Automated analysis of microplastics based on vibrational spectroscopy: Are we measuring the same metrics?

The traditional manual analysis of microplastics has been criticized for its labor-intensive, inaccurate identification of very small microplastics (less than 10 µm), and the lack of uniformity between instrumentation techniques. There are already three automated analysis strategies for microplastics based on vibrational spectroscopy: laser direct infrared (LDIR)-based particle analysis, Raman-based particle analysis, and focal plane array-Fourier transform infrared (FPA-FTIR) imaging. We compared the performances of these strategies in terms of their quantification, detection limit, size measurement, and material identification accuracy and analysis speed by analyzing the same standard and environmental samples. Unfortunately, the automated analysis results are not consistent in terms of quantification and material identification. The number of particles smaller than 60 μm recognized by Raman-based particle analysis is far greater than that recognized by LDIR-based particle analysis. Raman-based particle analysis has a submicrometer detection limit but should not be used in the automated analysis of microplastics in environmental samples because of the strong fluorescence interference. LDIR-based particle analysis provides the fastest analysis speed, but we suggest using a reliable detection limit of approximately 60 μm and manually cross-check between the material identification results and the reference database used. Misidentification could occur due to the narrow tuning range from 1800 – 975 cm-1 and dispersive artefact distortion of infrared spectra collected in reflection mode. FPA-FTIR imaging provides relatively reliable quantification and material identification for microplastics in environmental samples greater than 20 µm but might provide an imprecise description of the particle shapes. Optical photothermal infrared (O-PTIR) spectroscopy can detect submicron-sized environmental microplastics (0.5-5 μm) intermingled with a substantial amount of biological matrix; the resulting spectra are searchable in infrared databases without the influence of fluorescence interference, but the process would need to be fully automated.