Microfluidic Particle Sorter Employing Flow Splitting and Recombining

which increases the selectivity and recovery rates without using a liquid that .... Gravity-Driven Microfluidic Particle Sorting Device with Hydro...
0 downloads 0 Views 439KB Size
Download PDF
Anal. Chem. 2006, 78, 1357-1362

Microfluidic Particle Sorter Employing Flow Splitting and Recombining Masumi Yamada† and Minoru Seki*,‡

Department of Chemistry and Biotechnology, School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan, and Department of Chemical Engineering, Graduate School of Engineering, Osaka Prefecture University, 1-1 Gakuen-cho, Sakai, Osaka 599-8531, Japan

This paper describes an improved microfluidic device that enables hydrodynamic particle concentration and sizedependent separation to be carried out in a continuous manner. In our previous study, a method for hydrodynamic filtration and sorting of particles was proposed using a microchannel having multiple branch points and side channels, and it was applied for continuous concentration and separation of polymer particles and cells. In the current study, the efficiency of particle sorting was dramatically improved by geometrically splitting fluid flow from a main stream and recombining. With these operations, particles with diameters larger than a specific value move toward one sidewall in the mainstream. This control of particle positions is followed by the perfect particle alignment onto the sidewall, which increases the selectivity and recovery rates without using a liquid that does not contain particles. In this study, a microchannel having one inlet and five outlets was designed and fabricated. By simply introducing particle suspension into the device, concentrations of 2.1-3.0-µm particles were increased 60-80-fold, and they were collected independently from each outlet. In addition, it was demonstrated that the measured flow rates distributed into each side channel corresponded well to the theoretical values when regarding the microchannel network as a resistive circuit. Miniaturization of analytical devices has been eagerly pursued in the past decade for the development of microTAS and lab-ona-chip technologies. These miniaturized systems are composed of microunit operations,1 including sample pretreatment, separation, detection, and further utilization. Among these operations, size-dependent separation of particles is an essential technology, since various methods deal with small particles such as cells, organelles, synthetic particles, and biomacromolecules. In addition, concentration of such samples from a diluted solution is a useful and indispensable technique in the field of clinical diagnosis, environmental assay, and other biological applications.2,3 * To whom correspondence should be addressed. E-mail: seki@ chemeng.osakafu-u.ac.jp. Tel: +81-72-254-9296. Fax: +81-72-254-9911. † The University of Tokyo. ‡ Osaka Prefecture University. (1) Tokeshi, M.; Minagawa, T.; Uchiyama, K.; Hibara, A.; Sato, K.; Hisamoto, H.; Kitamori, T. Anal. Chem. 2002, 74, 1565-1571. (2) Cabrera, C. R.; Yager, P. Electrophoresis 2001, 22, 355-362. 10.1021/ac0520083 CCC: $33.50 Published on Web 01/06/2006

© 2006 American Chemical Society

Accurately fabricated microfluidic devices have the potential to aid in manipulating particles either actively or passively, since the sizes of small particles correspond well with those of typically used microfluidic devices.4 Microdevices are therefore being applied to the separation and selection of particles, especially biological particles such as cells and organelles. Miniaturization of fluorescence-activated cell sorter (FACS) systems is one of the most developed techniques for the active sorting of specific cells from a mixture.5-15 In FACS systems, particle suspension is continuously introduced into a microchannel, followed by hydrodynamic focusing, optical sensing, and physical sorting. Up to ∼100 particles/s can be sorted using electrokinetic flow control,5-8 switch valves,9-12 electrostatic force,13,14 dielectrophoresis,15 or optical manipulation.16 These systems, however, require fluorescence-labeled particles as well as optical sensing devices and sorting schemes, making them too complex for easy use. A simple system is certainly preferable if it must be incorporated into a microfluidic system as a unit operation. In this regard, various methods for the passive separation of particles have been developed that can be carried out in a miniaturized format and that do not require mechanical moving parts or outer sensing devices. Accordingly, dielectrophoresis,3,17,18 optical manipula(3) Wong, P. K.; Chen, C. Y.; Wang, T. H.; Ho, C. M. Anal. Chem. 2004, 76, 6908-6914. (4) Andersson, H.; van den Berg, A. Sens. Actuators, B: Chem. 2003, 92, 315325. (5) Fu, A. Y.; Spence, C.; Scherer, A.; Arnold, F. H.; Quake, S. R. Nat. Biotechnol. 1999, 17, 1109-1111. (6) Kunst, B. H.; Schots, A.; Visser, A. J. W. G. Rev. Sci. Instrum. 2003, 75, 2892-2898. (7) Emmelkamp, J.; Wolbers, F.; Andersson, H.; DaCosta, R. S.; Wilson, B. C.; Vermes, I.; van den Berg, A. Electrophoresis 2004, 25, 3740-3745. (8) Dittrich, P. S.; Schwille, P. Anal. Chem. 2003, 75, 5767-5774. (9) Fu, A. Y.; Chou, H. P.; Spence, C.; Arnold, F. H.; Quake, S. R. Anal. Chem. 2002, 74, 2451-2457. (10) Kruger, J.; Singh, K.; O′Neill, A.; Jackson, C.; Morrison, A.; O’Brien, P. J. Micromech. Microeng. 2002, 12, 486-494. (11) Wolff, A.; Perch-Nielsen, I. R.; Larsen, U. D.; Friis, P.; Goranovic, G.; Poulsen, C. R.; Kutter, J. P.; Telleman, P. Lab Chip 2003, 3, 22-27. (12) Studer, V.; Jameson, R.; Pellereau, E.; Pepin, A.; Chen, Y. Microelectron. Eng. 2004, 73-74, 852-857. (13) Takahashi, K.; Hattori, A.; Suzuki, I.; Ichiki, T.; Yasuda, K. J. Nanobiotechnol. 2004, 2, 5. (14) Yao, B.; Luo, G. A.; Feng, X.; Wang, W.; Chen, L. X.; Wang, Y. M. Lab Chip 2004, 4, 603-607. (15) Lee, G. B.; Lin, C. H.; Chang, S. C. J. Micromech. Microeng. 2005, 15, 447-454. (16) Wang, M. M.; Tu, E.; Raymond, D. E.; Yang, J. M.; Zhang, H. C.; Hagen, N.; Dees, B.; Mercer, E. M.; Forster, A. H.; Kariv, I.; Marchand, P. J.; Butler, W. F. Nat. Biotechnol. 2005, 23, 83-87.

Analytical Chemistry, Vol. 78, No. 4, February 15, 2006 1357

tion,19 hydrodynamic force,20 and physical filtration21-23 have been adopted to precisely separate or concentrate particles in microfluidic systems. On the other hand, hydrodynamic chromatography,24,25 acoustic manipulation,26-28 dielectrophoresis,29,30 optical manipulation,31,32 isoelectric focusing,2,33 magnetic field control,34-36 and aqueous two-phase partitioning37 have been combined with the stable laminar flow and utilized for the selection of specific particles or cells from a complex mixture. Although the methods mentioned above are carried out either in batch or in a continuous manner, it can be said that a method for continuous and passive particle manipulation not only facilitates incorporation into an integrated microfluidic system but also enables high-throughput processes. In addition, a method for purely hydrodynamic particle separation according to size, as is the case with hydrodynamic chromatography or flow field flow fractionation,38 is versatile and practical for various applications. From these points of view, several methods have recently been reported concerning continuous particle separation in a sizedependent manner utilizing hydrodynamic profiles in microfluidic devices.39-41 These methods are capable of particle sorting without any kind of outer field control or physical actuation. However, these methods require flow rate controls for liquids with and without particles, and the particle concentrations would be diluted. We have recently demonstrated a new method for continuous concentration and sorting of particles in a microchannel having

multiple branch points and side channels.42 By simply introducing particle suspension into the microchannel, particles are hydrodynamically concentrated, aligned onto sidewalls, and then sorted according to size without the help of outer field control. This method is simple and advantageous, since neither liquid without particles nor flow rate controls are needed. However, not all of the particles are aligned onto the sidewalls, making the recovery rate insufficient. In addition, contamination of the concentrated large particles by small particles cannot be avoided; that is, the selectivity is not high. In this study, we propose an improved microfluidic particle sorter that adopts flow splitting and recombining. These flow operations trigger movement of particle positions toward one sidewall, which enables perfect alignment of target particles onto the sidewall. Therefore, both recovery rate and selectivity can be substantially improved compared with the previously reported method. The scheme employed in this improved particle sorter utilizes the fact that a particle of a specific diameter cannot be divided into pieces, while fluid flow can be split and recombined according to the microchannel geometry. In the present study, the microdevices were designed based on the concept that a microchannel network works as a resistive circuit when liquid flow is continuously introduced into it. As model particles, 1.03.0-µm polymer particles were used, and their behaviors were observed.

(17) Huang, Y.; Ewalt, K. L.; Tirado, M.; Haigis, R.; Forster, A.; Ackley, D.; Heller, M. J.; O’Connell, J. P.; Krihak, M. Anal. Chem. 2001, 73, 1549-1559. (18) Gascoyne, P.; Mahidol, C.; Ruchirawat, M.; Satayavivad, J.; Watcharasit, P.; Becker, F. F. Lab Chip 2002, 2, 70-75. (19) Chiou, P. Y.; Ohta, A. T.; Wu, M. C. Nature 2005, 436, 370-372. (20) Yoshida, M.; Thoda, K.; Gratzl, M. Anal. Chem. 2003, 75, 4686-4690. (21) Carlson, R. H.; Gabel, C. V.; Chan, S. S.; Austin, R. H.; Brody, J. P.; Winkelman, J. W. Phys. Rev. Lett. 1997, 79, 2149-2152. (22) Vankrunkelsven, S.; Clicq, D.; Pappaert, K.; Ranson, W.; Tandt, C. D.; Ottevaere, H.; Thienpont, H.; Baron, G. V.; Desmet, G. Electrophoresis 2004, 25, 1714-1722. (23) Mohamed, H.; McCurdy, L. D.; Szarowski, D. H.; Duva, S.; Turner, J. N.; Caggana, M. IEEE Trans. Nanobiosci. 2004, 3, 251-256. (24) Chmela, E.; Tijssen, R.; Blom, M. T.; Gardeniers, H. J. G. E.; van den Berg, A. Anal. Chem. 2002, 74, 3470-3475. (25) Blom, M. T.; Chmela, E.; Oosterbroek, R. E.; Tijssen, R.; van den Berg, A. Anal. Chem. 2003, 75, 6761-6768. (26) Nilsson, A.; Petersson, F.; Jonsson, H.; Laurell, T. Lab Chip 2004, 4, 131135. (27) Petersson, F.; Nilsson, A.; Holm, C.; Jonsson, H.; Laurell, T. Analyst 2004, 129, 938-943. (28) Petersson, F.; Nilsson, A.; Holm, C.; Jonsson, H.; Laurell, T. Lab Chip 2005, 5, 20-22. (29) Fiedler, S.; Shirley, S. G.; Schnelle, T.; Fuhr, G. Anal. Chem. 1998, 70, 1909-1915. (30) Durr, M.; Kentsch, J.; Muller, T.; Schnelle, T.; Stelzle, M. Electrophoresis 2003, 24, 722-731. (31) MacDonald, M. P.; Spalding, G. C.; Dholakia, K. Nature 2003, 426, 421424. (32) MacDonald, M. P.; Neale, S.; Paterson, L.; Richies, A.; Dholakia, K.; Spalding, G. C. J. Biol. Regul. Homeostatic Agents 2004, 18, 200-205. (33) Lu, H.; Gaudet, S.; Schmidt, M. A.; Jensen, K. F. Anal. Chem. 2004, 76, 5705-5712. (34) Pamme, N.; Manz, A. Anal. Chem. 2004, 76, 7250-7256. (35) Han, K. H.; Frazier, A. B. J. Appl. Phys. 2004, 96, 5797-5802. (36) Furdui, V. I.; Harrison, D. J. Lab Chip 2004, 4, 614-618. (37) Yamada, M.; Kasim, V.; Nakashima, M.; Edahiro, J.; Seki, M. Biotechnol. Bioeng. 2004, 88, 489-494. (38) Giddings, J. C. Science 1993, 260, 1456-1465. (39) Huang, L. R.; Cox, E. C.; Austin, R. H.; Sturm, J. C. Science 2004, 304, 987-990. (40) Yamada, M.; Nakashima, M.; Seki, M. Anal. Chem. 2004, 76, 5465-5471. (41) Takagi, J.; Yamada, M.; Yasuda, M.; Seki, M. Lab Chip 2005, 5, 778-784.

PRINCIPLE The basic principle for particle concentration and separation is shown in Figure 1. In the figure, a liquid containing particles is continuously introduced in a left-to-right direction into a microchannel (main channel) having a branch point and a side channel. In microscale, laminar flow is dominant; so the flow rate profile would be almost parabolic when the flow is pressure-driven. As shown in Figure 1 a, when the ratio of the volumetric flow rate distributed into the side channel (Q1) to the total flow rate (Q0), a, is sufficiently small, particles whose diameter is larger than a specific value would never go through the side channel, even when the particles are flowing near the sidewall, and even when the cross-sectional size of the side channel is larger than the particle diameters. On the other hand, when the flow rate distributed into the side channel increases as shown in Figure 1b and c, the maximum size of the particles that can go through the side channels increases; larger particles can therefore go through the side channel if they are flowing near the sidewall. The virtual width of the flow distributed into the side channel, w1, determines the maximum particle size that can go through the side channel. It can be said that this method is a kind of filtration, in which the hydrodynamic profile determines the size of filtered particles. In our previous method, by repeating the flow state shown in Figure 1a or b, a large portion of the liquid flow was withdrawn from the main channel, and the concentration of target particles was increased. In addition, particles that did not go through the side channels were aligned onto sidewalls due to the stable laminar flow profile. Then, by increasing the flow rates into side channels stepwise as with Figure 1b or c, the aligned particles were sorted according to size. However, it was impossible to align all the introduced particles onto the sidewalls; it was impossible to

1358

Analytical Chemistry, Vol. 78, No. 4, February 15, 2006

(42) Yamada, M.; Seki, M. Lab Chip 2005, 5, 1233-1239.

Figure 1. Relation between particle behavior and relative flow rate distributed into a side channel at a branch point. The relative flow rate into the side channel is (a) small, (b) medium, and (c) large, i.e., a < a′ < a′′. The virtual region of the flow distributed into the side channel is dark-colored.

Figure 2. Schematic diagram of improved particle sorting in a microfluidic device. (A-D) are the enlarged parts of the main channel showing particle behavior: (A) before flow splitting, (B) after flow splitting, (C) after flow recombining, and (D) sorting of the aligned particles.

withdraw all the liquid flow from the main channel through the side channels. As such, particles not aligned onto the sidewalls could not be sorted. The recovery rate of the target particles was therefore not sufficiently high, and contamination of the concentrated and sorted larger particles by smaller particles was unavoidable. Therefore, an improved scheme has been proposed, as shown in Figure 2. First, by repeating the flow state shown in Figure 1a, so that target particles larger in diameter than specific values (large- and medium-size particles in the figure) cannot go through the first side channels, liquid flow without the target particles is

withdrawn from the main stream. By this flow splitting, a portion of the target particles is aligned onto one sidewall (Figure 2B). Then, by recombining this split flow into the main stream at the confluent points, the positions of the target particles move toward the other sidewall of the main channel (Figure 2C). As in the case of particles that are sufficiently small to go through the first side channels (small particles in the figure), they would be dispersed randomly in the main channel. Next, by repeatedly withdrawing liquid flow from the second side channels, which are located on the opposite side of the first side channels, as with Figure 1a, particles can be aligned onto the sidewall and concentrated (Figure 2D). That is, the flow splitting and recombining enable the perfect alignment of the target particles, which was impossible in our previous particle sorter. After that, concentrated and aligned particles are collected independently according to size, by increasing the distributed flow rates into the side channels (outlet channels) as shown in Figure 1b and c (Figure 2D). If flow splitting and recombining are not performed, some portion of the target particles would go through the outlet of the main channel (outlet 1 in the figure). In this improved scheme, however, the recovery rate of the target particles collected from a specific outlet channel ideally becomes 100%. Theoretically, when the sum of the volumetric flow rates split into the first side channels and the second side channels is larger than 100% of the initial flow rate, all particles will be aligned onto the sidewall as shown in Figure 2D. That is, if 50% of the liquid flow is split from the main stream and recombined through the first side channels, it is required that at least 50% of the flow be withdrawn from the second side channels. EXPERIMENTAL SECTION Microdevice Fabrication and Design. Microdevices were fabricated by irreversibly bonding a flat glass slide and a PDMS plate with a microchannel structure, as described previously.41,42 Figures 3 and S1 (Supporting Information) show the schematic diagram of the microchannel and a diagram of a resistive circuit that is an analogue of the microchannel structure. The device has one inlet and five outlets. The channel depth is almost uniform, ∼6 µm. The main channel connects inlet and outlet 1 and consists of several segments (15, 50, or 500 µm in width). The 50 first side channels branch off from the main channel, and then joined into the main channel at the confluent points downstream from the first branch points. The 80 second side channels and 3 outlet channels also branch off, and they are all connected to each outlet. The width of the first side channels is 5 µm, while the second side channels consist of narrow and broad channel segments 5 and 20 µm in width, respectively. By combining these channel segments, the lengths of the second side channels can be made uniform, while keeping the resistances at various values. The lengths of the narrow segments are shown in Table S1 (Supporting Information). It is well known that a microchannel acts as a resistive circuit when an incompressible Newtonian fluid is continuously introduced into the channel.41-45 The microdevice was therefore (43) Constantinescu, V. N. Laminar Viscous Flow; Springer-Verlag: New York, 1995; pp 109-137. (44) Knight, J. B.; Vishwanath, A.; Brody, J. P.; Austin, R. H. Phys. Rev. Lett. 1998, 80, 3863-3866. (45) Jeon, N. L.; Dertinger, S. K. W.; Chiu, D. T.; Choi, I. S.; Stroock, A. D.; Whitesides, G. M. Langmuir 2000, 16, 8311-8316.

Analytical Chemistry, Vol. 78, No. 4, February 15, 2006

1359

Figure 3. Schematic diagram of the microchannel design. (a) Whole microchannel, (b) an enlarged part, and (c) a diagram of a resistive circuit corresponding to the microchannel network. In (c), the broad line represents the main channel.

designed according to the concept that the volumetric flow rate Q, applied pressure P, and hydrodynamic resistance R are analogues of I, V, and R in Ohm’s law, respectively. In this study, the following equation was used to estimate the hydrodynamic resistance R of each segment of the microchannel:43

R∝

[

L l13l2

1-

192l1

∞

∑

π5l2 n)1,3,5

]

tanh(nπl2/2l1) n5

-1

(1)

where l1 and l2 are either the width or depth of the microchannel; however, l1 is the larger of these two values. The ratio of the distributed flow rate into a side channel at a branch point to the total flow rate introduced into the branch point, a, was calculated for each branch point (shown in Tables S1 and S2, Supporting Information), when regarding the microchannel structure as a resistive circuit as shown in Figure 3c. The hydrodynamic resistances of the branching and confluent points of the main and the side channels were neglected.44 Then the virtual widths w1 were calculated by assuming that the flow profile is parabolic in the width direction, and neglecting the flow rate distribution in the depth direction, although the actual flow rate profile in such a rectangular microchannel would not be perfectly parabolic. As a result, it was suggested that w1 for the first and second side channels was 0.63-0.89 and 1.02-1.03 µm, respectively, while those for outlet channels 2-4 were 2.55, 2.07, and 1.55 µm, respectively. Approximately 34.2% of the introduced liquid was expected to be split from the main channel 1360 Analytical Chemistry, Vol. 78, No. 4, February 15, 2006

Figure 4. Particle behavior at the confluent points of the main channel and the first side channels: (a) 1.0- and (b) 3.0-µm particles. The width of the first side channels is 5 µm.

through the first side channels. Also, ∼66.4% of the main stream would be removed from the second side channels. Since the sum of these values was larger than 100%, it was expected that almost all the introduced particles would be aligned on one sidewall, as shown in Figure 2D, if they do not go through the first and second side channels. Operation. Fluorescing polymer microspheres 1.0, 2.1, or 3.0 µm in diameter (R0100, B0200, R0300; Duke Scientific Corp.) were used as model particles. The coefficients of variation of the particle diameters were less than 5%. These particles were suspended in 0.5% (w/v) Tween 80 aqueous solution. The initial concentrations of the 1.0-, 2.1-, and 3.0-µm particles were 1.8 × 104, 4.7 × 103, and 6.8 × 103 per 1 µL of solution, respectively. The solution containing particles was introduced into the microchannel using a syringe pump; the 1.0- and 2.1-µm particles were mixed, while the 3.0-µm particles were used independently. The flow rate was constant for all experiments, 0.5 µL/min. A fluorescent microscope (IX70, Olympus Optical Co. Ltd.) and an ICCD camera (C240089, Hamamatsu Photonics K. K.) were used for the observation of particle behavior. Moving images were captured at each outlet, and the numbers of particles were counted for 10-120 s. For the flow rate measurement, silicone tubes with an inner diameter of 0.5 mm were inserted into each outlet, and the lengths of the collected liquid were measured. Experiments were carried out at least in triplicate for each condition. RESULTS AND DISCUSSION We first examined whether particle positions would move toward one sidewall by flow splitting and recombining. Figure 4 shows the photographs of particle behavior at the confluent points

Figure 6. Distribution ratios of particles into each outlet.

Figure 5. Particle behavior at the branch points for each outlet: (a) 2.1- and (b) 3.0-µm particles. The width of the three outlet channels is 15 µm, while that of the second side channels in this region is 5 µm.

of the first side channels and the main channel. As can be seen, a portion of the 1.0-µm particles flowed into the first side channels and then flowed out into the main channel (Figure 4a). On the other hand, the 2.1- and 3.0-µm particles did not flow into the first side channels (Figure 4b), although the cross-sectional size of the side channels (5 × 6 µm) was larger than the particle diameters. It was therefore demonstrated that the particles with diameter larger than ∼2 µm moved toward one sidewall in the main channel after the split flows were recombined in the main channel. The theoretical value of w1 for the first side channels was 0.63-0.89 µm; thus, it was suggested that particles with radii larger than w1 never go through the side channels. In a rectangular microchannel, possibly this virtual width is not uniform in the depth direction; however, the obtained results indicate that the influence of this ununiformity is negligible. Next, particle alignment onto one sidewall and size-dependent sorting were observed. The behaviors of the 2.1- and 3.0-µm particles at the branch points for each outlet are shown in Figure 5. As can be seen, after passing through the 80 branch points for the second side channels, 2.1- and 3.0-µm particles were aligned onto one sidewall in a single-file fashion. Then 2.1-µm particles primarily passed through outlet 4, while 3.0-µm particles primarily passed through outlet 3. The distribution ratios of particles at each outlet are shown in Figure 6. It was confirmed that particles larger in size than a specific value could pass through a specific outlet,

especially in the case of the 3.0-µm particles; ∼90% of the 3.0-µm particles went through outlet 3, while ∼67% of the 2.1-µm particles went through outlet 4. In this microfluidic particle sorter, the limitation size of the concentrated and separated particle lies between 1.0 and 2.1 µm; that is, particles smaller than 1.0 µm can go through the second side channels as well as the first side channels, and as such, they are not concentrated and aligned onto the sidewall. It can be said that the theoretical value of w1 for the second side channels was 1.02-1.03 µm, which corresponded well with the experimental results. However, ∼5% of the 2.1-µm particles was collected from outlet 5, although the number was small. This may be caused by the relatively large w1 for the second side channels; if this value is slightly decreased, the number can be decreased. This virtual width can be freely changed, by altering the microchannel geometries; for example, if it is necessary to concentrate and sort 1.0-µm particles, microdevices should be so designed that w1 becomes smaller than ∼0.5 µm. Although the recovery rates for 2.1- and 3.0-µm particles were high, there were particles that went through outlets 1 and 2. This may be due to the inherent distribution of the particle size, or other factors such as particle collision and particles’ Brownian motion. In addition, imperfection of the particle alignment should be considered for the particles that went through outlets 1 and 2. Since the theoretical value of the sum of the split flow rates through the first and second side channels is ∼100% of the initial value, it may be necessary to increase the amount of liquid withdrawn through the first or second side channels by increasing the number of channels. To examine whether the theoretical flow rate fit the actual value, and to calculate the concentration rate, the distributed flow rates were measured using a solution without particles. The measured flow rates, in comparison with the theoretical values, are shown in Table 1. There was a slight difference between the measured and calculated values. Such an error is unavoidable, since a slight error in the estimation of the channel geometry

Table 1. Comparison of the Distributed Flow Rates into Each Outlet between Measured and Theoretical Values (%) outlet

measured theoretical

1

2

3

4

5

24.43 ( 0.51 28.52

1.89 ( 0.03 2.39

1.42 ( 0.01 1.71

0.83 ( 0.02 1.00

71.43 ( 0.57 66.39

Analytical Chemistry, Vol. 78, No. 4, February 15, 2006

1361

potentially creates a severe difference. However, regarding that the microchannel geometry is relatively complex, it can be said that these values are reasonable. So it was demonstrated that the hydrodynamic resistance of the microchannel could be precisely calculated, and the distributed flow rates could be predicted. Considering that only ∼1.4 and ∼0.8% of the liquid flow were distributed into outlets 3 and 4, respectively, the concentration of the 2.1-µm particles in outlet 4 was increased ∼80-fold compared to the initial value, while the 3.0-µm particles in outlet 3 were concentrated ∼63-fold. These results are indicative of the capability of this microfluidic circuit, since particle concentration was dramatically increased without the help of outer fields. If there were no first side channels, ∼29% of the introduced particles collected from each side outlet would pass through outlet 1. In this regard, the validity of the flow splitting and recombining was successfully demonstrated. It is considered that when we use a microchannel without the first side channels but having two inlets, and introduce liquids with and without particles into the channel, perfect particle alignment will be achieved without performing flow split and recombination.40 However, the presented particle concentrator/sorter is capable of passively sorting particles without precisely controlling the flow rates, hence making it suitable for integration, and offering superiority over conventional particlesorting methods. CONCLUSIONS In this paper, we have described an improved scheme for particle concentration and size-dependent separation in microfluidic devices, employing effective flow manipulations based on the concept of hydrodynamic resistance of microchannels. This

1362

Analytical Chemistry, Vol. 78, No. 4, February 15, 2006

particle sorter is advantageous, in the sense that continuous introduction of particle suspension into a microchannel enables us to achieve simultaneous particle concentration and sizedependent separation. We consider that this study has successfully demonstrated the capability of microfluidic devices to easily and accurately manipulate small particles and expect this particle sorter being used for various biological, environmental, and industrial applications. ACKNOWLEDGMENT This research was supported in part by Grants-in-Aids for JSPS Fellows, Scientific Research (B) (16310101), and Priority Areas (A) (13025216) from the Ministry of Education, Science, Sports and Culture of Japan; by the Nanotechnology Project of the Ministry of Agriculture, Forestry and Fisheries of Japan; and by the Research Association of Micro Chemical Process Technology, Japan. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text (Tables S1 and S2, and Figure S1). This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review November 11, 2005. Accepted December 2, 2005. AC0520083