My Community

Unocupied states of molecules in VASP looks quite different than in some other DFT codes which use local basis set. In VASP all virtual states has energy not much above vacuum level and there is no clear evidence of non-bondig...pi-antibonding...sigma-antibonding structure typical for molecules, which can be seen in local-basis-set packages and which is natural from chemical point of view.

I know that Coopmans theorem does not stands for DFT and Unocuied states has (strictly speaking) no physical meaning in DFT. But still it's quite common to use DFT virtual molecular orbitals for estimation of some properties.

Is there any principial difference between virtual molecular obrbitals from VASP and from local-basis set DFT software (like Gaussian, Fireball).

especially the difference is very big for molecules with large ban-gap if virtual state is above vacuum level

for example

acetylene (C2H2) gap (LUMO under vacuum level)
vasp 6.78eV
local-basis 7.3eV

ethylene (C2H4) gap (LUMO under vacuum level)
vasp 5.74
local-basis 6.08

ethane (LUMO above vacuum level)
vasp 7.26 !!!!
local-basis -12.34 !!!

One of the biggest impacts on the relative LUMO energies are how the pseudopotentials are generated, which will vary from code to code. I don't think that the fact that VASP is a plane wave code should effect the relative LUMO energies since it's based on the variational principle.

modey3

I thing I have an answer already:

If I have small molecule in large vacuum box (cell), and if the first unoccupied antibonding state of the molecul is above vacuum level, than vacuum level (state of free electron in the vacuum) is LUMO of this system, not the first antibonding state of molecule.

So vasp get correct LUMO state (the vacuum level) of the system, which can not be get as a linear combination of atomic-like basis functions because they do not cover the empty space (vacuum) of the cell. This state however is not an antibonding state of molecule. Local basiss set program get the first unoccupied antibonding state as LUMO, which however is not real LUMO of the system if vacuum is considered.