asdf asdf

Detecting Atmospheric UF6 and HF as Indicators for Uranium Enrichment

Jens Bösenberg
Martin B. Kalinowski

Clandestine plutonium production can be detected at a distance of 100 km and beyond by atmospheric krypton-85 plumes. By contrast, it is much more challenging to measure a tracer that indicates unreported production of highly enriched uranium (HEU). It has not yet been proven that this is feasible, but many experts are confident that LIDAR (Light Detection And Ranging) could be used by IAEA (International Atomic Energy Agency) inspectors to detect UF6 (uranium hexafluoride) or HF (hydrofluoric acid) discharged into the atmosphere. Research projects have been funded to develop and test the required equipment. But there is severe doubt as to whether signatures are strong enough and measurement sensitivities low enough to make off-site detection possible. This paper compares realistic estimates of the source terms and the theoretically achievable sensitivities of DIAL (Differential Absorption Lidar) for UF6 and HF.

Atmospheric Releases from Uranium Enrichment Plants

Whenever estimates are used for uranium releases from an unreported enrichment plant, they refer back to the same study of Albright & Barbour (1997)1. That paper still remains the only extensive investigation on this topic. It is based on publicly available reports on historic releases from US enrichment plants. Table 1 presents the summary of that study for a reference facility that is defined by the annual production capacity of 25 kg HEU. That is one significant quantity or material for one weapon per year. 

Estimated uranium releases from reference facilities (kg/y)
Type of facility Maximal Central Minimal
UF6 production (conversion) 10 5 0,2
Gaseous diffusion2 7 4 0,04
Gas centrifuge 2 1 0,01
EMIS 150 90 35






Figure 1 shows the timeline of annual uranium releases from a diffusion type enrichment plant. Fairly high releases occurred in the first twenty years of its operation while the facility was comparatively clean in the following twenty years. Technical progress has resulted in much lower emissions. If a country has just mastered the enrichment technology, the uranium release probably provide a much better signature than later in the development cycle when the operator has had a chance to reduce emissions.


Estimated uranium release in Oak Ridge

Figure 1: Estimated annual uranium release to air from the uranium enrichment plant at Oak Ridge, Tennessee, USA3

Figure 2 shows the historic releases of the German centrifuge type uranium enrichment plant at Gronau, which is operated by Urenco. The first important observation is that the activity release rate remained almost constant over two decades while the production rate increased almost linearly. The license emission limit for alpha activity is 5.2 Mbq/a. The actual emissions reached a typical fraction of 1% of the limit, some 50 kBq/a. According to Urenco, 95% of the total alpha activity release consists of radon decay products. With a conversion factor of 25.3 kBq/g U-nat (0.568 kBq/g U-235), the estimated annual release rate is 0.1 g/a of uranium with natural abundance. The Urenco safety report is based on conservative assumptions, namely a mass release rate of 85 g/a. The annual release is caused solely by normal operation conditions, i.e. no accidental releases has occurred at the Gronau facility since 1985. The experience made at Gronau allows two conclusions: first, emissions from a large-scale uranium enrichment plant can be kept very low; second, inspectors can neither hope for an increased signal due to accidental release nor rely on a strong variability of the operational release levels.

Releases from enrichment plant in Gronau

Figure 2: Time series of alpha and beta activity releases from the German uranium enrichment plant at Gronau

It is of interest to compare the Gronau facility with the Urenco enrichment plant at Capenhurst. The alpha activity release of that facility is reported to be 0.60 MBq in 2002, 0.74 MBq 2003 and 0.57 MBq in 2004. This shows that the Capenhurst emissions are more than ten times larger than those at Gronau. Obviously, significant differences can be found between individual plants of similar design. This reinforces the conclusion that no general principles can be found to predict the uranium emissions of enrichment plants.

Releases of Gaseous Tracers UF6 and HF

Gaseous emissions from an uranium enrichment plant could be either UF6 or HF. Below, the possible densities of these trace gases at the exhaust stack shall be assessed.

The uranium enrichment plant operates on the gaseous uranium compound UF6. In humid air, this molecule reacts with water according to the chemical equation

UF6 + 2 H2OUF2O2 + 4 HF

Since this is a fast reaction, it is likely that UF6 is removed either inside the building or very soon after emission. If this is true, the only reliable indicators for remote detection of a uranium enrichment plant could be UF2O2 (uranyl fluoride) and HF.

UF2O2 is particle-bound and can probably be detected only in air filters. UF2O2 could possibly be identified with Lidar based on Raman scattering.

HF can very effectively be contained by filters in the stack. As a result, HF above the detection limit (0.02 mg / m³) has never been observed in the stack exhaust of the Gronau plant. The conservative assumption in the Gronau security report for the emission rate is 320 g HF per year.

In view to these data, the annual emission rate of UF6 is likely between 0.1 and 10 g / a. It would be very optimistic from the perspective of an inspectorate to assume a total emission of 100 g UF6 per year. A typical total volume rate of a single exhaust is 104 m³ / h. This results in a particle density at the stack of

n(UF6) = 2.8*1015 m-3

i.e. Mass ratio of 1*10-10

When further assuming the total conversion of UF6 to UF2O2 and HF, the molecular density would be

n(HF) = 1.1*1016m-3

i.e. Mass ratio of 4.2*10-10

Reduction of Concentrations During Atmospheric Transport

During atmospheric transport away from the source, the plume undergoes dilution, and the concentration of the tracer is eventually reduced towards background levels. With a source term for a uranium enrichment plant as stated above, the trace gas concentrations at a distance of hundreds of kilometers can be estimated. Atmospheric transport simulations have shown that after 24 hours transport, the plume concentrations are highly variable. Typically, 90% of the plume is characterized by a dilution in the range of 1012-1014. This part of the plume extends to ~104 km², and after one day the plume centre has located 100-500 km from the source.

Assuming a total emission of 100 g UF6 per year, this translates into continuous emission of 0.27 g per day or 0.07 g per six hours interval. If the release from six hours forms a plume with a dilution volume of 1012 m³, the concentration of UF6 will be

n(UF6) = 0.2 m-3

If the UF6 is completely converted to UF2O2 and HF, the molecule concentration is

n(HF) = 0.8 m-3

For any realistic scenarios, the chemical instability of HF must also be factored in. Further, any other possible industrial sources of HF have to be taken into account. Accordingly, the derived concentration estimate is very optimistic.

Summary on Releases from Enrichment

The assessment of indicator releases from uranium enrichment plants can be summarized as follows: the gaseous indicators are UF6 and HF. While UF6 is converted prior to release or within minutes after the release, HF is more likely to be found in the exhaust air of a uranium enrichment plant. However, HF is also known to be chemically unstable. In addition, it should be taken into account that there are plenty of other industrial sources for HF.

Any particle-bound chemical compound that contains both uranium and fluorine would be a perfect indicator for enrichment activities because there are no natural or other anthropogenic sources for this combination. The most likely compound that is emitted from enrichment facilities and that contains uranium and fluorine is uranyl fluoride (UF2O2). In view of the above discussion, UF2O2 must be considered the only reliable signature for enrichment plants. Even then, it is to be expected that detectable concentrations occur only close to the release point. Enriched uranium could also be a highly selective indicator for an enrichment plant. However, it has never been detected in any by-pass filter samples taken at the Gronau enrichment facility in the past two decades.

Remote Sensing of Gases from Enrichment

Remote detection of trace gases in the atmosphere it best done by using spectroscopic techniques based on the specific absorption characteristics of each molecule. Active remote sensing is preferred in order to ensure that the method is as sensitive and independent from external light sources as possible, and lasers are suitable because of their excellent spectral properties. The best choice for remote trace gas detection is the Differential Absorption Lidar (DIAL) method, and its basic properties will be explained in the following.

Principle of Lidar

Figure 3: Principle of lidar

DIAL is a special application of the lidar technique, where lidar stands for light detection and ranging in analogy to the well-known radar technique. It works in a very similar way, namely by transmitting short pulses of electromagnetic radiation (light) into the atmosphere and analyzing the backscattered light that is collected in a receiving telescope. The data thus gained includes range information that is acquired by measuring the time it takes the light pulse to travel from the transmitter to the scattering volume and back to the receiver. This is shown in Figure 3. The received power is described by the so-called lidar equation

where P(R) is the received power from range R; C is an instrument constant; ß(R) is the total backscatter coefficient at range RK and a(r) is the total extinction coefficient, including extinction by scattering and absorption. Backscatter in the lower part of the atmosphere originates from both molecular scattering by the air molecules and particle scattering from omnipresent aerosols. Therefore a detectable signal can practically always be obtained provided that transmitted power and receiver sensitivity are adequate. However, there is a principal problem with the quantitative evaluation of such lidar returns: Equation (1) contains two unknowns, ß and a, but only one measurement P. So additional information is needed to find an unambiguous solution.

P(R) * R2=C * ß(R) * exp[-2 * ∫0R a(r) dr]      (1)

DIAL is an elegant way to solve this problem. Here, two measurements are made for two adjacent wavelengths, where one, called “online,” is chosen at the center of an absorption line of the gas under study, while the second wavelength, called “offline,” is chosen in a region of negligible absorption. When the wavelength difference is sufficiently small, the backscatter coefficient is practically the same because backscatter is always a slow function of wavelength. Hence, the ratio of the two signals becomes independent from the unknown backscatter coefficient, and only the difference in absorption remains to be determined. It is then possible to determine the gas concentration cgas in a given range interval ΔR from the so-called DIAL equation, which contains a total of four measured signals, at two ranges and at two wavelengths:

ln(Pon, R2 * Poff, R1 / Pon,R1 * Poff, R2) = -2 * (αon- αoff) * ΔR = -2 * cgas * ( σon- σoff) * ΔR       (2)

on- αoff) is the differential absorption between on- and offline, and (αon- αoff ) * ΔR is called differential optical depth Δ τ.

The DIAL method is generally applicable to all gases, in particular when they show narrow isolated absorption lines. It is well established in atmospheric research, where it is used for the quantitative retrieval of the vertical distribution of, e.g., ozone, sulfur dioxide, water vapor, or hydrocarbons. Several instruments have been built and operated reliably from various platforms, including small trucks and aircraft. The most accurate measurements were made for water vapor, and a detailed error analysis established a detection limit of Δτmin = 0.002 for the differential optical depth, provided that high quality components are used in the detection chain and operating conditions are well controlled. With this limit for the detection of the differential optical depth it is easy to derive detection limits for any trace gas concentration when the absorption cross section is known as well as the range interval in which the trace gas is present:

Δcmin = Δτmin/(Δ σ * ΔR)

where Δcmin is the minimum detectable concentration; Δ σ is the differential absorption cross section of the gas under study; and ΔR is the range interval where the gas is present and the DIAL signals can be evaluated.

In summary, with DIAL it is possible to detect practically any gaseous component with high sensitivity, in particular at distances up to several km, from various platforms, and, if required, in an automated way. However, the instrument has to be designed specifically for each application, adapting the laser and detector systems to the most suitable wavelength and to the spectral resolution requirements as well as to the operating conditions. No serious problems are expected because basic building blocks are available and the methodology is well established.

Estimation of the Detection Limits for UF6 and HF

As explained above, the major gaseous components to look for in the uranium enrichment process are UF6 and HF. Therefore it is useful to estimate the detection limits for the concentration of these gases in remote sensing with DIAL.

From the literature it appears that the absorption cross section for UF6 for the strongest useful band is

Δ σ = 3 * 10-24

From this the minimum detectable number density for UF6 molecules is estimated as

Δcmin, UF6 = 7 * 1020 m-3

corresponding to a mixing ratio by volume of about 30 ppmV (parts per million by volume).

Smaller number densities may be detected when a longer path length can be used. This could be achieved by using a path along the plume rather than across the plume. Pathlengths on the order of 100 m to 1,000 m may thus be possible, by which the detection limit could possibly be reduced to

Δcmin,UF6, UF6= 1.4 * 1018m-3 or 50 ppbV

For HF the strongest useful absorption lines appear to be in the 2.5 µm spectral region, where an absorption cross section of ?s HF = 7 * 10-22 m² is reported in a region with small interference with other common trace gases. At the stack exit, for a range interval of only 1 m, the minimum detectable concentration is

Δcmin,HF= 3 * 1018m-3 or 100 ppbV

Again, smaller number densities may be detected when a longer path length can be used, by which the detection limit could possibly be reduced to

Δcmin,HF= 6 * 1015m-3

This corresponds to a mixing ratio of 0.2 ppbV, which demonstrates that DIAL is a very sensitive method for trace gas detection.


As explained above, estimations of UF6 or HF concentrations during normal operation and during accidental releases cover a very broad range. For UF6 it is rather unlikely that releases at enrichment plants exceed the detection limit of remote sensing with Differential Absorption Lidar. For well operated, clean establishments UF6 emissions are apparently at least three orders of magnitude below the detection limit of lidar, but “dirty” enrichment plants and conversion plants that produce this gas may emit sufficient UF6 to be detectable in plumes close to the release point.

Considering the short lifetime of UF6 in the atmosphere, it is more promising to look for the chemical reaction product HF, although this is less specific as an indicator of undeclared enrichment activities. However, for HF the sensitivity of DIAL is much better because stronger absorption lines can be used, so that the low levels originating from UF6 emissions under normal operating conditions, estimated as about 1 ppbV, may be detectable.

The question whether or not remote sensing of trace gases used in enrichment plants is feasible can ultimately be answered once field measurements have been conducted under realistic conditions, i.e. by making concentration measurements for UF6 and HF in the vicinity of clean and, if possible, also “dirty” enrichment plants.


At the time of writing, Jens Bösenberg was head of the Laser Remote Sensing Group at the Max-Planck-Institut für Meteorologie in Hamburg, Germany; boesenberg [at] dkrz [dot] de.


Martin B. Kalinowski is Carl Friedrich von Weizsäcker Professor for Science and Peace Research at the University of Hamburg and chair of the Independent Group of Scientific Experts; martin [dot] kalinowski [at] uni-hamburg [dot] de.

  1. 1. David Albright and Lauren Barbour, Source Terms for Uranium Enrichment Plants, US Support Program for the IAEA, Draft, August 1997.
  2. 2. Because of uncertainties about the minimum size of a gaseous diffusion plant able to produce weapon-grade uranium and the magnitude of uranium releases in such a facility, these release estimates were derived for a plant with an annual output of 20,000 SWU/yr, which is enough to make 125 kg of weapon-grade uranium per year.
  3. 3. Source: ChemRisk for Oak Ridge Health Studies, Oak Ridge Dose Reconstruction. Draft Task 6 Report Uranium Releases from the Oak Ridge Reservation, February 1997, quoted in Albright & Barbour, op.cit.