Fig. 1: Integrating DFT calculations and MLIP model for surface and adsorption analysis.

a Depiction of the optimized iridium (Ir) metal surface (slab model) simulating iridium grown on palladium (Pd) cubic seeds (\({{\mbox{Fm}}}\bar{3}{\mbox{m}}\)) and (001) surface. This panel includes the calculated electrostatic potential across the slab used to determine the work function (\(\phi={V}_{{vac}}-{E}_{F}\)), where \({V}_{{vac}}\) is the vacuum level, \({E}_{F}\) is the calculated Fermi level, and \({V}_{{avg}}\) is the average electrostatic potential across the slab. b Adsorption of H₂ and OH* on metal surfaces, illustrating various molecule configurations and adsorption sites identified via the Delaunay triangulation algorithm. A pre-trained MLIP model (CHGNet) was fine-tuned with additional DFT calculations to estimate total energies and to rapidly screen configurations of adsorbates and adsorption sites. Lower energy structures were re-evaluated with precise DFT calculations to identify the most favorable adsorbate-adsorbent configurations. c Comparative analysis of the relative Fermi levels and adsorption energies of H₂ and OH* on various metal surfaces, all conforming to Pd (001) symmetry. d Summary of the features of various metals and their bimetallic alloys based on DFT-calculated electron-giving tendencies and H₂/OH* adsorption strengths. More negative Fermi levels and adsorption energies indicate stronger electron accepting and adsorption capabilities, respectively. Features of the superior metal in bimetallic catalysts were chosen to represent the combined properties. e Predicted HOR activity sequence derived from DFT calculations for single metals and bimetals, prioritizing electron acceptance followed by H₂ and OH* adsorption strength.