The Quantum Theory of the Emission and Absorption of Radiation Author(s): P. A. M. Dirac Reviewed work(s): Source: Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathematical and Physical Character, Vol. 114, No. 767 (Mar. 1, 1927), pp. 243-265 Published by: The Royal Society Stable URL: http://www.jstor.org/stable/94746 . Accessed: 31/08/2012 10:44 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. . The Royal Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathematical and Physical Character. http://www.jstor.org Emizssionand Absorptionof Radiaction. 243 (2) Observationshave beelnmade of the electricfields anid field changes associated with 18 distantand 5 near thunderstorms.The suddenchangesof fielddue to distantlighiningdischarges(> 8 km.) werepredominantly negative in sign,those due to near discharges( < 6 km.) predominantlypositive. The relativefrequenciesof positive and negativechangeswere 1: 5 in the former case and 43: 1 in the latter. The steadyelectriefields.belowthe 5 nearstorms wereall stronglynegative. (3) It is shown that these results indicate that the thundercloudswere bi-polarin natureand that the polaritywas generally,if niotalways,positive, the upper pole being positive and the lower pole negative. It is douibtful if any active stormsof oppositepolaritywereobservedat all. (4) The electricmomentsof the chargesremovedby 82 ligltningdischarges have been measutred.The mean value is 94 coulomb-kilometres. The Quantum ,Theoryof the Ei'mssion and Absorption of Radiatton. and Institutefor By P. A. M. DIRAC, St. John'sCollege,Cambridge, Theoretical Copenhagen. Physics, (Commlunicated byN. Bohr,For.Mern. R.S.-ReceivedFebruary 2,1927.) ? 1. Introduction andSu.*mnary. The new quantumtheory,based on the assumptionthat the dynamical variablesdo notobeythecommutative law ofmultiplicationi, has bynowbeen to forma fairlycomplete developed sufficiently theoryofdynamics. One can treatmathematically the problemof any dynamicalsystemcomposedof a ofparticles withinstantaneous forcesactingbetween number it them,provided is describable andonecaninterpret bya Hamiltonian themathematics function, physically bya quite definite generalmethod. On the otherhand, hardly has beendoneup to thepresenton quantumelectrodynamics. The anything ofa systemn in whichtheforcesare propaquestionsofthecorrecttreatment oftheproduLction of gatedwiththevelocityoflightinsteadofinstanitaneously, fieldbya movingelectron, an electromagnetic and ofthereactionofthisfield on theelectronhave not yet been touched. In addition,thereis a serious the theorysatisfyall the requirements in miaking of the restricted difficulty 244 P. A. M. Dirac. principleof relativity,since a Hamiltonianfunctioncan no longerbe used. This relativityquestionis, of course,connectedwiththe previousones, and it will be impossibleto answerany one questioncompletelywithoutat the same time answeringthem all. However,it appears to be possible to build up a fairlysatisfactorytheoryof the emissionof radiation and of the reactionof the radiationfieldon the emittingsystemon the basis of a kinematicsand dynamicswhich are not strictlyrelativistic. This is the main object of the present paper. The theoryis non-relativisticonly on account of the time beingcountedthroughout as a c-number, insteadofbeingtreatedsymmetrically with the space co-ordinates. The relativityvariationof mass with velocity is taken into accountwithoutdifficulty. The underlying ideas ofthe theoryare verysimple. Consideran atominteracting with a fieldof radiation,whichwe may suppose fordefiniteness to be confinedin an enclosureso as to have onlya discreteset of degreesof freedom. Resolvingthe radiationintoits Fouriercomponents, we can considerthe energy and phase of each of the componentsto be dynamicalvariablesdescribingthe radiationfield. Thus if Er is the energyof a componentlabelled r and Or is the corresponding phase (definedas thetimesincethewave was in a standard phase),we can suppose each Er and ?r to forma pair of canonicallyconjugate variables. In the absence of any interactionbetweenthe fieldand the atom, the wholesystemof fieldplus atom will be describableby the Hamiltonian H = ErEr + Ho (1) equal to the total energy,Ho beingthe Hamiltonianforthe atom alone, since the variablesErn Or obviouslysatisfytheir canonical equations of motion _ all allH = ao = ? (3r aEr Whenthereis interactionbetweenthe fieldand the atom,it could be takeninto account on the classical theoryby the additionof an interactiontermto the Hamiltonian(1), whichwould be a functionofthe variablesofthe atom and of the variablesEr, Orthat describethe field. This interactiontermwould give the effectoftheradiationon the atom,and also the reactionofthe atom on the radiationfield. In orderthat an analogous methodmay be used on the quantum theory, it is necessaryto assume that the variables Er, Orare q-numberssatisfying the standard quantum conditions OrEr- ErOr= ,h,etc., whereh is (27)times the usual Planck's constant,like the otherdynamicalvariables of the problem. This assumption immediatelygives light-quantumpropertiesto Emissionand Absorptionof Radiation. 245 the radiation." For if vris the frequencyof the componentr, 27VrOr is an angle variable, so that its canonical conjugate Er/27vrr can only assume a discreteset of valuesdiflering by multiplesof h, whichmeansthat Er can ofthequantum(2wh)'Vr If we nowadd an changeonlybyintegral multiples interaction term(takenoverfromtheclasicaltheory)to theHamiltonian(1), to therulesof quantummechanies, theproblemcan be solvedaccording and resultsforthe actionofthe radiation we wouldexpectto obtainthe correct andtheatomi ononeanother. It willbe shownthatweactuallygetthecorrect ofradiation, and the correctvaluesfor laws forthe emissionand absorption Einstein'sA's and B's. In the author'sprevioustheory,t wherethe energies ofradiationweree-numbers, and phasesofthecomponents onlytheB's could of on reaction the atom the radiation could notbe taken and the be obtained, into account. It will also be shownthat the Hamiltonianwhichdescribesthe interaction waves can be made identicalwith the of the atom and the electromagnetic of of the atom with an assembly for the interaction Hamiltonian the problem of particlesmovingwiththe velocityof lightand satisfyingthe Einstein-Bose statistics,by a suitablechoiceof the interactionenergyforthe particles. The numberof particleshavingany specifieddirectionof motionand energy,which can be used as a dynamicalvariable in the Hamiltonianforthe particles,is wave in the equal to the numberof quanta of energyin the corresponding Hiamiltonianforthe waves. There is thus a completeharmonybetweenthe wave and liglit-quantumdescriptionsof the interaction. We shall actually the light-quantumpoint of view, and show that the build up the theoryfromrl Hamiltoniantransformsnaturallyinto a formwhich resemblesthat for the waves. The mathematicaldevelopmentofthe theoryhas been made possibleby the author's general transformation theory of the quantum matrices.: Owing we are allowedto use thenotion to thefactthatwe countthetimeas a c-number, of the value of any dynamicalvariable at any instantof time. This value is * Similar assumptions have been used by Born and Jordan ['Z. f. Physik,' vol. 34, p. 886 (1925)] forthe purpose of takingover the classical formulaforthe emission of radiation by a dipole into the quantum theory,and by Born, Heisenberg and Jordan ['Z. f. Physik,' vol. 35, p. 606 (1925)] for calculating the energy fluctuations in a field of black-body radiation. t ' Roy. Soc. Proc.,' A, vol. 112, p. 661, ? 5 (1926). This is quoted later by, loc. cit., I. I 'Roy. Soc. Proc.,' A, vol. 113, p. 621 (1927). This is quoted later by loc. cit., II. An essentially equivalent theory has been obtained independently by Jordan ['Z. f. Physik,' vol. 40, p. 809 (1927)]. See also, F. London, I Z. f. Physik,' vol. 40, p. 193 (1926). 246 P. A. M. Dirac. " according i bya generallsedinatrix a q-nunber, capableofbeingrepresenited matrixschemes,sox e of whichmay have continuous ranges to manydifferent to involvecertain. and may requirethematrixelements ofrowsand columns, (ofthetypegiven'bythe a functions*). A matrixschemecan kindsofinfinities ofintegration ofthedynamical be foundin whichanydesiredset ofconstants orin whicha setof bydiagonalmatrices, arerepresented thatcommute system by matricesthat are diagonalat a variablesthat commuteare represented of values of a diagonalmatrix the elenments diagonal The timie.t specified valuesofthatq-number.A anyq-nuimber are thecharacteristic representing valhes or will haveall characteristic co-ordinate momentum in general Cartesian setofcharactervariablehasonlya discrete from- o to + oo, whilean actioni to denotethe to use letters make it unprimed aJ rule istic values. (We shall dynamicalvariablesor q-numbers,and the same lettersprimedor multiply oreigento denotetheircharacteristic functions values. Transformation primed of the characteristic valuesandnotoftheq-numbers are functions functions so theyshouldalwaysbe written interinsofprimedvariables.) themselves, ofthe canonicalvariablesib, rh, thematrixrepreIff(i, ) is anyfunction schemeinwhichtheLat timetarediagonal at anytimetinthematrix sentingf downwithoutanytrouble,sincethematrices reprebe written matricesmnay at tine t are known, the L and kcthemselves senting naimely, ik (44 - i k 8 WVi), oftheL and_, we can at H is givenias a function ThusiftheHamiltonian thematrixH(i' i"). We can thenobtainthetransformation oncewritedowxn to a nmatrix scheme(M)in whichthe (i'/') say, whichtransform-s function, as ( '/oe')mustsatisfy is a diagonalmatrix, theintegral equation Hamiltonian Wa)*('' dH 1/R#d>'(4/7,)-A (3) levels. Thisequation valies W(c') are theenergy ofwhichthe characteristic waveequationfortheeigenfunetions ('foc'),whichbecomes, is justSchrddinger's ofthe differential an ordinary equationwhenH is a simplealgebraicffunction * Loc. cit. II, ? 2. One can have a matrix seliem in which a set of variables that commute are at all times representedby diagonal miatricesif one will sacrificethe condition that the matrices must fromsuch a scheme to one satisfythe equations of motion. The transformationftunction in which the equations of motion are satisfied will involve the time explicitly. See p. 628 in loe. cit., 11. Jr Emtissiontand Absorptionof Radiation. 247 ofthespecialquations(2) forthematrices L andkoroaccount representing Lk and 7. generalforia in themllore Equation (3) maybe Nwritten JH ( ')d"('i2J-h a(1) ,(3') in which it canbe appliedto systems forwhich theHamiltonian involves the timeexplicitly. Oniemnay havea dynamical system specified IH which by a' Hamiltonian cannot beexpressed as an algebraic ofanysetofcanonical function variables, canallthesameberepresenited bnlt which H(S'ua).Sncha problem bya matrix canstillbe solvedbythepresent sinceonecanstilluseequation method, (3) toobtain theenergy levelsandeigenfunctions. Weshallfind thattheHamiltonianwhich describes theinteraction ofa light-quantum andanatomic is system ofthismore general type,sothattheinteraction canbetreated mathematically, onecannottalkaboutan in-teraction although in theusuLal energy potenitial sense. It should beobserved is a difference thatthere a light-wave between andthe de Broglie orSchrddinger waveassociated withthelight-quanta. the Firstly, is alwaysreal,whilethede Broglie light-wave waveassociated witha lightina definite quantum directioni mustbetakentoinvolve animnaginary moving exponential. A moreimportant difference is thattheirintensities areto be in different interpreted ways. Thenumber oflight-quanta perunitvolume associated witha monochromatic light-wave equalstheenergy perunitvolume ofthewavedivided bytheenergy ofa singlelight-quantum. (27rh)v Onthe otherhanda monochromatic de Broglie waveofamplitude a (multiplied into theimaginary exponential factor)mustbe interpreted a2 lightas representing quantaperunitvolume forallfrequencies. Thisis a specialcaseofthegeneral rule forinterpreting the matrixanalysisj accordingto which,if (i'/') or ' of an atomic t (ik') is theeigenfunction in thevariables 4 of the state system (orsimple particle), I (a4') value L', [or !ia (i') 2 I2 iS theprobability of each4 having the dil' dt' ... is theprobability of each4 lyingbetween thevaluesik' andE.k'+ d<'k whenthe k.havecontinuous ranges ofcharacteristicvalues] ontheassumption thatallphases ofthesystem areequally probable. Thewavewhoseintensitv is to be interpreted in thefirst ofthesetwoways appears inthetheory onlywhen oneisdealing withanassembly oftheassociated particles theEinstein-Bose satisfying statistics.Thereis thusno suchwave associated withelectrons. * Loc. cit.,H. ??6, 7. P. A. M. Dirae. 248 Systems. ofIndependent ofa, A,ssembly ? 2. The Perturbation We shall now considerthe traisitions producedin aniatomic systemby an arbitraryperturbation. The iethod wreshall adopt will be that previously givenby the author,twhichleads in a simpleway to equationswhichdetermine the probabilityofthe systembeingin any stationarystate ofthe unperturbed the probablenumnber systemat any time.i:- This, of course,givesimmirediately of systemsin that state at that time for an assembly of the systenms that are independentof one anotherand are all perturbedin the same way. The object of the presentsectionis to show that the equations forthe rates. of changeof these probablenumberscan be put in the Hamiltonianformin a simple m'anner,which will enable furtherdevelopmentsin the theoryto be made. systemand V the perturbing Let H0 be the Hamiltonianforthe unpert-urbed variablesand may of the can be functiorn dynamical an arbitrary which energy, ormaynotinvolvethetimeexplicitly,so thatthe Hamiltonianfortheperturbed forthe pertuLrbed must systemn systemis H = Ho + V. The eigenfunctions satisfythe wave equation i a?at (Ho+ V)D6, rardr is the solutionof this equation where(H10-H-V) is an operator. If that satisfiesthe properinitialconditions,wherethe Q.'s are the eigenfunctions forthe unperturbedsystem,each associatedwithone stationarystate labelled. by the suffixr,and the ar's are functionsofthe time only,then Ia 12 is theprobability of the systembeingin the state r at any time. The a,'s must be normalised initially,and will then always remain normalised. The theorywill applydirectlyto an assemblyof N similarindependentsystemsif we nm-ultiply each of these ar's by N-!'so as to make X,.Ia7 12 = N. We shall now have that ar 12 iS the probablenumberof systemsin the state r. The equationthat determinesthe rate of changeofthe a,'s is? (4) ihd7r =EsrsVas wherethe V,,'s are the elementsofthe matrixrepresetiungT. imaginary equation is z.q -Ssas -tazr - zq7Ts ATse The conjugate ~~(4/) t Loc. cit. I. (1926)] I The theory has recentlybeen extended by Born [' Z. f. Physik,' vol. 40, p. 167 be that states the may in stationary adiabatic changes so as to take into account the in not is used This transitions. extension the as well as the by perturbation produced the present paper. ? Loc. cit., I, equation (25). Emissionand Absorptionof Radiation. 249 Ifweregard conjugates, equations (4) and(4') take a,.andiha,* as canonical the Hamiltonianformwith the HamiltonianfunctionF1 = rsar*V,T.as, namely, da}. I WIl da dt i- i"Faa' - * da7* aF, ihR --q dt aar We cantransform to thecanonical variables transNr, jr bythecontact formation 4 N le-4i4Ah a_* =N * aC. Thistransformation makesthenewvariables N,.and0,. real,Nrbeingequal to aa,.?*=1 1ar2,the probablenumberof systemsin the stater, and Orlh beingthephaseoftheeigenfunction thatrepresents them. TheHamiltonian F1nowbecomes ()S)h FJ.= ZrsVr.sN);1Ns2 and the equationsthat determinethe rate at whichtransitionsoccurhave the canonical form WI ~~aF, q 8?>r aN.r A mnore convenientway ofputtingthetransitionequationsin theHamiltonian formn may be obtainedwiththe help ofthe quantities b = a e-w1th b_-* a,.*%XYtIh W,.beingthe energyofthe state r. We have Ib,.12equal to Ia. numberof systemsin the state r. For brwe find 12,the probable ih b,.= WrbI.+ ihare- iwrtIl i (WI-W - 'Wrbr W.)tlh + Es V,.8bse t/h,so that v.s is a constaint withthe help of (4). If we put Vr, Vres(W whenV does not involvethe timeexplicitly,thisreducesto ih br Wrbr+ Esvrsbs -ESHI'ibS) (5) whereHrS-- WI.01.8 + v,.,,whichis a matrixelementof the total Hamiltonian H -ll + V withthe timefactoret(Wr-Ws) t/hreimoved, so that H,.,is a constant whenH does not involvethe time explicitly. Equation (5) is ofthe same form as equation (4), and may be put in the Hamiltonianformin the same way. It shouldbe noticedthat equation (5) is obtaineddirectlyif one writesdown the Schrbdinger equationin a set of variablesthat specifythe stationarystates of the unperturbedsystem. If these variablesare Rh,and if H(i'4") denotes VOL. CXIV.-A. 5 P. A. M. Dirac. 250 H in the (4) sIheine,this a matrixelementof the total Hamiiltonian equation would be Schri5dinger ik4a ( ')lat > H (g'i") qg/"), (6) fromthe previousequation(5) onlyin the like equation(3'). This differs r beingthereusedto denote; a stationary state instead a singlesuffix notation, of a setof numeric values i forthe variables k, ud ,beingnsed instead also equation(5r), canstillbe usedwhen of:4 ( '). Equation(6), and therefore whichcai notbe expressedas an is of themoregeneraltyype theHamiltonian of a st of canoniIalvariables,btutcan still be represented algebraicfuinction by a inatrix H (g't") or HI,,. Wenowtak-eb,and ihb to be canonically conjugatevariablesinsteadof equation-nvill (5) and its conjugateimaginary The equiation ar and ihaj' Hamiltoniian function form, with the nowtaketheHamiltonian. F rsbr (7) HrsbXsb we makethecontacttransformation as before, Proceeding N_' N eb-.=N_ bi rcl'e (8) to the new canonicalvariablesN., 0,, whereN, is, as before,the probable F in the stater,and 0ris a -newphase. The Hamiltonian iiumberofsystemis now become will F s= ,., N j Nje2Cic0,-)/h forthe ra.tesof changeof N, and 0, will take the canonical aandthe equLations fornm Ni a Or aNr. be written m-ay ?heHamiitoonian F - ?,.WrN, -F Eq,, S2Nj Ns2 e'(&0 )/h (9) and the is the total properenergyof the assembly, The firsttermE,.WrNN If dueto theperturbation. as theadditionalenergy secondmaybe regarded ld increaselinearlywith the time, tih perturbatio is zro, the phases 0, wvo whilethe previous phases +,. would in this case be cons-tants. theEinstein-Bose Sttistics. ? 3. The Perturbation ofan Assemblysatisfying of a perturbato thepreceding sectionwe can describe theeffect According ofindependent systems tionon an asse-mbly by ineansof canonicalvariables ofthetheorywhich and Hamiltonian equationsofmotion. The development EEmission and Absorptionof Radiation. 251 naturallysuggestsitselfis to make thesecanonicalvariablesq-nunmbers satisfying the usual quantumconditionsinsteadof c-numbers, so that their Hamiltonian equationsof motion become truequantuimequations. The Hamiltonian functionwill. now providea Schr6dinger wave equation,which.mustbe solved and interpreted in the usual manner. The in-terpretation will give not merely theprobablenuLmber of systemsin anystate, but the probabilityof any given distributionof the svstemsamong the various states, this probabilitybeing, in fact,equal to the square of the modulusof the normalisedsolutionof the wave equation that satisfiesthe appropriateinitial conditions. We could, of course, calculate directlyfromelementaryconsiderationsthe probabilityof any given distributionwhen the systemsare independent,as we know the probabilityofeach systembeingin any particularstate. We shallfindthlatthe probabilitycalculated directlyin this way does not atgreewith that obtained from. the wave equationexceptin the specialcase whenthereis onlyone system in the assembly. In the generalcase it will be shownthat the wave equation leads to the correctvaltLeforthe probabilityof any given distribut-ion when the systemsobey the Einstein-Bosestatisticsinstead of beingindependent. We assumethe variablesb,.,ihbY l of ? 2 to be canonicalq-nLmbers satisfying the quantumconditions b.i. ihb2.*-i7bl.*. b2.= h bqb2- b;. -_ 1, or and - bsbr 0, brbS b -bs -bs*br 0 - bs-b,, br*bs,* (s 4 0, qr). now be writtenin the quantumform The transformation equations(8) m-wust b. = (N2.-+ 1) e-01)1 b_t - e- (1o2N,i>1 (Ni.+ 1)U, Nr]e6i&/h_ 6ioGlh 3 . also be canonical variables. These equations in orderthat the Ni., n0may show that the N. can have only integralcharacteristicvalues not less than zero,t which provides us with a justificationfor the assumptionthat the of systems variablesare q-numbersin the way we have chosen. The nuimbers statesare now ordinaryquantumniumbers. in the different t See ? 8 ofthe author'spaper 'Roy. Soc. Proc.,'A, vol. 111,p. 281 (1926). What are valuesthata q-numbercan take are heregiventhemoreprecise therecalledthe c-number valuesofthatq-number. nameofthe characteristic s 2 P. A. M. Dirac. 252 (7) nowbecomes TheHamiltonian 2rsb7*Hr bei8z F = - (Ns + lhHrs (N, +Er$HrsNri 1-A 1) e- i0s11 (11) rs)!ei(0-)h in whichtheHrsarestillc-numbers.We maywritethisF in theformcorreto (9) sponding F - ErWrNr+ ErtvrsNr" (Ns+1 - ci(Or@)/h rs)k (11') in whichit is again composed of a proper energytermErWrN,and an inter- term. actionenergy The wave equationwrittenin termsofthe variablesNr ist ik N2%,Ns' (N1%, ...) = Fs (N1', N2%,N3' ), (12) to meanih a/aNrt. in F beinginterpreted whereF is an operator,each Oroccurring If we apply the operatore+iG0Ih to any functionf(N1', N2', ... Nr, ...) of the variables N'I N ', ... the result is e ? i,hf(Ni', N2', ... Nr' -) eFb/IN,'f(N1', N2', =f(N1', ... Nrt'..) N2', ... Nr' T 1, ... If we use thisrulein equation(12) and use the expression(11) forF we obtain: ih Tt+ (Nl', N2', N3' -rbs HrsNr (Ns' + 1. - 3rs)l (N1,'N2'... Nr' 1,... Ns'+ 1, ...). 2 (13} We see from the right-handside of this equation that in the matrixrepresenting F, the term in F involving ei(, - 0.)!h will contributeonly to those matrix elements that refer to transitions in which N, decreases. by unity and N. increases by unity, i.e., to matrix elementsof the type F (N', N2' ... N,' ... Ns'; N', N2'... Nr'-1 ... Ns' + 1 ...). If we find,a, solution6 (N,', N2' ...) of equation (13) that is normalised[i.e., one forwhich 1] and that satisfiesthe proper initial conN2' 1+ (N,', N2 ...)12 of that distribution in *w) 2 will be the probability N2' ditions,then I 4 (N1%, in ... at state time. in state 1, N2' any 2, whichN1' systemsare Considerfirstthe case whenthereis onlyone systemin the assembly. The probabilityof its being in the state q is determinedby the eigenfunction that the label r of the stationarystates takes the t We are supposingfordefiniteness values 1, 2, 3, .... i When s = r,# (N.,, N2'... Nrt'- 1... N*' + 1) is to be taken to mean # (N1'N2' ... N,'...). aind Absorptionof Radiation. lErntsswon 253 in whichall the N"s are pnt equal to zero exceptNq', \vhichis t! {q}. When it is substitutedin the left-handside of (13), all the termsin the sumumation on the right-handside vaanishexceptthoseforwhichr-q, and we are leftwith 4i(N1' N2', ...) puLtequal to unity. This eigenfunction we shall denoteby ih k Hq44} at whichis the same equationas (5) with Q {q} playingthe part ofb,. This establishes the fact that the presenttheoryis equivalentto that of the preceding sectionwhenthereis only olnesystemin the assembly. Nowtake thegeneralcase ofanlarbitrarynumberofsystemsin the assembly, and assume that they obey the Einstein-Bosestatistical mechanics. This requiresthat, in the ordinarytreatmentof the problem,only those eigenfunctionsthat are symmetricalbetweenall the syste:msinust be taken into beingby themselvessufficient to give,a complete account,these eigenfunctions quantumsolutionof the problem.t We shall now obtain.the equation forthe rateof chang,e ofone of thesesymnietricaleigenfunctions, and show that it is identicalwith equation (13). for the If we label each systenm with a numnber n, then the Hamniltonian a.ssemblywIll be H = H (n), whereH (n) is the H of ? 2 (equal to Ho + V) of expressedin terms the variablesof the nth system. A stationarystate of the assemblyis definedby the numbersrl, r2... rn... whichare the labels ofthe stationarystatesin whichthe separatesystemslie. The Schr8dinger equation forthe assemblyin a set of variablesthat specifythe stationarystates will be, of theform(6) [withHA instead of H], and we can write it in the notationof equation (5) thus: ihb(r1r2...) Esl,s... 1A(rlr2 1SS2 *.-) b(s1s2...), (14) whereHA(rlr2 ...; ss ...) is the general matrix of HA [with the time elem.ent factorremoved]. This matrixelementvanisheswhenmorethan onlesI, differs fromthe corresponding when sm differsfrom rmand every rn; equals Hrnl,;,, other sn equals r ; and equals SH ,. when every s. equals r, Substituting these values in (14), we obtain ihb(r1r2... ) =ms&frJ'rsk (r1Lr2 +m}1n+1 ) Y} XnHr)trnb(1'rr2...). (15) We mustnowrestrictb (r1r2...) to be a symmetrical functionofthe variables to r2 order obtain the Einstein-Bose statistics. This is permissible ... in rl, since if b (r_r2' at any time,then equation (15) showsthat -") is symmetrical t Loc. cit.,I, ? 3. P. A. M. Di-ac. 254 at thattime,so that b (rr2 ...) willremain b(r1r2...) is also symmetrical symmetrical. state inthestater, Thena stationary ofsystems LetN,.denotethenumber maybe specified eigenfunction ofthe assemblydescribable by a symmetrical by the numbersN1,N2 . N... just as well as by the numbersre, r.2... and we shall be able to transformequationl(15) to the variablesN1, N2 We cainnotactuallytake the new eigenfunction b (N1, N2. ) eqaal to the premultiple of the vious one b (r.1r2...), but must take one to be a numuerical with respect to its other in orderthat each may be correctlynoimlnalised in fact, respectiVevariables. We musthave, 2 Sr,,?,...16(rl ...) 12 I lb(N1, - N2 .) 12- 2 fo 2 equal to the suzmof ib (rLr2 . and hence we must take Ib (N1, N2 r, m2... suchthatthereare N1 ofthemequal to 1,N2 allvaluesofthenumnbers equal to 2, etco. Thereare iN!/N12!N2! ... termsin thissum,whereN of systems,and they are all equal, sinee b (rr ...) is a is the total inumnber Hence we musthave of its variablesr1,r2 .... funictionl symmetrical ) b (N1 N2 ...) (N !/N1 ! N2! b (rr2 ...) ...). left-handside will becomne If we make this suLbstitutiont in equation (15), theG b (r,.r2** rL1 8 ,/in-1 The term ih (N1 ! N2 ! ... /N!)- b (N1,N2 ...) * *) on the right-handside will becolme in the fist summamtion !.. .(Ns+ J1)! .. I/N!'H,sb (N,, NN 2 ... [N ! I (N2;-l) .. - I . Ns+ 1..), (iG) f r,,,and s fors,. This termmust be s,uLime'dfor wherewe have Arritten?or forr takingeach ofthe values mustthenbe summ-ed s and a11valuesof exceptr, hus each term(16) gets repeatedby the summationprocessui Ail i1, r2 T... it occursa total ofNr times,so that it contributes Nr [NI 1Ns . (N -i.) ! ... (Ns+ i) N-1(s + I)` (NL! N to the :right-handside of (15). IN!t]-!T-.;36, (N,, N2..N1.- -N- .-I ..N + ,sb(N, r.) Finally, the term XE.ff ( n_9 Hrff,binb ErNHrr* b (rl2 Er2Ns1Hm.,_ (N1,! N2! ... ...+Ns + 1...) ...) becomes /N !)- b (N1.,N ...) Hence equation(15) becoines,witlhthe reuoval of the factor(N,.! N2 !..../N IN)1 ihb (N1,NS ...) = N; (Ns+l)i Ns + 1 ..) + E)N7HTb (N,I 2 .. 15. (N1, N2 ... N-1 ... (17) Fnission a6-nd A bsorptionof Radiation. 255 whichis ideniticalwith (13) [exceptforthe fact that in (17) the primeshave been omittedfromthe N's, whichis permissiblewhenwe do not requireto refer to the N's as q-numbers]. We have thus establishedthat the Hamiltonian ofa perturbation (11) describesthe effect on an assemblysatisfying theEinisteinBose statistics. ? 4. TheReaction onthePerturbing oftheAssemnbly System. Up -tothe presentwe have consideredonly perturbations that can be represented by a perturbingenergyV added to the Hamiltonianof the perturbed system,V beinga functiononlyof the dynamicalvariablesof that systemand perhaps of the time. The theorymay readily be extendedto the case when the perturbationconsistsof interactionwith a perturbingdynamicalsystemn the reaction of the perturbedsystem on the perturbingsystem being taken int-oaccount. (The distinctionbetweenthe perturbingsystemaand the perturbed systenm is, of course,not real, but it will be kept up forconvenience.) We now considera perturbingsystem,described,say, by the canonical variablesJ,, , the J's being its firstintegralswhenit is alone, interacting ofperturbedsystemswithno mutualinteraction,that satisfy withan assenmbly the Einistein-Bosestatistics. Thi-etotal Hamiltonianwill be of the formi IT = Hp(J) + E.1- (n) whereH. is the lam iltonianof the perturbingsysterm (a functionof the J's only)and H (n) is equal to the properenergyHo (n) plus the perturbation energy V(n) ofthe nthsystemofthe assembly. H (n) is a functiononlyofthe variables of the nthsystemofthe assemxbly and of the J's and wv's, and does not involve the time explicitly. The Schr6dingerequation corresponding to equation (14) is now hb ( nlf2...e EJ sI>2 .. HT( l 2 ;J SIS2 ...*)b(J ,802...) in whichthe eigenfunction b involvesthe additionalvariablesJ'. The matrix elementHT (J', rLr2... ; J", s1s2...) is now always a constant. As before,it vanishes when more than one sn,differsfromthe corresponding,. When fromn and differs other qrn every it equals reduces to H (J'r ; J"tsn), sM, s. r?,, whichis the (J'r; J"s.n) matrix element(with the time factorremoved)of H = Ho +- V, the proper energy plus the perturbationeniergyof a single systemof the assembly; while when every sn equals rf, it has the value Hr (J') aj j + - H (J'r,; J"rqb). If, as before,we restrictthe eigenfunctions P. A. M. Dirac. 256 in the variablesr_,r2 *.., we can again transformto the to be symmnetrical variables N1,N2 ... whichwill lead, as before,to the result ihb(JtnNt, 2' *...=Hr (J) b(J', Y, N2/.. t + SJ?,sNi~~2b(JFfl N/r:N 2 (Y rs) J'r,3s) . .Ni ..Ns, ^ + I . ) (18) to the Hamiltonianfunction corresponding This is the Schrddinger equLation = Hp (J) + rsIfrsNP (Ns + I--s)-l et(0-)/, (19) ofthe J's and w's, being such that whenreprein whichH,Js is now a fuLnction sented by a m-atrixin the (J) schemeits (J' J") elementis H (J'r; J's). (It shouldbe noticedthat HIS,still commuteswiththe N's and O's.) F Thus the nteraction of a perturbing system and an assemiblysatisfyingthe Einstein-Bosestatisticscan be describedby a Haamiltonianof the form(19). to (11') by observingthat the matrix We can put it in the forn corresponding H of the element (J'r; J"s) is composed sumiiof two parts, a part that comes r and fromthe properenergyHo, which eqaals W, when Jk"= Jj, and s energyY, the j-uteraction vanishes otherwise,and a part that comes fronm be denotedby v (J'r; J"s). Thus we shall have whichmnay HRIS = i. ks + V2ss. wherev,. is that functionof the J's anidw's which is representedby the tiatrix whose(J'J")elementis v (J'r; J"s),and so (19) becomes F ci Hp (J) + EIW,N, + sVsNr (Ns + 1 - A j(OO9!b. (20) oftheproperenergyofthe perturbinlg system The Hamiltonianis thusthe sunm and the perturbaHp(J), the properenergyof the perturbedsystemsY,WVA7,Nr )/h tionl energy E,. ,v;s.AJ (Ns + 1-- &,s)Ae( ? 5. TheoryofTransitions in a Systemfomn One StatetoOthersoftheSame Energy. Beforeapplying the results of the precedingsectionsto light-quanta,we shall considerthe solutionof the problempresentedby a Hamiltonianof the type (19). The essentialfeatureofthe problemis that it refersto a dynamical systemwhich can, under the influenceof a perturbationenergywhich does fromone state to othersof not involve the time explicitly,make tranlsitions the same energy. The problemof collisionsbetweenan atomicsystemand an electron,whichhas beentreatedby Born,*is a specialcase ofthistype. Born's methodis to finda periodlicsolu-tionof the wave equation whichconsists,in so far as it involvesthe co-ordinatesof the collidingelectron,of planiewaves, * Born.'Z. f. Phvsik.'vol 38. D. 803 (1926). and Absorptionof Rad'ation'. Er-mission 257 systenm, whichare representing theincident electron, approaching theatomic or diffracted in all directions. ofthe ThesquareoftheamplituLde scattered is thenassumed byBorn wavesscattered in anydirection withanyfrequency in thatdirection with to be theprobability oftheelectron beingscattered the corresponding energy. in any simplemanner Thismethoddoesnotappearto be capableofextension from onestatetoothers -tothegeneral problem ofsystenms thatmaketransitiors andcertain ,ofthesameenergy.Alsothereis at present no verydirect way toapplytoa non-periodic ofinterpreting ofa waveequation a periodic solution that physical phenomenon suchas a collision.(Themoredefinite mnethod it being will nowbe givenshowsthat Born'sassulmptionis not quiteriglht, of a certain to the by factor.) necessary multiply thesquare amplitude is tofind a non-periodic ofsolving Analternative a collision method problem of whichconsists Solultion ofthewaveequation initially simply planewaves overthewholeofspacein thenecessary direction withthenecessary moving oftimewavesmoving torepresent electron.In course frequency theincident mustappearin orderthatthewave equationmayremain in otherdirections oftheelectron inanydirection beingscattered satisfied.Theprobability with ofthecorresponding anyenergy willthenbedetermined bytherateofgrowth of thesewaves. The waythe mathematics is to be harmonic component is by thismethod interpreted quitedefinite, beingthesameas thatofthe of ?2. beginning ofa system tothegeneral Weshallapplythismethod which makes problem ofthesameenergy undertheactionofa transitions from onestateto others oftheunperturbed system perturbation. Let Ho be theHamiltonian and thetimeexplicitly. V thepertuLirbing whichmustnotinvolve If we energy, takethecaseofa continuous bythefirst rangeofstationary states,specified thenmethod of integrals, motion, then,following oXksay,oftheunperturbed ? 2, we obtain ih&(a') V V(ooc")d%". a(c'), (21) ofthesystem beingin a state corresponding toequation (4). Theprobability . doc2' forwhicheachOCk at anytimeisIa (Oc)12do-.' liesbetween oC'andOCk7'+ daCk' initialconditions. theproper whena (oc')is properly andsatisfies normalised valueofa (o') the initial Ifinitially take wemust thesystem is inthestateoc0, a . a (oc'- oc0).We shallkeepa0 arbitrary, as it wouldbe to be oftheform a (oc')inthepresent case. Fora first approximation inconvenient to normalise 258 P. A. A. Dirac. fora (o'c)in theright-hand wemnay substitute sideof(21)itsinitialvalue. This gives (Mo')eiXW(a')-W(a?)It/h ih4 (a') = a0V(a'a?) = oc?v is a constantand W (o') is the energyofthestateoc'. Hence wherev(M'oJ0) iha (c.')= a (o' - 4o) + aov (c,'c() e[W() -W(CI)]t/h-I (22Y For valuesof theock'such that W (a') differs appreciably fromW (MO), a (a') is a periodic function ofthetimewhoseamplitude is smallwhentheperturbing corresponding to thesestationary energy V is small,so thattheeigenfunctions statesarenotexcitedto anyappreciable extent. Ontheotherhand,forvalues of the cXk'such that W (a') - W (o!2) and OCa'# Oko forsome k, a (oc')increases withrespectto thetime,so thattheprobability ofthesystembeing uniformly in thestatea' at anytimeincreases withthesquareofthetime. proportionally theprobability ofthesystem Physically, beingin a statewithexactlythesame being properenergy as the initialproperenergyW (ac?)is of no importance, infinitesimal. We are interestedonly in the integralof the probability through a smallrangeofproperenergy valuesabouttheinitialproperenergy, withthe which,as we shallfind,increases linearly withthetime,in agreement ordinary probability laws. We transform fromthe variablesrl,,o2 ... oa to a setof variablesthatare arbitrary independent functions ofthea's suchthatoneofthemis theproper W,Y -'Y2. yu1. The probability at anytime energyW, say,thevariables ofthesystemlyingina stationary stateforwhicheachYk lies between yk'andto is now from the (apart normalising factor)equal Yk' + dyk' d *dy2' ... d-'l'i a (') 12 a ( (X4l X2 * - (WI, Ti' X ) - dW'. Yu-i') (23) For a timethatis largecompared withtheperiodsofthesystemwe shallfind by valuesof thatpractically the wholeof theintegralin (23) is contributed W' verycloseto WO W (a?). Put a (a') - a (W', y') and a (cl', OC2'.X -')/a(W', Yl YU-1') J (W' y ) Thenfortheintegral in (23) we find,withthehelpof(22) (provided Yk' ? Y 0 forsomek) Ja(WI, J (W', y') dW' y) 12 a?I2' J Iv (W, y' ; - 2 = WO,yO)12J (W', Y/) [ei(WWO)tIh 12J (W',y') 1a012iv (W,y'; w?,yo) [1-cos th_1l dW" 1] [ei- (W'-VO) -WO)2 ~~~~~~~~~(WI (W'-W?) t/h]/(W'-WO)2 . dWI 2 ao?12t/h. lv(W?+hx/t,y'; WO,yO)12J(WO+hx/t,y')(1-cosxv)/x2. dx, Ermtssionand Absorption(f Radliaction6. if one makes the substitution(W'-W?)t/h reducesto 21a?12t/h. v (WO,y' ; W?, - x. For large values of t this ye)12J(W0,y') =27r I ao 12tlh. 25 9 ! v (W?, (1coS X)/X2. ? y' ;W? dx 2J( ') The probabilityper unit time of a transitionto a state forwhicheach yk,lies betweenY,' and yk'+ dyk>'is thus (apart fromthe normalisingfactor) 2 y') dyl'. d2 ' ... dyu-1l, . I v (WO,y' ; W , yO)12J(WO, a0 12/h 1ao (24) associatedwiththat elementwhichis proportionalto the square of the nmatrix energy. transitionofthe perturbing To apply this resultto a simplecollisionproblem,we take the ac'sto be.the componentsof mnomentum PZ, pg, p, of the collidingelectronand the y's to be 0 and q4,the angleswhichdetermineits directionof motion. Tf,takingthe relativitychange of mrasswith velocity into accotint, we let P denote the equal to (px2 + py2+ resultantmomentumn, pz2)Y, and E the energy, equal to ofthe electron,in beingits rest-mass,we findforthe Jacobian (m2C4+P2c2)A2, jaPi uPz)- Psin 0. a(,,) E C2 Thus the J (W0,y') ofthe expression(24) has the value J(W', y') = E'P' sinn0/c2 (2) whereE' and P' referto that valuefortheenergyofthe scatteredelectroiwxtiich makesthe total energyequal the initial energyWo (i.e., to that value required by the conservationof energy). We must now interpretthe initial value of a (c'), namely,a' 8 (or'--), whichwe did not normalise. Accordingto ? 2 the wave functionin termsofthe variables c,.is b (oc')= a (oc')e siW'tohSO thatits initialvalue is ao 8 (act-_ cc0)e-iWt a0a(P.' - Px0) g (pr - p,Y) g (P' - Pz) e-iV't function* If we -usethe transformation (x'lp') =--(27rh)-3/2eiV3!Xtpx'x'/l7 rule and the transformation - 4 (x') --^ |(x'lp') + (p') dpx' dpy' dpz' wave functionin the co-ordinatesx, y, z the value we obtain forthe iniiitial e-iW'tIh a0 (22rh)1-2etxp.~P9x'Ih . The syinbol x is used for brevity to denote x, y. z. 260 P. A. Ml. Dirac. This correspondsto an initialdistributionof IaQ12(27h)3 electronsper unit volume. Since theirVeloeityis POC2/Eo, the numuber per tmittime strikinga to theirdirectioni unit smface at right-angles ofmotionis Ia 2POc2/(2uh)3 E0. Dividingthisintothe expression(24) we obtain,witlhthe help of (25), P (p'; p) 2 (26) sin 0' dO'db'. 4n2 (2nh)2 E' This is the efectivearea that in-st be hit by an electronin orderthatit shall d withthe energyF. This result be scatteredin the solid angle sin 0dO' d+' differsby the factor(27rh)2/2mE'. P'/POfromBorn's.* The necessityforthe factorP'/P' in (26) could have beei predictedfrointhe prin ipie of detailed betwecenthe directand balancing,as the factor jv (p'; pO) 12 iS symnmetrica-l Teverseprocess s.t ? 6. Applicationto Liyht-Quancota. We shall now apply the theoryof ? 4 to the case whet the systemsof the assembly are light-quanta,the theorybeing applicable to this case since lightquanta obey the Einstein-Bjosestatisticsand have no muttualitteraction. A is in a stationarystate whenit is movingwithconstantmomenlight-quantu-m tum in a straightline. Thus a stationarystate r is fixedby tthethreecomponentsof momentumof the iqliht-quantum and a variable that specifiesits state of polarisationv.We shall workon the ass-umption that thereare a finite n mberof thesestationarystates,lyingveryclose to one another,as it would be inconvenientto use continuousranges. The interactionofthe light-quanta with an atomic systemwill be describedby a Hamiltonianof the form(20), in which Hp (J) is the Hamiltonian for the atomic system alone, and the coefficients v., are forthe presentunknown. We shall show that this fort forthe Hamiltonian,with the v,, arbitrary,leads to Einstein's laws forthe emissionand absorptionofradiation. The light-quant-iium has the pecuLliaritythat it apparenLtly ceases to exist its wheniit is in one of its stationarystates,narnely,the zero state, in wvhich momentum,and therefore also its energy,are zero. WVen , light-quantumn is absorbedit can be consideredto jumnpinto this zero state, and wvhenone is the zero state to one -inwlich it is emittedit can be consideredto j'umpfrom-L * In a more recent paper ('Nachr. Gesell. d. Wiss.,' Gottingen,p. 146 (1926)) Born has obtained a result in agreementwith that of the prese it paper fornon-relativitymechanics, 'by using an interpretation of the analysis based on the conservation theorems. I amn indebted to Prof. N. Bohr for seeing an advance copy of this work. t See Klei and Rosseland, 'Z. f. Physik,' vol. 4, .p 46, equation (4) (192l1). Emissionand Absorptionof Radiation. 261 inevidence, so thatit appears tohavebeencreated.Sincethere physically is. oflight-quanta nolimittothenumber thatmaybecreated inthisway,wemust thattherearean infinite oflight-quanta number inthezerostate,so, suppose thattheN0 oftheHamiltonian (20)is infinite. We mastnowhave0G,the toN., a constant, variable canonically conjugate since 00= aF/aNo WWo+ termsinlvolving No0 or (No + Ir! and W0is zero. In orderthattheHamiltonian (20) mayremain finite it is, forthe coefficients necessary to be infinitely small. We shallSeuppose vo0. v,.O, thattheyareinfinitely smallin sucha way as to makevToNo! andvorNo inorder thatthetransition finite, probability coefficients maybefinite.Thus. weput vrO (No + 1)Ar6 -i yr vo.NOeio/Il - 12*.. where vrandvr*arefinite and conjugate imaginaries. Wemayconsider the onlyof theJ's and w's of theatomicsystem, since v, and v* to be functions e-ioo'h and N0oeiOoI arepractically theirfactors therate (N0 - 1)-2constants, ofchangeofNobeingverysmallcompared withNo. TheHamiltonian (20), nowbecomes F (J)+ ErWrN+ rEr o[v,NA 4/+ v,*(N.1)r-H e-i&/l HPp + Er -0'Es o vrsNr(Ns + 1 -8 rs) et -Gi/h (27); The probabilityof a transitionin which a light-quantumin the state r is. absorbedis proportional to the square ofthe modulusofthat matrixelementof the Hamiltonianwhich refersto this transition. This matrix elementmustl come fromthe term v7NIe"OIhin the Hamiltonian, and must thereforebe proportionalto Nrt'whereN,' is the numberof light-quantain state r before. the process. The probabilityof the absorptionprocess is thus proportional to Nr'. In the same way the probabilityof a light-quantumin state r being emittedis proportionalto (Nr'+ 1), and the probabilityof a light-quantum in) state r being scatteredinto state s is proportionalto Nr'(N8'+ 1). Radiative processesof the more generaltype consideredby Einstein and Ehrenfest,tin whichmorethan one light-quantum take part sininltaneously, are not allowed on the presenttheory. To establisha connectionbetweenthe numberoflight-quantaper stationary state and the intensityof the radiation, we consideran enclosureof finite volume, A say, containingthe radiation. The number of stationarystates for light-quantaof a given type of polarisationw:hosefrequencylies in the t 'Z. f. Physik,' vol. 19, p. 301 (1923). 262 P. A. M. i)irac. range v. to v,- dVranidwhosedirectionof notion lies in the solid anlgledcx. The energy aboutthe directionof motionforstate r willnowbe Avrl2dv.do,/c3. of the light-quantain these stationarystatesis tlhusN' 2nhv,..Av2dv,dCw)r/c3. whereIJ is the intensityper unit frequency This inust equal AcmJ,dvrdor, about the state r. Hence ra:ngeof the radia-tioni 1 - Nr (27h)v7/C2, (28) to I. and (N' --?1) is proportionalto T2+ (2Ah)v 3/c2. so t'llmt NJ. is proportional to We thus obtaiinthat the probabilityof an absorptionprocessis proportional of an emission -er unit trequency range, and that It.,thtencident intermsity process is proportionalto 1,+ (2nh)v3/C2, which are just Einstein's laws. a light-quantutm ofa processin wNThich is scattered In thesamewaytheprobability a state 'rto a state s is proportio1alto 1, [LI+ (27ch)V73/c2], whichis Pauli's fro:m ofradiationby a meiectron.t' law forthe sea-ttering for EiiiissionantdAbsorption. ? 7. The Probability Coeficietlds fromnthewave We,shallnowconsidertheinteractionofan atomand radiationi and point of view. We resolve the radiatior into its Fourier comiponents, is f-inite. Let each niumber yery but be their component large suLpposethat labelledby a suffixr, and supposethereare a. componentsassociated witlhthe type of polarisation per unit solid ai-gle per -unitfreradiation of a definiite quencyrangeabolt the componentr. Each componentr can be describedby a vector potential 1c, chosemso as to inake the scalar potential zero. The termto be added to the flamiltonianwill now be, accordingto perturbation the classicaltheorywithneglectof relativitymechanics,C-41X. r. X,rwhereXT is the componentof the total polarisationof the atom in the directionof K2, whichis the directionof the electricvectorof the componentr. nical We can, as explainedin ? 1, supposethe fieldto be describedbythe ca mo varialblesNr, 02,of whichN, is the numberofquanta of energyof the componentir,anad0, is its canonicallyconijugatephase, equLalto 27chv,timesthe 0r of ? 1. We shall now have c, ar cos 0,/h,where ar iS tne amplitudeof 97 whichcan be connectedwithN, as follows:-The fow of energyper uinit KIC areta per unit time forthe componentr is 17rc1 a 2Vr2. Hence the intensity - Llie ratio of stimtulatedto spontaneous emission in the present theoryis just twice its value in Einstein's. This is because in the present theory either polarised component of the incident radiation can stimulate only radiation polarised in the same way, while in applies also Einstein's the two polarised components are treated together. This remnark to the scatteringprocess. t Pauli, 'Z. f. Physik,' vol. 18, p. 272 (1923). Emissionand Absorptionof Radiationt. 263 pet unit frequfency range of the radiationin the neighbourhoodof the coma2 2. Coinparingthis with equation (28), we obtain 2 (hvr/cajr)`Nj)and hence a,, ponent r is T2227r-1 K2.= 2 (hvr/5cc)` N37Cos Or/ht. The T:amiltonian forthe wholesystemnof atom plus radiationwould now be, accordingto the classical tlleory, F Hp (J) -Lr (27ihVr)N, + 2c1 , (hv,/Cn) XrNr!cos 0,/h, (29) whereH, (J) is the Halmiltonianforthe atomlalone. On the quanturm theory we must make the variablesNr anid 01 canionicalq- iumberslike the variables We mnst nlow replace the N7, cos 0,/hin (29) Jk, w1kthat describethe atoi. by the real q-nuimber 6i rl + 6---i0rjhN rlh N- t (.4 -1- Ci9rII2} ~iNfiNj e'Io+~zrl2Nj-1~{ e%f e (N2. ,sothat the liamiltonian(29) becomnes XF t1P (2 ) + + hA C ES (2J;71hVr) N2.X >2;r (V)r/(Tr)~ 2'X {Nq.gi8 I + (Nr+ v. = v.*-h (h2/Gr)' cvr i!t}. (30) This is of the form(27), -with and )~ C Xr (31) (r, s X0). 0 The wave pointofviewis thusconsistentwiththe light-quantunm point ofview and gives values for the unkniowninteractioncoefficientvrs in the lightquantum theory. These val-uesare not such as would enable one to express the interactionenergyas an algebraic functionof canonicalvariables. Since the wave theorygives vr.s. 0 forr,as 0, it wouldseemto showthat thereare no directscatteringprocesses,b-utthis may be due to an incompletenessin the presentwave theory. We shall nowshowthat the Hamiltonian(30) leads to the correctexpressions forEinsteirn's A's and B's. We mnust firstmodifyslightlythe anal is of ?5 so as to applyto thecase whenithesystemhas a largenumberofdiscretestationary states instead of a continuousrange. instead of equation (21) we shall now have ih a (Vc) -V (oa'oc")a If the systemis initiallyin the state oo,we musttake the initialvalue of a (') to be 8 whichis now correctlynormalised. This gives fora firstapproximnation whichleads to ih c (z)=V ih a (x')-A ieLW('>) (cx'ot0) _-- v(c4'x? + v (o'0?.) ; W eiL '-W(? - w(ol?)] t| 1L - P. A. II. Dirac. 264 to the variablesW, y, to (22). If, as before,we transform corresponding yO) Y2 Yff-1bwe obtain(wheny' a (W'y') = v (W', y'; w _WO). TO) L[-Ce(W'-W0)t/h]/(Wf The probabilityof the system being in a state forwhich each yk equals Ykt t is Yw, Ia (W' y')12. If the stationarystates lie closetogetherand ifthe timYle is nottoo great,we can replacethissumbytheintegral(/AW)-ij' a (W'y') 2 dW', whereA.Wis theseparationbetweenthe energylevels. Evaluatingthisintegral as beforewe obtainforthe probabilityper unit timeof a transitionto a state forwhicheach YkT Yk' (32) . V (WO,y'; W0, 0) 12. 2n/hAW'I In applyingthis resulltwe can take the ys to be any set of variablesthat are independentof the total properenergyW and that togetherwith W define a stationarystate. We now returnto the problemdefinedby the Hamiltonian(30) and considler an absorptionprocessin whichthe atom jumps fromthe state JOto the state fromstate r. We talkethe variables JXwiththe absorptionofa light-quantum y' to be the variables J' of the atom togetherwith variablesthat definethe directionof imotion and state of polarisationof the absorbed quantuml,burt not its energy. The matrixelemeictv (W0, y'; W0, TO) is now h1/C C: /,g) /2Xv(J?J')Nro? ( Mr whereX)r(JJX)is theordinary(J?J')matrixelementofX. probabilityper unit time of the absorptionprocessis Hence from(32) the. 27vc hvlk (J0Xr ('JoJ) 1 2N0 z,7w jcr comesfrom To obtain the probabilityforthe processwheiithe light-quantnm anydirectionin a solidangledc, we musti.aultiplythisexpressionbythenusmber of possible directionsforthe ligbt-quantumin the solid angle d4, which is dt) grAW/27rh.This gives Tcl a X (J2J) Nr0 2 1 - X (J?J') 12I for the absorption with the help of (28). Hence the probabilitycoefficient value forEiThin the usual X with is agreement 12, Xr(J0J') process 1/272cvr2. A for in the matrix mechanics. The agreemel stein's absorptioncoefficient manner. verified the same in be coefficients may emission the Emission and Absorptionof Radciation. 265 The presenttheory,since it gives a properaccountof spontaneousemission, must presumablygive the effectof radiationreactionon the emittingsystem, and enable one to calculate the natural breadthsof spectrallines, if one can overcomethe mathematicaldifficulties involvedin the generalsolutionof the wave problemcorresponding to the Hamiltonian(30). Also the theoryenables one to understandhowit comesaboutthatthereis no violationofthelaw ofthe conservationof energywhen, say, a photo-electron is emit;tedfroman atom underthe action of extremelyweak incidentradiation. The energyof interactionofthe atom and theradiationis a q-numiber that does not commutewith the firstintegralsof the motionof the atom alone or withthe intensityof the radiation. Thus one cannot specifythis energyby a c-numberat the same timethat one specifiesthe stationarystate ofthe atom and the intensityofthe radiation by c-numbers. In particular,one cannot say that the interaction energytends to zero as the intensityof the incidentradiation tends to zero. There is thus always .an unspecifiableamount of interactionenergywhich can supplythe energyforthe photo-electron. I wouldlike to expressmy thanksto Prof.Niels Bohr forhis interestin this workand formuch friendlydiscussionabout it. Summary. The problemis treated of an assembly of similar systemssatisfyingthe Einstein-Bose statistical mechanics, which interact with another different system,a iHamiltonianfunctionbeing obtainedto describethe motion. The theoryis applied to the interactionof an assemblyof light-quantawith an ordinaryatom, and it is shownthat it gives Einstein's laws forthe emission and absorptionofradiation. waves is then considered, .The interactionof an atom with electromagnetic and it is shownthat if one takes the energiesand phases of the waves to be q-numberssatisfyingthe proper quantum conditionsinstead of c-numbers, treatthe Hamiltonianfunctiontakes the same formas in the light-quantum ment. The theoryleads to the correctexpressionsforEinstein'sA's and B's. VOL. CXIV.-A. T