CCL:G: DFT and dispersion

 Sent to CCL by: "Dr. Lars Goerigk" [lars.goerigk a chem.usyd.edu.au]
 Dear Arne,
I guess it depends entirely on your system to decide whether a D correction with M06-2X has an impact on the geometries or not. As I do not know the system, I cannot say how much of the intramolecular dispersion effects are covered by M06-2X. I also do not have much experience with M06-2X-D3 geometry optimizations. However, I assume that a system of more than 50 atoms should show significant intramolecular dispersion effects, that of course also influence the relative energies between conformers.
Personally, I would always include a dispersion correction to be on the safe side. Maybe you could run a test calculation on one of the systems with M06-2X and another dipsersion-corrected functional (TPSS should be in general faster, of course). Then you can maybe see the effect on your system and can make a final decision.
But maybe other users, who are reading this, have already done such a comparison and can give you also some recommendations. 
 Cheers,
 Lars
 On 28/10/2011 4:15 PM, Arne Dieckmann adieckma(-)googlemail.com wrote:
 
 Sent to CCL by: Arne Dieckmann [adieckma(-)googlemail.com]
 Dear Lars,
 thanks for your immediate reply. I know about the differences between the
 Minnesota functionals and the DFT-D3 method. This is why I wanted to augment
 their mid-range dispersion with the long range dispersion your former group
 developed. Do you know if TPSS with D-corrections produces significantly
 different geometries than M06-2X without D-corrections? How much do geometries
 of molecules (>= 50 atoms) change in general? I am just wondering if it will
 be sufficient to optimize with M06-2X and apply the single point D-corrections.
 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?
 Thanks again,
 Arne
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 Dr. Arne Dieckmann
 Houk Research Lab
 University of California, Los Angeles
 email: adieckma+*+googlemail.com
 On Oct 27, 2011, at 5:28 PM, Dr. Lars Goerigk lars.goerigk{}chem.usyd.edu.au
 wrote:
 
 Sent to CCL by: "Dr. Lars Goerigk" [lars.goerigk(~)chem.usyd.edu.au]
 Dear Arne,
 M06-2X was fitted to noncovalently bound systems in their equilibrium geometries
 and due to that fit it somehow incorporates medium-range dispersion
 interactions. However, as also discussed in the literature, it is expected that
 the asymptotic long-range behavior is still not correct and that it can also
 benefit from dispersion corrections.
 On the issue DFT-D3 in combination with the Minnesota functionals you can refer
 to PCCP 2011, 13, 6670. There, the parameters for DFT-D3 for those functionals
 were published and discussed.
 DFT-D2 parameters for the Minnesota functionals were published by the Martin
 group in JPCA 2008, 112, 12868. The s6 had to be smaller than unity, though,
 which still does not give the correct asymptotics, opposed to DFT-D3, where it
 was fixed to unity.
 If you have the Gaussian code available you could try to implement DFT-D3 (based
 on the code Stefan Grimme made available on his website).
 If you do not have the code, you would have to use DFT-D2. Please note that in
 G09 Rev A, DFT-D2 only works with B97-D and omega-B97X-D.
 That bug was fixed however and in Rev. B01 and higher you can now also use
 DFT-D2 for those functionals, which are mentioned in Grimme's DFT-D2 paper from
 2006 (i.e. also for B3LYP, TPSS, PBE, BP86).
 You have to use the IOp(3/124=3) then. In combination with IOp33(3=3) you also
 get a printout to check, if the correct s6 value was applied.
 Other functionals are not supported and I do not know if it is possible to read
 in an s6 specified by the user (maybe you should contact the Gaussian support to
 find that out). IOp(3/124=1)  adds the dispersion correction to any functional,
 but only with an s6=1, which is too large for the Minnesota functionals.
 If you can't use M06-2X, you could try TPSS in combination with D-corrections.
 Based on publications and my own experience, this gives very good structures.
 Another possibility would be to get ORCA, which is for free. I am not 100% sure
 if DFT-D3 is already implemented; have a look at the ORCA website for that.
 You could also check if other programs have already implemented DFT-D3.
 I hope these information could help you.
 With best wishes,
 Lars
 --
 Dr. Lars Goerigk
 School of Chemistry
 The University of Sydney
 Sydney, New South Wales 2006
 Australiahttp://www.ccl.net/chemistry/sub_unsub.shtml>;
 
 
 --
 Dr. Lars Goerigk
 School of Chemistry
 The University of Sydney
 Sydney, New South Wales 2006
 Australia