Accelerated Discovery of Earth-Abundant Oxygen Evolution Catalysts Through Surface-Informed Active Learning

The development of efficient electrocatalysts for the oxygen evolution reaction (OER) remains a challenge in alkaline water electrolysis. Discovering optimal compositions for complex multi-element electrocatalysts is time-intensive and resource-consuming due to the vast compositional space. Here, we propose a hybrid computational strategy that integrates active learning (AL)-assisted screening with density functional theory (DFT) calculations to accelerate the development of multielement alloy electrocatalysts. Our approach employs 15 cycles of active learning, with 20 NiCoFe (NCF) surface configurations evaluated per cycle, totaling 300 DFT calculations. To minimize computational cost, we utilized scaling relationships and calculated only *O and *OH adsorption energies, while the AL featurization incorporated surface configuration information. This methodology filtered the compositional space of Ni_xCo_yFe_100-x-y alloys and identified the optimal composition of Ni₅₅Co₁₀Fe₃₅. Experimental validation demonstrated that the optimized NCF exhibits accelerated OER kinetics and practical activity, as evidenced by a low Tafel slope of 43.07 mV dec^-1 while maintaining 98.68% of its initial performance over 24 h at 10 mA cm^-2. This study demonstrates the effectiveness of the hybrid AL-DFT strategy for resource-efficient composition screening within the NiCoFe ternary alloy system and highlights its potential for application to other multicomponent catalyst systems following appropriate system-specific model development and validation.