Deformable stochastic boundaries in molecular dynamics

Abstract

A new approach to studying localized chemical events in condensed phases is developed. The approach provides a simple and convenient method for reducing the total number of solvent particles explicitly included in simulations of localized processes while decreasing spurious edge effects. Both energy flow across the boundary and density fluctuations in the simulation region are included; this makes it possible to treat nonequilibrium processes, such as thermal gradients and endothermic or exothermic chemical reactions. The essential element of the approach is the introduction of a soft boundary force and a stochastic buffer region. For simple liquids, the boundary force is determined from the solvent equilibrium structure (radial distribution function) and is readily incorporated into conventional molecular dynamics algorithms. The methodology is illustrated by application to liquid argon in spherical and cubical simulation regions; comparison with standard molceular dynamics results show excellent agreement for structural, dynamic, and thermodynamic properties.