ARTICLE pubs.acs.org/ac
Imaging Time-Resolved Electrothermal Atomization Laser-Excited Atomic Fluorescence Spectrometry for Determination of Mercury in Seawater Alain Le Bihan, Jean-Yves Cabon, Laure Deschamps, and Philippe Giamarchi* Universite Europeenne de Bretagne, Brest University, UMR CNRS 6521 CEMCA, 6 Avenue Le Gorgeu, 29285, Brest Cedex 3, France ABSTRACT: In this study, direct determination of mercury at the nanogram per liter level in the complex seawater matrix by imaging time-resolved electrothermal atomization laser-excited atomic fluorescence spectrometry (ITR-ETA-LEAFS) is described. In the case of mercury, the use of a nonresonant line for fluorescence detection with only one laser excitation is not possible. For measurements at the 253.652 nm resonant line, scattering phenomena have been minimized by eliminating the simultaneous vaporization of salts and by using temporal resolution and the imaging mode of the camera. Electrothermal conditions (0.1 M oxalic acid as matrix modifier, low atomization temperature) have been optimized in order to suppress chemical interferences and to obtain a good separation of specific signal and seawater background signal. For ETA-LEAFS, a specific response has been obtained for Hg with the use of time resolution. Moreover, an important improvement of the detection limit has been obtained by selecting, from the furnace image, pixels collecting the lowest number of scattered photons. Using optimal experimental conditions, a detection limit of 10 ng L1 for 10 μL of sample, close to the lowest concentration level of total Hg in the open ocean, has been obtained.
he determination of metals or metalloids by atomic fluorescence spectrometry began during the 1960s;1,2 their determination through their electrothermal atomization and excitation by laser (ETA-LEAFS) appeared during the 1970s.3,4 For the majority of elements, this technique is specific when nonresonant fluorescence can be used, and very low detection limits may be obtained with the use of very low volumes of water sample (i.e., a few attomoles level with the use of 520 μL of injected volume5). This technique is monoelemental, but with the use of modern equipments, it has become more robust, rapid, and can be eventually automated. In the case of seawater, 0.1 and 14 ng L1 detection limits (DLs) have been, respectively, obtained for lead6 and silver.7 In previous papers on aluminum,8,9 lead,9,10 and iron11 within our “Lyopo” program12 (http://www.univ-brest.fr/lyopo/), we had shown that the use of time resolution could also improve the limit of detection by lowering noise level, particularly in the presence of a saline matrix. Good specificity and sensitivity permitted us to carry out analysis of metals without a separation or/and preconcentration stage in a complex matrix. For some metals, such as mercury or cadmium, the only usable fluorescence (with a single laser apparatus) is resonant fluorescence that is much more complicated in the presence of scattering phenomena. In this study, our objective was to show that by combining temporal resolution and imagery development, it was possible to reach very low limits of detection for a direct analysis in a saline matrix, like seawater (salinity ∼35 g L1), by imaging time-resolved electrothermal atomization laser-excited atomic fluorescence spectrometry (ITR-ETA-LEAFS) with the use of only one laser.
T
r 2011 American Chemical Society
Because of its very low level concentration in seawater (a few nanograms per liter), mercury is generally determined by cold vapor atomic absorption13,14 or fluorescence15 spectrometry coupled with an amalgamation preconcentration step on a gold trap. For these techniques, very low detection limits have been obtained; however, they require an important sample volume (∼100 to 900 mL), an oxidative pretreatment step followed by a reductionamalgamation preconcentration step, and consequently, the use of various reactants (KBrO3, KBr, HCl, H2SO4, NH2OHCl, ...) to be purified. This two-stage analysis is generally time-consuming. It is also dependent on the various chemical forms of mercury, the chemical pretreatment having to lead to quantitative formation of Hg atoms; therefore, these analytical protocols are more subject to biased measurements of total mercury concentration. For the determination of mercury by ETA-LEAFS, the only usable excitation wavelength with one laser is 253.652 nm. Because metal fluorescence emission occurs at the same wavelength, the scattering light intensity (diffuse reflection plus back scattering) must be subtracted from the total signal. If scattering is very low, it can be determined from a second measurement near 253.652 nm, but this procedure increases the noise and, consequently, the detection limit; moreover, it is time-consuming. Some two-laser more sophisticated devices have been described1619 Received: March 7, 2011 Accepted: April 28, 2011 Published: April 28, 2011 4881
dx.doi.org/10.1021/ac2005982 | Anal. Chem. 2011, 83, 4881–4886
Analytical Chemistry that permitted researchers to access the energy level of 62 351 cm1 [5d106s(2S)7] and to measure the intensity of the 546.074 nm fluorescence line; the detection limits obtained were in the 1200 ng L1 range in the case of water, soil, or silver after mineralization. Scattering matrix problems are solved, but there is an increase of the noise due to blackbody emission. Because the 253.652 nm mercury fluorescence line has a theoretical radiative lifetime of 125 ns,20 we propose in this work to minimize scattering problems by using the time resolution detection and imaging mode of the camera. With an appropriate time shift between the laser pulse and the fluorescence acquisition, scattering problems can be greatly attenuated. Moreover, the use of the imaging charge-coupled device (ICCD) camera detector in imaging mode permitted us to select image zones containing the lowest proportion of scattered photons. In a first part of this work, we have optimized experimental conditions21,22 with a more convenient classical graphite atomic absorption spectrometer in order to obtain a low atomization temperature of Hg before vaporization of the seawater matrix. In a second part, we have optimized the temporal shift and demonstrated the interest of the imaging mode measurement for ETA-LEAFS.
’ MATERIAL AND METHODS Electrothermal Atomic Absorption Spectrometry. For the experimental electrothermal atomic absorption spectrometry (ETA-AAS) study, a Perkin-Elmer SIMAA 6100 electrothermal atomizer working in the one-element monochromator mode was used for atomic absorption measurements. Pyrolytic graphite coated tubes equipped with a pyrolytic platform were used. Samples were delivered to the furnace using a Perkin-Elmer AS800 autosampler. The light source was a Hg electrodeless discharge lamp operating at 180 mA using the 253.7 resonance line. The inert gas was argon. Laser-Excited Atomic Fluorescence Spectrometry. The excitation source used is a Continuum Nd/YAG laser coupled with an optical parametric oscillator (OPO). The third harmonic of the laser (wavelength = 355 nm, energy = 350 mJ, frequency = 10 Hz) pumps the OPO.11,12 The output signal is doubled with the FX1 module. This leads to a 2251675 nm working range with output energy between 1 and 30 mJ depending on the wavelength. In UV region (doubled signal: 225365 nm), the temporal pulse width is 34 ns and the pulse width at half-height is 3 pm. The laser beam is directed to the atomization furnace through a pierced mirror. To attenuate the spatial heterogeneity of the beam, the thermal lens effect (due to the variation of argon refractive index vs temperature), and scattering from furnace walls, a convergent lens (f = 10 cm) is placed in the laser beam before the mirror. Fluorescence is directed into a spectrometer Acton SP 750 (AN: f/9.7) equipped with a 50 grooves mm1 grating for the spectrum exploration and a 3600 grooves mm1 grating for the analysis. The spectrograph output is connected to an ICCD detector (Princeton Imax-512 T) with a 512 512 pixel array, leading to a maximal resolution of 7 pm pixel1 (3600 g mm1, slit 10 μm). The time gating is controlled by a DG 535 time delay unit (Stanford Instruments, gate width g2 ns, 1 ns time delay). The Winspec software (V 2.5.2) controls the spectrograph, the delay generator, and the detector for spectra acquisition. With this system, to acquire 10 images/s with a good sensitivity and a low noise pixels have to be binned per group of 7 7. Coupled to the laser device, the graphite furnace used for
ARTICLE
the electrothermal atomization is a Varian 800Z equipped with an autosampler; cups are polyethylene made. To avoid any contamination from the atmosphere, all the experimental setup is located in an air-filtered (1 μm) conditioned room. Reagents. The calibrating solutions were prepared by dilution from a 1 g L1 Hg(NO3)2, 0.5 M in HNO3 Merck standard. Ultrapure water from a Millipore Mro-MQ system was used. Seawater samples collected in the Atlantic ocean were acidified to pH 2 with Suprapur HCl (Merck). Pd(NO3)2 was Merck palladium modifier. Oxalic acid was Suprapur grade (Merck). Dilutions were carried out with calibrated Gilson Pipetman pneumatic syringes.
’ RESULTS AND DISCUSSION Electrothermal Atomic Absorption Spectrometry Optimization. On the contrary to other metals, due to its high volatility,
Hg is practically entirely lost at the drying step in water medium, the percentage of losses depending on the nature and on the different treatments of the graphite surface. Therefore, in order to avoid losses of Hg at this stage, various noble metals, like Pd, Au, Ir, Rh, ...,23,24 have been previously used as chemical modifiers to delay its atomization for its determination in water samples. Due to the presence of the important saline seawater matrix, the determination of elements by ETA-AAS is more difficult. Indeed, the vaporization of salts may generate spectral and chemical interference effects. The major part of this matrix is sodium chloride that is vaporized at about 1200 °C and generates a very important background absorption signal (BG) below 300 nm. This problem can be minimized for nonvolatile elements by eliminating the major part of this matrix with the use of a pretreatment step at about 1200 °C before atomization. This is not obviously possible in the case of a very volatile element like Hg. Consequently, experimental conditions have to be chosen both to stabilize Hg in the atomizer at the drying step and to obtain a good separation of its atomization signal and background absorption signal. This may be obtained by using an adequate chemical modifier permitting a relatively low atomization temperature. In this work, we have chosen to examine the behavior of two noble metals, Au and Pd, as chemical modifiers, these metals having been previously used for the determination of Hg in water23,24 or seawater.25 On the other hand, we have examined the behavior of oxalic and hydrofluoric acid that have been previously used as modifiers to promote a low atomization temperature for the determination of midvolatile elements like Cd or Pb in order to obtain a good separation of atomic signal and background absorption signal in seawater.2628 Au and Pd Modifiers. As previously shown, the efficiency of Pd and Au modifiers depends on their concentrations and their different chemical forms, like chloride, nitrate, or metallic form, the metallic form being generally obtained through the use of a reducing agent, like ascorbic acid, or after an in situ electrothermal prereduction of the corresponding salt in the atomizer, at temperatures between 600 and 1900 °C for Pd23 or between 600 and 1500 °C for Au (this study). In seawater a Pd(NO3)2 ascorbic acid mixture has been previously used for the determination of Hg;25 however, the presence of a background signal generated by decomposition products of the modifier at the atomization stage and by the simultaneous vaporization of seawater matrix does not appear adequate for Hg determinations by LEAFS, due to important scattering phenomena. Consequently, 4882
dx.doi.org/10.1021/ac2005982 |Anal. Chem. 2011, 83, 4881–4886
Analytical Chemistry
ARTICLE
Figure 1. Atomization peak of 1 mg L1 Hg in water (0), seawater (1), þ 1 μg Au (2), þ 1 μg Pd (3), þ 0.1 M oxalic acid (4), þ 10 M HF acid (5); BG seawater (6). Tatom = 1350 °C.
we have examined the potentiality of Pd or Au for the determination of Hg in seawater with the use of an in situ prereduction step of Pd(NO3)2 and H(AuCl4). For this purpose, 10 μL of HAuCl4 or Pd(NO3)2 solution was pretreated in the graphite furnace at 1000 °C for 30 s, in order to produce their corresponding metallic species in the atomizer before introducing 10 μL of seawater sample. As observed, pretreated Pd(NO3)2 increases significantly the integrated absorbance of Hg obtained in nonmodified seawater. On the other hand, pretreated HAuCl4 had practically no influence on the integrated absorbance of Hg as compared to the signal obtained in nonmodified seawater. Au and, particularly, Pd induce an important delaying effect of the atomization of Hg, even for a low 1 μg mass of Pd, but the vaporization of the seawater matrix starts before the end of the atomization signal of Hg (Figure 1), these conditions being not optimum for the determination of Hg by LEAFS due to scattering phenomena. Oxalic or Hydrofluoric Acid. Because noble metals do not lead to a sufficient separation of atomization and background signals for the determination of Hg by LEAFS, we have examined the use of oxalic acid and hydrofluoric that modify both the atomization signal of Hg and the background absorption signal of seawater matrix to an oxide or a fluoride matrix. As shown in Figure 2, the atomization signal of Hg is practically recovered for a concentration of oxalic acid about 0.1 M, concentration that is generally required to hydrolyze MgCl2 at the drying step. Indeed, in the absence of any chemical modifier, MgCl2 present in seawater matrix is partially decomposed at the drying step to MgOHCl then to MgO at about 600 °C, by releasing HCl in the atomizer that may induce strong chloride interference effects. However, the case of Hg is different from Cd or Pd27,28 and a too high oxalic acid concentration in seawater leads to an important interference effect, probably by promoting losses of Hg through its reduction on carbon compounds at a lower temperature. Similarly to oxalic acid, a reduction of the interference effect is also observed in the
presence of HF, but for much higher HF concentrations (above 10 M). As previously observed, HF modifies MgCl2, CaCl2, and NaCl salts to their corresponding fluoride salts at the drying step, avoiding also HCl generation at the atomization step of Hg consecutively to the hydrolysis of MgOHCl. A higher HF concentration does not improve Hg recovery but leads only to a decrease of the background absorption signal by removing HCl at the drying step, leaving a less absorbing NaF matrix in the atomizer. From this study, it appears that different types of modifiers may improve Hg recovery in seawater. However, the integrated absorbance, shape, and time appearance of the atomization signal are highly dependent on modifier nature, form, and concentration (Figures 1 and 2). Consequently, the atomic signal is more or less well-separated from the matrix vaporization that may lead to a noncorrectable background absorption signal. From our study, both the use of 0.1 M oxalic acid as chemical modifier and a low atomization temperature that leads to a good separation of background and specific signals (Figure 1) appears optimum for the determination of Hg with the use of our Perkin-Elmer apparatus and, by minimizing salt scattering, for LEAFS measurements. Using these optimized ETA-AAS experimental conditions, a limit of detection of about 5.7 μg L1 (20 measurements) was obtained for Hg by injecting a 10 μL seawater sample. Imaging Time-Resolved Electrothermal Atomization Laser-Excited Atomic Fluorescence Spectrometry Optimization. Access to the lowest DL requires us to maximize the signalto-noise ratio, noise and signal intensity having therefore the same importance. Concerning the signal, the method sensitivity depends (apart from the technical characteristics of the apparatus and of the optical setup) on experimental atomization conditions and on excitation and emission wavelength physical characteristics. For atomization relative similar experimental conditions to that optimized for ETA-AAS were used (except for atomization temperature, set to 800 °C for the Varian atomizer that has different furnace heating characteristics). Indeed, to avoid 4883
dx.doi.org/10.1021/ac2005982 |Anal. Chem. 2011, 83, 4881–4886
Analytical Chemistry
ARTICLE
Figure 2. Variations of the integrated absorbance of Hg with the logarithm of the concentration of oxalic acid or HF (Tatom = 1350 °C).
backscattering phenomena from the vaporized matrix, it was not possible to perform an atomization from the graphite tube above 800 °C, due to seawater matrix vaporization. Moreover, different from ETA-AAS measurements, a charring step at 300 °C was necessary in order to eliminate completely the degradation products of oxalic acid that induced a decrease of the atomization signal of Hg. The optimum signal was obtained with the use of a temperature ramp from 300 to 800 °C during 3 s and without argon flow. Then, specific parameters for LEAFS measurements, i.e., pulse energy, time delay t1 (between pulse intensity maximum t0 and beginning of data acquisition), and acquisition gate port Δt have been optimized. Study of the Noise. For each atomization, the total noise is the convolution of five types of noise: First is the noise of the detection unit in the experimental conditions of gain and acquisition time. Second is the noise generated by furnace thermal emission depending on the emission wavelength, the atomization temperature, and acquisition gate port (Δt). In the case of Hg analysis at 253.652 nm, this noise is negligible below 1200 °C. Third is the noise induced by the scattering of the excitation beam. In the case of Hg analysis, this noise is predominant. This scattering noise is the convolution of the noise induced by the variations of the excitation beam and the noise induced by the measurement chain. For a given scattering level, the scattered intensity ID is proportional to the intensity of the laser radiation; consequently, ID decreases with the time delay t1 according to the following pseudo-Lorentzian equation: ID ðt1 Þ ¼ ID ðt0 Þ
1 1 þ aðt1 t0 Þ2
The pulse-to-pulse variation of the laser beam intensity is about 8%, and the temporal width at half-maximum is only 34 ns, but some photons reach the camera a few microseconds after. Consequently, the scattering noise decreases with t1. Fourth is the noise induced by other possible sources of nonspecific emission at the observation wavelength. In the case of Hg analysis, there is no
other emission to be considered around 253.652 nm. Fifth is the noise induced by nonuniform distribution of the excitation beam due to the spatial and temporal laser beam instability and its convolution with the noise induced by the nonuniform instantaneous distribution of atoms in the furnace. The use of an energy pulse greater than the saturation level can reduce this noise. The ability to detect the presence of mercury with only one measurement depends on the importance of these five types of noise. For a signal intensity 3 times higher than the standard deviation of the resulting noise, there is a probability of presence of Hg higher than 99.7%. It corresponds to the DL measured from one atomization. From only one atomization, to accept a risk β of 5% (risk of error considering that the sample does not contain any Hg), the signal has to be 4 times higher than σ (1.33 DL). Finally, to calculate the limit of detection, as defined by IUPAC, it is necessary to repeat 20 measurements; consequently, the repeatability of injection and atomization can be also considered in this case as another type of noise. Pulse Energy Optimization. To minimize scattering, the pulse energy has to be the lowest possible. However, to obtain a good repeatability of measurements, it is necessary to reach the saturation level to excite all mercury atoms present in the furnace. As shown in Figure 3, a minimum energy level of 30 μJ is needed for 10 μg L1 of mercury, but the best repeatability is obtained for 50 μJ. Temporal Resolution. For a given number of Hg atoms to be measured in the analyzed sample, another given number of excited atoms are obtained at the initial time t0. This number of excited atoms and the number of emitted photons decrease if the measurement is shifted at t1 and if the integration gate Δt decreases. Because the decrease is exponential, the measured intensity varies as follows: I t1 , Δt ¼ ðI 0 et1 =τ Þð1 eΔt=τ Þτ Consequently, the measured signal depends on the time delay t1 between the pulse and the acquisition start, on the integration gate Δt, and on the radiative lifetime τ. 4884
dx.doi.org/10.1021/ac2005982 |Anal. Chem. 2011, 83, 4881–4886
Analytical Chemistry According to NIST tables, the fluorescence lifetime of the 253.652 nm emission line is close to 125 ns. We have estimated this value under argon flow (one drop of mercury introduced in
Figure 3. Fluorescence intensity vs pulse energy for 10 μg L1 Hg in seawater.
Figure 4. Signal, noise, and S/N variations vs time shift (1050 ns) for the same number of excited atoms and for the same integration gate for 15 and 35 ns radiative lifetime values.
ARTICLE
the furnace at 20 °C); however, it appears that in the presence of the seawater matrix, this value depends on the gas flow composition and on its atomization temperature. In fact, the existence of Hg2* excimer is well-known.29,30 The formation of Hg2N* excimer by a laser irradiation at 253.652 nm31 has been highlighted by the observation of fluorescence bands at 335 and 485 nm (with the 50 grooves/nm grating). This excimer formation contributes to decrease the number of excited Hg atoms able to emit a photon and induces also a decrease of the apparent radiative lifetime. During the atomization performed in the gas-stop mode, generally used to increase the residence time of atoms in the furnace and to decrease the detection limit, there is some entrance of nitrogen in the furnace. Consequently, we measured a radiative lifetime of only 15 ns for an atomization step performed at 700 °C in the presence of nitrogen and 35 ns under argon; these values indicate the persistence of excimer species in the furnace, even under argon. This excimer is less stable at higher temperature, and we have measured a radiative lifetime of 100 ns for an atomization at 1300 °C (Varian atomizer). However, these higher atomization temperatures are not usable for Hg determination in seawater by LEAFS due to the scattering phenomena induced by the vaporization of seawater salts (mainly NaCl). In Figure 4, two curves present the signal variations versus the time delay t1 (1050 ns), for the same number of excited atoms and for the same integration gate but for two different values of radiative lifetimes (15 and 35 ns). In the same figure are also presented the variations of the noise (standard deviation of 50 measurements) versus the time shift t1. In this figure is also plotted the variation of the signal-to-noise ratio versus the time shift t1 for the same scattering signal ID at t0 and for the two radiative lifetimes. After 13 ns time shift, and for the realistic selected value of ID, the noise becomes nearly constant and results from the electronic treatment. The figure shows also that a longer radiative lifetime permits us to obtain a better signal-to-noise ratio when the best time shift t1
Figure 5. Furnace image (black, low emission; white, high emission): left panel, beam scattering (t1 = 10 ns); right panel, mercury fluorescence. The optimal area is highlighted by the red dotted line. 4885
dx.doi.org/10.1021/ac2005982 |Anal. Chem. 2011, 83, 4881–4886
Analytical Chemistry of 1314 ns is selected. Moreover, the optimal t1 increases when the initial ID is higher, and consequently, the signal-to-noise ratio (S/N) decreases and the DL increases. The acquisition gate port Δt is as usually fixed at 3τ corresponding to 95% of the photons reemitted after t1, because a longer time increases the thermal or cosmic noise. Optical Setup and Imaging Mode. To reduce the ID intensity of the scattered signal, an optimized optical setup is required, i.e., furnace windows oriented at 45°, a 2 mm diaphragm placed just before the furnace, use of a converging lens (f = 10 cm) to avoid the illumination of the furnace walls (35 mm diameter), and attenuation of reflection by positioning a photon trap in the beam path after the furnace. Using these experimental conditions and P = 50 μJ, the beam scattering due to the optical setup is negligible at t1 = 13 ns. The attenuation of the beam scattering can be also obtained by using the imaging mode of the spectrophotometer. As only one wavelength reaches the ICCD, the resulting image is issued from the convolution of the object lightened by the laser beam scattering and of the emission of mercury atoms by the optical elements. This image can be easily observed by using a very large slit of 3 mm. Figure 5 presents the image due only to beam scattering (left) for a time delay t1 of 10 ns and the image obtained in the presence of 1 μg L1 of mercury (right). The optimal area must contain the smallest beam scattering and the highest Hg emission. The optimal area (red dotted line in Figure 5) comprised between 90 and 120 pixels (X-axis) and between 34 and 45 groups of 7 pixels (Y-axis) and has been used for all the measurements to reduce the part of scattering in the total noise. However, a small proportion of scattering is still remaining in this area that cannot be eliminated. Using these experimental conditions, and for a pulse energy of 50 μJ, the S/N value is increased by about 30%.
’ CONCLUSION With the use of our optical setup, oxalic acid as modifier, an optimized electrothermal program, and the imaging mode mathematical treatment, we have obtained very low detection limits from only one atomization signal, i.e., 30 ng L1 in peak height mode, 10 ng L1 in peak area mode, for 20 impulsions with an injection of a 10 μL seawater sample. From a practical point of view, these detection limits do not permit us to determine mercury in seawater for the lowest concentration level found in the literature for open ocean, but it allows a rapid control of the seawater quality for pollution monitoring purposes. From a general point of view, this study shows that the association of temporal resolution and ETA-LEAFS allow us to solve the major part of problems arising from interferences or beam scattering and to determine elements that cannot be determined with only one laser excitation. Moreover, the use of imaging spectrometry, which was not reported yet in the literature for the ETA-LEAFS technique, allowed us by selecting a working area on the CCD to reduce the noise induced by scattered photons and, consequently, to improve the limit of detection of Hg. In the ITR-ETA-LEAFS technique, the use of temporal resolution permits us to determine rapidly metallic elements in highly salted matrix, with both good sensitivity and specificity. The detection limits reported in this work, the robustness, and
ARTICLE
the ease of analysis could be greatly improved if manufacturers would develop devices specifically optimized for this purpose.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ REFERENCES (1) Winefordner, J. D.; Vicker, T. J. Anal. Chem. 1964, 36, 161–165. (2) Winefordner, J. D.; Staab, R. A. Anal. Chem. 1964, 36, 165–168. (3) Fraser, L. M.; Winefordner, J. D. Anal. Chem. 1971, 43, 1693–1696. (4) Bolshov, M. A.; Zybin, A. V.; Zybina, L. A.; Koloshnikov, V. G.; Mayorov, I. A. Spectrochim. Acta, Part B 1976, 31, 493–500. (5) Stchur, P.; Yang, K. X.; Hou, X.; Sun, T.; Michel, R. G. Spectrochim. Acta, Part B 2001, 56, 1565–1592. (6) Cheam, V.; Lechner, J.; Sekerka, I.; Desrosiers, R.; Nriagu, J.; Lawson, G. Anal. Chim. Acta 1992, 269, 129–136. (7) Gornushkin, I. B.; Smith, B. W.; Winefordner, J. D. Spectrochim. Acta, Part B 1996, 51, 1355–1370. (8) Le Bihan, A.; Lijour, Y.; Giamarchi, P.; Burel-Deschamps, L.; Stephan, L. Spectrochim. Acta, Part B 2003, 58, 15–26. (9) Burel, L.; Giamarchi, P.; Stephan, L.; Lijour, Y.; Le Bihan, A. Talanta 2003, 60, 295–302. (10) Le Bihan, A.; Lijour, Y.; Giamarchi, P.; Burel-Deschamps, L.; Stephan, L. Anal. Bioanal. Chem. 2003, 375, 815–819. (11) Cabon, J. Y.; Giamarchi, P.; Le Bihan, A. Anal. Chim. Acta 2010, 664, 114–120. (12) Giamarchi, P.; Burel, L.; Stephan, L.; Lijour, Y.; Le Bihan, A. Anal. Bioanal. Chem. 2002, 374, 490–497. (13) Bloom, N. S.; Crecelius, E. A. Mar. Chem. 1983, 14, 49–59. (14) Cossa, D.; Fileman, C. Mar. Pollut. Bull. 1991, 22, 197–200. (15) Knox, R.; Kammin, W. R.; Thomson, D. J. Autom. Chem. 1995, 17, 65–71. (16) Resto, W.; Badini, R. G.; Smith, B. W.; Stevenson, C. L.; Winefordner, J. D. Spectrochim. Acta 1993, 5, 627–632. (17) Pagano, S. T.; Smith, B. W.; Winefordner, J. D. Talanta 1994, 41, 2073–2078. (18) Baker, C. L.; Smith, B. W.; Bolshov, M. A.; Winefordner, J. D. Spectrosc. Lett. 1996, 29, 1497–1512. (19) Ezer, M.; Elwood, S. A.; Jones, B. T.; Simeonson, J. B. Anal. Chim. Acta 2006, 571, 136–141. (20) Furh, J. R.; Martin, W. C.; Musgrove, A.; Sugar J. NIST Atomic Spectra Data Base, line data, version 2 [Online]; Wiese, W. L., Ed.; National Institute of Standards and Technology: Gaithersburg, MD, 2002. (21) Le Garrec, H.; Giamarchi, P.; Cabon, J. Y.; Le Bihan, A. Anal. Chim. Acta 1998, 368, 59–78. (22) Le Garrec, H.; Giamarchi, P.; Cabon, J. Y.; Appriou, P.; Le Bihan, A. Anal. Chim. Acta 1997, 350, 171–182. (23) Bulska, E.; Kandler, W.; Hulanicki, A. Spectrochim. Acta, Part B 1996, 51, 1263–1270. (24) Krata, A.; Jedral, W.; Bulska, E. Spectrochim. Acta, Part B 2007, 62, 269–272. (25) Bermejo-Barrera, P.; Moreda-Pineiro, J.; Moreda-Pineiro, A.; Bermejo-Barrera, A. J. Anal. At. Spectrom. 1998, 13, 777–786. (26) Cabon, J. Y. Spectrochim. Acta, Part B 2002, 57, 513–524. (27) Cabon, J. Y.; Le Bihan, A. Spectrochim. Acta, Part B 1996, 51, 1245–1251. (28) Cabon, J. Y.; Le Bihan, A. J. Anal. At. Spectrom. 1992, 7, 383–387. (29) Philipps, F. S. Proc. R. Soc. London 1913, A89, 39. (30) Rayleigh, L. Proc. R. Soc. London 1927, A114, 620. (31) Phaneuf, R. A.; Skonieczny, J.; Krause, L. Phys. Rev. A 1973, 8, 2980–2989.
4886
dx.doi.org/10.1021/ac2005982 |Anal. Chem. 2011, 83, 4881–4886
