Characterization and identification of microplastics using Raman spectroscopy coupled with multivariate analysis.

The manufacture and use of plastic products have resulted in the release and spread of a massive amount of microplastics. Identifying and quantifying microplastics is challenging due to their small size and complicated composition. Although vibrational spectroscopy has been applied to analyze microplastics, its reliability and throughput are limited by the challenges to distinguish the pending alterations manually and the lack of a spectra-based automated microplastic classification model. The present study applied Raman spectroscopy coupled with multivariate analysis to develop a new and robust analytical method to comprehensively interrogate the spectral profiles of seven microplastic references and real microplastic samples post-exposure to environmental stresses. Besides identifying unique Raman peaks of individual microplastics, their whole spectra were separated by principal component analysis (PCA) and linear discriminant analysis (LDA). Support vector machine (SVM) classification achieved an accuracy rate of over 98% for polypropylene, polyethylene terephthalate, polyvinyl chloride, polycarbonate, polyamide, and over 70% for high-density polyethylene and low-density polyethylene. Real microplastic samples from the breakdown of snack boxes, mineral water bottles, juice bottles, and medicine vials were also matched to their chemical components by SVM with an overall sensitivity, specificity, and accuracy of 98.1%, 99.4%, and 99.1%, respectively. Additionally, post-exposure to environmental stressors, 1D PCA-LDA score plots could still distinguish microplastic type, and the developed SVM classification achieved an accuracy of 96.75% in the real-world scenario. These findings prove Raman spectroscopy coupled with multivariate analysis as an ideal tool to distinguish the types and environmental exposure of microplastics, demonstrating great potential for microplastic automatic detection.