[molpro-user] DF-DFT-SAPT fails with PBE0 hybrid functional

[Andreas Hesselmann]

Dear Leonid,

it is true that in JCP 122 (2005) 014103
we also show results using the PBE0AC
functional combined with the ALDA kernel
for the response calculations. These
are displayed in e.g. table III with the
acronym PBE0AC/ALDA. In this table you can
see, however, also two columns denoted
as DeltaPBE0AC and DeltaLPBE0AC, the former
using a hybrid-exact-exchange-ALDA kernel instead,
and the latter a local exact exchange potential
in the monomer calculations. You can see in both
cases that this has a significant effect on the dispersion 
interaction energy which is underestimated with
the PBE0AC/ALDA combination (simply due to the larger 
HOMO-LUMO gap with hybrid funcs). This is why this has 
been switched off now in DF-DFT-SAPT calculations with 
Molpro. 

If you insist on using it, you can use
the option NLEXFAC=0.0 in the program
so that Molpro thinks your input orbitals
originate from a standard GGA functional.

Best wishes,
Andreas





Am Mon, 2 Jun 2014 14:19:12 +0200
schrieb Leonid Shirkov <leonid.shirkov at gmail.com>:

> Dear Professor Hesselmann,
> 
> thank you very much for your answer.
> 
> I hope you will not mind me quoting your email here, since it may be
> interesting for other people.
> 
> "in DF-DFT-SAPT the xc kernel is implemented only with local xc
> functionals, so if you want to use the PBE0 functional in conjunction
> with DF-SAPT you have to use a local exact exchange potential. An
> example is provided in the examples directory of Molpro.
> As regards the fitlevel, as far as I know it is just for testing
> purposes and it should usually be set to the largest number."
> 
> My misunderstanding might stem from the notations used in the
> above-mentioned article (Phys. Chem. Chem. Phys., 2007, 9, 1680–1687).
> The manual gives an example for a functional called "lpbe0"
> while in the article they use both "pbe0" and "lpbe0" with
> DF-DFT-SAPT. Would it possible to get a sample input for the
> functional named "pbe0" in the article?
> The manual gives a sample for "pbe0" only for DFT-SAPT with no DF.
> 
> I have experienced another problem with DFT-SAPT (no DF). In order not
> to multiply the threads on the forum, I am attaching the output with
> the error
> (tgz archive file). It appears when I try to use local xc functional
> withoug DF for some dimers for some SAPT corrections in some basis
> sets. It just gives "NaN"
> value for some geometries, though for other it seems to work well.  I
> tried to use denser thresholds, but the results were the same.
> 
> 
> Best regards,
> Leonid
> 
> On Wed, May 21, 2014 at 5:02 PM, Andreas Hesselmann
> <andreas.hesselmann at fau.de> wrote:
> > Dear Leonid,
> >
> > this is related to our e-mail exchange from the summer time
> > last year, so check out what I wrote you then.
> >
> > Best wishes,
> > Andreas
> >
> >
> >
> > Am Mon, 19 May 2014 01:38:30 -0700
> > schrieb Leonid Shirkov <leonid.shirkov at gmail.com>:
> >
> >> The paragraph 35.5 of the manual says:
> >>
> >> "Currently only the ALDA xc-kernel is implemented for the case SAPT
> >> LEVEL=3 and SAPT FITLEVEL=3.
> >> This means that a corresponding SAPT calculation would be
> >> uncompatible with hybrid-DFT
> >> monomer orbitals/orbital energies..."
> >>
> >> It may be the answer to my question. However, it does not explain
> >> how the authors of the above mentioned article
> >> managed to use PBE0 with DF.
> >>
> >> _______________________________________________
> >> Molpro-user mailing list
> >> Molpro-user at molpro.net
> >> http://www.molpro.net/mailman/listinfo/molpro-user
> >
> >
> >
> > --
> > ---------------------------------------------------
> > Andreas Hesselmann
> > Lehrstuhl für Physikalische und Theoretische Chemie
> > Universität Erlangen
> > Egerlandstraße 3
> > 91058 Erlangen / Germany
> > Phone:  +49 9131/85-25021
> > E-Mail: andreas.hesselmann at chemie.uni-erlangen.de
> > --------------------------------------------------
> >
> >
> > _______________________________________________
> > Molpro-user mailing list
> > Molpro-user at molpro.net
> > http://www.molpro.net/mailman/listinfo/molpro-user



-- 
---------------------------------------------------
Andreas Hesselmann
Lehrstuhl für Physikalische und Theoretische Chemie
Universität Erlangen
Egerlandstraße 3
91058 Erlangen / Germany
Phone:  +49 9131/85-25021
E-Mail: andreas.hesselmann at chemie.uni-erlangen.de
--------------------------------------------------