Anal. Chem. 2000, 72, 4447-4457
Direct In Situ Measurements of Labile Inorganic and Organically Bound Metal Species in Synthetic Solutions and Natural Waters Using Diffusive Gradients in Thin Films Hao Zhang* and William Davison
Environmental Sciences, IENS, Lancaster University, Lancaster LA1 4YQ, U.K.
The emerging technique of DGT (diffusive gradients in thin films) is shown to be capable of performing new speciation measurements in situ in natural waters. In DGT, metals are bound to a resin layer after passing through a well-defined diffusion layer. Cd was measured in solutions containing glycine, EDTA, and fulvic (FA) and humic acids (HA) by atomic absorption spectroscopy (AAS), anodic stripping voltammetry (ASV), and DGT. DGT measured similar labile fractions to ASV, with detailed differences being consistent with a thicker diffusion layer allowing more dissociation of labile complexes and a slower diffusion of FA and HA complexes through the gel. When single measurements are made in complex solutions with DGT, precise quantification is impossible due to uncertainties concerning the distribution of species with different diffusion coefficients. A new procedure was proposed based on the advantage of DGT of being able to control the pore size of the diffusive gel layer. Small (inorganic) species diffuse freely through all gels but larger FA and HA (organic) complexes diffuse less freely in more constrained gels. When measurements were made on known solutions of Cu and FA or HA, it was possible to quantify the inorganic and organic species separately. They agreed well with predictions made using the WHAM speciation code. Multiple DGT units were also deployed in situ in a stream with high dissolved organic carbon (14.6 mg/L). The systematic differences between the devices with different gel compositions enabled determination, for the first time, of the in situ concentrations of both labile inorganic and organic species in natural water. A single DGT device with a constrained gel can be used to quantify inorganic species directly, providing absolute accuracy is not required. This ability of DGT to measure well-defined fractions of metals in situ using a simple device gives it considerable potential as a regulatory tool. A direct speciation measurement may be preferable to modeling approaches which require diverse input data that are difficult to determine. Most trace metals are essential to life, but excess concentrations adversely affect water quality due to their toxicity. Whether * Corresponding author: lancaster.ac.uk
(fax) +44 1524 593985; (e-mail) h.zhang@
10.1021/ac0004097 CCC: $19.00 Published on Web 08/17/2000
© 2000 American Chemical Society
acting as toxicants or nutrients, the bioavailability of trace metals is related to their precise chemical form, which depends on their concentration, the solution pH, and the concentrations of ligands, including dissolved organic carbon (DOC).1 Depending on the relative importance of the transfer kinetics across the cell membrane and mass transport and dissociation kinetics in solution, the biological response of an organism may be related to the free-ion activity of a metal ion or to the concentration of labile metal species in solution.2,3 It is important, therefore, to have techniques that can measure both these quantities, as different metal ions are likely to exhibit different behavior due to their kinetic characteristics. The success of the free-ion activity model in predicting biological response for selected metal ions has focused attention on direct determination of free-ion activities using ion selective electrodes and competitive binding techniques, including the use of cathodic stripping voltammetry and exchange resins.4-8 Providing the stability constants of all labile metal species are known, along with the total concentrations of all complexants, it is theoretically possible to calculate free-ion activity from total labile species using speciation codes. If it was possible to measure the total labile inorganic species, calculation of free-ion activities would be quite accurate. Ultrafiltration and/or dialysis offer this possibility, but the distinction is inexact, and both techniques have problems when using membranes with the necessary small pore sizes.9 Moreover, it is impossible to use conventional filtration to fractionate all labile species, including organic complexes, due to the inclusion of colloids and inert metal complexes. Even if such fractionation were possible, precise characterization of the organic components and their metal binding properties is still a problem. (1) Tessier, A.; Buffle, J.; Campbell, P. G. C. In Chemical and Biological Regulation of Aquatic Systems; Buffle, J., DeVitre, R. R., Eds.; Lewis: Boca Raton, FL, 1994; pp 197-230. (2) Morel, F. M. M.; Hering, J. G. Principles and Applications of Aquatic Chemistry; Wiley: NewYork, 1993. (3) Hudson, R. J. M.; Morel, F. M. M. Limnol. Oceanogr. 1990, 35, 1002-1020. (4) van den Berg, C. M. G. Mar. Chem. 1985, 16, 121-130. (5) Donat, J.; Lao, K.; Bruland, K. Anal. Chim. Acta 1994, 284, 547-571. (6) Zhang, H.; van den Berg, C. M. G.; Wollast, R. Mar. Chem. 1990, 28, 285-300. (7) Sunda, W. G. Mar. Chem. 1984, 14, 365-378. (8) Fortin, C.; Campbell, P. G. C. Int. J. Environ. Anal. Chem. 1998, 72, 173-194. (9) Lead, J. R.; Davison, W.; Hamilton-Taylor, J.; Buffle, J. Aquat. Geochem. 1997, 3, 213-232.
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000 4447
J ) CD/∆g
(3)
The time-averaged mean concentration in solution during the deployment time is obtained by combining eqs 2 and 3 into 4. As for voltammetry, calculation of C requires knowledge of the diffusion coefficients, D, of the measured species
C ) M∆g/(DtA)
Figure 1. Schematic representation of the concentration gradient of metal species in a DGT assembly in contact with aqueous solution (C is concentration and DBL is diffusive boundary layer with thickness of δ).
Voltammetric techniques measure a current that is proportional to the flux of labile metal species to the electrode. The flux, J, is directly related to the concentration of all labile species by eq 1
J ) D1C1/δ + D2C2/δ + D3C3/δ + ...
(1)
where D1, D2...... are the diffusion coefficients of an individual metal species of concentration C1, C2...... and δ is the effective reaction layer thickness.10 Although this flux of total metal species can be readily measured, it cannot be used to determine the total concentration of labile species, because each species may have a different diffusion coefficient. Even in a simple situation, where there is only a single ligand, L, and metal complex, ML (M + L ) ML), if the total metal concentration or proportion of ML is not known in advance, concentrations of the individual species cannot be calculated. In the above example it was assumed that the diffusion coefficients of M and ML were known, but in an unknown solution, such as natural water, the diffusion coefficients of ML will not be known in advance. A new speciation tool known as DGT (diffusive gradients in thin films), which measures fluxes through a known area, A, and thickness, ∆g, of high porosity hydrogel, has recently emerged.11,12 When used to measure trace metals it employs a layer of Chelex resin as a binding agent at the back of the diffusive gel layer (Figure 1). When DGT is deployed, metals are concentrated on the resin. The measured mass of accumulated metal, M, provides a flux, J, directly from the known deployment time, t, and area of the exposed gel, A (eq 2). During deployment, a steady state is rapidly achieved (5 20
6.18 × 10-6 4.43 × 10-6 6.16 × 10-6
6.28 × 10-6 4.50 × 10-6 6.20 × 10-6
1.15 × 10-6 0.37 × 10-6 1.92 × 10-6
0.60 × 10-6 0.18 × 10-6 1.19 × 10-6
facturing process which resulted in a slightly more restricted gel did not occur until all these measurements were completed.16,21 Another type of gel (RG), comprising an acrylamide monomer cross-linked with bis-acrylamide (BDH Chemicals), has a much smaller pore size and does not allow free diffusion of metal ions. Note that the pore structures of all acrylamide gels are highly dependent on monomer and cross-linker concentrations, the polymerization temperature, and the concentrations of catalysts and initiators. These were rigorously controlled throughout this work. The third type of gel was 1.5% pure agarose (AGE) (Biofinex, Switzerland) The same resin gel was used for all DGT devices. It consisted of 2 g of Chelex 100 (Na form, 200-400 mesh) set in 10 mL of APA gel.12 Less catalyst and initiator were used to prolong the setting process and allow the resin to settle, by gravity, to one side of the gel, as a plane of approximately close-packed beads. All gels were hydrated in ultrapure water, MilliQ (MQ), for at least 24 h before use, to allow them to establish a new stable dimension. By changing the water several times during hydration, any impurities, such as unreacted reagents, were removed by diffusion and the pH was lowered to neutrality. Diffusive gels were soaked for a further 24 h in 0.1 M NaNO3 solution to eliminate diffusional artifacts present at low ionic strengths.16 In all experiments, the diffusive gels were 0.8 mm thick and the resin gels 0.4 mm thick after hydration. Gel holders based on a simple tight-fitting piston design with a 2-cm diameter window (DGT Research Ltd.)19,22 were used for all measurements. A layer of resin gel was placed on top of the piston, with the side containing the plane of close-packed resin beads facing upward. A layer of diffusive gel was placed on top of it, followed by a 135-µm-thick, 0.45-µm pore size cellulose nitrate membrane. Care was taken to exclude air bubbles between each layer, by keeping the layers wet during assembly. The front cap was then pressed down tightly, to form a good seal between the cap and the membrane surface. The membrane has been found to behave as an extension of the diffusive gel layer and to allow the free diffusion of ions.16 Experiments. Natural ligands of fulvic acid (FA) and aquatic humic acid (AHA) were extracted from streamwater. They were both supplied by Jamie Lead who, along with co-workers, has published detailed characteristics, including their nominal molecular weights determined by ultracentrifugation as 2400 (FA), and 6300 (AHA).23 All laboratory measurements were performed on solutions containing 0.1 M NaNO3 (BDH Analar) and 2 mM tris, which (21) Alfaro-De la Torre, M. C.; Beaulieu, P. Y.; Tessier, Anal. Chim. Acta. 2000, in press. (22) Zhang, H.; Davison, W.; Gadi, R.; Kobayashi, T. Anal. Chim. Acta 1998, 370, 29-38. (23) Lead, J. R.; Hamilton-Taylor, J.; Nesketh, N.; Jones, M. N.; Wilkinson, A. E.; Tipping, E. Anal. Chim. Acta 1994, 294, 319.
4450
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
buffered the pH to 7.8. Two metals (Cu and Cd) and four ligands including two synthetic ligands (EDTA, glycine) and two natural ligands (FA and AHA) were chosen for investigating the speciation capabilities of DGT and to test our proposed theory. Solutions with at least three different concentrations of metal ion were prepared with each ligand. DGT assemblies were placed in wellstirred plastic containers that held 4 L of solution. They were deployed for various times from 2 to 8 h at a constant known temperature (15-22 °C). Measurements were also made on each solution by ASV and ETAAS. The concentrations of inorganic and organic species in each studied solution were calculated using the speciation model WHAM.24 Analysis. After deployment, the resin gel layer was carefully removed from the DGT assemblies and placed in 1.5-mL vials. One milliliter of 1 M HNO3 was added to the resin gel and left for a day before analysis. A pipetted aliquot of the eluate was diluted appropriately for analysis by electrothermal atomic absorption spectroscopy, ETAAS (Perkin-Elmer Zeeman 4100 2L). The concentration of metal ions in the eluent prior to dilution, Ce, was calculated. This was used to calculate the mass of metal ion in the resin gel layer, M (eq 13)
M ) 0.8Ce (1 + 0.16)
(13)
The volume of the resin gel layer, of 0.16 mL, must be added to the original volume of the added eluent, in this case 1 mL. In practice, only a fraction of the bound metal is eluted. The ratio of the bound to eluted metals has been determined to be 0.8 for both Cd and Cu used in this work, irrespective of metal concentration and with a precision of better than 5%.12 The measured mass was substituted into eq 5 to calculate the concentration in the deployment solution. Total concentrations of metals in laboratory-prepared solutions or filtered natural waters were measured by ETAAS directly. Anodic stripping voltammograms were measured at a Hg drop electrode using an Ecochemie Autolab PSTAT10, DAC 124 instrument coupled to a Metrohm 663 VA electrode stand. Cu was deposited at -0.4 V and Cd at -0.8 V versus a Ag/AgCl electrode. In both cases, the deposition time was 60 s. After a 5-s rest period, differential pulse voltammetric scans were performed from -0.35 to 0 V for Cu and -0.75 to -0.4 V for Cd. Field Measurements. Three sets of three standard DGT devices, with 0.8-mm diffusive gels of AGE, APA, and RG, were deployed in situ in a humic rich stream, Greenhole Beck, in the North of England. The flow of the water was reasonably fast without obvious turbulence. The pH was 7.5, and the concentration of DOC was 14.6 mg/L. To determine the thickness of the diffusive boundary layer (DBL) at the surface of the DGT device, (24) Tipping, E., Comput. Geosci. 1994, 20, 973-1023.
Table 2. Concentrations of Cd (nM) in the Presence of Synthetic (EDTA, glycine) and Natural Ligands (FA and HS) Measured by ETAAS, ASV, and DGT Techniques ligand (concentration) EDTA (1 × 10-7 M) glycine (1 × 10-7 M)
FA (1 ppm, 4.2 × 10-7 M)
HA (1 ppm, 1.6 × 10-7 M)
ETAAS
ASV
DGT
DGT/ ASV
% free Cda
105.6 126.3 139.9
16.9 31.8 49.5
18.7 33.9 52.3
1.11 1.06 1.06
20.6 20.8 36.9
45.5 85.2 129.3
44.7 83.2 128.5
44.9 82.4 128.1
1.00 0.99 1.00
329.1 521.4 693.5
280.0 475.0 640.0
278.4 465.9 622.8
0.99 0.98 0.97
83.9 85.9 87.2
78.0 238.8 403.6
66.0 220.0 386.0
58.8 208.2 368.9
0.90 0.95 0.96
86.3 87.1 87.8
100 100 100
a The predicted percentage free Cd is calculated from known stability constants25 for synthetic ligands and using the WHAM speciation code24 for natural ligands.
further sets of DGT units, with two different gel-layer thicknesses (0.4 and 1.2 mm) of APA gel were deployed in triplicate. The thickness of the DBL was calculated using the procedure previously described.22 All the DGT devices were deployed at the same time and retrieved 35 h later. The temperature of the water was monitored (11 ( 0.8 °C), and water samples were taken at the beginning, middle, and end of the deployment. The water samples were filtered immediately on-site through 0.45-µm cellulose nitrate membranes. An aliquot of the filtered sample was acidified, and the rest was kept at 4 °C. Measurements by ASV were carried out on the filtered, unacidified samples within 24 h of sampling. The acidified samples were analyzed by ICPMS (Varian Ultramass) to obtain total dissolved trace metal concentrations. RESULTS AND DISCUSSIONS Comparison with ASV. In these initial tests of the speciation capabilities of DGT, the first priority was to compare the measurements with ASV for a range of speciation conditions. Theoretically, DGT should measure only labile species, such as ASV.12 However, whereas the effective time of measurement for typical ASV conditions (diffusive layer thickness ) 10 µm, diffusion coefficient ) 5 × 10-6 cm2s-1) is about 0.1 s (eq 5), for DGT with a diffusive layer thickness of 0.9 mm it is about 13.5 min. Table 2 shows concentrations of Cd measured directly by ETAAS, ASV, and DGT in solutions containing different ligands. While metal concentrations are higher than natural background levels, they allow for significant complexation with EDTA and humic substances. The molar concentrations of metals and fulvic and humic substances were similar, but multiple binding sites on each molecule provide an effective ligand excess. DGT measurements were made using the open-pored APA gel. The results were calculated using eq 4, assuming that the value of D for simple metal ions applies. ASV results were calibrated using simple metal solutions, implying that a single value of D can be used for all complexes. Means of triplicate measurements are presented. The relative precision of all measurements was within 5%. The ratios
of the concentrations measured by DGT to those measured by ASV are also provided. The percentage of free metal not bound by the ligand has been calculated from known stability constants25 for EDTA and glycine and using the speciation code WHAM for the humic substances.24 While the EDTA and glycine estimates should be fairly accurate, calculations using WHAM serve only as a guide, as testing of the code’s predictive capabilities is still underway.8 The effect of tris buffer was tested by performing measurements with and without it and found to be negligible in all the solutions. Glycine is such a weak complexing agent, that there is effectively no binding in the prepared solution. In this case, the concentrations measured by all three techniques were indistinguishable, confirming the accuracy of the measurement procedures. EDTA is a relatively strong binding agent. It complexes most (63-80%) of the Cd. The complex is sufficiently strong that the rate of dissociation of metal ions is too slow for them to be measured by ASV or DGT. Although the ASV and DGT measurements are similar, there is a slight systematic difference, with the ratio of concentrations measured by DGT and ASV always being greater than one. This result is consistent with DGT being able to measure species that are slightly less labile, due to its longer measurement time. In solutions containing fulvic and humic acids, ASV and DGT measurements were, again, very similar. Generally these measurements of labile metal were 10-15% lower than the ETAAS measurements of total metal, reasonably consistent with the extent of complexation predicted by WHAM. The lower values observed by DGT and ASV reflect both binding by strong complexes or sites to form inert complexes and binding by weak complexes or sites to form labile complexes which will be measured, but at lower sensitivity due to a smaller diffusion coefficient. The measured mass for this open-pored gel, oM, is given by eq 14 o
M ) [(oDMCinorg) + (oDHSCorg)](At)/∆g
(14)
oD , oD M HS
are the diffusion coefficients of the free metal ions and humic substances in the open-pored gel, respectively. HS is taken to be a generic term that can embrace FA and HA, and oD HS is taken to be the same as that of a metal complex with HS. Cinorg and Corg are taken to be the total equilibrium concentrations of all labile inorganic and organic complexes of the measured metal ion. As Corg was unknown, a total labile concentration was calculated using a value of oDM for all species. If there is complexation by organic species, use of a higher value of D than oD HS will result in underestimation of total labile metal species. A similar argument applies to ASV measurements. For the humic acids, in particular, the DGT measured concentrations were systematically lower than the ASV measured concentrations (DGT/ASV ) 0.9-0.96). This is opposite to the expected effect if lability is a factor, as shown above for EDTA. Generally, the majority of binding sites on humic substances can be expected to be weak, so most species would be expected to be labile on the time scale of ASV and DGT. That is, lability would not be expected to be an issue. However, larger molecules, such as humic substances, diffuse more slowly than free metal ions. (25) Pettit, G.; Pettit, L. D. SCsDatabase, IUPAC, Academic Software, 1999.
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
4451
Therefore, in calculating concentration, use of the diffusion coefficient for the free metal ion is likely to underestimate the amount of labile metal measured by DGT. There will be a similar underestimation for ASV, which is calibrated with simple inorganic solutions. The diffusive gel of DGT retards the diffusion of humic substances more than the aqueous diffusive layer of ASV, as demonstrated by lower diffusion coefficients in the gel.16 Consequently, DGT measured values are lower than those for ASV and a ratio of DGT/ASV < 1 is reasonable. Labile Inorganic and Organic Measurements by Multiple DGT. Accurate measurement of the distribution of species in complex solution using conventional ASV or DGT with an openpored gel is usually impossible. This is because the extent of complexation by organic compounds with smaller diffusion coefficients than the free metal ion is unknown. Carefully selecting the deposition potential of ASV at a rotating disc electrode can provide Cinorg directly.5 Use of multiple DGT devices containing diffusive gels with different pore sizes opens up the possibility of measuring both Cinorg and Corg (eq 10). The mass of Cu accumulated on the resin gel of each DGT device was measured by ETAAS and calculated using eq 13. The diffusion coefficients of Cu and humic substances in the three different gels measured using a diffusion cell16 (Table 1) were used to calculate MDGT/(KDinorg) and Dorg/Dinorg. Plots of the first term against the second were linear in all tested solutions containing different metal concentrations and different ligands. Examples of DGT measurements in solutions of fulvic and humic acid at three different concentrations of Cu are given in Figure 2. The straight lines for each Cu concentration indicate the reliability of the DGT speciation measurement. The concentrations of labile inorganic species (RSD < 5%) and organic species (RSD < 10% except for one slope) in each solution were obtained from the intercept and slope. The effect of the filter, with different values of diffusion coefficients to the gels, has been neglected for this work. Although it could be accommodated in the calculations, it is sufficiently thin for its effect to be very small. Total metal concentrations were measured by ETAAS. Measurements were also made using ASV. The resultant labile concentrations embrace both inorganic and organic species, as discussed for Table 2. The speciation code WHAM was used, with the total concentration of metal ions and the concentration of FA or AHA as inputs, to predict the distribution of species in all the tested solutions. In these known, laboratory-prepared solutions, the multiple DGT measurements should give the correct distribution of species, as all parameters are measured directly. Assumptions only relate to the applicability of using mean values for the diffusion coefficients of FA and HA. The WHAM calculation is a predictive model that requires testing. The concentrations of labile inorganic and organic Cu species measured by DGT and predicted by WHAM, together with the ASV-labile concentrations and the total dissolved Cu concentrations, are shown in Figure 3. Interpretation of the DGT measurement does not rely on the total concentration of metal measured by ETAAS. The total labile concentration measured by DGT, represented by the overall height of the bars (Figure 3a,b), is obtained from summing the independently measured concentrations of labile inorganic and organic metal species. Except for the solutions with the highest concentration of copper (experiment C), the agreement with the 4452 Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
Figure 2. Results of experiments where duplicate DGT devices with three different gel compositions were deployed in three different solutions of Cu(NO3)2 with (a) 1 mg/L of fulvic acid (FA) and (b) 1 mg/L of humic acid (HA), respectively. The measured mass accumulated by DGT, M, was converted to the y axis function (eq 10) and plotted against the ratio of the diffusion coefficients for (a) FA and Cu2+ and (b) of HA and Cu2+ in each gel composition. The slopes and intercepts of each line provide estimates of the concentrations of organic and inorganic species, respectively.
directly measured total metal is excellent. This indicates (a) that virtually all metal species must be labile in these solutions, (b) the good precision of the DGT and ETAAS procedures, and (c) that the procedure of using multiple DGT devices containing gels with different calibrated values of the diffusion coefficients is valid. Differences at high copper concentrations might reflect the presence of inert copper complexes which are not able to dissociate during diffusion through the gel, and consequently, they cannot be measured by DGT. Simple equilibrium considerations would not predict the presence of inert complexes at high concentrations of Cu, but the solution could have been affected by aging. The solution with the highest Cu concentration was
Figure 3. Comparison of different measurements of Cu in three solutions, A, B, and C, in the presence of (a) FA and (b) HA. “Total” was the analytical measurement on the solution made by atomic absorption spectroscopy. “ASV” was measured by anodic stripping voltammetry, assuming that all species in solution have the diffusion coefficient of free metal ions. Inorganic and organic labile species were measured separately using DGT devices with different gels and were also calculated using the WHAM speciation code using the “Total” concentration as an input parameter.
prepared by adding Cu to the same solution used for experiments A and B. Ligands and Cu were present in the solution for a few days before experiment C was carried out. Inert Cu complexes may form and coagulation of humics may occur during that time. The ASV labile concentration is higher than the labile inorganic concentration measured by DGT, which is not surprising, as the ASV measurement does not exclude labile organic species. ASV cannot be assumed to measure only labile inorganic species. The ASV labile concentration was obtained conventionally by standard addition of Cu2+ to the solution. The diffusion coefficient of Cu2+ is much greater than that of Cu-humics complexes, so the ASV concentration would be expected to be closer to Cinorg measured by DGT than total Cu. Although this is not so (Figure 3), clearly ASV neither provides a measurement of labile inorganic or total labile species. The WHAM predictions agreed well with the direct measurements of labile inorganic and organic concentrations for solutions
of Cu and FA (Figure 3a). WHAM also successfully predicted Cu speciation in the solution of HA with the two highest concentrations of Cu (Figure 3b). At the lowest Cu concentration, however, WHAM underestimated the extent of organic complexation. With both FA and HA, the extent of organic complexation of Cu was underestimated at high Cu concentrations and overestimated at low Cu concentrations. Fortin and Campbell8 have observed discrepancies between model speciation predictions by WHAM and directly measured values for Zn and Cd in FA solutions. Direct Measurement of Labile Inorganic Metal by a Single DGT Unit. DGT units fitted with the gel with the smallest pore size (restricted gel, RG) can potentially be used for measuring labile inorganic concentrations directly. The restricted gel allows simple metal ions to diffuse through with little retardation, but it appreciably retards the diffusion of humic substances and minimizes their transport. An estimate, therefore, of labile inorganic metal species can be made by assuming that DGT measurements Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
4453
Table 3. Concentrations of Total Labile Cu (inorganic plus organic, CTL) and Labile Inorganic Cu (Cinorg) Obtained by Using Multiple DGT Devices Compared with Labile Inorganic Cu Measured Using DGT Devices with Single Restricted Gels (C sinorg) in two sets of experiments with different ligands (FA and HA)a ligand
experiment series
CTL (×10-6 M)
Cinorg (×10-6 M)
C sinorg (×10-6 M)
Cinorg/CTL (%)
overestimation by C sinorg (%)
FA FA FA HA HA HA
A B C A B C
0.482 0.884 1.894 0.595 1.093 1.935
0.250 0.366 1.016 0.130 0.195 0.416
0.270 0.408 1.090 0.147 0.227 0.513
51.9 41.4 43.6 21.8 17.8 23.8
8.0 11.5 7.3 13.1 16.4 11.3
a The percentage of labile inorganic species (C inorg/CTL) in each solution and the error (over estimation) associated with using the single gel DGT ((C sinorg - Cinorg) × 100/Cinorg) were also calculated.
made using this gel, calibrated with the value of D in the gel found for free metal ions, effectively exclude labile and inert organic species. This is equivalent to saying that the diffusion of humic substances, rDHS, in the restricted gel is so slow that the term rD C HS org in eq 15 can be neglected r
M ) [(rDMCinorg) + (rDHSCorg)]At/∆g
(15)
rM
is the measured mass using the restricted gel and rDM the diffusion coefficient of the free metal ion in the restricted gel. As the diffusion properties of humic substances in the restricted gel are known, it is possible to test this hypothesis directly by calculation. For hypothetical solutions where humic complexes accounted for 10, 50, and 90% of the labile metal ion, the percentage estimates of inorganic species using DGT with the restricted gel would be 90.4, 52, and 13.6. With the more freely diffusing fulvic substances, they would be 90.8, 54.1, and 17.4%. Clearly, the estimation is good when complexation by organic substances is moderate or minor, but errors can be substantial if organic complexes dominate the solution. Labile inorganic concentrations of Cu in all test solutions obtained from multiple DGT devices were compared with those recalculated using only the data provided by the single DGT units fitted with the restricted diffusive gel (Table 3). Single DGT measurements using the restricted gel provide good estimates of labile inorganic species for all studied solutions, as the inorganic species in those solution are g20% of the total Cu concentration. In Situ Measurement of Labile Inorganic and Organic Cu Species. Multiple DGT devices with three different diffusive gels and two thicknesses were deployed in a stream for 35 h, and the masses of metals accumulated on each DGT unit were determined using ICPMS. The diffusive boundary layer (DBL) thickness was calculated from data obtained by DGT devices with two different gel layer thicknesses using the procedure described by Zhang et al.22 The measured thickness of the DBL, of 0.4 mm, was similar to those found in lakes.22,26 Calculations of M/(KDCu) and DFA/ DCu were made after incorporating the DBL thickness into the diffusive layer of each DGT device and assuming that the organic ligands in the stream are dominated by fulvic acid (FA). The results are presented in Figure 4 as a plot of M/(KDCu) versus DFA/DCu. The reproducibility of the whole data set, as indicated by r2, is poorer than that for the controlled laboratory experiments,
(26) Beaulieu, P. Y. M.S. Thesis, University of Quebec (INRS-Eau), 1999.
4454 Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
Figure 4. Results of in situ deployment where triplicate DGT devices with three different gel compositions were used in Greenhole Beck. The measured mass accumulated by DGT, M, was converted to the y axis function (eq 10) and plotted against the ratio of the diffusion coefficients for FA and Cu2+ (assuming 100% humic substances are fulvic acid). The slope and intercept of the line provided estimates of the concentrations of organic and inorganic species, respectively.
but it is remarkably good for in situ measurements, considering the uncontrolled hydrodynamics and temperature and the difficulties associated with field deployment. The concentrations and standard deviations of labile inorganic (11.8 ( 0.7 nM) and labile organic species (18.2 ( 3.0 nM) were obtained from the intercept and slope of the regression line, respectively. To our knowledge, this is the first direct in situ measurement of both labile inorganic and labile organic trace metal in a natural water. The concentration of labile inorganic Cu was also estimated using only the DGT units with the restricted gel. It agreed reasonably well with the value measured using the multiple DGT devices with different pore sizes of gels, as shown in Figure 5. This demonstrates the potential of using a single DGT device, with a restricted gel as an in situ speciation tool. Its simplicity and accuracy provide a monitoring tool for regulation purposes and a speciation tool for bioavailability studies. The total labile concentration of Cu (sum of labile inorganic and organic) measured by DGT is about 10% less than the total
Table 4. Values of Input Parameters Used for Calculating Labile Inorganic and Organic Species in Greenhole Beck Using WHAM (Example W1 in Figure 6) inputs
values
(M) Mg2+ (M) Na+ (M) K+ (M) Cl- (M) NO3- (M) SO42- (M) total labile Cu (M) Al (M) Fe(II) (M) Fe(III) (M) pH total DOC (g/L) Humic DOC (g/L) FA (g/L) HA (g/L) alkalinity (mequiv/L)
3.0 × 10-4 2.5 × 10-4 2.4 × 10-4 2.8 × 10-5 7.6 × 10-5 1.3 × 10-5 3.6 × 10-5 3.15 × 10-8 3.7 × 10-6 4.3 × 10-6 1.5 × 10-5 7.5 1.46 × 10-2 8.76 × 10-3 1.75 × 10-2 0 3.15
Ca2+
a
Figure 5. Comparison of different measurements of Cu in Greenhole Beck. “TD” is total dissolved, measured by atomic absorption spectroscopy of filtered samples. “ASV” was measured by anodic stripping voltammetry, assuming that all species in solution have the diffusion coefficient of free metal ions. Inorganic and organic labile species were measured separately using DGT devices with different gels and were calculated using the WHAM speciation code using the total DGT (inorganic + organic) concentration and the data shown in Table 4 as the input parameters. “DGT-RG” represents the inorganic species estimated using only the DGT devices with the most restricted gel.
dissolved concentration measured by ETAAS on filtered samples (Figure 5). This may be due to the presence of inert Cu complexes and/or colloidal forms in the streamwater, which are not available for DGT measurement. Great difficulty was encountered with the ASV measurements due to insufficient supporting electrolyte and pH buffer in the water samples. A small amount of NaNO3 (10-2M), as supporting electrolyte, had to be added to the samples to make the measurement possible. Consequently, speciation in the samples may have been changed. The pH change during deoxygenation, due to a lack of buffer in the samples, could significantly affect metal speciation. By tolerating these artifacts and making calibrations using Cu2+, results were obtained by conventional ASV, but they can only be considered as estimates rather than reliable measurements, and therefore, they have not been compared with the DGT results. Some of the problems mentioned above may be overcome by using gel-coated microelectrodes with fast-scan ASV.27 The concentrations of labile inorganic and labile organic species were predicted for Cu using the speciation model WHAM.24 The input data, including concentrations of FA or HA, major cations and anions, Al, Fe(II), and Fe(III), are summarized in Table 4. The concentrations of major cations were determined by flame photometry and atomic absorption spectrophotometry. (27) Tercier, M.L; Buffle, Anal. Chem. 1996, 68, 3670-3678.
measured measured measured measured measured measured measured measured by DGT measured estimateda estimateda measured measured assuming 60% of total DOC assuming 100% of humic DOC assuming no HA measureda
Taken from Peters.28
Dionex ion exchange chromatography was used for analyzing anions. The concentrations of Fe(II) and Fe(III) were adopted from the results in a similar local stream.28 The concentration of humics was calculated from DOC concentration by assuming 60% of the DOC is humic substances, as previously found in these waters.28 The total labile concentration of Cu measured by DGT was used for the WHAM calculation. The total dissolved concentration is not an appropriate input parameter, as WHAM is an equilibrium model. The total dissolved metals in natural waters often include a fraction of inert complexes or colloidal forms which are not in equilibrium with other labile species. Assuming all ligands in the streamwater can be represented by FA, as assumed for DGT calculations, the concentrations of labile inorganic and organic metal predicted by WHAM are shown in Figure 5. The predicted labile inorganic concentration was two times higher than the concentration obtained by deploying multiple DGT devices in the stream. It is not entirely reasonable to assume that FA is the only organic ligand in the streamwater when ratios of FA/HA have been suggested to be 9:1.29 We have assumed alternatively, therefore, that the organic ligands consist of 10% HA and 90% FA and have recalculated the DGT results and the WHAM predictions. The concentrations of labile inorganic and organic Cu measured by DGT did not change significantly when 10% HA was introduced into the calculations compared with the no HA (100% FA) situation. The diffusion coefficients of FA are only about twice those of HA in all gels (Table 1). Consequently, the effect of 10% HA on the interpretation of the DGT measurements is negligible when 90% of the organic ligands is FA. WHAM predictions were significantly affected by introduction of HA into the input parameters. The labile inorganic concentration calculated by WHAM was halved when 10% HA was introduced (Figure 6). It then agreed well with the DGT measurement, perhaps fortuitously, in view of the dependency of the predictions on the assumed water chemistry. Predictions were also made after eliminating Fe(III) (28) Peters, A. J. Ph.D. Thesis, Lancaster University, 1999. (29) Tipping, E.; Hilton, J.; James, B. Freshwater Biol. 1988, 19, 371-378.
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
4455
Figure 6. Comparison of the DGT measured distribution of inorganic and organic Cu species in Greenhole Beck with predictions using the WHAM speciation code with different inputs: W1, inputs as shown in Table 4; W2, as W1 except Fe(III) ) 0; W3, as W1 except HA was considered with the assumption of HA ) 10% humic DOC; W4, as W3 except assuming all DOC is humics material (humic DOC ) total DOC); W5, as W3 except Fe(III) ) 0.
from the input data. Such information is commonly unavailable, and the solution form can easily be overestimated due to the presence of colloids.30 The speciation was very different from that obtained with Fe(III) inputs (Figure 6), with organic complexation being dominant, as usually predicted for Cu. The concentration of humic substances is another important parameter for model prediction. It is not simple to obtain the individual concentrations of FA and HA in natural waters, and estimations made from DOC measurements are inherently inaccurate. Speciation models such as WHAM depend on an accurate set of input parameters, including ligand type and concentration, major cations and anions, and trace metals with a strong affinity for the ligands. Without performing none-routine measurements of input data, such as FA/ HA ratios and total labile metal, they cannot be expected to reliably predict speciation, even if there are no problems with the mathematical description of the system and parametrization. Multiple DGT provides a direct measure of total labile metal, and its estimation of inorganic species is much less sensitive to FA/ HA ratios. GENERAL APPRAISAL Environmental chemists can regard the complete quantitative description of the chemical species present in a natural water as a legitimate scientific goal, while for other scientists it is a (30) Hassellov, M.; Lyven, B.; Haraldsson, C.; Sirinawin, W. Anal. Chem. 1999, 71, 3497-3502.
4456
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
necessary precursor to their work. Geochemists need such information to understand mineral formation and dissolution processes, and the role of solution speciation in biological uptake is a preoccupation of both biologists and regulators concerned with environmental risk. These practitioners require a simple procedure that provides an accurate description of the solution speciation. This work has begun to explore the possibility of using DGT to do this. No single speciation technique can measure all individual species, so reliance has to be placed on measuring single species or sets of species and using calculations to estimate the remaining species. DGT can be used to distinguish between small and large species, which we have ascribed to inorganic and organic species present in natural waters. This operational definition is based on their mobility, calibrated by simple inorganic solutions or by fulvic or humic acids. Only labile species are measured, which is appropriate for speciation calculations based on equilibrium models. The measured total labile inorganic species can therefore be used as an input parameter, along with major ions and cations, in a speciation program, allowing the calculation of the concentration of all species. Apart from some notable exceptions, such as metal sulfide complexes, the stability constants for metal inorganic complexes are reasonably well-known and tested, making this an accurate procedure. A logical future extension of this work will be to undertake parallel measurements with ion-selective electrodes and chemical competition methods to test the validity of the DGT procedure. While, theoretically, the single DGT measurements can never be as accurate as those using multiple gels, pragmatically, they offer great promise, as illustrated in Table 3. Measurement using a single device would clearly be favored by regulators who may need to make many measurements and therefore seek simplicity. A single measurement with a restricted gel, accompanied by pH and approximate solution composition, inferred from conductivity and degree of hardness, may be the simplest set of measurements that can be made to give a reasonable estimate of the inorganic species distribution. Use of multiple gels allows determination of labile metal associated with large molecules. While this could include inorganic colloids, they are unlikely to be of substantial concentration at the low molecular weights of FA and HA used as calibrations.30 In freshwater particularly, FA and HA are likely to be dominant. Complexes with strong binding ligands, either as single organic ligands or strong binding sites of humic substances, will not be measured, as they will not be labile.13 DGT is probably the only technique that estimates, directly, total, labile, organic, metal complexes. If a measurement is made with only an open-pored gel, both inorganic and organic labile complexes are measured. Like ASV, if calibration is performed with inorganic solutions, effectively using D for the free metal ion, the total labile concentration will be underestimated due to the lower diffusion coefficient of organic complexes. However, this measurement may still be appropriate for assessing bioavailable metal in a solution. When membrane transport is slow, the concentration of the free ion in solution determines the rate of uptake.31 However, with fast membrane (31) Campbell, P. G. C. In Metal Speciation and Bioavailability in Aquatic Systems; Tessier, A., Turner, D. R., Eds.; Wiley: New York, 1995; Chapter 2.
transport, uptake is governed by the product of the concentration of each species and their diffusion coefficient. This is exactly what DGT measures directly with the open-pored gel, as it is a physical surrogate of fast membrane transport. This type of DGT measurement therefore provides, directly, the effective bioavailable metal concentration for the worst possible biological situation of fast membrane transport. Moreover, it is easier to make the measurement by DGT rather than ASV because of DGT’s simplicity, robustness, and independence of medium characteristics such as ionic strength and surfactants. It could be argued that none of these speciation measurements are necessary. Models that include the interaction between humic substances and metal ions such as WHAM24 and NICCA32 can be theoretically combined with simple measurements, and the complete distribution of species can be calculated. This approach is reasonable if the input parameters are well-known and the solution well-defined. But for most routine data available for natural waters, the predictive capability is limited and not fully tested. Free Cu ion concentrations calculated using the NICCA model for the studied streamwater were about 3 orders of magnitude lower than those calculated using WHAM. Whereas WHAM can consider humic binding to Fe, NICCA, as supplied, cannot, but (32) Kinniburgh, D. G.; van Riemsdijk, W. H.; Koopal, L. K.; Borkovec, M.; Benedetti, M. F.; Avena, M. J. Colloids Surf., A 1999, 151, 147-166.
NICCA includes strong binding sites while WHAM (present available model V) does not. The examples of Figure 6 show that precise measurement of DOC, major ions, and the metal of interest are insufficient to constrain the predicted species. More detailed characterization of the FA/HA components of the DOC are required, and other trace metals must be measured. Furthermore, simple filtration may lead to gross overestimation of labile metal, as it embraces inert and colloidal species. Rather than strive for such a sophisticated set of input data for the full modeling approach, it may be more sensible to target sets of species with DGT and use simpler models.
ACKNOWLEDGMENT We thank the Natural Environmental Research Council for financial support, Jamie R. Lead for providing fulvic and humic substances, Adam J. Peters for analyzing major cation and anion concentrations in streamwater samples, and Bernie Simon for providing DOC analysis in streamwater samples.
Received for review April 10, 2000. Accepted June 30, 2000. AC0004097
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
4457
