[Wien] calculation quality

[Peter Blaha]

No, the atomic number is not really a good parameter.
The needed value of RKmax depends much more on the type of
valence electrons and an "old" rule said:
for s,p wavefunctions you need RKmax 6-7
     d                                7-8
     f                                8-9

So a "heavier" element, which has only s,p electrons (eg. S, Se, Te) needs
a much samller RKmax than a "light" Fe atom.

To some extend we try to "correct" this automatically in a compound and the
setup make a sphere of sp-atoms smaller than the sphere of 3d or even 4f elements.

Of course, within a row in the periodic table, your arguments would be correct,
i.e. a O or F atom needs a larger RKmax than Be or B. Thus for a CO molecule
I'd make the C sphere smaller than the O-sphere (this is NOT done yet by setrmt).

H, which has only s-electrons, is even more special and a much smaller RKmax is enough.

k-points: sure, the size of the unit cell matters and for one and the same material
the k-mesh scales with volume (eg. when creating a supercell).
But other properties may be even more important. For certain magnetic metallic
materials you may need 100-10000 times more k-points than for an insulator,
although the volume might be identical.

There is NO other choice than: start with a realistic (but small) k-mesh and test
the result of the desired properties with respect to k-mesh.

What "realistic, but small" means, needs a bit experience, but there were many
discussions about that in the mailing-list or in WIEN2k-workshops.

Pablo de la Mora schrieb:
> Dear Wien-users,
> There are two parameters that meassure the quality of a calculation; RxK
> and the number of k-points.
> 
> With respect to RxK; for hydrogen R is quite small but it also has one 
> nuclear charge, so a small RxK parameter is acceptable.
> So my question is: would a more complex parameter be more meaningful?
> Such as
>         RxK/n
> where n is the atomic number?
> The idea behind is that the nuclear potential increases with n, so a
> larger R is needed and R/n would be more or less constant.
> 
> With respecto to the number of k-points (Nk) with the size of the cell 
> the number of k-points should be reduced to maintain the same quality, 
> so this parameter could be modified to VxNk.
> 
>        Pablo de la Mora
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------