The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <xavier.gonze@uclouvain.be>]

Dear Manuel,

On 15 Apr 2009, at 12:34, Manuel Pérez Jigato wrote:

> Dear all
>
> Regarding the extrapolar method, i would like to find out
>
> - Is the approximate method to determine the best value of gwencomp
> based on the completeness relation and plasma frequency, implemented  
> in abinit?

You can monitor the effect described in the paper using abinit, but it  
is not automatic. At present,
some people proceed as follow :
- take a reasonable value of gwencomp, like 2 Hartree, and plot a few  
(e.g. three) values of the
 eigenenergy (or gap) as a function of nband
- on the basis of the curvature of this property, either increase of  
lower the value of gwencomp (e.g. 1 Ha or 3 Ha)
Usually one finds that these two curves constitute a bound on the  
converged value (so, somehow, an error bar).
This is quite convenient ...

> - According to the paper, the suggested value is 2 Hartree.
> I guess that when the number of bands is very high, you need to  
> increase gwencomp?

No, gwencomp can be left unchanged, whatever the number of bands.
> - Is the extrapolar method always benefitial? ie, are there no  
> requirements/conditions for a particular system to require a smaller  
> number of bands (than the standard nband using the full continuum)?

No, until now, we have always seen it to be beneficial, for bulk  
solids, isolated molecules as well as nanowires.

> thanks a lot

Regards,
Xavier

> --
> Manuel Pérez Jigato
> KU Leuven