Predicting External Quantum Efficiency of Red Phosphorescent Organic Light-Emitting Devices by Machine Learning

As the core material that affects the performance of OLED devices, the demand for performance improvement of OLED materials is urgent. However, due to the molecular diversity of organic compounds, traditional methods of development are inefficient and expensive. Recently, the machine learning (ML) approach has attracted increasing attention in the field of organic luminescent materials, which can learn from the existing results and provide the relation between the input features to the output performance. By constructing the relationship between molecular structure and device performance of OLED, the efficiency of material development is higher and guidance for material design is better provided through the identification of molecular key factors. In this work, we attempt to use the ML approach to explore the quantificational relation between the external quantum efficiency (EQE) of red phosphorescent organic light-emitting devices (OLEDs) and their material and device structural factors, aiming to screen out the key factors governing the EQE and predict EQE values directly by molecular structure. We established the dataset based on over 1000 device data from experiment, and reduced the number of molecular descriptors to below 35. Currently, the root mean squared error (RMSE) of test set has been lowered to 3.21%. These results provide essential guidance for material screening and experimental device optimization. On this basis, we further designed new red host materials, which have higher EQE, up to 27.56%. It has reached the first-class level of commercial materials.