Anal. Chem. 2000, 72, 2222-2228
On-Site Analysis of Arsenic in Groundwater Using a Microfabricated Gold Ultramicroelectrode Array Rosemary Feeney and Samuel P. Kounaves*
Department of Chemistry, Tufts University, Medford, Massachusetts 02155
Increasingly, arsenic is being found in drinking water in many parts of the world such as Bangladesh, India, England, and Thailand and in the United States in California, Oregon, Massachusetts, Maine, and New Hampshire.1-6 Much of the increase in reports about arsenic contamination stems from the calamity in Bangladesh. It has been reported that about 44 districts in Bangladesh have arsenic contamination, and a population of 3580 million people are at risk of arsenic toxicity.3 This has been called one of the biggest outbreaks of poisoning this century. Environmental arsenic contamination occurs mainly from industrial processes such as smelting of other metals, application of arsenical pesticides/herbicides, and power generation from coal or geothermal sources.7 Other common uses are in the semiconductor and electronic industries. In contrast, the source of arsenic
in places such as Bangladesh and New Hampshire has been attributed to geochemical mechanisms involving weathering bedrock or from the oxidation of arsenic in the aquifer sediment upon contact with oxygenated waters. Arsenic, has long been an EPA Priority Pollutant and is an important environmental concern because of its toxicity to a broad spectrum of organisms at the parts-per-billion (ppb) level. Arsenic can exist in four oxidation states, -3, 0, +3, and +5. In oxygenated waters, As5+ species are predominant and more thermodynamically stable. However, under slightly reducing conditions and/or lower pH, As3+ species become more stable. Moreover, up to 10% of total arsenic has been found to exist as As3+ in uncontaminated surface and deep ocean waters.8 It is the chemical form of arsenic that determines its toxicity. The trivalent inorganic forms, such as arsenic trichloride, arsenic trioxide, and arsine, are highly toxic and are 60 times more poisonous than the metal, its pentavalent salts, or organoarsenic compounds.8-10 There is a large body of evidence to support the cancer-causing nature of ingested arsenic and its connection to large numbers of people suffering from liver, lung, and kidney disease as well as adverse dermal effects such as hyperkeratosis and depigmentation.7,11 Such evidence has lead the U.S. EPA to consider lowering the present drinking water standard for arsenic from 50 to 5 µg/L.11 There are several EPA-approved methodologies for the determination of arsenic in groundwater. These include inductively coupled plasma atomic emission spectrometry (ICP-AES), graphite furnace atomic absorption spectrometry (GFAAS), and hydride generation atomic absorption (GHAA). Such technologies are laboratory-based and time-intensive and can lead to large capital cost for multisample analysis. In light of the number of contaminated sites still to be identified and remediated, rapid and lowcost field screening and monitoring systems are needed. Such techniques are an attractive alternative. However, to date there are no commercially available portable instruments that can be routinely used as such. Voltammetric stripping techniques are readily amendable for field screening, providing accurate measurements of low concentrations with rapid analysis times and a low-cost/-weight instru-
* Corresponding author: (e-mail)
[email protected]. (1) Peters, S. C.; Blum, J. D.; Klaue, B.; Karagas, M. R. Environ. Sci. Technol. 1999, 33, 1328. (2) Dhar, R.; Biswas, B.; Samanta, G.; Mandal, B.; Chakraborti, D.; Roy, S.; Jafar, A.; Islam, A. Curr. Sci. 1997, 73, 48. (3) http://www.dainichi-consul.co.jp/english/arsenic/as72.txt. (4) http://www.epa.gov/region02/superfnd/site_sum/02002096.htm. (5) Mitchell, P.; Barr, D. Environ. Geochem. Health 1995, 17, 57. (6) Welch, A. H.; Lico, M. S.; Hughes, J. L. Gound Water 1988, 26, 333.
(7) Chang, L. W.; Magos, L.; Suzuki, T. Toxicology of Metals; CRC Press: New York, 1996. (8) Cullen, W.; Reimer, K. Chem. Rev. 1989, 89, 713. (9) Mabuchi, K.; Lilienfeld, A. M.; Snell, L. M. Arch. Environ. Health 1979, 34, 312. (10) Morita, M. Pure Appl. Chem. 1992, 64, 575. (11) Smith, A. H.; Hopenhayn-Rich, C.; Bates, M. N.; Goeden, H. M.; HertzPicciotto, I.; Duggan, H. M.; Wood, R.; Kosnett, M. J.; Smith, M. T. Environ. Health Perspect. 1992, 97, 259.
Rapid on-site analysis of arsenic in groundwater was achieved with a small battery-powered unit in conjunction with a microfabricated gold ultramicroelectrode array (AuUMEA). The sensor, consisting of 564 UME disks with a unique gold surface created by electron beam evaporation, was demonstrated to be highly sensitive to low-ppb As3+ using square wave anodic stripping voltammetry. The influence of the square wave frequency, pulse amplitude, and deposition potential on the arsenic peak stripping current was investigated. Varying those theoretical parameters yielded results surprisingly similar to those for the thin Hg film case. The performance of the Au-UMEA was evaluated for reproducibility and reliability. Three stability tests showed an average relative standard deviation of 2.5% for 15 consecutive runs. Limits of detection were investigated, and 0.05 ppb As3+ could be measured while maintaining a S/N of 3:1. Interference studies were performed in the presence of 50-500 ppb of Cu2+, Hg2+, and Pb2+. On-site analysis of groundwater containing arsenic was performed with a small battery-powered potentiostat. Quantification was done through standard additions, and these results were compared to the standard EPA methodology.
2222 Analytical Chemistry, Vol. 72, No. 10, May 15, 2000
10.1021/ac991185z CCC: $19.00
© 2000 American Chemical Society Published on Web 04/13/2000
ment. These methods provide an efficient and reliable way to analyze for As3+ at the ppb levels found in natural waters. In general, the methods for determining As3+ electrochemically have involved cathodic or anodic stripping voltammetry using a pulsed waveform.12-23 More recent papers describe analysis of As3+ at low-ppb levels using mercury electrodes15-18,24 with the addition of copper or selenium.20 Macrosized gold film electrodes13,14,21 and solid gold electrode substrates19,22,25 have also been used. This paper reports on field-portable instrumentation, which uses microfabricated gold arrays, and its demonstration at a contaminated site in central Maine. These arrays offer several benefits such as uniform ultramicroelectrode geometries, cost efficiency, and applicability for use in field-portable or in situ instrumentation. In addition, the electron beam-evaporated gold surface of these ultramicroelectrodes (UMEs) appears to offer unique characteristics for the formation of monolayer metallic films. Use of single ultramicroelectrodes for arsenic determinations has been very limited.25-27 However, ultramicroelectrode arrays (UMEAs) possess inherent advantages for arsenic determinations, such as low noise level, amplification of the signal while keeping UME behavior, background current rejection, and potential incorporation into field-portable instrumentation. Moreover, the beneficial use of UMEAs for heavy metal analysis has been well documented in the literature.28-31 The gold ultramicroelectrode array (Au-UMEA) was used to demonstrate a quick and reliable way to determine As3+ in aqueous solutions using square wave anodic stripping voltammetry (SWASV) after deposition of an As0 monolayer. The Au-UMEA provides a unique gold surface created by electron beam evaporation. The linear dynamic range of the Au-UMEA and its response were characterized, the SWV parameters and the electrolyte solution composition were optimized, and interference studies were performed by addition of mercury, lead, and copper to solutions containing arsenic. A portable battery-powered device was used to demonstrate rapid on-site analysis at a site known to have significant arsenic concentration. Results were compared to EPA methodology. (12) Huang, H.; Dasgupta, P. K. Anal. Chem. Acta 1999, 380, 27. (13) Sun, Y.; Mierzwa, J.; Yang, M. Talanta 1997, 44, 1379. (14) Viltchinskaia, E.; Zeigman, L.; Garcia, D.; Santos, P. Electroanalysis 1997, 9, 633. (15) Li, H.; Smart, R. Anal. Chim. Acta 1996, 325, 25. (16) Greulach, U.; Henze, G. Anal. Chim. Acta 1995, 306, 217. (17) Zima, J.; van den Berg, C. Anal. Chim. Acta 1994, 289, 291. (18) Sadana, R. Anal. Chem. 1983, 55, 304. (19) Bodewig, F.; Valenta, P.; Nurnberg, H. Fresenius′ Z. Anal. Chem. 1982, 311, 187. (20) Holak, W. Anal. Chem. 1980, 52, 2189. (21) Davis, P.; Dulude, G.; Griffin, R.; Matson, W.; Zink, E. Anal. Chem. 1978, 50, 137. (22) Forsberg, G.; O’Laughlin, J.; Megargle, R.; Koirtyohann, S. Anal. Chem. 1975, 47, 1586. (23) Myers, D.; Osteryoung, J. Anal. Chem. 1973, 45, 267. (24) Kotoucek, M.; Vasicova, J.; Ruzicka, J. Mikrochim. Acta 1993, 111, 55. (25) Hua, C.; Jagner, D.; Renman, L. Anal. Chim. Acta 1987, 201, 263. (26) Rievaj, M.; Mesaros, S.; Bustin, D. Chem. Pap. 1994, 48, 91. (27) Huiliang, H.; Jagner, D.; Renman, L. Anal. Chim. Acta 1988, 207, 37. (28) Herdan, J.; Feeney, R.; Kounaves, S. P.; Flannery, A.; Storment, C. W.; Kovacs, G. T. A.; Darling, R. B. Environ. Sci., Technol. 1998, 32, 131. (29) Belmont, C.; Tercier, M.; Buffle, J.; Fiaccabrino, G. K.-H., M. Anal. Chim. Acta 1996, 329, 203. (30) Wang, J. Anal. Chem. 1995, 67, 1481. (31) Olsen, K.; Wang, J.; Setiadji, R.; Lu, J. Environ. Sci. Technol. 1994, 28, 2074.
EXPERIMENTAL SECTION Apparatus. Characterization experiments were perform with cyclic voltammetry (CV) and square wave anodic stripping voltammetry using a model 263 potentiostat/galvanostat (EG&G PAR, Princeton, NJ) interfaced to a DEC p420-SX microcomputer using the model 270 software (EG&G). All voltammetric experiments were carried out using a three-electrode cell which included a Ag/AgCl (saturated)//(3 M NaCl) reference electrode (BAS, West Lafayette, IN) to which all potentials are referenced, a Pt wire counter electrode, and the working electrode, which consisted either of the Au-UMEA or of a single 10-µm-diameter solid gold microelectrode (Cypress System, BAS). On-site analysis was achieved with a portable CS-3000 potentiostat/galvanostat system (Cypress Systems, Lawrence, KS) connected to a laptop computer. The potentiostat can operate for up to 6 h on battery power. The software allows for programmable waveforms, methods development, and SW frequencies of up to 1000 Hz. Data acquisition is at a 1-MHz throughput. The overall physical dimensions of the unit used are 28 × 28 × 5 cm and it weighs only 5.4 kg, while the electrometer is 15 × 6 × 11 cm and weighs 1 kg (more compact units are also available from Cypress Systems). A shield cable for the leads minimize noise in the system. An inverted polarizing microscope (Metaval-H, Jenoptik Jena GmbH) equipped with a video imaging system was used to record all optical observations in laboratory. Surface morphologies were recorded with a Dimension 3000 atomic force microscope (Digital Instruments, Santa Barbara, CA) and a JXA-840 scanning microanalyzer (JEOL, Peabody, MA). Microlithographic Fabrication of the Au-UMEA. Only a brief general description of the microfabrication processes for the gold ultramicroelectrode array chips is given here. Starting with a standard 5-in. silicon wafer, a 5000-Å oxide layer was thermally grown over the entire silicon wafer surface. Sequential electron beam evaporation steps were used to deposit an adhesion layer of titanium (100 Å) followed by a gold layer (5000 Å) in a specific pattern to be composed of the ultramicroelectrode surface, traces, and bond pad. A second layer of titanium was deposited on top of the gold to ensure adhesion to the insulating layer. Photolithographic stenciling was done to outline the UME pattern and an argon ion beam etch to transfer the stencil. Reactive rf-diode sputter deposition was used to deposit a 5000-Å SiO2 insulating layer. Reactive ion etching was performed to remove the SiO2 in the desired pattern for the UMEs and the bond pad. Additionally, an argon ion beam was used to remove the titanium layer on the top surface to expose the disks and the bond pad. The finished wafer was cleaned and diced into 3.4 × 3.1 mm chips that were glued onto a custom-designed printed circuit board (CFC, Waltham, MA) with an epoxy (EpoTek 905, Epoxy Technology, Billerica, MA). Electrical connection between the chip and the circuit board was made by a 1.25-µm gold wire (99.99% Williams Advanced Materials, Buffalo, NY). The gold wire was protected by a globtype epoxy (Orion Research, Inc. Beverly, MA) that was cured for 2 h at 50 °C. Figure 1 shows an SEM image of the entire array chip (A) and an AFM image of an individual UME disk (B). Each ultramicroelectrode is circular and 12 µm in diameter with a center-to-center spacing of 58 µm and is recessed by ∼0.5 µm. Analytical Chemistry, Vol. 72, No. 10, May 15, 2000
2223
Figure 2. Cyclic voltammograms of 6 mM ferricyanide in 0.1 M KNO3 at a scan rate of 1 mV/s using a gold UMEA and a single 10µm-diameter gold UME. Current for the single UME is multiplied by 100.
Figure 1. (A) SEM (20 keV, 30×) image of a typical UME array, with 564 gold UME disks with a center-to-center spacing of 58 µm, and the gold bond pad. (B) AFM image of an individual 12-µmdiameter UME disk.
There are 564 gold UME disks on the chip with a total electroactive area of 6.37 × 10-4 cm2. Reagents and Parameters. All solutions were prepared with 18 MΩ‚cm deionized water from a Barnstead Nanopure System (Barnstead Co, Dubuque, IA). All glassware was stored in 8 M HNO3 for a week and rinsed thoroughly with 18 MΩ‚cm water. A 10 ppm As3+ working solution was prepared from a 1000 µg/ mL arsenic AAS standard solution (Alfa Aesar) and diluted with 5% HNO3. Acids were of trace metal purity (Fisher). All other solutions were prepared with ACS grade reagents. Prior to all experiments, the Au-UMEAs were conditioned by cycling the potential 10 times between 0 and 1.5 V at 100 mV/s in 0.1 M H2SO4. Unless otherwise stated, a frequency (f) of 150 Hz, pulse amplitude (Esw) of 25 mV, step height (∆Es) of 2 mV, and deposition potential (Edep) and time (tdep) of -0.4 V for 80 s were chosen as the SWASV parameters. A conditioning potential and time (30 s at 0.55 V) was added to renew the gold surface. For the analysis in groundwater, a Nafion coated Au-UMEA was used to protect the surface. The Nafion film was applied with a microsyringe using a 5% Nafion 117 solution (Fluka) and allowed to dry for 1 h. The array was then place in a 120 °C oven for a few minutes to cure the Nafion film. Electrochemical characterizations indicated that As3+ does not adsorb on the Nafion. RESULTS AND DISCUSSION UME Array Characteristics. The general UME response of the Au-UMEA was first characterized and compared with that of 2224 Analytical Chemistry, Vol. 72, No. 10, May 15, 2000
a single 10-µm-diameter solid gold ultramicroelectrode using cyclic voltammetry in a solution containing 6 mM K3Fe(CN)6 and 0.1 M KNO3. Figure 2 shows the sigmoidal responses obtained for each electrode at a scan rate of 1 mV/s and is indicative of the diffusion-controlled process dependent on the size and geometry of the ultramicroelectrode.32 In this CV, the current for the UMEA was 500 times greater than that of the single UME, demonstrating the amplification of the current signal and the benefit of using an array. The separation between the forward and reverse scans for the UMEA shows that at this slow rate, even though there is insufficient separation between the UMEs to prevent some shielding effects due to the partial overlap of the diffusion layers of adjacent electrodes,33 the behavior and benefits of the array of UMEs are still maintained. Studies of several gold UMEAs support the assumption that these gold arrays give reproducible responses. It is important to note that, unlike a solid UME which has to be regularly polished, the microfabricated UMEA’s surface is not treated in any way after microfabrication and is used as is. It is thus remarkable that the monolayer stripping response is the same as that of the polished UME. The stripping signal for As3+ using a Au-UMEA and a single Au-UME was also compared using SWASV. The deposition time for the UME was 180 s whereas the time for the UMEA was 80 s. The stripping peaks were Gaussian and approximately 3.3 nA and 2.078 µA in height, respectively. Characteristics of Deposited Arsenic. Being a semiconductor, As0 is effectively nonconducting and does not easily electrodeposit on itself.21 Thus, after the first layer covers the electrode surface, the resistance of the monolayer is sufficient to prevent further electron transfer at the applied potential. This means that an ASV analysis can only be performed within this “window” of monolayer coverage. However, as shown below, this limitation can be overcome and a reasonable dynamic range is accessible. For the 80-s deposition times used here, only a fraction of the array surface is covered with As0. The amount of As0 deposited on the UMEA was estimated from the stripping charge, which was obtained by linear scan voltam(32) Heinze, J. Angew. Chem., Int. Ed. Engl. 1993, 32, 1268. (33) Feeney, R.; Herdan, J.; Nolan, M.; Tan, S. H.; Tarasov, V. V.; Kounaves, S. P. Electroanalysis 1998, 10, 89.
Table 1. Comparison of the Responses for the UME and the UMEA with Various SW Amplitudesa UME
UMEA
Esw (mV)
peak potential (mV)
stripping peak ht (nA)
peak potential (mV)
stripping peak ht (µA)
25 50 75 100
229 213 176 156
2.07 2.84 3.21 4.29
227 202 159 152
2.08 3.12 3.39 3.13
a These studies were done in 2 M HCl. Step height, 4 mV; deposition time, 80 (UMEA) and 180 s (UME); initial potential, -0.4 V.
Figure 3. Effect of the deposition potential on the peak stripping current of 100 ppb As3+ in 2 M HCl with SWASV parameters: Esw ) 25mV, f ) 150 Hz, ∆Es ) 2 mV, and tdep ) 80 s (b). Theoretical curves are calculated from the equation Edep ) E1/2 - (0.059/n) log(i/id - i) for the best fit (s), with E1/2 ) - 0.24 V with a slope of 150 mV, and with the same E1/2 and id for n ) 3 (- - -).
metry at a scan rate of 100 mV/s. The stripping peak of arsenic using a 80-s deposition time, and 100 ppb As3+ in 2 M HCl resulted in a total charge of 141 nC. Using the assumptions of a densely packed atomic layer, an atomically flat electrode surface, and an atomic radius for As0 of 1.39 × 10-8 cm, the calculated total charge should be 504 nC. However, from the stripping experiment, the measured charge would imply that only ∼16% of the UMEA surface is covered by the monolayer of As0. Other papers in the literature support the formation of an arsenic monolayer on gold.21,34 Further, the analytical utility of monolayer peaks for quantification of Cu and Se at gold electrodes has been demonstrated.35,36 One way to confirm a monolayer deposition on the electrode surface is to vary the tdep where submonolayer coverage occurs. For a given Edep, the relationship between tdep and the stripping current remains linear until the surface deposition exceeds that of a monolayer. For As, once the monolayer is formed, there is a rigid leveling and eventual decrease in stripping current due to the insulating properties of As0. This experiment was performed with arsenic at a Au-UMEA in 2 M HCl at deposition potential -0.4 V. Linearity was observed in the region 7 s > tdep< 120 s and the characteristic nonlinear form at 120 s > tdep < 500 s. This result further lends weight to the contention that a monolayer of As on the Au-UMEA surfaces does not dissolve in the Au surface as clearly indicated by the “saturation plateau”. To optimize and better characterize the deposition of arsenic at the Au-UMEA, a plot of peak stripping current vs deposition potential was made (Figure 3). The experimental data (b) was collected by applying deposition potentials (Edep) in the range from 0.0 to -0.6 V, in a solution containing 100 ppb As3+ and 2 M HCl. Each point is an average of three runs. The theoretical curves were calculated using the relationship Edep ) E1/2 - (0.059/n) log(i/id - i). For the best fit (s) we obtained an E1/2 of approximately -0.24 V and a slope of 0.150 V equal to n ) 0.4 electron. The theoretical curve for a 3-electron reduction for the (34) Dinan, T. E.; Jou, W. F.; Cheh, H. Y. J. Electrochem. Soc. 1989, 136, 3284. (35) Bonfil, Y.; Brand, M.; Kirowa-Eisner, E. Anal. Chim. Acta 1999, 387, 85. (36) Andrews, R. W.; Johnson, D. C. Anal. Chem. 1975, 47, 294.
same E1/2 and id is also shown (- - -). It can be concluded that the electron-transfer process of this system is slow and approaches electrochemical irreversibility as evidenced by a less steep slope. This is supported by the square wave voltammograms obtained during arsenic deposition. In all experiments performed, only a minimal reverse current peak was seen. In this way, only a minimal amount of arsenic undergoes the re-reduction during the application of the square wave pulse and is a sign of a sluggish electrontransfer process. In Figure 3, a Edep more anodic than ∼0 V yields no significant accumulation of arsenic on the UMEs. As the deposition potential is made more cathodic, the amount of arsenic deposited increases. Eventually, a sigmoid-shaped curve with a plateau should be obtained. However, for Edep more cathodic than -0.4 V, there is a decrease in the deposition efficiency because the reduction of As3+ starts to compete with the reduction of water and the associated production of H2, further blocking the surface and decreasing the current. It is also possible that at more cathodic potentials some of the elemental arsenic could be converted to arsine (As3-). Therefore, a deposition potential of -0.4 V was chosen as it yielded the highest current response with least interference from other reactions. Influence of SWASV Parameters on the Arsenic Stripping Peak. According to the theory for SWASV at a thin mercury film electrode,37 an increase in the pulse amplitude should cause a shift of the stripping peak to more negative values and an increase in the stripping net peak current (∆ip), indicating an increase in the reversibility of the system. In addition, for a thin film it predicts that ∆ip ∝ f; thus an increase in SW frequency should result in a linear increase in stripping peak height and ∆ip should have a constant peak half-width n(W1/2) of 100 mV. Even though a monolayer of deposited As0 on the solid Au-UMEA is used in this case, the results when varying Esw and f are surprisingly similar to those for the thin Hg film case. Current work in our laboratory on the theory of SWASV of a monolayer at solid electrodes suggests that these two cases may indeed give very similar responses. Table 1 shows the effects of varying the SW pulse amplitude on the stripping peak for the UMEA and UME for 100 ppb As3+ in 2 M HCl. For both electrodes, the monolayer arsenic peak potentials became more negative with increasing SW amplitudes and the stripping peak heights increased for amplitudes of e100 mV. However, a leveling-off effect was noticeable in the UMEA data, suggesting an optimum square wave amplitude of 50-75 mV. (37) Kounaves, S.; Chandresekhar, P.; Osteryoung, J. Anal. Chem. 1987, 59, 386.
Analytical Chemistry, Vol. 72, No. 10, May 15, 2000
2225
Figure 4. SWASV of 100 ppb As3+ in solutions containing (A) 4, (B) 3, (C) 2, and (D) 1 M NaCl, with 0.01 M HNO3. Parameters were Esw ) 25 mV, f ) 150 Hz, ∆Es ) 2 mV, and tdep ) 80s.
For SWASV at a thin-film mercury electrode, increasing the frequency, which is equivalent to using a smaller pulse width, results in a thinner diffusion layer and an increase in the flux/ current and leads to greater sensitivity. Varying the SW frequency from 35 to 210 Hz resulted in a 6-fold linear increase of ∆ip (R2 ) 0.997, x, y intercept ≈ 0, 0) as predicted by thin-film theory. A frequency of 150 Hz was chosen for the analytical procedures because it provided a reasonable enhancement of the stripping peak with minimal noise. Calibration and Linear Regression Plots. To optimize the stripping signal ∆ip, several supporting electrolytes were evaluated with the Au-UMEA. These included the following: a 0.002 M acetate buffer (pH 4.5), 0.05, 0.1, and 1 M H2SO4, and 0.1-4 M HCl. For all but HCl, broad non-Gaussian and nonreproducible arsenic stripping peaks were obtained. The stripping peaks for As3+ with the best Gaussian shape, reproducibility, and sensitivity were obtained in a solution with low pH and chloride concentration of >0.5 M. Increasing the chloride concentration by addition of HCl resulted in a significant increase and cathodic shift of the stripping peak. This increase can be attributed to either the increase in the chloride concentration, the ionic strength, or the decrease of the pH. All other parameters were kept constant. Of these factors, ionic strength should have little influence when UMEs are used. The above results demonstrate that low pH and a chloride concentration greater than 0.5 M is the best stripping medium for arsenic. The more cathodic stripping potentials are a direct consequence of the more facile oxidation of As0 from the gold substrate. However, as the chloride concentration is increased from 0.5 to 2 M, the pH does not change significantly, which suggests that the chloride concentration is the major influence. To confirm this, studies were performed using 100 ppb As3+ in a 0.01 M HNO3 solution with the addition of NaCl to give 2, 3, and 4 M Cl-. As shown in Figure 4, Ep shifts cathodically as the concentration of chloride is increased, but no significant increase in ∆ip is observed. Similar results were also obtained in a lower pH solution containing 1 M HNO3. There was a significant difference between ∆ip obtained with the 2 M hydrochloric and the sulfuric or sodium acetate media. The peaks were broad and at more positive potential probably due to slow reaction kinetics. In HCl, Gaussian-shaped peaks were obtained with the smallest width at half-height (e.g., 112 mV). 2226 Analytical Chemistry, Vol. 72, No. 10, May 15, 2000
Figure 5. Voltammograms for standard additions of (A) 0-500 and (B) 0-150 ppb As3+ in 2 M HCl. In both cases, SWASV parameters were Esw ) 25 mV, f ) 150 Hz, ∆Es ) 2 mV, tdep ) 80 s, tcond ) 30 s, and Econd ) 0.5 V.
This is indicative of faster electron kinetics. In chloride media at concentrations of >0.1 M, the chloride ions strongly complex arsenic to form AsCl3. In the sulfuric media, the SO42- ions are bulkier and less complexing, leading to a lower coordination number with the As3+. Thus more energy is necessary for dissociation. Either way, As3+ appears to prefer an excess of chloride ions in solution when SWASV is being performed. Concentrations higher than 2 M HCl offered no significant increase in ∆ip, and there was also more irreproducibility that may be due to adsorbed chloride on the gold substrate. Calibration plots were made using the gold UMEA in 2 M HCl and standard additions of arsenic. Each point was an average of three consecutive runs. Two separate calibration experiments were performed. Both, for 0-500 (Figure 5A) and 0-150 ppb (Figure 5B), displayed excellent linearity. The arsenic stripping peaks were Gaussian and were between 0.1 and 0.2 V in the voltammogram. The peak at -200 mV (Figure 5B) was confirmed, by standard addition, to be Pb contamination. Combining these two plots yields a linear regression coefficient of 0.991 and a calibration slope of 0.044 µA/ ppb. Each plot was achieved with a different Au-UMEA and is evidence of the reproducibility of these electrodes. The limit of detection (LOD) was also investigated for these UMEAs using both statistical and experimental approaches. Twenty runs in a blank HCl solution were performed. The mean and standard deviation of the blank signals were used in conjunction with the slope from the previously obtained calibration plot, to calculate a detection limit (at a 95% confidence level).38 A theoretical limit of 0.013 ppb was calculated. In comparison, an experimental limit was measured in solutions of 0.01 and 0.05 ppb As3+ that were carefully prepared in 2 M HCl. A peak was measured in the 0.05 ppb solution. Standard additions made to this solution yielded a linear relationship (Figure 6). The two concentration values are comparable and lend credence to the measured minimum concentration. However, if 10 times the standard deviation of the blank is used as a more reasonable limit of quantification (LOQ), the theoretical value would be 0.13 ppb. (38) Skoog, D. A.; Holler, F. J.; Nieman, T. A. Principles of Instrumental Analysis, 5th ed.; Harcourt Brace College Publishers: Philadelphia, 1998.
Figure 6. Voltammograms for standard additions of 0.0500-0.3 ppb As3+ in 2 M HCl. Parameters were Esw ) 25 mV, f ) 150 Hz, ∆Es ) 2 mV, and tdep ) 300 s. Linear calibration results: R2 ) 0.990, a slope of 4.40 µA/ppb, and a y-intercept equal to 1.33. Three runs were performed at each concentration and the S/N was at least 3:1.
Figure 7. Normalized current signal for a solution containing 100 ppb As3+ in 2 M HCl in the presence of increasing concentrations of Pb2+ (diagonal), Hg2+ (cross-hatch), and Cu2+ (solid black).
It should be noted that a tdep of 300 s was used instead of the tdep of 80 s used in the calibration plots. Therefore, the slightly lower value in the experimental case is understandable. In either case, the LOQ associated with this technique is at or near 0.1 ppb. For analytical utility in natural samples, values in the lower range of 1-5 ppb are necessary. The Au-UMEA sensor could reliably measure such low ppb levels. Overall, the data show excellent linearity from low ppb and extending to higher concentrations. Stability of the Stripping Peaks. Careful electrode conditioning was found to enhance the quality and reproducibility of the peak. In chloride-containing solutions, the conditioning potential should not exceed 0.55 V since at more positive potentials, chloride can adsorb or an oxide can form and alter the surface of the gold electrode. Three stability tests were done with two different UMEAs that had been previously used to determine the precision the measurements. In the first study, 30 consecutive runs were made using 100 ppb As3+ in 2 M HCl. For this sample, ∆ip slowly decreased with each successive run. The first 10 runs showed a relative standard deviation of 1.9%. By the 30th run, ∆ip had decreased by 30%. The array was examined with a microscope after this experiment and ∼50 UMEs had deposits on them. Repeated use in high-chloride media can lead to fouling of the electrode surface by the formation of a layer of gold chloride on some of the gold UMEs but only after several weeks of use. Two other stability tests were performed with 100 ppb As3+ in 2 M HCl using the same parameters as above. The values for ∆ip were slightly lower and gave relative standard deviations of 1.3 and 3.2%, respectively, for 15 consecutive runs. The third study, with the higher relative standard deviation, was complicated by the presence of a small amount of mercury and its incomplete removal from the gold surface. This added to the diminution of the arsenic peak. A conditioning potential of 0.55 V was used; however, under these conditions, it did not give more reproducible peaks. Gold is known to dissolve in mercury, which may lead to the degradation of the gold surface after several runs. The mercury stripping peaks seemed to be more stable but also had a decreasing trend. These studies indicate that a relatively stable signal for arsenic can be obtained using 2 M HCl. The UMEAs showed similar trends and relatively stable signals for arsenic were obtained.
Interference Studies. Interference studies were performed in 2 M HCl to examine the effect other metal cations may have on the arsenic monolayer stripping peak. Lead, copper, and mercury were chosen for several reasons. Lead and mercury compete for sites on the gold surface without forming intermetallic compounds. It has been shown in the literature that copper and arsenic form intermetallic compounds, which severely hinder determinations of arsenic. Lead and copper are also very common contaminants in many water systems. The results for three different Au-UMEAs, normalized for comparison and each value being the average of three runs, are shown in Figure 7. The relative standard deviation for 15 runs with the solution containing only arsenic was 2.5%. The peak stripping current, ∆ip, decreased for all the interfering metals additions, but not to the same degree. For lead, which is typically deposited at a more negative potential, only a fraction of lead was reduced compared to arsenic. This was supported by the data for concentrations of e200 ppb Pb. There was only a slight decrease of ∆ip, demonstrating that the deposition of arsenic was not affected. Only when lead was 5 times the concentration of arsenic was there significant interference with the arsenic deposition process. With respect to mercury, which has a stripping potential more positive than arsenic, at 50 ppb only slight decrease of ∆ip was observed. However, at equal and greater concentrations than arsenic, mercury deposition began to interfere significantly with arsenic deposition. Moreover, the arsenic signal decreased proportionately as the Hg concentration increased, which can be attributed to mercury competing for sites on the gold surface. Hg is known to have a strong interaction with gold, and at the negative potentials there would be a stronger affinity for mercury to deposit in preference to arsenic. The decrease in the arsenic signal can be attributed to the competing deposition of the metals and perhaps the partial dissolution of the gold into mercury, rather than the amalgamation of Hg and As. In this way, an increase of mercury in the solution leads to a greater percentage of it being deposited, thus limiting the available sites for arsenic deposition. For Cu, the interference is complicated because there is an interaction between the copper and arsenic.15,18 The arsenic stripping peak diminished quickly even with just 50 ppb Cu present. The shape of the stripping peak remained similar. Cu, Analytical Chemistry, Vol. 72, No. 10, May 15, 2000
2227
Figure 8. Voltammograms for the standard additions performed onsite with the battery-powered potentiostat and Nafion-coated AuUMEA. Parameters: Esw ) 50 mV, f ) 200 Hz, ∆Es ) 2 mV, and tdep ) 60 s. Sample was acidified to 2 M HCl.
like Hg, can be deposited at more positive potentials than arsenic and it too would compete for surface sites. The decreased value of ∆ip is most likely correlated to the poor deposition efficiency when a small amount of copper is in solution. At 200 and 500 ppb Cu, the arsenic stripping peak was minimal. If copper was predominately deposited on the surface, some arsenic would have to deposit in or on copper. An additional peak was not found in the voltammogram within this potential range. The copper stripping peak, however, was broad, which may be an indication of a complexation. On-Site Analysis. An on-site analysis using this field-portable potentiostat and the Au-UMEAs was performed on a site containing arsenic. The site is a municipal landfill and is located in central Maine. Previously, it was remediated for the removal of DMF. However, arsenic levels were found to exceed the drinking water limit of 50 µg/L and believed to be naturally occurring with the mechanism being dissolution of natural background arsenic from saturated zone materials under the landfill. The water had been found to be anoxic with high levels of mostly iron (60 mg/L). Other heavy metals were nondetectable. Upon air oxidation, the arsenic precipitates out of solution. Over a few months, the arsenic levels have remained fairly constant.39 The sensor described here was used and compared with current EPA techniques (GFAAS). Water samples were taken from an influent pipe at the on-line treatment facility. Groundwater was clear with very little particulate matter; therefore no filtering was necessary. The sample was immediately acidified to 1 M HCl and capped. Water temperature and pH were 12 °C and 6, respectively. A second sample was taken for laboratory-based analysis. This sample was acidified with nitric acid (trace metal grade) to a pH
