Environ. Sci. Technol. 2007, 41, 1615-1621
Degradation of Disinfection Byproducts by Carbonate Green Rust CHAN LAN CHUN, RAYMOND M. HOZALSKI, AND WILLIAM A. ARNOLD* Department of Civil Engineering, University of Minnesota, 500 Pillsbury Drive SE, Minneapolis, Minnesota 55455-0116
Disinfection byproducts (DBPs) in drinking water flowing through corroded iron or steel pipes may encounter carbonate green rust (GR(CO32-)), a mixed Fe(II)/Fe(III) hydroxide mineral and potent reductant. This research was performed to investigate the kinetics and pathways of the degradation of selected halogenated DBPs in the presence of GR(CO32-). Trichloronitromethane was rapidly degraded to methylamine via sequential hydrogenolysis followed by nitroreduction. Haloacetic acids reacted solely via sequential hydrogenolysis. Trichloroacetonitrile, 1,1,1-trichloropropanone, and trichloroacetaldehyde hydrate were transformed via hydrolysis and hydrogenolysis. Chloroform was unreactive over 300 h. The buffer identity affected reductive dehalogenation rates of DBPs, with faster rates in MOPS buffer than in carbonate buffer, the latter being representative of the buffer in drinking water systems. GR(CO32-) was unstable in both buffers and transformed to magnetite within 48 h. Thus, slower reacting compounds (half life >3 hours) were transformed by a combination of minerals. Reductive dehalogenation kinetics were influenced by DBP chemical structure and correlated with oneelectron reduction potential.
Introduction In water distribution systems, oxidizing agents present in drinking water, such as oxygen and chlorine, lead to the formation of corrosion products on unlined iron or steel pipe surfaces (1-5). Corrosion products are a mixture of ferrous and ferric iron minerals including goethite, magnetite, siderite, and green rust, and other minor constituents including natural organic matter and trace metals (1-3, 6). Corrosion products readily interact with drinking water constituents including disinfectants (1, 6, 7) and disinfection byproducts (DBPs) (8). For example, the loss of free chlorine was partially attributable to reaction with tubercles collected from iron pipes in Columbus, Ohio (6, 7). Valentine et al. (1) reported that reactions of free chlorine and chloroamines with the organic matter in corrosion products led to formation of the DBP chloroform (trichloromethane; TCM). Recently, degradation of chlorinated DBPs, specifically, trichloroacetic acid (TCAA) and trichloronitromethane (TCNM), by corrosion products obtained from cast iron pipes has been reported (8). These reactions may influence the fate of DBPs in drinking water distribution systems. Because the corrosion products in water mains are heterogeneous materials, experiments with field-collected * Corresponding author phone: (612) 625-8582; fax: (612) 6267750; e-mail:
[email protected]. 10.1021/es061571f CCC: $37.00 Published on Web 01/27/2007
2007 American Chemical Society
materials do not give insight into the importance of different minerals in the fate of DBPs. Thus, it is useful to investigate reactions of DBPs with individual minerals in relatively simple systems. Previously, we reported multiple loss mechanisms for DBPs in the presence of aqueous Fe(II), magnetite, or Fe(II) adsorbed onto goethite or magnetite, including reductive dehalogenation, hydrolysis, and sorption (9). Each of these reductants has been detected in iron water mains (15). In the presence of Fe(II) adsorbed onto magnetite (1 mM Fe(II) and 0.8 g Fe3O4/L), 9% of 1,1,1-trichloropropanone (1,1,1-TCP) was adsorbed onto magnetite and the remainder was degraded to TCM and 1,1-dichloropropanone (1,1-DCP) via hydrolysis (81%) and hydrogenolysis (replacement of a halogen by hydrogen; 10%), respectively (9). In the reductive dehalogenation of DBPs, Fe(II) adsorbed onto iron minerals consistently had greater reactivity than either aqueous Fe(II) or structural Fe(II) present in magnetite (9). This paper extends prior work by investigating reactions of DBPs with green rust, a mixed Fe(II)-Fe(III) hydroxide mineral formed in suboxic iron water mains via corrosion reactions (6, 10). Green rust ([FeII(1-x)FeIIIx(OH)2]x+‚[(x/n)An-‚(m/n)H2O]x-, where x ) the ratio FeIII/Fetot (1/3-2/3), A ) n-valent anion (Cl-, Br-, CO32-, or SO42-), and m ) amounts of interlayer water) has a pyroaurite-type structure in which positive charged brucite-like layers alternate with interlayers of anions and water molecules (11). Carbonate green rust (GR(CO32-)) was initially found inside the corrosion film on water pipes (10). In a water distribution pipe from Columbus, Ohio, Tuovinen et al. (6) reported GR(CO32-) and GR(SO42-) to be the major corrosion products. In addition to iron water mains, green rusts have been observed in reducing environments including anoxic soils and sediments (12, 13) and zerovalent iron permeable reactive barriers (14-17). Green rusts have the capability to attenuate redox-active contaminants including halogenated organic compounds (18-20), inorganic ions (18-23), and heavy metals (24, 25). Green rusts may also contribute to the degradation of halogenated DBPs in drinking water distribution systems. In this research, batch experiments were performed to investigate the pathways and kinetics of the degradation of halogenated DBPs in the presence of synthetic GR(CO32-), the most common green rust in iron water mains.
Experimental Section Additional details for the following sections are provided in the Supporting Information (SI). Chemicals. Chemicals, suppliers, and purities are listed in the SI. DBPs used as starting materials in experiments include TCNM, dichloronitromethane (DCNM), chloronitromethane (CNM), nitromethane (NM), trichloroacetonitrile (TCAN), dichloroacetonitrile (DCAN), tribromoacetic acid (TBAA), dibromoacetic acid (DBAA), bromoacetic acid (BAA), TCAA, dichloroacetic acid (DCAA), 1,1,1-TCP, trichloroacetaldehyde/chloral hydrate (TCAh), and TCM. GR(CO32-) Synthesis and Characterization. Synthesis of GR(CO32-) was performed in an anaerobic chamber (97% N2/3% H2, Coy Laboratory Products) using deoxygenated Milli-Q water at 22 ( 3 °C. GR(CO32-) was prepared using a modification of the method of Williams and Scherer (24). After synthesis, centrifugation, and flash drying using acetone/ argon sparging, X-ray diffraction (XRD; Co KR ()1.789 Å) radiation, continuous scans from 5 to 80° 2θ at 2.4° 2θ/min, Philips Xpert) confirmed that the bluish-green product was GR(CO32-), and no other iron oxides were detected. Samples were mixed with glycerol to minimize oxidation during VOL. 41, NO. 5, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1615
FIGURE 1. X-ray diffraction patterns of fresh GR(CO32-) and aged GR(CO32-). Results are shown for (a) 25 mM MOPS buffer and (b) 50 mM carbonate buffer. G, M, and S are used to label GR(CO32-), magnetite, and siderite, respectively. The initial pH of each buffer is 7.5. The fresh GR(CO32-) and aged GR(CO32-) were flash-dried using acetone under Ar sparging and then mixed with glycerol for XRD analysis. Note that the fresh GR(CO32-) was not exposed to either MOPS buffer or carbonate buffer. handling and XRD analysis (26). The flash-dried GR(CO32-) was used in reaction experiments within 24 h. Batch Experiments. DBP degradation experiments were carried out in 123 mL serum bottles containing a 2.4 g/L GR(CO32-) suspension buffered at pH 7.5 with Ar-sparged 25 mM MOPS buffer or 50 mM carbonate buffer at 22 ( 3 °C. All batch experiments were conducted in duplicate. GR(CO32-) Aging Experiments. The stability of GR(CO32-) was investigated by incubating GR(CO32-) in buffered, Arsparged water and monitoring the changes in mineralogy and reactivity over time. Mineralogical changes were monitored using XRD. Release of Fe(II) via dissolution of GR(CO32-) during the aging experiments was measured using the Ferrozine method (27). The effects of aging on GR(CO32-) reactivity were investigated by determining the degradation rate of a probe DBP (TCAN) in the presence of aged GR(CO32-). The batch experiments were performed as described above except that the GR(CO32-) was pre-aged in buffer for up to 48 h prior to introduction of TCAN. Analytical Methods. With the exception of the haloacetic acids (HAAs), NM, and methyl amine (MA), all DBPs and their degradation products were detected and quantified using gas chromatography (GC) with electron capture detection (Trace GC, ThermoQuest) after a 0.5 mL suspension sample (0.1 mL for halonitromethanes) was extracted with 1.0 mL of methyl-t-butyl ether. The mixture was vortexed for 1 min and the supernatant was transferred to an autosampler vial. For the HAAs, NM, and MA, suspension samples were filtered though a 0.2 µm PTFE Gelman Acrodisk syringe tip filter to remove iron mineral particles and to stop the reaction (see SI for details). Capillary electrophoresis was used for analysis of the HAAs (28). NM was analyzed using headspace GC with flame ionization detection (29). MA was detected using high-performance liquid chromatography (Waters LC Module 1 Plus) after derivatization of the filtered samples (30). The suspension pH was measured before and after reaction. The conductance of each buffer was measured using 1616
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 5, 2007
a conductance/TDS meter (model 72, Engineered Systems & Designs). Data Analysis. The overall loss and individual hydrolysis and reductive dehalogenation pseudo-first-order rate constants of the DBPs were determined by fitting the experimental data using Scientist for Windows (v. 2.01, Micromath Research). One-electron reduction potentials for the dehalogenation half reactions (E1H) were determined via free energies computed using computational chemistry software as described by Chun et al. (9).
Results GR(CO32-) Aging in Aqueous Solution. XRD patterns of the synthesized GR(CO32-) were comparable to those reported in previous studies (24, 31). The surface area measured by three-point BET gas adsorption with N2 was 37.1 ( 9.4 m2/g, which is similar to the value reported by Williams and Scherer (24). Due to its transient nature, the aging of GR(CO32-) in water in the absence of DBPs was investigated prior to conducting batch experiments with DBPs. The intensity of the GR(CO32-) peaks substantially decreased over the first 7 h of exposure to MOPS buffer (Figure 1a), but no new peaks were observed during this period. By 24 h, the GR(CO32-) peaks had disappeared and new peaks, consistent with those of magnetite (FeIIFeIII2O4), started to form. The crystallization of magnetite continued with time. No other crystalline minerals, such as siderite, were observed. The color of the suspension changed from bluish-green to dark green in the first 7 h which implies a change in the Fe(II)/Fe(III) ratio in GR(CO32-). Ruby et al. (32) observed different colors of GR(CO32-) depending on ratio of Fe(III) to total Fe. The color changed to black over longer times which indicates the formation of magnetite. Similar results were observed for GR(CO32-) aging in 50 mM carbonate buffer. The effect of aging on GR(CO32-) reactivity was also investigated using a probe compound, TCAN (Figure 2). In
FIGURE 2. (a) Effect of GR(CO32-) aging on the first-order reduction rate constant of TCAN by 2.4 g/L GR(CO32-) in the presence of 25 mM MOPS buffer (b) and 50 mM carbonate buffer (O). Error bars represent standard deviations for duplicate samples. (b) Aqueous Fe(II) concentration dissolved from GR(CO32-) before injecting TCAN in each reactor. Error bars represent standard deviations. Error bars are the standard deviations of duplicate samples. The initial pH of each buffer is 7.5. the MOPS-buffered system, the pseudo-first-order reductive dehalogenation rate constant (kr) for TCAN remained constant for aging times (i.e., pre-exposure times prior to initiating reaction with TCAN) up to 3 h, but decreased by 83% after aging the GR(CO32-) for 48 h in MOPS buffer. The kr in carbonate buffer was relatively stable for aging times up to 10 h, but decreased by 63% after 48 h. The kr in carbonate buffer was consistently less than that in MOPS buffer (even though transformation of GR(CO32-) into magnetite occurred more rapidly in MOPS buffer). The aqueous Fe(II) concentrations in the reactors as a function of pre-exposure time to buffer are shown in Figure 2b. The concentrations were measured immediately prior to the injection of TCAN (experiments of Figure 2a and b). In MOPS buffer, the aqueous Fe(II) concentration increased to a maximum of ∼1 mM. The Fe(II) concentration then decreased with continued incubation time. The increase in aqueous Fe(II) concentration was due to dissolution of GR(CO32-). The Fe(II) concentrations in the carbonate buffer remained relatively stable (60-150 µM) throughout the aging period. Reaction Kinetics. Reaction rate constants for the DBPs are summarized in Table 1. In most cases, the contribution of reductive dehalogenation to the overall degradation rate was much greater than that of hydrolysis. The pseudo-firstorder reductive dehalogenation rate constant (kr) for trichlorinated DBPs followed the trend TCNM > TCAN > TCAA > TCP ∼ TCAh . TCM in both MOPS and carbonate buffer. Note that for compounds that reacted slowly (i.e., TCP, TCAh) the rate constants represent composites for a mineral phase that was initially GR(CO32-) but transformed to magnetite over the timescales in which the DBP reacted. For HAAs, the
brominated compounds reacted more rapidly than their chlorinated analogues, consistent with previous research with Fe(0) (28, 33). The rates of DBP reductive dehalogenation, however, were dependent on the buffer. The reductive dehalogenation rate constants for reactions in MOPS buffer were greater than for those in carbonate buffer for all of the DBPs studied. The ratio of the reductive dehalogenation rate constant in MOPS buffer to that in carbonate buffer (RM/C), however, varied for the individual DBPs. Hydrolysis rate constants in carbonate buffer were consistently greater than those in MOPS buffer. The pH increased to 7.9-8.1 in carbonate buffer (due to incomplete buffering of the carbonate/hydroxide release upon oxidation of the green rust and/or consumption of protons by the DBP reduction), while it remained relatively constant (near 7.5) in MOPS buffer. Reaction Pathways. Despite the influence of buffer identity on reaction kinetics, the reaction pathways of the DBPs in the presence of GR(CO32-) were similar in both buffers. The degradation of TCNM by GR(CO32-) proceeded via sequential hydrogenolysis followed by reduction of the nitro-group: TCNMfDCNMfCNMfNMfMA (Figure 3). There was no loss of TCNM or its daughter products in GR(CO32-)-free controls. TBAA and TCAA were also transformed via sequential hydrogenolysis (Figure S1 in the Supporting Information): TBAAfDBAAfBAA and TCAAfDCAAfchloroacetic acid (CAA). In a separate experiment with BAA as a starting material, BAA was slowly reduced to form acetate (Figure S2), but CAA was not degraded over 150 h. The remaining DBPs, TCAN, 1,1,1-TCP, and TCAh, were transformed via parallel pathways of hydrolysis and sequential hydrogenolysis in the presence of GR(CO32-). In both buffer systems, TCAN was degraded to DCAN via hydrogenolysis and to trichloroacetamide (TCAM) via hydrolysis. DCAN was hydrolyzed to dichloroacetamide (DCAM) and/ or reduced to CAN (Figure S3). The distribution of products, however, differed for the two buffers. In MOPS buffer, reductive dehalogenation of TCAN (94%) was predominant over hydrolysis (6%). In the carbonate buffer system, reduction (46%) and hydrolysis (54%) occurred at comparable rates. For DCAN, the rates of formation of reduction and hydrolysis products were similar in MOPS buffer, while only the hydrolysis product was observed in carbonate buffer. Plots showing the degradation of 1,1,1-TCP and the simultaneous formation of two major products, TCM (via hydrolysis) and 1,1-DCP (via hydrogenolysis), are included as Figure S4. In MOPS buffer, the contribution of hydrolysis (87%) to the overall degradation of 1,1,1-TCP was much larger than that of hydrogenolysis (13%). Likewise, TCAh was degraded to DCAh via hydrogenolysis (69%) and to TCM via hydrolysis (31%). Also, trichloroethanol, the product formed via reduction of the aldehyde group, was detected in trace amounts. No loss of TCM was observed over 300 h.
Discussion GR(CO32-) Stability and DBP Reductive Dechlorination. Aging results (Figures 1 and 2) indicated that the lifetime of GR(CO32-) was 3 h in MOPS buffer and 10 h in carbonate buffer. This transformation of GR(CO32-) in our buffered solutions (after Ar sparging) occurred very rapidly compared to GR(CO32-) stored as a dried powder in an anaerobic chamber (stable over a 2-week period) (24). One explanation for the observed rapid transformation in our experiments is reaction with oxygen that remained in solution following deoxygenation or that leaked in during the experiments, because green rusts are rapidly transformed in the presence of dissolved oxygen (31). The initially deoxygenated buffer and control reactors without GR(CO32-) show traces of oxygen (
