Rapid Structural Classification of Post-Consumer Polyolefins by Multimodal Measurement Correlations

Infrared (IR) spectroscopy is the rapid, workhorse technique used by recycling facilities for materials classification; however, as currently used it does not provide complete identification of many recycled materials. While techniques such as multidetector chromatography and small-angle scattering can measure chain composition, topology, and conformation in the laboratory, they are too slow to be of use in a sorting facility. We will outline our approach and present recent results from our effort to develop machine learning tools that leverage the limited chemistry of polyolefins to connect the infrared chemical signature to structural parameters from more time-consuming analytical techniques. We will also describe developments in automated instrumentation to generate reference IR and SAXS datasets to enable development of the machine learning models. Such models will allow IR classification of polyolefin materials with unprecedented fidelity at production speed and will provide FAIR datasets for future development of improved commercial algorithms for advanced sorting and processing equipment.