The Importance of Phase Equilibria and Diffusion in High-Throughput and Autonomous Materials Discovery

Abstract Machine learning has accelerated the prediction and screening of crystal structures, while automation has increased the number of compositions and processing conditions that can be explored experimentally. However, identifying a structure near the zero-temperature convex hull or detecting its signature in a synthesis product does not, by itself, establish materials discovery. In this perspective, we argue that finite-temperature phase equilibria and transformation kinetics should guide these eff orts from the outset. CALPHAD, combined with diff usion data, provides a framework for selecting compositions and processing conditions capable of producing the required phase constitution, including metastable states whose formation and persistence depend on the processing path. Rigorous characterization is then needed to establish what was produced and whether its properties and stability meet the intended requirements. Progress should be measured by reproducible processing windows that yield the required phase constitution, microstructure, and properties under manufacturing and service conditions.