Related Experiment Video
Updated: Jan 12, 2026
Multimodal Analysis of Microplastics in Drinking Water using a Silicon Nanomembrane Analysis Pipeline
Published on: June 13, 2025
Integrating MetalAquaDect SERS platform: Machine-learning assisted real-time monitoring of sub-2mg/L microplastics
Jiamiao Wang1, Yan Liu2, Zheyu Shen3
1College of Metrology Measurement and Instrument, China Jiliang University, No. 258 Xueyuan Street, Qiantang District, Hangzhou, Zhejiang Province 310018, China; College of Energy Environment and Safety Engineering, China Jiliang University, No. 258 Xueyuan Street, Qiantang District, Hangzhou, Zhejiang Province 310018, China; College of Optical and Electronic Technology, China Jiliang University, No. 258 Xueyuan Street, Qiantang District, Hangzhou, Zhejiang Province 310018, China.
Abstract:
Using the AquaDect platform and SERS technology, we conducted a qualitative chemical study of microplastics and nanoplastics in aqueous solutions. Microplastics and nanoplastics of different sizes (60-100μm, 20μm, 5μm, and 100 nm) and types (polystyrene (PS), polyethylene (PE) with different colors, polypropylene (PP), and polymethyl methacrylate (PMMA)) were used in the research. The Raman enhancement effect of microplastics and nanoplastics in aqueous solution was studied by analyzing the different volume ratios and concentrations of the sample and silver sol. Silver nanoparticle sol and microplastics were characterized by scanning electron microscopy. The SERS activity of the substrate was tested using the Raman probe molecule Rhodamine 6 G. The results show that the Raman signals detected in both pure water and lake water are well enhanced. The analysis method based on SERS not only overcomes the limitations of detecting microplastics and nanoplastics in liquids, but can also detect microplastics down to 2 mg/L. To further explore the spectral characteristics and enhance the classification accuracy, a combination of dimensionality reduction techniques-principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), uniform manifold approximation projection (UMAP), and latent Dirichlet allocation (LDA)-was applied to visualize and differentiate the spectral data. In the classification task of the dataset using machine learning, the performance of various models was evaluated and compared. Moreover, LightGBM demonstrated the best performance, achieving an accuracy of 0.94 and an F1-score of 0.94. This model provides an effective solution for detecting residual microplastic pollutants in the environment using the AquaDect platform.
