Bridging Theory and Experiment: Machine Learning Potential‐Driven Insights into pH‐Dependent CO₂ Reduction on Sn‐Based Catalysts

Abstract Sn‐based materials are among the most promising catalysts for CO ₂ reduction reaction (CO ₂ RR) to formic acid. However, the complex electrochemistry‐induced surface reconstruction under negative potentials has hindered the precise elucidation of the structure‐performance relationship. Herein, machine learning potential (MLP) is employed to accelerate molecular dynamics (MD) simulations, and pH‐field coupled microkinetic modelling is perfromed to unravel the pH dependence of CO ₂ RR at the reversible hydrogen electrode (RHE) scale. Encouragingly, the developed MLP reveals that SnO ₂ adopts a nanorod‐like morphology, accurately reproducing experimentally observed reconstruction phenomena. Additionally, SnS ₂ prefers to form a rougher surface. Leveraging the precisely determined reconstructed surface, the exciting pH‐dependent behavior of Sn‐based catalysts is highlighted: the increase of pH will cause a left‐shift in the CO ₂ RR volcano and ultimately enhance the catalyst's activity. Most importantly, the excellent agreement between the theoretical simulations and our subsequent experimental measurements validates the accuracy of the simulations in terms of turnover frequencies, providing a clear benchmarking analysis between experiments and the MLP‐MD‐assisted pH‐field coupled microkinetic modelling. This work not only offers a valuable MLP‐based approach for studying surface reconstructions, but also provides new guidance for the design of high‐performance complex catalysts for CO ₂ RR.