The film, not the molecule, decides how deep-blue TADF hosts rank on gap and thermal robustness

Deep-blue host materials for thermally activated delayed fluorescence (TADF) are usually ranked on a few molecular numbers, on the assumption that the isolated molecule anticipates the film. We test that assumption for three phosphine-oxide hosts, S1, S3 and S4, by reading protocol-locked descriptors of the dilute solution, the evaporated film, the powder and the bulk side by side, each with its uncertainty, next to a separate density-functional layer. The descriptors fall into families that rank the hosts differently: film-gap and thermal descriptors place the multi-arm S4 host first, emission and order descriptors place the compact carbazole S1 host first, and band-tail and voltammetric descriptors do not separate the two. The split appears between the molecule and the film: the two hosts emit alike as dilute molecules, yet the S4 film absorption edge lies 0.33 eV higher. Calculations validated against the measured infrared and NMR spectra show frontier levels that move far more with the functional than between hosts. Film gap and thermal robustness must therefore be measured and assessed separately from emissive quality.