Restrictions: (avoided) crossings between excited-states

The numerical calculation of excited-state gradients has a number of advantages, but also a possible problem: If the step size is chosen too large, or if there are close-lying excited-states, then the order of the excited states can change. For such cases, the excited-state gradient method to estimate relative RR intensities is not reliable: If states with different electronic character (but of the same symmetry) are close in energy, this will cause an avoided crossing. If the numerical derivatives are, in this case, computed w.r.t. the adiabatic states, they will probably not reflect the true situation. Especially if the coupling matrix elements between the two excited states is small, the spectroscopic properties often behave as if there is no avoided crossing, i.e., according to the diabatic states. Such cases should be handled with extreme care, since it is often not possible in advance to see whether the adiabatic or the diabatic picture should be invoked.

Because of the possible (avoided) crossings the user must make sure that always enough excited states are calculated to include the state(s) of interest. E.g., if the resonance Raman intensities are required for the first excited state, also some higher excited states have to be included in the excitation calculation, as the first excited state at the ground-state equilibrium might be higher in energy for displaced structures.