Electron probe microanalysis
EPMA
Light Element Analysis
Modified 4/7/14
What’s the point?
What are the problems analyzing light elements?
How precise/accurate are such analyses?
Which Elements?
Ka X-ray Data
Element
Be
B
C
N
O
F
Z
4
5
6
7
8
9
Wavelength Å
114
67.6
44.7
31.6
23.6
18.3
Energy (eV)
109
183
277
292
525
677
Edge (eV)
112
192
284
400
532
687
We are concerned here primarily about the K lines of elements Be-F,
although many of the concerns affecting their analysis is also true for
L and M lines of heavier elements that fall in this low wavelength
(low energy) realm.
(Now, >2012, Li is doable…see last slide)
Issues for Low eV X-rays
• Decreased X-ray generation (low fluorescence yields vs high Auger
production)
• Absorption of these weak X-rays by mass in “path length” in material
• Interferences by L and M lines of higher Z elements, as well as higher
order lines
• Wavelength shifts and changes in shape of peaks, due to chemical effects
(e.g., valence electrons involved in X-ray generation), and sometimes
crystallographic orientation (polarization)
• Possible errors in matrix correction due to poorly known values especially
of m.a.c.s
• Low E0 operation beneficial (decreased range, thus decreased path
length), but then coating (and contamination) and thin film effects magnified
Long Wavelength Spectrometry
TAP and stearate crystals, or the newer layered synthetic (“pseudo
crystals”) have been the two options up to ~2012. Each category has
positive and negative features:
• TAP and stearate: better spectral resolution (avoid interferences),
but peak shape differences accentuated and there are lower count
rates; stearates somewhat less stable.
• LSMs: poorer spectral resolution (interferences unavoidable),
higher orders suppressed, and there are much higher count rates.
• Diffraction Gratings: JEOL (and Cameca?) have looked at
diffraction gratings, though only JEOL is commercially selling for a
very special spectrometer. Possibly combining LSM with groves laid
in it.
Fluorescence Yields
The yields of K lines of B-F are
<0.05, as well as the L and M
lines of the higher Z elements
that fall below 1 keV (longer
than 12 Å).
2006 comment: We are discussing this in one context--light
elements -- but need to view the wider issue too.
While the above may be strictly correct, it is also somewhat
disingenuous. Recall the F.Y. means only the fraction of x-rays
vs fraction of Auger electrons being produced by inner shell
ionizations…it says nothing about the actual number of x-rays
you can count on your spectrometers! At 7 kv and 20 nA I can
easily generate 50,000 counts a second of B Ka -- in pure B.
Fluorescence Yields
The yields of K lines of B-F are <0.05, as well as the L and M lines of
the higher Z elements that fall below 1 keV (longer than 12 Å).
…it says nothing about the actual number of x-rays you can count on
your spectrometers!
This will be a function of the “efficiency” of the crystal, the sin theta
position, and the ionization efficiency of the P10 gas for the x-ray.
Example: The chart would imply that Hf* La counts>>Hf Ma counts.
But at 18 keV, 20 nA,
* Hf: Z=72
Hf Ma (TAP) yields 8362 cps
Hf La (LIF low pressure) yields 844 cps
Hf La (LIF high pressure) yields 2509 cps
why?
Fluorescence Yields…aren’t everything
Line
counts
xtal
Sin 
Gas P
Overvoltage
MAC Hf
by Ar
Hf Ma
8362
TAP
.293
Low P
10.8
870
Hf La
844
LIF
.390
Low P
1.9
124
Hf La
2509
LIF
.390
High P
1.9
124
Absorption
X-rays with low energies (e.g. < 1
keV) have increased problems with
absorption:
• within the sample and standard
(compare emitted from generated C
Ka in B4C matrix, Fig. 1)
• by any intentional (e.g. C-coat) or
unintentional (contamination,
oxidized) thin film or coating on
sample and standard
• by windows or diffracting crystal
(though these are the same for both
standard and sample and thus ‘cancel
out’)
Fig 1: from Bastin and Heijligers, 1992, Present and future of
light element analysis with electron beam instruments.
Microbeam Analysis, 1, 61-73.
M.A.C.s
An additional complication is
that many mass absorption
coefficients for use with light
elements are not known with
great accuracy, as shown in
the adjacent table of m.a.c.s
for O Ka X-rays, which show
the ‘best’ determinations by
Bastin and Heijligers (1992),
two pre-eminent researchers
in this field.
Bastin and Heijligers, 1992, Present and future of light element analysis
with electron beam instruments. Microbeam Analysis, 1, 61-73.
Peak Shifts and Shapes
C Ka
The electrons involved with the
transitions yielding the C ka X-ray
are valence electrons. Differences in
bonding are reflected in differences
in the shapes of the peaks (including
shifts of the maxima).
Fig 18.1, from Reed 1993, p. 275
Here the Fe La and Lb peak shapes
and intensities are functions of the
bonding (valence states). Attempts
have been made to utilize this for
determining valences of Fe and Mn
compounds; results are complicated.
(From Meeker and Lowers, Standards for the analysis of geological and
ceramic materials, Slide 813, NIST-MAS Workshop, April 2002.)
Deconvolution of F peaks
This pair of figures provides some explanation why these low energy
peaks behave as they do, by deconvolution into the (apparent) related
peaks that the spectrometer does not have enough resolution to separate
out. Fluorite (left) and F-topaz (right).
Fig 4, Fialin et, 1994, Microbeam Analysis
Measuring Light Elements:
peaks or integrals?
The previous slide demonstrates a complication for light element
analysis: a simple measurement of intensity at the nominal
(=standard’s) peak position, may result in an incorrect quantitative
analysis. Clearly, the whole area under the peak is the true
representation of the X-ray’s intensity. However, to measure the whole
area (the integral) is a time-consuming task — particularly as you must
measure from the peak maximum to minimum, and the count rate on the
latter is low, so to achieve high precision, the count must be repeated
many times. One does not want to do it very often.
A time-saving alternative procedure is to first spend some time
acquiring both peak counts and integrals (=wave scans) of the standard
and each typical phase. A mathematical ratio, the area peak factor,
APF, is then calculated which can then be invoked within the matrix
correction. Then, for each unknown, only the peak (and background)
needs to be measured.
Area Peak Factors
I
APF =
I
integral
unk
peak
unk
peak
std
integral
std
I
×
I
If APFs are to be used,
that option can easily be
turned on in the
Analytical Options menu
at the very top of the
PfW window.
Prior to that, either APFs
determined for element
pairs (left) or for a
specified phase (under
Elements/Cations) must
be determined and
entered into the run.
C Ka
APF compilations
Bastin and Heijligers, 1992, Present and future of light element analysis
with electron beam instruments. Microbeam Analysis, 1, 61-73.
Crystal types and peak shapes
Bastin and Heijligers, 1992, Present and future of light element analysis
with electron beam instruments. Microbeam Analysis, 1, 61-73.
In many cases,the
synthetic crystals
are ‘better’ than
conventional
stearate crystals:
the count rates are
much greater
(compare top 2
spectra). And the
synthetics greatly
dampen out the
higher order
reflections
(compare higher
order Nd
reflections, bottom
left vs right).
Crystal types and peak shapes
5 Nb La 28.6 A
5 Nb Ln 31.1 A
5 Nb LL 32.6 A
Nb La n=5
Nb Ln n=5
Nb LL n=5
6 Nb La 34.3
Nb La n=6
Bastin and Heijligers, 1992, Present and future of light element analysis
with electron beam instruments. Microbeam Analysis, 1, 61-73.
In many cases,the
synthetic crystals
are ‘better’ than
conventional
stearate crystals:
the count rates are
much greater
(compare top 2
spectra). And the
synthetics greatly
dampen out the
higher order
reflections
(compare higher
order Nd
reflections, bottom
left vs right).
But not always...
F Ka
In some cases, the ‘traditional’
crystal is preferred over the
synthetic. The count rates are
lower, but critical first order
interferences can be avoided. One
case is the measurement of F in
biotite, where Fe L peaks are
present (not to mention n>1 Mg,
Fe and Al). In this case, I find the
TAP crystal preferable.
I need to revise this out-of-date
slide! I now always use PC0
(45Å LSM) which gives enough
separation from Fe L peaks, and
plenty of F Ka counts.
Crystallographic effects
Issue is orientation between polarized X-rays from the sample (here,
sample is rotated around one axis) and the analyzer crystal. Worst
case is hexagonal compounds; rhombohedral Boron does not show
these effects.
(From Meeker and Lowers, Standards for the analysis of geological and
ceramic materials, Slide 813, NIST-MAS Workshop, April 2002.)
Interferences
Mz?
This older listing by an
EDS manufacturer of
elements that fall in the
“light element” region is
somewhat misleading ...
unintentionally, of
course, as it most
certainly predates the
1990s. You can see that
there are some L lines
that fall in the region, but
there is no indication of
any M lines....
Mz = Mz = M5 – N2,3
Interferences (as shown by Virtual WDS)-1
B
C
N
Fortunately, Virtual
WDS provides an
indispensible
service for light
element analysis,
by showing all the
potential
interferences.
Green indicates
first order
interferences, and
blue higher order
ones.
Interferences (as shown by Virtual WDS)-2
O
F
Note the large
number of
interferences for O
and F ka. This
increases the
difficulty of
analysis of those
elements.
PHA can help
Higher order peaks interferring with the long wavelength
(light energy) peaks can be largely eliminated (though not
100%) by setting the PHA to differential mode and a window
to cut out the higher order lines (here, 2Cr and 3Ni.)
Fig 18.4, from Reed 1993, p. 278
Some (random) particular issues
• Be: rarely analyzed (heavily absorbed); polarization
• B: Interference by Cl L, Mo, Nb, Zr M lines; specular
reflectance of Si L with PC3 (but not PC2)
• C: diamond not good standard
• N: interference by Ti L
• O: Interference by 2Na K; just above C edge, heavily
affected by carbon contamination or coat thickness
• F: Interference by 3P Ka, Fe L, Ce Mz; mobility in
apatite (orientation)
Boron
3000
B Ka/10
Si L sp.ref.
2500
Mo Mz
1500
B-free Mo5Si3 appears to have
3.82 wt% B, or 19.7 atomic
%B – a major error
1000
500
Wavelength (A)
80
70
60
0
50
Counts
2000
Example of pathological
interferences on B ka: Mo Mz
falls very close to the peak, and
the strong spectral reflectance
of the Si La line greatly
complicates determining the
background.
With PfW, we can define
curved backgrounds, and
subtract the interferences
within the matrix correction,
yielding excellent results.
Oxygen
In “normal” geological epma, oxygen is not a
critical element, and is input into the matrix
correction by stoichometry. However, there
are instances where oxygen cannot be defined
by stoichometry, and must be explicitly
determined.
Nash (1992) evaluated oxygen in a variety of
minerals and glasses at 15 keV, and found that
the precision of O in silicate glass is 0.6%
relative, comparable to that of Si. “Optimum
results occur when standards similar in
composition to unknowns are used and care is
taken in standard and sample preparation. The
procedure is particularly advantageous in the
analysis of hydrous glass and minerals in
which conventional microprobe analyses
yield totals of less than 100% because O
attached to H remains unmeasured.”
Fe3O4 and Fe2O3
If oxygen is to analyzed, it is
critical to carbon coat both
standards and unknowns at
the same time.
Nash, 1992, Analysis of oxygen with the electron
microprobe: applications to hydrated glass and
minerals. American Mineralogist, 77, 453-457.
Carbon contamination
Hydrocarbons in the chamber inevitably will
find their way to the beam, which apparently
cracks them, and a layer builds up with time.
This thin layer of carbon has a strong absorption
of oxygen Ka as well as other light elements.
(Traditional) Anti-contamination
Curve A shows a time-dependent build up of carbon on a polished brass
sample. Anti-contamination devices that could be used to reduce C build up
include a cold finger located a few mm from the beam above the sample (b),
and an air jet -- a microtube that shoots a thin stream of air at the point of
beam impact on the sample surface (c: apparently oxidizes the hydrocarbon
and yields a gas that is pumped away). The optimal procedure is to use both
(d).
Fig 18.4, from Reed 1993, p. 283
Best “Anti-contamination Device”
is Oil-free Vacuum Pumps!
Other coatings
Thin layers or coatings (such as oxide layers on the surface of
metals like Al, Ti, Mn etc) similarly can affect X-ray
production, and then absorption of X-rays in the matrix
below.
We will discuss “thin film” analysis shortly.
“Soft X-ray Emission Spectrometer”
Within the past 2-3 years, JEOL has introduced the SXES which
permits EPMA of hitherto inaccessible element, Li.