[Pw_forum] problem with convergence during SCF-procedure

[Alexej Mazheika]

On 10/23/08, Paolo Giannozzi <giannozz at democritos.it> wrote:
>
> Alexej Mazheika wrote:
>
> > Yes, really. I use the version 4.0.1, there are such options in this
> > version.
>
>
> well, no. More exactly, there are, but they are fake. Try
>
>    diagonalization='whatever funny algorithm you can think of'
>
> and the code will work. Of course it will not use whatever funny
> algorithm you inserted there, but will use the default, i.e.
> Davidson diagonalization (with parallel subspace diagonalization).
> Only the following keywords do something different from the default:
> 'cg', 'cg-serial', 'cg+serial', 'david-serial', 'david+serial'.
> The documentation is rather unclear on this point.
> Anyway: you used Davidson, and that's fine.
>
> Paolo
>
> --
>
> Paolo Giannozzi, Democritos and University of Udine, Italy
> _______________________________________________


I use the program PWGUI for the building of my input files. There is the
keyword 'diis' in PWGUI, and so I thought that
such algorithm is present in Espresso-4.0.1 too. During the
calculations it does not write anything about 'unrecognized word' both
in output file and in Linux-console, so I really considered that the
'diis' algorithm is embedded in Espresso. Then it means that the
changing of occupation
and mixing data helped to solve the problem with covergance. That's fine.

Alex Mazheika, Research Institute for phys. and chem. problems, Minsk,
Belarus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20081025/7e640ffa/attachment.html>