Since my post last week delving into the question of how big the HEU particle found on something North Korean would have to have been for the US to date it to 3.5 years old, I’ve had a number of interesting discussions on the subject—both in the comments section to this blog and elsewhere. Since the subject is topical I wanted to summarize some of the key open questions.

How accurate is mass spectrometry?
Generally, people seemed to buy my mathematical analysis—the issue was whether my estimate for the accuracy of mass spectrometry (the one unknown parameter in my model) was too pessimistic. This is important because, for a given error bar on the particle’s age, the more accurate the measuring device the smaller the particle need be.

Importantly, I am assured by a number of knowledgeable people that the particular mass spectrometry technique I discussed (TIMS) can be made more accurate than in LaMont and Hall’s experiments on which my analysis was based. So, my estimate of the size is almost certainly something of an overestimate—how much of one is unclear. If anyone can point me in the direction of an article about TIMS with enough quantitative data to be useful, I’d be grateful.

Of course, we’re flying blind to some extent since we don’t know what method of mass spectrometry was used. Russell Leslie points out that accelerator mass spectrometry (AMS) is capable of much more accurate measurements than TIMS. Indeed, H. Lime even provides a link to a picture of LLNL’s AMS facility.

They might well be right. The one note of caution I’d sound is that this is exceptionally tricky experimental work and developing an experimental protocol can be a tough, time-consuming job. Having access to AMS technology doesn’t automatically mean it can be applied to this problem without a lot of prior effort—and I just don’t know whether the required effort has yet been applied.

Indeed, Vitaly Fedchenko pointed me in the direction of the IAEA’s Research and Development Programme for Nuclear Verification 2008–2009 which states that

Age-dating of high-enriched uranium (HEU) particles. This is needed to help deduce the origin of HEU particles found in environmental samples, but at the present time, no method is sensitive enough to measure the Pa-231 and Th-230 daughter isotopes. Candidate methods are accelerator mass spectrometry and resonance ionization mass spectrometry.

Although, as Vitaly also noted, just because the IAEA doesn’t have this capability does not mean that USIC lacks it as well.

What is really required here is a detailed literature survey of mass spectrometry techniques and their accuracies—but that is more like a research project than a blog post. If any reader wants to undertake it, however, you can be assured of somewhere to publish it…

How big are uranium particles?
A second related question is about the typical size distribution of uranium particles on swipe samples. After all, if massive particles are relatively common then the discussion above becomes more-or-less moot. Indeed, Mark Gubrub suggests that the size distribution was typically quite broad. Vitaly also sent me an interesting analysis of the physics of how large particles might form.

…there was a cool paper I saw some time ago, Informativeness of Morphology Analysis of Uranium Microparticles from Industrial Dust at Nuclear Plants by Vladimir Stebelkov (sorry that the English there is a little too Russian, but that adds some flavor for connoisseurs). On the second page Stebelkov mentions in passing that “Uranium molecules and uranium dust particles could form uranium-contained layers during operation of the plant”, and then illustrates it with pictures. Stebelkov was collecting his material from “working areas and ventilation shafts”.

The other article I attach discusses different thing, but also mentions that “The particles formed from uranium hexafluoride are highly hygroscopic and little is known about their long-term stability. Before being collected on swipes they might have been exposed to high temperatures, a high humidity or sunlight. All these environmental factors could have altered their morphology and composition…” and then “In some cases a uniform film of uranium was detected instead of single particles. The fact that the UO2F2 particles are highly hygroscopic could explain this phenomenon”.

So I would speculate, without having done a thorough analysis of this, that (a) it may not be so unusual for spherical particles to melt into uniform films on some surface; (b) that process may be facilitated if the surface well contaminated with particles leaves the controlled environment of the plant and gets exposed to moisture, heat, light, etc.

I guess, it would then not take much to shatter such “uranium film” into relatively large splinters, big enough to help the US labs out with age determination.

What was the enrichment level of the uranium particle?
I didn’t discuss this issue much in my original post—but it’s worth mulling on. Dating relies on the presence of U-234, which makes up only 50 ppm of natural uranium. However, being lighter than U-235 it accumulates in the product streams during enrichment. Thus, by the point uranium is 93% enriched in U-235 it is also about 1% enriched in U-234. This was what I used for my calculations.

But, all we actually know is that the particle was HEU; this covers everything from 20% U-235 upwards. The less enriched it was, the less U-234 it contained and the larger it would have to have been for dating purposes. Uranium that is 50% enriched in U-235 contains only 0.43% U-234. A particle of this material would have to be about twice as massive as a particle enriched to 93% to be dated with an equivalent accuracy.

My assumption—that the particle was 93% enriched—is certainly favourable from a USIC perspective (in contrast to my assumption about mass spectrometry accuracy). The dating would be harder to the extent that the particle was less highly enriched.

Moreover, in an operating enrichment plant, particles of all enrichment levels up to the maximum level are present. Therefore, even if the hypothetical North Korean plant was producing weapons-grade HEU (or something close to it) a good slice of luck would have been involved if the one particle that USIC found was sufficiently highly enriched to contain enough U-234 for dating.

Was the particle an outlier?
Yes.

I got this one wrong before. I wrote then that:

Moreover, even if USIC did find one huge particle, basing key policy choices on just one particle is foolhardy. There is an interesting reference in LaMont and Hall to “discarding outliers”. In other words, they occasionally obtained a result that was so wrong (many standard deviations away from the mean) that it was attributed to human error and discarded. If you have just one particle for analysis it is impossible to know whether your one result is one of these outliers.

However, as Josh points out in the comments, the US did have other particles and just one, so far as we know, was dated to 3.5 years. So, it was an outlier. This might be because it was genuinely much younger than the rest, or because of a SNAFU—it is simply impossible to tell. But it does give reason for caution.

Conclusions
There’s something here for everyone. The discussions on mass spectrometry accuracy and the possibility of finding large particles tilt the analysis more in USIC’s direction. The discussion of enrichment levels and outliers do the opposite.

As much as I risk piles by sitting very firmly on the fence let me summarise it this way: USIC’s claim is not impossible, but there are definitely reasons to be cautious. Most crucially, I would urge against making key policy decisions on the basis of one particle.