Either accurate singlet-triplet gaps or excited-state structures: Testing and understanding the performance of TD-DFT for TADF emitters

The energy gap between the lowest singlet S1 and triplet T1 excited states ∆EST is a key property for thermally-activated delayed fluorescence (TADF) emitters. Time-dependent density functional theory (TD-DFT) is widely used to predict this gap and study TADF in general, despite its well-known shortcomings concerning charge-transfer (CT) states. Furthermore, polar CT states strongly interact with their environment, whose treatment is another issue when in quantum-chemical excited-state methods. Addressing these two major challenges, this work explores the performance of Tamm-Dancoff-approximated (TDA)-DFT for the calculation of a large set of ∆EST values included in the recent STGABS27 benchmark set. To this end, we employ a wide range of approaches to include orbital relaxation, environmental effects, and structural relaxation. Surprisingly, the best-performing approach for the specific task of predicting ∆EST heavily relies on error cancellation to mimic environmental and orbital-relaxation effects, using functionals with a low amount of Fock exchange of ≈ 10 % (e.g., TPSSh) in combination with ground-state structures. Generally, however, this approach is not very useful as it provides systematically wrong excitation energies, excited-state structures and state characters. To some extent, similar issues are observed with all studied TDA-DFT approaches. We thus conclude that for the description of CT states in dielectric environments, TDA-DFT is not competitive with the recently presented ROKS/PCM approach regarding robustness, accuracy, and computational efficiency.