ABINIT Discussion Forums

Hello,

I'm using abinit 6.2.3.

I have a problem with the integrated density of states in the
DOS output file (prtdos 2). The top of the DOS output file gives
the fermi energy, but the integrated DOS value at that fermi
energy is not equal to the total number of electrons in the cell.

I have been using "GGA.fhi" pseudo potentials.
Also I tried a spin-polarized calculation, but with the same result.

Shouldn't the integrated DOS at the fermi energy be equal to
the total number of valence electrons in the system?
(I mean: in the unit cell each Na contributes 1 electron,
each O 6 electrons, each C 4 electrons etc.).

Why is there a discrepancy of a few electrons between the
integrated DOS at the fermi level, and the total number of
treated electrons in the cell?

Is the calculation of the fermi energy incorrect?
Or am I doing something completely wrong here?

Regards,
Rob.

Hi Rob,

Can you send more information please? Your input and which pseudopotential you used would be helpful.

Is the discrepancy significant? If it is an integer, it probably means that your pseudopotential have more (or less) core electrons than you thought.

Also, the option for band occupation (occopt) might influence the determination of the Fermi level. If you used occopt=1 for a metal, you will have a problem.

Simon

Thank you for your reply.

I have done an example calculation here:
http://skku.homeip.net/lahaye/abinit/NumberOfElectrons/
which hopefully demonstrates my issue.

This is NaTaO3, but I use 8 unit cells in my supercell and replace one oxygen
atom a by C atom, to simulate a kind of doping effect.
Hence, I have a supercell with 40 atoms: 8 x Na, 8 x Ta, 23 x O, 1 x C.

From the "fhi" pseudo potentials I can see:
Na has 1 valence electron, Ta has 5 electrons, O has 6 el., and C has 4 el.
Therefore, in my supercell I should have 8x1 + 8x5 + 23x6 + 1x4 = 190 electrons.

Now I want to highlight the troublesome excepts of the DOS output file nofo_DOS:

# nsppol = 2, nkpt = 75, nband(1)= 130
# Tetrahedron method
# For identification : eigen(1:3)= -0.872 -0.867 -0.866
#
# Fermi energy : 0.02039816
#
[...snip.....From Spin-up DOS:...]
1839 0.01950 28.6539 95.9629
1840 0.02000 31.8590 95.9805
1841 0.02050 0.1147 96.0000
1842 0.02100 0.1125 96.0000
[...snip.....From spin-dn DOS:...]
1839 0.01950 28.6539 95.9629
1840 0.02000 31.8590 95.9805
1841 0.02050 0.1147 96.0000
1842 0.02100 0.1125 96.0000
[...snip...]

I have done a spin polarized calculation and for both, Spin-up and Spin-dn,
I find 96 electrons around the indicated fermi level, that is 192 electrons.

Why does this DOS calculation have 192 electrons, whereas there should
be only 190 electrons ???

Am I doing something wrong, or am I misinterpreting the DOS data file?
Help is much appreciated.
Thank you!

Rob.

The problem probably comes from the option for the occupation. Since you have a metal (DOS not 0 at E_F), you should use occopt=7.
See:
http://www.abinit.org/documentation/hel ... tml#occopt

This will probably solve your problem. Your ecut is also pretty low, but I assume who knew that.

Good luck,

Simon

ok, thanks for the very complete details and files. Your occupations and output seem to indicate 95 bands occupied, whereas the DOS goes to a different number of electrons at the Fermi energy (note: your files are different from the example you cite in the post - don't know where your 0.02039816 comes from)

Try with prtdos 1 for a reference (you can rerun from the same WFK) - in some cases the tetrahedron method has issues if 3 or 4 k-points at the vertices of a tetrahedron are degenerate in energy (usually for very sparse or very symmetric grids). This should be dealt with cleanly, however - I have never seen any problems.

Did you set "charge" at any point? This would explain the added electron if the WF come from a charged run and are reused for the DOS, or something like that...

keep looking... Apart from the DOS did you see any other signs of charge non-conservation?

Matthieu