Statistical evaluation of single sparks

Single spark evaluation (SSE) needs time resolved multispectral detection of individual emission intensities from the spark plasma. The classification and statistical evaluation of these data, typically 300 data per second and spectral channel have to be processed. Experiments with metal samples have shown that the single spark pulse height distributions can be satisfactorily approximated by a Gaussian function if the respective element is homogeneously distributed within the sample. If there are non-homogeneities of the analyte concentration in the sample, such as inclusions, than significant deviations from the Gaussian distribution are observed. The integrals over the Gaussian and non-Gaussian parts can be used to quantify the respective components of the analyte. This procedure could be succesfully applied to calibrate the contents of soluble and insoluble Al in a set of steel samples. SSE allows also the calculation of correlations between pulse height distributions of different emission lines. One finds a range from anticorrelation to correlation for various line combinations. This information can help to select reference lines for improving the precision. (C) 1998 Elsevier Science B.V.