ZnO polar surfaces

[giacomo giorgi]

Dear all,
I am facing a result I am not able to understand. I am trying to reproduce at least qualitatively the bandgap of some polar surfaces.
For bulk ZnO I get results according very well with previous theoretical works (gap ~ 0.75 eV)

BUT
For the ZnO surfaces 0001

keeping the experimental lattice parameters fixed I optimized both the surfaces O and Zn terminated in the same 8 atom slab

Zn O
1.00000000000000
2.81424 -1.6248 0.000
0.00000 3.2496 0.000
0.00000 0.0000 26.2084
4 4
Selective dynamics
Direct
0.333 0.666 0.36835 T T T
0.666 0.333 0.46605 T T T
0.333 0.666 0.56807 T T T
0.666 0.333 0.66582 T T T
0.666 0.333 0.36612 T T T
0.333 0.666 0.46657 T T T
0.666 0.333 0.56764 T T T
0.333 0.666 0.66777 T T T

Then I did band calculations. Both adding LDIPOL=true, IDIPOL=3
and not adding them I get a value for the gap of about 1.2 eV (!!)

To the best of my knowledge the two surfaces should be metallic!
I also attach the INCAR for the geometry relaxation

INCAR

SYSTEM = ZnO 0001

# xc
GGA = PE
VOSKOWN = 1
ISTART = 1
#surface dipole
IDIPOL = 3
LDIPOL = .TRUE.

#electronic relaxation
LREAL = Auto
PREC= High
ENCUT = 500
ENAUG = 700

ISMEAR = 0
SIGMA = 0.05

NELMIN = 4
NELM = 200
ISIF = 2

IBRION = 1

#accuracy of the calculation
EDIFF = 1.E-5 # for bulk 1.E-6
EDIFFG = -0.02 # for bulk -0.005

#ionic steps
NSW = 500



The error of course is macroscopic, but I do not figure out the reason... For sure I am missing details about charge compensation between O- and Zn-terminated surfaces.

Must I split the calculation in 2 calculations, freezing bottom layers?

Thanks in advance.
Best,
G

[giacomo giorgi]

Has anybody studied polar surfaces?

Best,
G

[admin]

please read eg.
Competing stabilization mechanism for the polar ZnO(0001)-Zn surface by G. Kresse, O. Dulub, and U Diebold, PRB 68, 245409 (2003)

[giacomo giorgi]

Dear administrator. Thanks. Maybe I got the point. You suggest me to saturate the frozen O-bottom layer with fractional H atoms (i.e. H.5) without adding DIPOL (LDIPOL, IDIPOL) corrections, don't you?
Best,
G