From global optimization to transition state search: an automated workflow for surface reaction barriers

Abstract

Reaction barrier calculations present the major bottleneck in the systematic exploration of surface reaction networks in heterogeneous catalysis via atomistic simulations. Each transition-state search introduces a high-dimensional configuration space of initial and final-state combinations that must be explored to identify the minimum-energy pathway with the lowest activation barrier. Typically, this task is delegated to human chemical intuition as user input to, e.g., nudged elastic band (NEB) calculations, since its automation remains challenging. Here, we introduce NEBscape, a fully automated workflow for transition-state searches that requires no human intervention. Initial and final states are generated via global optimization and systematically combined through atom mapping, symmetry alignment, and heuristic selection to yield a tractable set of NEB calculations. We benchmark NEBscape against diverse reactions and model surfaces from the OC20NEB database, where it identifies, on average, lower transition-state energies and produces minimum-energy pathways with high chemical fidelity to the target reactions.