0
Article ? AI-assigned paper type based on the abstract. Classification may not be perfect — flag errors using the feedback button. Tier 2 ? Original research — experimental, observational, or case-control study. Direct primary evidence. Sign in to save

IntegratedSERS and Machine Learning Workflow forNanoplastic Detection on a Plasmonic Membrane

Figshare 2026
Amauri Horta-Velázquez, Erika Rodríguez-Sevilla (24362117), Angelica Hernandez-Rayas (24362120), Miguel A. Vallejo (24362123), Eden Morales-Narváez (1530190)

Summary

Scientists have developed a new tool that combines a special light-based sensor with artificial intelligence to detect tiny plastic particles (nanoplastics) in water quickly and accurately—catching amounts as small as 0.02 micrograms per milliliter. This matters because nanoplastics are increasingly found in our water and food, and being able to easily and reliably test for them is a key step toward understanding and eventually reducing our exposure to these potentially harmful particles.

Polymers

Nanoplastics pose increasing health risks, necessitating sensitive and reliable detection methods. Surface-enhanced Raman spectroscopy (SERS) offers high sensitivity and molecular fingerprinting capabilities, but faces challenges in variability and data interpretation complexity, particularly for large analytes such as micro- and nanoplastics. Here, we propose an analytical framework that combines a SERS-active plasmonic membranenanopaper functionalized with gold nanorods with a machine learning pipeline for the automated and semiquantitative detection of nanoplastics. The membrane format enables simultaneous collection and concentration of PMMA nanoplastics from aqueous samples. Our fully automated machine learning pipelineintegrating principal component analysis (PCA), Isolation Forests, K-means clustering, and an ExtraTrees classifier achieving 95% accuracyenables interpretable, semiquantitative detection of PMMA nanoplastics without manual spectral analysis. Additionally, we incorporated an interpretability algorithm that identifies the vibrational modes driving the machine learning classification, yielding chemically validated and trustable predictions. After processing and interpreting the data with machine learning, semiquantification becomes feasible through peak-intensity calibration curves, with an estimated limit of detection of 0.02 μg mL–1. This workflow demonstrates the feasibility of integrating a SERS-active membrane with a machine learning workflow to streamline sampling, detection, and automated data interpretation. This marks an important advancement toward the development of field-deployable SERS-based platforms for easy and user-friendly nanoplastic monitoring.

Share this paper