The ABINIT Users Mailing List ( CLOSED )

[Masayoshi Mikami <mmikami@rc.m-kagaku.co.jp>]

Dear Yu,

We have little information about your crystal model,
so my reply may be rather general ... In principle,
the VCA should be applied for chemically similar ions,
typically, Sr and Ba etc. I heard that someone VCA applied
to "Zn and Mg" for "Zn_1-x Mg_x O" --  and it failed --.
It is quite understandable; Zn (with d shallow core) and
Mg (without d) are not similar at all.
Thus, we should not use the VCA for heterovalent ions in general.

Yet, I come to think that the VCA for heterovalent ions might work
for some limited cases. One **weird** example ;-) is,
        http://dx.doi.org/10.1002/pssa.200669547
Here I used ABINIT. Kindly read the above paper for details.
In this case, an artificial cation (Z=13.5, ionic charge=+3.5
as the simple average of Al3+ and Si4+) was treated in the VCA
implemented in the ABINIT.
However, I should reiterate that this kind of "luck" may not be
expected for any kinds of heterovalent solid solutions, in general.
One should start from some supercells with assumptions of
the (positional) configuration of heterovalent ions. (I did them first.)

About the VCA implementation, the following paper may be referred:
        Ph. Ghosez et al., AIP Conf. Proc. 535, 102 (2000)
(You can find papers on typical VCA implementations by Google/
 Google Scholar). There are several ways of averaging.
ABINIT gives just pseudopotential-averaging.
Profl Andrew Rappe's Group has other implementations,
i.e. eigenvalue/eigenfunction-averaging...
(if I remember correctly... please check their papers on PRB.)

In ABINIT,
VCA Examples can be found in Test_v3/t90-t95 of v.4.6.5,
or, tests/v3/Input/t90-t95 of v.5.2.3.  You might want to
see the "deliberately unphysical example" in v3/t95;
the reference output is archived in the ABINIT src.tar.gz ball.
you will see "how the values are averaged."

Hope this helps & Good luck (indeed),
Masayoshi

On 2006/10/26, at 0:14, Chol-Jun Yu wrote:

> Dear ABINIT users,
>
> recently I am interested in solid solutions. I decided to use the  
> VCA method.
> So I generated the pseudopotential of the virtual atom, composed of  
> the
> component atoms, using fhi98PP software. Then, when I use it in  
> order to
> calculate the property of the solid solution in ABINIT, there is a  
> problem.
>
> Is it possible if the atomic number is a decimal fraction?
> At this time, how can be estimated the Ewald energy, atomic mass,  
> number of
> bands in ABINIT?
>
> Best regards,
>
> Yu.
> --
>
> Computational Materials Engineering(CME)
> Center for Computational Engineering Science(CCES)
> Institute of Mineral Engineering(GHI)
> RWTH Aachen
> Mauerstrasse 5
> D-52064 Aachen, Germany
> Tel: ++49 241 80 94969
> Fax: ++49 241 80 92271
> e-mail: yucj@ghi.rwth-aachen.de
> homepage: http://www.ghi.rwth-aachen.de/www/