[gpaw-users] LCAO calculations on metallic slabs

[Ulrik Grønbjerg]

Hi Ask and Lasse,
thanks for your help. Most of the modifications were actually stuff that I
usually use in my finite-difference calculations, which I had removed in
order to find the cause of the problem. So it seems that the major issue was
that the width of my Fermi distribution was too small.

However, my calculations actually converged to the weird energies, so the
SCF cycle did not rectify it in this case. This also caused the BFGS
algorithm to squeeze the atoms very close together.

Again, thank you very much for the help - you'll hear from me again if I
cannot get your fix working for me :)

Best regards,
Ulrik

2011/10/3 Ask Hjorth Larsen <askhl at fysik.dtu.dk>

> Hi
>
>
> On Fri, 30 Sep 2011, Ulrik Grønbjerg wrote:
>
>  Dear fellow GPAW users,
>> I would like to do calculations on slab containing Pt and Ca, but I
>> encounter two, probably related, problems. First of all, the energies are
>> obviously wrong, with total energies of +100 eV for the slab, in contrast
>> to
>> -350 eV that I get from regular GPAw calculations. The other problem is
>> that
>> the LCAO calculation will often fail in finding the Fermi level at some
>> point. I have attached the script setting up the system, and the script I
>> use to run the LCAO calculations.
>>
>> I am not especially familiar with LCAO calculations, so I might very well
>> be
>> missing something basic, and any help will be greatly appreciated.
>>
>> Best regards,
>> Ulrik Grønbjerg Andersen
>>
>
> I think the crazy energies in the beginning have to do with the ordering of
> states defined in the PAW setup being different from that defined among the
> basis sets, resulting in bad initial occupations which however should be
> rectified during the SCF cycle.  This is something I'll probably have time
> to fix in November.  It happens whenever the same angular momentum channel
> is present more than once on a given PAW setup (so also for the standard
> setups for K, Ti, ...).
>
> And thanks, Lasse, for figuring out a solution.
>
> The FermiDirac occupations object also has a maxiter parameter which can be
> raised, although that should not be necessary under normal circumstances.
>
> Regards
> Ask
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20111003/d8a0f055/attachment.html