Electrokinetic Fluid Control in Two-Dimensional Planar Microfluidic

Aug 25, 2007 - transfer region, control channels were placed on both sides of the first-dimension channel, and the electroki- netic flow from these co...
0 downloads 0 Views 452KB Size
Anal. Chem. 2007, 79, 7485-7491

Electrokinetic Fluid Control in Two-Dimensional Planar Microfluidic Devices Margaret A. Lerch and Stephen C. Jacobson*

Department of Chemistry, Indiana University, Bloomington, Indiana 47405-7102

We present microfluidic device designs with a twodimensional planar format and methods to facilitate efficient sample transport along both dimensions. The basic device design consisted of a single channel for the first dimension which orthogonally intersected a highaspect ratio second-dimension channel. To minimize dispersion of sample moving into and through the sample transfer region, control channels were placed on both sides of the first-dimension channel, and the electrokinetic flow from these control channels was used to confine the sample stream. We used SIMION and COMSOL simulations of the electric fields and fluid flow to guide device design. First, devices with one, two, and four control channels were fabricated and tested, and four control channels provided the most effective sample confinement. The designs were evaluated by measuring the sample stream widths and concentration to width ratios as a function of the electric field strength ratio in the control channels and first-dimension (1D) channel (EC/E1D). Next, both a single open channel and an array of parallel channels were tested for the second dimension, and improved performance was observed for the parallel channel design, with stream widths as narrow as 120 µm. The ease with which fluids could be introduced into both the first and second dimensions was also illustrated. Sample plugs injected into the planar region were confined as effectively as sample streams and were easily routed into the planar region by reconfiguring the applied potentials. Often, complex samples cannot be adequately separated by a single separation technique. For a one-dimensional separation, the peak capacity or number of components that can be separated by high-efficiency techniques is 100-250. Separation techniques, however, can be combined to form multidimensional separation systems which can be used to improve the peak capacity. If the two separation methods are orthogonal, the peak capacity for a two-dimensional system is the product of the peak capacities of the two separation methods,1-3 and therefore, peak capacities on the order of 5000-10 000 become possible. Recently, several * Corresponding author. E-mail: [email protected]. (1) Giddings, J. C. Anal. Chem. 1984, 56, 1258A-1264A. (2) Giddings, J. C. J. High Resolut. Chromatogr. Chromatogr. Commun. 1987, 10, 319-323. (3) Giddings, J. C. Unified Separation Science; John Wiley and Sons: New York, 1991. 10.1021/ac071003y CCC: $37.00 Published on Web 08/25/2007

© 2007 American Chemical Society

groups have integrated two-dimensional separations on microfluidic devices. Separations that have been serially coupled include micellar electrokinetic chromatography (MEKC) and capillary electrophoresis (CE),4,5 open channel electrochromatography and CE,6 isoelectric focusing (IEF) and CE,7,8 isotachophoresis and CE,9 and capillary gel electrophoresis (CGE) and MEKC.10 Typically, a constant volume of the first-dimension effluent is sampled into the second dimension at regular intervals. Because of the sampling requirement, the second dimension must run approximately 10 times faster than the first dimension, imposing restrictions on the separation techniques that can be suitably coupled. Alternatives to serially coupled designs are planar designs where the first dimension propagates along a line in the separation space after which the second dimension is run orthogonally to the first dimension. This format mimics a traditional planar gel and maintains independent time frames for the two separations. To date, devices consist of a single channel for the first dimension and a series of parallel channels for the second dimension. IEF and CGE have been coupled for the analysis of proteins and peptides,11-13 and DNA mutations have been analyzed with CGE and temperature gradient gel electrophoresis.14 One of the early designs was fabricated in multiple poly(dimethylsiloxane) (PDMS) layers11 where the channels for the second-dimension separation were not physically put in place until the first-dimension separation was complete. Precise manual alignment of the layers partway through analysis increases the total analysis time and leads to poor reproducibility. Physical barriers such as air gate capillaries have also been used to physically and electrically isolate the two dimensions.13 Other groups have offset the interconnecting (4) Rocklin, R. D.; Ramsey, R. S.; Ramsey, J. M. Anal. Chem. 2000, 72, 52445249. (5) Ramsey, J. D.; Jacobson, S. C.; Culbertson, C. T.; Ramsey, J. M. Anal. Chem. 2003, 75, 3758-3764. (6) Gottschlich, N.; Jacobson, S. C.; Culbertson, C. T.; Ramsey, J. M. Anal. Chem. 2001, 73, 2669-2674. (7) Herr, A. E.; Molho, J. I.; Drouvalakis, K. A.; Mikkelsen, J. C.; Utz, P. J.; Santiago, J. G.; Kenny, T. W. Anal. Chem. 2003, 75, 1180-1187. (8) Wang, Y. C.; Choi, M. N.; Han, J. Y. Anal. Chem. 2004, 76, 4426-4431. (9) Kaniansky, D.; Masar, M.; Dankova, M.; Bodor, R.; Rakocyova, R. J. Chromatogr., A 2004, 1051, 33-42. (10) Shadpour, H.; Soper, S. A. Anal. Chem. 2006, 78, 3519-3527. (11) Chen, X. X.; Wu, H. K.; Mao, C. D.; Whitesides, G. M. Anal. Chem. 2002, 74, 1772-1778. (12) Li, Y.; Buch, J. S.; Rosenberger, F.; DeVoe, D. L.; Lee, C. S. Anal. Chem. 2004, 76, 742-748. (13) Tsai, S.-W.; Loughran, M.; Karube, I. J. Micromech. Microeng. 2004, 14, 1693-1699. (14) Buch, J. S.; Rosenberger, F.; Highsmith, W. E.; Kimball, C.; DeVoe, D. L.; Lee, C. S. Lab Chip 2005, 5, 392-400.

Analytical Chemistry, Vol. 79, No. 19, October 1, 2007 7485

channels to minimize the dispersion caused by electric field distortions around the channel junctions,12,14 but band broadening was still 2 times larger than anticipated. In addition, planar formats have been used for free flow electrophoresis (FFE),15-17 which is a continuous flow, one-dimensional separation. For these separations, a series of parallel, hydrodynamically driven flows is used to control sample delivery into the separation region. To create high-speed integrated devices, we are developing devices which rely on electrical, not mechanical, actuation to control fluid flow. Also, by improving transfer efficiency between dimensions, higher separation efficiencies may be achievable. Toward this end, we have been investigating how to design and operate microfluidic devices with a planar format for twodimensional separations. To be most effective, the sample should transfer from the first to the second dimension without additional band broadening. In the basic design, sample from the first dimension entered into the top of a planar region, and additional control channels were placed adjacent to the first-dimension channel to minimize dispersion of the sample stream while moving into and through the sample transfer region. Once the sample from the first dimension was in the sample transfer region, sample could be routed into the second dimension. Initially, we evaluated the number of control channels needed to adequately confine sample flow and the optimum electric field strength ratio to operate the device. Next, we considered two designs for the planar second dimension, one with a single open channel and a second having an array of parallel channels. Device design was supported with SIMION and COMSOL simulations of the electric fields and fluid flows. To evaluate flow control in these devices, the primary criteria used were the sample stream widths traversing the sample transfer region as a function of the electric field strength ratio in the control channels and first-dimension channel (EC/E1D). Second, the influence of confinement on sample dilution in the sample transfer region was evaluated. With the preferred device design and operating conditions, gated injections into the sample transfer region and planar region were examined. Additionally, the performance of devices made from PDMS and glass (PDMS/ glass) and entirely from glass (glass/glass) was compared. EXPERIMENTAL SECTION Microfluidic Devices. PDMS/glass and glass devices were fabricated using standard micromachining techniques, and fabrication details are included in the Supporting Information. The hybrid PDMS/glass devices were used for a majority of the experiments because they were easier and less expensive to fabricate, allowing a number of different designs to be evaluated. The PDMS/glass and glass/glass devices performed similarly, and results from their comparison are also discussed in the Supporting Information. Buffers and Samples. The designs with four sets of one, two, and four control channels were evaluated with a buffer containing 10 mM sodium tetraborate. A second buffer, 20 mM N-[2hydroxyethyl] piperazine-N′-[2-ethanesulfonic acid] (HEPES) and 1 mM sodium tetraborate, was used in the other experiments. A (15) Raymond, D. E.; Manz, A.; Widmer, H. M. Anal. Chem. 1994, 66, 28582865. (16) Raymond, D. E.; Manz, A.; Widmer, H. M. Anal. Chem. 1996, 68, 25152522. (17) Fonslow, B. R.; Bowser, M. T. Anal. Chem. 2005, 77, 5706-5710.

7486 Analytical Chemistry, Vol. 79, No. 19, October 1, 2007

Table 1. Channel Lengths and Applied Potentials for Sample Stream Confinement and Injection Experiments on the Parallel Channel Device applied potentials (kV) channel and reservoir

channel length (mm)a

sample stream confinementb

injection into sample transfer regionc

injection into planar regionc

sample buffer 1 waste 1 control A control B control C control D waste 2 buffer 2d waste 3d 1D

14.2 17.2 17.2 12.8 13.0 13.8 13.0 8.5 2.5 10 4

2.0-5.5

1.93 3.14 0.761 3.04 3.07 0.0 0.0 1.01

1.93 3.14 0.761 1.72

3.04 3.07 0.0 0.0 1.01

2.10 1.22 3.04 0.0

a Dimensions are for the parallel channel device shown in Figure 2b. b Sample confinement experiments are shown in Figures 4 and 6. c Injection experiment is shown in Figure 8. d Channel lengths are from the 1D channel to the end of the planar region.

10 µM solution of disodium fluorescein was prepared in both buffers as a test sample, and for the glass/glass devices, a 10 µM solution of rhodamine B was also used as a sample. Buffer components and samples were obtained from Sigma-Aldrich Co. (St. Louis, MO). Microchip Operation. All flow within the microfluidic devices was electrokinetically driven with potentials provided by two highvoltage power supplies, each with four independent output channels. The high-voltage power supplies amplified the output of an analog output board (PCI-6713, National Instruments Corp., Austin, TX) and were controlled through LabView (National Instruments Corp.). To determine the optimum conditions for sample confinement in the sample transfer region, electrodes were placed in the sample, control A-D, and waste 2 reservoirs. The fluorescent sample was transported continuously across the sample transfer region from the 1D channel to the waste 2 channel. The width of the sample stream was varied by adjusting the potential applied to the sample reservoir relative to the control reservoirs. Table 1 lists the potentials used for the sample confinement experiments in the parallel channel device. Gated injections18 were used to introduce small plugs of sample into the 1D channel and, subsequently, the sample transfer region. For the gated injections, potentials were applied to the buffer 1 and waste 1 reservoirs in addition to those reservoirs used previously. Under loading conditions, the fluorescein solution flowed from the sample channel into the waste 1 channel. The electric field strength in the buffer 1 channel was greater than in the sample channel, preventing sample from entering the 1D channel. To make an injection, the potential at the buffer 1 reservoir was lowered to match the potential at the waste 1 reservoir for 0.05, 0.1, or 0.2 s. During this time, sample migrated into the 1D channel until the potential at the buffer 1 reservoir was returned to the original loading value. While confining the sample in the vertical direction, no potentials were applied to the buffer 2 and waste 3 reservoirs. To move a sample front or plug into the planar region, the applied potentials were reconfigured so that the primary flow was (18) Jacobson, S. C.; Koutny, L. B.; Hergenro¨der, R.; Moore, A. W., Jr.; Ramsey, J. M. Anal. Chem. 1994, 66, 3472-3476.

Table 2. Parameters for COMSOL Simulations viscosity, η density, F electrical conductivity, σ ζ-potential electric field, x and y components diffusion coefficient, D concentration, c0

1 × 10-3 Pa‚s 1 × 103 kg/m3 0.1565 S/m -0.05 V Ex_emdc, Ey_emdc 4 × 10-10 m2/s 1 mol/m3

from the buffer 2 reservoir to the waste 3 reservoir. High-voltage relays (K81C245, Tyco Electronics Corp., Harrisburg, PA) were placed in-line between the high-voltage power supply and the microfluidic device to control potentials applied to the buffer 2, waste 3, and control B and D reservoirs. While the sample was being transported from the 1D channel into the sample transfer region, no potentials were applied to the buffer 2 and waste 3 reservoirs. To launch the sample into the planar region, the highvoltage relays were toggled, applying potentials to the buffer 2 and waste 3 reservoirs and removing potentials from the control B and D reservoirs. Simultaneously, potentials at the other reservoirs were adjusted as needed. To illustrate, the potentials applied to the parallel channel device for injection into the sample transfer and planar regions are listed in Table 1, and the potentials applied to the open channel device for sample confinement and injections were nominally the same as those applied to the parallel channel device. An inverted optical microscope (IX71, Olympus America, Inc., Center Valley, PA) equipped with a short arc Hg lamp and a frame transfer charge-coupled device (CCD) camera (Cascade 512B, Photometrics, Tucson, AZ) was used to take images. IPLab software (BD Biosciences Bioimaging, Rockville, MD) was used to drive the camera and to acquire images. The images were processed by subtracting a background image and normalizing to an image of the entire device filled with the sample solution. Line profiles were extracted from the images at the midpoint of the sample transfer region using OriginPro 7.5 (OriginLab Corp., Northampton, MA), and the baseline width (4σ) and concentration to width ratio were determined. Simulations. Initially, the equipotential lines in the sample transfer region were calculated using SIMION. A two-dimensional geometry file was created for these simulations, and the dimensions were proportional to the device dimensions. Electrodes were substituted in the planar region for fluidic channels, and the electrode spacing and potentials applied to the 1D, control A-D, and waste 2 electrodes were varied. The equipotential lines within the planar region were then visualized to determine the appropriate number of control channels and their spacing. More recently, electric fields and fluid flow within the planar microfluidic devices were modeled using COMSOL Multiphysics (COMSOL, Inc., Burlington, MA). Models were created in two dimensions, with the same dimensions and geometry as the fabricated microfluidic devices. To simulate electroosmotic flow, the Stokes Flow and the Conductive Media DC models were coupled, and to compare the simulations with the experimental results, the Convection and Diffusion model was also included. The parameters used in the simulations are listed in Table 2. The buffer conductivity was for a 10 mM sodium tetraborate buffer, and the value of the ζ-potential was a conservative estimate for

Figure 1. Schematic of the microfluidic device with the open channel design. The planar region was 1.55 mm wide and 12.5 mm long, and the sample transfer region is highlighted with a gray box. For the parallel channel design, the open channel was replaced with an array of 16 parallel channels. Figure 2 shows transmitted light images of the sample transfer region for both designs.

the PDMS/glass devices. The direct UMFPACK solver was used to solve the models once they had been finely meshed. The concentration profiles generated in COMSOL were exported as text files to OriginPro 7.5, and the baseline width (4σ) and concentration to width ratios were determined. RESULTS AND DISCUSSION Microfluidic Device Design. The basic microfluidic design used in this study is schematically depicted in Figure 1. All microchip designs used four sets of one, two, or four control channels positioned on each side of the 1D and waste 2 channels. In the designs with two and four control channels, each set of control channels was combined into a single channel, which in turn was connected to the control A, B, C, or D reservoir. Figure 2 shows transmitted light images of the sample transfer regions for the open and parallel channel designs with four sets of four control channels. Initial work focused on the design in Figure 1, in which the planar region consisted of a single open channel. In order to estimate an appropriate number of control channels and the channel-to-channel spacing, SIMION calculations were performed to model the equipotential lines within the planar region (results not shown). Ideally, control channels would line the entire length of the planar region; however, fabrication and operation of such a device would be impractical. Assuming similitude between the electric and velocity fields,19 a calculation of the electric field distribution within the device would provide a reasonable approximation of sample transport. We observed in these simulations that the equipotential lines became increasingly parallel as the number of control channels was raised from zero to four. In addition, the equipotential lines empirically had the best shape when the distance between the 1D channel (or waste 2 channel) and the first control channel was 80 µm center-to-center and all control channels were spaced 240 µm center-to-center. More recently, these simulations were refined using the microelectromechanical systems module of COMSOL Multiphysics. The electric field lines generated in the designs with zero, one, two, and four control channels are shown in Figure 3. Only (19) Cummings, E. B.; Griffiths, S. K.; Nilson, R. H.; Paul, P. H. Anal. Chem. 2000, 72, 2526-2532.

Analytical Chemistry, Vol. 79, No. 19, October 1, 2007

7487

Figure 2. Transmitted light images of the sample transfer region for (a) the open channel design in Figure 1 and (b) the parallel channel design. Both designs had four sets of four control channels labeled 1, 2, 3, and 4. The control channels were combined into a single channel which was used to connect to the control reservoir. Scale is the same for both images.

the field lines emanating from the 1D channel are depicted to emphasize the effect of the number of control channels. With fewer control channels, the field lines near the 1D channel extend farther in the lateral direction and into the planar region, and increasing the number of control channels from one to four reduced the curvature of these electric field lines in the sample transfer region. With the use of the dimensions obtained from the simulations, three microfluidic devices were fabricated in PDMS/glass with one, two, and four control channels, and transmitted light images of these three designs are shown in Figure S1 in the Supporting Information. In addition, the buffer 2 and waste 3 channels at the ends of the planar region branch into a number of equally spaced channels for uniform distribution of the electric field within the planar region (see Figure 1). Simulations were used to determine the number and spacing of these channels. Flow Control in the Sample Transfer Region. For these devices to function properly, lateral dispersion of a sample band entering the planar region from the 1D channel must be kept to a minimum. The three channel layouts shown in Figure 3b-d were compared to determine how many control channels were necessary to create a narrow and uniform sample flow across the sample transfer region. The potential applied to the sample reservoir was adjusted to obtain a range of confinement ratios, which were the ratios of the electric field strengths in the control 7488 Analytical Chemistry, Vol. 79, No. 19, October 1, 2007

Figure 3. Simulations of the electric field lines in the sample transfer region between the 1D channel and waste 2 (W2) channel for (a) zero, (b) one, (c) two, and (d) four control channels with EC/E1D ) 2. Only the electric field lines emanating from the 1D channel are shown.

channels relative to the 1D channel (EC/E1D). Figure 4 shows fluorescence images of sample being transported through the sample transfer region. In Figure 4b-d, the confinement ratios (EC/E1D) were 1.0, 2.9, and 4.8, respectively, and as expected, the sample stream narrowed with increasing EC/E1D. The width of the sample stream was measured at the midpoint of the sample transfer region for different values of EC/E1D and is plotted in Figure 5a for the three designs. Of the three designs, the best confinement was seen with four control channels. In addition, with increased numbers of control channels, the sample stream had a more uniform width, which can also be seen from the electric field lines in Figure 3. The sample streams in Figure 4 not only narrowed with increasing EC/E1D but also became more diluted. Therefore, we monitored the concentration to width ratio of the sample stream as a function of EC/E1D to determine appropriate operating conditions. Concentration was determined by the relative fluorescence intensity of the sample in the sample transfer region compared to the 1D channel and was normalized for the excitation profile used for the fluorescence imaging. Figure 5b shows the variation of the concentration to width ratio with EC/E1D for the same three microfluidic devices. The concentration to width ratio had a maximum at EC/E1D ) 2, and the confinement ratio at which the maximum occurred was considered the optimum. Because the four control channel design produced the highest maximum of the three designs and the most uniform flow shape within the sample transfer region, we devoted our attention to devices with four control channels.

Figure 4. (a) Transmitted light image of the sample transfer region between the 1D and waste 2 (W2) channels and fluorescence images of sample confinement in the sample transfer region with EC/E1D equal to (b) 1, (c) 2.9, and (d) 4.8. These images were obtained from the device with the parallel channel design and four control channels (Figure 2b) although only two control channels are shown in (a). Arrows depict flow direction, and the dashed line indicates the midpoint of the sample transfer region where stream widths were measured. Scale in (a) applies to all images.

We initially investigated a device with an open channel design for the second dimension but also wanted to evaluate a device in which the second dimension was composed of an array of 16 parallel channels spaced 100 µm center-to-center. A close-up of the sample transfer region can be seen in Figure 2b. In addition, the channel connecting the control channels to the reservoir was widened to minimize the potential drop from the reservoir to the sample transfer region. The parallel channel design was compared to an open channel design of similar dimensions (1.55 mm wide × 12.5 mm long planar region, Figure 2a). As seen in Figure 6, the sample stream widths in the parallel channel design were narrower than those in the open design, and higher values were obtained for the concentration to width ratio. The optimum focusing ratio was EC/E1D ) 3. Also, Figure 6 shows good agreement for results obtained from two devices with the same design. The difference in performance between the two designs can be seen in the shape of the sample streams in the sample transfer region. In Figure 7, parts a and b, sample flow in both designs is shown with EC/E1D ) 2. The edges of the sample stream in the open channel design (Figure 7a) were slightly bowed, whereas those in the parallel channel design were straighter (Figure 7b). This small but distinct difference in shape accounted for the differences in width. To better understand why the designs

Figure 5. (a) Variation of the sample stream width (4σ) with the ratio of the electric field strengths in the control and 1D channels (EC/ E1D) for one, two, and four control channels and (b) variation of the concentration to sample stream width ratio with EC/E1D ratio for one, two, and four control channels.

produced flows of different shape, the two devices were modeled with COMSOL Multiphysics. The electric field lines generated within the planar region of the two designs are shown in Figure 7, parts c and d. Differences in the field lines emerging from the control channels in the two designs were observed. Those in the open channel design (Figure 7c) curved away from the 1D channel and extended into the planar region, whereas the field lines in the parallel channel design were parallel to the sample flow path. The computer simulations closely mirrored experimental observation, and the differences in the shape of the electric field lines accounted for the improved performance in the device with the parallel channels. Comparison of Experiments and Simulations. As described for Figures 3 and 7, COMSOL Multiphysics was used to simulate the electric fields generated within several microfluidic device designs to better understand fluid flow. We were also interested in how well the COMSOL models predicted fluid flow within the microfluidic devices. To evaluate this, the confinement experiments shown in Figure 6 were simulated. Theoretical stream widths and concentrations were measured at the midpoint of the Analytical Chemistry, Vol. 79, No. 19, October 1, 2007

7489

Figure 7. Fluorescence images of sample confinement in the sample transfer region for the (a) open and (b) parallel channel designs with EC/E1D ) 2. COMSOL simulations of the electric field lines in the (c) open and (d) parallel channel designs for the same experimental conditions as in (a) and (b), respectively. Scale in (a) and (b) is the same.

Figure 6. (a) Variation of the sample stream width (4σ) with EC/ E1D ratio and (b) variation of the concentration to sample stream width ratio with EC/E1D ratio for the open (solid symbols) and parallel (open symbols) channel designs. Two devices of each design were used, and error bars were (σ for three measurements. The solid and dashed lines are the COMSOL simulations for the open and parallel channel designs, respectively. The star symbols and arrows in (a) indicate sample plug widths for the gated injection.

sample transfer region in both the open and parallel channel designs. The simulation results are plotted as lines in Figure 6 for a comparison between the experiments and simulations. For both the open and parallel channel designs, excellent agreement was observed in sample stream widths (Figure 6a) and concentration to width ratios (Figure 6b). These results gave us confidence that the COMSOL simulations could be used in a predictive manner for future device designs. Injections into the Sample Transfer and Planar Regions. Similar to the sample stream experiments described above, a sample plug must travel through the sample transfer region without being distorted. Initially, gated injections into and through the sample transfer region were evaluated for both the open and parallel channel designs. As seen in Figure S3 in the Supporting Information, sample plugs were observed to move directly across the sample transfer region, not deviating from the expected path and shape. The average plug widths at the midpoint of the sample 7490 Analytical Chemistry, Vol. 79, No. 19, October 1, 2007

transfer region were 138 ( 2 µm with EC/E1D ) 3.9 for the parallel channel design and 189 ( 7 µm with EC/E1D ) 2.5 for the open channel design, and both matched the data for the stream width measurements (see Figure 6a). Injections of a sample front into the planar region were also evaluated, and the images in Figure S4 in the Supporting Information show how efficiently sample can be transferred from the first to second dimension. To illustrate fluid manipulation in both dimensions, injections into the sample transfer region and planar region were combined. After the sample plug arrived at the midpoint of the sample transfer region, the potentials were reconfigured to send the plug horizontally into the planar region. The plug was injected into the planar region using the potentials listed in Table 1. The potentials at the buffer 1, sample, and waste 1 reservoirs were left in the loading configuration to prevent additional sample from entering the sample transfer region. Additionally, a small potential was applied to the waste 2 reservoir to pull sample back from the sample transfer region, and simultaneously, a higher potential was applied to the control A and C reservoirs to push the sample front cleanly into the planar region. The control B and D reservoirs had no potential applied. The results of the injection process are shown in Figure 8. In Figure 8, parts b and c, the plug moves vertically from the 1D channel to the midpoint of the sample transfer region, at which time the potentials were reconfigured to move the sample horizontally along the planar axis (Figure 8df).

channels were the most effective in keeping the sample stream narrow as well as maximizing the concentration to width ratio. The primary advantage of this design was the ease with which sample could be transferred from the first dimension to the second for both the open and parallel channel designs. Interestingly, the parallel channel design showed better confinement than the open channel design, due in part to the proximity of the parallel channels to the sample transfer region. Currently, the planar region is only 1.55 mm wide, but simulations suggest the entire design can be easily scaled. Once the design parameters have been optimized, these devices should be readily applicable to twodimensional separations of complex biological samples. ACKNOWLEDGMENT This work was supported in part by Indiana University and by center grants from the Indiana 21st Century Research and Technology Fund and from the National Center for Research Resources at the National Institutes of Health (RR018942). Figure 8. (a) Transmitted light image of the sample transfer region in the parallel channel design and fluorescence images of a sample plug dispensed into the sample transfer region at (b) 0.21, (c) 0.84, (d) 2.52, (e) 4.20, and (f) 5.88 s after entering the sample transfer region. Arrows indicate direction of flow. Scale in (a) applies to all images.

In conclusion, this work illustrated fluid handling capabilities in a microfluidic device with a planar format. The control channels were an integral feature of this design, minimizing the dispersion of sample traversing the sample transfer region. Four control

SUPPORTING INFORMATION AVAILABLE Details for microfluidic device fabrication, comparison of PDMS/glass and glass/glass devices, and injection into the sample transfer and planar regions. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review May 16, 2007. Accepted July 17, 2007. AC071003Y

Analytical Chemistry, Vol. 79, No. 19, October 1, 2007

7491