The nature of Pu-bearing particles from the Maralinga nuclear testing site, Australia

The high-energy release of plutonium (Pu) and uranium (U) during the Maralinga nuclear trials (1955–1963) in Australia, designed to simulate high temperature, non-critical nuclear accidents, resulted in wide dispersion µm-sized, radioactive, Pu–U-bearing ‘hot’ particles that persist in soils. By combining non-destructive, multi-technique synchrotron-based micro-characterization with the first nano-scale imagining of the composition and textures of six Maralinga particles, we find that all particles display intricate physical and chemical make-ups consistent with formation via condensation and cooling of polymetallic melts (immiscible Fe–Al–Pu–U; and Pb ± Pu–U) within the detonation plumes. Plutonium and U are present predominantly in micro- to nano-particulate forms, and most hot particles contain low valence Pu–U–C compounds; these chemically reactive phases are protected by their inclusion in metallic alloys. Plutonium reworking was observed within an oxidised rim in a Pb-rich particle; however overall Pu remained immobile in the studied particles, while small-scale oxidation and mobility of U is widespread. It is notoriously difficult to predict the long-term environmental behaviour of hot particles. Nano-scale characterization of the hot particles suggests that long-term, slow release of Pu from the hot particles may take place via a range of chemical and physical processes, likely contributing to on-going Pu uptake by wildlife at Maralinga.

. X-ray µ-tomography (a,d,e) and µSXRF imaging (b-c, f-i) of hot particles from Maralinga. In (a) (Potatohead) the low-X-ray cross-section areas within the particles (black) correspond to pores, whereas in (d) and (e) (Bruce) they correspond to Al-oxide rich domains. Potatohead displays a U-rich grain (purple circle in a,b,c) attached on the outside of the particle; in (e), a U-rich, Pu-poor inclusion is highlighted in yellow. Potatohead and Bruce were further characterised by µSXRF-fluorescence tomography (Fig. 1a,d,e), µ-X-ray Absorption Near Edge Structure (µXANES) (Fig. 2), µ-Extended X-ray Absorption Fine Structure (µEXAFS) (Fig. S4) and µ-X-Ray Diffraction (µXRD) (Fig. S5, S6) to provide information on Pu-U oxidation states and identify mineral phases present in the particles (details in SI-Hot particle composition).

Scientific Reports
Potatohead: single Pu form. This particle was found to closely resemble the particle studied by Ikeda-Ohno 2 : XANES spectra show that Pu is present mainly as Pu(IV) (Fig. 2) and the bond lengths derived from EXAFS shell-by-shell fits (Table S2; Table S3). The U XAS results also show that a U-rich, idiomorphic grain on the outside of the particle, circled in purple in Fig. 1a-c, contains a high proportion of U(VI) in the form of uranyl (short U=O bond distances (1.76-1.78 Å) from EXAFS data, Table S4, analyses #483-484 and #486; XANES spectra showing a significant edge shift to higher energy relative to the bulk grain, and a uranyl-characteristic band at ~ 17. 18-17.19 keV, Fig. 2d).
Bruce: various forms of Pu. µXRD data reveal a complex and heterogeneous phase make-up, including (Pu,U)O 2±x + (U,Pu)O 2±x , bromellite (BeO), and Al-Fe and Al-Mg alloys (Fig. S6). Together, the Pu-XANES and electron microscopy results provide strong evidence for the presence of low valence Pu compounds in Bruce. The Pu µXANES spectra (Fig. 2a,b) are shifted to a lower energy compared to Pu(IV), indicating contribution of a lower valence state, but the white line is less intense and broader than that expected for Pu(III) compounds 10,11 . These spectra can be fitted as a linear combination of Pu(0) + Pu(IV). FIB-SEM results (Figs. 3,4) indicate that Pu occurs in at least three different forms within Bruce, with the first two likely associated with low valence Pu rather than Pu(III): (i) carbide phases, (ii) a minor constituent in Fe-Al alloys, and (iii) Pu-U oxides. The presence of Pu carbides has not been reported in environmental particles to date. Dyker and Bertrand 12 note that carbon exists in metal-like state in some carbides, an observation that is confirmed by the XANES spectrum of uranium carbide (UC) resembling a combination of metallic U and UO 2 spectra 13 ; indeed, the U XANES spectrum obtained from Bruce indicates a lower formal oxidation state of U compared to UO 2 , and is consistent with a significant proportion of U present in carbide and/or in metallic-like (alloy) form (Fig. 2d). In the absence of published XANES data on Pu-carbides, we calculated the XANES spectra of known compounds using the abinitio code FDMNES 14,15 ; these simulations confirm that the white lines of Pu-C compounds are at lower energy than in PuO 2±x (Fig. 2c).
CeresI-III + Chip. These four particles consist predominantly of metallic Pb (Figs. 3g-i, 5). Plutonium and U are present in particulate form (arrows in Figs. 3g-i, 5), with different Pu:U ratios (different colours in Fig. 1h-k; this is confirmed by the correlation plots shown in Fig. S2g). XANES data indicate that Pu exists mainly as Pu(IV), and the U XANES spectra closely match those of uraninite (i.e., UO 2+x , Fig. 2d).
µSXRF-fluorescence tomography data further highlight the differences between Potatohead and Bruce (Fig. 1a,d,e; full movies in SI). The concentrations of U and Pu are generally positively correlated in Potatohead (Fig. S2g), with the exception of the U-rich grain on the outside (Fig. 1a-c), but they are decoupled in Bruce: www.nature.com/scientificreports/ the framework is Pu-rich, but Bruce also contains particles of a U-rich phase varying in size from ≤ 2 to 10 µm (Fig. 1d,f,g; Fig. S2g). Both particles contain volumes characterised by low total X-ray attenuation (Fig. 1a,d,e); FIB-SEM data (see below) reveal that these volumes correspond to empty pores in Potatohead, but consist of light elements (Al-oxide-rich domains) in Bruce.

Dissection of the hot particles reveals extreme textural complexity
The characteristic X-rays of light elements are fully absorbed by less than a µm of U/Pu-rich material, which limits the information that can be gathered via non-destructive X-ray fluorescence. Hence, FIB-SEM was used to dissect Potatohead, Bruce, CeresI and CeresIII (Figs. 3, 4, 5; a gallery of images is included in SI Fig. S7). High resolution SEM imaging and X-ray spectroscopy of the freshly exposed surfaces reveal the significant complexity of the internal textures of the grains and the micro (< 5 µm) to nano-particulate (1-450 nm) nature of the Pu-rich phases; we follow the definition of nano-particle of Salbu et al. 1 .
Potatohead. The (Pu/U)O 2±x phases in Potatohead exist as particles that are up to ~ 1 µm in diameter, and are embedded in an Al-rich matrix (Fig. 3a-c). The FIB-SEM imaging also shows that the porosity observed via X-ray tomography consists of abundant spherical pores with sizes down to ~ 1 µm. The morphology of the pores is indicative of gas entrapment; (Pu,U)O 2±x is highly volatile compared to (U,Pu)O 2±x , with boiling at 1 bar observed at temperatures as low as 2300 K and compositions as rich as 50 mol% UO 2 16 . All compositions in the  Bruce. FIB-SEM revealed a level of chemical and textural heterogeneity not previously recognised in Maralinga hot particles (Figs. 3d-f, 4). The textures indicate the coexistence of at least two immiscible polymetallic melts during the formation of the particle in the explosion environment: spherical inclusions of Pu-U-Fe(Al)rich composition float in an Al-oxide-rich matrix (Fig. 3d, SEM-EDX maps in Fig. S8). The Al-oxide-rich domains correspond to the low-density areas in the X-ray tomograph (Fig. 1d,e), and also contain numerous, sub-µm-sized inclusions of a Pu-U-rich phase (light spots in the Al-oxide rich domains in Fig. 3d,e). Aggregates   (Fig. S11). This internal texture is consistent with formation via cooling of melts. In this case, the melts are dominated by Pb, which was a major constituent of the featherbeds used at Taranaki 18 . CeresI shows no evidence of weathering, but CeresIII displays a thin (5-10 µm thick) rim of Pb-O replacing metallic Pb (Figs. 3g,h, 5c,k), indicating localised oxidation (weathering). Inclusions (< 2 µm) of Pu-U-carbide are scattered throughout CeresIII (Figs. 3i, 5a); within the Pb-O weathering rim, the Pu-rich particles are finer (< 1 µm) and consist of Pu-U-oxides (Figs. 3g,h, 5b). Two types of porosity are present: the Pb-O weathering rim contains numerous cracks that extend into the metallic lead core (red arrows in Figs. 3g,h, 5c); and cracks surrounding Pu-U-carbide inclusions within unaltered metallic Pb are suggestive of radiation damage (Fig. 3i).

Discussion
Evolving mobility pathways. Liberation of Pu from hot particles was proposed to play a significant role in explaining the continuous exposure of animals at Maralinga 6 . Previous micro-characterization of Maralinga hot particles was limited to non-destructive µm-scale studies via PIXE and µXRF/µXAS 2,5,7 . Potatohead is similar in texture (fragile, highly porous) and composition (Pu present mainly in (Pu,U)O 2±x like form) to the particle studied by Ikeda-Ohno 2 ; however, our multi-scale and multi-method characterization of diverse hot particles invites a revisit of the implications of earlier results for the fate of Pu at Maralinga: our results reveal a great variability in the nature and internal make-up of hot particles; they provide direct evidence for decoupling of U and Pu geochemistry during weathering at Maralinga; and identify that some particles contain a significant amount of Pu in low-valence state, which is unexpected for particles that survived for ~ 30 years in the environment. Here, we start by discussing the significance of these results with respect to the chemical and physical factors, summarised in Fig. 6, that contribute to Pu liberation from the hot particles. Such observations provide a mechanistic foundation for predicting the future evolution of the particles and likely exposure pathways 6 . Finally, we emphasize how the nature of the particles is a direct result of the source material and their mode of formation 1 , which is revealed to be via cooling of polymetallic melts by textural and compositional data, and we discuss the broader implications of these results for high temperature sub-critical nuclear incidents.
Physical breakdown. The different particles have contrasting bulk composition and internal make-up; these differences probably reflect differences in the source term of individual tests 7 and correlates with the loca- www.nature.com/scientificreports/ tions of the particles (Fig. S1). Maralinga hot particles have long been recognised to be friable; for example, Ikeda-Ohno's 2 particle broke into smaller pieces during gentle handing, as did our Chip particle. The FIB-SEM data reveal that this fragility can result from a combination of factors. (i) Bruce and Potatohead are intrinsically friable as a result of their chemical and physical heterogeneity down to the nm-scale. This phase heterogeneity results in joints and fractures providing points of mechanical weakness within the particle. Potatohead also features a high amount of porosity (Figs. 1a, 3a). (ii) Further mechanical weakening and fracturing are caused by chemical weathering, which induces fracturing in the Pb-rich particles (Figs. 3g,h, 5c). (iii) Finally, radiation damage also contributes to fracturing and increasing porosity (Fig. 3i). 239 Pu (weapons grade Pu used during Vixen B consisted of ~ 95% 239 Pu; Table S1) generates 3 × 10 9 α-particles g −1 s −1 ; the porous halos around the Purich particles extend radially around 5 µm (Fig. 3i), suggesting that each 5 meV alpha particle travels ~ 5 µm in the Pb-rich matrix causing damage along its path 19 . Phase A also shows internal fracturing (Fig. 3e) and porosity (Fig. 4f), that could be due to relaxation of internal stress during cooling, chemical weathering, or radiation damage. Mechanical weathering is expected to dominate in deserts; the Maralinga nuclear test site is an arid environment (monthly mean temperatures range from 13 to 25 °C), with cool winters with overnight frosts (minimum − 3 °C), hot summers with temperatures frequently exceeding 40 °C, and large diurnal fluctuations in temperature 6,18,20 . In this context, the fact that Pu is mostly hosted in phases with particle size that range from ~ 5 to < 100 nm means that physical weathering will release particles that can travel in dust (particles less than 7-10 µm are identified as a respirable risk) or in groundwater. In addition to increasing mobility, the reduction in size of Pu particles increases the radiological risks significantly, as the probability of exposure increases, since α-emitters such as Pu need to be located at the surface of the particle to cause an immediate radiological risk; hence, the effective dose increases as the size of the particles decreases. The increasing surface area with decreasing particle size also enhances the susceptibility to chemical weathering.

Chemical weathering.
Our results confirm the mostly refractory nature of Pu in the hot particles 1 . This is in contrast with U, which shows significant in-particle mobility. Some of the U in Potatohead is in the form of uranyl (U(VI)O 2 2+ ) (Fig. 2d), and µSXRF maps show evidence for some U redistribution in Potatohead in the www.nature.com/scientificreports/ form of (i) a rim enriched in U relative to Pu (arrow in Fig. 1c) and a U(VI)-rich crystal growing on the outside of the particle (Figs. 1a-c, 2d). In general, U(IV) phases are insoluble at room temperature, and U mobility is mainly related to U(VI) uranyl complexes (e.g., aqua, carbonate, sulfate, and chloride 21 ). Similarly, Ikeda-Ohno et al. 2 suggested that U was leached out of the single particle they studied, based on µSXRF data showing Ca + U enrichment on its periphery. Based on µSXRF and µXANES (predominance of (Pu,U)O 2±x ) data, Potatohead is similar to Ikeda-Ohno et al. 's 2 particle; however, the FIB-SEM data show that Potatohead does not show the core-shell structure inferred by Ikeda-Ohno et al. 2 . Since the high-resolution SEM images do not show evidence of hydrous weathering, even along grain boundaries or inside pores (Fig. 3b,c), U oxidation and mobilization likely must result from limited capillary action through this porous particle. Calcium most probably reflects precipitation of calcite ('calcrete') 22 on the outside of the particle, rather than Ca substitution of U/Pu during (U,Pu) O 2±x leaching. Similarly, EDS maps of the CeresIII particle show an enrichment in Ca-C-Si on the outside of the particle, reflecting a thin coating of calcite (Ca,C) and some silicate/clay mineral (Si) (Fig. 5e-l). A core-shell style chemical weathering was observed in CeresIII (Figs. 3g-i, 5), replacing the Pb alloy with a Pb-O phase. This oxidation increased porosity, and Pu speciation changed from low-valence Pu in the core to Pu-O compounds in the weathered rim, concomitant with a decrease in size of the Pu-rich particles. However, despite these changes in Pu speciation, it appears that little or no Pu was released from the particle (e.g., Fig. 5l).
Further, a significant portion of Pu is present in low valence form in Bruce and CeresIII. Although uranium carbide (UC) has been found in depleted U particles from Kosovo and Kuwait 23 , this is the first report of Pucarbide and Pu-bearing alloys in hot particles retrieved from the environment. Low valence Pu-U-carbide phases are typically pyrophoric at µm-grain size in contact with water or molecular oxygen 24 , yet these phases were retained despite ~ 30 years exposure to an arid environment and then another ~ 30 years stored at ambient conditions. These phases were probably protected by their inclusion in Fe-Al alloy (Bruce) and metallic Pb (CeresIII), as well as by the presence of impurities. In CeresIII, Pu-U-carbide exists within metallic Pb, but a Pu-U-oxide is predominant within the oxidised rim (Fig. 5c). The properties of actinide carbides are strongly dependent on minor or trace elements in their crystal structure 24 ; this may further stabilize the carbide phases. Indeed, there was no evidence of oxidation of the Pu-U-carbide phases in Bruce following exposure to atmospheric, ambient conditions for around 10 months (Fig. S12). Hence, Pu-rich phases can persist in chemical states that are far from equilibrium with their environment (water, air), as a result of the physical and chemical make-up of the hot particles.

Ingestion and modification by stomach acids. Aside from mechanical breakdown and chemi-
cal weathering by exposure to moisture and atmosphere, hot particles may be transformed via ingestion by animals 6,20 . Leaching tests using 0.16 M HCl to simulate stomach acids found that just ~ 0.20% of 239+240 Pu was leached in ~ 200 h from a millimeter-sized particle with a glassy structure from Ground zero, Semipalatinsk, former USSR 1 , in line with the expected refractory nature of Pu hot particles. However, similar tests on six particles recovered from Taranaki (containing < 20 wt% PuO 2 ; other metals present in significant concentrations were U, Fe, Cr) found solubilities varying from 1 to 96% over 40 days 5 . Note that compared to the simple dissolutions tests conducted by 1,5 , stomach acids also contain enzymes that will affect hot particle behaviour, but the FIB-SEM results can explain the wide range in behaviour in HCl-leaching experiments: (i) although (Pu,U) O 2±x would be poorly soluble in these tests, dissolution of the Al-oxide-rich domains binding the individual µm-to nm-sized (Pu,U)O 2±x grains in Potatohead-type particles may release the Pu-rich grains, some of which have sizes < 0.45 µm; similarly dissolution of Al-and Fe-rich phases in Bruce-type particles would release Pu-rich nanoparticles. Dissolution of the Pb/PbO-rich particles is limited by the low solubility of PbCl 2 (s). Fig. 6, the weathering processes discussed above may result in the release of Pu-rich particles ≤ 5 µm from the hot particles. As Pu exists in micro-to nano-particulate form in all studied particles, despite their different chemistries, Pu liberation does not require a chemical process; simple mechanical breakdown will release nanoparticles. All released Pu-rich particles cause an inhalation risk, and the finer fraction (nanoparticles) can be transported into the groundwater during rainfall events. Nano-particlefacilitated transport of Pu in groundwater was first identified at the Nevada test site 3,25 and more recently at (i) the weapon-producing sites of the Mayak Production Association, Urals, Russia 26 ; Rocky Flats, Colorado, USA 27 and Hanford, Washington, USA 28 ; (ii) at the nuclear power sites at Sellafield/Windscale, UK; Chernobyl, Ukraine; and Marcoule, France 28 ; and (iii) in rivers waters related to the Grimsel test site, Switzerland 28 . The arid environment at Maralinga is conductive to both dust and water-borne nano-particulate transport; although rainfall is limited (~ 240 mm/year), it is often of a convective nature and characterised by high-intensity, shortduration events (≥ 3 days with > 10 mm rainfall per year 29 ). Although direct evidence of nano-particle transport is still lacking at the ultra-dilute conditions at Maralinga, direct imaging of the nature of the Pu-rich phases in the hot particles and their behaviour during weathering suggest that this may be a primary cause for wild-life exposure 6 .

Particulate transport. As summarised in
From fire to dust. Once liberated from the hot particles, the environmental behaviour of Pu is governed by complex processes 3,30 involving solubility, hydrolysis, complexation and sorption (with both inorganic and organic ligand and phases 31,32 ) and nano-particle (colloid) formation reactions (all possibly catalysed by microbiota 33 ). Yet, the potential for Pu to migrate through the soil environment and enter the food chain, and the resulting risk to biota (Fig. 6), can be estimated using radioecological models. These models need to consider not only the amounts of radioactive material released to the environment following an accident, but also the physical and chemical characteristics of the contaminant and their potential changes through time in order to determine the long-term impacts arising from contamination 1,3,34-37 . Based on non-destructive micro-analytical www.nature.com/scientificreports/ characterization, hot particles from sub-critical nuclear incidents across the globe are chemically and texturally heterogeneous (Table S5). This heterogeneity is hindering their inclusion in radioecological models that are used to predict long-term risks 38 . At Maralinga, the particles contain Pu (and U) in the form of high temperature, anhydrous phases, that are far from equilibrium with respect to environmental conditions. Textural and phase relationship considerations reveal that all studied particles formed via cooling of polymetallic melts resulting from fissile material mixing with the hot detonation environment. The Pb, Fe and Al present in the particles reflect the composition of the individual devices and detonation characteristics. Unfortunately, no information is available in public records on the specifics of the designs and materials of individual tests. Most Pu is hosted in nano-phases that crystallised during the cooling of these polymetallic melts, and, consequently, the micro-to nano-particulate nature of the Pu in these hot particles, regardless of their bulk composition, is an intrinsic result of their formation via cooling of micro-droplets of polymetallic melt 17 (Figs. 3d, 4). The hot, anhydrous micro-environment under which the particles condense in the explosion cloud also accounts for the crystallization of phases that contain Pu (and U) in low valence state (carbides; Pu in Fe-(Al)-alloys). Sub-solidus reactions (e.g. Bruce; arrow in Figs. 3e, 4c,d) and weathering (CeresIII, Figs. 3g,h, 5) further contributed to the generation of fine Pu-rich nanoparticles (< 100 nm) subsequent to cooling.
The recognition of the nature and internal make-up of the hot particles has important consequences for the cycling of Pu in the arid environment at Maralinga. As summarised in Fig. 6, liberation of micro-to nanoparticulate Pu is promoted over time via mechanical breakdown, facilitated by the heterogenous and/or highly porous nature of the particles and thermal cycling in the environment. In addition, near-surface particles have a high probability of ingestion by soil biota and higher animals. Dissolution via digestion of the reactive, metallic matrix of the particle (Fe-, Al-, or Pb-rich alloys) could result in the further release of Pu-rich nanoparticles. Compared to the parent hot particles, these micro-(≪ 5 µm) and nanoparticles have additional toxicity risks due to their small sizes, the low-valence state of Pu and their mobility in dust and groundwater.
The Vixen B trials at Maralinga were designed to simulate sub-critical nuclear incidents, i.e. accidents where energy is released from external fires or conventional explosives. Since these tests, the world has documented a few instances of such incidents (Table S5, SI-Previous studies), including the B-52 accidents that resulted in the conventional detonation of thermonuclear weapons near Palomares, Spain 39 and Thule, Greenland 39,40 ; and the explosion of an armed nuclear missile and subsequent fire at the McGuire Air Force Base, USA 9 . Hot particles from these sub-critical nuclear incidents share the chemical and textural heterogeneity of the Maralinga particles as a result of the similar mode of formation from polymetallic melts. In particular, the distinctive texture of Potatohead (Fig. 3a-c) is remarkably similar to that of the particles from Palomares 39 and Thule 40 ; and hot particles from the McGuire Air Force Base accident have µSXRF images and Pu-U correlations similar to either Bruce and Potatohead 9 . In summary, hot particles released via high energy sub-critical incidents acquire their compositions and textures at high temperature within the explosion cloud, and this mode of formation sets the scene for their long-term environmental behaviour 1 . We note that hot particles released following high energy failure of containment in nuclear reactors, such as Chernobyl 17,41,42 and Fukushima 43 , also show heterogeneous compositions, with U ± Pu associated with Zr and other metals from the reactor cladding [43][44][45] . These particles share a formation via cooling of high temperature melts, however in this case melting occurs in the nuclear fuel rather than in the explosion cloud. In addition, aerodynamic silica fallout particles generated in near surface critical nuclear tests were found to contain a heterogeneous distribution of elements, related to environmental and device material being incorporated and fractionated in the fireball [46][47][48] .
Between 1950 and 1988 alone, there were more than 230 recorded nuclear weapon accidents, including at least 10 with documented release of radioactive particles into the environment 49 . The risks of such incidents are only increasing as international treaties such as the Intermediate-Range Nuclear Forces Treaty and NewSTART come to an end. Yet, there is currently no international best practice for the inclusion of Pu-U rich hot particles released during sub-critical incidents in environmental impact assessment or risk characterization. The new observations on the hot particles from the Maralinga tests provide a clear explanation for the heterogeneous behaviour of different hot particles with respect to chemical and physical weathering 5,41 that has hindered predictive modelling. Bulk characterization of the hot particles provides limited information about the nature and heterogeneity of the particles 2 . This issue is alleviated by the use of FIB-SEM; nano-scale chemical and textural characterization exposes the diversity of hot particles; allows identification of the different weathering pathways in historic particles; and provides the fundamental information for predicting the future behaviour of the particles and the radioecological risk to humans and non-human biota. www.nature.com/scientificreports/