Passive Conductivity Detection for Capillary Electrophoresis

Sensitive detection methods for microchips are required, but the development and availability of ..... We also thank mister Oliver Noverraz for techni...
0 downloads 0 Views 197KB Size
Download PDF
Anal. Chem. 2004, 76, 3126-3131

Passive Conductivity Detection for Capillary Electrophoresis Xiaoxia Bai,† Zhiyong Wu,† Jacques Josserand,† Henrik Jensen,†,‡ Helwig Schafer,§ and Hubert H. Girault*,†

Laboratoire d’Electrochimie Physique et Analytique, Institut de Chimie Mole´ culaire et Biologique, Ecole Polytechnique Fe´ de´ rale de Lausanne, CH-1015 Lausanne, Switzerland

A passive electrochemical detection principle that can be applied to capillary electrophoresis is presented. The separation electrical field is used to generate a potential difference between two electrodes located along the channel. For constant-current electrophoresis, the generated signal is proportional to the resistance of the solution passing between the two electrodes. Contrary to conductivity detectors that are ac driven and need to be decoupled from the separation field, the passive detection directly takes advantage of the separation field. The signal is simply measured by a high-impedance voltmeter. The detection concept has been validated by numerical simulations showing how the magnitude of the signal is related to the ratio between the electrode distance and the length of the sample plug. As a proof of the principle, this detection concept has been demonstrated by the electrophoretic separation of three alkali ions on a polymer microchip. Based on preliminary results, a detection limit of 20 µM and a dynamic range of up to 3 orders of magnitude have been achieved. Capillary electrophoresis (CE) is a method that can be easily carried out on microchips since it is relatively easy to implement experimentally without compromising speed and separation efficiency. Sensitive detection methods for microchips are required, but the development and availability of effective detectors has lagged behind the improvement of microchip technology. During the past decade, laser-induced fluorescence (LIF) has been the detection method of choice for microfluidic devices. Recently, mass spectroscopy (MS) has received considerable interest, in particular for proteomic analysis. However, except for offering high sensitivity and powerful information, both LIF and MS require bulky off-chip control instrumentation that highly compromises the benefits of miniaturization and portability. Therefore, alternative detection methods with high sensitivity are highly desirable to meet the growing needs of microfluidic systems and compact devices. Compared to the above-mentioned techniques, electrochemical detection is sensitive, low cost, independent of optical path length, and easy to microfabricate and miniaturize. The suitability of electrochemical detection for microchip electro* Corresponding author. Fax: +41 21 693 3667. Email: [email protected]. † Institut de Chimie Mole ´ culaire et Biologique. ‡ Current address: Department of Analytical Chemistry, The Danish University of Pharmaceutical Sciences, DK-2100 Copenhagen, Denmark. § Metrohm Ltd., Ch-9101 Herisau, Switzerland.

3126 Analytical Chemistry, Vol. 76, No. 11, June 1, 2004

phoretic systems has already been proven by numerous research reports.1-11 Compared to the amperometric and potentiometric detections, conductometric detection is in principle a universal detection method for most modes of CE, since the signal is obtained from the bulk of the solution and not from the electrode surface and therefore not linked to a specific species (redox potential). Intrinsically, it is less sensitive to interference effects and has so far seen many applications for the detection of inorganic and small, charged organic species on microchip devices. The two principal conductivity detection modes in a CE microchip are contact and contactless. Contact detection involves placing the detection electrodes in galvanic contact with the solution (using on-column or end-column configurations). Proper use of contact conductivity detection requires an electrical isolation (decoupling) of the detector from the high separation voltage, since the current associated with the separation voltage is usually several orders of magnitude larger than the one measured at the electrochemical detector. However, a proper decoupling is not trivial and is therefore frequently problematic for most of the electrochemical detection modes. A significant contribution to conductivity detection for CE was made independently by Zemann12,13 and Fracassi da Silva14 in 1998 by the introduction of the capacitively coupled contactless conductivity detector. This very old method, which used to be called oscillometry (see, for (1) Woolley, A. T.; Lao, K. Q.; Glazer, A. N.; Mathies, R. A. Anal. Chem. 1998, 70, 684-688. (2) Rossier, J. S.; Roberts, M. A.; Ferrigno, R.; Girault, H. H. Anal. Chem. 1999, 71, 4294-4299. (3) Hilmi, A.; Luong, J. H. T. Anal. Chem. 2000, 72, 4677-4682. (4) Guijt, R. M.; Baltussen, E.; van der Steen, G.; Frank, H.; Billiet, H.; Schalkhammer, T.; Laugere, F.; Vellekoop, M.; Berthold, A.; Sarro, L.; van Dedem, G. W. K. Electrophoresis 2001, 22, 2537-2541. (5) Wang, J. Talanta 2002, 56, 223-231. (6) Rossier, J.; Reymond, F.; Michel, P. E. Electrophoresis 2002, 23, 858-867. (7) Zeng, Y.; Chen, H.; Pang, D. W.; Wang, Z. L.; Cheng, J. K. Anal. Chem. 2002, 74, 2441-2445. (8) Hebert, N. E.; Kuhr, W. G.; Brazill, S. A. Anal. Chem. 2003, 75, 33013307. (9) Vandaveer, W. R.; Pasas, S. A.; Martin, R. S.; Lunte, S. M. Electrophoresis 2002, 23, 3667-3677. (10) Backofen, U.; Matysik, F. M.; Lunte, C. E. Anal. Chem. 2002, 74, 40544059. (11) Martin, R. S.; Ratzlaff, K. L.; Huynh, B. H.; Lunte, S. M. Anal. Chem. 2002, 74, 1136-1143. (12) Zemann, A. J.; Schnell, E.; Volgger, D.; Bonn, G. K. Anal. Chem. 1998, 70, 563-567. (13) Mayrhofer, K.; Zemann, A. J.; Schnell, E.; Bonn, G. K. Anal. Chem. 1999, 71, 3828-3833. (14) Fracassi da Silva, J. A.; do Lago, C. L. Anal. Chem. 1998, 70, 4339-4343. 10.1021/ac035462k CCC: $27.50

© 2004 American Chemical Society Published on Web 04/29/2004

Figure 1. Fabrication process of the microchip.

example, ref 15 for a review of conductometry and oscillometry), relies on the detection electrodes that are not in direct contact with the solution. It has the important advantages of easy alignment of the electrodes with the capillary, better electrical isolation, and long-term robustness.16-20 The decoupling has proven to be very successful even if high-frequency stimulation is needed to implement the method (the detector is also being sensitive to the amplitude of the applied voltage). In the present work, a passive conductivity detection concept for CE is described and demonstrated using a polymer microchip with integrated microelectrodes along one of the walls of the chip. Similar to the methodology of the reported wireless electrochemiluminescence detector,21 the CE separation electric field is used to perform the detection, which highly simplifies the experimental setup as no decoupling is needed. Moreover, the use of a highimpedance voltmeter also decreases the instrument cost. EXPERIMENTAL SECTION Materials and Chemicals. Poly(ethylene terephthalate) (PET) sheets (100-µm-thick Melinex) were purchased from Dupont (Geneva, Switzerland). 2-(N-Morpholino)ethanesulfonic acid (MES) and L-histidine (His) were obtained from Sigma (St. Louis, MO). The standard ionic solutions (KCl, LiCl, NaCl) with a concentration of 0.1 M were from Fluka (Buchs, Switzerland). Milli-Q water (Millipore, Bedford, MA) was used to prepare all fresh solutions, which were filtered (0.2-µm filter, Nalge, Rochester, New York) before use. Microchip Fabrication. Fabrication of the microchip is similar to previously published work.22 The PET sheet is photoablated by a UV laser excimer (argon fluor excimer laser at 193 nm, Lambda Physik LPX 2051, Go¨ttingen, Germany) generating a channel with a trapezoidal cross-section shape. The fabrication process is briefly illustrated in Figure 1. Carbon ink is pasted into two parallel microchannels, which are photoablated on one side of the PET substrate and dried at 80 °C for 30 min to form two (15) Pungor, E. Oscillometry and conductometry; International Series of Monographs in Analytical Chemistry 21; Pergamon Press: Oxford, U.K., 1965; Vol. 21. (16) Pumera, M.; Wang, J.; Opekar, F.; Jelinek, I.; Feldman, J.; Lowe, H.; Hardt, S. Anal. Chem. 2002, 74, 1968-1971. (17) Tanyanyiwa, J.; Hauser, P. C. Anal. Chem. 2002, 74, 6378-6382. (18) Wang, J.; Pumera, M.; Collins, G.; Opekar, F.; Jelinek, I. Analyst 2002, 127, 719-723. (19) Wang, J.; Chen, G.; Muck, A. Anal. Chem. 2003, 75, 4475-4479. (20) Tanyanyiwa, J.; Abad-Villar, E. M.; Fernandez-Abedul, M. T.; Costa-Garcia, A.; Hoffmann, W.; Guber, A. E.; Herrmann, D.; Gerlach, A.; Gottschlich, N.; Hauser, P. C. Analyst 2003, 128, 1019-1022. (21) Arora, A.; Eijkel, J. C. T.; Morf, W. E.; Manz, A. Anal. Chem. 2001, 73, 3282-3288. (22) Roberts, M. A.; Rossier, J. S.; Bercier, P.; Girault, H. H. Anal. Chem. 1997, 69, 2035.

Figure 2. (A) Schematic description of the setup; (B) equivalent circuit of (A). HV is the high-voltage source for CE separation, V represents the high-impedance voltmeter, VD represents the potential difference between the detection electrodes, ICE represents the current in the main separation channel, ID represents the current in the detection loop and RD represents the resistance of the solution between the detection electrodes.

conductive tracks (the electrodes have a width of 50 or 100 µm and a depth of 55 µm). A polyethylene/poly(ethylene terephthalate) (PE/PET) sheet (35 µm thick, Morane, Oxon, U.K.) is thermally laminated over the carbon microbands at 135 °C and 2 bar, for protection. Then, the main separation channel with a reservoir at each end is photoablated on the other side of the polymer substrate and perpendicular to the carbon tracks with an appropriate depth (i.e., 45 µm) in order to expose the two carbon microbands at the bottom of the channel, as shown in Figure 1. The main separation channel is about 45 µm deep, 50 µm wide at the top with a typical length of 6 cm (efficient separation length of 4 cm). A double-T with a typical center-tocenter distance of 100 µm is designed for injection purposes. The obtained micro network is also laminated by a PE/PET layer. CE and Conductivity Detection Setup. A high-voltage power supply (Spellman) and a self-made switching box are controlled by PC through a Labview VI program (National Instrument, Austin, TX). An electrokinetic floating injection is used to inject the sample in the double-T: typically, a potential of 500 V is applied between the sample and sample waste reservoirs for 20 s while the buffer and water reservoirs potentials are kept floating (no pinch). After that, a potential of 2 kV is applied along the main channel during separation, while the sample reservoir potentials are floating. According to safety consideration, the high-voltage power supply has to be handled carefully to avoid electrical shock. A digital voltmeter (model 2000 Multimeter, Keithley Instrument Inc., Cleveland, OH) controlled by a PC is connected to the microelectrodes and is used as the detector. This voltmeter can make DCV measurements from 0.1 µV to 1000 V. We mainly work in the 10-V range, with an input resistance bigger than 10 GΩ. The setup is schematically shown in Figure 2. Data acquisition of the detector is initiated immediately after separation and posttreated. Numerical Model. The finite element simulation has been performed with the numerical software Flux-Expert (Simulog23), (23) Simulog; [email protected]; 35 Chemin du Vieax Chene, 38240 Meylan Zirst, France.

Analytical Chemistry, Vol. 76, No. 11, June 1, 2004

3127

Table 1. Parameters Used in the Simulationa D/m2 s-1

t0/s

∆x/µm

10-10, 10-9 (ref)

30

150-700 (ref 200)

V/mm/s 4 (ref), 1

σ/s /σb

σelec/σb

lelec/µm

0.6(ref), 0.1

1000

50, 100 (ref)

R ∆x is the center-to-center distance between the two electrodes, and lelec is the length of the electrodes. σelec is the electrical conductivity of the electrode. The reference values used in most of the calculations are given in parentheses.

Figure 3. Methodology of the simulation. Two geometries A and B have been used, a moving Gaussian distribution of conductivity is imposed, and a zero potential at the end of the studied channel is imposed as boundary condition. A constant-current value is adapted to ensure an electrical field of 300 V/cm. One can note as shown by this schematic drawing that the minimum value of σ is not zero, but σb as indicated by eq 2.

in a 2-D Cartesian form. It is operated on a Unix workstation (Silicon Graphics Octane2, 1 Gb RAM). An ohmic model has been used to study the influence of the electrode position on the detection signal. The typical geometry is a vertical cross section of a microchannel with two microelectrodes at its bottom, as shown in Figure 3B. The electric field is solved by using the Laplace equation:

div(j) ) div(- σ(x,t)gradφ)

(1)

where j is the electrical current density, φ the electrical potential, and σ(x,t) the electrical conductivity of the solution. The solution conductivity is ensured both by the presence of a buffer (uniform contribution) and by a moving species (featuring one of the species that have been separated by the electric field). Its conductivity distribution is expressed as a Gaussian function24 added to the uniform buffer conductivity contribution:

σ(x,t) σb

)1+

(

)

σ/s (x - vt)2 exp σb 4Dt0

(2)

where σb is the electrical conductivity of the buffer and σ/s is the relative increase of the sample electrical conductivity (σs ) σb + σ/s where σs is the electrical conductivity of the sample), D is the diffusion coefficient of the sample species that is studied, t0 is the diffusion time of the sample before the detection section, t is the traveling time of the plug over the electrodes, and v is the velocity of the corresponding species (which is assumed to be the sum of both electroosmotic and electrophoretic flows). σb is arbitrally fixed to 1, and the corresponding current value is adapted to ensure an electrical field of 300 V/cm. A two-dimensional model is assumed. The migration effects on the plug shape are neglected (stacking effect and deviation of the electrical current lines by the electrodes inducing a distortion of the plug). The electrical conductivity contribution of the other species (other than the considered species and the buffer) is also neglected. In addition, the diffusion of the plug during its traveling above the electrodes is neglected (t , t0). (24) Girault, H. H. Electrochimie Physique et Analytique; Presses Polytechniques Universitaires Romandes: Lausanne, 2001.

3128 Analytical Chemistry, Vol. 76, No. 11, June 1, 2004

Two limiting cases have been studied: the first one (Figure 3A) corresponds to a very high charge-transfer resistance on the highly resistive electrode surface (all the electrical current remains in the channel) while the other one (Figure 3B) assumes no electrochemical charge-transfer resistance of the conductive electrodes (maximum bypass of the electrical current in the electrodes). In both cases, the voltage difference between the two electrodes is calculated by integrating the local electrical potential over the electrode length. The simulated channel (detection zone) is 1 mm long and 45 µm deep with two electrodes located on its bottom, symmetrically to the position (1.5, 0) mm. The electrode width (case B) is 100 µm with a depth of 55 µm. The diffusion in the initial part of the channel (4 cm) is taken into account by the “4Dt0” and σ/s /σb in eq 2. t0 is fixed to 30 s where 20 s corresponds to the diffusion during injection and 10 s to the electrophoresis process until the starting position for detection (initial position of the Gaussian plug; i.e., x ) 0). A zero potential value is defined as the boundary condition at the end of the microchannel. The parameters used in simulation are described in Table 1. A transient algorithm is used to calculate the evolution of the potential when the plug (σ profile) is moving through the studied geometry. After calibration, a mesh size of 10 µm in the channel, (2 µm close to the electrode edges) and a time step of 0.02 (for high D) or 0.005 s (for low D) are used for all the calculations. RESULTS AND DISCUSSION Principle of the Method. As shown schematically in Figure 2, a high voltage applied along the separation channel results in the establishment of a potential difference, VD, between the two parallel microelectrodes. This potential difference is proportional to the strength of the electric field and to the distance between the two electrodes. The measured potential difference VD is also proportional to the local solution resistivity between the detecting electrodes. When the separation channel is only filled with the buffer solution, the magnitude of VD between the detection electrodes leads to the buffer-dependent baseline signal. When a sample plug of different conductivity is introduced into the separation channel and passes through the electrodes, the variation in VD enables the detection of the plug. When the sample is more conductive than the running buffer, a negative signal is observed compared to the baseline and vice versa. From a practical point of view, a high-impedance voltmeter is needed to avoid noticeable current passing through the detection circuit (ID f 0). Configuration of the Detection Electrodes. The microelectrode design is an important parameter for the performance of the conductivity detector, e.g., the size of the electrodes and the

Figure 4. Distribution of the current lines around the electrodes in the microchannel (geometry B).

Figure 6. Effect of the electrode distance on the peak-over-baseline ratio of the simulated potential signal. The width of the electrodes is 100 µm. The full signs correspond to the calculations done on geometry A (no current deviation): the triangles correspond to the case of a big plug (high D), the dot points to that of a small plug (low D); the open signs correspond to the calculations done on geometry B (deviation of the current lines); two sets of simulation results correspond to {σ/s}/{σb} ) 0.6 and {σ/s}/{σb} ) 0.1, respectively. The open square points correspond to that of the corresponding experimental values. Notice that the vertical dotted line at 150 µm separates the data into two parts: one corresponds to the potential difference obtained between the integration on the electrodes; the other corresponds to that obtained between the local potential at the center of the electrodes. It has been proved that the potential difference between the potential integration on electrodes is identical to that between the local potential at the center of electrodes. The data below 100 µm are only observed by using geometry A, since for geometry B, the effect of the current deviation on the potential distribution in the detection zone cannot be neglected.

Figure 5. Effect of the electrode distance on the simulated potential difference signal. The case of no current deviation by electrodes (geometry A) is considered. The potential difference is obtained between the potential integration on each electrode. The width of electrodes is 100 µm. (a) A small plug generated by using low D is imposed; (b) a big plug generated by using high D is imposed.

distance between them. It has to be pointed out that the distance mentioned in the present work refers to the center-to-center distance of the electrodes. In principle, decreasing the dimension of the electrodes can increase the detection resolution and limit the noise level. This is the reason microelectrodes are used, i.e., with a depth of 60 µm and a width of 50 or 100 µm. The proportionality between the detection baseline and the electrode distance is confirmed experimentally (data not shown), when the electrode distance is ranging from 100 to 300 µm (the width of electrode band is 50 µm). At the same time, the peak magnitude is reduced, leading to a noticeable decrease of the peak-overbaseline ratio (the square points in Figure 6). Therefore, a small electrode distance is advantageous for improving the signal sensitivity and the separation resolution. Systematic study of the effect of electrode configuration is time-consuming due to the

microfabrication process, therefore, numerical simulations are applied to demonstrate the principle and to quantify the effect of the electrode distance. Finite Element Simulation. As mentioned above, the main feature of the present concept is that the contact detection electrodes are passive, with only a negligible Faradaic current passing through the detection circuit. However, as shown in Figure 4, when conductive electrodes (e.g., metallic) are considered, the current lines in the channel are perturbed by the electrodes, which bypass the solution resistance. Compared to the case of highly resistive electrodes (geometry A in Figure 3), the fraction of the current in the channel is decreased, leading to a smaller potential drop between the electrodes and therefore a lower baseline. It is confirmed by simulation, which also shows a slightly affected peak-over-baseline ratio of the signal (the open triangle and dot points in Figure 6). A thin semi-insulating monolayer film over the electrodes could also reduce the interfacial (electron-transfer) kinetics. Experimentally, the carbon ink electrode is more resistive than a metal electrode, leading to a situation between the two limiting cases described in the model (insulating and conductive). Figure 5 presents the effect of the electrode distance on the detected signal (geometry A). When this distance is increased, the baseline is increased due to the increased fraction of the sampled potential. At the same time, the peak amplitude remains constant (Figure 5a) when the electrode distance is larger than Analytical Chemistry, Vol. 76, No. 11, June 1, 2004

3129

the plug (small plug corresponding to D ) 10-10 m2/s leading to a length of ∼200 µm). In this case, the detection window sees more buffer than the plug. As a consequence, the relative potential contribution of the sample becomes negligible leading to a huge decrease of peak-over-baseline ratio as observed in Figure 6 (the full dot points). The observed flat signal in Figure 5a indicates that the plug has totally entered the zone between the centers of the two electrodes. Consequently, when the electrode distance is too big compared to the plug length, the resolution of a separation will be sacrificed (when more than one species are passing the detection zone). Figure 5b shows the case of a big plug (corresponding to D ) 10-9 m2/s leading to a length of 1 mm). When the distance between the electrodes is increased, both the baseline and the absolute height of the peak are increasing. Since the maximum detection window is smaller than the plug length, the electrodes mainly sense the plug leading to proportionality between the detected peak and the electrode distance. Therefore, as shown in Figure 6 (the full triangle points), the peak-over-baseline ratio decreases slowly versus the electrode distance. The values of the peak-over-baseline ratios observed in experiments are relatively lower than those for the simulated results, due to a lower value of σ/s /σb. Simulations with 10% instead of 60% give a peak-overbaseline ratio of 0.1 instead of 0.4, which is closer to the experimental values as shown in Figure 6. Both approaches, experiment and simulation, show that a smaller electrode distance results in higher peak-over-baseline ratio of the signal. Even if a sharper decrease is obtained experimentally, it could be due to the experimental detector noise (which is amplified for a large electrode distance) and the decreased sensitivity (induced by the boundary anomalies25). In addition, the distortion of the plug over the electrodes has to be taken into account (deviation of the electric field and therefore of the migration component of the flux). Moreover, due to the locally lower electric field on electrodes, a retention of the plug can be induced. This effect can be solved by using smaller electrodes or using semi-insulating monolayer films over electrodes. Those phenomena, which are not studied in the present work, have to be considered to improve the detected signal. Based on geometry A, Figure 7 represents the effect of σ/s /σb on the peak-over-baseline ratio, when the detector mainly senses a small sample plug by using a small electrode distance (50 µm). When σ/s /σb is zero (inset A of Figure 7a), meaning only the buffer passes through the detection zone, the potential difference between two electrodes is constant and no peak is observed. When σ/s is equal to σb (σ/s /σb ) 1, inset B), the total conductivity of the plug is twice the buffer conductivity, leading to a peak-overbaseline value of 0.5. When σ/s is much bigger than σb (i.e., σ/s /σb ) 100), the sample plug can strongly cancel the local ohmic drop during the peak time (inset C corresponding to P/B ) 0.99 for σ/s /σb ) 100). Consequently, the peak-over-baseline ratio tends to 1, leading to a nonlinear relationship between the ratio of σ/s /σb and the ratio of P/B. However, when the ratio of σ/s /σb is quite small (from 0.02 to 0.3), a linear relationship between P/B and σ/s /σb is observed with a correlation coefficient of 0.997. One should notice that this linear relationship also applies to the (25) Mikkers, F. E. P.; Everaerts, F. M.; Verheggen, T. J. Chromatogr. 1979, 169, 1-10.

3130 Analytical Chemistry, Vol. 76, No. 11, June 1, 2004

Figure 7. Effect of σ/s/σb on the peak-over-baseline ratio. The calculations have been done based on geometry A (Figure 2A). The electrode distance is 50 µm. A small plug generated by using low D is imposed. (a) σ/s/σb is in a range from 0 to 10; (b) σs/σb is in a range from 0 to 0.3. The inset shows three curves A, B, and C representing the detected potential difference between the two electrodes by using different σs/σb values, which correspond respectively to the points A (σ/s/σb f 0), B (σ/s/σb ) 1), and C (σ/s/σb f ∞). For the inset, the normalized potential difference by the baseline (corresponding to buffer) is employed, i.e., UN ) {U}/{Ub}, where U is the detected potential difference, Ub the potential difference corresponding to the baseline, and UN the normalized potential difference.

Figure 8. Electrophoregram of cation separation. The microchip has a total length of 6 cm and an effective length of 4 cm. Electrodes 100 µm wide with a distance of 200 µm are used for detection. The background electrolyte is 20 mM MES/His solution at pH 6; the sample is a mixture of 1 mM K+, Na+, and Li+ in the electrolyte. Electrokinetic floating injection is performed at 500 V for 25 s, and CE separation is performed at 2 kV.

measured peak value P versus σ/s . This range (σ/s /σb: 0.02-0.3) defines the ideal linear range of the detector. Separation and Detection Using Chip Electrophoresis. An electrophoregram corresponding to a separation of cations has been obtained (Figure 8) using the system described above fabricated on a polymer chip. Three cations, potassium, lithium, and sodium, have been completely separated in 15 s using MES/ His electrolyte. The retention behavior is similar to what has been detected by the conventional conductivity detector using face-toface microelectrode design, which has been reported elsewhere.26 Similar to the simulated signal, the negative signal refers to the more conductive sample plug compared to the background

electrolyte solution. The unstable baseline at the beginning of the experiments has not been clarified, it tends to decrease over time. For six repeated experiments, the relative standard deviations (RSD) of the migration time and the signal intensity are 2.6 and 5.3%, respectively. However, compared to the obtained repeatability on one chip, the reproducibility between different chips is comparatively worse mainly due to the variation of EOF due to the energy fluctuation of the photoablation laser. A RSD of 11.7% corresponding to migration time and 13.7% corresponding to peak intensity between five chips have been obtained. The PET microchips are quite robust even though they are used as disposable chips due to the low cost. To maintain the long-term reproducibility of chips, they have been kept in dry condition when not in use. A RSD of 7.5% is observed for one chip during one week. The obtained detection limit (LOD) for potassium ion is 20 µM based on the preliminary results. This LOD is comparable to those obtained by contactless detection mode that mainly falls into the micromolar range16,18,27 and is lower than that obtained by contact mode, which is in the millimolar28 and submillimolar29 range. The calibration behavior of the detector has also been characterized. The linear response range is from 50 µM to 1 mM with a correlation coefficient of 0.98 and an average RSD of 4% (n ) 4) concerning the peak intensity; a dynamic range of up to 3 (26) Bai, X.; Roussel, C.; Jensen, H.; Girault, H. H. Electrophoresis 2004, 25, 931-935. (27) Lichtenberg, J.; de Rooij, N. F.; Verpoorte, E. Electrophoresis 2002, 23, 3769-3780. (28) Prest, J. E.; Baldock, S. J.; Bektas, N.; Fielden, P. R.; Treves Brown, B. J. J. Chromatogr., A 1999, 836, 59-65. (29) Guijt, R. M.; Baltussen, E.; van der Steen, G.; Schasfoort, R. B. M.; Schlautmann, S.; Billiet, H. A. H.; Frank, J.; van Dedem, G. W. K.; van den Berg, A. Electrophoresis 2001, 22, 235-241.

orders of magnitude is thus obtained. Future work will aim at further improvement of the microelectrode configuration to optimize both the LOD and the dynamic range of the detector. CONCLUSION The proof of principle of the passive electrochemical detection has been shown by experiments and numerical simulations. As the detection electrodes are placed along the flow, the separation electrical field is used to generate the detection signal. Due to the absence of detection current, decoupling between CE and electrochemical detection is avoided. This simplification of the electrical design can be very useful for complex microsystems and compact microdevices that incorporate the CE technique. The major advantage of the present detection method is that it can be easily multiplexed either in the separation channels, for example, to follow the separation process, or in the side channels to control the conductivity of the solution in the injection channels. A small electrode distance is desirable to improve the detection sensitivity. A detection limit of 20 µM is obtained for potassium based on preliminary results. Further improvement of the electrode configuration (i.e., smaller and nearer) has to be done to decrease the LOD. ACKNOWLEDGMENT The authors thank Metrohm (Herisau, Switzerland) and the Swiss Commission for Technology and Innovation (CTI: 4421.2) for financial support. We also thank mister Oliver Noverraz for technical support. Received for review December 11, 2003. Accepted March 16, 2004. AC035462K

Analytical Chemistry, Vol. 76, No. 11, June 1, 2004

3131