Evaluation of spectrum overlap correction in energy‐dispersive x‐ray spectrometry using the digital filter deconvolution procedure: Application to selected interferences encountered in the microprobe analysis of minerals

Abstract

AbstractAn evaluation was undertaken to characterize the accuracy and effectiveness in the deconvolution of energydispersive x‐ray spectra of the overlap correction procedure based on a digital ‘top hat’ filter and least‐squares fitting of library spectra using an algorithm developed by Statham. This deconvolution procedure was applied to x‐ray spectra recorded by electron microprobe to represent interferences that range in severity from minor to severe, specifically overlap between (i) Cr/Mn K lines, (ii) Ba L and Ti K lines, (iii) Au/Pt L and M lines and (iv) Bi/Pb L and M together with S K lines. A novel calibration procedure was used to simulate spectra and so avoid the requirement to obtain and analyse a large number of mineral standards. This procedure involved superimposing x‐ray spectra from individual elemental standards counted for a range of count times suitable for testing the deconvolution program. The results demonstrated that reliable deconvolution could be achieved down to <1% Cr in a manganese matrix, 2.5% Ba in a titanium matrix, <2.5% Au in a platinum matrix (based on either L or M lines) and about 2.5% Bi in a Bi–Pb–S matrix using the M lines of Bi and Pb. Within the fitting error calculated by the deconvolution program, bias from expected values was not detected excepted when Bi, Pb and S were fitted to the simulated spectrum of bismuthite (Bi2S3). Small discrepancies in the Bi and S values were then observed. Overall, it was concluded that in cases of severe overlap, numeric data from the spectrum deconvolution program were a more reliable indicator of the presence of an element than visual examination of the original spectrum.