Pore-scale evaluation of uranyl phosphate precipitation in a model groundwater system


Corresponding author: C. J. Werth, University of Illinois at Urbana-Champaign, Department of Civil and Environmental Engineering, 205 N Mathews, Urbana, IL 61801, USA. (werth@uiuc.edu)


[1] The abiotic precipitation of uranium (U(VI)) was evaluated in a microfluidic pore network (i.e., micromodel) to assess the efficacy of using a phosphate amendment to immobilize uranium in groundwater. U(VI) was mixed transverse to the direction of flow with hydrogen phosphate (HPO42−), in the presence or absence of calcium (Ca2+) or sulfate (SO42−), in order to identify precipitation rates, morphology and types of minerals formed, and effects of mineral precipitates on pore blockage. Precipitation occurred over the time scale of hours to days. Relative to when only U(VI) and HPO42− were present, precipitation rates were 2.3 times slower when SO42− was present, and 1.4 times faster when Ca2+ was present; larger crystals formed in the presence of SO42−. Raman backscattering spectroscopy and micro-X-ray diffraction results both showed that the only mineral precipitated was chernikovite, UO2HPO4 · 4H2O; energy dispersive X-ray spectroscopy results indicate that Ca and S are not incorporated into the chernikovite lattice. A pore-scale model was developed, and simulation results of saturation ratio (SR = Q/Ksp) suggest that chernikovite is the least thermodynamically favored mineral to precipitate (0 < SR < 1) compared to uranyl hydrogen phosphate and Na-autunite (13 < SR < 40), and uranyl orthophosphate and Ca-autunite (when Ca2+ is present; SR > 105). Fluorescent tracer studies and laser confocal microscopy images showed that densely aggregated precipitates blocked pores and reduced permeability. The results suggest that uranium precipitation with phosphate as chernikovite is rapid on the time scale of remediation for the conditions considered and can block pores, alter fluid flow paths, and potentially limit mixing and precipitation.

1. Introduction

[2] Contamination of groundwater and soil with uranium has resulted from anthropogenic activities such as uranium mining and milling and processing of uranium for nuclear weapons and energy programs [Wolbarst et al., 1999; National Research Council (NRC), 2010], as well as from natural processes in rock, soil, and water. Uranium contamination causes health risks and presents serious public health concerns [Zimmerman et al., 2007; NRC, 2008]. A great deal of resources has been directed toward identifying efficient in situ strategies to remove soluble U(VI) from groundwater, including promotion of in situ biological reduction (see summaries in Fang et al. [2009] and Dunham-Cheatham et al. [2011]), and abiotic reactions including adsorption and mineral precipitation [Morrison and Spangler, 1992; Naftz et al., 1996; Gabriel et al., 1998; United States Environmental Protection Agency (USEPA), 2000; Simon and Biermann, 2005; Wellman et al., 2008; Hiemstra et al., 2009]. Among these approaches, abiotic precipitation has received considerable attention due to its success in uranium removal from soil and groundwater [Arey et al., 1999; USEPA, 2000; Fuller et al., 2002; Simon et al., 2008].

[3] A variety of U(VI) mineral precipitates has been observed at sites where anthropogenic activities and/or weathering resulted in the release of uranium to the environment, and in associated laboratory studies. Near several uranium mines (Apex Mine, Nevada, and Midnite Mine, Washington) where sulfate is present, the mineral uranopilite [(UO2)6SO4(OH)10 · 12H2O] was found [Frost et al., 2007]. Near the Coles Hill U deposit in Virginia, Ba uranyl phosphate (Ba meta-autunite, [Ba(UO2)2(PO4)2 · nH2O]) is present [Jerden and Sinha, 2003; Jerden et al., 2003; Jerden and Sinha, 2006]. In the vicinity of a secondary uranium ore deposit at Koongarra, Australia, both Mg meta-autunite [Mg(UO2)2(PO4)2 · nH2O] and nanocrystals of metatorbernite [Cu(UO2)2(PO4)2 · 8H2O] were identified [Murakami et al., 1997, 2005]. At the Hanford site in Washington, several silicon-containing U(VI) minerals were identified in the vadose zone directly above the water table, i.e., uranophane-type phases [Ca(UO2)2(SiO3OH)2(H2O)5] [Um et al., 2010] and sodium boltwoodite [Na(UO2)(SiO3OH) · 1.5H2O] [Catalano et al., 2004], as well as a copper-containing mineral below a discharge pond, i.e., metatorbernite [Cu(UO2)2(PO4)2 · 8H2O] [Arai et al., 2007]. The facultative anaerobe Rahnella sp. Y9602, isolated from soil at the Oak Ridge Field Research Center, was used to release orthophosphate by hydrolysis of glycerol-3-phosphate and promote abiotic precipitation of chernikovite [UO2HPO4 · 4H2O] [Beazley et al., 2009]. In the presence of Ca2+, chernikovite can form the more stable mineral Ca meta-autunite [Ca(UO2)2(PO4)2 · nH2O] via ion exchange [Martinez et al., 2007], i.e., the replacement of hydrogen in two unit cells of chernikovite with Ca2+. Among the aforementioned mineral phases, those containing phosphate have received the most attention because they are stable under oxic conditions (e.g., in contrast to UO2), their precipitation rates are generally faster, and/or their solubilities are generally lower near circumneutral pH compared to nonphosphate-containing minerals; they are the focus of this effort.

[4] Phosphate has been amended as natural apatite materials (e.g., phosphate rock, bone meal, and bone meal charcoal), including the mineral hydroxylapatite (HAP) [Ca (PO4)3OH], in batch and column studies [Fuller et al., 2002; Arey et al., 1999; Simon et al., 2008] and in a permeable reactor barrier (PRB) wall [Fuller et al., 2003; USEPA, 2000]. In batch and column studies with low uranium concentrations (e.g., <7 mg U(VI) g HAP−1), immobilization occurred through U(VI) sorption to the HAP or natural apatite materials (e.g., bone char) [Fuller et al., 2002, 2003; Simon et al., 2008], not U(VI) precipitation. However, for higher U(VI) concentrations (e.g., 5500 mg U(VI) g bone char−1), and in the absence of dissolved carbonate, X-ray diffraction (XRD) results indicated that uranium precipitated with phosphate as chernikovite, autunite, and/or saleeite [Mg(UO2PO4)2 · 10H2O] [Fuller et al., 2002, 2003; Simon et al., 2008]. Uranium precipitation can be inhibited in the presence of dissolved carbonate due to the formation of highly soluble uranyl carbonate complexes [e.g., Ca2UO2(CO3)3(aq)] [Fuller et al., 2003]. In Fry Canyon, Utah, a PRB wall containing a natural apatite material (i.e., bone char) was used to remove U(VI) in the presence of approximately 4.8 mM total dissolved carbonate [USEPA, 2000]. Uranium in groundwater was reduced from as high as approximately 4.7 mg L−1 to as low as 0.06 mg L−1. Sorbed uranium concentrations in recovered bone char were as high as 690 µg U(VI) g−1. XRD and XAS (X-ray adsorption spectroscopy) were used to analyze reactive barrier wall material posttreatment, and no uranium precipitates were observed. In batch studies, similar results were obtained using the same materials and conditions, but precipitates were observed when carbonate was removed [Fuller et al., 2003]. This indicates that uranium removal occurred primarily by U(VI) sorption through surface complexation with apatite surfaces, and U(VI) precipitation as phosphate minerals is inhibited by the presence of dissolved carbonates.

[5] To overcome phosphate limitations associated with solid phase apatites, aqueous phase phosphate has been directly used to precipitate uranium from water in batch studies, amended to groundwater and uranium impacted Hanford sediments in batch and column studies [Fuller et al., 2002], and amended to a uranium impacted groundwater aquifer at Pacific Northwest National Laboratory's Hanford site (300 area) in Washington state [Wellman et al., 2007; Vermeul et al., 2009]. In the batch study containing only water, phosphate was added by first equilibrating water with apatite, and then the aqueous phase with phosphate was separated from apatite and amended with equimolar amounts of uranium [Fuller et al., 2002]. Chernikovite precipitation occurred when phosphate and uranium concentrations each exceeded 100 µM. In batch studies with Hanford groundwater and sediments, multiple phosphate sources were evaluated, including sodium salts of orthophosphate, pyrophosphate, tripolyphosphate, trimetaphosphate, and hexametaphosphate and calcium salts of dihydrogen phosphate, hydrogen phosphate, and pyrophosphate [Vermeul et al., 2009]. The sodium salts of orthophosphate, pyrophosphate, and tripolyphosphate resulted in the most uranium sequestration, and sequestration increased in the presence of calcium chloride. In column studies containing Hanford groundwater and sediments, polyphosphates and calcium were sequentially flushed through the column and used to sequester uranium. Posttreatment analysis of the sediments using short-wave UV radiation (254 nm) showed fluorescent green precipitates were present, indicative of uranium phosphate minerals. Similar to batch and column studies, uranium sequestration was also observed in Hanford 300 area groundwater after polyphosphate addition. Mineral precipitation was indicated by changes in permeability, suggesting pore blockage, but no mineralogical data are available to assess removal mechanisms. In none of these studies, mineral precipitation rates were evaluated, and the effect of phosphate mineral precipitation on pore blockage was not further evaluated. Hence, it is not clear if phosphate addition under mixing limited conditions in groundwater resulted in uranium precipitation, or if mixing limitations prevented widespread precipitation and resulted in pore blockage.

[6] The objectives of this work were to quantify uranyl phosphate precipitation rates at the pore scale under representative flow, mixing, and geochemical conditions, the corresponding effects of selected groundwater ions on these rates, the type and morphology of uranium minerals that precipitate, and the resulting pore blockage. Uranium precipitation with phosphate was promoted along a transverse mixing zone in a microfluidic pore network (i.e., micromodel) in the presence and absence of calcium and sulfate ions. Precipitation rates were quantified using brightfield reflected microscopy images, and mineral phases and compositions were evaluated in situ using Raman backscattering spectroscopy and ex situ using micro-X-ray diffraction (µ-XRD), environmental scanning electron microscopy (SEM), and energy dispersive spectroscopy. Pore blockage was evaluated using scanning confocal microscopy and breakthrough of a fluorescent tracer. The effects of calcium and sulfate were examined because they are common groundwater constituents, and because they are found in U(VI) minerals identified at several important field sites. A pore-scale reactive transport model with U speciation was used to evaluate the effects of pore-scale reactive transport on dominant U species in the pore space, and the degree of saturation with respect to different U(VI)-phosphate mineral phases. To our knowledge, this is the first study of uranium phosphate precipitation in a flow cell amenable to direct visualization and analysis of mineral precipitates and is the first direct comparison of experimental precipitates with pore-scale reactive transport model predictions. The unique approach allows us to test the hypotheses that the solution conditions (i.e., the presence of Ca2+ and SO42−) affect the type of minerals formed during precipitation, and the precipitation rate, morphology, and effects on flow and mixing.

2. Materials and Methods

2.1. Materials

[7] Uranyl nitrate (UO2(NO3)2, Fluka), sodium phosphate dibasic (Na2HPO4, Sigma-Aldrich), calcium chloride (CaCl2, Sigma-Aldrich), sodium sulfate (Na2SO4, Sigma-Aldrich), and fluorescein (70%, Sigma-Aldrich) were ACS reagent grade (>99.0% purity) unless otherwise noted. All solutions were prepared in water treated with the Barnstead Nanopure purification system (18.2 MΩ cm).

2.2. Micromodel Fabrication

[8] Standard photolithography techniques were used to fabricate all micromodels in silicon, as described by Chomsurin and Werth [2003] and Willingham et al. [2008]. The etched micromodel and pore structure are illustrated in Figure 1; the pore structure consists of a homogeneous array of cylinders arranged in a staggered pattern. Individual cylinders are 300 µm in diameter and spaced 180 µm apart in pore bodies and 40 µm apart in pore throats, resulting in a porosity of 0.39. The entire micromodel contains two inlet channels each with an inlet hole (A and B) drilled completely through the silicon using a 1.25 mm diameter diamond plated drill, a 2 cm × 1 cm area containing the pore network, and an effluent channel with an outlet drilled through the silicon wafer. The pore structure and inlet and outlet channels were etched to a depth of 35 µm. A 500 µm thick sheet of Pyrex glass was anodically bonded to the micromodel surface to create the flow channels. Nanoports (Upchurch Scientific) were attached to the back of the micromodel and held by adhesive. Teflon tubing (250 µm inside diameter) delivered fluid from two Milligat Flowpro pumps from solution reservoirs to each micromodel inlet via three-way PTFE (Polytetrafluoroethylene) valves.

Figure 1.

(a) Schematic of micromodel with labeled inlets and outlets and (b) pore structure with a homogeneous array of pores. The pore structure in (a) is 2 cm long by 1 cm wide. The individual grains in (b) are 300 µm in diameter.

2.3. Micromodel Experiments

[9] Precipitation experiments were performed using three different sets of inlet solutions listed in Table 1. Uranyl nitrate (UO2(NO3)2, 100 µM) and sodium phosphate dibasic (Na2HPO4, 100 µM), were prepared in nanopure water. Calcium chloride (CaCl2, 100 µM) was dissolved in a 100 µM UO2(NO3)2 solution, and sodium sulfate (Na2SO4, 100 µM) was dissolved in a 100 µM Na2HPO4 solution to yield combined solutions. The concentration of U(VI) was chosen to represent those found at the Oak Ridge Field Research Center (84–210 µM) [Wu et al., 2006]; it exceeds maximum concentrations found at the Hanford site 300 area (4.2 µM) [Zachara et al., 2005] and the Fox Canyon site (23 µM) [USEPA, 2000]. HPO42−, Ca2+, and SO42− were added in a 1:1 molar ratio with U(VI). For comparison, Ca2+ concentrations include values from 0.25 to 4.2 mM at the Hanford site [Moody et al., 1996; Simon et al., 2008; Zachara et al., 2005] and 2.5 to 41 mM at the Oak Ridge site [USEPA, 2000]. For SO42−, concentrations include values from 0.25 to 0.61 mM at the Hanford site [U.S. Department of Energy, 1992; Zachara et al., 2005; Vermeul et al., 2007] and a value of 2.13 mM at the Oak Ridge site. Hence, experimental Ca2+ and SO42− concentrations in this work are relatively low compared to the U.S. Department of Energy (DOE) sites. All inlet solutions were maintained in bottles open to the atmosphere at 31°C±1°C, and micromodels were kept at room temperature (21°C±0.5°C); the heated solutions were pumped through tubing that allowed the solutions to cool to 21°C before entering the micromodels. This approach ensured that dissolved gas concentrations were below saturation at 21°C and eliminated the evolution of gas bubbles in the micromodels. This approach also resulted in total dissolved carbonate concentrations (i.e., Tot(CO3)) less than 0.11 mM, which are low compared to the values up to and exceeding 10 mM that can occur at some DOE sites such as Hanford and Rocky Flats [Zachara et al., 2005; Bostick et al., 2003]. The relatively low Ca2+ and Tot(CO3) concentrations evaluated in this work allow a detailed analysis of uranium precipitation in the absence of other minerals. Higher Ca2+ and Tot(CO3) concentrations are evaluated as part of our modeling effort to explore how these constituents may affect mineral precipitation.

Table 1. Experimental Matrix With Inlet Conditions
Experimental NamePrecipitation Conditions
Inlet AInlet B
UP100 µM UO2(NO3)2, pH 4.5100 µM Na2HPO4, pH 6.7
UCaP100 µM UO2(NO3)2, 100 µM CaCl2, pH 4.1100 µM Na2HPO4, pH 6.7
UPS100 µM UO2(NO3)2, pH 4.5100 µM Na2HPO4, 100 µM Na2SO4, pH 7.3

[10] During precipitation experiments, inlet A was purged with UO2(NO3)2 alone or in combination with CaCl2. Inlet B was purged with Na2HPO4 alone or in combination with Na2SO4. Experiments with UO2(NO3)2 and Na2HPO4 are referred to as UP; those with CaCl2 added are referred to as UCaP; and those with Na2SO4 added are referred to as UPS (Table 1). Precipitation in each experiment was allowed to occur continuously for approximately 28 h. The volumetric flow rate was 75 µL h−1 through each syringe, corresponding to a Darcy velocity of 0.71 cm min−1 in the micromodel pore structure. This is similar to flow velocities observed in sands with high conductivities, such as at the 300 area Hanford site. In all images, fluid flow is from left to right, and the fluid residence time is approximately 65 s in the pore structure. All experiments were performed in duplicate.

2.4. Micromodel Visualization

[11] Experiments were imaged with a Nikon microscope (Eclipse Ti) equipped with 4X, 10X, and 40X inverted objectives, a digital camera (Photometrics Coolsnap HQ2), and an automated stage (Märzhäuser Wetzlar). The computer software NIS-Elements was used to control the microscope components and image acquisition. For example, the software was programmed to execute 26 × 3 stage movements and 78 image-capture events with the 10X objective (0.6 µm pixel resolution) at each time point and to montage the individual images into a complete image of the entire pore network. Using the same microscope, precipitates were analyzed through the glass cover slip using Raman backscattering spectroscopy for mineral identification. A laser of wavelength 532 nm was manually focused on an area with precipitate prior to each scan. The scattered light was collected by the focusing objective and passed through a Horiba Jobin Yvon LabRam-HR spectrometer with 1600 mm−1 gratings and captured by an air-cooled carbonate compensation depth. Collected spectra are an average of 10 or more exposures, each with a 10 s acquisition time. The Raman spectra were processed using the software program LabSpec. Precipitates were analyzed ex situ using a Rigaku µ-XRD system. X-rays were generated (MicroMax 007HF operated at 35 kV and 25 mA) from a rotating Cr target (λ = 2.2910 Å) and focused through a 300 µm diameter collimator. The X-ray beam cannot evaluate minerals through the glass cover slip, so the micromodel was cut across the outlet channel and the glass in this region with fixed mineral precipitates was broken free and mounted on the stage of a Rigaku D/Max Rapid II goniometer for analysis. Diffraction data recorded on a 2-D image plate were integrated between 10° and 160° 2θ to give powder traces. Exposure time for the samples was 10 min each. Comparison with database patterns (International Centre for Diffraction Data (ICDD), Pennsylvania) was carried out using JADE v9.2 (Materials Data Inc., California). The same sample was evaluated using a FEI Helios Nanolab dual-beam focused ion beam/SEM equipped with energy dispersive X-ray spectroscopy (EDXS). Samples were coated with carbon prior to analysis.

[12] Laser confocal imaging was used to determine whether mineral precipitates spanned the depth of the microfluidic pore structure and potentially blocked flow. The pore networks for UP, UCaP, and UPS experiments were filled with the fluorescein dye solution to provide contrast, and one upgradient pore and one downgradient pore were imaged using a Zeiss LSM710 multiphoton laser-scanning confocal microscope (Carl Zeiss MicroImaging, Inc.). Slices of 1 µm thick were acquired with a 488 nm Ar ion laser for excitation and detected at 500–570 nm with a 40X long working distance water immersion objective (approximately 1 µm resolution). They were montaged to obtain 3-D images of the precipitates.

[13] Tracer tests were conducted with a fluorescein dye (10 µM) in order to determine whether mineral precipitates completely blocked pores. The dye was injected through inlet A at 100 µL h−1, and nanopure water was injected through inlet B at 50 µL h−1. When precipitates are not present, the fluorescent tracer occupies approximately two thirds of the pore structure. When precipitates block the pores, the fluorescent tracer is confined to the pore structure above the zone of precipitation along the centerline of the micromodel pore structure. The fluorescent dye was imaged using fluorescent microscopy and an ET-GFP (Green Fluorescent Protein) filter.

2.5. Image Analysis

[14] Mineral precipitates are darker (lower in intensity) than silicon etched flow channels containing water and silicon grains. Shadows extending 0–10 µm around each grain were also present and of similar intensity to mineral precipitates. A blank pore structure was created numerically, and grain sizes were adjusted to match grain plus shadow dimensions in the individual images; image subtraction was then used to eliminate silicon grains and shadows from further evaluation. Image thresholding of gray-scale intensity was used to distinguish mineral precipitates from etched flow channels containing water, and the number of pixels with precipitates was summed. Thresholding was performed over the unit region (480 µm × 480 µm) surrounding each cylinder using the Otsu [1979] method, which minimizes overlap of gray-scale intensity distributions for mineral precipitates and water, and then the threshold values obtained from the Otsu method were decreased by 0%–10% to correct for any remaining shadow effects (the intensity of the shadow was slightly higher than for precipitates). The estimated error in precipitation area from uniformly adjusting threshold values from the Otsu method over the entire image was smaller than 5%, compared to the cases with manually adjusted threshold values from the Otsu method for each unit region. Total area was determined by multiplying the number of pixels by individual pixel area, and total volume was determined by assuming precipitates occupied the entire pore depth. It is unlikely that all precipitates spanned the entire pore depth, and observed gray-scale variations suggest packing density among mineral precipitates varied. Hence, estimated areas and volumes are semiquantitative and must be interpreted with caution.

[15] Precipitation rates were relatively constant over much of the experimental time and were approximated by linear regression of precipitate area versus time data from duplicate experiments. The areas of individual crystalline precipitates were determined in approximately five pores along the micromodel length in each unique precipitation experiment. The range of area values that individual crystalline precipitates occupied was determined based on the visual inspection of more than 40 crystals for each experiment. ImageJ [Rasband, 1997–2011] was used to count the area of each crystal, and an average area and standard deviation were determined.

[16] The transverse distance of continuous precipitates in the horizontal center of each pore was measured to determine if the precipitate width changed from pore to pore along the length of each micromodel. The edges of the continuous precipitate were selected manually, and the ImageJ [Rasband, 1997–2011] was used to measure the distance.

2.6. Pore-Scale Reactive Transport

[17] Pore-scale reactive transport was simulated to determine initial distributions of reactive solutes and saturation ratios (SRs) for each experimental system in the micromodel. Since only the initial distributions were simulated, the reactive transport model output was not used to calculate the amount of precipitation that occurred or to distribute precipitate in the micromodel pore structure. Fluid flow and reactive transport were solved in two dimensions because concentration gradients in the vertical dimension (i.e., with micromodel depth) are eliminated by fast diffusion over very short length scales [Willingham et al., 2008]. The lattice-Boltzmann method (LBM) was used to solve fluid velocities in the pore space as described in our previous work [Knutson et al., 2005]. Fixed pressure boundary conditions were imposed along the inlet (x = 0) and outlet (x = L), and the resulting steady flow field was scaled to the Darcy velocity used in the experiment. A finite volume method (FVM) was used for reactive transport and is an extension of the model we developed to simulate biofilm growth [Knutson et al., 2005], bimolecular reactions [Willingham et al., 2008, 2010], and heterogeneous reaction of calcium carbonate [Yoon et al., 2012]. In this study, we assume homogeneous reactions are instantaneous, and the reactive transport equations in terms of the total concentration of the primary species inline image in the liquid phase (Ωp) are given by

display math(1)
display math(2)

where u is the steady velocity field from LBM, D is the diffusion coefficient, υji are the stoichiometric coefficients, Cj and Ci are the solute concentrations of primary and secondary species in the liquid phase, respectively, and Neq is the number of secondary species (subject to equilibrium reaction). In this study, the same diffusion coefficient for all species was used to avoid the violation of charge balance [Lichtner, 1985; Liu et al., 2011]. Fixed concentrations and a zero gradient boundary condition were used at the inlet and outlet, respectively. The primary species (i.e., component species) in the model include UO22+, H2PO4, Ca2+, SO42−, H+, CO32−, NO3, and Na+. The secondary species concentrations (Ci) are given through the mass action law as functions of the primary species concentrations,

display math(3)

where γi is the activity coefficient, and Ki is the equilibrium constant. The activity coefficient is computed from the extended Debye-Hückel (D-H) equation. All constant values in the D-H equation are from Helgeson and Kirkham [1974] and Helgeson et al. [1981]. Once total concentrations of primary species are solved (equations (1) and (2)), equation (3) was solved simultaneously using the Newton-Raphson method. Accuracy of calculation was verified by comparing the speciation results for three different solutions in Table 1 over a range of pH values (4–9) with those calculated using MINEQL+ [Schecher and McAvoy, 1999] and Geochemist's Workbench [Bethke and Yeakel, 2009].

[18] The SR is defined as the ratio of ion activity product (Q) to the saturation product (Ksp) and is given by

display math(4)

[19] An SR value less than one indicates that precipitation is not thermodynamically favorable. Aqueous reactions and stability constants used in this study are listed in Table A1. These were obtained from the EQ3/6 thermodynamic database [Wolery, 1992], the Nuclear Energy Agency database for uranium [Bernhard et al., 2001; Guillaumont and Mompean, 2003], and Dong and Brooks [2006].

[20] We identified six uranium solid phases and HAP [Ca5(PO4)3(OH)] as possible precipitates for the experimental conditions evaluated in this work. The six uranium solid phases are chernikovite, uranyl hydrogen phosphate (UHP), uranyl orthophosphate (UOP), Na-autunite, Ca-autunite, and metaschoepite. Two other less thermodynamically stable forms of calcium phosphate were initially evaluated because they may form in preference to HAP, i.e., amorphous calcium phosphate (Ca3(PO4)2 · xH2O) and octacalcium phosphate (Ca4H(PO4)3 · 2.5H2O) [Van Kemenade and De Bruyn, 1987; Castro et al., 2012], but were not further considered because SRs for these solids were always much less than one under conditions considered in this work. The log Ksp values of these minerals were obtained from a recent review and updated experimental data [Grenthe et al., 1992; Gorman-Lewis et al., 2008a, 2008b, 2009; Zhu et al., 2009]; they are listed in Table A2. In some cases, multiple Ksp values were available in the literature; either the most recent, the medium, or a range was considered. For example, log Ksp values for chernikovite varied from −22.73 [Van Haverbeke et al., 1996] to −25.1 [Johnson, 2000]; a log Ksp value of −24.2 was used because it was recommended by Grenthe et al. [1992] and recently used by Guillaumont and Mompean [2003]. The log Ksp value of −25.1 was also evaluated to determine maximum SRs for this mineral. The log Ksp values for UHP, UOP, and Ca-autunite were reported as mean values (−25.52, −49.36, and −48.36, respectively), plus or minus two standard deviations from reported experimental data [Gorman-Lewis et al., 2009]. The mean and standard deviation values were used in this work. Single values of log Ksp were used for Na-autunite and metaschoepite; they are from the literature and were used without consideration of errors. For HAP, two log Ksp values obtained from the synthetic HAP dissolution experiments were used (−53.28 from Zhu et al. [2009] and −58.0 from Valsami-Jones et al. [1998]).

2.7. Model Application Scenarios

[21] Simulations were performed in the first three pore bodies to reduce computational time. SRs are highest at this location because concentrations are highest near the inlets; this provides the maximum initial SRs and allows for direct comparison to minerals that precipitated in micromodel experiments. A uniform grid spacing of 5 µm is used for both the LBM and FVM, which results in 60 grid cells across a grain and 7 grid cells across a pore throat. A grid spacing of 5 µm was previously used to simulate different biochemical reactions in an identical pore structure with adequate accuracy [Knutson et al., 2005; Willingham et al., 2008, 2010; Yoon et al., 2012]. Inlet conditions evaluated by the model are listed in Table 2. These include the experimental conditions and two additional conditions that consider either a higher concentration of Ca2+ (10 mM) or a higher concentration of Tot(CO3) (5 mM) compared to the experimental UCaP conditions (i.e., Tot(Ca) = 0.1 mM and Tot(CO3) = 10−2 mM). These two additional cases were simulated to evaluate thermodynamic favorability of mineral precipitation under Ca2+ and total carbonate concentrations more representative of DOE field sites as described previously. Inlet concentrations of total primary species are computed based upon both measured and calculated pH values of inlet solutions. Calculated values were determined from inlet solution concentrations in Table 1 and carbonate species concentrations based on equilibration of influent solutions with atmospheric CO2 at 31°C. Differences in measured and calculated values are likely due to insufficient equilibration of influent solutions with the atmosphere and small errors in pH measurement. We will show that these differences only marginally affect calculated SR values. The diffusion coefficients for all aqueous species are assumed to be equal and were set to 1.0 × 10−5 cm2 s−1 as a reference value. Yoon et al. [2012] demonstrated that the diffusion coefficient can affect mineral precipitation rates. Consequently, two additional D values of 0.3 × 10−5 and 2 × 10−5 cm2 s−1 were evaluated; this represents the upper and lower bounds of D values for elements of interest in radionuclide transport [Lerman, 1979].

Table 2. Total Concentration of Inlet Reactants Used in Simulations
Total ConcentrationaUP Experiment (Measured)UP Experiment (Calculated)aUCaP Experiment (Measured)UCaP Experiment (Calculated)aUCaP High Ca2+ CaseUCaP High Tot(CO3) CaseUPS Experiment (Measured)UPS Experiment (Calculated)a
Inlet AbInlet BcInlet AbInlet BcInlet AbInlet BcInlet AbInlet BcInlet AbInlet BcInlet AbInlet BcInlet AbInlet BcInlet AbInlet Bc
  1. a

    Total concentrations of the primary species were computed using equation (2), and aqueous speciation reactions are listed in Table A1. Units are in the molar concentration (M).

  2. b

    Inlet A contains UO2(NO3)2 in all experiments, in addition to CaCl2 in the UCaP experiment.

  3. c

    Inlet B contains Na2HPO4 in all experiments, plus Na2SO4 in the UPS experiment.

Tot(UO2)1.00E-4 1.00 E-4 1.00 E-4 1.00 E-4 1.00 E-4 1.00 E-4 1.00 E-4 1.00 E-4 
Tot(H)2.00 E-51.13 E-41.63 E-42.67 E-58.00 E-51.13 E-41.63 E-42.67 E-58.00 E-51.13 E-47.50 E-35.00 E-32.00 E-51.00 E-51.63 E-42.64 E-5
Tot(H2PO4) 1.00 E-4 1.00 E-4 1.00 E-4 1.00 E-4 1.00 E-4 1.00 E-4 1.00 E-4 1.00 E-4
Tot(CO3)1.02 E-51.06 E-41.02 E-56.32 E-51.02 E-51.06 E-41.02 E-56.32 E-51.02 E-51.06 E-45.00 E-35.00 E-31.02 E-56.30 E-51.02 E-56.30 E-5
Tot(Na) 2.00 E-4 2.00 E-4 2.00 E-4 2.00 E-4 2.00 E-4 2.00 E-4 4.00 E-4 4.00 E-4
Tot(NO3)2.00 E-4 2.00 E-4 2.00 E-4 2.00 E-4 2.00 E-4 2.00 E-4 2.00 E-4 2.00 E-4 
Tot(Ca)    1.00 E-4 1.00 E-4 1.00 E-2 1.00 E-4     
Tot(SO4)             1.00 E-4 1.00 E-4
pH (31°C)4.456.73.837.

3. Results and Discussion

3.1. Precipitate Morphology

[22] Brightfield reflected microscopy images for one of the UP experiments at two time points during precipitation are shown in Figure 2. Uranium and phosphate mixed along the horizontal center region of the micromodel pore structure due to diffusive mixing and precipitated in this region. Minerals formed first on the influent side of the micromodel pore network and later over the entire length of the pore network. Minerals formed more quickly upgradient because reactant concentrations were greater (i.e., greater SRs); diffusive mixing downgradient diluted the concentrations, making precipitate formation slower [Zhang et al., 2010]. Upgradient at early time (7.5 h), and along the entire length of the micromodel at late time (27.5 h), precipitates appear as dense agglomerations of small crystals in the center of pore bodies. Less precipitation occurs in pores directly adjacent to the middle pore due to diffusion-limited mixing across the horizontal centerline. The same behavior was observed for mineral precipitates in the UCaP and UPS experiments (not shown).

Figure 2.

Brightfield reflected microscopy images (10X objective) of precipitate in the UP system after (a) 7.5 h and (b) 27.5 h.

[23] High-resolution brightfield microscopy images of crystalline precipitates in individual pores of the UP, UCaP, and UPS experiments are shown at similar times in Figure 3. A thick zone of agglomerated crystals is present along the horizontal center of each pore for UP and UCaP. Away from the thick precipitate zones, and throughout the pore for UPS, individual crystals are present. These appear smaller for UP and UCaP compared to UPS and more amorphous for UCaP compared to UP and UPS. At least 40 individual crystal sizes were determined in at least five pores along the horizontal center zone of each micromodel experiment. Individual crystal sizes in the UP, UCaP, and UPS systems are 54±37, 72±45, and 170±101 µm2, respectively; a plot of cumulative frequency distributions for individual crystal areas are in Figure A1. Average values increase from UP, to UCaP, to UPS, and the wide distribution of crystal sizes in each experiment results in significant overlap. The combination of crystal size results and images suggests, as hypothesized, that the presence of Ca2+ and SO42− affects crystal size and morphology.

Figure 3.

High-resolution (40X objective) brightfield microscopy images showing crystal sizes in single pores of the (a) UP, (b) UCaP, and (c) UPS systems. The pores are located near the inlet, and images were taken after approximately 12–15 h. Individual images are 200 µm in width.

3.2. Precipitation Rates

[24] Amounts of precipitate over time in the UP, UCaP, and UPS experiments are shown in Figure 4. Fluid residence time in the micromodel is approximately 1.1 min. The reactants diffuse across the center mixing line and reach steady-state conditions relatively quickly with concentration gradients in the horizontal plane of the experiment. Hence, the concentration profiles in the micromodel are relatively constant in time and only change slowly as precipitates block pores. As a result, reactions that occur are limited by the kinetics of chemical reaction. For much of the precipitation period, the area with precipitates increases linearly with time. However, at late time some deviation is evident for UP and UCaP, as the amount of precipitate starts to level off as steady state is approached. Assuming a constant rate of precipitation, linear regression yields precipitation rates of 28,300±7100 pixels h−1 for UP, 44,100±6300 pixels h−1 for UCaP, and 14,300±1400 pixels h−1 for UPS. Results presented here support our hypothesis that the presence of Ca2+ and SO42− affect uranium precipitation rates; they also demonstrate that Ca2+ enhances the kinetics of uranium precipitation, while SO42− inhibits it for the conditions tested.

Figure 4.

Area containing mineral precipitates versus time for the three replicate experiments. Slopes are shown and represent mineral precipitation rates. UP (rectangles), UCaP (triangles), and UPS (circles) are shown.

3.3. Mineral Type and Composition

[25] Representative backscattering Raman spectra for the UP, UCaP, and UPS systems during precipitation are shown in Figure 5. All the three spectra are very similar and match the published spectrum for chernikovite, shown in the inset of Figure 5 [Dorhout et al., 1989]. The exceptions are two broad peaks at 515 and 1350 cm−1 and an intense, sharp peak at 521 cm−1, which arises from the silicon component of the micromodel [Zhang et al., 2010]. The broad peaks are present in all experiments and are associated with chernikovite Raman spectra and may represent a less crystalline component of this mineral [Zhang et al., 2010]. The peaks at 192 and 830 cm−1 are assigned to the UO2 symmetric bending and stretching modes, respectively, whereas the peak at 1000 cm−1 is assigned to the PO4 symmetric stretch [Dorhout et al., 1989]. We note that the molecular formula for UHP differs from chernikovite by a single water molecule, and no Raman reference spectrum is available for UHP. Hence, while the Raman results do not rule out precipitation of UHP, they do suggest formation of chernikovite. The µ-XRD results for mineral precipitates formed in the UPS experiments and a reference chernikovite pattern (diffractogram) are shown in Figure 6. The patterns are very similar, indicating that only chernikovite is present. The experimental pattern contains some additional weak and shoulder peaks that are not present in the reference pattern; however, all these additional peaks are predicted in a diffraction pattern calculated from the crystal structure [Morosin, 1978]. The µ-XRD pattern for UCaP (not shown) also matches the chernikovite reference, indicating that this is the only mineral present in all the three experiments. We note that XRD reference patterns are available for a range of uranium phosphate compositions, and none of these, with the exception of chernikovite, match the experimental data satisfactorily. This includes phases which in place of the 4 waters of crystallization in chernikovite had 2 (ICDD pattern 13-0061), 1 (33-1428), and 1.7 (47-0176) waters of crystallization. Hence, it appears that only chernikovite is present in all the three experiments, in contrast to our hypothesis that the presence of Ca2+ and SO42− affects the type of mineral precipitate.

Figure 5.

Raman spectra for the (top) UPS, (middle) UCaP, and (bottom) UP systems. Inset: Raman reference spectra for chernikovite from Dorhout et al. [1989].

Figure 6.

(top) µ-XRD pattern for mineral form in the UPS system and (bottom) comparison to reference pattern for chernikovite.

[26] Since the presence of Ca2+ and/or SO42− affected the rate of mineral precipitation, crystal size, and crystal morphology, but only chernikovite was detected in all the three experiments, EDXS was used to determine if Ca or S substituted into the chernikovite crystal structure. A typical EDXS spectra for UCaP and three corresponding ESEM (Environmental Scanning Electron Microscope) images are shown in Figure 7. The EDXS spectrum shows that only U, P, O, Si, Na, and C are present, not Ca. The Si peak is from the glass coverslip that is fixed to the mineral precipitate. The Na peak is very small and is likely due to Na+ adsorption onto the chernikovite surface. The C is carbon coating. Peak heights for other EDXS spectra of UCaP vary slightly but always show the same elements. EDXS results for UPS are similar (results not shown); U, P, O, Si, Na, and C were identified, but no S was detected. Hence, Ca and S do not appear to substitute into the chernikovite crystal structure. The corresponding ESEM images for UCaP (Figure 7) illustrate the complex morphology of chernikovite; it is comprised of thin plates stacked together in a dense framework or of thin plates connected at irregular angles in an open framework. Dunham−Cheatham et al. [2011] observed thin plates of uranyl phosphate precipitates under conditions similar to UP. The effects of Ca2+ and SO42− on uranium precipitation rates will be discussed later.

Figure 7.

EDXS spectrum for UCaP, with three inset SEM images of crystalline precipitates.

3.4. Model Results

[27] The distribution of dominant soluble uranium species and pH without consideration of precipitation reactions are shown in Figure 8 based on the experimental measurements of pH. The dominant uranium species are UO22+, UO2OH, UO2HPO4, and (UO2)2(CO3)(OH)3. All but (UO2)2(CO3)(OH)3 are primarily present below the horizontal centerline of the pore structure where the uranium concentration is higher. For all the three systems, UO22+ has much higher concentrations than any other uranyl complexes below the horizontal centerline where the pH is lower (pH<6). At the low pH, anion (OH, CO32−, HPO42−) concentrations to form uranyl complexes are very low, resulting in low concentrations of uranyl complexes compared to UO22+. The (UO2)2(CO3)(OH)3 is primarily present above the centerline of the micromodel where the pH and therefore CO32− are higher. The maximum concentrations of each soluble uranium species are within a factor of two (and often identical) for UP and UPS. These concentrations are within a factor of four (and often within a factor of two) compared to UCaP. For UCaP (compared to UP and UPS), the maximum concentration of UO22− is higher, and the maximum concentrations of UO2OH and (UO2)2(CO3)(OH)3 are lower because of the lower measured pH value. If the calculated pH values are used (i.e., those based on equilibration with atmospheric CO2 at 31°C), then concentrations of UO22+, UO2OH, UO2HPO4, or (UO2)2(CO3)(OH)3 would be almost identical in each experiment because aqueous uranium species concentrations with sulfate and calcium are relatively low.

Figure 8.

Calculated distribution of dominant soluble U(VI) species including the uncomplexed uranyl ion, uranyl hydroxide, uranyl phosphate, uranyl carbonate complexes, and pH for (a) UP, (b) UCaP, and (c) UPS. For all cases, the experimental conditions (not the calculated conditions) in Table 2 were considered.

[28] The distribution of SR values that correspond to the experimental conditions simulated in Figure 8 is computed to compare experimental results with thermodynamic predictions as shown in Figure 9. Maximum SR values are located near the horizontal centerline of the pore structures, where the reactants mix. Maximum SR values for the experimental conditions simulated in Figure 9 are listed in Table 3. Also listed are maximum SR values when calculated pH values are used, and maximum SR values when the range of Ksp values and diffusion coefficients in the literature are considered. Use of either set of pH values results in only very small changes in maximum SR values, so those based on experimental pH values are discussed hereafter. Based on the range of available literature values for Ksp used in this work, maximum SR values for chernikovite change by an order of magnitude but change by less than 50% for other minerals (and often much less). Hence, different SR values will only be discussed for chernikovite. Use of different diffusion coefficients does not change the maximum SR values (Table 3), but the mixing zone width does change in proportion to the magnitude of the diffusion coefficient. Since our modeling focus is on the maximum SR values, and not the precipitation kinetics, only results with the reference diffusion coefficient value (i.e., 10−5 cm2 s−1) are discussed further.

Figure 9.

Distribution of SRs of six different U(VI)-phosphate minerals for (a) UP, (b) UCaP, and (c) UPS. For all cases experimental conditions in Table 2 were considered.

Table 3. Maximum SR
SRaSRbSR (−2σ/+2σ)cSR (−2σ/+2σ)cSRSR (−2σ/+2σ)cSR
  1. a

    SR is based on the Ksp value from Grenthe et al. [1992]. Maximum SR values in parentheses were obtained with lower (0.3 × 10−5 cm2 s−1) and upper (2.0 × 10−5 cm2 s−1) molecular diffusion coefficients, while a reference value (1 × 10−5 cm2 s−1) of diffusion coefficient was used for all other cases.

  2. b

    SR is based on the Ksp value from Johnson [2000].

  3. c

    SR values are based on the mean and standard deviation of Ksp values from Gorman-Lewis et al. [2009]. σ is the standard deviation of each SR.

  4. d

    Exp: measured pH value was used to calculate pCO2, and speciation equations in Table A1 were used to obtain initial distribution of species.

  5. e

    Calc: pH and total carbonate concentration are calculated based on speciation reported in Appendix A. pCO2 is assumed in equilibrium with atmosphere (pCO2=10−3.5).

  6. f

    For UCaP experimental system, the maximum SR for HAP was less than 1. E-5 with a Ksp value of −58.0 from Valsami-Jones et al. [1998].

UP_exp (meas)d0.70 (0.70)9.6114.54 (12.38/18.73)8.54E+5 (8.16E+5/9.36E+5)16.28 0.56
UP_exp (calc)e0.89 (0.89)12.2718.58 (15.81/23.94)4.30E+5 (4.11E+5/4.71E+5)16.76 0.35
UCaP_exp (meas)f0.97 (0.97)13.3920.26 (17.25/26.10)6.77E+5 (6.47E+5/7.42E+5)14.792.33E+6 (2.17E+6/2.50E+6)0.26
UCaP_exp (calc)0.89 (0.89)12.2718.57 (15.81/23.93)4.30E+5 (4.11E+5/4.71E+5)16.752.24E+6 (2.09E+6/2.40E+6)0.34
UCaP_high Ca2+ case0.9312.8219.50 (16.60/25.13)6.40E+5 (6.12E+5/7.01E+5)13.671.95E+8 (1.82E+6/2.09E+6)0.33
UCaP_high Tot(CO3) case0.00760.100.16 (0.14/0.21)1.52E+1 (1.45E+1/1.66E+1)0.582.24E+4 (2.09E+4/2.40E+4)0.10
UPS_exp (meas)0.64 (0.64)8.8013.32 (11.33/17.15)7.82E+5 (7.46E+5/8.57E+5)37.01 0.72
UPS_exp (calc)0.88 (0.88)12.1218.34 (15.61/23.63)4.25E+5 (4.05E+5/4.66E+5)33.27 0.35

[29] The maximum SR values of experimental simulations reported in Table 3 for chernikovite, UHP, and Na-autunite are all less than 40, whereas those for UOP are greater than 105. Maximum SR values for chernikovite, UHP, or UOP are not significantly different in all the three experiments. Maximum SR values for Na-autunite vary by a factor of approximately two between different experiments; this range of variation is small relative to differences in the maximum SR values for different minerals. When Ca2+ is present at 0.1 mM, the maximum SR value for Ca-autunite is greater than 106. Hence, UOP and Ca-autunite are much more thermodynamically favored to precipitate. The ranges of the maximum SR values for UHP and Na-autunite are 13–21 and 14–37, respectively. The ranges of the maximum SR values for metaschoepite and HAP are 0.5–0.7 and <10−5, respectively. Thermodynamic favorability of mineral precipitation depends on the solubility product (Ksp). For chernikovite, the maximum SR values are all less than 1 when the recommended literature value of Ksp is used, and between 8 and 14 when the lowest Ksp [Johnson, 2000] available is used. These results indicate that chernikovite is among the least thermodynamically favored mineral to precipitate in the micromodel experiments, even though it is the only one identified with Raman spectroscopy and µ-XRD spectroscopy.

[30] The maximum SR values for the two additional simulations performed with elevated Ca2+ (i.e., 10 mM) or total carbonate (i.e., 5 mM) are also listed in Table 3. For [Ca2+] = 10 mM, there is no change in the relative favorability of minerals that will precipitate, although maximum SR values for HAP and Ca-autunite increased from <10−5 to 0.004 and 2.24 × 106 to 1.95 × 108, respectively, due to the higher Ca2+ concentration. For Tot[CO3] = 5 mM, there is also no change in the relative favorability of minerals that will precipitate, although the maximum SR values for chernikovite and UHP dramatically decrease to 0.0076 and 0.16, compared to experimental values of 0.97 and 20.3, respectively. Also, the maximum SR value for Ca-autunite decreased by two orders of magnitude to 2.24 × 104. The lower SR values for chernikovite, UHP, and Ca-autunite are mainly due to the formation of uranyl carbonate complexes in solution (i.e., UO2CO3(aq), (UO2)2CO3(OH)3, Ca2UO2(CO3)3). Hence, for Ca2+ and Tot(CO3) concentrations more representative of actual field sites, chernikovite is still among the least thermodynamically favored mineral to precipitate, and HAP precipitates, which can sequester uranium [Wellman et al., 2008], also remain unfavorable. We note that for higher Ca2+ and PO43− concentrations (i.e., 1 mM each) and higher pH (i.e., 9.8) (results not shown), HAP precipitation will be favorable and can represent a sink for uranium removal by sorption.

[31] Precipitates of chernikovite were recently observed by XRD, SEM (or transmission electron microscope), and/or EXAFS (Extended X-ray adsorption fine structure) in batch studies under uranium and phosphate concentrations similar to those used in the UP system of this study [Singh et al., 2010; Dunham-Cheatham et al., 2011]. In those studies, thermodynamically more favorable solid forms other than chernikovite were not observed as in this study. These results suggest that the formation of UOP is not kinetically favorable during experimental periods in this study. At room temperature, meta-autunite minerals such as Ca-autunite and Na-autunite have been synthesized by ion exchange of Ca or Na with the hydrogen ion in chernikovite, and reaction times on the order of a few days or less are indicated [Sowder et al., 2001; Wellman et al., 2005]. However, very high concentrations of cations (approximately 2 M) were used to obtain the more crystalline autunite. As described earlier, the EDXS spectrum for precipitates in our work shows no Ca and only very small peaks of Na. This suggests that under low Na+ or Ca2+ concentrations, ion exchange of the proton in chernikovite did not occur or was not stable during precipitation (<30 h). Jerden and Sinha [2003] also stated that the formation of the autunite group minerals takes 2–3 weeks at room temperature. Singh et al. [2010] also observed metastability of chernikovite formed within a day over a 1 year experimental period in a batch system. Longer-term experiments are needed to assess if autunite is formed over longer time periods and in the presence of higher Na+ or Ca2+ concentrations.

3.5. Effects of Ca2+ and SO42− on Uranium Precipitation

[32] Several additional studies have investigated the impact of Ca2+ on uranium transformation at room temperature. Sowder et al. [1996] found that at 25°C, schoepite was slowly transformed to Ca-autunite for 94 days after schoepite was in a solution with 10 mM Ca2+ and PO43−. Later, Sowder [1998] reported that Ca-autunite was synthesized from chernikovite in a high concentration Ca2+ solution (>0.01 M) by ion exchange at 22°C. Another study investigated the impact of Ca2+ on U(VI) sorption and found that the presence of Ca2+ reduced the sorption of U(VI) onto quartz [Nair and Merkel, 2011]. While these and aforementioned experiments indicate that Ca2+ in solution can promote the transformation of uranium from one mineral form to another or inhibit uranium sorption from solution, they do not provide insights into the effects of Ca2+ on uranium precipitation rates. Instead, our modeling results can provide insights. When Ca2+ is present (i.e., UCaP), maximum SR values for both chernikovite and UHP are greater compared to the UP case. These results suggest that the higher precipitate rate for UCaP compared to UP may be related to greater values of SR for chernikovite and/or UHP.

[33] Although we could find no supporting studies on the effects of SO42− on uranium precipitation rates, there is extensive work regarding the effects of ions on calcium phosphate precipitation. In batch experiments under pH static conditions, Cao et al. [2007] observed that SO42− reduced the rate of hydroxyapatite precipitation under supersaturated conditions. Aqueous SO42− concentrations decreased at early time (within 10 min) during precipitation and then increased within an hour to the end of experiment (24 h). As a result, the final SO42− concentration in the batch experiment was similar to the starting concentration, but the HAP precipitation rate was lower compared to that in a control experiment without SO42−. The radius of SO42− is smaller than that of phosphate; the authors proposed that the replacement of SO42− into the HAP lattice at early stage decreased precipitation rates, but this replacement was not stable such that PO43− replaced SO42− in the lattice over time. We propose that a similar mechanism may be occurring in chernikovite for UPS.

3.6. Potential Pore Blockage Due to Precipitation

[34] Tracer test results for a clean micromodel and for the UP, UCaP, and UPS experiments after precipitation are shown in Figure 10. In the blank micromodel (i.e., no precipitate present), the tracer occupies two thirds of the pore structure, consistent with the 2:1 ratio of tracer to clean water flow rate. In micromodels with precipitate, the tracer is restricted to half of the micromodel for part of the pore structure, and then breakthrough across the precipitate zone occurs. This breakthrough occurs earliest in the pore structure for UPS, later in the pore structure for UCaP, and latest in the pore structure for UP. The maximum transverse widths of continuous precipitate as a function of longitudinal distance in the pore networks of the three experiments are shown in Figure 11, along with laser confocal images of precipitate in a pore halfway down the length of the micromodel in the UP and UPS experiments. The transverse width is relatively constant over the entire pore length for UP but decreases midway in the pore network for UCaP and earlier in the pore network for UPS. Laser confocal images indicate the UP pore is completely blocked, whereas the UPS pore is not, especially further downgradient from the inlet. These results indicate that tracer breakthrough across the micromodel centerline occurs when precipitates exist as isolated crystals that do not connect across a pore body, and that faster precipitation (e.g., faster precipitation in UP compared to UPS) results in more complete pore blockage along the length of the micromodel. A primary concern of pore blockage is that mixing will be restricted, and this will limit the extent of mineral precipitation that occurs in the subsurface. It is also possible that pore blockage will decrease because precipitates will dissolve due to undersaturation of the surrounding fluid [Tartakovsky et al., 2008; Katz et al., 2011; Zhang et al., 2010; Yoon et al., 2012].

Figure 10.

Tracer test results for (a) blank micromodel, (b) UP, (c) UCaP, and (d) UPS systems.

Figure 11.

(a) Maximum transverse width of the precipitate line versus distance in the micromodel. (b and c) Montages of laser confocal images of mineral precipitates in individual pores from the (b) UP and (c) UPS systems halfway down the length of the micromodel.

4. Conclusions

[35] We determined from pore-scale experiments and spectroscopic evaluation that uranium and phosphate, each at 100 µM, will mix transverse to flow and react to form only chernikovite in a time scale of hours to days, even in the presence of Ca2+ and SO42−. We also determined that the presence of Ca2+ and SO42− affects the rate of precipitation, but only within a factor of three, and the shape and size of crystals formed. However, EDXS results indicate that Ca and S are not incorporated into the chernikovite lattice. We determined from pore-scale modeling that chernikovite is the least favored mineral to precipitate and that UOP and Ca-autunite are the most favored. Consistent with literature results, it appears that chernikovite is kinetically favored during the time scale (hours to days) of our study. Last, we determined that rapid chernikovite precipitation blocks pores and diverts flow. In the field, rapid chernikovite precipitation and flow diversion are possible when phosphate is amended to groundwater. Permeability changes were observed at the Hanford area 300 site after phosphate addition [Vermeul et al., 2009], and this was attributed to hydroxyapatite precipitation. Our results suggest that pore blockage due to chernikovite precipitation may also be occurring and adversely affecting mixing and continued precipitation of U(VI). Further work is required to evaluate the extent of uranium that can be sequestered in the field with phosphate addition, the prevailing mineralogy at longer time scales, and the stability of precipitated uranium minerals when phosphate addition is discontinued.

Appendix A:  

[36] In Appendix A, the supplemental information includes a plot of cumulative frequency distributions for individual crystal areas in the UP, UCaP, and UPS systems (Figure A1), a list of speciation reactions and stability constants used in this work (Table A1), and minerals and their solubility products available in the literature (Table A2).

Figure A1.

Cumulative frequency distribution of individual crystal areas.

Table A1. Speciation Reactions and Stability Constants
Reactionlog Ka
  1. a

    From Guillaumont and Mompean [2003] unless otherwise noted.

  2. b

    Dong and Brooks [2006].

  3. c

    Martell and Smith [2001].

Uranium Species
2UO22++2H2O=(UO2)2 (OH)22++2H+−5.62
UO22++2PO43−+5H+=UO2(H2PO4) (H3PO4)+45.05
Other Species
Table A2. Minerals and Their Solubility Products at 298 K and I=0 M
ReactionNamelog Ksp
Reference ValueOther Values
  1. a

    Grenthe et al. [1992].

  2. b

    Johnson [2000].

  3. c

    Gorman-Lewis et al. [2009]. Plus or minus two standard deviations from experimental data were used to compute maximum SR values.

  4. d

    Gorman-Lewis et al. [2008b].

  5. e

    Zhu et al. [2009].

  6. f

    Valsami-Jones et al. [1998].

UO2HPO4 · 4H2O(s)=UO22++PO43−+H++4H2OChernikovite−24.20a−25.1b
UO2HPO4 · 3H2O(s)=UO22++PO43−+H++3H2OUHP−25.52c−0.11/+0.07c
(UO2)3(PO4)2 · 4H2O(s)=3UO22++2PO43−+4H2OUOP−49.36a, c−0.04/+0.02c
NaUO2PO4 · xH2O(s)=UO22++Na++PO43−+xH2ONa meta-autunite−23.64a 
Ca(UO2)2(PO4)2 · 3H2O(s)=2UO22++Ca2++2PO43−+3H2OCa meta-autunite−48.36c±0.03c
UO3 · 2H2O(s)+2H+=UO22++3H2OMetaschoepite5.6d 
Ca (PO4)3OH(s)=5Ca2++3PO43−+OHHAP−53.28e−58.0f


[37] We thank Bruce Arey at the EMSL PNNL for help with SEM and EDS. We also thank Thomas Johnson and Craig Lundstrom at the University of Illinois at Urbana-Champaign for helpful suggestions regarding SEM and EDXS analyses. This work was supported by the Environmental Remediation Science Program (ERSP), DOE, grant DE-SC0001280. Support for MFF was also provided by a University of Illinois Graduate Student Fellowship. A portion of these experiments were conducted in the William R. Wiley Environmental Molecular Sciences Laboratory, a U.S. DOE national scientific user facility sponsored by the DOE's Office of Biological and Environmental Research and located at PNNL. H.Y. is supported as part of the Center for Frontiers of Subsurface Energy Security, an Energy Frontier Research Center funded by the DOE, Office of Science, Office of Basic Energy Sciences under award DE-SC0001114. Sandia National Laboratories is a multiprogram laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the DOE's National Nuclear Security Administration under contract DE-AC04-94AL85000.