Tracking UF6 cylinders

I’ve written a little bit about efforts by the Agency to safeguard enrichment plants (‘A new safeguards approach for enrichment plants’, 22 November 2007). What's prompted the development of new approaches is the construction of commercial scale gas centrifuge enrichment plants outside Europe. The present safeguards methodology dates back almost 30 years.

On 5-7 November 1980 a meeting was held in Marlow, England. Represented around the table were the Agency, Euratom, Australia, Japan, the United States and URENCO (Germany, the Netherlands and the United Kingdom). They decided to establish the Hexapartite Safeguards Project (HSP) comprising separate working groups that were tasked to figure out in detail how safeguarding could be done.

The result was an agreed paper with the rather long title Inspection Activities Associated with Limited Frequency Unannounced Access Model Applied to Gas Centrifuge Type Enrichment Plants (February 1983). This paper introduced the Limited Frequency Unannounced Access (LFUA) scheme, which includes managed inspector walkthroughs of the cascade area.

Weaknesses of the HSP scheme
The agreed approach contained a number of weak spots. In particular:

• The project participants did not find it necessary to provide for the monitoring of UF6 flows inside the cascades and associated piping. This was simply not seen as relevant.
• Participants felt that there was no need to agree to LFUAs at research and development facilities, simply because the throughput is generally very low.
• Participants also felt that there was no need to monitor the cascade perimeter as a closed boundary (to seal it off completely). The introduction of the LFUA inspector scheme was seen as a sufficient deterrent, and perimeter containment and surveillance therefore completely unnecessary.
• There are no provisions for detecting undeclared throughput. The project concluded that material accountancy and item verification outside the cascade hall, together with LFUAs inside the hall, were a sufficient deterrent for the plant operator. At least as long as the total plant capacity did not exceed 2,000,000 kg/SWU/year. If that logic would hold water today, the enrichment facility in Natanz would need to grow 170 times before the HSP safeguards would not serve as a deterrent.

In 2005, the IAEA met to consider ways to shore up the old HSP approach. Three working groups had been working on the problem, and the following is their main recommendations in respect to excess production and clandestine HEU production.

Addressing excess production

• Employ a resident inspector at the facility to verify flow changes in near real time;
• Put in place a mailbox system that will provide inspectors with real-time process inventory listings;
• Think about options for remote and/or unattended monitoring of UF6 cylinders;
• Sharing of operator load cell and mass spectrometer information on a continuous basis; and
• Additional containment and surveillance measures in states were unannounced inspections are not possible (since the presence of inspectors are telegraphed to the operator long before they actually show up at the gate).

Addressing clandestine HEU production

• Beef up LFUAs by including visual observation, environmental sampling, non-destructive assay measurements and destructive assay sampling from the cascades themselves;
• Installing enrichment process monitors on the cascade header product piping;
• Containment and surveillance within the cascade hall;
• Routine Design Information Verification;
• Interim inspections; and
• Precise weighing using the inspector’s load cells, non-destructive assay measurements, destructive-assay sampling and environmental sampling as appropriate.

Tracking the cylinders
The task of accounting for and controlling the UF6 cylinders that move through the facility is critically important for safeguards purposes. Once the content of cylinders are emptied into the cascade, the cylinder is disconnected and moved to a cooling down area for a while. It may thereafter be hooked up to the cascade again to, say, collect the tails. Some of the weight is lost when volatile gases and some UF6 is boiled off before the cylinder is attached to the process. This may amount to 80 per cent of the facility’s total material unaccounted for.

The current nuclear material accounting and control procedures for tracking UF6 cylinders typically involve manually entering cylinder data into logbooks or computer databases. In a small facility, this is seldom a problem. But as the facility grows, and more cylinders are shuffled around in it, there might be human errors in the data entry process. The IAEA uses conventional seals and weight checks to establish cylinder content so that it has continuity of knowledge, but this may become labour intensive. The Agency spends a considerable time on verifying inventories, shipments, and receipts of UF6 cylinders.

To make things more effective, a number of proposals have been worked up, including the use of RF tags to remotely monitor individual cylinders. Tagging cylinders this way would make it easier to detect when an undeclared cylinder is placed in the feed autoclave.

Diverting material into the MUF
This does not stop an operator from trying to divert materials into material unaccounted for. If the plant is small, this is a pretty hopeless endeavor. If, for instance, a facility is being fed with 48 metric tonnes of uranium hexafluoride gas per year, and this gas is transported in one tonne drums, the total measurement uncertainty in the feed would be about 100 grams. The measurement uncertainty in the low enriched product, shipped of in five one tonne drums, would be about 16 grams. And the measurement uncertainty in the tails, shipped off in 43 containers, would be about 92 grams.

All calculations assumes that the data in the International Target Values 2000 is correct, and I'm working with the relative combined standard uncertainty of the measurement. The figures would be even lower if ten ton cylinders were used, like the 48X cylinders apparently being used by Iran. Look at this video if you are interested in knowing what the UF6 looks like inside the cylinder. And this video if you want to know more about the UF6 phases.

As a rule of thumb, any deviation of more than three times the measurement uncertainty (also known as three 'sigma-MUFs') would have a 95 per cent detection probability, assuming a five per cent false alarm rate. In other words, a difference of 0.05 kilograms between the measured amount of UF6 and the declared amount would raise alarm bells with the Agency, and would warrant further investigation.

A small facility trying to divert material this way could squirrel off about one gram worth of weapons grade uranium per year. That is not a lot by any measure.