The Langevin method and Hubbard-like models

Abstract

We reexamine the difficulties associated with application of the Langevin method to numerical simulation of models with non-positive definite statistical weights, including the Hubbard model. We show how to avoid the violent crossing of the zeroes of the weights and how to move those nodes away from the real axis. However, it still appears necessary to keep track of the sign (or phase) of the weight. © 1989.