[Wien] Converging Gmax

[Laurence Marks]

One thing to add to what Peter said. If you are not also changing the
sampling in case.in0, primarily the FFT sizes, the results you are
getting may not be fully consistent and this could be giving some
artifacts.

On Mon, Oct 26, 2009 at 10:53 AM, Peter Blaha
<pblaha at theochem.tuwien.ac.at> wrote:
> A larger GMAX improves both, the representation of the Coulomb- and the
> XC-potential (energy)
> in the interstitial. We know that for H-containing systems large GMAX can be
> essential.
>
> For most other compounds, a GMAX increase is not necessary (although it
> should never be "wrong"),
> because the Coulomb part is usually well described, while for GGAs the
> XC-part might not.
> For those reasons we introduced an IFFT-FACTOR in lapw0, which only affects
> this term, but not
> the Coulomb part. Usually it is enough to converge the XC-part with this
> factor.
>
> ad 1) A large GMAX is always better.
> ad 2) I don't know.
> ad 3) Never use "dstart again". Of course one can continue with a larger
> GMAX (or FFT-factor)
> using the previously converged density.
>
> David Tompsett schrieb:
>>
>> Dear All,
>>
>> I have a query about converging energies with respect to Gmax. I am
>> working on a d-metal metallic system.
>>
>> I have been attempting to converge the difference in total energy between
>> a non-magnetic and antiferromagnetic spin configuration. The calculation is
>> done with the GGA using RKmax=9 and RMT's of 2.1 and 8000Kpt in the full
>> Brillouin zone.
>>
>> With these parameters (which I tested for convergence) I have then tried
>> to converge the energy difference with respect to Gmax and found some
>> behaviour that I can not understand. All calculations were converged with
>> -ec 0.00001.
>>
>> For the non-magnetic calculation the energies go as:
>> Gmax        Energy
>> 12            -28765.981815
>> 14            -28765.984603
>> 16            -28765.983740
>> 18            -28765.983085
>> 20            -28765.982506
>> 22            -28765.982092
>> 24            -28765.981842
>> 26            -28765.981561
>>
>> So this indicates large changes of order of 0.01Ry as Gmax is changed.
>> This seems like an extraordinarily high Gmax for convergence
>>
>> In contrast for the antiferromagnetic calculation we find little variation
>> as we change Gmax:
>> Gmax         Energy
>> 12            -28765.982610
>> 14            -28765.982595
>> 18            -28765.982645
>> 26            -28765.982645
>>
>>  From this I have a few questions:
>> 1) Why would a non-magnetic calculation take such a larger Gmax to
>> converge? Is Gmax=26 reasonable?
>>
>> 2) Why might the convergence of the AFM calculation be much faster in
>> Gmax?
>>
>> 3) When converging Gmax is it acceptable to reuse the charge density
>> (*.clm*) files from the previous calculation? ie. should I run dstart again
>> before each new value of Gmax?
>>
>> Thank you for any help,
>> David.
>>
>
> --
> -----------------------------------------
> Peter Blaha
> Inst. Materials Chemistry, TU Vienna
> Getreidemarkt 9, A-1060 Vienna, Austria
> Tel: +43-1-5880115671
> Fax: +43-1-5880115698
> email: pblaha at theochem.tuwien.ac.at
> -----------------------------------------
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>



-- 
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60208, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
Web: www.numis.northwestern.edu
Chair, Commission on Electron Crystallography of IUCR
www.numis.northwestern.edu/
Electron crystallography is the branch of science that uses electron
scattering and imaging to study the structure of matter.