Environ. Sci. Technol. 2005, 39, 8045-8051
Sorption and Desorption of Arsenic to Ferrihydrite in a Sand Filter S O R E N J E S S E N , † F L E M M I N G L A R S E N , * ,† CHRISTIAN BENDER KOCH,‡ AND ERIK ARVIN† Institute of Environment & Resources DTU, Technical University of Denmark, DK-2800 Kgs. Lyngby, and Department of Natural Sciences, KVL, The Royal Veterinary and Agricultural University, Thorvaldsensvej 40, DK-1871 Frederiksberg C, Denmark
Elevated arsenic concentrations in drinking water occur in many places around the world. Arsenic is deleterious to humans, and consequently, As water treatment techniques are sought. To optimize arsenic removal, sorption and desorption processes were studied at a drinking water treatment plant with aeration and sand filtration of ferrous iron rich groundwater at Elmevej Water Works, Fensmark, Denmark. Filter sand and pore water were sampled along depth profiles in the filters. The sand was coated with a 100-300 µm thick layer of porous Si-Ca-As-contaning iron oxide (As/Fe ≈ 0.17) with locally some manganese oxide. The iron oxide was identified as a Si-stabilized abiotically formed two-line ferrihydrite with a magnetic hyperfine field of 45.8 T at 5 K. The raw water has an As concentration of 25 µg/L, predominantly as As(III). As the water passes through the filters, As(III) is oxidized to As(V) and the total concentrations drop asymptotically to a ∼15 µg/L equilibrium concentration. Mn is released to the pore water, indicating the existence of reactive manganese oxides within the oxide coating, which probably play a role for the rapid As(III) oxidation. The As removal in the sand filters appears controlled by sorption equilibrium onto the ferrihydrite. By addition of ferrous chloride (3.65 mg of Fe(II)/L) to the water stream between two serially connected filters, a 3 µg/L As concentration is created in the water that infiltrates into the second sand filter. However, as water flow is reestablished through the second filter, As desorbs from the ferrihydrite and increases until the 15 µg/L equilibrium concentration. Sequential chemical extractions and geometrical estimates of the fraction of surface-associated As suggest that up to 40% of the total As can be remobilized in response to changes in the water chemistry in the sand filter.
Introduction Arsenic in drinking water poses a serious threat to the health of millions of people. A decade ago, it was estimated that high mortality risks from cancer diseases may develop due to an intake of drinking water containing As in concentrations of 50 µg/L over a long period (1). More recent studies have pointed to the possibility of poisonous effects to humans at * Corresponding author phone: 45 45 25 21 69; fax: 45 45 88 59 35; e-mail:
[email protected]. † Technical University of Denmark. ‡ KVL The Royal Veterinary and Agricultural University. 10.1021/es050692x CCC: $30.25 Published on Web 09/13/2005
2005 American Chemical Society
lower concentrations (2-5). In many groundwater resources, arsenic occurs in elevated concentrations, and concentrations above the recommended 10 µg/L WHO drinking water guideline (6) have been reported in many countries (7). Among a suite of As water treatment techniques proposed in the literature, techniques involving the precipitation of iron and immobilization of the arsenic by sorption onto the Fe(III) precipitates have received much attention. During aeration and filtration, which are key treatment processes in many centralized water treatment facilities, part of the Fe(III) precipitate adheres to or nucleates on the surfaces of filter sand grains. Thus, the filter sand slowly is transformed into grains consisting of a center of the original sand grains (dominantly quartz) now covered with voluminous Fe(III)dominated precipitates. The precipitate constitutes a significant part of the solids in the sand filter at any time and contributes both a high specific surface area of high reactivity and a large microporosity. Poorly ordered ferrihydrite has been observed to constitute a substantial part of the precipitates formed in such treatment processes (8, 9) and in other places where groundwater containing ferrous iron is aerated (10, 11). In an early stage of the precipitation of pure ferrihydrite, the precipitate consists of nanometer-sized ferrihydrite crystallites. The ferrihydrite crystals subsequently grow as the crystallites aggregate by Van der Waal forces and as Fe-O-Fe bonds form between the iron octahedra of the aggregated crystallites (12). During ferrihydrite precipitation, solutes in the water that sorb strongly to ferrihydrite surfaces (e.g., SiO2) can decrease further the Fe-O-Fe polymerization by blocking up the ferrihydrite crystallite surfaces (10), thereby maintaining a high specific surface area and reactivity of the precipitate. Both As(III) and As(V) form inner-sphere surface complexes on ferrihydrite (12-14). As(V) sorption decreases gradually with increasing pH from 6 to 9 (14-16). In many studies it has been observed that, in the presence of competitive solutes, As(III) sorption increases with increasing pH from 4 to 8 (16, 17). Although many have studied arsenic removal by sorption to iron oxides for water treatment purposes, only few have elucidated arsenic desorption, i.e., the reversibility of the sorption reaction. Fuller et al. (18) reported complete reversibility of As(V) coprecipitated and adsorbed to preformed ferrihydrite. Arsenic bound to the precipitates in the sand filter is presumably in the +V oxidation state. In natural water, As(V) sorbs more efficiently to iron oxides than As(III) (see, e.g., ref 19) and the presence of manganese oxides may oxidize As(III) possibly preadsorbed to iron oxides (20).
Materials and Methods Field Studies. Filter sand and water samples were collected from Elmevej Water Works, Fensmark, Denmark, during April to June 2004. At the water works, the raw water is aerated during a 3 m fall from a tray into the reaction basin/ clarification pool, from where the water flows through three parallel open sand filters (filters F1, F2, and F3). The filter sand was replaced in 1980. From the top, the filter sand consists of 60-70 cm of quartz sand (grain size 1-3 mm), followed by a 10-20 cm foundation layer of coarser gravel. The sand filters are backwashed every fourth week to maintain a high permeability. During backwash, the sand filter is purged with air added through tubes installed in the filter bottom, followed by water from the clean water tank, which is led backward through the filters and into an exterior sedimentation basin. VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8045
Elmevej Water Works abstracts a Ca-HCO3 type of groundwater with naturally derived arsenic concentrations up to 30 µg/L, extracted from the fractured upper 10-20 m of a limestone overlaid by 20-60 m of Quaternary deposits of fluvial sand and clay tills. Until the year 1998, the Fe, Mn, and As contents of the source groundwater were 2 mg/L, 50-100 µg/L, and 10-20 µg/L, respectively. After 1998, a different source of the groundwater was used, containing iron in a concentration of 0.2 mg/L and manganese below 10 µg/L. In this water, the concentration of arsenic is between 20 and 30 µg/L. At a production rate (Q) of 20 m3/h, corresponding to 70-100% of the normal Q, the residence time (V/Q) in the reaction basin was 50 min. Filter velocities (Darcian velocity) were 1.3 m/h during most sampling, except for one instance where filter velocities were varied between 0.7 and 4.4 m/h. Samples of the filter sand were collected after drainage of the filter. Material representing the upper 1 cm was collected with a plastic bag from a circular area of approximately 50 cm diameter. Samples from deeper parts of the filter were collected using a stainless steel sand bailer (OD ) 8 cm) with a polyethylene (PE) flap. The cored hole in the drained filter sand did not collapse between consecutive sampling depths, and undisturbed samples could therefore be collected. Each sample represents a depth interval of 10 cm. Pore water samples in the sand filter were collected through a 1 m PE tube mounted onto a 1.5 m stainless steel tube (OD/ID ) 10/7 mm) with a 2 cm long filter intake (polypropylene) near one end. The sampling rate was 30-40 mL/min, controlled by a peristaltic pump. Pore water samples were collected from 5-70 cm depths in the filter sand layer by pushing the steel tube progressively deeper into the filter sand. Consecutive samples were taken with 15-25 min intervals. Samples representative for water immediately prior to infiltration into the filter sand (i.e., water that had completed the aeration-reaction basin flow path) were collected 2-4 cm above the filter sand layer from the homogeneous water mass appearing on top of the filters during operation, using the same equipment as for pore water samples. Samples of raw water, filter effluent, and water from the clean water tank were collected from the nearby (100 m) extraction well or from taps at the water works using PE tubing and/or sterilized syringes inserted directly into the water stream. Field analysis for the EC, pH, [O2], and temperature of these samples was conducted in a flow cell mounted with WTW electrodes (196 series for pH and EC, Oxi320 for [O2]). All sampling was conducted between 3 and 6 days after filter backwash. All water samples were filtered through 0.10 µm cellulose nitrate filters (Satorius) into 20 or 50 mL PE vials. Separation of arsenic species and analysis for alkalinity, [Fe(II)], and [H2S] were conducted in the field immediately following filtration. Alkalinity was measured by Gran titration (21). [Fe(II)] and [H2S] were measured using a Dr. Lange LP 2 W spectrophotometer and the Ferrozine (22) and the methylene blue (23) methods, respectively. As(III) and As(V) were separated immediately after filtration by passing samples through an As(V)-retaining cartridge (24). [As(V)] was determined indirectly as the difference between [As(tot)] and [As(III)] in filtered samples. Samples for cation analysis were acidified to below pH 2 by addition of 1% (v/v) 7 M HNO3 (25). All sand and water samples were cooled in the field to 5-10 °C. Within 8 h after sampling, samples for As and cation analysis and filter sand samples were stored at 7 °C and samples for anion analysis at -18 °C. Tests showed that soluble As in nonacidified samples containing Fe was not removed from solution by Fe precipitation. Reagents. Reagents were prepared from Milli-Q water mixed with reagent-grade chemicals from Merck, Baker, or 8046
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005
Sigma. However, for sample acidification, a 7 M HNO3 solution was prepared from Baker Suprapur concentrated HNO3. FeCl2 Addition. A 0.1 M FeCl2 solution was prepared in a 25 L glass bottle with a conical stopper, preflushed with N2(g). The FeCl2 solution was added through a stainless steel tube (ID ) 3 mm) to the water stream between two serially connected filters. To minimize iron precipitation before the water infiltrated the filter sand, the added FeCl2 solution was mixed with the water stream in a bucket kept below the water surface. The rate of addition was 68.7 mL/min, yielding an Fe(II) concentration of 3.65 mg/L (3.45 mg/L was measured). Equilibration with Filter Sand. At the water works, production takes place at night while during the day the filters remain filled with stagnant water. To elucidate the oxygen consumption and As concentration within the filters during shutdowns, pore water samples were collected from 40-60 cm filter depths at the end of a minimum 8 h shutdown period. In addition, two 2 L glass bottles were filled completely with filter sand from F1 and F2 separately. The bottles were capped below the water surface, hindering trapping of air bubbles, and then equilibrated for 24 h at the water works filter hall (8-10 °C). At the end of the 24 h period, the bottles were quickly opened and 60 mL samples were withdrawn using a syringe. Water samples were handled as described above for pore water samples. However, O2 in the samples was measured by flushing a small glass flow cell mounted with a WTW electrode. Laboratory Work. Sequential extractions of the filter sand samples were carried out in a five-step wet chemical procedure. The step 1 and step 2 extractants were 1 M MgCl2 (pH 8) and 1 M NaH2PO4 (pH 5), targeting ionically adsorbed and ligand-exchangeable As, respectively (26). Step 3 consisted of a 1 h extraction with 20 mM HONH3Cl in 25% CH3COOH, targeting manganese oxides (27) and As coprecipitated by them. The step 4 extractant was 1 M HCl, targeting residual manganese oxides, poorly crystalline iron oxyhydroxides and carbonates, and As coprecipitated with these phases (26). Step 5 consisted of extraction with 6 M HCl for 1 h, targeting crystalline iron oxyhydroxides and the As coprecipitated with them. For further details refer to ref 26. Specific Surface Area (SSA). The SSA of the filter sand samples was determined by the N2-BET multipoint method using a Micromeretics GEMINI. Prior to analysis, the filter sand samples were air-dried for 72 h at 25 °C and then dried by flushing with dry N2(g) for an additional 8 h using a Micromeretics FlowPrep 060. Microscopy. For observations with light and electron microscopy, thin sections were prepared after filter sand samples were embedded with epoxy resin. Most observations were done on samples of the filter sand that were collected from 20-30 cm depth. However, also brief observations on samples from 0-1 and 50-60 cm depths were conducted. The composition of the precipitate covering the filter sand quartz grains was determined using the ThermoNoran energy-dispersive X-ray (EDX) detection system on a Philips XL40 scanning electron microscope. Spots for analysis were selected using the BSE detector. Whole-grain morphology was investigated using Pt-coated grains and a Philips XL20. Mo1 ssbauer and Infrared (IR) Spectroscopy and Powder X-ray Diffraction (pXRD). Grains from filter F1 (∼5 grains) from the depth intervals 0-1, 20-30, and 50-60 cm were treated with a Branson sonifier to remove the precipitate from the sand grains. The large grains were removed from the fines by settling in water, and the fines were dried in the air. These powder samples were investigated by pXRD using a Siemens D5000 equipped with Co KR radiation and an incident beam monochromator, IR spectra using a PerkinElmer FT-2000 spectrometer, and Mo¨ssbauer spectra measured between room temperature and 5 K using a conven-
TABLE 1. Water Chemistry at Elmevej Water Works during Production sample ID
EC (µS/cm)
[O2] (mg/L)
pH
[As(tot)] (µg/L)
[As(III)] (µg/L)
[As(V)] (µg/L)
As removal (%)
[Fe(II)] (mg/L)
[Fe(tot)] (mg/L)
raw water infiltrate watera filter effluent clean water tank
638 635 627 610
0.06 7.74 5.66 7.91
7.23 7.60 7.52 7.58
26 25 14 14
21 21 0.0 0.0
5 4 14 14
4 46 46
0.16 0.11 0.00 0.00
0.18 0.12 0.01 0.00
a
Infiltrate water corresponds to water sampled at 0 cm filter depth, i.e., water that has completed the aeration-reaction basin flow path.
tional constant-acceleration spectrometer and cryostats. The velocity of the Mo¨ssbauer spectrometer was calibrated using a 12.5 µm foil of natural Fe at room temperature. Desorption Kinetics. Desorption kinetics at 10 °C were investigated in a cooled room and with treated groundwater obtained from Haslev Water Works, Haslev, Denmark. This water (HW) had a composition very similar to the groundwater extracted at Elmevej Water Works, but contained arsenic below 1 µg/L. A 100 g sample of air-dried (72 h, 25 °C) filter sand was added to an aliquot of HW in a 2 L acidwashed blue-cap bottle, producing a final solid-to-liquid (S/ L) ratio of 77.6 g/L. The suspension was then agitated by back-and-forth strokes at a rate of 120 min-1. At agitation rates not faster than this, the sand in the bottle resided motionless at the bottom and the water did not slop around. Sample aliquots of 25 mL were collected using a sterile syringe with a 30 cm Teflon tube on the tip. Samples were immediately filtered through 0.20 µm (Satorius) filters. To flush the filters, the first 10 mL was returned to the blue-cap bottle. The remaining 15 mL was collected in 20 mL PE vials and stored at 7 °C for later As analysis. Analysis. [As] was determined by hydride generation and atomic absorbance spectroscopy (AAS) at 194.2 nm in a flow injection system (FIHG-AAS; see, e.g., ref 28), using a PerkinElmer 5000 with a deuterium background corrector. The apparatus detection limit was 1.5 µg/L. After sample acidification by 2 M HCl, As(V) was reduced by addition of 3% (v/v) NaBH4 in 0.1% (v/v) NaOH. The released arsine was carried by N2(g) to the atomizer, consisting of a quartz tube heated to 1000 °C. [Ca], [Mg], [Na], [K], and [Fe(tot)] were determined by flame AAS using a Perkin-Elmer A-Analyst 200. LaCl3 [3% (v/v)] was added to Ca and Mg samples before measurement. [Mn] was determined by GF-AAS on a Perkin-Elmer 800 with autosampler AS 800. [NH4] and [NO3] were determined on a Tecator FIAstar 5010, using the Tecator ASN 50-01/84 and the ASN 62-02/83 methods, respectively. [SO4] and [Cl] were determined on an HPLC instrument with a VYDAC column. [Si] was determined on a spectrophotomer using a molybdosilicate method (modification after 4500-Si D; 29).
Results Changes in Water Chemistry during Treatment. Elmevej Water Works receives reduced groundwater containing Fe(II), H2S, and NH4 (0.16 mg/L, 77 µg/L, and 0.84 mg/L, respectively) and with a pH of 7.23 (Table 1). The concentrations of the major cations, Ca, Mg, Na, and K, are 60, 30, 24, and 3.5 mg/L, respectively. The Si concentration is 9.3 mg/L. The major anions, SO4, Cl, and HCO3, are present in the concentrations 8, 18, and 360 mg/L, respectively. The concentrations of the major ions remain unaffected during the water treatment. Following the aeration-reaction basin flow path, [O2] increases to 7.74 mg/L while Fe(tot) and Fe(II) concentrations decrease to 0.12 and 0.11 mg/L, respectively. The total As concentration of the raw water is 26 µg/L, with 21 µg/L As(III) and 5 µg/L As(V). Arsenic in the clean water is 100% As(V). During the water treatment, less than 1% of the As(III) forms complexes with H2S, as calculated by PHREEQC-2 (30)
FIGURE 1. Arsenic (a), iron (b), and manganese (c) concentrations in the pore water of a sand filter (filter F1) versus depth. Filled triangles indicate the total concentration of As, Fe, or Mn, and open circles indicate As(III) or Fe(II) concentration. Open squares indicate As(V) concentration. using equilibrium constants from Wilkin et al. (31). Similar calculations for possible complexation by carbonate (32, 33) have not been performed due to the lack of equilibrium constants from the literature. Preliminary investigations show that the PO43- concentration in all samples, including samples of pore water, is low (
