Decoding Carbon Dot Synthesis: Machine Learning-Guided Analysis of Reaction Conditions Governing Fluorophore vs Nanoparticle Formation

Carbon dots (CDs) are fluorescent nanoparticles whose success stems from accessible bottom-up synthetic methodologies, using readily available substrates as precursors. Despite considerable interest in CDs, uncertainty surrounds their luminescence origin as molecular fluorophores (MFs) formed during thermal treatment may be the true source of fluorescence. The realization that MFs coexist with CD formation has intensified focus on purification protocols, yet reaction conditions have remained comparatively neglected due to the extensive parameter space. In this work, the synthetic conditions governing CD formation were systematically investigated. Ubiquitous citric acid (CA) and ethylenediamine (EDA) served as our carbon and nitrogen sources, respectively. Both hydrothermal and microwave heating methodologies were employed. The resultant products were subjected to meticulous examination, with particular emphasis upon their fluorescence properties, quantified through quantum yield measurements, and their size distributions, assessed via high-pressure size exclusion chromatography. Machine learning algorithms were employed to establish correlations between synthesis parameters and the size and fluorescence characteristics of the resultant products. Reaction mixtures of CA/EDA 1:1 at pH 8.5, subjected to microwave heating at 205 °C for 25 min, yield predominantly MFs, whereas reaction mixtures of CA/EDA 1:2 at pH 4.5, subjected to hydrothermal treatment at 260 °C for 15 h, yield CDs. Beyond furnishing guidance for tuning reactions toward either MF or CD formation, these findings permit elucidation of the underlying synthetic mechanisms: mildly acidic conditions favor polymerization and subsequent carbonization into CDs at elevated temperatures, while mildly basic conditions impede reactivity, arresting the synthetic pathway at the molecular level.