Date post: | 15-Aug-2015 |
Category: |
Technology |
Upload: | lccausp |
View: | 11 times |
Download: | 0 times |
Solving QCD:
from BG/P to BG/Q
Attilio Cucchieri
http://lattice.ifsc.usp.br/
Instituto de Fısica de Sao Carlos – USP
3WHPC–LCCA April 2015
Quarks: in the Heart of Matter
Interior of the atom: nucleus, made up of nucleons, made up of
quarks
There are 6 types (flavors) of quarks, corresponding to the 6 types
of leptons in nature (e.g. electron, neutrino)
Hadrons (e.g. protons, neutrons, mesons) are made up of quarks
3WHPC–LCCA April 2015
Quantum Chromodynamics
Strong interaction btw. protons e neutrons
is residue of interaction btw. their quarks.
Nucleons are made up of 3 quarks of dif-
ferent colors
The proton is a color-neutral bound
state of quarks interacting through the
exchange of (massless) gluons
Protons and Neutrons:
99% of the mass of the bound state comes from the
interaction!
⇒ we are not star dust, we are (virtual) gluons!
3WHPC–LCCA April 2015
Free the Quarks!
Why was a fractionary electric charge never
observed?
Answer: quarks are confined inside hadrons!
Confinement: it would take infinite energy to separate
the quarks that constitute a hadron
Quarks have flavor, and color, but no freedom
3WHPC–LCCA April 2015
(Usual) Quantum Field Theory
QED Lagrangian:
L = −1
4Fµν F
µν + ψ (i γµDµ −m)ψ
where
Dµ ≡ ∂µ−i eAµ , Fµν ≡ ∂µAν − ∂νAµ
Perturbative calculation: Feynman
diagrams for electron scattering; it
is possible to infer the redefinition
of m and e to obtain finite results
3WHPC–LCCA April 2015
Quantum Chromodynamics (QCD)
QCD Lagrangian is just like the one of QED:
quarks (spin-1/2 fermions)
gluons (vector bosons) / color charge⇔
electrons
photons / electric
charge
But: gauge symmetry is SU(3) (non-Abelian) instead of U(1)
L = −1
4F aµν F
µνa +
6∑
f=1
ψf,i(
i γµDijµ −mf δij
)
ψf,j
where [a = 1, . . . , 8; i = 1, . . . , 3; T aij = SU(3) generators]
F aµν ≡ ∂µAaν − ∂νA
aµ + g0 fabcA
bµA
cν
Dµ ≡ ∂µ − i g0Aaµ Ta
Note: g0, mf are bare parameters.
3WHPC–LCCA April 2015
Gluons Have Color
Note: F aµν ∼ g0 f
abcAbµAcν
⇒ QCD Lagrangian contains terms with three and four
gauge fields in addition to quadratic terms (propagators)
LψψA = g0 ψ γµAµ ψ ⇒ quark-quark-gluon ver-
tex
LAAA = g0 fabcAµa A
νb ∂µA
cν ⇒ three-gluon ver-
tex
⇒ gluons interact with each other (have color charge),
determining the peculiar properties and the nonpertur-
bative nature of low-energy QCD
⇒ Running coupling αs(p) instead of α ≈ 1/137
3WHPC–LCCA April 2015
Photons vs. Gluons
Photons do not interact directly with one another
⇒ lightsabers (Star Wars) could not possibly work...
3WHPC–LCCA April 2015
QCD vs. QED
QCD (strong force) vs. QED (EM force)
quarks, gluons
SU(3) (3 “colors”)
mq, αs(p)
electrons, photons
U(1)
me, α ≈ 1/137
q− −q +
3WHPC–LCCA April 2015
QCD on a Lattice (I)
Kenneth Geddes Wilson (June 8, 1936 – June 15, 2013)
Lattice used by Wilson in 1974 as a trick to prove confinement
in (strong-coupling) QCD
[Confinement of quarks, Phys. Rev. D 10, 2445 (1974)]
3WHPC–LCCA April 2015
QCD on a Lattice (II)
Three ingredients
1. Quantization by path integrals ⇒ sum
over configurations with “weights” ei S/~
2. Euclidean formulation (analytic continua-
tion to imaginary time) ⇒ weight becomes
e−S/~
3. Discrete space-time ⇒ UV cut at mo-
menta p ∼< 1/a ⇒ regularization
Also: finite-size lattices ⇒ IR cut for small momenta p ≈ 1/L
The Wilson action
is written for the gauge links Ux,µ ≡ eig0aAbµ(x)Tb
reduces to the usual action for a→ 0
is gauge-invariant
3WHPC–LCCA April 2015
Numerical Simulations
Monte Carlo methods (Ulam, 1940’s): statisti-
cal description of the possible configurations of
a system, which is simulated on a computer.
Useful in
designing/analyzing experiments
studying the theory of stochastic
(statistical) systems
doing calculations in quantum field theory
3WHPC–LCCA April 2015
Lattice QCD Results: Confinement
May observe formation of flux tubes
Linear Growth of potential between quarks, string breaking
3WHPC–LCCA April 2015
Confinement: the Elephant in the Room
Do we understand confinement?
⇒ we know what it looks like,
but do we know what it is?
Millenium Prize Problems (Clay Mathematics Institute, USA/UK)
Yang-Mills and Mass Gap: Experiment and computer simulations sug-
gest the existence of a mass gap in the solution to the quantum versions
of the Yang-Mills equations. But no proof of this property is known.
3WHPC–LCCA April 2015
Lattice QCD at the IFSC–USP
The only LQCD group (A.C. & T. Mendes) in South America.
1. Study of qualitative aspects of QCD: infrared behavior
of propagators and vertices, related to color confinement
and to color deconfinement (at high temperature).
2. Development of methods: determination of the strong
coupling constant αs(p) to be applied to the full QCD
case, lattice implementation of different analytic ap-
proaches (linear covariant gauge, background gauge).
3. Development of algorithms: gauge fixing, global mini-
mization, matrix inversion, evaluation of eigenvalues.
3WHPC–LCCA April 2015
Results with the BG/P
New lower bound for the smallest nonzero eigenvalue of the
Landau-gauge Faddeev-Popov matrix in Yang-Mills theories.
First candidate for a lattice configuration belonging to the
common boundary ∂Ω ∩ ∂Λ.
First estimate of the distance of a minimal-Landau-gauge
configuration A ∈ Ω from the boundary ∂Ω.
First concrete explanation of why lattice studies do not observe
an enhanced ghost propagator in the deep infrared limit.
First evaluation of the Bose-ghost propagator (of the
Gribov-Zwanziger theory).
First numerical manifestation of BRST-symmetry breaking due
to restriction of gauge-configuration space to the Gribov region.
3WHPC–LCCA April 2015
Solving QCD!
The building blocks of QCD (in a
given gauge) are:
Propagators: gluon, quark,
ghost.
Vertices: three-gluon, four-
gluon, ghost-gluon, quark-
gluon.
0
0.5
1
1.5
2
2.5
3
3.5
4
0 0.5 1 1.5 2
D(p
2 )
p
4D Results
Challenges:
breaking of rotational symmetry;
connect to the continuum: for example, for the ghost-gluon vertex
Γabcµ (q, s; k) = −ig0f
abc qν Γνµ(q, s; k) we have
Γνµ(q, s; k) = δνµa(q, s; k)− kν sµ b(q, s; k) + qν kµ c(q, s; k)
+kν qµ d(q, s; k) + qν qµ e(q, s; k)
3WHPC–LCCA April 2015
3-Step Code
main()
/* set parameters: beta, number of configurations NC,
number of thermalization sweeps NT */
read_parameters();
/* U is the link configuration */
set_initial_configuration(U);
/* cycle over NC configurations */
for (int c=0; c < NC; c++)
thermalize(U,NT);
gauge_fix(U,g);
evaluate_propagators_and_vertices(U);
3WHPC–LCCA April 2015
Parallelization
• We need a parallelized code in order to simulate at
very large lattice volumes V .
• Communication is required in each of the three
steps.
• Each node gets a contiguous block of v = V/N
lattice sites (local lattice).
• Communication is required only for sites on the
boundary of the local lattice.
• 4D simulations → high granularity due to the
surface/volume effect.
3WHPC–LCCA April 2015
Weak and Strong Scaling on BG/Q
V Nodes HB Micro Gfix GluonProp CG
642 × 322 32 494.9 54.7 0.0044 0.041 0.0081
643 × 32 64 496.3 62.1 0.0049 0.041 0.0088
644 128 496.8 59.2 0.0047 0.050 0.0084
643 × 128 256 499.4 63.0 0.0050 0.041 0.0090
642 × 1282 512 499.7 56.4 0.0046 0.042 0.0083
644 128 496.8 59.2 0.0047 0.0050 0.0084
644 256 256.3 37.9 0.0029 0.0028 0.0055
644 512 134.6 27.3 0.0020 0.0018 0.0040
644 1024 74.4 22.5 0.0016 0.0012 0.0035
644 512 2943.6 218.5 0.0171 0.0179 0.0239
Weak (with 5 different lattice volumes) and strong (with 4 different volumes) scaling:
time (in seconds) for 3 different updates of local variables and for the evaluation of
the gluon propagator and the time (in seconds) for one conjugate gradient iteration.
Link and site variables are SU(2) matrices. The last row is for the BG/P.
3WHPC–LCCA April 2015
Conclusion
Still lots to be understood inside a Proton!!
Keep the supercomputers coming!
3WHPC–LCCA April 2015