Multivariate Analysis of Micro-Raman Spectra of Thermoplastic Polyurethane Blends Using Principal Component Analysis and Principal Component Regression

Abstract

Probing the specific hydrogen-bonding behavior of thermoplastic poly-urethane (TPU) blends using vibrational spectroscopies remains the sin qua non for understanding the link between hydrogen-bonding and phasesegregation behavior. However, current literature holds to more traditional univariate approaches when studying the morphologically interesting normal molecular vibrations of TPUs. In the present study, multivariate analysis, including principal component analysis (PCA) and principal component regression (PCR), is used to scrutinize the relevant Raman bands acquired from a binary mixture of analogous TPU copolymer blends. Considering the near identical behavior of selected spectral regions, PCA was capable of isolating linear and nonlinear composition-dependent trends on PC-scores plots. From here, the PC scores, extracted from wavelengths comprising the carbonyl stretching region (1681–1764 cm−1), CH2 deformations (1380–1500 cm−1), aromatic stretch from the hard segment (1617 cm−1), and amide II mixed band (1540 cm−1), were used to explicitly predict the mole fraction of hard segment present in each blend using PCR. Spectral preprocessing, wavelength selection, and variable scaling were major factors in PCR accurately predicting the weight fraction of each copolymer in spite of the clearly evident, blend-specific spectroscopic behavior.