Machine learning-accelerated atomistic kinetic Monte Carlo simulations to predict embrittlement in FeCrAl alloys

Abstract FeCrAl alloys are a potential fuel cladding material for existing and next generation nuclear reactors.However, they are limited by irradiation induced embrittlement caused by the formation of Cr rich α′ 15 precipitates. To understand the mechanisms of this precipitate formation and growth, a machine learning surrogate model was developed and used to perform atomistic kinetic Monte Carlo simulations. These simulations demonstrate the effect of alloy composition, irradiation, and temperature effects on the formation and growth kinetics of Cr rich precipitates, addressing open questions in the literature. Minimizing Cr content with higher Al content was found to reduce the rate of precipitate formation. The predicted Cr precipitate size and concentration showed good agreement with experiments. The precipitate kinetics suggested that rather than a nucleation and growth model, Cr enrichment occurs uniformly though the material resulting in nanometer size precipitates.