Error in GW of Aluminum

[Vahid Askarpour]

Dear VASP Users,

When applying GW to aluminum (which can be done with ABINIT), I get the following error after the first iteration:

internal error in CDER_BETWEEN_STATES_ROTATED: optical matrix elements not calculated 1 259 17

Here are my input files (which, with minor changes, work fine for silicon):

KPOINTS
K-Points
0
Gamma
8 8 8
0 0 0

POSCAR
Fm-3m Al
4.0479
0.0 0.5 0.5
0.5 0.0 0.5
0.5 0.5 0.0
Al
1
Direct
0.0 0.0 0.0

INCAR(for generating WAVECAR and WAVEDER)
System = Al
ISMEAR = 1 ; SIGMA = 0.1
NBANDS=400
LOPTICS=.TRUE.
NEDOS=2000

INCAR(for GW run)
System = Al
ISMEAR = 1 ; SIGMA = 0.1
NBANDS=400
ALGO=GW
NOMEGA=56
LOPTICS=.TRUE.
LSPECTRAL=.TRUE.
LWAVE=.TRUE.
NELM=10
NEDOS=2000
NPAR=8

For metals, of course, one needs more kpoints and bands.

Can the GW routine of VASP be actually applied to metals or is the above error a bug? I have used both vasp5.2.2 and vasp5.2.11.

Thanks,
Vahid
<span class='smallblacktext'>[ Edited ]</span>

[alex]

Add ISPIN=2 and retry.

[Vahid Askarpour]

I added ISPIN=2 to both INCAR files and still get the same error after the first iteration.
<span class='smallblacktext'>[ Edited Thu Jul 07 2011, 12:51AM ]</span>

[Vahid Askarpour]

It is possible to achieve convergence by:
1. Changing ISMEAR=1 to ISMEAR=-1
2. Removing the NELM tag. This tag causes the code to hang after the first iteration. Instead, repeatedly calling vasp 4-5 times will give the desired result.
<span class='smallblacktext'>[ Edited Mon Jul 11 2011, 02:20PM ]</span>