Dynamics and Performance of Fast Linear Scan Anodic Stripping

The dynamics of fast linear scan (LS) ASV for the simul- taneous detection of Cd, Pb, and Cu was investigated at various scan rates (0.5-10 V/s) and a...
0 downloads 0 Views 196KB Size
Anal. Chem. 1996, 68, 1639-1645

Dynamics and Performance of Fast Linear Scan Anodic Stripping Voltammetry of Cd, Pb, and Cu Using In Situ-Generated Ultrathin Mercury Films H. Ping Wu

Research Center, YSI, Inc., P.O. Box 279, Yellow Springs, Ohio 45387

The dynamics of fast linear scan (LS) ASV for the simultaneous detection of Cd, Pb, and Cu was investigated at various scan rates (0.5-10 V/s) and at different metal ion concentrations (50-800 nM) utilizing ultrathin mercury films (9 nm) at a conventional size (d0 ) 1 mm) electrode. Results of the investigation show that when the thin films were utilized, diffusion of metals through the mercury film was not the rate-limiting step of the stripping process at moderate to fast scan rates (0.5-10 V/s). A fairly linear relationship between the peak height and scan rate was observed at scan rates (0.5-10 V/s) beyond the upper limit of the theoretical model for the behavior of LS-ASV. In addition, peak width at half-height (b1/2) as low as 33 mV was achieved at 0.5 V/s. The behavior of LS-ASV in terms of peak width at these scan rates is thus different from what the theoretical model of LS-ASV would have predicted. For the ultrathin mercury films, at least two additional factors, kinetics and concentration, have significant effects on practical LS-ASV. Experimental results show that the stripping process of Cu was primarily kinetic-controlled for fast scans, while those for Cd and Pb were dependent on both scan rates and concentrations. The ultrathin mercury film resulted in a significant enhancement of the ratio of signal-to-baseline slope (ip/ ∆ib, a ratio used to measure the effectiveness of discrimination of the peak signal against the steep sloping baseline in LS-ASV) for Cd and Pb stripping peaks, but only a slight enhancement for Cu stripping peaks. The optimal performance of LS-ASV in terms of sensitivity, peak width, and enhancement of the ip/∆ib ratio for the three metals was achieved at 2 V/s. Because of the high reproducibility of the background currents of the stable in situ MTFs, background subtraction was carried out at 2 V/s with little hysteresis. This feature, combined with the enhancement of the ip/∆ib ratio at the fast scan rate of 2 V/s, allowed for the detection of sub-ppb levels of Cd, Pb, and Cu at a deposition time of 2 min. Over the years, linear scan voltammetry has been one of the major methodologies for anodic stripping voltammetry.1-3 Theoretical and experimental studies include the modeling of linear (1) Wang, J. Stripping AnalysissPrinciples, Instrumentation and Applications; VCH: Deerfield Beach, FL, 1985. (2) Vydra, F.; Stulik, K.; Julakova, E. Electrochemical Stripping Analysis; Wiley: Chichester, Sussex, U.K., 1976. (3) Barendrecht, E. Stripping Voltammetry. In Electroanalytical Chemistry; Bard, A. J., Ed.; Marcel Dekker: New York, 1972; Vol. 2, p 53. 0003-2700/96/0368-1639$12.00/0

© 1996 American Chemical Society

scan anodic stripping voltammetry (LS-ASV),4 the in situ formation of ultrathin films for ASV,5-7 ASV by microelectrodes,8-11 and ASV with very fast scan rates10,11 and with background subtraction.10 Most of the early applications of LS-ASV1-4,8,9 have been limited to slow scan rates in order to avoid the broadening of stripping peaks and poor discrimination against the steep sloping baseline due to the charging background currents. Deposition times have usually been in the range of 5-10 min to achieve low detection limits. For those studies dealing with fast scan rates, most of the experiments were carried out in relatively thick mercury films with ex situ-generated mercury thin films (MTFs).10-11 Strategies of background subtraction, and combining fast scan with background subtraction, have been reported for reducing background currents and enhancing sensitivity. However, because of the relatively thick films and the potential restructuring of the mercury surface conditions with ex situ-generated MTFs, significant hystereses after background subtraction have been observed.10 Thus, understanding the dynamics of fast LS-ASV with thin mercury films to achieve the optimal performance for the analysis of heavy metals remains a challenge. In our initial investigation in this area, our laboratory conducted a series of studies of ASV with a goal of measuring trace amounts of heavy metals of Cd, Pb, and Cu simultaneously at low detection limit. Using optical microscopy and simple linear scan voltammetry with in situ-generated MTFs, it was shown in our previous paper7 that, immediately after deposition, MTFs closely resemble a thin-film state with fine microdroplets evenly distributed on a glassy carbon electrode (GCE) surface. It was also shown that the mercury surface can undergo a rapid structural change under potential change perturbation, depending on the total loading of mercury on the GCE substrate. In addition, with the use of these in situ-generated, very thin mercury films on GCE, excellent sensitivity and peak resolutions of Cd, Pb, and Cu were demonstrated at a fast scan rate of 2 V/s with very reproducible background currents from run to run. The key advantage of utilizing linear scan voltammetry for ASV in combination with ultrathin mercury films is that the metals are stripped as a surface effect, and thus the peak heights are directly proportional to the scan rate. In addition, the capability of determining peak charges from integration of the current under stripping peaks can be used (4) De Vries, W. T J. Electroanal. Chem. 1965, 9, 448. (5) Florence, T. M. J Electroanal. Chem. 1970, 27, 273. (6) Frenzel, W. Anal. Chim. Acta 1993, 273, 123. (7) Wu, H. P. Anal. Chem. 1994, 66, 3151. (8) Wong, D. K. Y.; Ewing, A. G. Anal. Chem. 1990, 62, 2697. (9) Wehmeyer, K. R.; Wightman, R. M. Anal. Chem. 1985, 57, 1989. (10) Baranski, A. S. Anal. Chem. 1987, 59, 662. (11) Nomura, S.; Nozaki, K.; Okazaki, S. Electroanalysis 1991, 3, 617.

Analytical Chemistry, Vol. 68, No. 9, May 1, 1996 1639

to overcome the potential kinetics complication during the stripping process.12 However, measurement of the peak heights in analytical methodology would be preferred because of its simplicity. The purpose of the present investigation was to optimize the performance of fast LS-ASV with regard to the sensitivity, detection limit, peak resolution, and discrimination against the sloping baseline. During our initial investigation,7 it was found that the peak widths were narrower than ASV theory would have predicted for linear scan voltammetry at the scan rate of 2 V/s. Furthermore, it was shown that very sensitive ASV signals were obtained at low metal ion concentrations and that the ratio of signal-tobaseline slope increased significantly with scan rates. The background currents were so reproducible from run to run that background subtraction could be carried out with almost no hysteresis. These findings prompted further investigations on the dynamics of fast LS-ASV, as described herein. With the high reproducibility and the very thin film nature, we have further studied ASV with in situ-generated MTFs, investigating the scan rate dependence of ASV peak currents, peak charges, and peak widths from moderate to fast scan rates (0.5-10 V/s). We have also explored the effects of reducing the scanning background currents by background subtraction. Using a combination of highly reproducible in situ-generated MTFs and background subtraction, a very low detection limit of sub-ppb level for Cd, Pb, and Cu was demonstrated at a short deposition time of 2 min. EXPERIMENTAL SECTION Reagents and Solution Preparations. HNO3 (doubly distilled) and KNO3 (99.999% pure) were obtained from GFS Chemicals (Columbus, OH). Certified AA standard solutions of Pb, Cd, Cu, and Hg ions of 1000 ppm concentration were purchased from Fisher Scientific (Pittsburgh, PA). Standard solutions of appropriate concentrations were prepared by dilution of the concentrated AA standard solutions. An Eppendorf pipet (0.1-1 mL) and an Oxford pipet (1-5 mL) equipped with disposable pipet tips were used for appropriate dilutions. All solutions were prepared with water from a NANOPURE water system from the Barnstead/Thermolyne Corp. (available from Fisher Scientific). All ASV experiments were carried out in a medium that was 0.1 M in KNO3 and 0.03 M in HNO3 (designated “nitrate medium”). Electrochemical Cell and Experiments. Experimental procedures and the electrochemical system were the same as previously reported.7 The Hg2+ concentration was 0.1 mM in the nitrate medium. The acidity and the controlled Hg2+ concentration assured that the mercury loading was constant and reproducible at about 115 µC/mm2, as reported before.7 For the study of scan rate dependence in linear scan (staircase) voltammetry, the scan rates were 0.5, 1, 2, 5, and 10 V/s, all with a filter constant of 50 µs. The voltage resolution in all five scan rates was 1 mV/ step. The deposition and rest times were 2 min and 30 s, respectively, for all experiments. Except for two test runs, all ASV experiments were conducted without deliberately deoxygenating the solution. In some cases, analyses of ASV signals were carried out with HYPERPLOT software (JHM International, Columbus, OH), which provided a way for determining peak heights, peak charges, and peak widths at half-height. (12) Florence, T. M. Analyst 1992, 117, 551.

1640

Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

Figure 1. LS-ASV voltammograms of Cd, Pb, and Cu at 50 nM each with in situ MTF in 0.1 M KNO3 + 30 mM HNO3 + 0.1 mM Hg2+ solution at the following scan rates: (a) 0.5 V/s (20×), (b) 1 V/s (10×), (c) 2 V/s (5×), (d) 5 V/s (2×), and (e) 10 V/s. Table 1. Ratio of Signal-to-Baseline Slope at 50 nM Each of Cd, Pb, and Cu Cd

Pb

scan rate (V/s)

ip (µA)

∆ib (µA)

ip/ ∆ib

0.5 1 2 5 10

0.400 0.852 1.88 5.51 10.8

0.21 0.262 0.364 0.762 1.15

1.91 0.509 3.25 1.08 5.16 2.47 7.24 7.44 9.37 14.7

ip (µA)

∆ib (µA)

Cu ip/ ∆ib

0.172 2.96 0.233 4.84 0.306 8.07 0.436 17.1 0.539 27.3

ip (µA)

∆ib (µA)

i p/ ∆ib

0.699 1.36 2.39 4.55 6.165

0.505 0.729 1.01 3.11 6.47

1.38 1.87 2.37 1.46 0.953

RESULTS AND DISCUSSION Ratio of Signal-to-Baseline Slope with Scan Rates. Shown in Figure 1 are typical LS-ASV voltammograms for a solution containing 50 nM each of Cd, Pb, and Cu at scan rates of 0.5, 1, 2, 5, and 10 V/s. Currents associated with the voltammograms at 0.5, 1, 2, and 5 V/s have been multiplied by 20, 10, 5, and 2, respectively, so that the apparent currents of all voltammograms are normalized to the same perspective as that of the scan at 10 V/s. It can be seen that the stripping peaks of all three metals shifted to more positive potentials as the scan rate increased, with Cu having the largest shifts over the range of scan rates. For example, the potential shifts for Pb and Cd at 10 V/s with respect to Ep at 0.5 V/s are 22 and 24 mV, respectively, but that for Cu is 54 mV for the same scan rate difference. One of the problems for ASV with simple linear scan voltammetry has traditionally been the high background due to charging current, exhibiting as a steep sloping baseline which made it difficult to quantify peak signals. Figure 1 shows that the problem of steep sloping baseline is significantly reduced when employing fast scan rates at ultrathin mercury films. The steep slope of the baseline at the peak was quantified by the current increment (∆ib) in the potential range (∆Eb), across the baseline of which the peak height was measured, and is listed in Table 1, along with the peak heights ip. Thus, the ratio of ip to ∆ib is a measure of how effective the peak signal ip is discriminated against the sloping baseline when sampling the signals. It can be seen from Table 1 that both the peak height ip and the current increment ∆ib increased with increasing scan rates for all three metals. However, the ratio of these two currents (ip/∆ib) improved greatly with increasing scan rates for Pb and Cd up to 10 V/s, but to a much lesser extent for Cu. While the Cd and Pb peaks are narrow and very well resolved at scan rates up to 10 V/s, the Cu peak is broadened dramatically upon increasing the scan rate.

Figure 2. Peak heights and peak charges versus scan rates at five different concentrations of Cd, Pb, and Cu in 0.1 M KNO3 + 30 mM HNO3 + 0.1 mM Hg2+ solution: 9, Cu peak height; b, Pb peak height; 2, Cd peak height (solid lines); 0, Cu charge; O, Pb charge; 4, Cd charge (dashed lines) for all plots.

The data in Figure 1 and Table 1 clearly demonstrate that fast scan voltammetry enhances significantly the ip/∆ib ratio for LSASV with conventional electrodes, an effect which is similar to the enhancement of signal-to-background ratio reported previously by others for microelectrodes.10,11 For analyses of Cd and Pb, the scan rate can be extended to 10 V/s with the electrode conditions described herein to achieve this enhancement of the ip/∆ib ratio. However, because of the difference in the behavior of Cu, if the analysis includes this metal, the optimal scan rate for simultaneous determination appears to be at 2 V/s. This scan rate provides the best performance in terms of sensitivity, peak resolution, and ratio of signal-to-baseline slope for the analysis of all three metal ions when utilizing ultrathin mercury films. Dependence of Peak Heights on Scan Rates and Concentrations. Experiments were also carried out at various concentrations of all three metals (200, 400, 600, and 800 nM) and at various scan rates (0.5, 1, 2, 5, and 10 V/s) in order to examine the effects of variation of these parameters on the peak height, peak charge, and peak width at half-height (b1/2) of the stripping peaks. The peak heights and charges of the three metal ions are plotted as a function of scan rates at five different concentrations in Figure 2.

Note that the peak current is on the left axis, while the peak charge is on the right axis of each plot. Several trends of the three metal ions can be seen from this figure. First, peak heights of Cd, Pb, and Cu are linearly proportional to the scan rate up to 2 V/s. However, while the peak heights of Cd and Pb are only slightly underlinear at scan rates greater than 2 V/s (and only at concentrations of 400 nM and higher), the Cu peak height deviates severely from linearity for scan rates higher than 2 V/s for all concentrations. Second, the peak charges (determined by integration of the stripping peaks) of each of the individual metal ions are not the same at the five different scan rates. For Cd and Pb at concentrations of 200 nM and below, the peak charges increase with scan rates and level off at 5 V/s. At concentrations of 400 nM and higher, however, the peak charges increase with scan rate and begin to level off at a lower scan rate (2 V/s). For Cd and Pb, the difference between the low and the high charges at different scan rates is as much as 35% for concentrations of 200 nM and below, becoming smaller at concentrations higher than 200 nM. On the other hand, the Cu peak charges increase with scan rate up to 2 V/s and then decrease at higher scan rates. Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

1641

Table 2. Concentrationsa of Three Metals inside the Mercury Film amalgam concentrations added aqueous concn (nM)

peak charge (µC)

50 200 400 600 800

0.057 0.218 0.401 0.552 0.783

Cd (MHg)

peak charge (µC)

0.042 0.16 0.23 0.41 0.58

0.070 0.268 0.507 0.723 1.062

Pb (MHg)

peak charge (µC)

Cu (MHg)

0.051 0.20 0.37 0.53 0.78

0.068 0.245 0.467 0.676 0.966

0.050 0.18 0.34 0.50 0.71

a Calculated from the stripping charges, which are the maximum values from Figure 2.

The fact that peak charges increase with scan rates, especially for Cd and Pb, appears to be a result of the fact that staircase voltammetry, as opposed to the true linear scan voltammetry, was used for the experiments. It has been well known that stripping of the surface substance is very sensitive to the waveform of the linear scan voltammetry, as reported by He.15 Since the stripping process studied here resembles that of the surface wave due to the nature of very thin mercury film, some of the stripping charge could be lost during the staircase scanning, depending on how fast the current within one potential step is being sampled. For example, the time resolutions for scan rates of 0.5 and 5 V/s are 2 and 0.2 ms, respectively, for 1 mV/step. If the surface charging currents are the same for both potential steps, then sampling 10 times faster at 5 V/s than at 0.5 V/s after each potential step could likely give up to 35% higher peak charge at low concentrations. An experiment conducted in a deoxygenated solution showed the same behavior, indicating that the oxidation of amalgam by oxygen is not likely to have caused the lower charges at slower scan rates. In the case of Cu, the situation is more complicated. At scan rates below 2 V/s, the peak charges seem to behave in the same manner as the Cd and Pb. However, at higher scan rates, the decreasing charges with scan rates may be due to the slow electron transfer kinetics of Cu during stripping. This hypothesis is consistent with the fact that the peak widths for Cu are broadened significantly at higher scan rates (see later discussion). Thus, it seems very possible that not all Cu metal inside the mercury is oxidized at scan rates of 5 V/s and higher because of the slow kinetics. Table 2 lists the concentrations of the three metals inside the mercury, calculated from the measured metal charges assuming a homogeneous thin film of 9 nm thick on a 1-mm-diameter GCE disk7 [(0.05 cm2)π(0.9 × 10-6 cm) ) 7.1 × 10-9 cm3 ) 7.1 × 10-12 L]. It can be seen from Table 2 that most of the amalgam concentrations determined by this method were found to be between 0.1 and 0.8 M, except for those corresponding to the 50 nM aqueous concentration. The metal ions are therefore preconcentrated by at least 5 orders of magnitude (0.1-0.8 M), which is significantly higher than for normal electroanalytical chemistry (millimolar range). The calculated Cu concentration inside the mercury thin film appears to exceed greatly the solubility of Cu in mercury (0.004 M after converting from atom %, as shown in ref 3). Two possibilities exist to explain the effect: (1) The Cu (13) Chovnyk, N. G.; Vashchenko, V. V. Russ. J. Phys. Chem. 1963, 37, 276. (14) Kemula, W. Pure Appl. Chem. 1967, 15, 283. (15) He, P. Anal. Chem. 1995, 67, 986.

1642 Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

Table 3. ASV Peak Widths at Half-Height A. At 50 nM of Each Species

B. At 0.5 V/s Scan Rate

scan rate (V/s)

Cd (mV)

Pb (mV)

Cu (mV)

concn (nM)

Cd (mV)

Pb (mV)

Cu (mV)

0.5 1 2 5 10

41 41 42 43 44

38 37 40 42 42

33 38 49 64 83

50 200 400 600 800

41 40 42 43 44

38 38 38 38 38

33 33 35 35 36

metal might have oversaturated inside the mercury film during the short deposition period of 2 min (as opposed to a slower equilibrium process). (2) The solubility of Cu in mercury might be different in the presence of other metals such as Pb and Cd. In addition, the possible deposition of Cu outside Hg would have resulted in the formation of intermetallic compounds such as CuPb and Cu-Cd, which have not been observed in the ASV voltammograms in this study. Thus, the nature of the possible Cu/Hg surface was not resolved in this work, and further experiments are needed to clarify the issue. Scan Rate Dependence of Peak Widths. Table 3 lists peak widths at half-height (b1/2) of the three metals at the lowest concentration of 50 nM (Table 3A) and at the lowest scan rate of 0.5 V/s (Table 3B). The entire profile of b1/2 dependence on scan rates and concentrations is plotted in Figure 3 for all three metals. The following two points are clear from Table 3 and Figure 3. First, peak widths of all three metal ions are quite narrow at 0.5 V/s, a scan rate which is an order of magnitude higher than 0.033 V/s (2 V/min), the theoretical limit of scan rate for the optimal LS-ASV behavior.4 These peak width characteristics of the three metals indicate that the diffusion of metal atoms/molecules through mercury is rapid, and the mercury surface indeed functions as a thin film, even at scan rates of 0.5 V/s or higher. Second, within the experimental error and the voltage resolution of (1 mV, Figure 3 clearly shows a consistent increase of peak width as a function of concentration. The peak widths of Cd and Pb at 50 nM increased only slightly upon increasing the scan rate, but the scan rate dependence of b1/2 is significantly affected by concentrations as the stripping peaks are broadened more and more at higher and higher concentrations. For Cu, however, while the peak is broadened significantly as the scan rate is increased, the overall peak broadening is much less affected by concentration increases than for the other two metals. Interpretation of Results of Peak Heights and Widths. The information demonstrated in Figures 2 and 3 has provided the basis of a three-dimensional picture of the dynamic LS-ASV behavior of the three metal ions as a function of scan rates and concentrations. The entire range of scan rates (0.5-10 V/s) studied here is beyond the upper limit of the theoretical model for LS-ASV. At these scan rates, the relationship between peak heights and scan rates is almost linear for Cd and Pb while good peak resolution is maintained. In the case of peak width versus scan rates, the LS-ASV behavior here is seemingly different from the theoretical model for thin mercury films. For example, a scan rate of 33 mV/s is the maximum for the model to achieve the lowest possible peak width (b1/2) of 38 mV.4 In contrast, the empirical peak widths of Cu, Pb, and Cd found in this study are 33, 38, and 40 mV, virtually identical to the theoretical value, even though the scan rate of 500 mV/s is much more rapid than for the model. Florence also reported the Pb peak to be consistently

Table 4. Characteristic Parameters of the Three Metals parameters DHg

(×105

cm2/s)

τ ) l2/DHg (ms)c k′ (cm/s)a

Cd °C)a

2.0 (25 2.45 (20 °C)b 5 × 10-5 0.6, 1 M KNO3

Pb

Cu

°C)a

2.1 (25 1.06 (25 °C)a 1.9 (20 °C)b 4.8 × 10-5 9.4 × 10-5 0.2, 1 M KCl 0.045, 1 M KNO3

a Data from ref 2. b Data from ref 13. c Calculated from D Hg (first row data) and l ) 9 nm. Note: DHg is the diffusion coefficient of metals inside mercury. τ is the normal diffusion time of metal through the mercury film of 9 nm. k′ is the rate constant of electron transfer across the mercury-solution interface.

Figure 3. Peak width at half height (b1/2) versus concentrations in 0.1 M KNO3 + 30 mM HNO3 + 0.1 mM Hg2+ solution at the following scan rates: 9, 0.5 V/s; b, 1 V/s; O, 2 V/s; 4, 5 V/s; and 0, 10 V/s for all three metals.

smaller than 38 mV using ultrathin mercury films, although the scan rate was 50 mV/s.16 Such difference from the model can be at least partially explained if the deposition of an ultrathin mercury film is assumed, while the theoretical model is based on a much thicker film. Indeed, our previous paper indicated that the thickness of the mercury coating on GCE is about 9 nm (9 × 10-7 cm),7 almost 3 orders of magnitude thinner than the lowest value of 4 µm in the model.4 As shown in Table 4, diffusion coefficients for diffusion of the three metals through mercury are on the order of 2 × 10-5 cm2/s. For the fastest scan rate of 10 V/s examined here, the time scale in each voltammetric step is 0.1 ms. This value is 3-4 orders of magnitude larger than the normal diffusion time (10-5 ms) of the individual species inside the mercury, as shown in Table 4. Thus, for the ultrathin mercury films, there should be more than enough time for metal atoms/ molecules to diffuse out of the mercury film before a new step of (16) Florence, T. M. J. Electroanal. Chem. 1970, 27, 273.

potential arrives. This behavior indicates that the diffusion through the mercury layer is not a rate-limiting step, even at a scan rate that is at least an order of magnitude higher than that predicted for an optimal behavior of LS-ASV signals. The point here is not to question the validity of the model for thicker films, but to point out that ASV performance can be better than predicted by the model when using linear scan voltammetry at ultrathin mercury films. The ultrathin mercury films and fast scan rates studied here have shown that there are two factors important to the prediction of LS-ASV behavior that have not previously been emphasized: kinetics and concentration. For Cu peaks, the kinetic factor plays an important role in the stripping process, as supported by four observations from this study: (i) the severe deviation from linearity of the peak height versus scan rates at scan rates higher than 2 V/s, (ii) the dramatic increase of peak width with scan rates, (iii) the decreasing peak charges with scan rates at scan rates higher than 2 V/s, and (iv) the large shifts of peak potentials with scan rates. Note from Table 3A that the peak widths for Cd and Pb increased only slightly at 50 nM over the range of 0.5-10 V/s scan rates, but those for Cu widened dramatically, from 33 to 83 mV, over the same scan rate range. Since diffusion is apparently not a rate-limiting step with ultrathin film condition, the strong dependence of Cu peak height, charge, and width on scan rates suggests that the fast stripping process is complicated by slow electron transfer kinetics. This behavior is also consistent with the rate constant of Cu, which is an order of magnitude smaller than that of Cd and Pb, as shown in Table 4. The concentration factor is demonstrated as the interdependence of peak widths on a combination of scan rates and concentrations for Cd and Pb. Note from Table 3B that the peak widths for all three metals are nearly constant at the moderate scan rate of 0.5 V/s, even when the concentration is varied from 50 to 800 nM. However, as the scan rates increase, the peak widths show increasing dependence on the metal concentrations for Cd and Pb. Such behavior has neither been predicted nor been demonstrated by any theories of ASV methodology. We believe that this dependence is related to the abnormally high amalgam concentrations, as shown in Table 2. Two scenarios might account for this behavior: (i) The formation of the weak intermetallic compounds, such as Cu-Cd and Cu-Pb,1 might have slowed down the diffusion inside the mercury film, as the three metals were highly packed inside the mercury. (ii) The diffusion coefficients of the metals inside the mercury may be strongly dependent on concentrations, making the metal diffusion much slower at higher amalgam concentrations. Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

1643

To explore further these possibilities, additional ASV experiments have been carried out at 2 V/s in three binary systems: Pb and Cu, Cd and Cu, and Pb and Cd. The results of these experiments showed that scenario (i) does not appear to be very likely for the following reasons: (1) Even though the amalgam concentrations were high, there was no additional Cu-Pb peak observed between the Cu and Pb peaks in the binary system of Cu and Pb, and also in the tertiary Cd-Pb-Cu system. (2) The peak heights and peak widths of the individual metals in the binary systems of Cd-Pb, Pb-Cu, and Cd-Cu were similar to one another at the same concentrations and similar to those of the tertiary Cd-Pb-Cu system at a scan rate of 2 V/s. For example, the peak widths and heights of Cd would not have been the same in the Cd-Cu and Pb-Cd systems if the supposedly weak bonding between Cd and Cu had interfered with the stripping process of Cd in the Cd-Cu system. In addition, peak widths and heights of Pb were very similar in the Pb-Cu and Pb-Cd systems. On the other hand, scenario (ii) is quite likely for the following two reasons: (1) The peak widths of Cd and Pb increased steadily with concentration at a fixed scan rate of 2 V/s, and they are similar to the relevant peak widths in two binary systems (Cd in Cd-Pb and Cd-Cu, Pb in Pb-Cu and Pb-Cd systems). (2) The amalgam concentrations of Cd and Pb in the systems used to measure the quoted diffusion coefficients were in the nanomolar range,13 which is at least 6 orders of magnitude lower than the concentrations encountered here (Table 2). Thus, it seems that the use of the quoted diffusion coefficients to interpret these data for high amalgam concentrations is questionable. The following scenario is more likely: The diffusion coefficients become smaller and smaller at higher and higher concentrations of metals inside the mercury, causing the slowing down of the metal diffusion and the broadening of the Cd and Pb stripping peaks. For Cu, similar abnormal diffusion behavior may also be occurring, as indicated in Figure 3, but the effect of this phenomenon has been masked by the greater effect of slow electron transfer kinetics. Background-Subtracted LS-ASV. As demonstrated previously, one of the advantages of using in situ-generated MTFs for LS-ASV is the greater reproducibility of the surface conditions of the mercury-coated electrode.7 This feature, in turn, has made it possible to reduce the background charging currents and the problem of steep sloping baseline when background subtraction is employed. Shown in Figure 4A are the ASV voltammograms for solutions containing 10, 60, and 110 ppb of Cd, Pb, and Cu, respectively, and Figure 4B for 10 ppb of Cd, Pb, and Cu after background subtraction. Note that, after background subtraction, the three stripping peaks are almost symmetrical, and the variation of background currents is insignificant compared to the signals. Note also the small peak at about +0.33 V due to the formation of Hg2Cl2 caused by the trace amount of chloride leached out from the reference electrode.7 The value of utilizing background subtraction in LS-ASV has been demonstrated previously10,14 as a strategy for reducing the charging currents. However, some difficulties were experienced. The original subtractive stripping analysis was conducted at two independent mercury drop electrodes which were characterized by virtually identical drop size.14 This method was later changed to subtracting the background currents for the same electrode in the absence of analyte.10 However, even with this refinement, the potential structural changes of the film at the same electrode 1644 Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

Figure 4. (A) LS-ASV voltammograms of Cd, Pb, and Cu in 0.1 M KNO3 + 30 mM HNO3 + 0.1 mM Hg2+ solution at 2 V/s at 2-min deposition time and 30-s rest period: (a) background scan, (b) 10 ppb each of three metal ions, (c) 60 ppb, and (d) 110 ppb. (B) Background-subtracted LS-ASV of the 10 ppb run obtained from (b) subtracting (a) in (A).

seemingly increased the chances of changes of background currents from run to run, and the method has not been as effective, as was demonstrated above in Figure 4B. In the previous attempts at background subtraction, hystereses were likely due to variations of surface structure of the mercury films. For the ultrathin, very reproducible mercury films described above, hysteresis is not expected to be a significant problem. Indeed, under the latter conditions, it has been demonstrated that background subtraction can be implemented with little hysteresis, and that the ratio of signal-to-background for the ASV peaks is further enhanced by this method. With the successful application of background subtraction in combination with ultrathin mercury films on GCE, LS-ASV can be used to detect sub-ppb levels of Cd, Pb, and Cu using a deposition time of only 2 min. Figure 5 shows the LS-ASV voltammograms of Cd, Pb, and Cu at concentrations as low as 0.25 ppb. In panel A, the Cd, Pb, and Cu peaks are seen as small bumps on top of the steep sloping baseline when 0.25 ppb concentration is added to the background. After background subtraction, however, distinct peaks are clearly resolved, as shown in panel B. The concentration of 0.25 ppb represents the detection limit for these three metal ions at a deposition time of 2 min and at 2 V/s scan rate. The detection level was confirmed by determining the sum of the standard deviations of three separate runs of seven ASV experiments. The noise observed in Figure 5A was apparently caused by the electronic noise (236 Hz in the FT of a voltammogram7) of the system, and that in Figure 5B was doubled in intensity by the subtraction procedure. It can be

CONCLUSIONS

Figure 5. (A) LS-ASV voltammograms of Cd, Pb, and Cu in 0.1 M KNO3 + 30 mM HNO3 + 0.1 mM Hg2+ solution at 2 V/s: (a) background scan, (b) 0.25 ppb each of three metal ions added, (c) 0.50 ppb, (d) 0.75 ppb, and (e) 1.00 ppb. (B) Background-subtracted ASV voltammograms obtained from subtracting the background scan (a) in panel A: (a) 0.25 ppb, (b) 0.50 ppb, (c) 0.75 ppb, and (d) 1.00 ppb. (C) After 15-point smoothing for the corresponding voltammograms in panel B. Except for scan a in each panel, all of the scans in the three panels have been displaced from original positions appropriately for demonstration.

The behavior of LS-ASV as a function of both scan rates and concentrations has been studied utilizing ultrathin mercury films. Experimental results reported here have provided the basis of a three-dimensional picture of the three metal ions of Cd, Pb, and Cu in terms of peak heights, peak charges, and peak widths at half-height. The very thin nature of the mercury film made diffusion so rapid that a fairly linear relationship of peak heights versus scan rates was obtained while good peak resolution was maintained. In the case of peak width versus scan rates, peak width as narrow as 33 mV (as opposed to the limit of 38 mV associated with theoretical the model for LS-ASV) was obtained at a moderate scan rate of 0.5 V/s. Also, it has been demonstrated that when ultrathin mercury films are utilized, kinetics and concentration factors have a significant effect on the behavior of LS-ASV at fast scan rates. The stripping process of Cu was shown to be primarily kinetics-controlled, while the peak widths of Cd and Pb at fast scan rates were more strongly dependent on the concentrations studied. This strong dependence of peak widths on concentrations suggests the likelihood of lower diffusion coefficients for metals inside mercury at abnormally high amalgam concentrations than previously measured. One important aspect of LS-ASV on ultrathin mercury films was the exceptionally high preconcentration efficiency for very low metal ion concentrations, which significantly enhanced the sensitivity of the ASV signals, and thus the ip/∆ib ratio. Fast scan LS-ASV was shown to enhance further this ratio, with the effect being particularly large for Cd and Pb and moderate for Cu. The peak charges obtained at moderate scan rates (0.5-1.0 V/s) appear to be significantly smaller than those obtained at fast scan rates (2-10 V/s), especially for Pb and Cd and for concentrations below 200 nM, due to the staircase voltammetry (as opposed to the true linear scan voltammetry). The performance of ASV signals in terms of peak heights, peak charges, and peak widths has been shown to be optimal at 2 V/s for the simultaneous analysis of Cd, Pb, and Cu. The advantage of the high reproducibility of background currents using the in situ-generated MTFs is once more demonstrated here. Such reproducibility has made it possible for background subtraction with little hysteresis. The strategy of background subtraction was used to achieve the subppb level of detection limit for the three metal ions studied here utilizing a 2-min deposition time and a 30-s rest period. ACKNOWLEDGMENT

removed, however, by a 0.5-ms filtering (instead of 0.05 ms) in the potentiostat, or by a 15-point smoothing as shown in panel C. The background subtraction technique works very well for Cd and Pb concentrations at or close to the detection limits of these metals. For Cu, however, there is a small distortion on the net baseline current after the subtraction. For concentrations at and higher than 10 ppb, the distortion becomes insignificant. This distortion appears to be caused by chloride, as indicated by the increasingly large Hg2Cl2 peak at +0.33 V.7 Overall, the performance of the background-subtracted ASV is excellent for the detection of low levels of these three heavy metal ions.

The author thanks John McDonald for his review of and inputs to the work and Bob Gleason for reviewing the manuscript. Discussions with Larry Anderson on several technical issues in the manuscript have been invaluable.

Received for review August 29, 1995. Accepted February 18, 1996.X AC950879E

X

Abstract published in Advance ACS Abstracts, April 1, 1996.

Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

1645