Mixture analysis using PFG-NMR (DOSY) data is, for many chemists, a valuable and increasingly popular technique where the NMR signals of different species are separated according to their diffusion coefficients. Where NMR signals overlap, however, it is often difficult to extract the spectra of pure components from experimental data. In such situations, it can often be helpful to use multivariate methods, which exploit all the available signal covariance, to resolve the spectra of the components of a mixture. The best-established and by some way the quickest such method, DECRA (Direct Exponential Curve Resolution Algorithm), unfortunately requires that data conform to a pure exponential decay as a function of gradient strength squared, while experimental data typically deviate significantly from this. If this deviation is known, the performance of DECRA can be greatly improved for components with similar diffusion coefficients by adjusting the choice of gradient strengths used.
Copyright 2007 John Wiley & Sons, Ltd.