Environ. Sci. Technol. 2005, 39, 4974-4982
Determination of MeHg in Environmental Sample Matrices Using Hg-Thiourea Complex Ion Chromatography with On-line Cold Vapor Generation and Atomic Fluorescence Spectrometric Detection CHRISTOPHER W. SHADE* AND ROBERT J. M. HUDSON Department of Natural Resources and Environmental Sciences, University of Illinois at Urbana-Champaign, W-512A Turner Hall, 1102 South Goodwin Avenue, Urbana, Illinois 61801
A novel system for mercury speciation analysis using high-pressure ion chromatography (IC) has been developed and validated. Its chemistry permits separation of the two most abundant forms of Hg in natural waters, soils, sediments, and biotasmonomethyl Hg (CH3Hg+) and mercuric Hg (Hg2+)son the basis of the difference in charge of their respective thiourea (SdC(NH2)20) complexes. Once separated, both species are converted to Hg0 on-line and quantified by cold-vapor atomic fluorescence spectrometry (CVAFS). A column containing thiol-functionalized silica resin installed in the sample loop of the IC system traps Hg2+ and CH3Hg+ from prepared sample solutions without retaining interfering sample matrix components. The resulting matrixindependent chemistry permits external calibration of the system and a high sample throughput (∼6 samples per hour). The system’s accuracy has been validated with environmentally relevant reference materials. Figures of merit for the system, an average precision of ∼2.5% and an absolute detection limit of 2, Hg2+ (and by implication CH3Hg+) is readily trapped on thiolfunctionalized resins, even from 0.2 M TU solutions (Figure 1), yet readily eluted by solutions with the same TU concentration at pH < 1 (10, 31, 32). These unique characteristics of TU make possible the coupling of on-line sample preconcentration using thiolfunctionalized resins to IC-based MeHg and HgII separation. Such preconcentration is important for obtaining low detection limits because, without it, only a fraction of a typical sample can be injected into the analytical system. Thiolresin preconcentration is infeasible with Cys-, 2ME-, or chloride-based LC systems, because it is difficult to dissolve enough ligand to elute the Hg species from the columns (Figure 1). While preconcentration of Hg complexes onto RP C18 columns has been demonstrated with dithiocarbamates, recovery was nonquantitative (21, 23), and none of the other ligands used in HPLC form Hg complexes that are sufficiently hydrophobic to trap in this way. The remainder of this paper describes the components and operation of our new system for on-line separation and quantification of MeHg and HgII, referred to as HgTU/ IC-CVAFS or more briefly as HgTU/IC, and documents its performance with solvent-extracted digests of reference sediments and biological tissues.
Experimental Section System Instrumentation and Hardware. The analytical system (Figure 2) comprises three main subsystems: (i) a high-pressure LC system to separate aqueous MeHg and HgII complexes across an ion chromatography column; (ii) a lowpressure, cold-vapor generation system with on-line photooxidation of MeHg to HgII followed by reduction of HgII to Hg0 by alkaline SnCl2; and (iii) an AFS Hg detection system that measures Hg0 transferred from the eluant into the carrier gas stream. All commercially available components selected for use in the aqueous portions of the system are able to withstand the high acid content of the eluant (∼1 M HCl) and strong-acid cleaning steps. PEEK tubing and fittings are used throughout the high-pressure end of the system and PTFE tubing and fittings for the low-pressure end. The most important custom-made elements of the system, the thiol resin used in the trapping column (TT) and the gas-liquid separator (GLS), are relatively simple to fabricate. Note also that by removing the speciation column, the system can be used as a flow-injection, cold vapor AFS system to analyze for HgT. System Carrier/Reagent Fluids. The functions of the principal aqueous and gas streams (see inset table in Figure 2) dictate both their composition and their point of entry into the flow path of the analytical system. Eluant. The aqueous TU/HCl eluant has both a very low pH and a high complexing capacity, causing the dominant Hg species to be charged MeHgTU+ and Hg(TU)42+ complexes that can be separated chromatographically and that have little tendency to adsorb to system components. The eluant 4976
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 13, 2005
also contains acetic acid, which behaves like a hydrophilic organic solvent in the high-pressure, low-pH end of the system, improving the elution characteristics of the thiol concentrator column, but ionizes under the alkaline conditions of the cold-vapor generation system, thereby preventing its volatilization into the carrier gas. Oxidant. The aqueous H2O2 solution (final concentration 4.5% m/v) enters the eluant stream immediately after the speciation column via a PTFE “T”-type mixer. The mixture then passes through a 7-m section of PTFE capillary tubing wound directly around the bulb of a low-pressure UV lamp (8 W at λ ) 254 nm), where H2O2 photolysis yields OH radicals that rapidly oxidize MeHg to HgII (37). Antioxidant. Antioxidant is injected after PCO to quench excess oxidant prior to reduction by SnII. The aqueous sodium ascorbate solution (14% m/v) is made up either from the Na-salt or from food grade ascorbic acid (12.5% m/v) and adjusted to pH > 8 with NaOH. Reductant. In the final flow injection step, the aqueous stream mixes with sufficient SnII to chemically reduce all HgII to Hg0. The presence of thiourea in the eluant necessitates the use of highly alkaline, rather than acidic, SnCl2 solutions (32). Carrier Gas. Detection of Hg0 formed in the reduction step requires it to be transferred from the aqueous stream to the gas stream via a gas-liquid separator. The UHP-Ar gas stream is then dried through a Nafion gas drying tube with a 200 mL min-1 countercurrent flow of dried compressed laboratory air. Other Reagents. Sodium borate (0.1 M) and sodium citrate (0.75 M) (Fisher Certified ACS grade) buffers included ascorbic acid (1% m/v) to preserve resin thiols. Solvent extractions were performed using toluene (Fisher Certified ACS grade) and back-extracted with acidic TU (1% m/v in 0.5% v/v HCl). Concentrated, bromine-based oxidant (HOBr) was synthesized by mixing 20 parts by volume concentrated HCl with 1 part 30% m/m H2O2, allowing 1 min for reaction, and then adding 20 parts of cleaned KBr (1 M). This low-blank HOBr stock was used to oxidize aqueous MeHg in place of BrCl (38). Reagent Cleanup. To maintain a low baseline, reagents were cleaned of contaminating HgII when necessary. The eluant was cleaned by adsorption of Hg(TU)42+ complexes onto a column of highly-sulfonated, 100% DVB resin located on-line between the HPLC pump and the injection valve. A clean bed of resin is typically effective for at least 3 days of use and can be regenerated with 50 mL of 1.5 M KBr in 30% v/v aqueous ethanol. Citrate, KBr, borate, and ascorbate stock solutions were purified by stirring batches with mercaptopropyl silica gel for 20 min and then filtering through an acid-cleaned 0.45-µm HVLP membrane. Hg0 in the alkaline SnCl2 reductant was removed by vacuum filtration and sparging with N2 gas. A stream of N2 was bubbled through the SnII solution during analytical runs to prevent sensitivity changes due to oxidation by atmospheric O2. The H2O2 solution used (Figure 2) contained little contamination. Preparation of Thiol Trap. The mercaptopropyl silica gel (MPSG) used in the thiol trap was synthesized using wellknown methods (34-36) with minor modifications. Briefly, silica gel (Davisil grade 633, 40- to 63-µm) was activated by refluxing in concentrated HCl for 3 h and then rinsed with DI water until the decanted water was neutral. MPSG was synthesized by refluxing 5-g of the activated silica gel in a N2 atmosphere with 10 mL of mercaptopropyltrimethoxysilane (Aldrich) in 50 mL of dry toluene for 6 h. Next, the MPSG was rinsed sequentially with toluene, ethyl alcohol, and ethyl ether before being dried under vacuum. The gel was transferred into small glass vials, which were then flushed with N2 and stored in a freezer until use. Using a 1-mL plastic syringe, MPSG slurried in 1% m/v ascorbic acid solution was packed
FIGURE 2. System schematic. Hardware component labels denote: [HPLC] High pressure, single-piston pump with pulse-dampener (Chromtech Series III); [CC] Cleaning column (Alltech, 4.6 × 50 mm) packed with sulfonated, 100% DVB resin (Jordi Associates); [HPIV] High-pressure injection valve (Perkin-Elmer/Rheodyne EV750-102) with [SL] 100-µL sample loop and [TT] thiol trap/preconcentration column (custom); [ICC] Ion chromatography column with mixed-mode resin (Dionex Ionpac CG4A, 4 × 50 mm); [UV-PCO] PTFE tubing, ∼7 m, wound directly around an 8-W, low-pressure UV bulb; [PP] Console drive peristaltic pump (Masterflex LS) with mini-cartridge pump head (Ismatec) and Norprene pump tubes; [GLS] Thin-film diffusion, gas-liquid separator (custom-built borosilicate glassware); [AFS] Atomic fluorescence spectrometer with mass flow controller (Tekran model 2500); [A-D/PC] Analog-digital converter with output connected to a personal computer running PeakSimple software (SRI Instruments); [NDT] Nafion gas drying tube (Perma Pure MD-050-24P). To stabilize its signal, the AFS draws its power from an uninterruptible power supply that was connected to a line conditioner. Carrier/Reagent Fluids Table: Flow rates in mL min-1; Mixing ratios of HCl, HAc, and H2O2 in % v/v (mL per 100 mL solution), solids in % m/v. Sources: (A) Acros, (F) Fisher, (B) J.T. Baker, (S) Soloray Foods. Aqueous solutions prepared from ultrahigh-purity (18-MΩ) water (Millipore, Milli-Q system). into a high-pressure microcolumn made of either PEEKlined stainless steel (Alltech, 2.1 × 30 mm) or borosilicate glass (Omnifit, 3.0 × 25 mm) and fitted with PEEK frits. Sample Loading and On-line Concentration. Samples were injected into the system via either the sample loop (SL) or the thiol trap (TT). For SL-injection, standards and samples were prepared in the same matrix-matched, eluant-compatible solutions, for example, acidic TU. To prepare for TT-loading, MPSG columns were rinsed with 2.5 mL of borate buffer to set the pH to ∼9 and aliquots of sample or standard in acidic TU were neutralized with excess citrate buffer. The resulting pH 3-5 solutions were delivered into the TT via a glass syringe (Hamilton Gastight) and then followed with 1 mL of high-purity water. Samples up to 5 mL in volume can be rapidly loaded; larger volumes are possible if desired. Extraction and Digestion Procedures. MeHg was leached from 0.1- to 0.2-g samples of reference tissue (DOLT-2 and DORM-2) and 0.5- to 1.0-g samples of reference sediment (BCR 580) by slurrying with 5 mL of KBr/H2SO4 (18% m/v in 5% v/v acid) plus 1 mL of 1 M CuSO4 and shaking for 1 h (13).
Next, 7 mL of toluene was added, and the mixture was shaken for 1 h. An accurately measured aliquot (>4 mL) of the toluene layer was back-extracted into 5 mL of acidic TU by shaking for 10 min. The back-extract was passed through an acidrinsed, PTFE-membrane syringe filter prior to analysis. For HgT analysis, solid samples were digested using a modification of Method 1631 (38). Briefly, freeze-dried samples of tissue or sediment were weighed into 40-mL, borosilicate glass vials with fluoropolymer-lined caps (I-CHEM Certified 200 series), combined with 7 mL of a concentrated HNO3/concentrated H2SO4 (7: 3) mixture, and boiled gently on a hot plate for 3 h. The acid digest was then transferred to a volumetric flask, made up to 25 mL with 10% v/v HOBr, and heated in an oven at 60 °C for g4 h. Aliquots of the final, oxidized digest were diluted (10% v/v) with ultrapure water and analyzed using SL-injection. Facilities and Cleaning Protocols. Critical steps in sample and standard preparation were performed on HEPA-filtered, laminar flow benches. The lab is supplied with HEPA-filtered, conditioned outside air and maintained at a positive pressure VOL. 39, NO. 13, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4977
FIGURE 3. Chromatogram of repeated SL-injections of standards containing ∼200 pg of MeHg and HgII.
FIGURE 4. Chromatogram of SL-injected, sediment leachate (BCR 580 digested with 2 M HBr w/o solvent extraction). relative to the rest of the building. All glassware used for sample preparation was cleaned successively in 2% v/v aqueous Micro-90 detergent (overnight at 20 °C), 6 M HCl (overnight at 20 °C), 5% v/v HOBr (8 h at 60 °C), and 1% v/v HCl (8 h at 60 °C). Pre-cleaned glass vials (I-Chem Certified 200 series) were cleaned further using the last two steps.
Results and Discussion Hg Speciation Analysis Using HgTU/IC. Chromatograms of aqueous standards containing MeHg and HgII exhibit baseline separation (Figure 3). At the observed peak separation of 5-6 min, samples that do not contain excessive levels of HgII may be analyzed at a rate of 5-6 per hour. Peak separation may be increased by reducing the acid content of the eluant, permitting the separation to be optimized for different sample types. While we have not specifically investigated them, other mono-organomercurials may elute with MeHg in this system. Thus, samples where their presence is suspected should be analyzed using another method. Chromatograms of sediment leachates also exhibit baseline separation (Figure 4) and underline the importance of MeHg eluting first. If MeHg eluted second, tails of large HgII peaks, such as evident here, would overlap with the MeHg peak and interfere with its quantification. The elution order 4978
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 13, 2005
FIGURE 5. System MeHg oxidation efficiency. Exactly equimolar HgII + MeHg solutions were prepared by oxidizing one of a set of paired MeHg standards at 5, 10, 15, 20, and 25 ng L-1 and combining each with the matching, unoxidized MeHg standards. 5 mL volumes were TT loaded. The standards (made in 20 mL of 0.5% v/v HCl) were (i) oxidized by adding 400 µL of HOBr stock and digesting overnight at 50 °C and (ii) neutralized by adding 50 µL of 30% m/v NH2OH‚HCl. is most important when making simultaneous direct measurements of MeHg and HgII, because MeHg is often 1 or more orders of magnitude lower in concentration. Such applications will be emphasized in subsequent work. Cold Vapor Generation. Efficient Hg cold vapor generation for AFS analysis requires an interference-free chemistry that couples complete MeHg oxidation with effective HgII reduction. Several PCO schemes were examined, including UV (21), persulfate with a Cu catalyst (28), persulfate/UV (29), BrCl/UV, and permanganate/UV. Their efficacies were assessed by comparing baseline fluorescence signals for a MeHg-spiked eluant (data not shown). The combination of H2O2 with UV-irradiation yielded the highest signal and made a negligible contribution to the HgII blank, which is essential because all strategies for reagent cleanup involve the use of electron-donor ligands that cannot be used with strong oxidants. Quenching excess oxidant from the UV-PCO process is necessary for SnII to efficiently reduce HgII. Absent an antioxidant, SnII had to be added at such high concentrations that a gray-black precipitate, possibly SnS(s), formed in the aqueous stream from the reaction of SnII with TU breakdown products. The presence of the solid altered sensitivity and eventually blocked flow. The use of higher peroxide levels eliminated the precipitate but decreased system sensitivity by consuming more SnII. Use of hydroxylamine as the antioxidant, as in Method 1631 (38), was investigated but caused a gas phase to form that destabilized the baseline. Aqueous ascorbate solution quenched the byproducts of photooxidation and produced no interfering solid or gaseous phases. Thus, ascorbate permits high sensitivity at a lower [SnCl2], thereby both reducing the risk of Sn precipitation and stabilizing the day-to-day sensitivity of the system. The efficiency of MeHg photooxidation in the UV-PCO step was evaluated by comparing peak areas in chromatograms of injections containing precisely equal amounts of MeHg and HgII. The mean ratio of MeHg:HgII peak-areas from the seven injections was 1.000 ( 0.056 (Figure 5), and neither the deviation from a 1:1 slope nor the intercept were significant. These results demonstrate both complete breakdown of MeHg in the UV/H2O2 photooxidation step and the system’s ability to simultaneously quantify both MeHg and HgII.
TABLE 1. Methyl- and Total Hg in Certified Reference Materials Analyzed by HgTU/IC with SL or TT Loading MeHg (µg g-1)
N
HgT (µg g-1)
SLa TTa CVb
Dogfish Liver (DOLT-2) 0.734 ( 0.041 5 2.20 ( 0.15 0.730 ( 0.038 3 0.693 ( 0.053 2.14 ( 0.28
SLa TTa CVb
Dogfish Muscle (DORM-2) 4.29 ( 0.03 3 4.34 ( 0.04 4.10 ( 0.18 3 4.47 ( 0.32 4.64 ( 0.26
SLa TTa CVb
Polluted Estuarine Sediment (BCR 580) 0.0714 ( 0.0015 3 135 ( 1.3 0.0718 ( 0.0045 2 0.0755 ( 0.0037 132 ( 3
a
(1 SD of measurements.
b
N 3
3
3
Most recent certified values (95% CL.
FIGURE 6. pH-dependence of MeHg (b) and HgII (O) retention by MPSG column. Solutions containing ∼100 pg of each Hg species in 2% m/v TU were adjusted to the indicated pH, loaded onto the column (Alltech) on-line, and eluted normally (pH ≈ 0). (-) Calculated % Hg retained ) rSi-SH/(rSi-SH + rTU) × 100 with rL from Figure 1. Sample Loading and On-line Preconcentration. Sample loop injection represents the most convenient sample introduction method, but brings with it the potential for large matrix influences upon cold-vapor generation. For example, we have found that different solution chemistries produce slightly different, albeit consistent, sensitivities; thus sample and standard solutions must be matched. This can be complicated by sample matrix ligands introduced during sample preparation (see below). Although isolation of MeHg from the matrix via distillation or solvent extraction is a common solution to this problem, only 100 µL out of a typical 5-mL volume of sample extract can be injected for analysis. The thiol trap permits complete introduction of such extracts, providing a 50-fold lowering of detection limits. The effective pH range of the MPSG resin in the thiol trap was tested by loading it with a MeHg/HgII standard in pH-adjusted, TU solutions (Figure 6). The resin quantitatively retained (at pH g 3) and eluted (at pH ≈ 0) both forms of Hg, in agreement with other work on thiol-based preconcentration (10, 31) and with the model calculations (Figure 1). For routine work, samples are buffered with sodium citrate to pH 3-4 before loading onto the thiol trap (Figure 6). Although the resin should get stronger with increasing pH up to the pKA of the thiol groups (∼9.5), the pH of the loading solution was kept as low as is practicable to avoid sorption problems. When eluting from the thiol-trap, peak widths were equivalent to (Omnifit) or narrower than (Alltech) directly injected samples (data not shown). Flow Injection System for HgT Analysis. Removing the speciation column and using the system for HgT analysis gave acceptable accuracy for oxidized digests of a variety of samples (Table 1). The use of the UV-photochemical oxidation system may allow flow-injection analysis of digests with less dilution than is recommended in Method 1631, thus allowing low detection limits despite a small sample injection volume (100 µL). Also, because the system completely oxidizes MeHg (see above), addition of BrCl or HOBr after acid digestion should not be necessary, thus lowering reagent blanks. Calibration, Linearity, and Precision. Because the system sensitivity is equal for both HgII and MeHg, it may be calibrated with standards containing either or both. Peak areas from a set of MeHg standards containing up 5000 pg Hg appeared to be fitted well (r2 ) 0.9994) using simple, unweighted linear regressions (data not shown). However,
FIGURE 7. Weighted residuals (actual - predicted)/σAi for 17 MeHg calibration curves (N ) 95). Lines represent modeled coefficient of variation (σi/Ai) of peak area. Curves with MeHgmax < 200 pg fitted using a linear model; remaining curves fitted using Ai ) S‚Ksat‚MeHgT/(Ksat + MeHgT) where MeHgT ) MeHgSTANDARD + MeHgBLANK. Inverse-square weighting (wi ) 1/σAi2; see eqn. 1) applied in both cases. Fitted MeHgBLANK in pg (and number of curves with each value) are: 0 (3), 0.3 (2), 0.6 (6), 0.9 (2), 1.2 (1), 2 (2), and 5 (1). Statistical analyses were conducted using SAS v9.0. careful calibration of an analytical system requires understanding the relationship of the random error component of the signal on its magnitude so that regression residuals may be properly weighted. To this end, we conducted a global analysis using this curve along with 16 additional curves of mixed standards (containing equal amounts of MeHg and HgII) obtained over 10 months using both sample introduction methods. Plots of Studentized-residuals and tests of heteroscedasticity using SAS (White test) demonstrated that modified inverse-square weighting was necessary, implying that the residuals approach a constant relative error, rather than the constant absolute error assumed in unweighted linear regression (Figure 7). The residual weighting scheme corresponds to modeling the variance in measurement of the ith measured MeHg peak area (Ai) as:
σAi2 ) (1.25 + 0.025‚Ai)2
(1)
At the average system sensitivity, this translates into a SD of ∼0.1 pg at the detection limit and a 2.5% CV at higher MeHg. This estimate of system precision agrees closely with the CV VOL. 39, NO. 13, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4979
TABLE 2. Representative Systems for MeHg Analysis and the Absolute Detection Limit (DL), and Quantitative Sample Preconcentration Capability (Q) of Each ligand ethyl
GC
SPDCa
HPLC
Cys 2ME
HPLC HPLC
ClBrTU
IC GC IC
a
FIGURE 8. Matrix-independence of measured MeHg in DOLT-2 after digestion+solvent extraction. MeHg per volume of injected sample (0, 9) and per dry mass of tissue (O, b). Extracts loaded via SL (9, b) or TT (0, O). Certified value (-); 95% confidence limits (- -). of 2.6% and 2.5% for MeHg and HgII, respectively, for seven replicate sample loop injections (Figure 3). In the weighted regression, curves with mixed standards were linear below ∼200 pg Hg, but the peak overlap that occurred for mixed standards containing >200 pg Hg necessitated the use of a saturation equation for these curves (Ksat ) 104.0(0.2 pg). Much less curvature (Ksat ) 105.0(0.2 pg) was observed in the MeHgonly standard curve even at high concentrations. The negative departure from linearity reaches ∼10% at 1000 pg for mixed standards and ∼5% at 5000 pg for MeHg-only standards. Because the non-linearity is caused by the running together of chromatographic peaks, the easiest way to minimize or eliminate it is by either working at levels of 50% (data not shown). Although solvent extraction successfully eliminates matrix influences on analysis by ethylation/GC (39, 40), this is not always the case for HPLC-based methods. For example, solvent-extracts of reference materials contained matrix interferences that precluded external calibration of the Hg(2ME)/RP-HPLC-CVAFS system (17, 18). To demonstrate the matrix-independence of the HgTU/IC system, quantities of DOLT-2 CRM ranging from 50 mg, as is typical, to 500 mg were digested, solvent-extracted, and analyzed. Measured MeHg concentrations in the extracts were directly proportional to the mass of tissue digested over the range 0.050.5 g with both SL- and TT-loaded samples (Figure 8). Thus, no matrix interference with analysis is evident for solventextracted samples. The DOLT-2 results together with analyses of MeHg in dogfish muscle and marine sediment reference materials verify the accuracy of the HgTU/IC-CVAFS system (Table 1). These CRM results are unbiased (41), and the measurements obtained using SL- and TT-loading were not significantly different from each other (P < 0.05). Our MeHg recoveries - mean 97% and range 91-105% - compare well with those obtained - mean 95% and range 90-103% - using the same leaching/solvent extraction procedure and analysis by ethylation/GC-CVAFS for similar reference materials (40). 4980
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 13, 2005
separation
detector
DL (pg)
Q
ref
CVAFS ICP-MS CVAAS CVAFS ICP-MS ICP-MS CVAFS ICP-MS ICP-MS ECD CVAFS
