CCL: Can CASSCF energy be a descriptor for the stability of active spac

Dear Mariusz,
 Thank you for the answer.
 You are right, extending the active space is the most elegant approach
 here, so maybe I should give additional thoughts on this. In the meantime I
 will look at the paper you cited.
 Thank you very much,
 Best regards.
 On Mon, Aug 23, 2021 at 10:11 PM Mariusz Radoń mariusz.radon]_[uj.edu.pl
 <
 owner-chemistry[A]ccl.net> wrote:
 >
 > Sent to CCL by: =?utf-8?B?TWFyaXVzeiBSYWRvxYQ=?= [mariusz.radon],[
 > uj.edu.pl]
 >
 >
 > > On 23 Aug 2021, at 08:58, Muhammed mbtemiz3_-_gmail.com
 > <owner-chemistry(a)ccl.net> wrote:
 > >
 > > Dear Nuno,
 > >
 > > Thank you for the answer.
 > >
 > > You are right, having a balanced AS is not an easy thing. And a
 matching
 > AS is needed for all spin states. However my problem is a bit more involved
 > than that. I have a matching AS for all spin states, sure, but there is an
 > additional solution - which does not include the orbitals I am after - with
 > a lower energy for one of the spin states. What I am wondering is, does the
 > lower energy solution tell me anything about the stability of the
 > wavefunction.
 > >
 > > To simply put, does the lower energy - wrong AS for this particular
 > problem - solution describe the system better and should I always aim for
 > lower energy solutions?
 >
 > Dear Muhammed:
 >
 > I agree with your point that you should not use a CASSCF solution yielding
 > wrong final orbitals in your active space (i.e. different character of
 > orbitals than for other spin states and incompatible with the nature of
 > your chemical problem), like it happened in your system for the doublet
 > state. Even if it gives a lower variational energy at CASSCF level than the
 > “right” solution.
 >
 > In this sense, yes: you should not always aim lower energy solutions, but
 > rather you should aim compatible and “chemically reasonable”
 solutions for
 > different spin states. This is also true if you go to post-CASSCF
 > treatment, e.g. CASPT2 or NEVPT2 (which you probably should do in order to
 > account for remaining correlation effects).
 >
 > Enlarging the active space (so that it includes both orbitals which tend
 > to rotate with each other) may be the most elegant solution to problems of
 > this sort. However, if you say that this is not feasible in your case, I
 > would prefer to stick to the common active space (identical character of
 > active orbitals for all states in which you are interested).
 >
 > > From my experience, problem similar to yours are quite common for
 > transition metal complexes when some chemically relevant ligand-based
 > orbitals occasionally rotate into metal outer-core orbitals (3s or 3p for
 > first-row metals). This simply means that correlating these outer-core
 > orbitals at the CASSCF level leads to lower variational energy, but it does
 > not mean that the such obtained solution represents a better starting point
 > for subsequent CASPT2/NEVPT2 calculations (especially if the orbital occurs
 > only in one state). Sometimes the problem can be solved by extending the
 > active space with the problematic outer-core orbitals, but this is not
 > always possible. Some discussion of a particular case where such issues
 > were observed and how the problem was solved, you can find in our paper:
 > 10.1021/acs.jctc.8b00200, page 4012/4013.
 >
 > Best wishes,
 > Mariusz
 >
 >
 >
 > --
 > Mariusz Radon, Ph.D., D.Sc.
 > Assistant Professor
 > Faculty of Chemistry, Jagiellonian University
 >
 > Address: Gronostajowa 2, 30-387 Krakow, Poland
 > Room C1-06, Phone: 48-12-686-24-89
 > E-mail: mradon(a)chemia.uj.edu.pl (mariusz.radon(a)uj.edu.pl)
 > Web: https://tungsten.ch.uj.edu.pl/~mradon
 > ORCID: https://orcid.org/0000-0002-1901-8521
 >
 >
 >
 > -= This is automatically added to each message by the mailing script =->
 >
 >
 --
 Muhammed Buyuktemiz
 Chemistry Department, Gazi University
 +90 554 844 11 25