On-Chip Free-Flow Magnetophoresis: Continuous Flow Separation of

Anal. Chem. 2004, 76, 7250-7256

On-Chip Free-Flow Magnetophoresis: Continuous Flow Separation of Magnetic Particles and Agglomerates Nicole Pamme† and Andreas Manz*,‡

Imperial College London, Department of Chemistry, London SW7 2AY, U.K.

The separation of magnetic microparticles was achieved by on-chip free-flow magnetophoresis. In continuous flow, magnetic particles were deflected from the direction of laminar flow by a perpendicular magnetic field depending on their magnetic susceptibility and size and on the flow rate. Magnetic particles could thus be separated from each other and from nonmagnetic materials. Magnetic and nonmagnetic particles were introduced into a microfluidic separation chamber, and their deflection was studied under the microscope. The magnetic particles were 2.0 and 4.5 µm in diameter with magnetic susceptibilities of 1.12 × 10-4 and 1.6 × 10-4 m3 kg-1, respectively. The 4.5-µm particles with the larger susceptibility were deflected further from the direction of laminar flow than the 2.0-µm magnetic particles. Nonmagnetic 6-µm polystyrene beads, however, were not deflected at all. Furthermore, agglomerates of magnetic particles were found to be deflected to a larger extent than single magnetic particles. The applied flow rate and the strength and gradient of the applied magnetic field were the key parameters in controlling the deflection. This separation method has a wide applicability since magnetic particles are commonly used in bioanalysis as a solid support material for antigens, antibodies, DNA, and even cells. Free-flow magnetophoretic separations could be hyphenated with other microfluidic devices for reaction and analysis steps to form a micro total analysis system. A vast number of reactions in genomics, proteomics, and clinical medicine involve molecular recognition events between single strands of DNA, between antibodies and antigens, or between receptors and cells. Such reactions usually require a number of steps that must be performed sequentially, for example, some form of isolation, washing, or purification. Recently, the application of biomagnetic particles to these procedures has become very popular. This is because they are simple to use, the operating conditions are mild, and their efficiency is large.1,2 Such * To whom correspondence should be addressed. E-mail: [email protected]. fax: +49 231 1392 300. † Current address: NIMS, ICYS, 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan. ‡ Current address: ISAS, Bunsen-Kirchhoff Strasse 11, 44139 Dortmund, Germany. (1) Bosnes, M.; Deggerdal, A.; Rian, A.; Korsnes, L.; Larsen, F. In Scientific and Clinical Applications of Magnetic Carriers; Ha¨feli, U., Ed.; Plenum Press: New York, 1997; pp 269-285.

7250 Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

microparticles usually consist of an iron oxide core, surrounded by a polymer shell. When coated with, for example, an antibody, the particles can attach themselves to particular targeted biomolecules or cells. By the application of an external magnetic field, the magnetically labeled cells or biomolecules can be retained in the reaction vessel while the remainder of the reaction mixture is removed. However, many bioanalytical procedures require a large number of sequential reaction and washing steps. This can be very time-consuming when using a batchwise method. To overcome the limitations of batch separation processes, a number of approaches for continuous flow separation of magnetic material from nonmagnetic material have been investigated. Zborowski et al. described the separation of magnetically labeled cells in a quadrupolar magnetic field.3 Hartig et al. used a separation chamber similar to a free-flow electrophoresis device.4,5 They applied a fanned out, inhomogeneous magnetic field across the chamber. On a test tube scale they separated magnetic particles from nonmagnetic particles. Another approach for continuous flow magnetic separation is split flow thin (SPLITT) fractionation.6 In this technique, two flow streams, one containing magnetic particles the other containing a buffer, were introduced into a central channel and were separated into two outlet streams further downstream. Magnetic particles were dragged away from their stream into the buffer stream by means of an external magnetic field. Miniaturization of magnetic separation devices has gained attention in recent years due to the anticipated integration of such devices into micro total analysis systems.7-9 The separated magnetic particles carrying the biomolecules or cells of interest could then be used for downstream analysis and testing. Miniaturization has the potential to offer a fast and highly efficient separation because the magnetic force on a particle is higher the closer the particle is to the magnet surface. Furthermore, sample (2) Dynal, A. S. Biomagnetic Techniques in Molecular Biology, 3rd ed.; Dynal Corp.: Oslo, Norway, 1998. (3) Zborowski, M.; Sun, L. P.; Moore, L. R.; Williams, P. S.; Chalmers, J. J. J. Magn. Magn. Mater. 1999, 194, 224-230. (4) Hartig, R.; Hausmann, M.; Schmitt, J.; Herrmann, D. B. J.; Riedmiller, M.; Cremer, C. Electrophoresis 1992, 13, 674-676. (5) Hartig, R.; Hausmann, M.; Cremer, C. Electrophoresis 1995, 16, 789-792. (6) Fuh, C. B.; Tsai, H. Y.; Lai, J. Z. Anal. Chim. Acta 2003, 497, 115-122. (7) Reyes, D. R.; Iossifidis, D.; Auroux, P. A.; Manz, A. Anal. Chem. 2002, 74, 2623-2636. (8) Auroux, P. A.; Iossifidis, D.; Reyes, D. R.; Manz, A. Anal. Chem. 2002, 74, 2637-2652. (9) Verpoorte, E. Electrophoresis 2002, 23, 677-712. 10.1021/ac049183o CCC: $27.50

© 2004 American Chemical Society Published on Web 11/18/2004

and reagent consumption can be kept to a minimum. Several groups have investigated the use of magnetic particles in microfluidic devices as a solid support for reactions such as mRNA isolation,10 immunoassays,11 DNA hybridization,12 protein analysis,13 and recently the retention of magnetically labeled cells.14 In these approaches, the particles were stopped in flow by an external magnet and the reaction was performed on the particle bed. The continuous flow separation of magnetic particles in microfluidic devices has so far received very little attention. The separation of magnetic particles has been achieved on an analytical scale by field-flow fractionation.15 This method, however, is not readily integrated with downstream applications for the separated particles. For the preparative separation of magnetic particles, two methods have been suggested. A device for continuous flow separation of particles in an H-shaped channel was theoretically described by Lee et al.16 Berger and co-workers published the design and fabrication of a microfabricated magnetic cell sorter, in which they employed a magnetic field gradient generated by miniaturized magnetic wires in combination with hydrodynamic forces. However, no separation of magnetic particles or magnetically labeled cells was demonstrated with this device. The reported devices for microfluidic continuous flow separation were not employed for the separation of mixtures of different types of magnetic particles. For a powerful separation system, it would be beneficial not only to separate magnetic particles from nonmagnetic material but also to separate magnetic particles with different properties from each other. Here, we demonstrate a technique for separating magnetic particles from each other as well as from nonmagnetic material in continuous flow on a microfluidic chip (on-chip free-flow magnetophoresis). The device features a separation chamber over which a laminar flow was generated by a number of inlet and outlet channels (Figure 1). A magnetic field was applied perpendicular to the direction of flow. Magnetic particles were dragged into this field and thus deflected from the direction of laminar flow, depending on their size and magnetic susceptibility. In this paper, we demonstrate the differentiation of different kinds of single particles as well as single particles and particle agglomerates. THEORY Free-flow magnetophoresis is a method for separating magnetic particles from each other in continuous flow. The central component is a rectangular flat separation chamber over which a laminar flow is generated in the x-direction by a number of inlet and outlet channels (Figure 1). Perpendicular to the direction of (10) Jiang, G.; Harrison, D. J. Analyst 2000, 125, 2176-2179. (11) Hayes, M. A.; Polson, N. A.; Phayre, A. N.; Garcia, A. A. Anal. Chem. 2001, 73, 5896-5902. (12) Fan, Z. H.; Mangru, S.; Granzow, R.; Heaney, P.; Ho, W.; Dong, Q.; Kumar, R. Anal. Chem. 1999, 71, 4851-4859. (13) Choi, J.-W.; Oh, K. W.; Thomas, J. H.; Heineman, W. R.; Halsall, H. B.; Nevin, J. H.; Helmicki, A. J.; Henderson, H. T.; Ahn, C. H. Lab Chip 2002, 2, 2730. (14) Furdui, V. I.; Kariuki, J. K.; Harrison, D. J. J. Micromech, Microeng. 2003, 13, S164-S170. (15) Gascoyne, P. R. C.; Vykoukal, J.; Weinstein, R.; Gandini, A.; Parks, D.; Sawh, R. In Micro Total Analysis Systems 2002; Baba, Y., Berg, A. v. d., Eds.; Kluwer Academic Publishers: Nara, Japan, 2001; pp 323-325. (16) Chronis, N.; Lam, W.; Lee, L. In Micro Total Analysis Systems 2001; Ramsey, J. M., Berg, A. v. d., Eds.; Kluwer Academic Publishers: Monterey, CA, 2001; pp 497-498.

Figure 1. Concept of free-flow magnetophoresis. Magnetic particles are pumped into a laminar flow chamber; a magnetic field is applied perpendicular to the direction of flow. Particles deviate from the direction of laminar flow according to their size and magnetic susceptibility and are thus separated from each other and from nonmagnetic material.

laminar flow, i.e., in the y-direction, an inhomogeneous magnetic field is applied, which forms a magnetic field gradient over the separation chamber. A mixture of magnetic particles and nonmagnetic particles can be injected continuously into the system through the sample inlet channel. The nonmagnetic particles are not influenced by the magnetic field and leave the chamber at the exit opposite the sample inlet. Superparamagnetic particles, however, become magnetized and are hence dragged into the inhomogeneous magnetic field. Magnetophoretic separation as described here is based on deflection of the particles into the y-direction. The applied magnetic field may also have a component in the z-direction. In this approach, the z-component is neglected due to the low height of the separation chamber. The deflection of the magnetic particles, udefl (in m s-1) can be described as the sum of two vectors: the vector for the magnetically induced flow velocity on the particle, umag, and the vector for the velocity of the hydrodynamic flow uhyd:

udefl ) umag + uhyd

(1)

The magnetically induced flow, umag, is the ratio of the magnetic force, Fmag (in N), exerted on the particle by the magnetic field to the viscous drag force:

umag )

Fmag Fmag ) Fdrag 6πηr

(2)

where η is the viscosity of the medium (in kg m-1 s-1) and r the particle radius (in m). The magnetic force, Fmag, is proportional to the magnetic flux density, B (in tesla) and the gradient of the magnetic field, ∇‚B, (in tesla m-1) of the externally applied field. In a homogeneous field with ∇‚B ) 0, the net force on a magnetic particle is zero. The magnetic force is also proportional to the particle volume, Vp, (in m3) and the difference in magnetic susceptibility between the particle and the fluid, ∆χ (dimensionless):17 (17) Hatch, G. P.; Stelter, R. E. J. Magn. Magn. Mater. 2001, 225, 262-276.

Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

7251

Fmag )

∆χ‚Vp ‚(∇‚B)‚B µ0

(3)

where µ0 is the permeability of a vacuum (in H m-1). When inserting eq 3 into eq 2, it can be seen that, for a given magnetic field and a given viscosity, the magnetic velocity, umag, is dependent on the size and the magnetic characteristics of the particle. It is proportional to the square of the particle radius and to the magnetic susceptibility of the particle:

umag ∝ r2χp

(4)

Hence, particles that are either different in size, r, or different in their magnetic susceptibility, χp, will be deflected from the direction of laminar flow to a different degree. Thus, it should be possible to separate particles. EXPERIMENTAL SECTION Microchip Design and Fabrication. The microchip design is shown in Figure 2. The main component of the chip was a 6 mm × 6 mm separating chamber supported by 13 square posts of 200 µm × 200 µm. The 16 buffer inlet channels and 1 sample inlet channel were all 100 µm wide and evenly spaced. On the opposite site of the chamber, there were 16 outlet channels, also 100 µm wide and evenly spaced. The sample inlet was 10 mm long. The buffer inlet system was branched to enable the use of a single buffer inlet reservoir and to evenly spread the liquid over the entire width of the separation chamber. The outlet system was branched in a similar way to facilitate the application of negative pressure over the outlet channels. Although using this outlet branching network would ultimately recombine any separated particles in the outlet reservoir, the design required only one pump and was chosen for proof of principle. Fabrication of the glass microchip was performed in-house using a direct write laser lithography system (DWL, Heidelberg Instruments, Heidelberg, Germany). The chip design as shown in Figure 2 was patterned onto a wafer of soda lime glass coated with chromium and photoresist (Nanofilm, Westlake Village, CA). After photodevelopment and chrome-etching, the structure was wet etched to a depth of 20 µm. Subsequently, the remaining photoresist and metal layers were removed. Access holes were drilled, and the structure was thermally bonded to a 1-mm-thick soda lime glass cover. Pipet tips were glued around the inlet holes as reservoirs with epoxy glue. A fused-silica capillary (150-µm i.d., 375-µm o.d., Composite Metal, Worcester, U.K.) was glued into a ferrule (Vespel, GVF 16-004, SGE Europe, Milton Keynes, U.K.) and then glued to the outlet reservoir. A 5-mL syringe was connected to the capillary via Tygon tubing (1.01-mm i.d., 1.78-mm o.d., Cole Parmer) and a short piece of PTFE tubing (0.3-mm i.d., 1.58-mm o.d., Waters, Elstree, U.K.). Flow was controlled by applying a withdrawal rate of typically between 50 and 200 µL h-1, corresponding to between 0.1 and 0.4 mm s-1 in the separation chamber, using a syringe pump (PHD 2000, Harvard, Kent, U.K.). The microchip was mounted onto the stage of an inverted microscope (DMIRB, Leica, Milton Keynes, U.K.) equipped with a CCD camera (CX-999P, Sony, Weybridge, Surrey, U.K.) for 7252 Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

Figure 2. (a) Design of the microfluidic chip. The 6 mm × 6 mm central chamber was supported by 13 pillars (white diamonds). Channels leading into and out of the chamber were 100 µm wide. The structure was 20 µm deep. (b) Photograph of the etched and bonded microchip.

Figure 3. Photograph of the free-flow magnetophoresis setup showing the glass chip with fluid reservoirs and pump connection as well as the magnet assembly.

observation and video filming. A photograph of the microchip with reservoirs and connecting tubing is shown in Figure 3. Chemicals. Deionized water obtained from tap water via a water purification system (Elga Ltd., Buckinghamshire, U.K.) was used throughout. Chemicals were obtained from Sigma (Dorset, U.K.) of analytical grade purity. Solutions were filtered through a 0.2-µm syringe filter (Whatman) prior to use. Magnetic Particles and Magnets. Superparamagnetic particles (Dynabeads) with a diameter of 4.5 µm coated with epoxy functional groups were obtained from Dynal Biotech (Oslo, Norway) as a suspension of 4 × 108 particles mL-1. The particles had a magnetic susceptibility of χ ) 1.6 × 10-4 m3 kg-1. Superparamagnetic particles with a diameter of 2 µm and chloromethylstyrene surface groups were obtained from Micromod (Rostock, Germany) as a suspension with 6.6 × 109 particles mL-1. These particles had a magnetic susceptibility of χ ) 1.12 × 10-4 m3 kg-1. The particles were pretreated as follows. A 10-µL aliquot of stock suspension was mixed with 990 µL of a glycine saline buffer

Figure 4. Modeling of the magnetic field generated by an assembly of three NdFeB magnets as typically used for free-flow magnetophoresis experiments. The position of the separation chamber in the magnetic field is indicated by the white rectangle.

(100 mM glycine, 150 mM NaCl, pH 8.3). The mixture was incubated for 30 min at 45 °C in a water bath and vortexed frequently to resuspend the particles. The glycine acted as a blocking agent for the reactive functional groups on the surface of the particles. The particle suspension was then diluted 1:100 in glycine/saline buffer and stored at 4 °C. Prior to experiments, particle suspensions were further diluted in a buffer consisting of PBS/BSA (0.1%), pH 7.4 to a concentration of 1 × 104 particles mL-1. Nonmagnetic particles, Polybead polystyrene microspheres with a diameter of 6 µm, were purchased from Polysciences (Warrington, PA) as a suspension of 109 beads mL-1. They were diluted in PBS buffer to a concentration of 1 × 106 particles mL-1. Small, strong, permanent neodymium-iron-boron (NdFeB) magnets (Magnetsales, Swindon, U.K.) were used to generate a magnetic field over the separation chamber. These magnets were cylindrical with diameters ranging from 4 to 6 mm and lengths ranging from 3 to 5 mm. Typically, assemblies of several magnets were used to increase the magnetic flux density, and a steel backplate was used to direct the flux toward the separation chamber. The flux density of the magnet assembly was ∼500 mT on the magnet surface. The assembly was placed on top of the microchip to one side of the chamber, as shown in Figure 3. The magnetic field over the separation chamber for a typical setup was modeled (Figure 4) using MagNet software (Infolytica Corp., Abingdon, U.K.). Typical Running Conditions. The chip was flushed for 10 min each with deionized water, NaOH solution (0.5 mM), deionized water, and then buffer solution (PBS/BSA (0.1%), pH 7.4) before each experiment. The magnet assembly was placed on the microchip, and the particle reservoir was filled with particle suspension. At flow rates between 50 and 200 µL h-1, the particles were drawn into the separation chamber and their flow path was observed through the microscope and videotaped. Images were analyzed using VirtualDub freeware. After experiments, the chip was flushed with deionized water. RESULTS AND DISCUSSION Flow Pattern. To visualize the flow regime within the microfluidic device, the sample reservoir was filled with water whereas the buffer reservoir was filled with black ink. As can be seen in Figure 5, a laminar flow regime was established within the separation chamber. Liquid from the sample inlet left the

Figure 5. Visualization of the laminar flow regime within the microfluidic device. The sample reservoir was filled with water; the buffer reservoir was filled with black ink. Fluid from the sample inlet was observed to flow straight through the separation chamber and to leave via the directly opposite exit. The white dots in the separation chamber are the supporting posts (compare to Figure 2).

separation chamber via the outlet directly opposite the sample inlet. The black ink from the buffer inlets left the separation chamber via all outlets except the outlet opposite the sample inlet. Differentiation of Several Kinds of Single Particles. The deflection of three kinds of single particles was investigated. A magnet assembly consisting of three NdFeB magnets was placed on top of the microchip toward one side of the separation chamber as shown in Figure 3. The magnets had diameters of 6, 5, and 4 mm, respectively, and lengths of 4, 3, and 5 mm, respectively. A steel backplate was attached to the largest magnet. A photograph of the whole separation chamber is shown in Figure 6a with the position of the magnet assembly being indicated. Figure 6a is in fact a combination of four separate photographs, since the field of view of the microscope, even with the lowest magnification objective (2.5×), covered only about a quarter of the chamber area. Different particle suspensions were consecutively introduced into the device at a flow rate of 150 µL h-1, which resulted in a flow of ∼0.3 mm s-1 in the separation chamber. The flow path was captured on video and could then be drawn into the photograph as shown in Figure 6a for each type of sample. First, a suspension of nonmagnetic 6-µm polystyrene particles was added to the sample reservoir. As expected, the particles were not deflected and followed the laminar flow regime to exit the chamber through the outlet opposite the sample inlet channel (Figure 6b). The chip was flushed and the particle reservoir was filled with a suspension of 2-µm-diameter magnetic particles. The particles were deflected toward exit 4 as indicated in Figure 6c. Not all the beads took exactly the same flow path, but the path indicated was that taken by the majority of particles. Finally, the sample reservoir was filled with a suspension of the 4.5-µm-diameter Dynabeads. These larger beads with a larger magnetic susceptibility were deflected further than the 2-µm particles and left the microstructure via exit 6 (Figure 6d). The forces exerted on the magnetic particles, Fmag, were estimated using eq 2. The 2.0-µm particles were deflected toward outlet 4, i.e., by 1.3 mm, whereas the 4.5-µm particles were deflected toward outlet 6, i.e., by 2.0 mm. Hence, the magnetic forces on the 2.0- and 4.5-µm particles were approximately 1.7 and 5.4 pN, respectively. This corresponds well with the forces expected from the simulated magnetic field (see Figure 4) that were estimated to be between several piconewtons to a few hundred piconewtons, depending on the position of the particle in the chamber. The deflected particles followed an undulating path rather than a straight line from the inlet to the exit. This undulating flow path Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

7253

Figure 6. (a) Photograph of the separation chamber. The position of the magnet assembly is indicated by the gray box. Consecutively, at a flow rate of 150 µL h-1, particles were introduced into the sample reservoir and their deflection path was observed. From the video footage obtained, their paths were drawn onto these photographs. (b) Nonmagnetic 6-µm polystyrene particles were not influenced by the magnetic field and left the separation chamber via outlet 1. (c) The 2.0-µm-diameter magnetic particles featuring a magnetic susceptibility of χ ) 1.12 × 10-4 m3 kg-1 were deflected toward outlet 4. (d) The 4.5-µm-diameter magnetic particles with χ ) 1.6 × 10-4 m3 kg-1 were deflected toward outlet 6.

arose from the inhomogeneous magnetic field, which competed with the hydrodynamic flow in the chamber. Upon entering the separation chamber, the particles were initially deflected considerably from the direction of laminar flow. This could be explained by the 90° junction between the inlet channels and the separation chamber. Sample and buffer must fill the space between the inlet channels, and they can only do this by deviating from the straight flow direction (compare Figure 5). After the first millimeter or so, the particles were observed to flow along more or less in the direction of the applied hydrodynamic flow, up to about half of the chamber, with minimal deflection toward the magnet. Only once the particles were in the second half of the chamber, where the magnetic flux gradient was stronger (see Figure 4), was a pronounced magnetic deflection observed. Not all particles followed exactly the same flow path into exclusively one exit. A distribution in deflection was observed such that the majority of the particles (>75%) flowed into one exit and the remaining particles flowed into neighboring exits (Figure 7). Two factors may account for this relatively wide deflection: (1) the irregular flow pattern at the 90° junction between the channels and the separation chamber as discussed above and (2) the 7254 Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

Figure 7. Distribution of particle deflection as obtained in an experiment with the 4.5-µm Dynabeads at a flow rate of 200 µL h-1 and a magnetic assembly similar to that in Figure 6. Whereas the majority of the particles (∼75%) left the separation chamber via exit 5, some particles were deflected toward exits 4, 6, and 7.

parabolic flow velocity profile of the hydrodynamic flow over the height of the separation chamber. The flow velocity in the middle is typically 50% higher than the average velocity of the liquid. Hence, a particle mainly flowing through the chamber at middle height would flow faster than a particle that diffused up- and

Figure 8. Photograph of the separation of single particles and a particle agglomerate consisting of three particles. The single particles were deflected to outlet 3, whereas the agglomerate was deflected toward outlet 5.

downward.18 Differences in particle velocity result in differences in deflection path and hence a distribution of the outlet taken by the particles. Although the laminar flow regime was disturbed slightly around the posts that support the separation chamber, these posts were observed to have a negligible effect on the deflection distribution. Single Particles and Particle Agglomerates. The behavior of single particles and particle agglomerates was also investigated. The 2-µm diameter particles were found to agglomerate when left standing for a few days. Although this nonspecific binding of beads to each other is generally undesirable, it was used in this experiment to demonstrate the different deflection in free-flow magnetophoresis of single and agglomerated particles. A particle agglomerate is larger in size than a single particle; thus, the magnetic deflection of the agglomerate is larger than that of a single particle (see eq 4). At a flow rate of 50 µL h-1, a mixture of single particles and particle agglomerates was introduced into the separation chamber and a magnetic assembly similar to the one described for previous experiments was used. The single particles were found to be deflected toward outlets 2 and 3, whereas the agglomerates were deflected further. The photograph in Figure 8 shows two single beads heading toward exit 3 and an agglomerate consisting of three particles on its way toward exit 5. Agglomerates consisting of two particles were deflected toward exits 3-5, with a majority of these particles entering exit 4. Differentiation between single particles and particle agglomerates was also observed in experiments with the 4.5-µm particles. Another observation made for particle agglomerates was that they would often flow faster through the 20-µm-deep chamber than single particles. This difference was not so pronounced for agglomerates of two particles. Agglomerates of three or four particles, however, were found to flow ∼30% faster than the single particles. This behavior has an averse effect on the separation of single and agglomerated particles. The faster the agglomerate travels, the less time it has to interact with the magnetic field; i.e., the less deflection it will undergo. The faster velocity of the agglomerates was likely to be caused by the hydrodynamic flow profile over the height of the microchip. The large agglomerates could not diffuse as close to the chamber walls as the smaller (18) Chmela, E.; Tijssen, R.; Blom, M. T.; Gardeniers, H. J. G. E.; Berg, A. v. d. Anal. Chem. 2002, 74, 3470-3475.

particles could. Hence, the agglomerates were exclusively carried along in the region of the flow profile that features higher velocities.18 Variation of Flow Rate. For a given magnet position and a given type of magnetic particle, the chosen flow rate had a great effect on the observed magnetic deflection, udefl. At excessively high flow rates, the hydrodynamic flow vector, uhyd, outweighed the magnetically induced flow velocity, umag, and no deflection was observed. However, at low flow rates, the magnetic force became too large with respect to the hydrodynamic flow. Particles were no longer observed to flow smoothly but were observed to flow in the intermittent manner of a stop-and-go flow. This resulted in a large deflection but also in widely distributed flow pathways. At even lower flow rates, the particles would become totally stuck to the ceiling of the separation chamber, due to the partly upwardly directed magnetic force from the magnet assembly on top of the microfluidic chip. Within a certain window, typically between 0.1 and 0.4 mm s-1, an optimum velocity was found, in which the particles were observed to flow smoothly through the separation chamber. The lower the velocity, the more the particles were deflected. The optimum flow velocity for the 2.0-µm particles was generally lower than the optimum flow velocity for the 4.5-µm magnetic particles. The larger particles with the higher susceptibility would experience a larger magnetic deflection (see eq 4) and could thus tolerate higher flow velocities. Variation of Magnetic Field. The design and position of the magnetic field, namely, the magnetic flux density as well as the magnetic flux gradient, were the key factors for obtaining a sufficient magnetic force on the particle. NdFeB was chosen as a magnet material due to its high flux density of up to 500 mT on the magnet surface. Lower flux values proved inefficient for magnetic separation. Care had to be taken not to place the magnet assembly in too close proximity to the sample reservoir, as the particles would then become magnetized in the sample reservoir and thus form chains and agglomerates. The best results were obtained when the magnet assemblies were placed more downstream the separation chamber (see Figure 4). Discussion. In many different experiments it was observed that magnetic particles were deflected in the free-flow magnetophoresis chamber depending on their size and magnetic susceptibility, whereas nonmagnetic particles were not influenced by the magnetic field. There are two challenges to be met for on-chip free-flow magnetophoresis: (1) the distribution of the deflection path for a given particle population was rather large and would typically spread over three to four exit channels; (2) the somewhat low throughput of ∼12 particles min-1 caused by the rather slow flowrates (mm s-1) and the rather low particle concentrations (104 particles mL-1). To improve these issues, a number of things could be changed. The distribution in deflection was partly attributed to the 90° junction between the inlet channels and the separation chamber as well as between the chamber and the outlet channels. A tapered channel junction would allow for a smoother and more reproducible introduction of the particles into the separation chamber. The chip design could be further improved by narrowing the width of the separation chamber and then placing the magnet assembly next to the separation chamber rather than on top of the device. Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

7255

The parabolic flow profile over the 20-µm height of the separation chamber resulted in some variation of the flow velocities of the particles through the separation chamber. To minimize this effect, electroosmotic flow (EOF) rather than hydrodynamic flow could be employed. EOF features a flat pluglike flow profile. Flow rates in the order of millimeters per second were necessary to see a pronounced effect on particle deflection. At faster flow rates, the hydrodynamic force outweighed the magnetic force on the particle and thus no deflection was observed. A more powerful and more sophisticated magnetic field could be employed to improve this. A field with a larger magnetic flux density or a larger magnetic flux gradient would result in a larger force on the magnetic particles (see eq 3). As can be seen from Figure 4, a large gradient and a large flux only occurs over a fairly small part of the chamber. The area of the separation chamber could be used more efficiently by applying a uniform and high flux gradient over the x-y plane of the entire chamber. The field gradient in z-direction should be kept to a minimum to avoid an upward pull of the particles. With such an improved magnetic field, faster flow rates could be used and hence an increase in the throughput of particles could be achieved.

7256

Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

CONCLUSIONS Free-flow magnetophoresis is a powerful method for miniaturized continuous flow particle separation that could be applied in many areas of biochemical analysis. Particles coated with biomolecules such as antigens, antibodies, or single-stranded DNA could be used in proteomics and genomics, respectively. Cells could also be labeled with magnetic particles and separated from nonmagnetic material in continuous flow. The separation method could be utilized inside a microTAS device and combined upstream with bioassays and downstream with sample analysis or collection of isolated species. ACKNOWLEDGMENT The authors thank Prof. Aric Menon for discussion and Jan C.T. Eijkel as well as Alexander Iles for valuable comments on the manuscript. N.P. thanks Asahi Kasei Corp. for financial support. Received for review June 4, 2004. Accepted September 22, 2004. AC049183O