HSE06 calculation for 2D materials

[ashima29]

I am trying to calculate the HSE06 band structure ,but on doing the scf step in the HSE calculation after the PBE calculations the energies of my struture are coming out to be highly positive, however the PBE results were fine (energies were negative). The INCAR I am using is the following :

SYSTEM= Materials
ISTART =1
ISPIN=1

#Electronic Relaxation
ISMEAR = 0
SIGMA = 0.05
NELM = 60
EDIFF = 1E-05
LORBIT=10
NEDOS=2000
IVDW = 1

# HSE calculations:
LHFCALC = T
HFSCREEN = 0.2
ALGO = Damped
AEXX = 0.25
TIME = 0.4
PRECFOCK= normal

# Global parameter
LREAL=.FALSE.
PREC= N
LWAVE= F
LCHARG =F
IDIPOL = 3
LDIPOL = .TRUE.
LVTOT = .TRUE.
LVDW = .TRUE.

I am using the 9X9X1 K-mesh for the scf calculation.

[ruva17]

Hello. I am having the same issue. Did you end up finding a solution to this problem?
Best regards

[alexey.tal]

Hi,

Please provide the input files for your problem.

[ruva17]

Hi,

Please provide the input files for your problem.

Hello,
I am attaching a .zip containing the input files, the CONTCAR and the OUTCAR from the relaxed structure, as well as the corresponding INCAR and KPOINTS files for the bands calculation. I also include a screenshot of the OSZICAR file from the HSE06bands calculation before I interrupted it.

A few notes: I know the KPOINT grids from the relaxation and bands calculations are different. Currently I am using a mesh with low resolution just to test the correctness of the parameters. These calculations still should converge relatively quickly. I believe the problem stems from the fact that the initial energy is positive in the bands calculation.
I used the exact same INCAR (minus the vDW correction) for just the 2D semiconductor and it worked perfectly.
I am running on VASP 5.4.1 with 36 cores.

Thanks in advance

[alexey.tal]

Dear ruva17,

If I understood correctly, you tried restarting your HSE06 calculation from charge density that was calculated with PBE but the k-mesh was different. When you change the k-points grid the orbitals from the WAVECAR file become essentially incompatible and cannot be used as an initial guess for restarting the calculation.

You should try starting your HSE calculation with orbitals calculated in GGA.
Also, the dipole corrections sometimes make it difficult to converge the calculation, so you can pre-converge your GGA run without LDIPOL.

[ruva17]

Hello alexey.tal,

Thank you so much for your response. While I agree that indeed the difference in the k-mesh makes the WAVECAR not completely compatible, using the file as provided from the PBE calculation has been a good initial guess for the other calculations using only the Janus structure. I think the cause of error lies within the calculation of a heterojunction instead of an isolated system. Also, I tried calculating without LDIPOL but the error persisted. However, I managed to solve the problem and make the calculations converge by changing the diagonalization algorithm. I used the next setup:
#HSE06
LHFCALC = .TRUE.
HFSCREEN = 0.2
GGA = PE
AEXX = 0.25
AGGAX = 0.75
AGGAC = 1.0
ALDAC = 1.0
ALGO = Normal
IMIX = 1 ; AMIX = 0.4
It is a bit slower, but converges to a seemingly correct result, so I'm happy with it.
Thanks again for the help.
Best regards,
ruva17