Designing High-Entropy Alloys: From Physical Metallurgy to Machine Learning Enabled Discovery

High-entropy alloys (HEAs) are characterized by having more than one principal element, in contrast to a traditional alloy with only one base element; this provides a much wider composition space compared to traditional alloys; hence, HEAs have been the most studied metal alloys in the last couple of decades. This review links the fundamentals of these materials to the data-driven methods now used to design them. The first part of this review describes the fundamentals of these materials, including the thermodynamic and kinetic basis of these materials and an entropy-based classification. The subsequent parts of this review describe the synthesis routes of these materials via liquid-state, solid-state, gas-state, and other processing routes. The choice of processing route can have a significant impact on the resulting material, including the phase(s) formed, the chemical homogeneity, and contamination of the material. We have systematically reviewed the mechanical, corrosion, thermal, electrical, and soft magnetic properties of these HEAs. In the final section, we discuss how artificial intelligence (AI) and machine learning (ML) can be used to solve the problem of combinatorial explosion of the design space of HEAs: supervised classifiers for phase prediction, regression and ensemble models for hardness, strength, and elastic moduli, active learning and Bayesian optimization for closed-loop discovery, high-throughput screening with CALPHAD, and machine-learned interatomic potentials. We conclude with continuing constraints such as lack of data, noisy labels, and poor encoding of process history, and areas of development including transfer learning, generative design, and self-driving laboratories.