Scaling cluster expansion models to large and challenging composition spaces

Cluster expansion models are foundational tools in the study of chemically disordered systems owing to their complete representation and exceptional speed. Yet, the study of high-entropy alloys (HEAs) presents significant computational challenges owing to the explosive growth of elemental interaction permutations in large compositions spaces. Here, we develop an efficient active learning workflow that has permitted chemical complexity scaling to match the state-of-the-art with our parameterization of a 6-element cluster expansion model. We showcase our workflow's ability to navigate dynamical instability and thermodynamic metastability in an example system, FCC NiFeCrAlTiSi, which was chosen as an earth-abundant alloy in which to explore multi-passivator effects for corrosion resistance.