YITP seminarTakashi Umeda (YITP, Kyoto Univ.)1 A new approach to QCD thermodynamics on the lattice Takashi Umeda (YITP, Kyoto Univ.) for WHOT-QCD Collaboration YITP seminar, Kyoto, Japan, 17 Dec /31 This talk is (partly) based on arXiv: [hep-lat] T.U, S. Ejiri, S. Aoki, T. Hatsuda, K. Kanaya, Y. Maezawa, and H. Ohno (WHOT-QCD Collaboration)

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)2 Contents of this talk /31 Our aim is to investigate QCD Thermodynamics with Wilson-type quarks QCD Thermodynamics with Wilson-type quarks Brief review on Lattice QCD at finite T (zero μ) Brief review on Lattice QCD at finite T (zero μ) Why do we need “Hot QCD with Wilson-type quarks” ? Why do we need “Hot QCD with Wilson-type quarks” ? Why is “Hot QCD with Wilson-type quarks” difficult ? Why is “Hot QCD with Wilson-type quarks” difficult ? How do we overcome the difficulties ? How do we overcome the difficulties ? - We propose “T-integration method” - Test with the SU(3) gauge theory Summary and Outlook Summary and Outlook

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)3 Introduction /31 Physics in Lattice QCD at finite temperature Physics in Lattice QCD at finite temperature Phase diagram in (T, μ, m ud, m s ) Phase diagram in (T, μ, m ud, m s ) Transition temperature Transition temperature Equation of state ( e, p, s,...) Equation of state ( e, p, s,...) Heavy quarkonium Heavy quarkonium Transport coefficients (shear/bulk viscosity) Transport coefficients (shear/bulk viscosity) Finite chemical potential Finite chemical potential etc... etc... These are important to study - Quark Gluon Plasma in Heavy Ion Collision exp. - Early universe - Neutron star - etc... quantitative studies qualitative studies

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)4 Hot QCD on the lattice /31 Finite T Field Theory on the lattice 4dim. Euclidean lattice 4dim. Euclidean lattice gauge field U μ (x)  periodic B.C. gauge field U μ (x)  periodic B.C. quark field q(x)  anti-periodic B.C. quark field q(x)  anti-periodic B.C. Temperature T=1/(N t a) Temperature T=1/(N t a) Input parameters : β(=6/g 2 ) (lattice gauge coupling) (Nf=2+1 QCD) am ud (light (up & down) quark masses) am s (strange quark mass) am s (strange quark mass) N t (temperature) N t (temperature) (*) lattice spacing “a” is not an input parameter (*) lattice spacing “a” is not an input parameter a=a(β, am ud, am s ) Temperature T=1/(N t a) is varied by a at fixed N t

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)5 Fermions on the lattice /31 Lattice QCD Path integral is carried out by Monte Carlo Integration Path integral is carried out by Monte Carlo Integration QCD action is defined on the lattice QCD action is defined on the lattice Fermion doubling problem Fermion doubling problem naive discretization causes 2 4 doublers naive discretization causes 2 4 doublers Nielsen-Ninomiya’s No-go theorem Nielsen-Ninomiya’s No-go theorem  Doublers appear unless chiral symmetry is broken Staggered (KS) fermion Staggered (KS) fermion - 16 doublers = 4 spinors x 4 flavors (“tastes”) - 16 doublers = 4 spinors x 4 flavors (“tastes”) - Remnant U(1) symmetry - Remnant U(1) symmetry - Fourth root trick : still debated - Fourth root trick : still debated - Numerical cost is low - Numerical cost is low and...

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)6 Fermions on the lattice /31 Wilson fermion Wilson fermion - adds the Wilson term to kill extra doublers - adds the Wilson term to kill extra doublers - breaks chiral symmetry explicitly  additive mass renorm. - breaks chiral symmetry explicitly  additive mass renorm. - Improved version (Clover fermion) is widely used. - Improved version (Clover fermion) is widely used. - Numerical cost is moderate - Numerical cost is moderate Domain Wall fermion Domain Wall fermion - 5dim. formulation - 5dim. formulation - Symmetry breaking effect m res  0 as N 5  ∞ - Symmetry breaking effect m res  0 as N 5  ∞ - Numerical cost is high - Numerical cost is high Overlap fermion Overlap fermion - Exact chiral symmetry - Exact chiral symmetry - Numerical cost is very high - Numerical cost is very high Wilson-typefermions

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)7 Recent lattice calculations of EOS RBC-Bielefeld: N t =4,6,8 Staggered (p4) quark pion mass ~ 220MeV, N f =2+1 pion mass ~ 220MeV, N f =2+1 Phys. Rev. D77 (2008) Phys. Rev. D77 (2008) MILC: N t =4,6,8 Staggered (Asqtad) quark pion mass ~ 220MeV, N f =2+1 pion mass ~ 220MeV, N f =2+1 Phys. Rev. D75 (2007) Phys. Rev. D75 (2007) Wuppertal: N t =4,6 Staggered (stout) quark pion mass ~ 140MeV, N f =2+1 pion mass ~ 140MeV, N f =2+1 JHEP 0601 (2006) 089 JHEP 0601 (2006) 089 CP-PACS: N t =4,6 Wilson (MFI Clover) quark pion mass ~ 500MeV, N f =2 pion mass ~ 500MeV, N f =2 Phys. Rev. D64 (2001) Phys. Rev. D64 (2001) /31 Hot-QCDCollab. (2007 ～ )

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)8 Contents of this talk /31 Our aim is to investigate QCD Thermodynamics with Wilson-type quarks QCD Thermodynamics with Wilson-type quarks Brief review on Lattice QCD at finite T (zero μ) Brief review on Lattice QCD at finite T (zero μ) Why do we need “Hot QCD with Wilson-type quarks” ? Why do we need “Hot QCD with Wilson-type quarks” ? Why is “Hot QCD with Wilson-type quarks” difficult ? Why is “Hot QCD with Wilson-type quarks” difficult ? How do we overcome the difficulties ? How do we overcome the difficulties ? - We propose “T-integration method” - Test with the SU(3) gauge theory Summary and Outlook Summary and Outlook

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)9 Problems in QCD Thermo. with KS fermions /31 Many QCD thermo. calc. were done with KS fermions. Phase diagram Phase diagram N f =2 massless QCD  O(4) critical exponets N f =2 massless QCD  O(4) critical exponets KS fermion does not exhibit expected O(4) scaling KS fermion does not exhibit expected O(4) scaling (Wilson fermion shows O(4), but at rather heavy masses) (Wilson fermion shows O(4), but at rather heavy masses) Transition temperature (crossover transition in KS studies) Transition temperature (crossover transition in KS studies) KS results are not consistent with each other KS results are not consistent with each other MILC : 169(12)(4)MeV(*) Phys. Rev. D71 (2005) MILC : 169(12)(4)MeV(*) Phys. Rev. D71 (2005) RBC-Bi : 192(7)(4)MeV Phys. Rev. D74 (2006) RBC-Bi : 192(7)(4)MeV Phys. Rev. D74 (2006) Wuppertal : 151(3)(3)MeV Phys. Lett. B643 (2006) 46 Wuppertal : 151(3)(3)MeV Phys. Lett. B643 (2006) 46 (*)T c at m q =0 (*)T c at m q =0 EOS EOS KS results are not consistent with each other KS results are not consistent with each other MILC & RBC-Bi are consistent ( N t =4,6,8 ) MILC & RBC-Bi are consistent ( N t =4,6,8 ) Wuppertal ( N t =4,6 ) Wuppertal ( N t =4,6 )

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)10 EOS with KS fermions /31 M.Chen et al. (RBC-Bielefeld) Phys. Rev. D77 (2008) Y.Aoki et al. (Wuppertal) JHEP 0601 (2006) 089.

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)11 EOS with KS fermions /31 RBC-Bi vs Wuppertal for pressure p/T 4

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)12 EOS with KS fermions /31 N t is small yet ? N t is small yet ? rooted trick ? rooted trick ? flavor symmetry violation ? flavor symmetry violation ? other systematic errors ? other systematic errors ? We have to study the QCD-EOS with Wilson-type fermions !! RBC-Bi vs Wuppertal for energy density e/T 4

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)13 Contents of this talk /31 Our aim is to investigate QCD Thermodynamics with Wilson-type quarks QCD Thermodynamics with Wilson-type quarks Brief review on Lattice QCD at finite T (zero μ) Brief review on Lattice QCD at finite T (zero μ) Why do we need “Hot QCD with Wilson-type quarks” ? Why do we need “Hot QCD with Wilson-type quarks” ? Why is “Hot QCD with Wilson-type quarks” difficult ? Why is “Hot QCD with Wilson-type quarks” difficult ? How do we overcome the difficulties ? How do we overcome the difficulties ? - We propose “T-integration method” - Test with the SU(3) gauge theory Summary and Outlook Summary and Outlook - Hot QCD requires huge computational cost computational cost  e.g. EOS calculation  e.g. EOS calculation - Wilson-type quarks requires large lattice cutoff simulations large lattice cutoff simulations

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)14 for large volume system Lattice QCD can not directly calculate the partition function however its derivative is possible high temp. low temp. with p ⋍ 0 One can obtain p as the integral of derivative of p Integral method to calculate pressure p/T 4 T=0 subtraction /31

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)15 In case of N f =2+1 QCD there are three (bare) parameters: β, (am ud ) and (am s ) β mqmq The physics (observables) should be kept along the integral path.  Line of Constant Physics (LCP) defined at T=0 Inaccuracy of the LCP is a source of systematic error in EOS. Integral on the path is carried out numerically. T=0 subtractions are necessary at each point. low T (small 1/a) p 0 ≃ 0 high T (large 1/a) p(T) parameter space Line of constant physics (LCP) integral path xxxx x x x x x x /31

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)16 Numerical cost for EOS calculations /31 In the EOS calculation, T=0 calculations dominate in spite of T>0 study. T=0 calculations dominate in spite of T>0 study. Search for a Line of Constant Physics (LCP) Search for a Line of Constant Physics (LCP) T=0 subtraction at each temperature T=0 subtraction at each temperature T=0 simulations are time consuming. - N t is sufficiently large (e.g x24 at T=0, 24 3 x6 at T>0 ) - N t is sufficiently large (e.g x24 at T=0, 24 3 x6 at T>0 ) - small Dirac eigenvalue (larger cost for D -1 (x,y)) - small Dirac eigenvalue (larger cost for D -1 (x,y)) (cost at T=0) = (5~20) x (cost at T>0) (cost at T=0) = (5~20) x (cost at T>0) Even with the Staggered fermions, EOS at N t =8 is the best with current computer resources. EOS at N t =8 is the best with current computer resources.

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)17 Further problems in Wilson-type quarks /31 CP-PACS, Phys. Rev. D73 (2006) GeV 2.0GeV 3.0GeV lattice cutoff Nonperturbative improvement of Wilson fermions : clover coefficient c sw by the Schrodinger functional method Large uncertainty of c sw at 1/a < 2GeV at 1/a < 2GeV

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)18 Further problems in Wilson-type quarks /31 RBC & Hot-QCD, Lattice 2008 Residual quark mass is not well controlled at 1/a < 2GeV (at typical L s ) 1.5GeV Residual quark mass m res in Domain Wall fermion RBC & HOT-QCD Collab. gave up N t =8, L s =32 Domain Wall project.  N t =8, L s =96 project on progress  N t =8, L s =96 project on progress

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)19 Further problems in Wilson-type quarks /31 Coarse lattice generally causes various problems.  1/a > 2GeV is safe to calculate physics at T=0 & T>0.  1/a > 2GeV is safe to calculate physics at T=0 & T>0. overlap fermion JLQCD, TQFT(YITP) 2008

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)20 How large Nt is safe ? /31 safe region ? integral method needs low T (p=0) T vs 1/a at various Nt (3fm/a) 3 = KS fermion results KS fermion results are not sufficient are not sufficient to finalize QCD-EOS to finalize QCD-EOS in lattice QCD in lattice QCD EOS calc. is very costly EOS calc. is very costly many T=0 simulations many T=0 simulations Wilson-type fermions Wilson-type fermions needs larger 1/a needs larger 1/a Situation for T c calc. Situation for T c calc. is similar to the EOS is similar to the EOS Phase diagram study Phase diagram study needs more cost !! needs more cost !!

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)21 Contents of this talk /31 Our aim is to investigate QCD Thermodynamics with Wilson-type quarks QCD Thermodynamics with Wilson-type quarks Brief review on Lattice QCD at finite T (zero μ) Brief review on Lattice QCD at finite T (zero μ) Why do we need “Hot QCD with Wilson-type quarks” ? Why do we need “Hot QCD with Wilson-type quarks” ? Why is “Hot QCD with Wilson-type quarks” difficult ? Why is “Hot QCD with Wilson-type quarks” difficult ? How do we overcome the difficulties ? How do we overcome the difficulties ? - We propose “T-integration method” - Test with the SU(3) gauge theory Summary and Outlook Summary and Outlook

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)22 Fixed scale approach to study QCD thermodynamics /31 Temperature T=1/(N t a) is varied by N t at fixed a(β, m ud, m s ) safe region ? integral method needs low T (p=0) (3fm/a) 3 = fixed scale approach Advantages Advantages - LCP is trivially exact - LCP is trivially exact - T=0 subtraction is done - T=0 subtraction is done with a common T=0 sim. with a common T=0 sim. (T=0 high. stat. spectrum) (T=0 high. stat. spectrum) - easy to keep large 1/a - easy to keep large 1/a at whole T region at whole T region - easy to study T effect - easy to study T effect without V, 1/a effects without V, 1/a effects Disadvantages Disadvantages - T resolution by integer N t - T resolution by integer N t - program for odd N t - program for odd N t - (1/a)/T = const. is not suited - (1/a)/T = const. is not suited for high T limit study for high T limit study

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)23 T-integration method to calculate the EOS /31 We propose a new method (“T-integration method”) to calculate the EOS at fixed scales Our method is based on the trace anomaly (interaction measure), and the thermodynamic relation. T.Umeda et al. (WHOT-QCD) arXiv: [hep-lat]

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)24 Simulation parameters (isotropic lattices) /31 We present results from SU(3) gauge theory as a test of our method plaquette gauge action on N s 3 x N t lattices plaquette gauge action on N s 3 x N t lattices Jackknife analysis with appropriate bin-size Jackknife analysis with appropriate bin-size To study scale- & volume-dependence, we prepare 3-type of lattices. we prepare 3-type of lattices. (1) β=6.0, V=(16a) 3 1/a=2.1GeV 1/a=2.1GeV (2) β=6.0, V=(24a) 3 1/a=2.1GeV 1/a=2.1GeV (3) β=6.2, V=(22a) 3 1/a=2.5GeV 1/a=2.5GeV

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)25 Simulation parameters (anisotropic lattice) /31 Anisotropic lattice is useful to increase Temp. resolution, we also test our method on an anisotropic lattice a s ≠ a t we also test our method on an anisotropic lattice a s ≠ a t plaquette gauge action on N s 3 x N t lattices plaquette gauge action on N s 3 x N t lattices with anisotropy ξ=a s /a t =4 β=6.1, ξ=4 V=(20a s ) 3 V=(20a s ) 3 =(1.95fm) 3 =(1.95fm) 3 1/a s =2.0GeV 1/a s =2.0GeV 1/a t =8.1GeV 1/a t =8.1GeV - EOS calculation - static quark free energy free energy V=(20a s ) 3 =(1.95fm) 3 =(1.95fm) 3 V=(30a s ) 3 =(2.92fm) 3 =(2.92fm) 3 V=(40a s ) 3 =(3.89fm) 3 =(3.89fm) 3 - critical temp.

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)26 Trace anomaly ( e - 3p )/T 4 on isotropic lattices /31 beta function : G.Boyd et al. (’96) lattice scale r 0 : R.Edwards et al. (’98) (1) β=6.0, 1/a=2.1GeV, V=(1.5fm) 3 (2) β=6.0, 1/a=2.1GeV, V=(2.2fm) 3 (3) β=6.2, 1/a=2.5GeV, V=(1.5fm) 3 dotted lines : cubic spline Excellent agreement Excellent agreement between (1) and (3) between (1) and (3)  scale violation is small  scale violation is small 1/a=2GeV is good 1/a=2GeV is good Finite volume effect Finite volume effect appears below & near T c appears below & near T c  volume size is important  volume size is important V=(2fm) 3 is necessary. V=(2fm) 3 is necessary.

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)27 Trace anomaly ( e - 3p )/T 4 on aniso. lattice /31 (1) ξ=4, 1/a s =2.0GeV, V=(2.0fm) 3 (2) ξ=1, 1/a=2.1GeV, V=(2.2fm) 3 dotted lines : cubic spline beta function : obtained by r 0 /a s fit r 0 /a s data H.Matsufuru et al. (’01) r 0 /a s data H.Matsufuru et al. (’01) Anisotropic lattice is useful Anisotropic lattice is useful to increase Temp. resolution. to increase Temp. resolution. is required in SU(3) gauge theory. in SU(3) gauge theory. T.R.Klassen (’98)

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)28 Pressure & Energy density /31 Integration Integration is performed with the cubic is performed with the cubic spline of (e-3p)/T 4 spline of (e-3p)/T 4 Cubic spline vs trapezoidal inte. Cubic spline vs trapezoidal inte. yields small difference ~ 1σ yields small difference ~ 1σ Our results are roughly Our results are roughly consistent with previous results. consistent with previous results. Unlike the fixed N t approach, Unlike the fixed N t approach, scale/temp. is not constant. scale/temp. is not constant.  Lattice artifacts increase  Lattice artifacts increase as temperature increases. as temperature increases. Our fixed scale approach with “T-integration method” works well !!

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)29 Transition temperature at fixed scale /31 T-dependence of the (rotated) Polyakov loop the (rotated) Polyakov loop and its susceptibility and its susceptibility No renormalization is No renormalization is required upto overall factor required upto overall factor due to the fixed scale. due to the fixed scale. Rough estimation of Rough estimation of critical temperature critical temperature is possible. is possible. T c = 280~300 MeV T c = 280~300 MeV at β=6.1, ξ=4 at β=6.1, ξ=4 (SU(3) gauge theory) (SU(3) gauge theory)

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)30 Static quark free energy at fixed scale /31 Static quark free energies at fixed scale at fixed scale Due to the fixed scale, Due to the fixed scale, no renomalization constant no renomalization constant is required. is required.  small thermal effects in V(r)  small thermal effects in V(r) at short distance at short distance (without any matching) (without any matching) Easy to study Easy to study temperature effect of V(r) temperature effect of V(r) without scale & volume effects without scale & volume effects color singlet static quark free energy V(r)

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)31 Toward full QCD calculations and new ideas at T>0 &μ>0 /31 There are many projects on high statistics full QCD at T=0. There are many projects on high statistics full QCD at T=0. PACS-CS, JLQCD, MILC, RBRC, etc... - some basic quantities at T=0 are studied - some basic quantities at T=0 are studied - T=0 config. are open to the public (by ILDG) - T=0 config. are open to the public (by ILDG) our method requires no additional T=0 simulation !! our method requires no additional T=0 simulation !! We have already generated T>0 configurations We have already generated T>0 configurations using CP-PACS/JLQCD parameter using CP-PACS/JLQCD parameter (N f =2+1 Clover+RG, 1/a=3GeV, pion mass ~ 500MeV) (N f =2+1 Clover+RG, 1/a=3GeV, pion mass ~ 500MeV) Our final goal is to study thermodynamics on Our final goal is to study thermodynamics on the physical point (pion mass ~ 140MeV) the physical point (pion mass ~ 140MeV) with N f =2+1 Wilson quarks (PACS-CS) with N f =2+1 Wilson quarks (PACS-CS) or exact chiral symmetry with N f =2+1 Overlap quarks (JLQCD) or exact chiral symmetry with N f =2+1 Overlap quarks (JLQCD) We are looking for new ideas to study other physics on our config. We are looking for new ideas to study other physics on our config. ( density correlations, J/psi suppression, finite density...) ( density correlations, J/psi suppression, finite density...)

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)32 Backup slides

YITP seminarTakashi Umeda (YITP, Kyoto Univ.) 33 (e-3p)/T^4 our (a2),(i2) results vs Nt=4,6,8 in Ref[9]

YITP seminarTakashi Umeda (YITP, Kyoto Univ.) 34 pressure our (a2),(i2) results vs Nt=4,6,8 in Ref[9]

YITP seminarTakashi Umeda (YITP, Kyoto Univ.) 35 pressure our (a2),(i2) results vs continuum limit in Ref[9]

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)36 Pressure & Energy density /31

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)37 Pressure & Energy density /31 G.Boyd et al. (’96)

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)38 EOS on an anisotropic lattice /31 Anisotropic lattice is useful Anisotropic lattice is useful to increase Temp. resolution. to increase Temp. resolution. Results are roughly consistent Results are roughly consistent with previous & isotropic results with previous & isotropic results Additional coefficients are Additional coefficients are required to calculate (e-3p)/T 4 required to calculate (e-3p)/T 4 is required in SU(3) gauge theory. in SU(3) gauge theory. T.R.Klassen (’98) beta function : obtained by r 0 /a σ fit r 0 /a σ data H.Matsufuru et al. (’01) r 0 /a σ data H.Matsufuru et al. (’01)

YITP seminarTakashi Umeda (YITP, Kyoto Univ.)39 Recent lattice calculations for Tc RBC-Bielefeld: Nt=4,6,8 Staggered (p4) quark pion mass ≥ 140MeV, Nf=2+1 pion mass ≥ 140MeV, Nf=2+1 MILC: Nt=4,6,8 Staggered (Asqtad) quark pion mass ≥ 220MeV, Nf=2+1 pion mass ≥ 220MeV, Nf=2+1 Wuppertal: Nt=4,6,8,10 Staggered (stout) quark pion mass ~ 140MeV, Nf=2+1 pion mass ~ 140MeV, Nf=2+1 DIK: Nt=8,10,12 Wilson (Clover) quark pion mass ≥ 500MeV, Nf=2 pion mass ≥ 500MeV, Nf=2 WHOT-QCD: Nt=4,6 Wilson (MFI Clover) quark pion mass ≥ 500MeV, Nf=2 pion mass ≥ 500MeV, Nf=2 RBC-HOT: Nt=8 Domain Wall quarks pion mass ~ 250MeV?, Nf=2+1 pion mass ~ 250MeV?, Nf=2+1 /31