Probing Electric Fields Inside Microfluidic Channels during

Jul 29, 2004 - Fast-scan cyclic voltammetry (FSCV) at carbon-fiber microelectrodes was used in microfluidic channels. This method offers the advantage...
18 downloads 12 Views 114KB Size
Download PDF
Anal. Chem. 2004, 76, 4945-4950

Articles

Probing Electric Fields Inside Microfluidic Channels during Electroosmotic Flow with Fast-Scan Cyclic Voltammetry Samuel P. Forry,† Jacqueline R. Murray,† Michael L. A. V. Heien,† Laurie E. Locascio,‡ and R. Mark Wightman*,†

Department of Chemistry, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599-3290, and Analytical Chemistry Division, NIST, 100 Bureau Drive, Stop 8394, Gaithersburg, Maryland 20899-8394

Fast-scan cyclic voltammetry (FSCV) at carbon-fiber microelectrodes was used in microfluidic channels. This method offers the advantage that it can resolve electroactive species not separated in the channel. In addition, this method provides a route to investigate the distribution of applied electrophoretic fields in microfluidic channels. To probe this, microelectrodes were inserted at various distances into channels and cyclic voltammograms recorded at 300 V/s were repeated at 0.1-s intervals. The use of a battery-powered laptop computer and potentiostat provided galvanic isolation between the applied electrophoretic field and the electrochemical measurements. In the absence of an external field, the peak potential for oxidation of the test solute, Ru(bpy)32+, was virtually unaltered by insertion of the microelectrode tip into the channel. When an electrophoretic field was applied, the peak potential for Ru(bpy)32+ oxidation shifted to more positive potentials in a manner that was directly proportional to the field in the channel. The shifts in peak potential observed with FSCV enabled direct compensation of the applied electrochemical potential. This approach was used to explore the electrophoretic field at the channel terminus. It was found to persist for more than 50 µm from the channel terminus. In addition, the degree of analyte dispersion was found to depend critically on the electrode position outside the channel. Microfluidics provides a way to miniaturize chemical analysis systems and allow rapid analyses that are inexpensive and generate little chemical waste. The use of channels with widths of a few micrometers and lengths of a few millimeters allows separations of chemical mixtures in a few seconds.1 The microchannels can be formed in plastic substrates with the use of simple molds.2 However, because the volumes of the microchannels used in such work are quite small, sensitive methods of detection must * To whom correspondence should be addressed. E-mail: [email protected]. † University of North Carolina at Chapel Hill. ‡ NIST. (1) Jacobson, S. C.; Culbertson, C. T.; Daler, J. E.; Ramsey, J. M. Anal. Chem. 1998, 70, 3476-80. 10.1021/ac049591s CCC: $27.50 Published on Web 07/29/2004

© 2004 American Chemical Society

be used. For this reason, many microfluidic devices employ laserinduced-fluorescence detection. Attractive alternatives to fluorescence detection are electrochemical techniques that can be readily miniaturized for use in microfluidic devices.3-8 The most common electrochemical detection scheme in microfluidic channels has been constant potential amperometry at single electrodes, although band electrodes in series at multiple potentials,9 multiple bands at the same potential,4 and sinusoidal voltammetry at individual electrodes10 have been used to provide more information. In microfluidic devices, the primary method of transport is electroosmotic flow. This is used to direct substances through an array of channels to accomplish sample introduction as well as electrophoretic separation.11,12 Fields exceeding 100 V/cm may be employed to generate the electrophoresis. Indeed, the applied electrophoretic field can generate a potential difference sufficient for oxidation of analytes that could undergo electrochemiluminescence, allowing optical detection,13 and changes in the electric field can be used to measure conductive analytes.14 However, the field complicates the use of traditional electrochemical methods because, even with electrical isolation, the electric fields that are (2) Anderson, J. R.; Chiu, D. T.; Jackman, R. J.; Cherniavskaya, O.; McDonald, J. C.; Wu, H.; Whitesides, S. H.; Whitesides, G. M. Anal. Chem. 2000, 72, 3158-64. (3) Woolley, A. T.; Lao, K.; Glazer, A. N.; Mathies, R. A. Anal. Chem. 1998, 70, 684-8. (4) Gavin, P. F.; Ewing, A. G. Anal. Chem. 1997, 69, 3838-45. (5) Lacher, N. A.; Garrison, K. E.; Martin, R. S.; Lunte, S. M. Electrophoresis 2001, 22, 2526-36. (6) Vandaveer, W. R.; Pasas, S. A.; Martin, R. S.; Lunte, S. M. Electrophoresis 2003, 23, 3667-77. (7) Klett, O.; Bjorefors, F.; Nyholm, L. Anal. Chem. 2001, 73, 1909-15. (8) Liu, Y.; Vickers, J. A.; Henry, C. S. Anal. Chem. 2004, 76, 1513-7. (9) Martin, R. S.; Gawron, A. J.; Lunte, S. M. Anal. Chem. 2000, 72, 3196202. (10) Hebert, N. E.; Kuhr, W. G.; Brazill, S. A. Anal. Chem. 2003, 75, 3301-7. (11) Seller, K.; Fan, Z. H.; Flurl, K.; Harrison, D. J. Anal. Chem. 1994, 66, 348591. (12) Ermakov, S. V.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 2000, 72, 35127. (13) Arora, A.; Eijkel, J. C.; Morf, W. E.; Manz, A. Anal. Chem. 2001, 73, 32828. (14) Bai, X.; Wu, Z.; Josserand, J.; Jensen, H.; Schafer, H.; Girault, H. H. Anal. Chem. 2004, 76, 3126-31.

Analytical Chemistry, Vol. 76, No. 17, September 1, 2004 4945

used to generate electroosmotic flow can interact with the electrochemistry.3 To avoid such interactions, the downstream electrode for electrophoresis and the electrochemical system should be decoupled. Various decoupling schemes have been explored, but most commonly the detecting electrode is placed further downstream than the ground electrode of the high-voltage system. This allows the channel potential to dissipate before it reaches the detection system.15 Placing the detecting electrode external to the channel minimizes electrical interactions, but band broadening can occur resulting in reduced separation efficiency.16 In the present work, fast-scan cyclic voltammetry at carbonfiber microelectrodes was used within polystyrene (PS) microfluidic channels. To prevent electrical cross talk, electrical isolation of the potentiostat from the electrophoretic power supply was used.10,16 Cyclic voltammetry has the advantage that overlapping compounds can be resolved by their potential differences.17 We examined the voltage shift of the cyclic voltammetric peaks recorded at different distances into the channel to provide a direct measure of the local field resulting from the ohmic drop due to the applied electrophoretic voltage. Prior work has shown that this effect can be overcome by placing the reference and working electrodes in proximity.7 Here we show that the shift in the peak potentials in cyclic voltammograms provides a direct measure of the ohmic drop and that it can be used to compensate for the effects of the electrophoretic field without decoupling it from electrochemical measurements. More importantly, this approach allows the local fields and their deviation from linearity to be examined. We found that the electrophoretic field at the channel terminus was nonlinear, similar to that found at the end of a capillary electrophoresis column18 and that its distribution effects band broadening in ways that were not appreciated previously. EXPERIMENTAL SECTION Chemicals. Tris(2,2′-bipyridyl)dichlororuthenium(II) hexahydrate (Ru(bpy)32+, Sigma-Aldrich, St. Louis, MO), and tris(hydroxymethyl)aminomethane (TRIS, Fisher Scientific, Fair Lawn, NJ) were used as received. Tris(2,2′-bipyridine)iron(II) (Fe(bpy)32+) was provided by Professor Holden Thorp (University of North Carolina at Chapel Hill). Electrode Fabrication. Disk-shaped carbon-fiber microelectrodes, insulated in glass, were fabricated as described previously.19 Briefly, carbon fibers (T650 (3-µm radius) Thornel, Amoco Corp. Greenville, SC) were inserted into glass capillaries, and the glass was heated until soft and then pulled to yield a tight seal around the carbon fiber. The glass insulation was typically ∼2 µm thick at the tip. Excess carbon fiber protruding beyond the glass seal was trimmed with a scalpel. The electrode tip was then dipped in epoxy (Epon 828 with 14 wt % m-phenylenediamine, Miller Stephenson, Danbury, CT) to fill any cracks between the (15) Woolley, A. T.; Lao, K.; Blazer, A. N.; Mathies, R. A. Anal. Chem. 1998, 70, 684-8. (16) Martin, R. S.; Ratzlaff, K. L.; Huynh, B. H.; Lunte, S. M. Anal. Chem. 2002, 74, 1136-43. (17) Caudill, W. L.; Ewing, A. G.; Jones, S.; Wightman, R. M. Anal. Chem. 1983, 55, 1877-81. White, J. G.; Soli, A. L.; Jorgenson, J. W. J. Liq. Chromatogr. 1993, 16, 1489-96. Park, S.; McGrath, M. J.; Smyth, M. R.; Diamond, D.; Lunte, C. E. Anal. Chem. 1997, 69, 2994. (18) Matysik, F.-M. Anal. Chem. 2000, 72, 2581-6. (19) Kawagoe, K. T.; Zimmerman, J. B.; Wightman, R. M. J. Neurosci. Methods 1993, 48, 225-40.

4946 Analytical Chemistry, Vol. 76, No. 17, September 1, 2004

Figure 1. Schematic of the microfluidic device and the associated electrodes. The HV power supply was connected to earth ground whereas the electrochemical system/lap top computer was battery operated with its common isolated from earth.

carbon fiber and the glass. The electrode was polished at a 45° angle on a polishing wheel (Sutter Instruments, Novato, CA). Microfluidic Devices. Channels for the microfluidic devices were fabricated using the hot imprinting method.20 Briefly, a negative relief silicon template (trapezoidal channel cross section, ∼20 µm deep and 20-50 µm wide) with a simple cross geometry (Figure 1) was fabricated by standard lithographic methods. PS was heated (115 °C) to near its glass transition temperature and pressed at 17 MPa (2500 psi) against the silicon template for 20 min. This transferred the channel pattern to the PS quickly and without damaging the master. The top for the microfluidic devices, also fabricated from PS, contained four holes for the solution reservoirs (4-mm diameter) formed with a mechanical drill. The top was clamped to the base and heated (103 °C for 8 min in a GC oven) to yield a tight seal. The length of the microfluidic channels in each device varied with the exact aligning of the tops (holes) and bottoms (channels), with ∼1 cm short arms (Figure 1, point 5 to reservoirs 1-3) and either a ∼4- or ∼2 cm separation channel (Figure 1, point 5 to reservoir 4). Procedures. The microfluidic device was placed on an inverted stage microscope (Axiovert 100TV, Zeiss, Germany) equipped with a CCD camera (SenSys, Photometrics, Tucson, AZ) for fluorescent observation of reagents within the device. Fluorescent imaging enabled the development of injection protocols (data not shown). Wire leads (silver-plated copper wire) were placed in each reservoir and connected to a programmable highvoltage power supply (µTk, Micralyne, Edmonton, AB, Canada) to generate electroosmotic flow and enable gated injections.12 Using Kirchhoff’s rules for resistive networks, the resistance of each solution-filled arm of the microfluidic device could be calculated from the measured resistance between reservoirs.11 The exit reservoir was always held at the ground potential of the highvoltage power supply. The voltages of the other three reservoirs were calculated based on the channel resistances so that the potential at the cross (position 5, Figure 1) was held constant, defining the magnitude of the applied electrophoretic field. Gated injections were utilized, which allowed the potential at the cross (position 5) and in the exit reservoir to remain constant. Time 0 was taken as the beginning of the injection. The working electrode was mounted on an x-y-z micromanipulator (Exfo Burleigh, Victor, NY) and positioned in the channel (20) Martynova, L.; Locascio, L. E.; Gaitan, M.; Kramer, G. W.; Christensen, R. G.; MacCrehan, W. A. Anal. Chem. 1997, 69, 4783-9.

Figure 2. Injection and FSCV detection. Ru(bpy)32+ and Fe(bpy)32+ (both at 2.5 mM) were electrophoretically injected (1 s) and allowed to migrate down a ∼4-cm channel in an electrophoretic field of 100 V/cm. The running buffer contained 25 mM TRIS at pH 7.4. The WE was inserted 50 µm into the end of the channel. The color plot shows the set of background-subtracted cyclic voltammograms (v ) 300 V/s, 10 Hz collection) that is unfolded at the switching potential and plotted vertically as a function of the acquisition time. The current intensity is shown in false color. The current from 1.00 to 1.03V for (A) Fe(bpy)32+ and 1.23 to 1.26 V for (B) Ru(bpy)32+ from each voltammogram was averaged to give the current vs time traces above the color plot. (C) The background-subtracted cyclic voltammogram is the average of those collected from 82.3 to 82.8 s.

terminus at the exit reservoir (Figure 1) with the beveled surface opposing the bottom of the channel at a 45° angle. The insertion distance into the channel was estimated by optical microscopy. Determining the location of the entrance to the channel caused an uncertainty of (5 µm in the reported positions. A chloridized Ag wire served as the electrochemical reference electrode and was placed in the exit reservoir. The electrochemical system was operated under battery control so that its circuit was isolated from the high-voltage power supply (Figure 1) in a way similar to methods employed previously.10,16 Electrochemical current was measured with a locally constructed, battery-operated picoammeter, which was interfaced to a battery-operated laptop computer via data acquisition, PCMCIA cards (ADC6062 and ADC6024, National instruments, Austin, TX). Potentials were generated with locally written software (LabVIEW 6.0, National Instruments) that allows fast-scan cyclic voltammetry (FSCV).21 All voltammograms were recorded at 300 V/s and were repeated at 10 Hz. RESULTS AND DISCUSSION Microfluidic Injections with Cyclic Voltammetric Detection. To evaluate detection by FSCV in the microfluidic device, Ru(bpy)32+ in 25 mM TRIS buffer, pH 7.4, was chosen for injection because it can be detected both by electrochemical oxidation and by photoluminescence. Following a 1-s injection with an applied electrophoretic field of 100 V/cm, photoluminescence established that Ru(bpy)32+ eluted from the ∼4-cm-long channel in ∼80 s (data not shown). Figure 2 shows FSCV detection following coinjection of 2.5 mM Ru(bpy)32+ and 2.5 mM Fe(bpy)32+ with the microelectrode inserted 50 µm into the channel. Fe(bpy)32+, which was selected since it should have electrophoretic properties similar (21) Michael, D. J.; Travis, E. R.; Wightman, R. M. Anal. Chem. 1998, 70, 586A92A.

to those of Ru(bpy)32+, has a peak anodic potential (Epa) of 0.88 V versus Ag/AgCl, whereas Ru(bpy)32+ has an Epa of 1.10 V versus Ag/AgCl. The data are presented in a color plot that allows all of the cyclic voltammetry data to be examined simultaneously.21 In this plot, the ordinate is the applied voltage and the abscissa is the time of acquisition. The current, from which the background has been subtracted, is encoded in false color. The color plot reveals that the two metal complexes are not separated by electrophoresis. However, a cyclic voltammogram recorded 82 s after injection reveals that both components are present and that they coelute (Figure 2C). The current extracted from successive voltammograms at the peak potential for Fe(bpy)32+ shows that it has a full width at half-maximum (fwhm) of 3.3 s (Figure 2A). Similar analysis of the current at the peak potential for Ru(bpy)32+ reveals a fwhm of 4.6 s (Figure 2B). The differences in dispersion are attributed to differences in the chemical interactions of the metal complexes with the walls of the channel. As demonstrated in Figure 2, FSCV provides an additional dimension that can be used to resolve electroactive substances that are not separated in the channel. However, this figure also demonstrates the interaction of the applied electrophoretic field with the electrochemistry. In the example shown, the values for Epa for both substance are shifted ∼130 mV from their location in a traditional electrochemical cell. Preliminary studies indicated that this shift was a function of the distance of insertion of the microelectrode into the channel. Therefore, we examined the electrochemistry of Ru(bpy)32+ within the channel and at the terminus in more detail to determine the effects on the cyclic voltammetry of the channel’s restricted dimensions as well as the applied electrophoretic field. Analytical Chemistry, Vol. 76, No. 17, September 1, 2004

4947

Figure 3. Diagram of potential gradients in an electrochemical cell. In this drawing, distances are not to scale; d represents the space between the working and reference electrode. In addition, all of the electrochemical potential drop is shown between the working electrode (WE) and the edge of the double layer or outer Helmholtz plane (OHP), although there is a potential drop at the reference electrode (RE) as well. Eapp,1.: potential applied between the WE and RE in the absence of an external field. Eeff.: potential drop between the WE and the adjacent solution after imposition of the applied electrophoretic field. Eeff + EiR: potential necessary in the presence of an applied electrophoretic field to accomplish the same electrolysis as without the applied electrophoretic field. Eapp,2: potential applied between the WE and the RE to compensate for the external electric field. Note that Eapp,2 ) Eapp,1 + EiR.

Fast-Scan Cyclic Voltammetry Inside a Microfluidic Channel. The device was filled with 5 mM Ru(bpy)32+ in 25 mM TRIS buffer, pH 7.4. Cyclic voltammograms recorded in the exit reservoir, which contained the reference electrode, were identical to those recorded in a conventional electrochemical cell. When the electrode tip was inserted in the channel, the separation in peak potentials (∆Ep) increased (∼42 mV, data not shown). Because the electrode has dimensions similar to those of the channel, it partially blocks the channel, increasing the resistance between the working and reference electrodes and thus increasing the peak separation. However, Epa shifted by only 3 mV when the electrode was inserted into the channel. With further insertion, Epa was independent of distance into the channel. Cyclic Voltammetry Inside a Microfluidic Channel with an Applied Electrophoretic Field. In a conventional electrochemical cell, most of the potential drop is at the electrode interfaces, with only a slight potential gradient in solution. Under

these conditions, the potential drop across the double layer of the working electrode is shown schematically in Figure 3 as Eapp,1.22 However, when another electric field is applied across the solution between the working and counter electrodes, it lowers the potential difference between the working electrode and the solution,7 resulting in a decreased potential for electrolysis (Eeff in Figure 3). This is because the potential at the solution side of the outer Helmholtz plane no longer has the same potential as the reference electrode. The net result is that the peaks in a cyclic voltammogram will appear to have shifted. The solution potential near the working electrode differs from that at the reference electrode by an amount we term EiR that arises from the ohmic drop in solution due to the applied field that generates electrophoresis. If a potential of equal magnitude is added to the originally applied potential, Eapp,2 ) Eapp,1 + EiR, the effects of the electrophoretic field are offset. The addition of this potential makes the potential difference between the working electrode and the solution the same as it was, Eapp,1, in the absence of the applied electrophoretic field. The potential of peaks within a cyclic voltammogram will be shifted from their original value, Ep, to Ep + EiR. For a linear potential gradient between the reference and working electrodes as shown in Figure 3, the value for EiR is dVapp, where d is the insertion distance (in cm) and Vapp is the applied electrophoretic field. These expectations were confirmed experimentally. The shape of the cyclic voltammogram and the position of Epa recorded at an electrode within the channel did not change when the high-voltage power supply was connected to the microchip with 0 V applied, establishing that the two systems were electrically isolated. However, when the electrophoretic voltage was increased, the peaks in the cyclic voltammogram for Ru(bpy)32+ shifted to more positive potentials, but the separation in peak potentials did not change. Only when an appropriate voltage was added to the applied potential could the complete cyclic voltammogram be obtained (Figure 4A). The peak heights did not increase, indicating that convection induced by the electrophoretic flow has a negligible effect on the current with the rapid scan rate employed. At a fixed location within the channel, Epa was found to change linearly with applied electrophoretic field (Vapp). For example, with the electrode tip 100 µm within the channel, the observed relationship was linear with Epa ) 0.011Vapp + 1.10; r2 ) 0.9999. The intercept of the regression analysis gives Epa in the absence of an applied electrophoretic field as expected, and in this example, the slope approximates

Figure 4. Effect of applied electrophoretic field on the cyclic voltammetry at a carbon-fiber microelectrode of 5 mM Ru(bpy)32+. The buffer contained 25 mM TRIS at pH 7.4. Scan rate, 300 V/s. (A) Voltammograms recorded 100 µm into the channel with a r ) 3 µm carbon fiber polished to an elliptical shape with a major radius of 5 µm measured with an optical microscope. Dotted voltammogram, no applied field. Solid line voltammogram, 50 V/cm. Dashed line voltammogram, 100 V/cm. (B) Voltammograms recorded 100 µm into the channel with a r ) 3 µm carbon fiber polished to an elliptical shape with a major radius of 15.5 µm. Dotted voltammogram, 25 V/cm applied field. Solid line voltammogram, 75 V/cm. Dashed line voltammogram, 100 V/cm. 4948 Analytical Chemistry, Vol. 76, No. 17, September 1, 2004

the insertion distance in centimeters. FSCV offers the advantage in these experiments of continuous monitoring of the voltage shifts arising from the ohmic drop in solution. In contrast, voltage shifts monitored with hydrodynamic voltammograms require construction from multiple amperometric recordings, each at a different potential.16 For electrodes with a larger electroactive area, the cyclic voltammograms showed distortion that increased with the applied electrophoretic field (Figure 4B). This was not apparent with electrodes beveled at angles of 45° yielding an elliptical surface with a major radius of 8.5 µm but was apparent with smaller bevel angles. This is in part due to the higher faradaic current with the larger electrodes that increases the electrochemical ohmic drop. However, an additional contributor is the nonuniform voltage gradient across the solution at the electrode tip. For example, with an applied electrophoretic field of 100 V/cm, the potential difference across the 16-µm surface, the long dimension of the beveled tip that is parallel to the electrophoretic field, is 160 mV. Thus, at each applied potential in a cyclic voltammogram, the voltage drop between the electrode and solution varies across its surface, resulting in a broad voltammogram. Consistent with this, the amount of distortion was found to increase with a decrease in the bevel angle. Measurement of Channel Voltage Gradients. From measurements at different insertion distances, it was confirmed that shifts in Epa provide a direct measure of the electrophoretic field within the channel. With a fixed electrophoretic voltage, the value of Epa shifted to more positive values as the electrode was inserted deeper into the channel. For example, with insertion distances between 50 and 150 µm and with an applied electrophoretic field of 100 V/cm, Epa ) 117d + 1.06, r2 ) 0.9988. Taking into account the uncertainty of (5 µm associated with locating the entrance to the channel, the calculated value of Epa at d ) 0 is approximated by the intercept to fall between the values of 1.02 and 1.12 V, very close to the experimental value of 1.10 V. Thus, the interaction between the applied electrophoretic field and the electrochemical measurement is predictable with a slope that approximates the electrophoretic field. This allows an offset value to be calculated for a particular electrophoretic field and insertion distance that will restore the appropriate electrochemical potentials. Although the voltage offsets with applied electrophoretic fields are linear with insertion distance when the electrode was inserted far into the channel, the field was found to be nonlinear near the channel exit (Figure 5). In this figure, the solid lines are from the regression analysis of values obtained deep within the channel (50-150 µm), while the points are the experimentally measured values of Epa within the channel and the exit reservoir. This figure reveals that the electrophoretic field extends quite far into the waste reservoir. Even 50 µm from the channel exit, a voltage shift of 100 mV is found for Epa in the cyclic voltammogram of Ru(bpy)32+. This effect also has been seen at the exit of a capillary during electrophoresis.18 This phenomenon arises from the finite resistance at the channel terminus that causes the electrophoretic field to disperse into the exit reservoir in a nonlinear fashion. This finding explains why shifts in amperometric responses occur with (22) Bard, A. J.; Faulkner, L. R. Electrochemical Methods, 2nd ed.; John Wiley and Sons: New York, 2001.

Figure 5. Anodic peak potential for the oxidation of Ru(bpy)32+ as a function of electrode distance within the microfluidic channel with an applied external field. Voltammograms were recorded in three different external fields. The designation 0 µm refers to the edge of the exit reservoir with positive insertion distances indicating electrode placement in the channel. Lines are the linear regression result from data obtained at insertion distances of g50 µm. Other conditions as in Figure 4A.

Figure 6. Electrode location and band dispersion. 5 mM Ru(bpy)32+ was electrophoretically injected (1 s) and allowed to migrate down a ∼2-cm-long channel in an electrophoretic field of 100 V/cm. Other conditions as in Figure 2. (A) Detection 50 µm within the channel. (B) Detection in the exit reservoir, 50 µm from the channel end with the electrode insulation resting on the bottom plate. In both cases, the current from a 30-mV range at the peak potential for Ru(bpy)32+ was averaged to give the current vs time trace.

changes in applied electrophoretic field strengths3 even with microelectrodes located external to microchannels. Potential offsets such as these are difficult to diagnose with amperometric detection but certainly affect amperometric sensitivity. Electrode Location and Band Dispersion. To further evaluate analyte band dispersion in relation to the location of the microelectrode in the microfluidic device, Ru(bpy)32+ was again chosen for injection, but in microfluidic devices with ∼2-cm-long separation channels to lower peak tailing. Figure 6A shows FSCV detection of an injection of 5 mM Ru(bpy)32+ with the microelectrode inserted 50 µm into the channel and an appropriate offset to account for the voltage shift caused by the 100 V/cm applied electrophoretic field. Elution is at 30 s with a fwhm of 2.8 s. Surprisingly, with the microelectrode in the exit reservoir, 50 µm outside the channel exit, and with its insulation resting on the bottom plate of the device, little additional band broadening was Analytical Chemistry, Vol. 76, No. 17, September 1, 2004

4949

seen (Figure 6B). The elution time was 32 s and the fwhm was 3.0 s. The fwhm of the peak recorded external to the channel was degraded less than in a previous report.16 However, in the prior work, comparison of the in-channel and external configurations was hampered because of a poorly defined area of the electrode in the latter position. In cyclic voltammetry, the dimension over which electrolysis occurs is defined by the length of the diffusion layer (δ), which is equal to22

δ ) (2DRT/nFυ)1/2 In this equation, D is the diffusion coefficient, RT/nF ) 0.059 V for this work at room temperature, and υ is the scan rate. Thus, in these measurements, the sampled region extends ∼0.5 µm from the electrode and provides a local view of dispersion at the microchannel terminus. In the measurement of Figure 6B, the electrode insulation was touching the bottom plate of the microchannel device. In this location, external to the microchannel, the electrophoretic field still exists (as seen in Figure 5) at this surface so that electroosmotic flow occurs and dispersion is quite low. Indeed, when the electrode was raised from the bottom plate by ∼10 µm, the peak dramatically widened and the peak concentration was lower. Thus, away from the surface, dispersion greatly increased, presumably due to mixing with the large volume of the exit reservoir.

4950

Analytical Chemistry, Vol. 76, No. 17, September 1, 2004

CONCLUSIONS The combined use of microelectrodes and fast-scan cyclic voltammetry provides a convenient way to monitor elution of substances from microfluidic devices and to examine the electrophoretic fields at the channel terminus. The peak shift in cyclic voltammograms provides a direct observation of the effects of applied electrophoretic fields on potentials within microfluidic devices. The magnitude of the shifts is determined by both the strength of the field and the distance of insertion of the voltammetric electrode into the microfluidic channel. The measured shifts can be used to adjust the applied electrode potentials for detection by cyclic voltammetry under electrophoretic conditions. Characterization of the potential at the channel terminus reveals a nonlinear electrophoretic field that could interact with amperometric detectors located in this region. ACKNOWLEDGMENT The authors thank Collin McKinney and John Peterson of the UNC Electronics Shop, Department of Chemistry, University of North Carolina, for the design and construction of the batteryoperated picoammeter. Funding for this research was provided by NSF Grant CHE-0096837.

Received for review March 17, 2004. Accepted June 11, 2004. AC049591S