Multiple time scale method for atomistic simulations

Abstract

A novel multiple time scale approach is proposed which combines dynamic and static atomistic methods in one numerical simulation. The method is especially effective for modeling processes that consist of two distinct phases: the slow phase when atomic equilibrium positions barely change and the fast phase associated with a rapid change of the system’s configuration. In this case, the slow phase can be effectively modeled using static energy minimization while molecular dynamics (MD) can be applied when specific dynamic effects have to be captured. Compared to direct MD simulations, the new method allows for computational cost savings, and eventually simulation timescale extension, since the major part of the simulation can be modeled as static, without the need to follow vibrations of individual atoms and comply with the critical time step requirement of molecular dynamics. As a result, this approach may allow for modeling loading velocities and strain rates that are more realistic than those currently attainable through direct MD simulations. The fundamental issues in developing this method include the correlation between the MD time scale and quasi-static step-like procedure as well as finding effective criteria for switching between the static and dynamic regimes. The method was inspired by and is applied to simulations of atomic-scale stick-slip friction. Possible applications of the new method to other nano-mechanical problems are also discussed.