A pattern recognition approach in X-ray fluorescence analysis

Abstract

In many applications of X-ray fluorescence (XRF) analysis, quantitative information on the chemical components of the sample is not of primary concern. Instead, the XRF spectra are used to monitor changes in the composition among samples, or to select and classify samples with similar compositions. We propose in this paper that the use of pattern recognition technique in such applications may be more convenient than traditional quantitative analysis. The pattern recognition technique discussed here involves only one parameter, i.e., the normalized correlation coefficient and can be applied directly to raw data. Its computation is simple and fast, and can be easily carried out on a personal computer. The efficacy of this pattern recognition approach is illustrated with the analysis of experimental XRF spectra obtained from geological and alloy samples.