Machine learning assisted screening of doped metals phosphides electrocatalyst towards efficient hydrogen evolution reaction

Transition metals (TM) doped metal phosphides usually exhibits promising reactivity towards acidic hydrogen evolution reaction (HER). However, the experimental screening of highly active TM-doped metal phosphides catalyst is time-consuming and challenging. In this study, a density functional theory combined machine learning (DFT-ML) framework is proposed to accelerate the screening and predicting TM-doped metal phosphides-based HER electrocatalysts. In this framework, the ML database is constructed using critical catalyst features and DFT-calculated adsorption energy of HER intermediates. Also, local average electronegativity of the adsorption site and the surrounding atoms as catalyst feature is proposed to describe the reaction sites in this ML model. Using the HER energetics on the state-of-art highly active Pt (111) as benchmark catalyst model, a set of 10 potential active HER catalysts is predicted. By performing the H* adsorption Gibbs free energy change analysis on these ML-predicted catalysts, six promising TM-doped metal phosphides HER catalysts are determined in the sample space. This study provides a facile and effective approach for the quick screening of high-performance HER electrocatalysts.