Environ. Sci. Technol. 2004, 38, 2919-2927
Preservation of As(III) and As(V) in Drinking Water Supply Samples from Across the United States Using EDTA and Acetic Acid as a Means of Minimizing Iron-Arsenic Coprecipitation PATRICIA A. GALLAGHER, CAROL A. SCHWEGEL, AMY PARKS, BRYAN M. GAMBLE, LARRY WYMER, AND JOHN T. CREED* U.S. EPA NERL Microbiological and Chemical Exposure Assessment Research Division, Cincinnati, Ohio 45268
Seven different treatment/storage conditions were investigated for the preservation of the native As(III)/ As(V) found in 10 drinking water supplies from across the United States. These 10 waters were chosen because they have different As(III)/As(V) distributions; six of these waters contained enough iron to produce an iron precipitate during shipment. The waters were treated and stored under specific conditions and analyzed periodically over a span of approximately 75 days. Linear least squares (LLS) was used to estimate the change in As(III) and As(V) over the study period. Point estimates for the first and last analyses days and 95% confidence bounds were calculated from the LLS. The difference in the point estimates for the first and last day were then evaluated with respect to drinking water treatment decision making. Three primary treatments were evaluated: EDTA/AcOHtreatment and AcOH treatment as well as no treatment. The effect of temperature was explored for all treatments, while the effect of aeration was evaluated for only the EDTA/ AcOH treated samples. The nontreated samples experienced a 0-40% reduction in the native arsenic concentration due to the formation of Fe/As precipitates. The Fe/As precipitates were resolubilized and shown to contain elevated concentrations of As(V) relative to the native distribution. Once this Fe/As precipitate was removed from solution using a 0.45 and 0.2 µm filter, the resulting arsenic concentration (As(III) + As(V)) was relatively constant (the largest LLS slope was -1.4 × 10-2 (ng As g water-1) day-1). The AcOH treatment eliminated the formation of the Fe/As precipitate observed in the nontreated samples. However, two of the AcOH water samples produced analytically significant changes in the As(III) concentration. The LLS slopes for these two waters were -5.7 × 10-2 (ng As(III) g water-1) day-1 and -1.0 × 10-1 (ng As(III) g water-1) day-1. This corresponds to a -4.3 ng/g and a -7.8 ng/g change in the As(III) concentration over the study period, which is a 10% shift in the native distribution. The third and final treatment was EDTA/AcOH. This treatment eliminated the Fe/As precipitate that formed in the * Corresponding author phone: (513)569-7833; fax: (513)569-7757; e-mail:
[email protected]. 10.1021/es035071n Not subject to U.S. Copyright. Publ. 2004 Am. Chem. Soc. Published on Web 04/06/2004
nontreated sample. The LLS slopes were less than -7.5 × 10-3 (ng As(III) g water-1) day-1 for the abovementioned waters, corresponding to a 0.6 ng/g change over the study period. One of the EDTA/AcOH treated waters did indicate that using the 5 °C storage temperature minimized the rate of conversion relative to 20 °C storage.
Introduction Arsenic is present predominately as arsenite [As(III)] and arsenate [As(V)] in ground and surface waters used as drinking water supplies. The sampling and treatment of arsenic in drinking water is a growing area of research because of the new maximum contaminant level (MCL) for arsenic (10 ng/ g) under the Safe Drinking Water Regulations (1). The distribution of As(III) and As(V) in the source water used for drinking water consumption is one factor which influences the treatment removal efficiency. As(V) (pKa ) 6.9) is generally anionic at drinking water pHs (6-8), while As(III) (pKa ) 9.3) is predominately neutral. The anionic form of As(V) facilitates its removal using most arsenic removal technologies, while the neutral charge associated with As(III) limits its removal efficiency(2). This difference in removal efficiencies often dictates the need to determine the distribution of As(III) and As(V) prior to developing a removal strategy. High concentrations of As(III) may imply the need for a preoxidation step in the overall treatment process. Under the new arsenic regulation, some systems will need to modify their existing systems to meet the lower MCL, while other utilities will need to install new technology. In either case, it is advantageous to determine the distribution of arsenic within the source water prior to implementing a treatment strategy. The preservation of As(III) and As(V) in natural waters has been investigated by a number of scientists. Cherry et al. (3) utilized redox reagents common to natural waters such as O2 and H2S and Fe(III) and concluded that the rates of reaction associated with these conversions should be sufficiently slow to allow sample shipment to the laboratory. Tallman et al. (4) came to the same conclusion. Feldman (5) used ascorbic acid, Van Elteren et al. (6) used lower pHs and temperatures, and Hall et al. (7) used lower temperature and acid in an attempt to preserve the native As(III)/As(V) distribution. In all cases the conversion was minimized but not eliminated. Cheam et al. (8) used sulfuric acid and polyethylene or Pyrex storage containers and concluded that As(III) and As(V) could be stabilized at room temperature. Andreae et al. (9) recommended freezing the natural water sample to ensure speciation based integrity. To further complicate the problem, Ahmann (10) and Hambsch (11) have reported that the conversion between As(III) and As(V) can take place via microbiological activity. In summary, the literature is not in complete agreement with respect to preserving the aqueous As(III)/As(V) distribution, while lower temperatures (5 °C) and some type of acidification is generally recommended to retain the original concentrations. In some natural waters it becomes necessary to distinguish aqueous phase arsenic (AsAq) from particulate/colloidal bound arsenic (AsPPT), because the existence of the latter complicates the preservation issue. Hall et al. (7) and Rassler et al. (12) both report colloidal arsenic (not trapped by 0.45 µm filter and not in the aqueous phase), while Rassler (12) utilizes ultrafiltration to further fractionate the (AsPPT) in waters high in iron, manganese, and sulfur. The geological coexistence of iron and arsenic in groundwater results in the potential for a “FeAsOOH” (“FeAsOOH” is used in this paper VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2919
FIGURE 1. Drinking water sampling, processing, and storage flow diagram. to represent arsenic containing iron precipitates) coprecipitate to form after sample collection. This implies a decrease in the aqueous phase arsenic (AsAq), and the total arsenic within the sample must then be considered to be the sum of AsAq and AsPPT. Pierce (13) and DeVitre (14) have studied the interaction of amorphous Fe oxides with arsenic. Aggett (15), Borho (16), Anderson (17), and Boyle (18) all reported arsenic preservation problems in iron-rich waters. Aggett (15) indicated that a sample pH of 2 was essential in holding the iron in solution. Anderson (17) indicated that delays in filtration as small as a few hours could influence the aqueous vs particulate bound arsenic fractions. Gallagher et al. (19) indicated this time can be as short as 10 min for natural and distilled waters fortified with iron at a pH 8.0. Since this precipitation reaction can take place during shipment, samples speciated in the field may not agree with the samples speciated in the laboratory. Therefore, to preserve these types of samples for accurate arsenic speciation analyses, the impact of coprecipitation of the Fe/As must be understood. 2920
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
The ability to quantify As(III) and As(V) is routine, but the ability to preserve the native point of sampling distribution of As(III)/As(V) until the sample can be shipped and analyzed in the laboratory is a somewhat more challenging problem. Clifford (20), Ficklin (21), and Le (22) have developed in-field speciation approaches in order to eliminate the shipment induced preservation issues, but this requires a skilled person to be present during sampling and some knowledge of the matrix is required in order to ensure the capacity of the fieldseparation column is not exceeded. Another approach to the analysis is to preserve in-field and ship it to the laboratory for analysis. Pursuing this approach, Gallagher et al. (19) utilized a combination of acetic acid (AcOH) and EDTA to retard the “FeOOH” precipitation, stabilizing the As(III)/As(V) ratio. Bednar et al. (23) also reported the use of EDTA as a means of preserving the aqueous arsenic in groundwaters and acid drainage. Bednar et al. (23) compared laboratory speciation to field based speciation cartridges using 71 groundwaters which were filtered and preserved in the field
with EDTA (but no additional acid) prior to speciation. Finally, Clifford et al. (24) utilized AcOH/EDTA to eliminate “FeAsOOH” formation during in-field speciation of arsenic. The use of AcOH/EDTA as an As(III)/As(V) preservation treatment in drinking water was evaluated (see Figure 1). This differs from the work of Bednar et al. (23), which compared two speciation techniques using one in-field preserved sample; this manuscript compares different infield preservation and storage techniques using a single laboratory speciation based analysis. The unique aspect of this research is that the stability of the individual arsenic species are assessed in time, which provides a foundation for establishing preservation and holding time criteria for arsenic speciation based analysis in drinking water supplies. Ten waters with varying iron concentrations and varying natural As(III)/As(V) distributions were studied for over 2 months using seven different storage conditions. The goal was to develop an improved means for preserving As(III)/ As(V) in iron rich source waters, which will aid treatment engineers in meeting the lower arsenic MCL within the Safe Drinking Water Regulations.
Experimental Section Reagents. The A.C.S. certified nitric acid [HNO3] was purchased from Fisher Scientific (Pittsburgh, PA). The liquid chromatography (LC) mobile phase consisted of glacial acetic acid [CH3COOH] (Fisher Scientific, A.C.S. certified), ammonium nitrate [NH4NO3] (Fisher Scientific), ethylenedinitrilotetraacetic acid tetrasodium salt dihydrate [EDTA] (J. T. Baker Chemical Co., Phillipsburg, NJ), and ammonium hydroxide [NH4OH] (trace metal grade, Fisher Scientific). The mobile phase contained 50 mM glacial acetic acid, 50 mM ammonium nitrate, 25 mM ammonium hydroxide, and 500 µg/g EDTA. The glacial acetic acid and nitric acid were also used to adjust the pH of samples and to resolubilize the “FeAsOOH” from the particulate on the 0.45 µm filter (Pall Gelman Sciences, Ann Arbor, MI). The 0.2 µm filters were Fisherbrand (Fisher Scientific). The standards were prepared from arsenite [As(III)] and arsenate [As(V)] purchased from Spex CertiPrep (Metuchen, NJ). All standards were verified against NIST 1643c based on total metal. All sample dilutions were carried out using a Mettler (Columbus, OH) PM1200 (0.001 g) or a Mettler AG204 (0.0001 g). All total metal measurements were collected using Ge as an internal standard and U.S. EPA Method 200.8. The 1 L sample collection bottles were made out of highdensity polyethylene. These samples were shipped to the laboratory overnight in coolers. The 125 mL bottles used for sample storage during the study were made from low-density polyethylene. The Target DP autosampler vials (National Scientific Company, Duluth, GA) were acid cleaned in 10% HNO3 and rinsed with 18 MΩ water (Millipore, Bedford, MA) prior to use. Instrumentation. The LC system used was an Agilent 1100 (Palo Alto, CA) with an inert injection needle. The chromatographic column was a Hamilton (Reno, NV) PRP-X100 (Peek, 4.6 mm × 25 cm) with the associated guard column. The mobile phase contained 50 mM glacial acetic acid, 50 mM ammonium nitrate, 500 µg/g EDTA, and 25 mM NH4OH. The pH of the mobile phase was 4.5. The pump was operated in the isocratic mode at 1 mL/min with a 100 µL injection volume. The analysis time was 7.5 min. To monitor instrument drift, a 100 µL injection of a 20 ng/g As(V) solution in the mobile phase was introduced postcolumn via a LabPro Rheodyne valve (Rohnert Park, CA). The valve was externally controlled by the Agilent 1100 LC. The ICP-MS used was an Agilent 4500. The RF power was 1200 W, and the cool and auxiliary argon flows were 14.8 and 0.9 L/min, respectively. The nebulizer was concentric (Preci-
sion Glassblowing, Englewood, CO) and was operated at a flow of 1.1 L/min. All chromatographic data collection was completed using single ion monitoring of m/z 75. The Agilent 1100 and the Agilent 4500 were interfaced (without any end user modifications) such that unattended operation was routine (19). The iron and manganese concentrations were determined with an Agilent 7500c collision cell ICP-MS. Sampling and Treatment Schemes for Arsenic Preservation Study in Well Waters. The well water samples were collected from the Southwest, Midwest, and Western parts of the United States where arsenic occurrence in groundwater is prevalent. The samples were collected with variable native distributions of As(III)/As(V) and considerable variation in native iron. Figure 1 is a schematic for the sampling, laboratory processing, and storage conditions used in this study. Three 1 L Nalgene containers were sent to each site. One of these was the control (no treatment), the second was the EDTA/ AcOH treatment, and the third was an acetic acid only treatment (AcOH). The EDTA/AcOH 1 L container was shipped to the sampling site with 10 mL of 8.7 M AcOH inside the container. Attached to this container was a 15 mL Nalgene bottle which contained 10 mL of 50 000 µg/g EDTA. (The EDTA cannot be added directly to the 8.7 M AcOH because of solubility issues). The sampling technician was given instructions to fill the 1 L container 3/4 of the way with sample, pour in the EDTA solution, and then finish filling the bottle with sample. The 1 L Nalgene bottles used for the AcOH treatment were shipped with only the 10 mL of 8.7 M AcOH. In all cases the sampler was told to fill the container trying to minimize headspace and ship it overnight back to the laboratory in a provided cooler. Samples were received in the laboratory by 10 a.m. the next day. Once received in the laboratory, samples were inverted several times (to ensure that a more representative aliquot could be taken for samples containing the “FeAsOOH” precipitate), and a 100 mL sample was taken and acidified for a total arsenic determination (see AsTotal, Figure 1). The remaining 900 mL sample was filtered through a 0.45 µm filter followed by a 0.2 µm filter. The 0.45 µm filter was first extracted with 1 M AcOH, then treated with EDTA, and pH adjusted prior to IC-ICP-MS analysis. After the 1 M AcOH extraction the 0.45 µm filters were extracted with 10% HNO3 (wt/wt) (to ensure complete dissolution of the iron/arsenic on the filter), and the 10% HNO3 extract was analyzed for total arsenic only. The 0.2 µm filter was extracted with 10% HNO3 only. (The nitric acid extracts were not speciated, but rather a total arsenic was determined in order to conduct an arsenic mass balance.) The sum of the arsenic extracted from the filters is referred to as AsFilter in Figure 1. The arsenic which passed through the 0.45 and 0.2 µm filter is referred to as AsDissolved in Figure 1. From a mass balance perspective, the ng of arsenic associated with the AsTotal should equal the sum of the ng of arsenic associated with AsFilter plus AsDissolved. Approximately 900 mL of the filtered samples were prepared for each of the three treatments. Each 900 mL sample was partitioned into separate 125 mL Nalgene containers (see lower half of Figure 1). The three 125 mL containers for the nontreated samples were stored at 5 °C (N-5-N), 20 °C (N-20-N), and (N-20-F); 20 °C samples were held at room temperature in the dark to mimic the lighting conditions within a walk-in refrigerator, and the third was stored at 20 °C and filtered through a 0.45 µm filter prior to analysis each day (N-20-F). The filtering provided an ongoing check to see if additional “FeAsOOH” was forming during the 60+ days of storage. The three 125 mL EDTA/AcOH treated samples were stored at 5 °C (E-5-N), 20 °C (E-20-N), and 20 °C with an aeration step on day one (E-20-A). The aeration step was in-house air bubbled though the sample VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2921
for 20 min. and was included to mimic a sampling condition in which the dissolved oxygen (which may cause the oxidation of native As(III)) may be slightly altered because of poor sampling. A comparison of the 5 °C and 20 °C EDTA/AcOH treatment samples with the nontreated samples was used to see if temperature was a factor in preservation. Finally, the AcOH treated sample was stored at 5 °C (H-5-N). A comparison of this to the EDTA/AcOH was used to determine whether EDTA is a factor in the preservation. The sampling code and its meaning is outlined at the bottom of Figure 1. Sample Analysis and Standard Preparation Prior to ICICP-MS. The refrigerated samples were sampled in the walkin refrigerator or removed for approximately a 5 min period and then returned so that the temperature did not appreciably change. After sampling, the headspace of the sample was reduced (by collapsing the plastic bottle) to minimize the potential of oxygen from the headspace causing an oxidation of As(III) to As(V) after numerous samplings. Aggett et al. (15) demonstrated that multiple samplings produced fresh and increasing headspace above the sample that in time may induce an oxidation of As(III) to As(V) within a sample. Each sample was analyzed (samples were diluted in eluent such that their concentration was within the calibration range) in duplicate on each sampling day using a staggered analysis sequence in order to ensure that instrument drift did not affect any treatment preferentially. A mixed As(III) and As(V) standard was made up in 10% eluent and run after every four samples. The standard was prepared fresh daily and verified to contain equal amounts of As(III) and As(V) via peak area ratios.
Results and Discussion Nontreated Samples. Figure 2a is a plot of As(III) concentration in well water #1 over a 78 day period, while Figure 2b is a plot of the As(V) concentrations for well water #1 over that same time period. Figure 2a clearly indicates that the As(III) concentration associated with the EDTA/AcOH ([) and AcOH (b) treatments is approximately 18 ng/g higher than the nontreated (9) water throughout the study period. This difference is caused by the coprecipitation of arsenic with the iron in the nontreated samples and the subsequent removal of the “FeAsOOH” by the initial filtration with a 0.45 µm filter (see Figure 1). This reduction in the initial arsenic concentration was observed for all samples containing the iron precipitate (waters 1-5 and 10). Twenty-five to forty percent of the total arsenic was lost in these samples due to the formation of “FeAsOOH”. Clearly, the need for a preservation agent is indicated by the loss of dissolved arsenic in these iron rich waters. The initial loss of the arsenic (18 ng/g) to the formation of “FeAsOOH” during the sample transport must establish a new equilibrium between the iron and arsenic for the nontreated samples. This new equilibrium is evident by the relatively stable As(III) concentration in Figure 2a and by the lack of additional “FeAsOOH” formation throughout the study period. The sampling design utilizes the filtering of sample N-20-F as a secondary means of verifying that additional “FeAsOOH” is not formed during the study period. The 0.45 µm filters associated with sample N-20-F did not contain detectable quantities of arsenic, indicating that little if any additional “FeAsOOH” is formed from day 1 to day 76. Figure 2a,b indicates some subtle changes in the distribution of As(III) and As(V) in the aqueous phase during the study period. Some of these changes are treatment specific. For instance, the combination of Figure 2a,b indicates a 1.5 ng/g conversion of As(III) to As(V) over the 76 day study period for the nontreated sample held at 20 °C (0, N-20-N). This same trend is not observed for the nontreated sample held at 5 °C (9, N-5-N). In general, across all waters used in 2922
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
FIGURE 2. A. Preservation of As(III) in well water #1 as a function of time using seven different storage conditions. The mean of daily replicate determination is plotted. Error bars based on relative percent difference were determined but for clarity only plotted for the point with the largest RPD (t ) day 5). Slope [4.4 × 10-3 (ng As(III) g-1) day-1] and y-intercept [66.7 ng As(III)/g] are calculated by linear least squares (LLS) based on pooled data from EDTA/ acetic acid treatments. The 95% confidence bound for the EDTA/ AcOH LLS is represented by a dashed line. B. Preservation of As(V) in well water #1 as a function of time using seven different storage conditions. The mean of daily replicate determination is plotted. Error bars based on relative percent difference were determined but for clarity only plotted for the point with the largest RPD (t ) day 1). Slope [-1.2 × 10-3 (ng As(V) g-1) day-1] and y-intercept [1.7 ng As(V)/g] are calculated by LLS based on pooled data from EDTA/ acetic acid treatments. The 95% confidence bound for the EDTA/ AcOH LLS is represented by a dashed line. a Axis is used for acetic acid and no treatment samples. b Axis is used for EDTA/acetic acid treatment only.
TABLE 1. An Investigation of Oxidation State Stability for Arsenic in Time Using a Linear Least Squares Approach To Estimate Distribution Changes in Drinking Water Matrices Preserved with EDTA/AcOHa AsTotal vs AsChromatography sample (days analyzed) water 1 (76 days) water 2 (78 days) water 3 (78 days) water 4 (76 days) water 5 (77 days) water 6 (79 days) water 7 (98 days) water 8 (83 days) water 9 (71 days) water 10 E-5-N (78 days) water 10 E-20, n ) 2 (78 days)
percent As total AsTotalb chromatographedc (ng/g) (% ( 2σ)
linear least squares (LLS) LLS equationd,e (y ) m*x + b) As(III)
As(V)
m ) -1.2 × 10-3 b ) 1.7 m ) -2.7 × 10-3 b ) 2.1 m ) -3.3 × 10-3 b ) 0.7 m ) 7.5 × 10-3 b ) 2.2 m ) 4.1 × 10-3 b ) 1.8 m ) -1.1 × 10-2 b ) 36.8 m ) -1.4 × 10-2 b ) 34.2 m ) -1.6 ( 10-2 b ) 37.9 m ) 3.3 × 10-3 b ) 0.8 m ) 9.9 × 10-3 b ) 5.6 m ) 6.5 × 10-2 b ) 5.2
71.9
97 ( 1.9
35.0
94 ( 2.8
20.8
82 ( 2.2
53.1
99 ( 0.8
64.2
98 ( 1.3
38.6
93 ( 0.7
m ) 4.4 × 10-3 b ) 66.7 m ) -5.1 × 10-3 b ) 31.2 m ) -1.4 × 10-2 b ) 17.5 m ) -1.4 × 10-2 b ) 51.0 m ) -2.5 × 10-2 b ) 63.2 ND
35.7
93 ( 0.9
ND
38.1
96 ( 1.6
ND
19.1
105 ( 3.8
36.6
96
36.6
95 ( 7.4
m ) -1.3 × 10-3 b ) 18.7 m ) -4.2 × 10-3 b ) 28.8 m ) -6.5 × 10-2 b ) 29.5
predicted difference f (ng/g) As(III)
As(V)
predicted % As(III) day 1g % As(III)
final dayh % As(III)
0.3 ( 0.7 -0.1 ( 0.4 97.5 ( 0.9 97.6 ( 0.9 -0.4 ( 0.7 -0.2 ( 0.2 93.8 ( 1.0 94.4 ( 1.1 -1.1 ( 0.4 -0.3 ( 0.1 96.0 ( 0.6 97.2 ( 0.6 -1.1 ( 0.9 0.6 ( 0.4 95.9 ( 1.3 94.8 ( 1.4 -1.9 ( 0.9 0.3 ( 0.3 97.3 ( 0.3 96.7 ( 0.3 ND
-0.9 ( 0.8
ND
ND
ND
-1.4 ( 0.5
ND
0.3 ( 0.4
ND
-1.3 ( 0.5
ND
ND
-0.1 ( 0.3 0.2 ( 0.3 95.8 ( 1.4 94.6 ( 1.4 -0.3 ( 0.6 0.8 ( 0.6 83.7 ( 1.8 81.7 ( 1.9 -5.0 ( 0.6 5.0 ( 0.8 84.8 ( 2.8 70.5 ( 2.9
a The linear least squares (LLS) and the chromatographic percentage are calculated by pooling all the EDTA treatment conditions. The only exception to this is water #10 in which the E-5-N treatment is not pooled with the other EDTA treatments. ND ) no detection. b AsTotal is determined on an acidified sample from the 1 L sample prior to filtration. c The data for the 3 EDTA/AcOH storage conditions have been pooled to calculate an average chromatographic recovery relative to AsTotal for the last analysis day. The 2σ is calculated based on n ) 3. For water #10, the data for the 20 °C storage conditions have been pooled (n ) 2); therefore, the E-5-N percentage is based on n ) 1. d The species specific LLS calculation is based on determining the average for each treatment from each analysis day, and then the 3 averages/day were used to calculate the LLS terms. e The slopes are reported as (ng As g water-1) day-1. The y-intercepts are for t ) 0 which corresponds to the first day of the study. f The predicted difference is calculated from the LLS estimate for the first and last analyses days. The predicted difference is reported as ng As/g water. The concentration differences are reported with 95% confidence intervals. g This is a predicted percentage for As(III) on day 1. It is calculated using a percentage based LLS. The predicted percentage is reported with a 95% confidence interval. h This is a predicted percentage for As(III) on the final day. It is calculated using a percentage based LLS. The predicted percentage is reported with a 95% confidence interval.
this study, the nontreated samples held at 20 °C produced changes in the As(III)/As(V) distribution earlier in the time study than the samples held at 5 °C. These treatment specific changes are relatively small in Figure 2 (compared to the remaining 9 waters used in the study) as some of the nontreated samples associated with other waters produced complete inversion of the distribution during the study period. This treatment specific variation observed for nontreated samples prohibits the pooling of data across treatments for statistical purposes. While, the distribution changes observed in Figure 2 would not affect the engineering treatment decision making process for the nontreated samples associated with water #1, the initial loss of arsenic to “FeAsOOH” would alter the decision making. Therefore, changes in the native concentration and/or the native distribution can affect the treatment decision making process. EDTA/AcOH Treatment. The across treatment variability observed for the nontreated samples in Figure 2a,b was not observed for the EDTA/AcOH treated samples. The EDTA/ AcOH treatments produced little variability across storage conditions; therefore, the data was pooled and a single linear least squares (LLS) line was calculated for the EDTA/AcOH treated waters. The error bars provide an estimate of the replicate variation as a relative percent difference (RPD) for the given EDTA/AcOH sample treatment set (n ) 3). For graphical clarity purposes only the largest RPD is reported in Figure 2a,b. The slope and the y-intercept for the LLS line in Figure 2a are 0.0044 (ng g-1) day-1 and 66.7 ng/g, respectively. The slope of this line is an estimate of the interconversion rate of As(III) to As(V) over the study period. The 95% confidence bound associated with the EDTA/AcOH LLS line is represented by a dashed line in Figure 2a,b.
In Figure 2a, a predicted As(III) change of 0.3 ng/g was calculated using the LLS slope and an 80 day period. The slope of the LLS line for As(V) in Figure 2b is -0.0012 (ng g-1) day-1 resulting in a predicted change in As(V) of -0.096 ng/g using an 80 day period. These estimated changes are analytically insignificant given the maximum within day duplicate variation of 1.5 ng/g (2.2% RPD) for a sample concentration of 66 ng/g As(III). From a drinking water engineering decision making perspective, this change would not alter the treatment decision making process. AcOH Treatment. The AcOH treatment compares very well with the EDTA/AcOH treatment in Figure 2a. The “FeAsOOH” formation observed in the nontreated sample is eliminated using the AcOH treatment which lowers the pH of the sample. Aggett et al. (15) and Hall et al. (7) have used pH adjustment as an indirect means of preserving the arsenic by reducing the “FeAsOOH” formation. An inspection of Figure 2b indicates a 1 ng/g rise in the As(V) concentration during the study period for the AcOH treated samples (b, H-5-N). Four out of the six waters which formed the “FeAsOOH” precipitate for the nontreated samples produce a plot very similar to Figure 2b in that the concentration of As(V) gradually diverges in time relative to the LLS line calculated for the EDTA/AcOH samples. The increase in As(V) associated with the four waters was between 1 and 7 ng/g over the study period. The 7 ng/g rise in As(V) (in water #5) coincided with a decrease of 8 ng/g in As(III). This change would be considered analytically significant but probably would not change the treatment decision given that its initial As(III) concentration was 62 ng/g. A more complete comparison for the AcOH treated samples with respect to the distribution changes in time can be found in Table 3. VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2923
TABLE 2. An Investigation of Oxidation State Stability for Arsenic in Time Using a LLS Approach To Estimate Distribution Changes in Drinking Water Matrices without Preservationa AsDissolved AsFilter sample (pH) water 1 (pH ) 7.3) water 2 (pH ) 7.5) water 3 (pH ) 7.9) water 4 (pH ) 7.4) water 5 (pH ) 7.6) water 6 (pH ) 7.3) water 7 (pH ) 8.6) water 8 (pH ) 7.5) water 9 (pH ) 7.3) water 10 (pH ) 7.1)
Fe (ng/g) 1290
Mn (ng/g) 120
mass balance percent recoveryb
percent lost to filterc
96.0
percent differencef
percent distributiond As(III)
As(V)
LLS for totale
24.9
68.3
31.7
m ) 4.8 × 10-3 b ) 51.3 m ) -1.3 × 10-3 b ) 22.4 m ) -8.1 × 10-3 b ) 13.0 m ) -98 × 10-3 b ) 33.2 m ) -6.7 × 10-3 b ) 39.2 m ) -7.3 × 10-3 b ) 36.7 m ) -1.2 × 10-2 b ) 36.8 m ) -1.4 × 10-2 b ) 38.5 m ) 1.2 × 10-2 b ) 19.4 N-5-N m ) 1.1 × 10-2 b ) 26.6 N-5-N
1200
7.9
100.0
36.0
21.9
78.1
720
3.6
88.5
26.9
29.3
70.7
1420.1
19
99.6
37.3
35.7
64.3
1780
14
101.1
40.5
55.1
44.9
150
3.7
92.1
g
g
g
31
2.7
102.8
g
g
g
16
0.2
101.9
g
g
g
160
10
103.2
g
g
g
780
130
96.1
21.9
10.8
89.2
N-5-N As(III)
N-20-N As(III)
0
-3.7
14.0
14.4
-49.8
14.6
24.7
19.4
-84.0
-43.1
ND
20.9
1.4
100
ND
ND
-4.9
8.8
-2.4
-1.5
a The nontreated water produced large variations across treatment conditions; therefore, the species specific data could not be pooled to determine a species specific linear least squares (LLS). However, the variation associated with the sum of As(III) + As(V) was small allowing the sum of the species to be pooled across treatment. The pH for each water analyzed is reported in parentheses under the sample name. The days analyzed for each water are the same as previously reported in Table 1. ND ) no detect. Mn and Fe concentration were determined using collision cell ICP-MS. b Percent recovery is determined by adding the arsenic concentration on the filter (AsFilter) to the chromatographable arsenic (AsDissolved average), and this total is divided by the total arsenic determined after acidification (AsTotal). c Percent loss to filter is the arsenic concentration on the filter (AsFilter) divided by the total arsenic determined after acidification (AsTotal). Arsenic concentration on the filter (0.45 and 0.2 µm) is the sum of the arsenic found after a 1 M AcOH extraction plus the arsenic found after a 10% HNO3 extraction. d The distribution of As(III) and As(V) on the filter is determined after an extraction of the filter with 1 M AcOH. (n ) 1, because the entire 1 L sample was filtered). e The LLS for the nontreated samples were calculated based on the sum of As(III) + As(V). Given that the LLS is not a species specific analysis it was possible to pool the data across treatments. f Percent difference is calculated by taking the percent As(III) on the last day minus the percent As(III) on the first day for N-5-N and N-20-N. g No “FeAsOOH” precipitate was observed when these samples were received in the laboratory.
The seven samples prepared from each of the sampling sites were analyzed over time, and the results were summarized for each water in a format similar to Figure 2. Figure 2 only represents the data from well water #1 of the 10 waters used in the study. A complete graphical representation of the data set would be too cumbersome for publication; therefore, Tables 1-3 contain the summarized data for each sample.
Data Summary for All Ten Waters using an EDTA/AcOH Sample Storage Treatments Aqueous vs Solid-Phase Arsenic: Mass Balance between Total and Chromatographable Arsenic. Table 1 summarizes the data for the EDTA/AcOH samples. One of the potential problems in a metal preservation study is the loss of aqueous phase metal due to precipitation or adsorption to the container wall. To assess whether the coprecipitation of arsenic (observed for nontreated samples in Figure 2) affects the data through time, a mass balance comparison was made for all data sets, which is reported in Tables 1-3. The mass balance was calculated between the total arsenic determined after acidification (second column) and the chromatographable (dissolved /aqueous) arsenic within the sample. The chromatographable arsenic should estimate the percentage of arsenic in the dissolved (aqueous) phase because the IC column has an in-line 2 µm filter which would remove precipitates. The third column of the Table reports the mass balance based on the measured chromatographic concentration on the last day and the total arsenic. The percentages of the chromatographable arsenicals range from 2924
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
93 to 105% for 9 out of the 10 waters, with water #3 being an outlier at 82%. The reason for the low recovery associated with water #3 is unknown. This mass balance percentage can be estimated for any remaining day by utilizing the LLS equations in columns four and five. These data along with the relatively constant chromatographic concentration determined using the LLS indicates that the arsenic remains in the dissolved or aqueous phase throughout the study period. A second means of confirming that “FeAsOOH” was not formed during the sample transport period was obtained by analyzing the 0.45 and 0.2 µm filters used on the first day (see Figure 1). These filters contained undetectable quantities of arsenic for the EDTA/AcOH samples. This implies that the arsenic was received in the laboratory in the aqueous phase (AsDissolved, Figure 1). Estimating the Interconversion of Arsenic using LLS. The LLS slope and y-intercept are reported in columns four and five for As(III) and As(V). These equations are used to predict a concentration at t ) 0 and on the final analysis day. The predicted concentration differences for As(III) and As(V) are reported in columns six and seven, respectively. Included with the point estimates in column six and seven are the 95% confidence bounds for the difference. The largest predicted change for the first 9 waters was observed for water #5 at -1.9 ng/g As(III). This represents a 3% decrease relative to the initial As(III) concentration, and this predicted concentration change would be considered insignificant analytically. The last two columns of the Table 1 contain the predicted percentage of As(III) on the first and last day and associated 95% confidence bounds. A review of these data
TABLE 3. An Investigation of Oxidation State Stability for Arsenic in Time Using a Linear Least Squares Approach To Estimate Distribution Changes in Drinking Water Matrices Preserved with AcOHa linear least squares (LLS) LLS equationd,e (y ) m*x + b)
AsTotal vs AsChromatography AsTotalb
sample percent AsTotal (days analyzed) (ng/g) chromatographedc(%) water 1 (76 days) water 2 (78 days) water 3 (78 days) water 4 (76 days) water 5 (77 days) water 6 (79 days) water 7 (98 Days) Water 8 (83 days) water 9 (71 days) water 10 (78 days)
As(III)
As(V)
m ) 9.0 × 10-3 b ) 3.1 m ) 4.7 × 10-3 b ) 3.6 m ) 1.7 × 10-2 b ) 1.1 m ) 4.9 × 10-2 b ) 3.7 m ) 9.1 × 10-2 b ) 3.6 m ) -9.8 × 10-3 b ) 35.6 m ) -1.5 × 10-2 b ) 36.0 m ) -1.1 × 10-2 b ) 37.8 m ) 3.1 × 10-3 b ) 0.9 m ) 4.9 × 10-3 b ) 5.7
74.3
94.8
35.0
99.4
20.8
82.7
53.8
97.2
66.1
99.9
41.8
89.3
m ) -2.8 × 10-2 b ) 66.3 m )m ) -2.3 × 10-2 b ) 32.5 m ) -2.8 × 10-2 b ) 16.7 m ) -5.7 × 10-2 b ) 48.7 m ) -1.0 × 10-1 b ) 61.7 ND
38.6
97.7
ND
38.6
97.4
ND
19.7
100.7
37.7
93.4
m ) -2.7 × 10-3 b ) 17.9 m ) -6.3 × 10-3 b ) 29.2
predicted difference f (ng/g) As(III)
As(V)
predicted % As(III) day 1g % As(III)
final dayh % As(III)
-2.1 ( 2.3 0.7 ( 0.9 95.6 ( 1.8 94.5 ( 2.1 -1.8 ( 2.2 0.4 ( 1.4 90.0 ( 5.1 88.6 ( 6.4 -2.1 ( 0.8 1.3 ( 0.4 93.5 ( 2.6 85.5 ( 3.2 -4.3 ( 1.2 3.7 ( 1.0 92.9 ( 2.7 85.7 ( 3.3 -7.8 ( 1.7 6.9 ( 2.1 94.3 ( 4.5 83.5 ( 5.5 ND
-0.8 ( 1.4
ND
ND
ND
-1.5 ( 1.0
ND
0.3 ( 0.4
ND
-0.9 ( 0.5
ND
0.7 ( 1.0
-0.2 ( 0.5 0.2 ( 0.3 95.3 ( 1.9 94.2 ( 2.0 -0.5 ( 0.9 0.4 ( 0.9 83.6 ( 2.3 82.5 ( 2.5
a The linear least squares (LLS) and the chromatographic percentage are calculated for the AcOH treatment condition. ND ) no detection. AsTotal is determined on an acidified sample from the 1 L sample prior to filtration. c The percent AsTotal chromatographed was calculated for the AcOH storage condition for the last analysis day. [(AsDissolved/AsTotal)*100]. d The species specific LLS calculation is based on a single point determination of As(III) or As(V) for each analysis day. e The slopes are reported as (ng As g water-1) day-1. The y-intercepts are for t ) 0 which corresponds to the first day of the study. f The predicted difference is calculated from the LLS equation for the first and last analyses days. The predicted difference is reported as ng As/g water. The concentration differences are reported with 95% confidence intervals. g This is a predicted percentage for As(III) on day 1. It is calculated using a percentage based LLS. The predicted percentage is reported with a 95% confidence interval. h This is a predicted percentage for As(III) on the final day. It is calculated using a percentage based LLS. The predicted percentage is reported with a 95% confidence interval. b
indicates the predicted interconversion of As(III) and As(V) over the study period for the first 9 waters is analytically insignificant. The percentage change is less than 2%. These changes would not influence the treatment decision making process. The findings are consistent with Gallagher et al. (19) and Bednar et al. (23) in that EDTA stabilizes the As(III)/ As(V) distribution by inhibiting undesirable precipitation formation. Water #10 in Table 1 is the only water in which the results of the analyses following the EDTA/AcOH treatments (E-5N, E-20-N, E-20-A; see Figure 1) were not pooled together for a single LLS calculation. The reason for this is that the concentration of the samples held at 20 °C were noticeably lower than those of the samples held at 5 °C during the study period. This becomes obvious by looking at the predicted difference and confidence bounds over the study period for As(III) (column six) and As(V) (column seven). The predicted decrease of 5.0 ng/g for As(III) represents a 17% (5.0/29.5) change in the As(III) concentration, while the other waters and the E-5-N treatment for the water #10 produced predicted changes in the 0.5-6% range. The last two columns of the table contain the predicted percentage of arsenic for the first and last day and associated 95% confidence bounds. A review of these data indicates that the predicted interconversion of As(III) and As(V) over the study period for water #10 (held at 20 °C) is analytically significant and may influence the treatment decision making. These results demonstrate that for long-term storage, a 5 °C storage temperature is preferred to 20 °C. It is worth noting that the sample design included a sample which would be aerated after it had returned to the laboratory (E-20-A). This aeration step was included to simulate a sample taken in a manor in which the dissolved oxygen may change in the sample due to improper sample collection. The E-20-A data was pooled with the E-5-N and E-20-N because of the low variability across all three treatments. These preliminary
findings indicate that a small amount of aeration does not produce As(III)/As(V) distribution shifts.
Data Summary for all Ten Waters using a Nontreated Sample Storage Condition Percentage and Distribution of Arsenic Removed by Filtration. Table 2 is a summary of the nontreated samples for the 10 drinking water supplies used in this study. In columns two and three, the iron and manganese concentrations are reported, respectively. The iron and manganese concentrations are reported because these metals can influence the water chemistry associated with arsenic. The iron concentrations range from 16 to 1780 ng/g, while the manganese concentrations range from 0.2 to 130 ng/g. The nontreated samples produced large variations in As(III)/As(V) distribution with respect to treatment. Therefore, the data sets (N-5-N, N-20-N, N-20-F) could not be pooled together for statistical purposes. This variation is most obvious for water #7 in which the N-5-N sample treatment produces little change in the As(III) concentration over the study period, while the N-20-N indicates a complete conversion to As(V). The fourth column of the table reports an arsenic mass balance as a percent recovery, which is the sum of the chromatographable arsenic plus the arsenic found on the filter divided by the total arsenic determined on an acidified nonfiltered sample. The recoveries range from 88.5 to 103%. This indicates that nearly all the arsenic can be accounted for between the filter and the aqueous phase. The percentage of arsenic lost to precipitation during shipment is reported in the fifth column as a “percent lost to filter”. Relative to the total arsenic, the percent lost to the filter ranged from 0 to 40%. However, the percent loss to filter for waters containing the “FeAsOOH” precipitates ranged from 25 to 40%. The loss of aqueous arsenic to the filter influences the concentration of arsenic and for this reason can affect VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2925
the treatment decision making. To evaluate the oxidation state of the arsenic which precipitated, the “FeAsOOH” precipitate on the filters was resolubilized using 1 M AcOH. The resolubilzed “FeAsOOH” was analyzed by IC-ICP-MS. The distribution of As(III)/As(V) associated with the arsenic on the filter is reported in columns six and seven as a percentage. The arsenic associated with the filter always contained a higher percentage of As(V) relative to the aqueous phase from which its was removed. If the distribution on the filter is compared to the y-intercepts reported in Table 1 the distribution difference (As(III)/As(V)) between the aqueousand solid-phase arsenic becomes apparent. This difference may be generated by the preferential removal of As(V) by the “FeOOH”. Aqueous Phase Stability in Time after Precipitate Removal. The “FeAsOOH” formation, which occurred during shipment, was estimated by the percent lost to the filter. To determine if additional “FeAsOOH” was forming during the study, a sample held at 20 °C was filtered using a 0.45 µm filter just prior to analysis (N-20-F). The analysis of these filters did not detect additional arsenic indicating that additional “FeAsOOH” was not formed. A second method for determining the concentration of aqueous phase arsenic is to calculate a LLS for the combination of As(III) + As(V). The result should be an estimate of the aqueous phase arsenic throughout the study. The slope and y-intercept of the LLS for each water is reported in column eight of Table 2. The largest change is 1.4 × 10-2 (ng As g water-1) day-1 which corresponds to a 3.0% change. Both means of estimating the stability of aqueous arsenic during the study period indicated that the aqueous arsenic concentration after the initial filtration is relatively constant (i.e. no additional loss to filtration or absorption to the walls of the container). Stability of As(III)/As(V) Distribution for Nontreated Samples. The last two columns of Table 2 compare the distribution of As(III) on the first day to the last analysis day, as a percent difference between the two. A positive number indicates that the distribution has shifted toward As(III), while a negative number indicates a change in the distribution toward As(V). The percent difference ranges from -84 to +25 for the N-5-N samples (column nine). Even with this extreme range its worth noting that 6 out of the 10 waters produced distribution changes of less than 5%. For waters #6-9 the distribution change for N-5-N compares favorably with the EDTA/AcOH. (Table 1) However, the N-20-N for waters #6 and #7 produced a 21 and 100% difference in the As(III) concentration over the study period, much higher than that experienced in the EDTA/AcOH experiments. A comparison of columns nine and ten clearly indicates that the preservation of As(III)/As(V) distribution in time for nontreated samples is very sample and treatment dependent. It also clearly indicates that the initial distribution of As(III)/As(V) in nontreated samples can undergo a larger change over the study period. This unpredictable change in distribution, coupled with the initial loss of arsenic to the formation of “FeAsOOH” (6 out of 10 waters), makes the use of no preservation agents an unsuitable experimental protocol and will result in questionable data; drinking water treatment operators cannot make decisions based on these data.
Data Summary for All Ten Waters using an Acetic Acid Sample Storage Treatment Aqueous vs Solid-Phase Arsenic: Mass Balance between Total and Chromatographable Arsenic. Table 3 summarizes the preservation data for the AcOH treated samples without the EDTA. The first two columns identify the water and the total arsenic concentration after acidification. The third column reports a percentage calculated from a ratio of chromatographable arsenic to the total arsenic. This per2926
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
centage is calculated on the last day of analysis and is a means of checking to make sure the arsenic has not precipitated out of solution during the study period. The percentages of chromatographable ranges from 89 to 101% for 9 out of the 10 waters, with water #3 being an outlier at 83%. The reason for the relatively low percentage for water #3 is unknown, but it is consistent with Table 1. Similar to the EDTA/AcOH treated samples, the 0.45 and 0.2 µm filters did not contain measurable concentrations of arsenic; therefore, the AcOH treatment inhibited the formation of “FeAsOOH” in waters #1-5 and 10. These are the same waters which produced a “loss to filter” in Table 2. In addition, daily estimates of the chromatographable percentage can be calculated using the LLS parameters in columns four and five. The combination of undetectable arsenic concentration on the filters and the associated LLS slopes both indicate that the “FeAsOOH” precipitation observed for nontreated samples has been eliminated by the AcOH. Estimating the Interconversion of Arsenic using LLS. The LLS slopes associated with waters containing iron (water 1-5 and 10) are all greater for the AcOH treatment relative to the EDTA /AcOH treated waters. These equations are used to predict a concentration at t ) 0 as well as on the final analysis day. The predicted concentration differences for As(III) and As(V) are reported in columns six and seven, respectively. Included with the point estimates in columns six and seven are the 95% confidence bounds for the difference. The largest predicted change was observed for water #5, with a drop in the As(III) concentration of -7.8 ng As(III)/g water. This represented a 12.6% change relative to the initial As(III) concentration. Water #4 also produced a change of greater than -4 ng /g water in the As(III) concentration. In both cases, the reduction in As(III) concentration coincides with an increase in the As(V) concentration. These predicted concentration changes are analytically significant. This distribution (As(III)/As(V)) change may affect treatment decision making in systems which require little additional arsenic removal to meet the arsenic MCL. The last two columns of the table contain the predicted percentage of As(III) for the first and last days. Each point estimate in columns eight and nine has a 95% confidence bound associated with it. The predicted interconversion of As(III) and As(V) for waters #3-5 range from 7 to 11%, while the others are very comparable to the EDTA/AcOH data in Table 1. All of these waters (waters #3-5) produced “FeAsOOH” precipitates in the nontreated samples; therefore, the addition of the EDTA treatment aids in the preservation of the native distribution relative to AcOH treatments in waters #3-5. The distribution changes observed for waters #3-5 are much greater than those reported in Table 1 and may influence the treatment decision making.
Acknowledgments The authors would like to thank Tom Sorg from U.S. EPA’s NRMRL for providing the well water samples as well as his knowledge in treatment technologies. In addition, the authors would like to thank the participants in the Arsenic Technology Demonstration sites for providing well water samples. The U.S. Environmental Protection Agency through its Office of Research and Development funded and managed the research described in this paper. It has been reviewed in accordance with the Agency’s peer and administrative review policies and approved for publication. Mention of trade names or commercial products does not constitute endorsement or recommendation for use. Amy Parks and Bryan M. Gamble are Oak Ridge Research Fellows.
Literature Cited (1) Code of Federal Register 40CFR, Section 141.23(a)(4)(i). (2) McNeill, L. S.; Edwards, M. J. AWWA 1997, 89, 75.
(3) Cherry, J. A.; Shaikh, A. U.; Tallman, D. E.; Nicholson, R. V. J. Hydrol. 1979, 43, 373. (4) Tallman, D. E.; Shaikh, A. U. Anal. Chem. 1980, 52, 199. (5) Feldman, C. Anal. Chem. 1979, 51, 664. (6) van Elteren, J. T.; Hoegee, J.; van der Hoek, E. E.; Das, H. A.; de Ligny, C. L.; Agterdenbos, J. J. Radioanal. Nucl. Chem. Lett. 1991, 154, 343. (7) Hall, G. E. M.; Pelchat, J. C.; Gauthier, G. J. Anal. At. Spectrom. 1999, 14, 205. (8) Cheam, V.; Agemian, H. Analyst 1980, 105, 737. (9) Andreae, M. O. Anal. Chem. 1977, 49, 820. (10) Ahmann, D.; Roberts, A. L.; Krumholz, L. R.; Morel, F. M. M. Nature 1994, 371, 750. (11) Hambsch, B.; Raue, B.; Brauch, H.-J. Acta Hydrochim. Hydrobiol. 1995, 23, 166. (12) Rassler, M.; Michalke, B.; Schramel, P.; Schulte-Hostede, S.; Kettrup, A. Fresenius J. Anal. Chem. 1998, 362, 281. (13) Pierce, M. L.; Moore, C. B. Water Res. 1982, 16, 1247. (14) De Vitre, R.; Belzile, N.; Tessier, A. Limnol. Oceanogr. 1991, 36, 1480. (15) Aggett, J.; Kriegman, M. R. Analyst 1987, 112, 153.
(16) Borho, M.; Wilderer, P. J. Water SRT-Aqua 1997, 46, 138. (17) Anderson, R. K.; Thompson, M.; Culbard, E. Analyst 1986, 111, 1153. (18) Boyle, D. R.; Turner, R. J. W.; Hall, G. E. M. Environ. Geochem. Health 1999, 20, 199. (19) Gallagher, P. A.; Schwegel, C. A.; Wei, X.; Creed, J. T. J. Environ. Monit. 2001, 3, 371. (20) Clifford, D.; Ceber, L.; Chow, S. XI AWWA WQTC Proceedings, Norfolk, VA, December 1983. (21) Ficklin, W. H. Talanta 1983, 30, 371. (22) Le, X. C.; Yalcin, S.; Ma, M. Environ. Sci. Technol. 2000, 34, 2342. (23) Bednar, A. J.; Garbarino, J. R.; Ranville, J. F.; Wildeman, T. R. Environ. Sci. Technol. 2002, 36, 2213. (24) Clifford, D. A.; Ghurye, G. AWWARF Water Quality Technology Conference and Exhibition, Nov. 5-9, 2000, Salt Lake City, UT.
Received for review September 29, 2003. Revised manuscript received February 10, 2004. Accepted February 16, 2004. ES035071N
VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2927
