How Well Can B3LYP Heats of Formation Be Improved by Dispersion Correction Models?

Yuwei Zhou,Jianming Wu,Xin Xu
DOI: https://doi.org/10.1007/s00214-015-1801-9
2016-01-01
Theoretical Chemistry Accounts
Abstract:In the present work, we have compiled six datasets for heats of formation (HOFs) of hydrocarbons of different chemistries, involving Set 1 for 21 n-alkanes up to n-C 32 H 66 , Set 2 for n-C 7 H 16 and its branched isomers, Set 3 for 36 polycyclic saturated hydrocarbons, Set 4 for 41 C 6 H 8 isomers of rings, alkenes, alkynes and cumulenes, Set 5 for 41 benzene-based compounds and Set 6 for 66 radicals. We have performed intensive high-level G4 calculations, which provide the reference data when the experimental data are not available or not reliable. Attention has been paid on how the dispersion-corrected methods, B3LYP-D2, B3LYP-D3 and B3LYP-D3BJ, can improve over the widely used density functional, B3LYP. It was found that, although the dispersion-corrected methods can eliminate the original B3LYP errors to a significant amount, they are unable to cure all B3LYP deficiencies besides the lack of van der Waals interactions. While D3 still underestimates the molecular stabilities as does B3LYP, D3BJ overestimates their stabilities. The more advanced D3 and D3BJ models do not necessarily outperform the simpler D2 model. Except for the PSH set, B3LYP-D3BJ always gives larger errors than does B3LYP-D2 for the other five datasets, showing room for further improvement.
What problem does this paper attempt to address?