“…Even hyperspectral imaging, an extremely data-rich technique, struggles to distinguish the subtle shifts between subclasses of PE . The use of multiple spectroscopic methods, such as NIR or Fourier transform infrared (FTIR) and Raman measurements, coupled with multivariate data analysis has been shown to identify key properties such as density and crystallinity in POs. ,, FTIR/mid-IR (MIR) can use peak ratios to provide a quantitative value for the branching content, which can then be used to infer trends in crystallinity and density. ,,− However, in their current states, both Raman and MIR are impractical for independent use in large-scale sorting operations . A combination of NIR and factor analysis has demonstrated some ability to distinguish between various classes of POs, including HDPE and LDPE blends. ,, Additionally, both Kim et al and Sato et al have demonstrated that partial least-squares regression (PLSR) of NIR spectra may be able to predict properties such as densities, crystallinity, and melting points of PEs made by a single manufacturer. , However, in both papers, the authors appear to use manufacturer-provided properties and assume they remain constant despite thermal processing.…”