The ABINIT Users Mailing List ( CLOSED )

[Ilija Dukovski <Ilija.Dukovski@MearsCorp.com>]

Great, thanks a lot.
I almost started writing a completely new routine on the anaddb level
and your answer saves me lots of time. 
Again, I am aware that I'm replicating previous post, I'm sorry for
that, but I think the gkks (or deformation potentials if you like) are
of interest to a larger semiconductors community. 
Can you please send me Xavier's contact.

Thanks a lot, 
Ilija Dukovski


On Fri, 2006-06-02 at 15:30 +0100, Matthieu Verstraete wrote: 
> Hi,
> 
> The gkk file format can be read out from the source code for mrggkk. The 
> simplest thing for you is to use anaddb, though. It already does the 
> reading in of everything, and can calculate the g_kn k'n'(q). You'll 
> probably need to intervene in the calls to 07ddb/normsq_gkq. The full 
> matrix elements are present, independently of the FS. It's only in the 
> routines called in 07ddb/normsq_gkq that the weights for "closeness" to 
> the FS are added. There is also an option for keeping the full band 
> dependency and not adding the weights, but I have not tested it at all. 
> Note that the quantity I play with in normsq_gkq is a kind of squared 
> matrix element:
> if
> 
> g_kn k'n' idir,iatom = <psi_k+q,n' | H'_q,idir,iatom | psi_k,n >
> 
> then the matrices contain
> 
> gamma_idir,iatom,idir',iatom' = g_kn k'n' idir,iatom g_kn k'n' idir',iatom'
> 
> which still depends on q, k, and n,n'
> 
> In normsq_gkq this is
> 1) either averaged over the FS bands or not
> 2) multiplied by the phonon displacement vector or not (then turned back 
> into a cartesian basis).
> 
> The bottom line is the same for everybody: my code is messy but commented. 
> Please also check the subroutine structure and so on - try to understand 
> what it does before asking me for details...
> 
> The ascii file is an incidental output, which should serve at some point 
> to compare with pwscf output. Another future project.
> 
> I have the impression I've been through this before on the forum. Please 
> search it before posting. There are also other people working on band 
> dependent stuff in the elphon, so try to avoid multiple implementations: 
> contact Xavier with your exact project.
> 
> Matthieu
> 
--