The detection of low‐intensity peaks in energy dispersive x‐ray spectra from particles

A critical step in the processing of energy dispersive (EDS) x‐ray spectra from the automated scanning electron microscopy (ASEM) analysis of particles is the detection and identification of elemental peaks. Since there are often several hundred to several thousand spectra for each ASEM analysis, it is important that this step operate rapidly and with a minimum of interaction between analyst and the program. For peaks with large peak‐to‐background (P/B) ratios, most peak‐find or peak‐fitting methods do a reasonable job even when the spectra have low signal‐to‐noise (S/N) ratios. The detection of peaks with small P/B ratios is much more problematical. Peak identification and fitting procedures may not work well on low‐intensity peaks, particularly in spectra with low S/N ratios. In this study, three procedures for identifying x‐ray peaks, with small P/B ratios in spectra with varying S/N ratios, were evaluated. The first procedure was the identification of peaks by human analysts. The results from the analysts were then used to set a benchmark for the performance of two computer‐based procedures that included three different qualitative peak identification methods, and one quantitative analysis procedure. The success of the qualitative methods in finding small peaks varied widely. In general, the quantitative analysis procedure performed as well as the best human analysts and was better than the qualitative methods.