HSE06 calculation - improving the convergence

[giacomo giorgi]

Dear All,
I am trying to make calculation at HSE06 level for some ZnTe:O alloys. The systems are essentially quite small (no more than 8 atoms). I did the first run calculation at DFT level (PAW/PBE) and got easily the optimized structures. Then on these preconverged structures I want to make a calculation at HSE06 level in order to investigate the bandstructure.
I am a little bit "desperate" because it is very hard to reach convergence for the single-point calculation at HSE06 level....
So Please I would be very grateful if someone can give me useful hints.

The case here reported regards the Zincblend structure of ZnTe (8 atoms)

On the preconverged DFT calculation I do at first a single point calculation at HSE06 according to this INCAR


SYSTEM = ZnTePureBULKMODULE

# xc
GGA = PE
ISTART = 1

#electronic relaxation
#IALGO = 48
LREAL =.False.
PREC= High

ENCUT=500.0
ENAUG=605.0

#ISMEAR = -5 for tetrahedron k-smearing w/ Blochl corrections
#ISMEAR = N for Methfessel-Paxton order N
#SIGMA = 0.05 eV of Temperature smearing
ISMEAR = 0
SIGMA = 0.005

LWAVE = .TRUE.
LCHARG = .TRUE.
IBRION = 0
ISYM = 0
LHFCALC = .TRUE.
HFSCREEN = 0.2
ALGO = Damped
TIME = 0.4
PRECFOCK= F
NKRED = 2
AEXX = 0.25

#electronic RELAXATION
NELMIN = 4
NELM = 500

ISIF = 3
ISPIN=2

#accuracy of the calculation

EDIFF = 1.E-5
EDIFFG = -0.005

#ionic steps
NSW =0

---------------------------------


POSCAR
Zn Te
1.00000
6.185921848 0.000000000 0.000000000
0.000000000 6.185921848 0.000000000
0.000000000 0.000000000 6.185921848
4 4
Selective Dynamics
direct
0.000 0.000 0.000 T T T
0.000 0.500 0.500 T T T
0.500 0.000 0.500 T T T
0.500 0.500 0.000 T T T
0.250 0.250 0.250 T T T
0.250 0.750 0.750 T T T
0.750 0.750 0.250 T T T
0.750 0.250 0.750 T T T


and these the last few lines of the OUTCAR


.....
N E dE rms(c)
ROT: 1 -0.647379715456E+02 -0.34493E-01
ROT: 2 -0.647379713290E+02 0.21656E-06
gam= 0.382 g(H,U,f)= 0.831E-05 0.323E-06 0.000E+00 ort(H,U,f) = 0.239E-04-0.419E-06 0.000E+00
DMP: 3 -0.314213155466E+02 -0.21173E-04 -0.70403E-05 8064 0.863E-05 0.235E-04
N E dE rms(c)
ROT: 1 0.364040213376E+02 0.10114E+03
ROT: 2 0.382984703276E+02 0.18944E+01
ROT: 3 0.391352639087E+02 0.83679E+00
ROT: 4 0.395386843007E+02 0.40342E+00
ROT: 5 0.397123188562E+02 0.17363E+00
ROT: 6 0.397575037125E+02 0.45185E-01
ROT: 7 0.395694244117E+02 -0.18808E+00
ROT: 8 0.388366584330E+02 -0.73277E+00
ROT: 9 0.364295447337E+02 -0.24071E+01
ROT: 10 0.263890858532E+02 -0.10040E+02
............
.............
.............
ROT: ** -0.516110559187E+04 0.95270E+01
ROT: ** -0.517063255011E+04 -0.95270E+01
ROT: ** -0.516110559187E+04 0.95270E+01
ROT: ** -0.517063255011E+04 -0.95270E+01
ROT: ** -0.516110559187E+04 0.95270E+01
ROT: ** -0.517063255011E+04 -0.95270E+01
ROT: ** -0.516110559187E+04 0.95270E+01
ROT: ** -0.517063255011E+04 -0.95270E+01
ROT: ** -0.516110559187E+04 0.95270E+01
ROT: ** -0.517063255011E+04 -0.95270E+01
ROT: ** -0.516110559187E+04 0.95270E+01
ROT: ** -0.517063255011E+04 -0.95270E+01
ROT: ** -0.516110559187E+04 0.95270E+01
ROT: ** -0.517063255011E+04 -0.95270E+01
gam= 0.382 g(H,U,f)= 0.609E+03 0.305E+02 0.101E+01 ort(H,U,f) =-0.205E+03 0.151E+01-0.774E+01
DMP: 4 0.345010323198E+03 0.37643E+03 -0.22397E+03 8064 0.641E+03-0.211E+03



As you can see, it is very hard to get the convergence at HSE06 for this system, for which, at variance, I would expect an "easier" convergence ( 8 atoms in cubic ZB cell)

Can someone give me some suggestions to improve the convergence of my calculation?
Which are the parameters I should have to "play with"?

Please help me!
Thanks in advance for any kind suggestion!

Giacomo

[giacomo giorgi]

Can anybody help me? Thanks

[admin]

I cannot reproduce the behavior as you report. Between DMP:3 and DMP:4 there are only 6 ROT steps.
I used your POSCAR and INCAR, PAW_PBE potentials and automatic Monkhorst 2 2 2 k-point mesh.

[Lee]

I have a similar problem with convergence. Did you solve that problem?
<span class='smallblacktext'>[ Edited Sun Apr 01 2012, 01:40PM ]</span>