JEOL Probe Users Listserver
Moderator: Ellery Frahm, [log in to unmask],
Electron Microprobe Lab, University of Minnesota
Post a message: send your message to [log in to unmask]
Unsubscribe: send "SIGNOFF PROBEUSERS" to [log in to unmask]
On-line help and FAQ: http://probelab.geo.umn.edu/listserver.html
Did you standardize with the same beam size as was used on the
unknown glasses? I do not like using a 30 micron beam because X-rays
generated outside the area that is actually analyzed by the detector
can enter the analytical area and fluoresce additional X-rays that are
not accounted for in the F correction. I would try a 10 micron beam
and run the analyses on the unknowns at 3 nA.
When you estimated H2O by difference from 100 %, did you include
the H2O in the ZAF program or did the program reduce the data not
"knowing" about the extra H2O? The latter could lead to errors
particularly in Na-Si corrections due to the missing O and the
somewhat different average atomic number (depending in part on whether
the program calculates the Z with missing elements as having an atomic
number of zero).
Analysis of O is dependent on the C thickness because of the
strong absorption of O Ka by C. If possible, the standards and
unknowns should be coated together. The problem could be tested in
part by doubling the coating and repeating the analysis for O.
As I understand it, the O analysis technique has been able to
work on rhyolite glasses. Have other andesitic glasses given
reasonable results for H2O in the past? I wonder if this particular
sample is not adequately grounded. What does the beam current and the
sample current look like during the analysis, and is the sample
current similar to that on andesite standards?
I have learned from Youxue Zhang and his students here over the
years that FTIR on real andesitic glasses is very challenging because
of the large background induced by Fe. I wonder if there could be
serious errors in the FTIR data.
> Probe folks,
> A researcher in our lab (an experimental petrologist) is trying to
> analyze a few of his samples (experimental andesitic melts/glasses
> containing between 3% and 6% water), and he is trying to (1) get
> good totals and (2) correlate measurements of the water content
> using FTIR and oxygen analysis in the microprobe. He is getting a
> linear fit between water contents from the FTIR and microprobe, but
> his totals are low on the probe when we analyzes his samples as both
> oxides and metals. He is an experienced user, and we've gone
> through all of the usual procedures for testing the causes of low
> totals -- nothing.
> The FTIR measured total water contents ranging of 3.37, 4.12, 4.54,
> 5.25, 5.45, and 5.88 to 5.9 wt%. When we analyze the glasses with
> the electron microprobe, assuming a deficit from a 100% total is
> attributed to H2O, the calculated water contents are 5.29, 6.45,
> 7.12, 7.75, 7.56, and 9.21 wt. %, respectively. So his totals are
> about 5% low, occasionally more, causing overestimation of water.
> Here is what he has tried (based on the notes he's given me):
> 1) Periodically aligned beam and checked for astigmatism on
> 2) Beam conditions: 10 nA current, 30 micron spot size.
> 3) Always analyzed Na first (monitored counts rates, found Na
> counts did not drop for at least 30 seconds).
> 4) Analyzed for oxygen using three different standards: enstatite,
> pyrope, and almandine -- all yielded similar results.
> 5) Counted the area under the peak for oxygen.
> 6) Adjusted the detector settings for oxygen: PHA Gain: 32, High V:
> 1708, Base L. (V) 0.70, Window: 2.5, Mode: Diff. Set up this way to
> filter out high-order Al and Na peaks. WDS scans indicated these
> peaks were eliminated by this filtering method but had little effect
> on improving the water contents of the glasses when compared to an
> “unfiltered” method.
> 7) Set oxygen backgrounds based on WDS scans and software overlap
> database: Back +: 12.000 mm, Back -: 6.500 mm.
> 8) Total counting time on oxygen: ~44 seconds. Other elements: 10
> seconds for peak, 5 seconds for background.
> 9) Elements present in samples (and analyzed for): Si, Ti, Al, Cr,
> Fe, Mn, Mg, Ca, Na, K, O, H (known from FTIR)
> 10) Full WDS scans (including light elements) revealed no missed
> elements that could have led to low totals.
> 11) Used basaltic glass to standardize MgO, CaO, FeO. Rhyolitic
> Glass (basaltic glass) for Al2O3. SiO2 Glass (Quartz, basaltic
> glass, rhyolitic glass) for SiO2. Benitoite for TiO2. Chromite for
> Cr2O3. Mn-Hortonolite for MnO. Omphacite (albite) for Na2O. K-Spar
> for K2O. Standards in brackets were also tries, and produced worse
> results. Standardized as oxides, and then measured the unknowns as
> a metal. Oxygen was standardized as a metal on the enstatite
> standard, and analyzed as a metal on the unknowns.
> 12) Sample thicknesses were 100 microns or greater.
> 13) Two carbon coat thicknesses tried: about 50 and 200 angstroms
> -- produced similar results, so charging isn't a problem.
> 14) Glasses were free of quench crystals, so the interaction volume
> should be homogeneous.
> 15) Moving elements to different spectrometers had no effect on
> improving the results.
> 16) Using these conditions, analyses on two different basaltic
> glass standards yielded matching elemental concentrations. These
> glasses were relatively dry (a few tenths of a wt.% H2O) compared to
> the specimens in question.
> 17) This technique was also performed on hydrous rhyolitic glasses
> with 1.3, 3.3, 4, 5, and 6 wt. % H2O (determined by FTIR) and
> produced the similar water contents (1.06, 3.11, 4.74, 5.65, and
> 7.24, respectively) and overall chemistry.
> The only real issue I see above is the benitoite as a Ti standard
> because it has CL, but the TiO2 content of these glass samples isn't
> high enough to account for the difference. Because these glasses
> are amorphous (presumably), even most of my usual "exotic"
> explanations for low totals are ruled out, like X-ray polarization
> between the crystals in the sample and the dispersing crystal. I'd
> like to have a better explanation for these low totals other than
> "well, something is causing the electrons and/or X-rays to be more
> strongly absorbed or behave in an unexpected way" -- that's not very
> So I need to get some fresh ideas or find out if others have had
> similar problems with hydrous andesitic melts. Anyone have an ideas
> about what to try next or what the problem might be (even if we
> can't fix it)?
> Ellery E. Frahm
> Research Fellow & Manager
> Electron Microprobe Laboratory
> University of Minnesota - Twin Cities
> Department of Geology & Geophysics
> Lab Website: http://probelab.geo.umn.edu
> Personal Website: http://umn.edu/~frah0010