Related Experiment Video
Updated: Apr 22, 2026
03:33
Author Spotlight: Development and Application of SERS Flexible Substrates Using Synthesized AgNPs
Published on: November 17, 2023
3.2K
Deep Learning-Aided SERS Detection of Microplastics in Water Samples with a Hierarchically Porous Gold Sponge
Zhenli Sun1, Yiyan Zhang1, Jingjing Du2,3
1MOE Key Laboratory of Resources and Environmental System Optimization, College of Environmental Science and Engineering, North China Electric Power University, Beijing 102206, China.
Analytical Chemistry
|April 21, 2026
Summary
This study introduces a novel, pretreatment-free method for detecting microplastics (MPs) in water using surface-enhanced Raman scattering (SERS) and deep learning. The advanced technique accurately identifies various plastic types in complex environmental samples.
Area of Science:
- Environmental Science
- Analytical Chemistry
- Materials Science
Background:
- Microplastics (MPs) are ubiquitous environmental contaminants in aquatic systems.
- Conventional MP detection methods are often labor-intensive, requiring extensive sample pretreatment and lacking specificity for mixed polymer types.
- Accurate and efficient monitoring of MPs is crucial for assessing environmental risks.
Purpose of the Study:
- To develop a pretreatment-free analytical method for sensitive and accurate identification of microplastics in complex water matrices.
- To integrate an electrostatically functionalized surface-enhanced Raman scattering (SERS) substrate with an interpretable deep learning framework.
- To validate the method's performance in real-world environmental samples.
Main Methods:
- Fabrication of a hierarchically porous gold sponge substrate functionalized with poly(diallyldimethylammonium chloride) for electrostatic enrichment and size-selective capture of MPs.
- Utilizing embedded gold nanoparticles to generate plasmonic hotspots for enhanced Raman signal amplification.
- Development of a modular binary convolutional neural network (CNN) framework with a one-vs-rest architecture for MP classification, employing Grad-CAM for interpretability.
Main Results:
- The integrated SERS-deep learning platform achieved high precision (0.9896) in classifying five representative MP types (polytetrafluoroethylene, polypropylene, polystyrene, polyvinyl chloride, and polyethylene terephthalate).
- Grad-CAM analysis identified characteristic Raman bands for each polymer, confirming the CNN model's chemical interpretability.
- The method demonstrated successful validation in complex matrices, including urban tap water and natural surface waters impacted by heavy metals.
Conclusions:
- The developed platform offers a sensitive, adaptable, and pretreatment-free approach for identifying MPs in diverse and challenging water samples.
- This technology holds significant potential for practical and routine environmental monitoring of microplastic pollution.
- The combination of advanced SERS substrates and interpretable deep learning provides a powerful tool for environmental analysis.
