High Statistics Calculations of Quenched Qcd Spectrum Using Various Quark Sources

We present the results for the hadron spectrum calculated on 400 configurations using point source, wall source and 8-cubic sources, in quenched QCD with Wilson fermions at β = 6.0 and K = 0.155 on a 24 3 × 54 lattice. The results for the ground state masses obtained with three types of quark sources agree well with each other. Masses of the first excited states appear consistent with experimental values within large errors. Recent high statistics simulations of quenched QCD have produced high quality data for the masses of light hadrons[1–5]. However, mass results obtained with various types of quark sources often do not agree with each other[6,1,2]. Masses of the first excited states also often appear unexpectedly heavy[2,3,7]. In order to resolve these problems, the QCD-PAX Collaboration has carried out a high statistics quenched calculation with Wilson quarks at β= 6.0 on a 24 3 × 54 lattice. Taking the last configuration of our previous work[1] as a starting configuration, we have generated new 400 configurations which are separated by 1000 heat bath sweeps. We choose K = 0.155: This K corresponds approximately to the strange quark mass, because the m π /m ρ appears to be 0.7. We use three types of quark sources: point source, wall source and 8-origin cubic sources of a size 7 3 (with a random Z(3) element on each cube). We call the third source multi-cube source. Hereafter we abbreviate them to p-, wand m-sources, respectively. Point sinks are used for all sources. Configurations are fixed to Coulomb gauge before quark propagators are solved for the wand the m-sources. Although we have 600 samples for the p-source in total, we report the results obtained by using the new 400 configura-* Talk presented by T.Yoshié at Lattice '93 tions, in order to compare results with three types of sources on an equal footing. Masses of the ground states are determined by global fits to data for [t min , t max ] for which the χ 2 /dof of the fits are less than 1.5. We take account of time correlations in the fits. Errors are estimated by single elimination jack-knife procedure. We have checked that errors do not change significantly even if we change the bin size in the jack-knife error analyses. In addition, we have estimated errors which depend on the choice of fitting ranges. Here we define the fitting-dependent upper …