Fermi energy of semiconductor

[simly]

For very accurate DOS and total energy calculations in semiconductors, ISMEAR=-5 should be used.
However, I've noticed that ISMEAR=-5 and ISMEAR=0 gives a difference of more than 0.5 eV for fermi energy of the same system. In general, which ISMEAR value generates a more accurate fermi energy?

Thank you

[dtrinkle]

The Fermi energy also depends on what smearing value you use. The Fermi energy for a semiconductor can technically be anywhere in the gap--if your gap is (say) 1eV, a range of 0.5eV doesn't mean anything is wrong.
The Fermi energy is such that there are N electrons in the system. If you mean to get the HOMO out, don't rely on the Fermi energy--go find the highest occupied orbital energy instead.

[admin]

furthermore, in all periodic systems, the energy zero is not defined. There may be a (small, but non-negligible) constant shift in the energies in 2 different calculations. In a system with a gap, you can check at which E the DOS drops to zero (this is consistent with the concept of computational solid state physics (not experiment!) that E-fermi can be set to the highest occupied level (HOMO). If you choose this approach, please mind that the energy grid on which the DOS is evaluated is chosen fine enough (NEDOS in INCAR). Please mind that for the ISMEAR=0 calculation, the smearing witdh gives an additional incertainty to determine this energy.