A Statistical Treatment for Quantitative SEM-EDS Analysis as Applied to Automated Analysis of Many Particles

"INTRODUCTIONGomez and Austin (1,2) have recently described calibration procedures for quantitative chemical analysis of particles by energy dispersive spectroscopy using particles mounted and polished in a resin matrix and scanned by an SEM under computer control. Their software is specifically designed for the analysis of Si, Al, Ca, Na, K, Fe and S in coal particles, and in this case the resin is loaded with iodine (3) to enable the exposed cross-section of an embedded coal particle to be differentiated from the matrix. However, the basic ideas of their procedures could be extended to any system of particles where it is desired to measure inorganic species in particles containing light elements such 0, C, H and N which are not directly detected. In these circumstances it is necessary to perform a quantitative determination of the measured components, so that the equivalent volume of light elements in the excited sample volume can be estimated by difference using ZAF and other corrections applied to the raw count intensities.Part of their procedure involves a new technique (2) for estimating background spectrum intensities; the intensity for a particular energy window is then subtracted from the total measured intensity for that window to enable the concentration of the corresponding element to be calculated. In addition, corrections for overlap of peak spectrums are applied. For example, Figure 1 shows some typical spectrums for known materials for the system they used, demonstrating a wide Gaussian spread about the theoretical peak energy. Since background spectrum and peak overlap corrections require a knowledge of the composition of the material, as does the application of the ZAF corrections, it is necessary to have an iterative solution of the equation sets, starting with estimates of the composition based on initial count rate corrected solely for dead time and detector efficiency."