389 23
measurement conditions, but a low total can also reveal the
presence on an unexpected elemental constituent that is not
in the list of elements analyzed. For example, an oxidized
inclusion in a metallic alloy will typically have an oxygen
mass fraction of approximately 0.3, leading to a sharp
decrease in the analytical total to ~ 0.7 (70 %) if oxygen is not
recognized and included in a standards-based analysis of that
location compared to the surrounding metallic region. Thus,
even for conventional analysis of ideal specimens, the raw
analytical total conveys useful information and should always
be inspected.
When a specimen deviates from the ideal flat condition and
geometric factors affect the analysis, the raw analytical total
gives a direct indication, providing a standards-based–matrix
correction factor protocol is being used. Note that “standard-
less analysis” does not provide this critical information on the
raw analytical total if this protocol reqwuires a forced normal-
ization to unity mass fraction (100 wt %) since the electron
dose information is not considered. “Standardless analysis”
schemes that use a locally measured elemental spectrum to
establish the dose relationship to the vendor’s standard inten-
sity database or which use the peak-to-background method
(see below) can provide a meaningful analytical total.
. Figures 23.11a, b shows the calculated normalized concentra-
tions as a function of the raw analytical total for Mg and Fe in
K411 from the suite of spectra obtained from the various
geometric shapes. Note that for this data set, the raw analytical
total varies from 0.03 to 1.30 mass fraction (3–130 weight wt
%). For this particular composition (K411 glass), the RDEV for
Mg and Fe is within a range of 10% relative when the analytical
total is in the range 0.8–1.2 mass fraction (80–120 wt %).
Different compositions are likely to have different sensitivities
to deviations in accuracy, but the general experience is that
when the raw analytical total ranges from 0.9 to 1.1 mass frac-
tion (90–110 wt %), the impact of the geometric factors on the
analysis will be minimized.23.4.2 The Shape of the EDS Spectrum
A second powerful indicator that can alert the careful analyst to
the possible impact of geometric factors on an analysis is the
shape of the EDS spectrum. The shape of the X-ray continuum
(bremsstrahlung) background from an ideal flat specimen has
distinctive properties. Consider the spectrum of pure boron,
selected because of the absence of significant characteristic
peaks above the energy of boron (0.185 keV), as shown in. Fig. 23.12. A small peak of oxygen that arises from the inevi-
table surface oxide on the boron can be seen in this spectrum, as
well as the artifact silicon peak from the absorption and internal
fluorescence of the silicon window support grid and the silicon
dead layer of the detector. Otherwise, the spectrum consists
Location 10 Norm conc relative error
S 0.2727 -49%
Fe 0.7273 56%
Raw Total 0.0391Location 3Norm conc relative error
S 0.5352 0.13%
Fe 0.4648 -0.15%
Raw Total 0.9892Location 9Norm conc relative error
S 0.3654 -32%
Fe 0.6346 36%
Raw Total 0.5244FeS 2
Element Ideal mass concentration
S 0.5345
Fe 0.4655Location 7Norm conc R
S 0.0381 -93%
Fe 0.9619 107%
Raw Total 0.1121Location 11 Norm conc RDEV (%)
S 0.5015 -6%
Fe 0.4985 7%
Raw Total 0.8773Errors in normalized analysis
(k-ratio protocol with CuS and Fe, DTSA-II)EDS. Fig. 23.10 Fragment of pyrite (stoichiometric FeS 2 ) analyzed at various locations; conditions: E 0 = 20 keV; DTSA-II calculations with Fe and CuS
as standards, followed by normalization
23.4 · Useful Indicators of Geometric Factors Impact on Analysis