Electronic and Geometric Effects of Dual-Atom Catalysts on C−C Coupling during CO2 Reduction: Insights from Density Functional Theory and Interpretable Machine Learning

Dual-atom catalysts (DACs) offer unique opportunities for promoting C ₂₊ formation in CO ₂ electroreduction; however, the roles of electronic and geometric structures in governing C − C coupling remain unclear. Herein, we systematically investigate eighty fourth-row transition-metal dimers anchored on N-doped graphene using density functional theory combined with interpretable machine learning. Our results demonstrate that both *CO and *OCCO adsorption strongly depend on the metal identity and metal–metal distance, exhibiting volcano-type or linear relationships that lead to regular variations in C − C coupling energies. To establish quantitative relationships between coupling energy and the electronic and geometric structures, we construct interpretable machine-learning models based on physically motivated descriptors. Feature screening and SISSO analysis identify the number of metal valence electrons and the metal−metal distance as the dominant factors governing coupling energies. The resulting three-dimensional descriptor achieves good predictive performance, enabling rapid evaluation of DAC activity for the C − C coupling process. Overall, this study provides fundamental insights into the structure-activity relationships of DACs and offers a practical strategy for designing catalysts with enhanced C ₂₊ selectivity.