Data-centric framework for crystal structure identification in atomistic simulations using machine learning

Abstract

The spatial complexity of cross-scale atomistic simulations renders them unsuitable for simple human visual inspection. Instead, specialized structure characterization techniques are required to aid interpretation. These have historically been challenging to construct, requiring significant intuition and effort. In this article the authors introduce a data-centric framework that favors the employment of machine learning over heuristic rules of classification. It is demonstrated that the data-centric framework outperforms all of the most popular heuristic methods while introducing a systematic route for generalization to new crystal structures.