Design of Thermally Activated Delayed Fluorescence Materials with High Intersystem Crossing Efficiencies by Machine Learning-Assisted Virtual Screening

Efficient intersystem crossing (ISC) and reverse ISC (RISC) processes are of vital significance for thermally activated delayed fluorescence (TADF) materials to achieve 100% internal quantum efficiency. However, it is challenging to rapidly predict the ISC/RISC rates of large amounts of TADF materials and screen promising candidates because of their flexible molecular design. Here, we perform virtual screening of 564 candidates constructed from 20 unique building blocks linking in D-A, D-pi-A, and D-A-D (D') configurations using the established machine learning models of GBRT and RF-GBRT-KNN with the Pearson's correlation coefficients (r) of 0.89 and 0.82, respectively. Novel descriptors of Delta ELL, Polar, and Delta ETT for predicting ISC/RISC rates were proposed, and nine TADF molecules with the predicted ISC and RISC rates of >7 x 10(7) and 2 x 10(5) s(-1), respectively, were revealed. We provide an efficient approach to predicting ISC and RISC rates of TADF molecules on a large scale, elucidating important building blocks and architectures to design high-performance optoelectronic materials for experimental explorations.

Automated summary

This study utilizes machine learning models to predict the ISC and RISC rates of TADF molecules, proposes novel descriptors for prediction, and identifies candidate TADF molecules with high ISC and RISC rates.