Identification of common textile microplastics via autofluorescence spectroscopy coupled with k-means cluster analysis

Abstract

Microplastics are an emerging anthropogenic pollutant risk with a significant body of research dedicated to understanding the implications further. To generate the databases required to characterize the impact of microplastics on our environment, and improve recovery and recycling of current plastic materials, we need rapid, in-line characterization that can distinguish individual polymer types. Here, autofluorescence spectroscopy was investigated as an alternative characterization method to the current leading techniques based on vibrational spectroscopy. It was confirmed that the autofluorescence of seven common textile polymers (acrylic, polyester, nylon, polyethylene, polypropylene, cellulose/cotton, wool) arose due to the cluster-triggered emission phenomenon. Both simulated polymer aging via photooxidation and dyeing of the polymers were found to affect the resultant autofluorescence spectra. A total of 1485 spectra from 39 unique sample groups (polymer type, colour, and degree of photooxidation) were analysed via machine learning (k-means cluster analysis). Correct identification of the polymer type was achieved in 71% of the cases from only eight input values (normalized intensity values at three autofluorescence emission wavelengths, the total autofluorescence emission intensity, the sample RGB colour values, and the sample shape). This represents a significant step towards automated polymer identification at the sub-second time scales required for the in-line characterization of microplastics.