[CP2K:2040] Re: optimisation problem of F compound

Jörg Saßmannshausen jorg.sassm... at strath.ac.uk
Thu Apr 23 21:25:05 UTC 2009


Dear Fawzi,

thanks for shedding some light into it (which is easy if it is sunny :-) :-)

I am playing around with that problem now for some time. Previously I used 
BLYP (on a larger system but with the same compound in it) and I got a 
imaginary frequency which was again movement of the F atom.
So I cut down the system and followed Axel's suggestion to use PBE. As I could 
not find a PBE optimised basis set for F, I tried the two I mentioned. With 
both I had the imaginary frequencies.

I agree with the forces, I think that was what my gutt tried to tell me. I got 
the feeling the cluster will be very busy tonight ;-)

Greetings to Berlin and thanks!

Jörg

On Donnerstag 23 April 2009 Fawzi Mohamed wrote:
> On 23-apr-09, at 22:35, Jörg Saßmannshausen wrote:
> > Dear Fawzi,
> >
> > I actually did:
> >
> > cutoff energy
> > 400  -183.5957955549
> > 450  -183.5944280905
> > 500  -183.5937679158
> > 550  -183.5935371009
> > 600  -183.5938000945
> >
> > one of the things I learned from Rachel at the workshop :-)
> >
> > For me it looked like that 500 seems to be ok, so I decided to go
> > for that
> > value.
> > However, these values were for the DZVP-ALLELECTRON basis set. I did
> > not run a
> > test for the aug-DZVP-GTH, simply assuming as I did not change the
> > potential
> > only the basis set the difference would be not that big?
>
> normally it is a good assumption to assume that the basis set does not
> influence much the cutoff, but switching from GAPW (allelectron) to
> GPW does change things, and a lot, gapw does not put the whole density
> on the grid, so for it the cutoff is not critical.
>
> > Or was that a wrong
> > assumption? What cutoff would you recommend?
>
> I haven't calculated with F, but from the high exponent in the basis
> (10.7079832847), and also for p functions yo might need more I would
> try 800.
> Think that you need accurate forces, on *everything* and if you have
> just one F looking just at the total energy might be misleading, if
> you want to be sure do some steps of MD with HF or something like
> that, and look at the energy conservation...
>
> > Thanks for the prompt reply!
>
> you are welcome.
>
> > All the best from a wet Glasgow!
>
> and greetings from sunny Berlin :)
>
> Fawzi
>
> > Jörg
> >
> > On Donnerstag 23 April 2009 Fawzi Mohamed wrote:
> >> I would say that the cutoff of 500 might be too small for the F
> >> element. Have you tested the cutoff needed for it?
> >>
> >> Fawzi
> >>
> >> On 23-apr-09, at 18:12, Jörg Saßmannshausen wrote:
> >>> Dear all,
> >>>
> >>> I am struggling to optimise a molecular structure with a F atom in
> >>> it.
> >>> Whereas I manage to get the optimisation converged, the frequency
> >>> calculation clearly indicates I am not at the local minimum.
> >>> Replacing
> >>> the F atom in the molecule with H (and leave everything else the
> >>> same)
> >>> results in a converged structure. I got the gutt-feeling that the
> >>> problem is the basis set / pseudopotential I am using for F. Would
> >>> somebody mind to give me a helping hand here?
> >>>
> >>> I have enclosed the following files:
> >>> - mol file of the frequency calculation (so you see what I mean)
> >>> - force_eval section
> >>> - subsys section
> >>> - fluor-geo.inp
> >>> - fluor-freq.inp
> >>>
> >>> I had no problems with the molecule using a different program, so
> >>> I am
> >>> quite confident about the structure. Note: it is a cation, I left
> >>> the
> >>> anion out for computational reasons.
> >>>
> >>> Thanks for your help!
> >>>
> >>> Jörg
>
> 


-- 
*************************************************************
Jörg Saßmannshausen
Research Fellow
University of Strathclyde
Department of Pure and Applied Chemistry
295 Cathedral St.
Glasgow
G1 1XL

email: jorg.sassm... at strath.ac.uk
web: http://sassy.formativ.net

Please avoid sending me Word or PowerPoint attachments.
See http://www.gnu.org/philosophy/no-word-attachments.html



More information about the CP2K-user mailing list