Joshua PollackOver-Enrichment at Natanz

Deep in the last IAEA safeguards report on Iran (GOV/2010/46, para. 7) is a mention of a curious finding at the Fuel Enrichment Plant (FEP), the larger of Iran’s two centrifuge enrichment facilities at Natanz. FEP isn’t supposed to enrich uranium past 5% U-235 content, but inspectors’ routine environmental samples have for the first time turned up “a small number of particles” that ran as high as 7.1%. U-235.

Questioned about this anomaly, the Iranians provided a “possible explanation” that was “not inconsistent with the Agency’s findings.” What was this explanation? The report doesn’t say, but we can make an educated guess: reflux.

(Warning: more nerdy technical than usual.)

The first clue is buried in a footnote in the report: “In this regard, it should be noted that Iran’s explanation refers to a known technical phenomenon associated with the start-up of centrifuge cascades.” After some scratching around, what turns up is a paper by Houston Wood and Stephanos Tongelidis, “Gas Centrifuge Cascade Study for Maximum Assays During Start-Up.” (Not online, sorry.) As the authors explain it, when UF6 gas is fed into a cascade during start-up, it’s done slowly at first, gradually ramping up to the “operational” rate:

During start-up of the separation process using gas centrifuges, the concentration of 235U may rise above the desired production level before coming back down at steady state conditions. As uranium feed initially enters the facility, the flow rate will increase until it reaches the design flow rate. This transient condition will produce an enrichment “spike” in the gas centrifuges.

The reason for the spike is the limited amount of UF6 gas in the cascade and the duration of its stay in the cascade. The less gas in the centrifuges and the longer it remains there, the more separation the machines will perform on it. FAS’s very helpful primer on separation describes this issue in terms of “reflux”:

[D]uring a fixed time, a machine can enrich a large amount of material to a small degree or a small amount of material to a large degree. The greatest separation occurs at total reflux — i.e. letting the centrifuge run with a fixed charge of UF6 but without adding or removing any material.  Clearly, this does not produce any separated product.

That’s the extreme case. Unless the flow of feed material into the cascade is actually stopped at some point, of course, the condition of reflux is not total. But the basic point is clear enough: the less feed material introduced, the higher the levels of enrichment, other things being equal.

(Caveat lector — some authors seem to use “reflux” to mean recycling the product through the cascade. That’s something different.)

Is the Explanation Plausible?

It seems so. Long story short, Wood and Tongelidis develop a model of a cascade meant to produce 3-6% U-235, develop a set of scenarios, and find that “expected peak enrichment levels due to the time transient [are] between 8-20% 235U.” That makes small traces of 7.1% U-235 seem comprehensible, certainly.

In the last few reports, we can see the Iranians stopping and starting feed on some seemingly troublesome cascades. The transient condition presumably takes place on each occasion that feed is (re)started. On at least one of these occasions, then, UF6 traces from the time of the reflux “spike” seem to have gotten out of the cascade and onto surfaces in the hall.

When that might have been is not clear, but it’s certainly notable that sometime between August and November 2009, the lowest levels of U-235 enrichment detected in environmental samples at FEP fell sharply, from 0.37% to 0.19% U-235. This seems to be the flip side of reflux: as the product stream gets more enriched, the tails stream gets more depleted. Of course, this could have been evidence of an earlier startup episode than the one that seems to have left the 7.1% U-235 traces.

It’s also possible that these weren’t the traces of startups, but of slowdowns during operations.

We’re still left to ask, why would traces of UF6 have wound up on surfaces in the centrifuge hall? A variety of actions could have done it: opening the cascade, dumping its contents, withdrawing samples from it, or switching out a product container. Why any of these things would be done during the “time transient” after startup, or during a slowdown of the feed, is unclear. Perhaps it testifies to the Iranians’ ongoing struggles to master a complex and temperamental technology.

There are more sinister explanations as well. It’s at least possible that the operators have been deliberately experimenting with higher levels of enrichment here and there, perhaps to create “noise” or find out what can or cannot be detected by inspectors. This seems unlikely, though: the Iranians are already enriching to nearly 20% at the adjacent PFEP facility, so what would be the point of mucking around beyond 5% at FEP?

It pays to be suspicious, but never forget Clark’s Law. For now, the bottom line is that it is indeed possible to identify a “known technical phenomenon associated with the start-up of centrifuge cascades” that could have created the 7.1% U-235 traces.

Update. The characterization of reflux has been corrected slightly, compared to the original post. Thanks to Webmaster Greg B. for jumping in to make the fix.


  1. Tom Bielefeld (History)


    Even though it’s not online, could you give us the complete reference of the Wood/Tongelidis paper? Would be much appreciated!


    • joshua (History)

      See the Proceedings of the 47th Annual Meeting of the Institute for Nuclear Materials Management (2006).

  2. Ano N. Ymous (History)

    Stuxnet? Could it have caused a spike if it managed to overdrive some centrifuges but they got shut down before (catastrophic) physical damage occurred? Particles could have escaped when those centrifuges were opened for inspections and repairs.

    • joshua (History)

      I suppose. It’s all pretty speculative. Why a computer attack would affect one unit (A26) so much more than another (A24) is very unclear to me. I’m not convinced that it did in fact attack FEP, although it’s a plausible choice of target.

  3. anon (History)

    Interesting but is there a point? There is no limit to enrichment level specified in the NPT or any other agreement Iran has signed.

    It could legally enrich to 80 or 99 percent. Under safeguards.

    • joshua (History)

      “Under safeguards” is the key phrase.

      Safeguards at a given facility are tailored to its design, which includes the level of enrichment of the product. The declared level at FEP is 5%.

  4. A Complete Stranger (History)

    It is worth remembering that the TRUE units for enrichment power of a centrifuge (or cascade etc.) is SWU-KG/yr. (I’ve tried to emphasize the kilogram part of that by making it capitalize.) We are always discussing how many SWUs each centrifuge represents but we forget the trade off in product quantity and enrichment when we leave off the kilograms. Thus if a 0.8 SWU-kg/yr machine produces 1 kg in a year, it has expended 0.8 separative work units on it while if the same machine only produces 0.5 kg in a year, it has expended 1.6 separative work units on that product. Smaller feed rates imply greater enrichment right from the units.

    • joshua (History)

      Well said.

  5. Mark Lincoln (History)

    Are there any Siemens S-7 PCLs in use at Natanz?

  6. Gridlock (History)

    I thought your ad server was being cute until I realised it’s your choice of illustration…

    Anyway, I look forward to breathless and uninformed news stories touting this finding until it becomes ‘conventional wisdom’ that the IAEA found HEU at Natanz.

    • joshua (History)

      20% U-235 or higher is HEU. Not 7.1%.

      (I recognize that you grasp this, Gridlock, but want to make sure your comment is not misunderstood by casual readers.)

  7. bradley laing (History)

    AP TEHRAN, Iran – A months-long delay in starting up Iran’s first nuclear power plant is the result of a small leak, not a computer worm that was found on the laptops of several plant employees, the country’s nuclear chief said Monday

Pin It on Pinterest