Neutron tomography of a highly irradiated spallation target rod

We performed tomographic investigation of the most heavily perturbed (thus highly radioactive) rod and its pristine/unirradiated replicate from the target No. 12 of the Swiss neutron spallation source (SINQ). The tomographic dataset reveals the 3D re-distribution of the lead filling inside the irradiated Zircaloy tube. The change in the linear attenuation coefficient of both the lead filling and the Zircaloy tube of the irradiated rod (due to the presence of the entrapped spallation products) in comparison with the pristine/unirradiated material is quantified. The dataset provides valuable input for the enhancement of safety and efficiency of future spallation targets at SINQ.


Introduction
The Swiss neutron spallation source (SINQ) [1] at the Paul Scherrer Institut (PSI) produces free neutrons by means of a target bombarded by an intense 590 MeV proton beam. The SINQ target comprises more than 300 hermetically-sealed Zircaloy tubes filled with lead up to 90 percent of its inner volume (see Fig. 1). The desired original distribution of lead filling inside the Zircaloy tube is achieved by lead melting and solidification [2]. The distribution of the pristine lead filling in the fully welded tube is then checked by neutron radiography using standard scintillator-camera detector.
The SINQ source is a continuous spallation source and it experiences a large number of short operational interruptions (in the order of fifty thousand over the standard two years of its operational lifetime). These include intentional pauses for the production of ultracold neutrons (so-called UCN-kicks) [3] and unintentional beam interruptions in the production of protons. Even though only a fraction of the incoming proton beam energy is deposited at SINQ [4], the operational pauses lead to cycles of melting and solidification of the lead filling (lead melting point is 600.6 K) in the most heavily thermally loaded rods [5]. The volumetric fraction of the lead inside the rods is only up to 90% which allows for the thermal expansion/contraction of lead during melting and solidification cycles (lead density in molten state 10.66 g/ cm 3 , lead density at 25 K 11.34 g/cm 3 ). Though, it has been recently demonstrated using neutron radiography of inactive target rods subjected to heating/cooling cycles [6] that such repeated melting/solidification process leads to redistribution of the lead filling inside the Zircaloy tubes.
In this process, the centre part of the target rod gets progressively filled up fully with the melted/solidified lead filling, while the solid lead parts of the filling at the ends of the tube shrink progressively towards the centre of the rod. This poses a potential risk to the evolution of cracks in the irradiated Zircaloy tubes due to the build-up of hydrostatic pressures in fully filled up parts of the rods and, therefore, to the safety of the target [7]. The detailed knowledge about the distribution of lead inside the irradiated target rods is therefore of crucial importance for both the safe operation of the source and for the optimization of the source efficiency.
The SINQ targets at the end of its lifetimes represent very highly radioactive samples (dose rate ~ 100 Sv/h). The non-destructive investigation of the inner structure of highly radioactive materials using neutrons is demanding and is routinely performed only at a very few facilities worldwide (e.g. [8][9][10][11]). In Paul Scherrer Institute, such samples can be imaged at the NEUTRA thermal neutron imaging beamline [12] by means of NEURAP-a dedicated set-up for highly radioactive materials. Applications of NEURAP for imaging of radioactive materials have been recently reviewed by Lehmann et al. [13]. It should be noted that all but one [14] of the applications in the above mentioned review are limited to 2D radiography. In this pioneering work we present the first attempt for tomographic investigation using a reasonably high number of projections with the goal of the visualization of the lead distribution in the most perturbed rod from target No. 12 of the Swiss neutron spallation source (SINQ).

Experimental
The SINQ target No. 12 has been operated in the years 2016 and 2017 and received the total proton current integral of 5,239 mAh. During that time the target experienced 23,435 UCN-kicks and 13,504 additional unintentional short proton accelerator operational pauses. In September 2020, five rods were removed from the target in a large dedicated hot cell (ATEC) and placed into the NEURAP sample capsule for neutron radiographic investigation. Positions of the investigated rods in the target No. 12 are clearly marked in Fig. 1 left. The radiographies showed major redistribution of lead in the first four rods (Nos. 1-4) while the rod No. 5 exhibited hardly any apparent redistribution of the lead filling.
Based on the radiographic investigations, the rod No. 2 (originally denominated Z4-5) exhibited the largest apparent redistribution of the lead filling (see the arrow in Fig. 2) making it likely the most perturbed rod of the target No. 12 and was thus selected as the sample of interest for the neutron tomographic investigation. The dose rate of the target rod No. 12 measured shortly before the tomographic investigation was equal to 345 mSv/h.
For the tomographic investigation, the target rod No. 2 has been placed at bottom of the NEURAP aluminum sample capsule. In addition, a pristine (unirradiated) Zircaloy replicate rod was placed on the top of the rod No. 2. For the purpose of the assessment of the spatial resolution, a broken piece of spatial resolution test object (gadolinium Siemens star [15,16], was fitted on the top of the pristine target rod. The tomographic investigation of the two target rods was performed at NEUTRA beamline at measuring position No. 2 (L/D = 365) in April 2021 (see Fig. 3). A set of 27 dysprosium based imaging plates [17] of 40 × 250 mm in size were utilized for the investigation. In the first step, the images of all 27 imaging plates without any sample present in the beam were acquired for the subsequent image normalization. The following temporal sequence was consistently applied for the data acquisition: (1) 15 min pre-erasure, (2) 20 min neutron exposure, (3) 15 min erasure, (4) 90 min self-exposure The sample capsule has been lowered to the beam position in such a manner that the entire length of the irradiated sample and a lower part of the pristine sample were in the available field of view. The samples were then subsequently scanned in thirty-seven angular positions that were evenly distributed along 180 degrees rotation. The following interleaved acquisition scheme [18] has been applied-0:20:180, 10:20:170, 5:10:175. Taking into consideration the decay time of Dy-165 m (half-life of 2.334 h), it was assured that the same imaging plate was not used more than once within 24 h. The acquisition sequence was completed with an image of the entire pristine sample with gadolinium Siemens star for the check of the spatial resolution.
The total experimental time for the tomographic investigation was approximately 32 h. All the original images were of the size of 9920 × 1440 pixels. The pixel size was equal to 25 µm.  The original projection images were normalized by the corresponding images of the imaging plates without samples. As the imaging plates could not be placed in the imaging plate reader with sufficient (sub-pixel) reproducibility the images were registered to each other based on the area outside the sample using affine registration routine in Avizo software [19]. The examples of the original and the normalized image are shown in Fig. 3.
After the normalization all the 37 projection images were manually prealigned and cropped to the size of 7236 × 744 pixels. In the next step, the stack of the normalized images was registered using StackReg routine in Fiji software [20]. Due to the fact that the 37 projections are still far from satisfying the sampling theorem, the stack of images was binned by factor 4 × 4 leading to 37 images of 1809 × 186 pixels in size and of the resulting pixel size of 100 µm.
The dataset has been reconstructed using a standard filtered back projection algorithm using Muhrec software [21] (Hamming filter, cut-off = 0.5) thus providing the 3D map of linear attenuation coefficients [22]. Wavelet-FFT ring artifact removal algorithm [23] was applied during the reconstruction. The reconstructed dataset was post-processed using ISS edge preserving filter [24] using Kiptool software [25].

Results and discussion
As this is the first tomographic investigation of highly radioactive samples using the NEURAP insertion device that is based on reasonably high number of projections, the spatial resolution of the technique is discussed first here. The visual assessment of the image of a part of the gadolinium Siemens star [15,16] at the top of the pristine sample revealed approximately 120 µm spatial resolution for a single spoke size (240 µm line pair)-see Fig. 4. At the same time, a Zircaloy edge response function from the 3D dataset was evaluated to be approximately 300 µm (10-90% of the edge response distance). Figure 5 shows the reconstructed vertical slice of the tomographic dataset from approximately axial position of the both target rods as well as the reconstructed horizontal/axial slices of both the rods. The observed attenuation coefficient of pristine lead (Ʃ pristine_lead = 0.21 cm −1 ) does not differ significantly from that of the pristine Zircaloy (Ʃ pristine_Zircaloy = 0.20 cm −1 ). The average linear attenuation coefficient of lead filling of the highly radioactive sample changes significantly along the horizontal position of the target (see Fig. 6). The highest attenuation coefficient  Fig. 5 bottom right) which represents more than 57 percent increase with respect to the pristine material. This is consistent with the fact that the centre of the highly radioactive rod received the highest dose of protons and therefore is expected to exhibit the highest level of spallation products (e.g. hydrogen) exhibiting superior neutron cross-section than that of the lead. Likewise, the increase in linear attenuation coefficient is observed also in the case of the highly radioactive Zircaloy tube (Ʃ irradiated_Zircaloy = 0.25 cm −1 ). However, this increase is relatively lower than the corresponding relative increase in the linear attenuation coefficient of lead (25 percent for Zircaloy versus more than 57 percent for lead). This observation is, however, consistent with the lower proton capture cross-section of zirconium than that of the lead. The linear attenuation coefficient of aluminum in the NEURAP's 16 mm-diameter sample capsule equals (Ʃ aluminum = 0.08 cm −1 ).
Thanks to the relatively low noise level in the reconstructed datasets several material phases can be easily rendered. Figure 7 presents the 3D renderings of the distinguishable material phases in the highly radioactive target rod. The presence of water (blue phase in Fig. 7 top left)likely due to the insufficient drying of the sample capsule during the preceding decontamination procedure-was unexpected. Several droplets of the decontamination liquid can be observed both on the outer wall of the Zircaloy tube as well as on the inner wall of the 16-mm-diameter aluminum sample capsule.
The shape of the Zircaloy rod has been revealed in 3D and its thickness analysis can be thus performed at an arbitrary cross-section. The approximately 100 µm larger thickness of the rod in the centre in comparison with the thickness at Fig. 6 The linear attenuation coefficient of lead in the highly activated SINQ spallation target rod plotted as the function of the horizontal position of the rod Fig. 7 3D renderings of the identified material phases: blue-water droplets in the sample capsule, dark green-Zircaloy tube, yellow-lead filling, black-porosity in the lead filling. The size of the visualized bounding boxes is equal to 12.6 × 12.3 × 117 mm the rod's ends had been observed and was confirmed by the tactile measurements performed in the ATEC hot cell.
Regarding the shape of the redistributed lead filling, the tomographic dataset revealed that approximately 50 mm of the central part of the highly irradiated rod has been fully filled with the lead while the lead filling shrunk approximately 10 mm one side of the rod. The defects/pores in the lead filling-otherwise undetectable by radiography onlywere revealed in the tomographic datasets as well (see Fig. 7 bottom right).
In general, the presented investigation offers the first NEURAP based tomographic dataset of highly radioactive material which allows a quantitative analysis based on the derived linear attenuation coefficients (LAC). The lower than expected values of the derived linear attenuation coefficients are to be expected due to the fact that no scatter correction [26,27] could be applied during acquisition. The observed increase in LAC of Zircaloy due to the irradiation corresponds well with the expected level of implanted spallation products (e.g. 6600 appmH in the SINQ target No. 11) [7].
The homogeneous distribution of linear attenuation coefficient along the perimeter (see Fig. 5 bottom right) suggests that no areas of inhomogeneous hydrides accumulation were present in the radioactive target rod. This is consistent with the recent results based on the high-resolution neutron imaging [28,29] of SINQ target rod material [30]. As much as this diminishes the concerns about the extent of the local embrittlement, the large area of the rod fully filled with the lead together with the observed increase in the rod's thickness in its centre part gives reasons for safety concerns [7].

Conclusions
The first tomographic investigation of highly radioactive (345 mSv/h) SINQ spallation target rod using a reasonably large number of projections was performed. The obtained neutron tomography dataset is pioneering in numerous ways and goes qualitatively beyond any tomography of highly radioactive object made so far using the NEURAP technique. It reveals in 3D the re-distribution of the lead filling inside the rod as well as the precise shape of the deformed target rod (including the approximately 100 µm increase in the thickness of its centre part). Thanks to the applied image normalization trustworthy values of the linear attenuation coefficients of the materials are derived. The change in the linear attenuation coefficient of the lead filling-probably thanks to the presence of the entrapped spallation products-in comparison with the pristine lead was quantified. The dataset will be used in the future as a valuable input for possible design improvements of future SINQ targets both in the direction of its safety and its efficiency.
Funding Open Access funding provided by Lib4RI -Library for the Research Institutes within the ETH Domain: Eawag, Empa, PSI & WSL.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.