Mistakes encountered during automatic peak identification in low beam energy X-ray microanalysis.

Automated peak identification in electron beam excited X-ray microanalysis with energy dispersive X-ray spectrometry (EDS) is subject to occasional mistakes even on well-separated, high-intensity peaks arising from major constituents. The problem is exacerbated when analysis conditions are restricted to operation in the "low beam energy scanning electron microscopy" (i.e. "low voltage scanning electron microscopy" or LVSEM) regime where the incident beam energy is 5 keV or less. These low beam energy microanalysis conditions force the analyst to use low fluorescence yield L-shell and M-shell peaks rather than higher yield K-shell and L-shell peaks typically selected for elements of intermediate and high atomic number under conventional high beam energy (>10 keV) conditions. Misidentifications can arise in automated peak identification procedures when only a single energy channel is used to characterize an EDS peak. The effect of the EDS measurement process is to convolve the closely spaced Lalpha-Lbeta and Malpha-Mbeta peaks into a single peak with a peak channel shift of 20 eV or more from the Lalpha or Malpha value, which is typically sought in an X-ray database. An extensive list of problem situations encountered in low beam energy microanalysis is presented based upon observed peak identification mistakes as well as likely troublesome situations based upon proximity in peak energy. Robust automatic peak identification requires implementation of peak fitting that utilizes the full peak shape. (c) 2007 Wiley Periodicals, Inc.