Title: Conformational hierarchies of weakly bonded systems: Accuracy of dispersion corrected DFT

Speaker: Alexandre Tkatchenko, Humboldt Fellow at the Fritz Haber Institute in Berlin, Germany

Date/Time: Monday, March 17, 2008, 4:30pm-5:30pm

Location: CSRI Building, Room 90 (Sandia NM)

Brief Abstract: It is well known that attractive van der Waals (dispersion) interactions, important for stabilization of, e.g., molecular crystals, biomolecules and physisorption, are badly described by state-of-the-art (semi)-local xc-functionals in DFT, but naturally included in post-Hartree-Fock (HF) methods or empirically in force field simulations.
We employ a semi-empirical C_6/R^6 correction for DFT total energies and obtain correction parameters for different DFT functionals by fitting to a database of high quality ab initio calculations.

In contrast to previous studies, our careful analysis reveals that non-empirical PBE and PBE0 functionals are by far the best choices for applying the dispersion correction when compared to empirical approaches such as revPBE, BLYP or B3LYP. We argue that such behavior is due to a better description of the short-range van der Waals interaction by PBE and PBE0 functionals.

By fitting the parameters of the C_6 correction close to the basis set limit, the reported values should be transferable to any implementation of DFT irrespective of the basis set employed for the description of the wavefunction.

We assess the accuracy and impact of the correction for conformational energy hierarchies of (i) formamide and N-methylacetamide dimers, (ii) Ala_2 and Ala_4, and (iii) infinite polyalanine conformers, comparing to published post-HF results for (i) and (ii).

Even though the relative energies are not changed for small organic molecule dimers and Ala_2 compared to DFT-GGA, the impact of dispersion on the conformational hierarchy of larger systems is substantial, dramatically influencing the relative isomer stability for both Ala_4 and infinite alanine helices and reaching up to 7.4 kcal/mol per residue for an infinite alpha-helix.

Alexandre Tkatchenko, Volker Blum, Matthias Scheffler

CSRI POC: Anatole von Lilienfeld-Toal, (505) 845-7442