Unravelling the origin of bifunctional OER/ORR activity for single-atom catalysts supported on C2N by DFT and machine learning

Abstract

Designing high-performance bifunctional oxygen evolution/reduction reaction (OER/ORR) catalysts is a newly emerging topic and these catalysts have wide applications in metal–air batteries and fuel cells. Herein, we report a group of (27) single-atom catalysts (SACs) supported on the C₂N monolayer as promising bifunctional OER/ORR catalysts by theoretical calculations. In particular, Rh@C₂N exhibits a lower OER overpotential (0.37 V) than the IrO₂(110) benchmark with good ORR activity, while Au and Pd@C₂N are superior ORR catalysts (with an overpotential of 0.38 and 0.40 V) to Pt(111) and their OER performance is also outstanding. More importantly, we discover the origin of the bifunctional catalytic activity by density functional theory (DFT) calculations and machine learning (ML). Using DFT, we find a volcano-shaped relationship between the catalytic activity and ΔG_O, and finally link them to the normalized Fermi abundance, a parameter based on the electronic structure analysis. We further unravel the origin of element-specific activity by ML modelling based on the random forest algorithm that considers the outer electron number and oxide formation enthalpy as the two most important factors, and our model can give an accurate prediction of ΔG_O with much reduced time and cost. This work not only paves the way for understanding the origin of bifunctional OER/ORR activity of SACs, but also benefits the rational design of novel SACs for other catalytic reactions by combining DFT and ML.