how to get the "total" electron density

[Matthias]

Hi Axel,

On Mar 13, 8:21 pm, Axel <akoh... at gmail.com> wrote:
> hi!
>
> thanks for the answers. this is very encouraging.
>
> On Mar 13, 6:12 am, Fawzi Mohamed <fa... at gmx.ch> wrote:
>
> > On Mar 13, 2008, at 9:23 AM, Matthias wrote:
>
> > > Hi Axel,
>
> > > indeed, the all-electron density is available with GAPW, but only in
> > > Fourier space as it is needed for the calculation of the X-ray
> > > diffraction spectrum. In principle one could perform a back
> > > transformation to real space using FFT and print it as a cube file.
>
> ok. can you give me a few hints on which code specifically
> to look at, so that i can try hacking it together?

OK, I will unitize the X-ray routine in such a way that one can
request the build of the all-electron density in Fourier space
independently.

>
> > > However, for numerical reasons you would apply in this way a low pass
> > > filter to the all-electron density and thus you would not obtain the
> > > true all-electron density in real space.
>
> > actually I think that this is the correct thing to do, we should
> > maybe do something like this also instead of the stride (or at least
> > do some mean).
>
> yep. visualisation doesn't really need much accuracy, but analysis
> does and preserving the integral of the property would be very helpful
> in the particularl analysis that i have in mind.
>
> an alternate option would be to incorporate the analysis directly
> into cp2k (i'm currently thinking of the code fromhttp://theory.cm.utexas.edu/bader/which looks like it
> could be merged with rather moderate effort) and thus at least
> avoid the issue of having to read and write huge formatted files.

A long time ago there was already a request to apply the Bader
analysis for wavefunctions obtained with CP2K/QS. Inter alia for this
reason CP2K provides the optional printout of the orbitals in
Cartesian representation. However, that is a pretty inefficient way to
interface two codes as you correctly noted, since one has to deal with
huge formatted files and there are also pitfalls due to the code
specific ordering of the atomic orbitals .. in a word: awful. It would
be better to pack the Bader code into one module with a clean and
simple interface, if this is possible and allowed (licence?).

cheers,

Matthias

> cheers,
>    axel.
>
> > direct collocation of the gaussians (and also stride as now done) do
> > not conserve the total density when the cutoff is not high enough
> > (basically always with GAPW).
>
> > ciao
> > Fawzi
>
> > > cheers,
>
> > > Matthias
>
> > > On Mar 13, 2:30 am, Axel <akoh... at gmail.com> wrote:
> > >> hi everybody,
>
> > >> would there be a way to get the total _electron_ density from
>
> > >> quickstep GPW/GAPW calculations?
>
> > >> i actually found a flag to output a 'total density', but that turned
> > >> out to be the valence electron
> > >> density plus the (smeared out?) core charge density, with the odd
> > >> side
> > >> effect that the cores were
> > >> negative and the electrons positive. i guess it does not really
> > >> matter, but...
>
> > >> for all-electron GAPW calculations, this information should - in
> > >> principle - be available.
> > >> the question is, would it be much effort to put it into a cube file?
>
> > >> for pseudopotential based calculations, be it GPW or GAPW one could
> > >> just use the
> > >> density that was "removed" during pseudopotential generation.
>
> > >> any comments or suggestions would be highly appreciated.
>
> > >> cheers,
> > >>    axel.