Authors: Marcel Swart, Miquel Solàb and F. Matthias Bickelhaupta
Abstract: Several recent studies (J. Phys. Chem. A 2004, 108, 5479; J. Comput. Chem. 2007, 28, 2431) have shown impressive results when replacing the non-empirical PBE density functional by the empirical OPBE or OLYP functionals, i.e. replacing the PBE exchange functional by Handy and Cohen’s OPTX functional. To investigate the origin of the improvements, we have placed constraints from the non-empirical PBE exchange functional on the empirical OPTX exchange functional, and tested the performance of the resulting constrained functionals for several characteristic chemical properties. The performance of the new functionals is tested for a number of standard benchmark tests, such as the atomization energies of the G2 set, accuracy of geometries for small molecules, atomic exchange energies, and proton affinities of anionic and neutral molecules. Furthermore, the new functionals are tested against a benchmark set of nucleophilic substitution SN2 reactions, for which we have recently compared DFT with high-level coupled cluster CCSD(T) data (J. Comput. Chem. 2007, 28, 1551). Our study makes clear that the performance depends critically on the number of constraints, and on the reference data to which the constrained functionals are optimized. For each of these properties studied, there is at least one functional that performs very well. Although a new promising functional (MLffOLYP) emerged from the set of constrained functionals that approaches coupled-cluster accuracy for geometries and performs very well for the energy profile of SN2 reactions, there is no one of the newly constructed functionals that performs equally well for all properties.
Open Access item.
Click below PDF icon for free download.
This is an Open Access item. Click above PDF icon for free download.