Combining first-principles calculations and machine-learning to model phase stability and defects in complex ceramic materials

The successful development and manufacturing of new materials, for applications ranging from thermal protection barriers in aerospace engineering to semiconductors for extreme environments, depends on computational thermodynamics to predict phase stability and defect formation. Thermodynamic models must incorporate entropy, which is particularly important at high temperatures for multi-element materials. Descriptors and thermodynamic models based on the thermodynamic density of states extracted from ensembles of ordered calculations have previously been used to predict the synthesizability of new materials such as high entropy carbides. These methods are now being extended and combined with machine-learning to predict phase-stability in high-entropy rare-earth silicates for thermal and environmental barriers in gas turbines, and to investigate defect formation in multi-component semiconductors.