LDA+U problem (band gap not consistent)

[superyoyo]

Dear All,

I wonder if anyone have found this behavior of LDA+U or GGA+U when apply to the oxide materials (e.g., TiO2, SrTiO3) that CBM composed of d states of metal. I found that the band gap, let say at gamma, is not quite equal when perform self-consistent calculation and non self-consistent (use CHGCAR from self-consistent; ICHARG = 11)

For example, SrTiO3 case

I use U = 4 applied to d states of Ti-d, the band gap from scf (just look at gamma of K7x7x7) is 2.65 eV while the band gap from non scf (make a band structure plot or just calculate at gamma) is 2.25 eV!!.

It is 0.4 eV difference. I tested LDA+U in another system such as ZnO (where d states of Zn is in VB) and do not find this problem.

Could anyone shed me some light if there is some physics behind and which value of band gap is making sense in the case of SrTiO3.

Thanks in advance!!

[forsdan]

Have you set LMAXMIX = 4 in the INCAR file for the SrTiO3, as stated in the manual?

"The CHGCAR file also contains only information up to LMAXMIX for the on-site PAW occupancy matrices. When the CHGCAR file is read and kept fixed in the course of the calculations (ICHARG=11), the results will be necessarily not identical to a selfconsistent run. The deviations can be (or actually are) large for L(S)DA+U calculations. For the calculation of band structures within the L(S)DA+U approach it is strictly required to increase LMAXMIX to 4 (d elements) and 6 (f elements)."

http://cms.mpi.univie.ac.at/vasp/vasp/P ... _tags.html

Cheers,
/Dan

[superyoyo]

Hi Dan
Thanks for your quick reply. I know what I missed.
I tried setting the LMAXMIX = 4 and everything seems making sense now.

Cheers!