Optimizing the computational efficiency of surface tension estimates in molecular dynamics simulations

Use of molecular dynamics (MD) as the “frontend” in a multi-scale simulation requires high accuracy and therefore demands significant computer resources, particularly for obtaining interfacial properties. In this work we investigate the computational requirements for estimating the surface tension of a planar interface, the accuracy of which depends on the errors due, respectively, to finite system size and statistical uncertainty. By investigating the parameters affecting the two errors, we derive the optimal strategy for minimizing the computational time for a given desired total error and apply the estimations to two modifications of the Lennard-Jones potential. The presented formalism may be used as a guide for optimizing the efficiency of similar calculations using MD.