The ABINIT Users Mailing List ( CLOSED )

[Fabien Bruneval <fabien.bruneval@cea.fr>]

Dear Manuel,

From my experience, the number of bands you have to use in the 
screening calculation and then in the sigma calculation is usually 
rather similar.
I would say that the number of bands you have to use in sigma is at 
least equal to the number of bands in screening.
In general, the number of bands needed depends strongly on the type of 
material and on the size of the cell.

Advertisement:
If you find that the convergence vs this parameter is too slow, you may 
try the new technique I coded in Abinit-5.6.x, using keywords:
gwcomp      1
gwencomp   1.0 or 2.0 Hartree
Details can be found in PRB 78 085125 (2008).

I hope this helps.

Fabien


Manuel Pérez Jigato wrote:
> Dear all
>
> there are two parameters in the GW band structure (GW1) which use the 
> same syntax,
> "nband", and appear in two places, screening and also self-energy.
>
> convergence has to be sought for both, number of bands for screening 
> and number of bands for self-energy calculation.
>
> My question is: do i have to consider any link/relation between the 
> two parameters
> when i carry out my tests, ie in case the first time it appears in 
> computation of the quasiparticle energy, which is screening, if you 
> have determined already that
> your nbands has to be really high (500?), should i expect that my 
> tests for nbands in self-energy be totally independent and can use a 
> fixed small value for
> the other one (50-100) in order to do fast tests, or, else,..
> i should expect that even for a test to be meaningful, the first nband 
> (screening)
> has to be set to some high number (400-500)?
>
> is this known?
>
> thanks
>
> Manuel Perez Jigato
> KU Leuven
>
> Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm