Dear Pierre,
Which density functional do You
use? Some of them might give quite reasonable results, whereas other might not,
especially if your molecule contains long conjugated chain. 1 eV does not seem
to be too few for me, but in no way I could say that I am high-valued
professional in the field. There are sometimes artificial bound-electron states,
then electron is placed in very high levels...
HF eigenvalues are non-naturally positive in most cases,
because at HartreeâFock level electron correlation is not treated. CI and
coupled-cluster methods solve this by using many-determinant wavefunctions,
describing electron correlation via admixing of the excitated states to the
ground state studied, so results You obtained from them (especially from CCSD(T)
) seem to be OK in most cases (if You have used basis set that is sufficiently
large). Is the bound electron on some very high level also in CCSD(T) and CI
case? If yes, this may indeed be the effect of artificial binding...
Regards,
Igors Mihailovs,
Institute of
Solid State Physics,
University of Latvia