ML-guided design and screening of chalcogenide catalysts for hydrogen evolution reaction

Abstract

Transition metal chalcogenides are regarded as the promising electrocatalysts for the hydrogen evolution reaction (HER). However, based on the larger chemical composition space of transition-metal single atom and chalcogenides, the design and screening of excellent HER electrocatalysts remain the challenges. Herein, a machine learning (ML) model was proposed to predict the HER performance of single-atom chalcogenide catalysts, and used to screen the excellent electrocatalysts by combining with density functional theory calculations. The results show that the ML model can predict the HER catalytic activity well. The band gap of support materials is identified as the most important descriptor of single-atom chalcogenide catalysts for HER. Sn@CoS and Ni@ZnS exhibit excellent catalytic activity towards HER, and even outperform the current most efficient Pt catalysts. The hydrogen adsorption free energies of Sn@CoS and Ni@ZnS are 0.04 eV and −0.05 eV, respectively. Both Heyrovsky and Tafel reaction mechanisms are responsible for the HER of Ni@ZnS catalyst. The HER of Sn@CoS catalyst is mainly controlled by the Heyrovsky mechanism. Sn@CoS and Ni@ZnS are considered as the promising electrocatalysts for the HER. This study can provide a competitive tool to predict the activity trends and to accelerate the catalyst design and screening for other catalytic reactions.