Author Topic: Best Practices for EPMA  (Read 1261 times)

John Donovan

  • Administrator
  • Emeritus
  • *****
  • Posts: 3223
  • Other duties as assigned...
    • Probe Software
Best Practices for EPMA
« on: June 01, 2023, 09:29:16 PM »
This is the presentation I gave just now for the China EPMA conference hosted by Institute of Mineral Resources, Chinese Academy of Geological Sciences.

The title is Best Practices in EPMA Microanalysis.

See attached pdf and remember to login to see attachments. And of course I am always happy to answer any questions.
John J. Donovan, Pres. 
(541) 343-3400

"Not Absolutely Certain, Yet Reliable"

John Donovan

  • Administrator
  • Emeritus
  • *****
  • Posts: 3223
  • Other duties as assigned...
    • Probe Software
Re: Best Practices for EPMA
« Reply #1 on: June 02, 2023, 05:03:02 PM »
The new dead time calibration and correction expression results shown in the talk can be found in the paper linked here:

https://epmalab.uoregon.edu/pdfs/Donovan-etal-DeadTime-2023.pdf
John J. Donovan, Pres. 
(541) 343-3400

"Not Absolutely Certain, Yet Reliable"

John Donovan

  • Administrator
  • Emeritus
  • *****
  • Posts: 3223
  • Other duties as assigned...
    • Probe Software
Re: Best Practices for EPMA
« Reply #2 on: June 07, 2023, 12:59:42 PM »
In case anyone is interested in seeing the actual talk I gave on "Best Practices in EPMA Microanalysis" with the audio and PowerPoint animations, here is a link to an .mp4 file of the 48 min presentation for the Chinese Academy of Geological Sciences:

https://probesoftware.com/download/Presentation%20of%20John%20Donovan_China_EPMA_2023.mp4

Thanks to Zoe Zhang for the file.
« Last Edit: June 07, 2023, 01:08:06 PM by John Donovan »
John J. Donovan, Pres. 
(541) 343-3400

"Not Absolutely Certain, Yet Reliable"

John Donovan

  • Administrator
  • Emeritus
  • *****
  • Posts: 3223
  • Other duties as assigned...
    • Probe Software
Re: Best Practices for EPMA
« Reply #3 on: June 15, 2023, 07:05:20 AM »
Here is the talk I gave yesterday at the IUMAS 8 meeting titled "The Holy Trinity of Microanalysis: Standards, K-Ratios and Physics". See attached pdf.

Please remember to login to see attachments.
« Last Edit: June 15, 2023, 08:19:14 AM by John Donovan »
John J. Donovan, Pres. 
(541) 343-3400

"Not Absolutely Certain, Yet Reliable"

Probeman

  • Emeritus
  • *****
  • Posts: 2771
  • Never sleeps...
    • John Donovan
Re: Best Practices for EPMA
« Reply #4 on: July 08, 2023, 10:49:28 AM »
This is a preliminary draft set of procedures for checking various instrument calibrations.  Some of these methods will be familiar to some of you, others may not.   We are seeking feedback and suggestions for improving these various calibrations.

These initial procedures were compiled by Mike Matthews, John Donovan, Aurelien Moy and Scott Boroughs and we welcome input from all others.

In this document we will discuss methods to check the calibration of our EPMA instruments. Some of these instrument calibration parameters are dependent on the manufacturer building the instrument properly, for example, that the column is perpendicular to the sample stage and centered with respect to the various WDS and EDS spectrometers. Others calibration parameters may change as the instrument and its detectors age over the lifetime of the instrument, for example, the WDS dead time calibrations.

1.   Sample, stage, beam and other geometric calibrations
a.   Check that the sample is level with respect to the light optics. Using a flat sample with a top referenced sample holder, move to three different stage positions approximately 1 to 2 cm apart, ideally forming a roughly equilateral triangle, in order to record the Z stage optical focus (and X/Y stage positions), and calculating the sample tilt from this data. If these three Z stage optical focus positions are always within the optical depth of field, the sample can be considered level with respect to the column.
b.   Check that the beam is central down the column by “rocking the beam focus”. For example, acquire a low mag image (SEM: at a short working distance) with all dynamic focus features turned off, and check that the area of defocus is concentric around the center of the image.
c.   Check that the stage is perpendicular to the electron beam and note: one cannot check that the stage is perpendicular to the electron beam using the light optics because if the stage is tilted, the optical focus will remain in focus over the stage top surface at all X/Y positions. Therefore, we must utilize an alternative method, for example, check that while electron imaging, an object (a small hole for example) remains centered as the Z stage axis is adjusted up and down.

2.   Spectrometer alignment and effective take-off angle for WDS and EDS spectrometers
a.   Check that all spectrometers yield similar k-ratios within measurement statistics. Using the method of “simultaneous k-ratios”, for example measuring Si Ka on SiO2 (primary standard) and Mg2SiO4 (secondary standard) on all TAP or PET Bragg crystals, one should obtain the same k-ratios within measurement statistics. For LiF or PET Bragg crystals we can utilize Ti Ka on TiO2 (primary standard) and SrTiO3 (secondary standard) for example.

3.   Dead time, picoammeter calibrations and Bragg diffraction symmetry checks. Details for the following three calibration procedures can be found in this recent publication: (https://doi.org/10.1093/micmic/ozad050).
a.   Check that each spectrometer yields the same k-ratios at different count rates (beam currents) by utilizing the method of “constant k-ratios” as described here by measuring k-ratio intensities for both the primary and secondary standard at different beam currents, e.g., both Ti metal and TiO2 at 10 nA, both at 20 nA, both at 30 nA, etc. up to say, 200 nA.
b.   Once the dead times are constant within measurement statistics over a range of beam currents (count rates), the picoammeter linearity can also be checked by utilizing only a single primary standard at one beam current, say 10 nA, and calculating k-ratios using a secondary standard measured over the same range of beam currents. Note that both 3a and 3b tests can be performed using the same dataset!
c.   Finally, using simultaneous k-ratio measurements (on multiple spectrometers tuned to the same emission line), check that all spectrometers yield similar k-ratios. There are several geometric/alignment issues that can mask each other: for example are the spectrometers properly centered around the column? Are the Bragg crystals diffracting symmetrically. Both of these can yield different effective takeoff angles for each spectrometer/crystal combination.  But if these effective takeoff angles are determined with reasonable accuracy, these various effective takeoff angles could be stored in one’s spectrometer calibration file for use in quantitative matrix corrections.

4.   Electron beam landing energy calibration
a.   As we all know, the Duane-Hunt limit can be utilized to check the accuracy of the electron beam accelerating voltage, however, it is best to test this limit “blind”, by having a colleague acquire the EDS spectra with sufficient statistics, but not at a nominal beam energy, for example: 15.2 keV or 14 .9 keV or 20.5 keV. Then, without knowing the purported beam energy in advance, one attempts to determine the Duane-Hunt limit for the spectrum in question. Always use a well grounded conductive (pure metal such as Cu or Au) sample and count a sufficiently long time to accurately determine the Duane-Hunt limit, and do not be misled by a few continuum coincidence events that will yield photons greater than the operating voltage. These continuum coincidence events can be reduced by counting at a relatively low beam current, for example 10 or 20 nA, and counting for a sufficiently long time, e.g., 1000 seconds.

5.   Miscellaneous test and calibration procedures. An example of these procedures can be found here: https://epmalab.uoregon.edu/reports/Additional%20Specifications%20New.pdf
a.   Check spectrometer reproducibility testing using the “half-intensity” method. First determine the spectrometer position which yields one-half the maximum peak intensity (on either side of the peak) for each pair of elements on each crystal at each end of the spectrometer range. Then move the spectrometer to each of these “half-intensity” positions and record the intensities. Repeat 100 times. The peak intensities shall vary by less than 0.6% (+/- 0.3%) with 99% confidence levels from the previous set and the one-half the maximum intensities shall vary by less than 1.2% (+/- 0.6%) at 99% confidence levels without a backlash or re-peak procedure. 
b.   Repeat the procedure with a crystal flip in between each “half-intensity” position. Verify that the intensities measured vary less than 2% (+/- 1%) with 99% confidence levels from the previous set without a backlash or re-peak procedure.
c.   Check beam stability over time. Beam current stability should be 0.1% or less per hour (+/- 0.05%) and 0.6% or less per 12 hours (+/- 0.3%) and 1.0% or less in 24 hours (+/- 0.5%) as measured at 15KeV and 10nA while repeatedly inserting the faraday cup approximately once per minute.
d.   Check for stray beam using an aperture mounted in a sample holder. Stray beam measured using a 100 micron W or Mo aperture target in a Ti target block to produce W or Mo Lα and Ti Kα k-ratios (both EDS and WDS) less than 0.0001 (0.01wt% or 100ppm) using a 100 nA beam and at operating voltages from 5 KeV to 30 KeV.
e.   The auto focusing reproducibility must be tested by performing the following test: 100 repeated auto-focuses that reproduce the stage Z position within 1 um each time on a static flat polished carbon coated Cu sample (dark blue color).
« Last Edit: July 09, 2023, 09:53:33 AM by Owen Neill »
The only stupid question is the one not asked!

Scott B.

  • Global Moderator
  • Graduate
  • *****
  • Posts: 9
Re: Best Practices for EPMA
« Reply #5 on: July 09, 2023, 07:20:11 PM »
Some suggestions for using PFE to automate some of the testing and calibration procedures mentioned above.

For checking autofocus reproducibility using Automate! in PFE...
  • Pick your point and focus it carefully.
  • Duplicate that point 100 times in Automate!
  • Either record the X/Y/Z position of the test point, or export the locations to .POS file
  • Check "Confirm unknown positions"
  • Check "Confirm all positions in sample"
  • Check "Use ROM autofocus" and "On Unks"
  • Set "Automate Confirm Delay (sec)" to 1 sec.
  • Run Selected Samples and walk away.
  • When the run is finished, export the POS file again. This will now have the new autofocus Z-axis values.
  • Open the pre-autofocus and post-autofocus files in excel, then compare the initial Z-axis value to the post confirmation values via a formula to evaluate what the max/min/stdev/etc/etc is between the before and after.

You can use a similar procedure for checking beam current stability...

  • Create a dummy method with at least one analyzed element. I suggest setting the background type to MAN, so you don't waste time and wear and tear on spectrometer movements.
  • Set your beam current at whatever value you want to test.
  • Set your count time on your dummy element to whatever you want for a measurement interval.
  • Pick a point on a robust material (pure metal, sample holder, etc) if you're using a high current or if this will be a long test. Can also defocus the stage substantially to spread out the beam.
  • Duplicate this point as many times as needed to make the test run as long as you want.
  • Critical step... go to Acquire! --> Acquisition Options and select "Do Not Set Conditions During Acquisition" in the Miscellaneous Options column. This will prevent PFE from adjusting the current back to what you set if it drifts!
  • Run your automation.
  • Go to Analyze! and export the time/date and current data. I right click on the sample and choose "Export Selected Samples to User Specified Format" and select Measured Beam Current in nA, Measure (2nd) Beam Current in nA, and Analysis Date and Time.
  • Graph up your data and run your statistics in Excel or your spreadsheet of choice...

For spectrometer reproducibility on half peaks (or peak center depending on your instrument's factory specs)...

  • Create a method with the elements of interest on all the channels you want to test. As mentioned above, it's best to test 2 elements at each end of the spectrometer travel range, and perhaps even one in the middle. For example, sphene (I know it's titanite, but I just like the way sphene rolls off the tongue...) is a good option for testing PET. Ti @ 88 mm, Ca @ 108 mm, and Si @ 228 mm. Albite or olivine for TAP (Si, Na/Mg, and O). I'm sure there's lots of other options, and don't forget to use those 2nd order lines if necessary. Be sure to choose a standard that can withstand repeated abuse without TDI or beam damage affects that will obscure the spectrometer drift.
  • Do a wavescan for all the elements on each channel in order to find the position for your "half intensity" on the side of the peak (Skip this step and just peak as normal if you're testing peak center reproducibility).
  • Assign the peak positions for each element as the half intensity position on one side of the peak (as above, skip and peak as normal for peak center reproducibility). You can do this directly in Plot! with your mouse, or type it in manually in "Elements/Cations" dialogue in Acquire! or Analyze!
  • If you want to test crystal flips, you'll need to do two sets of elements with one on a different crystal. You can make one of them an MAN element, with short count time, that's not too far from the last position in your real test elements, so all it's really doing is forcing a flip once per analysis and not wasting time driving the spectros and counting. Of course, it doesn't even need to be present in the sample either.
  • Pick a point on the standard of interest, and duplicate it as many times as you want the test to run.
  • Set your analytical conditions. I suggest using a rather gentle attack, with a big beam and moderate current, to avoid any affects of beam damage.
  • Run your list of samples in Automate!.
  • From Analyze!, export the data via right click "Export Selected Samples to User Specified..." and select "On-Peak Count Intensities".
  • Graph up your data and run your statistics in Excel or your spreadsheet of choice...

Lastly, if you want to check stage/beam reproducibility (i.e. the ability to hit a small target over and over)...

  • Find suitable objects/locations near the corners of your stage travel, that won't move, can allow precise location of crosshairs, and can be imaged clearly at high magnification (2000X plus). Spec of dust, crack, tiny mineral grain, etc.
  • In Automate!, ---> Digitize, add sample and location for each point. Be sure to make a separate sample for each location.
  • To duplicate your positions for longer term testing, Select your four samples and hit "Export Selected Samples (to *.POS)", then immediately Import them back in via "Import from ASCII (*.POS File)". You'll now have 8. If you plan to do lots of repeats, you can select the 8 and export/import. Then select 16 and export/import, and go on till your list is as long as you want.
  • Now go to "Acquisition Options" in Acquire! and select "Acquire Automated Images on Unknowns" and "Confirm Only".
  • In Acquire! hit the "Analytical Conditions" button. Set your "Magnification (Imaging)" value to whatever you'd like of sufficient magnification for good focus but also high spatial resolution. Something like 2000-5000X works well for me, but YMMV.
  • In Acquire!, hit the "Imaging" button, and set your imaging parameters to whatever count time, imaging channel (BSE, SEI, etc), and resolution you want, then hit "Close".
  • In Automate! select "Confirm Unknown Positions" up at the top right, then "Confirm All Positions in Sample" about halfway down on the right. And set your "Automate Confirm Delay" to 1 second.
  • Run your automation and walk away.
  • When finished, go to "Output" (in main PFE window top menus) and select "Save Images to BMP and GRD Files". PFE will spit out all the images to the folder where the MDB file is saved.
  • Open up all the images in a viewer or some kind, then you can scroll/click through them like a flip book and see how the images wander around your selected target point.

This test will show an aggregate effect of both stage reproducibility and beam drift (charging, temp changes, electrical instability, etc).

If you're a real nerd, you can even make an animated gif to send to your service engineer to report a stig drift that was otherwise hard to prove...



Hope this helps reduce some of the tedium involved in properly evaluating machine performance.
« Last Edit: July 09, 2023, 09:33:28 PM by Scott B. »

Probeman

  • Emeritus
  • *****
  • Posts: 2771
  • Never sleeps...
    • John Donovan
Re: Best Practices for EPMA
« Reply #6 on: October 18, 2023, 06:01:05 PM »
After today's Probe Software application webinar hosted by Anette von der Handt I just wanted to emphasize a couple of diagrams that were not shown but might help to explain some of the methods presented today. First regarding major element analyses:



and here regarding trace element analyses:



Only recently have I learned to appreciate these "rules of thumb" in setting up my analytical approaches.
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 2771
  • Never sleeps...
    • John Donovan
Re: Best Practices for EPMA
« Reply #7 on: October 30, 2023, 10:26:48 AM »
I asked Google Bard the following question: How can we improve accuracy in electron probe micro analysis?  And it responded:

Quote
Electron probe microanalysis (EPMA) is a powerful technique for determining the elemental composition of materials at a microscopic scale. However, the accuracy of EPMA measurements can be affected by a number of factors, including:

    Sample preparation: The sample must be carefully prepared to ensure that it is smooth, flat, and free of contaminants.
    Beam current and energy: The beam current and energy must be chosen carefully to avoid damaging the sample or producing excessive X-ray absorption.
    X-ray detection: The X-ray detector must be properly calibrated and aligned to ensure accurate peak identification and quantification.
    Data analysis: The data analysis software must be used correctly to account for various correction factors, such as matrix effects and absorption.

Here are some specific tips for improving the accuracy of EPMA measurements:

    Use a higher accelerating voltage to increase the X-ray penetration depth and reduce absorption effects.
    Use a lower beam current to reduce sample damage and avoid excessive X-ray absorption.
    Use a longer counting time to improve the signal-to-noise ratio and reduce the uncertainty in the peak intensity measurements.
    Use a wavelength-dispersive spectrometer (WDS) instead of an energy-dispersive spectrometer (EDS) for improved peak resolution and reduced background noise.
    Use a well-characterized set of standards to calibrate the X-ray detector and quantify the elemental concentrations in the sample.
    Use a data analysis software that includes corrections for matrix effects, absorption, and other factors.

In addition to these general tips, there are also specific techniques that can be used to improve the accuracy of EPMA measurements for certain types of samples. For example, to improve the accuracy of trace element analysis in quartz, it is recommended to use a higher accelerating voltage (e.g., 20 kV) and a lower beam current (e.g., 50 nA).

By following these tips, it is possible to achieve high accuracy in EPMA measurements for a wide range of materials.

So not too bad really, but not up to date...  e.g., blank corrections for improving trace accuracy:

https://probesoftware.com/smf/index.php?topic=29.msg565#msg565

or curved background fitting for off-peaks:

https://probesoftware.com/smf/index.php?topic=1378.msg9999#msg9999

not to mention quantitative interference corrections, TDI corrections, etc., etc.
The only stupid question is the one not asked!

Mike Matthews

  • Global Moderator
  • Professor
  • *****
  • Posts: 141
Re: Best Practices for EPMA
« Reply #8 on: October 31, 2023, 10:34:04 AM »
I’d question 2 of those statements:
  • Increasing the kV increases the electron penetration depth, not the x-rays. It does increase the mean emitted x-ray path length so the absorption will be higher not lower. For very soft x-rays a high kV can reduce the measured intensity drastically.
  • The beam current has no effect on the degree of absorption.

Probeman

  • Emeritus
  • *****
  • Posts: 2771
  • Never sleeps...
    • John Donovan
Re: Best Practices for EPMA
« Reply #9 on: October 31, 2023, 12:13:42 PM »
I’d question 2 of those statements:
  • Increasing the kV increases the electron penetration depth, not the x-rays. It does increase the mean emitted x-ray path length so the absorption will be higher not lower. For very soft x-rays a high kV can reduce the measured intensity drastically.
  • The beam current has no effect on the degree of absorption.

Yeah on the 2nd point I completely concur with you that beam current has no effect on absorption.   Though of course increasing beam energy does effect absorption. And that comes into the 1st point, because with regards to the first point, increasing the keV does (thought not in some cases as you stated, for low energy emission lines) increase the x-ray production for two reasons.

First because if one is at a low overvoltage e.g., Fe Ka at 10 keV, then increasing to 15 or even 20 keV will increase x-ray production due to more efficient ionization of the x-ray edge.

Second, increasing the keV increases the interaction volume size, and can increase increase the number of atoms interacted with and therefore (sometimes) increase intensity.  This is sometimes utilized for increasing trace element sensitivity for high energy emission lines.
« Last Edit: October 31, 2023, 04:34:37 PM by Probeman »
The only stupid question is the one not asked!