The topological susceptibility in the large
Transcription
The topological susceptibility in the large
New Frontiers in Theoretical Physics Cortona@GGI, 17-20 May 2016 The topological susceptibility in the large-N limit of Yang–Mills theory Marco Cèab in collaboration with M. García Veracd , L. Giustief , S. Schaeferc a) Scuola Normale Superiore b) INFN, Sezione di Pisa c) NIC, DESY, Zeuthen d) Humboldt-Universität zu Berlin e) Università di Milano-Bicocca f) INFN, Sezione di Milano-Bicocca 19 May 2016 Chiral symmetry breaking chiral symmetry: SU(3) L ×SU(3)R quark condensate = ψ̄ψ Pseudoscalar mesons [PDG 2014] - spontaneous χSB Meson Quark content Mass [MeV] SU(3)L × SU(3)R → SU(3)V + π d̄ À 139.57018(35) p π0 d d̄ − ̄ 2 134.9766(6) - 8 massless NG bosons: π− d ̄ 139.57018(35) π, K, η pseudoscalar + K s̄ 493.677(16) mesons K0 ds̄ 497.614(24) quark masses M = diag(m , md , ms ) K̄ 0 K− - explicit χSB - pseudo NG bosons: M2π = ( F 2 )mq Marco Cè – SNS & INFN, Pisa 19 May 2016 sd̄ s̄ 497.614(24) 493.677(16) η cos θη8 + sin θη0 η′ − sin θη8 + cos θη0 957.78(6) Àp η8 = (d d̄ + ̄ − 2ss̄) 6 Àp η0 = (d d̄ + ̄ + ss̄) 3 θ ≃ −11.4° 547.862(18) 1 Chiral symmetry breaking flavour singlet sector: U(1)B × U(1)A Pseudoscalar mesons U(1)B : baryon number - conserved [PDG 2014] Meson Quark content Mass [MeV] π+ π0 π− d̄ À p d d̄ − ̄ 2 139.57018(35) 134.9766(6) d ̄ 139.57018(35) K+ K0 K̄ 0 K− s̄ ds̄ sd̄ s̄ 493.677(16) 497.614(24) 497.614(24) 493.677(16) η cos θη8 + sin θη0 547.862(18) η′ − sin θη8 + cos θη0 957.78(6) Àp η8 = (d d̄ + ̄ − 2ss̄) 6 Àp η0 = (d d̄ + ̄ + ss̄) 3 θ ≃ −11.4° U(1)A - no parity partner in nature - broken explicitly by quark masses - the η′ is too heavy to be the 9thppseudo NG boson: Mη′ < 3Mπ [Weinberg 1975] η′ The U(1)A problem: Why is the much heavier than other pseudo Nambu–Goldstone bosons? Marco Cè – SNS & INFN, Pisa 19 May 2016 1 Topological charge The topological charge ∫ Q= d4 q() ∈ Z q() = 1 64π 2 εμνρσ Fμν ()Fρσ () - divergence of a gauge-dependent current q() = ∂μ Kμ () - U(1)A is broken by an anomaly ∝ 2Nƒ q(), a purely quantum effect Marco Cè – SNS & INFN, Pisa 19 May 2016 2 Topological charge The topological charge ∫ Q= d4 q() ∈ Z q() = 1 64π 2 εμνρσ Fμν ()Fρσ () - divergence of a gauge-dependent current q() = ∂μ Kμ () - U(1)A is broken by an anomaly ∝ 2Nƒ q(), a purely quantum effect θ-term in YM (or QCD) Lagrangian: θ-vacua L= 1 4g20 Fμν ()Fμν () − θq() - imaginary in Euclidean space: sign problem for θ ̸= 0 - strong CP problem: neutron EDM ⇒ |θ| < 10−10 Marco Cè – SNS & INFN, Pisa 19 May 2016 2 Topological charge The topological charge ∫ Q= d4 q() ∈ Z q() = 1 64π 2 εμνρσ Fμν ()Fρσ () - divergence of a gauge-dependent current q() = ∂μ Kμ () - U(1)A is broken by an anomaly ∝ 2Nƒ q(), a purely quantum effect θ-term in YM (or QCD) Lagrangian: θ-vacua L= 1 4g20 Fμν ()Fμν () − θq() - imaginary in Euclidean space: sign problem for θ ̸= 0 - strong CP problem: neutron EDM ⇒ |θ| < 10−10 The topological susceptibility is the two-point function of q() ∫ χ= Marco Cè – SNS & INFN, Pisa d4 〈q()q(0)〉 19 May 2016 2 The Witten–Veneziano mechanism A mechanism to solve the U(1)A problem based on the vanishing of the anomaly in the large-N limit [Witten 1979; Veneziano 1979] - anomalous chiral Ward identity ∂μ A0μ = 2mP0 () − 2Nƒ q() - N → ∞: U(1)A is restored, the η′ is a NG boson - in Yang–Mills theory, χYM ̸= 0 and O(1) in large-N - dynamical quarks are an O(1/ N) effect - no θ dependence in QCD with m = 0 quarks ⇒ χ = 0 ⇒ The η′ gets a mass M2η′ = O(1/ N) The WV formula, valid in the chiral and N → ∞ limits F 2 M2η′ 2Nƒ = χYM - links Mη′ in QCD with χ measured in YM theory - is a non-perturbative test of the quantization of QCD Marco Cè – SNS & INFN, Pisa 19 May 2016 3 large-N chiral perturbation theory Simultaneous expansion in mq , 1/ N and p2 [Di Vecchia, Veneziano 1980, Kaiser, Leutwyler 2000] M2η′ = 2mq F2 + 2Nƒ τ F2 + O m2q , 1 N2 , p2 - M2η′ linear in expansion parameters mq , 1/ F 2 ∝ 1/ N - τ = limN→∞ χYM is a low-energy constant - fitting LO χPT with real world [Veneziano 1979] Fπ2 2 + O( 1/3 ) Mη + M2η′ − 2M2K = (180 MeV)4 = χYM 6 N→∞ - N → ∞ YM theory is not physical Marco Cè – SNS & INFN, Pisa 19 May 2016 4 large-N chiral perturbation theory Simultaneous expansion in mq , 1/ N and p2 [Di Vecchia, Veneziano 1980, Kaiser, Leutwyler 2000] M2η′ = 2mq F2 + 2Nƒ τ F2 + O m2q , 1 N2 , p2 - M2η′ linear in expansion parameters mq , 1/ F 2 ∝ 1/ N - τ = limN→∞ χYM is a low-energy constant - fitting LO χPT with real world [Veneziano 1979] Fπ2 2 + O( 1/3 ) Mη + M2η′ − 2M2K = (180 MeV)4 = χYM 6 N→∞ - N → ∞ YM theory is not physical ⇒ measure χYM in lattice-discretized large-N YM theory N→∞ - using lattice SU(3) YM [Cè, Consonni, Engel, Giusti 2015] χYM = (180.5(43) MeV)4 = χYM + O 1 32 N=3 N→∞ - O 1 N2 corrections are likely small Marco Cè – SNS & INFN, Pisa 19 May 2016 4 Topological charge on the lattice (I) Problem: define the topological charge on the lattice ⇒ naïve discretization of q() qL () = 1 64π 2 εμνρσ Fμν ()Fρσ () () given in terms of plaquettes (clover definition) - Fμν Fμν () = 1 4 P ν μ with P (U) projecting over su(3) Lie algebra Marco Cè – SNS & INFN, Pisa 19 May 2016 5 Topological charge on the lattice (I) Problem: define the topological charge on the lattice ⇒ naïve discretization of q() qL () = 1 64π 2 εμνρσ Fμν ()Fρσ () () given in terms of plaquettes (clover definition) - Fμν Fμν () = 1 4 P ν μ with P (U) projecting over su(3) Lie algebra - qL () need multiplicative renormalization [Allés et al. 1995] - | − y| → 0: singularities in 〈qL ()qL (y)〉 ⇒ qL () does not satisfy anomalous chiral Ward identities Marco Cè – SNS & INFN, Pisa 19 May 2016 5 Topological charge on the lattice (II) Definitions of Q satisfying Ward identities: - fermionic definition, using a Dirac operator satisfying Ginsparg–Wilson [Ginsparg, Wilson 1982], as Neuberger–Dirac operator DN [Neuberger 1998] qN () = − ▶ ▶ ▶ ̄ 24 tr γ5 DN (, ) ∑ index theorem: QN = 4 qN () = − 2̄ Tr γ5 DN = n+ − n− ∑ no renormalization of 4 〈qN ()qN (0)〉 satisfy anomalous chiral Ward Identities [Giusti, Rossi, Testa, Veneziano 2002; Giusti, Rossi, Testa 2004; Lüscher 2004] ▶ computationally very expensive Marco Cè – SNS & INFN, Pisa 19 May 2016 6 Topological charge on the lattice (II) Definitions of Q satisfying Ward identities: - fermionic definition, using a Dirac operator satisfying Ginsparg–Wilson [Ginsparg, Wilson 1982], as Neuberger–Dirac operator DN [Neuberger 1998] qN () = − ▶ ▶ ▶ ̄ 24 tr γ5 DN (, ) ∑ index theorem: QN = 4 qN () = − 2̄ Tr γ5 DN = n+ − n− ∑ no renormalization of 4 〈qN ()qN (0)〉 satisfy anomalous chiral Ward Identities [Giusti, Rossi, Testa, Veneziano 2002; Giusti, Rossi, Testa 2004; Lüscher 2004] ▶ computationally very expensive - spectral projectors [Giusti, Lüscher 2009] - naïve qL () at positive YM gradient flow times Marco Cè – SNS & INFN, Pisa 19 May 2016 [Lüscher 2010] 6 The Yang–Mills gradient flow The gradient flow is the solution of the initial value problem [Lüscher 2010] d dt Vμ (t, ) = −g20 ∂μ, SW [V(t)] Vμ (t, ) Vμ (0, ) = Uμ () - proportional to the negative gradient of YM action p - Vμ (t, ): configurations smoothed within a radius 8t - universality: (multi)local fields at t > 0 are finite and do not require renormalization [Lüscher, Weisz 2011] - numerical simulations: integrate numerically ODE with structure-preserving Runge–Kutta method - scale setting with energy density E(t): t0 reference scale t 2 〈E(t)〉t=t0 = 0.3 t0 = (0.176(4) fm)2 in YM theory Marco Cè – SNS & INFN, Pisa 19 May 2016 7 Universality of gradient flow definition Result in [Cè, Consonni, Engel, Giusti 2015]: - in the continuum, using small-t expansion ∂t q(t, ) = ∂ρ ρ (t, ), density-chain expression of Q, OP expansion, t-independence of χ ∫ ∫ d4 〈q(t, )q(t, 0)〉 = d4 〈q()q(0)〉 - on the lattice, with Neuberger definition qN (), ∑ ∑ 〈qN (t, )qN (t, 0)〉 = lim 4 〈qN ()qN (0)〉 lim 4 →0 Marco Cè – SNS & INFN, Pisa →0 19 May 2016 8 Universality of gradient flow definition Result in [Cè, Consonni, Engel, Giusti 2015]: - in the continuum, using small-t expansion ∂t q(t, ) = ∂ρ ρ (t, ), density-chain expression of Q, OP expansion, t-independence of χ ∫ ∫ d4 〈q(t, )q(t, 0)〉 = d4 〈q()q(0)〉 - on the lattice, with Neuberger definition qN (), ∑ ∑ 〈qN ()qN (0)〉 〈qL (t, )qL (t, 0)〉 = lim 4 lim 4 →0 →0 - Universality at t > 0 ⇒ naïve discretization qL (t, ) Marco Cè – SNS & INFN, Pisa 19 May 2016 8 Universality of gradient flow definition Result in [Cè, Consonni, Engel, Giusti 2015]: - in the continuum, using small-t expansion ∂t q(t, ) = ∂ρ ρ (t, ), density-chain expression of Q, OP expansion, t-independence of χ ∫ ∫ d4 〈q(t, )q(t, 0)〉 = d4 〈q()q(0)〉 - on the lattice, with Neuberger definition qN (), ∑ ∑ 〈qN ()qN (0)〉 〈qL (t, )qL (t, 0)〉 = lim 4 lim 4 →0 →0 - Universality at t > 0 ⇒ naïve discretization qL (t, ) Cumulants CL,n (t) of the topological charge given by the naïve definition at strictly positive flow time have a finite and unambiguous continuum limit that satisfies anomalous chiral Ward identities lim CL,n (t) = lim CN,n = Cn , →0 Marco Cè – SNS & INFN, Pisa →0 19 May 2016 (t > 0) 8 Autocorrelation Topological observables are known to be changed slowly by the Monte Carlo updates PBC Q(t0 ) 10 - Q is changed through lattice artifacts 5 - topological charge freezing in the continuum 0 - and for larger N [Del Deb- bio, Panagopoulos, Vicari 2002] −5 Q τint ∝ ecN −10 Monte carlo history of Q at t0 , SU(6) lattice at ≃ 0.1 fm, PBC Marco Cè – SNS & INFN, Pisa −1 - applies also to χ 1 2 χ = TV Q 19 May 2016 9 Autocorrelation Topological observables are known to be changed slowly by the Monte Carlo updates PBC OBC 10 - our solution: open boundary conditions (OBC) in the time direction [Lüscher, Schaefer 2011] Q(t0 ) 5 τint ∝ −2 (N?) Q 0 - Q is no more an integer −5 - time translation invariance is lost −10 Monte carlo history of Q at t0 , SU(6) lattice at ≃ 0.1 fm, OBC Marco Cè – SNS & INFN, Pisa - excited states - O() discretization effects - ⇒ modify χ definition 19 May 2016 9 Topological susceptibility with OBC Definition based on correlators of topological charge on a ∑ time-slice: q̄(t, 0 ) = ⃗ q(t, ) C̄(t, Δ) = 1 (T−2d−Δ)V T−1−d−Δ ∑ χ(t) = 〈q̄(t, 0 )q̄(t, 0 + Δ)〉 0 =d +r ∑ C̄(t, |Δ|) Δ=−r - d fixed in physical units to avoid excited states from open boundaries (and boundaries cut-off effects) p d = 7.5 t0 ≃ 1.3 fm - r fixed in physical units to avoid summing over noise p r = 6.9 t0 ≃ 1.2 fm - systematics checked to be negligible Marco Cè – SNS & INFN, Pisa 19 May 2016 10 (preliminary) Numerical results [Cè, Garcia Vera, Giusti, Schaefer to appear] Marco Cè – SNS & INFN, Pisa 19 May 2016 11 Simulation results N = 4, 5, 6, three lattice spacings - gradient flow up to t0 [fm] 7.4 ×10−4 0.05 0.07 0.08 0.09 0.10 2 t e(t, ) SU(4) SU(5) 7.2 SU(6) = 0.1125 N N−1 2 t=t0 7.0 - t0 used as a scale: t02 χ(t0 ) t02 χ 6.8 - ∼ 1 % statistical error on the single lattice result - no clear 2 or 1/ N2 scaling 6.6 6.4 6.2 6.0 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 2 / t0 Marco Cè – SNS & INFN, Pisa 19 May 2016 12 Simulation results N = 4, 5, 6, three lattice spacings - gradient flow up to t0 [fm] 7.4 ×10−4 0.05 0.07 0.08 0.09 0.10 2 t e(t, ) SU(4) SU(5) 7.2 SU(6) = 0.1125 N N−1 2 t=t0 SU(3) 7.0 - t0 used as a scale: t02 χ(t0 ) t02 χ 6.8 - ∼ 1 % statistical error on the single lattice result 6.6 6.4 6.2 6.0 0.00 0.05 0.10 0.15 0.20 2 / t0 0.25 0.30 0.35 - including SU(3) results [Cè, Consonni, Engel, Giusti 2015] Marco Cè – SNS & INFN, Pisa 19 May 2016 12 Discretization effects vs. N t-dependent discretization effects can be studied taking the ratio χ(t)/ χ(t0 ) at different flow times t [fm] 0.05 1.05 0.07 0.08 0.09 0.10 1.00 0.95 χ(t)/ χ(t0 ) 0.90 - very small errors, smaller than marker size - universality in the continuum 0.85 lim χ(t)/ χ(t0 ) = 1 0.80 →0 0.75 0.70 0.65 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 2 / t0 fit with 1 + c1 (2 / t0 ) + c2 (2 / t0 )2 Marco Cè – SNS & INFN, Pisa - t-dependent discretization effects nearly N-independent - true also for SU(3) data 19 May 2016 13 Continuum and N → ∞ limit - combined fit with [fm] 7.4 ×10−4 0.05 0.07 0.08 0.09 0.10 c0,0 + c1,0 2 / t0 + c0,1 1/ N2 SU(4) SU(5) 7.2 SU(6) SU(3) 7.0 t02 χ 6.8 6.6 6.4 6.2 6.0 0.00 0.05 0.10 0.15 0.20 2 / t0 0.25 0.30 0.35 - using only two finest lattices - 2 / t0 discretization effects mainly fixed by SU(3) data - χ2 dof = 0.90 - ∼ 1.5 % statistical error (on a d = 4 observable!) Our (preliminary) result: t02 χN→∞ = 6.98(11) × 10−4 Marco Cè – SNS & INFN, Pisa 19 May 2016 14 Comparison with other results Our (preliminary) result: t02 χN→∞ = 6.98(11) × 10−4 , χN→∞ = (182.3(7)(44) MeV)4 5 % correction with respect to t02 χN=3 = 6.67(7) × 10−4 , [Cè, Consonni, Engel, Giusti 2015] χN=3 = (180.5(5)(43) MeV)4 Status of the art ≳ 10 years ago [Review by: Vicari, Panagopoulos 2009] - cooling definition of topological charge - coarser lattice spacings - statistical errors ≳ 6 % p - scale setting with σ = 440 MeV χN→∞ = 0.0200(43)σ 2 = (165(9) MeV)4 2 χN→∞ = 0.0221(14)σ = (170(3) MeV) 4 χN→∞ = 0.0248(18)σ 2 = (175(3) MeV)4 Marco Cè – SNS & INFN, Pisa 19 May 2016 [Lucini, Teper 2001] [Del Debbio, Panagopoulos, Vicari 20 [Lucini, Teper, Wenger 2005] 15 Conclusions & Outlook Conclusions: - we measured the topological susceptibility of SU(N) YM theory in the N → ∞ limit ▶ ▶ with an O(1 %) statistical accuracy and systematics under control - we obtained this result with open boundary conditions in the time direction ▶ they work to avoid topological charge freezing with N Marco Cè – SNS & INFN, Pisa 19 May 2016 16 Conclusions & Outlook Conclusions: - we measured the topological susceptibility of SU(N) YM theory in the N → ∞ limit ▶ ▶ with an O(1 %) statistical accuracy and systematics under control - we obtained this result with open boundary conditions in the time direction ▶ they work to avoid topological charge freezing with N Outlook: - 1 % accuracy should be sufficient for the future - verify the Witten–Veneziano relation on both sides: compute Mη′ on the lattice for N > 3, with dynamical fermions ▶ ▶ ▶ exponential degradation in S/N of correlators topological charge freezing with dynamical fermions long term project Marco Cè – SNS & INFN, Pisa 19 May 2016 16 Thanks for your attention! Backup Quantum Chromodynamics QCD is the theory of strong interactions L= 1 4g20 [Fritzsch, Gell-Mann 1972] Fμν ()Fμν () - SU(3) gauge group, imaginary-time YM Lagrangian - gluons Aμ : 8 vector bosons, g20 : bare gauge coupling - asymptotic free theory [Gross, Wilczek 1973; Politzer 1973] - confined spectrum: mass gap, glueballs Marco Cè – SNS & INFN, Pisa 19 May 2016 i Quantum Chromodynamics QCD is the theory of strong interactions L= 1 4g20 [Fritzsch, Gell-Mann 1972] Fμν ()Fμν () + ψ̄()(γμ Dμ + M)ψ() - SU(3) gauge group, imaginary-time QCD Lagrangian - gluons Aμ : 8 vector bosons, g20 : bare gauge coupling - asymptotic free theory (if Nƒ < ( 11/2 )N) - confined spectrum: mass gap, mesons, barions - quarks ψ: 3Nƒ Dirac fermions, mass matrix M - Nƒ : number of flavours (Nƒ = 2 or 3 light quarks) - massless quarks M = 0 ⇒ global symmetry of the action SU(Nƒ )L × SU(Nƒ )R × U(1)B × U(1)A (at classical level) Marco Cè – SNS & INFN, Pisa 19 May 2016 i The dilute instanton gas A solution of the U(1)A problem proposed by ’t Hooft using instantons: dilute instanton gas approximation [’t Hooft 1976] - Semiclassical approximation - Free energy dependence on θ F(θ) = − ln Z(θ) = −VA(cos θ − 1) - Arbitrary normalization A ⇒ no prediction for χYM 2F 1 d 1 Q2 = =A χYM = 2 V V dθ θ=0 - No IR bound on instantons size - Expected to be valid at high temperature Marco Cè – SNS & INFN, Pisa 19 May 2016 ii Higher moments Higher moments of the probability distribution of the topological charge Q in Yang–Mills theory - to confront with dilute instanton gas model prediction - higher cumulants are obtained deriving F(θ) 1 2n con (−1)n+1 d2n = Cn ≡ Q F(θ) V V dθ2n θ=0 DIG: Cn = A large-N: Cn ∝ N−2(n−1) - we define the ratio between the 4th and the 2nd cumulant con 2 C2 Q4 Q4 − 3 Q2 R≡ = = C1 Q2 Q2 DIG: R = 1 Marco Cè – SNS & INFN, Pisa large-N: R ∝ 19 May 2016 1 N2 iii Lattice regularization (I) Euclidean YM theory is discretized on a four-dimensional hypercubic lattice with lattice spacing [Wilson 1974] ν μ - gauge field defined on links between lattice sites: Uμ () = μ + μ̂ - plaquettes: Uμν () = ∈ SU(3) ν μ - finite volume V = L4 with periodic boundary conditions - 18 × 4 × L4 real variables Marco Cè – SNS & INFN, Pisa 19 May 2016 iv Lattice regularization (II) under gauge transformation g() ∈ SU(3) Uμ () → g()Uμ ()g−1 (+ μ̂) Uμν () → g()Uμν ()g−1 () Wilson plaquette action [Wilson 1974] ∑ ∑ 1 1 − Re tr Uμν () SW [U] = β N μ<ν β= 2N g20 - gauge invariant at finite lattice spacing ¦ © - Uμ () → exp Aμ ()T ∑ 1 SW [U] → 4 F ()Fμν () + O 2 2 μν 4g0 - naïve continuum limit: lim→0 SW = SYM - O 2 discretization artefacts, = (g0 ) - continuum limit → 0 ⇔ critical point g0 → 0 (β → ∞) Marco Cè – SNS & INFN, Pisa 19 May 2016 v Path integral Monte Carlo YM theory path integral ∫ 1 〈O〉 = D[U]O[U] exp{−SYM [U]} Z - V = (10)4 lattice: 320 000 integrations Marco Cè – SNS & INFN, Pisa 19 May 2016 vi Path integral Monte Carlo YM theory path integral ∫ 1 〈O〉 = D[U]O[U] exp{−SYM [U]} Z - V = (10)4 lattice: 320 000 integrations Markov chain Monte Carlo - importance sampling: generate n gauge field configurations {U } with pdf dp = D[U]e−SYM [U] ∑n n→∞ - Ō = (1/ n) =1 O[U ] −−−→ 〈O〉 Cabibbo–Marinari updates [Cabibbo, Marinari 1982] - local updates: sweep all the links on the lattice - heat bath + overrelaxation sweeps Configurations are correlated - critical slowing down: τint ∝ −2 - topological charge changed through lattice artefacts: → 0 ⇒ topological freezing Marco Cè – SNS & INFN, Pisa 19 May 2016 vi Universality of lim→0 χ(t) (I) - In the continuum, to all order of perturbation theory, using ∂t q(t, ) = ∂ρ ρ (t, ) is valid the small-t expansion ∫ t 〈q(t, )O(y)〉 = 〈q()O(y)〉 + ∂ρ dt ′ ρ (t ′ , )O(y) , 0 where ρ (t, ) = 1 ε 8π 2 μνρσ tr Fμν (t, )Dα Fασ (t, ) is a gauge-invariant, pseudovector dimension-5 field - No local d < 5 fields with ρ transformation properties ⇒ not singular at t = 0 〈q()O(y)〉 = lim 〈q(t, )O(y)〉, t→0 ( ̸= y) or 〈q(t, )O(y)〉 = 〈q()O(y)〉 + O(t), Marco Cè – SNS & INFN, Pisa 19 May 2016 ( ̸= y) vii Universality of lim→0 χ(t) (II) - With ∑ O = q, using the density-chain expression 4 q() = ∑ 5 20 −m z 〈P51 (z1 )S12 (z2 )S23 (z3 )S34 (z4 )S45 (z5 )〉F the leading order divergence in 〈q()P51 (z1 )S12 (z2 )S23 (z3 )S34 (z4 )S45 (z5 )〉 for → zj , computable in perturbation theory, is →0 q()Sj (0) −−→ c()Pj (0) where c() ∝ ||−4 for → 0 - Applying small-t expansion q(t, )Sj (z)O(y) = ∫ t q()Sj (z)O(y) + ∂ρ dt ′ ρ (t ′ , )Sj (z)O(y) , 0 c() = ∂μ μ () and do not contribute to the integrated correlation function Marco Cè – SNS & INFN, Pisa 19 May 2016 viii Universality of lim→0 χ(t) (III) - No other d ≤ 4 gauge-invariant pseudoscalar operators implies ∑ 〈qN (t, 0)qN ()〉 = finite, lim Zq 4 (t > 0) →0 - Replacing qN () with density-chain expression 4 ∑ 〈qN (t, 0)qN ()〉 = − m5 20 ∑ 〈qN (t, 0)P51 (z1 )S12 (z2 ) . . . S45 (z5 )〉 z the r.h.s. is finite for t > 0 - This implies that Zq = 1 and therefore lim 〈qN ()qN (0)〉 = 〈q()q(0)〉, →0 Marco Cè – SNS & INFN, Pisa 19 May 2016 ( ̸= 0) ix Universality of lim→0 χ(t) (IV) - Using small-t expansion again and t-independence of χ, ∑ ∑ 〈qN ()qN (0)〉 = lim 4 〈qN (t, )qN (t, 0)〉 lim 4 →0 →0 - Since qN () and the naïve discretization qL () have the same naïve continuum limit, at t > 0 thanks to universality ∑ ∑ 〈qN ()qN (0)〉 = lim 4 〈qL (t, )qL (t, 0)〉 lim 4 →0 →0 - By iterating the procedure, it extends to higher cumulants lim CN,n = lim CL,n (t), →0 Marco Cè – SNS & INFN, Pisa →0 19 May 2016 (t > 0) x Lattices details N β T/ L/ t0 / 2 [fm] L[fm] 4 10.92 11.14 11.35 64 80 96 16 20 24 2.9900(7) 4.5207(8) 6.4849(16) 0.098 0.080 0.067 1.58 1.60 1.60 5 17.32 17.67 18.01 64 80 96 16 20 24 3.0636(7) 4.6751(8) 6.8151(17) 0.097 0.079 0.065 1.55 1.57 1.56 6 25.15 25.68 26.15 64 80 96 16 20 24 3.0824(4) 4.8239(9) 6.9463(13) 0.097 0.077 0.065 1.55 1.55 1.55 Physical units using Marco Cè – SNS & INFN, Pisa p t0 = 0.17 fm 19 May 2016 xi Reference flow-time The reference flow-time t0 is defined by t 2 〈E(t)〉t=t0 = 0.3 - Easy to measure with great statistical accuracy - Our result p 8t0 = 0.941(7) r0 - using FK measured in quenched QCD for scale setting t0 = (0.176(4) fm)2 Marco Cè – SNS & INFN, Pisa 19 May 2016 xii