NELECT and Converged Integrated DOS Not the Same

Message

Modey3 · #1 Post by **Modey3** » Sat Jun 27, 2009 1:05 pm

Hello,

I have a serious problem. When I try to calculate the DOS for FCC Al the NELECT and converged integrated DOS are not the same. Using a USPP, each Al atom has 3 valence electrons. There are 4 Al atoms in my supercell. That gives 12 electrons total in my supercell, NELECT = 12. However, the integrated DOS converges to 24 electrons. What is the problem? Any help would be appreciated. Thanks

Luke

forsdan · #2 Post by **forsdan** » Sat Jun 27, 2009 1:16 pm

You are probably looking at the last value of the integrated DOS, which should be equal to the number of bands in your calculation. In your case you apperantly have NBANDS = 24 (verify this by looking in the OUTCAR file). Please check the integrated DOS at the Fermi level instead which should give you 12. The rest is just unfilled bands.

Cheers,
/Dan

Modey3 · #3 Post by **Modey3** » Sun Jun 28, 2009 11:19 pm

thanks forsdan for your response. Why are the "unfilled" bands occupied. There should not be any occupation above the Fermi Level. Thanks

Luke
[ Edited Mon Jun 29 2009, 01:26AM ]

forsdan · #4 Post by **forsdan** » Mon Jun 29, 2009 5:23 am

I never said that the unfilled bands are occupied. I pointed out that you'll get the DOS for all bands in your calculations. The last value in the integrated DOS column will therefore be equal to the number of bands. If you check the integrated DOS at the Fermi level this will be equal to the number of electrons in your systems, since you only get filling up to here.

EDIT: Just a small modification to my answer. I mostly perform spin-polarized calcaulations so I've been referring to that the integrated DOS per spin component should be equal to the number of bands. The total integrated DOS across all bands in the calculations, taking into account both spin components, will of course be equal to 2*NBANDS. Checking the total integrated DOS at the Fermi level will give you the number of electrons in the system.

Cheers,
/Dan

[ Edited Mon Jun 29 2009, 08:16AM ]

Modey3 · #5 Post by **Modey3** » Mon Jun 29, 2009 1:26 pm

forsdan,

The problem is that according to my DOSCAR, the energy levels above the Fermi Energy are occupied. This should not happen. The Fermi Energy given by the OUTCAR is correct because from the DOSCAR, the Integrated DOS = NELECT at the Fermi Level.

According to the VASP manual. The Integrated DOS is the total states per supercell. Is the manual wrong?

forsdan · #6 Post by **forsdan** » Mon Jun 29, 2009 2:57 pm

The manual entry

http://cms.mpi.univie.ac.at/vasp/vasp/node69.html

is perfectly fine. The integrated DOS N(e_i) can be written as "integral n(e) de" from minus infinity to e_i, where n(e) is the DOS. This corresponds to the number of states within the energy range (-infinity,e_i]. Integrating the DOS up to the Fermi level gives you the total number of electrons in your system. Then you can of course continue the integration over the unfilled bands as well and as a limiting value you will obtain 2*NBANDS. In that case the integrated DOS value will contain both filled and unfilled states.

The only thing in the DOSCAR file that specifies the filling of the energy levels is the Fermi level. So if the integrated DOS at the Fermi level correctly gives you NELECT (as it should), what makes you believe that the higher energy levels are filled?

Cheers,
/Dan

[ Edited Mon Jun 29 2009, 05:11PM ]

Modey3 · #7 Post by **Modey3** » Mon Jun 29, 2009 4:31 pm

Apparently, the integrated DOS must be the total available states not just occupied states. So when then the integrated DOS converges to a value that must mean a band gap is reached.

Luke