Environ. Sci. Technol. 2001, 35, 1959-1966
Determination of Silver Speciation in Natural Waters. 2. Binding Strength of Silver Ligands in Surface Freshwaters RUSSELL T. HERRIN, ANDERS W. ANDREN, MARTIN M. SHAFER, AND DAVID E. ARMSTRONG* Water Chemistry Program, University of WisconsinsMadison, 660 North Park Street, Madison, Wisconsin 53706
Three competing ligand methods were compared to determine characteristics of Ag(I) complexation by dissolved and colloidal ligands present in three rivers and one sewage treatment plant effluent. Iminodiacetate groups on Chelex-100 resin (used in batch and column experiments) and diethyldithiocarbamate (DDC) were used as competing ligands. Results of batch Chelex and DDC competition experiments show good agreement with regard to relative extent of Ag binding by natural ligands among the three river systems. Results of both methods also show a possible correlation between extent of Ag(I) complexation and organic matter concentration and/or Fe concentration. Fraction of Ag(I) associated with Chelex in both batch and column Chelex experiments was similar in each of the four systems tested, indicating that lability of Ag complexes does not change significantly on time scales ranging from seconds to 24 h. Results of Chelex and DDC competition were compared using a model based on a hypothetical single natural ligand. Under the experimental conditions used, this model quantified Ag(I) complexes with log Kcond values from approximately 12 to 14. For the three rivers studied, ligands with silver-association characteristics similar to those of reduced sulfur groups (log K ) 14-16) present at subnanomolar concentrations likely dominate Ag(I) speciation in these systems. A weaker ligand (e.g., log Kcond < 12) at concentrations >0.7 nM dominated Ag(I) speciation in the treatment plant effluent. This may result from elevated concentration of metals that compete for reduced sulfur groups rather than from a lower total concentration of these groups.
Introduction Compared to metals such as lead, copper, and mercury, concerns regarding environmental hazards related to silver (Ag) have been raised only recently. Silver is a relatively rare crustal element, present at concentrations of approximately 0.1 mg/kg (1). Silver is generally present in aquatic systems at subnanomolar levels. Even in a historic Ag mining area, subnanomolar concentrations were found in surface waters (2). Higher Ag concentrations in aquatic environments usually have their origin in waste streams from commercial or industrial processes. * Corresponding author e-mail:
[email protected]; phone: (608)262-0768; fax: (608)262-0454. 10.1021/es001510w CCC: $20.00 Published on Web 04/05/2001
2001 American Chemical Society
Silver is found in nature in the monovalent (Ag(I)) or metallic (Ag(0)) oxidation state. Studies have shown that free or hydrated Ag(I) exhibits toxic effects on a variety of aquatic organisms ranging from zooplankton (3) to rainbow trout (4). Concentrations of Ag(I) required to cause toxic effects can be quite low; phytoplankton grown in medium containing subnanomolar concentrations of Ag(I) caused toxic effects in zooplankton to which they were fed (3). Questions remain, however, as to the applicability of these results to natural waters where a range of complexing species exist, and it is possible that only a small fraction of Ag(I) is present as a free or hydrated ion. While it is unclear whether Ag-natural ligand complexes are in all cases less toxic or bioavailable than the free ion, complexation to ligands of different silver binding strengths and concentrations is likely to affect the toxicity of Ag(I). Differing affinities for particle surfaces are also likely to lead to species-dependent Ag fate and transport properties. Competing ligand techniques have been used to determine the effective binding strength and concentration of important complexers of Ag and other metals in surface waters and effluents (5-8). For the unique chemical properties of Ag(I), two specific competing ligand techniques that have shown promise are addition of diethyldithiocarbamate (DDC) followed by extraction of the Ag-DDC complex into chloroform (5, 6) and equilibration with Chelex resin followed by separation of Chelex-bound and aqueous-phase silver (9). We report the results of competing ligand experiments performed on three rivers and one effluent stream using both DDC and the functional group on Chelex resin as competing ligands. Results from the two competing ligands are compared to determine characteristics of the Ag ligands present in the four systems. Experimental parameters necessary for the investigation of silver speciation in natural waters are determined. Concentrations of species of potential importance to Ag(I) speciation are compared to competing ligand experiment results. Finally, the two types of competing ligand experiments are used to construct a simple model that allows for calculation of a single stability constant and concentration for natural ligands present in a given system.
Experimental Section Sampling Sites and Collection. Samples were collected from four aquatic systems: Black Earth Creek (BEC), near Madison, WI; the Madison Metropolitan Sewerage District effluent stream (MMSD), 5 mi from the outfall of the publicly owned treatment work (POTW); the Black River (BlkR), north of Black River Falls, WI; and the Mississippi River (MissR), north of LaCrosse, WI. All systems were sampled once during spring or summer 1999 at baseflow conditions, with the exception of BEC, which was sampled twice during springtime event flows in 1999. Sites were chosen to reflect a wide range in aquatic variables. BEC is a small stream, fed primarily by groundwater inflows, that is relatively low in dissolved and colloidal organic matter. Its aquifer contains large quantities of carbonate minerals. MMSD effluent consists of an undiluted stream of treated sanitary sewage water and stormwater. The plant is a secondary treatment facility that uses ultraviolet light irradiation as a disinfection method (10). The Black River is the northernmost river studied and, like many northern Wisconsin streams, has high concentrations of dissolved and colloidal organic matter, a high percentage of wetland area in its watershed, and a low conductivity as compared to southern Wisconsin streams. It is situated in a historic ironmining region and contains high concentrations of filterable Fe. The Mississippi River is unique among the streams tested VOL. 35, NO. 10, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1959
for its multiple uses; it serves as a shipping channel, a receiving stream for several POTWs, and a fishing and recreational boating resource. In addition, it flows through a major metropolitan area (Minneapolis-St. Paul, MN) upstream of the sampling site. As a result, it integrates many different types of anthropogenic point and nonpoint influences. Samples from these systems were collected from shore by peristaltic pump. Tubing was deployed from shore on a telescoping pole that allowed sampling to be performed approximately 6 m from shore. Within limits of this maximum distance, samples were collected as close to the center of the stream as possible and from approximately 15 cm below the surface. Sample weights, fittings, and tubing were constructed exclusively of Teflon, and pumphead tubing was C-Flex (ColeParmer, Inc.). Filtered samples were pumped through polypropylene cartridge filters with a 0.4-µm pore size rating (Meissner, Inc.). Filters, tubing, and other equipment that contacted samples were soaked for more than 24 h in mineral acids in a cleanroom environment. Bottles used for competing ligand experiments or concentration analysis of Ag and other trace metals were cleaned by a similar method (see refs 9 and 11). At least 1 L of streamwater was passed through the tubing and filter before samples were collected. Trace metal clean procedures, including double-bagging of bottles and equipment in clean polyethylene bags; clean hands-dirty hands procedures performed by workers wearing polyethylene gloves and new Tyvek suits; and elimination of metal sampling equipment were practiced during sampling (11, 12). Samples for use in competing ligand experiments and for determination of pH, conductivity, dissolved and colloidal organic carbon (DOC), a suite of trace element filterable concentrations, and sulfide concentration (13) were filtered on-site. Unfiltered samples were collected for determination of suspended particulate matter (>0.4 µm) and a suite of total trace metal concentrations. Filtered water collected for use in competing ligand experiments was composited in an acid-cleaned 2.5-L glass bottle and then divided among 500mL glass or Teflon bottles. Samples were stored at 4 °C until experiments were initiated. In all cases, competing ligand experiments were initiated within 24 h of sample collection. To avoid possible photoreduction of Ag(I) to Ag(0), exposure of samples to light was minimized during collection and storage. Samples for pH and conductivity determination were collected in polyethylene bottles and analyzed soon after collection. Dissolved organic carbon samples were collected in glass vials that had been ashed at 450 °C for 8 h and capped with Teflon septa that had been sonicated in ultrapure water. Samples for DOC analysis were stored frozen until analysis. Filtered and unfiltered samples for trace metal analysis were collected in exhaustively acid-cleaned Teflon bottles. Immediately after collection, concentrated HNO3 (Ultrex grade, J. T. Baker) was added such that concentration of HNO3 in the sample was 0.15 M (1% v/v concentrated HNO3). Samples for sulfide analysis were collected in 300-mL glass BOD bottles with an opaque polymer covering and stabilized with approximately 0.5 mL of 2 M zinc acetate and approximately 0.4 mL of 6 M sodium hydroxide. These samples were stored at 4 °C until preconcentration and analysis. Samples for SPM determination were collected in acid-washed polyethylene bottles and filtered through preweighed polycarbonate tracketched (PCTE) filters. If filtrations were not performed promptly after collection, the solution was stored at 4 °C until filtration could be performed. Competing Ligand Experiments. Three complementary procedures were performed to quantify the effective formation constant for silver complexation, the effective concentration, and the dissociation kinetics of Ag-natural ligand (AgL) complexes in the systems studied. Chelex resin, 1960
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 10, 2001
equilibrated with samples for 24 h in batch experiments, was used as a competing ligand of moderate strength with respect to silver complexation (log Kcond(Ag-Chelex) ) 7.5 for chelation of one Ag(I) ion by two Chelex-bound iminodiacetate groups; 9). Kinetic lability of Ag(I) complexes was studied by comparing 24-h batch Chelex experiments with column experiments in which resin contact time with a differential volume of sample was approximately 6 s. If a subset of Ag(I) complexes dissociated significantly more slowly than 6 s, less Ag(I) will associate with Chelex in column experiments than in batch experiments. Resin was cleaned with nitric acid solution and converted to the Ca2+ form (9). The diethylammonium salt of DDC was used as a strong competing ligand (log Kf(Ag-DDC) ) 9.1) (6). Batch and column Chelex experiments were performed on both unspiked samples and samples to which an aqueous AgNO3 standard was added. Sufficient standard was added to increase the silver concentration by 0.465 nM. For all systems except MMSD effluent, this spike represented a 2040-fold increase in filterable Ag concentrations. Unbuffered batch samples were allowed to equilibrate for 15 min with spiked Ag(I), after which 0.25 g of Chelex in the Ca2+ form was added to 500 mL of sample; this corresponds to 0.3 mM resin-bound iminodiacetate functional groups. The Ca2+ form rather than the Na+ form was chosen because the Ca2+ form is less likely to affect solution pH (9). Samples for batch and column experiments were tightly capped, triple-bagged, and placed on an incubator/shaker (model 236, Fisher Scientific) for 23-25 h at 25 °C and 100 rpm. Samples were kept above the water bath during shaking. After being mixed, Chelex in batch samples was separated from the solution phase using all-Teflon or all-glass resin columns. In all experiments in which glass sample bottles were used, glass resin columns were used. Solution-phase samples were collected for pH determination and for Ag concentration analysis. The latter were promptly acidified to 0.15 M with concentrated HNO3 (Ultrex grade, J. T. Baker). Silver chelated with the resin was eluted using 15 mL of 2 M HNO3 followed by 15 mL of ultrapure water. The Ag concentration corresponding to the fraction of Ag associated with Chelex can be accurately determined as a result of the concentration factor afforded by the resin elution procedure. This is the case even when only a small fraction of available Ag(I) in spiked samples associates with the resin. Second elutions and mass balances indicated that this elution technique quantitatively removed Ag from the resin. Samples (∼500 mL) for column experiments were passed through a column of 0.5 g of Chelex in the Ca2+ form. Sample collection and elution procedures were the same as for batch experiments. For a given system, two batch experiments were performed on sample solution spiked with Ag standard, and two were performed with no Ag standard. One or more unspiked and spiked column experiments were performed as well. For all systems tested, except MMSD effluent, Ag concentrations in all fractions of unspiked were negligible as compared with spiked values. As was the case during sample collection, exposure of samples to light was minimized during competing ligand experiments (14). Experiments such as those described above were also performed on samples that were collected in Teflon bottles and thoroughly oxidized using UV light and hydrogen peroxide. After being collected, filtered samples were placed in an opaque box and irradiated with high-voltage mercury vapor lamps for at least 24 h, after which 100-300 µL of a 3% solution of H2O2 (0.06-0.18 mM H2O2 in sample) was added, and the bottles were again exposed to UV light for 24 h or longer. Subsamples were collected for determination of organic carbon (OC) concentration, and H2O2 addition (0.06 mM) and irradiation were repeated if the OC concentration was greater than 1 mg/L. Competing ligand experiments were
initiated once OC concentrations were less than 1 mg/L. For UV-oxidized and unoxidized river water samples, all fractions of unspiked samples were below quantitation limits for silver. Procedures for experiments in which DDC was used as a competing ligand were similar to those of Miller and Bruland (5), although titration with Ag was not performed and separation of Ag from the organic phase was more similar to the technique of Bruland et al. (15). Two aliquots (250 mL) of filtered sample were added to separate 500-mL Teflon or glass separatory funnels with 1 mM DDC and an AgNO3 spike that increased [Ag]tot by 0.465 nM. Two similar samples were prepared in which no DDC was added. Samples were allowed to equilibrate for 15 min. A short equilibration time was used because DDC is hydrolyzed very quickly in aqueous solution at circumneutral pH. In addition, mass transport should be much faster when the ligand is not resin bound. Ten milliliters of chloroform (CHCl3) was added to each, and each funnel was shaken vigorously for 2 min. Organic and aqueous phases were allowed to separate for 15 min, after which the organic phase was drained into a 125-mL glass or Teflon separatory funnel to which 4 mL of 7.5 M HNO3 (Ultrex grade, J. T. Baker) was added. These funnels were shaken vigorously for 2 min to destroy the Ag-DDC complex and extract the silver into the acidic aqueous phase. Phases were allowed to separate for 15 min. The organic phase was discarded, and the acid solution was retained for dilution and Ag concentration analysis. A sample of the original aqueous phase was acidified and retained for Ag concentration analysis. Results of Benoit (16) have suggested that contamination concerns with regard to trace-level Ag may not be as important as for copper, lead, and other trace metals, although precautions must be taken to avoid Ag contamination. In addition to bottle-cleaning and clean sampling procedures, all experiments and silver analyses were performed in a trace-metal clean laboratory. This room is kept at a pressure higher than ambient atmospheric pressure with HEPA-filtered air, and room air is recycled through laminar flow hoods with HEPA filters. While particle counts are not determined frequently, procedure blanks and unspiked sample concentrations indicated that room air was clean enough to prevent sample contamination. Loss of Ag to sample containers may represent a more important concern than Ag contamination. Wen et al. (17) observed losses of over 50% of Ag spiked into river water during 2-month storage of unacidified samples in Teflon bottles. To assess sorption losses to bottles and other equipment in our experiments, Ag mass balances were routinely performed. Three Ag concentrations were determined in constructing mass balances: Ag remaining in the aqueous phase, Ag in Chelex eluate, and Ag associated with the bottle in which the experiment was performed. Bottle-associated Ag was determined after the experiment was complete by rinsing the sample bottle with Milli-Q and adding 20 mL of 1 M HNO3. The bottle was triple-bagged and placed on the incubator/ shaker for 12 h or more. The concentration of Ag in 1 M HNO3 was determined and considered to be bottle-associated Ag. Mass balances performed on unoxidized river water samples stored without acidification in Teflon indicated that unacceptable amounts of Ag were associated with bottle surfaces. For the three river systems, less than 70% of total Ag was recovered, and 90% of recovered Ag was associated with the bottle surface. These results are in good qualitative agreement with those of Wen et al. (17). Samples for copper analysis taken from highly sulfidic porewater systems have also shown poor recoveries when stored in Teflon or polyethylene bottles, probably as a result of sorption of copper sulfide species to bottle walls (18). It is possible that Ag was lost by a similar mechanism. Recoveries were much better in UV H2O2-oxidized samples and in unoxidized samples
stored in glass and separated in glass resin columns. In these cases, over 80% of Ag was recovered in the combined aqueous and eluate phases, and Ag concentrations in bottle rinses were below limits of detection (19, 20). As a result, all samples to be used unoxidized in Chelex experiments were stored and processed exclusively in glass until acidification was acceptable. Analytical Methods. Silver concentrations in samples from competing ligand equilibration experiments were analyzed by graphite furnace atomic absorption spectrometry with Zeeman background correction (GFAAS; Perkin-Elmer). Ammonium dihydrogen phosphate was added as a matrix modifier. Detection limits were improved by performing multiple depositions and drying steps before atomization. When sample concentrations were expected to be greater than 1 nM, two deposition steps were performed for each analysis. When concentrations were expected to be less than 1 nM, three depositions were performed. For two-deposition analyses, the limit of detection was 0.3 nM, and for threedeposition analyses, the limit of detection was 0.17 nM (based on 3 times the standard deviation of blanks, n g 19). Sample matrix effects were evaluated by performing spike recoveries. For two-deposition samples, spikes increased sample silver concentrations by 2.6 nM and mean recovery was 100.4% (s ) 4.9%, n ) 16). For three-deposition samples, spikes increased silver concentrations by only 0.13 nM; as a result, while mean recovery was acceptable (98%), recoveries were much more variable (s ) 36%, n ) 26). Spike concentrations were different because the spike was added to the sample on the L’vov platform (i.e., the autosampler added the Ag spike), and a standard of lower concentration was added by the autosampler for three-deposition analyses. This led to a lower spike concentration. In all analytical runs, duplicate or triplicate analyses were performed on a subset of samples. Triplicates generally agreed to within 10%. An exception to these acceptable analytical recoveries was noted in aqueous samples that had been extracted with CHCl3. Samples extracted in this manner provided very low and variable spike recoveries. Analyses of Ag back-extracted from CHCl3 into HNO3 solution however provided excellent recoveries. Results from these back-extracts were therefore used for modeling calculations. Analyses of filterable and total trace metal concentrations in sample solutions prior to spiking were performed by inductively coupled plasma mass spectrometry (ICP-MS; VG Plasmaquad II STE). Samples were prepared and analyses were performed in a clean room environment. Instrumental parameters and quality control measures were similar to those of Shafer et al. (10). Dissolved organic carbon concentrations were measured by high-temperature combustion (Shimadzu TOC-5000) following acidification and purging to remove inorganic carbon. Suspended particulate matter (SPM) concentrations were determined by filtering the sample solution through a preweighed, 0.4-µm PCTE filter, drying the filter for 24 h or more at 50 °C, and weighing them. Samples for sulfide analysis were preconcentrated by precipitation of Zn(OH)2-ZnS floc and redissolution in acid, similar to the method of Rozan et al. (21). Sulfide concentration was determined spectrophotometrically by the procedure of Cline (13). This method detects all available sulfide in the sample but does not detect organic thiols. Concentrations were corrected for absorbance blanks. A sulfide standard in ultrapure water was preconcentrated, redissolved, and analyzed with the samples. The concentration estimate for this spike was the same to two significant figures as its actual concentration. Matrix spike experiments performed on a lake water sample provided spike recoveries (spike ) 50 nM sulfide) from 84 to 120%. VOL. 35, NO. 10, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1961
TABLE 1. Selected Characteristics of Systems Studied systema
sampling date
pH
conductivity (µS cm-1)
DOC (mg L-1)
[Ag]
