Anal. Chem. 2000, 72, 5285-5289
Voltammetry on Microfluidic Chip Platforms Joseph Wang,* Ronen Polsky, Baomin Tian, and Madhu Prakash Chatrathi
Department of Chemistry and Biochemistry, New Mexico State University, Las Cruces, New Mexico 88003
Microfluidic chip devices are shown to be attractive platforms for performing microscale voltammetric analysis and for integrating voltammetric procedures with on-chip chemical reactions and fluid manipulations. Linearsweep, square-wave, and adsorptive-stripping voltammograms are recorded while electrokinetically “pumping” the sample through the microchannels. The adaptation of voltammetric techniques to microfluidic chip operation requires an assessment of the effect of relevant experimental variables, particularly the high voltage used for driving the electroosmotic flow, upon the background current, potential window, and size or potential of the voltammetric signal. The exact potential window of the chip detector is dependent upon the driving voltage. Manipulation of the electroosmotic flow opens the door to hydrodynamic modulation (stopped-flow) and reversedflow operations. The modulated analyte velocity permits compensation of the microchip voltammetric background. Reversal of the driving voltage polarity offers extended residence times in the detector compartment. Rapid square-wave voltammetry/flow injection operation allows a detection limit of 2 × 10-12 mol (i.e., 2 pmol) of 2,4,6trinitrotoluene (TNT) in connection with 47 nL of injected sample. The ability of integrating chemical reactions with voltammetric detection is demonstrated for adsorptive stripping measurements of trace nickel using the nickeldimethylglyoxime model system. The voltammetric response is characterized using catechol, hydrazine, TNT, and nickel as test species. The ability to perform on-chip voltammertic protocols is advantageous over nanovial voltammetric operations that lack a liquid-handling capability. Coupling the versatility of microfluidic chips with the rich information content of voltammetry thus opens an array of future opportunities. Microfabricated microfluidic analytical devices, integrating multiple sample-handling processes with the actual measurement step, are of considerable recent interest.1,2 For obvious reasons, such devices are referred to as “lab-on-a-chip” devices. The advantages of such analytical microsystems, including speed, minimal sample/reagent consumption, miniaturization, efficiency, and automation, have been well documented.1-3 Such miniaturized systems have been coupled with various detectors, ranging from (1) Jakeway, S.; de Mallo, A. J.; Russell, E. L. Fresenius J. Anal. Chem. 2000, 366, 525. (2) Freemantle, M. Chem. Eng. News 1999, (Feb 22), 27. (3) Hadd, A.; Raymond, D.; Halliwell, J.; Jacobson, S.; Ramsey, M. Anal. Chem. 1997, 69, 3407. 10.1021/ac000484h CCC: $19.00 Published on Web 09/23/2000
© 2000 American Chemical Society
laser-induced fluorescence4 to mass spectroscopy.5 Nevertheless, new detection protocols are highly desired to meet the growing needs of microfluidic chip devices. The goal of the present work is to demonstrate, assess, and characterize various hydrodynamic voltammetric operations on microchannel chip platforms. The suitability of electrochemical detection for microchip electrophoretic systems has been demonstrated in connection with fixed-potential amperometric measurements.6,7 Cyclic voltammetry has been used recently for measurements in subnanoliter vials.8,9 Yet, scanning-potential voltammetric protocols have not been adapted to microfluidic chip platforms, despite their inherent advantages for ultrasmall environments and their rich information content. In the following sections, we will illustrate the ability to perform linear-scan, square-wave, and adsorptive-stripping voltammetric protocols while “pumping” the sample through the microchannels. We will also introduce new stopped-flow and reversed-flow electrochemical operations based on manipulation of the electroosmotic flow and will examine the effect of relevant experimental variables (particularly the driving voltage) on the background and analytical voltammetric response of the microfluidic chip. Such ability to perform different hydrodynamic voltammertic protocols on microchannel chips, and to integrate them with onchip chemical reactions, is advantageous over nanovial voltammetry,8,9 which lacks the sample preparation, liquid-handling, and fluid control or manipulation capabilities. Such coupling also enhances the power and scope of microfluidic chips, as it adds new dimensions of qualitative information, based on the redox activity of target analytes. EXPERIMENTAL SECTION Reagents. Catechol and hydrazine were purchased from Sigma. 2,4,6-Trinitrotoluene (TNT) was purchased from Radian International. The gold and nickel atomic absorption standard solutions (1000 mg/L) were purchased from Aldrich, as was the dimethylglyoxime (DMG). The electrophoresis buffer was a MES buffer (20 mM, pH 6.2) for the detection of catechol, a mixed borate/phosphate buffer (10 mM each, pH 8.0) for the detection of hydrazine, a 10 mM borate buffer (pH 10.5) solution for the detection of TNT, and a 10 mM ammonium buffer (pH 9.0) for stripping measurements of the Ni-DMG complex. Stock solutions were prepared daily in deionized water and filtered with a 0.45(4) Chiem, N.; Harrison, D. J. Anal. Chem. 1997, 69, 373. (5) Ramsey, R. S.; Ramsey, J. M. Anal. Chem. 1997, 69, 1174. (6) Woolley, A.; Lao, K.; Glazer, A.; Mathies, R. A. Anal. Chem. 1998, 70, 684. (7) Wang, J.; Tian, B.; Sahlin, E. Anal. Chem. 1999, 71, 3901. (8) Clark, R.; Hietpas, P.; Ewing, A. G. Anal. Chem. 1997, 69, 259. (9) Bratten, C.; Cobbold, P.; Cooper, J. M. Anal. Chem. 1997, 69, 253.
Analytical Chemistry, Vol. 72, No. 21, November 1, 2000 5285
Figure 1. Schematic of the integrated chip/voltammetric microsystem: A, glass chip; B, thick-film detector; S, B, and RB, sample, buffer, and running buffer reservoirs, respectively; W, working electrode; Sp, spacer.
µm filter (Gelman Acrodisk). Sample solutions were prepared by diluting the corresponding stock solutions in the running buffer solution. Apparatus. Electrochemical detection was performed with an electrochemical analyzer 621 (CH Instruments, Austin, TX) connected to a Pentium 166-MHz computer with 32 MB RAM. Details of the integrated chip/detection microsystem were described previously.7 The chip was fabricated by Alberta Microelectronic Co. (AMC, model MC-BF4-001, Edmonton, Canada) by means of wet chemical etching and thermal bonding techniques. It consisted of a four-way injection cross (connected to the three reservoirs), followed by a 72-mm-long/50-µm-wide reaction channel. The original waste reservoir was cut off by AMC, leaving the channel outlet at the end side of the chip, thus facilitating the end-column amperometric detection. A Plexiglas holder was fabricated for holding the separation chip and integrating the detector and reservoirs. Shortened pipet tips, inserted into the three inlet (reservoirs) holes on the chip (that were submerged into the corresponding buffer and sample reservoirs), were filled with 30 µL of the corresponding solutions. The ground connection was placed in the detection reservoir. A platinum wire was inserted into each of the reservoirs to provide connection to the highvoltage power supply. The electrochemical detector, placed in the waste reservoir (at the channel outlet side), consisted of an Ag/AgCl wire reference electrode, a platinum wire counter, and a screen-printed carbon working electrode. The screen-printed working electrode was placed at the channel outlet (see Figure 1). A gold-modified strip was used for the detection of hydrazine, a mercury-plated one for the detection of Ni-DMG, and an unmodified surface in connection to the detection of catechol. A 50-µm-thick Teflon spacer was used for controlling the distance between the channel outlet and the working electrode; the detector strip was held in place with the help of a plastic screw. Platinum wires were inserted into all the reservoirs for connection to the homemade highvoltage power supply. The power supply had an adjustable voltage range between 0 and (4000 V, while the polarities were controlled by a two-position switch. Screen-Printed Electrodes. The screen-printed electrodes were printed with a semiautomatic printer (model TF 100, MPM, Franklin, MA). An Acheson carbon ink (Electrodag 440B; Catalog No. 49AB90; Acheson Colloids, Ontario, CA) was used for printing the 0.3 × 2.5 mm rectangular working electrode strips. Details of the printing process and dimensions were described elsewhere.7 The working electrode for the detection of hydrazine was coated 5286
Analytical Chemistry, Vol. 72, No. 21, November 1, 2000
with gold by applying a square-wave pulse sequence between -0.2 and +0.75 V (vs Ag/AgCl), with a pulse width of 0.6 s for 30 min in a 0.1 M NaCl/1.5% HCl solution containing 300 ppm Au(III). The mercury-coated working electrode for the stripping detection of the Ni-DMG complex was preplated at a potential of -0.6 V for 20 min from a 0.1 M HCl solution containing 100 ppm Hg solution. Procedure. Sample Injections. Samples were injected electrokinetically by applying the desired voltage between the “sample” reservoir (s) and the ground (in the detection reservoir), with other reservoirs floating. The driving voltage was applied between the “running buffer” reservoir and the ground, with other reservoirs floating. The sample was electrokinetically loaded either continuously (for the scanning-potential experiments), by turning the high voltage on/off (during the hydrodynamic modulation), or by reversing the polarity at a desired driving voltage (for flow reversal). An applied potential of +1500 V (for 5 s) was used for loading the sample for the square-wave voltammetric experiment; a potential of +2000 V was applied continuously for the sample loading in the adsorptive-stripping analysis. Scanning Potential and Hydrodynamic Modulation Experiments. Linear-scan voltammograms were recorded by scanning the detector potential while holding the driving voltage constant at different values (0-4000 V). Stopped-flow voltammograms were recorded by turning the high voltage “on” and “off” in 1-s intervals during the voltammetric scanning. Flow Reversal Operation. The amperometric response was recorded with a time resolution of 0.1 s while applying 0.80 V vs the Ag/AgCl wire. A 20-s injection of the 0.5 mM catechol sample (in MES buffer, pH 6.2) was performed after the baseline stabilization. First, the driving voltage was kept constant at 1500 V until an increase was observed in the amperometric signal indicating the “arrival” of the sample plug. Then the driving voltage was switched from +1500 to -1500 V for 2 s and from -1500 to +1500 V for 3 s repeatedly by reversing the polarity switch of the high-voltage power supply. For comparison, an unmodulated amperogram for the injected sample plug was recorded while holding the driving voltage at 1500 V. Square-Wave Voltammetry. Square-wave voltammograms were recorded, using a flow injection protocol, with a driving voltage of 1500 V and an injection time of 5 s. Measurements were initiated when the analyte “plug” (from the 5-s sample injection) reached the detector. Square-wave parameters included a frequency of 30 Hz, ampltitude of 50 mV, and a step potential of 4 mV. Adsorptive Stripping Voltammetry. Adsorptive stripping voltammograms were recorded by applying a driving voltage of 2000 V, to both the nickel and ligand (DMG) reservoirs. Accumulation proceeded for 120 s at -0.7 V. The solution flow was stopped after 110 s, creating a quiescent solution. The stripping voltammogram was recorded while scanning the potential linearly from -0.7 to -1.2 V at a scan rate of 100 mV/s. RESULTS AND DISCUSSION Linear Scan Voltammetry. The potential window and background contributions have a profound effect upon the detection limit and scope of voltammetric analysis. Unlike static (batch) voltammetric experiments, where these characteristics depend primarily upon the electrode material and electrolyte solution, the background response and potential window of microfluidic elec-
Figure 2. Effect of the driving voltage on the linear-scan voltammetric background response over the 0.0 to -1.0 (A) and 0.0 to +1.0 V (B) ranges. Running buffer solution containing 20 mM MES (pH 6.2). Driving (running) voltage, 0 (a), 500 (b), 1000 (c), 1500 (d), 2000 (e), 3000 (f), and 4000 (g) V. Scan rate, 10 mV/s.
Figure 3. Effect of driving voltage on the voltammetric signal. Linearscan voltammograms of 5 × 10-4 M catechol in 20 mM MES buffer (pH 6.2). Driving voltage, 0 (a), 500 (b), 1000 (c), 1500 (d), 2000 (e), 3000 (f), and 4000 (g) V. Scan rate and running buffer, as in Figure 2. Insets display the resulting plots of E1/2 and current vs the driving voltage.
trochemical systems are also influenced by the high voltage used for driving the electroosmotic flow. Figure 2 assesses this voltage effect upon the background response over the cathodic (A) and anodic (B) regions of the thick-film carbon electrode detector. The end-column chip detector exhibits a favorable anodic potential limit (with the solvent decomposition reaction starting at potentials higher than ∼+0.78 V), along with a low residual current (at lower potentials) (B). Note that the exact potential window and the corresponding oxygen evolution current are strongly affected by the driving voltage. Such a dependence is attributed to the shift of the potential of end-column working electrodes with the driving voltage, reported earlier for conventional electrophoresis systems.10 Such a shift reflects the incomplete isolation of the detector from the flow-driving voltage (with the later altering the applied potential at the working electrode). The background current observed at lower potentials appears to be independent of the driving voltage. Apparently, the running potential has a negligible effect upon the double-layer charging and residual surface redox reactions. In contrast, the background response of the cathodic region (primarily the oxygen reduction contribution) is strongly affected by the high voltage driving the solution flow. The oxygen signal (peak shape turning to a sigmoidal one at high driving voltages) shifts ∼0.22 V upon raising the voltage from 0 to 4000 V. Such a gradual shift is accompanied by a 20% increase of the maximum current signal. The shift of the oxygen voltammetric signal is consistent with the shifts of the hydrodynamic voltammograms reported for conventional capillary zone electrophoresis.10b The increased oxygen reduction current reflects also the faster electroosmotic flow rate at higher driving voltages. Figure 3 displays linear scan voltammograms at 10 mV/s for a catechol solution flowing electroosmotically using different driving voltages ranging from 0 to 4000 V (a-g). The electrochemical detector displays nearly sigmoidal voltammograms (with a slight decrease of the maximum current at high potentials). The maximum current signal increases linearly 8-fold upon raising the voltage field, reflecting the decreased diffusion layer thickness at higher driving voltages (see inset for plot). Examination of the voltammograms reveals that the wave shifts linearly to a higher
potential upon increasing the driving voltage (from E1/2 of 0.400.68V; see plot). Such a shift of the analytical signal is similar to that described above for the oxygen background wave and is consistent with data obtained with conventional capillary electrophoresis systems.10 Attainment of thermodynamically relevant voltammetric data would thus require compensation of the electrical field effect (based on the linear relationship between the driving voltage and the potential shift)10a or complete isolation (decoupling) of the detector. Such on-chip attainment of useful information about the electrochemical characteristics of migrated substances represents the main advantage of the linear-scan voltammetric protocol. Obviously, it cannot compete with fixedpotential amperometry in terms of low detection limits. Hydrodynamic Modulation Voltammetry. Hydrodynamic modulation voltammetry (HMV) has been shown to be a powerful tool for solid electrode electroanalysis.11,12 The major advantage of this technique is that the measured current amplitude due to pulsations in the convection rate is free from most of the background contributions.12 While HMV has been performed earlier with conventional (macro) flow systems,11,12 there are no reports of analogous operations on microfluidic chips. In the adaptation of HMV protocols to microchip platforms, one should consider the different nature of the flow profiles (laminar vs flat) accrued from the different pumping mechanism (hydrodynamic vs electroosmotic, respectively). One convenient way to perform a HMV operation on a microchip platform (i.e., impose modulation in the analyte migration velocity) is to turn the driving voltage “on” and “off” while scanning the detector potential, in a manner analogous to stopped-flow HMV of conventional flow systems.11,12 Figure 4 displays typical conventional (A) and stopped-flow (B) voltammograms for the background electrolyte (a) and analyte/ catechol (b) solutions. The voltammogram of the blank solution displays no modulated response over most of the potential region examined (0.2-0.9 V; Figure 4B,a). Note the appearance of small current amplitudes above 0.9 V, reflecting the change in the solvent decomposition current upon switching the driving voltage
(10) (a) Matysik, F. M. J. Chromatogr., A 1996, 742, 229. (b) Wallenborg, S. R.; Nyolm, L.; Lunte, C. E. Anal. Chem. 1999, 71, 544.
(11) Blaedel, W. J.; Boyer, S. L. Anal. Chem. 1971, 43, 1538. (12) Wang, J. Talanta 1981, 28, 369.
Analytical Chemistry, Vol. 72, No. 21, November 1, 2000
5287
Figure 4. Linear-scan voltammograms for blank (a) and 1 × 10-3 M catechol (b) solutions in 20 mM MES buffer (pH 6.2), using normal flow (A) and stopped-flow (B) conditions. Scan rate and driving voltage (“on”), as in Figure 2.
(in accordance with Figure 1B). Conventional HMV protocols, in contrast, effectively correct for solvent decomposition background processes.11,12 Increased current oscillations (over the 0.4-0.7-V potential-dependent region), followed by a nearly plateau region are observed as the potential is scanned in the presence of the analyte (Figure 4B,b). The modulated response accounts for ∼33% of the nonmodulated limiting current (of Figure 4A,b). Very recently, Wang and Morris demonstrated an analogous modulation of the driving voltage for rejecting the fluorescence background at plastic microchip electrophoresis.13 The current amplitude of the on-chip stopped-flow operation is proportional to the analyte concentration. Both the modulated and nonmodulated maximal current signals increased linearly with the catechol concentration over the 5 × 10-4-2 × 10-3 M range (not shown; conditions, as in Figure 4). Flow Reversal. To facilitate various detection reactions, it is often desired to extend the residence time of the narrow analyte band within the detector compartment. A stopped-flow quiescent operation has been proposed recently for coupling NMR detection with microfluidic chips.14 However, such quiescent conditions are not attractive for various mass transport-dependent detection processes (e.g., electrodeposition, hybridization). Reversal of the flow direction has been used earlier for enhancing the response of conventional flow injection systems.15,16 Similar repeated reversals of the flow direction can be achieved with microfluidic chips by reversing the polarity of the driving voltage (in a manner analogous to control of the pump motion). Figure 5 compares the conventional (A) and reversed-flow (B) amperometric response to an injected 5 × 10-4 M catechol solution. The flow oscillations extend the residence time of the catechol sample plug in the detector compartment to nearly 100 s, as compared to ∼60 s employing the forward flow alone. Yet, the oscillated response decays rapidly at longer periods due to dispersion of the analyte band. (13) Wang, S.; Morris, M. D. Anal. Chem. 2000, 72, 1448. (14) Olson, D.; Lacey, M.; Webb, A.; Sweedler, J. Anal. Chem. 1999, 71, 3070. (15) Rios, A.; Luque de Castro, M.; Valcarcel, M. Anal. Chem. 1988, 60, 1540. (16) Wang, J.; Huiliang, H.; Kubiak, W. Electroanalysis 1990, 2, 127.
5288 Analytical Chemistry, Vol. 72, No. 21, November 1, 2000
Figure 5. Comparison of normal flow (A) and flow-reversal (B) amperometric signals. Response to 0.5 mM catechol in phosphate/ borate mixed buffer solution (pH 8). Separation voltage, 1500 V; injection voltage and time, 1500 V and 20 s, respectively; detection at +0.8 V using a gold preplated screen-printed electrode. (A) Analyte eluted out by applying a constant voltage between the running buffer reservoir and the detection reservoirs. (B) was recorded by repetitively turning the polarity switch of the high-voltage power supply from negative (2 s) to positive (3 s) during analyte elution
Figure 6. Square-wave voltammetry on a microchip platform. Increasing levels of TNT: (a) 0, (b) 20, (c) 40, and (d) 60 ppm in 10 mM borate buffer pH 10.5. Injection, 5 s at 1500 V. Running voltage, 1500 V. Voltammograms were recorded when the analyte plug, from the 5-s sample injection, eluted out. Square-wave conditions: frequency 30 Hz, ampltitude 50 mV, step potential 4 mV.
Rapid Square-Wave Voltammetry. The charging-current correction capability of pulse-voltammetric techniques holds great promise for sensitive detection in microfluidic devices. Fastscanning Osteryoung square-wave voltammetry17,18 is particularly attractive for monitoring rapid processes on microchip platforms. Figure 6 displays representative square-wave voltammograms for 47-nL samples containing increasing concentrations of the TNT explosive (20-60 mg/L, b-d). The potential scan was (17) Wang, J. Analytical Electrochemistry; VCH: New York, 1994. (18) Samuelsson, R.; O’Dea, J.; Osteryoung, J. Anal. Chem. 1980, 52, 2215.
initiated after the flowing sample zone entered the detector compartment. Well-defined peaks, corresponding to the formation and reduction of the hydroxylamine moiety, can be observed (despite the use of nondeaerated media). A detection limit of ∼10 mg/L can thus be estimated from the signal-to-noise characteristics of these data (S/N ) 3). Such concentration corresponds to 2 × 10-12 mol (i.e., 2 pmol) in the 47-nL injected sample. Lower detection limits of ∼1 mg/L were reported recently using laserinduced fluorescence detection19,20 or using fixed-potential amperometric operation.7 Yet, the square-wave operation increases the information content, and hence the selectivity, as it provides information on the electrochemical characteristics of the migrated substances. Adsorptive-Stripping Voltammetry. The ability of integrating chemical reactions with voltammetric detection was demonstrated for adsorptive-stripping measurements of trace nickel using the nickel-DMG model system. Adsorptive-stripping procedures, which rely on the formation and adsorptive accumulation of complexes of the target metal,19 have greatly expanded the scope of stripping analysis toward trace measurements of metals that cannot be electrolytically deposited. The adaptation of adsorptivestripping voltammetry to microchip platforms requires proper fluid control for mixing the metal analyte and complexing agent. For this purpose, the target nickel and DMG ligand were mixed at the channel intersection. The complexation reaction occurred along the reaction channel, while the nickel and DMG moved downstream. The resulting complex was accumulated on the mercury-coated thick-film detector. Figure 7 displays typical adsorptive-stripping voltammograms for increasing nickel concentrations in 500 µg/L steps (b-d), along with the background signal (a). Defined peaks, proportional to the nickel concentration, are observed for these trace levels, despite the 2-min short accumulation period, extreme potential, and presence of dissolved oxygen. Besides its analytical importance, Figure 7 illustrates the ability of voltammetry to probe the interaction of various components of the sample and running-buffer solutions. CONCLUSIONS Microfluidic chip devices appear to be ideal platforms for performing microscale voltammetric analysis. Such on-chip voltammetric analysis is advantageous over nanovial voltammetry,8,9 which lacks the sample preparation, liquid-handling, and fluid (19) Paneli, M.; Volgaropoulos, A. Electroanalysis 1993, 5, 535. (20) Wallenborg, S. R.; Bailey, C. G. Anal. Chem. 2000, 72, 1872.
Figure 7. Adsorptive-stripping voltammograms of (a) 0, (b) 500, (c) 1000, and (d) 1500 ppb Ni(II) in 10 mM ammonium buffer (pH 9.0) in sample reservoir. Reagent reservoir contains 2 × 10-5 M DMG. A running voltage of 2000 V was applied to both of the reservoirs. Linear scan rate, 100 mV/s.
control or manipulation capabilities. Such coupling also enhances the power of microchip devices as it adds a new dimension of information (based on redox activity) and opens up highly sensitive detection schemes. Further sensitivity enhancements would require proper attention to the challenge of oxygen removal on a chip platform. On-chip voltammetry distinctly differs from conventional voltammetric analysis due to the effect of the driving voltage upon the background and analytical signals. While this effect, and other performance characteristics, was illustrated in connection to an end-channel thick-film wall-jet detector, other detectors are expected to display similar trends. Work is in progress toward expanding the scope of voltammetric detectors in microanalytical systems. ACKNOWLEDGMENT Financial support from the US EPA (Grant 9V1097NAEX) and Sandia NL is acknowledged. Received for review April 27, 2000. Accepted August 21, 2000. AC000484H
Analytical Chemistry, Vol. 72, No. 21, November 1, 2000
5289
