Unstable cps on standard samples

[Rom]

Hi everyone,
could you help me to understand possible causes of unstable cps number on standard samples.
I noted that number of cps on my standards can change from set to set during one run.
At the picture we can see that all standards were measured 2 times (3 points each) during the session.
All conditions were exactly the same (15kV, 20na, 10mkm beam etc).
But Al on the first set gives around 726cps/1nA, on the second set - around 695cps/1nA which lead us to uncertainty more than 4%.
The same issue we can see for Mg etc.
In my opinion, this is not the beam focus or surface issue of standards   or carbon coating issue because other channels at the same time give quite stable results.
May be this is DT issue because the total counts (for Al on Al2O3) is 730*20=14600cps; for Mg on MgO is 460*20=9200cps? But for Al on albite is just 130*20=2600cps and we can see less dramatic but still large uncertainty.

Thank you!
   

[JonF]

All other things being equal, what you're probably seeing is variation caused by barometric pressure and relative humidity in your (probably air conditioned) laboratory. I'd bet that the air pressure in your area has increased between the first and second standard runs, causing the intensities to drop (or rather, the response of the GFPC detectors). Sealed Xe detectors (for the JEOL folk) aren't impacted by this*.

We can get pretty big swings of 50mbar or more over a few days (see below) and this can cause havoc on our standard intensities, meaning we have to re-standardise frequently.



Fortunately, PfE has a standard drift correction to deal with this but be cautious, I believe it will apply a linear correction between one set of standards to the next when the weather is anything but linear! It can be caught out by sudden drops in pressure (e.g. thunder storms)**.

For example, imagine we ran the standards on the 19th Feb and the 28th Feb, PfE will correct for the drift between these two dates (not that it would be much), but would be oblivious to the large drop in air pressure in between, causing the standards intensities (and the unknowns!) to increase. The answer is obviously to standardise much more frequently, but at some point you actually have to measure the unknown samples and stop using your probe as a rather expensive weather station!



 
*At least not that I've ever spotted. Does anyone have a TAP crystal using a sealed Xe detector? I've been curious whether the drop in sensitivity is a fair trade for increased stability over time for long maps.

**In lieu of PfE releasing an omniscient edition, I've thought about building a small weather station out of an arduino and some cheap air temperature, humidity and pressure sensors and logging all this, with the goal of ultimately being able to calibrate and correct for it in PfE. 

[Probeman]

Jon: assuming sp2 (TAP) is a flow detector I think you are on track with your barometric explanations.  Assuming there were no crystal flips in between the standardizations!

I mention crystal flips because it just so happened that we were recently contacted by Axel Wittmann who was seeing changes in intensities on his JEOL instrument and Anette chimed in.  He and Anette have said it's OK to share their recent experiences on the forum...

I am longing for those days of not having issues with measuring Si with our TAPJ crystal, when routine measurements were routine.

As soon as the instrument schedule permits, I will generate a higher resolution wavescan for Si and share it with you.

Al reproduces fairly well on that TAPJ crystal, I have less data for Mg because I am typically using the TAPH crystal for Mg. I can run a test, though to verify. At times, I have the impression that there is a drift issue. However, the somewhat shocking result is that if I try to measure a standard as unknown right after the acquisition of intensity data on the same standard (say, Wollastonite, or San Carlos Olivine), more often than not the Si values are far worse than 1 rel% off (typically 2 to 4 rel.%).

Our microprobe has two H-type spectrometer slots (and they are not cheap-enough to change to L; I noticed the general dislike of H-type spectrometer settings but for what it is worth, most of the time, they yield better results than our J-type spectrometers) and they are not the problem.

...things that may or may not be related to your issue but have contributed to problems for me with analyzing Si in the past. Not so much really for Si as an element but due to its position on the TAP and as it usually occurs in high abundances, it acts like a canary in the coal mine for spectrometer hardware issues.

1) First off, your problem of not reproducing standards as unknowns pings a faint memory. At my previous lab, I once had problems with the crystal flip mechanism on a spectrometer. It would flip but not fully and there is no warning or "flip completed" message on the JEOL side. I think it is just a physical stop. Software will think it is all a-ok. All kinds of drift issues because the crystal surface was ever so slightly inclined. So maybe, experiment with crystal flips or when you have the spectrometer open watch it do crystal flips.

2) If the instrument has been inactive over a long weekend or a break, I usually exercise the spectrometers first (have them go between 80 mm and 200 mm a few times). That seems to greatly improve stability for me because the peaking is not the first movement that happens when it is "cold"

I sometimes recognize that that crystal would indeed not flip correctly if I peak that spectrometer for a X-ray intensity WDS map on the JEOL-side. Ironically, my way to troubleshoot this issue is to activate Probe for EPMA and briefly peak the element there. This typically remediates this issue for the mapping. Now, if the crystal flip does occur as part of the spectrometer movement during spot analyses (I think such notifications appear if one follows the procedure closely), this could indeed be the reason why Si is so funky. Other elements would be affected as well, though, I think. I will check if Al is similarly variable. It is a little less obvious because its concentrations are mostly lower than Si. I will try the anorthite again in my next test. And I will definitely present this as a possibility to the JEOL engineer. This may also explain why the issue does not occur during the performance tests of the engineer; the JEOL software may not flip the crystal during the ten analyses on the same spot.

Exercising the spectrometers ! I recently took to do more extensive PHA scans because I noticed that there is quite some variation for the bias settings that noticeably affect the optimal window setting. So I think that the spectrometers get reasonably well exercised before the peaking is completed. I’ll try the spectrometer exercising next time, maybe it helps.

Ok, if you even noticed that it is not flipping correctly than I think the case is pretty much closed. I learned the hard way that work-arounds around hardware problems usually just prolong the pain. It is rare that the JEOL

In my case, they ended up replacing the crystal-flipping mechanism but that was not such a big deal except that they need to realign the spectrometer.

But I think a faulty crystal-flipper explains also that you see the problems intermittently as sometimes it probably flips correctly or the engineer checks everything and it works but only until you have to flip the crystal again.

It also would explain that it sometimes gets worse when you move to other samples. I assume that when the crystal does not come to the full rotation, subtle differences in sample tilt might affect the focusing power of the crystal. Depending on the crystal type and the spectrometer position you might get subtle variation in how much X-rays get focused into the detector. At least that is how it works in my head.

Fortunately, PfE has a standard drift correction to deal with this but be cautious, I believe it will apply a linear correction between one set of standards to the next when the weather is anything but linear! It can be caught out by sudden drops in pressure (e.g. thunder storms)**.

For example, imagine we ran the standards on the 19th Feb and the 28th Feb, PfE will correct for the drift between these two dates (not that it would be much), but would be oblivious to the large drop in air pressure in between, causing the standards intensities (and the unknowns!) to increase. The answer is obviously to standardise much more frequently, but at some point you actually have to measure the unknown samples and stop using your probe as a rather expensive weather station!

Just to add some further explanation regarding the standard intensity drift correction in PFE it should be pointed out that this correction is only applied in the construction of the k-ratio and therefore only affects quantitative analyses, not the raw data for particular samples.

In other words, raw intensities (net counts) are not affected by the standard drift correction. Instead, in the analysis of an unknown (or standard) element the program calculates what the standard intensity *would be* based on the time of acquisition of that sample by interpolating between standardizations (assuming there are standards acquired before and after the sample in question).

Also in addition to the Drift program which as Jon mentioned compiles all raw intensity measurements of a particular element on a particular spectrometer/crystal and on a particular standard for all runs in a folder (and sub folders if selected), one can also use the List Standard Intensities button in the Analyze! window to list all standard intensities in a run.  Then it is easy to see if there has been drift (or glitching) in a single run).

Here is an extended discussion on the standard intensity drift correction in PFE:

https://probesoftware.com/smf/index.php?topic=168.0

[Rom]

Thank you for thoughts.
1. I think it is not pressure problem because I can see the same problem in my several sessions. Sometimes the distance between measuring sets is just 3-5 hours. Moreover, I unsuccessfully tried to find dependency cps changing and atmospheric pressure in my region which can be esyly found on weather sites.

2. In this run I didn't use crystal flip (may be crystal has some play in the fastening?) - the only detector voltage could be changed (20-30V difference on the total level around 1700V) from Mg to Al adjusting at the discussing run.

What kind of tests for issue searching could you recommend? 

[Anette von der Handt]

That is a pretty big drift in the data for barometric pressure changes. Of course, it depends on the instrument's location but usually, I feel the data drift is more subtle even on my previous probe in the Midwest where weather changes come in all caps and with exclamation marks.

Do you have an idea based on the analyses in between the standard analyses if the drift is gradual or more sudden? How much time was between the first and the second standardization? I would suggest setting up an overnight or even better an over-the-weekend run where you measure Si and some other element (so that the spectrometer has to move forth and back) on a Si-metal or some other homogeneous Si-bearing standard. That way you can see is there a gradual drift or something else.

If this is a JEOL probe, I think what you are dealing with is a problem with the spectrometer belt on your spectrometer 2. Those darn belts can not just snap but occasionally they can "slip", either gradually or in increments. I had that before with similar data and when we opened the spectrometer it had some strange black oily contamination on a part of the belt. Never knew where it came from.

While it is annoying to replace the belt and realign the spectrometer, it is a straightforward fix. Might be worth just opening up the spectrometer and visually inspecting it while the spectrometer is moving. Maybe you can identify the problem right then.

*Edited to add: I think Andrew Locock in Alberta has a TAP on a Xe counter. Maybe he can weigh in on the stability question.

[Probeman]

What kind of tests for issue searching could you recommend?

Did you answer the question whether this is a flow or sealed detector?

If a flow detector it could be a contaminated detector from a bad bottle of P-10.   If a sealed detector it could be that the detector is getting pumped out and therefore unstable and needs to be replaced.

Try these things:

1. First perform some bias scans in the PHA dialog in PFE on both emission lines and check that you are using a high enough bias voltage but no more than 10 or 20 volts over what is necessary.

2. Try not changing the bias voltage for your two elements but setting the higher bias value for both elements. Sometimes changing the bias voltage between elements can introduce some instability. 

If you are running in INTEGRAL mode, then PHA peak shift won't be an issue. And everyone should use INTEGRAL mode except for very rare situations as described here:

https://probesoftware.com/smf/index.php?topic=1466.msg11636#msg11636

3. Take a look at the BiasChangeDelay in the [software] section of your Probewin.ini file and see what it is set to.  Sometime one needs to add a bit longer delay to let things settle on some detectors.

4. Also check the PHAFirstTimeDelay in the [pha] section of the Probewin.ini file, which is a delay applied only the first time the PHA bias is set for a spectrometer.

[Rom]

Did you answer the question whether this is a flow or sealed detector?

The detector is flow. It uses P-0 gas bottle.
I will check the gas bubbles and quality of tubes from bottle to detector and let you know.

1. First perform some bias scans in the PHA dialog in PFE on both emission lines and check that you are using a high enough bias voltage but no more than 10 or 20 volts over what is necessary.

The optimal bias voltage (baseline is next to 4V) for kaAl is 1718V, for kaMg is 1737V.

2. Try not changing the bias voltage for your two elements but setting the higher bias value for both elements. Sometimes changing the bias voltage between elements can introduce some instability. 

What bias voltage do you recommend for both lines - something around average 1728V?
Yes, I am running in INTEGRAL mode.

3. Take a look at the BiasChangeDelay in the [software] section of your Probewin.ini file and see what it is set to.  Sometime one needs to add a bit longer delay to let things settle on some detectors.

The BiasChangeDelay in the [software] section of our Probewin.ini file is 1.0.
Should I change it?

4. Also check the PHAFirstTimeDelay in the [pha] section of the Probewin.ini file, which is a delay applied only the first time the PHA bias is set for a spectrometer.

The PHAFirstTimeDelay in the [pha] section of the Probewin.ini file is 1.0.
Should I change it?

[Rom]

Hi Annete, thank you for your efforts to help!

Do you have an idea based on the analyses in between the standard analyses if the drift is gradual or more sudden? How much time was between the first and the second standardization? I would suggest setting up an overnight or even better an over-the-weekend run where you measure Si and some other element (so that the spectrometer has to move forth and back) on a Si-metal or some other homogeneous Si-bearing standard. That way you can see is there a gradual drift or something else.

It happens suddenly.
it can be after 3 or after 8 hours (that time - 3 hours).
I will try these kinds of tests, thank you.

If this is a JEOL probe, I think what you are dealing with is a problem with the spectrometer belt on your spectrometer 2. Those darn belts can not just snap but occasionally they can "slip", either gradually or in increments. I had that before with similar data and when we opened the spectrometer it had some strange black oily contamination on a part of the belt. Never knew where it came from.

It is a JEOL probe, I asked Lab manager to check the belt and spectrometer.

[Probeman]

Did you answer the question whether this is a flow or sealed detector?

The detector is flow. It uses P-0 gas bottle.
I will check the gas bubbles and quality of tubes from bottle to detector and let you know.

1. First perform some bias scans in the PHA dialog in PFE on both emission lines and check that you are using a high enough bias voltage but no more than 10 or 20 volts over what is necessary.

The optimal bias voltage (baseline is next to 4V) for kaAl is 1718V, for kaMg is 1737V.

2. Try not changing the bias voltage for your two elements but setting the higher bias value for both elements. Sometimes changing the bias voltage between elements can introduce some instability. 

What bias voltage do you recommend for both lines - something around average 1728V?
Yes, I am running in INTEGRAL mode.

3. Take a look at the BiasChangeDelay in the [software] section of your Probewin.ini file and see what it is set to.  Sometime one needs to add a bit longer delay to let things settle on some detectors.

The BiasChangeDelay in the [software] section of our Probewin.ini file is 1.0.
Should I change it?

4. Also check the PHAFirstTimeDelay in the [pha] section of the Probewin.ini file, which is a delay applied only the first time the PHA bias is set for a spectrometer.

The PHAFirstTimeDelay in the [pha] section of the Probewin.ini file is 1.0.
Should I change it?

Can we assume this is your JEOL 8200 instrument?  If so, that is an old instrument. Have the detectors been replaced or cleaned since the instrument was installed?

When trying to discover the source of a problem, one must attempt to "divide the problem".  For example, does this instability only occur on one spectrometer or on multiple spectrometers?  If only on one spectrometer, does it occur for both (or all 4) Bragg crystals?

Tests like this will help you to narrow down the problem.  As for the bias delays you will have to determine those values empirically. 

If you do have an old instrument you could very well need to add delays or at least utilize the same bias for all elements on a spectrometer.

Or it could be a mechanical problem as Anette suggests.  Again: divide the problem!

[Rom]

Yes, it is JEOL 8200.
I'll try to divide the problem, thank you.

[JonF]

That is a pretty big drift in the data for barometric pressure changes. Of course, it depends on the instrument's location but usually, I feel the data drift is more subtle even on my previous probe in the Midwest where weather changes come in all caps and with exclamation marks.

Slight tangent on this thread, but drift can be anything but subtle here!

This is from our Cameca last month, low pressure P10 detector looking at Mg in periclase (repeated standardisations). Our JEOL GFPC show similar trends (the Xe detectors don't). This data isn't particularly unusual for here, and the lab air temperature is controlled (not brilliantly in the Cameca lab, but there's no obvious correlation).

[Probeman]

That is a pretty big drift in the data for barometric pressure changes. Of course, it depends on the instrument's location but usually, I feel the data drift is more subtle even on my previous probe in the Midwest where weather changes come in all caps and with exclamation marks.

Slight tangent on this thread, but drift can be anything but subtle here!

This is from our Cameca last month, low pressure P10 detector looking at Mg in periclase (repeated standardisations). Our JEOL GFPC show similar trends (the Xe detectors don't). This data isn't particularly unusual for here, and the lab air temperature is controlled (not brilliantly in the Cameca lab, but there's no obvious correlation).

Just curious if anyone has tried the absolute pressure gauges described in this post:

https://probesoftware.com/smf/index.php?topic=1109.msg10233#msg10233

For regulating P-10 back pressure flow...

[Anette von der Handt]

Wow, that is quite the drift and a great data set. Thanks for sharing, this is very educational.

Slight tangent on this thread, but drift can be anything but subtle here!

This is from our Cameca last month, low pressure P10 detector looking at Mg in periclase (repeated standardisations). Our JEOL GFPC show similar trends (the Xe detectors don't). This data isn't particularly unusual for here, and the lab air temperature is controlled (not brilliantly in the Cameca lab, but there's no obvious correlation).

[Rom]

Does the direction right: less pressure - higher intensity? Could you explain why.
Thank you!

[JonF]

Does the direction right: less pressure - higher intensity? Could you explain why.
Thank you!

Yes, that's right, as the barometric pressure increases, the count rate from the GFPC (not sealed Xe!) detectors decreases.

I suppose this data is the counterpart to the excellent dataset acquired by Brian Joy (here: https://probesoftware.com/smf/index.php?topic=1109.msg10889#msg10889), only instead of varying the bias voltage to maintain a constant count rate and PHA peak position, I held the bias constant and monitored the resulting change in count intensity. 

As per the suggestion in that thread, I suspect atmospheric gases (H2O? O2? N2?) are backflowing up the open vent side of the GFPC, hence why you see a higher bias voltage is required on the vent GFPC as opposed to the gas inlet GFPC (JEOL EPMAs tend to be - but don't have to be - daisy chained with the P10 flowing through one spectrometer to the next, as opposed to Cameca EPMAs that have them all fed independently from a single point).
There's a couple of possible mechanisms that I can think of as to why this would reduce count rate: whether via a simple dilution of P10; whether the atmospheric gas is absorbing the X-rays and dissociating, preventing a cascade forming; whether the gas molecules are capturing the e- emitted from the Ar... lots of possible mechanisms! 

One other point to note is that I haven't seen this effect on the sealed Xe detectors, leading me to think that it is a physical factor in the detectors rather than an electronic issue (assuming the counting chain for the JEOL GFPC and Xe detectors are the same!).