Experimental determination of sample stream focusing with fluorescent dye

Transcription

1 Electrophoresis 2008, 29, Jay Taylor G. D. Stubley Carolyn L. Ren Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, Ontario, Canada Received December 20, 2007 Revised March 4, 2008 Accepted March 5, 2008 Research Article Experimental determination of sample stream focusing with fluorescent dye This study aims to develop an experimental method for determining the focused-sample stream width in microfluidic chips. The focused-sample stream width in the junction of a cross-linked microchannel is determined by the distance between the two streamlines dividing the focusing sheath streams and the sample stream, which should be insensitive to the sample properties. However, the commonly used FWHM intensity method is dependent on the dye properties and cannot be used to estimate the sample focusing effects very accurately. The proposed method is based upon the fact that the cross-stream concentration profile of the sample stream in a region downstream of the junction matches the Gaussian distribution. Therefore, the width of the focused sample stream can be estimated from the channel widths and the variance and maximum concentration of the experimentally measured concentration profile. To validate this method, fluorescein dye has been used to obtain the concentration profile. It has been shown that the proposed method is independent of the fluorescent dye properties. Comparison between the proposed method, FWHM method, and some reported analytical methods also validates the proposed method. Keywords: Electrokinetic focusing / Full width at half maximum / Gaussian distribution / Microfluidic chips DOI /elps Introduction Typical bioanalytical and chemical synthesis devices include a flow platform or chip integrated with optical components, sensors, flow control components, heating components, and signal processing components. The central component of these devices is the flow chip in which the sample volume is transported and processed continuously. There is considerable incentive to miniaturize these devices to achieve many benefits, including reduced sample volumes, portability, reduced processing times, and reduced cost [1, 2]. A resulting miniaturized device is often called a lab-on-a-chip device. The degree of miniaturization that can be obtained in a lab-on-a-chip device is often limited by the degree to which the flow chip can be miniaturized. Flow chips with flow channels in the microscale dimensions are called microfluidic chips. Since the performance of a lab-on-a-chip device will often be limited by the performance of the microfluidic chip, there is a strong need to develop efficient and optimal microfluidic chips. There is a corresponding need to develop Correspondence: Professor Carolyn L. Ren, Department of Mechanical and Mechatronics Engineering, University of Waterloo, 200 University Avenue West, Waterloo, Ontario, Canada N2L 3G1 c3ren@mecheng1.uwaterloo.ca Fax: experimental and analytical methods for rigorously evaluating the performance of the elements in a microfluidic chip and their configuration. The present work addresses this need for one particular element of a microfluidic chip, the continuous focusing element. The continuous focusing element is a critical element in the microfluidic chip devices used for the detection of biological materials [3, 4]. The purpose of the focusing element is to take a flow stream transporting a sample species (i.e. cells, DNA or proteins) and focus it into an ultra-thin flow stream. Typical applications include cell cytometry, cell sorting, and discrete sample injection. In the first two applications, a buffer fluid solution with a suspension of biological cells is focused into a stream with a width comparable to the dimension of the cells [5 8]. In the focused stream, the cells are transported single-file to the downstream detection elements for characterization, identification, and processing. In these applications, the determination of the focused sample stream width is very important for evaluating the chip design and performance for the following two reasons. First, if the sample stream width is too large there are chances that two cells pass the detection point at the same time, which are not differentiable. If it is too small, it is possible that multiple cells will agglomerate before passing the detection point. The analysis of DNA and proteins in lab-on-a-chip CE devices is an application, which requires the injection of a fluid solu-

2 2954 J. Taylor et al. Electrophoresis 2008, 29, tion with discrete bands of solution containing biological material into a detection channel [4, 9 11]. In this sample injection, the discrete bands are commonly created by focusing a sample stream and then instantaneously switching the flow path to create a square plug of sample flow in the detection channel. The separation resolution is directly dependent on the initially generated sample plug size and shape, which in turn relies on the focused sample stream width. Therefore, the evaluation of the focused sample stream width is critical to the design of microfluidic chips for DNA and protein separations. Figure 1 shows the plan view of an illustrative focusing element of cross-channel design in a microfluidic chip. The sample fluid flows from the supply reservoir to the outlet reservoir. At the focusing junction, sheath fluid streams are introduced on both sides of the sample fluid stream. The sample stream and sheath streams can be driven by applied pressure forces to create hydrodynamic focusing [7, 12, 13], or by applied electrical forces to create electrokinetic focusing [3, 8, 14]. For both applied forces, the result of introducing the sheath stream is to cause the sample stream width to decrease through the focusing junction and into the outlet channel. This narrowing of the sample stream width is illustrated in Fig. 1 by the position of the streamlines that divide the sample and sheath streams. As the flow in the outlet channel reaches a fully developed state, the width of the focused sample stream becomes constant (assuming negligible mixing between sample and sheath fluids) and, in a 2-D channel flow, is fixed by the relative volume flow rates of the sample and sheath streams [15]. The width of the focused sample stream determines the dispensing characteristics of cells for cytometry and the width and shape of the injected sample band for an injection procedure. While focused-stream width is directly dependent upon the relative volume flow rates, it is a more complex function of the applied potential forces (either pressure or electrical) used to create and control the flows in the focusing element. Yang et al. [15] presented a focusing model, which used the Helmholtz-Smoluchowski equation to relate the bulk velocities to applied electrical potentials, f, in their analysis of electrokinetic focusing. For a symmetrical geometry grounded at the outlet reservoir, f 4 ¼ 0, they approximate the mass conservation balance to show d W 4 ¼ 1 þ 2W 2 W ðl 4 þ L 1 Þðf 2 Þ L f L 2 þ 2L 4 1 f f 1 where d is the focused-stream width, W 1, W 2, and W 4 are the widths of the inlet channel, sheath flow channels, and the outlet channel, L 1 ; L 2 ; and L 4 are the lengths of the inlet channel, sheath flow channels, and the outlet channel, f 1 is the applied potential at the sample reservoir, and f 2 is the applied potential at the two sheath fluid reservoirs. Equation (1) provides a relatively compact function for the focused stream width in terms of the applied electrical potential functions and microfluidic chip geometry. To verify this function and similar functions based on approximate analyses, there is a strong need for a method of experimentally determining the width of the focused stream. The specific goal of the present work is to present and verify an experimental method for determining the width of the focused-sample stream. Fluorescence-based visualization techniques, where fluorescent dyes are injected into a stream to create an image of varying intensity (related to dye concentration) when viewed in a microscope, are commonly used in experimental studies of transport in microfluidic chips. These techniques have been widely used to observe sample focusing [3, 6, 16 18]. Figure 2 shows a typical image of flow in a focusing element obtained with the fluorescence-based technique. Upstream of the focus junction, the sample stream contains fluorescent dye and appears white in the image. Prior to entering the focus junction, the two sheath streams have no dye and appear black in the image. Downstream of the focus junction, the distinction between the sample stream and the two sheath streams is not clear. The grey color in the outlet channel indicates that the dye has spread from the sample stream into the sheath streams. Figure 3, which shows the variation of image intensity across the outlet channel width, W 4, at the downstream section A-A indicated in Fig. 2, confirms the spreading of dye from the sample stream to the sheath streams. (1) Figure 1. Schematic of the chip layout for sample focusing where the sample dye solution is located in the reservoir at the far left. Figure 2. Image of fluorescent intensity of typical sample focusing in a cross-linked microchannel.

3 Electrophoresis 2008, 29, Microfluidics and Miniaturization 2955 The spreading of the fluorescence dye is caused by two possible physical processes: molecular diffusion and electrophoretic migration. In molecular diffusion, random molecular motions of the sample and sheath fluids cause dye transport. This transport is especially significant in directions normal to the flow streamlines. The extent to which dye is transported from the centerline of the outlet channel to the walls of the outlet channel is determined by the diffusivity of the dye in the sample and sheath fluids, D. Molecular diffusion occurs in both hydrodynamic and electrokinetic focusing. In electrokinetic focusing, electrophoretic migration of a charged dye also occurs. Typically, a cationic dye will migrate towards the outlet channel centerline and an anionic dye will migrate towards the outlet channel wall. The extent of this migration depends upon the electrokinetic properties of the dye and the sample and sheath fluids. The challenge of using fluorescence-based visualization to estimate the focusing of a sample stream is the estimation of the position of the dividing streamlines from the cross-stream variation of intensity (or dye concentration), as shown in Fig. 3. Presently, the most common method for this estimation is the FWHM intensity method proposed by Jacobson and Ramsey [3]. The FWHM is the width between the two points on the intensity profile where the intensity is half of the maximum (or centerline) intensity. Figure 3 shows the FWHM for this illustrative profile. In their experimental study, Jacobson and Ramsey demonstrate that the FWHM depends upon many factors including dye diffusion and the electrokinetic properties of the dye and sample and sheath fluids. However, the width of the sample stream, as measured by the width between the dividing streamlines, should be independent of these properties. Therefore, while the FWHM is a good measure of the focusing of the fluorescence dye, it is not a measure of the width of the sample stream or an indicator of the position of the dividing streamline. The premise of the present work is that if the shape of the cross-stream intensity profile is assumed to be a Gaussian distribution then analysis based on conservation principles can be used to determine the width of the sample stream. Details of this analysis are presented in the Section 2. It will be shown that the width of the focused sample stream can be estimated from the channel widths and the variance and maximum intensity of the experimentally observed intensity profile. In Section 3, it will be demonstrated that experimentally observed profiles can be reasonably approximated by Gaussian distributions with equivalent variance and maximum values and that the estimated sample stream width is independent of molecular diffusion and electrokinetic migration properties. Comparisons will also be made to FWHM estimates. Analysis of advective-diffusive transport based on Gaussian profile distributions is well-established practice in air quality and atmospheric dispersion modeling [19]. The concentration plume created by a point source in a uniform velocity field will have a Gaussian profile distribution in the cross-stream direction if the concentration diffusivity is uniform, the transport domain is infinite in extent, and if advection transport dominates diffusive transport in the streamwise direction. It has been shown that the Gaussian profile can be a good approximation in non-uniform velocity fields and in finite domains [20, 21]. The electrokinetic flow in an outlet channel downstream of the focusing junction will have a uniform velocity field. In the region close to the focusing junction, the effect of the channel walls on the diffusion of fluorescence dye will be minimal. A measure of the ratio of advective transport to diffusive transport in the streamwise direction is the Peclet number, Pe d ¼ ul D, where u is the bulk fluid velocity, l is the characteristic length scale of the channel (usually the channel width), and D is diffusivity of the dye in the sample and sheath fluids. Advective transport will dominate diffusive transport when the Peclet number is greater than one. Typical Peclet number values for the flow in the outlet channel of focusing elements are in the range 1005Pe d Therefore, it is reasonable to base the analysis of transport of dye in a focusing element upon the Gaussian distribution approximation for the cross-stream intensity profile. 2 Model formulation Figure 3. Experimentally recorded fluorescent intensity profile at the A-A plane in Fig. 2 and a typical Gaussian curve. Figure 4 shows the details of the region near a symmetric focusing junction of a microchip. The width of the sample input channel is W 1 and the width of the outlet channel is W 4. The two sheath flow channels have width W 2 and are perpendicular to the sample and outlet channels. The height of all channels, h, is typically two- to ten-times smaller than the channel widths. All channels are very long compared to their widths.

4 2956 J. Taylor et al. Electrophoresis 2008, 29, The sample and sheath fluids are the same Newtonian fluid. At section 1 upstream of the focusing junction the fluorescence dye concentration, C 1, is very low (i.e. 100 mm). At sections 2 and 3 in the sheath flow channels, the dye concentration is negligible. Since the dye concentration is low, the bulk fluid-transport properties such as density, R, and electrical conductivity, l, are independent of dye concentration and can be assumed constant. Applied electrical potentials in the fluid reservoirs create a steady, low Reynolds number flow in the channels. The surface area to the volume ratio in all channels is large so there is sufficient heat transfer to ensure that Joule heating effects are negligible. At cross-sections well away from the junction, like sections 1 through 4 shown in Fig. 4, the velocity fields will be fully developed and uniform or plug-like [22, 23] under these conditions. At section 1 upstream of the junction, the sample stream flow has a velocity V 1. As indicated in Fig. 4 two dividing streamlines form at the upstream corner of the junction and separate the sample and sheath flows (assuming that mixing of the sample and sheath fluids is negligible, [15]). By section 4, the focused-sample flow width (or distance between the dividing streamlines) becomes constant, d, and the velocities of the sample and sheath fluids are the same, V 4. As demonstrated previously, diffusion of dye from the sample stream to the sheath streams causes a dye concentration profile at section 4, C 4 (y), where y is distance from the outlet channel centerline. A typical concentration profile in the outlet channel is shown in Fig. 3. The analysis that follows establishes a relationship between the focused sample stream width, d, the channel widths, W 1 and W 4, and the measured concentration profile C 4 (y). For the idealizations given above, the steady mass balance of the sample fluid flow between sections 1 and 4 implies that the sample-fluid mass flow rates at sections 1 and 4 must be equal, phw 1 V 1 = phdv 4,or V 1 V 4 ¼ d W 1 (2) In other words, the focusing of the sample stream is directly proportional to the flow speed up, which occurs between sections 1 and 4 due to the addition ofthe sheath stream flows [15]. Similarly, the steady conservation mass balance of the fluorescence dye between sections 1 and 4 implies that the flow rate of dye at these two sections must be equal, hw 1 V 1 C 1 ¼ ZW 4=2 W 4=2 hv 4 C 4 ðþdy y (3) where C 1 and C 4 are the concentration of the dye with a unit of mm. Equation (3), when combined with Eq. (2), reduces to ZW 4=2 d ¼ 1 C 4 ðyþdy (4) W 1 W 1 C 1 W 4=2 Equation (4), the fundamental result of the present work, shows that the width of the focused sample stream can be inferred from an integration of the measured dye profile at section 4, C 4 (y). Observations of dye concentration profiles or intensity distributions observed here and thoset of other workers, Jacobson and Ramsey [3], indicate that the concentration profile is well approximated by a Gaussian distribution C 4 ðyþ C 0 y2 4e 2s 2 (5) where C 0 4 is the centerline (maximum) concentration at section 4 and s is the profile variance (or SD). Substituting this approximation into Eq. (4) gives d p ¼ ffiffiffiffiffi C 0 4 s W 2p erf p 4 W 1 C 1 W 1 2 ffiffiffi (6) 2 s where the error function is defined as p rffiffiffi Z ffiffi 2 x 2 erf ðxþ ¼ e t2 2 dt. Equation (6) is the practical result of p 0 the present work. It provides a simple relationship between the width of the focused-sample stream, d, and two parameters that can be easily measured from the dye intensity profile, C 0 4 and s. The Gaussian distribution profile can also be used to estimate the FWHM based on determining the distance from the centerline at section 4 to the point where the concentration is half of the centerline concentration, C Manipulation of Eq. (4) leads to FWHM p ¼ ffiffiffiffiffiffiffiffiffiffi s 2 ln 2 (7) W 1 W 1 Equation (7) demonstrates that the FWHM depends directly upon the variance of the intensity profile. Figure 4. Illustration of sample focusing in a cross-linked microchannel and fully developed flow field downstream of the junction. 3 Results and discussion The primary finding of the present work, Eq. (6) is based upon several assumptions including the approximation of

5 Electrophoresis 2008, 29, Microfluidics and Miniaturization 2957 dye concentration profile by a Gaussian distribution and the approximation of transport processes in the dye conservation balance. Therefore, there is a need to verify that Eq. (6) provides an accurate and reliable estimate of the focused-stream width. In this Section, it is demonstrated that the Gaussian distribution is a reasonable approximation of the dye concentration profile, that Eq. (6) provides an accurate estimate of the focuse-stream width for numerically simulated flow in a microfluidic chip regardless of the dye transport properties, and that the model of Yang et al. [15], Eq. (1), agrees well with experimentally measured focused-stream widths based on Eq. (6). 3.1 Gaussian profile verification To validate the approximation of the dye concentration field with the Gaussian distribution, experiments were conducted to generate dye intensity profiles and corresponding concentration profiles. The experimental instrumentation consisted of a fluorescence microscope and CCD imaging system to record and analyze images of a fluorescent dye sample undergoing focusing in a microfluidic chip with a single focusing element, as shown in Fig. 1. The microfluidic chip used in experiments was fabricated by a standard soft-lithography process; the polydimethylsiloxane (PDMS) replica was bonded to a glass substrate. The symmetric chip had geometric dimensions of W 1 =50mm for sample channel width, W 2 = 100 mm for sheath channel widths, and L 1 = 0.01 m, L 2 = m, and L 4 = m for inlet, sheath flow, and outlet channel lengths. The depth of all channels was 20 mm. The bulk fluid for the sample and sheath streams was a sodium bicarbonate buffer solution. For the baseline set of experiments, the buffer solution concentration was 50 mm. Fluorescein dye was added to the sample stream with a concentration of 100 mm. No fluorescein dye was added to either of the sheath streams. An external high-voltage power sequencer (HVS 448, Labsmith, Livermore, CA) was used to control the applied potentials at the fluid reservoirs. The ratio of potentials applied at the sheath reservoirs to the sampler reservoir was varied to achieve a range of focusing widths in the outlet channel. Full details of the experimental methods including fabrication processes, flow control techniques, imaging techniques, and buffer and dye solution protocols can be found in Taylor [24]. Care was taken to ensure that profiles were measured at a section representative of the fully developed flow of the focused-sample stream and where the concentration profile was most apt to fit the Gaussian profile. The measurement section needs to be far enough downstream of the focusing junction that the sample and sheath stream flows are fully developed. In addition, the measurement section has to be sufficiently close to the focusing junction that the dye diffusion is not significantly influenced by the presence of the channel walls. In the present study, it was found that reliable and consistent data could be obtained if the measurement sections were located a short distance, mm, downstream of the location of the minimum width of the fluorescence dye profile (Fig. 2). Figure 2 shows the fluorescence dye image for a case where sheath to sample reservoir potential ratio was f 2 =f 1 ¼ 1. The cross-stream intensity profile measured at a section 150 mm downstream of the minimum dye profile width (located approximately 100 mm downstream from the start of the focusing junction) and the corresponding best-fit Gaussian profile are shown in Fig. 3. It is clear from this typical case that the Gaussian profile is an excellent match to the actual observed intensity profile. 3.2 Insensitivity to dye transport properties As mentioned in the introduction, one of the challenges of verifying estimates for the focused-stream width like Eq. (1) is that it is difficult to directly observe and measure the focusedstream width. Therefore, as part of the verification of the proposed estimate of stream width based on an observed dye concentration profile, Eq. (6), numerical simulations were used to generate realistic electrical potential, fluid velocity and dye concentration fields that could be directly observed. Steady 2-D simulations were obtained for a samplefocusing element with geometry as shown in Fig. 1. All dimensions were identical to those of the chip used for the experimental study except the lengths of the sample input stream, L 1, and sheath flow streams, L 2, which were both reduced to 0.4 mm. These inlet lengths could be reduced because the flow velocities in these channels are high enough that there is negligible upstream propagation or influence of dye. A bulk fluid with constant density, (R ¼ 998 kg=m 3 ), electrical conductivity (l = 0.2 S/m), and viscosity (m ¼ 0:001 kg=ms) was assumed. The nominal diffusion coefficient for the dye in the bulk fluid was D i = m 2 /s. The dye electrophoretic mobility and electrical charge were m ep = m 2 /Vs and z = 22 respectively. At the walls of channels a zeta potential, z, of 280 mv was set to represent the experimentally determined values for PDMS microchips [25]. Applied electrical potentials of f 1 =f 2 =f 3 =f 4 ¼ 1000 V=1000 V=1000 V=0 V were set at the sample, sheath flow, and outlet reservoirs. The channels walls were set as electrically insulated surfaces. The pressures at all the reservoirs were set to zero so that purely EOF was simulated. The elcectroosmotically generated slip velocity at the channel walls was based on the specified wall zeta potential. The dye concentration at the inlet sample reservoir was specified, C 1 ¼ 100 mm, and at the sheath flow reservoirs was set to zero. Negligible transport of dye was assumed at the channel walls. The numerical simulations were obtained with a fully second order finite element solver, COMSOL Multiphysics, on structured meshes. The highest allowable grid density was adopted in this study. The cross-stream dye concentra-

6 2958 J. Taylor et al. Electrophoresis 2008, 29, tion profile observed at a section downstream of the minimum dye profile width and the focused stream width as based upon the position of the dividing streamline were directly obtained from the simulated dye concentration and velocity fields. To test the insensitivity of the proposed method for estimating focused-stream width to dye transport properties, the diffusion coefficient of the dye in the bulk fluid was varied about the nominal value, m 2 /s,d i, m 2 /s. Figure 5 shows the variation with dye diffusion coefficient of the focused-stream width as directly observed from the simulations, as predicted by the focusing model of Yang et al. [15], and as predicted from the dye concentration profile, Eq. (6). As expected, the focused-stream width observed from the simulated velocity fields and as predicted by the model of Yang et al. do not depend upon value of the dye diffusion coefficient. The predictions of the model of Yang et al. are within 1% of the observed numerical simulations. The predictions of the focused-stream width based on the dye concentration profiles do not vary with dyediffusion coefficient and are nearly identical to the observed numerical simulations. Figure 5 also shows the FWHM concentration as the diffusion coefficient is varied. Notice that the FWHM values vary with diffusion coefficient and are significantly larger than the observed simulated focusedstream widths. This verification of the proposed method for predicting focused-stream width based on observed dye intensity or concentration profiles shows that the method is accurate and is independent of the dye transport properties when applied to numerically simulated flow and concentration fields. While the validation based on the numerical simulations is positive, the numerical simulated fields are based upon several approximations. As a final verification of the proposed method, it was applied to the estimation of observed focusedstream width in the focusing element described above. In these experiments, three buffer concentrations of the bulk fluid, 1, 10, and 50 mm, were used without varying other properties. The electroosmotic mobility of the buffer varies with the buffer concentration. Figure 6 shows the variation with focusing potential ratio, f 1 =f 2, of the focused-stream width as estimated from the intensity profile, Eq. (6), and as predicted by the model of Yang et al., Eq. (1), for all three buffer concentrations. Figure 6 also shows the variation of the FWHM with focusing potential ratio for buffer concentrations of 1 and 50 mm. The estimates based on the intensity profile and the predictions of Yang et al. agree very well. There is a weak variation in the estimated stream widths based on the intensity profiles with the buffer solution concentration. However, this variation is well within the experimental uncertainty of the measurements [24]. Figure 6 also shows that the FWHM is consistently larger than the focused-stream width, does not follow the same variation with focusing potential ratio as the focused-stream width, and is dependent on the dye transport properties. The developed method is independent of the dye transport properties, which is particularly important for the evaluation of the performance of a microfluidic focusing element. Figure 6. Comparison of the determined sample stream width between the method reported here, focusing model reported by Yang et al. [15], numerical simulation and FWHM method for different applied potential ratios. Figure 5. Comparison of the determined sample stream width between the method reported here, focusing model reported by Yang et al. [15], numerical simulation and FWHM method for different diffusion coefficients. 3.3 Further experimental verification 4 Concluding remarks In this paper, we have presented an experimental method for determining the focused-sample stream width in cross-linked microchannels. This method is based on the mass conservation principle and the assumption that the sample concentration in a region downstream the junction matches the Gaussian distribution. This method has been validated both numerically and experimentally. It has been found that this method is insensitive to the sample properties and can provide more accurate prediction of the focused-sample width as compared to the most commonly used FWHM method.

7 Electrophoresis 2008, 29, Microfluidics and Miniaturization 2959 The authors gratefully acknowledge the support of a Research Grant of the Natural Sciences and Engineering Research Council (NSERC) of Canada to C. L. Ren and G. D. Stubley and Ontraio Graduate Scholarship of Science and Technology to Jay Taylor. The authors have declared no conflict of interest. 5 References [1] Clayton, J., Nat. Methods 2005, 2, [2] Dittrich, P. S., Manz, A., Nature 2006, 5, [3] Jacobson, S. C., Ramsey, J. M., Anal. Chem. 1997, 69, [4] Harrison, D. J., Manz, A., Fan, Z., Ludi, H., Widmer, H. M., Anal. Chem. 1992, 64, [5] Schrum, D. P., Culbertson, C. T., Jacobson, S. C., Ramsey, J. M., Anal. Chem. 1999, 71, [6] Ren, L., Sinton, D., Li, D., J. Micromech. Microeng. 2003, 13, [7] Lee, G.-B., Hung, C-I., Ke, B.-J., Huang, G.-R. et al., J. Fluids Eng. 2001, 123, [8] Xuan, X., Li, D., Electrophoresis 2005, 26, [9] Xu, Z., Nakamura, Y., Hirokawa, T., Electrophoresis 2005, 26, [10] Zhuang, G-S., Li, G., Jin, Q-H., Zhao, J-L., Yang, M-S., Electrophoresis 2006, 27, [11] Huang, F-C., Liao, C.-S., Lee, G.-B., Electrophoresis 2006, 27, [12] Chung, S., Park, S. J., Kim, J. K., Chung, C. et al., Microsyst. Tech. 2003, 9, [13] Yang, S-Y., Hsuing, S.-K., Hung, Y.-C., Chang, C.-M. et al., Meas. Sci. Technol. 2006, 17, [14] McClain, M. A., Culbertson, C. T., Jacobson, S. C., Ramsey, J. M., Anal. Chem. 2001, 73, [15] Yang, R.-J., Chang, C.-C., Huang, S-B., Lee, G.-B., J. Micromech. Microeng. 2005, 15, [16] Lin, J.-Y., Fu, L.-M., Yang, R.-J., J. Micromech. Microeng. 2002, 12, [17] Fu, L.-M., Yang, R.-J., Lin, C.-H., Pan, Y.-J., Lee, G-B., Anal. Chim. Acta 2004, 507, [18] Ermakov, S. V., Jacobson, S. C., Ramsey, J. M., Anal. Chem. 1998, 70, [19] Arya, S. P., Air Pollution Meteorology and Dispersion, Oxford University Press, New York, NY [20] Macdonald, R. W., Griffiths, R. F., Cheah, S. C., Atmosph. Environ. 1997, 31, [21] Macdonald, R. W., Coulson, B. J., Slawson, P. R., Environ. Monit. Assess. 2000, 65, [22] Tabeling, P., Introduction to Microfluidics, Oxford University Press, New York, NY [23] Li, D., Electrokinetics in Microfluidics, Elsevier Academic, San Diego, CA [24] Taylor, J., The Design and Evaluation of a Microfluidic Cell Sorting Chip, MASc. Thesis, University of Waterloo [25] Sinton, D., Flow Visualization in Microchannels, PhD Thesis, University of Toronto 2003.