Data-Driven Phase Selection, Property Prediction and Force-Field Development in Multi-Principal Element Alloys

Multi-Principal Element Alloys (MPEAs) have brought a paradigm shift in the alloy design process and pose a significant challenge due to the astronomical and compositional space available for exploration. Since experimental and ab-initio methods are more suitable for targeted alloy design over a narrow composition range, data-driven methodologies have shown promise in the search for alloys with unique or improved properties. This chapter introduces the field of materials informatics by laying out the fundamentals of machine learning, i.e., the types of problems, dataset formulation, feature selection, and machine learning algorithms, as applied to materials science. It follows with a holistic review of the existing data-driven models targeted towards the prediction of phase selection, mechanical properties and ordering behaviour in MPEAs. It also discusses the methodology for the development of machine learning force fields that enable atomistic modelling of various phenomena, such as diffusion, phase transformations and mechanical deformation, in MPEAs.