Transition State Resolved Pathways for Surface Catalyzed Chemistry

We propose the first open, large-scale dataset of transition states for heterogeneous catalysis, enabling kinetics-aware machine learning on surfaces. It would contain 100,000 elementary pathways as 3D reactant–TS–product triplets with energies, structures, and other metadata. Reactions span >10 metal/oxide systems, >1,000 adsorbates, and diverse mechanisms. Data will be generated by a reproducible pipeline: charge-balanced enumeration around Open Catalyst adsorbates, growing string method refinement, and Hessian validation.By supplying TS geometries absent from existing gas-phase and equilibrium-focused surface datasets, the resource enables modern generative models to infer pathways in seconds, sharply reducing screening costs and enabling “reaction-in-the-loop” catalyst design.