Machine-Learning-Accelerated Screening of Single Metal Atoms Anchored on Mnps3 Monolayer as Promising Bifunctional Oxygen Electrocatalysts

Abstract

Searching for bifunctional oxygen electrocatalysts with good catalytic performance to promote the oxygen evolution/reduction reactions (OER/ORR) is of great significance to the development of sustainable and renewable clean energy. Herein, we performed density functional theory (DFT) and machine-learning (DFT-ML) hybrid computations to investigate the potential of a series of single transition metal atoms anchored on the experimentally available MnPS3 monolayer (TM/MnPS3) as the bifunctional electrocatalysts for ORR/OER. The results revealed that the interactions of these metal atoms with MnPS3 are rather strong, thus guaranteeing their high stability for practical applications. Remarkably, the highly efficient ORR/OER can be achieved on Rh/MnPS3 and Ni/MnPS3 with lower overpotentials than those of metal benchmarks, which can be further rationalized by the established volcano and the contour plots. Furthermore, the ML results showed that the bond length of TM atoms with the adsorbed O species (dTM−O), the number of d electrons (Ne), the d-center (ԑd), the radius (rTM), and the first ionization energy (Im) of the TM atoms are the primary descriptors featuring the adsorption behavior. Our findings not only suggest novel highly-efficient bifunctional oxygen electrocatalysts, but also provide cost-effective opportunities for the design of single-atom catalysts using the DFT−ML hybrid method.

Keywords: single-atom catalysts, MnPS3 Substrate, Oxygen Electrocatalysts, Density functional theory, machine learning