Mid-Infrared Spectra from Near-Infrared Spectra Using Partial Least-Squares

Abstract

In this novel application of a multivariate method, partial least-squares (PLS) was used to generate mid-infrared (MIR) spectra (rather than selected concentrations) from near-infrared (NIR) spectra. The NIR spectra were obtained by in-line monitoring of a molten polymer blend of polyethylene with polypropylene during extrusion. Off-line MIR spectra of blends were used to calibrate the PLS method. Then PLS was used to generate the MIR absorbance spectrum of a 50:50-by-weight blend not included in the calibration set from its NIR spectrum. The synthesized MIR spectrum agreed very well with a directly measured one. The exception was absorbance peaks which were so strong that they apparently represented responses that were nonlinear with respect to concentration. Although more evaluation work has yet to be done, these results are encouraging, and they indicate that NIR interpretation may readily borrow the strengths of MIR interpretation both qualitatively and quantitatively.