Computational and AI-Assisted Design of Dual-Metal Single-Atom Catalyst for Oxygen Reduction Reaction

The electrocatalytic oxygen reduction reaction (ORR) plays a crucial role in numerous energy and sustainability systems, such as fuel cells, metal-air batteries, and water electrolysers. It holds significant potential for renewable energy generation, transportation, and storage, heralding a cleaner and more sustainable future. Recent trends have shown increased use of single-atom catalysts (SACs), particularly metal-N4 moieties grown on graphene-based 2D materials, for enhancing ORR efficiency. However, the rational design of SAC for high-performance ORR faces challenges due to unclear structure-property relationships and the limits of conventional experimental trial-and-error approaches. In this study, we harnessed the power of the density functional theory (DFT) calculations, combined with cutting-edge machine learning (ML) techniques, to explore 144 SACs featuring dual interacting M1-N4 and M2-N4 moieties (M1, M2 = Mn, Fe, Co, Ni, Cu, Ru, Rh, Pd, Ag Ir, Pt, Au), denoted as M1-M2, grown on graphene. Of all the catalysts we examined, emerged as the top performer, achieving an impressive overpotential of 0.980 V in alkaline conditions — outperforming most previously reported SACs. Even more striking, 25 of the evaluated SACs surpassed the renowned Fe-N4 SAC in catalytic efficiency, including more economically viable alternatives like Fe-Ag. Venturing further into innovation, we developed three ML models that accurately predict the overpotentials of various M1-M2 SACs, showing their strong ability to capture the relationship between single-atom metal site properties and overpotential. These models provide useful navigation toolkits for the rational design of effective electrocatalysts. Our study sheds light on the path toward achieving efficient SAC-catalyzed ORR, contributing to a more sustainable and energy-efficient future.