CALCULATION OF BAND BOUNDARIES OF FEASIBLE SOLUTIONS OBTAINED BY MCR-ALS OF MULTIPLE RUNS OF A REACTION MONITORED BY NIR SPECTROSCOPY

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

Lisboa

Congreso; IX CAC Chemometrics in Analytical Chemistry; 2004

Multivariate curve resolution (MCR) methods, specifically alternating least squares (ALS), applied to spectral data from monitoring chemical reactions, provide the concentration profiles and the spectra of all the species involved in the processes. Because of so-called ambiguities, the overall resolution of a data set is only possible if selectivity or local rank conditions are present. This is not often the case found in data matrices from monitoring chemical reactions; a band of feasible solutions that have the same residual is obtained instead of a unique solution. This study calculates the band boundaries of feasible solutions for spectra and concentration profiles obtained by applying the MCR-ALS method to NIR data from monitoring the model reaction between phenyl glycidyl ether (PGE) and aniline. As the system is rank-deficient, the strategy of augmented matrices is used. Therefore band boundaries are calculated, for the first time, for various data matrices simultaneously. The bands were calculated using the method proposed by Tauler. The non-linear optimization was carried out by applying the non-negativity and closure constraints. Additionally, the effects caused in the amplitude of band boundaries by constraints and by the number and type of data matrices included in the simultaneous analysis are studied in detail. The results obtained in this work demonstrate that it has been possible to know the composition of the reacting mixture at any time of the curing process, and to consider possible errors associated to the rotational ambiguities present in the spectral and concentration profiles.