A symmetry-informed active-configuration search tool for equilibrium surface coverage prediction in materials discovery

Materials discovery can now screen heterogeneous catalysts at unprecedented scale. However, high-throughput atomistic workflows typically evaluate catalytic performance on configurations that do not correspond to equilibrium surface states at the operating temperature and pressure of interest. Determining the condition-aware coverage for a given adsorbate and catalytic surface is a challenging computational problem. First, the number of possible arrangements over the available adsorption sites grows combinatorially. Second, identifying the most stable configurations commonly demands expensive density functional theory (DFT) relaxations and energy evaluations. Here we present a symmetry-informed active-configuration search (ACS) tool for efficiently identifying equilibrium adsorbate coverages. We introduce exact symmetry reduction and accelerate configurational search and relaxation with a machine-learned interatomic potential (MLIP). Hydrogen adsorption on Pt(100) is used as a benchmark application. Symmetry-informed ACS reduces the candidate pool from 262,144 to 4,172 surface states and selects energetically competitive configurations for ionic relaxations. Replacing DFT with MLIP-based calculations provides a 10,000-fold speedup per relaxation without compromising the reliability of saturation-limit predictions. Overall, the tool reproduces exactly the established DFT saturation limit and yields further insights into equilibrium surface coverage as a function of temperature and pressure. More broadly, our results show how leveraging configurational symmetry and efficient deployment of MLIPs can overcome the combinatorial scaling bottlenecks that limit brute-force materials modelling. This provides a foundation for evaluating catalytic energetics on coverage-consistent surface models in materials discovery.