High-entropy alloys (HEAs) are of particular interest due to their unique combination of high strength, ductility, and fracture resistance. These properties are largely impacted by the phases and number of phases present in the alloy. Due to the enormous chemical design space, with typically five or six elements present in the alloys in addition to the various processing parameters, the prediction of the number of phases is challenging. While there have been multiple recent reports of phase predictions through data-driven models, an alternate approach is proposed and demonstrated here. In this work, phase prediction is performed using features derived from DFT-calculated pairwise interactions among constituent elements (incorporating bonding and structure) rather than relying solely on traditional elemental descriptors such as electronegativity. By focusing on atomic interaction features, our model provides a novel perspective on the phase stability in HEAs. This approach yields a predictive model for phase formation in HEAs that is comparable in accuracy to prior models while offering improved interpretability. By analyzing the contributions of various binary interactions to the prediction, the model provides insight into the atomic-scale factors influencing whether an alloy forms a single-phase solid solution (SS) or multiphase microstructure.
© 2025 The Authors. Published by American Chemical Society.