Overlapping PAW Spheres

[juanquij0te]

I am trying to understand some of the details of the projector functions used by VASP in the PAW method.

All of the radii that are listed in the POTCAR files are large enough that normally bonded atoms would have overlapping radii (e.g. H2 would have overlapping RCORE). While VASP is able to get good results for bond-lengths and energies for these systems in spite of overlapping radii, are the PAW PDOS still trustworthy?

Specifically, in the documentation for LORBIT it is claimed that "This information can be used to construct e.g. the partial DOS projected onto molecular orbitals" - is this still true for the PAW projection scheme (LORBIT >= 10)?

I see if I graph the pseudo wavefunctions (from POTCAR) for various elements that some of them show cutoffs (they drop suddenly to 0) that cannot be matched to any of the listed cutoff radii at the beginning of the POTCAR file (while others show no apparent cutoff). For example, for C, these cutoffs happen at 0.93 bohr? Is this the true cutoff used for the pseudo wavefunctions, and do the projector functions also cut to 0 at these values (or perhaps those functions shown to cut to 0 are the projector functions)?