GW method for metal

[Divyapratheesh]

I was doing GW method for a metal by following the procedure on the VASP Wiki ( using vasp 5.4.4). LOPTICS is given as false in the second step and WAVEDER file is not generated. Then I copied the WAVECAR and done the second step. LORBIT=11 is included in the INCAR as I am interested in the magnetic moment of the material. But I am getting following error in the output:
"= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 12 PID 27244 RUNNING AT compute11
= KILLED BY SIGNAL: 9 (Killed) "
Could you please explain the tags that are to be included in GW method for a metal? Also, is it necessary to set NELM= 1 in the INCAR for both the steps?

[henrique_miranda]

Could you share all the relevant input files for the different steps in your calculation?
It would also be useful if you share a bash script with the sequence of commands that lead to this issue.

[Divyapratheesh]

Sure sir. These are my input files for step 1

1.rar

.
INCAR file for the second step is also attached. Other input files are same as that of first step

INCAR.rar

.
Thank you

[henrique_miranda]

Ok, can it be that you are simply running out of memory?
Can you try running the same calculation with a smaller number of bands (NBANDS=48) and KPOINTS (2 2 2) I would say.
Note that with these parameters the calculation will not be converged at all, my suggestion is just to help in figuring out what the problem is.

[Divyapratheesh]

Ok. I will try.

[Divyapratheesh]

Now i got the output without showing error. W0001.tmp, WFULL0001.tmp ... these files are generated.

[henrique_miranda]

This likely means that you were running out of memory in your calculation.
You can gradually increase the k-point sampling and bands and see how far you can realistically go.

[Divyapratheesh]

Then, how can I understand that whether my outputs are correct or not? If I am not using enough number of kpoints and band, the results will be wrong right?
Also, for different values of NELM in the INCAR, I am getting different values of magnetic moment in the OUTCAR. Then, which value do I need to consider? (From different tutorials, it is seen that NELM can be taken as 1,2,3 or4)

[henrique_miranda]

You need to monitor the convergence of the final quantity you are interested in as a function of k-points and number of bands.
To get the initial KS states you should first run a standard SCF calculation without a very high number of bands just to get the correct charge density. Then on a second step you start from the previous charge density or WAVECAR and specify ALGO=Exact and increase the number of bands. In that case NELM=1 should be sufficient because your charge density has already been converged in the first step.