Anal. Chem. 1998, 70, 2167-2173
Square-Wave Voltammetry Detection for Capillary Electrophoresis Geoff C. Gerhardt,*,†,‡ Richard M. Cassidy,‡ and Andrzej S. Baranski‡
Canadian Food Inspection Agency, Saskatoon, Saskatchewan S7N 2R3, Canada, and Chemistry Department, University of Saskatchewan, Saskatoon, Saskatchewan S7N OWO, Canada
A square-wave voltammetry (SWV) detection method for capillary electrophoresis (CE) using end-capillary detection is investigated. Although sensitive detection methods have been reported using off-capillary electrochemical detection methods, these methods typically use fragile carbon fiber electrodes and complex decoupling apparatus to separate the high separation voltage from the detection electronics. While end-column detection systems are not usually as sensitive as off-column techniques, they require much simpler instrumentation and are more robust. By modifying the Osteryoung SWV method to accommodate the high-frequency sample rate required by CE, a detection limit of 5 × 10-7 M (S/N ) 10) was obtained for dopamine using end-capillary detection, which is comparable to sensitivities obtained using offcapillary detection with amperometric detection. Although capillary electrophoresis (CE) has made major inroads in the analytical laboratory since its introduction in the early 1980s, it has only seen limited use for trace analytical applications. This is primarily due to the lack of sensitive detection systems available for CE. UV detection has been the most popular detection method used, but, being a mass-dependent detection method, its sensitivity is severely compromised by the short light path lengths encountered in CE. Laser-induced fluorescence1 has been used to achieve sensitive detection but is limited to compounds which fluoresce or are amenable to derivatization with a fluorophore. Ewing et al. introduced2 amperometric electrochemical detection for capillary electrophoresis (CE-ECD) in 1987. To avoid the detrimental effects of the separation voltage, they used a conductive joint to decouple the separation voltage from the detector. Since this introduction, several other implementations of this off-capillary CE-ECD method have been reported.3-6 Although good sensitivities were reported for offcapillary methods, the construction of the decoupling joint is not straightforward and it is intended for use with fiber electrodes inserted into the capillary end which are less robust than disk electrodes. This coupled with the general fragility of the detection †
Canadian Food Inspection Agency. University of Saskatchewan. (1) Cheng, Y.; Dovichi, N. J. Science 1988, 242, 562. (2) Wallingford, R. A.; Ewing, A. G. Anal. Chem. 1987, 59, 1762. (3) Wallingford, R. A.; Ewing, A. G. Anal. Chem. 1988, 60, 258. (4) Zhong, M.; Lunte, S. M. Anal. Chem. 1996, 68, 2488. (5) Hu, S.; Wang, Z.; Li, P.; Cheng, J. Anal. Chem. 1997, 69, 264. (6) Park, S.; Lunte, S. M.; Lunte, C. E. Anal. Chem. 1995, 67, 911. ‡
S0003-2700(97)01115-3 CCC: $15.00 Published on Web 04/14/1998
© 1998 American Chemical Society
apparatus make off-column CE-ECD unsuitable for routine use. End-column detection has also been reported7-13 where a diskshaped electrode is aligned with the end of the separation capillary. Generally, when similar analyses were done with off-column and end-column detection, the latter has given poorer sensitivities while offering a much simplified and more robust detection system.6,13 We will present here an electrochemical technique called square-wave voltammetry (SWV) for CE-ECD with end-capillary detection. SWV should offer sensitivities comparable to those of off-capillary CE-ECD while maintaining the durability and simplicity of end-column detection. Since SWV is a scanning technique, it can also give qualitative information (i.e., shape of the voltammogram) to assist in peak identification. Although SWV has been used extensively by physical electrochemists and electroanalytical chemists14 in static solutions, it has seen only limited use in flowing systems such as liquid chromatography15-17 and flow injection.18,19 Because of the high separation efficiency of CE, implementation of SWV as a CE-ECD detection method requires special consideration of the SWV instrumentation and waveform/current sampling parameters to maintain an acceptable overall sample rate. To properly quantitate CE peaks in the order of 2 s, a sample rate of at least 5 Hz is required. The ultimate sample rate of the SWV technique is governed by the period of the dc ramp, since each dc ramp provides one data point (i.e., the height of the square-wave voltammagram peak). Previous implementations of SWV in flowing systems15-19 used dc scan periods of several seconds (0.2-1.0 Hz) and square wave frequencies of ∼300 Hz. To achieve the high sample rates required for CE detection, the dc scan time must be decreased to less than 100 ms. This requires that square-wave frequencies of 1 kHz or higher be used to maintain an adequate voltage resolution in the (7) Matysik, F. M.; Meister, A.; Werner, G. Anal. Chim. Acta 1995, 305, 114. (8) Fermier, A. M.; Colo´n, L. A. J. High Resolut. Chromatogr. 1996, 19, 613. (9) Fermier, A. M.; Gostkowski, M. L.; Colo´n, L. A. Anal. Chem. 1996, 68, 1661. (10) Wen, J.; Cassidy, R. M. Anal. Chem. 1996, 68, 1047. (11) Chen, M.; Huang, H. Anal. Chem. 1995, 67, 4010. (12) Lu, W.; Cassidy, R. M.; Baranski, A. S. J. Chromatogr. 1993, 640, 433. (13) Sloss, S.; Ewing, A. G. Anal. Chem. 1993, 65, 577. (14) Osteryoung, J. G.; Osteryoung, R. A. Anal. Chem. 1985, 57, 101A. (15) Samuelsson, R.; O’Dea, J. J.; Osteryoung, J. Anal. Chem. 1980, 52, 2215. (16) Kounaves, S. P.; Young, J. B. Anal. Chem. 1989, 61, 1469. (17) Roush, J. A.; Anderson, M. R. J. Liq. Chromatogr. 1993, 16, 3887. (18) Fung, Y. S.; Mo, S. Y. Anal. Sci. 1994, 10, 179. (19) Baranski, A. S.; Norouzi, P.; Nelson, L. J. Proc. Electrochem. Soc. 1996, 9, 41.
Analytical Chemistry, Vol. 70, No. 10, May 15, 1998 2167
voltammagram as well as to keep the step heights of the dc voltage ramp low. Square-wave frequencies of this magnitude require particular attention to the electrode-electrolyte system in order to minimize the charging current of the system. The following will report the development of a SWV method that accommodates the high sample rates required by CE for a sensitive and robust CE-ECD system. The neurotransmitters dopamine and epinephrine were chosen for this investigation of CE-SWV. These compounds can be reversibly oxidized at moderate voltages, have simple, one-peak SWV voltammograms, and are easily separated using CE. Also, many off-column CE-ECD methods have been reported2-6 for similar neurotransmitters, which allowed comparisons to be made.
Figure 1. Working/auxiliary electrode assembly schematic.
EXPERIMENTAL SECTION Chemicals. All solutions were prepared from water purified using an ULTROpure with a TFM membrane and NANOpure II with Type 1 ORGANICfree cartridge kit (Barnstead Corp., Dubuque, IA). Unless otherwise specified, all reagents were analytical grade. Dopamine and epinephrine were obtained from Aldrich Chemical Co. (Milwaukee, WI). Capillary Electrophoresis System. A computer-controlled CE system was constructed to inject samples and pressure-rinse capillaries and to perform separations automatically. The autosampler consisted of a circular sample tray controlled by a stepper motor; the tray accommodated 10, 10-mm × 32-mm glass vials. Vials were lifted and pressure-sealed to the capillary/ electrode assembly by a stepper-motor-controlled elevator. Optoelectronic devices were used to sense the position of both the elevator and the autosampler tray. A Spellman CZE2000 highvoltage power supply (Plainview, NY) was used, which accepted an analog input to control the separation voltage and provided an analog output to monitor the separation current. The capillary electrophoresis unit was enclosed in a Plexiglas box with a safety interlock switch located on the lid to prevent access to the system while the high voltage was turned on. An electronic interface for the autosampler was constructed and used in conjunction with a PCL-812 data acquisition card (B & C Microsystems, Sunnyvale, CA) to enable computer control of the stepper motors and sensors as well as the high-voltage power supply. Software was written using the programming language Delphi 3.0 (Borland, Scotts Valley, CA) to perform automated runs, enabling a series of separation or detection conditions to be studied in a controlled manner. Micrometer-Size Electrodes. The electrodes used for electrochemical detection, shown schematically in Figure 1, were prepared by sealing a 25-µm gold or platinum wire in a 10-µL micropipet tube (Canadawide Scientific, Ottawa, ON, Canada). Only about 1-2 mm of the end of capillary tube was melted around the wire, using a short exposure to a butane flame. The capillary tube was cut to a length of ∼2 cm, silver epoxy was tamped into the open end, and a 0.5-mm copper wire was slid into the capillary to make contact with the 25-µm electrode wire. The copper wire was soldered to the core wire of a shielded cable, and heat-shrinkable tubing was used to cover the glass electrode and exposed wire. A 3-cm length of 1/8-in.-o.d. × 1/16-in.-i.d. stainless steel tubing was filled with standard 5-min epoxy and slid over the electrode assembly, allowing the sealed electrode
to extend out ∼5 mm from the end of the stainless steel tube. Excess epoxy was wiped away, and this assembly was allowed to cure for 1 h. Contact was made between the stainless steel tubing and the coaxial cable shielding by extending the shielding over the stainless steel tubing and securing it with heat-shrinkable tubing. The glass electrode was ground using wetted 3200-grit Micro-mesh abrasive paper (Micro-mesh, Wilton, IA) to expose the 25-µm disk, followed by further polishing with wetted 6000grit Micro-mesh abrasive paper to provide a smooth electrode surface. This electrode was then secured in the CE-ECD cell shown in Figure 2 using a nut and ferrule. The stainless steel shielding provided adequate shielding of the working electrode, as well as acting as the auxiliary electrode in the potentiostat and terminating electrode for the high-voltage power supply. Electrochemical Cell. The electrochemical cell that was used for end-column electrochemical detection in CE is shown in Figure 2. This CE-ECD cell is more compact than the three-dimensional micropositioners that have been used by other workers to align the capillary and electrode. Once the capillary and electrodes are fastened in the cell, axial alignment can be achieved by adjusting the allen screws around the capillary or electrode. The distance between the capillary and electrode can be adjusted by screwing in or out the brass capillary holder. Alignment was usually done using a ×20 Hastings triplet (Edmond Scientific, Barrington, NJ). Better alignment could be achieved by removing the CE-ECD cell from the CE unit and using a stereomicroscope to view the capillary-electrode alignment, but we found that this degree of accuracy was not necessary for our applications. A distance between the capillary and electrode of ∼10-20 µm was found to be optimal. Once the electrode and capillary were axially aligned, the capillary could be removed and replaced routinely, requiring only reestablishment of the optimal capillary-electrode distance. Holes in the side of the CE-ECD cell accommodated a Ag|AgCl reference electrode and fluid connections. A Ag|AgCl reference electrode small enough to fit in the CE-ECD cell was constructed by wrapping glass fiber filter paper around a 0.5-mm Ag wire coated with AgCl and covering this assembly with heat-shrinkable tubing. This produced a cylindrical electrode with an o.d. of ∼1/8 in. that was compressible enough to friction-fit into one of the CE-ECD observation holes. The volume of the CE-ECD cell (∼500 µL) was refilled with the CE separation buffer before each run. A potentiostat and current follower similar to those described
2168 Analytical Chemistry, Vol. 70, No. 10, May 15, 1998
Figure 2. CE-ECD cell schematic.
by Baranski and Szulborska20 were used to control the potential of and to sense the current flowing through the working electrode. Data Acquisition/Control Hardware and Software. A Pentium-class personal computer using the Windows NT 4.0 operating system, equipped with a National Instruments AT-MIO16E (National Instruments, Austin, TX) high-speed interface card, was used to output an analog waveform to the working electrode and acquire current readings from the working electrode. The AT-MIO-16E and accompanying dynamic link libraries allowed waveform generation and current sampling to be synchronized, which was essential in interpreting SWV current response. The memory and CPU requirements of the computer were dictated by the nature of the data acquisition requirements. For routine CE-SWV analyses, a Pentium 133-MHz CPU with 32 MB of RAM was sufficient. When all raw data were stored rather than analyzed and discarded, a dual Pentium Pro 200-MHz CPU system with 64 MB of RAM was required. Software was developed using Delphi 3.0 to repeatedly apply a waveform to the working electrode and synchronously acquire, analyze, and store the current data. The data could be interpreted in real time, or stored data could be loaded and reanalyzed to generate electropherograms. The algorithms used to interpret the current response from each waveform cycle will be discussed later. Most of the waveform parameters could be modified from within the software, including the pre- and postscan voltage/time, square wave frequency/ amplitude, dc ramp initial/final voltages, and ramp time. Although the AT-MIO-16E card was capable of data capture rates of 500 kHz, a rate of 100 kHz was used. Higher rates were not found to be necessary and resulted in slower data processing times due to the higher volume of data. This software was integrated with the autosampler control software. Calculations of S/N and Peak Efficiency. The S/N ratio was determined by calculating the standard deviation (σn-1) of a representative section of the electropherogram baseline (∼10 (20) Baranski, A. S.; Szulborska, A. Electroanal. Chem. Interfacial Electrochem. 1994, 373, 157.
peak widths) near the dopamine peak and dividing this into the height of the dopamine peak. Peak efficiency was calculated as the number of theoretical plates (N), where N ) 5.54(tR/ W1/2), tR is the migration time, and W1/2 is the peak width at halfheight. CE-ECD Conditions. Fused-silica capillaries (40 cm, 25 µm i.d., Polymicro Technology, Phoenix, AZ) were used in these experiments. Separations of neurotransmitters were achieved using pH 7.0, 50 mM Na2HPO4-Na2B4O7, and a separation voltage of 30 kV. Before use, capillaries were rinsed with 0.1 M sodium hydroxide for 5 min, followed by the run buffer for 5 min. Sample injections were made electrokinetically by inserting the anodic end of the capillary in the sample solution and applying 5 kV for 10 s. Samples were diluted to the desired concentration using the CE separation buffer. SWV conditions: prescan pulse, -500 mV for 30 ms; postscan pulse, 1000 mV for 30 ms; dc ramp, 100500 mV in 56 ms; square wave, 2000 Hz, 50 mV; quantitation method, average peak current, first 30% of forward and reverse pulse current response rejected, 70 mV detection bandwidth, 15 points included in running average. RESULTS AND DISCUSSION Electrode Type and Positioning. To accommodate the 1-kHz or higher square-wave frequencies necessary for CE-SWV, micrometer-size electrodes were used. Although cylindrical carbon fiber electrodes inserted into the end of the separation capillary have been popular for the analysis of neurotransmitters, the use of cylindrical micrometer-size electrodes in CE requires rather complex apparatus to decouple the high separation voltage from the detection system. Even using the decoupling apparatus, a potential gradient along the length of the carbon fiber may exist.21 Although this may cause insignificant problems for fixed potential amperometric detection, it would cause broadening of the voltammagram peak in SWV. As discussed in the introduction, end-capillary detection with disk-shaped ultramicroelectrodes was (21) Lu, W.; Cassidy, R. M. Anal. Chem. 1994, 66, 200.
Analytical Chemistry, Vol. 70, No. 10, May 15, 1998
2169
chosen over decoupling techniques. Although the high separation voltage will still cause a shift in the electrode/solution potential near the exit of the capillary, the electrode surface is normal to the voltage gradient. This potential shift will be uniform across the surface of the electrode and can be accounted for by including an “offset” voltage to any potentials applied to the electrode. Although we found it difficult to construct carbon fiber disk electrodes, we were able to construct 10- and 25-µm Pt and Au disk electrodes relatively easily and reproducibly. Pt electrodes have been used for the detection of neurotransmitters,11 but we observed no SWV reaction on a 25-µm Pt electrode. This may have been due to slow kinetics of the oxidation of neurotransmitters on Pt. A well-defined SWV peak was observed for both dopamine and epinephrine using 10- and 25-µm Au electrodes. Although increasing S/N should accompany decreasing electrode surface area, we obtained better S/N using the 25-µm Au electrode. It may be that the smaller currents produced by the 10-µm Au electrode were more affected by extraneous noise. The 25-µm Au disk electrode was axially aligned with the outlet of the capillary and positioned 10-20 µm from the end of the capillary. Electrode-capillary separations greater than this began to broaden CE peaks, while smaller separations decreased sensitivity. By monitoring the SWV voltammogram peak with the separation voltage on and off, we found that, with an electrode positioned as indicated above, a shift of ∼-250 mV occurred in the voltammogram when a 30-kV separation voltage was applied (40-cm × 25-µm capillary, 50 mM, pH 3.0 phosphate buffer). To account for this, a 250-mV offset voltage was added to the waveform applied to the working electrode during electrophoretic separations. Any waveform voltages reported here do not include this offset voltage. It should be noted that, although several electrodes were used throughout this experiment, little difference was observed in their response. Typically, electrode response remained constant for several months without polishing, and electrode failure was due usually to a breakage of an electrical connection within the electrode due to repeated movement of the coaxial cable connected to it. Square-Wave Voltammetry Parameters. The SWV parameters that can be modified are the dc ramp initial/final voltages, sweep time, and square-wave amplitude and frequency. The dc ramp initial/final voltages are governed by the compounds being detected. We found that, for the neurotransmitters which were oxidized at ∼300 mV, a dc ramp from 100 to 500 mV accommodated the entire SWV peak. The sweep time was maintained constant throughout this experiment at 56 ms. At shorter sweep times, the SWV peak for the neurotransmitters became smaller and skewed, while longer sweep times compromised separation efficiency, since longer sweep times lower the overall sample rate. Increasing the square-wave frequency will increase the squarewave peak current (and hence the sensitivity), but this will be tempered by a higher charging/faradic current ratio.16 The solution resistance, electrode diameter, and stray capacitance of the system will limit the sensitivity gains obtained by raising the square-wave frequency. A series of square-wave frequencies were examined to determine the optimal frequency for the detection of neurotransmitters. A plot of square-wave frequency versus S/N showed that a frequency of 2000 Hz was the instrumental limit for this system. Above this frequency, excessive charging 2170 Analytical Chemistry, Vol. 70, No. 10, May 15, 1998
currents interfered with the measurement of the faradic current, decreasing the dopamine S/N. Thus, further studies of SWV detection used a 2000-Hz square wave with a dc ramp time of 56 ms to provide an overall sample rate of 20 Hz (signal averaging and pre- and postscan pulses lowered this to 5-10 Hz). Theoretically, the optimal square wave amplitude for an reversible system is 50/n mV.16 To determine whether reversibility or interference from charging currents may require a different square-wave amplitude for the test solutes studied, various amplitudes were investigated. Increasing S/N was observed with increasing amplitude until ∼60 mV, after which S/N plateaued and then began to decrease when amplitudes greater than ∼100 mV were used. Low-frequency noise (baseline drifting) was more pronounced when square-wave amplitudes above 60 mV were used. This was probably due to an increased charging/faradic current ratio, making the system more sensitive to changes in the electrolyte during a CE run. A square-wave amplitude of 50 mV was found to be optimal. SWV Data Analysis. Having optimized the SWV parameters to minimize the charging/faradic current ratio, we investigated techniques for the interpretation of the current response to maximize the S/N where a higher than optimal charging/faradic current ratio may exist. In the traditional Osteryoung squarewave method,16 the current is sampled at two points for each square wave, t1 (the end of the first pulse of the square wave) and t2 (the end of the second pulse of the square wave). The difference current [(current at t2) - (current at t1)] for each square wave is plotted versus dc ramp voltage to obtain a peak-shaped voltammogram for an electroactive species. The major advantage of SWV results from the fact that an electrode’s charging current will decay exponentially (ia e-t/RC), while faradic currents decay in proportion to t-1/2 (assuming linear diffusion).22 In the Osteryoung technique, the majority of the charging current will have decayed at the end of each pulse, allowing the faradic current to be sampled independently. The software we developed enabled us to sample the current across the entire square-wave period and use a selected portion of the forward and reverse current response to calculate the difference current. We found that it was advantageous to collect more current samples near the end of the forward and reverse pulses and use signal averaging to increase the S/N. When most of the current response was used to calculate the difference current (i.e., 0-20% of the initial current rejected), the S/N was lower, possibly due to the charging component of the current interfering with the measurement of the faradic current. When >30% of the initial current response was rejected, the advantage of signal averaging was reduced, and S/N decreased. A rejection ratio of 30% was optimal for the neurotransmitters studied. Other analyte/electrolyte systems may need to be optimized independently, as the optimal rejection ratio will depend on the frequency of the square wave, conductivity of the electrolyte, diameter of the electrode, and diffusion rate of the analyte; this was a relatively simple process, though, since the raw data from each electrophoretic run could be stored and reanalyzed. The peak height of the square-wave voltammogram (see Figure 3a) is normally plotted versus time to obtain time(22) Wightman, R. M.; Wipf, D. O. In Electroanalytical Chemistry; Bard, A. J., Ed.; Dekker: New York, 1988; Vol. 5, p 308.
Figure 3. CE-SWV quantitation techniques.
resolved quantitative information in a flowing system using SWV detection.16-19,21 When this method was used for CE-SWV detection, a drifting and noisy baseline was observed due to two phenomena. First, a general movement of the voltammogram up and down the current scale caused random noise (∼0.5 Hz), and second, the general shape of the voltammogram of the background electrolyte gradually changed over the length of a CE run, causing a general baseline drift. Although the exact cause of these phenomena was not investigated, it is likely that the solvent electrolysis products, created in the CE-ECD cell due to the separation current, changed the conductivity and composition of the small volume of terminating electrolyte in the CE-ECD cell; these electrolysis products could also have been migrating past the working electrode into the separation capillary. To reduce this noise, an algorithm was developed that determined the position of the center of the voltammogram peak and drew a baseline across the bottom of the peak using a user-defined peak width (see Figure 3). This algorithm was similar to that used by applications which interpret chromatographic data where peaks are identified by scanning the data for inflection points. Using both the peak height (Figure 3b) and the average peak current (Figure 3c) of the baseline-corrected voltammogram peak, baseline drifting was significantly reduced for a range of peak widths, but the average peak current method gave the best S/N when an optimal detection peak width was chosen. A peak width of 80 mV was found to be optimal. Narrower peak widths did not take advantage of signal averaging, while wider peak widths included weak peak data at the peak’s edges, which lowered the overall signal. Although we did not observe noticeable voltage shifts of voltammogram peaks within a run, it was not uncommon for a voltammagram peak to shift 10-20 mV throughout a day. This algorithm was able to accommodate these shifts and give consistent quantitation without user intervention. The detection algorithm was compact enough that it could be applied in realtime rather than requiring a postrun processing step. Various forms of signal averaging can be used to improve the S/N of continuous signals. In the case of CE-SWV, where sampling periods of 50-100 ms were used, care had to be taken to ensure that temporal resolution losses due to signal averaging did not compromise the separation power of the CE system. We found that a running average offered a better compromise between S/N and peak resolution than simple signal averaging. Figure 4 illustrates the effect the number of points included in the running
Figure 4. Effect of the number of points included in running average on dopamine S/N (O) and peak efficiency (×).
average had on the S/N and peak efficiency of dopamine. Peak efficiency begins to decline after more than ∼10 points are included in the running average, while S/N was a maximum at ∼20 points. For our applications, a running average of 15 points was chosen as an appropriate compromise between peak efficiency and S/N. Pre- and Postscan Voltage Pulses. A technique that was found to improve detection performance was adding pre- and postscan pulses to the applied SWV waveform. Kounaves and Young found that pretreatment of carbon fiber electrodes by repeatedly applying a 0-1500-mV triangular pulse for 60 s before use improved S/N by a factor of 2 for SWV detection in liquid chromatography.16 We also obtained significant improvement in sensitivity when, similarly, we preconditioned the Au electrode before each run. However, we found that, when a high-voltage potential pulse was incorporated at the end of the detection waveform, the run-to-run reproducibility and within-run stability of the system increased dramatically. Studies showed that a postscan pulse voltage of 1000 mV was found to be optimal and also that a prescan pulse was required to reequilibrate the electrode after the high-voltage postscan pulse. Although one would have expected that the reequilibration of the electrode should be done at the initial dc ramp potential, when prescan potentials from 200 to -1000 were examined, lower prescan equilibration voltages improved the S/N for the detection of neurotransmitters. A prescan potential of -500 mV was found to be optimal. The exact nature of this enhancement is not fully understood. It may be that the lower prescan potential provided a more amenable surface for the neurotransmitters to physically adsorb and preconcentrate, or it may condition the electrode surface to a state that is more suitable for electrochemical reaction. This more negative prescan potential also improved the stability of the system (less baseline drifting). Pre- and postscan pulses of 30-ms duration were found to be optimal. Longer pulse times did not provide significant improvements, while they increased the overall scan time (reducing the sample rate). CE-SWV of Neurotransmitters. A CE-SWV electropherogram of dopamine and epinephrine (5 × 10-6 M) using the optimized conditions is shown in Figure 5. Linear calibration curves were obtained for both neurotransmitters for concentrations in the range 0.50-50 µM (dopamine, y ) 0.846x - 0.692, r ) 0.9985; epinephrine, y ) 0.640x - 0.373, r ) 0.9995). Because Analytical Chemistry, Vol. 70, No. 10, May 15, 1998
2171
Figure 5. CE-SWV electropherogram of dopamine and epinephrine (5 × 10-6 M). Separation conditions, 50 mM Na2HPO4-Na2B4O7 at pH 7.0, 30 kV. SWV conditions: prescan pulse, -500 mV for 30 ms; postscan voltage, 1000 mV for 30 ms; dc ramp, 100-500 mV in 56 ms; square wave, 2000 Hz, 50 mV. Quantitation method: average peak current, first 30% of forward and reverse pulse current response rejected, 70-mV detection bandwidth, 15 points included in running average.
the current follower gain was optimized for maximum sensitivity, neurotransmitter concentrations above ∼70 µM did not give a linear response. The linear range of the system could be moved higher (sacrificing sensitivity) by lowering the gain of the current follower. Although the correlation coefficient for the calibration curve was good, we have found this to be a poor indicator of the linearity of a calibration curve. By plotting the response factor (peak area/dopamine concentration) versus dopamine concentration, reasonable linearity ((20%) was observed only in the 5-25
µM region. Although this linear region is quite narrow, we have found that it is not atypical to find calibration curves with good correlation coefficients to have more limited linearity when the response factors are plotted. Further work is intended to examine the source of this nonlinearity. A detection limit of 500 nM (1.5 fmol) for dopamine using CE-SWV was determined by injecting successively lower concentrations until a S/N of 10 was obtained. This detection limit compares favorably to detection limits of CE separations of neurotransmitters obtained using constant potential amperometric detection by Zhong et al.4 [600 nM, S/N ) 10 (120 nM, S/N ) 2 reported)], Hu et al.5 [67 nM, S/N ) 10 (20 nM, S/N ) 3 reported)], and Ewing et al.2 [0.7-1.3 fmol, S/N ) 10 (0.2-0.4 fmol, S/N ) 3 reported, as extrapolated from high concentration injections)]. A more qualitative illustration of the CE-SWV data is shown in a 3-D contour plot (see Figure 6). A comparison of Figures 5 and 6 illustrates the extent to which baseline drift can be eliminated using the quantitation techniques discussed above. The electropherogram shown in Figure 5 has only minimal baseline drift, while the 3-D contour plot shows a large initial baseline drift at potentials near the peak voltage of the neurotransmitters (note the difference in current scale between Figures 5 and 6). This 3-D contour plot illustrates the potential for analyte identification and calculation of peak purity on the basis of an analyte’s SWV response. CONCLUSION Previously, to obtain the sensitivities for CE-ECD reported here, constant potential amperometry using complex decoupling systems and cylindrical carbon fiber electrodes was used. This technique, because of the fragility of the electrode/decoupling system and the potential for electrode fouling in a nonpulsed
Figure 6. Three-dimensional CE-SWV electropherogram of dopamine and epinephrine (5 × 105 M). CE separation and SWV detection conditions are the same as in Figure 5. 2172 Analytical Chemistry, Vol. 70, No. 10, May 15, 1998
system, is unsuitable for routine use. Conversely, the CE-ECD system described here is both sensitive and robust. A potential limitation of CE-SWV may be that, due to its high-frequency ac waveform, only compounds with reversible electrochemical kinetics will respond. Further work is intended to examine the utility of CE-SWV for the separation and detection of other compounds. Recent results in our laboratory have shown that CE-SWV can be used to monitor the physical adsorption of hydrophobic compounds on Pt electrodes. This physical adsorption produces a peak-shaped SWV response similar to the faradic responses
shown here but requires only that the analyte be hydrophobic rather than electrochemically active. This aspect is currently under investigation and could lead to electrochemical detection methods for a variety of nonelectrochemically active analytes that are difficult to detect spectrophotometrically.
Received for review October 8, 1997. Accepted February 19, 1998. AC971115X
Analytical Chemistry, Vol. 70, No. 10, May 15, 1998
2173
