Universal framework for efficient estimation of stability in multi-principal element alloys

Predicting the synthetic accessibility of multi-principal element alloys (MPEAs) across the global chemical space remains a challenge. In this study, we show that the synthesizability of MPEAs across broad compositional and structural spaces can be predicted using a physical model that expresses the total energy of any MPEA as a linear combination of energies from lower-dimensional subsystems. The model is validated with a large computational dataset and supported by the experimental synthesis of multiple MPEAs, achieving mean absolute errors near or below 7 meV/atom on a density functional theory dataset of 135,791 MPEAs spanning 28 metals and up to ten components. Its accuracy is comparable to state-of-the-art deep learning models while maintaining interpretability through cluster-expansion theory. Moreover, we show that the stability of high-entropy alloys can be predicted using a linear combination of energies from lower-dimensional systems with low errors, indicating a flatter energy landscape at high compositional complexity.