ALGO=ALL gives eigenvalues with wrong ordering

[saeed_bahramy]

Hi,

I am trying to perform some slab calculations using METAGGA=MBJ. For convergence, I have tried quite a few settings, including ALGO=ALL, ALGO=D with small TIME step, ALGO=N with a small linear mixing factor, etc. So far, the only setting working is ALGO=ALL and LDIAG=.TRUE., however, the eigenvalues in this case are not properly orderd.

Apparently, I am not the only person having this problem. I have found similar questions on this forum and elsewhere, but no solution. I apprecite it if Admin or anyone else can explain how to fix this.

Thanks,
Saeed

[rightsuman]

Did anyone get a solution to this issue?
I am also facing a similar problem with heterostructure mbj calculations that has vacuum. The band structure has noise but the overall characteristics are proper

[henrique_miranda]

First a disclaimer: the MBJ functional is not very good at describing systems with a vacuum because of the definition of the functional itself.
There is recent work to address this issue: https://doi.org/10.1021/acs.jctc.9b01147
This new functional, however, is still not implemented in VASP.

For the particular issue you are observing, could you share your INCAR, POSCAR, POTCAR and KPOINTS files?

[rightsuman]

Hello Henrique,
Thanks for your reply. As the same consists of my ongoing work, can I send this to you over email?

[henrique_miranda]

If you don't want to share the POSCAR you are working on maybe you find the same issue with a different POSCAR?