First-principles adaptive-boost accelerated molecular dynamics simulation with effective boost potential construction methods: a study of Li diffusion in Si crystal

Abstract

First-principles molecular dynamics (FPMD) simulation is a powerful tool for studying the mechanical, chemical, and thermal properties of materials. However, because of the fundamental time-scale limitation of molecular dynamics and the high computational cost of density functional theory, many of the rare events that dominate these properties cannot be directly investigated. Several methods of accelerating FPMD calculations have been developed to analyze these rare events, but to the best of the authors’ knowledge, no quantitative computation of event rate and activation free energy has been achieved. Here, we have developed an accelerated FPMD based on FPMD and the adaptive boost method (Ishii et al. Phys. Rev. B 85, 064303 (2012)), which remove the fundamental drawback of FPMD: time-scale limitation. We also propose a novel boost potential construction algorithms that allows an accurate boost potential to be constructed with a limited number of FPMD samplings. In this study, to confirm the validity of the method, we compute the diffusion of a Li atom in a Si crystal. The temperature-dependent diffusivity was obtained and the activation enthalpy was calculated from an Arrhenius plot and compared with reported experimental diffusivities.