Improving the B3lyp Bond Energies by Using the X1 Method

Jianming Wu,Xin Xu
DOI: https://doi.org/10.1063/1.2998231
2008-01-01
Abstract:Recently, we proposed the X1 method which combines density functional theory method (B3LYP) with a neural network correction for an accurate yet efficient prediction of heats of formation [J. M. Wu and X. Xu, J. Chem. Phys. 127, 214105 (2007)]. In the present work, we examine the X1 performance to calculate bond energies. We use 32 radicals and 115 molecules to set up 142 bond dissociation reactions. For the total of 147 heats of formations and 142 bond energies, B3LYP leads to mean absolute deviations of 4.54 and 6.26 kcal/mol, respectively, while X1 reduces the corresponding errors to 1.41 and 2.45 kcal/mol.
What problem does this paper attempt to address?