U(VI) Bioreduction with Emulsified Vegetable Oil as the Electron Donor

Feb 11, 2013 - ABSTRACT: We conducted microcosm tests and biogeochemical modeling to study U(VI) reduction in contaminated sediments amended with...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/est

U(VI) Bioreduction with Emulsified Vegetable Oil as the Electron Donor − Microcosm Tests and Model Development Guoping Tang,†,* Wei-Min Wu,‡,§ David B. Watson,† Jack C. Parker,∥ Christopher W. Schadt,⊥ Xiaoqing Shi,#,▽ and Scott C. Brooks† †

Environmental Sciences Division, Oak Ridge National Laboratory, P.O. Box 2008, MS-6038, Oak Ridge, Tennessee 37831-6038, United States ‡ Department of Civil and Environmental Engineering, Stanford University, Stanford, California 94305-4020, United States § Center for Sustainable Development & Global Competitiveness, Stanford University, Stanford, California 94305-4020, United States ∥ Department of Civil and Environmental Engineering, University of Tennessee, Knoxville, Tennessee 37996, United States ⊥ Biosciences Sciences Division, Oak Ridge National Laboratory, P.O. Box 2008, MS-6038, Oak Ridge, Tennessee 37831-6038, United States # Department of Scientific Computing, Florida State University, Tallahassee, Florida 32306, United States ▽ School of Earth Sciences and Engineering, Department of Hydrosciences, Nanjing University, Nanjing, 210093, China S Supporting Information *

ABSTRACT: We conducted microcosm tests and biogeochemical modeling to study U(VI) reduction in contaminated sediments amended with emulsified vegetable oil (EVO). Indigenous microorganisms in the sediments degraded EVO and stimulated Fe(III), U(VI), and sulfate reduction, and methanogenesis. Acetate concentration peaked in 100−120 days in the EVO microcosms versus 10−20 days in the oleate microcosms, suggesting that triglyceride hydrolysis was a rate-limiting step in EVO degradation and subsequent reactions. Acetate persisted 50 days longer in oleate- and EVOthan in ethanol-amended microcosms, indicating that acetate-utilizing methanogenesis was slower in the oleate and EVO than ethanol microcosms. We developed a comprehensive biogeochemical model to couple EVO hydrolysis, production, and oxidation of long-chain fatty acids (LCFA), glycerol, acetate, and hydrogen, reduction of Fe(III), U(VI) and sulfate, and methanogenesis with growth and decay of multiple functional microbial groups. By estimating EVO, LCFA, and glycerol degradation rate coefficients, and introducing a 100 day lag time for acetoclastic methanogenesis for oleate and EVO microcosms, the model approximately matched observed sulfate, U(VI), and acetate concentrations. Our results confirmed that EVO could stimulate U(VI) bioreduction in sediments and the slow EVO hydrolysis and acetate-utilizing methanogens growth could contribute to longer term bioreduction than simple substrates (e.g., ethanol, acetate, etc.) in the subsurface.

■

year.15,16 While U(VI) bioreduction with rapidly consumed electron donors has been under extensive investigation for more than 10 years,1−8 there have been few studies on the use of slowly degraded complex substrates. Based on the relative abundance of representative operational taxonomic units and known physiologies of closely allied species or genera, Gihring et al.15 developed a conceptual model for EVO degradation and subsequent reactions during a field injection test (SI Figure S1). The first step in the degradation of vegetable oil involves triglyceride hydrolysis to glycerol and long chain fatty acids (LCFA, palmitic, oleic, and

INTRODUCTION

Numerous electron donors such as hydrogen, acetate, lactate, ethanol, methanol, and glucose have been tested to stimulate indigenous microbial communities for U(VI) reduction and immobilization in contaminated aquifers.1−8 Use of these rapidly consumed electron donors requires daily to weekly injection to maintain reducing conditions and prevent biogenic U(IV) from reoxidizing to more mobile U(VI) species.8−13 Slow release electron donors have been considered to maintain long-term reducing conditions in the subsurface with less frequent injection. For example, perchlorate was degraded for over two years in downgradient wells after a single edible oil emulsion injection in a field test;14 a one-time 2 h emulsified vegetable oil (EVO) injection at the DOE Oak Ridge Integrated Field Research Challenge (ORIFRC) site resulted in anaerobic conditions in a fast flowing aquifer for over a © 2013 American Chemical Society

Received: Revised: Accepted: Published: 3209

November 13, 2012 February 5, 2013 February 11, 2013 February 11, 2013 dx.doi.org/10.1021/es304641b | Environ. Sci. Technol. 2013, 47, 3209−3217

Environmental Science & Technology

Article

Modeling. Approximating EVO and LCFA as triglyceride and oleic acid (with molecular weight of 882.38 and 282.46 g/ mol), respectively, the hydrolysis reaction was

linolenic acid), which are subsequently degraded through fermentation and beta oxidation to acetate and hydrogen.17,18 While syntrophic acetogens are generally considered as LCFA degraders in anaerobic digesters,18 results from this EVO injection test indicated that Pelosinus likely catalyzed triglyceride hydrolysis and Desulforegula degraded LCFA in the presence of sulfate.15 The produced acetate and hydrogen stimulated the indigenous microbial community and resulted in bioreduction of Mn(IV), Fe(III), U(VI), nitrate and sulfate, and methanogenesis.16 Several biogeochemical models have been developed to simulate terminal electron accepting processes (TEAPs) coupled to the oxidation of simple electron donors at laboratory and field scales. 19−27 Modeling of in situ bioremediation with complex substrates requires consideration of multiple functional microbial groups and a comprehensive biogeochemical model to describe complex microbial and geochemical dynamics. In this work, we report results of microcosm tests using EVO to stimulate U(VI) bioreduction in contaminated sediments, and development of a comprehensive biogeochemical model to describe the results. A major objective of this work was to construct reaction networks for numerical implementation of the conceptual model,15 and assemble and test representative parameter values for its application to the field test, which was described in a companion article.28

triglyceride + 3H 2O → 3HOleate + glycerol

with a rate of

dm m = −khx dt km + m

(1)

with m as EVO concentration [M], kh as the rate coefficient [s−1], km = 1 × 10−6 M, and x as hydrolytic bacteria (Pelosinus 15 ) represented by C5H7O2N (M).30 The LCFA precipitation reaction was Ca 2 + + 2Oleate− = Ca(Oleate)2

with a solubility of 1.34 × 10−5 M.31 The anaerobic glycerol fermentation reaction was 20glycerol + 5HCO3− + NH4 + → C5H 7O2 N + 26H+ + 10H 2 + 13H 2O + 30acetate−

which was derived from the electron donor half reaction (glycerol/acetate), electron acceptor half reaction (H+/H2), and cell synthesis reaction (Supporting Information (SI) Table S1) based on energetics.30 The reaction was treated as irreversible and its extent was determined by thermodynamics. 20 Specifically, a growth index (GI) was calculated by

■

MATERIALS AND METHODS Microcosm Tests. An overview of experimental procedures for the microcosm tests is presented here. For further detail, the reader is referred to Zhang et al.29 Serum bottles (158 mL) were filled with a groundwater-sediment suspension (∼8.5 g solids: 130 mL) under a nitrogen atmosphere. The pH 6.7 groundwater contained 1.37 mM NO3−, 0.01 mM O2, 1.1 mM SO42‑, 2.6 mM Ca, and 4.9 μM U(VI). The sediment contained ∼0.649 mg/g total U, and ∼49 mg/g total Fe (both extracted by 10 N HCl in anaerobic glovebag overnight for three times). Groundwater and sediments were from Area 2 at the ORIFRC site. Bottles were sealed with butyl rubber stoppers and aluminum caps. EVO (Terra Systems, Inc., Wilmington, DE) contained 60% soybean oil, 0.3% yeast extract, 6% food grade surfactant, and 0.05% (NH4)3PO4 by weight. The EVO did not contain lactate, normally added to the company’s stock solution, in order to eliminate a rapidly utilized electron source. The chemical oxygen demand (COD) was 1618 g/L and the specific gravity was 0.93. EVO was added to the microcosms to achieve 720 mg COD/L in the aqueous phase. Na2SO4 was added at concentrations of 0, 3.85, or 7.7 mM to test the effect of sulfate. After all amendments had been added, the bottles were shaken to mix, allowed to settle for 12 h, and initial samples of the clear supernatant were withdrawn using a needle and syringe. The microcosms were incubated at ambient temperature (22−24 °C) without further mixing. Liquid samples were withdrawn periodically for the analysis of pH, COD, acetate, sulfate, and U in the aqueous phase. At the end of the tests, headspace gas was analyzed for CH4, CO2, and H2, the aqueous phase was analyzed for sulfide, Fe(II) and alkalinity, and sediment samples were analyzed for U and Fe content. The U(IV)/total U in sediment was determined by X-ray absorption near-edge spectroscopy (XANES).4 No U(IV) was detected in the sediments before the tests.

GI = 20log[glycerol] + 5log[HCO3−] + log[NH4 +] − 26log[H+] − 10log[H 2] − 30log[acetate−] − log K

with [] for activity, and log K as the equilibrium constant. The activity of biomass was assumed as 1. Biomass was produced when the reaction was thermodynamically favorable (GI ≥ 0). NH4+ concentration was set at 1 μM20 for the GI calculation. Since the stoichiometric coefficient for NH4+ was much smaller than that of the other species, the calculated GI values was much less sensitive to NH4+ concentration than to other species. Similarly, the reactions for LCFA degradation by sulfate reducing bacteria (SRB) and syntrophic acetogens were derived as 5oleate− + 2H 2O + 5HCO3− + NH4 + + 16.25SO4 2 − → C5H 7O2 N + 19.75H+ + 16.25HS− + 45acetate−

and 4oleate− + 51H 2O + 5HCO3− + NH4 + → C5H 7O2 N + 28H+ + 50H 2 + 36acetate−

Reactions for nitrate, Fe(III), U(VI), and sulfate reduction and CH4 production with acetate and hydrogen as electron donors were obtained from refs 20 and 26 as summarized in SI Table S2. The growth rate for microorganisms was mD mA kI dx = k maxx dt kD + mD kA + mA kI + mI

(2)

with x as the biomass (M), kmax as the maximum growth rate (s−1), kD, kA, and kI as the donor, acceptor, and inhibitor concentrations (M) associated with half of the maximum rate, and mD, mA, and mI as the concentrations of donor, acceptor 3210

dx.doi.org/10.1021/es304641b | Environ. Sci. Technol. 2013, 47, 3209−3217

Environmental Science & Technology

Article

Figure 1. Observed (data points, with ethanol and oleate microcosm test data from 29) and model calculated (curves) acetate (row 1), sulfate (row 2), and aqueous U(VI) (row 3) concentration with ethanol (column 1), oleate (column 2), and EVO (column 3) as the electron donors in the microcosm tests. Blue, red, and green are for tests with 0, 3.85, and 7.7 mM sulfate amendment. Dash curves = predictions with parameter values in SI Table S3 without lag time for acetate-utilizing methanogens. Continuous curves = prediction with lag time for acetate-utilizing methanogens of 10 days for ethanol microcosms, and 100 days for oleate and EVO microcosms. Data points and error bars represent the mean and range of duplicates.

with kFeCO3 = 1.67 × 10−12 Ms1− and SR as the saturation ratio.40 Parameterization. Reliable estimation of the many parameters in these processes was not likely to be feasible by fitting the experimental data using the nonlinear least-squares methods.41 Therefore, we used parameter values reported in ref 26, tested and extended these parameter values using previously published experimental data with ethanol and oleate29 and new data from this study with EVO. We conducted sensitivity analysis for important parameters such as LCFA-utilizing sulfate reducer growth, EVO hydrolysis, U(VI) reduction rate coefficient, initial biomass, etc., and compared alternative parametrizations with, for example, different functional groups assemblages, EVO hydrolysis formulas, with or without time lag for acetate-utilizing methanogens, Fe (III) hydroxide bioavailable limitation, Ca−U−CO3 aqueous species reactions, etc. The lab data were used primarily for testing rather than calibration for model application to the field test. As Jin and Roden parametrized a microbial physiology-based model for an ethanol microcosm test that used sediments from the same site,26 their parameter values were first tested using data from our previous ethanol microcosm tests.29 Because acetate-utilizing sulfate reducers, which were considered to be “inactive” in previous modeling efforts,26 were expected to grow in these sulfate-amended microcosms, this microbial group was included with a small initial biomass (10−7 M, or 0.0113 mg/L, a low end of ref 26, SI Table S3). Similarly, a H2-utilizing metal reducer group and a H2-utilizing methanogen group were added. A rate coefficient of 10−6 s−1 was assigned for U(VI) reduction using ethanol, acetate or H2 as the electron donor. Assuming that biogeochemical parameters for nitrate, Fe(III), U(VI), and sulfate reduction and CH4 production with acetate and hydrogen were identical in the ethanol, oleate, and EVO microcosms, we added LCFA degradation parameters

and inhibitor (M), respectively. For microbial growth with Fe (III) oxide as the electron acceptor, the growth rate was x /msurf, avail mD kI dx = k maxmsurf, avail dt kA + x /msurf, avail kD + mD kI + mI (3)

with msurf,avail as the microbial-available surface sites taken as the Fe(III) oxide surface sites associated with H+, i.e., msurf,avail = mHfo_wOH + mHfo_sOH.26 Biomass decay was approximated by dx /dt = kdecayx

(4)

−1 26

with kdecay =10−7 s . Barnett et al.32 applied the surface complexation model (SCM) developed for U(VI) sorption to synthetic ferrihydrite33 to sediment from Oak Ridge, TN. We used an updated version of the SCM34 to model sorption of U(VI) and other species (e.g., Fe2+, Ca2+, SO42‑, CO32−, etc.). Sorption of acetate and simple organic species were not included because no significant retardation was observed in acetate injection tests (e.g., refs 19,21,27). As site characterization suggested that goethite content was greater than ferrihydrite,35 we simulated HCl-extractable Fe(III) oxides as goethite following ref 26. Redox couples involved in microbially mediated redox reactions were uncoupled to account for redox disequilibrium.20 Phases other than uraninite have been reported to control the solubility of U(IV).36,37 However, as thermodynamic data were not available for these other phases, uraninite was used to describe the removal of U(IV) from the aqueous phase. Precipitation reactions of FeS and uraninite have been reported to be fast38,39 and were treated as equilibrium reactions. FeCO3 (siderite) precipitation was considered to be kinetic with a rate of d m/d t = kFeCO3(SR − 1)

(4) 3211

dx.doi.org/10.1021/es304641b | Environ. Sci. Technol. 2013, 47, 3209−3217

Environmental Science & Technology

Article

extract, which contributed to less than 1% of the total COD for the EVO. Accelerated removal of sulfate should be associated with EVO degradation. Acetate concentration started to increase in 30−40 days, peaked at ∼3.2 mM ∼100 days (Figure 1c), and was consumed in 180 days. The decrease of aqueous U concentration occurred concurrently or after the decrease of sulfate concentration (Figure 1f, i), indicating possible biotic reduction by SRB,44 and/or abiotic reduction by Fe(II) and/or sulfide.22,29 The aqueous U concentration decreased to ∼1 μM, which was above 0.126 μM, the EPA maximum contaminant level (MCL) for drinking water. The challenge of decreasing U concentration to very low levels in Ca and carbonate rich groundwater using bioremediation and the implications were widely known.45,46 However, bioreduction is still useful particularly in cases where remediation goals or exposure hazards are based on contaminant flux rather than concentration. The pH decreased slightly by the end of the experiments, while bicarbonate concentration nearly doubled and tripled in the EVO microcosms with or without sulfate amendment, respectively (SI Table S5). HCl extraction by the end of the tests yielded 9.8, 18.7, and 21.3 mg/g Fe(II) in the control, and EVO microcosms without and with sulfate amendment. The Fe(II) in the control was thought to be in the sediments at the beginning of the test due to previous bioremediation.47 About 10% the total U was in the aqueous phase and no U(IV) was detected at the beginning of the tests. XANES analysis4 of sediments from the microcosms with EVO amendment at the end of the tests confirmed that 56−59% of the total U(VI) was reduced to U(IV). In addition to hydrogen, up to 11.6 and 6.5 mM methane was produced in the microcosms without and with sulfate amendment, respectively, indicating methanogen growth. Acetate and sulfate concentration changes suggested that acetate was degraded continuously after sulfate was consumed (Figure 1c, f) and the methanogens might be mainly acetateutilizing methanogens. An electron balance analysis29 suggested that about 15% and 0.3% of the electrons were consumed for Fe(III) and U(VI) reduction, respectively; while sulfate reduction and methanogenesis account for 12 and 73% of the electrons in microcosms without sulfate amendment, and 46 and 36% of the electrons in the microcosms with sulfate amendment, respectively (SI Table S5). Comparisons with Previous Ethanol and Oleate Microcosms. Sulfate was completely removed about 10−20 days later in EVO microcosms than in the ethanol and oleate microcosms29 (Figure 1d vs e,f). Acetate concentration peaked in 10−30 days in ethanol and oleate microcosms, 70−90 days earlier than in the EVO microcosms (Figure 1a, b vs c). The faster sulfate and U removal and acetate accumulation in the ethanol and oleate microcosms than in the EVO microcosms indicated that EVO hydrolysis was a rate-limiting process for EVO degradation and subsequent reactions. This was inconsistent with anaerobic reactors,18 indicating that LCFA oxidation catalyzed by sulfate reducers was faster than by acetogens. Assuming acetate consumption rate parameters for metal and sulfate reducers to be identical for the ethanol, oleate, and EVO microcosms, the delayed slow acetate concentrations decrease in the oleate and EVO microcosms was likely due to slow growth of acetate-utilizing methanogens. Sulfide and oleate (or other LCFAs) have been reported to inhibit growth of acetateutilizing methanogens.48−51 However, sulfide and LCFA

to describe oleate microcosm results. An oleate-utilizing SRB group15 and an acetogen group (known LCFA degraders 18) were added with initial biomasses of 10−6 M, and growth rate coefficients of 6 × 10−6 and 10−6 s−1, respectively. The initial biomass and growth rate coefficient were set to be between estimates for multiple functional groups in ref 26 (SI Table S3). The fast growth rate for oleate-utilizing SRB was based on the fast acetate concentration rise in the oleate microcosms, which was close to that in the ethanol microcosms (Figure 1a vs 1b). Finally, a fermentative organism (Pelosinus 15) was added with an initial biomass of 10−6 M and a growth rate coefficient of 2 × 10−6 s−1 for the EVO microcosms. All parameter values are summarized in SI Table S3. We implemented the model using PHREEQC.42 Equilibrium U reactions are listed in SI Table S4. The reaction database and input files are available through the authors upon request. In addition to including oleate and EVO, our model differs from ref 26 in that our reactions combined cell synthesis with electron donor and acceptor reactions, that is, we used anabolic rather than catabolic reactions. Besides EVO related reactions, the proposed model advances from ref 29 in that (1) microbial mass was explicitly incorporated in the growth rate calculation, (2) a generic comprehensive SCM model with Ca−U−CO3 complexes was used to account for competition among U(VI), bicarbonate, sulfate, Ca, etc., (3) reactions for metal reduction and methanogenesis were included, and (4) we employed a consistent set of parameter values to describe all of the individual tests. Note that we sought a single reaction network and parameter set to apply across all microcosm conditions, and that could subsequently be applied to the field test,15,16,28 guided by our experiments and review of the literature rather than fitting the model to the data using nonlinear least-squares methods.

■

RESULTS AND DISCUSSION EVO Microcosm Experiments. Control Experiment. In the control microcosm experiments without amendments, no acetate was detected as expected (Figure 1a). Aqueous U concentration gradually increased from 12.1 to 15.2 μM and then stabilized (Figure 1g). Sulfate concentration increased slightly and then became stable (Figure 1d). The increase in both U and sulfate was likely due to desorption from the sediments.43 Nitrate was consumed in all microcosms prior to carbon source addition (data not shown) due to microbial denitrification involving naturally occurring organic material or abiotic reduction via Fe(II) present in the sediments. These observations indicate that the organic matter or Fe(II) in the sediment supported denitrification but not sulfate or U(VI) reduction over the 182 day experiment. Microbial Growth. Microbial metabolism in the microcosms was visually apparent in EVO-amended microcosms by day 10 as small (0.1−0.2 mm) black spots appeared in the sediments. A black layer also occurred on the surface of the sediment. The change in sediment color was likely due to formation of biogenic ferrous sulfide. The development of these spots was accompanied by increased acetate concentration, and decreased sulfate and aqueous U concentrations. After 56 days, the black coloration had spread through the sediment. Terminal Electron Accepting Processes. Sulfate concentration decreased shortly after EVO addition and was consumed in about 40 and 50 days for microcosms with and without sulfate amendment, respectively (Figure 1f). Sulfate reduction might have been supported by easily degraded yeast 3212

dx.doi.org/10.1021/es304641b | Environ. Sci. Technol. 2013, 47, 3209−3217

Environmental Science & Technology

Article

The surface complexation model predicted about 44% of U(VI) in the sorbed phase in the microcosms at the beginning of the tests (SI S1, Figure S2). This was less than our observations (∼90%). Furthermore, the model predicted decreasing aqueous U(VI) concentration with sulfate amendment, which was inconsistent with our observations (SI Figure S2c). However, the predicted distribution coefficient was similar to that in ref 54 in which the formation of Ca−U− CO3 complexes was considered to reduce U(VI) sorption at around neutral pH. Removing the Ca−U−CO3 complexes from the reaction database, the UO22+ activity (not shown) and U(VI) sorption (SI Figure S2g vs h) were predicted to increase by 2−2.5 orders of magnitude, and the predicted aqueous U(VI) concentration matched the observed increase with sulfate amendment well (SI Figure S2d). This result implies that substantial uncertainty existed in the predicted aqueous U concentration and suggests the need to reevaluate the SCM34 taking consideration of the Ca−U−CO3 complexes. With Ca−U−CO3 complex reactions in the database, the model generally predicted nearly complete reduction of both aqueous and sorbed U(VI) as the predicted sorption was small (SI Figure S3). Without Ca−U−CO3 complexes, the model predicted much more sorbed U(VI), and about 60% the sorbed U(VI) was not reduced in the case with no sulfate amendment due to the slow U(VI) reduction rate (SI Figure S4f). As we assumed that sulfate reducers contribute to U(VI) reduction, more U(VI) reduction was predicted in the microcosms with sulfate amendment due to more biomass accumulation for sulfate reducers (SI Figure S4f vs g). Fe(III) Reduction. The model predicted 4−25% of Fe(III) to be reduced, with the high value in the microcosms with sulfate amendment (SI Figure S5). This prediction was lower than observed (20−40%). likely because we used goethite (FeOOH) to approximate the Fe(III) hydroxides as described in the methods. As the Gibbs free energy of formation varied from −452 kJ/mol for amorphous FeOOH (FeOOHa) to −489 kJ/ mol for goethite,55 and the equilibrium constant log K differred from −1.0 to 4.891, the activity of ferric ion in equilibrium with FeOOHa is about 6 orders of magnitude greater that with goethite. As a result, the growth index was raised by approximately 900 (6 times 150 with 150 as the stoichiometric coefficient for Fe3+ in Fe reduction reduction reaction with acetate as the electron donor, FeRA, Reaction 7 in SI Table S2) to be greater than 0 in the experiment duration when we changed FeOOH to FeOOHa (Figure 2a). If amorphous Fe(III) hydroxide were used, Fe(III) reduction would be thermodynamically much more favorable, and the model would predict nearly complete reduction of Fe(III) (Figure 2b, SI S6f). This results demonstrated the challenge in predicting the extent of Fe(III) reduction based on thermodynamic control due to the uncertainty in the Fe(III) mineral composition and thermodynamic parameters. LCFA Oxidation. An initial sulfate concentration of 1.1 mM in solution at most supported sulfate reducers to consume 0.34 mM of the 0.92 mM oleate and yield 3.1 mM acetate (SRO, reaction 15 in SI Table S2). Sulfate-amended microcosms (3.85 and 7.7 mM) had sufficient sulfate as an electron acceptor for complete oleate oxidation to produce 8.3 mM acetate. The observed higher acetate peak concentrations in the oleate microcosms with more sulfate additions provided evidence that sulfate reducers were involved in oleate (LCFA) oxidation. After acetate consumption by metal and sulfate reducers, acetate concentration was predicted to plateau until it was

concentrations were expected to be low due to precipitation with Fe(II) and Ca, respectively. Another reason for low methanogenic activity could be that in the oleate and EVO microcosms, H2 partial pressure was too low to initiate fast growth of Methanosarcina, genes of which were found in the sediments.15 Compared with H2 partial pressure (up to 200 Pa) with ethanol as electron donor source, the low H2 partial pressure with LCFA (