NPAR and the projected magnetizations

[qian_liu1]

Hi,

I'm a user of vasp5.3.5. The vaspwiki shows that the spd- and site projected character of each band is not evaluated in the parallel version if NPAR is not equal to 1. But from my calculation, the spd- and site projected character of each band is evaluated in the parallel version even if NPAR is not equal to 1. However, the projected magnetizations are different with each other. I want to know how to set NPAR parameter to get the right projected magnetizations.

Thanks very much.

[henrique_miranda]

Could you please indicate a few more details about your question?
In particular, on what page in the wiki did you read this?
When you say that the magnetizations are different from each other do you mean the ones computed with NPAR=1 and NPAR/=1?
Could you share the INPUT, KPOINTS, POSCAR, and POTCAR files you used in your calculation?

[qian_liu1]

Thanks for your reply.

1、The websit of the wiki is : https://www.vasp.at/wiki/index.php/PROCAR.
2、From the website, the spd- and site projected character of each band is not evaluated in the parallel version if NPAR/=1 . I did my calculation with NPAR=1 and NPAR/=1 in the parallel version when LORBIT=11, both of the PROCARs contain the spd- and site projection information. The orbital projection is the same, but the three spin projections are different for the two case in the PROCARs. So i'd like to know how to set NPAR parameter to get the right spin projection.
3、The input files for the two case i've add in the attachment. The files named npar1 and ncore4 contain the input files for NPAR=1 and NPAR/=1 respectively.

[henrique_miranda]

1. Thanks for pointing out the page on which you found that. This might be slightly outdated so we might need to update the wiki.

2. You are computing a system with no net magnetization (see magnetization in the OUTCAR).
What you are observing in the PROCAR file are basically spurious results since there is no preferred orientation for the magnetization computed locally around each atom.
On each new run, you will obtain different results for the magnetization.

3. Note that you used NPAR in one INCAR file and NCORE=4 in the other. These two variables control the same behavior inside VASP:
wiki/index.php/NCORE
In particular NCORE = number-of-cores/NPAR.
I don't know with how many cores you executed the code but having NCORE=4 and running vasp with 4 mpi ranks means the same than NPAR=1.

[qian_liu1]

Thank you for your detailed explanation, I have understood it.