Contributed by +Jan Jensen
Steven Bachrach recently highlighted Kruse and Grimme's geometric counterpoise scheme (gCP) - an add on term that corrects for BSSE in the same spirit as Grimme's dispersion correction. One of the many interesting observations in this paper is that gCP- and dispersion-corrected HF minimal basis set calculations can yield very accurate binding energies.
The paper by Sure and Grimme investigates this particular issue further and presents the so-called HF-3c method. "3c" stands for the 3 different corrections: gCP, dispersion, and a short-range basis incompleteness (SRB) correction for "systematically overestimated bond lengths for electronegative elements (e.g. N,O,F) when employing small basis sets." HF-3c uses a minimal basis set for H, C, N, and O and a split valence basis for heavier elements. In all only 9 empirically adjusted parameters are needed.
Being an HF/minimal basis set calculation the method can be applied to rather large molecules, but will be slower than corresponding semi-empirical calculations on a single core. However, the HF-3c can of course be run in parallel. HF-3c interaction energies computed for the S22, S66, and X40 datasets are of similar accuracy to PM6-DH2 and significantly better than PM6.
HF-3c geometry optimizations and vibrational frequency calculations are found to significantly more numerically stable and offer more accurate structures (including protein structures) than PM6, as implemented in MOPAC2012. HF-3c thus offers an attractive alternative when MOPAC2012 is giving imaginary frequencies. If HF-3c can produces accurate transition state geometries, the method might also be a viable alternative to QM/MM calculations.
HF-3c is implemented in the ORCA package, which is available free of charge.
This work is licensed under a Creative Commons Attribution 3.0 Unported License.