Enhancing HER catalyst screening of modified MXenes through DFT and machine learning integration

Abstract

MXenes doped with non-metallic and transition metal elements exhibit remarkable potential as catalysts in the hydrogen energy. Nonetheless, efficiently identifying viable materials from a vast array of candidates remains a formidable challenge. Here, we conducted density functional theory (DFT) calculations to obtain the hydrogen adsorption free energy ($∆G_{\mathrm{H}}$) of 78 types of doped TiVCO₂ MXene catalysts. Then we employed machine learning models to categorize the $∆G_{\mathrm{H}}$ values of the 78 catalysts, resulted in an accurate model which only uses 7 readily available elemental features but has an impressive accuracy of 93.6%. Our model successfully predicting 5 TiVCO₂ catalysts doped with S with superior performance, subsequently validated through DFT calculations. This classification methodology not only evaluates the range of $∆G_{\mathrm{H}}$ effectively but also facilitates qualitative prediction and screening of catalysts, presenting a novel approach for catalytic systems with limited available data.