You get variable <S**2> because you use the unrestricted approximation. For a triplet, the exact value of <S**2> is 2; any value higher than that indicates contamination from higher states. There is no easy or cheap way out of this. You could use restricted open shell calculations, RO instead of the default U (not sure about Gaussian, I’ve done those in Jaguar), but in my experience, energies are less reliable even though you’re guaranteed pure spin states. For ground state calculations, you can project out most of the higher spins in the unrestricted solution, but again, I’m not sure the method is available for TDDFT. For accurate identification of “real” excited states, I’ve used CASSCF/CASPT2, but even there, the energies from unrestricted TDDFT can be as good or better. I did find the CASPT2 calculations good for evaluating which of the DFT excited states were “real”, and which were artifacts in the unrestricted approximation. Not published, just experience from a long project we never managed to finalize.
Standard TDDFT calculations calculate singlet->singlet excitations. In Gaussian, I can see that for each excitation, <S**2>=0.000.
While I calculate molecules like O2 which the ground state
is triplet, I need calculate triplet->triplet excitations by assigning
Thank you very much!
Confidentiality Notice: This message is private and may contain confidential and proprietary information. If you have received this message in error, please notify us and remove it from your system and note that you must not copy, distribute or take any action in reliance on it. Any unauthorized use or disclosure of the contents of this message is not permitted and may be unlawful.