recentering option for isolated system calculations

Axel akoh... at gmail.com
Wed Mar 5 13:03:10 UTC 2008



On Mar 5, 4:13 am, Juerg Hutter <hut... at pci.uzh.ch> wrote:
> Hi

hi juerg,

> the wavelet solver uses its own FFT routine. These
> routines are the same as the FFTSG package in CP2K.
> Yes, this is duplicating code, but it came pre-packed
> by the developers of the wavelet solver.
> The FFT library you choose in the input affects only
> the FFTs in the regular CP2K code, not the wavelet solver.

that is what i figured. what i don't quite understand is that
the documentation for the wavelet solver says that one
_has_ to use FFTSG for the rest of cp2k as well.

but coming back to the initial question, do you have
any comments on implementing a recentering of coordinates option?

cheers,
    axel.

>
> best regards
>
> Juerg
>
> ----------------------------------------------------------
> Juerg Hutter                   Phone : ++41 44 635 4491
> Physical Chemistry Institute   FAX   : ++41 44 635 6838
> University of Zurich           E-mail: hut... at pci.uzh.ch
> Winterthurerstrasse 190
> CH-8057 Zurich, Switzerland
> ----------------------------------------------------------
>
> On Tue, 4 Mar 2008, Axel wrote:
>
> > On Mar 4, 5:29 pm, Rad <rad.... at arl.army.mil> wrote:
> >> Axel,
>
> >> I am struggling with a large isolated system and I am using the MT
> >> poisson solver. Maybe I should try the wavelet based PS. Could you
> >> please let me know how to set up the section and if there are any
>
> > syntax and parameters is in the cp2k input reference manual.
>
> > BTW: it would be nice to have an explanation for the requirement
> > of FFTSG as preferred FFT, it seemed to work with FFTW3 as well and
> > i would have expected the code to crash or refuse the run, if
> > this was a "hard" requirement.
>
> >> caveats? (Teo mentioned that and I forgot the syntax) Do we have to
> >> have
> >> a vaccum of 20% for the wavelet method or is it only for the MT
> >> scheme?
>
> > i would consider 20% vacuum (how do you compute it?) to be far too
> > little for MT p-solver. you may get away with it if your system has
> > favorable error cancellation, but i would not depend on it.
>
> > formally, the MT p-solver requires a box of twice the size of a
> > sphere that encloses the total charge density of your molecule
> > (i.e. atoms _and_ electron density). i can see small errors crop
> > up on my (weakly interacting) test system starting from exactly
> > the %50 mark and then the errors became intolerably large at
> > having 40% vacuum.
>
> > cheers,
> >    axel.
>
> >> Thanks
> >> Rad


More information about the CP2K-user mailing list