Causal Effective Field Eqns from the Schwinger-Keldysh Formalism Richard Woodard University of Florida Two Common Questions 1. Aren’t QFT effective field eqns nonlocal? Yes! 2. Doesn’t this introduce acausality? • • Not unless gravity is dynamical In-out eqns are acausal, but from in-out BC In-in eqns are causal Linearized Examples ∂µ[(-g)½gµν∂ν ϕ(x)] - (-g)½ [ξR+m2] ϕ(x) -∫d4x′ M2(x;x′) ϕ(x′) = 0 (-g)½ eµaγa [∂µ-½A µbcJbc] ψ(x) – (-g)½m ψ(x) -∫ d4x′ [Σ](x;x′) ψ(x′) = 0 ∂ν [(-g)½gνρgµσ Fρσ(x)] -∫d4x′ [µΠν](x;x′) Aν(x′) = 0 Eg. M2(x;x′) for λϕ3 Best in position space M2(x;x′) = ∫d4p/(2π)4 Exp[ip·(x-x′)] M2(p) No integrals at 1 loop! [-i M2(x;x′)]1 loop = ½ (-iλ)2 [i∆(x;x′)]2 Full Nonlinear Eqns For in-out (in-in similar) • Γ[ϕ] = S[ϕ] - ∑Ν=2 1/N! ∫d4x1 ϕ(x1) . . . . . .∫d4xN ϕ(xN) ΓN(x1,...,xN) • ΓN(x1,…,xN) ≡ N-point 1PI function • δΓ/δϕ(x) = δS/δϕ(x) - ∑Ν=11/N!∫d4x1 ϕ(x1) . . . . . ∫d4xN ϕ(xN) ΓN+1(x,x1,…,xN) = 0 What Does It Mean? ϕ(x) = <Φ+|φ(x)|Φ−> φ(x) ≡ Quantum field operator ϕ(x) ≡ C-number soln of eff. field eqn with Pos. freq. parts of ϕ and f+ agree at time t+ Neg. freq. parts of ϕ and f- agree at time t- |Φ±> ≡ State centered on f± at time t± For in-out t+→+∞ and t-→-∞ For in-in t+ = t- and often finite What’s Wrong with In-Out? Great for scattering in flat space! • Exp[iΓ[ϕ] – iJ∞·ϕ] generates S-matrix • Cf. on-shell finiteness But silly for cosmology or evolution • • • • Initial singularity can’t have t-→-∞ Don’t know state for t+→+∞ No formal S-matrix (but Fermilab still works!) In-out ϕ(x) depends on future and isn’t real Introductory QFT Teaches Now Multiply by Conjugate Now Sum over States at t = t2 1. ∑Ψ |Ψ Ψ | = I 2. ∑Ψ Ψ⋆[ϕ−(t2)] Ψ[ϕ+(t2)] = δ[ϕ+(t2) - ϕ−(t2)] Schwinger-Keldysh Formalism 1. 2 fields ϕ± line endpts have ± “polarity” 2. Exp(iS[ϕ+] – S[ϕ−]) Ints all + or all – • • 3. 4. 5. 6. 7. + Interactions same as usual - Interactions conjugated T(B[φ])↔B[ϕ+] T-ordered ext. lines + Anti-time-ordered external lines are – 4 Propagators: i∆±±(x;x′) In-out N-point func. 2N S-K N-points Φ[ϕ±(t1)] surface ints at t = t1 if not free Eg. –iM2(x;x′) for λϕ3 (-iλ)2 2 ′ [i∆(x;x )] 1 Loop In-Out: ½ 1 Loop Schwinger-Keldysh: • • • • -iM2++(x;x′) -iM2+-(x;x′) -iM2-+(x;x′) -iM2--(x;x′) = ½ (-iλ)2 [i∆++(x;x′)]2 = ½ (-i λ) (+iλ) [i∆+-(x;x′)]2 = ½ (+iλ) (-iλ) [i∆-+(x;x′)]2 = ½ (+iλ)2 [i∆--(x;x′)]2 S-K Effective Field Eqns Γ[ϕ+,ϕ−] = S[ϕ+] – S[ϕ−] -½∫d4x∫d4x′∑±±ϕ±(x)M2±±(x;x′)ϕ±(x′) + O(ϕ3) 0 = δΓ[ϕ+,ϕ−]/δϕ+(x) (then set ϕ±=ϕ) = δS[ϕ]/δϕ(x) -∫d4x′ [M2++(x;x′) + M2+-(x;x′)] ϕ(x′) + O(ϕ2) • NB: M2-+(x;y) = M2+-(y;x) Get Props with Canonical Relation F. F. Int. × ϕ+(x)ϕ+(x′) = <Ω0|T[φ(x)φ(x′)]|Ω0> i∆++(x;x′) = i∆(x;x′) F. F. Int. × ϕ+(x)ϕ−(x′) = <Ω0 |φ(x′)φ(x)|Ω0> i∆+-(x;x′) = θ(t-t′) [i∆(x;x′)]* + θ(t′-t) i∆(x;x′) F. F. Int. × ϕ−(x)ϕ(x′) = <Ω0|φ(x)φ(x′)|Ω0> i∆-+(x;x′) = θ(t-t′) i∆(x;x′) + θ(t′-t) [i∆(x;x′)]* F. F. Int. × ϕ−(x)ϕ−(x′) = <Ω0|A[φ(x)φ(x′)]|Ω0> i∆--(x;x′) = [i∆(x;x′)]* Properties of [M2++(x;x′) + M2+-(x;x′)] 1 Loop λϕ3: [M2++ + M2+-] = -iλ2/2 {[i∆++(x;x′)]2 - [i∆+−(x;x′)]2} Fact 1: i∆++(x;x′) = θ(t-t′) i∆-+(x;x′) + θ(t′-t) i∆+-(x;x′(x;x′) [M2++ Fact 2: + 2 M +-] i∆-+(x;x′) = 0 for t′ > t = [i∆+- * ′ (x;x )] [M2++ + M2+-] = -λ2θ(t-t′) Im{[i∆+-(x;x′)]2]} Manifestly real (unlike in-out!) Causality of [M2++(x;x′) + M2+-(x;x′)] [M2++ + M2+-] = -λ2 θ(t-t′) Im{[i∆+-(x;x′)]2} Fact 3: i∆+-(x;x′) = <Ω0|φ(x′)φ(x)|Ω0> = ½ <Ω0|{φ(x′),φ(x)}|Ω0> + ½ <Ω0|[φ(x′),φ(x)]|Ω0> Fact 4: [φ(x′),φ(x)] = 0 for spacelike sep. [M2++ + M2+-] contributes only for x′µ on or within past light-cone of xµ Not even superluminal w/o derivative ints Blurring the Light-cone The Physics: • Metric operator gµν = ĝµν+ hµν sets light-cone • hµν allows propagation not possible in ĝµν The Math: • Derivative ints (h∂h∂h) push inf. off light-cone • Typical ∆c/c ∼ G2Rρσµν Rρσµν • Cf. Larry Ford’s work Conclusions 1. S-K field eqns nonlocal, but good initial value problem 2. No violation of causality without quantizing gravity 3. No superluminal propagation without derivative interactions 4. Quantum gravity blurs the light-cone