EXTENDED ABSTRACT: Machine learning (ML) and Quantitative Structure-Property Relationships (QSPR) are pivotal in material development, enabling rapid prediction of material properties. In this context, we use ML to explore QSPR of organic luminescent materials used in organic light-emitting diodes (OLEDs) for understanding the underlying mechanisms and accelerating the discovery of new promising emitters. Starting from the core properties of luminescent materials that affect the device performance of OLEDs, we have developed a series of QSPR models for thermally activated delayed fluorescent (TADF) emitters. These models include the proportion of horizontal orientation of transition dipole moments (TDM), electronic energy level structures, carbon-nitrogen bond dissociation energies, photoluminescence quantum yield, and more. This allows for high-throughput virtual screening and design of stable and efficient emitters. By employing molecular generation algorithms, we explored molecular structures within a vast chemical space and identified novel emitters with more innovative molecular structures. Additionally, we optimized density functional theory (DFT) calculation strategies through active learning. We also used low-cost computational features as variables to establish predictive models for DFT-calculated parameters, further accelerating the efficiency of DFT in material property labeling. These works provide guidance for the design of advanced organic emitters.

Keywords：Machine learning; OLED; organic emitter; active learning; density functional theory calculation.
REFERENCES:
[1] H Shi, Y Shi, ZLiang*, S Zhao, B Qiao, Z Xu, L Wang* and D Song*, Chem Eng J, 494, (2024) 153150,
[2] Y Shi, H Shi, Y Zhang, X Zang, Z Zhao, S Zhao, B Qiao, Z Liang, Z Xu, L Wang and D Song*, Adv Opt Maters, 12(5),
(2024) 2301768