cp2k and Bader analysis

[Axel Kohlmeyer]

On Sep 18, 9:14 am, Ondrej Marsalek <ondrej.... at gmail.com> wrote:
> On Fri, Sep 18, 2009 at 15:03, Axel Kohlmeyer <akoh... at gmail.com> wrote:

> OK, I thought it is safe to add the cores to their atoms (as they
> would be always localized in one of the partitions) to get the correct

it is a pretty good approximation, otherwise you could not
be using pseudopotentials in the first place.

the problem is to get the matching densities. the simple way
is to generate them and write them out when building the
pseudopotential itself, but the atomic code for GTH potentials
scared the hell out of me, and when doing it "empirically" you
have the problem that you cannot easily extract the radial
component of l!=0 orbitals as they have different orientations
in the all electron or pseudopotential calculations.

> total charge. But I have never needed AIM myself, so I have no
> practical experience with this.
>
> So doing Bader analysis on the valence density is no good, even with
> norm-conserving PP? I would only expect this to be a problem with USPP
> or the PAW transformation, but as I said, no experience in this.

it is not completely bogus to do AIM on valence densities, but
one has to be very, very careful. obviously the zero-flux
surfaces will be shifted, particularly if you have partitioning
between elements where a different number of electrons has been
put into the core. you can still extract trends and relative
differences, but absolute values will be bogus in a large number
of cases.

as for USPP, you actually do have the size of the augmentation charges
and thus you can place them on the grid. codes like quantum-espresso
do support this, and for GPAW it had been implemented into cp2k
even, but i don't know whether it still works.

cheers,
    axel.

>
>
>
> > matthias krack once implemented a feature that you can do this
> > from GPAW, but it turned out to be extremely slow and had artefacts
> > that threw the bader code from the group in texas off.
>
> > i have been toying with the idea of constructing density files
> > to add to the existing valence density after the fact, by taking
> > the difference from a single atoms all-electron and pseudopotential
> > calculation, and then adding them to the valence density cube.
> > but that scheme only worked for me to some degree for s-electrons.
>
> >> with orbitals, you would perhaps need to use some Cube file
> >> manipulation program to obtain the densities from individual KS
> >> obrbitals first.
>
> > why the individual KS orbitals? AIM only works on the total density,
> > IIRC.
>
> That's right, AFAIK. I was not sure wheter the wavefunction mentioned
> in the original post was not meant to be a KS wavefunction. Also, you
> might want to partition the spin density as well, which in ROKS is the
> density of the SOMO. I thought that perhaps someone has some use (that
> I am unaware of) for the partitioning of other orbitals as well.
>
> Ondrej
>
> > cheers,
> >    axel.
>
> >> Best regards from Prague,
> >> Ondrej
>
> >> On Fri, Sep 18, 2009 at 11:15, Jörg Saßmannshausen
>
> >> <jorg.sassm... at strath.ac.uk> wrote:
>
> >> > Dear all
>
> >> > I was wondering whether it is possible (at least in theory) to use cp2k to
> >> > generate the wavefunktion-file for Bader's AIM analysis.
> >> > So basically I would need an ASCII file, similar to the uploaded file (which
> >> > is the file for propene). Is that possible at all with cp2k or as we are
> >> > using plane waves here not possible?
>
> >> > The file location is:
> >> >http://groups.google.com/group/cp2k/web/bader-aim-file.wfn
>
> >> > All the best from Glasgow!
>
> >> > 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.
> >> > Seehttp://www.gnu.org/philosophy/no-word-attachments.html