Anal. Chem. 2002, 74, 921-925
Technical Notes
A Method for Screening Total Mercury in Water Using a Flow Injection System with Piezoelectric Detection Lisbeth Manganiello,† Angel Rı´os, and Miguel Valca´rcel*
Analytical Chemistry Division, Faculty of Sciences, University of Cordoba, E-14004 Cordoba, Spain
An automatic microgravimetric screening system based on piezoelectric detection and the use of acidic stannous chloride as reductant was developed for the fast detection and determination of total mercury in water. Reduced mercury is detected as an amalgam by using a gold-coated piezoelectric crystal, the sensor subsequently being regenerated by passing it through a peroxydisulfate solution. The gold-coated piezoelectric crystal is a highly efficient retention unit for the main soluble mercury species (inorganic, complexed, and organometallic) previously reduced to elemental mercury and is free of interferences from other metal ions. This detector exhibits good sensitivity: it allows the determination of mercury at sub-partsper-billion concentration levels (0.30-1.00 µg/L). The precision, expressed as relative standard deviation, was ( 2.7% (n ) 11; P ) 0.05) at 0.5 µg/L total mercury. The proposed method was successfully used as a rapid screening method for mercury monitoring in natural waters. Ensuring quality in natural waters entails monitoring characteristic parameters by using rapid analytical methods. Thus, pollution is detected as close as possible to its origin. The development of screening methods for analytical monitoring is one interesting alternative to the simple, rapid control of water quality with reduced costs and minimal delay between sampling and analysis.1 Heavy metals are among the most significant pollutants in natural waters. Within this group of pollutants, mercury causes very negative effects, because of its toxicity and accumulative tendency in living species. This element is involved in many industrial processes, which constitute the main origin of mercury in water. The determination of mercury in natural waters has traditionally been difficult because of the generally low levels of mercury existing in these types of samples. Suitable procedures include atomic absorption spectrometry by the cold vapor technique, the use of gold amalgamation traps with atomic absorption spectroscopy, and the more sensitive atomic emission spectrom* Corresponding author. Fax: 34 957 218606. E-mail:
[email protected]. † Permanent address: Chemical Research Centre, Faculty of Engineering, University of Carabobo, Ba´rbula, Carabobo, Venezuela. (1) Valca´rcel, M.; Ca´rdenas, S.; Gallego, M. Trends Anal. Chem. 1999, 18, 685-694. 10.1021/ac010919g CCC: $22.00 Published on Web 01/04/2002
© 2002 American Chemical Society
etry (AES) technique. The recently applied inductively coupled plasma mass spectroscopy (ICPMS) and, especially, atomic fluorescence spectroscopy (AFS) detection have further increased the ability to determine low mercury levels accurately.2,3 These techniques require expensive, sophisticated instrumentation. Screening systems providing yes/no binary responses for the detection of total mercury in water are very useful, because they can be used as laboratory “filters” to avoid the permanent use of complex instrumentation as well as to perform in situ measurements using portable equipment. There have been several attempts at using piezoelectric quartz crystals involving mercury as both reagent and analyte. Piezoelectric sensors have been frequently applied to the detection of gas-phase mercury and have also been used in combination with the cold vapor atomic absorption spectrometry technique. Suleiman and Guibault4 used them as detectors for sulfur dioxide, utilizing a mercury displacement reaction. The mercury vapor produced by bubbling sulfur dioxide through a mercury nitrate solution was detected as a mercury amalgam on a gold-coated piezoelectric crystal. Nomura and Fujisawa5 examined the electrodeposition of mercury between piezoelectric quartz crystal electrodes upon application of a potential of ca. 0.1 V, in which only silver(I) was found to interfere. Monitoring CO is an important operation in producing Portland cement. McCallum et al.6 used a piezoelectric crystal for this purpose. The carbon monoxide produced was passed through a drying tube to remove moisture and then through a tube packed with mercury oxide, where it reacted to form mercury vapor and carbon dioxide. This vapor caused a change in the crystal frequency, thus allowing control of the gas mixture combustion. Andersen et al.7 studied the behavior of the electrochemical quartz crystal microbalance as a stripping detector for mercury(II), which allows mercury to be quantified at the parts per billion level. Ho et al.8 found mercury (2) Bloom, N. S.; Fitzgerald, W. F. Anal. Chim. Acta 1988, 208, 151. (3) Haraldsson, C.; Westerlung, S.; O ¨ hman, P. Anal. Chim. Acta 1989, 221, 77. (4) Suleiman, A. A.; Guibault, G. G. Anal. Chem. 1984, 56, 2964-2966. (5) Nomura, T.; Fujisawa, M. Anal. Chim. Acta 1986, 182, 267-270. (6) Benmakroha, F.; Belakroum, M.; McCallum, J. Anal. Proceed. 1993, 30, 133-135. (7) Andersen, N. P. R.; Hansen, P. H.; Britz, D. Anal. Chim. Acta 1996, 329, 253-256. (8) Ho, M.; Guilbault, G.; Scheide, E. Anal. Chim. Acta 1981, 130, 141-147.
Analytical Chemistry, Vol. 74, No. 4, February 15, 2002 921
Figure 1. Manifold used for the determination of mercury in water: PP, peristaltic pump; IV, injection valve; RC, extraction coil (25 cm); W, waste; FC-PZ, flow cell and piezoelectric crystal; OC, oscillator circuitry; F, frequency counter; PC, interfaced personal computer.
vapor produced by a solution previously treated with a reducing agent caused an increase in the crystal mass. In this work, a microgravimetric screening system was developed for the determination of total mercury in water using piezoelectric detection. The method is based on the reduction of mercury species to elemental mercury using stannous chloride. Elemental mercury is detected as an amalgam on the gold-coated piezoelectric quartz crystal that is used, which provides good sensitivity and a high selectivity. Thanks to its simplicity, the proposed method can be used as a rapid alternative to the monitoring of mercury in natural waters for drinking EXPERIMENTAL SECTION Apparatus. Piezoelectric Detector. An AT-cut 10 MHz piezoelectric quartz crystal (14-mm diameter, 0.17-mm thick) coated with gold-plated electrodes on both sides was used. The quartz crystal was housed in a flow-through PEEK cell and clamped between two O-rings recessed into the housing; one crystal face was exposed to the sample in a 70-µL cell. The piezoelectric quartz crystal and PEEK flow cell were supplied by Universal Sensors, Inc. (Metairie, LA). A laboratory-made oscillator circuit was connected to the electrode via platinum foil. A galvanic insulation filter was incorporated into the oscillator circuit to attenuate electronic noise and improve the system baseline. The resonant frequency was monitored using a Hewlett-Packard HP-53181A/ 225 MHz frequency counter that was connected to a PC Pentium microcomputer via an HP-IB interface (Hewlett-Packard. Madrid, Spain). HP-34812A BenchLink software (HP BenchLink/Meter, Madrid, Spain) was used to acquire and store data. A Gilson Minipuls-3 peristaltic pump and a Rheodyne 5020 injection valve were also used (both provided by Scharlau, Barcelona, Spain). Reagents and Standards. All reagents that were used were analytical high purity grade. Ultrapure water was obtained from a Milli-Q System (Millipore Co., Bedford, MA). Mercury(II) chloride and methylmercury chloride (Riedel-de Hae¨n, Buchs, Switzerland) were used as standards. A stock solution (1.00 g/L) in 1% v/v HCl (Merck Suprapur grade, Darmstadt, Germany) was used daily to prepare standards by appropriate dilution in 1% v/v HCl. Organomercury solutions were protected against light and 922
Analytical Chemistry, Vol. 74, No. 4, February 15, 2002
stored at 4 °C in the dark. A 1.1% w/v solution of anhydrous stannous chloride (Sigma, Steinheim, Germany) in 3% v/v HCl was used as the reductant. Sample Preparation. The water samples that were used were collected from the river Guadalquivir in appropriate polyethylene containers and preserved with HNO3 at pH ca. 2. Samples were protected against light and stored at 4 °C in the dark.9 Samples were injected into the system following dilution with HCl in the required proportion to obtain the same acid concentration as the carrier solution (1% v/v HCl). Manifold Design and Procedure. The manifold that was used is depicted in Figure 1. The carrier solution (1% v/v HCl) was pumped through Tygon tubes. Samples and standards were injected via an injection valve (IV). The solution held in the sample loop (41.3 µL) was transferred by the acid carrier to a T-piece for mixing with the reductant, and then it was transferred to the reaction coil (RC). Mercury was reduced in a coil 25-cm long × 0.8-mm i.d. Finally, it was passed through the piezoelectric flow cell (FC-PZ), where the signal was detected, transmitted to the frequency counter (F) and then recorded (PC). Samples and calibration standards were injected into the flow once the base resonance frequency (Fb) leveled off and measurements were found to be ∆F ) Fp - Fb, where Fp is the maximum frequency during each run. Mercury was removed from the electrode by passing a 0.01 M ammonium peroxydisulfate/0.01M nitric acid solution over it for 5 min. The incorporation of the sampler into the manifold (Figure 1) allowed the sequential introduction of samples and the regenerated solution. RESULTS AND DISCUSSION The aim of this work was the development of a rapid method for the determination of mercury in natural waters using a piezoelectric detector in a flow system. The assembly used to this end can be considered an analytical screening system that allows one to detect the divalent state of mercury, the dominant form in most natural waters and organomercury compounds, the concentration levels of which are generally low relative to total mercury. (9) Pe´rez-Bendito, D.; Rubio, S. Environmental Analytical Chemistry Comprehensive Analytical Chemistry; Elsevier: Amsterdam, 1999; pp 97-105.
Table 1. Steps and Conditions of the Determination of Mercury by Using a Flow Injection System with Piezoelectric Detection step first second third
reaction with 1.1% w/v SnCl2 in 3% v/v HCl; mercury species are reduced to elemental mercury; microprecipitate is formed elemental mercury is adsorbed onto the surface of the gold electrode, causing the crystal mass to increase and the frequency to decrease oxidant solution (0.01 M ammonium peroxydisulfate/0.01 M nitric acid) is injected into the system to remove the mercury from crystal surface.
optimal values for the experimental variables flow rate temp injected volume reaction coil carrier solution
2 mL/min 22 ( 1 °C 41.3 µL 25 cm (0.8-mm i.d.) 1% v/v HCl
The sensitive detection of a 10 MHz AT-cut quartz crystal provides performance similar to that of the gravimetric principle.10 The crystal acts as a microbalance in which viscosity and density remain unchanged as the temperature is kept constant (22 °C ( 1). This suggests that the frequency change is caused by the small mass deposited on the quartz electrode surface. The signal used to detect mercury is provided by mercury amalgamated directly onto the gold electrode of the crystal. SnCl2 is a specific reductant for mercury in gravimetric methods.10 The frequency change results from the increase in weight of the crystal surface. Table 1 summarizes the process. The presence of various major compounds in natural waters was examined in order to check for potential interferences with the determination of the mercury. Water samples containing calcium and magnesium ions at high concentrations caused no interference; neither did other common ions, including Cu, Fe, Mn, and Pb. The gold electrode of the crystal is a highly selective adsorbing unit for mercury, and its detection is free of interferences. Optimization. Preliminary experiments showed that the chemical variables (viz., the concentrations of the reductant and the oxidant used to regenerate the sensor) and hydrodynamic variables (viz., the flow rate and coil lengths) could influence the detector frequency. The mercury concentration used in these experiments was 0.50 µg/L. Chemical and Hydrodynamic Variables. Reductant and Coil Length. These variables are closely related, because the length of the reaction coil should be appropriate for the reaction rate of the reductant. Various reductants, including ascorbic acid, sodium hypophosphite, and stannous chloride were studied. These three agents are recommended in the literature for reducing mercury species to metal mercury.10-12 Ascorbic acid required a rather long coil (2 m) to ensure completion of the reduction reaction, which increased the analysis time. Sodium hypophosphite proved to be (10) Skoog, D.; West, D.; Holler, J. Fundamentos de Quı´mica Analı´tica, 4th ed; Reverte´: Barcelona, 1997; pp 85-88, 852. (11) La´slo´, E. Gravimetric Analysis, Part II; Pergamon Press: Oxford, 1965; pp 64-65. (12) El-Ahraf, A.; Van Willis, W.; Vinjamoori, D. J. Assoc. Off. Anal. Chem. 1981, 64, 4 (1), 9-13.
an effective reductant and reacted rapidly (the reaction coil was only 25 cm long); however, it tailed to reduce organomercury compounds.12 Stannous chloride was found to be the most appropriate reductant for the determination of total mercury in water samples. Its high reducing power in an acidic medium and the mere dilution of the sample with HCl9 allowed the reduction of all mercury compounds (inorganic and organic) present in natural water to metal mercury. The fast response of the detector allowed the analyte concentration to be determined within a few minutes. We checked the efficiency of stannous chloride in acidic medium to reduce free inorganic species, mercury complexes with EDTA, and thiourea (used as model complexes in waters), as well as organomercury species (monomethylmercury, ethylmercury and phenylmercury). All of these groups of mercury species can be considered to be the most toxic mercury species present in waters. Other mercury-organic matter complexes can be expected to have a behavior similar to those observed with EDTA and thiourea. By using the proposed methodology, the contribution of mercury associated with a particular matter can be lost, because previous filtration of the sample is needed if a particular matter is observed. Otherwise, a previous oxidative digestion of samples will be carried out, such as what Wurl et al.13,14 and Elsholz et al.15 recently reported. For this microgravimetric application, a quantitative relationship between the relative shift in resonant frequency and the added mass was found that conformed to the Sauerbrey equation.16,17 Factors such as mechanical clamping and damping in the electrical circuit were avoided by using calibration curves.18 Flow Rate and Oxidant. The flow rate was found to affect amalgamation of mercury on the gold electrode of the crystal. This process step allowed the integration of detection, amalgam formation, and baseline restoration. Figure 2 shows the influence of the flow rate on the detector response (frequency change). The sensor response increased as the flow rate decreased; however, baseline restoration was slower at decreased flow rates. This was a result of stronger amalgamation of mercury on the gold surface of the electrode. A 0.01 M ammonium peroxydisulfate/0.01 M nitric acid solution was used to remove mercury from the electrode. The optimal flow rate was 2 mL/min. This flow rate allowed rapid restoration of the baseline. The selection of the oxidant was based on the studies of Nomura and Fujisawa.5 The optimal concentrations of oxidant and acid allowed efficient removal without dissolving the gold electrodes. Figures of Merit of the Proposed Method. Standard solutions containing various mercury species at concentrations between 0.30 and 1.00 µg/L were used to construct the calibration graphs for the determination of total mercury in water. The figures of merit of the proposed method for the different mercury species are listed in Table 2. The calibration graph (run from triplicate measurements at each point) was linear over the range 0.301.00 µg/L. The mercury complexes with oxygen and sulfur (13) Wurl, O.; Elsholz, O.; Ebinghaus, R. Talanta 2000, 52, 51-57. (14) Wurl, O.; Elsholz, O.; Ebinghaus, R. Anal. Chim. Acta 2001, 438, 245249. (15) Elsholz, O.; Frank, C.; Stachel, B.; Reincke H.; Ebinghaus, R. Anal. Chim. Acta 2001, 438, 251-258. (16) Sauerbrey, G. Z. Phys. 1959, 155, 206-222. (17) Sauerbrey, G. Z. Phys. 1964, 178, 457. (18) Janata, J. Principles of Chemical Sensors; Plenum Press: New York, 1989; pp 59-61.
Analytical Chemistry, Vol. 74, No. 4, February 15, 2002
923
Table 2. Figures of Merit of the Proposed Method for the Determination of Mercury in Water Using Standard Solutions Containing Different Mercury Species mercury species
HgCl2
methylmercury
Hg-EDTA
Hg-thiolsa
equationb (n ) 27)
S ) 23.222C + 0.737
S ) 21.604C + 0.895
S ) 23.447C + 1.028
S ) 21.347C + 1.837
parameters characterizing the overall response for the four species used as a model 0.997 ( 0.02 99.3 ( 0.3% 0.5 ( 0.1 0.30-1.00 12 0.25
regression coefficient R2 (from ANOVA test) std dev of residuals, Sy/x determination range (µg/L) throughput (samples h-1) detection limit (µg/L)c
a By using thiourea as representative ligand. b Dependent variable: S, measured signal [frequency difference ∆F (Hz)]. Independent variable: C, µg /L mercury. c Defined as the blank signal plus three times its standard deviation.
Table 3. Screening of River Water Samples for Total Mercury Content Using the Proposed Flow System with Piezoelectric Detection mercury content, µg/L added (µg/L)
found (µg/L)
error (%)
river 1, (spiked with Hg2+)
0.40 0.60 1.00
0.39 0.61 0.98
-2.5 -1.6 -2.0
river 2, (spiked with Hg2+)
0.50 0.70 1.00
0.51 0.70 1.04
0.7 -0.5 4.0
river 3, (spiked with Hg2+)
0.30 0.50 0.70
0.31 0.49 0.72
3.3 -2.0 2.8
river 4, (spiked with Hg2+)
0.30 0.50 0.70
0.48 0.69 0.92
-4.0 -1.4 2.2
river 1, (spiked with HgCH3+)
0.20 0.50 1.00
0.19 0.49 1.01
-1.0 -0.4 1.0
river 1, (spiked with Hg-EDTA)
0.40 0.60 1.00
0.38 0.58 0.97
-5.0 -3.2 -3.0
sample
Figure 2. Frequency (Hz) vs flow rate (mL/min) plot for a 0.50 µg/L standard solution of mercury (injected volume, 41.3 µL; coil length, 25 cm; coil i.d., 0.8 mm).
ligands, as well as with chloride, were studied in order to check whether the proposed method provided an overall response for mercury irrespective of the particular mercury species present in the sample. The sensitivity of method for each mercury species was 23.22 ∆F (Hz) µg-1 L-1 for inorganic Hg(II) species, 21.60 ∆F (Hz) µg-1 L-1 for methylmercury, 23.45 ∆F (Hz) µg-1 L-1 for Hg-EDTA complexes, and 21.35 ∆F (Hz) µg-1 L-1 for Hg-thiols (thiourea complexes mainly). Calculations showed the sensitivities (slopes) to be not significantly different. The null hypothesis H(slope1 ) slope2 ) slope3 ) slope4) was proved, and the statistic was found to be 0.19. Because this value is less than the tabulated critical value, for the χ2 distribution with k - 1 ) 3 degrees of freedom,19 7.81, the null hypothesis could not be rejected, so the sensitivities were not significantly different. The values showed the detector to be acceptably sensitive for the potential forms of mercury in natural water. Organomercury compounds, particularly monomethyl mercury, HgCH3+, can be formed through microbial processes, but organomercury concentration levels in natural
waters are generally rather low (1-10% of total mercury).20 The precision of the method, expressed as relative standard deviation, for 0.50 µg/L mercury (HgCl2) was ( 2.7% (n ) 11; P ) 0.05). The proposed method is, thus, a rapid, inexpensive alternative to the determination of total mercury. Analysis of River Water. The proposed method was applied to the determination of total mercury in natural waters. Four river samples were analyzed, none of which was found to contain mercury. To identify potential matrix effects of the water samples, they were spiked with Hg2+, methylmercury, and Hg-EDTA at variable concentrations. Table 3 gives the results obtained in the determination of mercury in river water. As can be seen, there were small differences between the concentrations found and those added (less than 4%). This was confirmed by the paired comparison test, which revealed no significant differences between the concentrations added and found. The computed t statistic was
(19) Rahman, N. A. Theoretical Statistics; Charles Griffin & Company, Ltd.: Bristol, 1978; pp 431-435.
(20) Salbu, B.; Steinnes, E. Trace Elements in Natural Waters; CRC Press: Boca Raton, FL, 1995; pp 162-163.
924 Analytical Chemistry, Vol. 74, No. 4, February 15, 2002
Figure 3. Reliability of the proposed screening system applied to water samples spiked with mercury (Hg2+) at concentrations near the detection limit (LOD).
0.63. This value is smaller than the tabulated tcrit value (2.20 for 11 degrees of freedom at the 95% confidence limit). The proposed method can, thus, be applied to real samples. Its low detection limit (see Table 1) allows one to check drinking water against the standards of the European Community and Spanish legislation (the highest value is 1 µg/L Hg).21 Reliability of the Proposed Screening System for Mercury Monitoring. Figure 3 testifies to the reliability of the proposed screening system for determining mercury at concentrations near the detection limit in water samples. A simple chemometric study involving 100 spiked test sample standards at various concentration levels was performed in order to identify right answers, false positives, and false negatives. The cutoff concentration was defined as the concentration corresponding to the blank signal plus two standard deviations.22 A false negative arose when a water sample was spiked with an analyte concentration lower than the cutoff value. The proportion of correct results was increased when the samples were spiked with analyte concentrations near or higher than the cutoff value. Mercury(II) chloride was the species studied, because it is the one most commonly found in natural waters.20 A systematic study at four concentration limit detection levels, 0.5, 1, 2, and 3 LOD (LOD, detection limit), was carried out by using 20 samples at each concentration level and calculating the reliability at a 95% confidence level. As can be seen, the proportion of false negatives decreased as the concentration mercury increased: it was 30% near the detection limit at 2 LOD. At 3 LOD, the proportion of correct results was 100%. No false positives were encountered. (21) Bolletı´n Oficial Espan ˜ol, BOE, 1990, no. 226, pp 27 495. (22) Gambart, D.; Ca´rdenas, S.; Gallego, M.; Valca´rcel, M. Anal. Chim. Acta 1998, 366, 93-102.
CONCLUSIONS This work demonstrates a simple, rapid screening method for the determination of total amount of the main mercury species present in natural waters (free Hg2+ and labile inorganic complexes, organomercury species and mercury associated with organic matter) using a gold-coated piezoelectric detector that provides selective, sensitive results. The method is free from interferences and provides a good detection limit (0.2 µg/L), comparable to those of more complex techniques such as cold vapor AAS (0.1 µg/L), which uses expensive instrumentation.18 The proposed screening system is very useful for obtaining yes/ no binary responses for monitoring mercury in water. Direct information is provided about the most toxic species of mercury present in natural waters. A prior oxidative digestion of the water sample is needed if the fraction of mercury associated with particular matter is to be also included in the final results. This prior treatment could also be of interest if the screening methodology were expanded for processing wastewater samples.
ACKNOWLEDGMENT Financial support from Spain’s DGI of MCyT (BQU2001-1815) is gratefully acknowledged. L. Manganiello also thanks Venezuela’s Consejo Nacional de Investigaciones Cientı´ficas y Tecnolo´gica (CONICIT) for the award of a doctoral fellowship.
Received for review August 14, 2001. Accepted October 22, 2001. AC010919G Analytical Chemistry, Vol. 74, No. 4, February 15, 2002
925
