Use of the L-lines in connection with the elements of the ®rst transition series entails a lot of problems. The corresponding X-ray peaks are composed of contributions of a group of several lines. Because of the incompletely ®lled 3d-shells, the intensity ratios of the lines and thus the overall peak shapes of the X-ray peaks vary with the type of bonding of the respective element. A special problem posed is that of high energy satellites, which could be strongly absorbed even by use of low electron energies. Crucial for reliable quantitative results is a careful choice of reference spectra for peak deconvolution and peak ®t.Nevertheless, with appropriate care results with an accuracy of around 10 ± 15 wt7 relative may be obtained. This has been tested by performing quantitative analysis on a haematite and magnetite specimen by use of the Fe-L lines.Key words: SEM; low voltage; X-ray analysis; elements of the ®rst transition series.Low voltage scanning electron microscopy (E 0 7 keV) offers both high spatial resolution and a greatly reduced X-ray generation depth. This enables the analysis of particles and phases with diameters in the submicrometer range. Although thin ®lm analysis generally can be performed with great success by use of higher energies and conventional thin ®lm analysis programs [1], low voltages may be preferable in case of non¯at specimens or specimens with several common elements in the ®lm and the substrate.While with the advent of ®eld emission scanning electron microscopy imaging at low voltages has become commonplace, use of low voltages in X-ray analysis is still rather rare. Whereas its advantages have been recognized early [2 ± 6], also its limitations have always been emphasized, and quantitative low voltage X-ray analysis is still regarded as applicable in special cases only [7,8].
Special ProblemsQuantitative analysis with the L-lines of the elements of the ®rst transition series has always been regarded as extremely dif®cult. The corresponding peaks are composed of the contributions of the L , L , L 1 and L and additional non-diagram lines. Because of the incompletely ®lled 3d-shells, the intensities of the L 1 and L lines are no longer negligible, but can be as high as that of the L and L lines. Measurable L 1 and L peaks entail an overlap with the L , -peak in case of energy dispersive spectra (see Fig. 1) and generally more frequent peak overlaps with the X-ray peaks of other elements.Other problems are caused by bonding effects. The occupation number of the 3d-shell and thus the intensity ratios of the lines and the overall peak shapes of the EDXS peaks will vary with the bonding of the respective element. As a consequence, simple peak integration could give wrong results in case of different type of bonding of the speci®c element in the unknown and the standard. Additionally, the choice of an appropriate standard is also important for the recording of reference spectra for peak deconvolution. This will require some preknowledge about the specimen. In fact, the errors caused by ...