HSE06 BAND Calculations

[pragnya_samal]

the error message is
"VASP internal routines have requested a change of the k-point set.
Unfortunately, this is only possible if NPAR=number of nodes. Please remove the tag NPAR from the INCAR file and restart the
calculation."
however, I am not using NPAR, its NCORE I am using in Vasp.6.4.3 19Mar24 version.
Please sugget what to change

[martin.schlipf]

NPAR (ranks parallelizing bands) is still calculated from your INCAR file even if we do not recommend to use it anymore.
NPAR = total ranks / (NIMAGE * KPAR * NCORE)
You can derive that from our documentation. This also explains how your parallelization setup maps to the physical architecture.

If you want more specific advice, please provide more details about your setup: How many nodes do you run on, how many cores per node, how many ranks did you run on and what INCAR, POSCAR, and KPOINTS file did you use?

[pragnya_samal]

I have uploaded all files, please have a look and let me know how solve this error.

[martin.schlipf]

There are two issues with your input:

• You set LKPOINTS_OPT in the input, but your file should be named KPOINTS_OPT not LKPOINTS_OPT. Setting the INCAR tag is not required, the presence of the file would be enough.
• NCORE is not compatible with this setup. The error message for 6.4.3 is a bit misleading here, since 6.5.0 the message is clearer

  A calculation of the band-structure with hybrid-functionals on the k-points read from KPOINTS_OPT file was requested with NCORE/=1. This is currently not supported.

  If you remove NCORE, the calculation would run but not use the k-points you want, because the file has the wrong name.

I would also like to give you some more general advice:

• Start with small calculations and work your way up. In your case, I can reproduce the issue with a simple Si bulk unit cell that takes no time to run. Move to bigger systems only when you are sure your setup works. This avoids burning compute time on setups that do not work. In your case you could start with a simple molecule O2 or CO2 in a box and then move to larger molecules as you gain more experience. Perhaps there is a smaller prototype of the molecule that you are really interested in that you can study as a intermediate step.
• Study what convergence parameters are important. In the spirit of the advice above, start with a loose setting that is just enough to capture the physics. Usually that means default ENCUT. In your case, your molecule is 13 Å wide but you put it into a 50 Å box. You could probably start with 20-25 Å and then when that is working increase your cell size. You may also use electrostatic corrections to improve the convergence with the cell size.
• It is often advisable to keep you input file more minimal and add additional tags as you need them. For example, you set a couple of tags in your INCAR file that are the default. This makes it harder to process what you want to do and can lead to errors, if you later want to set LVTOT and miss that it is already set to false. Some of the choices are also questionable. I do not know whether you need NBANDS = 400 for 112 occupied states, NCORE is also very large, please check this against our advice. You set SIGMA to its default value, which may be too large. You should converge that to smaller smearings. If you want to relax the structure at some point, I am also not sure your vdW correction setup makes sense.
• Finally, your calculation does not make a lot of sense from a physics perspective. A molecule in a large box will not have a band structure. If the eigenvalues spread over the k-points that would be an indication that your cell size is to small. Instead you can just take the eigenvalues that VASP already reports to the OUTCAR file. Without this, you can also switch to the Gamma point version of VASP which is about twice as fast.

[pragnya_samal]

Thanks a lot, actually I was going to do GW-BSE calculations and wanted to benchmark it with HSE06, thanks for all suggestions. I will work on that.