The ABINIT Users Mailing List ( CLOSED )

["Anglade Pierre-Matthieu" <anglade@gmail.com>]

Hi,

Please, have a look at my most recent posts to find a suitable method
to make your calculation converge.
When you double the size of your system, supposing the susceptibility
to be constant with respect to frequency, you get a 4 time more
difficult SCF problem to solve. That's the reasons large cells usually
require better SCF algorithm (as describe in some of my previous
posts).  You do not need a larger number of k-points. Because they
sample reciprocal space (that gets smaller as your cell gets bigger)
you need fewer in fact. You don't need to change occopt either. It is
not necessary to have so many bands and Ecut is not to be increased.

Best regards

PMA

On Thu, Jul 17, 2008 at 3:52 PM, brunusrib <brunusrib@netcabo.pt> wrote:
>
> Hi!
>
> I have made calculations in a single cell of a cubic phase of BaFeO3 and my
> calculations always converged using PAW or Troullier-Martins
> pseudo-potencials. I'm now doing the same calculation but using a supercell
> ( with 2 x BaFeO3  with one lattice parameter doubled) and I do obtain an
> oscillation  in the total energy (of several Ha) and the calculation  do not
> converge. Does anybody know what could be the problem? I've already tried to
> change occopt (from occopt 3 to occopt 2), I've checked the number of the
> bands that I am using in the calculations, I am using more than the double
> of the bands that are needed to accommodate  all the electrons, and tried
> iscf 5 and 3.
> I've also tested  Ecut  up to 45 Ha and ngkpt  up to  11 x 11 x 11.
>
> Thank you in advance
>
> Bruno Ribeiro



-- 
Pierre-Matthieu Anglade