Minimum one empty band is required for calculations.

Archana Sharma You can simply run a calculation without specifying the NBANDS tag in your INCAR. By doing so, you will get the default NBANDS value. Now for any physical quantities that you are interested in, you can just either double the NBANDS value or triple it to do a convergence test and see if your desired quantity is converged w.r.t NBANDS. Usually, a larger NBANDS value requires a much larger time to finish the calculation. Also, NBANDS value doesn't depend on the no. of kpoints up to my understanding.

Hope it helps. Thanks!

Adding to above answer, you can find similar discussions in PW_forum and other periodic DFT forums.

For instance: https://lists.quantum-espresso.org/pipermail/users/2017-November/039699.html

The minimum value for no of bands to be taken is sum of valence orbital (for s=1, p=3 and d=5; valence orbital depend on the pseudopotential used) multiplied by number of atom, this approach gives complete density of states. Although the number of Kohn-Sham states are enough to get proper dos and bands and its minimum value for running wannier.

Dear Archana Sharma

If you are doing the calculations without SOC, then NBANDS should be larger than half of NELECT, and for with SOC, NBANDS should be at least twice of NELECT, where NELECT is the number of electrons in the system.

Suppose, your NELECT = 20 (you can find it in the OUTCAR), so for without SOC, 20 electrons can be occupied in 10 bands, so valence band would contain 10 bands, but to keep the finite conduction band region too, we keep NBANDS higher than 10, so you can keep NBANDS as 20 here.

But, with SOC, we need one band for each electron, so the valence band should contain at least 20 bands, and to consider the conduction band region too, we can consider NBANDS = 40.

Hope it helps!