Professional Documents
Culture Documents
Microanalysis (EPMA):
Doing It Right and Doing It Wrong!
Dale E. Newbury
Surface and Microanalysis Science Division
National Institute of Standards and Technology
Gaithersburg, MD 20899
Quantitative EPMA
Two zeroth order conditions must be fulfilled:
1. The specimen must have no factors other than
composition differences affecting the measured x-ray
intensities. Geometric factors (surface irregularities)
must be eliminated by careful polishing (NO ETCHING).
2. The specimen must be homogeneous over the
interaction volume (dimensions ~ 1 micrometer) that
the beam excites. Heterogeneous specimens (e.g., a
thin film on a substrate) can be quantified, but advanced
calculational procedures are necessary.
k = isample/istd
Ca ka (first order)
k = isample / istd
For each element being analyzed, measure the same
X-ray peak in the unknown and a standard under
identical conditions of beam energy, known dose,
and spectrometer efficiency.
k = isample / istandard
= (Igenerated * f() * )sample /(Igenerated * f() * )std
Spectrometer efficiency cancels quantitatively
when the k-ratio is measured because exactly the
same X-ray energy is measured for both! The
efficiency does NOT cancel in standardless!
Diffraction Crystal
e.g., LiF, PET, multilayer
Focussing circle
n = 2d sin B
Flow Proportional
X-ray Detector
= X-ray wavelength
d = crystal plane spacing
B = Bragg angle
AlGaAs
Microcalorimeter EDS
(9 eV)
WDS (TAP)
(3 eV)
I / I0 = exp[ -(/) s]
I0
Error Bins
are 1% wide!
Frequency
50
40
30
20
10
0
-15 -10
-5
10
15
Relative error
Relative error
X-rays
Electrons
Holes
Rear Au electrode, ~20 nm
Inactive silicon (n-type), ~100nm
Window:
Be, BN,
C (diamond),
or polymer
0.1 - 7
m
Ice
- 1000 V
Microcalorimeter EDS
(9 eV)
WDS (TAP)
(3 eV)
39.9
+0.8%
59.9
60.5
+1.0%
79.8
79.7
-0.1%
These errors are well within the error distribution seen with WDS.
Relative error = 100%x(measured - true) / true
Beam energy: 20 keV
Standards: pure elements
Matrix corrections: NIST ZAF
Analysis:
Test all combinations of K, L, and M lines
Calculate oxygen by stoichiometry; not included in the
error distributions
Error bins
are 5% wide
Frequency
40
30
-50
50
100
40
Count
20
10
-100
-75
-50
-25
25
50
75
100
Ich = ( NA Ci /A) R
(Q / (dE/ds) dE
E0
5 10
-20
-keV
4 10
3 10
2 10
1 10
Green-Cosslett
Mott-Massey
Powell (1996)
-20
-20
-20
-20
0
0
2
4
6
Overvoltage (U = E/E
8
)
10
Beam
electron
Beam
electron
Fluorescence Yield
Ejected
electron
Auger
Auger Emission
0.5
0.4
0.3
X-ray
n
Characteristic X-ray Emissi
0.2
0.1
0
0
10
15
20
25
Atomic number
Ix Q *
30
35
40
Energy Loss
R = Fraction
of possible
ionization
actually
deposited
in target
X-ray Absorption
I / I0 = exp [-(/) s]
Secondary Fluorescence
X-ray, E > Ec
Incident Primary Electron, E
Scattered Primary Electron, E - E
c
Auger
Auger Emission
Ejected electron
X-ray
Spectrometer Efficiency,
Imeasured = Igenerated * f(x) *
k = isample / istandard
= (Igenerated * f(x) * )sample /(Igenerated * f(x) * )std
Spectrometer efficiency cancels quantitatively
when the k-ratio is measured because exactly the
same X-ray energy is measured for both! The
efficiency does NOT cancel in standardless!
EDS Detector
Spectrometer Efficiency
0.45 m diamond
7.6 m beryllium
Be (7.6 m)
0.5 m
1 m
Frequency
30
25
20
15
10
5
0
-150
-100
-50
0
50
Relative Error, %
100
150
Error bins
are 5% wide
Frequency
40
30
-50
50
100
Conclusions
What to do about standardless analysis?
1. If standardless analysis must be used for an unknown,
report a realistic error estimate for the systematic error (ask
manufacturer for an error histogram) or use the error
histograms reported above, with a suitable caution.
2. If a known standard similar to the unknown is available,
an error estimate can be based on analysis of the known.
3. Dont report a numerical concentration. Use category
names instead, for example:
Major: > 10%
Minor: 1 to 10%
Trace: < 1%