Help with Th-232 Spectrum

[Chase]

I am attempting to use this spectrum to create a detailed energy calibration but I'm running into a number of issues. I was hoping someone with a bit more experience could help me interpret this.

This is a Th-232 spectrum I just collected from a vintage thoriated camera lens. I feel pretty confident in my energy calibration for all peaks up to the Ac-228 916keV peak but that last Tl-208 2615 peak is giving me a headache. It certainly looks like I've got it correctly pinned down here, but if my energy calibration is correct for the high energy range, then my pair production single escape peak is in the wrong place. It should be at 2103 but instead it looks like it is at 2204. Similarly, the compton edge is at a higher energy than it should be by about the same amount. If I assume a linear energy response and drop my last calibration point, then the larger peak that I thought was Tl-208 actually lands on 2436keV.

Some potentially relevant details: I'm using a vintage NaI detector taken from an old laboratory monitor, unsure of exact crystal dimensions but probably a 2x2. Using the GS-PRO-V5. Decently well shielded setup with about 2.5 cm lead directly around the probe with an open end placed directly up against the lens. Then I have the entire apparatus, lens and all, surrounded by about 4 layers of lead apron previously used in a medical fluoroscopy lab. I'm in the front range in Colorado so there is a substantially higher than average background here primarily from natural terrestrial sources and cosmic, but I'm not sure how much that would impact the measurements.

I guess my main question is this: How linear do you expect your energy calibration curve to be and are there any potential contaminate peaks you know of that might be causing this confusion? Also, on a side note, how quickly does the energy response drift for these sorts of detectors?

Thorium232 Spectrum.png (25.37 KiB) Viewed 23259 times

[Rob Tayloe]

In the thread linked below (on SrI2(Eu) detectors, on the first page of the topic, I present energy calibration data for NaI, CZT, LaBr, and SrI detectors. For each detector I was using different hardware and software. The calibration was done using a variety of sources, mostly obtained from Spectrum Techniques in Oak Ridge, TN. The energy calibration results were presented using a spreadsheet (tabular and graphical). A least squares fit to the data (as well as examining the graph) can reveal linearity (or not!)

viewtopic.php?f=14&t=885

[ColoRad-o]

OldMantle0001.pdf

(803.85 KiB) Downloaded 874 times

Chase,
The degree of linearity can depend on the bias voltage (in my experience). My workflow generally is to calibrate with 3 sources (Co60, Cs137, Ba133), sometimes with a thorium lantern mantle for the 2614 peak. I have used PRA polynomial interpolation many times, but these days often fit to a cubic with Mathematica, then apply the cubic to the interpolated raw (AU) data and export the interpolated keV data for use in InterSpec.

[Chase]

The degree of linearity can depend on the bias voltage (in my experience). My workflow generally is to calibrate with 3 sources (Co60, Cs137, Ba133), sometimes with a thorium lantern mantle for the 2614 peak. I have used PRA polynomial interpolation many times, but these days often fit to a cubic with Mathematica, then apply the cubic to the interpolated raw (AU) data and export the interpolated keV data for use in InterSpec.

That is good info, thanks! Just out of curiosity, I'm sure some of this depends on the detector quality itself, but how often do you need to adjust your calibration curve? I'm finding that it seems to require a new calibration almost every time I plug in the spectrometer or else clear energy peaks such as Cs-137 end up about 20 keV or more off of what it should be.

I was thinking of making some sort of spreadsheet that I could use to build a detailed calibration curve and then take a quick spectrum with a Cs-137 check source to scale all of the values of the full calibration curve to the 662 peak prior to use each time but wasn't sure if you or anyone else had any suggestions on ways to approach this.

[Rob Tayloe]

Here is the compressed (using 7zip, a freely downloadable program) spreadsheet used to perform the energy calibration mentioned in my previous post in this thread. The energy calibration page is on the first tab. All one need do is substitute energy of referenced peaks and associated channel number associated with that peak (usually the largest value). The plot and fitted polynomial will be adjusted as the data is replaced.

SC SrI Energy Cal2105.7z

(1.02 MiB) Downloaded 845 times

The 7zip program can be obtained from -

https://7-zip.org/

[ColoRad-o]

Re: How often to re-calibrate. Drift of the spectrometer response with time appears to be ubiquitous, so I let it (scintillator+PMT) sit there running (ostensibly acquiring data on the average pulse shape, but really just warming up) for a couple hours. Sometimes I do a calibration run, then the main run, then another calibration run, to get a handle on drift. Ugh. I have casually looked for how the shift in peaks depends on, say, peak energies in AU, but have failed to discern a very useful pattern. Some have tried attaching Peltier coolers to PMTs, but that's more trouble than I want to go to.

The paper "An automated drift correction method for in situ NaI(Tl)-detectors used in extreme environments", by le Roux and Bezuidenhout [Applied Radiation and Isotopes 181 (2022) 110069] discusses ways to compensate drift by adding some fine gain control circuitry based on the ubiquitous 40K 1460 keV line. But they note, "Several elements within the NaI(Tl)-detector are dependent on temperature, such as the decay constant and light output of the scintillation material, multiplication factor and drift of the photomultipliers, and quantum efficiency (QE) of the photocathode (Moszy´ nski, 2006; Ianakiev, 2009). The optimal relative light yield of NaI(Tl) detectors typically occur at approximately 300 K with a decay time of 215 ns. This yield, however, decreases at both higher and lower temperatures (Knoll, 2010). Decreasing the temperature also leads to longer decay times peaking around 150 K (Sailer, 2012). Long-term drift can also occur in NaI(Tl)-detectors due to changes in the charging of the dielectric surfaces (TB-5 User Manual, 2021), as well as hysteresis of the gain (Tsankov and Mitev, 2006).

Not much joy from that paper to be had for amateur gamma spectroscopists.
DMW

[Chase]

Thank you everyone, all very useful information.
Rob, I'll spend some time with that spreadsheet. I'm not quite sure how it will apply to my calibration process just yet but it looks like it has some pretty in depth stuff.

ColoRad-o, I suspected that it was temperature dependent! Good paper for sure, I'll give it a more detailed read through when I have the chance. Sadly from my brief experiments it doesn't look like the correction will be a linear shift. When I applied a uniform shift to my calibration curve based on the Cs-137 peak, many of the peaks at different energies for Th-232 were off enough to really bug me. It might be worth investing in a check source with a couple peaks at different energies for a better scaling of the calibration. All that to say, apparently the corner where I keep my detector counts as an "extreme environment" haha! All joking aside, I do know that my room's ambient temperature is very far from consistent and sadly there isn't much I can do about that.

[ColoRad-o]

I didn't make clear that I calibrate by cramming 3 sources in ONE run--Ba133, Cs137, Co60. This saves time. Alas, the Co60 source is so new that its count rate caused too much pulse overlap. Thus I put all 3 into a small lead pig which I suspend near the detector. This keeps the count rate manageable. U238 has some low- and medium-energy peaks that can be helpful in a second calibration run. Th232 has a wealth of peaks too, as you well know. Sometimes I find sum peaks (e.g., from the two Co60 peaks) which provide a handy high-energy calibration peaks. Make sure not to use Compton edges, though since they are non-Gaussian peaks nor do I find very precise values. Am241 from a smoke detector gives a clean 59.5 keV line as well.--ColoRad-o