COMBINED DECONVOLUTION AND CURVE FITTING FOR QUANTITATIVE-ANALYSIS OF UNRESOLVED SPECTRAL BANDS

Curve-fitting programs usually only yield accurate parameters for unresolved spectral bands when the separation of neighboring bands exceeds their average half-width. By reduction of the width of each spectral feature In a heavily overlapped band multiplet by Fourier self-deconvolution (FSD), the conditioning of curve-fitting algorithms Is Improved to the point that the area of each band In the multiplet can be estimated with much Improved accuracy. The practical limitations of this approach are discussed in this paper. Most FSD programs are strictly applicable to Lorentzlan bands; the effect of increasing the Gaussian fraction of isolated bands on the accuracy of band area estimation has also been investigated. Finally, the application of combined deconvolution and curve fitting to synthetic band multiplets In which the relative peak absorbances and bandwldths are varied is shown to yield accurate estimates of band areas until the band separation approaches the width of the Instrument line-shape function, which Is determined by the truncation, exponentiation, and apodization of the Fourier domain array. © 1990, American Chemical Society. All rights reserved.