If you never have done EPMA before and plan to bring your samples to our lab,
you MUST make sure you have a MIRROR-LIKE POLISH on them. This is minimally
a 1 micron polish, and preferably sub-micron. This is known as a
"probe polish" and must be applied for proper, correct microprobe analysis
(see red paragraph below).
I have had too many new probe users from other
institutions show up with samples that have a "standard
petrographic polish" of 18 microns. This is not a polish for a microprobe. If
you are not sure, please contact us first, rather than show up and we have to
spend hours trying to improve the polish in a less than optimal setup.
Electron microprobe analysis -- or electron probe microanalysis --
or electron microbeam probe analysis -- is a technique developed by R.
Castaing in the late 1940's as his Ph.D. dissertation in Paris.
A beam of energetic electrons, of say, 15,000 volts (but there is nothing
sacred about 15 KeV; sometimes I use 7 KeV, other times 30 KeV) is focused
upon an object of interest. The electrons are scattered elastically,
resulting in an 'excitation volume' much larger than the actual diameter
of the beam -- this volume usually having a diameter on the order a several
microns.
A variety inelastic scattering events also occur (if
they didn't, I wouldn't be writing this), yielding information about the
chemistry and structure of the target. The products that we mainly are
concerned with give us chemical information at the ~micron spatial level:
X-rays
Back-scattered electrons
Light (cathodoluminescence)
and to some extent, secondary electrons, which tell us about surface
topography.
I shouldn't need to remind you, but we can look at x-rays (and other
packets of energy) from either their particle description (so many electron
volts of energy) or from their wave-like feature (wavelength, in say angstroms
or nanometers).
Electron microprobes are outfitted with wavelength dispersive spectrometers
(WDS), and in many if not most cases have
an energy dispersive one (EDS) as well.
WDS operates with a small fraction of the x-rays that escape from the
sample, reaching a crystal, and a particular wavelength will be defracted
at a particular angle (the Bragg angle) into a gas-filled amplification
tube, where a pulse is produced that is proportional to the energy of the
x-ray. A WDS spectrometer is tuned to one wavelength and sits there and
counts the numbers of x-rays. You might compare it to a serial information
source, getting one channel of information at a time.
EDS, on the other hand, looks at all the x-rays coming off the
sample, and stretching the analogy, is a parallel information gathering
device, collecting all the different wavelengths. The EDS detector is a
solid-state device, where electron-hole pairs are produced for each x-ray
that hits it, with the number of pairs a function of the total x-ray energy
divided by the energy needed to produce one pair. This small signal is
amplified, and then becomes one bit of information in a spectrum gathered
for the sample ('multi-channel analyzer').
WDS and EDS each have their strengths and their weaknesses (more below)
Standards and calibration: the quantitative analytical routine, in
a nutshell, is in two parts: first you determine the x-ray count rate on
standards (should be well characterized and homogeneous material), then
you collect x-rays on your unknown usually the same conditions, and then
ratio the two. That gives you the "k-ratio", which is roughly
equivalent to the weight fraction of your unknown, if the standard were
pure element. The second part is that there must be a matrix correction:
that is, the x-rays you collect reflect several processes, with a major
one being absorption as they exit the sample. This correction has several
formulations, with 'ZAF'' a standard type.
The correct matrix correction assumes that the
sample has mirror-like "probe polish", because inherent in the correction
for absorption within the sample, is the assumption that the X-RAY PATH
LENGTH through the sample is the same in all directions, to all 5 (our case)
spectrometers. If the sample has a ROUGH surface, then there is no way that
the matrix correction will properly correct for the absorption effect, which
in many cases is large (25-50% of the X-rays never make it out of the sample,
for one X-ray line, but perhaps only 5% for another X-ray line).
Good standards are essential, and you may be called upon to produce
them yourself if we do not have appropriate ones. Always ask first.
Images are very important, both for determining locations to analyze,
as well as documenting what you analyzed. Backscattered electron (BSE)
images are one main image collected, and are produced rapidly for areas
up to 1-2 mm in dimension. Larger areas are possible but might take several
hours. X-ray maps can also be obtained, and generally take significantly
longer to produce, due to the lower efficiency of X-ray acquisition by
WDS. Alternately, x-ray maps using EDS can also be obtained, but the thru-put
of x-rays in the channels per element is not much better than WDS, so again
it takes some time for larger areas.
BSE images can give important information
about the spatial relationships of adjacent phases, plus about zoning and
inclusions within phases. BSE intensity is, to a first approximation, a
function of the chemical composition: the brighter an area, the greater
the mean atomic number of that area relative to adjacent areas. Thus ilmenite,
magnetite and chromite will be brighter than most adjacent silicate minerals.
Apatite or zircon similarly. Olivine or pyroxene are brighter than adjacent
feldspars, though the Fe/Mg ratio is of course important.
Spatial resolution of a chemical analysis: because of the elastic scattering
of the electrons, the x-rays are produced in an 'interaction volume' that
can be several microns across and deep. This area is a function of two
factors: the composition itself, so that in high atomic number materials
the scattering is less than in lower atomic number ones; and the accelerating
voltage: the more energy the electrons have, the greater distance they
can scatter. It is possible to utilize minimal accelerating voltage, to
minimize the scattering. I may be possible, in specifica cases, to constrain
most of the electrons and thus most of the generated x-rays, to a layer
of say 1 micron. But this is not typical....everything is in the details,
like the exact composition of your sample. There are
some very nice Monte Carlo simulations that you can use to model the scattering
distance. (links on my 777 class page)
Light element analysis: this is somewhat complicated, and is do-able
in theory, and in reality in some cases. Appropriate standards, conductive
samples (no carbon coating necessary) and low accelerating voltages are
essential.
Trace element analysis: what do you mean by trace? We can measure down
to the hundred ppm level in many samples, depending on their nature and
how well they can stand up to high current. In some samples, 50 ppm or
below is possible, but again "it all depends on what you got".
Trace element mapping is virtually impossible.
But CL can tell you about spatial distribution of some trace elements...
Coating of specimens -- if your samples are not conductors, they need
to be coated, and ONLY coated in our lab if you are using our standards.
We have had too many problems with samples coated elsewhere.