Luminescent organic semiconducting doublet-spin radicals are unique and emergent optical materials because their fluorescent quantum yields (Φfl) are not compromised by spin-flipping intersystem crossing (ISC) into any dark high-spin states. The multiconguration nature of these radicals challenge their electronic structure calculations in the framework of single-reference density functional theory (DFT) and introduce room for method improvement. In the present study, we extend our earlier development of ML-ωPBE, a range-separated hybrid (RSH) exchange-correlation (XC) functional constructed using the stacked ensemble machine learning (SEML) algorithm, from closed-shell molecules to doublet-spin radicals. We assess its performance for an external test set of 64 radicals from five categories based on the original training set of 3,926 molecules. Interestingly, ML-ωPBE agrees with the first-principles OT-ωPBE functional regarding the molecule-dependent range-separation parameter (ω), with a small mean absolute error (MAE) of 0.0197 a0-1 but saves the computational cost by 2.46 orders of magnitude. This result demonstrates outstanding generalizability and transferability of ML-ωPBE among various organic semiconducting species. To further assess the predictive power of ML-ωPBE, we also compare its performance on absorption and fluorescence energies (Eabs and Efl) evaluated using time-dependent DFT (TDDFT), with nine conventional functionals. For most radicals, ML-ωPBE reproduces experimental measurements of Eabs and Efl with small MAEs of 0.222 and 0.121eV, only marginally different from OT-ωPBE. Our work illustrates a successful extension of the SEML framework from closed-shell molecules to open-shell radicals and willopen the venue for calculating optical properties using single-reference TDDFT.

Luminescent organic semiconducting doublet-spin radicals are unique and emergent optical materials because their fluorescent quantum yields (Φfl) are not compromised by spin-flipping intersystem crossing (ISC) into any dark high-spin states. The multiconfiguration nature of these radicals challenges their electronic structure calculations in the framework of single-reference density functional theory (DFT) and introduces room for method improvement. In the present study, we extend our earlier development of ML-ωPBE, a range-separated hybrid (RSH) exchange−correlation (XC) functional constructed using the stacked ensemble machine learning (SEML) algorithm, from closed-shell organic semiconducting molecules to doublet-spin organic semiconducting radicals. We assess its performance for a new test set of 64 radicals from five categories based on the original training set of 3,926 molecules. Interestingly, ML-ωPBE agrees with the first-principles OT-ωPBE functional regarding the molecule-dependent range-separation parameter (ω), with a small mean absolute error (MAE) of 0.0197 a0−1 but saves the computational cost by 2.46 orders of magnitude. This result demonstrates outstanding domain adaptation capacity of ML-ωPBE among various organic semiconducting species. To further assess the predictive power of ML-ωPBE, we also compare its performance on absorption and fluorescence energies (Eabs and Efl) evaluated using time-dependent DFT (TDDFT) with nine conventional functionals. For most radicals, ML-ωPBE reproduces experimental measurements of Eabs and Efl with small MAEs of 0.222 and 0.121 eV, only marginally different from OT-ωPBE. Our work illustrates a successful extension of the SEML framework from closed-shell molecules to open-shell radicals and will open the venue for calculating optical properties using single-reference TDDFT.

Luminescent organic semiconducting doublet-spin radicals are unique and emergent optical materials because their fluorescent quantum yields (Φfl) are not compromised by spin-flipping intersystem crossing (ISC) into any dark high-spin states. The multi-configurational nature of radical electronic structures challenges computational studies in the framework of single-reference density functional theory (DFT) and introduces room for method improvement. In the present study, we extended our earlier development of a machine-learned range-separated hybrid functional, referred to as ML-ωPBE, from closed-shell molecules to doublet-spin radicals, and assessed its performance for the original training set of 3,926 organic semiconducting molecules and an external test set of 64 organic semiconducting radicals from five categories. Interestingly, for this external test set, ML-ωPBE reproduced the optimal value of ω, the molecule-dependent range-separation parameter, from the first-principles OT-ωPBE functional with a small mean absolute error (MAE) of 0.0197 a0−1 and with a significant save of computational cost by 2.46 orders of magnitude. This result demonstrated excellent generalizability and transferability of ML-ωPBE among a variety of organic semiconducting species. To further assess the predictive power of ML-ωPBE on organic semiconducting radicals, we also compared its performance on experimentally measurable absorption and fluorescence energies (Eabs’s and Efl’s), evaluated using time-dependent DFT (TDDFT), with nine conventional functionals. ML-ωPBE reproduced experimental Eabs’s and Efl’s for most radicals in questions, with small MAEs of 0.222 and 0.121 eV, marginally worse from OT-ωPBE. Our work not only illustrated a successful extension of stacked ensemble machine learning (SEML) framework from closed-shell molecules to open-shell doublet-spin radicals, but also opened the venue for the calculations of optical properties these using single-reference TDDFT.