Computationally Efficient Curve-Fitting Procedure for Large Two-Dimensional Experimental Infrared Spectroscopic Arrays Using the Pearson VII Model

Abstract

Extraction of quantitative information from large sets of sequential time-series spectra (two-dimensional, 2D, data) using a curve-fitting procedure is investigated. The difference between any experimental spectrum and any curve-fitted spectrum is constructed in a least-squares manner and the Pearson VII model is selected as the general band-shape function for the infrared absorbance spectra. The starting point for the curve-fitting procedure is a 2D peak map. Next, an interior spectrum from the 2D array is chosen, initial guesses of the band parameters are provided by an initialization scheme, and curve fitting is performed. Once optimal values of the Pearson VII model band parameters are obtained from the starting spectrum in the 2D array, these are then used as reasonable and judicious initial guesses for subsequent curve fitting of the neighboring spectra in the time series. This increases computational efficiency enormously. By adjusting the band center positions for moving bands, and re-optimizing each subsequent curve-fitting calculation to account for changing band shapes, the entire 2D array is rapidly and efficiently modeled. A 2D experimental spectroscopic time-series array from a homogeneous rhodium-catalyzed hydroformylation reaction was successfully curve fitted in this manner. The procedure holds considerable promise for the curve fitting of massive time-series spectroscopic data, such as that arising from on-line process monitoring of expensive value-added specialty chemical and pharmaceutical syntheses.