Perepelitsa MIT Department of Physics 70 Amherst Ave Cambridge MA 02142 Abstract We present the path integral formulation of quantum mechani cs and demon strate its equivalence to the Schr57512odinger picture We appl y the method to the free particl ID: 22613
Download Pdf The PPT/PDF document "Path Integrals in Quantum Mechanics Denn..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
PathIntegralsinQuantumMechanicsDennisV.PerepelitsaMITDepartmentofPhysics70AmherstAve.Cambridge,MA02142AbstractWepresentthepathintegralformulationofquantummechanicsanddemon-strateitsequivalencetotheSchrodingerpicture.Weapplythemethodtothefreeparticleandquantumharmonicoscillator,investigatetheEuclideanpathintegral,anddiscussotherapplications.1IntroductionAfundamentalquestioninquantummechanicsishowdoesthestateofaparticleevolvewithtime?Thatis,thedeterminationthetime-evolutionj (t)iofsomeinitialstatej (t0)i.Quantummechanicsisfullypredictive[3]inthesensethatinitialconditionsandknowledgeofthepotentialoccupiedbytheparticleisenoughtofullyspecifythestateoftheparticleforallfuturetimes.1Intheearlytwentiethcentury,ErwinSchrodingerderivedanequationspecieshowtheinstantaneouschangeinthewavefunctiond dtj (t)idependsonthesysteminhabitedbythestateintheformoftheHamiltonian.Inthisformulation,theeigenstatesoftheHamiltonianplayanimportantrole,sincetheirtime-evolutioniseasytocalculate(i.e.theyarestationary).Awell-establishedmethodofsolution,aftertheentireeigenspectrumof^Hisknown,istodecomposetheinitialstateintothiseigenbasis,applytimeevolutiontoeachandthenreassembletheeigenstates.Thatis, 1Intheanalysisbelow,weconsideronlythepositionofaparticle,andnotanyotherquantumpropertysuchasspin. 2D.V.Perepelitsaj (t)i=n=1Xn=0exp[ iEnt=~]hnj (t0)ijni(1)This(Hamiltonian)formulationworksinmanycases.Inclassicalmechanics,however,theLagrangianformulationisknowntobeequivalenttotheHamiltonianone.Thus,weseekananswertotheabovequestionthatreliesonsomeanalogueoftheLagrangianaction.In1920,P.A.M.Diracmadeamysteriouscommenttothiseect,whichlaterinspiredRichardFeynman.Consideratrajectoryx(t)betweenaninitialpoint(x0;t0)andpossiblefuturepoint(x0;t0).Letthetransitionprobabilityamplitudeh (x;t)j (x0;t0)ibetheinnerproductofthewavefunctionintheSchrodingerpictureoftheparticleevaluatedatthesetwopoints.Feynmanhintedatthe\equivalence"oftheprobabilityamplitudeandtheexponentoftheclassicalactionofthetrajectoryexp[iS[x(t)]=~],where\equivalence"isnotyetwell-dened.Itwasnotuntil1948thatFeynman,asapost-doctoralstudentatPrinceton,formalizedthisconnection.Inhislandmarkpaper[4],Feynmanpresentedaformulationofquantummechanicsbasedonthisprinciple.Letagiventrajectoryx(t)beassociatedwithatransitionprobabilityamplitudewiththesameformasthatgivenbyDirac.Ofcourse,byquantummechanics,wecannotspeakoftheparticletakinganywell-denedtrajectorybetweentwopoints(x0;t0)and(x0;t0).Instead,wecanonlyspeakoftheprobabilityofndingtheparticleattheselocations,whichisrelatedtowavefunctionsj (x0;t0)iandj (x0;t0)i.Thatis,allthatcanbedeterminedistherelativeprobabilityoftheparticletakingonepathoranother.Feynman'sinsightwasthis-thetotaltransitionprobabilityamplitudecanbeobtainedbysummingtheamplitudesoftheparticlehavingtakenanyindividualpath.Ifthequantityh (x0;t0)j (x0;t0)icanbecalculatedinthemethodsuggestedbyFeynman,thetime-evolutionofthestatecanbedeterminedbyconsideringcon-tributionsfromallpossiblefuturestates,andtheproblemissolved.Below,theketsareeigenstatesofthepositionoperator,suchthatintegrationoverallxspanstheentirebasis.j (x;t0)i=Z1 1h (x0;t0)j (x0;t0)idx0j (x0;t0)i(2)Knownasthepathintegralformulationofquantummechanics,thismethodgivesthesameresultsasthosedictatedbytheSchrodingerpicture,butalsoilluminatessomeofthedeeperaspectsofquantummechanics.Inthispaper,wewillpresentthemethodusedbyFeynman.Thoughitispedagogicallybackward,wewillthendemonstrate PathIntegralsinQuantumMechanics3theuseofthemethodbeforeshowingitsequivalencetotheSchrodingerpicture.Wewilltheninvestigatethemethodasappliedtotheharmonicoscillator.Followingthis,wewillintroducetheconceptofEuclideanpathintegralsanddiscussfurtherusesofthepathintegralformulationintheeldofstatisticalmechanics.2PathIntegralMethodDenethepropagatorofaquantumsystembetweentwospacetimepoints(x0;t0)and(x0;t0)tobetheprobabilitytransitionamplitudebetweenthewavefunctionevaluatedatthosepoints.U(x0;t0;x0;t0)=h (x0;t0)j (x0;t0)i(3)IftheHamiltoniancarriesnoexplicittime-dependence,wecanrelabelthersttime-valuet0=0andworkonlywithelapsedtimet=t0 t0.Wewilloftenwrite(3)asU(x0;t;x0)toillustratethis.Thepropagatorabove,alongwithaninitialstateket,fullydescribestheevolutionofasystemovertime.Itisalsocustomary,asisdoneinSakurai[2],touseherethesymbolKinsteadofUandrefertoasthe\kernel"or\Feynmankernel".Thepathintegralmethod,asweareabouttosee,isanexplicitwaytoconstructthispropagator.Weconsiderpossibletrajectoriesx(t)ofaparticlemovingthroughatime-independentpotentialV(x)withendpointsxedat(x0;t0)and(x0;t0).2Aninnitecontinuumofsuchtrajectoriesarepossible,eachwithclassicalactionS[x(t)].Feynman[4]positsthatthecontributiontothepropagatorfromaparticulartrajectoryisexp[iS[x(t)]=~].Thatis,everypossiblepathcontributeswithequalamplitudetothepropagator,butwithaphaserelatedtotheclassicalaction.Summingoverallpossibletrajectories,wearriveatthepropagator.ThenormalizationconstantA(t)isindependentofanyindividualpathandthereforedependsonlyontime.U(x0;t;x0)=A(t)Xalltrajectoriesexpi ~S[x(t)](4)Equation(4)istheheartofthepathintegralformulation.Howthissummation(whichhasyettobewell-dened)istoarriveatthecorrectpropagatorisfarfrom 2Atthispoint,x(t)coulddescribeatrajectorythroughspacetimewithanynumberofdimensions.Forourpurposes,however,theparticlemovesalongonespatialdimension,thoughageneralizationtomoreisstraightforward. 4D.V.Perepelitsaobvious.Wenowdiscusssomesalientfeaturesof(4)anddiveintothetechnicaldetailsofasimpleexample.2.1TheClassicalActionHowcanitbethattheinnitesumabovedoesnotdiverge?Thedierentphasesarethekeytothis.FortrajectoriesbetweenwhichtheactiondiersbyS~,thecorrespondingcontributionswillcancel.Contributionstothepropagatorfrompossibletrajectoriesthrougharegionfarawayfromtheclassicalpathwould,intheaggregate,cancel.Denotetheclassicaltrajectoryxcl(t)asthetrajectorywiththeminimumvalueoftheactionS[xcl],whichisstationarytorstorderwithregardtodeviations.Onthemacroscopicsystem,thisisthetrajectoryobservedwithverylittleuncertainty.InvestigationofEquation(4)givesareasonforthis.Trajectoriesclosetotheclassicalonecausenorst-orderdeviationintheaction,andcontributewithcoherentphasetotheintegral.Trajectorieswithaction~morethantheclassicalactionareoutofphase,andinterferedestructivelywitheachother.Integratingovermoreandmoresuchtrajectoriesshouldcausetheircontributiontoaverageouttozero.Inthisway,theclassicaltrajectoryisqualitativelyimportant.Ingeneral,theregionofcoherenceisrelatedtothe\classical"natureofthesystem.Onthe(macro-scopic)classicalscale,~isafrighteninglysmallamount,makingtheprincipalcon-tributingtrajectoriesthoseinanarrowbandaroundtheclassicalone.Onthequan-tumscale,however,theactionissmallenoughthat~isenoughtoallowsignicantquantumdeviationsfromtheclassicaltrajectory.Intuitively,thiscorrespondstothefundamentaluncertaintyintheparticle'spositionatanygiventime.Shankar[3]brie\rygivesamorequantitativeargumentofthedierencesinthetwocases.2.2FreeParticleForconcreteness,weusethemethodabovetodeterminethepropagatorforthesim-plestofsystems-aparticlemovinginfreespacealongonedimension.Inthiscase,wewillactuallyevaluatetheintegralgivenabove\overallpossiblepaths",although,asMacKenzie[5]notesafterasimilarderivation,oftenmuchofthestudyofthepathintegralformulationisconcernedwithhowtoavoidjustthis.Letx(t)describeapotentialtrajectoryfrom(xa;ta)to(xb;tb).Wediscretizethetrajectorybydividingitintochunksoftimet=(tb ta) N,suchthattheintermediate PathIntegralsinQuantumMechanics5pointsare(x1;t1),:::,(xN 1;tN 1).WedothiswiththehopethatinthelimitasN!1,thismodelsacontinuouspath.3AsV(x)=0forafreeparticle,theactiondependsonlyonthevelocity,whichbetweenanytiandti+1=ti+tisaconstant.Wedenotetheactionbetweentiandti+1bySi=Zti+1tim 2_x(t)2dt=m 2xi+1 xi ti+1 ti2(ti+1 ti)=m 2t(xi+1 xi)2(5)Todescribeeverypathfrom(x0;t0)to(xN;tN),wesimplyvaryeachintermediatepointxiateachtiovertheentiredomain.Asthefxigeachtakeonacontinuumofvaluesfrom 1to1,thesumofthecontributionsfromeverypathish (xN;tN)j (x0;t0)i=A(t)Z1 1Z1 1exp"i ~m 2ti=NXi=1(x2i x2i 1)#dx1dxN 1(6)HereA(t)isthenormalizationconstantthatdependsonlyontheelapsedtimetN t0.Wecanevaluate(6)byintegratingoveronevariableatatime.Forconvenience,letk=im 2~t.IntegrationovertherstvariableyieldsZ1 1expk(x22 x21)+k(x21 x20))dx1(7)=Z1 1exp2kx21+kx1( 2x2 2x0)+k(x22+x20)dx1(8)=p p 2p kexpk 2(x2 x0)2(9)Weignoretheconstantterm,sinceitisabsorbedintothenormalizationconstantA.Afterafewmoreintegrations,apatternemergeswhichFeynmanandHibbs[1]illustratemoreexplicitly:integratingovertherstnspacetimepointsleavesafactorofk n+1(xn x0)2intheexponentafterevaluation.WeevaluateN 2moreintegrationsaftertherstone,andusethefactthatNt=(tN t0)torewriteexpression(6)conciselyas 3Notethattheconstructedaccelerationisadiscontinuousimpulsetrainofinstantaneousjumpsinthevelocity.SincetheLagrangiandoesnotvaryexplicitlywithx,andthatinthelimitthisvalueiscontinuousanyway,weignorethis. 6D.V.PerepelitsaU(xN;tN;x0;t0)=A(tN t0)expk N(xN x0)2(10)U(x;t;x0)=A(t)expim 2t~(x x0)2(11)ThevalueforAisobtainedbynormalizing(11)overallxwhileholdingtconstant.ThenalresultisU(x;t;x0)=r m 2i~texpim 2t~(x x0)2;(12)whichisinfactthepropagatorforafreeparticle.AfascinatingfeatureofthisresultisthatUiscomposedofthecontributionfromtheclassicalpathalone.Thisisnotalwaysthecase,andwewilldiscussthisphenomenoninSection3.2.3EquivalencetoSchrodingerPictureButisthisnewformulationreallyanequivalentpictureofnon-relativisticquantummechanics?Schrodinger'sequationindierentialformdenesthetime-derivativeofagiveninitialstatej (x0;t0)i.i~d dtj (x0;t0)i=^Hj (x0;t0)i= ~2 2md2 dx2+V(x0)j (x0;t0)i(13)Weattempttorecreatethisdierentialequationusingthepathintegralformulation.Consideraninitialstatej (x0;t0)i.Someinnitesimaltimelater4,andatsomeinnitesimalposition\raway,thestatecanbedescribedasfollows.j (x;t0+)i=Z1 1d\rj (x0+\r;t0+)iU(x0+\r;;x0)(14)TocalculatethepropagatorU=h (x0+\r;t0+)j (x0;t0)i,weneedtodeterminetheaction.Foragivenpositioncoordinatex0+\rattime,theparticle'svelocityandpositioncanbeapproximatedas1 (\r)andx+\r 2,respectively.Thepropagatoris 4Oureventualintentionistoallow!0torecreatetheinstantaneousderivative.Thus,wefeelnoqualmmakinganyjustifyingassumptionsthatareexactinthelimit. PathIntegralsinQuantumMechanics7U(x0+\r;;x0)=A()expi ~Z0dtm 2_x(t)2 V(x(t))(15)A()expi ~m 2\r2 Vx0+\r 2(16)Inthelimitas(andtherefore\r)becomezero,theapproximationabovebecomesexact.Therefore,weexpandseveralofthequantitiesin(16)inpowersofand\r.Theexponentialofthepotentialenergy,andthepotentialenergyitselfarei ~exph Vx0+\r 2i=1 i ~Vx0+\r 2+:::(17)Vx0+\r 2=V(x0)+\rd d\rV(x0)+:::(18)Wekeepthersttwotermsof(17)andonlythersttermof(18),whichhastheeectofdiscardingalltermssecondorderin(;\r)andhigher.Additionally,wecanexpandtheketintheintegrandj (x0+\r;tiaroundx0.Therstandthirdtermsaresignicant,butthesecondtermresultsinanoddfunctionwithinthe(symmetric)integrand,sowediscarditasin[1].j (x0+\r;t0)i=j (x0;t0)i+\rdj (x0;t0)i dx+1 2\r2d2j (x0;t0)i dx2+:::(19)Theapproximateexpressionfor(14)becomesj (x;t0+)i=A()1 i ~V(x0)Z1 1d\rj i+1 2\r2d2j i dx2expi ~m 2\r2(20)ItisstraightforwardtoevaluatethisGaussian.Integrating,andkeepingtherstordertermsintheinnitesimalsgivesj (x0;t0+)i=A()r 2~ im1 i ~V(x0)+~ 2mid2 dx2j (x0;t0)i(21)Forsmall,weseethatthersttermontherightmustequaltheexpressionontheleft,andthatthenormalizationconstantisnecessarilyA()=p m 2~i.Torecoverthetimederivative,wecanrearrange(21)andtakethelimitas!0ofi[j (x;t0+)i j (x;t0)i]=(~). 8D.V.Perepelitsai~d dtj (x;t0)i=V(x0) ~2 2md2 dx2j (x0;t0)i(22)Thisisjust(13),thetime-dependentSchrodingerEquation,derivedfromapathintegralformulationofquantummechanics.Thereareotherwaystoshowthisequiv-alence.Forexample,MacKenzie[5]beginswiththeSchrodingerpictureandshowsthatitcanbemassagedintotheFeynmanformulation.Ourdirection,whichisthesameasthatofFeynmanandHibbs[1],istoshowthatthestandardquantumtheorycanbereconstructedfromthe\rstprinciples"ofthepathintegralmethod.3HarmonicOscillatorNowlettheparticlebeinapotentialgivenbyV(x)=1 2m!2x2.Inprinciple,wecouldexplicitlycomputethepropagatorbyconsideringagenericinitialstateandapplyingtime-evolutiontoeachoftheeigenstates.Thiswayisalgebraicallybrutal,requiringthesummationofaninniteseriesofexpressionsinvolvingHermitepolynomials.Usingthetoolswehavedevelopedabove,wewillderivethepropagatorfortheone-dimensionalharmonicoscillatorwithasurprisinglackofintensecomputation,relyinginsteadonthetheoreticalbasisofthepathintegralformulation.3.1DerivationofthePropagatorWeapproachtheprobleminaslightlydierentmanner.Letxcl(t)and_xcl(t)betheclassicalpathanditsvelocity,respectively.Wecanrepresentanyotherpathasadeviationfromthisone,xcl(t)+y(t)and_xcl(t)+_y(t),withthecorrectboundarycon-ditions(yand_yarezeroattheendpointsofthepath).Let'srewritetheLagrangianforagenericpathinawaythatisolatestheclassicalLagrangian.Werelaxthetimedependence,sinceitisimplicit.L(xcl+y;_xcl+_y)=1 2m(_xcl+_y)2+1 2m!2(xcl+y)2=1 2m_x2cl+1 2m!2x2cl+(m_xcl_y+m!xcly)+1 2m_y2+1 2m!2y2(23)Thersttermin(23)isjusttheLagrangianfortheclassicaltrajectory.ThesecondtermiszerobytheEuler-Lagrangeequations(sinceonastationarypathxcl,this PathIntegralsinQuantumMechanics9termmustbezeroforanywell-behaveddeviationy).Wewillreturntothislater.ThethirdtermistheLagrangianforthedeviation.Theactionisseparableinthesensethatthepropagatorhasthefollowingform:U=A(t)expi ~S[xcl]Zallpathsexpi ~Zt=tNt=t01 2m_y2+1 2m!2y2(24)Itcanbeshownthattheintegraloverallpathsdependsonlyonelapsedtimetandnotonx(t0)orx(tN).Shankar[3]arguesthatthismustbetruesinceeverydeviationhasnoknowledgeoftheendpointsofthetrajectory(sinceyand_yarezerothere).Wethereforeabsorbthistime-dependentfunctionintothenormalizationfactor,andareleftwithjusttheclassicalaction.5Normalizingthisexpressionisnottrivial.FeynmanandHibbs[1]considerthedecompositionofadeviationy(t)intoitsFourierserieswithperiodtN t0,andthenintegrateoverthecomponents.Weusetheirresult,thoughthereareothermethods.U=r m! 2i~sin!texpim! 2~sin!t(x2+x20)cos!t 2xx0(25)Thisisthecorrectanswer.Later,whenweinvestigatetheconnectionbetweenthepathintegralformulationandstatisticalmechanics,wewillusethisresulttorecovertheenergyeigenvalues.3.2SeparableLagrangiansWehavediscoveredthatinsomecases,onlytheclassicaltrajectorycontributestothesummationin(4).Thereisastraightforwardreasonforthis.TheLagrangianforanygenericpathcanbeexpressedasatwo-dimensionalTaylorexpansionabouttheLagrangianoftheclassicalaction.L(xcl+y;_xcl+_y)=Lcl+1Xn=11 n!yd dx+_yd d_xnLcl=Lcl+ydLcl dx+ydLcl d_x+y2d2Lcl dx2+2y_yd dxdLcl d_x+_y2d2Lcl d_x2+:::(26) 5Thisisaknownresult,sowedonotderiveithere. 10D.V.PerepelitsaTheEuler-Lagrangeequationsontheclassical(stationary)trajectorydictatethatthen=1termiszero.Inanon-relativisticframework,thekineticenergytermintheLagrangiancontributesam 2_y2term.Additionally,inapotentialisquadraticinxand_x(asinthecaseoftheharmonicoscillatorandmanyotherelementarypotentials),theLagrangianseparatesexactly.Relation(26)showsthatinthesecases,thepropagatorissimplyrelatedtotheimaginaryexponentoftheclassicalactiondividedbythequantumofaction.UA(t)expi ~S[xcl](27)Thus,therearemanyproblemswhichcanbereducedtodeterminingtheclassicalaction,oncetherelationaboveisshowntobetrueforthesystem.Afterthis,thenormalizationofthepropagatorbecomestheonlychallengingtask.Although(27)isoftenagoodapproximation(andisevenexactinsomeimportantsystems),wemustbeawareofitslimitationsanduses.Thirdorhigherpowersofxor_xintheLagrangianwillannulit.However,itworksreadilyiftheLagrangiandependsonmorethanonetrajectory,andcanevenbesolvedinthecaseofatime-dependentinputintothesystemf(t).FeynmanandHibbs[1]provideamoreindepthdiscussionofthissolutionmethodandwhereitisapplicable.4EuclideanPathIntegralsOurapproachinthefollowingsectionispragmatic,focusingmoreonthefunctionalityofamethodthanonarigorousjustication.Inprevioussections,wehaverelatedthetransitionprobabilityamplitudeofthewavefunctionbetweentwopointsinspacetime.Thisnextsectionconcernsrevisingnotthemethodbutthesenseofspacetime.Farfromtheclassicaltrajectory,therapidlyoscillatingtermsin(2)cancausecon-vergenceissuesandaregenerallyunpleasanttodealwith.ObservethatinMinkowskispacetimewithonephysicalcoordinate,theproperdistance2goesasthenegativesquareofthetime t2.However,inEuclideanspacetime,thesignofthet2termispositive;theydierbyaphasefactori.Considerwhatwouldhappenifweintroducedafactorofiintotheexponentials-theoscillatingtermswouldturnintodecayingexponentials,whichhaveanentirelydierentphysicalmeaning.Withthisasourmotivation,letusanalyticallyextendthetimeparameterintothecomplexnumbersC.Inparticular,lettimehavenorealcomponent,andidentify PathIntegralsinQuantumMechanics11t= iwitharealparametercalledimaginarytime.ThisisformallyknownasaWickrotation,andrigorousjusticationofthisstepisnottrivial.Thefruitofsuchagambitwillhopefullysoonbemadeclear.Nevertheless,letusexaminetheformpropagatoroverasmalltimeinterval i.U(x0; i;x)Xallpathsexpi ~m 2(x0 x)2 i iVx+x0 2(28)Weseethatthepotentialenergyhas\rippedsignrelativetothekineticenergyterm!WeredenetheEuclideanactionSE[x()]ofatrajectoryinimaginarytimex()below.SE[x()]=Z21m 2_x2()+V(x())(29)TheEuclideanactionisavaluerelatedtothequantumtunnelingprocess.Wewillnotenterintomoredetailonthis;bothGrosche[6]andMacKenzie[5]giveamorein-depthtreatment.Withthis,wecandenetheimaginarytimepropagatorU(x0;;x)aswell.U(x0;;x)=A()Zallpathsexp 1 ~SE[x()](30)Thoughthephysicalmeaningbehindthisnewpropagatorisstillnotobvious,wecangainsomesenseofitsinterestingproperties.Whatsortofvaluecanthesumoverallpathsin(30)have?Nolongerdoestheclassicalactioncausethecontributiontothesummationtooscillate;eachpathisnowgivenanegativeexponentialweightbasedonitsEuclideanaction.Theclassicalpathhasthelargestcontribution,sincetheactionisaminimumthere.TrajectoriesfarawayfromtheclassicaloneseeanexponentiallydecreasedcontributionasSEgrows.AparticleobeyingtheEuclideanequationsofmotionexperiencesthepotentialintheotherdirection,whichisintuitivelycorrect-apaththrougharegionofhighpotentialdampensthecontributionofthatpaththroughthenegativeexponentialdependenceoftheaction.4.1StatisticalMechanicsWebrie\ryexploretheconnectionbetweentheEuclideanpathintegralandstatisticalmechanics.Consideranensemblesystematthermodynamicequilibriumwithordered 12D.V.PerepelitsaenergymicrostatesfEngforn=1;2;:::.ThepartitionfunctionZofstatisticalmechanicsencodesprobabilisticinformationaboutthesystem.Z=1Xn=1e En(31)Above,=1=kbTbetheinversetemperatureofthesystematagiventemperatureTwithBoltzmann'sconstantkb.Theformof(31)isreminiscentofthetime-evolutionconstructedfromenergyeigenstatesbackin(1),whichwehaveshownhasanequiv-alentexpressionastheintegraloverpropagatorstoallpossiblepointsin(2).Thenaturalquestioniswhetherwecanderivethepartitionfunctionfromapathintegralstandpoint.Itwouldseemthatreplacingthetimevariabletwiththequantity iwouldbeagoodrststep.Aswehaveseenabove,thismovesusintotherealmofEuclideanpathintegrals.TheanalysisbelowisbasedonthatofGrosche[6]andFeynmanandHibbs[1].Consideraquantumsystemwiththediscreteenergyspectrumgivenabove.Werewritetheimaginarytimepropagatorbetweenx0andxbydecomposingthebraandketintoabasisofeigenstatesandapplyingtime-evolutiontoeachone.U(x0;;x)=Xnhx0jnie iEn( i)=~hnjxi=Xnhnjxie iEn( i)=~hx0jni(32)Now,settingx0=xand=~andintegratingoverallxyieldsZdxU(x;~;x)=XnhnjZxdxe Enjxihxjni=Xne En=Z(33)Thisistherelationbetweentheimaginarytimepropagatorandthepartitionfunctionthatwearelookingfor.Asavividillustrationofthisequality,wereturntothecaseoftheharmonicoscillator,whosepropagatorwederivedinSection3:1.U(x;~;x)=r m! 2~sinh~!exp mwx2 ~sinh~!(cosh~! 1)(34)ThepartitionfunctionfortheharmonicoscillatorZistheintegralof(34)overallx,whichiseasytoevaluatesincethefunctionisaGaussian. PathIntegralsinQuantumMechanics13Z=ZdxU(x; i;x)=r m! 2~sinh~!Zdxexp24 x2 ~sinh~! m!(cosh~! 1)35(35)=r m! 2~sinh~!p s ~sinh~! m!(cosh~! 1)=1 p 2(cosh~! 1)(36)Expandingthishyperbolicfunctiongivesthedesiredresult.(p 2(cosh~! 1)) 1=(p e~! 2+e ~!) 1=(e~!=2 e ~!=2) 1(37)=e ~!=21 1 e ~!=Xn=0exp n+1 2~!(38)Thisistheexpectedpartitionfunction,completewiththeenergyeigenvaluesoftheharmonicoscillator!5DiscussionThismanuscripthasbeenintendedasabriefintroduction;theliteratureisricherinapplication,scopeandrigor.FeynmanandHibbs'[1]textbookclassicintroducesallofquantumtheoryfromthisperspective.MacKenzie[5]notesthatapathintegralformulationnaturallyleadstoaninvestigationoftheAharanov-Bohmeectusingthesemethods,andgoesontopresentstationaryperturbationtheoryusingFeynmankernels.Shankar[3]delvesintopathintegralsthroughphasespaceandcoherentstatespace,andappliesthemtoanumberofsophisticatedtopicssuchastheBerryphase.Sakurai[2]writesmoregenerallyaboutGreen'sfunctions.Grosche[6]presentsapaperstartlinginitsrigorandthoroughnessonthetopic.Othersourcesevendemonstratetheuseofpathintegralsinquantumeldtheory.Still,wehavemanagedtocoverquiteabitofgroundonthetopicofpathintegralsinquantummechanics.Ourformulationofquantummechanicshasbeendrivenbytwokeyprinciples.Startingfromtheprinciplethataparticletakesnowell-denedtrajectorybetweentwopointsatwhichitisobserved,wedenethetransitionprobabilityamplitudebetweentwopointsasasummationoverallpaths.ThiswasFeynman'sinsight.Then,weconnectthecontributiontotheamplitudefromanygivenpathwiththecorrespondingclassicalactionalongthepathinthemannersuggestedbyDirac. 14D.V.PerepelitsaWehavederivedthefreeparticlepropagatorinamannerthatfullydemonstratingthe\integrateoverallpossiblepaths"principle.WehaveshownequivalencetothestandardSchrodingerformulationofquantummechanics.Then,takingadvantageoftherestrictionsthattheclassicallyobservedLagrangianactionprovides,wehavederivedtheharmonicoscillatorpropagator,anddiscussedpotentialsforwhichthepropagatoriseasilysolvable.Afterthis,wehavedenedthepathintegralinEu-clideanspacetime,andshowedhowitisrelatedtothepartitionfunctioninstatisticalmechanics.Fromthis,wehaverecoveredtheenergyeigenvaluesoftheharmonicoscillator.ItissaidthatRichardFeynmancouldnotallowbelieveanyphysicshehadnotrederivedforhimself.Thoughouranalysishasbeenbrief,andwithlessemphasisonstraightforwardrigorthanonpragmatism,itisclearthatthescopeandsuccessofFeynman'smethodistrulyremarkable.AcknowledgmentsDVPwouldliketothankBrianPepperandEricFitzgeraldfortheiradviceandproofreadingduringthecreationofthismanuscript.References[1]R.P.FeynmanandA.R.Hibbs,QuantumMechanicsandPathIntegrals(McGraw-Hill,NewYork,1965)[2]J.J.Sakurai,ModernQuantumMechanics(Addison-Wesley,Reading,MA,1994)[3]R.Shankar,PrinciplesofQuantumMechanics,2ndEd.(PlenumPress,NewYork,NY,1994)[4]R.Feynman,Space-TimeApproachtoNon-RelativisticQuantumMechanics(1948)Rev.ModernPhysics.20[5]R.MacKenzie,PathIntegralMethodsandApplications(2000)arXiv:quant-ph/0004090v1[6]C.Grosche,AnIntroductionIntotheFeynmanPathIntegral(1993)arXiv:hep-th/9302097v1