Site-specific hydrogen adsorption on GaAs-supported dual-atom catalysts: first-principles insights and interpretable modeling

Understanding how local atomic environments influence hydrogen adsorption is important for the development of dual-atom catalysts (DACs). Here, we investigate the hydrogen adsorption thermodynamics of transition-metal pairs supported on two-dimensional GaAs using density functional theory (DFT) and machine learning. Of 100 initially constructed configurations, 83 were successfully optimized and evaluated for hydrogen adsorption at designated metal sites. Eight configurations exhibit near-thermoneutral hydrogen adsorption at the investigated TM' sites, with absolute Gibbs free energies below 0.01 eV. Electronic structure analyses reveal charge redistribution and orbital hybridization between the adsorbed hydrogen and metal centers. However, neither charge transfer nor the integrated crystal orbital Hamilton population alone exhibits a clear linear correlation with hydrogen adsorption free energy. To examine the combined effects of metal identity, local geometry, and electronic structure, we develop a two-stage regression model that combines an element-identity baseline with an XGBoost residual correction. The full-data fit yields an R² of 0.939 and a root mean square error of 0.075 eV. These results quantify associations between local adsorption-state characteristics and hydrogen adsorption thermodynamics in GaAs-supported dual-metal systems. The findings provide a basis for further investigation of site-specific hydrogen adsorption in dual-atom catalysts.