BorylXAT-DB: an IRC-validated transition-state database for machine-learning prediction of Lewis base-boryl radical C-Cl XAT barriers

BorylXAT-DB is an IRC-validated quantum-chemical transition-state database for Lewis base-coordinated boryl-radical mediated C-Cl halogen-atom transfer (XAT). It was constructed from a reaction space comprising 55 boryl radicals, 386 Lewis bases, and 179 chlorinated substrates. Thermodynamic filtering, automated transition-state searches, frequency analysis, and bidirectional IRC validation produced 9,237 validated transition states with complete reactant-transition-state-product mappings. The release includes two complementary data products: BorylXAT-DB.db, an ASE SQLite database for atomistic inspection and reaction-path reconstruction, and BorylXAT-DB.parquet, a flattened Parquet table for statistical analysis and machine-learning workflows. The dataset contains 50,057 ASE structures, including boryl radicals, Lewis bases, chlorinated substrates, reactant complexes, product complexes, carbon-radical products, and IRC-validated transition states. Each transition-state entry stores component identifiers, reactant and product mapping keys, activation free energy, reaction free energy, imaginary-frequency information, and IRC endpoint data. The release also includes metadata, checksum information, a schema summary, a minimal Python requirements file, and an example loading script.