Dataset for Geometric vs Global Knowledge in GNNs: A Controlled Leakage Study for Early–Late TS Classification of the Isomerization of the 2-Norbornyl Carbocation.

all_logs_structured.h5 is a structured HDF5 dataset containing quantum chemical reaction data extracted from intrinsic reaction coordinate (IRC) calculations. Each entry corresponds to a single chemical reaction and contains the optimized reactant, transition state (TS), and product geometries together with atomic forces, bond-distance matrices, IRC energy profiles, and reaction progress descriptors. The dataset was developed to support machine learning applications for transition-state classification, reaction mechanism analysis, and graph neural network (GNN) model development. Each reaction is stored as an individual record identified by a unique reaction identifier (e.g., TST0001). Field Description reaction Unique reaction identifier. dist_R_TS RMSD (Å) between the optimized reactant and transition-state geometries. dist_P_TS RMSD (Å) between the optimized product and transition-state geometries. reaction_fraction Reaction progress descriptor computed from the reactant and product distances to the TS. Values close to 0 correspond to early transition states, whereas values approaching 1 indicate late transition states. early_or_late Binary transition-state label (0 = Early TS, 1 = Late TS). reactant_geom Cartesian coordinates of the optimized reactant geometry stored as (atomic number, x, y, z). ts_geom Cartesian coordinates of the optimized transition-state geometry stored as (atomic number, x, y, z). product_geom Cartesian coordinates of the optimized product geometry stored as (atomic number, x, y, z). reactant_forces Atomic forces on the optimized reactant geometry (Hartree/Bohr). ts_forces Atomic forces on the optimized transition-state geometry (Hartree/Bohr). product_forces Atomic forces on the optimized product geometry (Hartree/Bohr). D_R Bond-distance (or connectivity) matrix for the reactant. D_TS Bond-distance (or connectivity) matrix for the transition state. D_P Bond-distance (or connectivity) matrix for the product. irc_energies Electronic energies along the intrinsic reaction coordinate (IRC) relative to the transition state. irc_rxcoords IRC reaction coordinates corresponding to the energy profile. Geometry Format Geometry arrays have the form [Atomic Number, x, y, z] Force Arrays Force arrays contain the Cartesian force acting on every atom Fx Fy Fz with one row per atom. Distance Matrices D_R, D_TS, and D_P store the pairwise interatomic distance matrices (Å) for the reactant, transition state, and product, respectively. These matrices can be used to reconstruct molecular graphs, identify bond-breaking and bond-forming events, and generate graph-based machine learning representations. IRC Data The IRC information consists of reaction coordinate values (irc_rxcoords) corresponding electronic energies (irc_energies) These arrays describe the minimum-energy pathway connecting reactants and products through the transition state.