XIcombinations of elements, for example, Ti and
Ba; S, Mo, and Pb; and many others, especially
when the peaks of major constituents interfere
with the peaks of minor or trace constituents.
Operator knowledge of the physical rules govern-
ing X-ray generation and detection is needed to
perform a careful review of software-generated
peak identifications, and this careful review must
always be performed to achieve a robust mea-
surement result.
After a successful qualitative analysis has been
performed, quantitative analysis can proceed.
The characteristic intensity for each peak is auto-
matically determined by peak fitting procedures,
such as the multiple linear least squares method.
The intensity measured for each element is pro-
portional to the concentration of that element,
but that intensity is also modified by all other ele-
ments present in the interaction volume through
their influence on the electron scattering and
retardation (“atomic number” matrix effect, Z),
X-ray absorption within the specimen (“absorp-
tion” matrix effect, A), and X-ray generation
induced by absorption of X-rays (“secondary flu-
orescence” matrix effects, F, induced by charac-
teristic X-rays and c, induced by continuum
X-rays). The complex physics of these “ZAFc”
matrix corrections has been rendered into algo-
rithms by a combined theoretical and empirical
approach. The basis of quantitative electron-
excited X-ray microanalysis is the “k-ratio proto-
col”: measurement under identical conditions
(beam energy, known electron dose, and spec-
trometer performance) of the characteristic
intensities for all elements recognized in the
unknown spectrum against a suite of standards
containing those same elements, producing a set
of k-ratios, where
kI= UnknownS/Itandard (1)for each element in the unknown. Standards are
materials of known composition that are tested to
be homogeneous at the microscopic scale, and
preferably homogeneous at the nanoscale.
Standards can be as simple as pure elements—e.g.,
C, Al, Si, Ti, Cr, Fe, Ni, Cu, Ag, Au, etc.—but for
those elements that are not stable in a vacuum
(e.g., gaseous elements such as O) or which
degrade during electron bombardment (e.g., S, P,
and Ga), stable stoichiometric compounds can be
used instead, e.g., MgO for O; FeS 2 for S; and GaP
for Ga and P. The most accurate analysis is per-
formed with standards measured on the same
instrument as the unknown(s), ideally in the same
measurement campaign, although archived
standard spectra can be effectively used if a quality
measurement program is implemented to ensure
the constancy of measurement conditions, includ-
ing spectrometer performance parameters. With
such a standards-based measurement protocol and
ZAFc matrix corrections, the accuracy of the anal-
ysis can be expressed as a relative deviation from
expected value (RDEV):RDEV()%%=×[]()Measured−True/True (^100)
(2)
Based on extensive testing of homogeneous
binary and multiple component compositions,
the distribution of RDEV values for major con-
stituents is such that a range of ±5 % relative cap-
tures 95 % of all analyses. The use of stable, high
integrated count spectra (>1 million total counts
from threshold to E 0 ) now possible with the sili-
con drift detector EDS (SDD-EDS), enables this
level of accuracy to be achieved for major and
minor constituents even when severe peak inter-
ference occurs and there is also a large concen-
tration ratio, for example, a major constituent
interfering with a minor constituent. Trace con-
stituents that do not suffer severe interference
can be measured to limits of detection as low as
C = 0.0002 (200 parts per million) with spectra
containing >10 million counts. For interference
situations, much higher count spectra (>
million counts) are required.
An alternative “standardless analysis” protocol
uses libraries of standard spectra (“remote stan-
dards”) measured on a different SEM platform
with a similar EDS spectrometer, ideally over a
wide range of beam energy and detector parame-
ters (resolution). These library spectra are then
adjusted to the local measurement conditions
through comparison of one or more key spectra
(e.g., locally measured spectra of particular ele-
ments such as Si and Ni). Interpolation/extrapola-
tion is used to supply estimated spectral intensities
for elements not present in or at a beam energy not
represented in the library elemental suite. Testing
of the standardless analysis method has shown
that an RDEV range of ±25 % relative is needed to
capture 95 % of all analyses.
High throughput (>100 kHz) EDS enables col-
lection of X-ray intensity maps with gray scale rep-
resentation of different concentration levels (e.g.,
. Fig. 7a). Compositional mapping by spectrum
imaging (SI) collects a full EDS spectrum at each
pixel of an x-y array, and after applying the quanti-
tative analysis procedure at each pixel, images are
created for each element where the gray (or color)
level is assigned based on the measured concentra-
tion (e.g.,. Fig. 7b).
Scanning Electron Microscopy and Associated Techniques: Overview