Missing electrons if spd-site projected

[physicalattraction]

I have calculated the DOSCAR of a 12 layer Ag(111) slab using LORBIT = 11. If I integrate the DOS given at the beginning of the DOSCAR file over energy, I find 178 electronic states, consistent with my NBANDS = 79. After this, the DOS is projected onto the 12 ions and onto the orbitals, in my case 9.

Question 1: There are no names above the orbitals, but I assume they are identical as the ordering in the PROCAR file, where their names are given. In this case, this would be " s py pz px dxy dyz dz2 dxz dx2". Is this correct?

Question 2: If I sum the DOS over all ions and over all orbitals and integrate over energy, I find only 126 electronic states. Where are the other electronic states?
* Do they possess neither s, p nor d character?
* Is the DOS split into so many small numbers, that I have a serious rounding error?
* Is there a misconception in my interpretation?

[riche]

Answer 1: correct.
Answer 2: The discrepancy is normal (I guess). Because the partial DOS is evaluated within a volume specified by WS radius written in POTCAR, which means you may get overlaps or voids. This is a very annoying problem. The sum of partial DOS is not identical to total DOS.

[forsdan]

Just for your information:

Graeme Henkelman's group at UT Austin has developed a tool to vasp to allow for DOS projection in the Bader volumes, please see:

http://theory.cm.utexas.edu/vtsttools/dos/

In theory this would allow for having a partial DOS summing up to the total DOS ( although I am still waiting for a response abot some differences occuring for this scheme).

Overall the results from the Bader projections give the same qualitative conclusions regarding the bonding nature in the materials as an approach with approriately choosen Wigner-Seitz radii does. So the latter is therefore still a valid method, even though I naturally perfer the former.

Cheers,
/Dan

<span class='smallblacktext'>[ Edited Tue Oct 19 2010, 07:33PM ]</span>