Machine learning-accelerated density functional theory optimization of PtPd-based high-entropy alloys for hydrogen evolution catalysis

High-entropy alloys (HEAs) have emerged as promising catalysts for the hydrogen evolution reaction (HER) due to their compositional diversity and synergistic effects. In this study, machine learning-accelerated density functional theory (DFT) calculations were employed to assess the catalytic performance of PtPd-based HEAs with the formula PtPdXYZ (X, Y, Z = Fe, Co, Ni, Cu, Ru, Rh, Ag, Au). Among 56 screened HEA(111) surfaces, PtPdRuCoNi(111) was identified as the most promising, with adsorption energies (E_ads) between −0.50 and −0.60 eV and high d-band center of −1.85 eV, indicating enhanced activity. This surface showed the hydrogen adsorption free energy (ΔG_H*) of −0.03 eV for hydrogen adsorption, outperforming Pt(111) by achieving a better balance between adsorption and desorption. Machine learning models, particularly extreme gradient boosting regression (XGBR), significantly reduced computational costs while maintaining high accuracy (root-mean-square error, RMSE = 0.128 eV). These results demonstrate the potential of HEAs for efficient and sustainable hydrogen production.