[Wien] ghostband

[Laurence Marks]

A much longer answer than you may have expected.

Isolated transition metal atoms (for certain the 1st row, perhaps
other rows as well) pose special problems in DFT with simple
functionals such as PBE. The functional does a bad job for the
d-electrons, and can end up making them (incorrectly) degenerate with
the s electrons (and maybe p's as well). In addition, depending upon
exactly what cell you use the degeneracy of the orbitals may not match
the "real" isolated atom configuration. (A simpler example of this is
the O atom which should be calculated in a tetragonal cell, not a
cubic one, in order to stabilize the correct spin states.)

What can easily happen is that at the start of an iteration, you may
have (for instance) just d-electrons occupied but the predicted states
(after lapw1) are s-electrons; or the other way around. You then get
big oscillations in what states are occupied. Since the -in1new switch
uses the states from the previous cycle these may be inappropriate for
the current cycle leading to QTL-B problems.

While the new mixer handles many things well, this is one case where
it can have problem. (Fixing this using a trust-region approach is
planned "sometime".) You may be able to fudge around the problem by
reducing the mixing parameter to (perhaps) 0.025 and ignore the
warnings that the mixer gives you. You can also use TEMP with a
reasonably large value (perhaps 0.01 Ryd). Probably best is to use the
PRATT mode in the mixer (not BROYD, this will be the worst) with a
conservative value of something like 0.05.

Getting an adequate number takes skill -- this is not simple and there
is no easy fix. Be aware that in the end your energy (and states) for
the isolated atom may be rather inaccurate (that's life).

On Sat, Dec 6, 2008 at 8:44 AM, shekoufe khosravi
<shekoufe.khosravi at gmail.com> wrote:
> Dear all
> I am trying to perform a calculation for isolated Co atom, but even using
> -in1new option, I face with "ghost band" error. To overcome this problem I
> used energy parameters same as similar calculation for Fe atom for which
> there is no problem, but this trick also didn't change any thing. It should
> be mentioned that this "Qtl-B" error appears after about 10 cycles. the
> initial in1 file was attached.
> Any suggestion is appreciated.
>
> _______________________________________________
> 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 to study the structure of matter.