Using energy-resolved neutron imaging, a U-20Pu-10Zr-3Np-2Am (weight percent) sample, an alloy researched for transmutation nuclear fuels [19], was characterized to obtain the bulk composition for comparison with mass-spectrometry. The sample consisted of a 20 mm long, 4.2 mm diameter U-20Pu-10Zr-3Np-2Am (weight percent) cast fuel slug contained in a double-walled steel container. The sample was prepared at Idaho National Laboratory where also mass spectrometry was performed (Table 1). The sample was mounted on a rotation stage and energy-resolved neutron imaging data was collected for 120 minutes per rotation for 65 rotations.
Considering only an active-area on the detector from pixels where the neutron beam traversed the sample volume (Figure 1, A), 50,000 (125 ⋅ 400) pixels required neutron transmission data analysis per sample rotation, resulting in 3.25 million total transmission-spectra fits. The SAMMY code [12] was used with cross section data for the isotopes 237Np, 238U, 239Pu, 240Pu, and 241Am obtained from the ENDF/B-VIII.0 data base [20]. For conventional transmission measurements, the sample is generally several tens of meters away from the detector, such that sample induced background can be neglected [21]. In contrast, for imaging measurements, the sample position is as close as possible to the detector to reduce blurring resulting from the divergent beam. However, this sample position impacts the background in imaging applications and therefore the sample induced background needed to be properly accounted for to obtain reliable areal and volumetric densities. To accomplish this, a Ta foil was mounted on the detector window with a thickness of 100 µm, leading to opaque resonances. This allowed for reliable determination of the background, including sample contributions, by the transmission values at the bottom of these resonance dips in the transmission data.
Figure 1 shows an example of the data analysis for single pixels with fitted transmission for a pixel within the slug in B) and outside the slug in C). The total computing time for the data analysis of the tomographic data-set processing 3.25 million spectra on a desktop PC using 64 ⋅ 2.4 GHz CPUs was 16 days. The bulk composition was determined from the average density of the reconstructed sample volume. Table 1 shows the comparison of the results with the results of the mass spectrometry. The maximum discrepancy for the four isotopes except 238U is 0.08 g/cm3, establishing the reliability of this method. The reason for the discrepancy of 1.8 g/cm3 for 238U lays in the properties of the 238U resonance in the dataset. The analyzed energy range of 0.2 to 9 eV only provided a single 238U absorption resonance at 6.67 eV. At a fractional density of ~9 g/cm3 and cross sections >>103 barn, this led to insufficient non-zero transmission at resonance energies for a reliable data fit. Furthermore, the absorption resonance at 6.67 eV for 238U has a comparatively narrow profile with respect to the other isotopes present and the resonance overlaps significantly in the non-zero transmission range with other resonances. Increasing the energy range for the analysis could mitigate this problem but was not feasible with the available resources for all 3.25 million datasets.
Table 1
Isotope densities from energy-resolved neutron imaging compared with those from mass spectrometry. For the neutron analysis, the factional density was computed by the average of the reconstructed sample volume, averaged over all voxels fully inside the sample with errors computed using the standard deviation of the voxel densities.
Isotope | Weight fraction from mass spectrometry [µg/g] | Fractional density from mass spectrometry [g/cm3] | Fractional density from neutron analysis [g/cm3] |
234U | <30 | 0.00 | |
235U | 1490 | 0.02 | |
236U | 113 | 0.00 | |
237Np | 24400 | 0.34 | 0.310(3) |
238U | 639000 | 9.01 | 10.8(2) |
239Pu | 166000 | 2.34 | 2.26(2) |
240Pu | 26400 | 0.37 | 0.369(3) |
241Am | 23000 | 0.32 | 0.314(3) |
Zr | 100000 | 1.41 | |
Total | 980433 | 13.82 | 14.09 |
The large absorption cross section for thermal neutrons of 239Pu made it impossible to utilize thermal neutrons for neutron tomography. Using the sample composition, the computed transmission at a thickness of 4.2 mm for 25 meV neutrons equates to 3.5%. For such opaque samples, the reconstruction leads to artefacts of the average voxel density, so-called ‘beam hardening’ [22]. However, at epithermal neutron energies (E > 0.4 eV) the cross-section drops such that significant fractions of the beam are transmitted and a reliable tomographic reconstruction is possible. Figure 2 shows a 3D rendering of the fuel slug resulting from neutron tomography by selecting time-of-flight neutrons only at the upper end of the thermal spectrum, with energies ranging from 0.1 to 0.2 eV. Several globular features are apparent. The tomographic reconstruction of the densities of the aforementioned isotopes allowed further investigation of these features and greatly reduced densities for all isotopes were found at the locations of the globular features. This would indicate that the features are either casting voids or Zr-rich metallic inclusions (Zr did not provide neutron absorption resonances in the 0.2 to 9 eV energy range used for the measurements). With the slices of the tomographic reconstruction providing data similar to elemental maps provided by X-ray microprobe, but within the bulk of a sample, non-destructively, and for samples that have to be in containers such as nuclear fuels, this characterization guides destructive post irradiation examination by identifying where this sample should be cut for further investigations.