Two-flavor full QCD configurations by CP-PACS

The two-flavor full QCD configurations in JLDG are generated by the CP-PACS Collaboration on the CP-PACS parallel computer at the Center for Computational Sciences, University of Tsukuba.

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 a renormalization-group improved gauge action and a mean field improved clover quark action at three values of beta = 6/g², corresponding to lattice spacings of a \approx 0.22, 0.16 and 0.11 fm, and four sea quark masses corresponding to m_PS/m_V \approx 0.8, 0.75, 0.7 and 0.6. The sizes of lattice are chosen to be 12³x24, 16³x32 and 24³x48 so that the physical spatial size is kept constant at La \approx 2.5 fm. The scale a is fixed by M_rho = 768.4 MeV.

beta	lattice	c_SW	a [fm]	La [fm]	m_PS/m_V for sea quarks #traj (thermalized)
1.80	12³x24	1.60	0.2150(22)	2.580(26)	0.807(1) 6250	0.753(1) 5000	0.694(2) 7000	0.547(4) 5250
1.95	16³x32	1.53	0.1555(17)	2.489(27)	0.804(1) 7000	0.752(1) 7000	0.690(1) 7000	0.582(3) 5000
2.10	24³x48	1.47	0.1076(13)	2.583(31)	0.806(1) 4000	0.755(2) 4000	0.691(3) 4000	0.576(3) 4000
2.20	24³x48	1.44	0.0865(33)	2.076(79)	0.799(3) 2000	0.753(4) 2000	0.705(5) 2000	0.632(7) 2000

Configurations are generated with the Hybrid Monte Carlo (HMC) algorithm. At the main coupling constants beta=1.8-2.1, runs are made with a length of 4000-7000 HMC unit-trajectories per sea quark mass. The additional runs at beta=2.2 are stopped at 2000 HMC trajectories per sea quark mass because the physical lattice size La turned out to be significantly smaller than the other three lattices. (Configurations at beta=2.2 are not stored in JLDG) For the inversion of the quark matrix during the HMC update we use the even/odd preconditioned BiCGStab algorithm.

For the clover coefficient c_SW we adopt a mean field improved choice defined by c_SW = (W_1x1)^-3/4 = (1-0.8412/beta)^-3/4 where for the plaquette W_1x1 the value calculated in one-loop perturbation theory is substituted. This choice is based on our observation that the one-loop calculation of W_1x1 reproduces the measured values well, with a difference of at most 8% in our simulations. We also note that the one-loop value of c_SW = 1+0.678/beta differs from our choice c_SW = 1+0.631/beta+... only by a few per cent. The leading scaling violation with our choice of the clover coefficient is O(g²a).

For more details, see

Light hadron spectroscopy with two flavors of dynamical quarks on the lattice
CP-PACS Collaboration: A. AliKhan, S. Aoki, G. Boyd, R. Burkhalter, S. Ejiri, M. Fukugita, S. Hashimoto, N. Ishizuka, Y. Iwasaki, K. Kanaya, T. Kaneko, Y. Kuramashi, T. Manke, K.-I. Nagai, M. Okawa, H.P. Shanahan, A. Ukawa, T. Yoshi\'e
Phys. Rev. D 65 (2002) 054505 [E: D67 (2003) 059901] (hep-lat/0105015)
Comparative study of full QCD hadron spectrum and quark potential with improved actions
CP-PACS Collaboration: S. Aoki, G. Boyd, R. Burkhalter, S. Hashimoto, N. Ishizuka, Y. Iwasaki, K. Kanaya, T. Kaneko, Y. Kuramashi, M. Okawa, A. Ukawa, and T. Yoshi\'e
Phys. Rev. D 60 (1999) 114508 (hep-lat/9902018)

Configuration ensembles

The table below gives a list of ensembles of configurations. A consecutive set of HMC trajectories for a given value of beta, kappa and c_SW is called an ensemble. A series name (A, B, C, etc) is given to distinguish multiple HMC runs. The step size dt is sometimes changed in the course of HMC evolution as indicated in the table below.

ensemble #	beta	kappa	c_SW	series	lattice	dt	initial traj	final traj	#config	size/config [MB]	total [GB]
1	1.80	0.1409	1.60	A	12³x24	0.0100 0.0330	210 710	700 6750	47 595	22.8	15
2	1.80	0.1430	1.60	A	12³x24	0.0080 0.0250	210 700	680 5450	47 472	22.8	12
3	1.80	0.1445	1.60	A	12³x24	0.0065 0.0200 0.0167	210 510 580	500 570 3830	30 7 322	22.8	9
	1.80	0.1445	1.60	B	12³x24	0.0065	210	3880	362	22.8	9
4	1.80	0.1464	1.60	A	12³x24	0.0033	260	1740	148	22.8	4
	1.80	0.1464	1.60	B	12³x24	0.0033	590	1630	104	22.8	3
	1.80	0.1464	1.60	C	12³x24	0.0033	600	1620	103	22.8	3
	1.80	0.1464	1.60	D	12³x24	0.0066	890	1420	54	22.8	2
ensemble #	beta	kappa	c_SW	series	lattice	dt	initial traj	final traj	#config	size/config [MB]	total [GB]
5	1.95	0.1375	1.53	A	16³x32	0.0313	110	7100	599	72.0	44
6	1.95	0.1390	1.53	A	16³x32	0.0200 0.0250	110 150	140 7130	4 682	72.0	50
7	1.95	0.1400	1.53	A	16³x32	0.0125 0.0156 0.0185	10 70 420	60 410 3420	6 35 294	72.0	25
	1.95	0.1400	1.53	B	16³x32	0.0125 0.0156 0.0185	10 70 480	60 470 3570	6 40 305	72.0	26
8	1.95	0.1410	1.53	A	16³x32	0.0100 0.0080	10 250	240 2500	24 223	72.0	18
	1.95	0.1410	1.53	B	16³x32	0.0100 0.0080	10 240	230 2500	22 226	72.0	18
ensemble #	beta	kappa	c_SW	series	lattice	dt	initial traj	final traj	#config	size/config [MB]	total [GB]
9	2.10	0.1357	1.47	A	24³x48	0.0200	205	4205	798	364.5	291
10	2.10	0.1367	1.47	A	24³x48	0.0160	215	4235	788	364.5	288
11	2.10	0.1374	1.47	A	24³x48	0.0143	205	4200	767	364.5	280
12	2.10	0.1382	1.47	A	24³x48	0.0075	205	1505	253	364.5	93
	2.10	0.1382	1.47	B	24³x48	0.0075	205	1835	320	364.5	117
	2.10	0.1382	1.47	C	24³x48	0.0075	205	1265	213	364.5	78

Naming of Ensembles and Configurations (for ILDG users)

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

mc://JLDG/CP-PACS/RCNF2/RC24x48-B2100K013820C1470

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

lfn://JLDG/CP-PACS/RCNF2/RC24x48-B2100K013820C1470-C-01265

Basename of the dataLFN is formmated as
RgClover spatial size x temporal size - \beta\kappa\c_sw - series - trajectory number

To get full list of markovChainURIs and LFNs, please use QCDml faceted navigation or follow this link.

Location of Configurations (for JLDG users)

Configurations are placed in the gfarm directory
/gfarm/public/ILDG/JLDG/CP-PACS/RCNF2/
and classified in terms of β and κ. For example, the directory
/gfarm/public/ILDG/JLDG/CP-PACS/RCNF2/RC12x24-B1800/K014090
contains configurations at β=1.800 and κ=0.14090. Filename is the same as basename of dataLFN. All ensemble and configuration XML files can be found in
/gfarm/public/ILDG/JLDG/CP-PACS/RCNF2/QCDml

Last update: 03 Sept. 2009, T.Yoshie