All electron calculations

[marci]

Dear Eva

according to your input, you changed the values of some parameters
(EPSFIT, EPSISO, EPSRHO0) related to the construction of the local
densities
for all electron calculations.
Why?
The result is that the local densities are not accurately described.

As a general rule, if you do not know what a parameter is really
doing,
better to trust the default values that are typically reasonable in
cp2k.

With your input and the default EPS... parameters I get the following
Mulliken charges
     1  O   -0.75837
     2  H    0.379019
     3  H    0.379019

I would like to add that it is not true that you need a higher cutoff
for all-electron calculations with the GAPW method.
In normal cases 250 Ry are sufficient,  because the hard part of the
density is not "collocated" on the PW grid.
However, when printing the total density to a cube file (only then),
the whole density is "collocated" on the grid, and
this is most probably not very accurate in the regions close to the
nuclei.
However, the cube files are  often printed for visualization,
and high accuracy at the atomic centers is not required.

kind regards
Marcella




On Dec 13, 4:25 pm, Eva Pluharova <pluha... at gmail.com> wrote:
> Hi,
>
> I am trying to perform all electron calculation of water molecule with
> print out of all electron density using CP2K version 2.2.366. Input
> files see in attachment.
>
> I obtained following results that I wouldn't expect:
>
> 1. Total electronic charge (G-space):      -10.0349568912
>     Total electronic charge (R-space):      -10.0349568912
>
> Why it isn't exactly -10.000000? In other systems that I have tried
> the discrepancy is significantly higher, for example CH3CONHCH3
> -40.0000 vs. -41.758.
>
> 2. Mulliken population analysis gives the following results:
> # Atom  Element  Kind  Atomic population                Net charge
>       1     O        1           7.943086                  0.056914
>       2     H        2           1.028457                 -0.028457
>       3     H        2           1.028457                 -0.028457
>  # Total charge                 10.000000                  0.000000
>
> Which aren't in agreement with BLYP/6-31g* (identical basis set)
> calculation in Gaussian:
>      1  O   -0.742711
>      2  H    0.371355
>      3  H    0.371355
> that make more sense.
>
> 3. Total electronic density printed to cube file doesn't integrate to
> 10, but something orders of magnitude larger.
>
> Any suggestions? Thanks.
>
> Best regards,
> Eva
>
>  H2O.inp
> 2KViewDownload
>
>  pos.xyz
> < 1KViewDownload