I finished investigating the influence of the NO bond length (symmetric stretch) onto excitation energies at ADC(3)/SVP level of theory! The result is as expected: due to the population of the π* orbital at the nitro group (see first post concerning nitrobenzene) in all lower excited-states, the influence of this bond length is quite large in the following way: along the symmetric NO stretching mode, the energy of the ground state largely increases, while the absolute energies of the excited states decrease (see Figure 1).
For the vertical excitation energies (ω) these effects add up and the ω of all states drop rapidly by about 0.1 eV/pm (Figure 2), which explains for the large difference between CC2 geometry with 125 pm NO-bonds and the CCSD geometry with 120 pm NO-bonds.
Now I know why and how strong excitation energies depend on the NO bond length, but one very important question remains: Why is the ADC(3) result for the excitation energy of the lowest excited singlet state (almost 4.1 eV) so far off from the experiment (3.65 eV)? From a third order method like ADC(3) one would expect 0.1-0.3 eV deviation at max, in particular for singly excited states. Maybe the reason is lying in the dynamics of the molecule, cause in the experiment, this bond vibrates like any other bond. I will have a look into this using born-oppenheimer ab-initio molecular dynamic in the next post.