Tue Oct 09, 2018 5:48 pm by mleipold
Hi Takuya and Eric,
Back when this happened to us, I did calculations on whether Praxair could even give a spec on Xe levels. From what I was told, they mainly use GC, with a LOD around 1ppm.
Calculations I sent to Praxair are below:
-------------------
"I *can* give [Praxair] an estimate on the regular Xe signal. During the Tuning process, we do measure the Xe131 signal. Then, by comparing that signal to the closest Tuning Solution isotopes, we can get at least a rough number.
----If Cs133 and La139 are 0.5ppb, then on a recent good Xe tuning:
Xe131 (21.18% of total Xe) is a Dual of 60,822 --> 287,167 @ 100% Xe
La139 (99.91% of total La) is a Dual of 898,504 --> 899,313 @ 100% La
Cs133 (100% total Cs) is a Dual of 1,023,182 --> 1,023,182 @ 100% Cs
Therefore, if total Cs and total La are each at 0.5ppb, then total Xe would be at approximately 0.15ppb in a regular tank.
From above: the Xe131 has a software QC limit of ~400,000 Counts, which is about 6x what I measured on the "good" tank. If so, then total Xe to allow pass of Tuning would be roughly (0.15ppb x 6 =) ~0.9ppb. Since labs are failing tuning and based on the Xe comparison I attached, I'd say we're well above that on the "bad" tank.
One thing I didn't think about until last week was whether Xe was being fully ionized in the Ar plasma. From looking up the first ionization energy of Xe compared to Ar, it seemed likely that (unlike Cs or the other Ln metals), Xe would *not* be fully ionized.
So, I asked Fluidigm, and they confirmed that Xe is only ~10% ionized in an Ar plasma. Therefore, we should multiply my previous numbers of a factor of 10. As such, ~1.5ppb for "good", >10ppb for these "bad" tanks would be the "new" values.
From what [Praxair] told me about the detection limit of [their] GC (usually ~1ppm), we're still ~100x below that for our max allowable.
---------------------
Mike
So, there's still a question whether they can even really *measure* down to the levels that we need it........and therefore, it's not clear whether they can even write a spec for it....'