The ABINIT Users Mailing List ( CLOSED )

[matthieu verstraete <Matthieu.Verstraete@ulg.ac.be>]

aha! Your problem may be a too large dilatmx. Try with 1.05 or 1.1 at the 
most:

I have had strange segfaults with too large values of dilatmx, because
the various G vector spheres are too different when you do the
relaxation (this is not a rigorous explanation - I don't know where
the bug is exactly). I have just remarked on this problem elsewhere on
the forum (forum.abinit.org)

On Tue, Feb 2, 2010 at 4:43 PM, Martin Haeufel
<martin.haeufel@googlemail.com> wrote:

> @Xavier, all ABINIT users: Question
>
> There's another thing I would like to point out:
>
> If you start with optcell 0 and then continue with optcell 2, you might
> end up with a different crystal symmetry.
Normally in all cases you can not break crystal symmetry: both the
stresses and the forces are symmetrized before being applied, and they
should not remove any symmetries. Optcell 2 should modify the cell in
an arbitrary way, but subject to a symmetrized strain which respects
all initial symmetries.

> Why is that? Well, in step 1 the structure symmetry might change,
> because it corresponds to energetically more favourable ionic positions.
> But in step 2 the symmetry is fixed to the input symmetry, which is now
> the output of step 1!
>
> Is that correct?
In step 1 as well the symmetry was imposed on the forces, so nothing
should be broken. This was enforced cleanly by Xavier only in 5.8 I
think.


> Let's think about the following situation: I have a crystal structure
> with correct symmetry, but I want perform (LDA+GW) calculations with a
> method that differs from the one that was used for the structural
> optimization (e.g. paw/uspp instead of norm-conserving pp).
more likely you start with PAW, then go to something more fancy that
is not implemented with PAW,
just importing the positions and lattice.

> Then the crystal structure should also be optimized with the new method.
your example kind of presumes that this is not possible. Otherwise
yes, you should be consistent between the methods used for the
relaxation and the rest.

> Am I right?
>
> What it is the suggested procedure in this case? So should one use
> optcell 2, and nothing else? Is there a way to use less cpu-seconds?
I would always use optcell 2 unless you want to fix a strain, eg for a
substrate effect.

If you have non zero forces (not just strains) you can use tolrff
instead of toldff, to reduce the number of scf cycles which will be
done...


Matthieu


> Martin Haeufel
> TU Munich, WSI (T33)
> www.wsi.tum.de
>
> Mohua Bhattacharya schrieb:
>> Hi Martin,
>> Thank you very much for your help. It worked. What confuses me that the
>> same input file with restartxf 1 and getwfk2 1 worked for the
>> optimization of a different structure with optcell 0 and optcell 1.
>> Thanks again.
>> Regards
>> Mohua
>



--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Prof. Matthieu Verstraete

Universite de Liège
Institut de Physique, Bat. B5, 3/7
Allée du 6 aout, 17
B- 4000 Sart Tilman, Liège
Belgium

Associate member European Theoretical Spectroscopy Facility (ETSF
http://www.etsf.eu)

Phone : +32 4 366 90 17
Fax   : +32 4 366 36 29

Mail : matthieu.verstraete@ulg.ac.be
         matthieu.jean.verstraete@gmail.com



-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Prof. Matthieu Verstraete

Universite de Liège
Institut de Physique, Bat. B5, 3/7
Allée du 6 aout, 17
B- 4000 Sart Tilman, Liège
Belgium

Associate member European Theoretical Spectroscopy Facility (ETSF
http://www.etsf.eu)

Phone : +32 4 366 90 17
Fax   : +32 4 366 36 29

Mail : matthieu.verstraete@ulg.ac.be
          matthieu.jean.verstraete@gmail.com