Hydrogen diffusion in zirconium cladding alloys with an inner liner as quantified by neutron radiography and nanoindentation

Embrittlement caused by hydride precipitates in zirconium claddings constitutes a significant risk factor for the integrity of spent nuclear fuel rods, especially during the handling and transport operations necessary before and after long-term intermediate dry storage. Zircaloy-2 is the most commonly used alloy for cladding tubes in boiling water reactors. It is typically manufactured with an inner liner to better cope with issues related to pellet-cladding interaction. Previous work showed that in clad samples possessing a liner, a significant portion of hydrogen tends to migrate towards the liner region during cool-down. In this work, the hydrogen precipitation and distribution after thermal exposures comparable to service was investigated in different commercial claddings for boiling water reactors. The resulting hydrogen distribution was quantified by means of high-resolution neutron imaging at the SINQ spallation neutron source at Paul Scherrer Institut in Switzerland, and the presence of hydrides was subsequently confirmed by metallography and nanoindentation mapping of selected samples. The results show that when a cooling rate of 30 C/h is applied, the vast majority of the hydrogen tends to accumulate at the liner/substrate interface when a liner is present. At lower cooling rates, the hydrogen tends to distribute more homogenously into the liner material, and the cladding bulk material is left completely hydrogen-free. Neutron radiography of samples after water quenching from homogenization temperatures between 370 C and 460 C revealed a significant amount of hydrogen trapped in the hydrides form at the liner/substrate interface at temperatures up to 400 C. Undissolved hydrides may act as nucleation seeds for hydrides upon cooling, contributing to the observed hydrogen segregation. A model that takes into account the interfacial properties of the region between the liner and the bulk material is proposed.


Introduction
Zirconium (Zr) alloys are one of the most common materials for fuel cladding and structural components in light-and heavy-water nuclear reactor cores because of their excellent mechanical properties, good corrosion resistance, and low neutron cross section [1][2][3]. Zircaloy-2 is one of the main Zr alloys used as fuel claddings in boiling water reactors (BWRs). Some of these fuel claddings are produced with an inner liner to better cope with problems related to pellet-clad interaction (PCI) and fission-gas corrosion [4]. Liners consist of a thin (typically 10% of the total clad thickness) layer of a more pure-grade Zr alloy metallurgically bonded to the inner surface [5].
Zircaloy-lined fuel cladding tubes were introduced by GE in the early 1990s [5] to counter PCI and the subsequent failure mechanism [6][7][8] in fuel for BWRs; they were shown to provide a major improvement in fuel reliability and operational flexibility. Despite the fact the frequency of fuel failures has been considerably reduced, an increased tendency of failed fuel rods to exhibit post-failure degradation in the form of longer cracks has been observed. One reason for this degradation is the precipitation of hydrides in the cladding, sometimes in remote locations from the initial perforation of the fuel rod. The location of the hydride precipitation is often associated with the stress increased by the fuel expansion and burn-up [9]. Later, an adjustment of Zr-lined barrier fuel cladding tubes was developed to delay the post-failure local hydride precipitation while retaining the benefit of PCI resistance.
Regardless of their PCI resistance, liners have been reported to act as a preferential nucleation site for hydrides. This phenomenon has been observed in post-irradiation examinations in both irradiated and unirradiated samples [10][11][12] in both Zircaloy-2 and Zircaloy-4 based clads. Hydride precipitation, growth, and orientation in nuclear cladding material is of high importance for the integrity of spent nuclear fuel (SNF) and have been studied extensively in zirconium alloys. Despite the large amount of research on Zircaloy-4 [13][14][15][16] and Zr-2.5Nb [17,18] alloys, research on Zircaloy-2 alloy [10,11,19] is more limited.
As an example, Zircaloy-2 with an ~80 μm thick inner liner of almost pure Zr was analyzed by Une et al. [20]. In that work, the precipitation morphology and habit planes of the Zr hydride phases were evaluated via electron back-scatter diffraction (EBSD). The authors identified δ-hydrides in both liner and bulk, with intra-granular hydrides being about 2 times more frequent in the liner region compared to the bulk material.
In a subsequent work, Valance et al. [21] observed strong precipitation of hydrides at the liner interface in contrast to an unchanged microstructure in samples without liner. Similar behavior was observed in irradiated Zircaloy-2 cladding tubes subjected to hydrides reorientation experiments [11], where Valance et al. observed very few hydrides precipitating in the cladding matrix when a liner was present. More recently, Duarte et al. [19] analyzed the diffusion of hydrogen into the liner of unirradiated Zircaloy-2 claddings with high resolution neutron radiography. The cooling rates applied were 0.3 and 10 • C/h, and a strong accumulation of hydrogen at the liner-bulk interface and inside the liner could be found. With the lower cooling rate, the accumulation of hydrides expanded more into the liner, and the concentration of hydrides was less pronounced directly at the interface compared to the test samples with higher cooling rate.
In the present work, the influence of the cooling rate on the hydrogen diffusion toward the inner liner was studied at rates of 3 and 30 • C/h. Moreover, the hydrogen distribution present in the samples prior to controlled cooling has been evaluated by water quenching from different homogenization temperatures. Besides high-resolution neutron radiography at the SINQ spallation neutron source at Paul Scherrer Institut (PSI), the phases present in the material were characterized by nanoindentation mapping, revealing the hardness variations at the interfacial region. The influence of the different cooling rates on the hydrogen migration and precipitation is analyzed with respect to the hydride precipitation inside the liner and at the inner liner/substrate interface of the respective Zircaloy-2 samples.

Materials
The materials employed in this study are three different types of commercial Zircaloy-2 cladding tubes supplied by Framatome (LTP2 and LTP) and Westinghouse (LK3/L). The sample dimensions are provided in Table 1. Of the three types of material, the LTP2 has no liner, whereas the LK3/L and LTP have an inner liner.
Zircaloy-2 cladding tubes contain Sn in concentrations between 1.20 and 1.70 wt.% and other alloying elements in lower concentrations, such as Cr (0.05-0.15 wt.%), Fe (0.15-0.20 wt.%), and Ni (0.03-0.08 wt.%) as described by the relative ASTM specification. The inner liner composition is a Fe-or Sn-microalloyed Zr alloy containing minor amounts of the alloying elements, with levels between 0.085 wt.% and 0.5 wt.%. The liner compositions were a purer grade of Zr with a microaddition of 0.3 wt.% Sn for the LK3/L sample and 0.4 wt.% Fe for the LPT sample.
From metallographic analysis using optical microscopy and a circular polarization filter (Fig. 1b), the liner shows a coarser microstructure (average diameter ~12 µm) compared to the bulk material (average diameter ~6 µm).

Hydrogen loading and concentration determination
Tubular samples approximately 15 cm in length have been enriched to hydrogen concentrations of 100 and 200 wppm. This was performed via diffusion of hydrogen into the sample material at high temperature in a Sievert-type apparatus, following a similar process to that described by Fagnoni et al. [22]. The procedure consists of evacuating the chamber containing the sample to 10 − 7 bar, then exposing the sample, which is heated to 400 • C, to an atmosphere of 3-7 mbar of pure hydrogen. The hydrogen absorption by the sample was then monitored via the decrease in hydrogen pressure measured in the chamber.
Following the hydrogenation process, the specimens were subjected to a homogenization heat treatment at a temperature of 400 • C for 24 h in a furnace with a protective Ar atmosphere with the aim of creating a uniform distribution of hydrogen throughout the specimens. After homogenization, multiple ring specimens of 5 to 20 mm in length were sectioned from the hydrogenated tube specimen, as shown in Fig. 2. Every 20 mm, a small section was extracted for destructive hydrogen quantification analysis by hot gas vacuum extraction (HVE). Results obtained by HVE are presented in Fig. 3, whereas the measured average hydrogen concentration is compared to the expected hydrogen concentration in Table 2.
Hot gas extraction values of the measured locations show 20% higher average hydrogen concentrations than that expected from the pressure drop in the hydrogenation chamber. Additionally, the hydrogen distribution along the length of each hydrogenated tube showed a consistent gradient in concentration, likely indicating the presence of a thermal gradient in the homogenization chamber. Ring samples with an axial length of 4.5 mm were selected from the central section of the rod, and used for heat treatment and subsequent examinations. The samples were selected from regions adjacent to locations examined using HVE to facilitate a good estimation of the hydrogen content in the imaged section. The estimated values of hydrogen content in the selected samples, as obtained by interpolation of the two nearest HVE locations, are reported in Table 3.

Heat treatment
To investigate the hydrogen diffusion within the samples and hydride precipitation, selected rings were subjected to a second heat treatment consisting of 8 h holding temperature at 400 • C. After this holding time, samples were directly exposed to different cooling rates: controlled furnace cooling at 3 and 30 • C/h and water quenching.
In a second part of the study, a sample of LTP material (with liner) was subjected to a repeated homogenization treatment for 8 h at temperatures of 370, 400, 430, and 470 • C. After each holding temperature, the sample was quenched in water and the radial hydrogen distribution assessed by neutron radiography.
All thermal treatments were conducted with a constant Ar flow and the chamber was purged with the inert gas immediately before loading the samples. A thin layer of black oxide was observed despite the protected atmosphere, but it was considered non-critical for the test purposes; thus, it was not removed before imaging.

Hydrides distribution characterization
The hydrides distribution within the samples before and after heat treatment was assessed by neutron radiography at the Swiss spallation neutron source, SINQ, at the POLDI beam line [23].
The neutron microscope used in this experiment is equipped with a  [24,25]. Both experimental procedure and post-processing follows the same methodology as developed by Gong et al. [12]. The axis of tube segments was positioned parallel to the neutron beam by adjusting the sample stage.The final projection image (I) was obtained through pixel-wise referencing of the sample image (I sample ) to an open beam image (I open ) and a dark current image (I dark ) using the following equation:    The individual images used for Eq. (1) were originally acquired as sequences of 30 images with exposure time of 100 s. The sequence was later average-filtered following the removal of 'outliers' seen as white spots. The neutron images were processed and statistically analyzed with the ImageJ software.

Nanoindentation mapping
Nanoindentation mapping was performed using an FT-I04 Femto-Indenter (FemtoTools AG, Switzerland) with a diamond Berkovich indenter. A high-speed nanoindentation technique was employed to map local mechanical properties using continuous stiffness measurement (CSM) indentation testing [26,27], in displacement control. Each indentation was conducted in ≈1 s with an oscillation frequency of 140 Hz and an amplitude which was linearly increased with increasing depth from 1 to 2.5 nm. The total time for each indentation including repositioning was ≈2.5 s. Each indentation was performed to a specified depth of 190 nm, so that a spacing of 2 μm between indentations could be used while still ensuring an indentation depth/spacing ratio of ≈10 to avoid any significant interaction between neighboring indents [28]. Hardness and reduced modulus were measured as a function of depth for each location, and representative values for each parameter were taken by averaging values from depths >120 nm to minimize the influence of indentation size effects.

High-resolution neutron imaging calibration
High-resolution neutron radiography was performed at the POLDI beamline at the Swiss spallation neutron source SINQ. The setup allows for a spatial resolution of 2.7 μm per pixel, with a resolving power estimated to be 9.6 μm [12]. The neutron flux at the beamline allows for an effective imaging window of approximately 500 pixels × 1200 pixels, corresponding to a sample size in the image plane of 1.35 mm × 3.24 mm.
The neutron absorption in the sample can be modelled using the Beer-Lambert Law, for which there is a logarithmic dependence between the transmission, T, through a substance and the product of the absorption coefficient of that substance ∑ total (x, y), and the distance the neutron travels through the material (l), as described by the following equation: Where the transmission T is defined as the ratio between the incident beam intensity in a specific location I 0 (x, y) and the transmitted beam intensity in the respective trajectory after having passed the sample material I(x, y). With all the variables but the local hydrogen concentration constant, the transmission is dependent only on the hydrogen-free transmission (T 0 ) and the hydrogen concentration (H) (multiplied by a sensibility factor (H sens ): Experimentally, the hydrogen sensibility has been calculated in Zircaloy-2 from the transmission values obtained from quenched and asreceived samples, where the hydrogen concentration has been assumed to be uniformly distributed in the radius and cladding wall thickness. Ring samples of axial length of 4.5 mm were selected from the hydrogenenriched rods, as illustrated in Fig. 3. The samples were subsequently subjected to homogenization heat treatment for 8 h at 400 • C followed by rapid cooling by quenching in water. The samples planar surfaces were then ground with 1000 Grit paper to remove the oxide layer formed during heat treatment with minimal influence on the final sample thickness. The samples were then subjected to neutron imaging, and the neutron transmission values obtained during the campaign were referenced to the hydrogen content measured by HVE.The minor differences in composition between the LTP, LTP2 and LK3/L cladding tube were neglected, as well as the difference in the hydrogen-free transmission in the liner region. The calibration curve obtained for the presented experiment follows Eq. (3): where 0.8949 is the hydrogen-free transmission value, and − 9.325 × 10 − 5 is the hydrogen sensitivity. When compared to the relationship obtained in a similar setup by Gong et al. for DX-D4 [12], a small difference in the hydrogen-free transmission can be noted. This difference is associated with the different alloy compositions between Zircaloy-2 and Zircaloy-4, and the respective liner composition. The hydrogen sensibility, which determines the slope of the calibration curve, can be seen to be very similar as it is solely determined by the hydrogen influence (Fig. 4). Given the consistency between the hydrogen sensibility observed in our experiment and that reported by Gong et al. [12], it is possible to conclude that the obtained calibration is sufficient for the intended purposes. Therefore, the obtained relationship between neutron transmission and hydrogen concentration can be utilized to assess the high hydrogen concentrations measured in the liner region.

Hydrogen distribution in cladding tubes by neutron imaging
High-resolution neutron images of unlined LTP2, Fe-containing liner LTP and Sn-containing liner LK3/L cladding samples, charged with a target concentration of 200 wppm in H, are shown in Fig. 5, Fig. 6, and  Fig. 4 was used. All graphs display the neutron transmission on the ordinate axis from the evaluated neutron image area and its corresponding hydrogen concentration is plotted on the abscissa. Neutron transmission curves were horizontally aligned for all samples by the inner edge. A contact threshold for each sample was considered to be a H-free transmission value of 0.8949. The first 30-80 μm drop in neutron transmission at the sample edges is a consequence of neutron refraction from the sample surfaces, and is primarily influenced by small variations in the rotation angle of the sample edge with respect to the beam [29,30]. This edge effect is particularly relevant in the samples with liners, as it extends into a considerable fraction of the liner and can affect the calculation of the hydrogen content.
As shown in Fig. 5, no significant changes are observed in the hydrogen content in the radial direction for the LTP2 clad tubes without liner. Transmission fluctuations corresponding to hydride clusters are clearly observable. The relatively good contrast means that the hydride clusters have a considerable depth within the sample, i.e. along the neutron traveling path, and the fluctuations become more visible in the slowly cooled sample. Hydrides tended to be longer and thicker when slower cooling rates were applied. The hydrogen content was homogeneously distributed in the case of the quenched samples without liner, and this was also confirmed by the neutron transmission plot - Fig. 5(b). Strong accumulation of hydride precipitates was observed at the interface between the liner and substrate in all cladding with liner, as shown in Figs. 6 and 7. The transmission profiles confirm that a strong hydrogen diffusion from the bulk to the liner interface is observed for all cladding materials under different cooling rates. Both cladding tubes with liners, LTP and LK3/L, showed the same tendency of hydride precipitation at the interface, leading to a local hydrogen concentration exceeding 2000 wppm H in the interfacial region.
Hydrides nucleate preferentially at the interface between the liner and the substrate. However, at lower cooling rates, an increased fraction of the hydrides appear to nucleate and grow in the internal liner region. In the samples cooled at 3 • C/h, the formation of a second hydrogen accumulation rim on the inner edge of the sample can also be speculated from the reduced transmission in the area where otherwise also the edge effect from neutron radiography images occurs. In any case, in the liner bulk persists a lower hydrogen concentration in comparison to the interface. Samples cooled at 30 • C/h present a characteristic hydrogendepleted zone that extends for about 200-300 μm from the liner into the substrate. The hydrogen-depleted zone is more pronounced in the LTP sample, where it extends for ~75% of the sample thickness, compared to the LK3/L sample, where it extends for ~50% of the sample thickness. The hydrogen-depleted zone extends to the entire thickness of all the samples when cooled at 3 • C/h. Samples quenched from the homogenization temperature of 400 • C present an increased hydrogen concentration at the liner/substrate interface. This accumulation is more pronounced in the LTP sample, Fig. 4. Calibration of the hydrogen concentration: Zircaloy-4 data were reported by Gong et al. [12]. Zircaloy-2 data (LTP2, LTP and LK3/L) were obtained by neutron transmission of the quenched samples (denominated 'Q' in the image above), and the hydrogen content measurements are reported in Table 3. where it reaches an apparent local concentration of 500 wppm, compared to the LK3/L sample, where it reaches a concentration of 400 wppm. The hydrogen concentration further in the bulk of the liner of the quenched samples goes back to lower values, close to those of the substrate. At the liner-air interface, the already mentioned edge effect occurs.
The hydrogen accumulation at the liner/substrate interface could be explained by an incomplete hydrides dissolution during the homogenization heat treatment (8 h at 400 • C) prior to quenching. To test this hypothesis, a sample of LTP cladding material enriched to 200 wppm H nominal has been subsequently subjected to quenching from 8 h holding temperatures at 370, 400, 430, and 460 • C. The resulting hydrogen distribution in the radial direction has been assessed by neutron radiography (Fig. 8).
This study shows how the homogenization treatment affects the dissolution of the hydrides. At temperature of 370 • C, all the hydrogen in the system is trapped at the liner-bulk interface. At increasing temperatures, a higher amount of hydrogen gets distributed into the sample, with complete uniform distribution achieved at temperatures higher than 430 • C.

Hydride distribution by optical microscopy and nanoindentation mapping
The hydride distributions were also investigated on selected samples by optical microscopy and nanoindentation mapping to confirm the presence of hydrides and their distribution along the cladding wall thickness. Differential interference contrast (DIC) micrographs (Fig. 9) of these samples show very similar hydride distributions to those observed by neutron radiography.
Nanoindentation mapping was performed over samples homologues to those investigated by neutron radiography/microscopy. Samples of  LTP and LTP2 were investigated in two conditions: as received and charged with hydrogen to 200 wppm, homogenized, and cooled at a rate of 30 • C/h. In the nanoindentation maps presented in Figs. 10 and 11, the state of the four different samples is clearly observable. The unlined, LTP2, as received sample shows mostly uniform substrate properties across the entire sample width, without edge effects noticed in the neutron radiography, with small variations corresponding to local anisotropy from individual grains. The as-received sample with a liner, LTP, shows similar properties in the substrate as the unlined sample in the majority of the cross section, but the presence of the liner is also clearly observed (Fig. 11) which was not observable in the neutron radiographs. The liner materials are observed to be considerably softer than the substrate material used for the bulk of the cladding tube. Liner materials presented an average hardness of ~1.5 GPa; relatively soft compared to ~2.5 GPa of the substrate material. The interface between the tube and the liner is clean with only minor variations from the majority phases in either hardness or reduced modulus at the boundary.
In the samples charged with hydrogen, the presence of hydride clusters elongated and oriented in the circumferential direction is clearly visible in the indentation property maps. These present a strong contrast to the substrate properties with a ~200% increase in hardness (H = 3.5 GPa) and a ~10% decrease in reduced modulus. The orientation and periodicity of the hydride clusters observed in the nanoindentation maps is in good agreement with the neutron radiographs.
In the LTP2 sample without a liner, the hydride clusters are distributed throughout the thickness of the cross-section. Whereas, in   Table 1. the LTP sample with a liner, the hydrides are concentrated at the linersubstrate interface, with some additional hydride clusters visible in the outer portion of the substrate beyond the hydrogen depleted zone. These outer clusters are also indicated in the neutron radiography profiles (Fig. 6).
As hardness and elastic modulus are interrelated properties, 2D histogram plots are utilized to display the statistical distributions of the obtained H and E values from the three different phases in the two different hydrided materials over the entire mapped regions simultaneously - Fig. 12.
Nanoindentation mapping allowed the identification of the hydrides as an additional phase to the substrate material and the liner material. The color of each pixel represents the number of indentations that are included within a range of H and E, which is defined as a 2D bin size. Values of only 1 indent are shown in light gray to minimize the visual impact of outliers, and higher indentation numbers are shown with a shaded gradient from darker red tones to yellow-white peaks at the highest values in each histogram. These are arbitrary units, depending on the number of total indentations performed and the bin size used.
The 2D histograms of reduced elastic modulus and hardness values offer a simple visual method to evaluate 'hot spots' in the indentation property space which statistically correspond to individual phases. As shown in previous work [31], these 'peaks' in the 2D histograms often take the form of elliptical, normal distributions which are elongated along the direction of the H/E ratio. These can be easily segmented using mixed Gaussian clustering algorithms to measure statistical average values and standard deviations. Values for each phase extracted using cluster analysis are given in Table 4.
In all samples, the Zr substrate phase appears as a typical example of one of these clusters with a well-defined "hot spot" with values which are consistent within the standard deviation between all four samples. In the hydrogen-charged LTP2 sample, the hydride phase appears as a lobe elongated from the Zr substrate cluster with its intensity diminishing at higher hardness. This suggests that the hydrides are relatively narrow and integrated within the substrate. In the lined LTP sample, the hydride phase is more discrete from the Zr substrate with its properties more uniformly distributed producing higher average hardness values. This reflects that it occurs as a mostly separate, concentrated phase at the interface between the substrate and liner. The liner phase is also easily segmented from the majority substrate phase as a cluster with lower hardness and slightly lower modulus than the substrate phase.

Discussion
The typical explanation for the hydrogen migration toward the liner region [12,32], relates to the difference in the diffusion of hydrogen and the terminal solid solubility for precipitation (TSSP) between substrate and liner material. The diffusion of hydrogen and terminal solid solubility, in fact, varies considerably with small difference in alloying elements, crystal structure, grain size, temperature, kinetics and local hydrogen concentration (e.g. [33][34][35][36][37][38][39]). The diffusion of hydrogen towards the liner could therefore be explained by a lower TSSP of the liner region with respect to the substrate. The TSSP-based model is able to predict the amount of hydrogen trapped in the liner region following controlled cooling. However, it fails to accurately predict the observed strong hydride accumulation at the liner/substrate interface. Difference in grain size might also account for some difference in chemical potential between liner and bulk material, however the influence of grain size on hydrogen diffusion in zirconium is minimal [39], suggesting that the grain size does not have a major influence over the observed The diffusion of hydrogen observed towards the liner interface from both sides, substrate and liner, can be explained by treating the interface between liner and bulk as a separate material with locally higher TSSP. The higher TSSP leads to a local increase of hydrogen and an increased precipitation of hydrides, as schematically represented in Fig. 13. The accumulation of hydrides at the liner interface indicates that in the bulk/ substrate interface there is a gradient of electrochemical potential, causing a locally higher availability of hydrogen in the interfacial area. This hypothesis is supported by the following observations: (i) During the initial hydrogen loading and continuous increase of hydrogen concentration in the cladding, followed by the 24 h homogenization at 400 • C and cooling to room temperature, hydrides form already at the liner/substrate interface, as indicated by the neutron imaging after quenching . (ii) The presence of residual hydrides at the liner/substrate interface enhances the growth of hydrides, supported by the diffusion of hydrogen from the substrate towards the liner. This effect can be seen by the increasing amount of hydrides directly at the interface with slower cooling. Thus, to model the hydrogen redistribution behavior of cladding material in presence of a liner, the liner/substrate interface can be considered as an additional layer with own properties. The reason of the different electrochemical potential of the interfacial region might be found in a higher amount of defects induced by the joining of the substrate and the liner. It is not known in how far any defects at the interlayer persist during the final heat treatment of the cladding which is recrystallized. If there is a higher lattice damage, this could lead to a higher density of trapping sites for hydrogen in the damaged areas, which would then have also consequences on the hydrides formation (e. g. [40]). Nanoindentation measurements of the cladding material in the hydrogen-free form (presented in Fig. 11a) show an increased hardness of the interlayer, which could be a sign of, for instance, a higher density of dislocations.  The interactions caused by the interlayer are summarized in Fig. 13. The driving force for a higher amount of hydrides in the interlayer is probably due to the a-priori higher hydrogen concentration (lower electrochemical potential for hydrogen), as result of a higher diffusion of hydrogen from substrate and liner into the interface. This diffusion leads to a local concentration increase of hydrogen and consequently precipitation at the interface starts. Diffusivity is maintained over larger time due to the higher TSSP of the interface, and precipitation in the direct vicinity of the interface happens due to the lower TSSP of the liner, as supported by the relatively reduced hydrides concentration inside the liner itself. With longer time, i.e. with lower cooling rate, there is a certain saturation of the interface with hydrides, and some of the hydrogen continues to diffuse further inwards the liner driven by the difference in TSSP between liner and substrate.
Finally, the temperature plateau of ~400 • C where the cooling started may have entailed a non-full dissolution of the hydrides which have formed and accumulated during the hydrogen loading process at the liner/substrate interface, thus providing nucleation seeds for hydrides formation at the liner interface during the experiment. However this hypothesis does not explain why the hydrogen accumulated at the interface during charging process prior to the homogenization process.

Conclusion
In this paper, hydrogen migration and precipitation in unirradiated Zircaloy-2 cladding at different cooling rates under the influence of a liner and the liner/substrate interface were investigated. The temperature plateau of 400 • C where the cooling started entailed a non-full dissolution of the hydrides which have formed and accumulated during the hydrogen loading process at the liner/substrate interface. The hydrogen/hydride concentration and the microstructure were analyzed with neutron radiography, optical microscopy, and nanoindentation maps. The combination of high resolution neutron radiography and nanoindentation mapping proved to be a very useful tool to identify and quantify the local concentration of hydrogen/hydrides along the clad wall.
Hydrogen distribution observed after controlled cooling experiments and quenching experiments suggests that the liner/substrate interface acts as a hydrogen sink at temperatures above the expected TSSP. The interfacial region between liner and clad bulk material may be considered as additional layer with lower electrochemical potential compared to the adjacent bulk materials. The lower electrochemical potential causes hydrogen trapping, leading to hydrogen accumulation at the interfacial area. Open questions remaining are the reason for the lower electrochemical potential at the interface, and whether the TSSP at the interface works synergistically in respect of amount of precipitated hydrides with the increased solubility. At least, at lower cooling rate the amount of hydrides reaching into the liner increases compared to faster cooling, so that a difference between TSSP of the liner and the substrate plays a certain role.
The results of this work constitute a step forward towards understanding the hydrogen behavior in zirconium-based clads in presence of an interfacial area and can be used to refine the modeling of hydrogen diffusion and precipitation in spent nuclear fuel claddings.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.