Noninvasive Polymer Reaction Monitoring by Infrared Emission Spectroscopy with Multivariate Statistical Modeling

Abstract

Infrared absorption and emission spectroscopy have been used to monitor the curing of a commercial paint product. Principal component analysis of the absorption data indicates that three factors are needed to explain the observed spectral/temporal variance. The interpretation of this finding in terms of changes in the physical state of the reaction mixture is discussed. A similar analysis of the emission data proved more difficult due to a nonlinear concentration/response relationship. A linearization step based on an approximate theoretical model is suggested. The absorption, linearized emittance, and raw emittance data are fit to a two-step sequential rate model using multivariate nonlinear optimization and error estimates derived by Monte Carlo calculations. Better agreement of the model parameters between the absorbance and emittance data is found after linearization, but it is found that linearization introduces large errors in the nonlinear parameter estimates. Comparisons of model parameters for the raw emittance data at different temperatures are made.