Dependence of Gamma point band gap on directions within HSE06

[bakulin]

Hallo all!

I try to performed band structure calculation within HSE06 methods for InAs bulk structure.
For all next calculations I added in INCAR the following ?standard? flags:

LHFCALC = .TRUE.
HFSCREEN = 0.2
ALGO = Damped
TIME = 0.4
PRECFOCK = Fast

1. I performed relaxation of cell volume with ISIF=7 and obtained CONTCAR:

InAs bulk
6.05800000000000
0.0000000000000000 0.5067455015472403 0.5067455015472403
0.5067455015472403 0.0000000000000000 0.5067455015472403
0.5067455015472403 0.5067455015472403 0.0000000000000000
In As
1 1
Direct
0.0000000000000000 0.0000000000000000 0.0000000000000000
0.2500000000000000 0.2500000000000000 0.2500000000000000


The following KPOINTS was used:

Automatic mesh
0
G
7 7 7
0. 0.

2. I did one self-consistent iteration with ISTART=1, ICHARG=1 and get ?good? CHGCAR which can be read by ICHARG=11. At this step I have 0.3029 eV band gap in Gamma point.

3. I perform band structure calculation:

INCAR include ICHARG=11, LORBIT=12, NBANDS=12 besides ?standard? flags.

If I use the following KPOINTS

Automatic mesh
50
L
rec
0.5000 0.0000 0.0000 L
0.0000 0.0000 0.0000 G

I get 0.5757 eV band gap. It is almost two times higher than in the second step, i.e. in case of self-consistence calculation.


MOREOVER, if I calculate two directions

Automatic mesh
25
L
rec
0.5000 0.0000 0.0000 L
0.0000 0.0000 0.0000 G

0.0005 0.0005 0.0000 G
0.5000 0.5000 0.0000 X

I get 0.4652 eV band gap.


If I use five directions

Automatic mesh
30
L
rec
0.5000 0.2500 0.0000 W
0.2500 0.2500 0.2500 L

0.2495 0.2495 0.2495 L
0.0000 0.0000 0.0000 G

0.0005 0.0000 0.0000 G
0.5000 0.0000 0.0000 X

0.5000 0.0005 0.0000 X
0.5000 0.2495 0.0000 W

0.4995 0.2505 0.0000 W
0.3750 0.3750 0.0000 K

Band gap is equaled 0.2993 eV.

It is very strain result that value of gap in Gamma depends from number of directions in KPOINT because band structure calculation is nonself-consistence one.

Somebody know why it happens and how to fight it?

PS. Small shifts from high symmetry k-point was done to avoid ?error in IBZKPT_HF: two k-points are equivalent this will cause problems in the HF routine?

[admin]

band structures must NOT be calculated with this approach if hybrid functionals are used, please have a look, this issue has been discussed in the forum before