Enhancing chemistry-intuitive feature learning to improve prediction performance of optical properties

Emitters have been widely applied in versatile fields, dependent on their optical properties. Thus, it’s of great importance to explore a quick and accurate prediction method for the optical properties. To this end, we developed a start-of-the-art deep learning (DL) framework by enhancing chemistry-intuitive subgraph and edge learning as well as coupling the prior domain knowledge for classic message passing neural network (MPNN) such that can more sufficiently capture the structure feature associated with the optical property from the limited dataset. Benefiting from the technical advantages, our model significantly outperforms eight competitive ML models used in five different optical datasets, achieving the highest accuracy to date in predicting four important optical properties (the absorption wavelength, emission wavelength, photoluminescence quantum yield and full width at half-maximum), showcasing its robustness and generalization. More importantly, based on our predicted result, one new deep-blue light-emitting molecule PPI-2TPA is successfully synthesized and characterized, which exhibits close consistence with our prediction, clearly confirming the application potential of our model as a quick and reliable prediction tool for the optical property of the diverse emitters in practice.