I am confused by the units used in the DOSCAR.
When I perform a simple simulation of, e.g., Ru with hcp structure and 2 atoms in the unit cell, I get a Fermi energy of 12.00756235 in line 5 of the DOSCAR. (ISPIN=2)
The corresponding lines are
11.997 0.8678E+00 0.8678E+00 0.7990E+01 0.7990E+01
12.024 0.8726E+00 0.8726E+00 0.8015E+01 0.8015E+01
So that makes 2*8 states. Each Ru atom has 8 electrons so this makes perfect sense.
However, when I perform a larger cell, e.g.a cell with 108 Ni atoms in a 3x3x3 fcc cell, I get a Fermi energy of 7.864. The lines in the DOS file are
7.860 0.3434E+00 0.3364E+01 0.1797E+02 0.1658E+02
7.877 0.3577E+00 0.2777E+01 0.1797E+02 0.1663E+02
Summing spin up and down and interpolating to E_F, I get
(0.013/0.017)*34.55+(0.004/0.017)*34.6 = 34.5617647059
So the integrated number of states is about 34.56.
The total number of electrons (using Ni_pv potential) however is 1728.
The manual says that the units are states/eV and number of states, defined extensively, so I would expect it to be 1728 at E_F.
So why is the integrated DOS at E_F not equal to the number of electrons in the second case, but it is in the first?
The problem seems not to be due to the cell consisting of identical atoms; for a Ni106ZrRu cell (for example), I get similar (small) numbers at E_F
BTW, interestingly,
1728 / 34.5617647059 = 49.9974470258
but this may just be numerology?