1)

x(τ)= √ε qk 1 + τ τk 1 τk τk 1 (qk qk 1) for τk 1 ≤ τ ≤ τk and k ≥ 1 , (2.5)

withtheboundaryconditions x(t′)= √εq0 ≡ x ′ ,x(t′′)= √

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.