Energy Dispersive Spectrometry (EDS)

1
UW- Madison Geology 777
Version Last Revised 8/21/08
Electron Probe MicroanalysisEPMA

• Energy Dispersive Spectrometry (EDS)

2
Whats the point?
UW- Madison Geology 777
Using X-rays to produce e-hole pairs (charges
proportional to X-ray intensity), which are
amplified and then digitized, put in a
histogram of number of X-rays counts (y axis)
versus energy (x axis). A solid state technique
with unique artifacts.
EDS spectrum for NIST glass K309
(Goldstein et al, Fig. 6.12, p. 356)
3
Summary
UW- Madison Geology 777

• X-rays cause small electric pulses in a solid
  state detector. Associated electronics produce
  instantaneously a spectrum, i.e. a histogram of
  count (number, intensity) vs the energy of the
  X-ray
• Relatively inexpensive there are probably
  50-100 EDS detectors in the world for every 1 WDS
  (electron microprobe)
• Operator should be aware of the limitations of
  EDS, mainly the specific spectral artifacts, and
  the poor spectral resolution for some pairs of
  elements

4
Generic EMP/SEM
UW- Madison Geology 777
Electron gun
Column/ Electron optics
Optical microscope
EDS detector
Scanning coils
SE,BSE detectors
WDS spectrometers
Vacuum pumps
Faraday current measurement
5
EDS assemblage
UW- Madison Geology 777
Goldstein et al fig 5.21
There are several types of solid state EDS
detectors, the most common (cheapest) being the
Si-Li detector. Components thin window (Be, C,
B) SiLi crystal, FET (field effect transistor
initial amp), cold finger, preamp, vacuum, amp
and electronics (single channel analyzer).
6
EDS Windows
UW- Madison Geology 777
Windows allow X-rays to pass and protect
detector from light and oil/ice. Be The most
common EDS detector window has been made of Be
foil 7.6 mm (0.3 mil) thick. It allows good
transmission of X-rays above 1 keV. It is
strong enough to withstand venting to atmospheric
pressure, and opaque to optical photons. Thin -
Ultrathin For transmission of light element
X-rays (lt1 keV), windows 0.25 mm thick of BN,
SiN, diamond or polymer are used. They must use
supporting grids to withstand pressure
differentials the grid (e.g., Si or Ni) takes up
about 15 of the area, but thin enough that low
energy X-rays pass through.
This plots shows the transmittance of X-rays thru
different types of window material. (Quantum BN
0.25 um, diamond 0.4 um). The higher the
transmission number, the better
Windowless Here there is no film, and there is
a turret that allows swapping with a Be window.
Difficult to use as oil or ice can coat the
detector surface. Not used much.
Goldstein Fig. 5.41, p. 318
7
How it works energy gap
A semi-conductor like Si has a fully occupied
valence band and largely unfilled conduction
band, separated by an energy gap (1.1 eV).
Incident energy can raise electrons from the
valence to the conduction band.
X-ray hits the SiLi crystal, producing a specific
number of electron-hole pairs proportional to
X-ray energy e.g. one pair for every 3.8 eV, so
for incident Fe Ka, 6404 eV, 1685 e-hole pairs
are produced. With a bias applied across the
crystal, the holes are swept to one side, the
electrons to the other, producing a weak charge.
Boron is important acceptor impurity in Si
and degrades it (permits thermal excitation
bad) Li is drifted in (donor impurity) to
counter its effects.
1.1 eV energy wasted in lattice vibrations,
etc
Goldstein et al, Fig 5.19
bias a voltage is applied between 2 points
e.g. one 1500 v, other -1500 v.
8
How it works inside the detector
UW- Madison Geology 777
X-rays are absorbed by Si, with photoelectrons
ejected. This photoelectron then creates
electron-hole pairs as it scatters inelastically.
The Si atom is unstable and will either emit a
characteristic Auger electron or Si ka X-ray. If
Auger, it scatters inelastically and produces
electron-hole pairs. If Si Ka X-ray,
it can be reabsorbed, in a similar process, or it
can be scattered inelastically. In either case,
the energy will end up as electron-hole pairs.
The result, in sum, is the conversion of all the
X-rays energy into electron-hole pairs -- with 2
exceptions.
Fig 9.5 Reed Fig 5.22 Goldstein
9
Artifacts Si-escape peak Si internal
fluorescence peak
UW- Madison Geology 777
There are 2 exceptions to the previous neat
explanation of how the Si(Li) detector
works.Si-escape peaks are artifacts that occur in
a small of cases, where the Si ka X-ray
generated in the capture of the original X-ray
escapes out of the detector (red in figure).
Since this X-ray removes 1.74 keV of energy, the
signal generated (electron-hole pairs) by the
incident X-ray will be 1.74 keV LOW. This will
produce a small peak on the EDS spectrum 1.74 keV
below the characteristic X-ray peak. Another
artifact is the Si internal fluorescence peak,
which occurs if an incident X-ray is absorbed in
the Si dead layer (green
region). This region is dead to production of
electron-hole pairs, but Si ka X-rays can be
produced here which then end up in the live
part of the detector, and result in a small Si ka
EDS peak.
Consider Ti
Fig 5.22 Goldstein et al
10
Artifacts Si-escape peaks Si internal
fluorescence peak extraneous peaks
UW- Madison Geology 777
The figure shows a real spectrum of a sample of
pure Ti metal -- but there are 7 peaks besides
the Ti Ka and Kb. At 1.74 keV below each, are the
respective escape peaks (blue arrows). Also
present is a Si internal fluorescence peak (green
arrow). The Fe and Cu peaks are from excitation
of metal in chamber or sample holder by BSE or Ti
X-rays. Note the sharp drop in the background
intensity on the high side of the Ti Kb peak (
Ti K absorption edge, red arrow). (2 Ti Ka and Ti
KaKb explained shortly.)
Note the scale of the spectrum the Ti Ka max is
1.3 million counts. These effects are generally
weak, but evident when you are looking for minor
elements.
Goldstein et al Fig 5.39,p. 316
11
Question Do all characteristic X-rays have
Si-escape peaks in a Si(Li) detector?
Why or Why Not?
UW- Madison Geology 777
Hint 1 Sr La does not, but Os Ma does Hint 2
Look up the characteristic energies of each Hint
3 Look up the absorption edge (critical
excitation) energy of Si Ka Hint 4 Compare the
numbers in 2 to number in 3. Which one will
greater than the one in 3? Would a Si Ka x-ray
produced in the sample, which then makes its way
thru the vacuum to the EDS detector, have enough
energy to knock out the inner shell (K) electron
of the Si detector crystal?
12
Signal processing
UW- Madison Geology 777
Si(Li) detector has no internal gain for Ca Ka
photon with 1000 e-hole pairs, the charge is
only 10-16 Coulomb (weak!) Directly coupled to
the Si crystal is a field effect transistor (FET)
that converts charge to voltage, followed by a
preamp. We need low noise, high gain
amplification, so the
Si detector and FET are cooled to about 100K with
liquid nitrogen (LN) to prevent noise (and
prevent diffusion of Li in detector--at least in
old ones). More signal gain provided by main
amplifier (signal now boosted to 1-10 volts)
where also RC (resistor-capacitor) circuits are
used to shape the pulse, to maximize signal/noise
ratio and minimize pulse overlap at high count
rates. Then ADC (analog to digital converter)
outputs data to the screen as a spectrum display.
gain electronic multiplication of signal
intensity
13
The first signals in the EDS detector
UW- Madison Geology 777
The set of electron-hole pairs produced by
the impact of the X-ray on the Si(Li) detector
produces a tiny charge (10-16 C), very quickly
(150x 10-9 sec). The FET(preamplifier) changes
the charge (capacitance) into a tiny voltage
(millivolts). These steps are shown in the first
half of (a) to the right. The output of the FET
is shown below at (b) where the x axis is time
and y is voltage. The jump represents the
presence of a voltage proportional to the number
of electron-hole pairs generated by each X-ray,
so Photon 2s jump is of a higher energy than
Photon 1s jump which is higher than Photon 3s
jump. At a certain point the FET reaches the
limit of the number of
charges it can hold, and then there is a reset or
zeroing back to some baseline where it starts
over. Following this are electronics to shape
the voltage into a pulse that can be counted.
Goldstein et al (1992), p. 297
14
UW- Madison Geology 777
Processing Time and Pulse Pileup Rejection
The user can tweak the time constant (T.C.)
which sets the time allocated in the electronics
to process each pulse (X-ray). In the top
figure, a short T.C. (1 ms) permits each pulse to
be counted correctly. A longer T.C. (10 ms) means
the gate is open longer and a second pulse can
enter and be incorrectly added this is pileup
and causes distorted spectra. Therefore, circuits
are added (4, bottom figure) to sense when
pileup occurs and to ignore that pulse.
Goldstein et al (1992), Fig. 5.24 and 5.25, p.
300
15
Dead Time
UW- Madison Geology 777
Deadtime is the period during which the
detector is busy and cannot accept/process
pulses. This can introduce error unless it is
accounted for, either by extending counting time,
or correcting for it in the software. In most
systems, the user sets the live time which is
the time during which counts are actually
counted, and the real time is automatically
determined by the electronics or
software. Optimal deadtime is in the 35-45
range. This optimizes both user/machine time and
moderate to high throughput of counts.
Goldstein et al (1992), Fig. 5.25 (p. 300)
and Goldstein et al (2003), Fig. 7.9 (p. 307)
16
Detector performance peak resolution (FWHM)
UW- Madison Geology 777
The characteristic X-rays generated in the
specimens are very close to lines, i.e. only a
few eV wide at most. However, the conversion of
X-ray to a pulse in the detector has several
variables (imperfections) that broaden the peak
to between maybe 135-200 eV, depending upon the
type of detector and how well maintained it is.
The narrowness of the peak is measured by the
width of the peak at one half the maximum
intensity of the peak -- this is what is termed
the FWHM.
In EDS detectors, it is usually measured at the
Mn Ka position, with values of 160 eV and below.
Modern (2005) one are quoted at lt130 eV.
Goldstein et al, Fig 5.34, p. 311
17
Why Mn Ka for EDS resolution?
UW- Madison Geology 777
EDS companies (their engineers mainly) do not
want to have to carry around an SEM or EMP to be
able to test, repair and calibrate an EDS system.
Instead they carry a small 1 diameter x 2 long
tube that fits over the end of the EDS snout.
Inside it is an Fe-55 isotope source (half life
2.7 yr) which emits an intense x-ray at 5.985 keV
which is only a few eV different than Mn Ka.
18
Spectral processingbackground correction
UW- Madison Geology 777
The characteristic X-rays that we need to
quantify ride atop the continuum, and the
continuum contribution to the characteristic
counts must be subtracted. (Top) Linear
interpolation (B-D) will be in error due to the
abrupt drop of continuum at the Cr K-absorption
edge (5.989 keV). B-C is possible but critically
dependent upon having good spectral resolution
(lt160 eV). A-B would be preferable. (Below) Doing
background fit of a complex stainless steel.
Goldstein et al Fig. 7.1,2, p. 367
19
Spectral processingbackground modeling or
filtering
UW- Madison Geology 777
Correcting for the background is done by either
of 2 methods developing a physical model for the
continuum, or using signal/noise filtering.
Modeling is based upon Kramers Law there is a
function describing the continuum at each energy
level, that is a function of mean atomic number,
and measured detector response.
20
Background Modeling
UW- Madison Geology 777
The spectrum of Kakanui hornblende (top left),
with superimposed calculated (modeled)
background, based upon Kramers Law. Bottom shows
after the background has been subtracted. Cu is
artifact (stray X-rays). Mn is actually present
at lt700 ppm.
Goldstein et al Fig 7.4, p. 372
21
Background Filtering
UW- Madison Geology 777
Theoretically Fourier analysis will separate out
the low frequency continuum signal and high
frequency noise from the medium frequency
characteristic peaks however, there is overlap
and the result is a poor fit. A better filter is
the top hat filter, where no assumptions are
made about the spectrum, and only the
mathematical aspects of signal vs noise are
considered.
22
Top Hat Filtering
UW- Madison Geology 777
This filter (top right) moves across the EDS
spectrum (with an optimally defined window, 2
FWHM Mn Ka320 eV), and assigns a new value for
the center channel based upon subtracting the
values in the left and right channel from the
center (value hk chosen to total area 0). Thus,
in the simple spectrum (bottom right), the center
channel (), when the left and right channels are
subtracted, leaves a value 0.
FWHM full width at half maximum.
Reed Fig 12.7 p. 174,Goldstein et al Fig 7.6, p.
374
23
More Artifacts Pulse Pile Up
UW- Madison Geology 777
true single X-ray peak, and piles it up with
all the other peaks from the elements actually
present. For 2 major elements, could be 3 sum
peaks for 3, 6. In reality,you only see 1 or 2
unless you zoom in to the background level.
Always consider their possible presence.
There is a short period of time (t0) during each
X-ray capture by the EDS detector, when the
detector can capture a second X-ray by mistake.
The electronics cannot distinguish this sum
peak from a
24
Fools even the pros
UW- Madison Geology 777
Even the fanciest, slickest EDS setup can fool
newcomers, not to mention experienced users.
Above, is a partial spectrum (major peaks) of a
commercial glass that has a lot of Si and O, plus
Na and Al. Notice the S (Sulfur) label over a
peak around 2.3 keV sure looks like it might be
Sulfur, right? It is NOT, rather it is a sum peak
of O Ka (.525 keV) Si Ka (1.74 keV). Previous
experience with this fake peak had taught me to
be skeptical
25
Sum Peaks
UW- Madison Geology 777
In qualitative analysis of silicates, there are
some combinations of element Ka peaks that fall
close to Ka peaks of elements possibly present,
as indicated in the table below
26
And More Artifacts
UW- Madison Geology 777
There is always a potential for stray X-rays
being detected. It thus pays for the EDS operator
to understand what the path is for the electron
beam and for the X-rays, and know what other
elements might show up unintentionally. This is
particularly true for EDS associated with TEM,
where specimens routinely sit on grids (Cu?) and
the high energy (200 keV?) electrons can go
through the specimen and hit a metal part of
column or chamber, with the resulting X-rays
finding a way back to the detector.
27
And More Artifacts
UW- Madison Geology 777
Another thing many SEM labs use gold or
palladium coating on specimens. These very thin
coats will produce definite x-ray peaks!
28
And beware of lazy peak IDs
UW- Madison Geology 777
Even the fanciest, slickest EDS setup can make
misidentificationsso the analyst cannot get lazy
and assume just because the expensive software
said something was there, it was there. I knew
Arsenic was possible (As La identified), but
unlikely, and rather Mg Ka was more likely. To
confirm it was NOT As, I cranked the accelerating
voltage up to 20 keV (the K shell binding energy
is 11.9 keV) and found there was NO As Ka x-ray.
Ergo, not As.
29
Artifical EDS spectrum
UW- Madison Geology 777
Artificial no background, no artifacts, and
assumes EACH element at 100 concentration.
Why, then, the two slopes?? Peak intensities of
elements from Si to Na decrease, and also from Si
to Zn -- why? (Hint 2 physical phenomena)
30
Artificial spectrum
UW- Madison Geology 777
The actual spectrum of pure elements, as
generated at the point of impact, would be one
steady decreasing curve from Na down to Zn,
following the red curve superimposed here.
Slope down from Si to Zn there are less and less
X-rays being produced because the accelerating
voltage is constant (e.g. 20 keV) and the
overvoltage is lower.
Slope down from Si to Na X-ray energies are
increasingly weaker, and are absorbed both within
the specimen and by the window.
31
Evolution of EDS spectrum from the specimen to
the monitor - 1
UW- Madison Geology 777
The spectrum on our monitor (d) is a result of
many things impacting the real spectrum generated
within the specimen (a). At instant of generation
within the specimen, there is only the Ka, Kb and
continuum. An instant later (b), as the X-rays
leave the specimen, two things can happen some
of the continuum X-rays above 5.464 keV are
absorbed, producing the drop in the continuum
there.
Simulation of element (say V) X-ray generation
and display
Goldstein et al Fig 5.53 (by R. Bolon) p. 330
32
Evolution of EDS spectrum from the specimen to
the monitor - 2
UW- Madison Geology 777
Also in (b) the lower energy continuum is
absorbed, causing the dropoff in the spectrum
there. When the X-rays hit the detector (c), Si
fluorescence peaks can result. And after signal
processing (d), the display will show peak
broadening, sum peaks, Si-escape peaks, further
decrease of intensity and low energy noise.
Simulation of element (say V) X-ray generation
and display
Goldstein et al Fig 5.53 (by R. Bolon) p. 330
33
EDS-WDS comparison
UW- Madison Geology 777
34
Further EDS details
UW- Madison Geology 777
There are several modern EDS companies, with most
producing very informative brochures that go into
the technical details of EDS hardware (and
software) For example Oxford Instruments
ltwww.osinst.com/ANLPDP174.htmgt has a nice
technical publication EDS Hardware Explained
available as a pdf.