A j-Walking Algorithm for Microcanonical Simulations: Applications to Lennard-Jones Clusters

The j-walking method, previously developed to solve quasiergodicity problems in canonical simulations, is extended to simulations in the microcanonical ensemble. The implementation of the method in the microcanonical ensemble parallels that in the canonical case. Applications are presented in the microcanonical ensemble to cluster melting phenomena for Lennard-Jones clusters containing 7 and 13 particles. Significant difficulties are encountered in achieving ergodicity when Metropolis Monte Carlo methods are applied to these systems, and the difficulties are removed by the j-walking method.