Perturbation and Time-Dependent DFT¶
This was very complex, so here are some takeaways:
- TD-DFT is generally better for excitations than CI due to the direct inclusion of correlations, but is naturally very functional/situation dependent.
- Charge transfer transitions are not well modelled using TD-DFT. This is due to self-interaction issues, causing lower energy excitations.
- TD-DFT works best with a bit of HF exchange (typically ~25%) - e.g. B3LYP, PBE0
- Certain functionals are better for Rydberg states, when they account for longer range behaviour.ß