Fitting interatomic potentials using molecular dynamics simulations and inter-generation projection genetic algorithm

Abstract

In this paper we propose a new algorithm to fit interatomic potentials. In the new algorithm, molecular dynamics simulations are applied to calculate the material properties which are used to match the experimental data during the fitting procedure. This includes the effect of atom relaxations in fitting calculations. An inter-generation projection genetic algorithm is used to optimize the fitting parameters until the error between the calculated and experimental material properties is within tolerance. This leads to a global optimal solution. The new algorithm significantly improves the accuracy and transferability of the fitted potential. It has been demonstrated by a numerical example of fitting potential of nickel.