Quantification of ternary microplastic mixtures through an ultra-compact near-infrared spectrometer coupled with chemometric tools

The ubiquitous distribution of plastics and microplastics (MPs) and their resistance to biological and chemical decay is adversely affecting the environment. MPs are considered as emerging contaminants of concern in all the compartments, including terrestrial, aquatic, and atmospheric environments. Efficient monitoring, detection, and removal technologies require reliable methods for a qualitative and quantitative analysis of MPs, considering point-of-need testing a new evolution and a great trend at the market level. In the last years, portable spectrometers have gained popularity thanks to the excellent capability for fast and on-site measurements. Ultra-compact spectrometers coupled with chemometric tools have shown great potential in the polymer analysis, showing promising applications in the environmental field. Nevertheless, systematic studies are still required, in particular for the identification and quantification of fragments at the microscale. This study demonstrates the proof-of-concept of a Miniaturized Near-Infrared (MicroNIR) spectrometer coupled with chemometrics for the quantitative analysis of ternary mixtures of MPs. Polymers were chosen representing the three most common polymers found in the environment (polypropylene, polyethene, and polystyrene). Daily used plastic items were mechanically fragmented at laboratory scale mimicking the environmental breakdown process and creating "true-to-life" MPs for the assessment of analytical methods for MPs identification and quantification. The chemical nature of samples before and after fragmentation was checked by Raman spectroscopy. Sixty three different mixtures were prepared: 42 for the training set and 21 for the test set. Blends were investigated by the MicroNIR spectrometer, and the dataset was analysed using Principal Component Analysis (PCA) and Partial Least Square (PLS) Regression. PCA score plot showed a samples distribution consistent with their composition. Quantitative analysis by PLS showed the great capability prediction of the polymer's percentage in the mixtures, with R greater than 0.9 for the three analytes and a low and comparable Root-Mean Square Error. In addition, the developed model was challenged with environmental weathered materials to validate the system with real plastic pollution. The findings show the feasibility of employing a portable tool in conjunction with chemometrics to quantify the most abundant forms of MPs found in the environment.