Use of Diffusive Gradients in Thin Films To Measure Cadmium

EMILY R. UNSWORTH, HAO ZHANG, AND. WILLIAM DAVISON*. Institute of Natural and Environmental Sciences,. Environmental Science Department ...
0 downloads 0 Views 145KB Size
Environ. Sci. Technol. 2005, 39, 624-630

Use of Diffusive Gradients in Thin Films To Measure Cadmium Speciation in Solutions with Synthetic and Natural Ligands: Comparison with Model Predictions EMILY R. UNSWORTH, HAO ZHANG, AND WILLIAM DAVISON* Institute of Natural and Environmental Sciences, Environmental Science Department, Lancaster University, Lancaster, United Kingdom

The performance of the technique of diffusive gradients in thin films (DGT) was characterized in well-defined systems containing cadmium with chloride and nitrate ions, simple organic ligands (nitrilotriacetic acid and diglycolic acid), and Suwannee river fulvic acid for the pH range 5-8. Cd was fully labile in all Cd, Cl-, and NO3- solutions tested (I ) 0.1 and 0.01 M), even at very low Cd concentrations (10 nM), consistent with there being no binding of Cd to the diffusive gel. Diffusion coefficients of Cd-nitritotriacetic acid (NTA) and Cd-diglycolic acid (DGA) species were measured and found to be ca. 25-30% lower than the equivalent coefficient for free metal ions. These values were used to calculate concentrations of labile Cd from DGT measurements in solutions of Cd with NTA or DGA. Cd-NTA and Cd-DGA species were found to be fully DGT-labile. DGT devices that used a diffusive gel with a reduced pore size, which retarded the passage of fulvic acid species through the gel, were used to estimate the proportion of Cd complexed by fulvic acid. These results were compared with predictions of the solution speciation from models with default parameter values. ECOSAT, incorporating the NICA-Donnan model, correctly predicted the magnitude of the binding and its pH dependence, while predictions from WHAM V (with humic ion binding model V) and WHAM 6 (with humic ion binding model VI) were less satisfactory at predicting the pH dependence. Reasonable fits to the data could be obtained from WHAM 6 when the effective binding constant log KMA was changed from 1.6 to 1.5, the value of ∆LK1 from 2.8 to 1.0 to minimize the dependence on pH, and the value of ∆LK2 from 1.48 to1.0 to decrease the strength of the strong bidentate and tridentate binding sites.

Introduction The toxicity and bioavailability of trace metals depends on both their total concentration and their chemical speciation (1). Accurate assessment of speciation in natural waters relies on the use of measurement techniques that are sensitive to the chemical form of the metal and on models of chemical equilibrium. Both these complementary approaches require validation in the laboratory under well-controlled conditions * Corresponding author phone: +44-1524-593-935; fax: +44-1524593-985; e-mail: [email protected]. 624

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005

to provide confidence in their application to natural waters. Because sampling and storage can induce changes in the nature and distribution of species in samples of natural water (2), measurement of speciation in situ is desirable. The technique of diffusive gradients in thin films (DGT) has the potential for measuring trace metal speciation in situ (3), but laboratory validation is still required. The computer speciation models WHAM (4) (incorporating ion binding model VI) and ECOSAT (5) (incorporating the NICA-Donnan model) are able to predict metal binding to both inorganic species and humic substances. The parameters for describing the binding of metals to humic substances used in these models have been optimized by fitting data, obtained in the laboratory, for metal binding to isolated fulvic and humic acids. Use of these models to predict the distribution of species in natural waters relies on the use of their optimized parameters, which are provided as the default values when the models are run without modification. The models provide a coherent framework for bringing together laboratory data obtained under a wide range of conditions. They therefore provide a convenient conduit for comparing new measurements with those of previous workers whose results have been included in the weighted regression, to obtain the default parameter values. Attempts to test speciation models have largely focused on comparisons of the modeled free ion concentration with experimental measurements of the free metal ions. Benedetti et al. (6) were able to model (NICA-Donnan) reasonably well the Cu2+ concentrations in a lake. Cd and Zn speciation has been measured by the Donnan membrane technique and compared with that modeled (NICA-Donnan) by Oste et al. (7). The model slightly overestimated the concentration of metal bound to humic acid and hence underestimated the free ion concentration. Dwane and Tipping (8) compared Cu2+ concentrations calculated from WHAM V to values measured by an ion-selective electrode during titrations of natural waters by base. They demonstrated how results depended on the proportion of dissolved organic matter assumed to be active fulvic acid. By comparing results for Cd, Ni, and Zn competition between a resin and natural DOC to models based on WHAM V, Christensen and Christensen (9) concluded that WHAM V “overestimated the pH dependence of metal-DOC complexation”. Comparison of predictions made with a speciation model with results from a dynamic measurement technique, such as DGT or anodic stripping voltammetry, is not straightforward. Species that are labile (dissociate within the time window of a few minutes) and mobile (capable of diffusing sufficiently quickly through the gel layer) contribute to the measured mass, as previously discussed in detail (10). Speciation models generally predict the distribution of individual species at true equilibrium in an undisturbed sample. Assumptions must therefore be made on how each species in a system contributes to the DGT measurement. By varying the pore size of the diffusive gel, it is possible to alter the discrimination between species according to their mobilities. The effect of pH and ionic strength on the measurement of Cd in simple inorganic solutions by DGT has been reported (11, 12). Measurements by DGT of Cd in the presence of complexing ligands has only been studied at pH 7.8 (13). These measurements demonstrated that DGT discriminates between labile Cd-glycine and nonlabile Cd-EDTA species. Cd-fulvate complexes were shown to be labile and measured by DGT. By using gels with different pore sizes, the slower diffusion of larger fulvic molecules was exploited to dis10.1021/es0488636 CCC: $30.25

 2005 American Chemical Society Published on Web 12/13/2004

criminate between species. Cd has been measured by DGT in situ in freshwaters (14-16) and seawater (17). The DGT measurement generally accounted for a high proportion of the filterable metal, consistent with complexation being relatively weak. This paper investigates more systematically the measurement of Cd and its complexes by DGT in solutions containing inorganic, synthetic organic, and natural organic ligands for a range of pH in DGT devices having gels with different diffusional characteristics. Simple inorganic solutions containing well-characterized complexes (diglycolic acid and nitrilotriacetic acid) were used to test the performance of the devices. Measurements were then made in solutions containing fulvic acid in the pH range 5-8, and results were compared to predictions from speciation models with default parameters. Binding parameters of the models were adjusted to gain insight into possible reasons for discrepancies.

Experimental Section Reagents. Buffered (5 mM) synthetic water solutions were prepared in 2-(N-morpholino)ethanesulfonic acid (MES) and 3-(N-morpholino)propanesulfonic acid (MOPS) buffers (Biomedical, BDH, Poole, U.K,) with the addition of sodium hydroxide to adjust the pH (18). MES (Aristar, BDH) was used for pH 5-6.5, and MOPS was used for pH 7-8. The ligands added were chloride (sodium chloride, AnalaR, BDH); diglycolic acid (Aldrich, Gillingham, Dorset, England); nitrilotriacetic acid (Aldrich); and Suwannee river fulvic acid (IHSS, University of Minnesota). Magnesium nitrate (1 mM, Aristar, BDH) was added to counter adsorption of cadmium to the container walls at high pH, and sodium nitrate (AnalaR, BDH) was used to adjust the ionic strength where required. Solutions were prepared in purified water (18 MΩ, Milli-Q, Millipore SA, Molshelm, France). Cadmium (1000 ppm SpectrosoL standard solution, BDH) was added to give a concentration of 1 × 10-8 M except for the Cd-Cl solutions, where the cadmium concentration was varied from 1 × 10-8 to 1 × 10-7 M. Deployment Procedure. Normal diffusive gel was prepared as detailed by Zhang and Davison (3); restricted gel (with a smaller pore size) (19) was supplied from DGTResearch.com. The DGT units were deployed (in triplicate for each gel) in plastic containers holding 2 L of continuously stirred solution. The solution temperature was measured at the start and end of the deployment period. If a difference was noted, the mean result was used to temperature-correct the diffusion coefficient. The DGT units were deployed in solution for 4 h, except for the fulvic acid solutions where the deployment time was 18 h. Some DGT units, to be used as blanks, were not deployed. During the deployment, aliquots of solution were sampled. After deployment the gel layers were separated and the resin-gel layer was placed in a 1.5 mL plastic tube. One milliliter of 1 M nitric acid (Aristar, BDH) was added, and the tube shaken and left overnight to extract the metal from the Chelex-100. Samples were analyzed by inductively coupled plasma mass spectrometry (ICP-MS) (Thermo X-7, Thermo Elemental, Cheshire, U.K.) with rhodium (SpectrosoL standard solution, BDH) as an internal standard. Concentrations measured by DGT (and blank-corrected) were calculated by the well-documented procedures and equations given elsewhere (3) with an elution factor of 0.8. Determination of Diffusion Coefficients. Diffusion coefficients were experimentally determined following the procedure outlined by Zhang and Davison (19). The diffusion cell used had two (50 mL) reservoirs separated by a disk of diffusive gel. At the start of the experiment (t0), one reservoir contained a high concentration (10-5 M) of the metal to be measured and the other reservoir had a zero concentration. The metal diffused from the high concentration side to the

low concentration side. The reservoirs were constantly stirred to minimize formation of a diffusive boundary layer. Previous work (20) has demonstrated that this is negligible for a 0.8 mm gel under these conditions. Aliquots were taken from the low concentration side at 10 min intervals and the concentration of metal was measured. Aliquots were also taken from the high concentration reservoir to maintain the same solution volume in each reservoir and to provide the total metal concentration. The diffusion coefficient was calculated by use of

D ) S∆g /ACT

(1)

S is the slope from a plot of mass of metal (nanograms) versus time (seconds), ∆g is the gel thickness (centimeters), A is the gel area (square centimeters), and CT is the mean metal concentration (nanograms per milliliter) measured in the reservoir with high metal concentration.

Results and Discussion Diffusion Coefficients. Plots of metal mass (nanograms) versus time (seconds) resulted in straight lines with R2 values of typically 0.999. The diffusion coefficient was calculated from the measured slope by use of eq 1. A potential source of error in this procedure is the determination of the concentration of metal in the diffusion cell reservoirs, but uncertainty was minimized by analyzing samples from both reservoirs (high concentration side was diluted) in the same ICP-MS run. The repeatability of the ICP-MS analysis was estimated from QC standards to be 0.5%. Duplicate determinations of the diffusion coefficients gave values of (3.99 ( 0.05) × 10-6 cm2‚s-1 for Cd-NTA at 22 °C and (4.72 ( 0.30) × 10-6 cm2‚s-1at 25.5 °C for Cd-DGA, 70% and 76%, respectively, of the free metal ion diffusion coefficient for Cd in simple inorganic solutions in the same gel (21). According to calculations made with ECOSAT in the solution containing NTA, 100% of the Cd was complexed, while in the solution containing DGA only 83% of the Cd was present as Cd-DGA species. The measured diffusion coefficient was therefore due to a combination of the diffusion coefficients of Cd and Cd-DGA species. By assuming the ratio of Cd/Cd-DGA remains constant throughout the diffusion layer, the diffusion coefficient of Cd-DGA can be calculated from simple proportionality:

DCd-DGA ) (Dm - (DCdfCd))/fCd-DGA

(2)

DCd-DGA is the diffusion coefficient of the Cd-DGA species and Dm is the measured diffusion coefficient. DCd is the diffusion coefficient of Cd free metal ions (6.175 × 10-6 cm2 s-1 at 25.5 °C). fCd is the fraction of Cd present as inorganic species (0.17), and fCd-DGA is the fraction of Cd complexed by DGA (0.83). DGT Deployments in Inorganic Solutions. The mass of metal accumulated by DGT deployed in inorganic solutions was converted to concentration by use of the previously published equation (3) and values of diffusion coefficients supplied by DGT Research Ltd (21).The cadmium concentration measured by DGT was plotted against the total cadmium concentration measured in the deployment solution (Figure 1). DGT measured all the cadmium present in solution over a range of cadmium concentrations (1 × 10-8 to 1 × 10-7 M), at intermediate (0.01) and high (0.1) ionic strengths, and at different chloride concentrations (1 × 10-2 and 5 × 10-2 M). Model predictions (from ECOSAT; see Table 1 for model input values) of the cadmium speciation in these solutions (Figure 2) indicate that DGT measured the free metal Cd2+ as well as the cadmium chloride (CdCl+, CdCl2(aq)) and cadmium nitrate (CdNO3+) species, all of which can therefore be termed DGT-labile. Any appreciable interactions VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

625

TABLE 1. Input Values for the Computer Speciation Models of the Cd-Cl System component

solution 1

solution 2

solution 3

Cd, mol L-1 Cl, mol L-1 Na, mol L-1 NO3, mol L-1 MES, mol L-1 pH pCO2, bar temperature, K

1 × 10-8 1 × 10-2 1.27 × 10-2 0 5 × 10-3 fixed at 6 3.55 × 10-4 298.15

1 × 10-8 1 × 10-2 1 × 10-1 8.78 × 10-2 5 × 10-3 fixed at 6 3.55 × 10-4 298.15

1 × 10-8 5 × 10-2 1 × 10-1 3.73 × 10-2 5 × 10-3 fixed at 6 3.55 × 10-4 298.15

experimentally determined diffusion coefficient for Cd-NTA. Had the diffusion coefficient of Cd free metal been used in these calculations, the concentration would have been lower, due to the lower diffusion coefficient of Cd-NTA. In contrast, even at the highest DGA concentration, the Cd-DGA species were calculated to comprise less than 50% of the total Cd species in solution. Therefore the diffusion coefficient of the free metal was used in the DGT calculations. The contribution of the Cd-DGA species would lower the combined diffusion coefficient (of the free Cd and Cd-DGA species) by less than 10%. The results for a ligand concentration of 10-5 M (mean ratio of DGT-measured concentration to the directly measured concentration for Cd-DGA of 0.99 ( 0.09 and for CdNTA of 0.98 ( 0.15) indicate that both Cd-DGA and CdNTA complexes are DGT-labile. There was no significant modification of lability due to pH (5-8) or ligand concentration (10-6-10-4 M). FIGURE 1. Solution Cd concentration calculated from DGT versus the concentration measured directly in solution for solutions with different Cd concentrations and ionic strengths. Error bars are the standard deviation of replicates (n ) 3) Solution 1 (b), I ) 0.01, [Cl] 1 × 10-2 M; solution 2 (O), I ) 0.1, [Cl] 1 × 10-2 M; solution 3 (4), I ) 0.1, [Cl] 5 × 10-3 M.

FIGURE 2. Computer model predictions of the species distribution in the solutions used in Figure 1. of cadmium with the diffusive gel would be expected to lower the diffusion coefficient. The effect would most likely be seen at low cadmium concentrations. However, all cadmium present in solution was measured by DGT, even at the lowest cadmium concentration of 1 × 10-8 M. DGT Deployments in Solutions with Simple Organic Acids. To investigate the lability of simple organic Cd complexes, DGT was deployed in solutions containing diglycolic acid (DGA) and nitrilotriacetic acid (NTA) as ligands. These represent small (compared to humic substances), simple organic acids, which form relatively weak (DGA) and strong (NTA) complexes with Cd, with known stability constants. Measurements were made for a range of pH and ligand concentrations. In all the solutions used, CdNTA complexes accounted for >99% of the cadmium, so the Cd-NTA concentrations were calculated by use of the 626

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005

Lability depends on whether the complex has time to dissociate as it traverses the diffusion layer (10, 22). The rate of dissociation of metal from the complex is governed by the stability constant of the metal complex and the rate of dissociation of water molecules according to the Eigen mechanism (23). As water exchanges very quickly in the hydration shell of Cd (24), the rate of complex dissociation is fast and so Cd complexes with DGA and NTA, which do not have exceptionally high stability constants, would be expected to be labile. DGT Deployments in Solutions Containing Fulvic Acid. The normal diffusive gel allows the passage of both inorganic and fulvic acid species, but measurement of the latter is ca. 5 times less sensitive due to their diffusion coefficient in the gel being ca. 5 times less than that of simple inorganic species (19). The restricted gel retards the passage of fulvic acid species more severely. The diffusion coefficient of the Cdfulvic complex is about 12 times smaller than that of simple inorganic species. When organic species do not dominate the solution composition, the cadmium measured with the restricted gel by use of the diffusion coefficient for inorganic species can be approximated to the concentration of inorganic (or nonhumic) species present (13). The proportion of Cd measured by DGT in the restricted gel for a range of pH values and with two fulvic acid concentrations are shown in Figures 3 and 4. The ratio of the DGT-measured concentration to the total metal concentration measured by ICPMS was between 0.68 and 0.83 for the lower concentration of fulvic acid (Figure 4), showing that fulvic complexes accounted for between at least 17% and 32% of the Cd present. At the higher fulvic acid concentration (Figure 3) the proportion of cadmium measured by DGT was lower, consistent with greater complexation with fulvic acid. The decrease in the DGT measured Cd with increasing solution pH, indicated increased formation of cadmium-fulvic acid species at higher pH values. The solution pH was maintained with either MES or MOPS buffer. These are unlikely to complex with Cd (25). They may potentially interact with the fulvic acid (26), but only at pH values above the buffer pKa. According to previous data (26), there may potentially be an

TABLE 2. Input Values for the Computer Speciation Models of the Cd-Fulvic Acid System component Cd, 10-8 mol L-1 Na, 10-3 mol L-1 Mg, 10-3 mol L-1 NO3, 10-3 mol L-1 MES, 10-3 mol L-1 MOPS, 10-3 mol L-1 fulvic acid, g L-1 pH, fixed pCO2, 10-4 bar temperature, K

1 0.30 1 2 5 0 0.012 5 3.55 298

1 0.90 1 2 5 0 0.012 5.5 3.55 298

FIGURE 3. Computer-modeled speciation (% inorganic species present) of the Cd-fulvic acid system [12 ppm fulvic acid, 10-8 M Cd, 5 × 10-3 M MES or MOPS buffer, and 10-3 M Mg(NO3)2] and experimental restricted gel DGT results (ratio of DGT concentration/ total metal concentration × 100). (Upper dashed line) WHAM V; (lower dashed line) WHAM 6; (s) ECOSAT NICA-Donnan; (b) DGT results; (O) corrected DGT result.

FIGURE 4. Computer-modeled speciation (% inorganic species present) of the Cd-fulvic acid system [2 ppm fulvic acid, 10-8 M Cd, 5 × 10-3 M MES or MOPS buffer, 10-3 M Mg(NO3)2] and experimental restricted gel DGT results (ratio of DGT concentration/ total metal concentration × 100). (---) WHAM V; (s s) WHAM 6; (- ‚ -) optimized WHAM 6 (with log K at 1.5, ∆LK1 at 1, and ∆LK2 at 1); (s) ECOSAT NICA-Donnan; (b) DGT result. effect in the solutions used in this work at pH 8, but the effect would be marginal at pH 6.5 and 7.5 and nonexistent at pH 7 and lower pH. The DGT results showed no discontinuity across the pH range, indicating that there is no effect that can be attributed to binding by the buffer.

1 2.07 1 2 5 0 0.012 6 3.55 298

1 3.46 1 2 5 0 0.012 6.5 3.55 298

1 1.90 1 2 0 5 0.012 7 3.55 298

1 3.33 1 2 0 5 0.012 7.5 3.55 298

1 4.32 1 2 0 5 0.012 8 3.55 298

Predicted Speciation Compared to DGT Measurements. Cadmium speciation in the presence of fulvic acid was modeled (see Table 2 for model input values) from both readily available versions of WHAM: version 1, incorporating model V (27), referred to here as WHAM V, and version 6, incorporating model VI (4), referred to here as WHAM 6. They both contain an inorganic speciation code coupled to a humic model for ion binding to humic or fulvic acid that is based on discrete sites and electrostatic effects. The main difference between the two is the addition of a small component of strong binding sites to Humic Ion Binding Model VI. This approach provides a better description of metals that bind strongly to humic substances, such as Cu. A further difference between model V and VI is that the proton and metal interaction parallelism has been relaxed in model VI (4). In model V metal binding was represented by the equilibrium constant KMHA for metal-proton exchange at a type A site, whereas in model VI the equilibrium constant KMA for metal binding to a deprotonated site is used. Model VI is still a discrete site model but takes a more distributional approach in describing binding sites for metals. Two extra terms are added to the metal-humic substance binding databases in model VI. These are ∆LK1, which is a distribution term that modifies log KMA, and ∆LK2, which is a term that modifies the strengths of bidentate and tridentate (strong binding) sites. One further difference between model V and VI is that the values for the equilibrium constants in the model VI databases were updated on the basis of experimental data that were not available at the time of producing model V. WHAM V predicts a low percentage of Cd-fulvic acid species at low pH that starts to increase substantially above pH 6.5 to reach ∼50% at pH 8 (Figure 3). However, WHAM 6 predicts ∼50% Cd-fulvic acid species at pH 5 and for complexation to increase to near 100% by pH 8. The percent inorganic species measured directly by DGT with the restricted gel falls between the two sets of model predictions, with WHAM 6 being closer at low pH and WHAM V being closer at high pH (Figure 3). The difference between the two model predictions is most likely attributable to the addition of extra parameters to define metal-humic substance binding in WHAM 6 and the different databases used to obtain the default binding parameters used in each program. A third set of model predictions were carried out by use of the ECOSAT (5) program. This couples an inorganic speciation code to the NICA-Donnan humic binding model (28). It uses a function for a continuous distribution of binding strengths about a mean value, rather than the discrete site approach used in WHAM. The electrostatic component is linked to a Donnan equilibrium (6). The ECOSAT results lie between those of WHAM V and WHAM 6 and are closer to the experimental results. To avoid problems associated with inconsistent databases (29, 30), NIST inorganic thermodynamic data (log K values) were taken from MINTEQA2 (version 4) documentation (31) and inserted into the databases of WHAM V, WHAM 6, and ECOSAT. Metal-DGA VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

627

FIGURE 5. Model fits (with parameter adjustments) to the experimental rDGT data shown in Figure 3. (a) Effect of changing the Cd log KMA value from 1.6 to 1.4 (with 1.6 being the default value provided in the WHAM 6 database). (b) Effect of changing the Cd ∆LK1 value from 0 to 4 (with 2.8 being the default value provided in the WHAM 6 database). (c) Effect of changing the Cd ∆LK2 value from 1.48 to 0 (with 1.48 being the default value provided in the WHAM 6 database). (d) Optimized WHAM 6 model fit (with log K at 1.5, ∆LK1 at 1, and ∆LK2 at 1) and the ECOSAT model with default values. thermodynamic data (32) and MES and MOPS thermodynamic data (18) (adjusted to zero ionic strength by use of the Davies equation), was also added into each of the model databases. Metal-fulvic acid binding data were used as provided with each program. Results from DGT with restricted gels will inevitably slightly overestimate the percentage of inorganic species in the presence of high concentrations of organic species, due to the diffusion coefficient of fulvic acid in the restricted gel being ca. 8% of the diffusion coefficient of the free metal (19). Consequently, the actual level of inorganic Cd species would be lower than that calculated from the DGT experiments. The error will be most marked when most of the Cd is present as fulvic acid species. This effect would be seen to the greatest extent at high pH, where 8% of the organic species concentration is a substantial amount compared to the concentration of inorganic species:

corrected finorg ) finorg - 0.08forg

(3)

Equation 3 was used to calculate a corrected fraction of Cd present in inorganic form from initial values of the fraction present in inorganic form (finorg) and the fraction present as organic complexes (forg) estimated from the measurements with the restricted gel. An iterative calculation was necessary to allow for forg progressively approaching the true answer. This calculation has assumed that all Cd-fulvic acid complexes are labile, which is likely to be the case (33). If even a fraction of them were nonlabile, the correction would be even smaller. As the effect on the estimate of finorg is small (Figure 3), the correction does not affect the conclusions regarding the model fits. 628

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005

In a second system, that had a lower concentration of fulvic acid (2 ppm), the proportion of Cd present as fulvic acid species was lower (Figure 4). Model predictions of the distribution of Cd species in this system from WHAM 6 were close to the experimental results for the solutions at pH 5 and pH 6. Above pH 6, both WHAM V and ECOSAT predicted only a slight increase in the proportion of Cd-fulvic acid species, as observed experimentally. By contrast, WHAM 6 predicted a substantial increase in the proportion of Cdfulvic acid species present. ECOSAT again provided the best overall fit to the data. The good agreement between DGT measurements and the NICA-Donnan model provides confidence in the DGT measurements, the model formulation, and the default parameters derived from model fits to previously measured data for Cd fulvic systems. While predictions by WHAM 6 of the proportion of Cd complexed by fulvic acid were reasonable (within 30% of the measurement), the dependence on pH was not as well predicted as for the NICA-Donnan model. To investigate the possible causes of the discrepancy, the effect on the model predictions of modifying the default metal-humic binding parameters was tested. The log KMA (FA) value for Cd2+ in the WHAM 6 database was varied between 1.6 (the default value provided in the program database) and 1.4. A standard deviation of 0.1 was quoted (4) for the spread of data used to produce the log K data provided in WHAM 6. Simulations were run with these log KMA (FA) values and the results are summarized in Figure 5a. Decreasing log KMA (FA) brought the predicted proportion of inorganic species closer to the experimental results, but the pH dependence of the model predictions was still substan-

tially greater than found experimentally. The log KMA (FA) value of 1.5 produced a close match to the experimental results at pH 5 and 6, but the model predictions were low at higher pH values. Decreasing the log KMA further to 1.4 produced model predictions that were higher than the experimental results except at pH 7 and above. The WHAM 6 model has two further metal-humic binding parameters: ∆LK1 and ∆LK2. The ∆LK1 parameter determines the spread of values about the median log K value and “principally influences the pH dependence of metal binding” (4). The effect of changing this value from 0 to 4 (2.8 being the default value provided in the program database) on the percentage of Cd complexed by fulvic acid, as a function of pH, is shown in Figure 5b. The percentage of Cd-organic species decreases and the slope increases as the ∆LK1 value is increased from 1 to 2. Above this value the trend changes, and although the percentage of Cd-organic species increases, the slope remains the same. All these changes affect the results at the low end of the pH range more than those at the high end of the pH range. The shallower slope observed when ∆LK1 ) 1 is closer to the slope of the experimental results, but the overall percentage of organic species predicted is too high. The third parameter that can be adjusted is ∆LK2, which modifies the strengths of bidentate and tridentate sites. Decreasing this value would be expected to increase the proportion of inorganic species. A higher percentage of inorganic metal species was indeed observed (Figure 5c), but this effect was much more pronounced below pH 7. The optimized model fit to the experimental data was obtained by slightly lowering the binding constant log KMA ) 1.5, lowering the value of ∆LK1 from 2.8 to 1 to minimize the dependence on pH, and lowering the value of ∆LK2 from 1.48 to 1 to increase the proportion of inorganic species (Figure 5d). When these values were used, the fit of WHAM 6 to the data in Figure 5 was also improved. This finding suggests that both the ∆LK1 value (pH dependence) and ∆LK2 value (affecting strong binding sites) currently selected as default values in WHAM 6 should be lower for modeling Cd binding to Suwannee River fulvic acid. For the Cd-fulvic acid systems used in this study the ECOSAT program produced the closest predictions to the experimental results, but whether this would be the case for other metal-organic systems needs further investigation. The different predictions obtained with the models used here could be due to a need for model calibration (to the particular system being studied) as well as the different approaches to modeling metal-humic interactions. Implications for Measurements in Natural Waters. When DGT is used in natural waters to measure Cd, precise interpretation in terms of species present is complicated because the ligands are unknown. However, the results from this work provide useful information to aid the interpretation. Simple inorganic species will be measured quantitatively, as they are labile and have diffusion coefficients similar to that of the free ion. If there are complexes with natural organic acids and even quite strongly complexing contaminant chelates, such as NTA, they will be also measured, as they too are labile. Their sensitivity would be a little lower than that of the free metal ion due to their slightly reduced rate of diffusion through the gel. As such complexes are unlikely to be the dominant species in solution, the error in the estimation of the total labile Cd introduced by disregarding the difference in diffusion coefficients will be small. In freshwaters fulvic acids usually dominate the dissolved organic matter and the complexation of metal ions (34). These complexes are sufficiently labile to be measured by DGT, but their diffusion in the normal gel is about 5 times less than that of the free metal ion, and so the DGT signal due to these complexes is substantially reduced compared to

simple inorganic solutions. Without a priori knowledge of the proportion of Cd complexed by fulvic acid, their contribution to the DGT measurement cannot be deduced. One way around this problem is to perform measurements on the restricted gel in the DGT device. This work has shown that the diffusion of fulvic complexes in the gel is sufficiently low that it is a good approximation to ignore their contribution to the DGT measurement. This approach is appropriate for Cd, as complexation by fulvic acid is unlikely to dominate the speciation. For this work the complexes accounted for 30-40% of the total dissolved Cd between pH 6.5 and 8 when fulvic acid was 12 mg L-1 and Cd was 10 nmol L-1. As Cd concentrations can be as low as 1 nmol L-1, a much greater degree of complexation might be expected, but this is more than offset by the presence of more strongly binding metals (Al, Fe, Cu) at much higher concentrations. Thus, Munksgaard and Parry (17) found in estuarine waters that labile Cd accounted for almost 100% of the total dissolved Cd, while in a pristine river water a value of 90% was reported (16). These measurements with fairly open-pored gels suggest that neither complexation by fulvic acid nor the presence of inert forms of Cd [complexes with very strong binding ligands, such as EDTA (nonlabile) and Cd attached to large colloids (nonmobile)] was important.

Acknowledgments We acknowledge the European Union (EU) for funding this work as part of the BIOSPEC project and Steve Lofts (Centre for Hydrology and Ecology) for technical help with the WHAM 6 program.

Literature Cited (1) Ure, A. M., Davidson, C. M., Eds. Chemical Speciation in the Environment; Blackie (Chapman and Hall): Glasgow, Scotland, 1995. (2) Reeve, R. N. Environmental Analysis; (ACOL) Wiley: Chichester, England, 1994. (3) Zhang, H.; Davison, W. Performance characteristics of diffusion gradients in thin films for the in situ measurement of trace metals in aqueous solutions. Anal. Chem. 1995, 67, 3391-3400. (4) Tipping, E. Humic ion-binding model VI: An improved description of the interactions of protons and metal ions with humic substances. Aquat. Geochem. 1998, 4, 3-48. (5) Keizer, M. G.; van Riemsdijk, W. H. ECOSAT; Department of Soil Science and Plant Nutrition, Wageningen Agricultural University, Wageningen, The Netherlands, 2002. (6) Benedetti, M. F.; van Riemsdijk, W. H.; Koopal, L. K.; Kinniburgh, D. G.; Gooddy, D. C.; Milne, C. J. Metal ion binding by natural organic matter: From the model to the field. Geochim. Cosmochim. Acta 1996, 60 (14), 2503-2513. (7) Oste, L. A.; Temminghoff, E. J. M.; Lexmond, T. M.; van Riemsdijk, W. H. Measuring and modeling zinc and cadmium binding by humic acid. Anal. Chem. 2002, 74, 856-862. (8) Dwane, G. C.; Tipping, E. Testing a humic speciation model by titration of copper-amended natural waters. Environ. Int. 1998, 24 (5-6), 609-616. (9) Christensen, J. B.; Christensen, T. H. The effect of pH on the complexation of Cd, Ni and Zn by dissolved organic carbon from laechate-polluted groundwater. Water Res. 2000, 34 (15), 3743-3754. (10) Scally, S.; Davison, W.; Zhang, H. In situ measurements of dissociation kinetics and labilities of metal complexes in solution using DGT. Environ. Sci. Technol. 2003, 37, 1379-1384. (11) Gimple, J.; Zhang, H.; Hutchinson, W.; Davison, W. Effect of solution composition, flow and deployment time on the measurements of trace metals by the diffusive gradients in thin films technique. Anal. Chim. Acta. 2001, 448, 93-103. (12) Peters, A. J.; Zhang, H.; Davison, W. Performance of the diffusive gradients in thin films technique for measurement of trace metals in low ionic strength freashwaters. Anal. Chim. Acta. 2003, 478, 237-244. (13) Zhang, H.; Davison, W. Direct in situ measurements of labile inorganic and organically bound metal species in synthetic solutions and natural waters using diffusive gradients in thin films. Anal. Chem. 2000, 72 (18), 4447-4457. VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

629

(14) Odzak, N.; Kistler, D.; Xue, H.; Sigg, L. In situ trace metal speciation in a eutrophic lake using the technique of diffusion gradients in thin films (DGT). Aquat. Sci. 2002, 64, 292-299. (15) Sangi, M. R.; Halstead, M. J.; Hunter, K. A. Use of the diffusion gradient thin film method to measure trace metals in freash waters at low ionic strength. Anal. Chim. Acta. 2002, 456, 241251. (16) Denney, S.; Sherwood, J.; Leydon, J. In situ measurements of labile Cu, Cd and Mn in river waters using DGT. Sci. Total Environ. 1999, 239, 71-80. (17) Munksgaard, N. C.; Parry, D. L. Monitoring of labile metals in turbid costal seawater using diffusive gradients in thin-films. J. Environ. Monit. 2003, 5, 145-149. (18) Perrin, D. D.; Dempsey, B. Buffers for pH and Metal Ion Control; Chapman and Hall: London, 1994. (19) Zhang, H.; Davison, W. Diffusional characteristics of hydrogels used in DGT and DET techniques. Anal. Chim. Acta 1999, 398, 329-340. (20) Scally, S. Ph.D. Thesis, Lancaster University, Lancaster, U.K., 2004. (21) DGT Research Ltd., http://www.dgtresearch.com. (22) Tusseau-Vuillemin, M. H.; Gilbin, R.; Taillefert, M. A dynamic numerical model to characterize labile metal complexes collected with diffusion gradient in thin films devices. Environ. Sci. Technol. 2003, 37, 1645-1652. (23) Eigen, M.; Wilkins, R. G. The kinetics and mechanism of formation of metal complexes. In Mechanisms of Inorganic Reactions; Advances in Chemistry Series 49; Gould R. F, Ed.; American Chenical Society: Washington, DC, 1965. (24) Morel, F. M.; Hering, J. G. Principles and Applications of Aquatic Chemistry; John Wiley and Sons: New York, 1993. (25) Soares, H. M. V. M.; Conde, P. C. F. L.; Almeida, A. A. N.; Vasconcelos M. T. S. D. Evaluation of n-substituted aminosulfonic acids pH buffers with morpholinic ring for cadmium and lead speciation studies by electrochemical techniques. Anal. Chim. Acta 1999, 394, 325-335.

630

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005

(26) Hosse, M.; Wilkinson, K. Determination of electrophoretic mobilities and hydrodynamic radii of three humic substances as a function of pH and ionic strength. Environ. Sci. Technol. 2001, 35, 4301-4306. (27) Tipping, E. WHAMsA chemical equilibrium model and computer code for waters, sediments and soils incorporating a discreet site/electrostatic model of ion-binding by humic substances. Comput. Geosci. 1994, 20, 973-1023. (28) Benedetti, M. F.; Milne, C. J.; Kinniburgh, D. G.; Van Riemsdijk, W. H.; Koopal, L. K. Metal ion binding to humic substances: Application of the nonideal competitive adsorption model. Environ. Sci. Technol. 1995, 29, 446-457. (29) Unsworth, E. R.; Jones, P.; Hill, S. J. The effect of thermodynamic data on computer model predictions of uranium speciation in natural water systems. J. Environ. Monit. 2002, 4, 528-532. (30) Twiss, M. R.; Erre´calde, O.; Fortin, C.; Campbell, P. G. C.; Jumarie, C.; Denizeau, F.; Berkelaar, E.; Hale, B.; van Rees, K. Coupling the use of computer chemical speciation models and culture techniques in laboratory investigations of trace metal toxicity. Chem. Spec. Bioavail. 2001, 13, 9-24. (31) Allison, J. D.; Brown, D. S.; Novo-Gradac, K. J. MINTEQA2/ PRODEFA2 User’s Manual Version 4 Supplement; U.S. Environmental Protection Agency: Washington, DC, 1999. (32) Powell, K. J., IUPAC Stability Constants Database (SCD), v 4.04; Academic Software, 1999. (33) Buffle, J. Complexation Reactions in Aquatic Systems: An Analytical Approach; Ellis Horwood: Chichester, U.K., 1988. (34) Tipping, E. Cation Binding by Humic Substances; Cambridge University Press: New York, 2002.

Received for review July 22, 2004. Revised manuscript received October 11, 2004. Accepted October 28, 2004. ES0488636