A fluorescent optical platform sensor based on ratiometric carbon dots integrated with machine learning for the identification and detection of organic solvents

Rapid identification of organic solvents is of great significance for chemical analysis, environmental monitoring and laboratory safety. In this study, ratiometric carbon dots (R-CDs) were integrated with machine learning to develop a portable fluorescence-based optical sensing platform for the rapid discrimination and recognition of common organic solvents. Based on the highly sensitive solvatochromic response of R-CDs fluorescence intensity ratio (I_590nm/I_502nm), a six-channel sensor array was constructed by preparing water/organic-solvent mixtures with different solvent fractions. And owing to the distinct fluorescence responses of R-CDs in solvents with different Relative Polarity (E_T^N) values, principal component analysis (PCA), hierarchical cluster analysis (HCA) and linear discriminant analysis (LDA) were employed to discriminate and identify five organic solvents with E_T^N values ranging from 0.355 to 0.762. Specifically, PCA and HCA were used for label-free analysis of the fluorescence signals, while LDA was employed for model training and prediction. Furthermore, by combining the optical sensing platform with self-developed computer software, a multifunctional system capable of solvent detection, recognition, and result visualization was established. The results showed that five organic solvents could be successfully distinguished with an identification accuracy of 100%, demonstrating the promising potential of the proposed optical sensing platform for organic solvent recognition.