DOS and the effect of NBANDS

[q365]

Dear VASP forum,

I am trying to calculate the full DOS for the d-orbitals of different transition metals near the fermi level such as a surface slab of Cr(110). In this case, I would like to calculate the full 3d orbital d-band from below the fermi level to the unoccupied portion above the fermi level. I've noticed that I need to increase NBANDS from the default program value in order to get the full d-band which makes sense but my question is what happens if NBANDS is too large? Is it possible that if I include too many BANDS that instead of just having 3d orbitals that I could also include 4d or other higher unoccupied orbitals in my calculation of the angularly resolved DOS? I ask because I notice that my number of states per atom will pass 10 d-states/atom if I increase NBANDS to higher numbers. If this is the case, is there a way to distinguish which d-orbital states are from which primary quantum number so that I can just single out the 3d states in my DOSCAR file?

Thank you very much!
q365

[admin]

if you want to plot the complete d-DOS (including the complete unoccupied part of the d-band) please simply count the number of states you need altogether and set NBANDS accordingly. Please don't forget to check which electrons are included as valence in your POTCAR in addition (s,p). The 4d shell should be much higher in energy than the 3d shell, so you also can set EMIN and EMAX to confine the energy range of the DOS.

[q365]

Thank you, thank makes sense. In the POTCAR file, I found the number of valence electrons listed as ZVAL.

If I were to calculate my DOS using a spin-polarized calculation with something like Ni, do I need twice the number of bands then when I have a non spin-polarized calculation? So instead of needing only 6 valence bands for each Ni atom (s and d valence bands) would I then need 12 valence bands to cover the same s and d valence orbitals?

Thank you very much!
q365

[admin]

no, NBANDS has to be set to the number of majority (spin up) bands. The band-indices are not merged for up and down states, there is a separate spin index to count the states.