An Efficient Replica Exchange Monte Carlo Method Using the Gaussian Ensemble for First-Order Transitions

摘要

We propose an efficient replica exchange Monte Carlo method to simulate first-order phase transitions. This is realized by the hybrid use of the canonical and gaussian statistical ensembles; the latter is an artificial statistical ensemble interpolating the microcanonical and canonical ensembles. Our method has been applied to a two dimensional ferromagnetic Potts model exhibiting the first-order transition. Configurations both below and above the transition temperature were demonstrated to be well sampled via sufficient exchange of states of replicas by the canonical and gaussian hybrid replica exchange method, while replicas could not cross the energy gap region associated with the latent heat of the transition by a standard canonical replica exchange Monte Carlo calculation.