[Wien] mBJ + SOC :: spin-orbit splitting reduction

[Martin Gmitra]

Dear Wien2k developers,

We are observing underestimating trend of spin-orbit coupling
splittings of Gamma_{15} bands in zinc-blende semiconductors.

For instance in GaAs, the Gamma_{15v} splits to Gamma_{8v} and
Gamma_{7v} by the values of about 0.29 eV, while experiment gives 0.34
eV. PBE calculations gives 0.32 eV. Similar one sees for Gamma_{15c}
which splits due to spin-orbit to Gamma_{8c} and Gamma_{7c} where mBJ
gives 0.156 eV and PBE 0.181 eV. This one can compare to 0.174 eV from
GW+SOC calculations [PRL 96, 086405].

In PRB 82, 205212 you present two implementations of mBJ, within
Wien2k and VASP. Could you please provide information whether both the
implementations provide the same spin-orbit splittings of Gamma_{15}
states?

In case they are the same, which part of the Becke-Johnson approach
can be responsible for such a reduction?

Best regards,
Martin Gmitra
Uni Regensburg