Density Functional Theory-Machine Learning Characterization of the Adsorption Energy of Oxygen Intermediates on High-Entropy Alloys Made of Earth-Abundant Metals

High-entropy alloys (HEAs) have emerged as promising electrocatalysts due to their high tunability. Among HEAs, those made of earth-abundant metals have shown high stability and corrosion resistance, making them attractive as low-cost alternatives to noble metal electrocatalysts. However, the catalytic characteristics of these HEAs remain largely unexplored, mainly due to computational challenges posed by the vast number of local binding environments on their surfaces. Here, we combine density functional theory calculations and machine learning (ML) regression models to reconstruct the distribution of adsorption energies of O* and HO* on HEAs containing CoFeNi-X, where X represents Mo, Mn, or Cr. Our ML models predict the adsorption energies on different HEA binding sites with reasonable accuracy despite the modest size of the training data sets. We find that although hollow binding sites are preferred for both O* and HO*, the elemental composition of the HEAs significantly influences the preferred binding site types, with Mo and Cr promoting bridge and on-top binding sites, particularly for HO*. We also find that while the scaling relationship between average adsorption energies of O* and HO* holds for equimolar HEAs, local disruptions to the scaling relationship can occur induced by specific stoichiometric changes. Our study also provides insight into the contributions of different chemical environments to the adsorption energy distribution, providing valuable guidance for the future design of HEA electrocatalysts