The role of nano-perovskite in the negligible thorium release in seawater from Greek bauxite residue (red mud)

We present new data about the chemical and structural characteristics of bauxite residue (BR) from Greek Al industry, using a combination of microscopic, analytical, and spectroscopic techniques. SEM-EDS indicated a homogeneous dominant “Al-Fe-Ca-Ti-Si-Na-Cr matrix”, appearing at the microscale. The bulk chemical analyses showed considerable levels of Th (111 μg g−1), along with minor U (15 μg g−1), which are responsible for radioactivity (355 and 133 Bq kg−1 for 232Th and 238U, respectively) with a total dose rate of 295 nGy h−1. Leaching experiments, in conjunction with SF-ICP-MS, using Mediterranean seawater from Greece, indicated significant release of V, depending on S/L ratio, and negligible release of Th at least after 12 months leaching. STEM-EDS/EELS & HR-STEM-HAADF study of the leached BR at the nanoscale revealed that the significant immobility of Th4+ is due to its incorporation into an insoluble perovskite-type phase with major composition of Ca0.8Na0.2TiO3 and crystallites observed in nanoscale. The Th LIII-edge EXAFS spectra demonstrated that Th4+ ions, which are hosted in this novel nano-perovskite of BR, occupy Ca2+ sites, rather than Ti4+ sites. That is most likely the reason of no Th release in Mediterranean seawater.


Results & Discussion
Bulk and microscale characterization. Concerning the raw BR sample Fe, Al, Ca, Ti, Na, Si, O and H, are the elements corresponding to solids detected by powder X-ray diffraction/PXRD (see Supplementary Information and Supplementary Fig. S1), while phases related to other elements of interest, such as actinides (namely Th), Cr, V, As, Pb, as well as REEs, are not detectable. The same applies for the seawater-and the acidtreated BR samples. However, Cr, along with the above major elements, is additionally indicated by scanning electron microscopy (SEM) equipped with an energy dispersive spectrometry (EDS), into an "Al-Fe-Ca-Ti-Si-Na-Cr matrix" appearing in a few μ m size (Fig. 1). Since all the micro areas of the raw sample, checked by SEM-EDS, follow the same spectral pattern and, therefore, it is argued that the raw BR is apparently rather homogeneous in terms of composition at the microscale. Taking into account the PXRD patterns, it could be assumed that this matrix is an admixture of nanocrystalline phases and potential amorphous and/or poorly crystalline phases.
The wavelength dispersive X-ray fluorescence (WDXRF) and, complementary, the inductively coupled plasma mass and optical emission spectroscopic measurements (ICP-MS/OES) of the studied BR proved that major Fe, Al, Ca, Si, Ti, Na and C, as well as significant volatiles (loss on ignition/LOI) comprise 86.5 wt.% and 13.6 wt.% of its chemical composition, respectively. In addition, when the trace elements' concentrations of BR are normalized to the average values of Greek bauxites (n = 16) 28 and of Upper Continental Crust (UCC) e.g. 28 and references therein , depletion in Ga (following the chemical behavior of Al) and enrichment -among others-in Cr (2403 μ g g −1 ), V (1081 μ g g −1 ), Ni (902 μ g g −1 ), As (164 μ g g −1 ), Pb (120 μ g g −1 ), Sc (114 μ g g −1 ) as well as in Cd, but also in Nb, Y, Ta, and REEs (Supplementary Table S3 and Supplementary Fig. S2) is pointed out. Moreover, high levels of Th (111 μ g g −1 ) have been determined using high-resolution (HR) γ -ray spectroscopy; rather higher than those reported in the literature 28,39 and references therein. The latter, and also minor U content (15 μ g g −1 ), are responsible for the radioactivity of the studied BR (355 Bq kg −1 for 232 Th and 133 Bq kg −1 for 238 U) with a total dose rate of 295 nGy h −1 (see Supplementary Table S4). It should be mentioned that the radioactivity of Greek BR due to Th is in the Scientific RepoRts | 6:21737 | DOI: 10.1038/srep21737 same range from other BR reservoirs around the world, with the exception of the Australian BR, reaching up to 1900 Bq kg −1 (see Supplementary Table S4 and references therein). The studied BR has higher Th radioactivity compared to the average of the studied Greek bauxites from Parnassos-Ghiona mines (avg. 182 Bq kg −1 ; avg. total dose rate: 162 nGy h −1 ; n = 10). The same stands for radioactivity due to U, whereas both materials exhibit similar values due to 40 K.

Leaching experiments.
Leaching experiments for the studied BR, in conjunction with sector field (SF) ICP-MS spectroscopy using Mediterranean seawater from Greece indicated significant release of V, relatively to seawater composition 40 , depending on S/L ratio (Fig. 2). Thus, V appears to be the most mobile element, perhaps due to its major association with rather soluble phases. This is in accordance with previous leaching NEN7341 tests concerning BR calcined in inert and reducing atmosphere 11 , as well as in good agreement with recent studies about V in Hungarian BR 1,7 . According to the latter studies, presenting X-ray absorption near-edge structure (XANES) spectroscopic data of vanadium (V K-edge), V in Hungarian BR is pentavalent associated with a Ca-Al-hydroxysilicate phase, corresponding to "hydrogarnet"-type phase. We can, therefore, assume that in Greek BR, having the same phase composition as the Hungarian BR, V is present most likely as VO 4 3− anions, known to be mobile in seawater. On the other hand, V in Australian BR has been reported as tetravalent and/or trivalent related to ilmenite and/or goethite 41 . However, ilmenite has not been detected in Greek BR so far using PXRD, SEM or TEM (see text below), whereas the actual presence of goethite, either at the microscale or at the nanoscale, cannot be confirmed with certainty. Traces of As and Cr have also been detected, relatively to seawater composition 40 , after prolonged exposure in lower solid-to-liquid (S/L) ratio, while Pb is practically immobile. Based on previous Cr K-edge XANES and TEM studies 7,41 Cr is associated with hematite showing less solubility. The most interesting point is that Th -related to radioactivity-seems to be immobile in seawater, at least after 12 months of leaching tests (Fig. 2). However, taking into account the PXRD study and the SEM-EDS investigation of the BR (see text above), it can be stated that there is no clear evidence of the Th-hosting phase(s) in the Greek BR at the microscale. That was finally realized combining the data yielded by both the scanning transmission electron microscopy-energy dispersive spectroscopy (STEM-EDS) and the electron energy-loss spectroscopy (EELS) together with the Th L III -edge EXAFS spectroscopic measurements on the studied BR (see text below). Analogous leaching experiments with the basic parent material (typical low-grade and high-grade Greek industrial bauxite) in Mediterranean seawater from Greece were carried out for comparative reasons; they showed negligible V, As, Cr, Pb, and Th release ( Supplementary Fig. S3). The effect of S/L ratio is also depicted, indicating that lower S/L ratio causes a relative increase on the solubilization of heavy metals in the studied bauxites, during the experiments. Additional acetic acid-leaching experiments with the BR indicated significant release of V, as in the case of seawater, and much higher release of Cr ( Supplementary Figs S4 and S5), a phenomenon also mentioned in the literature 31 . Nevertheless, it is worthy to note that the high release of REEs from BR ( Supplementary Fig. S5), giving facts for potential recovery technologies inasmuch future availability of REEs is of great concern due to monopolistic supply conditions, environmentally unsustainable mining practices, and rapid demand growth 42 . This is in line with previous relevant works for Greek BR with regard to acid-recovery of lanthanides and Y [21][22][23][24] . In either case, the negligible effect of S/L ratio is also presented, indicating that the percentage of released elements -including the REEs-from leached BR is not severely affected by the S/L, during the experiment. Similar leaching experiments with typical low-grade (Fe-rich) and high-grade (Fe-depleted) Parnassos-Ghiona bauxites, presented for the first time in the literature, have also indicated significant recovery of REEs ( Supplementary  Figs S6 and S7). In the case of bauxites, the mobility of REE in acid is due to the presence of light REE (LREE) minerals (mostly bastnäsite/parisite-group) as reported previously 27,28 . Despite the fact that the mobility of Th in acid-treated bauxite residue is relatively enhanced, as compared to its negligible mobilization in seawater environment, considerable amount of this actinide element seems to remain in Greek BR. This robustly proves that the aforementioned acid-insoluble as well as seawater-insoluble solid nanophase (not apparent in the SEM-EDS observations) is hosting immobile Th into the "Al-Fe-Ca-Ti-Si-Na-Cr matrix". The difference at the microscale between the initial (raw) and acid-treated "Al-Fe-Ca-Ti-Si-Na-Cr matrix" does affect mainly the lowering of Ca-Kα X-ray emission peak in the EDS spectra ( Fig. 1 and Supplementary Fig. S8) attributed to the loss of Ca-carbonate minerals and Ca-Al-hydroxysilicate phases.
Nanoscale characterization. Except for the aforementioned observed solid phases at the microscale (see text above) calcium titanium oxide (CTO) phases are expected to be produced due to the addition of lime during the Bayer process 43 , under low-T and low-P conditions. Regarding CTO in Greek BR, a calcium titanium oxide phase previously identified by bulk PXRD 32 , was arbitrarily reported as "perovskite" 20,43 . However, perovskite was detected by PXRD in BR from Zhengzhou Changcheng alumina plant in China 43,44 , from the Aughinish alumina plant in Ireland 43,45,46 , and from India 43,47 . On the other hand, Santini (2015) 48 in a similar study with regard to BR from USA, determined only sodalite-and cancrinite-phases but not any crystalline CTO and/or perovskite. Furthermore, conventional TEM "photographs" showed perovskite in BR from China 49 without any structural proof.
According to the STEM-EDS/EELS study (Figs 3 and 4 and Supplementary Figs S9-S12), the most motivating of the Ti-phases constituting the microscale "Al-Fe-Ca-Ti-Si-Na-Cr matrix" of the Greek BR is an unusual low-T and low-P perovskite-type phase e.g. 50 occurring in nanoscale (Fig. 3a,b). The major composition of this nano-perovskite has been determined by TEM-EDS to be Ca 0.8 Na 0.2 TiO 3 . Since several other minor elements (such as Ce, Nb, Zr, Cr and, possibly, Sr) might additionally be hosted as impurities (Fig. 3c) that might lead us to argue about a Ca-Na-(Ce-Nb-Zr-Cr)-nano-perovskite. This argument implies a minor contribution of Ce-loparite-type phase (Ce(Ti,Nb)O 3 ). The most important finding is that this nano-perovskite contains Th (Fig. 3c), which is estimated to be 700 μ g g −1 . The STEM imaging, generated by the high angle annular dark field (HAADF) detector, together with the STEM-EDS mapping (Fig. 3d), clearly showed the distribution of key elements (such as Na, Ca, Ti, and O) for the Th-hosting nano-perovskite (CTO) that co-exists with other neighboring mineral phases (hematite and clay-like phases; see text below) into the "Al-Fe-Ca-Ti-Si-Na-Cr matrix". In the depicted EDS maps Fe is exclusively attributed to the neighboring hematite, while Al is due to the clay-like phase. All the obtained selected area electron diffraction (SAED) patterns from nano-perovskite correspond to a perovskite structure [lattice parameters are a: 0.562 ( ± 0.02) nm, b: 0.752( ± 0.02) nm and c: 0.561 ( ± 0.02) nm, Pnma) with superlattice reflections (Fig. 3a)]. Since these reflections do not exist in the conventional perovskite we assume that they might be associated with the substitution of Na on A-site. It has recently reported that Na may incorporate into A-site of a ABO 3 and BST perovskite, when synthesized at high-T or at high-P 51-53 , but that has never been observed at low-T and low-P perovskites. According to recent discussion by Seki et al. (2014) 53 , Na-ions could be linked with tilting of TiO 6 octahedra, causing significantly stronger superlattice reflections in Ca-containing perovskites. Additional structural information on the novel Ca 0.8 Na 0.2 TiO 3 is obtained by EELS measurements (see Supplementary Information and Supplementary Fig. S9). As it has already been mentioned, the BR nano-perovskite occurs together with Th-free Ti-containing hematite particles (Fig. 4). The concentration of Ti in this Ti-containing hematite has been measured up to 5 wt.%. Small amounts of Si, Al, Cr (Fig. 4b) and, probably, P ( Supplementary Fig. S10) also exist in the Ti-containing hematite. Distinct Th-free Ti-oxides were characterized as anatase ( Fig. 4c and Supplementary Fig. S11). The anatase crystallites exhibit rounded shape with a smooth surface. When the STEM-EDS spectra of Th-free Ti-containing hematite and anatase are compared with the EDS spectrum of the Th-containing nano-perovskite the total absence of Th is obvious (Fig. 4d). Concurrently, EDS artifacts like escape peaks for Ti Kα and Ti Kβ are also present. The peak ratio of Th Mα and Th Mβ was found to be about 5:3, which is in good agreement with the experimental profile, although the second broad peak (3.15 KeV) may be partially affected by the Ti Kβ escape peak at 3.18 KeV (Fig. 4d). Besides, several planar defects, common found in anatase, have been demonstrated by high-resolution TEM (HRTEM) imaging (a,b of Supplementary Fig. S11). Subsequently, the STEM-HAADF and HR-STEM-HAADF observations   Supplementary Fig. S9). The almost visible narrow peak of Na is due to background noise. STEM-EDS data for anatase particle, including a BF image and SAED pattern as well as a STEM-EDS spectrum (c). The viewing direction for SAED pattern is [111]. A peak at 2.77 KeV is related to the Si escape peak for Ti Kα , while a peak at 3.19 KeV can be assigned as the Si escape peak for Ti Kβ . Comparison (d) between the (see Supplementary Fig. S11c,d, respectively) have proved, along with the STEM-EDS spectra, the absence of Th in all studied anatase particles. The observed AlOOH polymorphs corresponding to diaspore (α -AlOOH) contain Fe and Cr impurities (Fig. 4e). Moreover, Th-free clay-like phases (with major d-spacings corresponding to 0.31 nm, 0.28 nm, 0.26 nm and 0.17 nm; Fig. 4f) have also been confirmed as components in the "Al-Fe-Ca-Ti-Si-Na-Cr matrix". The obtained Debye-Scherrer ring patterns of clay-like particles suggest that this phase could be related to a zeolite-type material, which is rather doubtful due to its unusual chemical composition (Si: 19.87 wt.%; Al: 10.71 wt.%; Ca: 10.94 wt.%; Ti:20.63 wt.%; O: 37.81 wt.% as major elements and C, P and S as minor elements). We might also assume that this phase could be an admixture of several nano-phases. Iron and chromium peaks attributed to the presence of Ti-containing hematite crystallites, as these Fe-oxides, are often associated with clay-like phases and nano-perovskite into the "Al-Fe-Ca-Ti-Si-Na-Cr matrix" (Supplementary Fig. S12). Thus, the STEM-EDS/EELS study revealed that both the initial and acid-treated "Al-Fe-Ca-Ti-Si-Na-Cr matrix", as appeared at the microscale, are actually an aggregate of several Al-, Fe-, Ca-, Ti-, Si-, Na-, Cr-(and also S-, P-, Ce-, Nb-, Zr-, and maybe, Sr-) particles, including the Th-hosting nano-perovskite. There is no evidence for other phases hosting Th, neither at the microscale (see text above) nor at the nanoscale, except for the nano-perovskite, but the minor presence of element traces, in isolated parts of various BR phases, might not be excluded.

EXAFS spectroscopy and the local environment of Th. Details on the structural characteristics of
Th, assigned to the discovered low-T and low-P nano-perovskite (Ca 0.8 Na 0.2 TiO 3 ), have been obtained by X-ray absorption fine structure spectroscopy/XAFS (Th L III -edge XANES and, especially, EXAFS) spectra and data processing using the ATHENA 54 and EDA 55 software packages. The Th L III -edge bulk XANES results of the studied BR, together with bulk & micro-XANES of Greek industrial bauxite 27,28 in comparison with spectra of reference materials indicated that the valence of Th is typical 4+ (Supplementary Fig. S13). The experimental Th L III -edge EXAFS spectrum of the BR, which was measured in the fluorescence mode, is noisy above k = 6 Å −1 (Supplementary Fig. S14). This fact limits the accuracy of the analysis, as well as it reduces its resolution in the R-space. The EXAFS spectrum does not show any evidence for significant high frequency contributions either. This fact is compatible with the shape of the EXAFS spectrum Fourier transform (Supplementary Figure S14), which consists of a single broad peak located at 1.6 Å. Thus, the contribution of outer coordination shells around thorium is smeared out, and only analysis of the nearest environment can be performed. The first peak contribution into the total EXAFS spectrum was singled out by the Fourier filtering procedure and best fitted using two different approaches: the one-component Gaussian model 56 and the regularization method 57 . It is considerable that in the latter method, the radial distribution function (RDF) could have had an arbitrary shape and, thus, may possibly account for anharmonicity and strong disorder effects. In both models, the theoretical backscattering amplitude and phases shift functions for the Th-O atom pair were calculated by the ab initio real-space FEFF8 code 58 , employing a complex Hedin-Lundqvist exchange-correlation potentially accounting for inelastic effects. The results of the best fits, obtained within the two models (i.e., Gaussian and regularization method) in the k-space range from 2 to 6 Å −1 , are shown in the upper and middle images of Fig. 5, respectively, and the corresponding RDF's of them are given in the lower image of Fig. 5. The regularization method results in better agreement, suggesting a deviation of the RDF shape from the Gaussian form. The numerical values of structural parameters for both models are given in Table 1. For the sake of discussion, it is noteworthy that, due to the short k-range interval of the EXAFS spectrum, there is a strong correlation between the values of the coordination number N and the Debye-Waller factor σ 2 . The local environment of Th ions in the studied BR is not very far from that in Fe-depleted bauxite 27,28 , but is significantly more disordered as is evidenced by the larger value of Debye-Waller factor in the Gaussian model and broad RDF's in Fig. 5. As in the case of Fe-depleted bauxite 27,28 the Th ions in BR are coordinated by about 7-8 oxygen atoms. This coordination number of Th is smaller than that in Th-nitrate (N = 12), being in agreement with the difference in the mean Th-O distances. However, it is notable that in the case of bauxite, being the basic parent material of BR, Th is hosted in microscale anatase (TiO 2 polymorph) and there is no evidence for any perovskite phase. It is known that Th 4+ in the structure of CaSiO 3 perovskite, synthesized at high temperature (and occasionally at high pressure), may fundamentally occupy "large" Ca dodecahedral sites ( [12] Ca 2+ ) or "small" Si octahedral sites ( [6] Si 4+ ) 59 . On the other hand, in the structure of CaTiO 3 perovskite 60 Ti appears as [6] Ti 4+ , whereas Ca may be [8] Ca 2+ or even [7] Ca 2+ , compared to [12] Ca 2+ in the ideal perovskite structure. As one can see in the lower image of Fig. 5, the distribution of the Ca-O distances in CaTiO 3 perovskite agrees well with the shape of the Th-O RDF in BR, obtained by the regularization method. Consequently, the Th L III -edge EXAFS signal was simulated by the FEFF8 code for the case of Th-absorbing atom substituting calcium or titanium in CaTiO 3 perovskite. The calculated Th L III -edge EXAFS signals are compared with the experimental one for BR; see upper image of Fig. 6). In the frame of this simulation, the atom positions were fixed as in orthorhombic CaTiO 3 perovskite-type structure 60 , and all Debye-Waller factors were set to zero STEM-EDS spectra of Th-hosting nano-perovskite (a), Th-free anatase (b) and Th-free Ti-containing hematite (c). The Th Mα and Th Mβ peaks are demonstrated (red lines) together with the STEM-EDS artefacts of the escape peaks for Ti Kα (2.77 KeV) and Ti Kβ (3.18 KeV). The theoretical intensity ratio of Th Mα to Th Mβ is 5:3. A peak at 2.34 KeV KeV may be associated with S (S Kα ), while a peak at 3.7 KeV is assigned to Ca (Ca Kα ). STEM-EDS data for AlOOH phase with a diaspore structure, including a BF image and a SAED pattern as well as a STEM-EDS spectrum (e). The viewing direction for its SAED pattern is [100]. STEM-EDS data for Th-free clay-like phases, including a BF image and its Debye-Scherrer ring patterns as well as a STEM-EDS spectrum (f). A peak at 2.77 KeV is related to the Si escape peak for Ti Kα . The almost visible narrow peak of Na is due to background noise.
Scientific RepoRts | 6:21737 | DOI: 10.1038/srep21737 that explains smaller damping of the calculated EXAFS amplitude at larger k-values. As one can see (upper image of Fig. 6), the model of Th at the Ca site (Ca(Th)TiO 3 ) results in the overall amplitude and main frequency of the EXAFS signal close to the experimental one, whereas the model of Th at the Ti site (CaTi(Th)O 3 ) differs from the experiment significantly, in both amplitude and frequency. Unfortunately, the weak contribution of the outer shells ( Supplementary Fig. S14) in the experimental Th L III -edge EXAFS spectrum in BR does not allow us to make unambiguous conclusion on the Th location. However, taking into account the above argument about Th in Ca site of the nano-perovskite, in conjunction with the best-fit EXAFS results (R = 2.42-2.48 Å and CN = 7-8), we could presume that Th, hosted in the low-T and low-P novel Ca-Na-(Ce-Nb-Zr-Cr)-nano-perovskite of the studied BR, occupies Ca 2+ sites rather than Ti 4+ sites (lower image of Fig. 6).

Environmental and technological implications about the Greek bauxite residue (red mud). It is
herein stated that the novel low-T and low-P Th-hosting Ca-Na-(Ce-Nb-Zr-Cr)-nano-perovskite (Ca 0.8 Na 0.2 TiO 3 ) is the reason of the low Th release in acid medium, and subsequently of the Th immobility if the Greek bauxite   27,28 , obtained from the processing of the EXAFS signals using the EDA software package 55 . * This distance is the position of the RDF maximum. residue/BR (i.e., red mud) is exposed to Mediterranean seawater. In general, it is demonstrated that Greek BR, and perhaps huge quantities of BR around the world, do not consitute, under certain circumstances an environmental hazard due to actinide content and radioactivity. In turn, BR is a valuable alumina refineries' by-product to be used as secondary resource for a sustainable supply of critical & strategic elements (such as REEs) and, thus contribute to a more sustainable "modus operandi".

Conclusions
The nature of thorium in the bauxite residue/BR (the so-called "red mud") from Greek Al industry has been investigated in detail for the first time in the literature. The interest arises from the fact that huge quantities of that BR, showing relative radioactivity, had been deposited into the Mediterranean Sea in Greece. The chemical analysis and the HR γ -ray measurements proved that the observed higher radioactivity is higher compared to the parent Greek bauxite, mainly attributed to the presence of Th (111 μ g g −1 ; 355 Bq kg −1 for 232 Th). The marine environment might not be affected by this actinide element; leaching experiments have confirmed negligible Th release in Mediterranean seawater, at least after 12 months of interaction. In contrast, the experiments of the present study indicated that the mobility of V might be of potential risk. Characterizing the studied BR in microscale yielded no evidence of Th hosting phase into a homogeneous "Al-Fe-Ca-Ti-Si-Na-Cr matrix. However, the STEM study of the leached BR sample at the nanoscale showed that the immobility of Th can be attributed to the existence of an insoluble nano-perovskite with major composition of Ca 0.8 Na 0.2 TiO 3 . Subsequent study of Th L III -edge EXAFS spectroscopy revealed that the local environment of Th 4+ in the structure of nano-perovskite is occupying Ca 2+ sites, rather than Ti 4+ sites. We do consider that this structural peculiarity is related to the negligible Th release in the Mediterranean seawater.

Seawater and Acid-Leaching Experiments (SF-ICP-MS). Leaching experiments on the investigated
BR and bauxite samples were carried out using (a) Mediterranean seawater from the Gulf of Corinth (Greece), and (b) concentrated acetic acid (Merck), over a period ranged from 2 weeks to 1 year. Acetic acid was used, instead of typical TCLP, according to a recent relevant work on red mud 31 . The analyses of leachates for potentially hazardous metals and metalloids (Cr, V, Ni, As, Pb), Th, Ta, Nb, and REE, were performed using a SF-ICP-MS (Thermo Scientific Element 2/XR). For more details, please consult the Supplementary Information and the Supplementary Tables S1 and S2.

Electron Microscopy (SEM-EDS, STEM-EDS/EELS and STEM-HAADF). Details on the elemental
composition of BR were obtained on carbon-coated free surfaces and polished (in epoxy resin) solid samples using a Jeol JSM-5600 SEM equipped with an Oxford EDS. The observations were made with an accelerating voltage of 20 kV, a working distance of 20mm, a current beam of 1.5 nA, an active time of 20-100 s and a magnification from × 35 to × 3500. Pure metallic materials and minerals were used as standards. STEM-EDS/EELS and STEM-HAADF were carried out using an FEI probe-corrected Titan 80-300ST FEG TEM, equipped with an Oxford Instrument X-Max 80 mm 2 silicon drift detector and a Gatan Tridium imaging filter. SAED was used to identify mineral phases. For the detection of Th, we performed TEM-EDS with count rates of 5,000-8,000 counts sec −1 and dwell times of 60-180 sec (i.e., at total counts of 500,000-700,000), using a Gatan low-background Be specimen holder. The EELS measurements were made in TEM diffraction mode with an energy resolution of 0.9 eV.

X-ray Absorption Fine Structure Spectroscopy (XAFS). Bulk XAFS (EXAFS/XANES) spectroscopic
study of BR was performed at the Th L III -edge (16300 eV; energy was calibrated using a Y metal foil) in the fluorescence mode on powdered samples pressed with cellulose into pellets. Spectra were obtained at the SUL-X beamline of the ANKA Synchrotron Radiation Facility (KIT, Germany). Thorium compounds (Th(NO 3 ) 4 • 4H 2 O and ThO 2 ) and minerals containing Th impurities, such as zircon (ZrSiO 4 ), were used as reference materials. The XAFS spectra were analyzed using the ATHENA 54 and the EDA 55 software packages. The EXAFS signal corresponding to the first main peak in Fourier transforms (FTs) was isolated using the Fourier filtering procedure. The range of the back-FT was 1-3 Å for Th(NO 3 ) 4 • 4H 2 O and 0.8-2.5 Å for the samples. Thus, obtained EXAFS signals were simulated using two different approaches: the conventional Gaussian model 56 and the regularization method 57 . The theoretical backscattering amplitude and phase shift functions for Th-O atom pair were calculated by the ab initio FEFF8 code 58 using a complex Hedin-Lundqvist exchange-correlation potential accounting for inelastic effects.