>just a couple of questions and some suggestions:
I am trying to describe S0 state. This geometry was obtained by S0 state-specific CASSCF (12,11) optimization based on previously obtained PBE hybrid optimized geometry. That was done with def2-SVP basis set for both cases. The results of XMCQDPT calculations were pretty much the same for both geometries (obtained from DFT and CASSCF) for this conformation. I then decided to stick with this particular geometry to check different effects on the results.
Actually I have also optimized S1 state with state-specific for S1 CASSCF and performed S0,S1 state-averaged XMCQDPT with obtained geometry then to estimate the Stoke's shift and got the value of ~20 nm, which is pretty good. But overall error was the same for fluorescence spectra. I am interested in emission spectra too (maybe even more than in absorption), I just decided to clarify the situation with excitation first because it doesn't require MCSCF optimization for each calculation.
From the experiment there should not be any significant influence from the nature of side-chain substituents, and the other equilibrium structures may refer only to different conformations of this “side-chain hand”, as the main Pi-system is rigid. I suppose that for gas phase the conformation could significantly affect the spectral properties since the main Pi-system would interact with Pi-orbitals of succinimide fragment, for example. This is why I excluded Pi orbitals of succinimide fragment from active space.
I performed S0,S1 state-averaged PCM (water) calculation for this conformation. It actually led to a good result. But it was done in def2-SVP basis set (with def2-SVPD I had a problem with crashing of Firefly which is now solved thanks to you:)), and I think I would perform PCM calculation again later. The reason why I decided to stick without solvation modeling for now is that for some other systems for which I obtained a good agreement with experiment using a gas phase calculation I got a big error after addition of PCM water modeling. But on the other hands those were carboxylic anionic structures (e.g. fluorescein dianion which I will study extensively too), for which an explicit description of water molecules in proximity of anionic center is probably needed. I planned to create another topic related to PCM and absorbance/fluorescence spectra prediction later :) For that I am planning to optimize geometry with one ore two honest water molecules around ionic centers while also using PCM as the framework, and then to perform PCM calculations of such “solvates” with CASSCF/XMCQDPT. Does it make sense? As for the salvation of Cy3-NHS ester – in experiment it is considered that these structures (indocyanine green derivatives) greatly change spectral properties upon aggregation, magnitude of which is greatly affected by the nature of solvent, while in highly diluted solvents (or when aggregation is suppressed somehow) the results are close, as far as I know. And of course aggregates by no means can be considered as the same molecules, they are actually different substances from point of view. I highly doubt that there is data of gas phase spectral properties for them. I want to clarify also, that I am performing calculations for the reasons of learing right now, not that I have some particular interest in exact Cy3-NHS ester.
I will check the degeneracy of S1 and S2 by performing calculations with def2-SVPD basis set with (1,1,1,0) averaging of three states and NSTATE=12. I will also perform several calculations now to study the system further, and as soon as I got some meaningful set of results, I will report it here. I think that it really could help future non-experienced beginners.
I understood that it is better to stick with state-averaging calculations rather than with state-specific, thank you! But I still wonder how someone should understand when the ground and excited states are very different, so the averaging will really produce bad orbitals for both states? As far as I understand, equally good for several states orbitals from state-averaged calculation can be also called equally bad with the same level of integrity, isn't that so?
Finally, thank your for mentioning that XMCQDPT can run in parallel now, I really didn't know that, that's great news!!!