Answers are grounded in this paper and cite their sources.
Site-specific hydrogen abstraction from alcohols by OH radicals constitutes a chemically coherent but kinetically diverse reaction family in which rate coefficients depend on site location, local substitution, oxygen proximity, molecular topology, and reaction-path degeneracy. We assembled 42 site-specific abstraction channels from 11 alcohols, represented by 462 correlated rate coefficients evaluated on a common temperature grid. Molecule-level partitions were used throughout model development and evaluation. In the per-H formulation, reaction-path degeneracy was excluded from all predictors and applied only after prediction to recover the total channel rate. The Arrhenius-informed ElasticNet gives a test mean absolute error of 0.157 in log10 k(T), while XGBoost gives 0.081. Leave-one-alcohol-out analysis shows that site type contributes consistently across both models, while topology provides a reproducible contribution to XGBoost. Literature benchmarks using molecule-withheld 1-butanol and external 1-pentanol and 1-octanol further support transferability beyond the internal split. The resulting representation provides an interpretable description of site-specific alcohol + OH abstraction kinetics within the local environments represented by the present descriptor set.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: