Identifying early stage precipitation in large-scale atomistic simulations of superalloys

Abstract

A method for identifying and classifying ordered phases in large chemically and thermally disordered atomistic models is presented. The method uses Steinhardt parameters to represent local atomic configurations and develops probability density functions to classify individual atoms using naïve Bayes. The method is applied to large molecular dynamics simulations of supersaturated Ni-20 at% Al solid solutions in order to identify the formation of embryonic γ′-Ni3Al. The composition and temperatures are chosen to promote precipitation, which is observed in the form of ordering and is found to occur more likely in regions with above average Al concentration producing 'clusters' of increasing size. The results are interpreted in terms of a precipitation mechanism in which the solid solution is unstable with respect to ordering and potentially followed by either spinodal decomposition or nucleation and growth.