Machine learning assisted screening of non-metal doped MXenes catalysts for hydrogen evolution reaction

Abstract

Heteroatom doping is a promising strategy to enhance the hydrogen evolution reaction (HER) performance of MXenes. The combination of machine learning (ML) techniques and high-throughput density functional theory (DFT) calculations offers an efficient approach for screening and designing HER electrocatalysts. In this study, we systematically investigated the impact of non-metal (NM) single-atom doping on the HER activity of V₂C MXenes with different surface functional groups (O and S). Our results reveal how the NM dopants influence the electronic structure, particularly the pz orbital electron redistribution, which subsequently affects the Gibbs free energy of hydrogen adsorption (ΔG_H∗). Additionally, a universal descriptor, integrating both electronic and structural properties, was developed using ML to predict ΔG_H∗ and successfully captures the HER catalytic activity trends for a variety of NM dopants in V₂CO₂ and V₂CS₂. Notably, the descriptor can also be extended to doped V₂CSe₂ and V₂CTe₂ for HER catalysis. Among the doped MXenes, P–V₂CTe₂ outperforms platinum (Pt) in terms of ΔG_H∗, demonstrating exceptional potential for practical HER applications. Our study provides a comprehensive framework for the efficient exploration and design of high-performance MXene-based HER catalysts.