Contribution of energetically reactive surface features to the dissolution of CeO2 and ThO2 analogues for spent nuclear fuel microstructures.

In the safety case for the geological disposal of nuclear waste, the release of radioactivity from the 2 repository is controlled by the dissolution of the spent fuel in groundwater. There remain several 3 uncertainties associated with understanding spent fuel dissolution, including the contribution of 4 energetically reactive surface sites to the dissolution rate. In this study, we investigate how surface features 5 influence the dissolution rate of synthesised CeO 2 and ThO 2 , spent nuclear fuel analogues which 6 approximate as closely as possible the mineral structure characteristics of fuel-grade UO 2 but are not 7 sensitive to changes in oxidation state of the cation. The morphology of grain boundaries (natural features) 8 and surface facets (specimen preparation-induced features) were investigated during dissolution. The 9 effects of surface polishing on dissolution rate was also investigated. We show that preferential dissolution 10 occurs at grain boundaries, resulting in grain boundary decohesion and enhanced dissolution rates. A 11 strong crystallographic control was exerted, with high misorientation angle grain boundaries retreating 12 more rapidly than those with low misorientation angles, which may be due to the accommodation of 13 defects in the grain boundary structure. The data from these simplified analogue systems support the 14 hypothesis that grain boundaries play a role in the so-called "instant release fraction" of spent fuel, and 15 should be carefully considered, in conjunction with other chemical effects, in safety performance 16 assessements for the geological disposal of spent fuel. Surface facets formed during the sample annealing 17 process also exhibited a strong crystallographic control and were found to dissolve rapidly on initial contact 18 with dissolution medium. Defects and strain induced during sample polishing caused an overestimation of the dissolution rate, by up to 3 orders of magnitude.

1. Introduction 1 7 of Th was performed using high resolution ICP-MS (Element 2, ThermoScientific). Prior to analyses, all 1 samples were L HNO3 to keep all elements dissolved in solution. Solution 2 data are expressed as the normalised elemental leaching NL(Ce, Th) (g m -2 ) according to: 3 4 (Eqn. 1) 5 6 where mCe, Th is the total amount of Ce or Th released into solution and S/V is the surface area to volume 7 ratio. The normalised element leaching rate RL(Ce, Th) (g m -2 d -1 ) is determined by: The aqueous dissolution rate data derived from all experiments are given in Table 1. The dissolution rate  17 for CeO2 dissolved at 150°C was (6.34 ± 0.1) x 10 -3 g m -2 d -1 ; data are shown in Figure 1a. The initial 18 dissolution rates, between 0 to 7 days, appeared to be more rapid than between 7 and 35 days (Fig. 1a), 19 which may be due to solution saturation effects. Under the same dissolution conditions, but at 90°C, the 20 dissolution rate was significantly lower, at (7.26 ± 0.2) x 10 -5 g m -2 d -1 , but the same trend was found, with 21 an initial, rapid dissolution followed by a slower dissolution from 7 to 35 days (Fig. 1b, Table 1). Dissolution 22 of ThO2 exhibited very different behaviour to that of CeO2, as shown in Figure 1c. At 90°C, the dissolution 23 was initially rapid (between 0 and 7 days, Fig. 1c) at rate of (6.71 ± 0.5) x 10 -5 g m -2 d -1 , but subsequently, 24 the dissolution rate significantly decreased, giving an overall rate between 0 and 28 days of (5.23 ± 0.1) x 25 10 -7 g m -2 d -1 (Table 1). Rapid dissolution could result from the release of material from high energy surface 26 sites at the ThO2 surface, leading to saturation of the solution with respect to Th, to a point at which 27 amorphous ThOx(OH)y·H2O or even polynuclear Thx(OH)y species form. These amorphous phases are known 1 to be capable of re-crystallising; for example, Rai et al. 21 showed that ThO2(am) converted to crystalline 2 ThO2(cr) upon heating at 90 °C in acidic solution. Therefore, crystalline ThO2 may be able to form a 3 protective layer on the surface of ThO2, providing a barrier to further dissolution and giving rise to the 4 dissolution behaviour observed in Figure 1c.  Grain boundaries are a well-known feature of SNF; they typically contain volatile fission products (e.g. Cs 3 and I) and alloy particles of Mo, Tc, Ru, Rh and Pd 22 . The presence of these highly radioactive elements 4 necessitates a careful understanding of how grain boundaries behave during dissolution, and an evaluation 5 of their contribution to the dissolution rate of spent fuel. Figure 2 shows the typical microstructure 6 obtained for CeO2; annealing conditions were optimised to give grain 7 The contribution of grain boundary dissolution to the overall dissolution rate of CeO2 and ThO2 was 14 investigated in 0.01M HNO3 at 90°C, and also at 150°C for CeO2. Figure 3 shows VSI images of CeO2 before 15 dissolution (Fig. 3a) and after 3 and 7 days of dissolution at 150 °C (Figs. 3b and 3c, respectively). It was 16 found that after 3 days, dissolution was focused at grain boundaries and the pores between grains (Fig 3b). 17 1 surface of the grains became rough and pitted (Fig. 3c). Some of the grains dissolved at different rates, as 2 evidenced by the height contrast in different grains (Fig. 3b-c). In their analysis of CaF2 dissolution 3 (isostructural to CeO2, ThO2 and UO2) Godinho et al. 23 showed that the measured retreat rates of CaF2 4 grains depended upon the crystallographic orientation of the exposed planes. They concluded that the 5 {111} plane is the most stable and dissolved most slowly, while the {112} plane was the least stable, 6 dissolving up to 33 times faster than {111}. First principles calculations have shown that the surface stability 7 of CeO2, from the most to the least stable plane is in the order of {111} > {110} > {100} 24,25 , although it 8 should be noted that the {100} plane in such calculations is modelled and not real, due to the difficulties 9 associated with modelling the dipolar {100} plane. The results presented here are in agreement with these 10 findings; EBSD analysis of the grains in Fig. 3c showed that the most stable grain had a (111) surface.  After 7 days, grains and grain boundaries were clearly visible (Fig. 4b), and after 14 days entire grains 10 appeared to become separated from one another (Fig. 4c). After 21 days of dissolution, the grains 11 F 3d. It 12 is surprising that the material appeared to retain cohesion although the grains appeared to reduce in size 13 ( Fig. 4d); close inspection reveals that the grains were fractured between pores, giving rise to apparently 14 smaller grains. These data confirm that grain boundary dissolution in CeO2 is extensive, and that grain 15 boundaries may act as conduits for solution ingress, leading to dissolution and fracturing between pores. 16 This process is expected to contribute substantially to the overall dissolution rate. Analysis of the CeO2 sample used to provide solution data at 90°C, shown in Fig. 1b, was performed using 6 AFM and EBSD. Figure 5 shows the boundaries between several grains of different crystallographic 7 orientation, including grain boundary 'A' between surfaces of (025) and (001), and grain boundary 'B' 8 between surfaces of (001) and (356). EBSD analysis of these boundaries gave mean misorientation angles of 9 36.01° and 59.84°, respectively ( Table 2). The dissolution of these boundaries was monitored over a period 10 of 7 days (after which the surface became too rough to accurately measure) with reference to an inert 11 surface mask of constant height. The mean surface retreat rates were measured as 0.001 nm d -1 , 0. suggesting the removal of material from within. Furthermore, the dissolution was greatest for the high 12 misorientation angle boundary, compared to the low misorientation angle grain boundary, giving grain 13 boundary retreat rates of 0.017 µm d -1 and 0.014 µm d -1 , respectively (Table 2). After 7 days of dissolution, 14 grain boundaries appeared to become shallower as a result of enhanced grain surface retreat at this time, 15 especially for grain boundary B where the (356) surface dissolved very rapidly (Fig. 5). In summary, when 16 CeO2 samples were contacted with the dissolution medium, a rapid loss of material from grain boundaries 17 occurred, which is in agreement with the enhanced release of Ce into solution during this time (Fig. 1a). 18 Subsequently, surface retreat rates increased and the surface, or matrix, dissolution became the dominant 19 dissolution mechanism. Comparison to the aqueous Ce concentrations in Figure 1a shows that the 20 dissolution was less rapid after 7 days, confirming that grain boundary dissolution contributes significantly 1 to the initial dissolution rate, while surface controlled dissolution leads to slower dissolution rates. 2 3 Similar experiments were conducted to monitor the dissolution of ThO2 grain boundaries as a function of 4 crystallographic orientation and grain boundary misorientation ( Table 2). The dissolution behaviour of ThO2 5 grain boundaries at 90°C in 0.01M HNO3 was comparable to that of CeO2, whereby grain boundaries 6 preferentially dissolved and boundaries with high misorientation angles retreated more rapidly than those 7 with low misorientation angles. For example, a grain boundary between two grains with (103) and (506)  8 surfaces had a misorientation angle of 23.91° and a retreat rate of 0.007 µm d -1 , while another grain 9 boundary formed between grains with (416) and (506) surfaces with a mean misorientation angle of 56.05° 10 gave a retreat rate of 0.357 µm d -1 , more than twice that of the lower misorientation angle grain boundary 11 (Table 2). It should be noted that after 7 days of dissolution it was no longer possible to measure grain 12 boundaries in ThO2 due to the presence of a surface layer, giving further evidence to the hypothesis 13 discussed above, that a dissolution rate drop after 7 days (Fig. 1c) is due to the formation of a protective 14 layer that results from the transformation of amorphous ThOx(OH)y·H2O to ThO2(cr) precipitates. 15 16 Table 2. Grain boundary depths of CeO2 (corresponding to Figure 5  Surface facets comprising flat terraces separated by inclined steps were found on annealed grains of CeO2 22 (Fig. 2). These features were not identified on ThO2 as a result of the greater surface roughness, which 23 resulted from difficulties in polishing. EBSD analysis of these grains was not possible due to multiple 24 orientations arising from the facetted surfaces, however alternative geometric measurements were used to 1 determine the orientation of the facet features. B 2 surface, the best combination of planes can be found according to Maldonado et al. 25 : 3 4 (Eqn. 3) 5 6 Where and are the normal vectors that define the planes. Godinho et al. 23 showed that a dissolution 7 surface is only made of the most stable planes, as the less stable ones are more prone to dissolution. 8 Therefore, if we assume only the most stable planes are present at the surface, this method allows the 9 unambiguous definition of the intersection of two distinct planes. These facets ranged in height from 1.50 nm to 6.61 nm and had flat terraces of 24.06 nm (or multiples 3 thereof). These {111}/{-111} facets were themselves facetted, giving rise to a "zig-zag" edge, as shown in 4 Fig. 7a. These "mini-facets" were found to be perpendicular to the {-111} plane, suggestive of the plane 5 {511}. The facet structures were also observed to extend into the grain boundaries (Fig. 7b). Figures 7b -e  6 show detailed AFM images of another grain, which exhibited a ridge and valley-like morphology, with 7 stacked concentric facets, building ridges at the grain edges (giving rise to the "tooth-shaped" grains shown 8 in AFM profiles in  Table 3. It is clear that the addition of just 0.01M HNO3 resulted in 5 a significant increase in height for most facets, compared to the height prior to dissolution. Facet height 6 increases ranged between 1.30 nm and 3.91 nm. Two facets were observed to become shallower (facets 2 7 and 5, Table 3). With increasing acidity between 0.1M and 3M HNO3, facet height change was variable, with 8 some facets showing little change (Table 3), while others decreased in height and others increased 9 (suggesting addition of material to facets). This variability suggests that these surface sites are highly 10 dynamic, changing in response to the reaction medium, but with little observable trend. However, it is clear 11 that upon initial immersion in the reaction medium, instantaneous dissolution of the facets occurred. 12 13  High energy surface sites may also be induced through specimen preparation, leading to over-estimated 6 laboratory dissolution rates. Surfaces of CeO2 7 at 150°C in 0.01M HNO3. Dissolution data were compared to those for annealed surfaces. Figure 8 shows 8 VSI images of the surface of a polished monolith after 3 and 14 days of dissolution (Fig 8a-b, respectively). 9 During initial dissolution, the surfaces were rough and pitted, and after 14 days exhibited areas of high and 10 low topography, indicating further dissolution had occurred. It was found that the dissolution rate was an 11 order of magnitude greater for the polished surface than for an annealed CeO2 surface under the same 12 conditions ( Table 2). At 90°C the effect was similar, however the dissolution rate of the polished surface 13 was found to be 3 orders of magnitude greater than an annealed surface, with rates of (7.40 ± 0.2) x 10 -14 In the results presented above, we have observed that surface features act as energetically reactive surface 3 sites that transform during dissolution. These features can be classified into two categories: i) natural 4 surface features, i.e. grain boundaries; and ii) specimen preparation-induced features, i.e. surface facets 5 and polishing defects. The dissolution of these features in CeO2 and ThO2 is discussed below, with 6 comparison to the dissolution behaviour of UO2. It should be noted that while the chemical and redox 7 characteristics of the analogue materials investigated here are simple compared to those of spent fuel, it is 8 possible to draw comparisons between CeO2, ThO2 and spent fuel that focus only on the physical and 9 structural properties that give rise to the dissolution behaviour observed. As such, in the discussion that 10 follows, only microstructural surface features that affect dissolution are described, with a cautious 11 interpretation for the overall behaviour of spent fuel during dissolution. 12 13

Grain boundary dissolution 14 15
The results presented in this investigation give evidence that UO2 and spent nuclear fuel analogue grain 16 boundaries undergo extensive transformation during dissolution; material is rapidly removed from grain 17 boundaries in both CeO2 and ThO2, corresponding to rapid initial dissolution rates. It is hypothesised that 18 grain boundaries are effective sinks for atomic defect high energy sites 28 ; the greater the number of 19 defects, the greater the proportion of high energy surface sites for dissolution. In CeO2, it has been shown 20 that increasing the density of oxygen vacancy defects results in an increase in the dissolution rate. For 21 example, Horlait et al. 29,30 showed that for every 10% of Ln 3+ cations added to CeO2, for which charge 22 compensation through the formation of Ce 3+ occurred, the dissolution rate increased by 1 order of 23 conductive, which was attributed to a high concentration of oxygen interstitial ions arising from hypo-2 stoichiometric UO2+x. This suggests that defects, and especially those that are concentrated within grain 3 boundaries, may play a key role in the dissolution of UO2 and its analogues. In spent fuel, grain boundaries 4 are expected to contain more defects than laboratory-prepared UO2 or UO2 analogues, primarily due to the 5 accumulation of fission gas bubbles and metallic precipitates 33 , therefore the effects of such high energy 6 surface sites might be expected to be greater. 7 8 We have observed that crystallographic orientation of the grains plays an important role in the dissolution 9 of the grain boundaries of spent nuclear fuel analogues; grain boundaries with a high misorientation angle 10 were found to dissolve more rapidly than those with a low misorientation angle in the current study. We 11 hypothesise that high misorientation grain boundaries have a higher concentration of defects (or defect 12 clusters) than grain boundaries with low misorientation angles. Indeed, simulations of UO2 grain boundaries 13 have shown that different types of defect structure were present in grain boundaries, depending on the 14 misorientation angle 34 ; in grain boundaries with lower misorientation angles edge dislocations were the 15 most common defect, while in higher misorientation angle boundaries oxygen point defects dominated the 16 grain boundary structure. 17 18 Assessment of the safety of geological disposal of spent nuclear fuel requires detailed information on the 19 rates and mechanism of release of radionuclides. This is hypothesised to occur in two main stages: ii) the 20 so-called "instant release fraction" (IRF), which represents a rapid release of long-lived and geochemically 21 mobile radionuclides (e.g. 129 I, 36 Cl, 135 Cs, 99 Tc); and ii) the slow, long-term release of radionuclides from the 22 UO2 matrix 35, 36 . The IRF is considered to come from two regions of the spent fuel; the gap between the 23 cladding and the fuel, and the grain boundaries. However, the IRF rates are still largely unknown, and the 24 contribution of grain boundary dissolution is not fully understood; in fact there is some controversy in the 25 literature as to whether grain boundaries make any significant contribution to the IRF 36, 37 . The results 26 presented in the current work, which demonstrate an "instant release fraction" of Ce and Th from the 27 spent fuel analogues, which is directly linked to grain boundary dissolution, support the hypothesis that 1 grain boundaries contribute to the IRF in spent fuel, and suggest that crystallographic direction of the grains 2 and the density of defects within the grain boundary may play a role. However, it should be noted that the 3 IRF of spent fuel is largely governed by the complicated chemical composition of the grain boundaries; 4 therefore the extent to which structural defects and grain boundary misorientation between adjacent 5 grains contributes to the IRF is unknown in comparison to the chemical effects. Our results are also in 6 agreement with the hypothesis that a second, slower stage of spent fuel dissolution occurs; in both CeO2 7 and ThO2, dissolution rates were lower after the initial release. In ThO2 it was apparent that this second 8 stage of dissolution was impeded by the formation of a protective layer. A similar effect was found in 9 laboratory UO2 dissolution experiments, where secondary U-bearing alteration products formed a 10 protective layer, preventing further dissolution 38 . On the basis of the data and arguments presented here, it 11 is evident that grain boundary dissolution in spent fuel and spent fuel analogues requires further detailed 12 chemical and physical analysis and that geological disposal safety performance assessment should carefully 13 consider the contribution of grain boundaries to the dissolution rate. 14 15 16

Dissolution of specimen preparation-induced features 17 18
We show that dissolution occurs at facet edges, especially during initial contact with dissolution medium. 19 With increasingly aggressive dissolution media, the dissolution of these features does not show a particular 20 trend, but instead appears to experience a dynamic process, whereby facet heights constantly change in 21 response to the dissolution medium. We have also shown that each facet contains some aspect of the {111} 22 plane, which is the most stable plane in fluorite-type structures, suggesting that crystallographic 23 orientation also plays an important role in facet formation. 24

25
It is thought that facets form by a dislocation growth mechanism during annealing, where spiral-like 26 structures form around threading dislocations likely induced through surface preparation (e.g. polishing). 27 Each dislocation produces a step as it emerges at the surface 39, 40 . O'Neil et al. 17 and He and Shoesmith 40 1 described similar surface morphologies to those identified in the current study in UO2. Current-sensing AFM 2 analysis showed that UO2 grains with facets were highly conducting, while smooth grains were not. Raman 3 and EDX investigation of these features revealed a high degree of non-stoichiometry in the UO2 of facetted 4 grains, attributed to the incorporation of interstitial oxygen atoms to locations in the {110} direction, 5 accompanied by shifts in vacant sites in the {111} direction. It was concluded that these non-stoichiometric, 6 defect-containing features would be more vulnerable to dissolution than defect-free surfaces. Further 7 investigations are currently underway to understand the relative stability of different facet orientations, 8 their degree of non-stoichiometry and defect structures. It is clear that these high energy surface sites play 9 a role in dissolution, but the evidence presented here suggests that the influence on dissolution rate is not 10 as significant as that of grain boundaries. It is important to note that these features are present as a result 11 of specimen preparation and annealing, thus are likely to contribute to the potential over-estimation of 12 dissolution rates in the laboratory. These features are not expected to be present in spent nuclear fuel. 13 14 We found that polished surfaces of spent nuclear fuel analogues gave dissolution rates of up to three 15 orders of magnitude greater than for annealed surfaces. Polishing has been shown to introduce strain and 16 defects into oxide material surfaces, giving rise to high surface energy. For example, diamond paste 17 polishing has been shown to result in the formation of dislocation loops, other lattice defects and also high 18 surface strain 41, 42 . Thermally annealing the surface of CeO2 allowed the strain and defects to be relaxed due 19 to recovery processes during heating, lowering the surface energy, and thus lowering the dissolution rate. It 20 is possible that defects induced during polishing may act as nucleation sites for the observed facet 21 structures formed during annealing. These results show that the introduction of defects to the surface 22 through polishing can lead to a significant increase in the observed dissolution rate, demonstrating the 23 importance of careful specimen preparation for dissolution rate determination. Dissolution experiments were conducted on non-redox sensitive, isostructural UO2 and SNF analogues, 3 CeO2 and ThO2 to investigate the contribution of energetically reactive surface sites to dissolution, and to 4 determine whether their presence may lead an over-estimation of dissolution rates. Grain boundaries, 5 which are part of the natural texture of SNF, were shown to significantly enhance the dissolution rate, 6 dissolving preferentially in the initial stages of dissolution, supporting hypotheses that grain boundaries 7 contribute to the instant release fraction of spent fuel. A strong crystallographic control was exerted, with 8 high misorientation grain boundaries dissolving more rapidly than those with low misorientation angles in 9 both CeO2 and ThO2. It was hypothesised that different crystallographic directions can accommodate 10 different densities of defects, explaining the observations found. Further investigation is required to 11 ascertain the extent to which structural defects and grain boundary misorientation between adjacent 12 grains contribute to the instant release fraction of SNF, in comparison to the chemical effects. 13

14
In addition to the natural high energy surface sites found in grain boundaries, energetically reactive sites 15 were also found to be formed through sample preparation. Facet structures formed during annealing, likely 16 nucleated on defects sites on polished surfaces, also exhibited a strong crystallographic control (all 17 combined some aspect of the {111} plane), and upon introduction to dissolution media, they experienced 18 instantaneous dissolution. Finally, the effect of surface polishing on the dissolution rate was found to 19 increase dissolution rates by up to three orders of magnitude. This results from induction of strain and 20 defects in the surface during the polishing process. We have shown that defects induced through sample 21 preparation contribute to the dissolution rate; the dissolution from facets is low, therefore not likely to 22 significantly over-estimate long-term dissolution rates, however sample polishing without any further 23 treatment is likely to cause over-estimation of dissolution rates.