CHARGE CONSERVATION I. AND GAUGE INVARIANCE* BIALYNICKI-BIRULA Institute of Theoretical Physics, Warsaw University, Warsaw, Poland Abstract Direct relation between charge conservation and gauge invariance of quantum electrodynamics is exhibited. If one insists on the charge conservation at all levels of the theoretical description one is led to consider new types of propagators - chargeconserving propagators. Charge-conserving propagators after renormalization are free of ultraviolet divergencies even in the case of massive neutral vector meson theory without the introduction of indefinite metric. One class of charge-conserving propagators is particularly useful in the description of scattering phenomena because it leads to transition amplitudes which are free of all infrared divergencies. "I remember that when someone had started to teach me about creation and annihilation operators, that this operator creates an electron, I said, "how do you create an electron? It disagrees with the conservation of charge". R. P. Feynman, Nobel Lecture. 1. Introduction Charge conservation is perhaps the only conservation law which is universallyvalid! and yet the restrictions which it imposes on all physical processes are completely ignored in the field-theoretic description of electromagnetic processes involving charged particles. The basic concept employed in this description is that of the propagator or the vacuum matrix element of the product of field operators. Propagators represent quantum probability amplitudes for creation of charges from the vacuum and destruction of charges into the vacuum, so that they automatically carry within themselves the violation of local charge conservation. Real charges are, of course, indestructible. Creation and destruction of charges never takes place in experiments and therefore some modifications are necessary to make the propagator description of electromagnetic processes more realistic. As we shall K. Maurin and R. Rqczka (eds.), Mathematical All rights reserved. Copyright © by PWN Physics and Physical Mathematics, 39-52. - Polish Scientific Publishers, Warszawa 1976 D. Reidel Publishing Company, Dordrecht - Holland 40 CHARGE I. BIALYNICKI-BIRULA see later, these modifications lead to a theory which is also formally more appealing. Abstraction and simplifications must always be introduced, it is true, in a theoretical description of every real process, but these should not be made at the expense of the fundamental laws of nature. The compliance with the requirements of local charge conservation leads to a highly welcome property of the formalism, namely, gauge invariance becomes manifest at all levels of the theoretical description. In order to account for the charge conservation in the description of all processes one must take into consideration the additional currents that flow during every experiment in which beams of charged particles are produced. These additional currents flow between the source and the detector of charged particles and we will call them the compensating currents because they compensate for the charge lost by the source and gained by the detector. Taking into account the compensating currents, instead of creation and destruction one simply has the separation and the recombination of charges. The idea of compensating currents is in full accord with the observation that the natural state of all matter (at least under terrestial conditions) is an electrically neutral state. The form of the compensating current win, of course, depend on the conditions of the experiment, but we expect this form to have only a negligible influence on the results of the experiment, provided these currents flow far from the region in which the interaction of particles occurs. Compensating currents can be described at various levels of sophistication, ranging from classical mechanics to quantum field theory. In this paper we shall adopt the simplest approach assuming that all dynamical aspects of the compensating current can be neglected so that the current can be described by a set of four given c-number functions JlJ,(z) of space-time coordinates. Such an external current can be considered to be a part of the apparatus used for the initial preparation and the final measurement of the quantum system. Even this greatly simplified model is quite sufficient to produce renormalizability of neutral vector meson theory in the physical Hilbert space and the disappearance of all infrared divergencies. In what follows we shall restrict ourselves, for definiteness, to the study of electrons interacting with the electromagnetic field (pure CONSERVATION AND GAUGE INVARIANCE 41 quantum electrodynamics), but our method can be applied equally well to the description of processes involving other charged particles. Many of the ideas presented in Sees. 2-4 are not new. They can be found in Brandeis Lectures by Johnson and in the book by Schwinger [11], [15] and also in earlier works of the present author [2], [3], [5]. We give here a unified derivation which is best suited for the discussion of Sec. 5. We found it more convenient to construct quantum electrodynamics, with charge conservation built in, in two steps. In the first step (Sec. 2) we introduce the quantized electron field operators tp(x) and v;(x) but we treat the electromagnetic field as an external field, described by four functions .91iz) of space-time coordinates. In the second step (Sec. 3) we quantize the electromagnetic field with the help of a method first introduced by Feynman, In Sec. 4 we shall prove the existence of a very simple relationship between the new charge-conserving, gauge-invariant propagators and the old charge-nonconserving, gauge-dependent propagators. In Sec. 5 we discuss the role of charge-conserving propagators in the removal of ultraviolet and infrared divergencies. 2. External Electromagnetic Field Let us consider n electrons negatons and/or positons moving in a prescribed electromagnetic field which vanishes in the remote past and in the distant future. The vacuum matrix element·of the time-ordered product of electron field operators (1) T[XI' ... , Xn,Yn, = (tJ°utl T(tp(xI), ""Ylld] ... , tp(xn)1j1(Yn), ... , 1j1(YI) )tJin) has the interpretation of the probability amplitude of the process in which negatons first are being created from the vacuum at the points Yt, ... , Yn or positons created at the points Xl, ... , Xn and subsequently annihilated at the points Xl, ... , x; or positons annihilated at the points Yl, ... , Yn' The state vectors tJin and tJout describe respectively the state which is the vacuum state in the past and in the future. In the simplified version of the theory I1-0wunder consideration, the vacuum matrix element (1) can be expressed ([8], [13], [14]) as a sum t! I 42 I. BIALYNICKI-BIRULA CHARGE of products of the vacuum to vacuum transition amplitude C[d] and one-electron Feynman propagators KF[x, Yld] (2) T[XI' ... , X,,, y", ... , Ylld] = (-i)"C[d]KF[XI' .. ·,Ylld], where (3) KF[XI, ... , X,,, Y", ... , Ylld] L: == epKF[xt.Yitld], ... ,KF[x"'Yi"ld]. permi The introduction of an external current changes the probability amplitude T[xl> ... , x,,, y", ... , Ylld] only by the following phase factor (4) exp{-i~d4zdl'(z)JI'(z)}. The compensating current, which we want to use in formula (4), in the case of n electrons depends not only on z but also on the coordinates Xl, ... , X" and YI, ... , y" at which electrons appear and/or disappear. We shall indicate this dependence explicitly by writing JI'(z; Xl, ... , X",Yl, ... ,y,,). The conservation of charge requires that the compensating current satisfies the following inhomogeneous continuity equation (5) GI'JI'(z; Xl, ... , X"' Yl, ... , y,,) = e L (t5(4)(Z-Xi)- O(4)(Z-Yi»). + GA] = exp( -ieA(x») KF[x, yld = ••• , Xn, y", ... , Ylld, <P[x, yld] == -e ~ d;"d,,(;), y represents a special case of our more general expression: ~d4zd p(z)JI'(z; X, y). ... ,xn,y", X"l, •.. , Xn, Yl, ... , y,,)} X ···'Ylld]. Even though the phase factor (4) has no effect on the probabilities when external current is not dynamically coupled to electrons, it makes the charge conserving propagator (6) gauge invariant under gauge transformations of the vector potential of the external field --+ dl'(z) (11) JIl(z; x, y) = -e ~ d;"t5(4)(Z-;) y and then check that indeed J" obeys the right equation: J] (-i)"exp{ -i ~d4zdp(z)J:(z; dl'(z) KF [x, Yld]exp(ieA(y». x xC[d]KF[Xl, (7) 43 In order to see this it suffices to make the following identification The n-electron propagator in the presence of a compensating current will be denoted by T[Xl, ... , X"' Yo, ... , Ylld, J], T[x1, INVARIANCE x (9) i=l (6) GAUGE The gauge invariance is such an appealing theoretical feature, after all, all measurable quantities must be gauge invariant, that we believe that the modified propagators (6) are much more suitable for the description of real processes than ordinary propagators, which are always gauge dependent. The idea of introducing phase factors to construct various gaugeinvariant quantities from gauge-dependent object is an old one. It goes back to Dirac [6], [7] and Heisenberg (10)2who were the first to realize the importance of line integrals in the construction of physical quantities. A line integral <P[x, yld1 of the vector-potential evaluated along a certain trajectory in space-time, (10) I!""'n ,,.. AND C[d], from eq. (5) and from the following transformation property of KF'S under the change of gauge (8) ... ,X",y", CONSERVATION + G"A(z). This gauge invariance of T[Xl' ... , X"' Y", ... , Ylld, J] follows from the gauge invariance of the vacuum to vacuum transition amplitude (12) G"J"(z; x,y) = e(t5(4)(z-x)_!5(4)(z-y»). For n electrons one must introduce n trajectories. Extensive use of line integrals was made in the gauge-invariant formulation of quantum electrodynamics given by Mandelstam [12]. 3. Quantized Electromagnetic Field The transition from the external field case (first step) to the quantized field case (second step) can be accomplished [9] by expanding the external field propagators into power series in external potentials, selecting 44 CHARGE I. BIAL YNICKI-BIRULA pairs of potential vectors in all possible ways and replacing each pair, say dll(z) and d.(z'), by the Feynman photon propagator D:.(z-z'), , (13) == -gll.DF(z-z'), D:.(z-z') multiplied by -i. The full n-electron propagator, with all the effects of interaction via photon exchanges included, will be denoted by I[x 1, ..• , x,; Yn, ... , Ylld, J]. It can be written in the following compact form with the use of functional derivatives (14) I[xl, ... ,Xn,Yn'''',Ylld,J] = exp(~i 15) \ J ~ ~d D F ~ ~d shortened D:.(z-z') \d4 \d4 , = J zJ Z - 4. Equivalence ~ F ( ~dll(z) Dil• z- z') ~ -.-.,,,. -+ D:.(z-z')+ °llf.(z-z')+ ~ all ~diz) 0 = Theorem (19) 1'[Xl' ... , Xn, Yn, ... , hid, = (-i)"exp x exp{ X J] { - i ~d4zd Il(Z)JIl(Z; Xl, ... , Xn,Yn, ... , Yl)} x ;i H~~-if) DF( ~~ -if)} C[d] x KF[Xl, ... , Xn, YII' ... , Ylld]. Let us now choose the simplest solution of eq. (5) for the compensating current, namely the one which can be expressed in terms of one vector function a"(z) of space-time coordinates II 1'[Xl, ... ,XII,Yn, "',Ylld,J] which follows directly from the invariance transformations of the potential (18) 45 We will show in this section, that under certain natural assumptions on the compensating current, our current-dependent but gauge-invariant propagators become numerically equal to the gauge-dependent but currentindependent propagators of the standard theory" evaluated in a particular gauge. To this end we shall rewrite the current-dependent propagator (6) in the form o.fiz'-z), where fll(z) is any vector function of space-time coordinates." The f-dependent gauge terms in the photon propagator do not contribute to the charge-conserving propagators I[xl>"" Xn, Yn, ... , Yll .91 , J] as seen from the equation (17) INVARIANCE It thus follows that the charge-conserving propagators are invariant under two types of gauge transformations: of the external potential and of the free photon propagator. Free photon propagator in any gauge therefore can be used in the exponential operation in formula (14) leading to the same final result. The Feynman gauge, of course, is the simplest and for that reason we have chosen it in formula (13). It should be stressed, however, that our charge-conserving propagators depend on the form of the compensating current. This compensating current should in principle be appropriately chosen for each experimental situation. notation We will not consider here mixed electron-photon propagators, because the presence of real photons has no bearing on the question of charge conservation. Owing to our inclusion of the compensating currents, full electron propagators I[xl,"" Xn, Yn, ... , Ylld, 1] are invariant under the gauge transformations (7) of the external potential. This invariance in turn leads to the invariance under gauge transformations of the propagator (16) AND GAUGE ~ ~~ DF ~~ )T[Xl> ... , XII' YII, ""Ylld,J], where we have used the following ( CONSERVATION I[ ...ld+o~A, = - ~d4Z~A(z) all under the infinitesimal 1]-1'[ ... 1.91, J] JLA~ = 0, 1_, 1'[... 1.91,1]. (20) gauge JIl(Z;Xl, ""xII,Yn, ""Yl) =e L(all(z-Xi)-all(z-Yi»), i=l where a" obeys the equation (21) 0llall(z) = ~(4)(Z). The current described by all(z) has one point-like source of unit strength. The physical meaning of the formula (20) for the compensating current CHARGE I. BIALYNICKI-BIRULA 46 is that we have assumed the same universal mechanism for the production and the capture of all electrons by the apparatus. In order to transform the r.h.s. of eq. (19) to a more useful form we will use the following property of the Feynman n-electron propagators 15 (22) 8"(ld,,(z) KF[Xl, .'" x«. Yn, ... , hid] n = ie ~ (d(4)(z-xi)-d(4)(z-Yi»)KF[xl, ,.. ,Xn,Yn, .. "ydd], 1=1 which follows from the transformation law (8) of KF under the change of gauge (cf. also formula (18». From (20), (21) and (22) we obtain (23) L~ n -ie d4zd,.(Z)(a~(z-Xi)"':"ai.(Z-Yi»} x i=l x exp{ x 8A x8 Q •. ;i ~ d4z1d4Z2('dd~(Z1) A - ~ d4wla"(zl -Wl) x (l,.. ,) D=.(Z1-Z2) ( dd~(Z?) - ~ d4w2a:(Z2 -W2) x (ld!!~W2»)}C[d]KF[Xl, ... , Xn, Yn, .. " Ylld]. Choosing the gauge for the external potential in such a way that it obeys the condition (24) ~d4Zdiz)aI'(z-x) = 0 we obtain finally (25) I[x1, = ',., Xn, Yn, ... , hid, J] . (l~d d) _DF[a]_ (-l)nexp xKF[Xl, 2i (ld ... , Xn,Yn, ''',Ylld], AND GAUGE INVARIANCE 47 where the new photon propagator DF [a] is defined as follows (26) D=.[Zl - z2la] == ~ d4wl d4w2 X (d!(l(4)(zl -Wl)- 8,.aA(Zl-Wl))( - gAQDF(Wl- W2»)x x (d~d(4)(z2 - W2)- 8.a!!(Z2- w2»). X This propagator differs from the Feynman propagator (13) only by a gauge transformation (16). This completes the proof of the following equivalence theorem: Under the choice (20) of the compensating current the charge-con: serving, gauge-invariant propagator (14) becomes equal to the gaugedependent propagator of the standard theory evaluated in a special gauge which is defined by eqs. (24), (25) and (26). 5. Ultraviolet and Infrared Divergencies T[Xl, ... , Xn, Yn, .. " Y11d, J] = (-i)nexp{ CONSERVATION (ld C[d] x Our motivation for introducing charge-conserved propagators was the desire to restore local charge conservation and bring with it full gauge invariance of all objects considered in the theory. We will show now / that these new propagators have additional attractive features which only:indirectly follow from local charge conservation. The first property which we would like to discuss is the renormalizability of charge conserving propagators in electrodynamics with massive photons." It is known" that in the standard formulation massive electrodynamics can be renormalized only at the expense of enlarging the space of state vectors by including unphysical state vectors with zero and negative norms. More precisely the off-mass-shell values of ordinary propagators can not be made finite by renormalization in the Hilbert space of physical state vectors even though all renormalized S-matrix elements between physical states are finite. One usually does not view this fact to be a serious deficiency because off-mass-shell values of the propagators being gauge-dependent are not considered to have any physical significance. The nonrenormalizability of propagators in massive electrodynamics can be directly linked to the presence of the term ft-28,.8.i1F in the massive photon propagator. The presence of such terms increases the degree of ultraviolet divergencies in momentum-space integrals making the number of primitively divergent diagrams growing with the increas- I. BIALYNICKI-BIRULA 48 ing order of perturbation expansion. Charge-conserving propagators are renormalizable because all dangerous terms cancel in every order of perturbation theory. This cancellation follows directly from the property (17) of these propagators. In order to see this we will write the expression for the propagators in the massive theory in the form. (14) used before in massless electrodynamics (27) I[X1' "',Xn,Yn, where JF is chosen in the Feynman gauge, i.e. = J!.(z-z') .nx, yld , J] = - iexp ( ;i ~ b~ DF b~) C[d] exp ( - i ~ .91. J)KF[X ,yld]. Infrared divergencies arise from the interaction of charged particles with soft photons. In formula (29) this interaction is described by the soft part of the expression: (30) r b J bd D F b bd' We shall show now that soft photons are decoupled from the charges if the compensating currents do not flow to infinity. The strength of this coupling of the system to photons with four-momentum k is determined by the functional derivative: (31) b bsi,,(k) CONSERVATION AND GAUGE INVARIANCE 49 = 0, we obtain 32 () b = ~ d4z-- b bsi,,(O) bdiz) For one-electron propagator KF[x, Yld] we have the relation (33) ~ KFlx, Yld] bd,,(O) = -ie(x-y)"KF[x, yld], which is proved in the Appendix. The compensating current contributes the following term to the derivative (31) of (29) (34) ."! ._,exp( -i~ d.J). -g"oJF(z-z'). Thus the ultraviolet divergencies in the perturbation expansion of chargeconserving propagators in the theory with massive photons are not any worse than the divergencies of standard propagators in the massless theory. The second property of charge conserving propagators is the disappearance of infrared divergencies, provided the compensating current is properly chosen. For simplicity we shall consider one-electron propagator 1'[x, yld , J] in the presence of a compensating current (29) At k ""Y1Id,J] = exp( ~i ~ b~ JF b~ )T[X1, ... ,x", Yn, ... , Y11d, J], (28) CHARGE (' 4 = Jd ik ze .: b bdiz) = +i ~d4zJ"(z; x, y)exp( =i ~.91. J). For the line current (11) we obtain x (35) ~drze ~d~"b(4)(Z_~) = e(x- y)" y and the same result will hold for any superposition, with total charge e, of line currents flowing between the points x and y. On the other hand each current flowing in a bounded region of space-time between the points x and y can be regarded to be a superposition (perhaps continuous) of line currents. Therefore, the electron propagator in the pres. ence of such compensating currents decouples from photons with zero four-momentum: (36) "!_ (exp(-i~d'J)KF[X'YId])=O. This ends the proof of the disappearance of infrared divergencies since the vacuum polarization described by C[d] in the formula (29) is known not to contribute infrared-infinite terms. Appendix Our starting point will be the inhomogeneous Dirac equation obeyed by KF (A.1) [m-iy· o+ey· d]KF[x, Yld] = b(4)(X_Y). CHARGE I. BIALYNICKI-BIRULA 50 After differentiating functionally both sides with respect to .!III'(z) and integrating over z we obtain (A.2) [m-iy· 8+ey' = -eyI'KF[x, .!II(x)] ~ d4z LAIJ E-\ KF[x, YI.!ll] YI.!ll]· The same equation and the same (Feynman) boundary conditions are satisfied by -ie(x-y)I'KF[X, [6] [7] [8] [9] [10] [11] YI.!ll] and this establishes the equality (33). [12] [13] [14J [15] [16] [17] References * Earlier versions of this paper appeared as University of Pittsburgh preprints: Charge-Conserving Propagators (September 1972), Charge-Conserving Propagators and Ultraviolet and Infrared Divergencies (June 1973). This work has been partially supported by the NSF Grant No GF 36217. 1 Charge can also mean the barionic or leptonic charge. Energy, momentum and angular momentum conservation laws are not so universally valid because these quantities in general can not be unambiguously defined in the presence of gravitation. 2 The Dirac-Heisenberg line-integral method was rediscovered many years later and was unjustly given the name of later discoverers. 3 It is often convenient to impose restrictions of the functions fp(z) such that the photon propagators will obey proper boundary conditions (cf. [4]). 4 An extensive study of gauge-dependent or gauge-covariant propagators was made by B. Zumino [18]. 5 More detailed discussion of renormalizability is contained in the Ph. D. dissertation of A. Bechler [1]. 6 See for example [16], [17]. 7 There is a very lucid discussion of this question in a recent paper by D. Zwanziger [19]. Bibliography [1] A. Bechler, Acta Physica Polonica B5 (1974) 353, 453, 645. [2] I. Bialynicki-Birula, in: Proceedings of Seminar on Unified Theories of Elementary Particles, University of Rochester, Rochester, N.Y. (1963) 274. [3] I. Bialynicki-Birula, Bull. Acad. Polan. Sci. III 11 (1963) 483. [4] I. Bialynicki-Birula, Phys. Rev. D2 (1970) 2877. [5] I. Bialynicki-Birula and Z. Bialynicka-Birula, Quantum Electrodynamics, PWN and Pergamon Press, Warszawa-Oxford (1975) 352. a. [18] [19] CONSERVATION AND GAUGE INVARIANCE 51 P. A. M. Dirac, Proc. Cambro Phil. Soc. 30 (1934) 150. P. A. M. Dirac, Canadian Journal of Physics 33 (1955) 650. R. P. Feynman, Phys. Rev. 76 (1949) 76. R. P. Feynman, Phys, Rev. 80 (1950) 440. W. Heisenberg, Z. Phys, 80 (1934) 209. K. Johnson, in: Brandeis Summer Institute in Theoretical Physics 1964, vol. 2, Prentice Hall Inc., Englewood Cliffs N.J. (1965) 28. S. Mandelstam, Ann. of Physics 19 (1962) 1. J. Schwinger, Phys. Rev. 92 (1953) 1283. J. Schwinger, Phys. Rev. 93 (1954) 615. J. Schwinger, Particles, Sources, and Fields, Addison-Wesley, Reading, Mass. (1970) 254. W. Zimmermann, Commun. Math. Phys. 8 (1968) 66. W. Zimmermann, in: 1970 Brandeis University Summer Institute in Theoretical Physics, vol. 1, The M.LT. Press, Cambridge, Mass. (1970) 552. B. Zumino, J. Math. Phys. 1 (1960) 1. D. Zwanziger, Phys. Rev. D7 (1973) 1082.