This can be seen from the propagator, e.g. for U(1) [QED] we have Z Z 1 1 2 = d4 xAµ (x)(η µν ∂ 2 − ∂ µ ∂ ν )Aν (x) d4 x − Fµν 4 2 with the propagator 2 (ηµρ ∂ − We constrain the gauge directions, by introducing a gauge fixing term at each point x Ga (A) = 0 Now define ∆F P [A] by [FP = Faddeev-Popov] ∂µ ∂ρ )DFρν (x − y) = iδµν δ(x − y) i∆µν F ≡ DFµν However ηµρ ∂ 2 − ∂µ ∂ρ cannot be inverted, as multiplying both sides of the ∆F P [A] Z [dU]δ[Ga (AU )] = 1 equation by ∂ µ gives [δ[ ] is a product of δ functions at each space-time point.] 0 6= i∂ ν δ(x − y) where Aµ is the gauge transformed A, i.e. A contradiction! Alternatively we can say that the operator has zero eigenvalues. (In the canonical formalism, tackle this with the Gupta-Bleuler method.) To complete the functional integral, we must ‘factor’ out these gauge transformations i.e. “ Z [dA] ∼ Z [dĀ][dU] ” i AUµ = U(Aµ + ∂µ )U † g and [dU] is the Haar measure, invariant under group transformations, i.e. if U ′ = U0 U, then [dU ′ ] = [dU]. ∆F P [A] is invariant under gauge transformations as Z ∆F P [AU0 ]−1 = [dU]δ[Ga (AU0 U )] U ′ = U0 U Z ′ = [dU ′ ]δ[Ga (AU )] = ∆F P [A]−1 “gauge orbits” related by gauge transformations gauge inequivalent Aµ s ‘gauge orbits’ (related by gauge transformations) gauge inequivalent Aµ s (assumed to interest each gauge orbit only once) 105 Thus we have Z Z Z R 4 R 4 [dA]ei d xLY M (A) = [dA] ∆F P [A] [dU]δ[Ga (AU )] ei d xLY M (A) {z } | 1 Z Z R 4 ′ = [dU] [dA′ ]∆F P [A′ ]δ[Ga (A′ )]ei d xLY M (A ) 106 upon interchanging integrals, and then changing variables A′ = AU , ∆F P Now and LY M being gauge invariant. We see that we have factored out the (infinite) gauge transformation ‘volume’, R [dU]. Now throw it away (it cancels in ratios anyway). Z [dA]ei R d4 xLY M (A) ∼ Z [dA]∆F P [A]δ[Ga (A)]ei R d4 xLY M (A) (90) Ga (AU ) = Ga (Aµ + DµA Λ) δGa A b (D Λ) + O(Λ2 ) = Ga (A) + δAbµ µ which gives ∆F P [A]−1 G=0 = = This is now a well defined integral, but we now need to evaluate ∆F P [A] µ or where ∆F P [A]−1 = Z δGa A b [dU ] = [dΛ] around Λ ∼ 0 (Dµ Λ) ] [dΛ]δ[Ga (A) + b δAµ G=0 Z 1 1 h i dN xδ(N ) (M x) = δG det M det δA DµA Z G=0 [dU]δ[Ga (AU )] a a ∆F P [A]|G=0 ∼ det U As this factor is multiplied by δ[G (A)], the only solution of G (A ) = δG A D δAµ µ (92) Ga (A + O(Λ)) = Ga (A) + O(Λ) = 0 is Λ = 0 or U = I. See the picture, where we have assumed that we have a unique solution (otherwise Gribov Thus we have copies may exist). Z So for infinitesimal Λ, then U(Λ) = 1 + igΛata and i Aµ = Abµ tb AUµ = UAµ U † + U∂µ U † g = Aaµ + ∂µ Λa − igif abc Abµ Λc + . . . ⇒ AaU µ [dA]ei R d4 xLY M (A) ∼ Z [dA] det R 4 δG A Dµ δ[Ga (A)]ei d xLY M (A) δAµ (93) f We now wish to re-write the extra terms so as to have an Lef YM ≡ Aaµ + (DµA Λ)a + . . . 1. Replace where Ga (A) → Ga (A) − ω a (x) [Nothing changes in derivation] where ω a (x) is any scalar function. The LHS of eq. (93) is independent (DµA )ab ab c ab = δ ∂µ − ig(T ) Acµ c ab (T ) ≡ −if cab is the covariant derivative in the adjoint representation (cf eq. (83)) 107 (91) 1 R 4 a2 of ω a , so can integrate over a Gaussian e−i 2ξ d xω to give Z R 4 a2 R 4 i i a 2 [dω]δ[Ga (A) − ω a (x)]e− 2ξ d xω = e− 2ξ d x[G (A)] 108 Note: need ξ → 0 to recover original δ-function constraint, Ga (A) = 0 The gauge field propagator as given from the quadratic piece of the action in A is then Z Z 1 1 d4 xLA,quad = d4 xAaµ (x)(η µν ∂ 2 − (1 − )∂ µ ∂ ν )δ ab Abν (x) YM 2 ξ 2. Introduce anticommuting (Grassmann) fields c, c̄ Z R δG δG A DA c −i d4 xc̄ δA µ µ det Dµ = [dc̄dc]e δAµ (under Lorentz transformations) but are anticommuting. They can be i.e. diagonal in group space, so these indices can be suppressed (if wished) 1 ηµρ ∂ 2 − (1 − )∂µ ∂ρ DFρν (x − y) = iδµν δ(x − y) ξ used in internal lines in computing Feynman diagrams – the Fadeev- or c and c̄ (independent) cannot be physical fields as they are scalars Popov ghosts. µν i∆µν F (x − y) ≡ DF (x − y) = Z d4 p −ip·(x−y) −i e (2π)4 p2 + iǫ So finally we have Z [dA]ei R d4 xLY M (A) ∼ Z [dA][dc̄dc]ei R f d4 xLef Y M (A,c,c̄) (94) with f Lef Y M (A, c, c̄) = − 1 δG A 1 a 2 F − Ga (A)2 − c̄ D c 4 µν 2ξ δAµ µ (95) pµ pν η µν − (1 − ξ) 2 p (97) [check (see example sheet) 1 δρµ ∂ 2 − (1 − )∂ µ ∂ρ DFρν (x − y) ξ Z d4 p −ip·(x−y) 1 −i pρ pν = e −δρµ p2 + (1 − )pµ pρ 2 η ρν − (1 − ξ) 2 4 (2π) ξ p + iǫ p Z d4 p −ip·(x−y) −i 2 µν µ ν e −p η + (1 − ξ)p p = (2π)4 p2 + iǫ 1 1 + (1 − )pµ pν − (1 − ξ)(1 − )pµ pν ξ ξ = iη µν δ(x − y) 1 −1+ξ+ 1 −1 = 0 pµ pν coefficient: 1 − ξ + 1 − ξ ξ The obvious gauge condition (Lorentz) is Ga (A) = ∂ · Aa so f Lef Y M (A, c, c̄) δGa δAbµ 1 a2 1 = − Fµν − (∂.Aa )2 + ∂ µ c̄a DµAab cb 4 2ξ = ∂ µ δ ab or (96) ] Note the different sign to the KG equation ξ = 0 Landau gauge [∂.A = 0 as have δ function again] Feynman gauge U(1) [Simple looking propagator] ξ = 1 Feynman − t′ Hooft gauge SU(N) (98) 109 110 Note also that for QED as f abc → 0, then DµA → ∂µ is independent of gauge field A, so that the Faddeev-Popov ghosts decouple, and hence do not appear in the action. Finally: Physical amplitudes being gauge invariant do not depend on the gauge fixing term (and Faddeev-Popov ghosts). The generating functional is ¯ ζ] Z[J, η̄, η, ζ, Z R 4 a aµ a a a a a a a a = N [dA][dη̄dη][dc̄dc]ei d x[L0 (A,ψ,ψ̄,c,c̄)+Lint (A,ψ,ψ̄,c,c̄)+Jµ A +η̄ ψ +ψ̄ η +ζ̄ c +c̄ ζ ] Z δ δ δ δ δ = N exp i d4 zLint −i , −i , i , −i , i Z0gluon [J]Z0quark [η̄, η]Z0ghost [ζ̄, ζ] δJ δη̄ δη δζ̄ δζ Now use analogies to our previous scalar and fermion theories 22 Feynman Rules for Gauge Theories Including fermions, our final Lagrangian in the Lorentz gauge is 1 1 a2 / − m) ψ +∂ µ c̄a DµAab cb − (∂ µ Aaµ )2 + ψ̄(iD L = − Fµν |{z} 4 2ξ |{z} 22.1 Propagators 22.1.1 gluon fields (99) quark/ electron gluon/ photon SUc (3)/U (1) gauge groups We must now find the Feynman rules 1 Z0gluon [J] = e− 2 νb d4 xd4 yJ µa (x)i∆ab F µν (x−y)J (y) i∆µνab (x − y) ≡ DFµνab (x − y) = F p a, µ where 22.1.2 b, ν = i∆µνab (p) = F 1 1 Lint = − g 2f abe f cde Aaµ Abν Acµ Adν − gf abc (∂µ Aaν − ∂ν Aaµ )Abµ Acν 4 2 +g ψ̄γ µ ta ψAaµ − gf abc (∂µ c̄a )cb Acµ 111 p2 pµ pν −i η µν − (1 − ξ) 2 δ ab + iǫ p quark fields 4 4 i ij j Z0quark [η̄, η] = e− d xd yη̄ (x)iSF (x−y)η (y) Z d4 p −ip·(x−y) i iSFij (x − y) = e δ ij (2π)4 p/ − m + iǫ R and pµ pν ab d4 p −ip·(x−y) −i µν e δ η − (1 − ξ) (2π)4 p2 + iǫ p2 Z so L = L0 + Lint 1 1 L0 = − (∂µ Aaν − ∂ν Aaµ )(∂ µ Aνa − ∂ ν Aµa ) − (∂µ Aµa )2 4 2ξ +ψ̄(i∂/ − m)ψ + ∂ µ c̄a ∂µ ca R 112