GW0 band gap results different between 5.3.3 and earlier versions

[huangj3]

Dear all,

For some reason, I compared the GW0 band gap results between the latest 5.3.3 and earlier versions including 5.2.11, 5.2.12, and 5.3.2. The rest are the same and only 5.3.3 results are different. Tried to compare the 5.3.3 OUTCAR with the 5.3.2 OUTCAR and don't see any changes in default setting. Don't see any relevant changes on the new release page for 5.3.3 at http://www.vasp.at/index.php/news/44-ad ... -vasp532-2. So I don't have a clue what has been changed in 5.3.3 so that the band gaps are different. Using the following input files, the GW0 gap is 1.285 eV with 5.3.3 and 1.369 eV with 5.3.2 and earlier versions.

INCAR.1
SYSTEM = Si
ISMEAR = -5
ALGO = Fast
NBANDS = 192
LOPTICS = .TRUE.
ENCUT = 500.0
NPAR = 48
NEDOS =2000
LCHARG = .FALSE.

INCAR.2
SYSTEM = Si
ISMEAR = -5
ISTART = 2
ICHARG = 0
ALGO = GW0
NBANDS = 192
LOPTICS = .TRUE.
NEDOS = 2000
NELM = 4
NOMEGA = 48
!LORBIT = 11
ENCUT = 500.0
NPAR = 48
!LRPA = .TRUE.
EMIN = -6
EMAX = 19
LCHARG = .FALSE.

POSCAR
system Si
5.430
0.5 0.5 0.0
0.0 0.5 0.5
0.5 0.0 0.5
2
cart
0.00 0.00 0.00
0.25 0.25 0.25

KPOINTS
Automatic k mesh
0
Gonkhorst Pack
4 4 4
0 0 0

POTCAR is the PAW_PBE Si 05Jan2001 file.

Could any expert give a hint? Which result is the correct one? (I am not talking about the converged values and am just curious about the difference in the band gap values.)

Thanks a lot.

JSH

[huangj3]

Replaced the PAW_PBE Si 05Jan2001 potcar file with the latest PAW Si_GW 04May2012 potcar file. The GW0 Eg gap is 1.313 eV with 5.3.2 (and possibly 1.313 eV with earlier versions -didn't try for this potcar) and 1.180 with 5.3.3. So, what has been changed in 5.3.3 so that the band gaps are different and which one should we trust?

[huangj3]

Any comments or suggestions from the admin? Thanks.

[huangj3]

Never mind. Probably compilation error. The problem is gone. Also, the newest 5.3.5 is good, giving consistent GW0 results.
<span class='smallblacktext'>[ Edited Fri Apr 25 2014, 05:23PM ]</span>