>In addition, I'd recommend to consider MP2 natural orbitals
>or state-averaged CIS natural orbitals as the initial guess for MCSCF.
>This Forum is fully searchable through "Search" button on the left
>and there should already be some samples uploaded here.
>On Wed Sep 26 '12 3:43pm, sanya wrote
>>>In the refs manual of GAMESS-US, I found the following advises:
>>>1)One way to improve upon the SCF orbitals as starting MOs is to generate valence virtual orbitals (VVOs).
>>>However, the VVOs is not implented in firefly. Can I reuse the calculated VVOs by GAMESS-US in firefly?
>>Yes, the format of $VEC groups in GAMESS and FireFly is the same. But be careful with such options as d5=.t. in FireFly and its analog ispher=1 in GAMESS (be sure both are set true or false simultaneously, i.e., if you use ispher=1 in GAMESS, you must set d5=.t. in FireFly, and vice versa)
>>>2)Although they said "Pasting the virtuals from a MVOQ run onto the occupied orbitals of a SYMLOC run (both can be done in the same SCF computation) gives the best possible set of starting orbitals.", I'm confused about this because they also said " An alternative choice, not usually as good, are the modified virtual orbitals (MVOs)."
>>There are several ways of choosing starting MOs for CASSCF. You can use localized natural orbitals (Alex can describe this method better) or even use DFT MOs. You can simply reorder orbitals to bring all necessary MOs to the active space. If the problem has a stable MCSCF solution, all the choices lead to the same solution.
>>>The 3rd question: Can I use the converged DFT orbitals as the start orbitals for MCSCF?
>>It may be a good choice.