NEAR INFRARED SPECTROSCOPY AND MULTIVARIATE CURVE RESOLUTION-ALTERNATING LEAST SQUARES FOR IN SITU QUANTITATIVE MONITORING OF EPOXY RESINS REACTIONS. VALIDATION BY HPLC

M. GARRIDO; M. S. LARRECHI; F. X. RIUS

Marrakech

Simposio; IX International Symposium in Analytical Chemistry (KAC); 2006

Epoxy-amine polymers are often used in the industry and medicine because of their good mechanical and physical properties. In these applications, the curing process of the epoxy resins affects the network structure and, hence, determines the final physical properties [1]. Consequently, there is a need for a more detailed understanding of the structure-property relationship in order to control the curing economically and to optimise the properties of the final product. Near infrared spectroscopy (NIR) in combination with multivariate curve resolution methods, has demonstrated to have a great potentiality for monitoring epoxy resins reactions, so the concentration profiles of each species involved in the reaction and the corresponding pure spectra can be estimated [2]. In this study, the model reaction between phenylglycidylether (PGE) and aniline was monitored in situ by NIR. Five different process runs starting from different initial conditions, by modifying PGE/aniline molar ratio, were designed (1:1, 1.25:1, 1.5:1, 1.75:1, 2:1). Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS) was used to simultaneously analyze the five sets of NIR data in order to obtain the corresponding kinetic profiles and the only one spectral profile, in common for the five experiments. The recovered spectral profile was used as a calibration model to predict, at any time, the concentration of the species in new reactions eventually performed with any PGE/aniline molar ratio (between 1:1  and 2:1). The simple procedure consists of recording, at a certain time, a single spectrum for the reacting mixture and using the model to obtain the concentration of each species. The same samples were analyzed by both the proposed method and HPLC, and the results were statistically comparable (á = 5%).