Multimodal X-ray microanalysis of a UFeO 4 : evidence for the environmental stability of ternary U( V ) oxides from depleted uranium munitions testing †

An environmentally aged radioactive particle of UFeO 4 recovered from soil contaminated with munitions depleted uranium (DU) was characterised by microbeam synchrotron X-ray analysis. Imaging of uranium speciation by spatially resolved X-ray di ﬀ raction ( m -XRD) and X-ray absorption spectroscopy ( m -XAS) was used to localise UFeO 4 in the particle, which was coincident with a distribution of U( V ). The U oxidation state was con ﬁ rmed using X-ray Absorption Near Edge Structure ( m -XANES) spectroscopy as +4.9 (cid:1) 0.15. Le-Bail ﬁ tting of the particle powder XRD pattern con ﬁ rmed the presence of UFeO 4 and a minor alteration product identi ﬁ ed as chernikovite (H 3 O)(UO 2 )(PO 4 ) $ 3H 2 O. Re ﬁ ned unit cell parameters for UFeO 4 were in good agreement with previously published values. Uranium – oxygen interatomic distances in the ﬁ rst co-ordination sphere were determined by ﬁ tting of Extended X-ray Absorption Fine Structure ( m -EXAFS) spectroscopy. The average ﬁ rst shell U – O distance was 2.148 (cid:1) 0.012 ˚ A, corresponding to a U valence of +4.96 (cid:1) 0.13 using bond valence sum analysis. Using bond distances from the published structure of UFeO 4 , U and Fe bond valence sums were calculated as +5.00 and +2.83 respectively, supporting the spectroscopic analysis and con ﬁ rming the presence of a U( V )/Fe( III ) pair. Overall this investigation provides important evidence for the stability of U( V ) ternary oxides, in oxic, variably moist surface environment conditions for at least 25 years. long term environmental behaviour and health risk posed by depleted uranium particles depends critically on uranium speciation, which is of importance in managingand remediating contaminatedland. In particular, oxic and variably moist surface conditions areexpectedto promote oxidation and dissolution of U( V ) phases to form U( VI ) species. Here, we demonstrate the long term (>25 year) stability of UFeO 4 , under such conditions, formed by testing of depleted uranium munitions, using multi-modal X-ray microanalysis. The broader signi  cance of this study provides evidence for the environmental stability of U( V ) phases of relevance to environmental contamination by radioactive particles from nuclear fuel cycle and other activities.


Introduction
Radioactive and hot particles are introduced into the environment by a number of civil and military nuclear events, including nuclear power plant (NPP) accidents, effluent discharges from nuclear fuel reprocessing, nuclear weapons testing and acts of war. 1 In order to understand the long term environmental behaviour and health risk posed by these particles, information on physicochemical characteristics is required such as morphology, radionuclide inventory and major element speciation. These data can also provide information on the formation and origin of radioactive and hot particles. 2 In this study, information on the chemical speciation of U in a particle containing the ternary oxide UFeO 4 is established by multi-modal synchrotron X-ray microscopy. Ternary compounds in the U-Fe-O system are of interest in the interaction of uranium wastes with iron oxides 3 and as a component of corium in severe nuclear power plant accidents. 4,5 Iron is ubiquitous in structural components of nuclear reactor systems, and in particular, some modern nuclear reactor designs employ hematite as a sacricial barrier in core catcher systems. 6 Compounds of Fe and U have also been previously described as minority phases in some hot particles. 7 In the U-Fe-O system two ternary oxides are known, UFeO 4 and UFe 2 O 6 , the latter suggested to be stable only at high pressure. 3,[8][9][10][11] UFeO 4 crystallises in an orthorhombic system with space group Pbcn, 8,10,11 and a similarly structured ternary oxide in the U-Cr-O ternary system (UCrO 4 ) has also been synthesised and characterised. [11][12][13] The oxidation state of uranium in these compounds was rst inferred as U(V) by measurement of a small magnetic moment on the U atom, 12,14 and more recently veried by X-ray Absorption Spectroscopy and X-ray Photoelectron Spectroscopy, with supporting evidence from DFT calculations. 11,15 Pentavalent uranium disproportionates to U(IV) and U(VI) in aqueous systems, and as such is rarely found in geologic materials. 16 However, many compounds of U(V) have been characterised, including some rare examples of naturally occurring U(V) minerals, of which wyartite (CaU 5+ (UO 2 ) 2 (CO 3 )O 4 (OH)(H 2 O) 7 ) was the rst to be identied. 17 The most common co-ordination environment for U(V) is pentagonal bipyramidal, 18 although some structures containing U(V) in 8-fold 19 and distorted octahedral environments are reported. 16 The U sites in UFeO 4 and UCrO 4 are octahedral with differing extents of distortion. 8,[10][11][12] The high X-ray photon ux and small spot size achievable with modern microfocus synchrotron X-ray beamlines allows the use of localised X-ray absorption spectroscopy (XAS) techniques to probe the oxidation state and chemical environment of elements in radioactive and hot particles, which may not be amenable to regular preparation or characterisation methods. 20,21 The particle in this study was recovered from soils contaminated as a result of depleted uranium munitions test ring. 22 The use of these techniques offers a direct measure of the uranium oxidation state to demonstrate the presence of U(V) in UFeO 4 , and provides evidence on the environmental behaviour of this compound. Coupled with renement of micro X-ray diffraction data, and elemental analysis by microfocus X-ray uorescence spectroscopy (m-XRF), these techniques provide an integrated methodology for detailed chemical characterisation of radioactive and hot particles of a scale commensurate with, or greater, than the X-ray footprint.

Particle collection
Particles containing DU were sampled from the UK Ministry of Defence Eskmeals ring range, Cumbria, UK in November 2010 and separated using autoradiography and sample splitting. 23 Details of the soil sampling, preparation, and a comprehensive SEM imaging study of the uranium particulate morphology and composition, were published previously. 22,24 The particle of UFeO 4 selected for this study was from DU contaminated soil that has been exposed to the environment for at least 25 years and is representative of particulates of this phase which occur as a minor fraction of the U-bearing particulates in these soils. The UFeO 4 particles were identied from co-location of Fe Ka and U La emission in m-XRF maps and were not obviously identiable by morphology in our previous SEM analysis, 22,24 demonstrating the advantage of using high brilliance synchrotron radiation for wide area m-XRF analysis to select particles of interest.

Synchrotron X-ray micro-analysis
Particles were mounted on Kapton tape (area $ 1.3 cm 2 ) for microfocus X-ray characterisation experiments performed at the microXAS (X05LA) beamline at the Swiss Light Source. 20 The source spot size was 2 mm (v) x 5 mm (h), and the samples were mounted on an x-y-z stage at 25 to the incident beam to allow localisation of different areas of interest in the beam. All data were collected at ambient conditions. UFeO 4 particles appeared relatively abundant in this specimen, with 2 of the 10 particles selected for analysis conclusively identied as UFeO 4 (the others being uranium oxides or secondary alteration products previously described 22,24 ), with a similar abundance in other specimens. Here we report a detailed characterisation of a representative UFeO 4 particle.
X-ray uorescence (m-XRF) spectra were collected using a silicon dri detector (KETEK instruments) placed at 90 to the incident beam. 2D X-ray diffraction (XRD) patterns were recorded using a PILATUS 100K Hybrid Photon Counting (HPC) pixel array detector 25 mounted 46 mm behind the sample with a tungsten beamstop in place. The m-XRD setup was calibrated with respect to a silicon standard (NIST 640c), and the angular resolution was approximately 0.1 2q. Incident photon energy for m-XRF and m-XRD was 17.500 keV (l ¼ 0.70849Å).

Micro-XANES and micro-EXAFS
X-ray absorption near edge structure (m-XANES) spectroscopy was performed in uorescence mode across an energy range of 16.900 to 17.500 keV. Energy calibration was performed with respect to the K edge of yttrium foil (17.038 keV). m-XANES spectra of uranium reference compounds of different oxidation state were recorded to aid interpretation, including UO 2 (U 4+ ), U 0.5 Y 0.5 Ti 2 O 6 (U 5+ ), 19,21 U 3 O 8 (U 5.33+ ) 26 and UO 3 (U 6+ ). Reference compounds were prepared as 3 mm diameter pellets of ceramic powders distributed in polyethylene glycol (PEG). The edge shi from U 4+ to U 6+ standards was 3.2 eV and the energy resolution across the edge region was 0.1 eV, resulting in an oxidation state uncertainty of approximately 3%. Oxidation states were estimated by a calibration line established for a linear relationship of oxidation state and chemical shi, using the reference compounds (see Fig. S1 †).
Extended X-ray absorption ne structure (m-EXAFS) spectroscopy was performed in uorescence mode at the uranium L 3 edge. Data were collected across an energy range of 16.900 keV to 18.000 keV. Raw XAS data were processed using the program Athena 27 to remove the absorption edge background. EXAFS data were self-absorption corrected with an idealised composition of only UFeO 4 using the Troger algorithm implemented in Athena. 27,28 Theoretical backscattering path phase and amplitude functions were calculated using FEFF 6 and t to the data using the Artemis/IFEFFIT soware package. 27,29 Fits were performed to Fourier transformed R-space data with k-weights of 1, 2 and 3 to reduce parameter correlation.

Chemical imaging
Rastering of a sample in the X-ray microbeam allows for maps of spatially resolved chemical information to be constructed. Elemental distributions were mapped by monitoring regions of the XRF spectrum corresponding to emission lines of interest whilst the sample was moved in the beam. Phase distributions were similarly mapped by monitoring the intensity of Bragg reections corresponding to phases of interest, using the soware XRDUA. 30 The spatial distribution of uranium oxidation state was determined using a m-XAS mapping approach. 22,31,32 Maps of absorption co-efficient were constructed by m-XRF mapping divided by incident intensity (I 0 ) at two energies in the U L 3 XANES region (17.168 keV and 17.850 keV), normalised with respect to post-edge energy (17.500 keV). The estimated oxidation state was calculated from the per-pixel absorption coefficient with reference to a linear calibration relationship derived from uranium standard spectra (see Fig. S2 †). Maps of oxidation state at both energies showed good agreement and were averaged. It should be noted that both the average local structure of the absorber element and oxidation state determine the normalised absorption at each pixel, and, therefore, this approach affords a map of chemical speciation contrast. Nevertheless, by choosing the excitation energies with due care, and with validation using an independent technique, is possible to construct chemical speciation maps dominated by oxidation state contrast. Our choice of excitation energies is based on an earlier investigation, in which optimisation allowed differentiation of U 3 O 7 and U 3 O 8 by chemical speciation mapping, veried by m-XRD. 22 3 Results and discussion 3.1 Synchrotron X-ray chemical imaging X-ray chemical imaging was used to probe the spatial distribution uranium species in a set of DU particles recovered from contaminated soil on a UK ring range. 22 Areas containing uranium were localised using XRF mapping (Fig. 1a), and the distribution of U species analysed by oxidation state and XRD mapping ( Fig. 1c and d). In this study, a particle containing UFeO 4 was characterised aer rst being localised and identi-ed by this chemical imaging approach. Fig. 1d shows an approximately circular domain of UFeO 4 ($10 mm), with a similar shaped region of elevated uorescence intensity observed in both uranium and iron elemental maps ( Fig. 1a and b). These distributions suggest a spherical particle, which is a common morphology for residues formed from the ring of DU munitions against hard targets, due to the low melting point of metallic uranium. 23 Such particles have been observed previously in soils from this site, comprising primarily U 3  3.2 Microfocus X-ray diffraction and X-ray uorescence spectroscopy Fig. 2 shows powder diffraction data extracted from the DU particle in the centre of Fig. 1a; the pattern was obtained by summing per-pixel XRD data over the UFeO 4 particle area ($10 mm). These data show that the main phase present in this particle is UFeO 4 , with a minor contribution from the coassociated uranyl secondary alteration phase, which was initially modelled as meta-ankoleite, 33 although further analysis described below suggested this phase is actually the iso- U-Fe phases are thought to be produced by high temperature interactions (T $ 3000 C (ref. 34)) which arise on impact of DU munitions with steels in armour plate. Laves phases such as UFe 2 have been observed in DU residues, 6 and UFeO 4 can form as a minority high temperature oxidation product of this phase. 35 In this particle however, the lack of other UFe 2 oxidation products (such as UO 2 or FeO 35 ) suggests that UFeO 4 may form as a primary species. The presence of a UFeO 4 particle in these soils shows that this phase can persist in oxic, variably moist surface environment conditions, which may be expected Quantitative analysis of powder diffraction data was performed using a Le Bail intensity extraction method. 36 A low number of randomly oriented crystallites in the particle was evident as 'spots' rather than full rings in the 2D diffraction pattern (Fig. 2). Although the angular position of reections is unaffected, this limits a full structural analysis (e.g. by Rietveld renement) as the intensity of reections in the pattern are distorted. This has been previously noted as a problem in the renement of m-XRD data. 37 The Le Bail method removes the link between the model structure and peak intensities, and allows unit cell parameters to be rened without a structural model, independent of preferred orientation effects. However for low symmetry systems this approach may incorrectly resolve closely spaced peaks, as intensities are not constrained by a structural model. 38 To overcome this, the results of Le Bail tting are recommended to be compared with results from tting to a structural model, even if this is imperfect. 38 In this study good agreement between Le-Bail rened unit cell parameters (Table 1) and a limited Rietveld analysis (data not shown) was observed.
The pattern was adequately described (c 2 ¼ 6.48, R wp ¼ 13.7%, R p ¼ 6.43%) with contributions from UFeO 4 as the majority phase (95.9 wt%) with a minority presence of a secondary phase (4.1 wt%) modelled initially as meta-ankoleite. The good agreement of the rened and published unit cell values for UFeO 4 (ref. 7, 8 and 11) gives quantitative identication of this species in the particle, and allows correlation of our XAS data with the published structure.
Rened unit cell parameters (  39 which is isostructural with metaankoleite. These minerals have layered uranyl and phosphate polyhedral chains, with interlayers of water and cations of different size, which allows discrimination based on the unit cell size. Uranyl-phosphate-hydrate phases have been identied in other particles from this sample site at the Eskmeals range, linked to corrosion of DU oxide particles over extended periods of time in a waste disposal pit for contaminated timbers. 22 The co-location of minor amounts of chernikovite in this particle may suggest weathering of the UFeO 4 phase. However, the majority of the remaining particle is still composed of UFeO 4 , and the particle size is consistent with primary unaltered uranium oxide particles observed at this site, 22,24 indicating an extent of longer term environmental stability over at least 25 years of exposure to the surface environment. Due to the structural similarity between UFeO 4 and UCrO 4 , qualitative XRF spectroscopy was performed to conrm the identity of the primary U species. Fig. 3 shows that the particle is composed mainly of U and Fe, with only trace levels of Cr present. This result compares well with the rened unit cell parameters which are in good agreement with the presence of UFeO 4 only. To distinguish between chernikovite and metaankoleite, energy dispersive XRF spectroscopy is not useful as interference with U M emissions (U Mb ¼ 3339.8 eV) prevents conrmation of the presence of K (K Ka ¼ 3313.8 eV) in the sample, and secondary phase identication relies on rened unit cell parameters only.

X-ray absorption spectroscopy
3.3.1 XANES. XANES spectra of the UFeO 4 particle and U oxide standards are shown in Fig. 4. The sample spectrum shows closest agreement with the U(V) (U 0.5 Y 0.5 Ti 2 O 6 ) standard, although self absorption effects are evident, particularly in the dampening of white line and post-edge oscillation intensities. The rst derivative XANES spectra show that the inection point (B) is of similar position to that of U 0.5 Y 0.5 Ti 2 O 6 , which contains uranium in average U(V) oxidation state. A linear relationship between edge inection energy and oxidation state was established from standard spectra, and interpolated to calculate U oxidation state of +4.90 AE 0.15 in the UFeO 4 particle. The UFeO 4 rst derivative spectrum pre-edge feature (A) also appears similar in intensity to the U(V) standard, however selfabsorption artefacts may distort the magnitude of this peak.
Although damped by self-absorption, near edge structure suggests a U chemical environment distinct from that observed in UO 2 and UO 3 (Fig. 4). In particular, the multiple scattering resonance at an energy 10-15 eV greater than the white line observed in the UO 3   further direct evidence for the presence of U(V) in UFeO 4 , supporting the recent XAS and XPS studies of Guo et al., 11 and earlier inferences of Bacmann et al. 14 XAS mapping (Fig. 1c) provides a more rapid method than XANES analysis to determine the spatial variation in oxidation state. Good agreement between the oxidation state determined by XANES (+4.90 AE 0.15) and XAS redox mapping (particle average +5.20 AE 0.15) provides indication that U oxidation state throughout the particle is pentavalent. XAS mapping also reveals that the U(V) oxidation state is homogenous in the particle (Fig. 1c), and agrees well with the distribution of UFeO 4 determined by XRD phase mapping (Fig. 1d). Due to the coassociated chernikovite phase, incorporating the uranyl species, XPS could not be applied to reliably infer the U oxidation state in the particle.
3.3.2 m-EXAFS analysis. Calculation of backscattering path amplitude and phase shi was based on the crystal structure of UFeO 4 determined by Bacmann et al. and Guo et al. 8,11 The krange used for analysis was limited by energy dependence of the microbeam position 42 to 2-8Å À1 (Dk ¼ 6Å À1 ) as, at higher k, movement of the beam across U chemical gradients in the sample introduces additional oscillations into the data. The Fourier transform R space resolution for distinguishing individual scattering paths is equivalent to 1 independent data point (1 idp ¼ p/2Dk). 43 Using this criterion, 1 idp for this data is equivalent to 0.262Å, with a total of 9.73 independent data points (N idp ) in the R-range of 1.2-3.75Å (DR ¼ 2.55Å) used in tting. No backscattering paths from meta-ankoleite/ chernkovite were included in the t as the contribution of this phase to the composition of the particle was determined by XRD to be small (4.1%, Table 1). Table 1 Le Bail refined unit cell parameters for the two uranium phases identified by powder diffraction. Good agreement with the published unit cell values for UFeO 4 is observed, and unit cell parameters suggest that the second phase is chernikovite, which is iso-structural with metaankoleite. Uncertainty in the last figure of refined parameters is displayed in brackets  . 3 Qualitative XRF spectrum of the particle from Fig. 1 in which maximum counts for U La 1 (4.8 Â 10 5 ) and Fe Ka 1 (5.2 Â 10 4 ) compared to Cr Ka 1 (6.3 Â 10 2 ) indicate that the U ternary oxide phase is UFeO 4 rather than UCrO 4 . The excitation energy was 17.500 keV.   5 shows the k 2 -weighted EXAFS spectra and ts for k-and R-space from the UFeO 4 particle, with the t detailed in Table 2. The intensity in R-space is consistent with the published structure of UFeO 4 , with an intense and broad second peak arising from a number of scatterers in a complex second shell. Although the published structure of UFeO 4 indicates U in a distorted octahedral co-ordination, the R-space resolution offered by m-EXAFS (DR ¼ 0.262Å) analysis shows this as a single intense peak at 1.55Å (Fig. 5), corresponding to an average U-O distance of $2.15Å with phase correction. This shell was tted using backscattering phase and amplitude terms calculated for the middle path length (R 0 ¼ 2.155Å), with the path degeneracy (N) xed at 6 instead of 2. The EXAFS path length for this shell was rened to 2.148 AE 0.012Å (Table 1), representing an average of the U-O distances in the distorted rst shell geometry. This is agrees well with the mean crystallographic (R c ) U-O distance calculated from the published crystallographic structures of UFeO 4 (R c ¼ 2.148 and 2.168Å), 8,10 and that from a recent EXAFS investigation (R ¼ 2.148Å). 11 The second co-ordination shell in UFeO 4 is apparent as a broad peak in the m-EXAFS data in the region 2.5-3.7Å. The crystal structure shows this is expected to comprise 2 distinct O subshells and 3 Fe subshells. However, the close spatial relation of these paths and the limited number of available independent data points (N idp ¼ 9.53) mean it is not possible to resolve individual EXAFS parameters (s 2 , DR) for these paths. For the second shell paths in Table 2, changes in path length were described with a single scaling factor multiplied by the path length, and a single mean squared path length variation (s 2 ) was used. This two parameter model for the second shell allows a reasonable t to the data and extraction of useful chemical information from the rst shell, in particular the average U-O distance as discussed above. This value can be used to conrm the oxidation state of U in UFeO 4 by bond valence sum analysis.
We also considered the possibility that the U environment in UFeO 4 could be an average of U(V) as a result of plausible combinations of U(IV) and U(VI) environments. This involved modelling the EXAFS data using two U cores, initially congured as equal ratios of U(IV) and U(VI) in octahedral co-ordination to t the rst shell of the |FT k 2 c(k)|. The U IV O 6 environment was modelled with 6 Â d U-O ¼ 2.281Å and s 2 ¼ 0.003Å 2 , based on the environment in UTi 2 O 6 . 44 The U VI O 6 environment was modelled as: non-uranyl U(VI), with 6 Â d U-O ¼ 2.07Å and s 2 ¼ 0.003Å 2 ; or uranyl U(VI) with 2 Â d U-Oyl ¼ 1.798 A, 4 Â d U-O ¼ 2.275Å and s 2 ¼ 0.003Å 2 (using the mean distances for such environments determined from a comprehensive literature survey and analysis 45 ). When the proportions of the environments and path lengths were rened, the models converged to mean path lengths of $2.13Å, implying a single U environment. To develop further insight, we computed the k 2 c(k) and |FT k 2 c(k)| of plausible combinations of U(IV) and U(VI), using our initial models with FEFF 6 in the Artemis/ IFEFFIT soware package. 27,29 We compared the component and resultant calculations with that for the single U(V) environment determined in Table 2. The results of this analysis show that plausible bounding combinations of U(IV) and U(VI), charge compensated by Fe(III) and/or Fe(II), are unable to accurately approximate a single U(V) environment (Fig. S3 †). Thus we conclude that UFeO 4 incorporates U(V) in octahedral coordination, rather than a combination of U(IV) and U(VI) charge compensated by Fe(III) and/or Fe(II).
3.3.3 Bond valence sums. The bond valence sum method can be used to calculate element oxidation state, 46 based on the  principle that the bond length is a function of valence. The exponential parameterisation for cation bond valence (n i ) was used: where R ij is the measured bond length, R 0 is a reference bond length for unity valence and B is a constant. Values for R 0 (2.051 A) and B (0.57) were used from Burns et al., 18  Bond valence sum analysis may also be applied to bond distances calculated from the published crystal structures, 8,10 which yield a bond valence sum of 5.0, which is in excellent agreement with the oxidation state of U determined by XANES and EXAFS for this particle. A corresponding Fe valence may also be calculated using the published crystal structuresas the oxidation state of uranium has been determined as U(V), bond valence parameters for Fe(III)-O bonding were used (R 0 ¼ 1.

Conclusions
The presence of ternary U oxides in DU particles is indicative of intense interaction temperatures during impact, and the absence of other Fe and U oxide species in this particle suggests a primary formation mechanism for UFeO 4 . Importantly, the occurrence of UFeO 4 in environmentally aged demonstrates the medium term (>25 year) stability of this phase in the surface environment, which may not be expected for species containing U(V) in oxic, variably moist conditions. Studies of the UO 2 -Fe 2 O 3 -ZrO 2 ternary phase diagram, under conditions relevant to severe nuclear power plant accidents, and Fukushima Daiichi in particular, demonstrate the formation of UFeO 4 . 47,48 The evidence presented here for the long term environmental stability of UFeO 4 , may therefore be of considerable importance in predicting the evolution of hot fuel particles in the environment.
The oxidation state of uranium in UFeO 4 was determined as U(V) by microfocus synchrotron chemical imaging, m-XANES and m-EXAFS spectroscopies. Unit cell parameters of UFeO 4 were rened by Le Bail tting of powder XRD data, revealing values consistent with the structure of UFeO 4 determined by Bacmann et al. and Read et al. 8,10 This structure was used as an input to calculate EXAFS path amplitudes and phase shis, which were found to t well to the data, and agree well with the results of the recent EXAFS study of Guo et al. 11 Bond valence analysis of the EXAFS rened U-O bond and of the U and Fe sites in the original structure suggests a U(V)/Fe(III) couple, and conrms early studies of UFeO 4 in which U(V) was inferred, 12,14 and more recent U L 3 XAS and XPS studies. 11 This study demonstrates the utility of microbeam X-ray experiments to extract chemical information from challenging samples by a range of complementary analyses, which may be of interest in characterisation of secondary minerals, alteration products and other materials for which bulk samples are not available for conventional characterisation regimes. This approach is particularly suitable for radioactive and hot particles as it non-destructive, thereby preserving the limited sample for other complementary analyses and allowing safe containment of the material.

Conflicts of interest
There are no conicts to declare.