Mechanisms of Uranyl Sequestration by Hydrotalcite

Abstract

Since the advent of large-scale U mining, processing, and enrichment for energy or weapons production, efficient capture and disposal of U, transuranics, and daughter radionuclides has constituted an omnipresent challenge. In this study, we investigated uranyl (UO₂²⁺) sequestration by hydrotalcite (HTC) as a coprecipitation or surface adsorption reaction scenario. The master variables of the study were pH (7.0 and 9.5) and CO₂ content during the reactions (CO₂-rich, CO₂r vs CO₂-depleted, CO₂p). In addition, we compared the outcomes of U–HTC coprecipitation reactions between pristine salt precursors and barren U mine wastewater (lixiviant). Extended X-ray absorption fine structure spectra revealed that uranyl adsorbs on the HTC surface as inner-sphere complexes in CO₂r and CO₂p systems with U–Mg/Al interatomic distances of ∼3.20 and ∼3.35 Å indicative of single-edge (¹E) and double-edge (²E) sharing complexes, respectively. Partial coordination of uranyl by carbonate ligands in CO₂r systems does not appear to hinder surface complexation, suggesting ligand-exchange mechanisms to be operative for the formation of inner-sphere surface complexes. Uranyl symmetry is maintained when coprecipitated with Al and Mg from synthetic or barren lixiviant solutions, precluding incorporation into the HTC lattice. Uranyl ions, however, are surrounded by up to 3–5 Mg/Al atoms in coprecipitated samples interfering with HTC crystal growth. Future research should explore the potential of Fe(II) or Mn(II) to reduce U(VI) to U(V), which is conducive for U incorporation into octahedral crystal lattice positions of the hydroxide sheet.

Introduction

Efficient capture and containment of U, transuranics, daughter radionuclides, and other contaminants from U mining, processing wastewater, and nuclear accidents is an essential component to ensure an environmentally and socially responsible operational lifecycle of the U industry.¹⁻⁴ Although a number of conventional technologies such as ion exchange are available to address one or more contaminants, few technologies are able to efficiently capture a range of geochemically diverse contaminants often present and to rapidly convert them into a stable, solid form. Layered-double hydroxide (LDH) anionic clays, specifically hydrotalcite (HTC, {[Mg_1–x,Al_x](OH)₂}A_x/m–^m–·nH₂O, e.g., A = 1/2SO₄^2–, 1/2CO₃^2– or 1NO₃^–), have been investigated extensively over a wide range of pH, ionic strengths, and with various ligands (EDTA, citrate, carbonate, nitrate, and sulfate), to be effective U scavengers under controlled laboratory settings.⁵⁻¹¹ However, few studies have been able to test HTCs on actual, real-life U-containing samples. Recent research undertaken by CSIRO tested the in situ formation of HTCs as a broad spectrum repository for a range of cationic and anionic contaminant species on Heathgate Resources’ Beverley North (South Australia) in situ recovery (ISR) barren mine lixiviant¹² and on tailings pond water from ERA’s Ranger Mine (Northern Territory¹³). This research and related studies at laboratory^14,15 and industrial scale¹⁶ have demonstrated that the in situ HTC formation, initiated by the tailored addition of Mg or Al salts to produce a suitable M²⁺/M³⁺ molar ratio and the addition of base (OH^–), appears to incorporate a suite of contaminants either in the metal hydroxide layer or as interlayer anions to balance structural charge requirements. These HTCs have sequestered U at up to 1% (w/w) from an initially acidic ISR barren lixiviant containing ca. 20 mg/L U, highlighting the substantial concentration factor achievable.^12,13,17

The suite of cationic, neutral, and anionic uranyl species present across the pH spectrum (Figure S1) during neutralization raises questions as to the nature of the associations between uranyl and HTC and the specific role carbonate ligands may play in affecting the outcome of U solid-phase speciation. Despite the extensive research into the U–HTC reactions at the laboratory scale, to date (and to the best of our knowledge), the molecular-scale bonding environments of U in or on HTCs have not yet been reported. In this study, we present data on the solid-phase speciation of U adsorbed on or coprecipitated with HTC, in CO₂-enriched (CO₂r) or CO₂-depleted (CO₂p) atmospheres at pH 7.0 and 9.5. These results address the yet-unanswered questions about the speciation and fate of U during the reaction with HTC: Is U incorporated into the primary metal hydroxide layers? Is it adsorbed on its surfaces or alternatively does it occur in the HTC interlayers as anionic uranyl carbonate or other anionic complexes? How does atmospheric CO₂ influence the reaction outcomes? Finally, we propose four reaction schemes exploring the effects of CO₂ and pH on uranyl surface complexation and concomitant anion-exchange reactions.

Results and Discussion

Composition of the HTC Solids

The final Al, Mg, and U contents in the HTC solids (Table S3) varied as a function of pH and CO₂r/CO₂p environment. All samples were deficient in Mg relative to the intended M²⁺/M³⁺ ratio of 3:1, leading to an excess positive charge ranging from +0.5 (HTC3 CO₂p) to +0.29 (HTC1 CO₂r). Magnesium dissolved in the HTC3 samples at pH 7.0 causes the Mg/Al ratio to decrease from 2.73 in the pure HTC3 sample to 2.0 in the HTC3 CO₂p sample. As incorporation of Al and U into the solids was similar, one could not assume necessarily that either replaced Mg in the structure extensively, as U uptake was near 100% in nearly all cases. Less than the ideal Mg content in the HTC structure at pH 7.0 may thus have caused the permanent positive charge to increase from 0.37 to 0.50, assuming an intact oxyanion framework (Table S3).

Uranyl uptake from solution ranged from 72 to 100%. Considering only the chemically purer systems (HTC2 and HTC3), the uptake was in all cases essentially complete (97–100%). In the CO₂r atmosphere, only 72% of available U was sequestered by the barren lixiviant HTC1 solid as opposed to 100% uptake into the corresponding CO₂p sample. On the basis of the comparison between the coprecipitation systems of HTC2 and the adsorption systems of HTC3, U retention in the interlayer on the mineral surface and/or incorporation in the metal hydroxide sheets was hence significant.

Structure of the HTC Solids

The HTC1 and HTC2 CO₂r samples bear close resemblance to the prototypical HTC pattern (Mg₆Al₂CO₃(OH)₁₆(H₂O)₄^18,24), and their (003) and (006) spacings (7.85 ± 0.02 Å) are consistent with the mean interlayer distance for HTC intercalated with CO₃²⁵ (Figure S2). The HTC1 and HTC2 CO₂p samples and the pure HTC sample (HTC3 prior to adsorption tests) closely resembled each other, but their interlayer distances ((003) and (006)) occurred at lower deg 2θ and hence greater d-spacing (8.65 ± 0.10 and 8.46 Å, respectively) than for the CO₂r samples consistent with the mixed SO₄/NO₃ interlayers (8.8/8.1 Ų⁵), reflecting the exclusion of CO₂ from the synthesis liquors, the SO₄/NO₃ salts of Mg and Al used in their synthesis and the anion selectivity sequence for HTC and LDH phases determined previously.²⁶ The HTC interlayer distance is furthermore directly related to the Al content (ionic radii of octahedral Al³⁺ and Mg²⁺ are 0.535 and 0.72 Å, respectively²⁷), and the data (Table S4) reflect an expansion of the interlayer with decreasing Al content as previously reported.²⁸ The unit cell length a (=b) ranged between 3.052 (HTC2 CO2p) and 3.080 Å (HTC1 CO₂r) with a noticeable effect from CO₂r atmospheres and higher Al contents on a. The unit cell volumes of the HTC1 and HTC2 CO₂p samples were expanded in comparison to the HTC3 unit cell volume. This expansion cannot be explained on the basis of U incorporation because octahedral U(VI, 0.73 Å) has a comparable ionic/crystal radius to octahedral Mg²⁺ (0.72 Ų⁹).

The X-ray diffraction (XRD) peaks of the prominent (003) and (006) reflections were broad in all samples, indicating the presence of small crystallites. Application of the Scherrer equation to the (003) and (006) reflections indicated a mean crystallite size along the c-axis of 25.1 ± 15.0 nm with a range of 7.5 (HTC1, CO₂p) to 96.0 nm (HTC2, CO₂r, Table S4). Crystallite sizes in the ab plane (120 reflection) indicate variations between 23 nm (HTC1 CO₂r) and 366 nm (HTC3) consistent with previous reports.^30,31

In the barren lixiviant CO₂r HTC1 sample, aragonite, CaCO₃ (Pmcn³²) was identified. This is not unexpected, as the addition of CO₂ and NaOH to a solution containing calcium ions (5.7 mM) can be predicted to precipitate calcium carbonate. Aragonite is a known scavenger of UO₂²⁺ from solution,³³ and hence, extended X-ray absorption fine structure (EXAFS) data in this study might reflect mixed-surface U bonding environments stemming from HTC as well as from aragonite. Evidence for Al(OH)₃ (s) formation, for example, gibbsite, as a result of the low-to-high pH adjustment, was not found in the diffraction patterns.

First Ligand Shells of Uranyl

Complementary U L-III edge X-ray absorption near-edge spectroscopy (XANES) and EXAFS spectra were collected to identify and estimate neighboring atoms in the bonding environment of U in the HTC samples. All spectra contained a shoulder feature on the high-energy side of the white line indicative of multiple photoelectron scattering in the trans-, di-oxo uranyl moiety ({O=U=O}²⁺, UO₂²⁺).³⁴ The primary difference between XANES spectra could be ascribed to the presence or absence of CO₂ in the reaction (Figure S3a). The CO₂p samples (black spectra) relative to the spectra of CO₂r samples were phase-shifted to higher energy noticeable with the first EXAFS oscillation near 17 210 eV (Figure S3a, inset). This phase shift was most pronounced in synthetic, coprecipitated HTC2 samples at pH 9.5 and least pronounced in surface-adsorbed HTC3 samples at pH 7.0. No other discernible differences in the XANES were observed among the samples of CO₂p or CO₂r preparation, suggesting that similar phases were formed within each atmospheric environment (Figure S3b).

The EXAFS spectra of CO₂p and CO₂r samples (Figure 1a,b) revealed significant differences in the oscillatory structure of the sinusoidal wave functions. All wave functions deteriorated in data quality above 11 Å^–1, which could generally be ascribed to low U concentrations in the samples and increasingly less coordinated photoelectron scattering with distance from the central absorber.

Figure 1.

k³-Weighted U L-III EXAFS spectra of adsorption and coprecipitation samples for (a) CO₂r atmospheres and (b) CO₂p atmospheres: 1 = HTC3, pH 7.0; 2 = HTC3, pH 9.5; 3 = HTC2, pH 9.5; and 4 = HTC1, pH 9.5.

The EXAFS of all samples was dominated by a central sinusoidal wave function, which, in the corresponding radial structure functions (RSFs) (Figure 2a,b), led to a dominant peak near 1.4 Å (U–O_yl scattering, Δr + r uncorrected for phase shift). An equatorial ligand shell (O_eq) around uranyl was modeled at distances of ∼2.35 and 2.50 Å for all samples. Neither the distances nor the coordination numbers of O_eq atoms were significantly different among the CO₂r samples (N ± 21%, R ± 3% or 0.07 Å, see Table S5 for EXAFS accuracy determinations), which were distinct because of the interference of a prominent beatnode between 6 and 9 Å^–1 (Figure 2a). The beatnode arises from U–C photoelectron scattering (∼2.85–2.90 Å) due to carbonate ligands coordinating the central uranyl moiety. Nonlinear least-square shell fits (Table S6) confirmed the presence of 1.5 to 2.3 CO₃ ligands (±10%) in the first ligand shell of UO₂²⁺ at a distance of 2.88–2.91 Å. Thus, 1.5–1.6 C atoms in the HTC3-pH9.5 and HTC2-pH9.5 samples can be considered significantly fewer than that in the HTC3-pH7.0 and HTC1-pH9.5 samples (N_C–U ≈ 2.2–2.3). Coordination numbers smaller than 3 suggest the presence of binary mixtures between free uranyl, mono-, bis-, and tris-carbonate species³⁵ (Table S8).

Figure 2.

Corresponding RSFs following Fourier transformation (Bessel window function, β = 4) of spectra shown in Figure 1. (a) CO₂r atmospheres and (b) CO₂p atmospheres: 1 = HTC3, pH 7.0; 2 = HTC3, pH 9.5; 3 = HTC2, pH 9.5; and 4 = HTC1, pH 9.5. The solid lines show the Fourier transform (FT) magnitude and the dashed lines the extracted imaginary part.

Carbonate ions did not coordinate uranyl in the CO₂p samples and produced a concomitant smaller FT magnitude of shells between 1.8 and 2.8 Å (r + Δr, uncorrected for phase shift) in the CO₂p RSFs (Figure 2b). The ligand environment of O_eq atoms between CO₂p spectra differed significantly in the number of next-nearest O neighbors and appeared more disordered than their CO₂r counterparts. Fewer O_eq atoms (4.2 ± 0.9) were present in the HTC3-pH7.0 sample than in the HTC3-pH9.5 sample (7.2 ± 1.5) or the HTC2-pH9.5 sample (6.0 ± 1.3, Table S7). The magnitude of the Debye–Waller parameter (σ) of the fit for HTC3-pH9.5 reflected the greater disorder (unconstrained ∼0.015 Ų) in the CO₂p coprecipitation system. When fixed σ to 0.012 Ų (e.g., σ for cejkaite 1) or 0.007 Ų (e.g., σ for ruthefordine and HTC2-pH9.5, CO₂p), however, the summed N_Oeq dropped to 5.9–7.2, which is well within established accuracies for a sixfold coordinated equatorial ligand environment. Despite these differences in the coordination number, U–O_eq bond lengths were similar between the CO₂p samples and not substantially different from their CO₂r counterparts at ∼2.32 and ∼2.50 Å (Table S7). No evidence could be found for the formation of schoepite despite the slight oversaturated state of the initial conditions of the HTC3 CO₂p (pH 7.0) sample reacted with 10 μM uranyl at pH 7.0.

Second Shells and Higher Order Bonding Environments

The second shell bonding environments of the adsorption, HTC3, samples did not differ as a function of pH or due to the presence of CO₂. Depending on the Mg/Al–O, U–O_eq, and O–O edge distances, interatomic distances for single-edge (¹E) and double-edge (²E) sharing configurations will occur between 3.13 and 3.85 Å (Table S9). All of the adsorption samples show the presence of ∼0.5 Mg/Al neighbors at 3.21 Å and ∼1.0–1.5 Mg/Al neighbors at 3.38 Å (Tables S6 and S7). Arai et al.³⁵ fitted next-nearest Al neighbors at 3.30 Å on imogolite, an amorphous aluminosilicate mineral, whereas Hennig et al.³⁶ modeled next-nearest Al neighbors at 3.40–3.45 Å on montmorillonite, a generally poorly crystalline 2:1 type aluminosilicate. The magnitude of the coordination numbers (±20%, Tables S6 and S7) of Mg/Al neighbors suggests that single-edge (¹E) and double-edge (²E) sharing complexes are formed in approximately 1:1 to 1:2 ratio (¹E/²E) and that the presence of CO₃ in the ligand shell of uranyl did not inhibit the formation of these surface complexes.

The second shell and other higher order bonding environments of the coprecipitated samples (HTC1 and HTC2) were substantially more complex. Second shells appeared to have shifted away from the central absorber in the RSFs to ∼3.50 to 4.10 Å (r + Δr, uncorrected for phase shift, Figure 2). The EXAFS fit of the simplest precipitates, HTC2 CO₂r (pH 9.5) and HTC1 CO₂p (pH 9.5), was similar to the bonding environments of the adsorption samples (HTC3) with an additional Mg/Al neighbor near 3.60–3.65 Å (Tables S6 and S7). Coordination numbers ranged between 0.3 and 0.6 for Mg/Al neighbors near 3.20 and 3.65 Å and between 1.3 and 1.8 for the Mg/Al neighbor at ∼3.37 Å which are too high to consider that three different surface species (2 × ¹E + 1 × ²E) are present. Instead, the magnitude of the coordination numbers suggests that uranyl moieties are surrounded at the surface by Al/Mg(OH)₆ octahedra to different extents having between two and five next-nearest Mg/Al neighbors. If we consider a binary solution of the HTC2 CO₂r sample with 1.6 C atoms as part of the first ligand shell and a ∼50/50 mixture between uranyl tris-carbonate and CO₃-free uranyl(hydroxide) moieties (Table S8), the latter would hence appear to be coordinated by 4–5 Mg/Al neighbors (2 × 0.5 + 2 × 1.3 + 2 × 0.5 = 4.6, ±20%, see Table S9). A similar bonding environment, containing less C atoms, appears to be present in the HTC1 CO₂p sample, however, as a mixture of ¹E, ²E, and partially surrounded uranyl complexes (Table S7). It should be mentioned that while no meaningful fits were observed when the distance at 3.60–3.65 Å was modeled with the linear O–U–O–U multiple scattering path,³⁷ some of the EXAFS signals may have been nevertheless contributed by this type of photoelectron scattering (Figure S4).

The two remaining coprecipitates, HTC2 CO₂p and HTC1 CO₂r, presented the most complex EXAFS signals, likely because of (partial) cancellation effects between many different U–Mg/Al/U, multiple scattering events and uranyl adsorption on aragonite (HTC1 CO₂r). Magnesium/Al shells at 3.20 and 3.37 Å could not be fitted in these samples, resulting in either negative coordination numbers, Debye–Waller parameters, or both. The ΔR₂ segments shifted away in distance from the central absorber and, in the case of HTC2 CO₂p, was split into two ΔR₂ segments (a and b, Table S7). Tentatively, the fitted data for HTC2 CO₂p may show a well-incorporated uranyl moiety in the HTC structure, but this interpretation is problematic because of the persistence of the trans-, di-oxo moiety of uranyl (see Results and Discussion below) as seen in the XANES spectra. More distal (>4.0 Å) neighbors may represent either Mg/Al or O neighbors in the HTC structure (the fit shown in Table S7 used U–Mg/Al scattering paths). HTC1 CO₂r was the only sample in which a next-nearest U neighbor may have been present at 3.90 Å alongside Mg/Al at 3.61 and 3.72, and 3.90 Å being similar to the interatomic U–U distance in schoepite and clarkeite (both near 3.96 Å). The presence of aragonite, a phase concomitantly formed during HTC1 precipitation in the CO₂r atmosphere, combined with uranyl’s affinity for aragonite,³³ complicates the interpretation of the EXAFS further. The fit results of these latter two coprecipitates should hence be cautiously interpreted to show one or more (possibly many) bonding environments; however, the spectra are probably too complex (and noisy) for a nonlinear least-square shell fit approach to be conclusive. It is also quite possible that the EXAFS captured an intermediate, still evolving state which only over time will reveal a more distinct bonding environment; long(er)-term kinetic studies would be required to substantiate this. Structural evolution of precipitates can be explained through Ostwald ripening, for example, Gräfe and Sparks³⁸ have shown that zinc arsenate precipitates on goethite do not assume a stable coordination environment until several months after the reaction. Lead (Pb(II)) will nucleate only over time into hydroxide precipitates on oxide surfaces.³⁹

Uranyl Bonding Mechanism I: Coprecipitation (HTC1 and HTC2)

The basic building blocks of the HTC structure are near-perfect octahedra of Mg(OH)₆^4– and Al(OH)₆^3– which present two O–O edge lengths. These can be simply defined as “bonding” and “nonbonding” (Figures 3 and S5). The bonding O–O edge lengths (2.664 Å) run perpendicular to the chain of edge sharing octahedral units, whereas the nonbonding O–O edges (3.054 Å) run parallel to it and reflect the interatomic Mg/Al–Mg/Al distances (Figures 3 and S5). The U L-III XANES spectra show that the uranyl symmetry, that is, the trans-, di-oxo O=U=O bridge, was not lost in the coprecipitation samples HTC1 and HTC2. This creates distinct steric restrictions on the ability of (UO₂)(O_eq)_n to complement (or become incorporated into) the HTC structure because the apical, double-bonded O atoms are not shared with other polyhedra—including structures such as schoepite or clarkeite.^40,41 Recently, several research groups⁴²⁻⁴⁵ have shown that uranyl can be successfully incorporated into goethite (α-FeOOH), when U abandons the actinyl symmetry and assumes approximately octahedral symmetry. This is made possible when U⁶⁺ is reduced to U⁵⁺, for example, by Fe²⁺, and is immediately stabilized by other Fe^2+/3+ octahedra. In the present study, only the synthesis of HTC1 contained potential electron donors for U⁶⁺ to reduce to U⁵⁺, whereas the HTC2 precipitation system did not. Hence, U incorporation into HTC2 in any of the present experiments is implausible, but some degree of Al/Mg(OH)₆surrounding (without incorporating) a surface uranyl moiety appears plausible based on the XANES and EXAFS data presented here (HTC2 CO₂r, HTC1 CO₂p). Such a surface complex would have distinctly limiting effects on the crystal growth and expansion in the ab plane of the HTC.

Figure 3.

Presentation of ¹E and ²E surface adsorption complexes of uranyl on HTCs. The ²E surface complex is most likely formed in kink sites where coordination to more than one Mg/Al octahedron would be possible. Oxygen atoms in the equatorial plane of uranyl may be part of a carbonate ligand shell. The apical/actinyl O atoms (O_yl) are shown in red.

The adsorption of uranyl to surface sites therefore had to have two effects on HTC formation: (1) an interference effect if uranyl adsorbed on the bonding O–O edges or (2) a limited effect if uranyl adsorbed to the nonbonding O–O edge. Of the four precipitation systems studied here, the (012) and (120) reflections of HTC1s and HTC2 CO₂p diffraction patterns (Figure S2) indicate that the expansion of the precipitates in the ab plane was limited. The diffraction pattern of the HTC2 CO₂r sample, however, displays well-resolved (110, 113) reflections compared to its CO₂p counterpart indicative of less restricted crystal growth despite a similar extent of U retention (Table S3). Crystallite size calculations using the Scherrer equation confirmed that the HTC2 CO₂r sample (∼366 nm) was approximately five times larger than the HTC2 CO₂p sample (∼72 nm), further suggesting that complexation by CO₃ effectively inhibited uranyl from binding to the HTC growth sites (bonding O–O edges) while permitting its reaction at nonbonding O–O edges. Indeed, the U–Mg/Al distances at 3.47 and 3.71 Å (Table S7) suggest that the surface complexation at the 2.664 Å O–O bonding edge with both the shorter (2.29 Å) and longer (2.49 Å) U–O_eq ligands (Table S9) involved leads to a significant interference effect on HTC formation. The U–Mg/Al distances observed in the HTC3 (CO₂r) samples and the HTC2 CO₂r and HTC1 CO₂r samples all indicate that uranyl adsorption occurred predominantly on the nonbonding O–O edge involving partial ligand exchange (O/OH for CO₃). This ligand exchange is likely favored by the presence of SO₄/NO₃ in the interlayer, which would be exchangeable and thus consume carbonate released from uranyl (mono-, bis-, and tris-carbonates) and OH ions during inner-sphere complexation; such a reaction would thus favor not only the formation of the surface complex but also the concomitant SO₄^2–/2NO₃^––CO₃^2–/OH^– anion exchange.²⁶ We therefore propose (Schemes S1 and S2, Supporting Information) that the partial exchange of CO₃–SO₄ and OH–NO₃ should result in a net entropy increase because of the greater number of ligands in the interlayer and in solution. These proposed reaction pathways agree with previous thermodynamic and kinetic studies, indicating a significant entropy increase (113–149 J/(mol·K)) of an endothermic reaction (25–28 J/mol: SO₄–CO₃ exchange⁴⁶) between uranyl and HTC, and would be approximately consistent with the observed pseudo-second-order reaction kinetics^5,8,10 and incomplete anion exchange.⁴⁷

Uranyl Bonding Mechanism II: Surface Complexation (HTC3)

Surface adsorption of negatively charged uranyl carbonates would have been favorable under the reaction conditions because of the positive, pH-dependent surface charge of the LDH; the point of zero charge of HTC is approximately 12.2,²⁸ although it may vary substantially with composition (e.g., to as low as pH 9.2⁴⁸) which needs to be considered in the context of the polymetallic barren lixiviant samples. Schemes S1–S4 show proposed surface adsorption mechanisms of different uranyl species, which depend on pH (7.0 vs 9.5) and the presence/absence of CO₂; scheme S1 uses uranyl bis-carbonate (UO₂(CO₃)₂^2–) as the initial species because this species would be dominant under normal atmospheric conditions. In CO₂r suspensions, uranyl will occur dominantly as the tris-carbonate species at pH 7.0 and 9.5 (Figure S1). A bis-carbonate complex at pH 7.0 does not necessarily have to undergo OH–CO₃ ligand exchange to form an ¹E complex (scheme S1a,b); however, coordinated water molecules may become deprotonated in an initial step (scheme S1a) if they are not exchanged for surface-OH in the subsequent one (scheme S1b). The resulting 2H⁺ would be neutralized by the ligand-exchange reaction between surface- and uranyl-OHs, resulting one way or another in 2H₂O. The ²E complex for the same reaction conditions, however, requires the exchange of at least one CO₃ for OH and possibly the deprotonation of two coordinated water molecules of uranyl (scheme S1c). This reaction yields H⁺ and HCO₃ ions in solution, in which the H⁺ is neutralized by the release of 3OH ions during the subsequent OH ligand exchange between uranyl and the surface yielding H₂O, 2OH, and HCO₃ ions in solution. The OH and HCO₃ ions may possibly exchange for NO₃ and SO₄ in the interlayer; however, there is no prior research that has established the affinity of HCO₃ relative to SO₄, NO₃, or OH in the HTC interlayer. At pH 9.5 (CO₂r), CO₃–OH ligand exchange as an initial reaction step (scheme S2a) would be obligatory for either ¹E or ²E surface complexes to form (scheme S2b). The release of CO₃^2– ions into solution would provoke subsequent SO₄/NO₃–CO₃/OH exchange of interlayer anions (scheme S2c). Adsorbed uranyl species at pH 7.0 (CO₂r) were coordinated by ∼2.3 carbonate ligands, which was surprisingly higher than that at pH 9.5 (N_U–C ≈ 1.5) but is consistent with the dominant UO₂(CO₃)₃^4– species in solution at pH 7.0 and 1 atm of CO₂ (Figure S1).

At pH 7.0 (CO₂p), UO₂(OH)₂ (3%), UO₂(OH)⁺ (3%), (UO₂)₃(OH)₅⁺ (74%), and (UO₂)₄(OH)₇⁺ (20%) make up soluble uranyl species.⁴⁹ Although the initial concentration of uranyl nitrate was slightly supersaturated with respect to schoepite (S.I. ≈ 0.22), no evidence was observed for its formation nor for the formation of polynuclear uranyl species (e.g., (UO₂)₃(OH)₅⁺ or (UO₂)₄(OH)₇⁺). Scheme S3 relates the adsorption mechanism of the dominant (UO₂)₃(OH)₅⁺ species at pH 7.0 undergoing depolymerization into neutral UO₂(OH)₂⁰ (aq) species before adsorbing to the HTC surface as ¹E and/or ²E surface complexes. Reaction (scheme S3a) is probably driven by the adsorption of the neutral, UO₂(OH)₂ (aq, 3%), fraction on HTC as the dominant (UO₂)₃(OH)₅⁺ and (UO₂)₄(OH)₇⁺ species would be expected to be repelled by the positive surface charge. Six or seven H₂O molecules are generated in the reaction as a result of ¹E or ²E surface complex formation (with four equatorial OH ligands according to the EXAFS models), respectively, and the associated H₂O–OH ligand-exchange reactions of uranyl (assumed to be initially sixfold coordinated in the equatorial plane) and the surface. At pH 9.5 (CO₂p, scheme S4), UO₂(OH)₃^– (66%), (UO₂)₃(OH)₇^– (31%), and UO₂(OH)₂⁰ (3%) are the three dominant uranyl species.⁴⁹ Being negatively charged, the first two species are expected to be attracted to the positively charged surface where ligand exchange would subsequently result in ¹E and ²E surface complexes (scheme S4a,b). Previous research^5,6,50 has shown that U adsorption does not decline until the point of zero charge of HTC is approached with the extent of adsorption not being significantly different between pH 7.0 and 9.5, which is consistent with the findings of this work.

The modeled bonding environments, specifically of the first ligand shell, have provided important insights into the potential reaction mechanism(s) of the surface/interlayer reactions and are significant in the context of many previous sorption studies of uranyl with minerals of pH-dependent surface charge (e.g., ferrihydrite) at neutral to alkaline pH and the presence of carbonate ligands.^35,51−53 These previous studies have shown that adsorption declines significantly with an increase of uranyl carbonate species, whereas in HTC suspensions, the sorption reaction is probably limited by the point of zero charge but not uranyl speciation. The critical difference between HTC and ferrihydrite sequestration is the anion-exchange reaction (SO₄–CO₃ and NO₃–OH) in the final reaction steps shown in Schemes S1 and S2. Catalano and Brown,⁵³ however, have also shown that uranyl carbonate species will bond at Fe sites of montmorillonite surfaces while retaining between 0.6 and 1.5 CO₃ moieties. HTC removed uranyl from solution very effectively, even at the low concentrations tested here. Removal was largely complete whether the HTC formed in situ or was applied as an adsorbent. In the absence of a reductant, it is unlikely that uranyl could be properly incorporated into the HTC structure because of the steric constraints posed by the trans-, di-oxo, O=U=O moiety. Uranyl adsorption to the HTC–water interface did not appear to be limited by the presence of excess carbonate ligands, probably because of favorable anion exchange with interlayer NO₃ and SO₄. Future studies should consider the potential role of reductants such as Mn(II), Fe(II), or Co(II) to reduce U(VI) to U(V) to break the actinyl symmetry that hinders the incorporation of U into HTC and thereby (potentially) form a (more) stable repository material. In this respect, time-dependent EXAFS and XRD studies could provide valuable information into the long-term fate of uranium and HTC both from a contaminant (U) and from a mineral (HTC) perspective. The selectivity sequence of Cl^–, NO₃^–, OH^–, and HCO₃^– should be established to understand differences in CO₃^2– versus HCO₃^– selectivity in HTC/LDH interlayers.

Experimental Section

HTC Precipitation

Three distinct types of HTC were prepared for the purpose of this study (Table S1). Type 1 HTC (HTC1) represented an industrially relevant HTC product that was synthesized from the barren lixiviant of the Beverley North U mine under a CO₂-enriched (CO₂r) or CO₂-depleted/CO₂-poor (N₂, CO₂p) headspace (details of the barren lixiviant composition are provided in Table S2 of the Supporting Information section) with NaOH addition for a final pH of 9.5 (initial pH of the reaction was 1.3). Type 2 HTCs (HTC2) were also precipitated under a CO₂r or CO₂p headspace from a solution containing 110.7 mmol/L MgSO₄, 36.9 mmol/L Al(NO₃)₃, and 19.0 μmol/L uranyl nitrate (UO₂(NO₃)₂) with an initial pH of 2.8. HTC2s represent the controlled, synthetic counterparts to HTC1, thereby avoiding the potential complications of multiple polyvalent cationic species and calcium contamination. HTC1 and HTC2 were aged for 13 days at 358 K, with intermittent swirling, pH readjustment, and refilling of the synthesis vessels’ headspace with the appropriate gas (CO₂ or N₂, respectively). In all cases, the headspace volume was approximately equal to the solution volume.

HTC3 samples were precipitated in the absence of UO₂²⁺ from 30 mmol MgSO₄ and 10 mmol Al(NO₃)₃ in 1 L of doubly deionized (DDI) water. The pH was adjusted to 9.5 using NaOH, and the suspensions were allowed to age at 358 K. The pH was checked after 8 days but found unchanged. After 15 days, the solids were washed twice in DDI water, solids were allowed to gravity settle, the supernatant was decanted, and the suspended solids were divided into four weight-equivalent portions for UO₂²⁺ adsorption tests.

UO₂²⁺ Adsorption Tests

The pH of each HTC3 suspension was adjusted to either pH 7.0 or 9.5 with H₂SO₄ or NaOH solution (ca. 10 M) as required (see also Table S1). CO₂ (CO₂r) or N₂ (CO₂p) gas was bubbled into the suspensions for ca. 10 min before the final pH adjustment was made. To each test, 52 μL of 25 mM UO₂(NO₃)₂ solution was added. This produced an equivalent HTC/UO₂²⁺ ratio in the adsorption tests as in the HTC2 samples; however, conditions in the HTC3 CO₂p sample at pH 7.0 were thus slightly supersaturated (S.I. ≈ 0.22) with respect to schoepite ([UO₂]₈O₂(OH)₁₂[H₂O]₁₂). The pH of these suspensions was regularly rechecked and adjusted as necessary; the flask headspaces were refilled with the designated gas, loosely capped, and swirled occasionally while the samples aged at 358 K for 2 days.

At the end of their respective aging periods, the solids from HTC1, 2, and 3 were separated from their synthesis liquors by centrifugation, and a supernatant sample was taken to determine the concentrations of Al, Mg, and U in solution by inductively coupled plasma mass spectrometry and to calculate the composition of the HTC solids (Table S3). The solids were washed with DDI water, shock-frozen in liquid nitrogen, and freeze-dried until in powder form for XRD and X-ray absorption fine structure (XAFS) spectroscopy analyses.

XRD and XAFS Data Collection

Randomized powder samples of freeze-dried HTC1, 2, and 3 were prepared on zero-background slide holders and patterns collected on an Empyrean PANalytical X-ray diffractometer in a Bragg–Brentano mode with monochromatic Co Kα radiation. The patterns were indexed against the published HTC structures,¹⁸⁻²⁰ and the unit cell parameters were calculated from the fitted peak positions (WinXAS (v3.1²¹)) of the (003), (006), and (012) reflections (Table S4).

Dry, randomly oriented powders of HTC1, 2, and 3 were sealed into thin, steel sample holders with a X-ray transparent Kapton tape for XAFS measurements in fluorescence mode (CANBERRA germanium detector) on the 1.9 T wiggler XAS beamline (12IDB, section 9) at the Australian Synchrotron (Melbourne, VIC). Between 2 and 4 scans per sample were collected to improve the signal-to-noise ratio (additional details of the XAS data collection are provided in the Supporting Information section).

XAFS Data Interpretation

Standard XAFS data reduction procedures were applied as detailed in ref (22) using the programs Average (v2.0.6, energy corrections) and WinXAS (v3.1²¹), including inspection of beam damage, glitch removal, background correction, normalization, averaging, and EXAFS fitting. Spectra were converted from energy (keV) to wavevector (k) space (Å^–1) by assigning the binding energy, E₀, to the main inflection in the U L-III X-ray absorption edge. The oscillatory motion of the EXAFS was extracted using a cubic spline function of ≤7 knots over an average k-range of ∼2.4–14.6 Å^–1. Radial structure functions (RSFs) were calculated using a forward Fourier transform (FT) over a range of 3.20–14.50 Å^–1 and a Bessel windowing function (β = 4). Nonlinear least-square shell fits were performed after applying a back Fourier transform (BFT) over a specific range (ΔR) in the RSF. The first segment (ΔR₁) comprises the axial and equatorial ligand shells to approximately 2.50 Å (R + ΔR, uncorrected for phase shift) in CO₂p and up to 2.90 Å in CO₂r samples to include contributions from U–C photoelectron scattering. The second BFT segment (ΔR₂) ranged from ca. 2.90 to 3.45 Å for HTC3 samples and up to 4.10 Å (R + ΔR, uncorrected for phase shift) for HTC1 and HTC2 samples to model contributions from U–Al/Mg and U–U photoelectron scattering. The phase-shift parameter (ΔE₀) was correlated among all scattering paths to the same value and varied freely during the calculation, and the amplitude reduction factor (S₀²) was fixed to 1.0 for all samples. The number of permissible free-floating parameters was determined from the extended Stern equation (N_dpi = (2 × Δk × ΔR/π) + 2²³) as applied in WinXAS (see further details regarding EXAFS modeling in the Supporting Information section).

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsomega.7b01050.

• Additional details regarding the XAS data collection; details regarding the fitting procedure of the EXAFS data; additional figures (Figures S1–S5) and tables (Tables S1–S9) supporting the main findings; Schemes S1–S4 that are discussed in the main text of the manuscript; and additional references (PDF)

Author Present Address

^⊥ Departamento del Manejo de Suelos y Aguas, Instituto Nacional de Investigaciones Agropecuarias (INIAP), Estación Experimental Santa Catalina, Sector Cutulaghua (Cantón Mejía, Pichincha), Ecuador (M.G.).

The authors declare no competing financial interest.