Microfluidic Control Using Colloidal Devices

See allHide authors and affiliations

Science  07 Jun 2002:
Vol. 296, Issue 5574, pp. 1841-1844
DOI: 10.1126/science.1072133


By manipulating colloidal microspheres within customized channels, we have created micrometer-scale fluid pumps and particulate valves. We describe two positive-displacement designs, a gear and a peristaltic pump, both of which are about the size of a human red blood cell. Two colloidal valve designs are also demonstrated, one actuated and one passive, for the direction of cells or small particles. The use of colloids as both valves and pumps will allow device integration at a density far beyond what is currently achievable by other approaches and may provide a link between fluid manipulation at the macro- and nanoscale.

Microscale devices designed to accomplish specific tasks have repeatedly demonstrated superiority over their macroscale analogs (1, 2). The advantages of such devices are due largely to unique transport properties resulting from laminar flows and vastly increased surface-to-volume ratios (3) and have enabled microscale sensors (4) and fabrication schemes (5) not possible on the macroscale. Additionally, microfluidic processes may be easily parallelized for high throughput (6) and require vastly smaller sample volumes, a major benefit for applications in which reagents or analytes are either hazardous or expensive (7). The utility, speed, and performance of microsystems typically increase as the overall device size decreases. The need to pump and direct fluids at very small length scales, however, has long been the limiting factor in the development of microscale systems, thus generating a tremendous amount of interest in microfluidics development (8). As improved actuation techniques have become available, conventional valving and pumping schemes have been miniaturized, yet continue to dwarf microchannels and other chip-top features (9).

Recently, several approaches conceived explicitly for the microscale have been developed, including platforms based upon electrohydrodynamics (10), electroosmosis (11), interfacial phenomena (12, 13), conjugated materials (14), magnetic materials (15), and multilayer soft lithography (16). Although these microfluid handling techniques enable functional devices on microscopic length scales, they also impose unique constraints upon potential device capability, flexibility, and performance. To fully integrate multiple fluidic processes within a single microsystem, methods for microfluid handling must be developed that can accommodate fluids of complex and dynamic composition and are of comparable size to that of the processes into which they are embedded.

Development of devices that can function at these length scales has focused on complex fabrication schemes for intricate components such as gears (17, 18), cantilevers (19), and other microscale objects (20,21). The fabrication and actuation of these devices, however, has been limited to bulk environments external to the microfluidic geometries. Because no practical implementation scheme has been developed for their incorporation into functioning microfluidic systems, they have not realized their suggested potential as microfluidic pumps and valves. We avoid the complexities inherent in other approaches by using colloidal microspheres as the active flow-control element. These materials are an excellent choice because they can be synthesized at length scales easily transported through microfluidic networks and their surface chemistry may be readily altered. Colloidal silica, for example, is easily modified for dispersion in both aqueous and nonaqueous solvents and is well known for its biocompatibility, allowing its use in a host of microfluidic applications.

It has also been shown that colloids can be directly manipulated through the application of external fields. The processes of electrophoresis, dielectrophoresis, and magnetophoresis have all been used to control and influence the motion of small particles in solution. Particularly useful for our preliminary studies, however, is the technique of optical trapping, which has become popular because it allows the direct manipulation of individual colloids. This noncontact, noninvasive technique eliminates the need to physically interface to the macroscopic world and allows for the manipulation of complex asymmetric objects or multiple objects at once, as would be required for the actuation of a microfluidic valve or pump.

For multiple trap generation, we used a scanning approach in which a piezoelectric mirror is translated to rapidly reflect a laser beam in a desired pattern. If the piezoelectric mirror is scanned over the desired pattern at a frequency greater than that associated with Brownian time scales, a time-averaged trapping pattern is created. Through the application of this technique, known as scanning laser optical trapping (SLOT) (22, 23), we are able to rearrange microspheres into functional structures and subsequently actuate these structures to generate microfluidic pumping and valving. However, using an optical actuation scheme and transmission microscopy to monitor device performance requires a method for the creation of channels at single-micrometer length scales in a transparent housing. Because of this requirement, we applied soft lithography techniques, pioneered by the Whitesides group, which allow for the inexpensive fabrication of microfluidic networks in poly(dimethylsiloxane) (PDMS), an optically transparent elastomer (3, 24–26).

Working at microscopic length scales offers unique challenges for colloidal pump design. This is illustrated through calculation of the Reynolds number, Re = ρνD/η, where for colloidal length scales in aqueous solutions, ρ = 1 g/cm3, η = 0.01 g/cm·s,D ≈ 5 μm, and ν ≈ 5 μm/s, givingRe ≈ 10−5 ≪ 1 and corresponding to laminar flow. Under these circumstances, fluid flow is fully reversible, and pump designs that rely on centrifugal action, such as impeller-type approaches, are inappropriate. For this reason, our designs are based on positive-displacement pumping techniques that operate by imparting forward motion to individual plugs of fluid.

Our first design is a two-lobe gear pump in which small, trapped pockets of fluid are directed through a specially designed cavity fabricated in a microchannel by rotating two colloidal dumbbells or “lobes” in opposite directions. Over repeated and rapid rotations, the accumulated effect of displacing these fluid pockets is sufficient to induce a net flow. This motion is illustrated in Fig. 1, where clockwise rotation of the top lobe combined with counterclockwise rotation of the bottom lobe induces flow from left to right. In these experiments, each of the lobes consisted of two independent 3-μm silica spheres (Bangs Laboratories, Fishers, IN). To create these structures, we first maneuvered the colloids with the optical trap to a 3-μm-deep section of channel designed with a region of wider gap to accommodate lobe rotation. Once the particles were properly positioned, the laser was scanned in a manner such that a time-averaged pattern of four independent optical traps was created, one for each microsphere composing the two-lobe pump. By rotating the two traps in the upper part of the channel and the two traps in the lower part of the channel in opposite directions and offset by 90°, the overall pump and the corresponding fluid movement were achieved. Flow direction was easily and quickly reversed by changing the rotation direction of both top and bottom lobes.

Figure 1

Pump design illustrating lobe movement (the top pair rotate clockwise, the bottom counterclockwise). Also shown is 3-μm colloidal silica undergoing rotation at 2 Hz within a 6-μm channel. Frames are separated by two cycles to show movement of the 1.5-μm colloidal silica tracer particles. To create the pump, we first cured PDMS over a negative photoresist pattern created with conventional photolithography. The resulting elastomeric layer (∼1 mm thick) was then placed upon a base fabricated from a block of Plexiglas designed to allow easy colloidal fluid introduction to the channel network through a syringe pump (Kd Scientific, KDS200, New Hope, PA). Holes through the PDMS membrane were then aligned with the fluid ports in the base with which the PDMS layer seals tightly and reversibly. A cover slip placed on top of the channel network created a tight, reversible seal and a closed microfluidic system. The optical trap consisted of a laser (Spectra Physics Millennia V, Mountain View, CA; 532-nm wavelength, typically operated at 0.8 W for actuation) and a microscope (Optiphot 150, Nikon) with an oil immersion 100× objective, numerical aperture = 1.3 (CFN plan fluor, Nikon) used to create a tight laser focus of ∼1 μm and to observe the experiments. The piezoelectric mirror (Physik Instrumente, Karlsruhe, Germany) used to create the scanning trap was controlled by computer (Macintosh G3, Apple) with routines written in LabView (National Instruments, Austin, TX) and scanned at a rate of 50 to 100 Hz.

To aid visualization of flow and provide a means of estimating flow rate, we added tracer particles consisting of 1.5-μm silica spheres to the aqueous solution (Movie S1). The measured tracer particle velocities were determined as a function of the gear rotation rate both in the forward and reverse directions and indicate a maximum flow rate of 1 nl/hour. The dependence appears to be linear and independent of direction, as expected from the laminar nature of the flow (Fig. 2). From the measured tracer particle velocities, we estimate that pressure drops of ∼30 μm H2O are achievable at our fastest rotation rates.

Figure 2

Tracer particle velocity as a function of gear rotation rate for both forward and reverse directions. The line is a fit to guide the eye.

The gear pump design illustrates the success of positive-displacement pumping through the use of colloidal microspheres; however, its design may prove particularly harsh to certain solutions. Although we have been able to pump individual cells using the gear pump (Movie S2), concentrated cellular suspensions may be damaged by the aggressive motion of the meshing “gears” of the pump. We have therefore investigated a second approach that incorporates a peristaltic design also based on the concept of positive fluid displacement, effectively a pseudo two-dimensional analog of a three-dimensional, macroscopic screw pump. If instead of rotating the particles as in the gear pump, they are translated back and forth across the channel in a cooperative manner, fluid propagation can be achieved.

The colloidal movement required to direct flow with this approach is illustrated in Fig. 3. The optical trap moves the colloids in a propagating sine wave within which a plug of fluid is encased. Direction of the flow can be reversed by changing the direction of colloidal wave movement. Again, these experiments were performed with independent, 3-μm silica spheres; however, a larger number of particles was used in the experiments shown in Fig. 3 to represent a complete wavelength. Fabrication of these pumps required first maneuvering the colloids into the channel section. Once in place, the optical trap was scanned such that multiple independent traps were created, one for each colloid composing the peristaltic pump. Tracer particles were also used in these experiments and indicate that comparable flow rates could be achieved with this approach. The “snakelike” motion of this pump is best viewed as movie clips (Movies S3 and S4).

Figure 3

Three-micrometer colloidal silica used as a peristaltic pump, operating at 2 Hz, to induce flow from right to left within a 6-μm channel. Frames are separated by four cycles to show movement of the 1.5-μm tracer particle.

The physical, colloid-based, in situ positive-displacement pumping scheme of these two pumps has a number of advantages in addition to its diminutive size. Because colloidal particles are used, and depending on the design, the actuation scheme could be electrophoretic, magnetophoretic, or optical based. This range of actuation schemes will allow pumping of complex suspensions and nonpolar organic solvents, two fluid classes in which electrophoretic pumping techniques falter or fail.

To restrict and direct the flow of cells or colloids within microfluidic networks, we created two types of valve (Fig. 4) by using laser-initiated photopolymerization to first lock colloids into specific geometries. Once polymerized, these structures are positioned and, in some cases, actuated by the same laser used for their construction. Each valve consists of a 3-μm silica sphere photopolymerized to several 0.64-μm silica spheres that form a linear structure. For passive applications, the device was maneuvered into a straight channel and the 3-μm sphere was held next to the wall, allowing the arm to rotate freely in the microchannel. As the flow direction was changed (Fig. 4A), the valve selectively restricted the flow of large particles in one direction while allowing passage of all particles in the other. To actively direct particulates to one of two exit channels, the passive valve was maneuvered into a confining “T” geometry. As the valve structure was rotated about its swivel point by using the optical trap, the top or bottom channel was sealed, directing flow of particulates toward the open channel (Fig. 4B) (Movies S5 and S6).

Figure 4

(A) A passive colloidal valve, where arrows indicate the direction of fluid flow. In images 1 to 3, a flow of ∼2 nl/hour pushes the valve arm against the channel wall, allowing particulates in the flow to pass. In images 4 to 6, the flow is reversed, swinging the valve structure across the channel, restricting the flow of the 3-μm colloids but not the smaller 1.5-μm tracer particles. The aqueous solution used to photopolymerize the valve structure consisted of an aqueous monomer, 1.99 M acrylamide (Sigma-Aldrich) and cross-linker solution, 0.048 M bis-acrylamide (Sigma-Aldrich), containing both a photoinitiator, 0.0006 M triethanolamine (J. T. Baker, Phillipsburg, NJ), and a coinitiator, 0.0002 M riboflavin 5′ phosphate sodium salt (Sigma-Aldrich). This solution was pumped into the microfluidic channel network through a syringe pump. Initially fabricated in a deeper region of the network, the linear structures were optically maneuvered into an 11-μm-wide and 3.2-μm-deep straight section of the PDMS microfluidic channel. (B) An actuated, three-way colloidal valve. Images 1 to 3 show the valve structure in a position to direct 3-μm spheres to the lower channel with flow rates of ∼2 nl/hour. In images 4 to 6, the valve is moved downward to direct particulates to the upper channel.

We have shown here that colloidal particles can be used to fabricate true micrometer-scale microfluidic pumps and valves that are much smaller than those constructed with current approaches (16, 27, 28). Although the use of an optical trap provides a number of advantages, including the elimination of physical connection to macroscopic hardware and the ability to instantly alter device design or location in situ, actuation of these devices via other applied fields is certainly feasible. As discussed previously, appropriately selected colloids will also translate in applied electric and magnetic fields. Because of its versatility, a colloid-based approach to microfluidic flow generation and control may indeed prove a powerful technique for the creation of complex, highly integrated, microscale total analysis systems.

  • * To whom correspondence should be addressed. E-mail: joakey{at} (J.O.); dmarr{at} (D.W.M.M.)


Stay Connected to Science

Navigate This Article