Analysis of Thermal Phase Transition via Time-Resolved Infrared Spectra Using Partial Least-Squares Regression Modeling Parameters

Abstract

The thermal phase transition of dl-norleucine (crystalline) has been analyzed via time-resolved Fourier transform infrared (FT-IR) spectroscopy by using a double-scores plot calculated from partial least-squares regression (PLSR). In a previous study, the temperature-dependent spectra collected by time-resolved infrared spectroscopy suggested that another minor thermal phase transition would be present prior to the well-known major transition at ∼118 °C. Eigenvalue analysis by principal component analysis (PCA) suggested, however, that there were four distinguishable signal components, which was not consistent with the evaluation by the absorbance-change plot. In the present study, it has been noticed that a non-systematic response to the temperature increase is involved in the systematic spectral changes. To discriminate the temperature-independent (non-systematic) change from the temperature-related (systematic) changes, two scores of X- and Y-variables in the PLSR modeling have been comparatively used. The new evaluation has readily discriminated a non-systematic factor from basis factors, and the non-systematic factor has been attributed to fringes due to the optical setup used for the measurements.