Machine Learning-Guided Tailored Synthesis of Single Room-Temperature Phosphorescent Carbon Dots.

Phosphorescent carbon dots (CDs) have attracted considerable interest due to their unique photophysical properties and broad applicability. However, customizing their afterglow emission remains a major challenge. Here, we propose a machine learning-guided strategy for tailored synthesis of single room-temperature phosphorescent CDs from complex precursors, allowing programmable control over phosphorescence emission wavelengths and lifetimes. These single CDs are efficiently prepared via microwave-assisted heating by tuning precursor compositions and subsequent separation from the in situ confined domains. Experimental findings reveal that the nitrogen dopant from urea governs the energy-level structure of CDs, while a self-protected superficial amorphous rigid network formed by sodium hydroxide and urea can stabilize triplet excitons, collectively enabling precise regulation of phosphorescence emission. Leveraging machine learning to model these multivariate relationships, we achieve on-demand phosphorescent CDs with tunable emission wavelengths from 429 to 586 nm and lifetimes spanning from 3 to 2500 ms. The tailored CDs exhibited excellent performances in time-resolved multilevel encryption, mechanically flexible afterglow devices, and white LEDs with persistent afterglow and outstanding operational stability (T