introduction to lattice qcd ii a dedicated project
TRANSCRIPT
Introduction to Lattice QCD IIA dedicated project
Karl Jansen ��������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������
• Leading order hadronic contributionto the muon anomalous magnetic moment
– motivation– describe the computation– new method for the lattice– results
A success of Quantum Field Theories: Electromagnetic Interaction
Quantum Electrodynamics (QED)
coupling of the electromagnetic interaction is small⇒ perturbation theory 4-loop calculation
magnetic moment of the electron
~µ = gee~
2mec~S
~S spin vector
deviation from ge = 2: ae = (ge − 2)/2
ae(theory) = 1159652201.1(2.1)(27.1) · 10−12
ae(experiment) = 1159652188.4(4.3) · 10−12
1
The question of the muon
magnetic moment of the muon
~µ = gµe~
2mµc~S
deviation from gµ = 2: aµ = (gµ − 2)/2
aµ(theory) = 1.16591790(65) · 10−3
aµ(experiment) = 1.16592080(63) · 10−3
→ there is a 3.2σ discrepancy
aµ(experiment)− aµ(theory) = 2.90(91) · 10−9
2
Motivation
��������������������������������� �������������������������������������������������������������
���������������������������������������������������������������������������������
���� ��������
���������������������������������������������������� �����������
3
Why is this interesting?
→ new proposed experiments at Fermilab (USA) and JPARC (Japan) ≈ 2015
brings down error
σex = 6.3 · 10−10 → 1.6 · 10−10
challenge: bring down theoretical error to same level
⇒ possibility
aµ(experiment)− aµ(theory) > 5σ
• possible breakdown of standard model?
• aµ = aQEDµ + aweak
µ + aQCDµ + aNP
µ
aNPµ ∝ m2lepton/m
2new
→ muon ideal to discover new physics
for τ experimental results too imprecise
4
Is the lattice important?
→ size of various contributions to aµ
Contribution σth
QCD-LO (α2s) 5.3 · 10−10
QCD-NLO (α3s) 3.9 · 10−10
QED/EW 0.2 · 10−10
Total 6.6 · 10−10
⇒ non-perturbative hadronic contributions are most important
5
γ(q)µ(p′)
µ(p)
1
The contributions to gµ − 2
The vertex
– p, p′ incoming and outgoing momenta
– q = p− p′ photon momentum
• put muon in magnetic field ~B interaction Hamiltonian H = ~µ ~B
• measure interaction of muon with photon (magnetic field)
6
µ µγ⇒ a
QED(1)µ = α
2π ,
1
The simplest, QED contribution to gµ − 2
Schwinger, 1948
α QED fine structure constant
7
The leading order hadronic contribution, ahadµ
µ
γ
had
µ
• ahadµ contributes almost 60% of theoretical error
• needs to be computed non-perturbatively
• computation of vacuum polarization tensor
Πµν(q) = (qµqν − q2gµν)Π(q2)
8
Relation to experimental extraction
connection between real and imaginary part of Π(q2)
Π(q2)−Π(0) = q2
π
∫∞0ds ImΠ(s)s(s−q2)
ImΠ(s) related to experimental data of total cross section in e+e− annihilation
ImΠ(s) = α3R(s)
π−π ρ,ω φ J/ψ
0 0.5 1 1.5 2 2.5 3 3.5E [GeV]
1
2
3
4
5
R(E
)
important contributions
• ρ, ω (Nf = 2)
• Φ (Nf = 2 + 1)
• J/Ψ (Nf = 2 + 1 + 1)
9
How the data really look like
• demanding analysis of O(1000) channelsJegerlehner, Nyffeler, Phys.Rep.
10
Euclidean expression for hadronic contribution
ahadµ = α2
∫∞0
dQ2
Q2 F(Q2
m2µ
)(Π(Q2)−Π(0))
with F(Q2
m2µ
)a known function T. Blum
→ need to compute vacuum polaization function
Continuum:
Πµν(Q) = i∫d4xeiQ·(x−y)〈0|TJµ(x)Jν(y)|0〉
Jµ hadronic electromagnetic current
Jµ(x) =∑f ef ψf(x)γµψf(x) = 2
3u(x)γµu(x)− 13d(x)γµd(x) + · · ·
local vector current given is conserved
∂µJµ(x) = 0 ⇒ QµΠ(Q)µν = 0
eliminating the factor QµQν −Q2δµν
⇒ obtain Π(Q2)
11
Going to the lattice
• work with Nf = 2 twisted mass fermions
Stm =∑x χ(x) [DW +m0 + iµγ5τ3]χ(x)
• use the conserved lattice current
Exercize: derive conserved current
use vector transformation
δV χ(x) = iεV (x)τχ(x) , δV χ(x) = −iχ(x)τεV (x)
τ =
(2/3 00 −1/3
)= 1
61 + 12τ
3
Can we also use the local current?
12
Conserved lattice current
J tmµ (x) = 12
(χ(x)τ(γµ − r)Uµ(x)χ(x+ µ) + χ(x+ µ)τ(γµ + r)U†µ(x)χ(x)
)satisfying
∂∗µJtmµ (x) = 0
with ∂∗µ backward lattice derivative
13
Lattice vacuum polarization tensor
Fourier transformation of conserved vector current:
J tmµ (Q) =∑x e
iQ·(x+µ/2)J tmµ (x) , Qµ = 2 sin(Qµ2
)leading to
Πµν(Q) = 1V
∑x,y e
iQ·(x+µ/2−y−ν/2)〈J tmµ (x)J tmν (y)〉
from which we extract the vacuum polarization function
Πµν(Q) = (QµQν − Q2δµν)Π(Q2)
14
Fit to vacuum polarization function
Fit function
ΠM,N(Q2) = −59
∑Mi=1
m2i
Q2+m2i
+∑Nn=0 an(Q2)n
typically i = 1, 2, 3 , n = 0, 1, 2, 3 (systematic error)
0 10 20 30 40 50 60 70 80 90
Q2 [GeV
2]
-0.2
-0.15
-0.1
-0.05
Π(Q
2)
0 1 2 3 4
-0.2
-0.16
15
Do we control hadronic vacuum polarisation?(Xu Feng, Dru Renner, Marcus Petschlies, K.J.; Lattice 2010)
0 0.1 0.2 0.3 0.4
mπ2 [GeV
2]
0
100
200
300
400
500
600
700
aµ
had [
10
-10]
Jegerlehner 2008a=0.079 fma=0.063 fmquadratic in mπ
2• experiment: ahvp,exp
µ,Nf=2 = 5.66(05)10−8
• lattice: ahvp,oldµ,Nf=2 = 2.95(45)10−8
→ misses the experimental value→ order of magnitude larger error
• have used different volumes
• have used different values of lattice spacing
16
Dis-connected contribution
a graph representing dis-conected contributions
tsrc t
→ has been basically always be neglected
17
Can it be the dis-connected contribution?(Xu Feng, Dru Renner, Marcus Petschlies, K.J.)
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
q2 [GeV
2]
-0.12
-0.08
-0.04
0
Π(q
2)
disconnectedconnected
• dedicated effort
• have included dis-connected contributions for first time
• smallness consistent with chiral perturbation theory (Della Morte, Juttner)
18
Different extrapolation to the physical point
lattice: simulations at unphysical quark masses, demand only
limmPS→mπ ahvp,lattl = ahvp,phys
l
⇒ flexibility to define ahvp,lattl
standard definitions in the continuum
ahvpl = α2
∫∞0dQ2 1
Q2ω(r)ΠR(Q2)
ΠR(Q2) = Π(Q2)−Π(0)
ω(r) = 64
r2(
1+√
1+4/r)4√
1+4/r
with r = Q2/m2l
19
Redefinition of ahvp,lattl
redefinition of r for lattice computations
rlatt = Q2 · Hphys
H
choices
• r1: H = 1; Hphys = 1/m2l
• r2: H = m2V (mPS); Hphys = m2
ρ/m2l
• r3: H = f2V (mPS); Hphys = f2
ρ/m2l
each definition of r will show a different dependence on mPS but agree byconstruction at the physical point
remark: strategy often used in continuum limit extrapolations, e.g. charm quarkmass determination
20
comparison using r1, r2, r3
0 0.1 0.2 0.3 0.4
mPS
2 [GeV
2]
1
2
3
4
5
6a
µhvp [
10
-8]
mV
fV
1
21
A new result from the lattice
• experimental value: ahvp,expµ,Nf=2 = 5.66(05)10−8
• from our old analysis: ahvp,oldµ,Nf=2 = 2.95(45)10−8
→ misses the experimental value→ order of magnitude larger error
• from our new analysis: ahvp,newµ,Nf=2 = 5.66(11)10−8
→ error (including systematics) almost matching experiment
22
Anomalous magnetic moments, a check
0 0.1 0.2 0.3 0.4
mPS
2 [GeV
2]
2.2
2.4
2.6
2.8
a τhvp [
10-6
]
4.5
5
5.5
6
a µhvp [
10-8
]
11.21.41.61.8
a ehvp [
10-1
2 ] ahloe = 1.513 (43) · 10−12 LQCDahloe = 1.527 (12)·10−12 PHENO
ahloµ = 5.72 (16) · 10−8 LQCD
ahloµ = 5.67 (05) · 10−8 PHENO
ahloτ = 2.650 (54) · 10−6 LQCDahloτ = 2.638 (88)·10−12 PHENO
23
anomalous magnetic moment of muon
0 0.1 0.2 0.3 0.4
mPS
2 [GeV
2]
3.5
4
4.5
5
5.5
aµh
vp [
10
-8]
a=0.079 fm L=1.6 fma=0.079 fm L=1.9 fma=0.079 fm L=2.5 fma=0.063 fm L=1.5 fma=0.063 fm L=2.0 fmlinearquadraticmeasured
• have used different volumes
• have used different values of lattice spacing
• have included dis-connected contributions
⇒ can control systematic effects
24
Why it works: fitting the Q2 dependence
Fit function
ΠM,N(Q2) = −59
∑Mi=1
m2i
Q2+m2i
+∑Nn=0 an(Q2)n
i = 1: ρ-meson → dominant contribution ∝ 5.010−8
i = 2: ω-meson ∝ 3.710−9
i = 3: φ-meson ∝ 3.410−9
0 10 20 30 40 50 60 70 80 90
Q2 [GeV
2]
-0.2
-0.15
-0.1
-0.05Π
(Q2)
0 1 2 3 4
-0.2
-0.16
25
Why it works
0 0.1 0.2 0.3 0.4
mπ2 [GeV
2]
800
900
1000
1100
1200
mV
[M
eV
]
a=0.079 fma=0.063 fmPDG
• mV consistent with resonance analysis(Feng, Renner, K.J.)
• strong dependence on mPS
26
Ken Wilson award
27
Next steps
Fit function
ΠM,N(Q2) = −59
∑Mi=1
m2i
Q2+m2i
+∑Nn=0 an(Q2)n
add i = 4: J/Ψ, i = 5 . . .
• ETMC is performing simulations wit dynamicalup, down, strange and charm quarks→ unique opportunity
• avoids ambiguity with experiment comparison(what counts for Nf = 2?)
• generalized boundary conditions: Ψ(L+ aµ) = eiθΨ(x)→ θ continuous momentum→ allows to realize arbitrary momenta on the lattice
28
INT Workshop onThe Hadronic Light-by-Light Contribution to the Muon Anomaly
February 28 - March 4, 2011
Some sildes from
Lee Roberts (g-2) collaboration
• new experiment E989
• move Brookhaven equipment to Fermilab
29
What is moved
���������������������������������������������������� ����������
��������������������������
30
How and where is it moved
���������������������������������������������������� �����������
�����������������������������������������������������������
31
What is expected
���������������������������������������������������� �����������
��������������������������������������
�� �������������������������������������������������
�� ����������������������������������������������������������������
�� ����������������������������������������������������������������������������������
�� �����������������������������������
�� ������������������������������
�� �����������������������������������������������������������
32
What can be achieved
�����
���������������������������������������������������� �����������
�����
33
When will it happen
������������������������������������
���������������������������������������������������� �����������
34
The accuracy question
We need a precision < 1%
• include explicit isospin breaking
• include electromagnetism
• need computation of light-by-light contribution
• reach small quark mass → physical point
Much ado for young people
35
Simulation setup for Nf = 2 + 1 + 1Configurations available through ILDG
β a[fm] L3T/a4 mπ[MeV] status
1.9 ≈ 0.085 24348 300 – 500 ready1.95 ≈ 0.075 32364 300 –500 ready2.0 ≈ 0.065 32364 300 ready2.1 ≈ 0.055 48396 300 – 500 running/ready
643128 230 thermalizing643128 200 queued963192 160 planned
• trajectory length always one
• 1000 trajectores for thermalization
• ≥ 5000 trajectores for measurements
36
Next, α3s, contribution
µ(p)
γ(k) kρ
had + 5 permutations of the qi
µ(p′)
q1µq2νq3λ
termed: light-by-light scattering
involves 4-point function
Πµναβ(q1, q2, q3) =∫xyz
eiq1·x+iq2·y+iq3·z 〈jµ(0)jν(x)jα(y)jβ(z)〉
jµ electromagnetic quark current
jµ = 23uγµu− 1
3dγµd− 13sγµs+ 2
3cγµc
37
Momentum sources(Alexandrou, Constantinou, Korzec, Panagopoulos, Stylianou)
← following Gockeler et.al.
for renormalization: need Green function in momentum space
G(p) = a12
V
∑x,y,z,z′ e
−ip(x−y)〈u(x)u(z)J (z, z′)d(z′)d(y)〉
e.g. J (z, z′)=δz,z′γµ corresponds to local vector current
sources:
baα(x) = eipxδαβδab
solve for
DlattG(p) = b
Advantage: very high, sub-percent precision data (only moderate statistics)
Disadvantage: need inversion for each momentum separately
→ would need 3V inversions ...
38
Illustration of precision
with O(a2) effects
O(a2) effects pert. subtracted
with O(a2) effects
O(a2) effects pert. subtracted
use the momentum source method to attack the 4-point functionas needed for light-by-light scattering (P. Rakow et.al., lattice’08)
39
Summary
• lattice calculation of muon anomalous magnetic moment
• looked hopeless first ← order of magnitude larger error than experiment
• introduced new method → start to match experimental accuracy
• outlook
– include first two quark generations– include isospin breaking and electromagnetism– attack light-by-light scattering
40