ccl:g98 summary

 2 weeks ago I posted the following problem:
 > I did a UHF/LANL2MB SP calculation with G98 on a triplet state. After
 > SCF has
 > converged i found that the energie of the alpha HOMO was higher than the
 > alpha LUMO. The same for beta orbitals.
 > Alpha  occ. eigenvalues --   -0.22532  -0.21631  -0.08415
 > Alpha virt. eigenvalues --   -0.10304   0.00737   0.01977   0.03175
 > Beta  occ. eigenvalues --   -0.09456
 > Beta virt. eigenvalues --   -0.10780  -0.07132   0.02100   0.02998
 > What happened? Is it possible that the VShift has something to do with it?
 > I could not set it to zero even with SCF=(VShift=0). Also the symmetry of
 > the orbital is not found correctly at the end of the SCF.
 > Does anybody know why this happened?
 I got 2 responces:
 1. Perhaps you're studying transition metals. The energy inversion
    is very often In those cases when the vshift is activated.
    Did you try iop(5/10=-1)
    in the route section to disable vshift?
 2. There are a number of possibilities.  1) you may have failed to converge
    to the lowest triplet due to symmetry.  2) vshift has trapped you into
    a higher state 3) the triplet is quite high lying.  In looking at the
    eigenvalues the fact that you have a very negative "virtual" I
 would agree
    that 1) or 2) see most likley.
    From the checkpoint file you can try GUESS=(Read,Alter) to switch
    these orbitals.  Another alternative is to use STABLE=OPT which will
    actually solve for the stability of the wavefunction and re-optimize
    the SCF if it finds a lower state.
 I tried IOP(5/10=-1) to set VSHIFT=0 and found that the
 occ. orbitals were lower in energy than the virt. orbtials as one would
 like to have it. So  indeed it seemed to be a problem of VSHIFT with
 transitions metals.
 | Bernd Rieger                         Phone:089 289 13768           |
 | Physic Department T38, TU Muenchen   Fax:  089 289 12444           |
 | James-Franck-Strasse                                               |
 | 85748 Garching                email: brieger : at : |