Your methodology makes sense providing you are interested in
a ground state. If you are interested in excited states as well,
a better methodology would be to run CI, or, better, a several
iterations of state-averaged MCSCF using focas converger with
iexcit set to 3 i.e. to perform CISDT or MCSCF-CISDT. This would
allow to describe ground and single-electron excited states
approximately on an equal footing. Similarly, to properly catch
orbitals required to describe two-electron-excited states within
CASSCF one may use CISDTQ.
All the best,
On Sat May 18 '13 4:14am, Solntsev Pasha wrote
>Please, correct me if i am wrong. I have some high symmetric molecule with transition metal. I am thinking about CASSCF for the molecule. I did DFT calculations and found all orbitals which were proposed by group-theoretical analysis and used them as starting orbitals for following calculations. The active space is too large: 18e19o. Since i am going to use cistep=guga for CASSCF i decided to run cityp=guga for my active space (iexcit=2, nstate=16, wstate=1,-0) for my large active space to estimate the population of the orbitals. I am interesting mainly in ground state, but of course low lying excited states also will be considered. After cityp=guga run the population of the natural orbital in atomic basis was analyzed. I found that population of two the lowest orbital and two the highest orbitals within my active significantly differ from the other orbitals. Does it mean i can remove two valence and two virtual orbitals from the active space and thus contract my active space 18e19o->14e15o for the casscf run? Does this methodology make sense?