Investigation of Surface Passivation Mechanisms in CrCoNi Alloys via Interpretable Machine Learning

This study investigates the formation and stability of passivation films in medium-entropy CrCoNi alloy using an integrated approach combining first-principles calculations and Gaussian Approximation Potential (GAP). By performing extensive structural relaxations coupled with random occupancy of the alloy and multiple potential adsorption sites, the results based on surface reconstruction are analyzed. Our results demonstrate that disordered atomic structures in medium entropy CrCoNi alloy exhibit improved oxide film stability compared to ordered structures, with lower system energies correlating with more robust passivation layers. Furthermore, by constructing physically meaningful descriptors and employing symbolic regression in complex surface environments, simple yet highly correlated rules are obtained, providing insights into the passivation strengthening of CrCoNi alloy from both global and local perspectives. Our work paves the way for exploring surface regularities in complex medium-entropy alloys with the combination of quantum mechanical accuracy and machine learning efficiency.