Luminescent doublet-spin organic radicals are unique functional materials for applications in organic light-emitting diodes because they avoid energy-wasting intersystem crossing from singlet to triplet. However, electronic structure calculations of conjugated organic radicals are challenging and remain to be explored due to the potentially strong multi-configurational characters even at the ground state. In this work, we applied our stacked ensemble machine learning model, and predicted the molecule-dependent range-separation parameter (ω) in our exchange-correlation functional, ML-ωPBE, for over 50 doublet-spin conjugate organic radicals in 4 classes. Interestingly, we accomplished a small mean absolute error (MAE) of 0.0214 a0-1 when radical molecules are absent from the training set, demonstrating excellent generalizability and transferability of our model. In addition, we evaluated their photophysical properties using single-configurational time-dependent density functional theory (TDDFT), and compared the performance of ML-ωPBE with more than a decade of 10 functionals. Our ML-ωPBE exhibited outstanding behaviors in absorption and fluorescence energies for most organic radicals, with MAEs of 0.22 eV and 0.12 eV, respectively. Our work illustrated a successful extension of the ML-ωPBE approach to doublet-spin conjugate organic radicals.

Photophysical properties of organic radicals based on machine learned range-separated density functional

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.