Integrating Machine Learning With Constant‐Potential Simulation to Unravel Charge‐Transfer Mechanisms in Electrochemical Nitrogen Fixation

The electrochemical nitrogen reduction reaction (NRR) offers a sustainable approach to ammonia (NH₃) synthesis under mild conditions. To achieve scalable NH₃ production, discovering high-performance catalysts for the efficient NRR is crucial. For this purpose, the activity mechanisms of functional group-modified carborin/graphene-supported single-atom catalysts were systematically investigated using the grand-canonical fixed-potential method, which simulates operando constant-potential conditions. Among 144 candidates screened, Cr@NO₂-carborin/graphene and Cr@CHO-carborin/graphene are identified as the most promising NRR catalysts, with low limiting potentials of -0.220 V for the ^*N₂→^*N₂H step and -0.245 V for the ^*NH→^*NH₂ step, respectively. Furthermore, interpretable machine learning models revealed that the shift in potential of zero charge, induced by intermediate adsorption, serves as the key voltage-responsive descriptor governing charge transfer patterns and influencing the N₂ activation. These findings establish a paradigm shift from static electronic descriptors to dynamic interfacial property engineering, offering a universal framework for designing electrocatalysts for multi-electron reactions like NRR.