HSE DOS convergence problem

[apple]

hi all,
I am running a DOS calculation with HSE. The following is the INCAR file I prepared:
----------------------------------------------------
SYSTEM = Wurtzite GaN

Startparameter for this Run:
NWRITE = 1
ISTART = 0
ICHARG = 11

Electronic Relaxation 1
ENCUT = 300.0 eV
PREC = HIGH
EDIFF = 1E-04 stopping-criterion for ELM
EDIFFG = -1E-02 stopping-criterion for ELM

Ionic Relaxation
ISYM = 1
LCORR = T Harris-correction to forces
IBRION = -1
NSW = 0
ISIF = 2
POTIM = 0.1


DOS related values:
ISMEAR = 0; SIGMA = .01
RWIGS = 1.71368 0.82511
LORBIT = 2

Electronic Relaxation 2
ALGO = Damped
# LDIAG = T sub-space diagonalisation
# LREAL = F real-space projection


NBANDS = 26
TIME = 0.5
LWAVE = T
LHFCALC = T
HFSCREEN = 0.207
PRECFOCK= N
------------------------------------------------------------------
This is the second step (after the self-consistent step, which has yielded CHGCAR).
The job is not converging as shown below:
energy without entropy = 495.28100875 energy(sigma->0) = 495.28089000
energy without entropy = 107.10207758 energy(sigma->0) = 107.10207758
energy without entropy = 63.99151600 energy(sigma->0) = 63.99150613
energy without entropy = 63.20614408 energy(sigma->0) = 63.20613413
energy without entropy = 63.17681980 energy(sigma->0) = 63.17680985
energy without entropy = -24.28379926 energy(sigma->0) = -24.28379926
energy without entropy = -9.23295232 energy(sigma->0) = -9.23295232
energy without entropy = -10.55410634 energy(sigma->0) = -10.55410634

Can someone let me know why there is a problem and how to fix it. I would greatly appreciate it.
Thanks.

[riche]

What do you use Gaussian smearing? I thought we should use -5 and Gamma to get DOS.

[apple]

hi,
is the smearing=-5 setting necessary when performing DOS? I would appreciate your answer.

[admin]

Please note that DOS or band structure calculations using HSE must NOT be continued from an exixting CHGCAR file, since the non-local exchange is not determined by the charge density but by the density matrix and/or the KS-orbitals. (this has been discussed in the forum before)

[sankh]

Hi,

I do not really see any objection towards using Tetrahedron integration method with Blöchl's correction (i.e. Ismear = -5) to conduct DOS calculations at HSE level, but of course this cannot be used for HSE band structure calculations that requires explicit inclusion of chosen/desired high-symmetry path k-points in the KPOINTS file with "0" weights, on top of your converged k-mesh points with their weights from a previous SCF run, and continue another SCF run using gaussian or MP type of integrations.

However, it would be helpful if Admin kindly confirms whether or not Tetrahedron+Blöchl correction method should be used for HSE DOS calculations.

Thanks

[kirby]

Hi,

I am also concerned about whether or not Tetrahedron+Blo"chl correction method should be used for HSE DOS calculations. Or could I use the same settings as SCF runs and a denser K-point grid to calculate HSE DOS?