Application of mathematical procedures to background correction and multivariate analysis in inductively coupled plasma-optical emission spectrometry

In inductively coupled plasma-optical emission spectrometry, continuous progress is made in acquiring and processing more spectral information than one gross analyte line intensity and two background intensities measured at both sides of the line. Compared to direct readers and slew-scan spectrometers, the recently introduced systems with echelle optics and charge transfer device detection provide more signals at a time, not only with respect to line and background, but also with respect to multiple lines for multiple elements. PC systems are increasingly powerful to apply more sophisticated mathematics for data reduction. Procedures developed over the last decade, based on convolution, differentiation, Fourier transforms, correlation, expert systems, neural networks, principal component analysis, projection methods, Kalman filtering, multiple linear regression and generalised standard additions are reviewed. The first four procedures aim at continuum or structured background correction in every spectral window measured, without prior knowledge about the appearance of the background and the composition of the sample. The other procedures utilise the information that is present for the sample and readily available in the spectrum. In addition to background correction, they allow signal averaging, noise separation, multiple line and multiple element analysis. Multiple line procedures concentrate on qualitative analysis and line selection, whereas multicomponent analysis allows quantitative analysis. Based on the characteristics of ICP-OES, the multicomponent analysis procedures are preferred.