Mapping continuous potentials to discrete forms

Abstract

The optimal conversion of a continuous inter-particle potential to a discrete equivalent is considered here. Existing and novel algorithms are evaluated to determine the best technique for creating accurate discrete forms using the minimum number of discontinuities. This allows the event-driven molecular dynamics technique to be efficiently applied to the wide range of continuous force models available in the literature, and facilitates a direct comparison of event-driven and time-driven molecular dynamics. The performance of the proposed conversion techniques are evaluated through application to the Lennard-Jones model. A surprising linear dependence of the computational cost on the number of discontinuities is found, allowing accuracy to be traded for speed in a controlled manner. Excellent agreement is found for static and dynamic properties using a relatively low number of discontinuities. For the Lennard-Jones potential, the optimized discrete form outperforms the original continuous form at gas densities but is significantly slower at higher densities.