Two flavor full QCD configurations by JLQCD

The dynamical QCD configurations with 2 flavors of overlap fermions listed below were generated by JLQCD Collaboration on supercomputers IBM Blue Gene/L at High Energy Accelerator Research Organization (KEK).

• The JLQCD Collaboration
• KEK Supercomputer System
• High Energy Accelerator Research Organization (KEK)

Index

• Simulation parameters
• Configuration ensembles
• Naming of Ensembles and Configurations (for ILDG users)
• Location of Configurations (for JLDG users)

Simulation parameters

These gauge configurations are generated with an overlap fermion action and a renormalization-group improved gauge action accompanied by an extra Wilson fermion term. The extra Wilson fermion term fixes the topological charge of the system, and are considered as a part of the gauge action. The values of parameters are as follows; The gauge coupling: β= 6/g²=2.3, the negative mass of the Wilson kernel (domain-wall height): M₀=1.6, twisted mass of the topology fixing term: μ = 0.2.

For the overlap-Dirac operator, the sign function of the hermitian Wilson kernel is approximated by the Zolotarev's partially fractional approximation. The number of poles, N_poly, is set to 10 in this work. The Zolotarev approximation is applied to the region |H_W| > 0.108, and for the eigenvalues in the range [-0.108, 0.108] the eigenvectors are determined so as to treat the sign function of these modes exactly. The boundary conditions are periodic for all the directions.

The extra Wilson term is introduced to fix the topological charge of the system. The determinant of the hermitian Wilson kernel (H_W)² with the same mass parameter as in the overlap operator suppresses the near-zero modes of H_W, and forbids the change of the topological charge. To decrease the effect of high mode region of H_W, a determinant of twisted mass ghost is also introduced. The twisted mass parameter is set as μ=0.2 throughout this work.

The lattice size is 16³x32, corresponding to the physical spatial size of about 1.9 fm. We have six ensembles corresponding to six quark mass which roughly covers (m_s/6, m_s). The corresponding lattice spacing, a∼0.12 fm, is set by the Sommer's hadronic radius r₀=0.49 fm. All the currently available configurations have the topological charge 0.

Simulations were carried out with the hybrid Monte Carlo (HMC) algorithm. The mass preconditioning is applied to the overlap fermions together with the multi-time step. The trajectory length is τ=0.5. Two types of algorithms were applied. The one algorithm adopts the nested CG algorithm for the overlap solver, and the sign function of low-lying eigenmodes of H_W is treated exactly during molecular dynamics as well as the Hamiltonian calculations. The other algorithm uses 5-dimensional CG algorithm for the overlap solver without determination of low-lying eigenvalues of H_W. This effect is removed by applying the noisy Metropolis test after the usual Metropolis test of HMC, so the treatment of the low-lying modes in the sign function is finally made exact. These two algorithms are denoted '4D' and '5D' in the table of the next section.

lattice size	β	m0	μ	a [fm]	La [fm]	mud	Npoly	Q	τ
163x32	2.30	1.6	0.2	0.1184(21)	1.894(33)	0.015	10	0	0.5
						0.025	10	0	0.5
						0.035	10	0	0.5
						0.050	10	0	0.5
						0.070	10	0	0.5
						0.100	10	0	0.5

• Two-flavor QCD simulation with exact chiral symmetry
  JLQCD Collaboration: S. Aoki, H. Fukaya, S. Hashimoto, K-I. Ishikawa, K. Kanaya, T. Kaneko, H. Matsufuru, M. Okamoto, M. Okawa, T. Onogi, A. Ukawa, N. Yamada, T. Yoshie.
  Phys. Rev. D 78 (2008) 014508 (22 pages).
  [ PRD Server / e-Print: arXiv:0803.3197 / SPIRES ]

Configuration ensembles

The table below gives a list of ensembles of configurations. A consecutive set of trajectories for a given values of lattice size, β, m₀, μ, m_ud and Q is called an 'ensemble'. A 'series' name (a, b) is given to distinguish the series of configurations sequentially generated with different series of random numbers. Each configuration is separated by 10 trajectories of length 0.5.

lattice size	β	m0	μ	mud	Q	series	algorithm	traj. range	#config	total #config	size/config [MB]	total [GB]
163x32	2.30	1.6	0.2	0.015	0	a a b b	4D 5D 4D 5D	000042-000201 005001-005370 002042-002161 007001-007350	160 370 120 350	1000	75.5	75.5
163x32	2.30	1.6	0.2	0.025	0	a a b b	4D 5D 4D 5D	000042-000361 005001-005300 002042-002241 007001-007180	320 300 200 180	1000	75.5	75.5
163x32	2.30	1.6	0.2	0.035	0	a a	4D 5D	000042-000501 005001-005540	460 540	1000	75.5	75.5
163x32	2.30	1.6	0.2	0.050	0	a a	4D 5D	000042-000521 005001-005520	480 520	1000	75.5	75.5
163x32	2.30	1.6	0.2	0.070	0	a a	4D 5D	000042-000491 005001-005550	450 550	1000	75.5	75.5
163x32	2.30	1.6	0.2	0.100	0	a a	4D 5D	000042-000501 005001-005540	460 540	1000	75.5	75.5

Naming of Ensembles and Configurations (for ILDG users)

An ensemble is identified by markovChainURI named e.g. as

• mc://JLDG/JLQCD/RgOvrNF2/RgOvr_16x32_b2.30_rho1.60_mu0.20_mud0.015_Q0

A configuration is labeled by dataLFN named e.g. as

• lfn://JLDG/JLQCD/RgOvrNF2/RgOvr_16x32_b2.30_rho1.60_mu0.20_mud0.015_Q0-a-000042

Basename of the dataLFN is formated as
RgOvr_[spatial size]x[temporal size]_[beta]_[kernel mass]_[quark mass]_[topological charge]-series-trajectory number

To get full list of markovChainURIs and LFNs, please use QCDml faceted navigation.

Location of Configurations (for JLDG users)

Configurations are placed in the Gfarm directory

• /gfarm/public/ILDG/JLDG/JLQCD/RgOvrNF2/
  

and classified in terms of the basename of markovChainURI. Filename is the same as basename of dataLFN.

All ensemble and configuration XML files can be found in

• /gfarm/public/ILDG/JLDG/JLQCD/RgOvrNF2/QCDml/