In post-hartree fock (HF) calculation, such as CASSCF and CASPT2, selection of active orbital and active electron is difficult. The following steps are suggested based on some experienced that I have learnt from workshop on multireference configuration interaction (MRCI) by Hans Lischka and the Columbus team in Tianjin, China.
The most important point is that "both active orbitals and active electrons must involve in bond breaking/forming of your intersest".
Procedure:
1. The active space should contain all molecular orbitals (MOs) that have the same character as the "HOMO" or the orbital that you think is important and involving, says bond-breaking's orbital. For example, in a pi system, you should look at all orbitals with pi character (pi bond).
2. Active orbital should include MOs that are close to each other in energy.
3. Look at the occupation numbers of the natural orbitals. If they differ from 2 (fully occupied) or 0 (unoccupied), they signify it is an important MO that should be include to the active space.
4. Check the MOs of the starting orbitals with that of the RASSCF solution. They should be similar in the nature.
5. Look at twisted molecule geometries to highlight configurations that may become important in deformations.
6. For Doing a CASPT2 calculation, if there are states that are not included in the active space, but that are close in energy to the reference state, the CASPT2 may diverge and complain of intruder states. You should try to include the intruder state into the active space.
Read more:
https://www.molpro.net/info/2012.1/doc/quickstart/node18.html
https://sites.google.com/site/orcainputlibrary/cas-calculations/tutorial-active-space-selection
http://molcas.org/documentation/manual/node34.html
http://chemistry.stackexchange.com/questions/54108/typical-runtime-for-casscf-calculations
Rangsiman Ketkaew