Bayesian-optimized surface energy microstructure-informed model of active dissolution in CrMnFeCoNi high-entropy alloys

A comprehensive mathematical framework was developed to predict the corrosion behavior of the Cantor high-entropy alloy (CrMnFeCoNi) by quantitatively incorporating grain orientation, grain size, and the fraction of special grain boundaries. High-entropy alloy specimens with average grain sizes ranging from approximately 7 µm (fine-grained) to 67 µm (coarse-grained) were prepared through cold rolling followed by annealing and subsequently exposed to 3.5 % NaCl solution. The model, calibrated through Bayesian optimization, reproduced the experimentally observed corrosion trends with high fidelity. Specifically, the corrosion current density increased from approximately 1.5 × 10⁻⁶ A/cm² in the fine-grained condition to 2.5 × 10⁻⁶ A/cm² in the coarse-grained condition, despite the latter exhibiting a higher proportion of special (low-energy) grain boundaries. Detailed microstructural investigations demonstrated that no single factor—grain size, texture, or boundary fraction—was solely responsible for the corrosion response. Instead, crystallographic orientation, associated with surface energy, was identified as the dominant parameter, reconciling the opposing influences of grain size and special boundary fraction on corrosion resistance observed in this study. Electrochemical impedance spectroscopy confirmed active dissolution behavior, with charge-transfer resistance (R ct ) on the order of 10⁴ Ω·cm² and no indication of passive film formation, a finding corroborated by XPS analysis. Post-corrosion examination revealed that pitting occurred selectively within high-surface-energy grains, with pit diameters (∼10–75 µm) corresponding closely to the grain size. This observation aligned with the model’s prediction that preferential dissolution arises in specific crystallographic orientations. The proposed microstructure-informed modeling approach demonstrated an accuracy improvement of nearly eightfold (R 2 =0.88) over conventional empirical models, offering a reliable strategy for designing and controlling the dissolution behavior of high-entropy alloys and potentially other metallic systems.