Environ. Sci. Technol. 2005, 39, 8032-8037
Arsenic Removal from Bangladesh Tube Well Water with Filter Columns Containing Zerovalent Iron Filings and Sand OLIVIER X. LEUPIN,† S T E P H A N J . H U G , * ,† A N D A. B. M. BADRUZZAMAN‡ Swiss Federal Institute of Aquatic Science and Technology (Eawag), U ¨ berlandstrasse 133, CH-8600 Du ¨ bendorf, Switzerland, and Bangladesh University of Engineering and Technology, Dhaka, Bangladesh
Arsenic removal is often challenging due to high As(III), phosphate, and silicate concentrations and low natural iron concentrations. Application of zerovalent iron is promising, as metallic iron is widely available. However, removal mechanisms remained unclear and currently used removal units with iron have not been tested systematically, partly due to their large size and long operation time. This study investigated smaller filter columns with 3-4 filters, each containing 2.5 g of iron filings and 100-150 g of sand. At a flow rate of 1 L/h, these columns were able to treat 75-90 L of well water with 440 µg/L As, 1.8 mg/L P, 4.7 mg/L Fe, 19 mg/L Si, and 6 mg/L dissolved organic carbon (DOC) to below 50 µg/L As(tot), without addition of an oxidant. As(III) was oxidized in parallel to oxidation of corrosion-released Fe(II) by dissolved oxygen and sorbed on the forming hydrous ferric oxides (HFO). The open filter columns prevented anoxic conditions. DOC did not appear to interfere with arsenic removal. Manganese was reduced after a slight initial increase from 0.3 mg/L to below 0.1 mg/L. About 100 mg of Fe(0)/L of water was required, 3-5 times less than that for larger units with sand and iron turnings.
Introduction In Bangladesh, around 40 million people consume drinking water with arsenic concentrations exceeding the guideline values of the WHO (10 µg/L) and of Bangladesh (50 µg/L) (1). From the 9-11 million tube wells installed, over a fourth deliver water contaminated with arsenic (2). Chronic ingestion of arsenic causes skin lesions, melanosis, skin cancer, and internal cancers (3). Vietnam (4) and several other Asian countries are also heavily affected, and arsenic from various sources contaminates drinking water in many regions worldwide (5). Several methods are available to remove arsenic from drinking water, among them low-cost methods that have been applied and tested in West Bengal, with mixed results (6). Most of these tested methods used filters with commercially produced sorbents. No prefabricated sorbents are required with the application of chemical oxidants and alum * Corresponding author phone: +41-44-823-5454; fax: +41-44823-5210; e-mail:
[email protected]. † Swiss Federal Institute of Aquatic Science and Technology. ‡ Bangladesh University of Engineering and Technology. 8032
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005
or iron salts to induce precipitation (7-9). Natural iron concentrations in Bangladesh are generally too low for passive removal or simple sand filtration, such that additional aluminum or iron is necessary to remove sufficient arsenic and necessarily also the competing and more strongly adsorbing phosphate (10). No additional chemicals and the minimum mass of man-made material are required if metallic iron can be used for As(III) oxidation and formation of arsenicsorbing HFO. Despite quite numerous arsenic removal studies with Fe(0), the removal mechanisms remain unclear. One group of studies reported arsenic removal with metallic iron under anoxic conditions, with sorption of arsenic in unspecified form on the iron surface (11, 12). Bang et al. (13) found that As(III) can be reduced to As(0) on the surface of Fe(0) and that As(III) is more rapidly removed than As(V) under anoxic conditions. Under oxic conditions, As(V) was removed faster due to stronger sorption on HFO formed at the Fe(0) surface (13). Karschunke et al. (14) treated up to 1.19 m3 of Berlin tap water spiked with 500 µL-1 As(V) with 45.8 g of iron wool. Recently, it has been shown that As(III) is partly oxidized in Fe(II)-containing water during the oxidation of Fe(II) with dissolved O2 (10, 15). A very recent study shows that Fe(II) released from corroding iron filings in aerated water also leads to As(III) oxidation and to sorption of As(V) on the forming HFO (16). Munir et al. (17), Khan et al. (18), and Sutherland et al. (8) investigated locally built 3-Kolshi filters consisting of a first container with 3 kg of iron turnings and 2 kg of sand, a second container with 2 kg of sand and 1 kg of charcoal, and a third container serving as the recipient. 3-Kolshi units have treated up to 6 m3 of arsenic contaminated water with low phosphate concentrations over a period of 9 months to less than 10 µg/L arsenic (17). However, to our knowledge, these units have not been tested until breakthrough of the arsenic; thus, their capacity under various conditions remains unknown. The large size and long operation time of several months make systematic studies difficult. Problems with clogging or with pathogens can develop before breakthrough of the arsenic. Due to restricted flow in parts of the filters, anoxic zones are likely to develop around the iron turnings, such that the contribution of oxic and anoxic pathways for arsenic removal is not known. In a detailed laboratory study (16), arsenic-contaminated water was repeatedly filtered through a glass column containing sand and iron filings. After four filtrations, synthetic groundwater with initially 500 µg/L As(III) met the guideline value of 50 µg/L. In a first attempt at upscaling, a filter column made of four filters with 2.5 g of iron filings and 100-150 g of sand, followed by a fifth filter with sand, treated 36 L of synthetic groundwater with 500 µg/L As(III) and 2-3 mg/L P to below the limit of 50 µg As(tot)/L (16). To study the impact of natural water components (e.g., dissolved organic carbon (DOC), natural ferrous iron, intially anoxic conditions), removal experiments were conducted on a pilot scale in Bangladesh. The aim was to test arsenic removal from tube well water with well-aerated filter columns constructed from locally available materials, and to understand the relevant processes for the future optimization, upscaling and application of filters with iron filings and sand.
Materials and Methods Materials. All chemicals were reagent grade from Merck, Aldrich, or Fluka and were used as received. Iron Filings. Iron filings were freshly produced with a “bastard” hand file (Baiter AG, Switzerland) from cast iron 10.1021/es050205d CCC: $30.25
2005 American Chemical Society Published on Web 09/09/2005
TABLE 1. Composition of Synthetic Groundwater and of Tube Well Water and the Average Composition of Tube Well Water in Bangladesha
pHinitial HCO3- (mM) Ca (mM) Mg (mM) Si (mg/L) P (mg/L) As (µg/L) Fe (mg/L) Mn (mg/L) DOC (mg/L)
synthetic groundwater
water from tube well 2b
Bangladeshi groundwater (BGS)c
7.0 ( 0.05 8 2.5 1.6 20 2.0 and 3.0 500 0 0 0
7.1 ( 0.1 not measured 3.09 ( 0.2 1.52 ( 0.1 19.1 ( 0.2 1. 9 ( 0.1 441 ( 20 4.7 ( 0.1 0.30 ( 0.02 6.1 ( 0.5
7.0 ( 0.2 7.8 ( 2.7 1.9 ( 1.4 1.3 ( 0.8 19.2 ( 4.7 1.47 ( 1.48 199 ( 166 5.3 ( 4.8 0.57 ( 0.75 3.3 ( 2.8
a As obtained from the BGS database (pH, DOC, and HCO - are from 3 the special study areas, and all other values are from the national survey b data; only wells with As > 50 µg/L were used). Standard deviations of 2-5 measurements of tube well 2 are given. c Standard deviations of the tube wells in the BGS database are given.
water pipe (provided by BUET) that is typically used for the construction of tube wells. The filings consisted of 50-300 µm sized, irregularly shaped, and strongly folded iron flakes (16). For the laboratory experiments, St-37 steel was used [C e 0.17%, Si e 0.30%, Mn e 1.4%, P e 0.045%, S e 0.045%, N e 0.009% (19)]. Sand. In Bangladesh, locally available construction sand was washed several times to remove the silt and fine sand fractions, and then sterilized by boiling for 10 min. The sand had a bulk density of 1.5 g/cm3 and a measured porosity of about 40%. Merck laboratory quartz sand of analytical purity with a grain size of 0.2-0.8 mm was used as received for the laboratory experiments. Groundwater Composition. Water was collected from a tube well built in 1996 of 150 feet depth (well 2) in the village of Srinagar in the district of Munshiganj, 30 km south of Dhaka. Laboratory experiments were performed with synthetic groundwater with the main characteristics of Bangladeshi groundwater, except for the absence of iron and DOC, prepared as described previously (10, 16) (Table 1). Filtration Unit. To construct filter columns, 1.5-L poly(ethylene terephthalate) (PET) bottles were truncated and used upside down, with the sand and iron filings filled into the bottlenecks, as shown in Figure 1. The water flowed from one bottle to the next through approximately 2 mm diameter holes (covered by a piece of cloth to avoid loss of sand) in the caps. Two different arrangements of sand and iron were used, as also shown in Figure 1. The columns were suspended from a rope coiled around a bamboo rod and connected with a silicon tube to a polyethylene (PE) water container placed such that the water level in the top bottle and the container was equal. Due to the much larger cross section of the raw water container, the water levels in the filters changed only very slowly and were occasionally adjusted by refilling the tank or by lowering the filter column by rotating the bamboo rod to release more rope. Sampling. Samples (5-20 mL) were collected directly at the outflow of each filter and from the recipient and were acidified with 20 µL of 5 M HCl/mL of sample. As(tot) Analysis. Analysis was performed at EAWAG with an atomic fluorescence spectrometer (AFS) (PS Analytical Ltd., Kent, U.K.). Disodium citrate buffer solution (0.5 M, pH 5.0) and 0.7% NaBH4 was used for the hydride generation. To measure As(tot), 100 µL of sample, 100 µL of 10 M HCl, and 20 µl of reducing agent (0.6 M ascorbic acid and 3 M KI) were left to react for 60 min prior to 1:50 dilution with citrate buffer.
ICP-OES. An inductively coupled plasma optical emission spectrometer (Spectro Ciros CCD) was used to analyze P, Mg, Ca, Si, Fe, and Mn. Fe(II) Measurement. Fe(II) was sampled by directly collecting 5 mL of column effluent into vials containing 2 mL of phenanthroline and 1 mL of acetate buffer and quantified by UV-vis spectrophotometry (512 nm ) 11 000 M-1 cm-1). XRD. Powder X-ray diffractograms were measured with a Phillips Scintag XDS 2000 (Cu cathode 45 kV and 40 mA) from 2° to 60° 2θ. For the qualitative measurements, the samples were air-dried, homogenized, and mixed with an internal standard (Si). To analyze the coating on the sand and the scalings on the iron filings, samples were treated for 10 min with ultrasound. Small amounts of the suspended material were dried with a nitrogen stream on a glass disk and placed in the sample holder for XRD analysis. DOC and TOC Measurements. Samples (30 mL) were filled into vials containing 200 µL of 5 M HCl during the field campaign and stored at 5-10 °C in a refrigerator until analysis with a Shimadzu 5000 TOC analyzer.
Results and Discussion Collection, Transport, and Filtration of Well Water. After the water was pumped into 20-L PE tanks, oxidation of the natural Fe(II) started and the water turned brown within several hours. Filling the tanks headspace -free slowed the oxidation, but partial aeration took place during pumping. We made no attempts to avoid aeration, as water treatment in villages would involve the same steps. The water was transported to Dhaka within 1 h and each tank was used up within 24 h. During this time, the water turned brown and some precipitation took place, but most of the HFO remained in suspension. The flow rate of the columns was controlled by the water level in the first filter. The water levels in the next filters adjusted accordingly, until all filters reached an equal steady-state flow rate of 0.5-1.5 L/h, with an average of around 1L/h. The sand-iron filing layers in each filter remained immersed in water during the entire filtration, with the height of the water column above the sand varying between 2 and 20 cm. Four Filters Containing Sand Only (Column C1). The efficiency of sequential sand filtration was investigated with a filter column made of four filters containing sand without iron filings. The arsenic removal was clearly insufficient (Figure 2 and Figure S1, Supporting Information), with a reduction of 37% after the first filter (mainly by retention of the precipitates from naturally formed HFO) and of less than 5% in the following filters. In the recipient, an average removal of 43% was achieved. Compared to other studies with similar iron concentrations (4-6 mg/L), this is higher than the value expected from Roberts et al. (10) in passive removal experiments (ca. 15% removal) but lower than the arsenic removal achieved in Vietnam in simple sand filters (between 60% and 85% removal) (20). In contrast to arsenic, phosphate was strongly reduced (by 81%) after the first filter and by an average of 90% in the recipient. Analogous to the arsenic, most of the phosphate is removed by retention of HFO in the first filter, but phosphate outcompetes As(III) and As(V) due its much higher initial concentration. The relatively low iron and high phosphate and silicate concentrations make simple sand filtration an insufficient treatment option in Bangladesh for water of typical composition. Four Filters with Homogeneously Mixed Iron Filings and Sand (Column C2). C2 consisted of four filters containing 2.5 g of iron filings homogeneously mixed with the sand. To provide colorless HFO-free water, a fifth filter containing sand was added to remove the Fe(II) formed from corrosion in the last filter with Fe(0). The arsenic concentrations measured after filter 1 and to a lesser extent after filter 2 (Figure 2) show an initial decline and a minimum after 30 VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8033
FIGURE 1. Filter columns made of PET bottles were suspended from a bamboo rack. A 25 L water tank with raw water was connected to the first filter by 3 mm diameter silicon tubing. The flow rate was regulated by the water level in the first filter, which was adjusted by lifting or lowering the filter column. The treated water was collected in 5-10-L recipients. Two different filter designs were investigated: The sand and iron filings were either homogeneously mixed in the whole filter (top) or layered with a homogeneous mixture of sand with iron filings on a layer followed by a layer of only sand (bottom).
FIGURE 2. Arsenic concentrations during the filtration of well water in the filter effluents and the recipient of C2 (four filters with homogeneous mixtures of iron filings and sand and of one filter with sand). Also shown are the concentrations in the recipient of C1 (four filters with only sand, `). L, before increasing again. As shown in Figure 3, initially up to 3 mg/L iron left the first filter and 2.5, 1.5, and 0.5 mg/L iron leached from filters 2, 3, and 4, respectively. The water flowing out of each filter was colorless and turned visibly brown within 5-30 min in the next filter. Analogous experiments in the laboratory showed that the iron leaving the filters was predominantly (
