[Wien] force difference between LAPW and PAW method

[Peter Blaha]

If the disordered compound goes metallic ??? it could be a problem. The 
default WIEN2k is using TETRA, while VASP uses a large broadening.

Again, forget the 400 eV of the PAW calculation. It has NOTHING to do 
with our RKmax and our basis sets are usually more efficient.

In any case, RKMAX=8.5 is "more than overconverged", since your smallest 
sphere is probably Oxygen. I'd start such a calculation with RKMAX=5 and 
later check with 6 and/or 7.

Another possible problem: SiO2 might have small spheres for Si leading 
to quite some core leakage ??? Either use .lcore or put the Si-2p as 
semicore !


On 01/29/2016 12:04 PM, Pavel Ondračka wrote:
> Peter Blaha píše v Pá 29. 01. 2016 v 11:43 +0100:
>> This looks fairly large.
>>
>> I do NOT understand your statement of using "the same RKMAX" ???
>> PAW does not have RKMAX and their KMAX has NOTHING to do with ours.
>> What was your RKMAX and what are youre sphere sizes ???
>
> Sorry, the right word should probably be "equivalent RKMAX". As
> mentioned in the original email, the RKMAX was calculated from cutoff
> in EV using the scheme from userguide (sqrt(cutoff in Ryd) * size of
> smallest sphere). Specific values differed between cells, smallest
> spheres were around 1.5 and the RKMAX was around 8.0 (PAW cutoff of
> 400eV).
>>
>> k-mesh: not only the grid is important ! As far as I know, VASP
>> usually
>> uses a "huge" temperature broadening. What FERMI-method are you using
>> ?
>
> default
>
>> spin-polarization ??
>
> no
>
>> Is the PAW done with VASP and a fairly new release (accurate PAW
>> potentials ?)
>
> It was VASP and it was one of the latest versions. Will check for
> specifics.
>
>>
>> On 01/29/2016 10:50 AM, Pavel Ondračka wrote:
>>> Dear Wien2k mailing list,
>>>
>>> I've been doing some calculation on a moderately large (around 100
>>> atoms) amorphous-like SixTi1-xO2 cells produced by simulated
>>> annealing
>>> MD in a PAW software. According to my colleague who did the final
>>> force
>>> relaxation, the residual forces on atoms should be around 0.5
>>> mRyd/Bohr. However when I calculate forces in Wien2k I see forces
>>> of
>>> order of magnitude larger, eg. some as large as 30mRyd/Bohr.
>>>
>>> This are exactly the same calculations: both using PBE, same k-
>>> grid,
>>> same RKMAX (calculated from cutoff using the formula from UG) and
>>> I've
>>> spent a lot of time checking the force convergence with respect to
>>> all
>>> possible parameters.
>>>
>>> We did some tests with small cells (eg. 10 atoms) and there we can
>>> get
>>> a consistent results with max difference of forces around 10-20%
>>> which
>>> I find reasonable. I'm actually quite clueless about it. I've been
>>> suspecting user error at the beginning however I haven't found any
>>> so
>>> far. My current theory is that in this large cases without symmetry
>>> maybe the small differences between methods somehow sum up and
>>> hence
>>> the big difference in final forces, however I would like to hear
>>> your
>>> opinion in this matter.
>>>
>>> Best regards
>>> Pavel Ondračka
>>> _______________________________________________
>>> Wien mailing list
>>> Wien at zeus.theochem.tuwien.ac.at
>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>> SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.
>>> theochem.tuwien.ac.at/index.html
>>>
>>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300             FAX: +43-1-58801-165982
Email: blaha at theochem.tuwien.ac.at    WIEN2k: http://www.wien2k.at
WWW:   http://www.imc.tuwien.ac.at/staff/tc_group_e.php
--------------------------------------------------------------------------