• High-throughput screening of TM-MXene catalysts for ORR/OER. • Scaling relations between adsorption strength of intermediates. • TM’s physical properties and binding energy to MXene are key to catalytic activity. • Identification of η ORR/OER formulas using symbolic regression. • Data-driven machine learning strategy uncovers structure–activity relationships. Oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) are pivotal reactions in sustainable energy conversion, necessitating the utilization of highly efficient catalysts. This study systematically evaluated the ORR/OER catalytic performance of 239 transition metal-MXene (TM-M 2 CT 2 with T=O/F single-atom catalysts (SACs) by combining density functional theory (DFT) calculations with machine learning (ML) prediction. Pd-Nb 2 CO 2 , Pd-V 2 CF 2 , Rh-Zr 2 CF 2 , Pd-Cr 2 CF 2 , and Ni-W 2 CF 2 were selected as outstanding bifunctional catalysts. The eXtreme gradient boosting regression (XGBR) and random forest regression (RFR) models effectively predicted the catalytic performance of TM-M 2 CT 2 . Feature importance analysis revealed that the physical properties of the active center TM atom (the d-electron number and radius) and the binding strength between TM and the substrate M 2 CT 2 are crucial features influencing catalytic activity. The symbolic regression (SR) model established direct relationships between fundamental structural features and ORR/OER overpotentials. This study provides a data-driven strategy for accelerating catalyst screening and exploring rational design approaches, laying the foundation for experimental research.
Wu et al. (Fri,) studied this question.