Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
The axial nucleon charge gA
(and its renormalization constant ZA)
using the point-split axial vector current
operator on the lattice
Stanislav Kazmin
with Arwed Schiller and Holger Perlt
University Leipzig
25.11.2016
Stanislav Kazmin (University Leipzig)
Content
1. Introduction
2. Nucleon Axial Charge gA
3. Summary and Outlook
Stanislav Kazmin (University Leipzig) 1 / 19
Introduction – Basics
Why Lattice QCD?
� fundamental quantum field theory of quarks and gluons
� large energies: asymptotically free → perturbation theory
� small energies: coupling large → nonperturbative techniques
Basic steps towards lattice QCD
1 discretize Euclidean space-time by a hyper-cubic lattice Λ
� lattice volume: V = L3 × T
� lattice spacing: a
2 construct a discrete version of the QCD action
3 quantize QCD using Euclidean path integrals
4 calculate expectation values using Monte Carlo techniques
Stanislav Kazmin (University Leipzig) Sources: [1, 2] 2 / 19
Introduction – Basics
Fermion fields ψf (x)cα and ψf (x)cα
� live on lattice sites as Nc ⊗ Nd objects
Link variables Uµ(x)
� link variables Uµ(x) ∈ SU(3) as parallel gauge transporters
from point x to a neighboring point x + µ (Figure 1)
� connected to gauge fields Aµ(x) in the continuum by
Uµ(x) = exp(iaAµ(x)) (1)
x x + µUµ(x) x x + µU†µ(x + µ)
Figure 1: link variables on the links between two sites
Stanislav Kazmin (University Leipzig) Sources: [1, 3] 3 / 19
Introduction – Basics
Wilson plaquette U�(x)
� closed loops of Uµ(x) are gauge invariant
� smallest closed loop is the Wilson plaquette
U�µν(x) = Uµ(x)Uν(x + µ)U†µ(x + ν)U†ν(x) (2)
x µ
ν
a plaquette U�(x)
x µ
ν
b planar rectangle R�µν(x)
Figure 2: schematic Wilson plaquette U�(x) and planar rectangle R�(x)
Stanislav Kazmin (University Leipzig) Sources: [1] 4 / 19
Introduction – Basics
Integration on lattice
∫d4x →
∑x∈Λ
(3)
Parameters
� Inverse coupling β =6
g2(4)
� Hopping parameter κf =1
2(mf + 4)(5)
� work with 2 + 1 flavors → two light quarks with isospin
symmetry and the strange quark
Stanislav Kazmin (University Leipzig) Sources: [1, 3] 5 / 19
Introduction – Improved QCD Action
� gauge fields are tree-level Symanzik improved
� quark fields are clover improved and contain the Wilson term
� Df (x , y ) is the Dirac operator
S(U, ψ, ψ) = SF (U, ψ, ψ) + SG (U) (6)
SG (U) =β
3
∫d4x
(C0
∑µ<ν
Re Tr(1− U�µν(x)
)
+C1
∑µ<ν
Re Tr(1− R�µν(x)
))(7)
SF (U, ψ, ψ) =∑f
∫d4x
∫d4y ψf (x)Df (x , y )ψf (y ) (8)
Stanislav Kazmin (University Leipzig) Sources: [4, 5] 6 / 19
Introduction – Improved QCD Action
Dirac Operator Df (x , y ) with clover improvement
Df (x , y ) =1δxy − κf∑µ
(1− γµ)(Uµ(x)δx+µ,y − U−µ(x)δx−µ,y
)− cSWκf
∑µ,ν
1
2σµν Fµν(x)δxy (9)
� cSW is the clover (or Sheikholeslami-Wohlert) coefficient
� Fµν(x) is the field strength tensor on lattice
� inverse Dirac Operator is the quark propagator S(x , y )
S(x , y ) = D−1(x , y ) = 〈ψ(x) ψ(y )〉F (10)
� the quark propagates from point y to point x
Stanislav Kazmin (University Leipzig) 7 / 19
Introduction – Expectation Values
Expectation value of operator O
� path integral formalism as in continuum
〈O〉 =1
Z
∫DU Dψ Dψ O(U, ψ, ψ) e−S(U,ψ,ψ) (11)
� Grassmann numbers (anti-commuting numbers) allow to
integrate out the fermionic part
� fermion determinant det Df (x , y ) as an effective action
〈O〉 =1
Z
∫DU O(Sf ,U) e−S(U) (12)
S(U) = SG (U) + SeffF (U) (13)
SeffF (U) = −
∑f
Tr ln Df (14)
Stanislav Kazmin (University Leipzig) Sources: [6] 8 / 19
Introduction – Expectation Values
Configuration ensemble
configuration Ui is a set of link variables
Ui = {Uµ(x) | x ∈ Λ, µ = 1, . . . , 4} (15)
ensemble U is a set of configurations Ui
� generate ensemble U with a Hybrid Monte Carlo algorithm
� configurations Ui distributed according to the weight
W =1
Ze−S(U) (16)
� use importance sampling to obtain expectation values 〈O〉and errors σ(O)
Stanislav Kazmin (University Leipzig) Sources: [1, 7] 9 / 19
Introduction – Analysis Steps
1 configuration generation
� used configurations produced by the QCDSF
collaboration
2 measurements
� performed at the HRLN supercomputing system
� used the Chroma C++ package
� used self-written Chroma module for the vertex function
calculation from the point-split axial current operator
3 data analysis
� performed in Leipzig with self-written Julia code
Stanislav Kazmin (University Leipzig) Sources: [8–10] 10 / 19
Content
1. Introduction
2. Nucleon Axial Charge gA
3. Summary and Outlook
Stanislav Kazmin (University Leipzig)
Nucleon Axial Charge gA – Introduction
� axial vector current operator Aµ = ψ(x)γµγ5ψ(x)
� nucleon matrix element 〈N(p′, s ′)|Aµ|N(p, s)〉� proportional to the axial form factor GA(q2)
� proportional to nucleon beta decay rate
� axial charge is defined by gA = GA(q2 = 0)
� gA is a benchmark quantity in lattice QCD
� current lattice measurements usually underestimate the
experimental value gexpA = 1.2723(23)
Stanislav Kazmin (University Leipzig) Sources: [11–13] 11 / 19
Nucleon Axial Charge gA – Introduction
Axial vector current operator Aµ on the lattice
� local operator Alocµ (x) or point-split operator Aps
µ (x)
Alocµ (x) = Aµ = ψ(x)γµγ5ψ(x) (17)
Apsµ (x) =
1
2
(ψ(x)γµγ5Uµ(x)ψ(x + µ)
+ ψ(x + µ)γµγ5U†µ(x)ψ(x))
(18)
Motivation for Apsµ :
� at comparable pion masses estimates, that are closer to the
experimental value than for Alocµ
� renormalization constant is closer to one
Stanislav Kazmin (University Leipzig) Sources: [14] 12 / 19
Nucleon Axial Charge gA – Introduction
Ratio of correlation functions can be used to determine gA
R(t; τ) =G3(t; τ)
G2(t)= gA (19)
Correlation functions
G2(t) =∑x
e−ipx Γ(2)βα〈Nα(x)Nβ(0)〉 (20)
G3(t; τ) =∑x ,y
e−ipxe−ip′(x−y) Γ
(3)βα〈Nα(x)O(y )Nβ(0)〉 (21)
with the nucleon interpolators N and projectors Γ
0 t
d
u
u0 t
O(τ)d d
u
u
Figure 3: Two-point and three-point function of a nucleon
Stanislav Kazmin (University Leipzig) Sources: [15, 16] 13 / 19
Nucleon Axial Charge gA – Measurements
Performed measurements
L3 × T β a / fm t, mπ /MeV meas.
323 × 64 5.50 0.0740t = 13, 15, 17, 19, 21 ≈ 2600
mπ = 465
483 × 96 5.80 0.0558t = 15, 17, 19, 21, 23 ≈ 600
mπ = 427
� varied separation time t between source and sink at
symmetric point (κl = κs = κ)
� got curves gA(τ) for each t and gA as the mean value in the
range 0� τ � t
Stanislav Kazmin (University Leipzig) 14 / 19
Nucleon Axial Charge gA – Measurements
0 2 4 6 8 10 12 14 16 18 20 22τ = τ + 1
2 (23− t)
1.050
1.075
1.100
1.125
1.150
1.175
1.200
1.225
gA
gA = 1.125(11), t = 15
gA = 1.138(13), t = 17
gA = 1.149(15), t = 19
gA = 1.183(21), t = 21
gA = 1.194(26), t = 23
dotted lines to guide the eye
Figure 4: gA for different separation times t for β = 5.80
Stanislav Kazmin (University Leipzig) 15 / 19
Nucleon Axial Charge gA – Results
Excited states contribution
� gA depends on separation time t
� for small and large t values the contribution is large
� systematically analyzed with two different methods
1 Summation method
2 Global fit method
Global fit method
1 from G2 obtain the ground state mass and two excited state
masses Mi with a symmetric fit ansatz
2 insert the obtained masses into a global fit function which
involves all gA(t, τ) values
Stanislav Kazmin (University Leipzig) Sources: [17, 18] 16 / 19
Nucleon Axial Charge gA – Results
0 10 20 30 40 50 60 70 80 90t
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
G2
+G2
−G2
three mass fit
Figure 5: G2 three mass fit for β = 5.80
Stanislav Kazmin (University Leipzig) 17 / 19
Nucleon Axial Charge gA – Results
0 2 4 6 8 10 12 14 16 18 20 22τ = τ + 1
2 (23− t)
1.06
1.08
1.10
1.12
1.14
1.16
1.18
1.20
1.22
1.24
gA
t = 15t = 17
t = 19t = 21
t = 23
gfitA = 1.234(16)
Figure 6: Global fit method result for β = 5.80
Stanislav Kazmin (University Leipzig) 18 / 19
Content
1. Introduction
2. Nucleon Axial Charge gA
3. Summary and Outlook
Stanislav Kazmin (University Leipzig)
Summary and Outlook
� excited states contributions can be treated with global fit
method
� higher gA estimate than with the local operator for
comparable pion masses (β = 5.50, renormalized)
gpsA = 1.224(11) gloc
A = 1.1203(95) gexpA = 1.2723(23)
� Renormalization constant closer to one (β = 5.50)
Z MSA = 1.0212(12)(47) Z MS,loc
A = 0.8728(06)(27)
Outlook:
� Feynman-Hellmann approach could improve the measurement
as it does not involve three-point correlation functions
� extrapolate gA to the physical point
� measure ZA for β = 5.80 (expect to be even closer to one)
Stanislav Kazmin (University Leipzig) Sources: [13] 19 / 19
Thank You for Your Attention!
Stanislav Kazmin
with Arwed Schiller and Holger Perlt
University Leipzig
25.11.2016
Stanislav Kazmin (University Leipzig)
Sources
[1] C. Gattringer and C. B. Lang, Quantum chromodynamics
on the lattice: an introductory presentation, Lecture notes
in physics 788 (Springer, Heidelberg ; New York, 2010),
343 pp., ISBN: 978-3-642-01849-7.
[2] B. Jager, “Hadronic matrix elements in lattice QCD”,
(May 14, 2014),
http://ubm.opus.hbz-nrw.de/volltexte/2014/3741/.
[3] H. J. Rothe, Lattice Gauge Theories: An Introduction,
(World Scientific, Jan. 29, 1992), 397 pp., ISBN:
978-981-4602-30-3.
[4] B. Sheikholeslami and R. Wohlert, “Improved continuum
limit lattice action for QCD with wilson fermions”, Nuclear
Physics B 259, 572–596 (1985).
Stanislav Kazmin (University Leipzig)
Sources
[5] H. Matsufuru, Lattice QCD simulations with clover quarks,
Apr. 2012,
http://suchix.kek.jp/hideo˙matsufuru/Research/
Docs/Bridge/Clover˙HMC/note˙cloverHMC˙1.3.1.pdf.
[6] R. Horsley, “The Hadronic Structure of Matter – a Lattice
approach”, (Institut fur Physik, Humboldt-Universitat zu
Berlin, Jan. 18, 2000), 152 pp.
[7] H. Matsufuru, Introduction to lattice QCD simulations,
Sept. 7, 2007, http://www.rcnp.osaka-
u.ac.jp/%5C%0020%5C%003c%5C%0020matufuru/.
[8] The Julia Language (NumFocus project), (2016)
http://julialang.org/.
[9] HLRN, https://www.hlrn.de/home/view/Service.
Stanislav Kazmin (University Leipzig)
Sources
[10] R. G. Edwards and B. Joo, “The Chroma Software System
for Lattice QCD”, Nuclear Physics B - Proceedings
Supplements 140, 832–834 (2005),
arXiv:hep-lat/0409003.
[11] J. Zanotti, “Hadron Structure”, Seattle, USA, Aug. 6–24,
2012, http://www.int.washington.edu/PROGRAMS/12-
2c/week3/.
[12] M. E. Peskin and D. V. Schroeder, An introduction to
quantum field theory, (Addison-Wesley Pub. Co, Reading,
Mass, 1995), 842 pp., ISBN: 978-0-201-50397-5.
[13] K. A. Olive and Particle Data Group, “Review of Particle
Physics”, Chinese Physics C 38, 090001 (2014).
Stanislav Kazmin (University Leipzig)
Sources
[14] R. Horsley, Y. Nakamura, H. Perlt, P. E. L. Rakow,
G. Schierholz, A. Schiller, and J. M. Zanotti, “Improving
the lattice axial vector current”, Proceedings, 33rd
International Symposium on Lattice Field Theory (Lattice
2015) (2015).
[15] C. Alexandrou, R. Baron, M. Brinet, J. Carbonell,
V. Drach, P.-A. Harraud, T. Korzec, and G. Koutsou,
“Nucleon form factors with dynamical twisted mass
fermions”, (2008), arXiv:0811.0724 [hep-lat].
[16] D. B. Leinweber, R. M. Woloshyn, and T. Draper,
“Electromagnetic structure of octet baryons”, Physical
Review D 43, 1659–1678 (1991).
Stanislav Kazmin (University Leipzig)
Sources
[17] G. von Hippel, J. Hua, B. Jager, H. B. Meyer, T. D. Rae,
and H. Wittig, “Fitting strategies to extract the axial
charge of the nucleon from lattice QCD”, PoS
LATTICE2013, 446 (2014).
[18] B. Jager, T. D. Rae, S. Capitani, M. Della Morte,
D. Djukanovic, G. von Hippel, B. Knippschild, H. B. Meyer,
and H. Wittig, “A high-statistics study of the nucleon EM
form factors, axial charge and quark momentum fraction”,
(2013), arXiv:1311.5804 [hep-lat].
[19] S. Collins, “Proceedings, 34th International Symposium on
Lattice Field Theory (Lattice 2016)”, PoS LATTICE2016
(2016).
Stanislav Kazmin (University Leipzig)
Backup – Renormalization Constant ZA
� observables measured with lattice QCD must be
renormalized to compare them to real physical values
� many observables diverge in the limit a→ 0
Necessary steps for a lattice renormalization scheme
1 remove the ultraviolet divergence in the observables
� perturbatively: in general bad convergence
� nonperturbatively: used RI’-MOM scheme
(regularization independent momentum subtraction)
2 match a convenient continuum scheme
� common choice: MS scheme (minimal subtraction)
� typical comparison scale µR = 2 GeV
Stanislav Kazmin (University Leipzig) 1 / 48
Backup – Renormalization Constant ZA
Axial current renormalization constant ZA
� ZA = 1 for chirally symmetric actions and conserved current
� stays finite in the continuum limit a→ 0
Calculation steps
1 calculate Z RI’
A (p2;κi) in RI’-MOM scheme for quark masses
along the symmetric line κl = κs = κ
2 calculate Z RI’
A (p2;κsymcr ) by taking the chiral limit m → 0 of
Z RI’
A (p2;κi) at each p2
3 rescale the Z RI’
A (p2;κsymcr ) with a scale function RRI ′→MS(p2)
which lead to Z MSA (p2)
4 analyze the Z MSA (p2) curve and obtain the final Z MS
A
Stanislav Kazmin (University Leipzig) 2 / 48
Backup – Renormalization Constant ZA
RI’-MOM scheme
� applicable in a scale window
Λ2QCD � µ2
R �1
a2(22)
� needs gauge fixed configurations (used Landau gauge)
� ideally, Z MSA is scale independent but due to lattice artifacts
we see a dependence
� lattice artifacts considered to be small: linear fit ansatz
� choice of the fit range p2min . . . p2
max leads to systematic error
Stanislav Kazmin (University Leipzig) 3 / 48
Backup – Measurements
Performed measurements
κ 0.12090 0.12092 0.12095 0.12099 0.121021
mπ /MeV 465 439 402 343 290
cfg. 18 9 18 10 9
� β = 5.50, lattice size: 323 × 64, a = 0.0740 fm
� diagonal momenta in the range 0 ≤ p2 ≤ 10
� total number of momenta increased to 32 by using twisted
boundary conditions
pi =2π
L(ni + θi) , p4 =
2π
T(n4 + θ4 +
1
2) (23)
Stanislav Kazmin (University Leipzig) 4 / 48
Backup – Results
0 1 2 3 4 5 6 7 8 9 10p2
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
ZA
κ = 0.120900κ = 0.120920κ = 0.120950κ = 0.120990κ = 0.121021
2.1 2.3 2.5 2.7 2.9p2
0.930
0.935
0.940
0.945
0.950
0.955
0.960
Figure 7: ZA(p2) for different kappa values (zoomed region highlighted)
Stanislav Kazmin (University Leipzig) 5 / 48
Backup – Results
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5p2
0.85
0.90
0.95
1.00
1.05
1.10
ZA
ZMSA (p2)
fit: Z0A + B · p2
Z0A 1.0212(12)B −0.03106(33)dof = 11, χ2
red = 1.4082p2
min = 1.166 , p2max = 5.099
Figure 8: Z MSA (p2) curve and linear fit
Renormalization constant ZA result for β = 5.50
Z MSA = 1.0212(12)(47)
Stanislav Kazmin (University Leipzig) 6 / 48
Backup – Sequential Propagator Technique
0 x
a Calculate the ordinary quark
propagator Sψ(x , t; 0)
0
N(0)
x
N(p, t)
b Construct the sequential source
Sψseq(x , t; 0; p)
0 x
O(y ) , y
c Compute the sequential
propagator Σψ(0; y , τ ; p, t) with
a fermion matrix inversion and
insert the operator O
0 x
O(y ) , y
d Complete the correlation
function with the ordinary quark
propagator Sψ(y , τ ; 0)
Figure 9: Sequential propagator technique
Stanislav Kazmin (University Leipzig) 7 / 48
Backup – Sequential Propagator Technique
Ratio of correlation functions can be used to determine gA
R(t; τ ; Aµ; Γ(3)µ ) =
13
∑3µ=1 Im Gu−d3 (t; τ ; Aµ; Γ
(3)µ )
Re G2(t; Γ(2)µ )
= gA (24)
� polarized projector for the three-point function
Γ(3)µ =
1
2(1 + γ4)iγ5γµ (25)
� unpolarized projector for the two-point function
Γ(2)µ =
1
2(1 + γ4) (26)
Stanislav Kazmin (University Leipzig) 8 / 48
Backup – Sequential Propagator Technique
Two-point correlation function
G2(p, t) =
∫d3x e−ipxεabc εa′b′c ′⟨
Tr(
Γ(2)Su(x , 0)aa′)
Tr(
Sd(x , 0)bb′Su(x , 0)cc
′)
+ Tr(
Γ(2)Su(x , 0)aa′Sd(x , 0)bb
′Su(x , 0)cc
′)⟩(27)
� the tilde operation is defined by
S = (Cγ5 S Cγ5)T (28)
� only one-to-all propagators
� low effort calculation
Stanislav Kazmin (University Leipzig) 9 / 48
Backup – Sequential Propagator Technique
Three-point correlation function
Gψ3 (p, t; p′, τ ;O; Γ(3))
=
∫d3y e iqy
⟨Tr[Σψ(0; y , τ ; p, t)Oψ(y , τ) Sψ(y , τ ; 0)
]⟩(29)
� sequential propagator Σψ
Σψ(0; y , τ ; p, t) =
∫d3x Sψseq(x , t; 0; p)S(x , t; y , τ) (30)
� in sequential sources Sψseq skip disconnected terms as they
cancel each other in
Gu−d3 = Gu3 − Gd3 (31)
Stanislav Kazmin (University Leipzig) 10 / 48
Backup – Sequential Propagator Technique
Sequential sources Sψseq for the nucleon
Sdseq(x , t; 0; p)a′a = e−ipxεabc εa′b′c ′
×[Su(x , t; 0)bb
′Γ(3)Su(x , t; 0)cc
′
+ Tr(
Γ(3)Su(x , t; 0)bb′Su(x , t; 0)cc
′)](32)
Suseq(x , t; 0; p)a′a = e−ipxεabc εa′b′c ′
×[Sd(x , t; 0)bb
′Su(x , t; 0)cc
′Γ(3)
+ Tr(
Sd(x , t; 0)bb′Su(x , t; 0)cc
′)Γ(3)
+ Γ(3)Su(x , t; 0)bb′Sd(x , t; 0)cc
′
+ Tr(
Γ(3)Su(x , t; 0)bb′Su(x , t; 0)cc
′)](33)
Stanislav Kazmin (University Leipzig) 11 / 48
Backup – Sequential Propagator Technique
gA(t = 13) quark mass dependence
� no mass dependence measured (in the given error range)
� compared only one separation time t = 13
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40m2π /GeV2
1.05
1.10
1.15
1.20
gA gA(m2π)
fit: g0A + B ·m2
π
g0A 1.103(36)B 0.12(20)dof = 1, χ2
red = 0.5143
Figure 10: gA for different quark masses
Stanislav Kazmin (University Leipzig) 12 / 48
Backup – Plateau results for β = 5.80
Plateau values gA for β = 5.50 and β = 5.80
� chosen symmetrical time range τmin . . . τmax
� to remove contributions from excited states one has to fulfill
0� τ � t
� chosen the range according to this condition and to the good
fit condition χ2red ≈ 1
� gA is the mean value in the τ range
Stanislav Kazmin (University Leipzig) 13 / 48
Backup – Plateau results for β = 5.80
0 2 4 6 8 10 12 14 16 18 20 22τ = τ + 1
2 (23− t)
1.050
1.075
1.100
1.125
1.150
1.175
1.200
1.225
gA
gA = 1.125(11), t = 15
gA = 1.138(13), t = 17
gA = 1.149(15), t = 19
gA = 1.183(21), t = 21
gA = 1.194(26), t = 23
dotted lines to guide the eye
Figure 11: gA for different separation times t for β = 5.80
Stanislav Kazmin (University Leipzig) 14 / 48
Backup – Results for β = 5.50
0 2 4 6 8 10 12 14 16 18 20τ = τ + 1
2 (21− t)
1.075
1.100
1.125
1.150
1.175
1.200
gA
gA = 1.1364(64), t = 13
gA = 1.1591(93), t = 15
gA = 1.152(13), t = 17
gA = 1.189(22), t = 19
gA = 1.181(29), t = 21
dotted lines to guide the eye
Figure 12: gA for different separation times t for β = 5.50
Stanislav Kazmin (University Leipzig) 15 / 48
Backup – Results for β = 5.50
0 10 20 30 40 50 60t
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
G2
+G2
three mass fit
Figure 13: G2 three mass fit for β = 5.50
Stanislav Kazmin (University Leipzig) 16 / 48
Backup – Results for β = 5.50
0 2 4 6 8 10 12 14 16 18 20τ = τ + 1
2 (21− t)
1.06
1.08
1.10
1.12
1.14
1.16
1.18
1.20
1.22
1.24
gA
t = 13t = 15
t = 17t = 19
t = 21
gfitA = 1.200(12)
Figure 14: Global fit method result for β = 5.50
Stanislav Kazmin (University Leipzig) 17 / 48
Backup – Summation method
Summation method
� ansatz for the ratio (∆ is the energy gap between ground
state and first excited state)
R(τ, t) = gA + C1e−∆·τ + C2e−∆·(t−τ) + C3e−∆·t + . . . (34)
� sum over all ratio values R(τ, t) for each t (with cut tc)
S(t) =t−tc∑τ=tc
R(τ, t) (35)
S(t)t→∞−→ S(t) = gsum
A (t + 1− 2tc) + C sum (36)
� linear fit to S(t) yields gsumA
� still a systematic dependence on tc
Stanislav Kazmin (University Leipzig) 18 / 48
Backup – Summation method
13 14 15 16 17 18 19 20 21t
0
5
10
15
20
25
30
35
S(t)
tc = 1, gA = 1.209(31)
tc = 2, gA = 1.205(27)
tc = 3, gA = 1.200(25)
tc = 4, gA = 1.195(22)
tc = 5, gA = 1.189(18)
tc = 6, gA = 1.183(14)
Figure 15: Summation method fit for β = 5.50
Stanislav Kazmin (University Leipzig) 19 / 48
Backup – Summation method
15 16 17 18 19 20 21 22 23t
0
5
10
15
20
25
30
35
40
S(t)
tc = 1, gA = 1.243(41)
tc = 2, gA = 1.237(37)
tc = 3, gA = 1.227(34)
tc = 4, gA = 1.219(31)
tc = 5, gA = 1.209(27)
tc = 6, gA = 1.199(21)
Figure 16: Summation method fit for β = 5.80
Stanislav Kazmin (University Leipzig) 20 / 48
Backup – Feynman-Hellmann Approach
� additional effective term in the Lagrangian
L→ L + λO (37)
the theorem states for any hadron state H
∂E
∂λ|λ=0 =
1
2E〈H|O(0)|H〉 (38)
� measurement steps:
1 perform hadron spectroscopy for multiple values of λ
2 observe the linear behavior in the resulting energy shifts
about λ = 0
� advantage: no three-point correlation functions needed and
therefore less noticeable excited states contributions
� disadvantage: new configuration are needed for each
operator and each λ
Stanislav Kazmin (University Leipzig) 21 / 48
Backup – gA summary
exp
var
t=
13
t=
15
t=
17
t=
19
t=
21
t c=
0t c
=1
t c=
2t c
=3
t c=
4t c
=5 fit
t=
15
t=
17
t=
19
t=
21
t=
23
t c=
0t c
=1
t c=
2t c
=3
t c=
4t c
=5 fit1.08
1.10
1.12
1.14
1.16
1.18
1.20
1.22
1.24
1.26
1.28
1.30
gA
β = 5.50 β = 5.80
gA gsumA gA gsum
A
Figure 17: All obtained gA values and reference values
Stanislav Kazmin (University Leipzig) 22 / 48
Backup – gA summary
Figure 18: gA measurements from different collaborations, point-split
operator is used for “QCDSF Nf = 2 + 1”
Stanislav Kazmin (University Leipzig) Sources: [19] 23 / 48
Backup – RI’-MOM Scheme
consider a quark bilinear operator: O(z) = u(z)ΓOd(z) (39)
� renormalization condition at a scale µR
ZO
1
12Tr(〈u(p)|O(z)|d(p)〉〈u(p)|O(z)|d(p)〉−1
0
)|p2=µ2
R
m→0
= 1
(40)
� scale window: Λ2QCD � µ2
R � 1a2 (41)
� after some transformations and insertions
ZO(p) =12 Zq(p)
Tr(
ΛO(p)Γ−1B
)∣∣∣∣∣p2=µ2
R
m→0
(42)
Zq(p) =Tr(−i∑µ γµ sin(apµ)S−1(p)
)12∑µ sin2(apµ)
∣∣∣∣∣p2=µ2
R
(43)
Stanislav Kazmin (University Leipzig) Sources: [19] 24 / 48
Backup – RI’-MOM Scheme
� Born term is equal to the tree level matrix element
〈u(p)|O(z)|d(p)〉0 = ΓB (44)
� amputated vertex function
〈u(p)|O(z)|d(p)〉 = Z−1q ΛO(p)ΛO(p) = S−1(p)GO(p)S−1(p)
(45)
� propagator S(p)
S(p) =
∫d4x
∫d4y e−ip(x−y)S(x , y ) (46)
� vertex function GO
GO(p) =1
V
∫d4x
∫d4y
∑z
e−ip(x−y)〈u(x)|O(z)|d(y )〉
(47)
Stanislav Kazmin (University Leipzig) Sources: [19] 25 / 48
Backup – RI’-MOM Scheme
0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040m2π
0.932
0.934
0.936
0.938
0.940
0.942
ZA
ZA(p2 = 2.785, m2π)
fit: Z0A + sZ · m2
π
Z0A 0.93936(100)sZ −0.152(44)dof = 3, χ2
red = 0.0423
Figure 19: Chiral limit fit example
Stanislav Kazmin (University Leipzig) Sources: [19] 26 / 48
Backup – RI’-SMOM Scheme
� In RI’-MOM scheme:
p21 = p2
2 = µ2R , q = p2 − p1 = 0 (48)
� exceptional channel where q2 � µ2R
� asymmetric subtraction point
� chiral symmetry breaking effects decrease as 1/p2
� In RI’-SMOM scheme:
p21 = p2
2 = q2 = µ2R , q = p2 − p1 6= 0 (49)
� no exceptional channels
� symmetric subtraction point
� chiral symmetry breaking effects decrease as 1/p6
� some changes in the amputated vertex function are needed
Stanislav Kazmin (University Leipzig) Sources: [19] 27 / 48
Backup – Applications
Some applications of lattice QCD
� hadronic matrix elements
� confinement mechanism
� gluon self-interaction
� spontaneous chiral symmetry breaking
� . . .
Stanislav Kazmin (University Leipzig) Sources: [19] 28 / 48
Backup – Continuum QCD
QCD action in continuum
S(A, ψ, ψ) =1
2g2
∫d4x Tr (Fµν(x)Fµν(x)) +
+∑f
∫d4x ψf (x) (γµDµ(x) + mf )ψf (x) (50)
� fermion fields ψf and ψf with mass mf and flavor index f
� gauge field A
� gauge coupling constant g
� field strength tensor Fµν
� covariant derivative Dµ
Dµ(x) = ∂µ + iAµ(x) (51)
Stanislav Kazmin (University Leipzig) Sources: [19] 29 / 48
Backup – Continuum QCD
Gauge fields Aµ(x)
� represent the gluons
� elements of Lie algebra su(3)
� directional index µ
� N2c − 1 (= 8 for Nc = 3) generators TB of su(3)
Aµ(x) =8∑B=1
ABµ (x)TB (52)
� color components ABµ (x)
Stanislav Kazmin (University Leipzig) Sources: [19] 30 / 48
Backup – Continuum QCD
Field strength tensor Fµν(x)
� commutator [ Aµ(x) , Aν(x) ] does not vanish
Fµν(x) = −i[ Dµ(x) , Dν(x) ] (53)
Fµν(x) = ∂µAν(x)− ∂νAµ(x) + i[ Aµ(x) , Aν(x) ] (54)
� representation with generators and structure constants f ABC
Fµν(x) =8∑B=1
FBµν(x)TB (55)
FAµν(x) = ∂µAAν (x)− ∂νAAµ(x)− f ABCABµ (x)ACν (x) (56)
Stanislav Kazmin (University Leipzig) Sources: [19] 31 / 48
Backup – Continuum QCD
Fermion fields ψf (x) and ψf (x)
� live on lattice sites
� positional index x ∈ Λ
� physical spatial distance is ax
Note:
� all quantities are measured in lattice units a
� fermion fields, derivatives and masses are dimensionless
m → am (57)
ψ(x)→ a3/2ψ(x) (58)
ψ(x)→ a3/2ψ(x) (59)
Stanislav Kazmin (University Leipzig) Sources: [19] 32 / 48
Backup – Continuum QCD
Lattice derivative ∂µ
forward-backward difference as derivative
∂µψ(x) =1
2(ψ(x + µ)− ψ(x − µ)) (60)
Covariant derivative Dµ(x)
use link variables Uµ instead of gauge fields Aµ and discretize
Dµ(x)ψ(x) =1
2(Uµ(x)ψ(x + µ)− U−µ(x)ψ(x − µ)) (61)
Stanislav Kazmin (University Leipzig) Sources: [19] 33 / 48
Backup – Continuum QCD
Wilson loop U�(x)
� closed loops of Uµ(x) are gauge invariant
� smallest closed loop is the Wilson loop
U�µν(x) = Uµ(x)Uν(x + µ)U†µ(x + ν)U†ν(x) (62)
x Uµ(x)
Uν(x + µ)
U†µ(x + ν)
U†ν(x)
Figure 20: Wilson loop U�µν(x) starting at position x
Stanislav Kazmin (University Leipzig) Sources: [19] 34 / 48
Backup – Wilson QCD Action
Wilson gauge action SWG (U)
SWG (U) =β
3
∫d4x
∑µ<ν
Re Tr(1− U�µν(x)
)(63)
Wilson fermion action SWF (U, ψ, ψ)
SWF (U, ψ, ψ) =∑f
∫d4x
∫d4y ψf (x) DW
f (x , y ) ψf (y ) (64)
� Wilson Dirac operator DW
f (x , y )
� full lattice QCD Wilson action
SW (U, ψ, ψ) = SWF (U, ψ, ψ) + SWG (U) (65)
Stanislav Kazmin (University Leipzig) Sources: [19] 35 / 48
Backup – Wilson QCD Action
Wilson Dirac operator DW
f (x , y )
� contains naive fermion action discretization and Wilson term
� Wilson term removes doublers from the action
DW
f (x , y ) = 1δxy − κf∑µ
(1− γµ)(Uµ(x)δx+µ,y − U−µ(x)δx−µ,y
)(66)
� Wilson action ready for calculations
� Wilson gauge action: O(a2) artifact effects
� Wilson fermion action: O(a) artifact effects
� improvements useful
Stanislav Kazmin (University Leipzig) Sources: [19] 36 / 48
Backup – Action Improvements
Clover term
SSWF (U, ψ, ψ) = −cSW∑f
κf
∫d4x ψf (x)
1
2σµν Fµν(x)ψf (x)
(67)
� improve fermion action to order O(a2)
� improved action: SF = SWF + SSWF� clover coefficient cSW
� matrices: σµν = 12 [ γµ , γν ] (68)
� field strength tensor on lattice Fµν(x)
Fµν(x) =1
8(Qµν(x)−Qνµ(x)) (69)
� Qµν(x) is a sum of plaquettes
Stanislav Kazmin (University Leipzig) Sources: [19] 37 / 48
Backup – Action Improvements
Qµν(x) = U�µ,ν(x) + U�−µ,ν(x) + U�µ,−ν(x) + U�−µ,−ν(x) (70)
xµ
ν
Figure 21: clover term Qµν(x) starting at position x
Stanislav Kazmin (University Leipzig) Sources: [19] 38 / 48
Backup – Action Improvements
Tree level Symanzik improved gauge action
SG (U) =β
3
∫d4x
(C0
∑µ<ν
Re Tr(1− U�µν(x)
)+
+C1
∑µ<ν
Re Tr(1− R�µν(x)
))(71)
� include planar rectangles R�µν(x)
� constants depend on each other C0 = 1− 8C1
� in tree-level Symanzik improvement C1 = − 112
� possible to cancel order O(a2) artifacts and stay with O(a4)
Stanislav Kazmin (University Leipzig) Sources: [19] 39 / 48
Backup – Action Improvements
x µ
ν
a plaquette U�(x)
x µ
ν
b planar rectangle R�µν(x)
x µν
η
c extended rectangle R�µνη(x)
x µν
η
d extended bent rectangle R�µνη(x)
Figure 22: schematic Wilson loop U�(x) and rectangles R�(x)
Stanislav Kazmin (University Leipzig) Sources: [19] 40 / 48
Backup – Final QCD Action
S(U, ψ, ψ) = SF (U, ψ, ψ) + SG (U) (72)
SG (U) =β
3
∫d4x
(C0
∑µ<ν
Re Tr(1− U�µν(x)
)+
+C1
∑µ<ν
Re Tr(1− R�µν(x)
))(73)
SF (U, ψ, ψ) =∑f
∫d4x
∫d4y ψf (x)Df (x , y )ψf (y ) (74)
Df (x , y ) =1δxy − κf∑µ
(1− γµ)(Uµ(x)δx+µ,y − U−µ(x)δx−µ,y
)−
− cSWκf∑µ,ν
1
2σµν Fµν(x)δxy (75)
Stanislav Kazmin (University Leipzig) Sources: [19] 41 / 48
Backup – Expectation Values
Expectation value of operator O
� path integral formalism
� definition as in continuum
〈O〉 =1
Z
∫DU Dψ Dψ O(U, ψ, ψ) e−S(U,ψ,ψ) (76)
� partition function Z
Z = 〈1〉 =
∫DU Dψ Dψ e−S(U,ψ,ψ) (77)
measures:
Dψ Dψ =∏x∈Λ
∏f ,α,c
dψf (x)cαdψf (x)cα , DU =∏x∈Λ
∏µ
dUµ(x)
(78)
Stanislav Kazmin (University Leipzig) Sources: [19] 42 / 48
Backup – Expectation Values
� Grassmann numbers (anti-commuting numbers) allow to
integrate out the fermionic part
〈O〉 =1
Z
∫DU
∏f
det Df (x , y ) O(Sf ,U) e−SG (U) (79)
Z =
∫DU
∏f
det Df (x , y ) e−SG (U) . (80)
� in O fermionic part was integrated out
� O depends only on link variables and quark propagators
Quark propagator Sf (x , y )
Sf (x , y ) = D−1f (x , y ) = 〈ψf (x)ψf (y )〉F (81)
Stanislav Kazmin (University Leipzig) Sources: [19] 43 / 48
Backup – Expectation Values
� integral∫DU very high dimensional
� use Monte Carlo techniques for calculation
� have to deal with the fermion determinant
Fermion determinant det Df (x , y )
1 simplest solution – quenched approximation: det Df = 1
2 treat as an effective action and add to the gluon action
SeffF (U) = −
∑f
Tr ln Df∏f
det Df = e−SeffF (82)
S(U) = SG (U) + SeffF (U) (83)
3 other possibilities
Stanislav Kazmin (University Leipzig) Sources: [19] 44 / 48
Backup – Expectation Values
〈O〉 =1
Z
∫DU O e−S(U) (84)
Z =
∫DU e−S(U) (85)
this form calls for Monte Carlo technique
Monte Carlo basic simulation steps
1 configuration generation
2 operator measurement
3 data analysis
Stanislav Kazmin (University Leipzig) Sources: [19] 45 / 48
Backup – Configuration Generation
Configuration ensemble
configuration Ui is a set of link variables
Ui = {Uµ(x) | x ∈ Λ, µ = 1, . . . , 4} (86)
ensemble U is a set of Ncfg configurations Ui
� generate configuration ensemble U
� configurations Ui distributed according to the weight
W =1
Ze−S(U) (87)
� main difficulty:
� local link updates: many updates needed to uncorrelate
� global updates: low acceptance rate
� possible solution: Hybrid Monte Carlo algorithm
Stanislav Kazmin (University Leipzig) Sources: [19] 46 / 48
Backup – Configuration Generation
Hybrid Monte Carlo algorithm
� idea: do a step from U to U′ by performing a Hamiltonian
molecular dynamics in fictitious time
� use concept of pseudofermions
Basic steps:
1 start with configuration U
2 generate momenta p
3 evaluate the Hamiltonian H(U, p)
4 perform molecular dynamic in fictitious time t → U′, p′
5 evaluate the Hamiltonian H′(U′, p′)6 accept U′ according to Metropolis probability
P = min(1, eH(U,p)−H′(U′,p′))
7 if accepted: U := U′ an go to 1
Stanislav Kazmin (University Leipzig) Sources: [19] 47 / 48
Backup – Measurement
Importance sampling
� generated configurations Ui are distributed with 1Z e−S(U)
� expectation value estimator: 〈O〉 = 1Ncfg
Ncfg∑i=1
Oi (88)
� standard error: σO =
√〈O2〉 − 〈O〉2
Ncfg(Ncfg − 1)(89)
� often need bootstrap or jackknife method to obtain errors
use operators as two or three point correlators e.g. pion two point
correlation function with pion interpolator P(x) = d(x)γ5u(x)
O = 〈P(x)P(y )〉F = Tr (γ5Su(x , y )γ5Sd(y , x)) (90)
Stanislav Kazmin (University Leipzig) Sources: [19] 48 / 48