Khimich
khimich@kinet.chem.msu.ru
>Firefly uses old CI vectors save in CIVECTR file (provided this file
>exists and is valid) as an initial guess vectors in CI calculations.
>Actually, you can find this remark in your outputs. Thus, it is not
>surprising to get different states starting from the scratch and
>from converged CI vectors of MCSCF procedure (and especially of
>MCSCF procedure which was interrupted). Note, the MOs you are
>using are very far away from the optimal ones, so one can apriori
>expect dramatic changes of spins of the low-lying CI roots during MCSCF.
>Another tricky point is that exetyp=check runs save non-initialized
>dummy CI vectors in the CIVECTR file so that the restart from this
>vectors can result in weird irreproducible behavior. This potential
>issue was fixed in the recent builds of Firefly.