[Wien] band gap problem

[Peter Blaha]

> > I'm doing calculations on Mg2Sn, a semiconductor with a small indirect
> band
> > gap of about 0.3 eV. The calculations converge well, but the band gap is
> > underestimated significantly and valence and conduction band already
> overlap
> > slightly and I get a semimetal structure (attached).
> > I$B!G(Bm not sure how this influences the results (most important the
> dispersion
> > relation of the conduction band). How bad is it and is there a chance to
> > correct this somehow? Did I do something stupid?
>
> Most likely you observe a well-known failure of LDA/GGA : underestimation of
> band gaps for semiconductors by 50% or more. I'm afraid there is little you
> can do about it within wien2k. There are methods that go beyond LDA/GGA
> (e.g. GW method) and that correct for this. They are much more
> computationally demanding, however, and they are not (yet ?) available in
> wien2k.

A "cheap" solution is to use the GGA of Engel-Vosko. It is a "hidden"
option (put 25 instead of 13 in case.in0). This GGA gives much better
bandgaps, but it gives a terrible total energy (Don't use it for
volume optimization,...)

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