Dale E. Newbury and David S. Bright

Objective: To provide a logical, easily interpretable intensity scale for the presentation of quantitative compositional maps.

Problem: Compositional mapping is one of the most widely used methods of presenting information on heterogeneous microstructures measured with microprobe instrumentation, such as scanning electron microscopy/x-ray spectrometry, analytical electron microscopy/x-ray and electron energy loss spectrometry, Auger electron spectrometry, etc. Existing methods for displaying the compositional axis, especially those incorporated in commercial software systems, are only useful at the most rudimentary level of qualitative analysis with no useful quantitative information available at the pixel level. For example, x-ray mapping systems use a gray level or color value that is related to the raw x-ray count at the pixel location with no corrections for background or relative excitation effects. It is therefore not possible to make sensible comparisons of different elements in the same region or for the same element from different regions or different specimens.

Approach: The new compositional mapping method involves collecting intensity data and making two critical corrections at the individual pixel level. A background correction by simple interpolation from nearby background windows in the spectrum eliminates a critical artifact from measurements of trace (arbitrarily defined as less than 0.01 mass fraction) and minor (0.01 to 0.1 mass fraction) constituents, where the measured "peak" intensity may actually consist primarily of background contributions. Under electron excitation, the x-ray continuum forms a significant fraction of the excited x-ray spectrum, and the continuum intensity is proportional to the overall average atomic number. Thus, when uncorrected x-ray maps of minor and trace constituents are examined, the analyst may perceive apparent compositional contrast which is purely artifactual, arising only from changes in the other constituents and manifested through the continuum dependence on atomic number. The second correction involves the differential excitation as a function of characteristic photon energy. X-ray yield depends strongly on the ratio of beam energy (E0) to the critical excitation (Ec) energy, (E0/Ec)n, where n is in the range 1.5-1.7 Since the incident beam energy is fixed, the excitation varies strongly across the photon energy range, typically 0.1 keV to 12 keV. Moreover, the fluorescence yield (the fraction of inner shell ionization events leading to photon emission) depends upon atomic number and the atomic shell. Finally, the x-ray intensity actually measured is modified by the efficiency of the spectrometer, which in the case of energy dispersive x-ray spectrometry decreases strongly for photon energies below 4 keV. All of these effects can be compensated by determining the ratio of the background-corrected pixel intensity to the intensity measured from a pure element (or its equivalent calculated from a compound standard) under identical beam conditions, which yields the "k-value" of classic quantitation procedures. For quantitative display, the resulting k-value maps are encoded with a logarithmic intensity scale using primary colors which grade to pastels, selected such that the blue band covers the range k less than 0.01, green from 0.01 to 0.1, and red greater than 0.1 to 1.0.

Results and Future Plans: The "log 3-band" images have been found to enable the viewer, including non-specialists in the technique, to readily distinguish trace, minor, and major constituents while retaining excellent contrast sensitivity to compositional structures. We plan to extend the technique to incorporate matrix corrections at each pixel to create concentration maps. For other compositional techniques with a wider dynamic range, such as electron energy loss spectrometry in the analytical electron microscope, additional bands will be created to extend the display range to lower trace levels.

Web Contact micro@nist.gov