Reproducing bandgap of GaAs as PRB75 235102

[pumpkin]

Dear all,

I am now trying to reproduce the bandgap of GaAs result of Phys. Rev. B75 235102 by M.Shishkin and G.Kresse.

In this thesis, bandgap of GaAs for PBE was 0.49eV.
However, my result was around 0.62eV.
I don't know what is the reason of this difference.
I want to pick your brains what is improper of fatal.

[INCAR]
ISTART = 0
PREC = high
ALGO = Fast
ISIF = 2
ISMEAR = -1
SIGMA = 0.05

[KPOINTS]
Gamma centered k-point mesh
0
Gamma
8 8 8
0 0 0

[POSCAR]
GaAs
5.648
0.5 0.5 0.0
0.0 0.5 0.5
0.5 0.0 0.5
1 1
direct
0.0 0.0 0.0
0.25 0.25 0.25

[POTCAR]
The output of 'grep TITEL POTCAR' is as follows :
TITEL = PAW_PBE Ga 08Apr2002
TITEL = PAW_PBE As 06Sep2000

[OUTCAR]
E-fermi : 3.1217 XC(G=0): -9.1988 alpha+bet : -8.5920
add alpha+bet to get absolut eigen values

k-point 1 : 0.0000 0.0000 0.0000
band No. band energies occupation
1 -9.9310 2.00000
2 2.7801 1.99785
3 2.7801 1.99785
4 2.7801 1.99785
5 3.4054 0.00685
6 6.5419 0.00000
7 6.5419 0.00000
8 6.5419 0.00000

Eg(GaAs) = 3.4054 - 2.7801 = 0.6253eV

Thanks.

[pumpkin]

Dear all, this is self-reply

The possible reason is the relaxation of ions in the unit cell. Proper IBRION parameter would be necessary.

Thanks.