LDOS

[forsdan]

Dear VASP-users,

I'm trying to calculate the LDOS for a system with one B-atom in a supercell with 128 Fe-atoms. using PAW-potentials. The Fe-atoms have 8 valence electrons each while the Boron atom usues 3 valence electrons. I've used LORBIT = 12 in order to avoid to specify the RWIGS-tag. When I sum up the individual LDOS in the OUTCAR-file I find that about 90 electrons are missing. As far as I understand there will be a loss from the interstitial regions, but is it reasonable that roughly 10 % isn't accounted for ?

The same thing happens if I put one single B-atom in a large supercell (10x10x10 A) and calculate the the LDOS with LORBIT=12. The LDOS then accounts for 1.6 of the 3 electrons. I've also used LORBIT=2 together with different RWIGS-values. Then I manage to increase th total charge accounted for up to 2.7 instead. I 'm now wondering if the LORBIT=12 is expected to produce these low values or if the reason may lie in the POTCAR-files used.

Best regards

Dan Fors

[admin]

if LORBIT=12 is taken, all settings (RWIGS, projectors) are taken from POTCAR, i.e. the integration volume of the sphere centered at the atom is fixed. So if there is considerable charge (transfer) outside that sphere, this charge is not included in the integration volume.

[cacarden]

I am used to compute the DOS from contributions from LDOS (LDOS= Sum over atoms (LDOS(atom i) and compare it with the DOS computed from the actaul DOS table. If both curves belong to the same family ( i.e differ only by a constant) i supposed that the RWIGS-Values are good enough for representing the whole space. I have to say that I do that in hard-Ionic systems, where de KS states are well localaized around ions. Nevertheless, in covalent and metallics systems a good partition of the space in the bonding (in teratomic) regions is more critical. For me, the ideal situation is condensed in the basins defined by a topologial analysis of the electron density (AIM theory of Bafer).

The best,
Carlos Cárdenas