Molecular Simulation of Three-body Interactions for Vapor-liquid and Solid-liquid Phase Equilibria

The phase behavior of both pure systems and binary mixtures has been widely studied by molecular simulation [1], using techniques such as the Gibbs ensemble [2], Gibbs-Duhem [3] or histogram reweighing [4] algorithms. The common aim of many of these investigations is to accurately predict the phase diagram using effective intermolecular potentials, most notable the Lennard-Jones potential. In contrast, other studies [5-11] have used molecular simulation techniques, in conjunction with genuine twoand three-body intermolecular potentials [12-14], to determine the influence of various intermolecular interactions on phase behavior. These studies have concluded that three-body interactions have a significant influence on phase behavior. Threebody interactions decrease the density of the liquid phase of pure fluids [5,6,7,9] and they contribute significantly [6] to the vapor-liquid critical point. In binary mixtures [10,11], three-body interactions are required to obtain good agreement between theory and experiment for the pressurecomposition behavior. There is also some evidence [15] that three-body interactions have a pivotal role in the transition between the different global phase behavior types of binary mixtures. Previous investigations of three-body interactions on phase equilibria have been confined largely to pure fluids. Theoretical studies have been reported [16-22] which indicate three-body interactions are important in solid phases. However, the direct molecular simulation of solid-liquid equilibria [23-25] for both pure fluids and mixtures has mainly focused on predicting phase coexistence using an effective intermolecular potential. The solid-liquid phase transition is difficult to determine accurately using traditional molecular simulation techniques. The high densities mean that it is not practical to use the Gibbs ensemble [2] because of the difficulty of exchanging particles between the phases. Although this limitation is avoided by the Gibbs-Duhem [3] technique, it is not self-starting which means it requires prior knowledge of one pair of coexistence data. Therefore, its ability to predict the phase boundary largely depends on the accuracy of the starting point data. In this work, we employed a novel approach [26] for locating the solid-liquid phase boundary which combines elements of both equilibrium and non-equilibrium molecular dynamics techniques. The approach yields reliable calculations and it avoids the problems encountered in both Gibbs ensemble and Gibbs-Duhem methods.