Suppressing the Temperature Effect in near Infrared Spectroscopy by Using Orthogonal Signal Correction

Abstract

Temperature changes alter the near infrared spectra for liquid samples and substantially diminish the predictive ability of calibration models constructed from spectra that are not recorded under strictly controlled temperature conditions. The effect is especially significant in compounds forming hydrogen bonds, which exhibit shifts and intensity variations in their spectral absorption maxima and provide non-linear relationships between variables that require more complex models. In this work, we used the variable filtering method known as orthogonal signal correction (OSC) with a view to eliminate the spectral variability introduced by temperature change that is unrelated to the concentrations of the ingredients of an esterification reaction. Once the spurious spectral information was removed by the OSC algorithm, calibration models were constructed from spectra recorded over a broad temperature range in order to determine each individual component (PLS1) and the four reaction ingredients simultaneously (PLS2) in liquid samples. The OSC spectral treatment was found to provide more simple and accurate ( RSEP < 5%) predictive models that those obtained in its absence.