CCL: DFT and dispersion

["Stefan Grimme" <grimme%thch.uni-bonn.de>]

 Sent to CCL by: "Stefan  Grimme" [grimme _ thch.uni-bonn.de]
 >By the way, do you know how well DFT-D3 works with implicit solvation? I
 guess in this case the >parameters would have to be solvent    -dependent,
 right?
 One important comment on this point: DFT-D3 (and similarly other dispersion
 corrections) are constructed to yield accurate isolated molecule (gas phase)
 energies and geometries. These can be combined with any kind of solvation model
 and when this model accurately accounts for solvent-solute dispersion,
 everything is consistent. What could be done in principle as suggested (to
 modify the intramolecular dispersion to implicitly account for solvent-solute
 dispersion) is theoretically not justified in my opinion.
 DFT-D3 has been developed to reproduce CCSD(T)/CBS as closely as possible and
 should not be used as an empirical tool to balance inter- and intra-molecular
 effects.
 A further note in this context: this balance of intramolecular and
 intermolecular dispersion is also the reason why in some cases
 dispersion-UNcorrected DFT seemingly provides better results compared to
 experimental solution data than physically correct methods. But when one wants
 to get the right result for the right reason I would prefer to first treat the
 gas phase system as accurately as possible and then to add the best available
 condensed phase corrections. And this holds for DFT as well as WFT methods.
 Best wishes
 Stefan Grimme
 grimme-,-thch.uni-bonn.de