Influence of different sources of error on estimated kinetics parameters for a second-order reaction

A second-order reaction with a rate constant of 0.2 M−1 min−1 is simulated. The data consists of 100 UV/VIS spectra in a reaction mixture at 20-s intervals. The paper investigates the influence of three sources of error, namely instrumental noise, error in determining initial concentrations and error in calibration of pure standards, on the determination of rate constants from the resultant spectra. Twelve methods involving a mixture of multivariate and kinetic models for estimating rate constants are compared, in five groups (MLR, Rank Augmentation, Difference Spectra, Mixed Spectra and PCR). The behaviour of the error surface depends on the information required by each algorithm. Best results were obtained using rank augmentation, but these require two or more batches run under stable instrumental conditions. Traditional approaches based on MLR do not yield good results if there are errors in the initial concentrations or spectra, although using alternating least squares improves performance. A new approach based on difference spectra shows relatively low spread of errors in rate constant estimates. Methods based on principal component regression and mixed spectra provide accurate average estimates of rate constants but are imprecise.