Universal and Interpretable Descriptor to Design Dual‐atom Catalysts for Multi‐Type C–C Coupling with Ultrahigh C ₂₊ Yield

Abstract Dual‐atom catalysts (DACs) have demonstrated superior performance in the C─C coupling processes to generate multi‐carbon (C ₂₊ ) products due to the multiple metal sites of DACs. However, the rational design of DACs that exhibit high activity and specific selectivity for the C─C coupling process in the CO ₂ reaction remains a significant challenge due to the intricate inter‐metal interaction. Herein, a universal and interpretable descriptor is proposed for the emerging metal‐nitrogen‐carbon (M ₁ –M ₂ –N ₄ –C) dual‐atom catalysts in the CH ^* ─CH ^* coupling process, effectively evaluating the activity and selectivity for various C─C coupling (i.e., CH ^* ─CH ^* , CO ^* ─CO ^* , CHO ^* ─CO ^* , CH ^* ─CH ₂ ^* ). This defined descriptor, which consists of inherent atomic properties (electronegativity, d electron number), proves to facilitate high‐throughput screening of more than 400 graphene‐based DACs. Through this screening process, potential catalysts such as FeZr@NC are predicted to have superior activity and selectivity for C─C coupling with a C ₂₊ yield of 22.6%, outperforming the state‐of‐the‐art catalysts by a factor of 3 to 30. Importantly, this descriptor can also be extended to evaluate the activity and selectivity of DACs for other C─C coupling processes (i.e., CO ^* ─CO ^* , CHO ^* ─CO ^* , and CH ^* ─CH ₂ ^* ), with very small discrepancies between the prediction and computational results. By systematically integrating machine learning (ML) with physical insights, this work offers practical guidelines for the construction of universal descriptors for DACs based on inherent atomic properties and the logical design of advanced catalysts in C─C coupling.