Lattice Study of
Energy-Momentum Tensor
with Gradient Flow:
Thermodynamics, Correlations, and Stress
Masakiyo Kitazawa (Osaka U.)
In Collaboration with:
FlowQCD Collab.(R. Yanagihara, T. Iritani, Asakawa, Hatsuda, Suzuki) WHOT Collab.(Taniguchi, Kanaya, Ejiri, Suzuki, Umeda, Shirogane, … )
Workshop of Recent Developments in QCD and QFT, Nov. 10, 2017, Taipei
Energy-Momentum Tensor
One of the most fundamental quantities in physics
stress
energy momentum
Energy-Momentum Tensor
One of the most fundamental quantities in physics
: nontrivial observable on the lattice
Definition of the operator is nontrivial
because of the explicit breaking of Lorentz symmetry
①
② Its measurement is extremely noisy
due to high dimensionality and etc.
ex:
stress
energy momentum
Thermodynamics
direct measurement of expectation values
Fluctuations and Correlations
viscosity, specific heat, ...
Hadron Structure
• flux tube / hadrons
• stress distribution
Contents
1. Constructing EMT on the lattice 2. Thermodynamics
3. Correlation Function
4. Stress distribution in qq system 5. Summary
ー
(Yang-Mills) Gradient Flow
t: “flow time”
dim:[length2] leading
• diffusion equation in 4-dim space
• diffusion distance
• “continuous” cooling/smearing
Luscher 2010
Narayanan, Neuberger, 2006 Luscher, Weiss, 2011
(Yang-Mills) Gradient Flow
t: “flow time”
dim:[length2] leading
• diffusion equation in 4-dim space
• diffusion distance
• “continuous” cooling/smearing
Luscher 2010
Narayanan, Neuberger, 2006 Luscher, Weiss, 2011
(YM) Gradient Flow: Properties
• All (composite) operators are finite at t>0
• Quite effective in reducing statistical error.
• Flowed field ≠ original field
• Gradient flow is not an approximation method.
• Applications
• scale setting
• topological charge/susceptibility
• Structure Func. / PDF
All operators are renormalized at t>0 Safe a0 limit
Luscher,Weisz,2011
孔子
confucius
(論語、子路 13)
Miss the wood for the trees
小利を見ればすなわち大事成らず
見小利則大事不成
孔子
confucius
(論語、子路 13)
UV fluctuations Physics
Miss the wood for the trees
小利を見ればすなわち大事成らず
見小利則大事不成
Gradient Flow for Fermions
Luscher, 2013
Makino, Suzuki, 2014 Taniguchi+ (WHOT) 2016
Not “gradient” flow but a “diffusion” equation.
Flow for gauge field is independent of fermion field.
Divergence in field renormalization of fermions.
All observables becomes finite once Z(t) is determined.
Small Flow-time Expansion
Luescher, Weisz, 2011 Suzuki, 2013
remormalized operators of original theory
an operator at t>0
t0 limit
original 4-dim theory
Constructing EMT
Suzuki, 2013DelDebbio,Patella,Rago,2013
gauge-invariant dimension 4 operators
Constructing EMT 2
Suzuki coeffs.
Suzuki, 2013
Constructing EMT 2
Suzuki coeffs.
Remormalized EMT
Suzuki, 2013
EMT with Fermions
Makino, Suzuki, 2014
Contents
1. Constructing EMT on the lattice 2. Thermodynamics
3. Correlation Function
4. Stress distribution in qq system 5. Summary
ー
Thermodynamics
direct measurement of expectation values
QCD EoS
(Energy Density, Pressure)
• Rapid increase of e/T4 around T=150-200 MeV
• Crossover transition
• Low T: hadron resonance gas model / High T: perturbative QCD BNL-Bielefeld
2011
hadronic
QGP
QCD Thermodynamics
Statistical Mechanics
Changing lattice spacing 1/T and V change
e-3p
Integral Method
measurements of e-3p for many T
vacuum subtraction for each T
information on beta function
Boyd+ 1996
Numerical Simulation
Expectation values of T
mn SU(3) YM theory
Wilson gauge action
Parameters:
Scale from gradient flow
• Nt = 12, 16, 20-24
• aspect ratio 5.3<Ns/Nt<8
• 1500~2000 configurations
FlowQCD 1503.06516
MK+ (FlowQCD),
PRD94, 114512 (2016)
t Dependence
: strong discretization effect : over smeared
: Linear t dependence
clover+plaq clover
Double Extrapolation
Continuum extrapolation
FlowQCD, 2014: continuum extrapolation only WHOT-QCD, 2016: small t limit only
Note:
strong
discretization effect
Double Extrapolation
Continuum extrapolation
Small t extrapolation
FlowQCD, 2014: continuum extrapolation only WHOT-QCD, 2016: small t limit only
Note:
strong
discretization effect
Double Extrapolation
Fitting ranges:
range-1:
range-2:
range-3:
Black line: continuum extrapolated
Range1 Range2
Range3
Systematic error from the choice of fitting range
≈
statistical errorT Dependence
Error includes
statistical error
choice of t range for t0 limit
uncertainty in aLMS
total error <1.5% for T>1.1Tc
Excellent agreement with integral method
High accuracy only with
~2000 confs.
FlowQCD, PRD, 2016
N f =2+1 QCD Thermodynamics
• Nf=2+1 QCD, Iwasaki gauge + NP-clover
• mPS/mV ≈0.63 / almost physical s quark mass
• T=0: CP-PACS+JLQCD (ß=2.05, 283x56, a≈0.07fm)
• T>0: 323xNt, Nt = 4, 6, ... , 14, 16):
• T≈174-697MeV
• t0 extrapolation only (No continuum limit)
Taniguchi+ (WHOT-QCD), PRD96, 014509 (2017)
Fermion Propagator
t=0 gauge field
fermion field
• propagator of flow equation
• Inverse propagator is needed
t0 Extrapolation
”linear window” for Nt>6
Checked: fit range, a2/t term
Taniguchi+ (WHOT-QCD), PRD96, 014509 (2017)
N f =2+1 Thermodynamics
Taniguchi+ (WHOT-QCD), PRD96, 014509 (2017)
Agreement with integral method except for Nt=4, 6
No stable extrapolation for Nt=4, 6
Suppression of statistical error
Physical mass: Kanaya+ (WHOT-QCD), 1710.10015
no linear window
Chiral Condensate / Suceptibility
Chiral condensate decreases for T>Tc.
Chiral susceptibility has a sharp peak around T=Tc.
Subtracted condensate
Chiral susceptibility
Taniguchi+ (WHOT-QCD), PRD96, 014509 (2017)
Contents
1. Constructing EMT on the lattice 2. Thermodynamics
3. Correlation Function
4. Stress distribution in qq system 5. Summary
ー
Fluctuations and Correlations
viscosity, specific heat, ...
Why EMT Correlation Func.?
Kubo Formula: T
12correlator shear viscosity
Energy fluctuation specific heat
Hydrodynamics describes long range behavior of Tmn
EMT Correlator: Extremely Noisy…
Nakamura, Sakai, PRL,2005
Nt=8
improved action
~106 configurations … no signal
Nt=16
standard action
5x104 configurations
With naïve EMT operators
Conservation Law
t independent constant
Linear Response Relations
Specific heat
entropy density
Giusti, Meyer, 2011
enthalpy density
Derivation
Minami, Hidaka, 2012
Numerical Simulation
• SU(3) pure gauge
• Wilson gauge action / clover operator
• Ns/Nt=4
• Statistics: 18-20x104
β T=1.66Tc T=2.22Tc 483x12 6.719 6.943 643x16 6.941 7.170 963x24 7.265 7.500
on Bluegene/Q @KEK FlowQCD, arXiv:1708.01415
Euclidean Correlator @T=2.24Tc
t-independent plateau in all channels conservation law
small t region: artificial enhancement due to overlap of operators
linear response relations for 4411, 4141 channels
FlowQCD, arXiv:1708.01415
Mid-Point Correlator @T=2.24Tc
• (44;11), (41;41) channels : confirmation of LRR
• (44;44) channel: new measurement of cV
2+1 QCD:
Taniguchi+ (WHOT-QCD), 1711.02262
Contents
1. Constructing EMT on the lattice 2. Thermodynamics
3. Correlation Function
4. Stress distribution in qq system 5. Summary
ー
Hadron Structure
• flux tube / hadrons
• stress distribution
Stress
Pressure
force on a surface per unit area
Stress
Pressure
force on a surface per unit area
Stress
Generally, F and n are not parallel
Stress Tensor
v
F
Maxwell Stress
E
Parallel to field: Attractive Vertical to field: Repulsive
𝑇 = 1 2
𝐸
20 0
0 −𝐸
20
0 0 −𝐸
2Maxwell Stress
attractive eigenvector
(≈Line of electric field)
E
Parallel to field: Attractive Vertical to field: Repulsive
Coulomb force
Previous Studies
• action density
• color electric field
• (color electric field)2
q-qbar System in YM Theory
=Flux Tube Formation
Color electric field is
confined into a flux tube
Linear potential
This Study: stress
• gauge invariant!
• definite physical meaning!
• establish action thr. medium
Stress Distribution in qq System
beta=6.819 (a=0.029fm) R/a=16
t0 limit
No continuum limit
attractive eigenvector
R. Yanagihara+ (FlowQCD) to appear soon!
First visualization of distortion in space due to color charges
Preliminary
Stress Distribution in qq System
R. Yanagihara+ (FlowQCD) to appear soon!
Yang-Mills SU(3) Maxwell
Stress Distribution in qq System
R. Yanagihara+ (FlowQCD) to appear soon!
Yang-Mills SU(3) Maxwell
New insights into physics of confinement
qq force?
Don’t miss
Yanagihara’s talk
this afternoon
Summary
• EMT operator on the lattice is now available!
• Correctly renormalized operator
• Statistical error is suppressed thanks to gradient flow.
• The operator is applied to various analyses:
Many future studies
• transport coefficient
• EM/stress distribution in hadrons
• Flux tube: T dependence
EMT (Naïve Constructuion)
Z1, Z2, Z3 have to be determined non-perturbatively.
Accurate determinations of Z1, Z3: Giusti, Pepe, 2014-; BW, 2016
So far, only for pure gauge theory
multi-level algorithm
Gradient Flow Method
lattice regularized gauge theory
gradient flow
continuum theory
(with dim. reg.)
continuum theory
(with dim. reg.) gradient flow
analytic (perturbative)
measurement on the lattice
lattice regularized gauge theory
gradient flow
continuum theory
(with dim. reg.)
continuum theory
(with dim. reg.) gradient flow
Caveats
Gauge field has to be sufficiently smeared!measurement on the lattice
analytic (perturbative) Perturbative relation
has to be applicable!
lattice regularized gauge theory
gradient flow
continuum theory
(with dim. reg.)
continuum theory
(with dim. reg.) gradient flow
Caveats
Gauge field has to be sufficiently smeared!analytic (perturbative) Perturbative relation
has to be applicable!
measurement on the lattice