[gpaw-users] [gpaw-developers] problem of electronic convergence in BN or Graphene slab

[Ask Hjorth Larsen]

Hi

2013/1/14 Chengjun Jin <chej at fysik.dtu.dk>:
> Hi everyone,
>
> I have a structure composed of  a few layers of Graphene and a fews layer of BN and also bulk Cu. structure are attached.
>
> I try to relax the structure, however the electronic step doesn't seem to converge after 100 electronic steps.  the total energy is jumping quite a lot. output are attached.
>
> a couple of things I have tried.
> fd vs lcao
> lda vs gga
> occupation temperature
> mixing
> different Poisson solvers
> nbands
>
> None of these seems to improve the convergence.
>
> After a quick search on the mailing list, this kind of layer structure seems to be problematic to get converge.  However, I did a structure relaxation with ATK(LDA, lcao) and it get converged.
>
> So I am wondering if there is any sensitive parameter in GPAW to tune this convergence.
>
> Regards,
> Chengjun

Here's a script with one set of parameters such that it does converge:

  http://www.student.dtu.dk/~ashj/opendir/chengjun-CuBCN.py

The most important difference is that the mixing is very conservative.
 It would seem that something is not handled correctly in the
initialization, as the energy on the first iteration is crazy.

Please note that this belongs on the -users list, not the -developers.
 All developers are also users.  They subscribe to both.

Regards
Ask