Machine Learning Unveils Ce–O–Co Motifs in Perovskite with Lewis Acidic–Basic Microenvironment for Dual-Site Intramolecular Oxygen Evolution in Saline Water

Perovskite oxides are promising anodic electrocatalysts for alkaline saline water electrolysis, yet their performance is often limited by sluggish oxygen evolution reaction kinetics and competing chloride corrosion. Building on Pr0.1Sr0.9Co0.5Fe0.5O3 (PSCF), a candidate previously identified via our transfer learning paradigm, we further leveraged data-driven compositional searching of the A-site configuration, which led to an optimized catalyst with a proper Ce/Sr/Pr ratio (Ce-PSCF). The asymmetric Ce–O–Co motifs invoked a negative charge-transfer regime and activated the evolution of nonbonding oxygen (ONB) states, which reconciled the activity-stability trade-off and promoted a dual-site intramolecular lattice oxygen mechanism, as evidenced by promoted 18O–18O evolution. The electrolyzer with Ce-PSCF lowered the overpotential of the PSCF-based one by 100 mV at 300 mA cm–2 and exhibited 15-fold greater durability. Electrochemical analysis revealed a higher reaction order, promoted *OH coverage, and faster surface reconstruction on the Ce-incorporated electrocatalyst, indicating *OH-controlled chemical-step kinetics. Theoretical insight inferred a variation of the rate-determining step from O–O coupling to secondary OH– refilling with a reduced energy barrier from 0.68 to 0.54 eV. Moreover, ab initio molecular dynamics simulations visualized a dynamically strengthened affinity for OH– and significant repulsion of Cl–, arising from its intrinsic Lewis acidic and basic microenvironment.