Artificial intelligence applied to atomistic kinetic Monte Carlo simulations in Fe–Cu alloys

Abstract

Vacancy migration energies as functions of the local atomic configuration (LAC) in Fe–Cu alloys have been systematically tabulated using an appropriate interatomic potential for the alloy of interest. Subsets of these tabulations have been used to train an artificial neural network (ANN) to predict all vacancy migration energies depending on the LAC. The error in the prediction of the ANN has been evaluated by a fuzzy logic system (FLS), allowing a feedback to be introduced for further training, to improve the ANN prediction. This artificial intelligence (AI) system is used to develop a novel approach to atomistic kinetic Monte Carlo (AKMC) simulations, aimed at providing a better description of the kinetic path followed by the system through diffusion of solute atoms in the alloy via vacancy mechanism. Fe–Cu has been chosen because of the importance of Cu precipitation in Fe in connection with the embrittlement of reactor pressure vessels of existing nuclear power plants. In this paper the method is described in some detail and the first results of its application are presented and briefly discussed.