Improved Passive Treatment of High Zn and Mn Concentrations Using

Nov 14, 2008 - Caustic magnesia (“MgO”) has been used successfully as an alternative alkaline material to remove these metals almost completely fr...
2 downloads 8 Views 3MB Size
Download PDF
Environ. Sci. Technol. 2008, 42, 9370–9377

Improved Passive Treatment of High Zn and Mn Concentrations Using Caustic Magnesia (MgO): Particle Size Effects ¨ T T I N G , * ,†,‡,§ TOBIAS S. RO CARLOS AYORA,† AND JESUS CARRERA† Institute of Earth Sciences “Jaume Almera”, CSIC, Lluis Sole´ i Sabar´is s/n, 08028 Barcelona, Spain, and Sir Joseph Swan Institute, Newcastle University, Newcastle upon Tyne, NE1 7RU, United Kingdom

Received August 4, 2008. Revised manuscript received October 16, 2008. Accepted October 20, 2008.

High concentrations of divalent metals such as Zn, Mn, Cu, Pb, Ni, Cd, Co, etc. are not removed satisfactorily in conventional (calcite- or organic matter-based) passive treatment systems. Caustic magnesia (“MgO”) has been used successfully as an alternative alkaline material to remove these metals almost completely from water, but columns with coarse-grained MgO lost reactivity or permeability due to the accumulation of precipitates when only a small portion of the reagent had been spent. In the present study, MgO was mixed with wood chips to overcome these problems. Two columns with different MgO grain sizes were used to treat Zn- and Mn-rich water during one year. Performance was compared by measuring depth profiles of chemical parameters and hydraulic conductivity. The column containing 25% (v/v) of MgO with median particle size of about 3 mm displayed low reactivity and poor metal retention. In contrast, the column containing only 12.5% (v/v) of MgO with median particle size of 0.15 mm depleted Zn and Mn below detection limit throughout the study and had a good hydraulic performance. 95% of the applied MgO was consumed in the zone where Zn and Mn accumulated. The fine alkaline grains can dissolve almost completely before the growing layer of precipitates passivates them, whereas clogging is prevented by the large pores of the coarse inert matrix (wood chips). A reactive transport model corroborated the hypotheses that Zn(II) was removed due to its low solubility at pH near 10 achieved by MgO dissolution, whereas Mn(II) was removed due to rapid oxidation to Mn(III) at this pH and subsequent precipitation. The model also confirmed that the small size and large specific surface area of the MgO particles are the key factor to achieve a sufficiently fast dissolution.

Introduction Heavy metal-polluted discharges from active or abandoned mines and industrial sites are a major hazard for the * Corresponding author phone: +44 (0)191 246 4902; fax: +44 (0)191 246 4961; e-mail: [email protected]. † Institute of Earth Sciences “Jaume Almera”. ‡ Newcastle University. § Present address: Hydrogeochemical Engineering Research and Outreach group (HERO), Sir Joseph Swan Institute, third Floor, Devonshire Building, Newcastle University, Newcastle upon Tyne, NE1 7RU, United Kingdom. 9370

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008

environment, and may persist for decades or centuries after closure (e.g., 1,). “Active” treatment (controlled addition of chemicals to precipitate the metals) is straightforward but expensive (e.g., ref 2,), and may be impracticable at abandoned sites. Passive treatment systems may be a cost-effective alternative (e.g., ref 3,). Once installed, these systems do not need the continuous supply of reagents or energy, and they only require of an infrequent maintenance. Usually, passive systems are gravity-fed flow-through reactors which remove metals using geochemical processes such as calcite dissolution, bacterial sulfate reduction, sorption, ion exchange, etc. (1, 4). These processes, however, are insufficient to remediate high concentrations of divalent metals such as Zn, Mn, Cu, Pb, Ni, Cd, Co, etc. (tens to hundreds of mg/L, commonly found at metal mining and smelting sites) due to the following reasons: (i) under field conditions calcite dissolution only raises pH to values around 7, which are insufficient to precipitate divalent metals (5); (ii) these metals are toxic to sulfate reducing bacteria (6, 7) at the concentrations considered here; and (iii) sorption and exchange sites may saturate rapidly under high metal loads (8). Brucite (Mg(OH)2), produced from the hydration of magnesium oxide (MgO), buffers solution pH between 8.5 and 10 where solubilities of many divalent metal compounds are low (5). It is therefore an attractive alternative reagent for passive remediation systems. High purity MgO or Mg(OH)2 obtained from seawater evaporation and precipitation with sodium hydroxide is usually too expensive for application in wastewater treatment. Caustic magnesia produced from calcined magnesium carbonate is an alternative product containing mainly brucite (MgO) and minor amounts of lime (CaO), quartz (SiO2), magnesite (MgCO3), dolomite (MgCa(CO3)2), and calcite (CaCO3). The price of caustic magnesia is only slightly higher than that of lime or Ca(OH)2, and similar to or cheaper than that of other alkaline reagents commonly used to neutralize acidic waters, such as caustic soda, soda ash, or ammonia. Cost of caustic magnesia can be reduced if precipitator dust, a waste material from the calcination furnace, can be used. Producers of caustic magnesia are located around the world (9). Cortina et al. and Ro¨tting et al. (5, 10) conducted laboratory column experiments which demonstrated that caustic magnesia (in the following abbreviated as “MgO”) can be used to deplete concentrations of 75 mg/L of Zn, Mn, Cu, Pb, Ni, Cd, and Co to concentrations below 0.5-100 µg/L (depending on the metal). Nevertheless, reactivity of the columns decreased sharply due to coating of the reactive surfaces when only a small fraction of initial MgO had been used. Cortina et al. (5) showed that MgO use can be improved if the alkaline material is mixed with an inert matrix, quartz sand in their study. The present work is the continuation of these experiments. Here, MgO is mixed with wood chips instead of quartz sand to create a more reactive and more permeable substrate. The curved flakes of wood provide a matrix with higher porosity and larger pore size than gravel particles. The wood chips also offer a large surface area to which the MgO particles can adhere. Therefore, the grain size of MgO can be reduced, which increases specific surface area and accelerates reaction kinetics while permeability of the mixture remains high. The bigger pore size and the more separated grains acting as nuclei for precipitation should also reduce clogging problems. In the present study, the MgO-wood chip mixture was used to remove high concentrations of Zn and Mn(II), two 10.1021/es801761a CCC: $40.75

 2008 American Chemical Society

Published on Web 11/14/2008

FIGURE 1. (Top) evolution of pH at different sampling points in (left) the fine- and (right) the coarse-grained MgO column. (Bottom) column outlet concentrations of Mg, Ca, Zn, and Mn. Average flow rates during different phases of the experiment are given above the graphs. Outlet Ca concentration of the coarse-grained MgO column declined from 874 to 233 mg/L during the first two months (not shown). divalent metals that are very common in metal mine and industrial drainage. Most divalent metals (Ni, Cd, Co, Pb, Cu) were found to behave similar to Zn (5, 10). Moreover, Mn(II) was treated specifically because retention may be linked to oxidation and precipitation of Mn(III) phases. The chemical and hydraulic performance of two columns with different MgO grain sizes was compared in a one year laboratory experiment. The columns were then dismantled, and petrography and mineralogy of the residual solid phase was investigated. The experimental results were reproduced using a numerical reactive transport model to test the hypotheses on the geochemical metal elimination processes.

Materials and Methods Columns. The columns were fabricated from transparent acrylic glass (9.6 cm inner diameter, height 35 cm; see Supporting Information (SI) Figure S1). Sampling ports for 0.1 µm filtered water samples (Eijkelkamp Macro Rhizon) were placed at 3.5, 8.5, and 13.5 cm depth, pressure ports for measurement of hydraulic conductivity at 1.0, 3.5, 8.5, and 13.5 cm depth. Each column contained a perforated drain pipe and a 2.5 cm layer of quartz gravel (5-8 mm diameter) at the bottom, and 15 cm of reactive substrate. Samples of caustic magnesia were provided by Magnesitas Navarras S.A., Spain. Two columns with different MgO grain sizes were prepared. The fine-grained MgO column was filled with a mixture of 12.5% (v/v) caustic magnesia precipitator dust of median particle size ∼ 0.15 mm (“Magna PC”, for details see SI Table S1) and 87.5% (v/v) wood chips. The coarse-grained MgO column was filled with a mixture of 25% (v/v) caustic magnesia gravel of median particle size ∼ 3 mm (“Magna L”, SI Table S1) and 75% (v/v) wood chips. Assuming that both caustic magnesia samples have similar particle shapes, the total surface area of the reactive material in the fine-grained column was about 10 times greater than that of the coarse-grained column, even though it contained only half the mass of MgO (SI Table S1). Input solutions for the columns were prepared from commercial mineral water in order to simulate the ionic

composition of natural water. The solution was fed to the supernatant of each column using a Gilson “Minipuls 3” peristaltic pump at a median Darcy flow rate of about 0.1 m3/m2 · day (volumetric flow rate 0.5 L/day). It flowed downward through the reactive substrate and the column outlet at the drain pipe into an output container that acted as a sedimentation tank (for details see SI Figure S1). The experiments were started using pure mineral water (pH 8.3, Mg 9 mg/L, Ca 38 mg/L, Zn < 0.2 mg/L, and Mn < 0.2 mg/L; for details see SI Table S2). After two weeks of operation, 310 mg/L Zn as ZnSO4 · 7H2O were added and pH was adjusted to 6 with concentrated H2SO4 in order to resemble water from Zn smelting sites in areas with carbonate host rocks. After four months of operation, also 30 mg/L Mn as MnCl were added (SI Table S2). Water Sampling and Analysis. Water samples from the column supernatant, column outlet and output container (SI Figure S1) were taken at weekly intervals at the beginning of the experiments, and at monthly intervals toward the end. Depth profiles were taken every 1-2 months from the sampling ports within the substrate (SI Figure S1). pH was measured using a Crison glass combination electrode calibrated with buffer solutions of pH 2, 7, and 9. Redox potential was measured with a ThermoOrion Sure-Flow ORP platinum combination electrode which was checked for accuracy with 220 and 465 mV standards. Total (gross) alkalinity was measured using CHEMetrics “Titrets” test kits (range 10-100 or 100-1000 mg/L as CaCO3, accuracy approximately 5%). Aliquots of 4 mL were acidified with HNO3 and stored in the dark at 4 °C until analysis for major cations and S by ICP-AES (Perkin-Elmer Optima 3200 RL). Detection limits were 1 mg/L for S, 0.5 mg/L for Mg and Ca, and 0.2 mg/L for Zn and Mn due to the necessary dilution of the samples. Hydraulic Conductivity and Porosity. Hydraulic conductivity of the filling material was determined in depth intervals of 1-5 cm by measuring head-loss between the different pressure ports (Figure 1). During the measurements, flow rate VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9371

FIGURE 2. Depth profiles of pH and concentrations of Mg, Zn, and Mn in (left) the fine- and (right) the coarse-grained MgO column. The values at -10 cm depth represent the input container, those at -1 cm the supernatant, the gray shaded area from 0 to 13 cm the sampling ports within the column substrate, the values at 14 cm represent the column outlet, and those at 20 cm the output container.

through the column was temporarily increased to around 3 mL/s to obtain significant head differences. The hydraulic 9372

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008

conductivity K (cm/s) in the depth interval between two pressure sampling ports was calculated using Darcy’s Law:

K)

QL A∆h

(1)

where Q (mL/s) is flow rate, A (cm2) is tank cross-section perpendicular to flow, and L and ∆h (cm) are distance and head-loss between pressure ports, respectively. Several readings were taken to calculate mean values and standard deviations. Initial porosity was determined gravimetrically by weighing the columns before and after they had been saturated with water from the bottom. Solid Samples. After the end of the experiments, the substrate and precipitates of the columns were excavated from the top in a glovebox under N2 atmosphere. Samples were immediately frozen at -30 °C and freeze-dried. Thin petrographic sections on glass mounts were prepared with Struers Epofix epoxy resin and hardener, and polished with diamond paste from 3 µm down to 1 µm. Granular material and thin sections of the samples were observed under a JEOL JSM840 scanning electron microscope with oxford link energy dispersive system (SEM-EDS). Element maps were recorded using a Leica Cambridge Stereoscan S-360 scanning electron microscope with Oxford Link EDS. Samples for chemical analyses and identification of mineral phases were ground in a tungsten carbide mill for 1-2 min. Powder diffractograms were obtained using a Bruker D5005 X-ray diffractometer (XRD) with Cu LR radiation. Chemical composition of the newly formed precipitates was determined by ICP-AES after dissolution in HNO3 7.5 M. Total inorganic carbon was determined by measuring the volume of CO2 gas produced after addition of HCl 16% to a known mass of ground solid sample. Geochemical Modeling. A 1-D reactive transport model was created using the code RETRASO (11) to test the hypothesized mechanisms of Zn and Mn removal. A finiteelement mesh with 310 linear elements and a total length of 15 cm was used to represent the depth profile of the finegrained MgO column.

Results and Discussion Water Samples. Performance of the two columns differed considerably, as summarized by the time evolution of pH at several sampling points and Mg, Ca, Zn, and Mn column outlet concentrations (Figure 1; Zn and Mn are only given when these metals were added to the input solution). Both columns were operated for more than 1 year (54 weeks). Supernatant pH of both columns started near 9, but stabilized around 6.5 when ZnSO4 · 7H2O and H2SO4 were added after two weeks. In the fine-grained MgO column, outlet pH was near 10 throughout the study; outlet Mg concentration rose to 120 mg/L within 2 months and remained close to this value (Figure 1, left-hand side). Water samples from the column outlet were saturated with respect to brucite from the second week until the end of the experiment, indicating that this mineral controlled pH in the fine-grained column. Zn and Mn were always below detection limit at the column outlet. Ca concentration was 10-20 mg/L above the input value (38 mg/L), indicating that a Ca-bearing solid (probably calcite) also dissolved in the column. pH in the output container was near 8 at all times (maximum 8.65). The decrease with respect to the outlet value is probably due to CO2 uptake from the atmosphere and precipitation of magnesite (MgCO3). This is an advantageous property, because very alkaline waters usually have to be neutralized with additional chemicals before discharge to comply with regulations. While this is often necessary after treatment with lime, caustic soda, etc. which produce pH > 10, it is unnecessary for treatment with caustic magnesia.

FIGURE 3. Hydraulic conductivity profiles of (left) the fine- and (right) the coarse-grained MgO column with standard deviations (n: number of repetitions). The coarse-grained MgO column behaved very differently (Figure 1, right-hand side). Outlet pH started above 12 concomitant with high Ca concentrations of up to 874 mg/L and Mg concentrations below detection limit. This indicates that lime (CaO), which is present in small quantities in caustic magnesia (SI Table S1), initially controlled pH until it was exhausted (also observed by (5)). After 2.5 months, pH at the column outlet had decreased to 11, Ca concentration had declined to 233 mg/L, and Mg concentration started to increase, reaching maxima of about 100 mg/L from month 4 to 5. During the rest of the experiment, outlet Mg decreased gradually to 62 mg/L and pH to 7, whereas Ca concentration was similar to the fine-grained column. Brucite was near saturation only during the fourth month; it was supersaturated before (during lime control) and subsaturated afterward. This shows that brucite did not control pH despite the large quantity of caustic magnesia applied in the coarsegrained column. Zn and Mn were detected at the column outlet at the end of the fourth month, and increased steadily afterward. By the end of the study, half of the input Zn and two-thirds of the input Mn concentration already reached the outlet of the coarse-grained column. Depth profiles (Figure 2) provide further insights into the performance of the two columns. In the fine-grained column, pH and dissolved Mg attained near-maximum values in the upper 4 cm of substrate during the first 7 months of the experiment, while dissolved Zn and Mn were below detection limit at this depth. This shows that MgO dissolution was very fast in this column. During the rest of the study pH and Mg approached supernatant values at 4 cm depth but attained maximum values at 9 cm depth. This suggests that MgO had been exhausted by that time in the upper 4 cm of the substrate. Dissolved Zn and Mn showed a chromatographic segregation. Mn was more mobile through the substrate and precipitated at a higher pH than Zn. A thin rim of dark precipitates of Mn was observed visually ahead of the front of white precipitates of Zn (see photos in SI Figure S2, lefthand side). Both fronts moved downward through the column with time. The coarse-grained column was much less reactive. Gradients of pH and dissolved Mg, Zn, and Mn (Figure 2, right-hand side) were less steep than in the fine-grained column, showing that MgO dissolution was slower. Zn and Mn retention was poor. After 4.5 months the white precipitation front had moved halfway through the column, and the black precipitates had almost reached the bottom (SI VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9373

FIGURE 4. Major chemical composition of bulk solid samples (mg/g) from (left) the fine- and (right) the coarse-grained MgO column. TIC ) total inorganic carbon as HCO3-. Wood chips (organic carbon) and structural water are not represented.

FIGURE 5. (A) Scanning electron microscope image of a granular sample of the fine-grained MgO column from 2 cm depth. In the central area the outer crust of Zn-rich platy crystals has broken away and reveals the inner Mn-rich precipitates and several MgO grains. (B) EDS-element maps of Mg (green), Mn (red), and Zn (blue) of the central area of the SEM image superposed onto electron backscatter image (C). Figure S2, upper right). Two weeks later, Zn and Mn were detected at the column outlet, both increasing with time (SI Figure S2, lower right). Porosity and Hydraulic Conductivity. Initial gravimetric porosity was 71% for the fine-grained column, and 65% for the coarse-grained column. These values are higher than those obtained in granular substrates (maximum porosity usually about 40% for well-sorted spheres), because the curved flakes of wood form a matrix with higher porosity than round particles of the same size. The porosity of the fine-grained column was higher, since less MgO was mixed with the wood chips. Hydraulic conductivity K in the fine-grained column (Figure 3, left) was high throughout the substrate during the first two measurements, except for the depth interval 1-3.5 cm where K decreased with time. In the last measurement, K had also decreased considerably in the interval 3.5-8.5 cm. This is consistent with the water samples (Figure 2) and 9374

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008

visual observations (SI Figure S2), showing that the front of Zn and Mn precipitation was moving downward through the column. It suggests that precipitates were accumulating in an increasing volume of substrate over time. Despite the substantial decrease of hydraulic conductivity, the finegrained column did not clog at any time. Head loss across the entire substrate at normal flow rate was less than 3 mm by the end of the study. The coarse-grained column (Figure 3, right) had a high permeability throughout the substrate during the entire study, but this is consistent with the poor metal retention of this column. As only few precipitates formed and little MgO dissolved inside this column, hydraulic conductivity changed only marginally. Solid Samples. Excavation of the fine-grained column after the conclusion of the experiment revealed that the uppermost 1 cm consisted of loose wood chips covered with white gelatinous precipitates. They bore high Zn concentra-

FIGURE 6. Depth profiles of reactive transport model results for the fine-grained MgO column: Left: simulated (lines) and measured (symbols) pH and pe; middle: total dissolved concentrations of primary aqueous species as simulated (lines) and as measured after 12 months of operation (symbols); right: simulated total amount of precipitation (positive values) or dissolution (negative values) of minerals. TIC as HCO3-. tions with minor Mg and Mn (Figure 4, left), and XRD indicated the presence of a hydrated zinc hydroxycarbonate (Zn4CO3(OH)6 · H2O). From 1 to 6 cm depth the column contained cemented wood chips covered with dark gray precipitates and white spots. These samples had similar Zn concentrations, but higher sulfate and Mn. XRD displayed peaks of β-Zn(OH)2, hydrozincite (Zn5(CO3)2(OH)6), brianyoungite (Zn12(CO3)3SO4(OH)16), and Zn4CO3(OH)6 · H2O, consistent with the chemical composition of the samples (Figure 4, left). No Mnbearing phase could be identified by XRD in our samples. This had to be expected because Mn concentrations in the input solution and in the solid samples were 10 times lower than those of Zn. SEM images (Figure 5) revealed that Mn precipitated as a mass without visible crystal structure close to the MgO grains, whereas the outer precipitate layer was formed by Zn-rich platy crystals. As observed on the macroscale in the depth profiles of water samples, on the microscale Mn only precipitated in regions of very high pH close to the alkaline MgO grains, whereas Zn precipitated further away. Cortina et al. (5) identified manganite (γMnOOH), feitknechtite (β-MnOOH) and hausmannite (Mn3O4) in their MgO columns treating water with a much higher Mn concentration. They concluded that these Mn(III) compounds precipitate because the high pH in the MgO columns greatly accelerates oxidation of Mn(II) to Mn(III) (12, 13). Probably a similar mechanism is responsible for Mn removal in the MgO-wood chip columns. From 6 cm depth to the bottom, the column contained loose wood chips covered with brownish dust, apparently unreacted MgO. These samples had accumulated only small quantities of Zn and Mn. XRD indicated the presence of the phases found in the initial substrate except gypsum, although due to hydration brucite peaks were higher and periclase peaks were lower. This confirms that in the fine-grained column metal elimination was essentially complete above 7 cm depth, and MgO reactivity remained unaltered below. In the Zn- and Mn-rich samples up to 6 cm depth, about 95% of the initial MgO had dissolved. These results largely improve those obtained by Cortina et al. and Ro¨tting et al. (5, 10) who had used caustic magnesia with grain size 2-4 mm either as pure material or mixed with silica sand. In their experiments only 10-40% (w/w) of caustic magnesia had been consumed before the columns clogged or lost reactivity. The coarse-grained column contained (from the top) about 1 cm of loose wood chips covered with white gelatinous precipitates and about 2 cm of wood chips with little

precipitates. Below, the substrate showed heterogeneous zones of either cemented wood chips covered with dark gray precipitates and white spots, or of loose wood chips with MgO grains but little precipitates. SEM-EDS images (see SI Figure S3) show large MgO grains covered by a thin layer of Zn-rich precipitates with traces of Mn. Probably the precipitates slow down MgO dissolution critically when only a small portion of the coarse grains (average 22%) has reacted. This shows that the coarse-grained MgO mixture uses the alkaline reagent much less efficiently than the fine-grained MgO mixture. Reactive Transport Modeling. The fine-grained MgO column was simulated using the code RETRASO to test the mechanisms of Zn and Mn removal discussed above, and to explore quantitatively the behavior of the system under different conditions. The use of a reactive transport model is required due to the large number of reactions and the nonlinearity of coupled transport and reaction processes (14). Thermodynamic data for aqueous, mineral, and gas phases were taken from the WATEQ4F database (15), and data for hydrozincite was added according to Schindler et al. (16). A kinetic rate law for the dissolution of brucite was derived from data of various authors summarized by Pokrovsky and Schott (17). The dissolution rates shown in Figure 8 of their publication were approximated by the law r ) 10-4·aH+0.45mol · m-2·s-1

(2)

The amount m of mineral dissolved (mol/s) was calculated as m ) A r(1 - Ω)

(3)

where A is the reactive surface area (m2/m3 column) and Ω is the saturation of the solution. Similarly, the rate of calcite dissolution was simulated according to the kinetic rate law of Chou et al. (18). The value of A for both minerals will be discussed below. The autocatalytic oxidation of Mn(II) to Mn(III) was modeled after the kinetic law of Morgan (19), as cited in Diem and Stumm (20): -d[Mn(II)] ⁄ dt ) k′1[O2(aq)][OH-]2[Mn(II)] + k′2[O2(aq)][OH-]2[MnOx][Mn(II)] (4) where [Mn(II)], [O2(aq)] and [OH-] are dissolved concentrations (mol/L) of Mn(II), oxygen and OH-, respectively, [MnOx] is the concentration of Mn(III) reaction solid product (mol/L), k′1 is the rate constant for homogeneous oxidation (4 × 1012 VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9375

mol3 · L-3 · day-1) and k′2 is the rate constant for heterogeneous oxidation (1018 mol4 · L-4 · day-1). Hydrozincite, β-Zn(OH)2 and Manganite were set to precipitate in equilibrium. The input solution was set to be in equilibrium with atmospheric O2. The values of the initial reactive surface areas of dissolving MgO and calcite (A0) were calibrated to match the depth profiles of Mg and Ca concentrations measured during the first month of functioning of the column. Then, the surface area at any time (A) was updated using the volume of the residual mineral (V) following a 2/3 law: A ) A0(V ⁄ V0)2⁄3

(5)

The rest of properties of the pore water (concentrations, pH, pE) and the dissolved and precipitated solids calculated with the model match very well with observations at different points of the column (Figure 6). After one year of simulated operation, MgO has dissolved completely in the upper 3 cm of the substrate, whereas no MgO dissolution has occurred below 6 cm depth. Dissolved Mn(II) decreases at the MgO dissolution front, where the pH is high enough for fast oxidation to Mn(III). Mn(III) precipitates immediately as Manganite. The excellent agreement of the measured Eh with the pE value calculated from the Mn(II)/Mn(III) equilibrium confirms the hypothesis of a Mn(III) phase for the precipitated solid. Zn precipitates within the entire zone of MgO dissolution as β-Zn(OH)2, slightly delayed with respect to Manganite. Near the column inlet, a small amount of Zn precipitates as hydrozincite, redissolving previously precipitated β-Zn(OH)2. This pair of minerals buffers pH near 7. As the MgO dissolution front moves downward, Manganite forms a band within the column because it is virtually insoluble at alkaline or circumneutral pH (21). This explains why Mn and Zn are found together in the solid samples, even though they precipitate at different depths. The model was used to extrapolate the longevity of the fine-grained MgO-column. For 15 cm of substrate and a Darcy flowrate of 0.1 m/day as used in this experiment, the column would have been able to operate during almost three years (33 months) before Zn- or Mn-breakthrough. Accordingly, a column of 1 m substrate thickness would have been effective for almost 19 years. However, this will have to be corroborated in field trials. The model was also used to predict the maximum flowrate for the fine-grained column. For 15 cm of substrate, the column could have operated during one year at a Darcy flowrate of up to 0.3 m/day, three times the flow used in this experiment. In order to achieve higher flowrates or a longer durability, substrate thickness has to be increased. The model also explains why the coarse-grained column did not work satisfactorily. If MgO reactive surface is lowered by factor 50 with respect to the value in the model of the fine-grained model, MgO dissolution is too slow to reach equilibrium within the column length (not shown). Zn is then only partially eliminated, and outflow pH is too low for significant Mn removal. This confirms that the small size and large specific surface area of the MgO particles are the key factor of the much better performance of the fine-grained column.

Acknowledgments We thank Eva Peleri (ICP-AES), Antonio Padro (ICP-MS) and the SEM-EDS team from SCT-UB as well as Josep Elvira (XRD) and Jesus Parga (preparation of thin sections) from IJA-CSIC for technical assistance. This study was funded by the Spanish 9376

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008

Government projects REN-2003-09590-C04-02, CTM200628151-E/TECNO and CTM2007-66724-C02-01.

Supporting Information Available Figure S1: Schematic experimental setup; Figure S2: Photos of the two columns after 4.5 and 12 months of operation; Figure S3: SEM-EDS images of a solid sample from the coarsegrained MgO column. Table S1: Grain size distribution, calculated specific surface and chemical composition of the caustic magnesia samples used in the fine- and coarsegrained MgO column; Table S2: Mean major composition of input water. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Younger, P. L.; Banwart, S. A.; Hedin, R. S., Mine Water Hydrology, Pollution, Remediation; Kluwer Academic Publishers: Dordrecht, 2002; p 442. (2) Skousen, J. G.; Politan, K.; Hilton, T.; Meek, A., Acid Mine Drainage Treatment Systems: Chemicals and Costs. In Acid Mine Drainage: Control and Treatment, Skousen, J. G.; Ziemkiewicz, P. F., Eds. West Virginia University and the National Mine Land Reclamation Center: Morgantown, WV, 1995; pp 121-129. (3) Johnson, D. B.; Hallberg, K. B. Acid mine drainage remediation options: a review. Sci. Total Environ. 2005, 338 (1-2), 3–14. (4) Sheoran, A. S.; Sheoran, V. Heavy metal removal mechanism of acid mine drainage in wetlands: A critical review. Miner. Eng. 2006, 19 (2), 105–116. (5) Cortina, J. L.; Lagreca, I.; De Pablo, J.; Cama, J.; Ayora, C. Passive in situ remediation of metal-polluted water with caustic magnesia: Evidence from column experiments. Environ. Sci. Technol. 2003, 37 (9), 1971–1977. (6) Utgikar, V. P.; Tabak, H. H.; Haines, J. R.; Govind, R. Quantification of toxic and inhibitory impact of copper and zinc on mixed cultures of sulfate-reducing bacteria. Biotechnol. Bioeng. 2003, 82 (3), 306–312. (7) Cabrera, G.; Perez, R.; Gomez, J. M.; Abalos, A.; Cantero, D. Toxic effects of dissolved heavy metals on Desulfovibrio vulgaris and Desulfovibrio sp strains. J. Hazard. Mater. 2006, 135 (1-3), 40–46. (8) Gibert, O.; de Pablo, J.; Cortina, J. L.; Ayora, C. Sorption studies of Zn(II) and Cu(II) onto vegetal compost used on reactive mixtures for in situ treatment of acid mine drainage. Water Res. 2005, 39 (13), 2827–2838. (9) Kogel, J. E.; Trivedi, N. C.; Barker, J. M.; Krukowski, S. T., Industrial Minerals & Rocks: Commodities, Markets, and Uses. 7th ed.; Society for Mining, Metallurgy, and Exploration: Littleton, CO, 2006; p 1548. (10) Ro¨tting, T. S.; Cama, J.; Ayora, C.; Cortina, J. L.; De Pablo, J. Use of caustic magnesia to remove cadmium, nickel, and cobalt from water in passive treatment systems: Column experiments. Environ. Sci. Technol. 2006, 40 (20), 6438–6443. (11) Saaltink, M. W.; Battle, F.; Ayora, C.; Carrera, J.; Olivella, S. RETRASO, a code for modeling reactive transport in saturated and unsaturated porous media. Geol. Acta 2004, 2 (3), 235–251. (12) Kessick, M. A.; Morgan, J. J. Mechanism of Autoxidation of Manganese in Aqueous-Solution. Environ. Sci. Technol. 1975, 9 (2), 157–159. (13) Morgan, J. J. Kinetics of reaction between O2 and Mn(II) species in aqueous solutions. Geochimi. Cosmochim. Acta 2005, 69 (1), 35– 48. (14) Rezaei, M.; Sanz, E.; Raeisi, E.; Ayora, C.; Vazquez-Sune, E.; Carrera, J. Reactive transport modeling of calcite dissolution in the fresh-salt water mixing zone. J. Hydrol. 2005, 311 (1-4), 282–298. (15) Ball, J. W.; Nordstrom, D. K. User’s Manual for WATEQ4F with Revised Thermodynamic Database and Test Cases for Calculating Speciation of Major, Trace and Redox Elements in Natural Waters.; U.S. Geological Survey Water-Resources Investigation Report, 1991; 91-183. (16) Schindler, P.; Reinert, M.; Gamsja¨ger, H. Thermodynamics of metal carbonates 3. Solubility constants and free formation enthalpies of ZnCO3 and Zn5(OH) 6(CO3) 2 at 25 degrees. Helv. Chim. Acta 1969, 52 (8), 2327–2332. (17) Pokrovsky, O. S.; Schott, J. Experimental study of brucite dissolution and precipitation in aqueous solutions: Surface speciation and chemical affinity control. Geochim. Cosmochim. Acta 2004, 68 (1), 31–45.

(18) Chou, L.; Garrels, R. M.; Wollast, R. Comparative study of the kinetics and mechanisms of dissolution of carbonate minerals. Chem. Geol. 1989, 78, 269–282. (19) Morgan, J. J. Chemical Equilibria and Kinetic Properties of Manganese in in Natural Waters. In Principles and Applications of Water Chemistry; Faust, S. D.; Hunter, J. V., Eds.; Wiley: New York. 1967; pp 561-624.

(20) Diem, D.; Stumm, W. Is Dissolved Mn2+ being oxidized by O2 in absence of Mn-bacteria or surface catalysts? Geochim. Cosmochim. Acta 1984, 48 (7), 1571–1573. (21) Eary, L. E. Geochemical and equilibrium trends in mine pit lakes. Appl. Geochem. 1999, 14 (8), 963–987.

ES801761A

VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9377