Study of a generalized Metropolis decision rule in auxiliary field quantum Monte Carlo

Abstract

We consider a generalization of the standard Metropolis algorithm acceptance/rejection decision rule and numerically explore its properties using auxiliary field quantum Monte Carlo. The generalization involves a free parameter which, given a criterion for proposing attempted moves, can be used to tune the average acceptance rate in a particular way. Such tuning can also potentially change Monte Carlo autocorrelation times, and the combination of the changing acceptance rate and autocorrelation times raises the possibility of more efficient simulations. We explore these issues using primarily massively parallel quantum Monte Carlo runs of the ``test case'' two-dimensional Hubbard model, and discuss results and applications.