Possible bug in PAW_PBE Al_h

[nickjer]

I recently ran a band structure calculation on zincblende AlN using the PAW_PBE Al_h pseudopotential. In the bandstructure there appears 3 flat bands that appear very very noisy right above the top of the valence band. The rest of the bandstructure looks like it should except for those three bands that appear flat, but as I zoom in they appear like a rocky terrain.

I changed the pseudopotential from PAW_PBE Al_h to PAW_PBE Al and redid the entire calculation. The 3 ghost bands are now gone and everything looks right. I made sure to converge the cutoff energy and the k-points. I also added extra bands to be sure nothing was wrong.

You can see the odd results also in the DOSCAR, if you prefer not finding the band structure. The integrated DOS jumps from 8 to 14 almost instantaneously from one energy to the next where the ghost bands appear.

If I am doing something wrong then please let me know. Otherwise, I will just be working with the Al potential and not Al_h.

Also, the US_GGA Al_h does not have this problem. It seems limited to only the PAW_PBE Al_h.

INCAR used:
---------------------------------------------
SYSTEM=AlN
PREC=accurate
ENCUT=800
EDIFF=1E-5

ISPIN=1 ! non spin polarized calculations

ISMEAR=-5 ! use for semiconductors

LWAVE=.FALSE.
LCHARG=.TRUE.
ICHARG=2

ALGO=FAST
-------------------------------------

POSCAR used:
-------------------------------------
AlN zincblende structure
4.38
0.0 0.5 0.5
0.5 0.0 0.5
0.5 0.5 0.0
1 1
Direct
0.00 0.00 0.00
0.25 0.25 0.25
-------------------------------------

KPOINTS used:
-------------------------------------
Large Mesh
0
Monkhorst
11 11 11
0 0 0
----------------------------------------

[job]

On the vasp ftp site there is a short note that the PAW_PBE Al_h potential has been removed due to a problem with ghost states. And indeed the Al_h directory has been removed, except not from the tar archive which I guess most people use anyway.

Yes, I ran into this problem myself as well.

[nickjer]

Thanks job.