Académique Documents
Professionnel Documents
Culture Documents
btn ty> th (1134) Comparing with eq. (11.28), we see that this expression is just the matrix element of the time or- dered product T(Q(11)Q't2)..0(¢,)) between the ground state at f and the ground state at” Therefore the expression (11.34) is the ground state expectation value of a time-ordered product of co-ordinates. In field theory, these will become the Green’s functions. We shall indicate how WIJ ean be evaluated from eq. (11.33). To within a multiplicative face tor independent of J WUl~lim «719. TY Tomine or , wu ~ tim Sida oli EL gel. 2) + (Og) (11.35) rie In field theoretic applications, the multiplicative factors independent of J never matters, and we are allowed to be cavalier about it. From eq. (11.34) and the remark following it, we have TQ). QE_Wo ~ liom f day.n.d4,6Q", Ta Tenis DANG C12, 129420 4ylIye fy! Q. TD, where t > f2.. > fq, and (p denotes the ground state expectation value. Let us consider con- ining (7(Q(¢))...QU,) in f, analytically, from real to imaginary values f, = —ir,. Since sof Bet MASE, SMe q. tig’33) 5) )sulculation is posed as a problem, wi EES. Abers and BLM. Lee, Ganse theories a er -ntD, i depends on f — 1" only through e: | | | tte analytic continuation is effected by writing | cq tla’, where es (rt 1 Yat 1). » Thus the analytic continuation of (T(Q(t,)...QCt,))e may be written as OC)-LEeya-iey~ tin fidalatedate atrero| J S tae(a ist) ‘This suggests going over to an imaginary time, or Euclidean, formulation and defining We = flag) oxo| S ar[Lac(a The boundary condition to be imposed on (11,36) is that q approaches some constants as t + t=. Itis convenient, but not necessary, to take these constants to be zero. The connection between ' W{d} and Wy, (1 is that t soa} 5 (11.36) t 1 8"WeL) Well 51Cr)..51E,) aL owt aed (11.37) = o.r ‘where analytic continisation is implied on the right-hand side, Equation (11.37) is manifestly in- Es ‘dependent of the overall normalizations of W{J] and W, {J} which are independent of J. Finally, in order to illustrate the formal discussion, and especially the Euclidean formulation, wwe discuss a simple example. Consider a simple harmonic oscillator in one dimension, whose a Legrangian is, 14, )=3(q? = wg). (11.38) ‘The transformation matrix in the presence of external source J can be computed from eq. ego «im fF (11.29): » dg(t)) g Vine? (i ar{L(q(a), Go) + G7) (11.39) with the boundary condition q(t" ', q(t) = q. The integral can be worked out explicitly. The h enough hints, in Feynman and Hibbs, “Quantum Mechanics SERRE NE ETE TE ewee ES. Abors end BW. Lee, Gauge theories and Path Integrals”, p. 64. The answer is Ag’. 1g. 1Y = [eo/2ni sin wo TF)! expliQ’. #4. 0} 1.40 where Tsf-4 and Og’. t.g.0 2 [(q? + q"?) cos w T — 249") GOOF ino TTT 2a@ q He + q r a San f Jer) sin cae ~ dr + J Jeonin wot’ = nar 1 BoA ~ Senor) & f some) sin ee ~ nsin exo = ar aan We leave the derivation of eq. (11.41) as an exercise. The quantity W[J] defined in the remark following (11.33) is WIA] = 60,110, 07 = fag’ o8(a', €f da ba, 0a’, t'1g, #¥ a ) where “0” in (11.42) means the ground state, not the state with eigenvalue 0 for the coordinate; 0 is the ground stete wave function of the simple harmonic oscillator: bold. 7) = (wo/n)'* exp(—j eq? Jexp(~i jor) (11.43) $0 that the integrals over q and q’ are just Gaussian integrals. The result is oy 0 t Gof i , ~ (0, 110,07 = eo if eo fornor[s, exp (-io(o = nije} : (11.44) We will make the result more general by extending the limits on the integrals from —» to 4, “Thus, if we are interested in the effect on the oscillator of the force term for just the period ¢ to 1’, we may restrict J(7) to vanish outside of this interval. Finally, we shall write eq. (111.44) as [see, R.P. Feynman, Phys. Rev. 80 (1950) 440} iv. 7 wis) so 5 Jf a f do J(r) D,(r ~ (a) (11.45) where XG) (0(De-# + O(- 1oier) (11.46) ies40) An 42) ite; 43) 44) to 45) 46) / the term (5L,/5q)(dy/ar) is FS. Abersond BW. Lee, Gauge theories 6 Notice that yt Swi is? ayes - | =F a yy 47 D-H 51E08H0)| p57 KOBE) "| aia7) In the Euclidean formulation (11.36), we have Wet) = faq) exp {8,1} ae) where SUI= f dr bglr), bg = H(dg/dr)? + 302g? ~ Sn)q(a). 1.49) We expand q(7) around q(r9), writing 4(r) = qo(r) + (7), and then expand 5, in powers of y, SQ =Selq* f [ f(y 2 + art 4(2) +2 >| ar 11,50) we laeutt LLG) Fr] aso We wish to choose go(r) so that the term in (11.50) linear in y vanishes. If the boundary condition is taken to be q(r = #00) = 0, we require qo(r = +) = 0 also. Then the surface terms vanish when integrated by parts, and we require that qa(r) satisfy the classical «quations of motion: d ile bLp dr 8q(7) =o. ast (TY) g =aots) IICq is allowed to approach non-zero constants q, in the limits r+ 49, we may require qa > 0 and y+ q,. Then there will be surface terms equal to de(+=)q, in (11.50). However, from the seneral solution below it is evident that qo(¢) = 0 if qo(#) vanishes. | Now we insert eq. (11.50) into eq. (11.48) and perform the integral over paths, The term Encat in y has disappeared, so we write [Ee ey 2 Well ~ expi—Se(qa)} f TT ane) | - ‘The integration over y(7,) is just a number, independent of J, so we are left with WU] ~ exp{-S,(go)}- 1.52) Let us evaluate qo(J). From eq. (11.51) &e > (Ss ~o'] ain = 109, 1.33)0 ES. Abers.and BL: Lee, Gauge theories Define a Euclidean Green’s function Dy(r) by eo ar o] Dy) = 5(r) (11.54) with the boundary condition lim,.,., Dy(r) = 0. The solution is ay oe gw Ded=— f seh ctnaD: (LSS) and therefore alr) =~ f Delt ~ 0) (0) do. 1.56) [With other boundary conditions, the most gencral solution is (11.56) plus the generat solution to the homogeneous equation, namely, Ae-#* + Be". If qo approaches a constant at both + and ~=», A = B = 0, and it follows from (11.56) that Gol) = 0 alse.) Now in the definition (11.49) of Sy. we substitute coqo(x) from eq. (11.53), integrate by parts, to obtain, using eq. (11.56) Sx€qo)=—+ f Mdqdrdr=+ f dr f do d(r g(r - oso) as7) so that, from (11.52) Weld ~ exp {4 Jf drdo Heer - ono} (11.58) or . -5'W gL) Pe TT BIDET) | go We can get the propagator D.(t) by analytic continuation in r, by rotating counter-clockwise from real r to imaginary 1: Dt) =i Dg (it) (11.59) which yields eq. (J 1.46) immediately. Note that the functional integral in (111.36) is a well behaved Gaussian (or more precisely, Wisner-Hopf) integral. Our notation may be simplified by writing the real time, ground-state to ground-state amplitude (11.49) as mui ~ fraacresol FEC) =H" sing +4006000) | (11.603ES. Abers nd BAM Lee. Gauge thooriet n Then we can repeat the imaginary time analysis using (1.60) directly, to obtain eqs. (11.45) and (11.46) the ie in (111.60) serving to select the correct boundary contlition on the propagator: aa 7 dy expt-ivt) DEO a J Wut tie’ 55) Foc the path integra formulation of quantum theory, the basic papers are 1 PAM. Dirac, Physik. Z, Sowjetunion 3(1933) 64. 2. RP. Feynman, Rev. Mod. Phys. 20 (1948) 267, 3. RP. Feynman, Phys. Rev. 80 (1950) 440, + Theseate reprinted in 4.3, Schwinger (e.), Quantum Electtodynainics (Dover Publications, New York. 1958). iesibook i avilable on this subject: S: RP. Feynman and A.R. Hibbs, Quantura Mechanics and Path Integrals (MeGraw-ti, New York, 1965). a ‘er the path integral formulation of ea. (T1.1 1) see 5.C.Garrod, Rev. Mod. Phys. 38 (1966) 483. steako 7. RJ. Finkelstein, 3.8. Kvitsky and J.0. Mouton, Phys. Rev. D4 (1971) 2220 for an excellent review of this subject. arts, The effective action for the velocty-dependent potential wes fist obisined by 8.TD. Lee and C.N. Yang. Phys. Rev. 128 (1962) 888, | Bibfography | | | (Canonical transformations are discussed in any good book an classical mechanies. For exumple | 57) >) 8.1. Gotdstein, Classical Mechanics (Addison Wesley, Reading. Mas. 1989) ehapter 8 | ‘The problem of operator ordering ofa velocity dependent potential has been discussed in 1. T. Kawai Progress Theos. Phys 48 (1970) 2082: Kamo and Kawai, to be published; and references cited therein See ako 3) {eS De, Rew Mo, Phys. 29980 377 | 58) 1 kS. Chang dh Pye 13 97D 1, | fr dation of Strings equation ora ls-depenent pos, wih amount te dering sapere | | | assitlontan which corresponds tothe path integzal formulation This sction incorporates several useful rentark ofS. Coleman, D. Grossand S.8.T 12, Path integral formulation of field theory Physics — Where the Action Is, ‘Anonymous 59 | We have remarked that the generalization of the considerations in section 11 to many degrees of freedom is immediate. The transformation function is given by (11.24), which is a shorthand for A Mn Wot dpglty) lim TT TT @aatty TT 5 50) nw as) +44) X exp {i D ( D Paltaalty) — daly = att (pu, say) a2.y tle , 2ES. Aberzand BW, Lee, Gauge theories Eq. (12.1) can be applied to field theory. Consider a neutral scalar field 6(x). Let us subdivide space into cubes of dimension €? and label them by an integer a. We define the ath coordinate dealt) = G_h0) by (=> f Proce, ), ey where the integration is over the ath cell of dimension €®, We can also rewrite the Lagrangian as L= fd'x2> De 266.00, bal). 64250) where (0) is the average of ap(x, 1)/a¢ over the ath cell and ¢4.2, is the average value of the field in the neighboring cell « + s. The canonical momenta p, conjugate tog, are aL ake Pad) = a5 (8° 36,(0 The Hamiltonian is H= Dpba- b= De Hy, al). Ha = Mabe — Ly = KalFas bas Gas) We may now write the expression (12.1) as lim m0 Ta.) TT £ dnt) carne xexpl[i bebe {re @ = foo [Ss where we defined the momentum density conjugate to o(x, ¢) by a(x, 1) = a.L/agix, 1) Its coll average is just the #,(0) defined above. In field theory, all physical quantities are derivable from the vacuum-to-vacuum transition am- plitude in the presence of external sources. The physical vacuum is the ground state, and plays the same role as the state whose wavefunction is @9(q) in eg. (11.33) ‘This amplitude, which we shall call Iv{/], can be calculated from eq. (12.2) with a term: JePxJ(x, p(x, 0 added to the Lagrangian, in the limit f+ ©, + —e. That is tall Balt bl #OaAtj—1) dart) Canta) (rn. 5 2 ) ar] cvohf arf [res n? ie Doe, | 12.2)as am ‘s ES. Abers ond B.W. Lee. Gauge theorlet B wu) = fldor| Zan] exp] if atetnesrgce) — 900) Fick Jo9e68))] 23) ‘Theextra term fic 9? is simply a symbolic way of indicating how to rotate the time-integration contour to pick out the correct limit as indicated on the left-hand side of eq. (11.33). More on thiskater. Now it follows from eq. (11.34) and the discussion following it that 5rd BMC )BI(%2)-- In) ITP )G(%2) (pO = PE ky) 12.4) J0 where G is the n-point Green’s function, the vacuum expectation value of the time-ordered pro- duct of 1 fields. The fact that the Green’s functions may be defined by (12.4) was first dis- covered by Schwinger, and does not depend on the path-integral formula (12.3) for 1V[/]. How- ‘ever, eq, (12.3) provides not only a simple proof of (12.4), but an explicit formula for com- puting WU}. Fa, (12.4) gives the complete Green’s functions. In general, these inchide some contributions from disconnected vacuum to vacuum diagram, which are simply products of lower order Green's functions. The connected graphs are given by Cor awit Cu) WE Te pbhoD aa or, writing, WU] = exp{iZ[41 }, (12.6) Goon.) = (ir) 2 12.7) BI). BIO,) ‘The proof that the connected parts of the n-point function is given by eq. (12.7) is left as an exercise. When the Hamiltonian density takes the form (x) = PO) + (1G). VEO] (12.8) ‘The s-integrations can be carried out explicitly, and we obtain WUT ~ f db lexp ti [L260 + JO) dtx} 2.9) ‘here 2(x) is the Lagrangian density 20) = 306)? — M190), VO] When we discuss vector meson theories, the form (12.9) will be inadequate and we shail have {0 use the original form (12.3). Asan example, however, let us first consider a cese where (12.9) isapplicable,ES. Abers and BW. Lee, Gauge theories Let us concentrate, for definiteness, the case in which the Lagrangian is of the form £2,420, = 310,07? — 6") (12.10) and Ly = 249). The functional WJ] of eq. (12.9) is in general an ill-defined integral even in the “lattice” approxi- mation. Recent advances in axiomatic field theory indicate that if one can construct a well-be- haved field theory in the Euclidean space (x, ¢), obeying certain appropriate axioms, then there is.a corresponding field theory in the Minkowski space (x,, x) as the analytic continuation of the former as 7 = ix, which obeys the Wightman axioms. Thus any ambiguities should be resolved by appealing to the Euclidicity Postulare, namely that the Green's functions (12.5) are the ana- lytic continuation of those defined by the well-defined functional integral in the Euclidean field theory WU = ftastesn| Sorsar[ GE) + or +wre -24@-s6]} . Note that since -&; is bounded from below the quantity in the square bracket in the exponent above is also bounded from below. As we anticipate, the Euclidicity postulate determines the boundary conditions to be imposed on propagators. For the present problem it means that we may provide a camping factor for the functional integration by adding a term in 2,: L. > T1@yO? — wg? + eg?) as we did in eq, (12.3). First, consider the free field case: WU) = fldolexpti fotxt(00)* ~ $u%9? + Lies? +46) = lim sa 46, exe [Be De tb, Keote* Delete \] (2.11) Here, « labels space-time cells of dimension e*, and the matrix Kya is such that lim Keg = (-0? = 4? + ie) — v) fo where a + x and 6+ y ase + 0. The ¢-integrations in eq. (12.11) can be performed explicitly. We obtain tr WUT = lim — oT = lin rasa V2 where, of course, K~' is the inverse of K: Dh (Kg 6 7 wn [ He De Dev, KS] 8ES. Abers ond BW. Lee, Gage theories 7 10h 02.12) ) Ase 0, we have sroxi + 8x =y), ea so, with the definition the i SOs + Ap =), > ¢q.{12.12) may be written, in the continuum limit ¢ + 0, as (8 = uP ied gC — y) = Gr — yD, (2.13) ‘Therefore, neglecting an inessential multiplicative factor, we can write i WU] = exp(—3i fatx [atv Ae — yO) 12.44) where ) 2.15) is the Feynman propagator. Now we are ready to discuss the interacting case. Returning to eqs. (12.10) and (12.11), we write | WU ~ fldelexptifatxte, +2\¢) +401} rn 15 ) sewr[i farve,(t pr )ftasiowtifarete, +01} 18 i ~ exofifatee,(t juga) |w$0 fats for FOAL = YO. 02.16) Equation (12.16) is the basis for the Feymman-Dyson expansion of the Green's functions of this ‘theory, and when itis substituted in eq. (12.4), we obtain a formula which generates Green's functions. W{J] can be expanded in powers of 2,, for example, by simply expanding the ex- Ponential factor oobivalt fl) -3 Selma What corresponds to Wick’s theorem is simply the rule for functional differentiation’ LE NEES LLG TNT EYES. Abersand BW. Lee, Gauge theories 8 Fr Mr). The student should convince himself the rules outlined here are in fact the Feynman rules dis- cussed in the second volume of Bjorken and Drell. In fact, collateral reading of the first six section of Chapter 17 of this book is urged In order to quantize fermion fields by the method of path integrations, it is necessary to in- troduce the concept of anticommuting c-numbers. We shall forgo this though, because the incor poration of fermion fields presents no special problem in quantizing a gauge theory. Ingeneral, 2, isa function of ¢ as well as 4, and eq. (12.4) is inadequate. Just as in the one- dimensional example discussed in the preceding lecture, we shall see, the action of eq, (12.16) must then be replaced by an “effective action”, which contains a correction to the integral over the Lagrangian. In that case, the correct Feynman rules are modified, and cannot be directly read off the Lagrangian, Bibliography For the Luclidisn Field Theory ad its connection to the Minkowsky field theory, see 1.3.4. Schwinger, Pro. Nat. Acad. Se. 44 (1958) 956, 2. K. Symansi, in Proc. tater, School of Physics “Enrico Fermi” Course XLV, ed. R. Jost (Academic Pres, N.., 1968). 3.E. Nelson, Construction of Quantum Fields from Markov Files (to be published); The Free Maskor Fields (a be pablished. 4. Ostersatder and R Schrader, Phys, Rev, Letters 29 (1972) 1423. ‘The view of taking (12.1) asthe basis of quantization rather than (12.8) was frst expounded by’ $.L.D, Fedde'ev, Theoret. Math. Phys (1969) 3 Engl. rash. by Consultant Breau 1 (1968) 1}. ‘The idea tha the Green's function canbe obtained from vatition ofthe vseuumto-racum am amexterna source term i due to Sehwingee, Seo, for example 16.4, Sehwinge, Proe. Nat, Aead, Sei 37 (1951) 452. 7.D. Lurie, Particles and Fics (Interscience, New York, 1968) Chapter 10, and roferences cited therein, 13. The Yang+ field in the Coulomb gauge We wish to apply these path-integral methods to theories with gauge vector mesons. Indeed, it is in this case that the method becomes a powerful tool both to discover the correct Feynman rules and to study renormalizat the canonical Wick theorem methods become awkward. We shall study the three-component Yang-Mills field, although the generalization to other com ' Pact non-Abelian groups is immediate. In this section, we work out the canonical formalism in the Coulomb gauge, and construct the IV[J} function, starting from the basic equation (12.3). In later sections we shall study gauge-invariance and work out the Feynman niles in a more mani- festly covariant gauge. It is convenient to write out the Yang-Mills Lagrangia the first-order formulation, in which A, and F,, are treated as independent co-ordinates: Fyy° FY — 3B, “(QAP — YAH + gAH x AP). a3) Gold-face symbols, dots and crosses all refer to isovectors and operations among them; we write out the space-time vector indices explicitly.)dis. « seetio, ished) ed, an ward. com in 3). In ni vhich 13.) write ES. Abersend BLM. Let, Gate theories ” ‘The Lagrangian (13.1) is invariant under infinitesimal gauge transformations gl) > Ayla) # WC) XAG) = Fy Fy FX Eye 03.2) “The Euler-Lagrange equations, Fu = Quy ~ AA, HRA, x A, 13.3) and OF, +gA" X F,, = 0. 13.4) Equations (13.3) and (13.4) fogether are equivalent to the Euler-Lagrange equations of the sxcond-order formulation, in which 2, is written in terms of A¥ and a*A* only. Im classical field theory, one is given an initial configuration of fields in a space-tike hyperplane and then one tries to determine the fields at later times. Equations (13.3) and (13 4) can he sparated into two classes: those which specify the temporal evolutions of the fields are called equations of motion; the others are constraint equations. From (13.3) and (13.4), the equati of motion for A, and F.,, are Ay = For # (Vj + BA, X)AQ (13.5) 8,F oj =; +8A; X)Fy — 8A, X Fay (13.6) a, Next, let us determine the independent variables. Since 5.2/6(8,A,) = -F™ 3.7) F,, = -F°" are the momenta canonically conjugate to A,. Since Lis independent of 0,A°, A° does not have a conjugate momentum, and must be treated as a dependent variable. The constraint equations are Fy =2,A, — A, +8, X Ay (13.8) which defines F,, in terms of A, at equal times, and a HBA XIE eg =O 3.9) which tells us that not all the conjugate momenta F,, are independent (eq, (13.9) is analogous to V-E=0 in ordinary electrodynamics). It follows that net all the Ay can be treated as inde- pendent, and we are forced to impose a gauge condition, We choose the Coulomb gauge: Vp Ay = 0. 3.10) This is always possible because of the ge invariance of the second kind of the Lazr igs8 ES Abers ond BLM. Lee, Gauge theories Eq. (13.10) means that A must be transverse. Therefore, the longitudinal component Ft, of the canonical momentum F,, is not independent, but depends on the other degrees of freedom through the constraint equation (13.9). F4, and the transverse component FT, can be defined FucPh Fh, UR SVR. MOF =0. «3a Our task is now to express A, and Fl; in terms of the independent variables and construct the Hamiltonian, Let us write FL=-Vif, FA, ot E, ViFoi= -V7E (13.12) where E, is purely transverse. Therefore E, and the transverse components of A, are the indepen- dent variables conjugate to one-another. From the constraint eq. (13.9), we find that (W? +A, X VA = ZA, x E, 13.13) Equation (13.13) can be formally solved by introducing a Green’s function ®,, defined as the solution to (8759 +g cOPALV, DM (x, yA) = 545g (x ~ y). 13.14) Then f isa solution of (13.13) if Sy, D= a fPy DLC. y; Me tAg(. DELL, 1. (13.15) Considering D, to be an integral operator, we may write (13.15) as, f=gO.-A, x Ey ‘The function ®, has no closed form, but can be expanded in a power series in g. The first ap- proximation is just the Green’s function for the Lagrangian, and 500 te fe nix)! afte in analogy to the method for finding the Green’s function for H, +H’ where i is small and the Green's function for H,, is known. We obtain an equation for A, by taking the divergence of eq. (13.5) and using (13.10) and 3.1). 1 De, PAE 1 ecrgcy, —1 13.16 wai PARMA Bip oai Ch) (WP +E, xX VDA, = VE 3.17) which can be solved using ., since the operator in brackets is the same as in eg. (13.13): Able, 12 fly DEO, AVEO, O or A = 0.0%. (13.178) Now we construct the Hamiltonian density 9%sin the hid ea 131 the 14) fo) he 1m Ia) ) ES. Abers ond BW. Lee, Geuse tories 19 aA, =E-—-2 (3.18) H=E; a From (13.5), (13.11), (13.12) and (13.17a) we find that = =E,—V,ft(¥, +g, x) D, Vf =E, —[V,-(0, +24, x) ®VE, (13.18a) Because of (13.14), the operator in brackets operating on f'is explicitly transverse. From (13.18a) and (13.13) of (13.15), forse, forte; + aE, xX A): D,-V7f] " efoete = £-04) = f Px TEP + 09,081 and UF gy PHY — $F, (YAY ~ YAH + 2A" X AY = FFoy)? 3 (B,)? = FCB, — 94? — (By)? (13.19) where B,=3e*F,. So the Hamiltonian is A=} fatee +87 +09") 13.20) ‘The last term is like the familiar instantaneous Coulomb interaction which occurs in electro- dynamics when quantized in this gauge. Now we can write the Coulomb gauge generating functional W¢ {J} in terms of the indepen- dent co-ordinates and momenta, A, and E,, where T stands for “Transverse”, according to ea, (12.3): Wel] = fdEF) (dA Texp WO -HIR-AL Ay a32p where fis a function of Ef and AT as expressed in (13.15). [We write the source term with @ negative sign, so that the covariant version below will have +A¥J,.] The transverse field EY js difficult to compute with. Therefore we introduce a dummy variable EL by SloEF) = ftdEF }(4E*) TT 5(e4) a3. at tE, Ay — SER 4B and define three independent components E, by 1 (,-vd9,) et Fev het (3.23) aah80 ES. Aborsend BM. Lee, Gauge theories in an obvious notation. From (13.23). EL=VE, and therefore Slade? = fae 1977 40,5) where 9 is the Jacobian of the transformation from the three E, to ET, EL, and tdE,) = TT fl 1] de). To give 9a meaning, we should go back to the definition of (dE,] as a limit of an approxima- tioa with a finite number of lattice points. In the limit, 9+ ©, but in a way independent of the fields, so it is just a multiplicative factor in We [J] which doesn’t matter. The same construction works for A?. Therefore, Well = Sd, (AA,) TT 6¢¥,E, 507.4, expt fate LEA, -$ER1B2—L(V,*-Ay gh . (13.24) AL this point, we could examine (13.24) and obtain the Feynman rules in the Coulomb gauge. But they wouldn’t be covariant, and the Lorentz covariance of the S-matrix will not be obvious throughout the calculation. It isn’t useful to do calculations in the Coulomb gauge; the Coulomb gauge is the one in which the form IV{J] is most easily obtained from first principle. The S-matrix, of course, is covariant and gauge invariant, so it must be possible to find a more covariant-looking form of W,[/] than (13.24), In (13.24), f is a function of E and A given by eg. (13.15), We introduce f as a dummy variable by multiplying eq. (3.24) by the constant flansir— 20, -A,x ED (13.25) where by @ + we mean the operation in (13.15). Since (13.15) is equivalent to (13.13), we write (13.25) as flatldet m,509? +A, x VE 8 A,X E,) (13.26) where det Af, is the Jacobian of the transformation from f to (V? +8 A, X V)f.M, isa matrix in X=¥ space as well as isospin space: MEX, 3) = (V? 820 +g eEAFIV, EN — y) = V2[5S Cx — y) +g EMBL. ALOT, 66% — YQ) (13.27) where V°G(x, ») = 53(x — »). Now, eq. (13.24) becomes Wel = fIdA,1AE,1 (attaet a1, 11 819, TT 8(0,E,)8 19? +8 A, x Vp BAX El expli (Ep: Ay 4G + BE + (0,08 J,-A,}eel. (03.28)ES. Abers nd B.W. Lee, Gauge theories a Next we change variables from E, to F,), defined by For Bs Wh (13.29) Then, in (13.28), we write > (Eat) TT 6¢V, EDS CV? +g A, x VF ~8A,x E,] = (dFor] (a) TT 8(V,F,, + V°N8( Vf —g A, x F, ol (OF 5/1 (461 1] 8¢°V, Fy, +2 A, x F,,)8(0%F ~ g A, x Fy). (13.30) | ‘oxima Now we consider the {df} integration, using the last 5-function in (13.30). The Jacobian is re of th just det A*, an infinite constant which we drop (or absorb into the definition of Mo). Thus {I | {i ers Wel = fldA,]tdF det Me [1 8(V,A)5(VF,, +8 A, x F,)) i ! I). x expti fate Fay 0,4, — }F2, — H(V,A, ~ VA teAX A) -5,-A}}. 13.31) ie) {jd (13.24) ‘To obtain the exponent in (13.31), we have written in the exponent in (13.28) } i reaupe, (ER + (VAI? = (Ey — Vf = FR, | j vious | rulor) _ omitting the cross-term which vanishes upon integration over x. + Next we write the 6-function as an integral, using A, as the dummy variable: . amore aa, 1 1 by eq. TT 8,F, +A, x F,,) = (fs Sexptia, (VF — 8A, x F,)} I (13.25) | ~ flaa,lexptifarxF,, WA, x A, V/A.) (13.32) | ve 1 Finally, we write the term 4(,A, — V/A, +4 A,X A,)? In the exponent in (13.31) as | Slur exp tite, -F, — 18, (WA, VA, +A, X ADI) 3.33) | (13.26) | “hich isa standard Gaussian integral. Putting (13.33) and (13.32) into (13.31), and restricting J,, trix in | tbe zero, we obtain Well] = faa, i{4F,,14et a, TT 8¢9,A,) % expt fae (SP oy Feith Fy Fy Fy (VA/-V Ate AX A)HF, (2,A,~V,A, 4A, X A] Por Forte Ry By FF y (V,A,—VjAptg AXA) HE (2,A,~V,A,, 13.27) : = faa, } (dF, 1det afy [T A(T Aexp {i fale + J*-A,]} 13.34) _ Mere it not for the factor det Mc, (13.34) implies that one could get the Feynman rules directly ftom 2. The extra factor is analogous to the correction obtained in section 11 for a velocity-de- 13.28) Pendent potential FETE ae RE mee een: Pot iceme een eke ee SERED EIEN8 ES. Abersand B.W. Lee, Gauge theories How do we interpret det Me? From (13.27), det Mfg = det V? = dot (7421 3.35; where Log G(x, PAGO) VAC os = 55% — y) 13.36) Now det V? is an infinite constant, and det(/ + L) = exp Tr log(/ +L) exp fey dt Te L (vy a9) On Sadak Oye 80. 33) n ‘The trace inside the integral is over isospin indices only. We shall encounter Jacobiens like det Mg in the next few sections. Eq. (13.37) is a general formula for evaluating them. In our case det(4L) = ens [-2 Jorr..dy, far ofp A 01. OV),GGs1. xYT-A, 63, OV,0U0 23) LTAy, ge OV), Gye vl] : (13.38) where (7*),, = e®¢ and Tr means trace over isospin indices. Since (13.38) is a power series in the exponent, it is an effective correction in each order to the Feynman rules obtained from 2 alone. Bibliography The presentation inthis letute i 1. V.N. Popov and L.D. Fadde'v, Pert See ako 2. LD. Fadds'ey aed V.N. Popov, Phys, Leters 25B (1967) 29 to and inspzed by jon Theory for Gauge Invariant Fields, Kiev ITP report (unpublished). 14, Intuitive approach to the quan ization of gauge fields Equation (13.34) can be further simplified. We can perform the functional integration over Fi, and obtain WL = fldd,ldet Me F] aCVAWDexp Ci fa*L (x) + 3,60) AMAID a4 where 2(x) is the second-order Lagrangian: L0)= 4Q,A,— A, 2A, X AY)? Except for the factor det Mg Tl, 6(¥,A,(x)), eq. (14.1) is in the standard form for simple field theories(13.35) 13.36) 13.37) al 2X3). 13.38) to er 4.) ld BS Abersand 2: ee, Gauge Heories 8 wii ~ ftdolexp (i fats (26) +100)6(0)1 } 42) ‘The following intuitive argument due to Faddeev and Popov shows very clearly the raison d'etre for this extra factor. ‘The reason eq. (14.2) is not applicable to the gauge theory is that the quadratic part of the Losrangian SQA, ~ 3,A,)d'x = fdtxdty (B%gh — BI) — y) is singular, in the sense that the operator K#” which defines the quadratic form is singular and cannot be inverted. In fact, the operator K*” is essentially a projection operator for the transverse components of A,,. This means, in particular, that the Euclidean version of the functional integral of cq. (12.2) [see the discussion following eq. (12.10)] has no Gaussian damping factor with re- spect to the variation of the longitudinal component of A,,, and eq. (14.2) is meaningless at this elementary level even in the Euclidean formulation. More generally, the action is invariant under the gauge transformation A, ~ Ag where Af is the result of applying the element g of the gauge group G to the field A, : Ay (3) K(x, AO), Ke, y) AGL ve[a,-L+2 vex, ve)]u@). 143) To put it differently, the action is constant on the orbits of the gauge group, which are formed by all Af for fixed A,, and g ranging all over G. Thus, the path integral for the vacuum-to-vacuum anplitude W(J) diverges even in the Euclidean formulation, since for those variations of A, which are along the orbits, the action does not provide necessary damping, Faddeev and Popov pointed out that the amplitude W{J = 0] is therefore proportional to the “volume” of orbits T,dg(x), and this factor should be extracted before defining W{J].. In other words, for the gauge fields, the path integral is to be performed not over all variations of the gauge fields, but rather over distinct orbits of A, under the action of the gauge group. To implement the above idea, we shall choose a “hypersurface” in the manifold of all fields Which intersects each orbit only once. This means that if SA) = , a=1,2,..N 44) is the equation of the hypersurface, N being the dimension of the group, the equation SoA) = Must have a unique solution g for given A,,. We are going to integrate over this hypersurface, in- siead of integrating over the manifold of all fields. The conditions f,(A,) = 0 define a gauge; the Coulomb gauge f,(A,,) = V,A? is an example. Before proceeding further, let us patise here to review briefly a few simple facts about group Nepresentations. Let g, g€G. Then ge’eG, and O@)UG') = Ugg"). ‘The invariant Hurwitz, measure over the group G is an integration measure on the group space “hich is invariant in the sense thata4 . ES. Abersand B.W. Lee, Geuge theories If we parametrize Uig) in the neighborhood of the identity as U(g)=1 + iu-L + Ow), then in the neighborhood of the identity we may. always choose dg= [du,, gel Let us define the quantity 4,[A,] by Ayla) FLT dete) 11 SUALO) 47) The “naive” expression for the vacuum-to-vacuum amplitude is JSleA,) expt fa'x.209} (143) ‘We may insert the leftchand side of eq. (14.7) into the integrand of eq. (14.8) without changing anything: SElaeootaa,a,(ayt FL sicageorterrtifarxeta,eo). 149) Now, in the integrand of eq. (14.9) we can perform a gauge transformation on Ay(x) A,(x) > [A,()]*, Under the gauge transformation (14.3) the action and the metric are in= Variant, and one can verify easily from eqs. (14.5) and (14.7) that 4,fA,.] is gauge invariant: astag) = fT de) IT BUCA OD)! = f Tages cago) = fT ae" M1 ouzcag'oon = 4714, 1, or AAAg] = Oy, 1. (14.10 So, eq. (14.9) is equal to : TI gee) faa laytAy) TT SUA exp li fatx2 1A, 001} and we find that the integrand of the group integration is independent of g(x). This is the obser- vation of Fadeev and Popov, who saw that /TI, dg(.x) is simply an infinite fector independent of fields. Therefore, it can be divided out, and WIJ] may be defined as We) = fl2A,14A,1 TT BUA lexpti fatst2e + 34x) -A,001). a4atp Itis to the credit of Faddeey and Popov that they also gave the canonical derivation of eq. (13.#4) as discussed in the preceeding section, as well as this clegant argument. Before demonstrating the connection between eqs. (14.1) and (14.11) above, we shall compute A,LA,] SorryES. Abersond BW. Lee, Gauge theories 8s since the factor 4,fA,] is multipled by IT, 5{/(A,,(x))1 in eq, (14.11). it suffices to compute tne former only for A, which satisfies eq. (14.4), Now define Ml, by JAARON = FLAC + J oy DUAL Lx, y gptg) + O02). (04.12) (14.6) + -pyen from eq. (14.7) we find that © ap tad =f TTT taucosu,cagconn) = STH TMtau,cosem,u)) (471 ee A, satistying f,(A,) = 0,50 that AplA,] = det My = exp (Trin M,}. 14.13) ‘The hypersurface equation f, ~ 0 is just the gauge condition, and for the Coulomb gauge adopted in the preceeding section, we have TAA,)=V,Ap=0 (49) ad L(A EVAR + (995 — ge AE V,)uglx) + OlU) © 1 2 {gab ve 4 , Ce, May ~ 0? (B5Y ~ 2° AF) 8G — 9) ~ [A Phys aay which shows that eq. (14.1) is indeed a special case of eq. (14.11) above for f, = V,Ag The form of eq. (14.11) suggests using a wide range of gauges order than the Coulomb gauge. For the moment, we shall not ask what relations the Green's functions generated in such a gauge bear to those defined in the Coulomb gauge, but merely note the explicit form of 4, when the manifestly covariant Laudau gauge condition 4.10) FACS 0 is chosen. Equation (14.12) takes the form #AG(3) = 08A,0) + Pu + g0"(A, X u)] + Olu’) & SP sO that Mf, is given by 1 ULC, han = 2 Bap — Be apeAGO*)E — ¥) (14.15) 41) d when A, is restricted to aA, = 0. Therefore removing the trivial factor (1/g)0*, we have 3.34) i AL = det M, ~ exp{Tr Int] +L)) 4.16) Fe atiraua ancammine tear iee eeepersis kta Ne ARENAS 86 ES Abersond B.M. Lee, Gauge theores where a (% alLly, B= geyyeJ Dg — 2AG(2) —B%r — ya More explicitly we can write A, tex [2 tfers. dt THa™D pry — xa)t- AO )04D pls — x5) 1 Dy bv ~sot-axeroi} Here we have used the conventional notation (2 + iDp6x =») = 54 - y), The ie, € > 0, is chosen according to the Euclidicity postulate. The necessity of having the extra factor AyIl, 5 /(A,(x)} was first noted by Feynman. We can write eq. (14.11) as - 4 WAS = JlA,) FT BLAA,CODI exp lilSgee + fax3#0x) “A, C1) (14.18); where Bee= [etx —i Te In M,. In the case of the Landau gauge, it has been observed that the additional term in the effective ac tion -i Tr In Mf, can be viewed as arising from loops generated by a fictitious isotriplet of complex scalar fields ¢ obeying Fermi statistics, whose presence anc eractions can be described by the action S.5-- fatulaet(n) -,c6x) + g0%ct(x) A,X) x e(2)] ~ fatxaty D cf CUM LO, Yeote. That is, eq. (14.18) may be written as MUL = SOaA,1 [1 SLA,CODIf det Kael exp fits +s, + fax 20-A,co]], 4.19) In fact it is not difficult to show that the c- and ct-integrations could be carried out trivially if they were commuting c-numbers, yielding Sidct) Lac] expliS,) ~ (det My" = exp(—Tr in, }, and exp{—Tr In My} ~ exp[—Tr In +L} mexp{-TrL +4 Trl? Ltep1 an 18) 19) nee ities ereeer rman enerrerr renee oe gE ST TORRE TERA STM NN L ES. Aberond BW. Lee, Gauge theories 87 etre the terms ia the exponent may be viewed as arising from loops of the complex boson fields wif the ¢ are fermion fields, then the terms TrL" have to be multiplied byan extra ~ sign, so that wwe hove IF Tr ‘fiact det expGs,) ~ exp 7 Tr L-} Trt... + = exp(Tr In( + L)} ~ det My, ‘The Feynman rules for I, (J] of eq. (14.19) can be worked out in much the same way as we tid for a scalar field theory in section 12. The gauge boson propagator is determined from wal = ftaa,) [1 800"A,corexn[i fors| 2,4, 2A)? + JO asco) (14.20) A convenient way of computing eq. (14.20) is to write [1o[a"a,oo1 ~ Jim, exo [S: forsora,cor'} : [We have discarded ai ite constant 1, 77a.) Then we have 1 EUL = im, ftaA, exofif- fers A,e) [ ag + 980? (1 Jamo + fare P00-A,00]] im ex) fatrary JG) DEG = ys0)1,0) «a4ziy eno where the vector boson propagator Df” in this gauge is @ Qn \ pe —re= SE ete em pte f atk Dex — y= -S&e exp tik: (x — y)} and is four-dimensionally transverse. The rest of the Feynman rules can be derived as in the scalar case. They are recorded in the following fig. 14.1. In addition, the following rules must kept in mind: the ghost-ghost-vector vertex is “dotted”, the dot indicating which ghost line is differentiated; a ghost line cannot be dotted at both ends, a ghost loop carries an extra minus sign. 14.22) Beri (1422) Bibliography We have given the references to the work of Fuddeev snd Popov inthe preceeding setion; in adition, we cite |.NB. Konopleva and V.N. Popov, Kabbrovchaye Polya (Atomizdat, Moscow, 1972), in Russian ‘Qvansiaation of the gauge folds has been dlscussed sso in 2. RP. Feynman, Acta Phys. Polonies 26 (1963) 697. 3.B. DeWitt, Phys. Rev. 162 (1967) 1195, 1238. 5. Mandeltam, Phys. Rev. 175 (1968) 1580. IS-ES. Fradkin and IV. Taytin, Phys. Letters 308 (1969) 562; Phys. Rev. D2 (1970) 2641 6.8.7. Vetuman, Nock Phys. B21 (1970) 288. 7.6." Hooft, Nuel Phys. B33 (1971) 173.ES Abers end BW. Lee, Gauge theories vomets pase venners iS yrto) |-8°*[(que-Duvr/o? up rapysi" eS toa] eli aro,, peaweed | 16-7 guet(-p).on] 555 08,,) Sesto ie, dye-G5e 5p prowess [atthe 6 -2icipl fave? yaaa | py oa Fig. 14.1. Feynman rules inthe YangMils theory. Solid lines are vector mesons. Dashed fines ae scalar ghosts, 15. Equivalence of the Landaa and Coulomb gauges Formally, the S-matrix computed in the Landau gauge is the same as that computed in the Coulomb gauge, An element of the unrenormalized S-matrix is obtained from the corresponding Green's functions by removing single particle propagators corresponding to external lines, taking the Fourier transform of the resulting “amputated” Green’s function and placing external mo- menta on the mass shell. The demonstration we shall present is rigorous except that the S-matrix of a gauge theory is plagued by infrared divergences and may not even be defined. In fact this may be the reason why massless Yang-Mills particles are not seen in Nature. The point of pre- senting this demonstration is purely pedagogical: the spirit and the technique we espouse here will become useful when we discuss spontaneously broken versions of gauge theories. We shall first establish the connection between We[J} and W, [J]. Recall that [¢q, (14.1)] Wel = flddy1delAyl TT B19 Ays)exp[iSTA,I +ifers-a,} as.) where Ag = det Mc and that AAI TT agen TH sanagexy) = 1. (15.2) Inserting the left-hand side of eq. (15.2) in the integrand of the functional integration in eq. (154 we writeES. Abrsand BW. Lee, Gouge theories weld =f [det fda, 1ac(4,14, (4,1 TT 87,409) x [1 5@MALCNexPliSTA,] +i fateI-A,} We now make a gauge transformation of the integration variables A,(x): Ay(x) + L4,G)]#-? Recalling the gauge invariance of the action S, A, and the metric (da, oon we find that Weld) = {4A}, 14,1 [7 6@4A, @exPGSTA,]) x cla f TT ao SAP evi farssr ag} where Aj? is the gauge transform of A,, which satisfies a"A,, = 0, such that L-v,afe = v,fuea[t-a, + EUG Ue] UG) =0. «5.4) , In deriving eq. (15.3), we have used the fact that actA TI ase (1 sear = actAgif {IT Taxco} 0 a(viafe pMclAL Iu) oA] Ae! EAE? Now, we must find out Af? by solving eq. (15.4). It is possible to construct Af? in a power series in A, We leave it as an exercise to construct fist few terms in this expansion. For our pur- pose it suffices to note that f= (6,-1.5 ¥,)A,+0cap. The source J,, in the Coulomb gauge shall be restricted to ‘Therefore, we may write fate Akos fateh F(x; AQ) where FO AQ We can finally write down an equation for Mg in terms of W,. It is Yel = [ovofifarcste F,(« Hap) 15.6) Peace orm Stata sora conero[istay ifarco-as} 5.3) Wi), =0. (13.26) Ayo) + O(AR). ass)0 ES Abersond B.W. Lee, Gauge theories tions say that Green's functions in the Coulomb gauge are the same as those in the Landau gauge, when the source is suitably restricted [eq. (13.26)] except that one must take into account extra vertices between a source and fields, represented by the term Jerk, -A,) cs.) when one tries to construct Coulomb gauge Green’s functions by the Feynman rules of the Landay gauge. This connection becomes much simpler, if we go to the mass shell. In this case, we ought to compare only the terms having a pole in each of the external momenta, p,, when p} + 0. Of all the diagrams generated by the extra couplings of (15.7), only those in which the whole effect of the extra vertices can be reduced to a type of self energy insertion to the corresponding external line survive in this limit. The other corrections introduced by (15.7) will not contribute to poles of the Green’s functions at p? = 0, and therefore not to the S-matrix. Therefore in the limit Pj 0, the Coulomb gauge and the Landatt gauge (unrenormalized) S-matrix elements will differ by a factor o” where u is the number of external lines and a is a factor independent of n. Com- Paring the two-point Green’s functions in the two gauges C and L: Iris helpful to visualize eq. (15.3) or eq. (15.6) in terms of Feynman diagrams. These equa- | lim Dj, a0 we find Zl, o In general, unremormalized S-matrix elements in the two gauges C and L are related to each order by Se = o"S, = (ZolZ, "7S, so that the renormalized S-matrix element BES. = ZS is independent of the gauge chosen to compute it. In sum, what we have shown here is that W [J] is equal to the expression (15.3) which would be Iv, [/] except that the coefficient of J* is Af? instead of A,. For the S-matrix, the only con- sequence of this difference is that the renormalization constants attached to each external line depend on the gauge. Thus we have shown that the S-matrix can be calculated from V, [J], not just by the intuitive argument of section 14, but more formally, by obtaining W/,[J] from first principles, and then demonstrating the equivalence of Sp and Sy, As pointed out earlier, the only flaw in the above argument is that the singularity at p? = 0 is not in general a simple role. Bibliography ‘his sction i an explication ofthe ditcusson on the same subject in 1. VIN. Popov and L.D. Faddcev, Perturbation Theory for Gauge Invariant Flelds, Kiew ITP report (unpublished). A similar discus 2B. Zumino, J J as even for ath. Phys. | (1960) 1 ntwmy electrodynamics inthe operator fet theory language byES. Abers and BW. Lee, Gauge theories a 16, Generating functionals for Green's functions and proper vertices In this section we develop the formalism of generating functionals of connected Green's fune- tions and of proper vertices. This topic is slightly out of the main line of development of this re- sinw. However, many recent papers on spontaneously broken symmetry make use of this elegant tormalism for a very good reason: this formalism allows the discussion of the conditions for spon- taneous breakdown of symmetry which goes beyond the one based on the classical Lagrangian snd which is valid to all orders in perturbation theory. Let us go back to the discussion of section 12 on scalar fields. We-define the generating func- tional ZU] of connected Green’s functions by WU] = exp ZUI1} = flo] eke i fa*xt2tecol + 5-601} (6.1) cre @ and J are multicomponent fields and sources, respectively. ‘The first derivative of Z[J] with respect toJ, is 6zU1 30) ar —~ f(a o(xexp Gi fax 268) + x)-90)]}« (16.2) We give it a special name, (x): ZIT [BS (x) = (x). (16.3) 47 is the vacuum expectation value of @,(x) in the presence of J(x); ie., it is the classical field. ‘The value of eq. (16.2) when the external source is turned off (I(x) = 0) is the vacuum expec = tation value of the field @: BZUIIBIAX)| 20 16.4) Note that v is independent of space-time, since in the limit J = 0, the left-hand side of (16.4) is iranslationally invariant. - It turns out that higher derivatives of Z[5] at J = 0 are Green's functions of the field 6= 6 ~v jose vacuum expectation value vanishes, For example 5210) “ e081) ral (011009 —, 160) - v1explifatx2) : ral MOB O05 (dexpti fatx26x)} (165) as can be verified by differentiating eq. (16.2) with respect to J, and letting J + 0. More generally we have ZU) 8Y;, Cr) BU, q) = "UTE, C2) lino B,D" (16.6) (where the superscript ¢ denotes the connected part of the Green’s function) as can be shown by *uction, ES SS ete:92 ESS. Aborsond BW. Lee, Gauge theories We shall now define the Legendre transform P[®] of Z{J]. It is defined as PLO) = ZU) ~ fatxtx)- (x), 823) /4J, =o, (6 The meaning of eq. (16.7) is this: P is a functional of (x) as clefined by the right-hand side of the first equality. In it, Jis to be expressed in terms of @ by inverting eq. (16.3), which define ‘© asa function of J. The Legendre transform (16.7) is @ functional version of the well-known transformation fam liar in classical mechanics and thermodynamics. By differentiating eq. (16.7) with respect to ®,, we find that SP 401/60) = Z fary (6Z191/0,091 61,0918 09} ~ LOD Jay 0181 076960), or SL [D] /5@(x) = -J,(x). (16.8) Equation (16.8) is dual to eq. (16.3): by this we mean that the relation (16.3) which expzesses ® in terms of J is the inverse of eq. (16.8) which expresses J in terms of ®. This, in particular, means that eq. (16.4) can be written as SPO] /80,0x);,., = 0, (169) e., when J = 0, © takes the value ¥, and vice versa. Equation (16.9) is very important. It express: the vacuum expectation value v of the field @ as the solution to a variational problem: y is the value of © which exiremizes P{@]. What is the physical significance of ? To streamline our discussion, let us agree on the follow- ing convention: We will denote by subscripts i,j... any labels J or carry, including the space- time variable x. We will adopt the convention that summations and integrations are always to be carried out over repeated indices. Differentiating eq. (16.3) with respect to @, we obtain ary ee = by 6.10, From eq. (16.8) we .carn that 81,/5%, = -5°T [0] /5,5%,. 6.1) Define (XU = ~8°Z151/ 61,8, 6.13) and {X10} }, = 68°F 101 /5,50,, (16.13) Equations (16.10) and (16.11) mean that (XD a = Bre 16.14) SinceES Abersand B.W. Lee, Gauge theories NE = ON Fy = ~P°ZUN 8,845] 509 = AGL J gthe full propagator for the barred field, and J = 0 implies © = v, it follows that (XID = ¥] }y = 5101/50/59, |, is the inverse of the full propagator. ) Next differentiate eq. (16.10) with respect to J,. We obtain bel) or ayy, On Sab ba, Om = or 1 P28) iy Gary, Gary, fj SElel F agayan, OX WOK Viney {i sae te, (16.15) Now take the limit J = 0, @ =v. In this limit X-'[J = 0] is the full propagator, so that HT(01/50,50,50,|, = FG) (16.16) is the three-point proper vertex. A proper vertex (or one-particle irreducible vertex) is a Green’s ‘s+ function which cannot be made disconnected by cutting @ single internal propagator, and from which (by convention) full propagators corresponding to external lines are removed. The three- 4 ht funetion has no such disconnected graphs except corrections to the propagators, which are explicitly removed in (16.15). In general, the nth derivative of [at © = v is the n-point proper vertex: 51 /50,58, .. = TY... The proof of this statement proceeds inductively. Assume that 5”Z(J] /5J;8J)... can be expressed 454 sum of tree diagrams, each diagram consisting of proper vertices corresponding to b"P[] /5,54,..., internal lines corresponding to Ay connecting pairs of proper vertices, and external lines, In particular, 1 5"ZUy) fi arto} "LSE 58,.. dX + one-particle reducible terms. 6.17) \n Now, differentiate eq. (16.17) with respect to J,. Recall that 5s, 8, 5, 81,51, 5, (8/52, (16.18) The differential operator 8/5%,, when applied to the right-hand side of eq. (16.17) can act either on some X“!, or on some 5”1/5+,54.... In the former case, we have 16 Fa, OX Mer EM KM Pr TRIE4 ES. Abersand B.W. Lee, Gauge theories which amounts to adding a new external line to a newly created three point vertex, and in the sion. ce cei hatter s 16 or amen 18 Pay —_ 18), 50,60,.. OX Wi 50,5,5%,... which amounts to adding a new external line to what used to be an m-point proper vertex. In any |i: cease, when the differential operator of eq. (16.15) is applied to the right-hand side of eq. (16.14), |v we generate all tree diagrams for the (n + 1)-point Green’s function, and Lemay To) Tisih 5; UMA mo 15 aoe 2 4 one-particle reducible terms. 6.19) . Therefore, in the limit J= 0, ®= v, i SPIO] 89,88, |, = reer : i is the (n+ 1)-point proper vertex. Now our proof is complete, since the induction hypothesisis | jg true for n = 3, as shown in eq. (16.15). ° ‘The generating functional of proper vertice P[®] has the representation: a | Tlel= 5 a DE tspentg(® De (@ = v);, AO —v),, (16.20), with dot Pps (45), (16.21) Let us revert to the standard notation: a a aot eat Alp Because of the translational invariance T°) depends only on n—1 differences x,—x,,so thatits || fe Fourier transform P is defined as ° Feo iy (Pts Py QAO + ou +p,=(Tt fetvexotiog)) re in Bis 9 X qe 16.22. |g ro This means that four-momentum must be conserved at vertices. In discussing the implications of the condition (16.9), it is convenient to consider the case in which ® is a constant $ independent of space-time. Define the super-potential Y by PL = 9] = -2m)*6() VC), Vo) = FO O05 ON — ¥),, (8 — ¥),, 1G —¥),,,5 (16.23) serey 9, % 0) ES, Abertand BW. Lee, Gauge theoriet so that an (@) —— HBO) 1,04 0 5¢,,54,,-d0, “inf , iat is the negative of the N-point proper vertex evaluated at the point where all external momenta vanish. The condition (16.9) translates into avo) ed, (16.24) Furthermore, ev@)_ 5646, is positive semi-definite, since Az! behaves like (p? — m?) near p? = m? and it cannot have any other zero for p?< m?. Thus the vacuum expection value ¢ = v is the value of @ which minimizes (y (9). The discussion in section 2 suggests that ¢ = v must be the absolute minimum of Y, but we do not prove it here. When 2 is invariant under 657 6, ~ i0*LE8, it follows from the structure of eq. (16.1) that Z[J] is invariant under 7 d,- iL, AO), (16.25) and so on, and finally the superpotential ‘ (9) is an invariant function of ¢ under the above trans- formation. The analysis of section 2 on the potential V can now be applied verbatim to the super- potential Y, with —[4/,(0)I, of eg. (16.28) taking the place of Mj, of eq. (2.19). We find there- fore that the occurrence and the number of the Goldstone bosons discussed there are true to all orders of perturbation theory. We can construct Z(J], ![®] and{¢} in perturbation theory. For simplicity we shall con- sider the case of a single-component field. An effective way of expanding these quantities in a series is to write eq. (16.1) with a fictitious parameter a: p6i2i81} = fraptere[ifare{!200 + 309-400]] tifa tee (ta4 LL fares _ A ~exp[ifa iol? Ss) oo atyalApr— yO}, (16.26) and expand Z{J} in powers of @ and let a = 1 afterwards. Since each propagator is multiplied by @ and each vertex by a” when we use eq. (16.26) as the definition of Z, it follows that a Feynman diagram with £ external lines, / internal lines and V vertices is multiplied by the factor, a®*/-¥. ‘There is a topological relation that holds for any Feynman diagram. It is fer ee% ES Abers and BW. Le, Gauge theories Lel-Vvel where L is the number of loops (ie. the number of independent fourmmomentum integrations) in the diagram. Therefore the expansion in this fictitious parameter @ corresponds to expanding, 4 Green's function in the number of loops in the Feynman diagrams. The reason this expansion is preferable ever the expansion in powers of some coupling constant is that in the former uny syinmetry of the Lagrangian is preserved in cach order of perturbation theory since, effectively, 4 multiplies the whole Lagrangian, In contrast, if we were to split up the Yang-Mills Lagrangian into a free and perturbing parts to develop a perturbation expansion, for example, each part would not be separately gauge invariant and the consequences of gauige invariance of the Lagrangian might not manifest themselves in each order of perturbation series. (Recall that non: Abelian gauge transformations depend on the coupling constant.) In the Following we shall discuss explicit constructions of Z, P and in the first two orders of loop expansion for a simple mode! HOP ~ $ube — Prod" 6.2713 The method can be generalized easily to other models, Our discussion will not show that our con- | struction is in fact the expansion in the number of loops, but the interested student can convince himself of this fact by first referring to Nambu’s paper which shows that the loop expansion is also an expansion in the Planck constant and then noting that our method is an asymptotic eviluation of these quantities in hi Imagine thai eq. (16.1) is written in the Euclidean space as explained in section 12. Since the exponent in the right-hand side is bounded from above in this case, we are tempted to evaluate the functional integral by the method of steepest descent. We shall keep the Minkowsky notation for simplicity, but the ultimate justification of this method lies in the Euclidicity postulate, We shall expand the exponent on the righthand side of eq. (16.1): SOI + fats Jaen) = fate (260) +1696}, about a point ¢(x) = go(x): S161 + farses) = Stdol + f ote seeoce) + fate { ae +100] L(x) — bo(8)1 © fatwa y —2*Sto0)_ +5 Set Y1B)-001 [609-000] Faataigg (1828) and choose #9 so that the term linear in § ~ @y is missing from the expansion of eq. (16.28). This will be achieved provided 5S{b01/6b0(x) = —JOx) (16.29) which means that 69 is the solution of the classical (non-quantized) field equation in the presence of the external source J(x). For the Lagrangian (16.27), eq. (16.29) is (2? + HDGoLN) +d GROW) =O). (16.30) In any case, $9 8 obtained from eq. (16.29) as 2 functional of the external source J.ES: Abersand BM. Lee, Gage theories 7 Wheneg. (16.28) is substituted in eq. (16.1), we obtain exp(i ZVI) = exp€i Sldal +i fax JOoG0)} 1 _ 85601 2! 5o(5)660) ‘The lowest order approximation (which is one order lower than the steepest descent approxima tion) is obtained if we ignore the functional integral over g(x) altogether and set ZUI ¥ Slo) + faxJepo(x) = ZT (16.32) x Sesser ateas (91x) ~ ls COL) ~ dol Joan which isa functional of J only, because $6 is a functional of J. We can evaluate Z° explicitly by Fiegt solving for $o in eq. (16.30) and then substituting that ¢y in eq. (16.32). Equation (16.30) Abe solved in powers of A: dal) = — fay Aplx ~ ys? VO) — AL fey Aple — y; AO)? +. (16.33) where the use of Ae is dictated by the Euclidicity postulate. When eq. (16.33) is slbstituted in eq. (16.32), one finds that Z°[/] is the generating functional of Green’s functions in the tree- {i.€., no loop) approximation: 1 X A PUL = fated a ee undone; ftw PT fate toepaple, - wat) +. 16.34) We can see more readily that Z° is the tree approximation to Z if we compute P{] in this ap- j proximation, Since 8Z_ 82 _ ry, { 85160] BH 4 5) Hg, HO Sra5~ B70) S*” 5900) B10) 52), we have, to this order, Px) = dol. (16.35) Therefore, P[) can be computed to this order: Pb] = ZU] = faxsx) Oo) SLO] + fatxJGVO(x)} - fated) = S14]. (16.36) So, to this order, proper vertices are generated by the Lagrangian itself and Green's functions are | built up of these unmodified vertices by the rules of tree graphs. The superpotential Y (eq. (16.36)] is, to this order V(9) = -S(9) = V6) where @ is independent of space-time and V is the negative of the part of the Lagrangian which is independent of derivative of fields. That is, (9) is the potential of the field 9. This justifies the }e “super-potential” for8 ES. Abersand BW. Les, Gauge theories We can proceed further by applying the steepest descent method to the functional integral in eq. (16.31). This consists of neglecting terms higher than quadratic in (# — o) in the exponent of the integrand and performing the functional Gaussian integration, In this way we obtain fisetro| for ey S161 2 Bel5G00) 196) ~ $000) 160) — do00)1 pe oe ty ten Od __ Viet FSTGo1/G6(x)6640) P21 SEBO) * so that ZU] Z°LJ] + Fi Tr M{5*SlGo]/5G0(x)5G0(y)} = ZL]. (16.37) For the Lagrangian (16.27), for example, BAS/EG(x)B SLY) = (—3°—"?-3AG*XHCX ~ V), so that i 8s. i Ten Sago) 3 rn(1 —3h ei Gin --75 ) =e Ae Pd, MeO — POOR — 25) Ape, — 8H). (16.38) ME] = Z1V) — farsa, (16.39) where d(x) = ZIV BI) = balx) + eC) (16.40) and e(x) is given by 8 ia, ay 2°Slt0) Bx) 2 on" BbalEBGACN) Fortunately, it is not necessary to know the form of e(x) to construct P[1 to first order in eC), as we shall demonstrate presently. First, note that eG) = ZI) = Sto) + fatxICepalx) ste) Gov) = S14] + fadteJG9b(x) + OC) 6.40) = SI] + fatxsonwcn — faty suenroe Let us now construct P'[] to this order: |din ont 6.39) 6.40) ao. 6.4) ES. Abers ond B.W. Lee, Gauge theories » by virtue of eq. (16.29). Therefore to order e, we have from eqs. (16.37), (16.39) and (16.41) i & Mm(e] =S[@) +>Tr no. 16.42) 4 Dein 5HCE)EH) ‘ ‘The second term is the one-loop correction to the generating functional of proper vertices. ‘The super-potential Y can be evaluated explicitly from eqs. (16.38) and (16. 42). Recalling the definition of Y of eq. (16.23), we find By Qa it mL LOGO rs ey er 2, wl y" e—) 16.43 1? = pt tie eee ‘The terms for N = | and 2are divergent. However these terms are proportional to $* and ¢* and the divergences in these terms can be amalgamated with u3 and Ao. We may write (9) = 404? + PAO" +197) where (16.44) and u? and ) are defined as the value of the two- and four-point vertices at the point where all ex- ternal momenta vanish. Bibliography “Te idea of using generating furctionas for Green's funetions and proper vertices was originated by 1.3, Schwinger, Proe. Nat. Acad. Si. 37 (1951) 452,455, “The folowing paper contains the fst explicit construction of the generating functional for proper vertices by the Legendre 2G. Jona-Lasinio, Nuovo Cimente 34 (1964) 1790. “his pape also contains the derivation of the Goldstone theorem by this technique. Recent reviews of ths method may be found 3. LL, Leo, Chiral Dynamics (Gordon and Breach, New York, 1972). ‘HLM. Fried, Functional Methodsand Models in Quantum Field Theory (MT Press, Cambie, 1972). 5. Zumino, in Lectaes on Clementary Particles and Quentum Field Thoory, eds. S. Deser, M. Grist, 1, Pendleton (MIT Press, ‘Cambridge, 1970) Vol. 2, p. 438 et sa “Ihe observation that the expansion in the number of loops is equivalent tothe expansion in is due to 6.¥. Namba, Phys, Latiers 268 (1966) 626. TLS. Brown and D. Boulvare, ys. Rev. 172 (1968) 1628 “The evaluation of the one-loop corrections bythe steepest descent approximation is discussed in &.UN, Lee and. Zinn-lustin, Ph. Rev. DS (1972) 3121, Appendix B Foran extensive use ofthis meta in a recent literature, se, for example, 5.5. Coleman and E. Weinberg Phy. Rev, to be published.100 ES. Abers and B.W. Lee, Gauge theories 17. Renormalization in the a-model The formalism developed in the preceeding section is useful in discussing renormalization of sportaneously broken symmetry models and, in particular, the o-model. In the generic sense. the o-model is a model in which a symmetry is broken by a term of dimension one, i., by a term. prorortional to a boson field. s Asimple example of this kind of models is 4 £= F1Qy7)? + @,0)"] — $ud(o? +24) ~1Ag(o? +9)? + co= Laym +00 a7 Which is a two-dimensional generalization of the model discussed in the preceeding section. Exceg for the last term co, the Lagrangian (17.1) is the one studied in section 2, and it was noted there that this Lagrangian is invariant under a U(1) transformation of the fields o and 7. The salient features of this model are that the “almost” conserved current 4 A, = 10,0 ~ 08,7 «73 has a divergence proportional to the m-field 3) 24, = or 735 and that the o-field acquires a nonvanishing vacuum expectation value thanks to the last term in €4. (17.1). Equation (17.3) is a version of the PCAC condition, and for this reason the model is._| of some physical interest. q It pays to study first the classical solution of the Lagrangian (17.1). The potential is given by V(9, m) = 3A(0? +m)? + $u%(0? + 7) — co 744] (we drop the subscript 0 ond and for the moment). The minimum of the potential occurs at m= Oand o= wu where u(u? + uA) = 6, u being the vacuum expectation value of the o-field in this approximation. If we displace the fiel @ by the amount u and define s by s = 6 ~ w eq. (17.1) takes the form 2= 71,8) + Qn) — puis? — tude? — Ns? + 22)? — duds? + ws so that in this approximation the s-field represents a particle of mass #3: a= py? +3? and the m-field a particle of mass p2: wR we In this approximation, when ¢ = 0, i.¢., when the Lagrangian is invariant under the UCI) trans formation, either w = 0 or uz = u? + Mu? = 0 according to eq, (17.5). If 12 = 0, then u? > O in order: that u} = uf = u? > 0. This is the “usual” way the symmetry of the Lagrangian manifests itself: the particles corresponding to the fields o and w are degenerate. On the other hand, if u2 = 0, wenof se, the ™m aw min lis iby 17.4) sat 17.5) fie! 17.6) ES. Abers and B.W. Lee, Gauge theories 10 must have x? < 0 since Nu? > 0. The second case is the Goldstone mode of the symmetry with the field 1 playing the role of the Goldstone boson. In that case, u? = Mw, and p2 = —2y? > 0. For more thorough discussion of the o-model, see the monograph “Chiral Dynamics” by one of us We return to the discussion of the full solution, including radiative corrections. An important fact about the o-model is that the Green’s function of this model are generated by the generating fanetional of Green’s functions of the symmetric theory. The latter is given by expt Z(}} = fldo) fdr lexpti fate 2, q(x) +Jy(soCe) + JgCx mC] by Ug. J). Now, expand Z(J) in J about J, fo 79) 0. We have and J, 1 promos FE BT). BT (Xp OF g(V1)--- BF, 0p) _ 1 Wie. (17.10) ore.dy=O an Sido Lartsex)..5(x,)101)...10%,) exp i | fa*x.2(x)} — disconnected pieces, where s = ¢ ~ u, w being the vacuum expectation value of o so that Sidoi tan lscoexpti faty20)} = 0, and s(x) = a(x) 0, since in that case there is a comparison symmetric theory that makes sense. However, once P{®] is constructed in terms of A, n? and c there is nothing that stops us from expressing T{@) in terms of A, w and m3, where the last is defined as 2 = -A51(0) = 26(6)/56" |. ‘and taking the limit m2? ~ 0. Then eq, (17.23) reads 17.24) Equation (17.24) is the renormalized Goldstone theorem: if ¢ = 0 either w= 0, or m= 0. The { latter corresponds to the Goldstone mode. In this case the basic parameters of the theory can be 7.19) taken to be Aand w= », instead of A and —p2. ‘The moral of the above discussion is that the renormalizability of the o-model in the Goldstone jis | Made depends only on the renormalizability of the symmetric theory. The process of renormaliza- tion does not induce additional symmetry breaking, in the sense that the symmetric counter- terms exhibited in (17,20) suffice to remove infinities from the theory whether or not the sym- | metry is broken external'y (c # 0) or internally (uv # 0). © Later we will discuss a way of renormalizing the o-model without making explicit reference to the symmetric theory. This method makes use of the Ward-Takahashi identities. Let us derive them. The generating functional Z{3) in eq. (17.9) is invariant under the U(1) transformation of wn thet e 104 ES. Abersond BW. Lec, Gauge theories the external sources: Luwt (2): (22 sn )(te) arg} tks J.) \sind cos a/\y,, “gy iese as can be seen by making the change of integration variables epi oy _fcosé sin 8 y/o) 4 (2)*(ne cools) C728F sor Which leaves the scalar product J,o +J,a invariant. Therefore, Intl —_ recent dZ/d9 = 0, but we or Firs fra 8z0| ay "7 4 5 sing Joe (2 tae soy tt] 0. a2 .. ‘Substituting eqs. (17.23) and (17.24) into eq. (17.27), we find that oro} arto} at ‘ 5 = & (x) Son Ja {2.09 58,02) 7 ** 5664) which shows that I’ is an invariant functional of © under the UCL) transformation: yA o,\' 0 =sin 8\/&, 3 ( ey = (en? sin ( 2). (17.291 #,) “\sind cosa lo, Fea Note that the invariance of P under the transformation (17.29) is true whether p? > 0 or p? < 0. The renormalized P constructed according to the prescription above, this satisfies eq. (17.28) as The ‘we continue m2 to zero. AD by cou Equation (17.28) is the Ward-Takahashi identity for the generating functional of proper: vertices lis An infinite number of Ward-Takahashi identities is obtained if we differentiate eq. (17.28) with ql Dr respect to ®, and ®, repeatedly, and set , = 0, b, = 1. If we diffecentiate eq, (17.28) with re x! spect to @, and set , = 0, &, = u, we obtain the “eigenvalue” equation for u, eq. (17.23). If well the has differentiate it with respect to &, and &, and take the limit, we obtain aP tion AQMP?) — A?) = uP, g(0; 0, —P)- (17.3034 one de 4 ), ‘An important Jesson to be learned here is that the Ward-Takahashi identity for the generating {anctional for proper vertices is the same, whether of not the symmeiry is spontaneously broken: Itis satisfied by the generating functional constructed first in the symmetric theory and then co inued to the Goldstone mode by varying an appropriate parameter of the theory, on, [ope Bibliography The integra This sections bated on 1. BAW, Lee, Nuel Phys, B9 (1969) 649. ys