Isolated molecules & unoccupied orbitals: does the basis set matter?

[ferdl]

When comparing the DOS of isolated molecules as obtained by VASP with the results of other codes which use local basis sets (Siesta, Turbomole), it seems that the states match only up to a certain energy below the vacuum energy. For higher energies (but still below the vacuum energy), neither the number nor the position of the orbitals match necessarily.

I focused on the molecule 4-Pyridin-4-yl-phenylamine. Using parameters as comparable as possible, I see 4 unoccupied states below the vacuum energy with Siesta and Turbomole (at approximately the same energies). Using VASP, the DOS matches those results very nicely up to the LUMO+2. 3D-representations of the orbitals are consistent as well.

Between the states LUMO+2 and LUMO+3 (according to Siesta and Turbomole), VASP calculates an additional state. By looking at its 3d-representation I can see that this state looks like what is the LUMO+4 (the first state above the vacuum energy) in siesta and turbomole. A very quick test with another plane-wave code, abinit, finds a DOS very similar to what VASP finds.

To sum up, in this example I see

HOMO LUMO L+1 L+2 L+3 VacEn. L+4

in siesta and turbomole. Vasp finds:

HOMO LUMO L+1 L+2 ex L+3 VacEn.

where the peak "ex" corresponds to L+4 from siesta and turbomole, and VacEn is the vacuum energy. From 3d representations of the L+3 state I don't think they represent the same orbital.

Is it possible that plane-wave and local-basis set codes behave somehow differently concerning unoccupied orbitals? If so, why?

There is a related thread,
http://cms.mpi.univie.ac.at/vasp-forum/ ... php?4.6181
There, however, it seems that the calculation of states above the vacuum energy was the issue (which, I guess, could in principle be solved by setting NBANDS accordingly). I want to emphasize that I see (for at least three molecules i looked at so far) a different DOS also in a certain range of unoccupied states below the vacuum energy.

Thanks in advance,
Ferdinand

[physicalattraction]

For the simple H2O molecule, I find the same.

With Siesta: eigenvalues are -24.74, -12.70, -8.71, -6.22, 1.68, 4.09, 9.85 11.55 and 21 higher ones.
With Vasp: eigenvalues are -24.81, -12.62, -8.84, -6.74, -1.24, 1.43, 2.25, 2.85, 3.65 and 3.85.