Posted by Ignacio Fdez. Galván on October 15, 2010 at 10:39:51:
Hello,
I'd like to ask for some advice on optimizing specific electronic states. The key here is following the right state as the geometry changes and the states get mixed and cross. In those cases, just optimizing the "nth root" is not a good strategy. What else can one do (in particular, when there is no symmetry)?
In Molcas 6.4 there was the option CIselect in RASSCF, that would select roots according to CSF contributions, but last time I tried this selection was not passed to the gradient calculation, at least not for state-average calculations. It seems this option does not exist in Molcas 7.4, but I see the option HOME, does this serve the same goal? Nevertheless, it can happen that the orbital order changes during an optimization, or that they get "mixed" and lose their initial character, so just looking at the CSF contributions might not be the solution.
In some cases, it is possible to identify an state, or distinguish between some of them by checking properties such as dipole moment, Mulliken populations, oscillator strengths... but this is also not perfect, and these properties can change a lot as the geometry change.
What other strategy could one use? Is it possible to have something like a "wavefunction overlap" or "electron density overlap"?
Thanks.