Lattice QCD and transport coefficients

Transcription

1 International Nuclear Physics Conference, Adelaide, Australia, 13 Sep Cluster of Excellence Institute for Nuclear Physics Helmholtz Institute Mainz

2 Plan Strongly interacting matter at temperatures T = MeV probed in heavy-ion collisions: hadrons quark-gluon plasma state of matter for the first microsecond after Big Bang Thermal physics: β = 1/(kT ), A = 1 Z Tr {e βh A}, Z = Tr {e βh } Overview of equilibrium properties from lattice QCD Near-equilibrium (real-time) properties formalism vector channel for light quarks: dilepton rate, diffusion coefficient heavy-quark momentum diffusion coefficient pion quasiparticle in the hadronic phase

3 QCD phase diagram T heavy ion collider QGP hadronic non CFL gas liq CFL nuclear superfluid neutron star µ at µ B = 0: T transition = 155 ± 8 MeV from lattice simulations (crossover) e.g. from chiral susceptibility d 4 x ψ(x)ψ(x) ψ(0)ψ(0) [see e.g. review Soltz et al ]. Fig. from Braun-Munzinger, Koch, Schäfer, Stachel, Phys.Rept. 621 (2016) 76

4 Thermodynamic potentials Fig. from review by Soltz et al at T = 260MeV, p 1/2p SB: far from weakly interacting quarks and gluons; hadron resonance gas (HRG) model works well up to T = 160 MeV; HRG also describes well the fluctuations of conserved charges, 1 e.g. V Q2, B 2 and S 2.

5 Regularization of QCD on a lattice ν Uµ P µν µ Gluons: U µ (x) = e iag 0A µ(x) SU(3) link variables Quarks: ψ(x) on site, Grassmann Gauge-invariance exactly preserved; no gauge-fixing required. Imaginary-time path-integral representation of QFT (Matsubara formalism): Formally: Lattice QCD = 4d statistical mechanics system Starting point for Monte-Carlo simulations using importance sampling

6 Near-equilibrium properties Typical questions: What quasiparticles are there in the system? How fast does an external perturbation dissipate in the system? for long wavelength perturbations, the rate is parametrized by transport coefficients (shear/bulk viscosity, diffusion coefficients,... ) t large J 0(t, k) θ( t)µ(k) e Dk2 t What is the production rate of photons or dileptons?

7 Formalism Relation between the correlator }{{} and the computed on the lattice G(x 0, p) d 3 x e ip x J(x) J(0) = 0 spectral function : }{{} what we want to know dω 2π cosh[ω(β/2 x0)] ρ(ω, p). sinh[ωβ/2] in the low-t phase, J em i can excite e.g. an ω-meson-like quasiparticle. for J = J em i electromagnetic current, ρ(ω, 0) ω 0 6χ sdω χ s = d 4 x J em 0 (x)j em 0 (0) = static susceptibility of electric charge D= diffusion coefficient photon rate: dγ d 3 k = e2 f Q2 f ρ(k,k) 2(2π) 3 k e βk 1 numerically ill-posed inverse problem for ρ(ω, p) ω>0 0; 0 x 0 < β.

8 The inverse problem has many faces: here is one of them Linearity: n c i( ω) G(t i) = i=1 0 dω 2π ρ(ω) n i=1 c i( ω) cosh[ω(β/2 ti)] sinh[ωβ/2] } {{ } δ( ω,ω) choose the coefficients c i( ω) so that the resolution function δ( ω, ω) is as narrowly peaked around a given frequency ω as possible (idea behind the Backus-Gilbert method, [used in Robaina et al. PRD 92 (2015) ]) T ˆδ( ω, ω) λ = 1, ω/t = 4 λ = 0.5, ω/t = 5 λ = 0.01, ω/t = 5 λ = 0.002, ω/t = 5 Resolution function at ω = 4T for N t = 24, t i/a = 5, Resolution only improves slowly with increasing n Large, sign-alternating coefficients need for ultra-precise input data ω/t

9 Expected thermal changes in spectral functions Isoscalar vector channel: spectral fct. of J i = 1 2 (ūγ iū + dγ id) tanh(ω / 2T) ρ ii (ω,t) / ω SND e + e - --> π + π π 0 wk coupl T=260MeV N=4 SYM λ=oo intercept = 3 χ s D / T 1 3 / (2 π) ω / GeV presence of weakly coupled quasiparticles transport peak at ω = 0; is it really there at T 260MeV? SND hep-ex/ D = diffusion coefficient; χ s = static susceptibility.

10 The isovector vector channel at p = 0 Lattice QCD correlators (T = 0.8, 1.0, 1.25, 1.67 T c) Spectral functions 1GeV fit ansatz method: simultaneous fit to all temperatures, because ρ(ω) ω (# + # αs +... )ω2 with T -independent coefficients; exact sum rule: 0 π dω ω [ρ(ω, T ) ρ(ω, T )] = 0 constraint; Francis et al. (N f = 2, m π T =0 = 267MeV, T c = 203MeV), PRD93 (2016)

11 Model-dependence of the spectral function Alternative ansatz at T = T c Spectral functions 1GeV fit ansatz method: simultaneous fit to all temperatures, because ρ(ω) ω (# + # αs +... )ω2 with T -independent coefficients; exact sum rule: 0 π dω ω [ρ(ω, T ) ρ(ω, T )] = 0 constraint; Francis et al. (N f = 2, m π T =0 = 267MeV, T c = 203MeV), PRD93 (2016)

12 Lattice QCD vs. phenomenology ρ(ω, T ): lattice vs. pheno Phenomenological model. compare ρ(ω,t ) tanh(ωβ/2) 0 dω δ(ω, ω ) ρ(ω,t ) tanh(ω : lattice result is model-independent. β/2) shift of spectral weight from the ρ to low frequency region as T increases. Francis et al. PRD93 (2016) ; Rapp & Hohler, PLB731, 103 (2014).

13 Calculations of the diffusion coefficient inverse problem treated with the Maximum Entropy Method; D ρ(ω)/(χ sω) ω=0 comes out very small; stability of the results tested under variations in the procedure. N f = simulations, m π T =0 = 384MeV, Aarts et al. JHEP 1502 (2015) 186. See also N f = 0 continuum calculation using fit ansätze Ding, Kaczmarek, F. Meyer PRD94 (2016)

14 Selected recent results for the light-quark diffusion coefficient D N f =2+1, N f =2, N f =0, AMY hep-ph/ α s =0.3 D T N f =0, heavy-quark AdS/CFT T/T c lattice calculations yield very low values, D 1/(πT ); however, all results assume that no narrow transport peak is present: these methods would fail at very high temperatures. Except green point!

15 T > T c : Heavy-quark momentum diffusion coefficient κ Re Tr ( U(β, τ)ge k (τ, 0)U(t, 0)gE k (0, 0) ) dω cosh[ω(β/2 τ)] G(τ) = = ρ(ω) 3 Re Tr U(β, 0) 0 2π sinh[ωβ/2] color parallel transporters U(t 2, t 1) are propagators of static quarks (Lorentz) force-force correlator on the worldline of the quark. T κ = lim ω 0 ω ρ(ω), D = 2T 2 /κ. NNLO calculation available: ρ(ω) = smooth function ω g 2 ω 3. Result: 2πT D = Francis, Kaczmarek, Laine, Neuhaus PRD92 (2015)

16 Spectral function on the light-cone photon rate dγγ d 3 k D eff (k) = ρ V (k,k) k > 0 2χ qk lim ω 0 + ρii (ω,0) k = 0 ; dγ γ 3χ qω d 3 k = 2αemχq 3π 2 D eff (k) e βk 1. at k = 0, a narrow transport peak cannot be excluded large uncertainty on result for D. for k 2T, a more reliable result for D eff (k) is possible: spectral function expected to be smooth; fit ansatz: polynom up to ω = k 2 + π 2 T 2, perturbation theory beyond. N f = 0, analysis in the continuum; Ghiglieri, Kaczmarek, Laine, F. Meyer PRD 94, (2016)

17 The pion quasiparticle in the low-temperature phase

18 Chiral symmetry is spontaneously broken for T < T c: ψψ > 0. Goldstone theorem a divergent spatial correlation length m 1 π the limit m u,d 0. also: a massless real-time excitation exists: the pion quasiparticle. dispersion relation: [Son and Stephanov, PRD 66, (2002)] ω p = u(t ) m 2 π(t ) + p exists in T 100MeV: Two-loop chiral perturbation theory prediction for the pion quasiparticle mass u(t )m π(t ) [D. Toublan, PRD (1997)] Quasiparticle mass full: physical quark mass dashed: massless case.

19 key point: pion dominates parametrically the Euclidean two-point function of the axial charge density ( d 3 x e ip x τ ψγ0γ a 5 ψ) and its second 2 derivative at x 0 = β/2 0.6fm and p 300MeV inverse problem can be solved via the ansatz ρ A(ω, p, T ) = f 2 π(t ) (m 2 π(t ) + p 2 ) δ(ω 2 u 2 (T )(m 2 π(t ) + p 2 )) here m π(t ) and f π(t ) are determined from screening (=static) correlation functions; from time-dependent correlator: u = 0.75(2) and T = 0 : pion mass = 267(2) MeV T = 169MeV : quasiparticle mass = 223(4)MeV screening mass = 303(4)MeV. Simulation details: N f = 2 (no strange quark); lattice; Transition temperature T c 203MeV. How does this fit in with the success of the hadron-resonance gas model? Robaina et al. PRD 90 (2014) ; PRD 92 (2015)

20 Conclusion Significant progress in lattice QCD on near-equilibrium quantities: few-permille precision on correlation functions at small lattice spacings, even continuum in quenched approximation advanced weak-coupling calculations, effective field theories, exact sum rules,... provide crucial prior information on spectral function. many channels not discussed here: fate of quarkonium in the quark-gluon plasma, open-charm spectral functions, shear/sound channels,...

21 Backup slides

22 Thermal fluctuations and correlations Fig. from S. Borsanyi et al Light-quark number susceptibility: suggests that deconfinement occurs practically at the same temperature as chiral restoration. Successful predictions of the HRG.

23 Pion channel, continued: description of the lattice data GA(x0, T, p1) GA(x0, T, p2) GA(x0, T, p3) GA(x0, T, p4) GA(x0, T, p 5 ) ρ(ω,p) x0/a ω 1 3 d 3 x e ip x A a 0(x)A a 0(0) = 0 dω cosh[ω(β/2 x0)] 2π ρa (ω, p). sinh[ωβ/2] Ansatz : ρ A (ω, p) = a 1(p)δ(ω ω p) + a 2(p)(1 e ωβ )θ(ω c).