Parallelized sampling of the Gibbs ensemble

Abstract

An expression for the probability distribution of NVT-like sub-ensembles constituting the Gibbs ensemble is derived. Knowledge of this distribution makes it possible to carry out the simulation without the explicit exchange of real particles between the simulation boxes and to evaluate directly any Gibbs ensemble average from a series of independent simultaneous simulations (Monte Carlo or molecular dynamics) performed on a set of NVT-likt sub-ensembles with the fixed distribution of particles. An implementation of the method, which is tailored mainly for complex systems, is exemplified for the square-well fluid, and its efficiency and results are compared with those obtained from the conventional Gibbs ensemble simulations.