C-H bond dissociation enthalpy prediction with machine learning reinforced semi-empirical quantum mechanical calculations

We introduce a combined fast semi-empirical quantum mechanical and machine learning (SQM/ML) approach capable of matching the C-H bond dissociation enthalpies (BDEs) computed with the highly accurate (RO)CBS-QB3 method. The usefulness of our proposed SQM/ML model is corroborated by the fact that a single C-H BDE of a molecule is calculated in seconds and the mean absolute error amounts to only 1 to 2 kcal/mol. Graphical Abstract The combination of fast semi-empirical quantum mechanical with machine learning (SQM/ML) approach is proposed to reproduce the C-H bond dissociation enthalpies (BDEs) computed with the highly accurate (RO)CBS-QB3 method in seconds. Our SQM/ML models turn out to be able to compute a single C-H BDE of a molecule with an mean absolute error of 1-2 kcal/mol.