How to understand decomposed/projected DOS components not summing up to the total DOS?

[alpinnovianus]

Hi, when I use sumo to plot the projected DOS over elements and orbitals, I notice that the height of the projected DOS for each component does not sum up to the total DOS.
is this a known phenomenon?

I am wondering what makes up the total DOS beside the sum over each projected contributions?

added note: i also observe the gap between total dos and summed over PDOS contributions is larger for unoccupied conduction states, compared to the occupied valence states. i wonder if this is relevant information....

edit: tidying up the issue to be more precise.

Thank you.
alpin

[ajjackson]

They should do, unless there is some double-counting going on. Is it possible that the gap would be filled by orbitals that are not shown? (Try setting --legend-cutoff 0 to show all orbitals.)

Is this also the case if you compare the el_dos.dat and total_dos.dat files?

[utf]

Actually, I don't think the projections should sum to the total DOS.

The total DOS is calculated using the eigenvalues, whereas the projections depend on the radius of the projector functions used. By default, I think VASP uses the PAW sphere radius for projections. Hence it makes sense that the occupied states are described better than the unoccupied states.

The gap between the total DOS and sum of projections can be put down to electron density that occurs outside of the projector spheres.

[scanlond]

Yes, Alex is right. If you are concerned by this, you can change the RWIGS values in the INCAR to see what difference they make.

…

On Wed, 14 Sept 2022 at 14:48, Alex Ganose ***@***.***> wrote: Actually, I don't think the projections should sum to the total DOS. The total DOS is calculated using the eigenvalues, whereas the projections depends on the radius of the projector functions used. By default, I think VASP uses the PAW sphere radius for projections. Hence it makes sense that the occupied states are described better than the unoccupied states. The gap between the total DOS and sum of projections can be put down to electron density that occurs outside of the projector spheres. — Reply to this email directly, view it on GitHub <#175 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADSCO4XLQZLMV4U2Q5YSCUTV6HJSHANCNFSM6AAAAAAQCNFS2Y> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[alpinnovianus]

Actually, I don't think the projections should sum to the total DOS.

The total DOS is calculated using the eigenvalues, whereas the projections depend on the radius of the projector functions used. By default, I think VASP uses the PAW sphere radius for projections. Hence it makes sense that the occupied states are described better than the unoccupied states.

The gap between the total DOS and sum of projections can be put down to electron density that occurs outside of the projector spheres.

Thank you for the reply.

I know this isn't the VASP forum, but is it usually assumed (or known?) that the pDOS proportions relative to each other computed within the default mode to be similar or representative of the true proportions? (e.g. pDOS of A : pDOS of B in a compound AB)

i.e. whether we can at least compare and analyze the pDOSes relative against each other even though there's this gap of tDOS vs pDOS.

[ajjackson]

Ah, that makes sense! One more reason to take pDOS results with a pinch of salt...

[ajjackson]

I know this isn't the VASP forum, but is it usually assumed (or known?) that the pDOS proportions relative to each other computed within the default mode to be similar or representative of the true proportions? (e.g. pDOS of A : pDOS of B in a compound AB)

How do we define the "true proportions"? There's probably a better method than these spheres but it seems arbitrary at some level. Something involving Bader analysis, perhaps...