Hydrogen, Oxygen,\nand Lead Adsorbates on Al<sub>13</sub>Co<sub>4</sub>(100): Accurate\nPotential Energy Surfaces at Low Computational\nCost by Machine Learning and DFT-Based Data

Intermetallic compounds are promising materials in numerous\nfields,\nespecially those involving surface interactions, such as catalysis.\nA key factor to investigate their surface properties lies in adsorption\nenergy maps, typically built using first-principles approaches. However,\nexploring the adsorption energy landscapes of intermetallic compounds\ncan be cumbersome, usually requiring huge computational resources.\nIn this work, we propose an efficient method to predict adsorption\nenergies, based on a Machine Learning (ML) scheme fed by a few Density\nFunctional Theory (DFT) estimates performed on <i>n</i> sites\nselected through the Farthest Point Sampling (FPS) process. We detail\nits application on the Al<sub>13</sub>Co<sub>4</sub>(100) quasicrystalline\napproximant surface for several atomic adsorbates (H, O, and Pb).\nOn this specific example, our approach is shown to outperform both\nsimple interpolation strategies and the recent ML force field MACE\n[arXiv.2206.07697], especially when the number <i>n</i> is small, i.e.,\nbelow 36 sites. The ground-truth DFT adsorption energies are much\nmore correlated with the predicted FPS-ML estimates (Pearson R-factor\nof 0.71, 0.73, and 0.90 for H, O and Pb, respectively, when <i>n</i> = 36) than with interpolation-based or MACE-ML ones (Pearson\nR-factors of 0.43, 0.39, and 0.56 for H, O, and Pb, in the former\ncase and 0.22, 0.35, and 0.63 in the latter case). The unbiased root-mean-square\nerror (ubRMSE) is lower for FPS-ML than for interpolation-based and\nMACE-ML predictions (0.15, 0.17, and 0.17 eV, respectively, for hydrogen\nand 0.17, 0.25, and 0.22 eV for lead), except for oxygen (0.55, 0.47,\nand 0.46 eV) due to large surface relaxations in this case. We believe\nthat these findings and the corresponding methodology can be extended\nto a wide range of systems, which will motivate the discovery of novel\nfunctional materials.