[Wien] 'NN' - overlapping spheres during the volume optimization

[Stefaan Cottenier]

> I am working on a Fe-Ge compound with about 17% vacancies on one site. I
> have set this site as a fully occupied and I am currently trying to
> optimize the volume and the atomic coordinates.
>
> For the initial volume optimization run, I have reduced RMTs by 5%, and
> as a result I got an the "optimized" volume smaller than the original. I
> believe this is expected due to the reduced RMTs.

No, this is not related to the choice of RMT. RMT is a mathematical 
construct only. Whatever RMT you choose (within reasonable limits), you 
should get the same observable properties (e.g. the volume, in your 
case) if you perform well-converged calculations.

First estimate what is the minimal volume that you will ever want to 
calculate for this compound, and then take the RMTs such that they'll 
not overlap for this minimal volume.

> However, in reality,
> it would be expected to get a slightly larger volume for this new
> "stochiometric" phase.

Which functional are you using, and how large is 'slightly'? If you use 
e.g. LDA, it will usually underestimate the volume by several percent, 
making the volume for the stochiometric compound perhaps smaller than 
the experimental volume of the off-stochiometric compound.

You might have a look at http://prb.aps.org/abstract/PRB/v79/i8/e085104 
to get an idea about which functional is likely to provide the optimal 
volume for the elements you are using.

Stefaan