About the problem of the convergence of HSE06

[VASP001]

About the problem of the convergence of HSE06, the corresponding INCAR is as followed:

SYSTEM= Cr I Cl
NCORE = 6
Startparameter for this run
#NWRITE = 2 LPETIM=.TURE. # write-flag&timer
#INIWAV = 1 # INIWAV, default is 1,only useful when Istart=0
ISTART = 1 #job : 0-new 1-constant cut-off 2-constant cell size/shape.and cutoff read from WAVECAR
ICHARG = 2 #job : 0-from initial wf 1-from CHGCAR 2-atomic charge density , relax/DOS=2 EK=11
LCHARG = .T.
LWAVE = .T.
ISPIN = 2
MAGMOM = 6 6 -6 -6 12*1

######
ISYM = 2 #default 1-US-pp 2-PAW symmetry stuff ON(1 or 2) or OFF(0)
#####

PREC = Normal
#GGA = PE
#VOSKOWN = 1 #when GGA91 used for magnetic calculation
LORBIT = 11
ENCUT = 254 ev

Electronic Relaxation
NELMIN = 6
NELM = 166 # the max number of electronic SC
EDIFF = 1E-05 # stopping-criterion for ELM
LREAL = F # F-reciprocal T-real A-auto
#AMIN = 0.01

Ionic Relaxation
IBRION = -1 # ionic relaxation:0-MD 1-quasi-Newton 2-CG DOS/EK =-1
ISIF = 2
NSW = 0 # DOS/EK=0
EDIFFG = -0.01
POTIM = 0.6

DOS related values
ISMEAR = 0
SIGMA = 0.06
#NBANDS = 128

ALgorithm
ALGO = FAST

#VDW correction
#LVDW = .T.

#spin-orbital interaction
#LSORBIT = .TRUE.
#SAXIS = 0 0 1
#MAGMOM =
#GGA_COMPAT = .FALSE.
#LMAXMIX = 4
#ADDGRID = .TRUE.

#vdw
LVDW = .T.

LHFCALC = .TRUE.
HFSCREEN = 0.2
LMAXFOCK = 4
ALGO = Normal
TIME = 0.6
PRECFOCK = FAST

#LVTOT = .TRUE.
#LVHAR = .T.


The OSZICAR is as followed:



What's the problem ?

[admin]

Positive total energy indicates that the calculation goes wrong.
It will not recover to "normal" run, therefore it is not necessary
to run large number of electronic steps (166).

[jiri.klimes]

I'd change ALGO=Fast to IALGO=38 (Davidson) first, if that doesn't work, I'd change mixing as discussed in the manual.
BTW you define ALGO two times in INCAR, once in the middle, second after the HSE tag.

[guiyang_huang1]

ALGO=F can not be used for HSE06, although it does not show any warnings according to the vasp manual.

ALGO=N;IMIX=1;AMIX=...., or ALGO=Damped or All should be used. See the vasp wiki.



Using the WAVECAR from GGA calculation (with ISPIN=2 or without ISPIN=2) or HSE06 calculations with less k-point (if it can be converged) as input may be helpful.