Oxidation of UN-U2N3/UO2 composites: an evaluation of UO2 as an oxidation barrier for the nitride phases

Composite fuels such as UN-UO2 are being considered to address the lower oxidation resistance of the UN fuel from a safety perspective for use in light water reactors, whilst improving the in-reactor behaviour of the more ubiquitous UO2 fuel. An innovative UN-UO2 accident tolerant fuel has recently been fabricated and studied: UN microspheres embedded in UO2 matrix. In the present study, detailed oxidative thermogravimetric investigations (TGA/DSC) of high-density UN-U2N3/UO2 composite fuels (91-97 %TD), as well as post oxidised microstructures obtained by SEM, are reported and analysed. Triplicate TGA measurements of each specimen were carried out at 5 K/min up to 973 K in a synthetic air atmosphere to assess their oxidation kinetics. The mass variation due to the oxidation reactions (%), the oxidation onset temperatures (OOTs), and the maximum reaction temperatures (MRTs) are also presented and discussed. The results show that all composites have similar post oxidised microstructures with mostly intergranular cracking and spalling. The oxidation resistance of the pellet with initially 10 wt% of UN microspheres is surprisingly better than the UO2 reference. Moreover, there is no significant difference in the OOT (~557 K) and MRT (~615 K) when 30 wt% or 50 wt% of embedded UN microspheres are used. Therefore, the findings in this article demonstrate that the UO2 matrix acts as a barrier to improve the oxidation resistance of the nitride phases at the beginning of life conditions.


Introduction
After the Fukushima Daiichi disaster in 2011, the nuclear community has strived to engineer a successor to the standard uranium dioxide (UO 2 )-Zr fuel-cladding system within light water reactors (LWRs). This severe accident scenario demonstrated that the standard fuel system degrades rapidly in such extreme conditions [1]. From that point on, many studies have been carried out focusing on fabricating accident tolerant fuel (ATF) materials that can withstand accident conditions for a longer time. These new ATF materials have to maintain or enhance the fuel performance under normal and transient operating conditions, as well as during a potential design basis accident (DBA) and beyond-design basis accident (BDBA) [2].
Uranium nitride (UN) has been considered a promising ATF candidate to substitute UO 2 in LWRs mainly because of its higher uranium density, thermal conductivity, and similar melting point in comparison with UO 2 [3]. These improved properties would allow operating the reactor with a lower fuel centreline temperature, which provides the benefit of a higher margin for melting. Additionally, an improved fission density would enable higher burnup, larger power uprates, and longer fuel resident time [4][5][6] that are often labelled as advanced technology fuel attributes. However, the UN fuel has low oxidation resistance when in contact with the coolant water in the LWR system [7]. In such a situation, the nitride fuel readily reacts with the coolant and loses its structural integrity, resulting in fuel pellet oxidation, pulverisation, washout and relocation.
Composite fuel designs are considered to overcome the lower oxidation resistance of the UN fuel. The main idea of such composites is to combine the UN with a material that has a better oxidation resistance. This material would form a protective barrier around the UN and, therefore, prevent its oxidation. Several materials such as CrN and AlN [8], ZrN [9], U 3 Si 2 [10,11], and UO 2 [12][13][14] have been studied. Among these options, UO 2 shows great advantages since it is already used as a fuel in LWR, as well as it has a good oxidation resistance when in contact with the coolant water.
Many studies on UN degradation/oxidation have been performed since the 1960s [7,[15][16][17][18][19][20][21][22][23][24][25][26]. Nevertheless, there are few studies in the literature related to the degradation/oxidation of UN-UO 2 composite fuels [25,27], but none of them is related to UN microspheres embedded in UO2 matrix. Both previous studies used mixtures of UO 2 and UN powders to fabricate the composites. Watkins et al. [25] investigated the degradation behaviour of UN-UO 2 (5-10 wt%) sintered pellets (> 90 %TD) in high-pressure water. The samples were subjected to hydrothermal oxidation in a static autoclave filled with water at high temperatures (523-623 K) and pressures (up to 16.5 MPa), which are relevant to LWR operating conditions for a short duration. The authors concluded that grain boundary attack and spallation are the primary degradation mechanism in hydrothermal oxidation conditions. Shivprasad et al. [27] reported thermogravimetric analyses of UN-UO 2 (5-30 v%) in steam (62-83 %) during heating to 1273 K, and isotherm data for such composites at 623 K under 82 % steam during 12 h. In both thermogravimetric experiments, the addition of UO 2 delayed the oxidation onset temperature when compared to pure UN. So, the autoclave experiments have permitted the reproduction of an LWR-type environment, but have not provided useful information such as the oxidation onset temperature and the maximum reaction rate, as thermogravimetry has provided. Thus, both autoclave and thermogravimetric investigations are relevant and complementary.
An innovative UN-UO 2 accident tolerant fuel has recently been fabricated and studied by the authors of this work [28]: UN microspheres embedded in UO 2 matrix. Since this composite fuel is a new concept, there is no data regarding its oxidation resistance neither using thermogravimetric examination nor autoclave test. As a first approach, thermogravimetric investigations are essential to understand the material parameters which govern the oxidation reactions of this fuel.
In the present study, detailed thermogravimetric (TG) and differential scanning calorimetry (DSC) investigations of the oxidation reactions in a synthetic air atmosphere up to 973 K are reported and discussed. High-density UN-UO 2 (91-97 %TD) composites, within the limits for LWR application [29,30], were selected from our previous study [28] and are investigated here. This work also describes the influence of the amount (wt%) of each phase present in the sintered pellet, i.e. UO 2 , UN and α-U 2 N 3 , on the oxidation kinetics.

UN-UO 2 composite fuel fabrication
Uranium nitride (UN) microspheres were fabricated by the internal sol-gel process at Chalmers University of Technology [31,32]. The reported amounts of nitrogen, carbon and oxygen, as well as the associated errors in the triplicate measurements (given as one standard preparation, as well as the sintering procedures, can be found in our previous article [28]. The samples are identified in this paper as UN(X)-UO 2 (SC/FC), where X is the initial weight fraction (wt%) of UN microspheres in the composite, and SC/FC specifies which cooling rate was applied. Table 1 reports the sample identifications and the SPS parameters used in this study.

Oxidation experiments
The oxidation kinetics of the samples were assessed by thermogravimetric analyses (TGA) in a synthetic air flow of 20 mL/min, using a TGA Discovery equipment (TA Instruments) with a weight sensitivity of 0.1 µg. The samples were placed in a platinum pan without cover and heated up to 973 K at 5 K/min, held at this plateau for 10 min, and then cooled to room temperature. Each sintered pellet (~3.5 g, ~6 mm x 5 mm diameter x height) was cracked in small pieces, and about 45 mg were used in the TG experiments. The weight variation (%) was defined as the percentage due to the weight gain during the oxidation reactions. Also, oxidation onset temperature (OOT) was defined as the temperature at which 5 % of the total weight variation was observed [20], while the maximum reaction temperatures (MRTs) were obtained by analyses of the first derivatives of the weight variation curves (%/min). These considerations are suitable to adopt a consistent basis of comparison between each sample.
Three TG replicate measurements were performed for each sample to assess the standard deviations of all mean values of total weight variation (%), OOTs, MRTs, as well as the repeatability and reliability of the TG results. The standard deviations of the mean values were calculated by dividing the standard deviation of each variable (i.e. total weight variation (%), OOTs, MRTs) by the square root of the number of repeated measurements made [35], which is equal to three in our study. As-fabricated UO 2.13 powder and uncrushed UN microspheres were also oxidised to provide insights about their oxidation behaviours and data for comparison with UO 2 (SC) and UN (SC) samples.
Additionally, differential scanning calorimetry (DSC) analysis was performed in a Netzsch STA 449 F3 Jupiter equipment under a synthetic air flow of 40 mL/min.
Approximately 90 mg of UN(50)-UO 2 (FC) composite fuel was characterised by DSC to evaluate the thermodynamics involved during the oxidation of the three phases present in the sintered material, i.e. UO 2 , UN and α-U 2 N 3 . The material was also heated at 5 K/min up to 973 K, kept at this temperature for 10 min, and cooled to room temperature.

Characterisation
UN microspheres and UO 2.13 powder were characterised by scanning electron microscopy (SEM) and X-ray diffraction (XRD) to evaluate their morphologies and crystalline structures, respectively. The SEM used in all analyses was an SEM FEI XL30 with the INCA software. The external surfaces of both compounds were examined using the secondary electrons (SE) detector. Bulk morphologies of sintered UN microspheres and UO 2.13 powder, UN (SC) and UO 2 (SC) samples, respectively, were examined after standard metallographic preparation (grinding and polishing) of pellets that were hot mounted in a phenolic resin with carbon filler. Colloidal silica suspension (OP-U, 0.04 µm, Struers) was used to perform a final polishing (2 min) to enhance the contrast between the grains of UO 2 (SC).
The XRD scans of the raw materials were carried out in a Siemens D5000 diffractometer with Cu Kα radiation (Ni filter), 2θ ranging between 20º and 120º, with 2θ step of 0.02º and acquisition time of 9 s for each step. The UN microspheres were milled inside a glovebox filled with argon (< 250 ppm O 2 ) using a mortar and pestle, and then encapsulated in an air-tight XRD specimen holder (Bruker model A100B138-B141) to avoid any oxidation during the analysis.
After sintering, the external surfaces of the samples were ground to remove the residual graphite foil from the SPS experiments. XRD and elemental analyses had been performed before in sintered pellets and the results showed that the C contents were consistent with the initial materials [36]. Therefore, grinding the sintered pellet surface is enough to remove the bounded graphite paper from SPS.
A modified Archimedean method, with chloroform as the immersion medium, was used to measure the sintered densities of all samples [36]. Afterwards, the sintered pellets were cut longitudinally and milled to powder using a mortar and pestle in an argon-filled glovebox (< 250 ppm O 2 ) for XRD analyses, using the same equipment, special holder, and parameters. The theoretical densities (TD) were computed based on the weight fractions of each phase after SPS using the Rietveld method and the software MAUD [37]. Before all XRD measurements, the software MAUD were configured with the information from the Siemens D5000 diffractometer (e.g. wavelengths, goniometer radius). Additionally, an instrumental broadening determination (with yttrium oxide) was performed before calculating the lattice parameters and the weight fractions after sintering [38]. The considered TD reference values [39] for UO 2 , UN and α-U 2 N 3 were 10.96 g/cm 3 , 14.32 g/cm 3 and 11.24 g/cm 3 , respectively. SEM examinations were performed in the polished longitudinal cross sections, coated with carbon, using both secondary and backscattered electron (BSE) detectors.
SEM-BSE images were used to estimate the average grain sizes of the UO 2 phase in the composites using the software ImageJ [40,41]. The averages were computed based on 5 measurements per grain, considering 20 grains per image and 3 images per sample.
Brightness and contrast were adjusted to enhance the grain orientation contrasts. In total, 100 measurements per sample were collected and the average grain sizes, with 95 % confident intervals [35], were obtained from these values.
Post oxidised samples were examined using the same SEM equipment to analyse their surface morphologies. Energy-dispersive X-ray spectroscopy (EDS) was also used to identify the chemical elements present in the oxidised samples.   [42], with internal porosity that mainly contributes to the surface area of the material. The as-fabricated UN microspheres showed superficial and internal porosity with a density of ~52 %TD [28].

Phase and morphology of the raw materials and non-oxidised samples
The XRD patterns of the UO 2.13 powder and UO 2 (SC) sample, as well as the UN microspheres and UN (SC) sample, are presented in Fig. 2 (a) and Fig. 2 (b), respectively.
The computed lattice parameters were 0.5459 ± 0.0001 nm and 0.5475 ± 0.0001 nm (ref. 0.5470 ± 0.0001 nm [43]) for the UO 2.13 and UO 2 (SC) phases, respectively. The UO 2.13 data show broader peaks when compared to UO 2 (SC). This broadening can be attributed to peak overlap owing to the presence of secondary phases such as U 3 O 7 and U 3 O 8 [25]. Additionally, a lattice distortion as a result of excess oxygen in the UO 2 powder [43,44] can cause broadening. These higher oxide phases are intrinsically formed during the conversion process, where the UO 2.0 powder is pacified to UO 2+x [42,45]. Shifts in the position of the UO 2 peaks before and after sintering are observed in Fig. 2 (a). These shifts may be a result of the differences in the lattice parameters, as well as thermally induced residual stress in the UO 2 (SC) sample [46]. Previous studies [25,47] have also attributed the shifts to the variation in the lattice parameters caused by dissolved/excess of oxygen in the lattice, as well as to thermally induced residual stress.
Both samples had cubic phases with equivalent lattice parameters, but showed very low intensity peaks at (about) 28° (111) and 33° (200), which correspond to the UO 2 phase originated from oxygen impurity.  [25], as well as U 4 O 9 [43,44]. Low intensity peaks (*) are present in the nitrides at (about) 28° (111) and 33° (200), which correspond to the UO 2 phase originated from oxygen impurity. Fig. 3 shows SEM images of the non-oxidised sintered samples. UO 2 (SC) micrographs in Fig. 3 (a) reports a dense microstructure with small and rounded pores (< 2 µm) on the grain corners, characteristic of closed porosity at the end of the sintering process [49]. In contrast, the UN (SC) sample in Fig. 3  sample was still sintering at 1773 K after 3 min, as shown in our previous study [28]. This porosity in UN (SC), as a result of sintering coarse UN microspheres (when compared to UN powder [50]), was then originated from an incomplete pore rounding step during the last stage of sintering [49]. Both micrographs represent the sintered densities reported in Table 3: higher density for the oxide (97.5 %TD), and lower density for the nitride (83.8 %TD) pellets. The UN-UO 2 composite fuel microstructures described in Fig. 4 (a-d) show that a third phase, α-U 2 N 3 (sesquinitride), was formed during fabrication as a result of the interaction between UO 2 and UN [28]. Similar microstructural features were observed in a previous study of the reaction of UO 2 with carbon in the presence of N 2 at 1973 K [51].
Moreover, based on our previous article [28] and published data [12][13][14], the interaction between the UN and UO 2 does not depend on the sintering atmosphere and density/morphology/type of the UN phase. Yet, the interaction between the UN and UO 2 does not depend on the O/U ratio either. As demonstrated by [14], sintering UO 2.0 powder with dense UN pieces (125-1000 µm, 95.7-97.2 %TD) resulted in a second nitride phase. This behaviour may be due to the equilibrium phase of UO 2 , which describes that the oxide can be present as a hypo-stoichiometric phase (UO 2-x ) at the sintering temperature and low oxygen potential [52]. The amount of each phase in the sintered materials affected the oxidation kinetics of the samples (details are described in the upcoming sections).
Our previous study [28] showed that sintering the UN-UO 2 composites at low  microstructural features were observed near the pellet edges. A third phase indicated by the arrows, α-U 2 N 3 (sesquinitride), was formed in all UN-UO 2 composite fuels. The amount of UN microspheres and the cooling rates affected the amount and/or shape of the sesquinitride phase (more details can be found in [28]). Table 2 reports the average grain sizes of the UO 2 (SC) and UN (SC) samples, as well as of the UO 2 phase in the composites. The computed average grain size for UO 2 (SC) was 3.2 ± 0.2 µm, slightly different from a previous study (5.6 ± 1.6 µm) that used similar SPS parameters (1623 K, 40 MPa, 5 min) [47]. This difference may be due to an extra 2 min in the sintering time, since the authors found that the average grain size increases with the sintering time. UN (SC) sample's value was 3.1 ± 0.4 µm, lower than the result obtained by Johnson and Lopes [50]. The authors sintered UN powder by SPS at 1723 K (90 MPa, 3 min) and obtained an average grain size of 7.3 ± 0.2 µm. Since the expected sinterability of microsized powder is higher than coarse microspheres [49], this may be the reason why the authors found a larger average grain size. Regarding the composites, as a general rule, the higher the amount of nitride phases in the sample, the higher the average grain size of the UO 2 matrix.
Since N forms a solid solution with the UO 2 matrix, such as UO 2-x N x [28], the grain growth was enhanced with the increase of the UN initial amount [49]. On the other hand, by only changing the cooling rate, the average grain size did not differ significantly when the uncertainties are considered. Table 2 Average grain sizes of the sintered samples computed using ImageJ [40,41]  Composite fuels with sintered densities of 91-96 %TD were obtained using SPS in just 3 min of sintering time. From Table 3, the final α-U 2 N 3 /UO 2 wt% ratios for the samples UN (30) respectively. These results indicate that most of the initial N was retained in the sintered composites, as experimentally observed during the sintering of pure UN microspheres and a UN-UO 2 composite; no significant variation in the vacuum pressure at high temperatures (> 773 K) [28].

Oxidation behaviour of the raw materials and ATF composites
TGA measurements were carried out in triplicate to assess the uncertainties and reliability related to the weight variations (%), OOTs and MRTs. However, for the sake of better visualisation, only one result per sample is presented in the following figures in the text. All measurements are available in Appendix A, which shows good repeatability and reliability of the experiments. Therefore, the results obtained from the microanalyses using TGA can be used as a starting point to understand the macro behaviour of the samples.
Weight variation (%) measurements of the specimens oxidised in a synthetic air atmosphere up to 973 K are shown in Fig. 6, which indicates the differences between the nitride, composites, and UO 2 behaviours. The mass decreases observed at high temperatures (> 800 K), in both the nitrides and composites with 30 wt% and 50 wt% of UN microspheres, are also observed in previous studies [24,48]. According to these previous studies, the mass decreases can be interpreted as the release of nitrogen (as N 2 gas) from an intermediate oxynitride product that is formed during the oxidation of the nitride phases. We have not detected nitrogen in the SEM-EDS examinations; only oxygen and uranium as chemical elements were present in all the oxidised samples. Theoretical weight variations were obtained using the reactions described in Table 4, considering that the ultimate product in all experiments was U 3 O 8 and that the nitrogen exits the samples as N 2 gas [7,17,20,57]. These theoretical values were compared to the ones obtained from the TG results, with their deviations described in Table 5  The weight variation of the UO 2 (SC) sample (3.94 %) in Table 4, when compared to the theoretical value (3.95 %), indicates that the UO 2.13 powder reduced to stoichiometric UO 2 during sintering. Previous studies using spark plasma sintering [47] and conventional sintering [58][59][60]   Oxidation of α-U 2 N 3 present in the sintered samples.
x UO 2 + y UN + z U 2 N 3 + O 2 → U 3 O 8 + N 2 (4) * Overall oxidation reaction of the sintered samples. * x, y and z are related to the amounts of UO2, UN and α-U2N3, respectively, present in the sintered pellets ( Table 3).
The oxidation onset and maximum reaction temperatures, extracted from the weight variation curves (Fig. 7, Fig. 8, Fig. 9), are summarised in Table 5. The uncertainties related to both OOTs and MRTs are also described, showing a minimum variation of ± 1 K and a maximum of ± 9 K.   (Fig. 7 (a)) showed two oxidation steps: the first one is associated to a parabolic oxidation curve, while the second one is related to a sigmoid weigh variation curve. The first step corresponded to the formation of U 4 O 9 /U 3 O 7 on UO 2 powders, following a diffusion-controlled mechanism [61]. The results showed an OOT and MRT of 440 ± 3 K and 458 ± 2 K, respectively, and a weight gain of 0.86 % up to 523 K, which is in agreement with the formation of the U 3 O 7 /U 4 O 9 phases [57].
The second step, associated with a sigmoid curve, is interpreted as the oxidation of the U 3 O 7 /U 4 O 9 phases into U 3 O 8 by a nucleation and growth mechanism [62]. OOT and MRT of this step were 623 ± 3 K and 631 ± 3 K, respectively, with a total weight variation of 2.47 ± 1 %, which is in agreement with a previous TG analysis [63].  Table 5). In the UO 2 (SC) case, the oxidation followed a sigmoid curve in one major step: from UO 2 to U 3 O 8 .
The oxidation started with incubation time, at which a fine layer of U 3 O 7 was formed on the external surface, followed by a parabolic kinetic (~623 -680 K), and then a linear kinetic (~680 -750 K) until the complete oxidation into U 3 O 8 [57,61]. The OOT (570 ± 5 K) and MRT (704 ± 2 K) for UO 2 (SC) were higher than the UO 2.13 powder, since the oxygen chemisorption and the diffusion-controlled formation of U 3 O 8 are both proportional to the surface area [57,64,65]. Thus, the higher the surface area, or the smaller the particle size, the faster the oxidation process.
UN microspheres and UN (SC) weight variation curves are presented in Fig. 7 (c) and porous, which provides more surface area to be attacked during oxidation. Our ongoing work is focused on optimising the UN-UO 2 composite by using denser (80-95 wt%, based on previous studies [66][67][68]) and coated UN microspheres to avoid the interaction between UN and UO 2 , as well as to delay the oxidation onset temperature of the UN microspheres. The amount of UN microspheres affected the oxidation kinetics of the composite fuels, as reported in Fig. 8. For the same cooling rate, e.g. fast cooling (FC), the higher the UN content in the sintered composites, the higher the total weight variation. Additionally, there was a sesquinitride phase in the sintered composites, α-U 2 N 3 , that formed during fabrication as a result of the interaction between UN and UO 2 (Fig. 4, Fig. 5, [28]). The oxidation of this sesquinitride also contributes to increasing the total weight variations. As reported in Table 3, the total amount of UN + α-U 2 N 3 in UN(10)-UO 2 (FC), UN(30)-UO 2 (FC) and UN(50)-UO 2 (FC) were 5.0 wt%, 26.6 wt% and 46.2 wt%, respectively.
Regarding the grain sizes of the UO 2 phase of the composites (Table 2), the higher the initial weight fraction of UN microspheres, the larger the average oxide grain sizes in the final pellets. The initial oxidation rates, up to (about) 50 % yield, were faster when the matrix grain sizes increased. Quémard et al. [69] found that the oxidation rate up to ~ 50 % yield increased when the grain sizes diminished. This different behaviour might be because they were analysing pure UO 2 sample, as well as owing to additional complexity of the UN-UO 2 system: sesquinitride precipitates along and inside the UO 2 grain and UN embedded microspheres. Therefore, the initial oxidation rates in Fig. 8 depended not only on the average oxide grain sizes, but also on the nitride amounts in the sintered samples, which seemed to impact much more the rates than the grain sizes themselves. Fig. 8 also reports the influence of the initial amount of microspheres on the OOTs and MRTs. The OOT of UN(10)-UO 2 (SC) (593 ± 6K) was surprisingly 23 K higher than the reference UO 2 (SC). Additionally, the composite showed a slightly higher MRT and lower oxidation rate at the maximum point, indicating that the oxidation reaction was more moderate in the composite than in the UO 2 (SC) reference. A previous study [23] proposed that the formation of U-N-O compounds on the surface of nitrided uranium can resist the diffusion of oxygen ions into the inner layer. Also, Lu et al. [19] showed that the corrosion resistance of the U 2 N 3 phase is better than the U or UN phase structure films. Thus, the complexity related to the U-O-N system seems to retard the oxygen diffusion through the sample in some cases. Increasing the initial amount of UN microspheres from 10 wt% to 30 wt% and 50 wt% have not affected the OOTs. Also, these composites had similar OOTs when compared with the UO 2 (SC) reference. These results seemed to demonstrate that the UO 2 matrix provided a barrier against oxidation of the nitride phases. However, the reaction occurred approximately three times faster in UN(50)-UO 2 (FC) (0.84 ± 0.14 %/min) than in UN(30)-UO 2 (FC) (0.29 ± 0.01 %/min) at the maximum rates, since the amount of nitrides was the highest in UN(50)-UO 2 (FC). The influence of the cooling rates on the oxidation kinetics is reported in Fig. 9. The faster the cooling, the higher the nitride amounts in the sintered composite and, consequently, the greater the total weight variation. Samples UN(30)-UO 2 (SC) and UN(30)-UO 2 (FC) had a total amount of nitrides of 20.2 wt% and 26.6 wt%, respectively (Table 3). Concerning the OOTs reported in Fig. 9 (a), the cooling rate affected neither the chemisorption of oxygen onto the composites, nor the diffusion-controlled formation of U 3 O 8 during the early oxidation stage (< 523 K) [57,64,65]. Thus, the OOTs for the composites were quite similar to the UO 2 (SC) reference, when the uncertainties are considered, demonstrating that the UO 2 matrix retard the oxidation of the nitride phases. Nevertheless, the MRTs were affected by the cooling rate ( Fig. 9 (b)). In the case of UN(30)-UO 2 (FC), the oxidation reaction reached a maximum of 0.29 ± 0.01 %/min at 618 ± 3 K, while the slowly cooled sample showed a slightly lower reaction rate (0.26 ± 0.01 %/min) at a higher temperature (641 ± 4 K).
Therefore, the cooling rate affected directly the amount of UN + α-U 2 N 3 in the composites and, consequently, the oxidation kinetics: the faster the cooling, the higher the amount of nitrides and the reaction rate, and the lower the MRT. From the DSC data plotted in Fig. 10, two linear kinetic regions associated with exothermic peaks are observed: first, between 573 K and 650 K; and second, between 660 K and 685 K. At these regions, a proposed mechanism for the composite oxidation is that the reaction occurred via UO 2 grain boundary attack at fresh UO 2 surfaces that formed during cracking and nucleation/growth of U 3 O 8 after an incubation time [57,65,69,70].
Simultaneously to the grain boundary attack, the α-U 2 N 3 phase oxidised since it was present along and inside the UO 2 grains (Fig. 4). With the extent of reaction, the embedded UN microspheres were also reached by oxygen and resulted in a strong exothermic DSC peak at 673 K, which is characteristic of a highly exothermic nitride phase oxidation [15]. Therefore, the UO 2 addition was beneficial to retard and smooth the oxidation of the nitride phases.  not changed after its oxidation, which is known to have a complex kinetic above 473 K owing to a simultaneous formation of U 3 O 7 and U 3 O 8 [71]. Since the powder had an initial surface area of 5.33 m 2 /g, with most of it coming from internal porosity( Fig. 1 (a)), a net increase of ~36 % in volume due to its oxidation to U 3 O 8 [70] might have been compensated by porosity and, then, avoided macro-cracking. Conversely, the oxidised UO 2 (SC) sample showed a general surface roughening and spalling after oxidation, as illustrated by SEM-SE images of a fragment in Fig. 11. As previously presented in [57,61] 11 shows that macro-cracks were mostly formed at the UO 2 grain boundaries, with some micro-cracks inside the grains as well. Additionally, there are small fragments in the image that could have been generated during transportation, since the cohesive forces in the oxidised UO 2 samples are weak and may crack into powder when weak stress is applied [69].

Post oxidation morphology
UN microspheres and UN (SC) samples showed similar oxidation kinetics ( Fig. 7 (c,d)) and post oxidised microstructures (Fig. 11). As a general behaviour, the growth mechanism of the U 3    A previous article reports in-situ oxidation of UO 2 fuel pellet at 603 K and P O 2 = 265 Pa up to 5.5 h [69]. The authors demonstrated that two types of cracks occurred during oxidation: first, macro-cracks occurred at the grain boundaries after 30 min of isothermal oxidation, followed by micro-cracking at the cracked surfaces that enhanced spallation. They also showed that the sample was still oxidising when the pellet was completely cracked with a constant weight variation, i.e. no more reactive surfaces was created. At this final stage, nucleation and growth of U 3 O 8 proceeded until the total consumption of the UO 2 phase.

Conclusions
This article is presenting, for the first time, thermogravimetric experiments in triplicates concerning the oxidation resistance of high-density (91-97 %TD) UN-U 2 N 3 /UO 2 composite fuels in a synthetic air atmosphere bellow 973 K. The results show that the oxidation resistance of the composite with initially 10 wt% of UN microspheres is surprisingly better than the UO 2 reference. The OOT of UN(10)-UO 2 (SC) (593 ± 6 K) is about 23 K higher than the reference UO 2 (SC). Additionally, this composite shows a slightly higher MRT and lower reaction rate at the maximum rate, indicating that the oxidation reaction is smoother in the composite than in the UO 2 (SC). Additionally, there is no significant difference in the OOTs (~560 K) and MRTs (~613 K) when using 30 wt% or 50 wt% of embedded UN microspheres. These behaviours are a direct result of the improved oxidation resistance provided by the UO 2 phase, which acts as a protective barrier for the nitride phases. Nevertheless, the oxidation reaction occurs approximately three times faster in UN(50)-UO 2 (FC) (0.84 ± 0.14 %/min) than in UN (30) At these regions, a proposed mechanism is that the oxidation occurs via UO 2 grain boundary attack at freshly exposed UO 2 surfaces that are formed during cracking and nucleation/growth of the U 3 O 8 phase. During the grain boundary attacks, it is suggested that the α-U 2 N 3 phase oxidise since it is present along and inside the UO 2 grains. With the extent of reaction, the UN phase is also reached by oxygen and results in a strong exothermic DSC peak at 673 K.
The oxidised composites have similar morphologies, showing a corroded surface with mostly intergranular cracking and spalling as a result of the oxidation reactions. Furthermore, the nitrides and UO 2 (SC) microstructures have a general roughening and spalling with intragranular and some intergranular cracks. Conversely, the overall morphology of the UO 2.13 powder does not change after its oxidation.
The findings in this study demonstrate that the UO 2 matrix acts as a barrier to improve the oxidation resistance of the nitride phases as they would exist at the beginning of life conditions. Therefore, the use of the UO 2 fuel to protect the UN microspheres is promising and requires future developments/tests to be used as an accident tolerant fuel in LWR.

Appendix
All TGA experiments were performed in triplicate. The results for the raw materials, as well as for the UO 2 (SC) and UN (SC) samples, are plotted in Fig. A1. The coloured/bolded graphs are the data used in the article, and the grey curves indicate the replicates.