Artificial Decision Tree Guided Screening of Ultralong Room-Temperature Phosphorescent Cocrystals with Locally...

High-throughput screening of ultralong room-temperature phosphorescence (RTP) systems is essential for advancing photofunctional materials. Among molecular solids, charge-transfer (CT) cocrystals offer great potential due to the wide range of electron donors and acceptors with tunable energy levels. However, realizing RTP in CT systems remains difficult, as intricate orbital interactions in donor-acceptor pairs often lead to spin-forbidden singlet-to-triplet transitions. To overcome this challenge, we developed an artificial-decision-tree method guided by theoretical calculations and introduced an energy-based descriptor (E) to streamline the design of RTP-active cocrystals. This descriptor enables CT-dominated singlet emission alongside locally excited triplet emission, effectively promoting intersystem crossing. Applying this strategy, we predicted and experimentally validated new cocrystals (TP&1,2-TFP and TP&1,4-TFP), which achieved RTP lifetimes up to 1.223 s, the longest reported among current CT cocrystals. We further demonstrated the versatility of this strategy by identifying additional RTP-active systems (TP&1,2-TCP and TP&1,4-TCP). Owing to their tunable emission and time-resolved RTP, these materials were successfully employed in digital and visual information encryption and time-gated optical logic gate applications. Therefore, this work presents a robust framework that combines machine learning-driven prediction with experimental validation, offering a systematic and efficient alternative to traditional trial-and-error methods in the discovery of luminescent materials.