Advancing efficiency in deep-blue OLEDs: Exploring a machine learning-driven multiresonance TADF molecular design

Abstract

The pursuit of boron-based organic compounds with multiresonance (MR)-induced thermally activated delayed fluorescence (TADF) is propelled by their potential as narrowband blue emitters for wide-gamut displays. Although boron-doped polycyclic aromatic hydrocarbons in MR compounds share common structural features, their molecular design traditionally involves iterative approaches with repeated attempts until success. To address this, we implemented machine learning algorithms to establish quantitative structure-property relationship models, predicting key optoelectronic characteristics, such as full width at half maximum (FWHM) and main peak wavelength, for deep-blue MR candidates. Using these methodologies, we crafted nu-DABNA-O-xy and developed deep-blue organic light-emitting diodes featuring a Commission Internationale de l'Eclairage y of 0.07 and an FWHM of 19 nm. The maximum external quantum efficiency reached ca. 27.5% with a binary emission layer, which increased to 41.3% with the hyperfluorescent architecture, effectively mitigating efficiency roll-off. These findings are expected to guide the systematic design of MR-type TADF clusters, unlocking their full potential.