[gpaw-users] problem of electronic convergence

Ask Hjorth Larsen asklarsen at gmail.com
Thu Jan 17 16:14:43 CET 2013


Hi

This looks a bit nasty.  Try to converge it first with LCAO/dzp (also
dzp for Bi/F/N) and grid spacing 0.24 or so.

The energy is crazy in iteration 1.  If you can converge or
semiconverge an LCAO calculation, and then switch mode, then it might
contribute to solving the problem as the energy at that point should
be non-crazy.

When you have some parameters that work with LCAO you can return to fd mode.

Is the energy also crazy if you use the partial waves only?

(Crazy energy can be caused by populating the wrong states in
initialization, but I'm not sure exactly what happens.)

Regards
Ask

2013/1/17 Vladislav Ivanistsev <olunet at gmail.com>:
> yes, here is the output
>
>
> On Thu, Jan 17, 2013 at 2:35 PM, Ask Hjorth Larsen <asklarsen at gmail.com>
> wrote:
>>
>> Dear Vladislav
>>
>> Can you send a text output file showing it not converging?
>>
>> Regards
>> Ask
>>
>> 2013/1/17 Vladislav Ivanistsev <olunet at gmail.com>:
>> > Dear Ask,
>> > a while ago you have "magically" resolved a problem of electronic
>> > convergence in BN or Graphene slab. I'm also having a problem with
>> > convergence for several systems. One is attached. Could you please take
>> > a
>> > look and provide some explanation on how to converge similar
>> > calculations.
>> > I'm using GPAW 0.9
>> >
>> > Many thanks,
>> > Vladislav
>
>


More information about the gpaw-users mailing list