Broad application prospects near-infrared carbonized polymer dots combined with machine learning for the detection of Cu2+ in seawater and aquatic products

Abstract

A highly selective and sensitive fluorescence sensor, using the decision tree (DT) machine learning, was successfully fabricated for the quantitative detection of Cu2+, which is based on near-infrared carbonized polymer dots (r-CPDs). Using o-phenylenediamine and 1,3,5-benzene tricarboxylic acid as raw materials first to synthesize the r-CPDs (quantum yield of 32.9%), which pyrrole NH ring can specifically form metal chelates with Cu2+ hiding the electron leap result in the fluorescence burst, by a one-step hydrothermal synthesis method. Furthermore, the fluorescence sensor based on the r-CPDs was fabricated for ultra-sensitively detect Cu2+ in the range of 0.5–80.0 nM (R2 =0.9986) in aqueous environments and aquatic products with relative standard deviation (RSD) below 4.4% and a limit of detection (LOD) of 0.24 nM. Combined with the machine learning algorithm model, the r-CPDs fluorescence color changes accompanied by different Cu2+ concentrations were classified. A self-developed smartphone application equipped with 3D printing technology to prepare portable cartridges successfully applied to rapid real-time detection of trace Cu2+ in various practical samples. The experimental results show that the method has not only convenient for calculating but also accurate. Generally, this study displayed a perfect fusion of r-CPDs fluorescence sensors with machine learning which applied to the determination of Cu2+ in environmental and aquatic products.