Machine learning to predict aluminum segregation to magnesium grain boundaries

Abstract Magnesium alloys are good candidates for a number of applications due to their high strength-to-weight ratio, but other properties like corrosion resistance, formability, and creep are still a concern. In magnesium-aluminum alloys, Mg17Al12 phase precipitates at the grain boundaries (GBs) can have important implications on the mechanical and corrosion behavior. In order to better understand the effects, atomistic segregation of aluminum to GBs must be evaluated first. This study uses atomistic simulations to quantify aluminum segregation energetics for training a machine learning model. Aluminum atoms were iteratively placed at various atomic sites near 30 different 〈 0001 〉 symmetric tilt grain boundaries (STGBs) in magnesium. The results show how aluminum segregation is affected by GB structure and the local atomic environment. The ability to compute grain boundary physical properties of interest using machine learning techniques can have broad implications for the area of grain boundary science and engineering.