How to Accurately Predict the Fluorescent Property of Multiresonance TADF: Density Functional Theory or Machine Learning?

Organic thermally activated delayed fluorescence (TADF) emitters have attracted considerable attention in organic light-emitting diode (OLED) applications, owing to their potential for 100% exciton utilization. Among them, multiple-resonance (MR) TADF compounds have emerged as a particularly promising class of emitters, offering superior color purity and high photoluminescence quantum yields (PLQYs) compared to conventional donor-acceptor-type TADF materials. With the increasing demand for ultrahigh-definition display technologies, exemplified by the BT.2020 color gamut standard, the development of MR-TADF materials featuring narrowband emission and high efficiency has become a key research priority. In this study, we systematically assessed the predictive capabilities of density functional theory (DFT) and machine learning (ML) methods in determining the luminescence properties of MR-TADF materials. While ML models offer rapid prediction capabilities, their generalization ability remains limited, primarily constrained by the quality and size of the available training data set. In contrast, DFT-based approaches, although more computationally demanding, exhibit reliable accuracy and generalization performance when calibrated using appropriate regression models. Using a B3LYP-based calibration model, we further predicted the optical properties of the designed MR-TADF molecules and identified a promising candidate exhibiting deep-blue emission. Overall, the findings underscore the complementary roles of DFT and ML approaches in MR-TADF research and provide valuable theoretical guidance for the rational design of high-performance MR-TADF emitters tailored to next-generation OLED applications.