[Wien] 2-window semi core calculation of Ge3d fails

[Peter Blaha]

> My calculations run fine now. It is a little bumpy ride with the 
> semi-core treatment of the 3d hole but at least they are converging.
> I'm a little confused though, according to the instructions:
>> edit case.in2s and specify NE=9 (just for the Ge 3d with hole)
>> edit case.in2 with NE="all electrons+1"; but set EMIN such that the 3d-states of the Ge-hole atoms are cut-off.
> I get the impression that there are 9 (10-1) electrons in the shifted Ge 
> 3d levels and 'all electrons+1' in the other valence states. I 
> (mis)understood this to be 9 electrons too much. Why is it not 
> 'all-electrons+1-9' in case.in2?

"EF" is calculated with ALL states listed in case.energy, i.e. it includes 5 states (10 electrons) form
Ge-d/atom and you want to "add" one electron to the conduction band.
Using the EMIN value, you later on cutoff all Ge-d states and get the density without Ge-d but including
1 electron in the conduction band.

> 
> I also tried to put the 3d electron in the case.inm file instead of in 
> case.in2, doing this gave a warning about 'charged cell'. Can I ignore this?

Provided you put the proper background charge, you can ignore that.

The "ideal" calculation would put the extra electron into a Ge-p or f-state exactly on
the site where you have the Ge-d corehole, because this is what "happens" in
nature (dipol selection rule). In a bandstructure code you cannot do that so
easily, so you must compromise. If the first conduction band states are Ge-states
(which I would expect),
I definitely would put the electron into "case.in2"; if the lowest conduction band
states have a very different character (eg. localized TM-3d orbitals), putting them into
"mixer" might be more appropriate.

> 
> Finally I have a small general question about the Kmax cut off. In the 
> literature I see pseudopotential based calculations using around 20Ry 
> cut-off. In Wien2k calculations I usually only reach around 10-15Ry. Is 
> the pseudopotential based cut-off and the cut-off in Wien2k comparable 
> in some way?

the "APW+lo" basis set is a much better basis set than planewave-pseudopotentials,
thus you need usually a smaller PW-cutoff (unless you have very small spheres
so that the "augmentation" does not help that much).
Of course you have to pay a price: the APW+lo matrixelements are very complicated
(compared to PW-PP) and our basisfunctions are not kept fixed, but change (adapt)
within the scf-cycle, so certain optimizations of PW-PP codes are not possible.
-- 

                                       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/
--------------------------------------------------------------------------