[CPMD-list] Charges and population analysis.

[Juerg Hutter]

Hi

On Mon, 28 Jul 2003, Jacek Wojdel wrote:

> 	Dear CPMD users,
> I recently started using CPMD for optimization only (no MD) and I got
> stuck when it comes to properties calculations. At the moment I'm only
> interested in charges and population analysis. For charges, my
> input contains:
> &CPMD
>  PROPERTIES
>  RESTART COORDINATES VELOCITIES WAVEFUNCTION LATEST
>  LSD
> &END
> (...)
> &PROPERTIES
>  CHARGES
> &END
> The output contains almost reasonably looking ESP-fit charges, but
> Hirshfeld charges report just centers stripped of valence electrons:
>
>  ***********************************************************************
>       ATOM                 COORDINATES                      CHARGES
>                       X         Y         Z              INT       ESP
>     1   Co         0.0000    0.0000    0.0000          9.000     8.637
>     2   Fe         9.5620    0.0000    0.0000          8.000     1.878
>     3    C         3.3670    0.0000    0.0000          4.000    -1.664
> (...)
>     9    N         5.5639    0.0000    0.0000          5.000    -0.092

I was running a simple test case (water dimer) with USPP's and got
reasonable results for both, Hirschfeld and ESP charges.
There must be something special with your input I can't reproduce.

>
> I'm using Vanderbilt uspps, is this a source of the problem? Moreover,
> CPMD complains about the lack of electron velocities in restart file:
>
>  RV30| WARNING! NO WAVEFUNCTION VELOCITIES
>
> should I do a single step MD to get it right, and is it neccessary for
> calculating charges?

You don't need VELOCITIES for a property calculation. The warning can
be ignored.


>
> The other problem I have is when calculating Mulliken population, one
> needs to project the wavefunction, but the file ATWFN mentioned in the
> documentation is nowhere to be found. I tried to specify the basis
> functions on my own, but couldn't get it right, does someone have an
> example input file that I could learn from? (oh, with transition
> metals in an example please, I didn't succeed in extrapolating from
> the H2O input included in the documentation)

If you don't specify a basis, CPMD will use a default Slater type basis
with exponents from the Slater rules or where available from the
Clementi set.
If you want to specify your own basis the input strongly depends on
the type of basis functions you want to use.

The probably best way to do a Mulliken population analysis is to
use the pseudo-atomic wavefunctions generated from the pseudopotentials
in an atomic calculation. For USPP's you need to change Vanderbilt's
atomic code or to get the functions from one of the output/restart
files of that code.

regards

Juerg



> Any help will be appreciated,
> 	Jacek
> --
> +-------------------------------------+
> |from: J.C.Wojdel                     |
> |      J.C.Wojdel at tnw.tudelft.nl      |
> +-------------------------------------+
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://www.cpmd.org/mailman/listinfo/cpmd-list
>