Interface interactions in UN-X-UO 2 systems (X = V, Nb, Ta, Cr, Mo, W) by pressure-assisted diffusion experiments at 1773 K

UN-UO 2 composite fuel is considered an advanced technology fuel (ATF) option to overcome the low oxidation resistance of the UN fuel. However, the interaction between UO 2 and UN limits the performance of such composites. A possible way to avoid this interaction is to encapsulate the UN fuel with a material that has a high melting point, high thermal conductivity and reasonably low neutron cross-section. Amongst many candidates, refractory metals can be the ﬁrst option. In this study, detailed investigations in UN-X-UO 2 composite systems (X = V, Nb, Ta, Cr, Mo, W) were performed using SEM/FIB-EDS. The systems were heat-treated at 1773 K and 80 MPa for 10 min in vacuum using the spark plasma sintering method as a pressure-assisted diffusion apparatus. The results suggest that Mo and W are the most promising coating candidates to protect the UN fuel against interactions with UO 2 . Both metals are inert to N migration and preserve sharp interfaces with the nitride fuel. V, Nb, Ta and Cr strongly interact with UO 2 and UN and form their respective nitrides V 2 N/V 8 N, Nb 2 N, and Cr 2 N. The formation of TaN x was not observed but Ta reacts with UO 2 and forms two phases at the UO 2 -Ta interface (UTa 2 O 7 and Ta 2 O 5 ), while O from UO 2 + x diffuses throughout the Ta foil and oxidise the UN pellet via grain boundary attack. This oxidation mechanism also occurs at the V, Nb and Cr-UN interfaces. Our recent atomic scale modelling of the X-UN interfaces also proposes Mo and W as the optimal candidates. Therefore, these results validate the coating candidates for the UN fuel and may guide further experimental/modelling development in UN-X-UO 2 advanced technology fuel.


Introduction
Uranium nitride (UN) is a potential advanced technology fuel (ATF) candidate to substitute the standard UO 2 fuel in light water reactors (LWRs), mainly due to its higher uranium density and thermal conductivity [1] .However, the UN fuel has a low oxidation resistance when in contact with the coolant water, which is a major drawback for application in LWRs [2] .To overcome this disadvantage, UN-based composite fuels have been proposed to enhance the UN oxidation resistance by adding a compound that acts as a protective barrier against UN oxidation.Some materials such as ZrN [3] , U 3 Si 2 [4] and UO 2 [5][6][7][8] have already been studied.Amongst these materials, the UO 2 fuel can be considered a promising candidate due to its good oxidation resistance against water, as explored.Moreover, there is a need for a complementary ab initio approach to evaluate the interface reactions and diffusion behaviours in this composite system.Recently, our group evaluated the compatibility of UN with refractory metals (X = V, Nb, Ta, Cr, Mo, W) by density functional theory (DFT) calculations of interactions and diffusion at the X-UN interfaces [23] .The DFT calculations for the UO 2 -X interfaces are in progress and therefore not published yet.
In this article, we present detailed interface examinations in UN-X-UO 2 (X = V, Nb, Ta, Cr, Mo, W) composite systems using an innovative pressure-assisted diffusion experiment at 1773 K.In this developed setup, the spark plasma sintering (SPS) method was used as a pressure-assisted apparatus to guarantee contact between UO 2 -X and X-UN during the heat treatments and, therefore, to compensate for the swelling or shrinkage of the different phases during heating and cooling.Furthermore, this methodology can accurately simulate the conditions used during the fabrication of the composite fuels.Thus, 80 MPa was selected to provide the most severe sintering conditions used in our previous study on uncoated UN-UO 2 composite fuels [8] .
This work also presents a pioneering procedure to prepare the cross-sections for interface characterisation, which minimises cracks and spalls of the phases during cutting, grinding and polishing steps.The cross-sections containing the UO 2 -X and X-UN interfaces were analysed by scanning electron microscopy (SEM) coupled with a focused ion beam (FIB) and energy-dispersive Xray spectroscopy (EDS).
The findings in this study may be valuable to identify potential candidates to overcome the interaction between the UO 2 and UN fuels, and to suggest new insights on using the SPS method as a pressure-assisted diffusion apparatus for interface examinations at the same fabrication conditions.Furthermore, this work aims to provide experimental validation for the electronic structure modelling of the X-UN interfaces (X = V, Nb, Ta, Cr, Mo, W) already performed by our group and collaborators [23] and may encourage further experimental and modelling developments in such composite systems.

Raw materials
Uranium dioxide (UO 2 ) powder was provided by Westinghouse Electric Sweden AB, which had the following properties: O/U ratio of 2.13, fill density of 2.19 g/cm 3 , specific surface area (B.E.T.) of 5.33 m 2 /g, mean particle size of 20.20 μm, and 900 ppm of H 2 O.
Uranium nitride (UN) powder was fabricated from uranium metal by a hydriding-nitriding-denitriding method at KTH, which was based on previous fabrications [24] .Fig. 1 summarises the thermal profiles used in the hydriding, nitriding, and denitriding processes, as well as the atmospheres and heating/cooling rates of each step.The as-fabricated UN powder contained about 1550 ppm of oxygen and 5.4 wt% of nitrogen, which were measured by the inert fusion method [25] .

UN pellet fabrication
The UN pellets were fabricated by the SPS method at the National SPS Facility in Stockholm/Sweden.This technology is a fieldassisted sintering technique that uses low voltage and high current, combined with an applied pressure, to consolidate powders [ 26 , 27 ].The as-fabricated UN powder (3.0 g) was poured out in a graphite die (9.50 mm inner diameter) inside an argon-filled glovebox ( < 0.1 ppm O 2 ) connected to the SPS machine.The SPS chamber was depressurised to about 5 Pa to sinter the samples at 1973 K and 80 MPa, using the following thermal profile: heating at 100 K/min until 1773 K and then 50 K/min until 1973 K, held at this plateau for 10 min, and cooled to room temperature at 50 K/min.After sintering, the UN pellets were ground with SiC (grit 280) to remove the residual graphite from the surface.More de-Fig.1. Thermal profiles used in the hydriding (red), nitriding (green), and denitriding (blue) processes [24] .Argon was used during the denitriding process to avoid U 2 N 3 formation.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig. 2. Experimental setup for the pressure-assisted diffusion experiments.This system consisted of a polished UN pellet at the bottom of the graphite die, a polished metal disk in the middle, and finally the UO 2.13 powder on top.Both the UN and metal disks underwent polishing inside the SPS glove box ( < 0.1 ppm O 2 ) to minimise the formation of an oxide layer on the surfaces.tails about the SPS machine and sintering procedures can be found in our previous studies [ 8 , 25 ].

Pressure-assisted diffusion experiments
Fig. 2 illustrates the experimental setup for the pressureassisted diffusion experiments that consist of a UN pellet, a metal foil disk (W, Mo, Ta, Nb, V, or Cr), and the UO 2.13 powder.These materials were assembled inside an SPS graphite die (9.5 mm inner diameter).The UN pellet surfaces were ground and polished using consecutive diamond suspensions (9 μm, 3 μm, 1 μm) to ensure a flat surface of contact with the foils.A final polishing using a diamond paste (0.25 μm) was performed inside the SPS glove box ( < 0.1 ppm O 2 ) to minimise the formation of an oxide layer on the pellet surface.Likewise, both surfaces of the metal foil disks were polished inside the SPS glove box.Finally, 2.5 g of UO 2.13 powder was poured out on top of the metal foil disks, and the assembled system was then transferred to the SPS chamber to perform a heat treatment at 1773 K and 80 MPa for 10 min in vacuum ( ∼5 Pa).The samples were heated at 100 K/min until 1573 K and at 50 K/min until 1773 K, held at this plateau for 10 min, and then cooled at 50 K/min until room temperature.This thermal profile was adopted to use the most severe UN-UO 2 sintering scenario from our previous work [8] .
A pioneering methodology was used to prepare the UN-X-UO 2 interfaces for analysis.After the heat treatments, the samples were extracted from the SPS dies and directly hot mounted in a phenolic resin with carbon filler, without grinding the sample surface to remove the residual graphite paper from the SPS.Subsequently, the mounted samples were cut in half to expose the crosssections for grinding and polishing.We adopted this novel procedure, which is outlined in Fig. 3 , to minimise cracks and spalls at the interfaces UN-metal-UO 2 during sample preparation for analysis.

Characterisations
The nitride powders (U 2 N 3 + x and UN) and the UN pellet, as well as the UO 2.13 powder, were characterised by SEM and X-ray diffraction (XRD).The SEM used to examine these samples was an Fig. 3. Experimental steps from the heat-treated samples until the characterisations.The samples were mounted with the graphite foil from SPS to prevent cracks and spalls during cutting, grinding and polishing.FIB cross-sections were obtained in some specific samples to better examine the bulk interfaces.Fig. 4. SEM-SE images of the UO 2.13 , U 2 N 3 and UN powders, as well as a micrograph of a sintered UN pellet.The UO 2.13 powder showed a characteristic ex-AUC morphology [ 29 , 30 ], while the U 2 N 3 and UN powders presented a flaky morphology with particles varying from ∼3 μm to ∼30 μm.The UN pellet image shows a dense microstructure obtained by SPS (99.1%TD).
SEM FEI XL30 with the INCA software.Bulk morphology of sintered UN was examined after standard metallographic preparation (grinding and polishing) of a hot-mounted pellet.XRD scans of the nitrides and oxide powders were carried out in a Siemens D50 0 0 diffractometer with Cu K α radiation (Ni filter), 2 θ ranging between 25 º and 80 º, with 2 θ step of 0.02 º and acquisition time of 4 s for each step.The UN pellet was milled inside a glove box (Ar atmosphere) and encapsulated in an air-tight XRD specimen holder (Bruker model A100B138-B141) to minimise oxidation during the analysis.The lattice parameters were computed using the Rietveld refinement method and the software MAUD [28] .The morphologies and phases of the UO 2.13 , U 2 N 3 + x and UN powders presented in Figs. 4 and 5 , respectively, are in agreement with previous results in literature [29][30][31][32][33][34][35][36] .
The sintered densities (g/cm 3 ) of the UN pellets were measured using a modified Archimedean method with chloroform as the immersion medium [4] .The theoretical densities (TDs) were computed considering a TD reference for the UN as 14.32 g/cm 3 [37] .All UN pellets used in this work had sintered densities greater than 99.0%TD.
The polished cross-sections containing the interfaces were coated with carbon and examined by SEM coupled with an EDS detector.The SEM used for this purpose was a field emission gun SEM FEI Nova 200 with EDS detector Aztec Ultim (Oxford Instruments).FIB cross-sections were performed using a Zeiss Crossbeam 550 with an Aztec EDS system (Oxford Instruments).These FIB millings were performed at some selected interfaces of the regular cross-section samples.Fig. 3 illustrates the developed experimental steps from the SPS (after the diffusion experiments) to the FIB millings.The EDS measurements should be considered as qualitative for EDS chemical maps and line scans, and semi-quantitative for the area measurements (three regions per phase).Thus, the EDS results are useful to indicate and suggest what occurred in a sample during the heat treatment.

Phase and morphology of the materials
Fig. 4 reports the morphologies of the UO 2.13 , U 2 N 3 + x and UN powders, as well as a sintered UN pellet.From this point on in this manuscript, the U 2 N 3 + x phase is named U 2 N 3 for simplification reasons.The UO 2.13 powder morphology was characteristic of the industrial ammonium uranium carbonate (AUC) wet route [ 29 , 30 ].The nitride powder morphologies were flaky, with particle sizes varying from ∼3 μm to ∼30 μm.The SEM image of the UN pellet shows a high-density UN pellet (99.1%TD), which can be representative of all UN monoliths used in the experiments since all UN pellets were fabricated using the same UN powder source and SPS parameters.

Pressure-assisted experiments: interface interactions
as a lamellar-type structure, which is highlighted with arrows in Fig. 6 and visualised as the N-rich phase in the EDS map.This behaviour emphasises the need of coating the UN fuel before sintering.
As also described in our previous study [8] , oxygen from the UO 2.13 powder is available to interact with the other materials at high temperatures ( > 10 0 0 K).Additionally, it is reported that the H 2 O content present in UO 2.13 is eliminated at lower temperatures ( < 673 K) as water desorption and uranyl hydroxides decomposition.Thus, traces of H 2 O in the powder will not contribute as O and H 2 sources at the heat treatment temperature (1773 K) since H 2 O is eliminated at lower temperatures.
The regular cross-sections of all samples are summarised in Fig. 7 .The integrity of the interfaces between UO 2 and W, Ta, Nb, V and Cr were, somewhat, affected during the heat treatment.Moreover, there were interactions between the UN pellet and Ta, Nb, V and Cr metals.These results and the microstructures at the UO 2 -X and X-UN interfaces are presented and discussed in the subsequent sections.

UN-W-UO 2 system
Fig. 8 shows a cracked and spalled interface between UO 2 and W.These cracks might have formed during sample preparation or as a result of the thermodynamic relations between UO 2 , metal tungsten and (possible) tungsten oxides.A recent study reports EDS maps of U, W and O of a W-UO 2 cermet fabricated by SPS at 2123 K and 50 MPa in vacuum [20] .The interfaces between UO 2 and W in their study were also irregular/deformed.The EDS map of W in Fig. 8 shows that tungsten was present in the UO 2 phase around some crack regions (highlighted in the figure).Additionally, the O chemical map indicates the presence of oxygen in W near the interface.So, to have a better visualisation of this interface and to eliminate any effect of sample preparation (cutting, grinding, polishing) on the microstructure, FIB cross-sections were obtained at two different locations.These microstructures are reported in Fig. 9 .
A crack morphology similar to the one present in Fig. 8 is observed in the FIB cross-sections ( Fig. 9 ).This observation is relevant to eliminating the influence of sample preparation on crack formation.Additionally, the EDS maps of U, W and O indicate that these elements might be present around the cracks, as observed in the regular cross-section in Fig. 8 (EDS map of W).This result suggests that the thermodynamic relations between UO 2.13 and W are relevant at high temperatures.It seems that a ternary U-W-O phase was present during fabrication and left its residual shape (pores/crack) and trace compositions.Such assumption is in agreement with a previous study [40] that proposed a new fabrication process of UO 2 -W composite fuel, which consisted of a UO 2 pellet with a W channel on the UO 2 grain boundaries.The authors controlled the sintering atmosphere, based on thermodynamic calculations, to oxidise dispersed W particles to WO 3 above its melting point ( ∼1743 K).Then, liquid WO 3 diffused via grain boundary and, finally, reduced to W metal in H 2 at 1923 K.The authors report that about 1 at.% of W was observed in the UO 2 grains by EPMA analysis, indicating that W has a certain solubility in UO 2 during such liquid-phase sintering process.Likewise, our qualitative EDS chemical map of W in Figs. 8 and 9 might indicate some solubility of W in UO 2 .
The equilibrium oxygen potential curve for the UO 2.13 powder is higher than that of tungsten metal to WO 3 via WO 2 [ 40 , 41 ].This thermodynamic feature makes it possible to selectively oxidise the W metal foil to WO 3 in the presence of UO 2.13 powder.Moreover, we performed the heat treatments at 1773 K, which is slightly higher than the WO 3 melting point.What might have occurred is that W oxidised to WO 3 during heating and formed a U-W-O phase during cooling.Yang et al. [40] identified an U x WO 3 type cubic phase in XRD analysis, associating this phase to an A-cation deficient perovskite like U 0.07 WO 3 .Another recent study also found an Fig. 6.SEM-EDS examination of a regular cross-section of the reference sample, UN-UO 2 .As previously demonstrated [8] , the interaction between both fuels during fabrication forms a third phase, α-U 2 N 3 , in the UO 2 matrix.
anomalous third phase during W-UO 2 cermet fabrication at 1873 K by SPS (50 MPa, vacuum) [21] .The authors identified a cubic structure, U 0.1 WO 3 , based upon an electron diffraction pattern and justified its formation due to the availability of oxygen vacancies from UO 2 reduction in vacuum sintering.Yet, they observed uranium migration into the tungsten matrix in all experiments (from 1873 K to 2123 K), which was explained to be due to oxygen vacancies in UO 2 generated due to vacuum environment and the migration of uranium due to Fick's law of diffusion.
Since the vapour pressure of tungsten oxide is relatively high [ 40 , 42 ], some amount of material can evaporate during fabrication and leave the observed pores and cracks behind ( Figs. 8 and  9 ).This assumption is in agreement with a previously reported association relating pore formation and ternary U-W-O [40] .Additionally, the expansion (W to WO 3 ) and shrinkage (oxide decomposition) can cause cracks during the fabrication steps.Therefore, it seems that W may have oxidised to WO 3 during the heat treatment at 1773 K and, at the WO 3 near-melting temperature ( ∼1743 K), U dissolved in the tungsten oxide and then generated the residual cracks/pores during cooling due to the differences in density (swelling/shrinkage) and vaporisation of the phases.
A FIB cross-section of the W-UN interface in Fig. 10 shows a well-defined and sharp interface, but with some O contamination from the UN pellet surface.In this image, EDS chemical maps of U, W, O and N are reported, together with an EDS line measurement to qualitatively assess the chemical composition profiles across the interface.
The atomic compositions (at%) of U, W and N at the interface varied within a short range of distance ( < 0.5 μm), which indicates that W acted as a diffusional barrier against nitrogen diffusion.This observation is in agreement with our electronic structure calculations [23] .The DFT results indicate that additional energies from, for instance, high temperature or radiation, are required to form any tungsten nitride compound at the W-UN interface.Also, the possible interface phases (WN, W 2 N 3 , or WN 2 ) are chemically in equilibrium with both UN and W, and will not promote further interface reactions.Previous experimental studies on UN-W also showed good chemical stability between W and UN, even at high temperature (1773-2273 K) in Ar (with or without N 2 ) [14] or in high purity He [43] .However, at very high temperatures (2573-2993 K), W and UN interact and form a eutectic containing tungsten and liquid uranium as a result of UN dissociation [ 15 , 17 , 43 ].

UN-Mo-UO 2 system
The UN-Mo-UO 2 image in Fig. 7 shows a cracked UO 2 region (bulk), but with a sharp interface between UO 2 and Mo.This crack behaviour, which is also observed in previous studies [ 22 , 44-46 ], may be formed due to differences in the linear thermal expansion coefficients (e.g. at 1773 K) of UO 2 (14.6 × 10 −6 K −1 [47] ) and Mo (8.3 × 10 −6 K −1 [48] ), or due to differences in sintering shrinkage between the materials.At a higher magnification, as presented in the FIB cross-section in Fig. 11 , the interface is crack-free but irregular with a depth of less than 1 μm.The EDS qualitative line measurements in the figure indicate an abrupt change in the U and Mo concentration across the interface, within a gradient of about 0.5 μm.In a recent study, Cheng et al. [49] investigated the UO 2 -Mo interface of sintered composites (1473-1873 K) using high-resolution transmis-sion electron microscopy (HRTEM) and auger electron spectroscopy (AES).The authors reported no distinct reaction products at the interface, with the interdiffusion of U and Mo occurring to an extent of just several nanometres by Fick's law for solid diffusion.Therefore, a thin protective barrier of Mo in an UN-Mo-UO 2 composite fuel would be sufficient to avoid the interaction between UO 2 and UN.The FIB cross-section of the Mo-UN interface in Fig. 12 shows a microstructure without cracks or spall, but with oxygen contamination on the UN pellet surface.Similar to W-UN, the qualitative line measurements indicate a transition region at the interface with a width of less than 0.5 μm without forming a U-W-N ternary.These results are in good agreement with a recent study [50] that presents an innovative method to fabricate UN-Mo cermet fuels: sintered UN microspheres encased in a Mo matrix by the SPS method (1873-1973 K, 25-65 MPa, 10 min, vacuum).Also, the authors reported no interaction phase at the Mo-UN boundary even at 1973 K.However, they observed crack widths of 1-30 μm in the sample sintered at 1873 K, mostly due to gaps generated by the Mo particles only necking around the microsphere, and also crack widths less than 2 μm when sintering at 1973 K.
The linear thermal expansion coefficient of UN (9.6 × 10 −6 K −1 [51] ) is about 1.2 times greater than Mo (8.3 × 10 −6 K −1 [48] ) at 1773 K (our sintering temperature), which is lower than the ratio between UO 2 and Mo ( ∼1.8) at the same temperature.Thus, it is more probable to obtain cracks at the UO 2 -Mo interface than at Mo-UN, considering the same heating/cooling profile.As mentioned, Raftery et al. [50] observed cracks at the interfaces between UN and Mo.This behaviour may be due to their higher heating/cooling profile during SPS, 150 K/min, when compared to ours (heating: 100 K/min until 1573 K, and then 50 K/min until 1773 K; cooling: 50 K/min).Even at very high temperatures (e.g.20 0 0 K), the thermal expansion of UN is about 1.1 times the Mo value [ 48 , 51 ].So, the observed difference in the coefficients seems not to be an obstacle to fabricate a crack-free Mo-UN interface without interaction phases; it is possible to tune the sintering parameters and use e.g. a slower heating/cooling profile.It is important to mention that cracks might also form during sample preparation (cutting, grinding, polishing) and, in this case, a FIB crosssection is helpful to examine a fresh interface and eliminate that influence.
It was reported that UN and Mo do not interact even at temperatures up to 2473 K in the presence of N 2 (10%v and 67%v in argon).However, in vacuum, there is evidence of uranium liquidphase sintering due to UN dissociation into U metal and N 2 at high temperatures.Then, molybdenum can dissolve in U and form a ternary U-Mo-N with composition (about) 75 at.%Mo, 25 at.%U, and 1-2 at% N [14] .Moreover, liquid-phase sintering can also be observed at 2673 K when 58 wt% Mo in UN is used (eutectic composition, N 2 pressure of 0.85 atm) [15] .
The experimental investigations at the Mo-UN interface reported in this manuscript are in good agreement with our DFT calculations, which demonstrate that Mo is chemically compatible with UN [23] .Similar to that in the W-UN system, no formation of ternary Mo-U-N phases are expected when all stable nitrides are considered at the Mo-UN interface.Thus, both results (modelling and experiments) indicate that Mo can be used to efficiently avoid interactions between UO 2 and UN.

UN-Ta-UO 2 system
Fig. 13 demonstrates that Ta strongly interacted with UO 2 and formed two phases between UO 2 and Ta: one above the initial   [52] .This ternary was also identified in a previous study, which describes the reaction of an excess of UO 2 powder with Ta 2 O 5 powder at 1973 K in an induction furnace [53] .
It seems that Ta diffused into UO 2 during sintering and the oxygen, from the uranium dioxide powder, migrated towards the Ta foil.Thus, a fraction of oxygen stayed in the ternary and the great majority continued to diffuse into the Ta bulk.A previous study [54] reports no interaction between UO 2 and Ta when a UO 2.14 -Ta pellet containing 14 wt% (19.7 at.%) of Ta metal powder was sintered at 1673 K in a vacuum furnace for 3 h.Conversely, a crackfree eutectic phase with the composition of 17 wt% (23.4 at.%)Ta was obtained at 1873 K when UO 2.03 powder was used (instead of UO 2.14 ).The ternary phase observed in Fig. 13 contained about 23 at% of Ta, which is quite similar to the eutectic composition presented by [54] .
A different structure, UTa 3 O 9.5 (or U 2 Ta 6 O 19 ) [55] , can also be formed when UO 2 and Ta 2 O 5 powders react in the presence of HfO 2 (UO 2 :Ta 2 O 5 :HfO 2 = 1:2:2) at 1223 K [56] .Since our experiment was carried out at 1773 K and in absence of HfO 2 , as well as based on our semi-quantitative EDS results and previous articles [52][53][54] , the phase that was formed above the dashed line in the image is in good agreement with the UTa 2 O 7 ternary.
The EDS maps in Fig. 13 also reveals the formation of a tantalum oxide layer ( ∼50 μm) below the initial UO 2 -Ta interface highlighted in the figure.This oxide can be the result of Ta oxidation by the oxygen from the UO 2.13 powder.In fact, the oxygen atoms present at the UO 2 -Ta interface can be quickly transported into the tantalum foil with a diffusion coefficient (at 1773 K) of approximately 2.3 × 10 −9 cm 2 /s [57] .Such diffusion may be responsible for oxidising the Ta foil into tantalum oxide.From our EDS measurements at the regions (B), the oxide present is Ta 2 O 4.6 , which is similar to the Ta 2 O 5 stable oxide.A previous study [11] likewise observed a tantalum oxide phase between UO 2 and Ta after sintering a powder mixture of 95 wt% UO 2 -5 wt% Ta at 2423 K for 5 min.As discussed elsewhere [ 7 , 8 , 25 , 58 , 59 ], the oxygen from hyperstoichiometric UO 2 + x (also from UO 2 and UO 2-y ) can be mobile and then interact with the Ta foil to form the observed Ta 2 O 5 layer ( Fig. 13 ).Oxygen diffusion in Ta is about 160,0 0 0 times higher than in Ta 2 O 5 at 1773 K (1.4 × 10 −14 cm 2 /s [57] ).This slower diffusional behaviour in Ta 2 O 3 was not enough to stop the O diffusion because O can continue diffusing from Ta 2 O 5 at high temperatures, where small stoichiometry shifts to Ta 2 O 5-x ( x = 0.2-0.3) is possible.So, the gradient for oxygen diffusion throughout the Ta 2 O 5 -Ta interface could continue until the system was cooled down.
Fig. 14 reports the Ta-UN interface and the EDS chemical maps of U, Ta, O and N, together with their chemical compositions (at%) at two distinct regions (A) and (B).Macro cracks and spalls are observed on the UN pellet surface, as well as a thick oxide layer ( ∼45 μm) within the nitride pellet.Our results differ from a previous study [16] , which reports that Ta did not react with UN even after 66 h in contact at 1673 K in vacuum.This different behaviour might be related to a lower temperature used in the authors' work, and also because they used only Ta and UN.The UO 2 fuel, in our study, added complexity to the system due to the oxygen mobility throughout the Ta foil.
Higher magnification EDS maps of O and N in Fig. 14 indicate that the oxidation process occurred primarily via UN grain boundary attack and, then, evolved to intra-grain oxidation.As previously    discussed, the diffusivity of O in Ta may be considerably high to cross the Ta foil and reach the UN pellet's surface and oxidise it to UO 2 .This oxidation could cause the expansion of about 40% of the unit cell [ 33 , 39 ], which could contribute to crack the as-oxidised UN pellet surface.Yet, the differences in the linear thermal expansion coefficients could have caused the macro cracks during cooling, since the values at 1773 K are reported to be UO 2 > UN > Ta (14.6 x 10 −6 K −1 [47] , 9.6 x 10 −6 K −1 [51] , and 8.0 × 10 −6 K −1 [48] , respectively).Another possible explanation for the observed cracks at the Ta-UN interface would be the formation of a U-Ta-N ternary phase as a result of the interaction between the nitrogen available during the UN oxidation [39] .According to our DFT calculations [23] , the N atoms can diffuse from the UN pellet's surface into the tantalum metal and be trapped by the interface phase UTaN 2 forming N vacancies at the UN surface.This dynamic would reach thermodynamic equilibrium with the presence of the ternary.To verify this possibility, a FIB cross-section of the Ta-UN interface was obtained and is shown in Fig. 15 .The same pattern reported in the regular cross-section is observed in the bulk: macro cracks and spalling.This microstructure indicates that an oxide phase could be formed in the UN pellet and, due to the difference in the thermal expansion coefficients mentioned, the phases could crack and spall.Additional ab initio studies should be performed to consider the influence of O in the Ta-UN system.But it can also be possible that the UTaN 2 phase was formed at the Ta-UN interface as predicted by DFT and, due to either the interaction with O or thermal stresses during cooling, the ternary was dissociated and/or spalled.Considering any of these interaction mechanisms, Ta metal would not be a reasonable option to act as a protective barrier against N and O interdiffusion.

UN-Nb-UO 2 system
The UO 2 -Nb interface in Fig. 16 shows that the materials interacted and formed a cracked interface, with a transition layer of about 25 μm (interaction zone).According to the EDS chemical maps, there are two phases at this layer: an O-rich phase indicated by regions (A), and a N-rich phase represented in the regions (B).Semi-quantitative EDS measurements suggested that phase (A) is a uranium oxide phase with chemical composition UO 1.8 , which is in good agreement with the UO 2 phase.Phase (B) was identified as Nb 1. 7 N, which is similar to the stable Nb 2 N phase [60] .This N- rich phase was spread throughout the whole Nb foil, as visualised in Fig. 7 .
Niobium can interact with uranium oxides to form ternary phases [ 61 , 62 ] such as UNb 2 O 7 [63] , UNb 3 O 10 [64] , and UNb 4 O 12 [65] .However, amongst these oxides, only UNb 3 O 10 is stable below 1373 K [61] .Thus, what might have occurred at the UO 2 -Nb interface is the formation of a UNb x O y ternary during heating due to the availability of oxygen from UO 2.13 , followed by dissociation of this ternary into uranium and niobium oxides below 1773 K on cooling [61] .This mechanism would cause the observed cracks and phase segregation ( Fig. 16 ) at the end of the heat treatment.
The Nb-UN interface reported in Fig. 17 shows a strong interdiffusion of O and N across the interface.This phenomenon could be explained by oxygen diffusion from uranium dioxide throughout the Nb foil until reaching the UN phase.In fact, it is reported that the solubility of O in Nb can vary from ∼1 mol% at 1273 K to ∼5 mol% at 1773 K [66] .Thus, from the microstructure presented in the O chemical map, it seems that O diffused into UN via grain boundaries and formed a thick oxide layer ( ∼30 μm).According to the EDS results in regions (A), this phase is comparable to UO 2 .Conversely, N diffused into Nb and formed a N-rich phase in the metal, as represented in regions (B), with an estimated chemical composition of Nb 1. 6 N.This composition is similar to the phase found at the UO 2 -Nb interface ( Fig. 16 ).Thus, it appears that N diffused into the Nb foil and O into UN, forming Nb 2 N in the metal and UO 2 in UN, respectively.This experimental observation is in good agreement with our DFT calculations [23] , which proposes that UNbN 2 decomposes at the UN-Nb interface into UN and Nb 2 N, liberates N and allows N incorporation in Nb.Therefore, from both computational and experimental results, niobium metal is not a good coating candidate for the UN fuel since it strongly interacted with both UO 2 and UN fuels.

UN-V-UO 2 system
The UO 2 -V interface in Fig. 18 reveals a severe interaction between both phases.Qualitative EDS maps of U and V indicate that vanadium diffused into UO 2 via grain boundary attack, showing a fully connected V-rich solid skeleton filling the gaps between the UO 2 grains.A possible explanation for that is the formation of vanadium oxides such as V 2 O 5 and VO 2 , at the interface due to interaction between UO 2.13 and V.A previous study [67] reports that the oxidation of V at low temperatures (473-873 K) forms V 2 O 5 (major phase) and VO 2 .Yet, a mixture of V 2 O 5 (93-96%)-VO 2 (4-7%) was obtained by thermal oxidation of V metal at 763 K [68] .There are a few possibilities to form liquid VOx up to 1773 K [69] .For instance, the congruent melting of V 2 O 5 at 951 K reported in [69] is thermodynamically favourable to occur when UO 2 + x and V are in contact [ 40 , 41 ].Thus, liquid V 2 O 5 might have caused the UO 2 grain boundary attack.
Fig. 18 also shows that the N atoms from the UN pellet diffused throughout the V foil and reached the oxide side.Semi-quantitative EDS chemical compositions of the N-rich phases (A), (B), and (C) are reported in Fig. 18 .Phases (A) and (C) had a similar N/V ratio of ∼0.2, while phase (B) showed a higher nitrogen content (N/V ∼0.6).These phases (A) and (C) are more closely related to the stable V 8 N phase (N/V = 0.125), while the N-rich phase (C) seemed to be the stable phase V 2 N (N/V = 0.5).This N-rich compound is the same observed at the V-UN interface ( Fig. 18 ) and throughout the whole V foil ( Fig. 7 ).
As presented in Fig. 19 , V and UN strongly interacted by interdiffusion of O and N across the interface.Similar to Ta and Nb, the oxidation of UN seemed to occur via grain boundary attack followed by intra-grain oxidation.EDS results in regions (A) indicate that this oxide phase is UO 2 , and the thick ( ∼45 μm) N-rich phase formed at the interface (regions B) is the same N-rich phase (possibly V 2 N) observed at the UO 2 -V interface.The EDS chemical map    of O also shows a possible diffusion path of oxygen in the N-rich phase towards the UN fuel.
Therefore, it seems that N from UN diffused throughout V and formed a N-rich phase at the V-UN interface (possibly V 2 N), with O migrating in the opposite direction (towards UN).This V 2 N phase is also found in the V foil and at the UO 2 -V interface, where another N-rich phase (possible V 8 N) was present in the sample.Thus, an overall layered microstructure such as V 8 N/V 2 N/UO 2 (in the UO 2 pellet)-V 2 N/V 8 N/V (in the V foil)-UN/UO 2 (in the UN pellet) was observed ( Fig. 7 ).A layered structure was also identified by our DFT calculations [23] , which show a substantial driving force to increase the N content in the V bulk and forming vanadium nitrides at the V side.Thus, a N-rich phase can change to hypostoichiometric VN 1-x while releasing N atoms.These behaviours observed in the experiments and DFT calculations make the use of V unviable as a potential coating for the UN fuel., respectively.This Cr 2 N phase was spread throughout the whole V foil ( Fig. 7 ).

UN-Cr-UO 2 system
Fig. 20 shows that a CrO x phase was formed at the UO 2 -Cr interface.According to the EDS measurements at the regions (A), this phase is a chromium oxide with an O/Cr ratio of 1.8, which could be either Cr 2 O 3 or CrO 2 .It is reported that the thermal oxidation of Cr metal gives a single type of chromium oxide, Cr 2 O 3 , at a wide range of temperature (300-1373 K) [70][71][72] .Thus, it seems that the oxide phase formed at the UO 2 -Cr interface was the Cr 2 O 3 stable phase.
Cracks are observed at the UO 2 -Cr interface in both UO 2 and chromium oxide phases, probably due to tensile stresses generated by the metal oxidation, as well as due to the differences in the linear thermal expansion coefficient ( α) of Cr and UO 2 at e.g.1773 K ( αCr / αUO 2 ∼1.5) [ 47 , 73 ].Moreover, a N-rich phase is present at this interface, as observed in the regions (B) ( Fig. 20 ).According to our EDS measurements at the regions (B), this phase presented a Cr/N ratio of 2.1.This result suggests that this phase could be Cr 2 N [74] .
The Cr-UN interface in Fig. 21 shows a strong chemical interaction between the materials.As shown in the EDS map of N, a N-rich phase (regions A) was formed in the Cr foil as a result of the interdiffusion of N and O across the interface.Yet, the O chemical map confirms that oxygen migrated towards the UN pellet and formed an oxide phase there.The EDS measurements in the regions (A) indicate that this N-rich phase could be the same Cr 2 N observed at the UO 2 -Cr side.At the regions (B), the results suggest the formation of UO 2 in UN.
Previous studies on nitridation of Cr metal observed the formation of CrN [ 75 , 76 ], Cr 2 N [76][77][78] and CrN + Cr 2 N mixtures [ 76 , 77 , 79 ].These studies show that the atmosphere of the process is important to form one specific nitride, instead of another.For instance, in a low nitrogen activity environment, the formation of Cr 2 N is prioritised due to the higher thermodynamic stability of that phase [77] .On the other hand, by increasing the availability of nitrogen in the atmosphere, an initial CrN x phase changes from Cr + Cr 2 N to single-phase Cr 2 N, and then from Cr 2 N + CrN to single-phase CrN [76] .These previous studies corroborate our assumption that the N-rich phase observed in the regions (B) ( Fig. 20 ) and (A) ( Fig. 21 ) may be the Cr 2 N compound, since the nitrogen activity was limited to a certain extent during the heat treatment in vacuum.
Therefore, the Cr foil strongly interacted with both UO 2 and UN.This system also showed an overall microstructure of interdiffusion of O and N across the metal foil, possibly with the formation of Cr 2 O 3 at the UO 2 -Cr interface, as well as Cr 2 N at the Cr-UN side.This Cr 2 N is observed throughout the whole Cr foil ( Fig. 7 ).These experimental observations agree with our DFT calculations [23] .From the DFT results, the presence of a ternary phase U 2 CrN 3 at the Cr-UN interface provides a driving force to incorporate N atoms in the Cr matrix, allowing then the formation of chromium nitrides.This process is accompanied by the decomposition of U 2 CrN 3 into UN and Cr 2 N. Thus, based on both experimental and DFT results, Cr metal is not a good candidate for protecting the UN phase against the interaction with UO 2 .

Conclusions
Pressure-assisted diffusion experiments were used to analyse the interface interactions in UN-X-UO 2 ( X = V , Nb, Ta, Cr, Mo, W) composite systems at 1773 K and 80 MPa for 10 min in vacuum.The SPS method was used in this experimental setup as a pressureassisted apparatus to guarantee the contact between the phases during heating/cooling.Moreover, this methodology can be used in any multiphase system to provide the same conditions used to fabricate sintered composite samples.The cross-sections containing the UN-X-UO 2 interfaces were prepared using a pioneering procedure that consists of directly mounting the heat-treated samples, without removing the graphite paper from the SPS die, in order to minimise cracking and spalling during metallographic preparation.The mounted samples were cut longitudinally to expose the interfaces for grinding and polishing.The polished cross-sections were analysed by SEM coupled with FIB and EDS.
The results suggest that Mo and W are the most promising coating candidates to protect the UN fuel against interactions with UO 2 .Both metals are inert to N migration and preserve sharp interfaces with the nitride fuel.Previous DFT calculations performed by our group also point to this conclusion.The observed cracks at the UO 2 -W interface may be due to favourable thermodynamic relations between both materials and WO 3 at its near-melting temperature ( ∼1743 K), as well as due to thermal expansion mismatches and volatility of the compounds.Mo metal preserves a crack-free but irregular interface with UO 2 .Moreover, cracks in UO 2 (bulk) due to differences in the thermal expansion coefficients of UO 2 and Mo are observed.By tuning, for instance, the sintering temperature ( < 1373 K) and heating/cooling rates ( < 50 K/min), it seems possible to overcome these anomalies at W-UO 2 and Mo-UO 2 interfaces.
Ta, Nb, V and Cr strongly interact with both UO 2 and UN.At the UO 2 -Ta interface, two layered phases are observed: UTa 2 O 7 ( ∼6 μm thick), above the original UO 2 -Ta interface, and Ta 2 O 5 ( ∼50 μm thick) towards Ta bulk.At the Ta-UN interface, cracks are observed in UN together with its oxidation into UO 2 via grain boundary attack.Nb interaction with UO 2 generates an irregular/cracked interface with an interaction zone of ∼25 μm, where UO 2 and Nb 2 N are present.N diffuses into Nb and forms the Nb 2 N phase at the Nb-UN interface and throughout the Nb foil, with oxidation of UN also occurring via grain boundary attack.V diffuses in UO 2 probably via liquid V 2 O 5 through UO 2 grain boundaries and forms a fully connected V-rich solid skeleton around the UO 2 grains after cooling.N also diffuses in V and forms a layered structure along with the foil: V 2 N near the UN interface and V 8 N throughout the foil.Cr strongly interacts with UO 2 and UN, forming Cr 2 O 3 and Cr 2 N at the UO 2 -Cr and Cr-UN interfaces, respectively.Cr 2 N is also found throughout the whole Cr foil.
The results presented in this article are relevant to provide enough experimental data not only to validate our recent DFT calculations at the X-UN interfaces, but also to other groups that may study UN-X-UO 2 composites.Although the simulations consider the reactions at 0 K and in absence of oxygen, several aspects are well described by the modelling.DFT predicts well what is observed at the interfaces between UN and W, Mo, Nb, V and Cr.Ta metal behaves differently in the experiments probably due to the strong interaction between O and UN observed in the EDS maps.
The findings in this study may be valuable to identify potential candidates to overcome the interaction between the UO 2 and UN fuels, and to suggest new insights on using the SPS method as a pressure-assisted diffusion apparatus for interface examinations of multiphase systems.Furthermore, this work may encourage further experimental and modelling developments in UN-X-UO 2 advanced technology fuels.

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.

Fig. 6
Fig. 6 reports a regular cross-section of the UN-UO 2 reference

Fig. 5 .
Fig.5.X-ray diffraction patterns of the UO 2.13 , U 2 N 3 and UN powders, as well as of a sintered UN pellet.UO 2.13 broadening peaks can be attributed to peak overlap owing to the presence of secondary phases such as U 3 O 7 , U 3 O 8 , or U 4 O 9[32][33][34] .Broadening in U 2 N 3 can be a result of the continuous incorporation of N 2 during the nitriding process, which can cause structure distortions[ 35 , 36 ].

Fig. 7 .
Fig.7.SEM-EDS images of the UN-X-UO 2 (X = W, Mo, Ta, Nb, V, Cr) regular cross-sections.The interfaces between UO 2 and W, Ta, Nb, V and Cr were affected during the heat treatment.Moreover, there were interactions between the UN pellet and Ta, Nb, V and Cr metals, but any interaction with W and Mo.

Fig. 8 .
Fig. 8. Regular cross-section of the UO 2 -W interface showing some interface cracks and the presence of W around some pores/cracks in UO 2 (highlighted in the EDS map of W).

Fig. 9 .
Fig. 9. FIB cross-sections at two different positions on the UO 2 -W interface.The same porous/cracked interface is observed in W bulk.These images and EDS maps indicate that the thermodynamic relations between UO 2.13 and W are relevant during the heat treatment.

Fig. 10 .
Fig. 10.FIB cross-section at the W-UN interface showing the EDS qualitative chemical maps of U, W, O and N, as well as a line measurement across the interface.Sharp interface and short-range ( < 0.5 μm) chemical composition (at.%) are observed in the EDS maps and line measurements, respectively.

Fig. 11 .
Fig. 11.FIB cross-section at the UO 2 -Mo interface showing the EDS qualitative chemical maps of U, Mo and O, as well as a line measurement across the interface.The results show a crack-free and irregular interface, with a short-range variation ( < 0.5 μm) of the chemical composition (at.%) across the interface.

UO 2 -
Ta interface (dashed line) and another below this interface.According to the EDS measurements in the regions (A) and (B), highlighted in the figure, the phase (A) is a U-Ta-O ternary ( ∼6 μm thick) with chemical formula UTa 2.5 O 7.1 , which is similar in composition to UTa 2 O 7

Fig. 12 .
Fig. 12. FIB cross-section at the Mo-UN interface showing the EDS qualitative chemical maps of U, Mo, O and N, as well as a line measurement across the interface.The results show a sharp and crack-free interface, with a short-range variation ( < 0.5 μm) of the chemical composition (at.%) across the interface.

Fig. 13 .
Fig. 13.Regular cross-section at the UO 2 -Ta interface showing a severe interaction between Ta and UO 2 .Two phases are observed in the figure: one is represented by the regions (A) and another by the regions (B).Semi-quantitative EDS measurements in the regions (A) and (B) indicate that (A) is a ternary phase with a composition close to UTa 2 O 7 , and (B) is a tantalum oxide with a composition similar to Ta 2 O 5 .

Fig. 14 .
Fig. 14.Regular cross-section at the Ta-UN interface showing a cracked UN pellet surface.Higher magnification EDS maps of O and N show the oxidation extent in the UN pellet.It seems that the oxidation occurred initially via grain boundary attack, followed by intra-grain oxidation.Phases highlighted in regions (A) and (B) were identified as uranium dioxide and uranium nitride, respectively.

Fig. 15 .
Fig. 15.FIB cross-section at the Ta-UN interface showing the EDS qualitative chemical maps of U, Mo, O and N.This bulk microstructure also shows cracks at the interface and strong oxidation of the outer surface of the UN pellet.

Fig. 16 .
Fig. 16.Regular cross-section at the UO 2 -Nb interface showing a transition layer of about 25 μm (interaction zone).EDS semi-quantitative measurements suggest that phases (A) and (B) closer match UO 2 and Nb 2 N, respectively.

Fig. 17 .
Fig. 17.Regular cross-section at the Nb-UN interface showing a strong oxidation of the UN pellet (regions A) and the formation of a N-rich phase in Nb (regions B).EDS semi-quantitative measurements suggest that phases (A) and (B) are UO 2 and Nb 2 N, respectively.

Fig. 18 .
Fig.18.Regular cross-section at the UO 2 -V interface showing a severe interaction between the materials.EDS semi-quantitative measurements suggest that phases (A) and (C) are closely related to V 2 N, and phase (B) to V 8 N. The suggested V 2 N phase is spread throughout the whole V foil ( Fig.7).

Fig. 19 .
Fig. 19.Regular cross-section at the V-UN interface showing a huge interaction between the materials.EDS semi-quantitative measurements suggest that phases (A) and (B) are fairly related to UO 2 and V 2 N, respectively.Higher magnification EDS maps of O and N show an oxidation mechanism via grain boundary attack.

Fig. 20 .
Fig. 20.Regular cross-section at the UO 2 -Cr interface showing the formation of a chromium oxide layer.EDS semi-quantitative measurements suggest that phases (A) and (B) are best represented by Cr 2 O 3 and Cr 2 N, respectively.

Fig. 21 .
Fig. 21.Regular cross-section at the Cr-UN interface showing the severe interaction between both phases.EDS semi-quantitative measurements indicate that phases (A) and (B) are closely represented by Cr 2 N and UO 2, respectively.This Cr 2 N phase was spread throughout the whole V foil ( Fig.7).