[Wien] Meta-GGA

[Peter Blaha]

> Has anyone tried the meta-GGA options at all? By a process of elimination,
> I've worked out that it can be run after a normal GGA by doing "x
> mixer_vresp" to generate the case.vrespup/dn files, after copying case.inm
> to case.inm_vresp as described in the UG. However, except for a shift in
> the absolute energy is does not look like it's changed much if anything.
> Am I missing some additional information/files that are needed?

All meta-GGA implementations have only the meta-GGA Exc
functional implemented, but use the PBE-GGA potential (Vxc). Unfortunately
the formulaes for Vxc are so complicated that nobody has derived dE/dn for
these meta-GGA functionals as a closed expression. Most people use
meta-GGAs "a posteriory" using GGA densities (as we do), but I know that
some calculate Vxc with some tricks directly in the Hamiltonian (this
would be nice to have, since I'm much more interested in Vxc than in
Exc....).

If you really want to use M-GGA, please note that it is numerically much
more instable (E-tot will oszillate much more than in GGA) and you should
use GMAX=24. In addition, the "best" M-GGA is no longer PKZB, but TPSS
(see recent PRL/PRB ? by Perdew). You will have to look into the code
(SRC_lapw0, grep for PKZB or TPSS). In any case, this is NOT for
beginners, but more for people testing and developing new functionals
and thus hardly documented.


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