Unveiling high-performance hosts for blue OLEDs via deep learning and high-throughput virtual screening

Organic light-emitting diodes (OLEDs) hold immense potential for next-generation display and lighting technologies, creating an urgent need for the development of advanced materials that can enhance both efficiency and performance. While the progress in deep learning (DL) and high-throughput virtual screening (HTVS) have significantly accelerated in-silico material design in predicting target properties and generating new compounds, they often overlook efficiency metrics that can only be assessed at the device level. To address this gap, we incorporate experimental intuition into the design process of OLED materials using DL-based HTVS. This approach led to the discovery of a bipolar-type host material, K-15, which achieves an external quantum efficiency (EQE) of 30.8 %, marking a 20 % improvement over the benchmark molecule, 9,9′-(4-(Pyridin-2-yl)-1,3-phenylene)bis(9H-carbazole) (2CzPy). Our work highlights the power of integrating experimental insights with advanced computational techniques in identifying high-performance OLED host materials.