CCL: Unoccupied orbitals, eigenvalues

[Mikael Johansson <mpjohans~at~pcu.helsinki.fi>]

 Hello All!
Catalysed by the latest in the row of experimental observations of
 molecular orbitals (which nicely also measures the phase of the
 orbital)
 [1], I again started pondering on the reality of the _unoccupied_ MO's,
 this time especially between the qualitative difference (if any)
 between
 unoccupied orbitals with negative energies versus those with positive
 energies.
Positive orbital energies for occupied orbitals indicate the
 non-existence of the orbital, or at least the inability of the chosen
 computational method to give a proper description of it (a larger, more
 diffuse basis set will just throw the electron, "orbital",
 away from the
 molecule) [2].
Assuming for the moment that also unoccupied (virtual) orbitals have
 physical meaning, would then the ones with negative energies (if there
 are any for a give species) be "more real" than the ones with
 positive
 energies? I have not come across a disussion on this, or at least don't
 remember having done so.
 [1] J. Itatani et al., "Tomographic imaging of molecular orbitals",
 Nature
     432 (2004) 867-871.
 [2] I like the discussion in: N. Rösch, S.B. Trickey, J. Chem. Phys. 106
     (1997) 8940-8941.
 Merry Christmas,
     Mikael J.
     http://www.helsinki.fi/~mpjohans/