Environ. Sci. Technol. 2006, 40, 6432-6437
Effects of Carbonate Precipitates on Long-Term Performance of Granular Iron for Reductive Dechlorination of TCE SUNG-WOOK JEEN, ROBERT W. GILLHAM,* AND DAVID W. BLOWES Department of Earth Sciences, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1
Long-term column experiments were conducted to evaluate the effects of secondary carbonate minerals on permeability and reactivity of commercial granular iron treating trichloroethene (TCE). The results showed that carbonate precipitates caused a decrease in reactivity of the iron, and spatially and temporally varying reactivity loss resulted in migration of mineral precipitation fronts, as well as profiles of TCE, pH, alkalinity, calcium, and dissolved iron. In the columns receiving solutions of dissolved calcium carbonate, porosity gradually decreased in proportion to the source concentrations, as carbonate minerals accumulated. However, the rate of porosity loss slowed over time because of the declining reactivity of the iron. Thus, secondary minerals are not likely to accumulate to the extent that there is a substantial reduction in hydraulic conductivity. The reactivity of the iron was found to decrease as an exponential function of the carbonate mineral volume fraction. This changing reactivity of iron should be incorporated into predictive models for improved designs of iron permeable reactive barriers (PRBs).
Introduction Granular iron permeable reactive barrier (PRB) technology has gained considerable acceptance as an effective method for remediation of contaminated groundwater. Although the empirical evidence supports high expectations for long-term operation (1, 2), the ability to predict performance over a range of hydrogeochemical conditions remains weak. In particular, precipitation of secondary minerals may affect the system hydraulics (porosity and hydraulic conductivity) and reaction rates, leading to incomplete treatment of contaminants. Although the role of precipitates may be beneficial in some conditions (3, 4), many laboratory (5, 6) and field (7, 8) tests have shown that corrosion products and precipitates can have a negative effect on the performance of iron PRBs. However, the pattern of mineral precipitation and the degree to which it affects performance over time remains uncertain. Field observations are not available for a sufficient time period and are not sufficiently precise to clearly delineate the pattern of precipitation. Although a recent laboratory column study showed that mineral precipitation occurs as a moving front, rather than accumulating only near influent ends of the * Corresponding author phone: (519)888-4658; fax: (519)746-1829; e-mail:
[email protected]. 6432
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 20, 2006
columns (9), a more systematic evaluation of geochemical changes in iron PRBs and quantification of changing reactivity could make important contributions to the development of more efficient designs of iron PRBs. The purpose of this study was to develop a quantitative relationship between the degree of mineral precipitation and permeability and reactivity of iron and to examine the reaction processes that govern the performance of iron. Longterm column experiments were conducted, using simulated groundwater containing differing amounts of dissolved calcium carbonate. Carbonate minerals were the main focus of this study because bicarbonate/carbonate are the most common and abundant inorganics in groundwater and carbonate minerals are known to affect greatly the performance of iron PRBs (5, 6, 9).
Experimental Section Columns. Four columns, 2.54-cm I.D. and 50-cm long, were assembled to simulate groundwater flow through iron PRBs (see Figure S1, Supporting Information). Eleven sampling ports were located at 5-cm intervals along each column with two additional ports at 2.5 and 7.5 cm from the influent end. Ports for retrieving samples of the iron material for mineral examination were located on the opposite side of the columns. A fine and a coarse granular iron were used, both obtained from Connelly-GPM, Inc. (Chicago, IL). Characteristics of the iron materials are provided in Table S1 of the Supporting Information. Feed Solutions. Compositions of the feed solutions for each column are provided in Table S1. Column A was a control, receiving 10 mg/L (7.6 × 10-5 M) trichloroethene (TCE) in deionized water. Columns B and C received identical solutions (100 mg/L (1 × 10-3 M) CaCO3 + 10 mg/L (7.6 × 10-5 M) TCE) but differed in that column C contained the coarse iron material. Column D received 500 mg/L (5 × 10-3 M) CaCO3 + 10 mg/L (7.6 × 10-5 M) TCE. To prepare the dissolved calcium carbonate solutions, a precalculated amount of analytical grade CaCO3(s) was added to deionized water and was purged with CO2 gas until all solid was dissolved, followed by purging with oxygen-free N2 gas to adjust the pH to 6.4 ( 0.2. The feed solution for column A was deoxygenated by purging with oxygen-free N2 gas. A concentrated stock solution of TCE in methanol was then added to the feed solutions to achieve nominal concentrations of 10 mg/L. Operation. The feed solution was pumped at a flow rate of 0.4 mL/min using an Ismatec multichannel peristaltic pump (Model 78001-12). The flow velocity varied as a consequence of differences in porosity, but the initial values were close to 2.3 m/day. Manometers, connected near the inlet and outlet of the columns (Figure S1), were used to monitor the hydraulic head difference across each column. Average hydraulic conductivity values were calculated from the hydraulic gradient and the solution flux, using the Darcy equation. The rate of gas generation, measured using a sealed glass tube (Figure S1), was used to calculate the rate of iron corrosion (10). Solution samples were collected periodically for analyses of various organic and inorganic parameters. Columns were operated for about 2 years (about 3000 pore volume (PV)), with the actual operation period for each column listed in Table S1. The duration of the tests, combined with the accelerated flow rate, would be equivalent to approximately 40 years of operation at an assumed groundwater velocity of 10 cm/day. The tests were conducted simultaneously in a laboratory at ambient temperature (24 ( 2 °C). 10.1021/es0608747 CCC: $33.50
2006 American Chemical Society Published on Web 09/07/2006
FIGURE 1. TCE profiles over time and fits of the mixed-order (solid lines) kinetic model for (a) column A (D.I. H2O + 10 mg/L TCE), (b) column B (100 mg/L CaCO3 + 10 mg/L TCE), (c) column C (100 mg/L CaCO3 + 10 mg/L TCE, coarse iron), and (d) column D (500 mg/L CaCO3 + 10 mg/L TCE). The earliest TCE concentrations are also fitted with the first-order (dashed lines) kinetic model. All data points are plotted with respect to the initial residence time, and concentrations are normalized to the influent concentration. Equal symbols do not always correspond to profiles at the same time. Analytical Methods. TCE was analyzed using a liquidliquid pentane extraction method (11) and a Hewlett-Packard 5890 Series II gas chromatograph (GC) (DB-624 capillary column; J&W Scientific; 30 m × 0.538 mm I.D.; film thickness 3 µm) with a 63Ni electron capture detector (ECD). Samples were prepared for analysis by adding 2 mL of sample to 2 mL of pentane containing an internal standard (500 µg/L of 1,2dibromoethane) in a 5-mL glass screw cap vial with Teflonfaced septum. The method detection limit (MDL) was 1.0 µg/L. Concentrations of the dichloroethene (DCE) isomers and vinyl chloride (VC) were determined by a headspace method using a Hewlett-Packard 5890 Series II gas chromatograph (Hnu NSW-PLOT capillary column; 15 m × 0.53 mm I.D.) with an Hnu photoionization detector (PID). The MDLs were 2.9, 2.4, 1.3, and 1.3 µg/L for 1,1-DCE, transDCE, cis-DCE, and VC, respectively. For hydrocarbon gases, a headspace was created in the aqueous sample with a ratio of 2.5-mL headspace to 2.5-mL aqueous sample. After equilibration, a 250-µL sample taken from the headspace was injected onto a Hewlett-Packard 5790 gas chromatograph (GS-Q plot capillary column) with a flame ionization detector (FID). The MDLs were 0.5, 0.4, 0.6, 6.8, 3.2, 2.3, and 2.0 µg/L for methane and ethene, ethane, propene and propane, isobutane, n-butane, 1-butene, and acetylene, respectively. Chloride measurements were made using an ion chromatograph (Dionex ICS-2000) with an ion-eluent generator and a conductivity detector (MDL of 0.5 mg/L). Oxidation-reduction potential (ORP) was determined using a combination Ag/AgCl reference electrode with a
platinum redox sensor and a Markson pH/mV meter (Model 90). The potential reading was recorded immediately once the electrode equilibrated with the solution and then was corrected with respect to standard hydrogen reference electrode (SHE), giving an Eh value. Following the Eh measurement, the pH of the same sample was measured using a combination glass-Ag/AgCl reference electrode (Orion, Model 91-06). Total dissolved iron and calcium were analyzed using an inductively coupled plasma-optical emission spectrometer (ICP-OES) (Thermo Jarrell Ash Corp.). The MDLs were 3 µg/L for iron and 0.3 µg/L for calcium. At the end of the experiments, the carbonate precipitates that had accumulated within the columns were quantified using a modified Bundy and Bremner (12) method. Samples of solid materials (0.5 g) were collected from the solid sampling ports and were treated with 2 M HCl to dissolve resident carbonate minerals, generating CO2 gas. The CO2 gas evolved from the sample was trapped in a solution containing 2 M KOH. Alkalinity was determined using a HACH alkalinity kit (Model AL-DT). Full details of the procedures and analytical methods are provided in Jeen (10).
Results and Discussion TCE Profiles. Selected profiles of TCE concentration versus residence time (Figure 1) were the primary data used for evaluating the reactivity of the iron. In preparing Figure 1, the flow velocity was used to convert distance of the sampling VOL. 40, NO. 20, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6433
points from the influent end to residence time. The full data set is available in Jeen (10). In all columns, the TCE concentration profiles gradually migrated over time, indicating progressive passivation of the iron material. Passivation was also indicated by a decrease in the iron corrosion rate, calculated from the measured rate of gas generation (data not shown). However, the rates of migration of the profiles were slow relative to the pore water velocity and thus each profile reflects the degradation characteristics at the time the measurements were made. Previous studies have reported the kinetic constants for TCE degradation using the pseudo-first-order kinetic model described by Johnson et al. (13), for example:
d[TCE] ) -kSA,1‚Fa[TCE] dt
(1)
where kSA,1 is the specific reaction rate constant (L H2O m-2 iron h-1), Fa is the surface area concentration of iron (m2 iron L-1 H2O), and [TCE] is the concentration of TCE (mol L-1 H2O). This model provided a good representation of the data at very early time (dashed lines in Figure 1), within the first 100 pore volumes. The calculated initial rate constants (Table S2, Supporting Information) were within the range reported in the literature (13). As time progressed, the shapes of the profiles appeared to change from first-order to zero-order, and thus while the first-order model gave reasonable fits to the data only at early time, a mixed-order kinetic model (13-15) appeared to capture the major trends in the TCE profile over all times (solid lines in Figure 1 and kinetic parameters of Table S2). The form of the mixed-order model is
d[TCE] [TCE] ) -kSA,0‚S dt K1/2 + [TCE]
(2)
where kSA,0 (mol m-2 iron h-1) is the zero-order rate constant k0 (mol L-1 H2O h-1) normalized to the surface area concentration of iron S (m2 iron L-1 H2O), and K1/2 is the half-saturation constant (mol L-1 H2O). Equation 2 was integrated with respect to time and was fitted to the experimental data using a nonlinear least-squares optimization procedure (16) to simultaneously estimate the two kinetic parameters, k0 ()kSA,0‚S) and K1/2. Then, kSA,0 was then determined by dividing k0 by S. For the initial values of the surface area concentration of the iron, the physical surface area, measured by BET, was used for each column. While the coarse iron had a lower specific surface area than the fine material (2.08 vs 2.52 m2/g), this difference does not appear to be reflected in the k0 values of Table S2. Because of the spatial variability in the reactivity of the iron at a particular time, the assumptions inherent in eq 2 are not strictly met. Thus, while the nomenclature and definitions commonly assigned to the mixed-order expression are used, the physical interpretation of the parameters may not be accurate. Also, spatial and temporal changes in porosity because of mineral accumulation (in columns B-D) and gas generation (in all columns) may have also contributed to the changing shapes of the profiles. These changes were not accounted for in preparing Figure 1. Column A, which received only TCE in deionized water, showed migration of the profiles (Figure 1a). Although the reason is not clear, this may be a consequence of formation of ferrous hydroxide (Fe(OH)2) or magnetite (Fe3O4) and thus changes in surface conditions of the iron. For this column, however, no further migration of the TCE profiles was observed after 1719 PV, suggesting that the profiles had reached steady state. Columns B and C received the identical feed solution (100 mg/L CaCO3 + 10 mg/L TCE) and showed similar rates of migration of the TCE profiles (Figure 1b and 6434
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 20, 2006
1c), although column C contained the coarse-grained iron. Whereas column B was terminated after 2281 PV as a consequence of reduced permeability, in column C (coarse iron) the decline in permeability was substantially less and this column was continued to over 3000 PV. Carbonate precipitates appeared to be the main cause of passivation for both columns. For column D, receiving 500 mg/L CaCO3 + 10 mg/L TCE, the rate of migration of the profiles was much greater than in the other columns (Figure 1d), indicating a higher rate of passivation. However, 60% of TCE removal was sustained to the end of the experiment, indicating that carbonate precipitates had not completely passivated the iron. Degradation Products. The major chlorinated degradation product was cis-DCE, followed by 1,1-DCE, and VC. The concentrations of cis-DCE and 1,1-DCE increased with distance along the columns and then decreased toward the effluent end, and the concentration of VC increased toward the effluent end, suggesting that some portion of the TCE was degraded by sequential dechlorination (17). Throughout the period of operation, the sum of all chlorinated intermediates was generally consistent at less than 5% of the initial TCE concentration. Thus, most of the TCE was considered to proceed to ethane via β-elimination (17). The distribution of degradation products did not change substantially over the course of the experiments. The dominant non-chlorinated hydrocarbons were ethene and ethane, with substantially lower concentrations of methane, propene, propane, n-butane, 1-butene, and acetylene. In general, the proportion of ethene over ethane increased over time. Carbon mass balances, including the chlorinated degradation products, were between 70 and 96%. The chloride concentrations increased over the length of the columns, as TCE and the chlorinated products degraded. Generally, the rate of chloride production was slower at later times, providing further evidence of passivation of the iron. Chloride mass balances were between 82 and 119% along the length of each column, with no noticeable trend over time. Analytical uncertainties for each chlorinated compound and chloride were 10-15%. Incomplete analyses for hydrocarbons may be additional contributions to the uncertainties in carbon mass balances. Inorganic Species. For the columns receiving dissolved calcium carbonate, the pH values were closely related to other geochemical parameters. As an example, Figure 2 shows profiles of pH, alkalinity, calcium, and total dissolved iron, measured over time for column D. The pH did not increase significantly in the region where carbonate precipitation was occurring (e.g., up to 15 cm from the influent end at 18 PV, Figure 2a and 2b). Alkalinity increased adjacent to the influent end, particularly at early times. The feed solution contained almost equal amounts of H2CO30 and HCO3-. While carbonate precipitation kept the pH relatively low (
