[Pw_forum] filelph

[Paolo Giannozzi]

On Mar 5, 2007, at 13:49 , Amit Kumar wrote:

>  It implies  that we can't see the individual electron-phonon
>  matrix elements in new version.

no, it implies that results are written in a different way

>   I ran an identical program of electron phonon
>   calculation  using la2F flag in new the version as well as
>   using the old method (scf nscf without la2F).
>   Surprisingly, I'm getting different results  still using
>   the identical input parameters for both case.

in some earlier versions, a different scheme for broadening was
used in the calculation of the 'double-delta' integrals. Presently
a simple gaussian is used (ngauss=0), while Methfessel-Paxton
broadening of order one (ngauss=1) was used instead. I don't
see any good reason to use the latter instead of the former,
since all we need is a representation of the delta. This may lead
to some differences.

>   is it possible to run the el-ph part in parallel cluster????

you can run the phonon code in parallel. The el-ph part is
not parallelised but it doesn't cost much by itself

>   Is it possible to restart the jobs if it get killed in the mid  
> way???

you can, but it is a mess

>   Is it possible to calculate different reprensentations or
> phonon modes separetly?

in principle you can, but the code does not implement it

P.
---
Paolo Giannozzi, Democritos and University of Udine, Italy