Machine Learning-Predicted Ternary Molybdenum Chalcogenophosphides\nfor High-Efficiency Hydrogen Evolution Catalysis

The\nsearch for efficient and cost-effective alternatives to platinum-based\ncatalysts for the alkaline hydrogen evolution reaction (HER) remains\na formidable challenge, driving the need for innovative materials.\nIn this study, we employed machine learning-driven moment tensor potentials\nin conjunction with particle swarm optimization to predict a new family\nof ternary molybdenum chalcogenophosphides, specifically Mo<sub>2</sub>SP and Mo<sub>3</sub>SP. Our calculations show that these materials\nexhibit robust thermodynamic, kinetic, and thermal stability at room\ntemperature, as indicated by their low formation energies, phonon\ndispersion curves free of imaginary frequencies, and molecular dynamics\nsimulations. Advanced HSE06 calculations further confirm their metallic\nnature in the bulk phase. We also investigated their size-dependent\nnucleation behavior, finding that Mo<sub>2</sub>SP and Mo<sub>3</sub>SP possess distinct crystallization pathways, with Mo<sub>2</sub>SP showing a lower nucleation barrier. Notably, the (100) facet of\nMo<sub>3</sub>SP displays optimal hydrogen adsorption energies and\nlower water dissociation barriers compared to the (110) facet of Mo<sub>2</sub>SP, highlighting its superior catalytic efficiency in the\nVolmer step of HER. These findings provide a crucial foundation for\nthe development of high-performance, low-cost catalysts based on ternary\nmolybdenum chalcogenophosphides, offering a promising alternative\nfor sustainable hydrogen production.