Profiling the Photosensitizing Properties of Atmospheric Brown Carbon

Upon light absorption, some atmospheric brown carbons
(BrC) can
be converted to triplet excited states (3C*), which initiate
reactions with other atmospheric molecules in a process known as photosensitization.
While 3C* have emerged as critical atmospheric oxidants,
parametrizing their reactivities remains challenging due to the complex
speciation of photosensitizers (PSs) and their diverse reactivity.
In this study, we report machine learning models to predict and classify
the key properties of BrC governing photosensitization, including
triplet-state energy (ET), one-electron reduction potential
(Eground°), and intersystem crossing (ISC) quantum yield (ΦISC), based on molecular structures. The regression models effectively
predicted ET (R2 = 0.77) and
Eground° (R2 = 0.75), while the classification
model for ΦISC achieved an F1 score of
0.66. We first applied models on commonly used PSs in the literature
to profile their chemical reactivities toward typical atmospheric
substrates. Moreover, we employed the models on identified potential
PSs in biomass-burning organic aerosols (BBOA). Our findings provide
new insights into the speciation and reactivities of atmospheric PSs.