Prospects for energy dispersive (ED) analysis of light elements in the electron probe are assessed, with particular concern for the accuracy which can be achieved in quantitative analysis. A comprehensive study of the ED system and the low energy spectrum it produces has revealed several problems. These have been thoroughly investigated and techniques developed to overcome them. Reasons for the failure of a commonly used dead time correction system are discussed and the magnitude of errors in quantitative work are determined. An alternative correction method is developed and its accuracy confirmed. It is also shown that differences in methods used for signal discrimination in ED systems lead to different presentations of the low energy spectrum. These effects must be clearly understood before measurements of soft x-ray peak intensities are attempted. The problem of electronic noise counts which appear at low energy is discussed and a way of removing them from spectra is demonstrated, A new component of the spectrum background below 500 eV is identified and characterised, to provide a means of accounting for it in recorded spectra. Two remaining problems, although common to conventional ED analysis, are more difficult to solve in the soft x-ray region. The first concerns prediction of x-ray continuum, and here a promising technique based on the use of a reference standard is developed, extending it's range down to 0.1 keV. The second problem is deconvolution of overlapping peaks and a least squares fitting procedure is adopted for this work. The accuracy of ED analysis incorporating these techniques is demonstrated by comparison with wavelength dispersive analysis of selected specimens. Finally realistic sensitivity levels are measured 6for light elements at a range of probe voltages.
|Date of Award||1983|