Numerical tests of Langevin lattice QCD

Abstract

By numerical means, we address the question of whether systematic errors are controllable in Langevin simulations of lattice QCD. This is done by comparing Langevin updating algorithms, which have been corrected for the analytically calculable errors, to benchmarks set by other methods known to be free of such systematic errors. The updating step size required for the Langevin results to come close to the benchmarks is found to decrease rapidly as the mass of the dynamical quarks decreases. Similar behavior must be expected in other approximate algorithms for full QCD.