Integrating\nActive\nLearning and DFT for Fast-Tracking\nSingle-Atom Alloy Catalysts in CO<sub>2</sub>‑to-Fuel Conversion

Electrocatalytic\ncarbon dioxide reduction (CO<sub>2</sub>RR) technology\nenables the conversion of excessive CO<sub>2</sub> into high-value\nfuels and chemicals, thereby mitigating atmospheric CO<sub>2</sub> concentrations and addressing energy scarcity. Single-atom alloys\n(SAAs) possess the potential to enhance the CO<sub>2</sub>RR performance\nby full utilization of atoms and breaking linear scaling relationships.\nHowever, quickly screening high-performance metal portfolios of SAAs\nremains a formidable challenge. In this study, we proposed an active\nlearning (AL) framework to screen high-performance catalysts for CO<sub>2</sub>RR to yield fuels such as CH<sub>4</sub> and CH<sub>3</sub>OH. After four rounds of AL iterations, the ML model attained optimal\nprediction performance with the test set <i>R</i><sup>2</sup> of approximately 0.94 and successful prediction was achieved for\nthe binding free energy of *CHO, *COH, *CO, and *H on 380 catalyst\nsurfaces with an accuracy within 0.20 eV. Subsequent analysis of the\nSAA catalysts’ activity, selectivity, and stability led to\nthe discovery of eight previously unexplored SAA catalysts for CO<sub>2</sub>RR. Notably, the surface activity of Ti@Cu(100), Au@Pt(100),\nand Ag@Pt(100) shines prominently. Utilizing DFT calculations, we\nelucidated the complete reaction pathway of the CO<sub>2</sub>RR on\nthe surfaces of these catalysts, confirming their high catalytic activity\nwith limiting potentials of −0.11, −0.34, and −0.46\neV, respectively, which are significantly lower than those of pure\nCu catalysts. The results showcase the exceptional predictive prowess\nof AL, providing a valuable reference for the design of CO<sub>2</sub>RR catalysts.