Uranium (U) is a ubiquitous trace element occurring in low-temperature carbonate minerals, sediments and rocks. These carbonate materials containing U are extensively used for fundamental investigations, including geochronology, oceanic paleo-redox studies, and environmental proxies. The latter studies use 238U/235U isotope fractionation signature (δ238U) as a specific proxy for redox and coupled climate changes throughout Earth’s geological history1,2,3,4,5,6. In contrast, only a limited number of δ238U-studies constrain local environmental impacts after engineering activities, such as U mining and related redox perturbations7,8,9. Since considerable U isotope fractionation is expected, mainly during U(VI) reduction, systems that include abundant redox-active species, e.g., iron (Fe)-sulfides, have attracted the most interest. The occurrence of elevated U as U(IV) has been reported from a few U-rich ancient calcites10,11,12, but no U isotope fractionation analyses have yet been reported for these systems. It has been proposed that a high partition coefficient of U detected in anoxic systems might result from structurally incorporated U(IV) in secondary minerals13. However, this is yet unconfirmed as no coupled proof exists for the occurrence of structural U(IV) in calcite formed under reducing conditions, coupled with a high partition coefficient attested by water and mineral analyses.

Recent laboratory studies have aimed to better understand U isotope fractionation mechanisms that would link δ238U signatures to local and global redox conditions14,15. However, these links remain elusive and underexplored when applied to deep anoxic, microbial-mediated rock environments. Current estimates show that the deep biosphere hosts the majority of microbial life on Earth and a substantial amount of biomasses16. Since microbes catalyze the decomposition of organic substances, facilitating the formation of secondary minerals, such as calcite13,17 and Fe-sulfides18,19, and thus affect the environmental mobility of trace elements such as U, there is an urgent need for exploring microbially mediated U reduction and related U isotope fractionation in subsurface settings. A recent study has reported the formation of non-crystalline U(IV) species formed in a roll-front deposit after microbial activity based on the finding of positive δ238U signatures9. While the latter and other recent studies have focused on either ancient systems in various natural samples1,3,10,11,20,21,22,23, or laboratory experiments24,25,26 (relevant publications are shown in Supplementary Information (SI) and Supplementary Table 1), investigations of modern carbonates, involving analysis of δ238U signatures linked to U speciation in diluted environmental systems are scarce.

Herein, we aim to explore how δ238U is associated with U incorporation into calcite in a deep anoxic, igneous-rock fracture network-hosted groundwater aquifer at the Äspö Hard Rock Laboratory (HRL), south-eastern Sweden. Over 17 years, calcite has precipitated along with Fe-sulfides on borehole equipment in an isolated deep borehole section and immobilized large portions of U from groundwater during precipitation (Supplementary Fig. 1a–d). A recent study performed on these calcite specimens showed that U (along with some other trace elements) exhibits unexpectedly high distribution coefficients, implying that U redox transformation processes aiding co-precipitation with calcite have occurred in the borehole13. As both the source (groundwater) and sink (calcite) can be examined in this borehole section, and because the system is influenced by bacterial sulfate reduction, dissolved sulfide (HS-(aq)) and ferrous (Fe(II)(aq)) species, and Fe-sulfidic precipitates, this system is optimal for assessing U speciation and isotope fractionation related to a complex natural system that involves both abiotic and biotic redox mediators. This is an important complement to laboratory experiments that investigate more simplified systems with individual biotic and abiotic redox species14,15,27. In this study, we use a suite of micro-analytical synchrotron-based and laboratory techniques, U and Ca isotopic tools, and thermodynamic modeling to reveal U speciation, redox evolution, removal pathways, and related isotope fractionation in the deep subsurface.


Uranium speciation in calcite

Synchrotron micro-X-ray fluorescence (µXRF) mapping coupled to U LIII-edge micro-X-ray absorption near edge structure (µXANES) spectroscopy, as well as an extended X-ray absorption fine structure (EXAFS) technique, were used to reveal U distribution, U oxidation state and the local atomic environment in “Äspö-calcite” material. The U µXRF map recorded on a 200 µm × 300 µm area of one “Äspö-calcite” crystal is shown in Fig. 1a with µXANES spectrum collected on a 30 µm × 100 µm region of the U-rich rim, together with spectra of U(IV)-calcite and U(VI)-calcite (Fig. 1b), and Fourier-transformed (FT) EXAFS data for “Äspö-calcite” (Fig. 1c).

Fig. 1: µXRF and µXAS results.
figure 1

a U µXRF map of a calcite crystal from Äspö HRL. The area outlined with a dashed line was used for U LIII-edge XANES measurements (estimated ~20 ppm U13, on the right y-axis - relative signal intensity scale); b U LIII-edge µXANES spectra of Äspö-calcite, U(IV)-calcite11 and U(VI)-calcite20. The energy shift between U(IV) and U(VI) references is ~1.6 eV. Marked on the spectrum of ‘U(VI)-calcite’ with I and II are spectral features characteristic to uranyl(VI) species; c Magnitude of the Fourier-transformed EXAFS (black solid line) and best fit (colored rhombs) for the “Äspö-calcite” and single scattering paths for the fitted shells (see further details in SI and Supplementary Fig. 2c, d for k3 -weighted and the real part EXAFS data).

The “Äspö-calcite” XANES spectrum exhibits the position and shape of main resonances at nearly 17,176 eV and broader post-edge peaks around 17,216 eV, which perfectly agree with those for the “U(IV)-calcite” spectrum—a well-studied U(IV)-bearing calcite material collected from a high-U anoxic Mississippi Valley-type zinc ore deposit11. Distinct spectral features characteristic to U(VI), such as a main resonance peak shifted nearly 2 eV to higher energy at ~17,178 eV, a ‘shoulder’ feature at ~17,190 eV (marked as I), and a less intense peak with a maximum around 17,210 eV (marked as II), which are usually assigned to uranyl species28 can be distinguished on the spectrum of “U(VI)-calcite” used as a U(VI) reference and is described elsewhere20. The µXAS measurements at two different microfocus beamlines showed similar results (see Supplementary Figs. 2a,b), and k3-weighted χ(k) and the real part of the FT-EXAFS (Supplementary Figs. 2c,d). Additionally, the same calcite material was analyzed by U LIII-edge high-energy resolution fluorescence-detected XANES spectroscopy, which showed the bulk occurrence of U(IV) (Supplementary Fig. 3). Linear combination fit (LCF) analysis gave 94 ± 3% for U(IV) and 6 ± 3% for U(VI) fractions. Additional EXAFS fit was performed, including three U-O shells for the first intense FT-EXAFS peak, aiming to determine the possible contribution from the uranyl(VI) group. The yielded 1.74(2) Å is inconsistent with any reported axial U-O bond in uranyl(VI) compounds29. FT-EXAFS of “Äspö-calcite” is better described with a two U-O shell approach for the main peak with U-O1 at 2.22(2) Å and U-O2 at 2.82(3) Å, and nearly equal coordinating oxygens, 3.1(4) and 2.7(5), respectively (Table 1). The further backscattering peak is well fitted with equal contribution from carbonates: (U-C) at 3.30 Å and (U-O3) at 3.52 Å, indistinguishable from each other, and more distant low-intensity backscattering peak at 4.07(7) Å, referring to U-Ca shell.

Table 1 EXAFS fit results for the “Äspö-calcite” sample.

Uranium isotope fractionation signatures in groundwater and calcite

U isotope values (δ238U), U-series activity ratio (AR = 234U/238U), and the U concentrations measured for calcite and water samples are given in Table 2. Both borehole-section specific pairs of calcites (calcite1 and calcite2) and fracture water samples (FW1 and FW2) exhibit strong 238U depletion, at δ238U ~ –2‰, and thus, insignificant net fractionation Δ238Ucalcite1-FW1 of ~0.11‰ and 0.03‰ (Fig. 2), within the 0.12‰ uncertainty. A third calcite sample from a separate packed-off section in the same borehole (no corresponding water sample available) showed a similar value as δ238Ucalcite2. The 234U/238U AR were relatively high and showed small variations within 4.47–4.87 (Table 2).

Table 2 U content, δ238U and 234U/238U measured in calcite and fracture water (FW) samples, collected from borehole sections at Äspö HRL.
Fig. 2: δ238U in waters and solids.
figure 2

δ238U in fracture water (blue squares) and calcite (green rhombs) samples collected from the borehole equipment at 415 m depth in KA3105A borehole, Äspö. FW #22970 refers to calcite #22132 and FW #22968 refers to calcite #22125. For calcite sample #22134, no water sample is available. Dashed lines at –1.90‰ and –2.10‰ are plotted for reference relevant to the fluid-solid setting. Error bars represent δ238U = 0.12‰ as ±2σ. δ238U values for modern continental crust and seawater are –0.29(3)‰ (2σ) and –0.39(1)‰ (2σ), respectively90.

Calcium isotope composition (δ44Ca) determined for both calcite and borehole water samples revealed an average calcite-water difference Δ44/40Cac-w of –0.43‰ (Supplementary Fig. 4). Using the observed thickness (average 500 µm) and a volume of calcite (3.5 × 10−5 dm3) for one of the calcite samples and duration of 17 years, an estimated calcite growth rate of 2.5 × 10−8 moles/m2/sec can be calculated. Together with the Δ44/40Cac-w of –0.43‰ this is consistent with the surface kinetic model describing the relationship between calcite growth rate and Ca isotope fractionation, as well as with experimental data at higher growth rates30 (Supplementary Fig. 5). Our observed Δ44/40Cac-w and Ca2+/CO32− ratios are compared to predicted curves for different values of calcite super-saturation31 (Supplementary Fig. 6). The position of our data for the growth of calcite falls along the curve for σ = 0.5, in a good agreement with the calculated median value of 0.46 for the growth conditions in the boreholes. The consistency of our data shown in Supplementary Figs. 5, 6 indicates, on average, essentially continuous calcite growth over the 17-year period of the experiment at low calcite supersaturation. The well-preserved growth zones in calcite agree with the groundwater sampling events and the variation of Sr content in the calcite that reflects the change of water chemistry (increased fraction of Baltic Seawater) during the experiment, supporting the Δ44/40Cac-w data for the continuous growth of calcite13. However, this does not preclude short periods of more rapid calcite growth (or dissolution), involving the production of the high-U small-scale (tens of microns) laminations.


Uranium speciation in calcite

The µXRF analysis shows an uneven distribution of U within and between the different growth zones, from 2 ppm up to 20 ppm in the brighter part of a calcite sample with median U at ~9 ppm U. Relatively U-rich zonation that occurs at regular intervals can be attributed to water sampling activities when an inflow of pristine fracture water with a higher U content replaced flushed out stagnant water. This semi-closed cycle that follows flushing out of ten section volumes starts with high incorporation of U, and as partitioning of U(IV) is high, the calcite will become successively poorer in U with time, resulting in the varying U content in the precipitating calcite. Direct spectroscopic determination of U(IV) in calcite at a relatively high U content is in accordance with two previous studies that demonstrated the incorporation of U(IV) into ancient carbonates of various settings10,11. Since high U partition coefficients (pu ~ 100) have been determined for the calcite samples used in the current study13, we can here confirm, based on EXAFS results, the previous hypothesis that partitioning of U into calcite is governed by structural incorporation of U(IV). This observation is further supported by the absence of the peak at 3.83–3.86 Å relevant to the U-U backscattering peak observed in UO2. There is neither any evidence for the short U-O shell, which is normally resolved in an EXAFS spectrum as an intensive peak at 1.8 Å, nor for the longer U-O shell characteristic for equatorial bonds from 2.3 to 2.5 Å29. U-O backscattering peak of some uranyl(VI) species, e.g., U(VI) on calcium silicate hydrates32, may appear as a broad backscattered peak to some degree resembling the U-O in our EXAFS data. The clear evidence of U(IV) from µXANES data supported by LCF analysis rules out the possibility of U(VI) stabilization in the “Äspö-calcite”. Nearly 6% of U(VI) derived from LCF analysis can be referred to as statistical error due to spectral noise. EXAFS results for U-O1 and U-O2 shells agree well with the previously analysed U(IV)-calcite material for U-O1 at 2.21(3) Å and U-O2 at 2.78(2) Å11. The local coordination environment of U(IV) in calcite can be described with two models assuming different contributions of shorter (U-O1) and longer (U-O2) shells with ratios 4:2 and 3:3 that result in statistically indistinguishable mean U-O distances at 2.42 Å and 2.52 Å, respectively. Both calculated U-O values are in the range of U-O distances (2.35–2.49 Å) for some U(IV) compounds11 and notably larger than mean U-O distances (2.1–2.3 Å) for uranyl(VI) compounds33. A relatively high amount of Na (200–900 ppm) detected in the “Äspö-calcite” sample13 is theoretically enough to stabilize U4+ in the calcite through an earlier proposed charge compensation mechanism: 3Ca2+ = U4+ + 2Na+, that achieves local charge balance in the crystal lattice11. The EXAFS data is complemented by high-resolution field emission electron probe micro-analysis (FE-EPMA), which shows the relationship of several major and trace elements (Supplementary Fig. 7). The technique was described recently to detect coffinite-like U(IV) nanoclusters formed and preserved in ancient calcite34. No evidence of individual UO2 microphases or nanoparticles was found in the present study, supporting EXAFS results for U incorporation into the calcite crystal lattice as indicated by the above observations. Structural incorporation of U(IV) into calcite is evidently an effective mechanism for long-term immobilization of U from groundwater under anoxic conditions, even for exceedingly U-deficient groundwater near the solubility limit of UO2(am).

Redox evolution in borehole water

An estimation of U redox speciation in the borehole water, [U] ⁓ 10−9 mol/L, was done with the available geochemical data for the KA3105A borehole using PHREEQC35. The modeling shows a correlation between the calculated fraction of U(IV) and the reducing agents and their concentrations in the groundwater: either dissolved hydrogen sulfide (HS-(aq)) at 4.3 × 10−6 mol/L or dissolved ferrous (Fe(II)(aq)) species at 5.4 × 10−6 mol/L (Supplementary Table 2). PHREEQC reaction path calculations predict a redox-related evolution of U speciation based on the HS- concentration in the borehole water. Aquatic modeling of Äspö groundwaters have shown that thermodynamically stable U(VI) complexes are the main species that occur in this deep aquifer, with up to 4% of U(IV) present as hydrolyzed species13. The persistence of U(VI) under the reducing conditions (redox potential, Eh ~ –270 mV) dominating in fracture water can be attributed to the formation of thermodynamically stable U(VI) complexes, such as Ca2UO2(CO3)3(aq). There is virtually no available speciation technique for direct U redox determination at the extremely low U concentration ([U] ⁓ 10−9 mol/L) measured in the KA3105A borehole water that is near the solubility limit of UO2(am)36. Depending on the U concentration and main geochemical parameters (pH, Eh, ionic strength, dissolved inorganic carbon, and Ca), the U(VI) reduction may result in the precipitation of UO2, usually represented as amorphous hydrate, UO2·2H2O(am)36. The PHREEQC speciation calculations predict that in the KA3105A borehole water, U occurs predominantly as aqueous U(VI) species but with a minor fraction of aqueous U(IV) species13. Under this condition, the concentration of U(IV) is too low for UO2·2H2O(am) to precipitate, while the removal of U(IV) by co-precipitation with calcite is energetically favorable. The occurrence of high Fe(II)(aq) relative to dissolved U(VI) is expected to facilitate the reduction of U(VI) in borehole water37, where slow fluid exchange favors elevated concentrations of dissolved species that promotes precipitation of Fe-sulfides. Laboratory experiments show that U(VI)-carbonate complexes are more readily reduced to U(IV) on the surface of Fe(II) minerals38,39, by HS-(aq)40 or with Fe(II)(aq)37 making these routes thermodynamically feasible in stagnant water in the borehole setting. The simultaneous presence of both Fe(II)aq and HS-(aq) in the water, and limitations for measuring redox potential in the sulfidic system, e.g., fouling of electrode surface41, set certain complications to experimentally determine the redox effect from the individual species. Although measurements of the redox potential for this borehole section are not available, according to PHREEQC calculations, the redox potential for borehole water corresponding to the KA3105A section is found lower for the HS-/SO42- as compared with the Fe(II)/Fe(III) couple for low concentration of redox species. The reducing capacity of these two species can be understood from redox equations, where eight times less sulfide (Eq. 1) compared to ferrous iron are consumed for the reduction of 1 mole of U(VI) to U(IV) (Eq. 2). Thus, Fe(II)(aq) can be considered as a moderate reducer while HS-(aq) is a strongly reducing species42.

$$\begin{array}{c}{{{{{{\rm{Ca}}}}}}}_{2}{{{{{{\rm{UO}}}}}}}_{2}{({{{{{{\rm{CO}}}}}}}_{3})}_{3({{{{{\rm{aq}}}}}})}+0{{{{{{\rm{.25HS}}}}}}}^{-}+{{{{{{\rm{3H}}}}}}}_{2}{{{{{\rm{O}}}}}}={{{{{{\rm{UO}}}}}}}_{2}\cdot {{{{{{\rm{2H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{({{{{{\rm{am}}}}}})}\\ +0{{{{{{{\rm{.25SO}}}}}}}_{4}}^{2-}+{{{{{{{\rm{3CO}}}}}}}_{3}}^{2-}+{{{{{{\rm{2Ca}}}}}}}^{2+}+2{{{{{{\rm{.25H}}}}}}}^{+}\end{array}$$
$$\begin{array}{c}{{{{{{\rm{Ca}}}}}}}_{2}{{{{{{\rm{UO}}}}}}}_{2}{({{{{{{\rm{CO}}}}}}}_{3})}_{3({{{{{\rm{aq}}}}}})}+{{{{{{\rm{2Fe}}}}}}}^{2+}+{{{{{{\rm{6H}}}}}}}_{2}{{{{{\rm{O}}}}}}={{{{{{\rm{UO}}}}}}}_{2}\cdot {{{{{{\rm{2H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{({{{{{\rm{am}}}}}})}\\ +{{{{{{\rm{2FeOOH}}}}}}}_{({{{{{\rm{s}}}}}})}+{{{{{{{\rm{3CO}}}}}}}_{3}}^{2-}+{{{{{{\rm{2Ca}}}}}}}^{2+}+{{{{{{\rm{6H}}}}}}}^{+}\end{array}$$

The redox potentials for these two systems depend on pH, as well as on the concentrations of the reducing species, and therefore, the U(VI)/U(IV) ratio calculated according to either HS(aq) or Fe(II)(aq), will also depend on pH and on the concentrations of HS(aq) or Fe(II)(aq). For the investigated borehole section, the calculated Eh is slightly higher for the sulfide-sulfate system (Eh = −217 mV) than for the Fe(II)/Fe(III) system (Eh = −270 mV), resulting in a larger U(VI)/U(IV) ratio calculated for the sulfide redox system, even when the reducing capacity of the sulfide is larger than that of Fe(II)(aq) (Supplementary Table 2).

Uranium turnover and removal pathways in deep aquifers

Dissolved U concentrations at the site decrease in groundwater by more than 1 order of magnitude from the highest measured of 7 µg/L at 31 m to ~0.3 µg/L in 600 m deep borehole sections43. There is thus a systematic removal of U from the groundwater along its flow path, associated with continuous interaction of mobile U(VI) with redox fronts on deep fracture mineral surfaces that can be a key mechanism resulting in a decrease of U concentration in Äspö groundwaters from shallow to deep aquifers43,44,45.

The studied Äspö HRL borehole sections represent semi-closed systems where fracture water constantly flows through and partly resides as stagnant water. Hence, it is important to estimate the U mass balance for the interpretation of the U reductive removal and related isotopic fractionation. It should be noted that time-dependent groundwater compositions and U concentrations are not available for the 17-year time span of the experiment; hence it is not possible to provide chemical rate constants from our results. Previous laboratory studies have indicated that primarily uranyl non-carbonate species control the kinetics of the bio-reduction rate of U(VI) to U(IV)46. If the fracture water flow rate through the borehole was not subject to considerable variations, one can approximate the U mass balance during the 17 years of the experiment using simplified calculations, which do not consider the removal of solid phases by water flow or precipitation on borehole walls (Supplementary Table 3).

The U mass (in mg) that passed through the A:3 borehole was about 3.1 mg of U, of which 1.25 mg is removed in the borehole. The amount of groundwater passed through the A:3 section is estimated at 5900 L, which is 1000 times the volume of the A:3 section47. In the same way, the turnover U amounts can be estimated for A:2 and A:4 borehole sections resulting in 2 and 5.92 mg, respectively. The Fe turnover is nearly three orders of magnitude higher than U for each borehole section. The removal of dissolved U in the borehole water is accompanied by the decrease of dissolved Fe. The removal efficiencies for U and Fe vary in the borehole sections; 75% in A:2, 40% in A:3, and 12% in A:4 (Fig. 3a) and ca. 80% in A:2, 17% in A:3, and 89% in section A:4 (Fig. 3b), respectively. The joint decrease of U and Fe from the borehole water might be due to a coupled reduction of U(VI) with sulfides and subsequent incorporation into the structure of calcite, further based on observations of surface U clusters forming on Fe-sulfides mixed with calcite crystals.

Fig. 3: Uranium and iron variation in waters.
figure 3

a U and b Fe concentration in fracture water (blue squares) vs stagnant water (orange circles) in monitored A:2, A:3, and A:4 sections in KA3105A borehole based on analysis of the final water withdrawal (see Supplementary Table 3).

The laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) analysis of Fe-sulfide coatings formed on magnetite crystals shows that U is associated primarily with Fe-sulfide forming on the surface of magnetite48 (Fig. 4a) with U concentration reaching ~90 ppm in few spots forming on the mineral grain boundary (Fig. 4b, c). The finding of this relatively high amount of U on the surface of Fe-sulfide is remarkable, given that U concentration in fracture water is at ~10−9 mol/L, under dynamic conditions where fracture water continuously passes through the borehole.

Fig. 4: Elemental distribution within the mineral grain.
figure 4

a LA-ICP-MS map of U, S, and Fe in sulfidized magnetite (Äspö HRL sample label #22131_4); b U map with several elevated U spots (up to 90 ppm) identified on the grain borders; c Transect across grain borders marked as A and B shows the amount of U associated with S and Fe (for details see SI and Supplementary Table 4).

The environmental behavior of U is controlled by several processes, including complexation, adsorption, precipitation, and redox transformations, and often followed by structural incorporation in secondary minerals, e.g., earth-alkali metal carbonates, Al- and Fe-oxyhydroxides49. While both U(IV) and U(IV) are removed by carbonates primarily through structural incorporation as shown for calcite in this work and in previous studies11,20, Fe-oxyhydroxides may exhibit several U removal pathways, e.g., adsorption of U(VI) species onto mineral surface followed by the formation of UO2 precipitates50 or structural incorporation of various U redox species51,52. Based on the LA-ICP-MS results, U has accumulated on the surface of Fe-sulfide, which in turn is formed on magnetite formed during the corrosion of borehole equipment. Sulfidization of magnetite occurs due to microbially-driven production of HS-(aq), enhanced in stagnant water under relatively slow water inflow conditions. In Äspö groundwater, U occurs as soluble U(VI)-carbonate complexes, which can be readily removed by Fe-sulfides via selective adsorption followed by fast reduction to U(IV), as shown in laboratory batch experiments53,54, and detailed reaction mechanisms proposed elsewhere55. U speciation was not analysed for Fe-sulfide material due to technical reasons. It is reasonable to anticipate that U occurs as UO2 stabilized on the surface of Fe-sulfide, in accordance with several laboratory studies53,54,55,56,57. The stability of the precipitated UO2 depends on several factors, including the redox sink of the mineral phase, e.g., Fe-sulfides, the persistence of reducing and anoxic conditions, and a limited supply of reactants to maintain redox processes.

Evidence for microbial organic phase degradation

The elevated sulfide concentrations in the borehole can be explained by sulfate-reducing bacteria (SRB) degrading dissolved organic species in water and plasticizers included in polyvinylchloride (PVC) tape used to fasten the water collection equipment. Bis(2-ethylhexyl) phthalate (or DEHP, chemical formula – C24H38O4) is one of the most common plasticizers, which is added to make the plastics flexible. DEHP is known to be readily biodegraded by bacteria in water and sediments58 and to produce carbonate and sulfide while reducing dissolved sulfate in the water, contrary to PVC, which shows meager degradation rates59. The stoichiometric reaction shown in Equation 3:


Thus, the degradation of 1 mole of plasticizers results in 24 moles of HCO3 and nearly 16 moles of HS, providing the main source of carbonate and sulfides, and promoting precipitation of calcite, amorphous FeS, pyrite, and intermediate species between the latter. Assuming that HS-(aq) occurring at ~0.19 mg/L in the borehole section was constant since the installation of equipment, the total amount of sulfide produced and removed from the borehole section by the flowing groundwater during the last 17 years is 1.13 g (0.034 moles). The amount of tape used in borehole sections provides a sufficient source of DEHP to produce the calculated amount of sulfides. Other mechanisms that can produce sulfide involve the decomposition of dissolved organic carbon (DOC), including low-chain organic compounds present in the fracture water at corresponding depth60, such as acetates, and small amounts of ethanol used during drilling campaigns, that are more readily degraded compared to PVC. The degradation of PVC components, in addition to DOC from the incoming groundwater, would be theoretically enough to produce the observed sulfide concentration (0.19 mg/L) in the borehole section (see calculations in SI).

Isotope fractionation signatures in the groundwater-calcite system

Removal of U following reduction of U(VI) to U(IV) species is expected to result in fractionation of primordial 238U and 235U isotopes, usually detected both as enrichment of 238U (positive δ238U) in the reduced U(IV) phase accumulated on the surface of redox-active minerals at the rock-groundwater interface, and as depletion of 238U in the aqueous unreduced residual U fraction as shown for several geological settings7,8,14. The U decrease in fracture water versus depth implies that infiltrating U(VI) has undergone reduction, followed by immobilization of U(IV) through multiple removal pathways, as shown above. In the same way, the detected incorporation of U(IV) in the calcite, and the supposed reduction of U(VI) in the borehole water suggests that fractionation between 238U and 235U might occur, in agreement with several environmental and laboratory settings7,8,27,61. In the Äspö granitoid fracture network, reduction of U(VI) and subsequent precipitation of U(IV) possibly occur along water flow paths, resulting in preferential 238U removal over 235U with the lowest measured δ238U = −2.17‰. The evidence for similar δ238U values measured across roll-front mineral deposits suggests that a similar isotope fractionation pathway might occur in Äspö fracture water7. The depleted U that occurs in borehole KA3105A in the form of U(VI)-carbonate complexes is further reduced to U(IV) by HS-(aq) and Fe(II)(aq) species results in positive δ238U for longer interaction time15 and to less extent by FeS(nc) (results in depletion of 238U)14 formed in situ in borehole sections following calcite precipitation promoted by bacterial activity. The subsequent incorporation of U(IV) into calcite results in indistinguishable 238U fractionation between groundwater and calcite with Δ238Ucalcite-GW = (δ238Ucalcite – δ238UGW) ~0‰. The negligible net isotopic fractionation suggests either a kinetically controlled single-step process with no fractionation, or more complex multi-step fractionation routes, including multiple abiotic and microbial redox species, resulting in δ238U varying both in sense and magnitudes, but with a net zero effect. The set of different abiotic (HS-(aq), Fe(II)(aq), Fe-sulfides) redox species and SRB occurs in the borehole simultaneously and might produce contrasting isotopic signatures that complicate the interpretation of observed U isotope fractionation values9,15. Abiotic reduction of U(VI) by Fe(II)(aq) and HS-(aq) species has been reported to result in positive δ238U (up to 1.7‰), whereas reduction with FeS(nc) results in negligible fractionation15 to strong 238U depletion, down to –2.3‰, in the remaining aqueous U(VI) fraction, depending on U speciation14. The greatly varying δ238U obtained from the two latter studies for similar systems suggest that there are several key parameters, e.g., U/FeS ratio and U speciation, which depend on Ca content that defines U reduction and removal kinetics, and thus the extent of isotope fractionation. Fe(II)(aq) is a dominant redox species in the stagnant water in section A:3 with ~50% higher concentration (0.29 mg/L) than for HS-(aq) (0.19 mg/L) along with detected SRB (889–2444 cells/cm3) on tape and pipes47 that might be simultaneously involved into U redox process. The U(VI) redox pathways involving bacterial species may proceed through direct reduction of U(VI) on the surface of bacteria27,62,63 and indirectly by facilitating the formation of Fe-sulfides capable of substantial U removal53 that may result in δ238U fractionation of varying sense and magnitude15. Due to the relatively low amount of bacteria found on the equipment and in the water48 at Äspö, a direct association between U and bacteria cannot be unambiguously identified. The FeS removal route supports the previous finding of Fe-sulfides overlying calcite crystals and all parts of the borehole equipment at Äspö HRL47 with extreme 34S/32S fractionation measured in pyrite, thus providing isotopic evidence of SRB activity in the borehole sections64. Although the presence of calcite and Ca in groundwater is reported to suppress the removal of U(VI)-carbonates by silicate minerals65, Fe-sulfides are expected to hinder U remobilization by fast reduction of U(VI). While UO2(CO3)34-(aq) in a Ca-free system can be readily reduced by Fe(II)(aq) species66 in Ca-rich settings, the dominating Ca2UO2(CO3)3(aq) is a more stable complex that is directly identified in deep aquifer settings even under reducing conditions due to the occurrence of Fe(II)(aq)67. A minor effect from U(VI) reduction on FeS might be due to the fact that it is formed later in stagnant water under nearly steady-state conditions47. The reduction of U(VI) and subsequent mineralogical removal of U(IV) is not expected to result in notable isotope fractionation due to kinetic effects provided that U(VI) is readily reduced by microbial sulfides in cycles starting with each fracture water inflow event caused by the flushing of the borehole during groundwater sampling. The occurrence of up to 4% of dissolved U(IV) in water may facilitate energetically favorable U incorporation into the calcite structure11, with negligible δ238U changes for ambient systems that do not involve U redox processes3. Alongside δ238U, the 234U/238U (AR) isotopic disequilibrium data can complement the explanation of the observed effects and serve as an isotopic indicator of redox conditions in rock–water systems14. The elevated ARs may result from two major mechanisms, involving weathering of the fracture minerals and preferential leaching of alpha-recoiled U fractions7. Because significant weathering is not expected in deep anoxic settings where reducing conditions dominate, leaching of recoiled 234U atoms remains a more preferred pathway creating high AR68,69,70. It is likely that complex mechanisms have affected the observed AR, including multiple U remobilization events.

Environmental implications for subsurface anoxic aquifers

Deep aquifers feature reducing conditions that are largely controlled by organic matter, microorganisms, and abiotic species, such as Fe(II). The latter determines the redox processes involved, and after forming several Fe-oxyhydroxide and/or sulfide phases, alongside calcite13 it will act as a natural or engineered sink for various trace cations and anions, including natural and artificial radionuclides, e.g., fission products71 and transuranic elements25,72. Based on detailed geochemical and hydrogeological datasets and U isotopic data, we propose a model for U removal and evolution of isotope fractionation as summarized in Fig. 5. Contrary to environments that involve bulk Fe species, e.g., Fe-sulfides and Fe-oxyhydroxides, in low Fe settings the biotic reduction of U(VI) may result in the formation of UO2 nanoclusters on mineral or bacterial surfaces53,63. More rarely reported is U removal through the formation of U(IV) nanoclusters stabilized in a coffinite-like structure included in Ca carbonate crystals34.

Fig. 5: A tentative model for U removal and δ238U during U flow in granite fracture network in the Äspö HRL.
figure 5

The degradation of PVC plasticizer is promoted by SRB resulting in the formation of bicarbonate and sulfide. Formation of sulfides in water (HS-(aq) and subsequently as FeS precipitates) is followed by several U removal pathways, including U immobilization on the surface of FeS and pyrite and U(IV) co-precipitation with calcite through structural incorporation.

Here we report deep aquifer U removal by calcite through structural incorporation of U(IV) and U accumulation on the surface of Fe-sulfides formed on magnetite crystals as a promising process for the sequestration of trace contaminants in the subsurface. In engineered systems, this could be accomplished with low molecular organic matter to produce bicarbonate in addition to conventional biotic remediation strategies for U scavenging, where continuous nutrient injection must be provided to stimulate bacterial activity to maintain U in reduced form. Bacteria promote the degradation of organics that stimulate the formation of calcite and sulfides, followed by reduction of dissolved U(VI) and subsequent immobilization of U(IV).

While calcite is a thermodynamically stable phase under certain environmental settings and oxic conditions, FeS possesses enough sink to scavenge relatively small amounts of oxygen73 but will oxidize to form elemental sulfur and goethite as ultimate oxidation products during oxygen intrusions74. In the case of nuclear waste repositories, particularly those of low-level waste, cement used for grouting will provide alkaline conditions (pH ~10) that will keep the stability of calcite high, and in addition, in this type of anoxic, circumneutral to slightly alkaline settings, secondary calcite in open water-conducting bedrock fractures have been shown to sustain dissolution for hundreds of millions of years75,76,77. This is much longer than anticipated from laboratory studies and seawater settings78,79. Thus, the sequestration of heavy metals in carbonate minerals, such as calcite, mediated by biotic/abiotic redox species, may emerge as an important route to trap toxic wastes, e.g., radioactive species. Specific implications for these U removal processes are obviously deep subsurface repositories for spent nuclear fuel80. Although SRB enhances the production of sulfide that can be corrosive to copper canisters encapsuling the spent fuel81, we show that SRB can aid in the immobilization of U into secondary mineral sinks that can act in the long-term in subsurface anoxic environments.


Borehole setting and sampling description

The studied borehole, KA3105A at 415 m depth, was drilled in 1995 into Paleoproterozoic granitoid rock at the Äspö HRL, Sweden, a full-scale test-facility tunnel built and operated by the Swedish Nuclear Fuel and Waste Management Co. (SKB)82. The borehole was drilled into the tunnel wall sub-horizontally, with a slight downwards inclination, has a length of 34 m, and was equipped with inflatable pairs of rubber packers, connected by rods of Al running through the whole length of the boreholes. The packer system isolates borehole sections intersected by targeted individuals or clusters of water-conductive fractures from which water can be extracted at the tunnel wall through polyamide tubing for each individual borehole section. A pump is not needed for sampling as groundwater flows spontaneously due to the hydrostatic pressure in the borehole. As a consequence of tunnel excavation, the groundwater in the borehole has a marine signature (as seen in the δ18O and Cl compositions) and is composed of Baltic Seawater mixed with minor proportions of meteoric water and/or older more saline water residing in the bedrock fractures83. Calcite and Fe-sulfides crystals observed with no evidence for individual UO2 or Fe(III)-oxides on the borehole instrumentation after withdrawal of the packer at the end of a 17-year experiment (1995–2012), were scraped off for laboratory analysis (average thickness of the calcite encrustations 100 µm) (Supplementary Fig. 1a–d). First, laboratory microscopic analyses showed that calcite, Fe-sulfides, barites, and Al-oxyhydroxides (not analysed by other micro techniques) precipitate together on the borehole equipment. The groundwater in each packed-off section was composed of stagnant water, which was sampled initially when water started to be discharged from the borehole section and may have been affected by the borehole instrumentation, and fracture water collected after discharge of 10 section volumes (59 L in total) and hence represents the pristine groundwater residing in the borehole fractures. Both water types were sampled for chemical analysis by SKB before extraction of the borehole instrumentation in 201213,64. Chemical data for fracture water, collected by SKB from the KA3105A borehole sections, includes four sampling occasions during the 1995–2012 period.

Microfocus X-ray fluorescence (µXRF) and U LIII-edge X-ray absorption near edge structure (µXANES)

Selected crystals of calcite were analyzed with µXRF and U LIII-edge XANES spectroscopy. U LIII-edge µXANES was recorded at two different microfocus beamlines 4-BM of the National Synchrotron Light Source, NSLS-II, Brookhaven, USA, and microfocus spectroscopy beamline at I18 of the Diamond Light Source, DLS, Harwell, UK84. For both beamlines, sample preparation was done in a similar way. Calcite crystals were fixed in between two Kapton tape stripes and positioned at a 45° angle relative to the incident beam. The X-ray beam size at the sample position was focused to 2 µm (vertical) × 5 µm (horizontal) with Kirkpatrick-Baez mirrors. For µXRF measurements, an area of ~150 µm × 200 µm was selected on calcite crystal. Crystal scanning was performed with a step size of 3 µm, 0.2 s integration time, with a total measurement time estimated at 30 min. The energy was calibrated by measuring the peak in first derivatives of K edge XANES spectra of Y foil (Eabs = 17.038 keV) at I18 (DLS) beamline or Zr foil (Eabs = 17.998 keV) at 4-BM (NSLS-II) beamline. U LIII-edge µXANES spectra were collected in a fluorescence mode within the 17.000–17.500 keV range by scanning a Si <111> double-crystal monochromator. Several XANES spectra were recorded, 15 min each, until a desired spectral quality of the average of the scans was achieved. To minimize possible beam damage effects, each spectrum was collected at a new position on a U-rich rim area of 30 µm × 150 µm. Preliminary spectra analysis was done using PyMca software85, and normalized with the ATHENA program from DEMETER software package86. µXANES spectra collected at different microfocus beamlines were aligned by comparing the positions of the first inflection points determined for the same calcite material.

U(IV) and U(VI) references. Natural calcite sample reported by Sturchio et al.11 as used as U(IV) standard for XANES (as U(IV)-calcite). U(VI)-rich calcite material collected from speleothems was used as U(VI) reference (as U(VI)-calcite)20 Reference materials were used for U LIII-edge µXANES spectroscopy according to their availability at 4-BM (NSLS-II, USA) and I18 (DLS, UK) beamlines. The same calcite material was used for both beamlines and spectra were used for the energy correction (Supplementary Fig. 1a, b).

Extended X-ray absorption fine structure (EXAFS) measurement and data fitting

U EXAFS measurements were performed at the National Synchrotron Light Source II (NSLS-II), Brookhaven National Laboratory (USA) at beamline 4-BM (XFM). The XFM beamline utilizes an air-bearing fixed-exit Si <111> double-crystal monochromator87. Calcite crystal was packed into a double layer of 50 µm Kapton tape and mounted on a motorized three-axis sample stage, positioned at 45° to the incident beam. EXAFS spectra were recorded in a fluorescence mode using Canberra SXD-7 SDD coupled to Quantum Detectors Xspress3 DXPs. In total, 122 scans were recorded 20 min per scan in nine discrete spots. Data were extracted, and Fourier transforms were applied over the k-range 2.5–8.5 Å−1 using the ATHENA and ARTEMIS programs from the IFFEFIT program package86. The fit was performed in R space for the 1.00–4.10 Å range. The scattering paths were generated with the FEFF8.2 code implemented in ARTEMIS using calcite crystal structure (AMCS database no. 0017866). Two short U-O1 and U-O2, one U-C, and long U-O (denoted as U-O3) and one U-Ca single scattering paths (in total, five) were used to fit the EXAFS spectrum. A shell-by-shell modeling approach was followed in the EXAFS analyses. The coordination number of U-C, U-O3, and U-Ca were fixed at 6. The amplitude reduction factor was set to 1. Other parameters were varied.

Uranium isotope (238U/235U) measurement

Samples of calcite were weighed and dissolved in 0.5 M acetic acid and dried down. The samples were then treated with concentrated acetic acid and dried down, before finally bringing them up in 3 M nitric acid. In the case of the groundwater samples, a volume of 100 ml was used for U isotopic analysis. To correct for mass fractionation artefacts due to column chemistry and mass spectrometry, we used a 233U-236U mixed-spike, IRMM 3636 (obtained from European Commission Joint Research Inst.) with a 233U/236U of 1.01906(16) (2019 ECJRI certificate), to produce a 238U/236U ratio of ~20. The 233U-236U mixed-spike was added to the aliquots for U isotopic analysis prior to chemical separation, and the mixtures dried down and were taken up in 3 M HNO3, repeating this cycle to insure sample/spike equilibration. Uranium separation from the sample/spike mixtures was conducted using UTEVA resin (Eichrom Technologies) on small Teflon columns following the method described by ref. 88. 238U/235U is represented by δ238U, the deviation of the sample 235U/238U ratio in parts per thousand from that of the CRM145 U standard. The δ235U values (deviation of the 235U/238U from that of CRM145 in parts per thousand) and 234U/238U ratios (reported as an activity ratio relative to secular equilibrium, where 1 is secular equilibrium) of the samples were measured using a Neptune (Thermo Fisher), a plasma source multi-collector magnetic sector mass spectrometer housed at Lawrence Berkeley National Lab (Center for Isotope Geochemistry). The estimated external reproducibility (2σ) of the δ238U measurements is 0.1–0.15‰ based on replicate measurements of CRM112a. Further details of the mass spectrometry method can be found in ref. 14.

High-resolution laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS)

LA-ICP-MS mappings were performed at the MAGMA Lab of the TU Berlin using an Agilent 8900 ICP-MS/MS coupled to an Analyte Excite 193 nm excimer laser (Teledyne Photon Machines). Helium is used as a carrier gas with a total flow rate of 0.95 l/min (0.5 l/min cell flow, 0.45 l/min cup flow). After plasma ignition and initialization of Helium flow, m/z = 42 (e.g., 14N14N14N+) and m/z = 31 (e.g., 15N16O) were recorded to monitor the amount of air entrained in the interface tubing and ablation cell. After counts were stable and below 20000 cps, the Ar nebulizer gas and He carrier gas flows were tuned to achieve m/z = 248/232 ratios of <0.15% (ThO/Th) and m/z = 232/238 ratios of 100 ± 1% (Th/U) while ablating NIST 610 in line scan mode. Once stable plasma conditions were reached, an automatic lens tune was performed on NIST 610 to maximize sensitivity in the high mass range. Laser parameters for the tuning were 100 Hz repetition rate, 5 µm square spot size, 50 µm/s scan speed, and 2 J/cm2 fluence. The resulting gas and plasma parameters are summarized in Supplementary Table 4 and were used for the actual mappings. To reduce sampling time without compromising spatial resolution, the aerosol rapid introduction system ARIS (Teledyne Photon Machines) was used89. The single pulse response (SPR), i.e., the combined wash-in and wash-out, was monitored by ablating NIST 612 glass at 1 Hz in line scan mode and recording m/z = 238 (U) with a dwell time of 5 ms. By averaging the peak shape of 300 single pulses, the SPR was determined and was 30 ms (full width at 10% of the maximum intensity). For imaging, the elemental distribution within the target grains, three mass channels m/z = 34 (S), m/z = 57 (Fe), and m/z = 238 (U) were recorded with dwell times of 10 ms each. This resulted in signal-to-noise ratios of ~ 200, ~20,000, and ~2000, which were obtained from a single-line scan across one of the target grains. Scan speed and repetition rate were chosen to suppress aliasing artifacts and maximize spatial resolution for the given SPR, dwell times and the spot size of 5 µm. Grains were mapped using a raster of unidirectional scans with no overlap between lines. By applying the above parameters, mapping of a 250 × 250 µm2 area was completed within ~2 min. All the raw data processing was performed using the software HDIP (Teledyne Photon machines). The background was corrected using a cubic spline function. To obtain quantitative mappings, a multi-standard calibration was performed by using multiple line scans on reference materials NIST 612 and BCR-2G recorded before and after the mapping of individual grain. No drift occurred over the short measurement period of <30 min, which was evident from using NIST 612 as a drift monitor.