Gamma-ray spectroscopy measurements and simulations for uranium mining

AREVA Mines and the Nuclear Measurement Laboratory of CEA Cadarache are collaborating to improve the sensitivity and precision of uranium concentration evaluation by means of gamma measurements. This paper reports gamma-ray spectra, recorded with a high-purity coaxial germanium detector, on standard cement blocks with increasing uranium content, and the corresponding MCNP simulations. The detailed MCNP model of the detector and experimental setup has been validated by calculation vs. experiment comparisons. An optimization of the detector MCNP model is presented in this paper, as well as a comparison of different nuclear data libraries to explain missing or exceeding peaks in the simulation. Energy shifts observed between the fluorescence X-rays produced by MCNP and atomic data are also investigated. The qualified numerical model will be used in further studies to develop new gamma spectroscopy approaches aiming at reducing acquisition times, especially for ore samples with low uranium content.

its detection in ore samples of a few hundred grams with a low uranium concentration may require measurement times of several hours.Therefore, we plan to use more intense peaks at lower energy, such as self-fluorescence X rays of uranium, to reduce acquisition times.Fluorescence is induced by the main gamma radiations emitted in the ore by 214 Pb (242, 295, 352 keV), 214 Bi (609 keV) and 226 Ra-235 U (186 keV).After Compton scattering, these gamma rays produce a high continuum of lower energy photons in the K-edge region of uranium (115.6 keV), resulting in a large photoelectric absorption rate and, subsequently, in the emission of uranium fluorescence X-rays, like the most intense 98.4 keV line.In order to study the potential of these new approaches and to establish a numerical simulation model, a measurement campaign was carried out in the radiometric calibration station of AREVA Mines in Bessines, France [1], with a coaxial HPGe detector.This station holds seven cement blocks with increasing uranium contents, up to 1% weight fraction (i.e.10,000 ppm).The optimization of the MCNP model of the HPGe detector will be presented, as well as a comparison of different nuclear data libraries to explain missing or exceeding peaks in the simulation.Energy biases between the fluorescence X-rays produced by MCNP and atomic data will also be investigated.

II. THE RADIOMETRIC CALIBRATION STATION
AREVA Mines radiometric calibration station in Bessines aims to measure the counting rate due to gamma radiation emitted by seven independent cubic standard blocks with 70 cm edges (Fig. 1).The uranium content of each block has been determined by chemical analysis (Table I).These blocks allow the calibration of radiometric probes.Gamma-ray spectroscopy measurements and simulations for uranium mining The Nuclear Measurement Laboratory of CEA Cadarache has carried out a series of gamma-ray spectroscopy measurements on these standard blocks (Fig. 2).Results have confirmed the uranium content of each block, as shown further in Table II.
The gamma spectra acquired during this campaign (Fig. 3) will also be used to validate the MCNP numerical model and to identify new information (X or  peaks, Compton continuum, K-edge, etc.) that could be used to improve the assessment of uranium concentration.This campaign has been performed with a type N, coaxial HPGe detector, model GR1020, CANBERRA.

III. SIMULATION OF THE COAXIAL DETECTOR
The coaxial detector was modeled with the MCNP Monte Carlo computer code [2].The geometric model of the detector is based on the manufacturer scheme, on X-ray radiography, and on measurements performed with a multi-energy beam coming from a highly collimated 152 Eu gamma source (see Fig. 4), which allowed to precisely estimate the position and active area of the crystal.The calculated and experimental detector efficiencies are reported in Fig. 5.The measured efficiency was built with calibration point sources ( 241 Am 59.5 keV line, 109 Cd 88.0 keV peak, and 152 Eu gamma-rays from 121.8 keV to 1408 keV) at a distance of 50 cm from the detector head.Note that to obtain the good agreement observed in Fig. 5, a dead layer has been implemented in the model of the germanium crystal, in the back, external edge of the germanium cylinder, to reflect a deficit of charge collection in this region.A copper electrode connecting the cooling system and the germanium crystal is also described, as well as the external aluminum cover.The mass concentration obtained by gamma-ray spectroscopy using (1) is in good agreement with chemical analysis, as shown in Table II.The reported uncertainty includes the detector calibration and modeling uncertainty (2.6 %), the statistical uncertainty on the net area of the peak at 1001 keV (between 0.9 % and 22 %, depending on uranium concentration), the uncertainty on the 1001 keV gamma-ray intensity (1.32 %), the statistical uncertainty of MCNP calculations (less than 1 %), and the uncertainty of the MCNP modeling of the concrete blocks and measurement setup.Concerning this last, the main source of uncertainty concerns concrete density, which varies from 1.89 to 1.93 g.cm -3 depending on the measured block.The dispersion of the efficiency values calculated on the blocks with MCNP is equal to 0.97 %.This value has been retained to be the uncertainty of the modeling of the measurement scene at 1 σ.The global uncertainty is the quadratic sum of these uncertainties.Pb.The intensity and energy of each gamma ray are taken from the JEFF 3.1.1library, and only the lines with energy larger than 40 keV are modeled.Fig. 6 shows a good general agreement between simulated and experimental spectra, only small differences being observed for minor peaks.For instance, the 89.5 keV peak was initially absent from the simulated spectrum due to a missing emission line of 214 Pb at 89.8 keV in the JEFF 3.1.1library (Bi K2 Xray, intensity of 0.67% in ENSDF [3]), combined with an underestimation emission of 214 Bi at 89.2 keV (Po K3 X-ray, 0.085% in JEFF 3.1.1instead of 0.116% in ENSDF [3]).After implementing ENSDF data [3], the calculated 89.5 keV peak was found in satisfactory agreement with experiment (Fig. 7).
Fig. 6.Experimental and MCNP spectra of block B5 using ENSDF database.
We can also note in Fig. 7 that a peak is missing in the simulation at 110.4 keV, in the broad bump including both 110.4 and 111.3 keV uranium fluorescence X-rays (Kβ3 and Kβ1, respectively [4]).Indeed, MCNP manual [5] indicates that for elements with atomic number Z > 37, the X-ray emission Kβ3 is not simulated, and Kβ1 X-ray includes all transition intensities from layers L to K, which explains the absence of the 110.4 keV peak (Kβ3) and a higher intensity at 111.3 keV (Kβ1).In further studies, this energy shift will be corrected when processing MCNP output data to allow a better agreement with the experimental spectrum.

VI. CONCLUSION AND PROSPECTS
Seven calibration blocks with known uranium concentrations have been measured by high resolution gamma-ray spectroscopy at AREVA Mines calibration station, in Bessines, France, and simulated with MCNP computer code.
First, a detailed model of the coaxial HPGe detector has been carried out using calibration acquisitions with standard point sources of 109 Cd, 152 Eu, and 241 Am.Then, the uranium mass concentrations of the blocks (in ppm), originally determined by chemical analysis, has been confirmed by gamma-ray spectroscopy with the 1001 keV gamma ray of 234m Pa, in the 238 U radioactive chain.Finally, we obtained a very good agreement between the overall gamma spectra of the blocks calculated with MCNP and the ones measured in Bessines.Only minor discrepancies concerning small peaks have been observed and corrected, which were due to unprecise or missing nuclear data, as well as small bias in the energy of fluorescence X-rays produced by MCNP, which will be corrected in future studies.We will now use the validated numerical model to study new gamma-ray spectroscopy methods, in view to characterize uranium faster than with the reference 1001 keV line.Indeed, its intensity (0.84 %) leads to hours of acquisition time for ore samples of a few hundred grams with uranium concentration lower than 1 % mass fraction.Among the information that is not yet used, uranium self-fluorescence X-rays in the 100 keV region (K and K transitions) can be detected in a much shorter time.They are induced by gamma radiations of the ore itself and could be used to characterize uranium up to a depth of several centimeters, hence being appropriate for small ore samples.

Fig. 3 .
Fig. 3. Gamma-ray spectra of the seven calibration blocks of AREVA radiometric station in Bessines, France.

Fig. 4 .
Fig. 4. X-ray radiography and the modeling of the coaxial detector.

Fig. 5 .
Fig. 5. Experimental and MCNP efficiency of the coaxial HPGe measured with point sources at a distance of 50 cm from the detector entrance face.IV.VERIFICATION OF THE URANIUM CONTENT IN THE STANDARD CONCRETE BLOCKSThe global detection efficiency, including self-absorption in the standard cement blocks, has been calculated with MCNP with a first objective to determine the uranium mass concentration of each block, using the 234m Pa gamma ray at 1001 keV.Equation (1) allows converting the net area of this peak into uranium mass concentration (ppm).

:::
mass concentration of each block (ppm) ) 1001 ( k e V S n : Net area of the peak at 1001 keV measured on the gamma spectrum (number of counts, dimensionless number) Tc: Acquisition "live time" (i.e.real time corrected for dead time) of the gamma spectrum (sDetection efficiency calculated with MCNP for each block (number of count in the full-energy peak at 1001 keV per photon of 1001 keV emitted in the block) Radioactive period of 238 U (s) A N : Avogadro constant (6,023 . 10 23atoms per mole, Molar mass of 238 U (g) : Mass fraction of 235 U in natural uranium (0.72 %)

Fig. 7
Fig. 7 also shows a significant energy shift between calculation and experiment concerning uranium fluorescence

Fig. 8 .
Fig. 8. Energy shift between the energy of fluorescence X-rays calculated with MCNP and the theoretical values taken from [3].

TABLE II .
URANIUM MASS CONCENTRATIONS (PPM) OF BESSINES Th, 226 Ra, 214 Pb, 214 Bi, 210 Pb, 235 U, 231 Th, 227 Th, 223 Ra, 219 Rn, and 211 The simulation of the concrete blocks is based on a cement chemical composition given by AREVA Mines, including uranium concentrations (and densities) given in TableI.An isotropic gamma source uniformly distributed in the cement block is implemented in MCNP.The natural 238 U and 235 U radioactive chains are supposed in secular equilibrium.The source term retains only 14 predominant gamma emitters observed in the experimental spectra: 238 U, 234 Th, 234m Pa,230