Abstract

Quantitative detection of trace small-sized nanoplastics (<100 nm) remains a significant challenge in surface-enhanced Raman scattering (SERS). To tackle this issue, we developed a hydrophobic CuO@Ag nanowire substrate and introduced a multiplex-feature analysis strategy based on the coffee ring effect.

This substrate not only offers high Raman enhancement but also exhibits a high probability of detection (POD), enabling rapid and accurate identification of 50 nm polystyrene nanoplastics over a broad concentration range (1–10−10 wt%). Importantly, experimental results reveal a strong correlation between the coffee ring formation and the concentration of nanoplastic dispersion.

By incorporating Raman signal intensity, coffee ring diameter, and POD as combined features, we established a machine learning-based mapping between nanoplastic concentration and coffee ring characteristics, allowing precise predictions of dispersion concentration. The mean squared error of these predictions is remarkably low, ranging from 0.21 to 0.54, representing a 19 fold improvement in accuracy compared to traditional linear regression-based methods.

This strategy effectively integrates SERS with wettability modification techniques, ensuring high sensitivity and fingerprinting capabilities, while addressing the limitations of Raman signal intensity in accurately reflecting concentration changes at ultra-low levels, providing a new idea for precise SERS measurements of nanoplastics.