Acoustic Particle Filter with Adjustable Effective ... - ACS Publications

This article presents analysis and optimization of a microfluidic particle filter that uses acoustic radiation forces to remove particles larger than ...
0 downloads 0 Views 457KB Size
Download PDF
Anal. Chem. 2008, 80, 8447–8452

Acoustic Particle Filter with Adjustable Effective Pore Size for Automated Sample Preparation Byoungsok Jung,* Karl Fisher, Kevin D. Ness, Klint A. Rose, and Raymond P. Mariella, Jr. Lawrence Livermore National Laboratory (LLNL), Livermore, California 94551 This article presents analysis and optimization of a microfluidic particle filter that uses acoustic radiation forces to remove particles larger than a selected size by adjusting the driving conditions of the piezoelectric transducer (PZT). Operationally, the acoustic filter concentrates microparticles to the center of the microchannel, minimizing undesirable particle adsorption to the microchannel walls. Finite element models predict the complex twodimensional acoustic radiation force field perpendicular to the flow direction in microfluidic devices. We compare these results with experimental parametric studies including variations of the PZT driving frequencies and voltages as well as various particle sizes (0.5-5.0 µm in diameter). These results provide insight into the optimal operating conditions and show the efficacy of our device as a filter with an adjustable effective pore size. We demonstrate the separation of Saccharomyces cerevisiae from MS2 bacteriophage using our acoustic device. With optimized design of our microfluidic flow system, we achieved yields of greater than 90% for the MS2 with greater than 80% removal of the S. cerevisiae in this continuous-flow sample preparation device. The development of micro total analysis systems (µTAS) over the last two decades has provided innovative advances in analytical chemistry and biochemistry, clinical diagnostics, and biochemical threat detection.1,2 However, handling real-world, complex samples (e.g., blood, saliva, nasal washes, seawater, etc.) in laboratories and clinics often requires slow, manual sample purification or preconcentration techniques.3,4 These samples typically contain high concentrations of particles larger than 2 µm (cells, pollen, debris, etc.), which must be removed to analyze smaller targets such as bacteria or viruses. Filtration through a high surface area filter material is one of the most common techniques for these samples and is generally effective for removing particles as small as 100 nm. Researchers who attempt detection of potentially low titer targets (e.g., viruses for viral discovery), however, prefer to avoid these filters since there is a significant risk that the targets will adhere to the filter material and, therefore, escape detection * To whom correspondence should be addressed. E-mail: [email protected]. Fax: (925) 424-2778. (1) Auroux, P. A.; Iossifidis, D.; Reyes, D. R.; Manz, A. Anal. Chem. 2002, 74, 2637–2652. (2) Reyes, D. R.; Iossifidis, D.; Auroux, P. A.; Manz, A. Anal. Chem. 2002, 74, 2623–2636. (3) Ramsey, J. M. Nat. Biotechnol. 1999, 17, 1061–1062. (4) Crevillen, A. G.; Hervas, M.; Lopez, M. A.; Gonzalez, M. C.; Escarpa, A. Talanta 2007, 74, 342–357. 10.1021/ac8011768 CCC: $40.75  2008 American Chemical Society Published on Web 10/11/2008

by even high-performance assays.5 Thus, the ideal filtration system for real-world samples should enable high-throughput extraction of large particles (>2 µm) while reducing the filter surface area to minimize losses due to adsorption. Several approaches for accomplishing sample preparation within µTAS platforms6 have been investigated including on-chip filtration,7 microdialysis,8 affinity-based extraction,9 immunomagnetic beads,10 and acoustic focusing.11 Of these approaches, acoustic focusing offers a robust and high-throughput separation for real-world samples because the relatively low surface area filtering reduces clogging and surface adsorption. Acoustic focusing is particularly effective for manipulating relatively large (>2 µm) particles suspended in liquid. Previously demonstrated applications leveraging acoustic radiation forces include the following: acoustic control of particles,12,13 filtration or trapping of cells,14-16 separation of lipids from blood,17 and continuous separation of mixed microparticles or cells.18-20 Acoustic particle filters that maintain eukaryotic cell viability21 are also used in a variety of large-scale applications.22,23 Acoustic focusing in microfluidic devices is typically realized using a piezoelectric transducer (PZT) to generate acoustic (5) DeRisi, J. Private communication, 2008. (6) Mariella, R., Jr. Biomed. Microdevices. In press. (7) Yuen, P. K.; Kricka, L. J.; Fortina, P.; Panaro, N. J.; Sakazume, T.; Wilding, P. Genome Res. 2001, 11, 405–412. (8) Xu, N. X.; Lin, Y. H.; Hofstadler, S. A.; Matson, D.; Call, C. J.; Smith, R. D. Anal. Chem. 1998, 70, 3553–3556. (9) Yu, C.; Davey, M. H.; Svec, F.; Frechet, J. M. J. Anal. Chem. 2001, 73, 5088–5096. (10) Liu, R. H.; Yang, J. N.; Lenigk, R.; Bonanno, J.; Grodzinski, P. Anal. Chem. 2004, 76, 1824–1831. (11) Hawkes, J. J.; Limaye, M. S.; Coakley, W. T. J. Appl. Microbiol. 1997, 82, 39–47. (12) Nilsson, A.; Petersson, F.; Jonsson, H.; Laurell, T. Lab Chip 2004, 4, 131– 135. (13) Oberti, S.; Neild, A.; Dual, J. J. Acoust. Soc. Am. 2007, 121, 778–785. (14) Evander, M.; Johansson, L.; Lilliehorn, T.; Piskur, J.; Lindvall, M.; Johansson, S.; Almqvist, M.; Laurell, T.; Nilsson, J. Anal. Chem. 2007, 79, 2984–2991. (15) Hawkes, J. J.; Coakley, W. T. Sens. Actuators, B: Chem. 2001, 75, 213– 222. (16) Yasuda, K.; Haupt, S. S.; Umemura, S.; Yagi, T.; Nishida, M.; Shibata, Y. J. Acoust. Soc. Am. 1997, 102, 642–645. (17) Petersson, F.; Nilsson, A.; Holm, C.; Jonsson, H.; Laurell, T. Analyst 2004, 129, 938–943. (18) Hwang, S. H.; Koo, Y. M. Biotechnol. Lett. 2003, 25, 345–348. (19) Kapishnikov. S.; Kantsler, V.; Steinberg, V. J. Stat. Mech.: Theory Exp. 2006, P01012. (20) Petersson, F.; Aberg, L.; Sward-Nilsson, A. M.; Laurell, T. Anal. Chem. 2007, 79, 5117–5123. (21) Radel, S.; McLoughlin, A. J.; Gherardini, L.; Doblhoff-Dier, O.; Benes, E. Ultrasonics 2000, 38, 633–637. (22) Crowley, J. BioProcess Int. 2004, 46-49. (23) Gorenflo, V. M.; Ritter, J. B.; Aeschliman, D. S.; Drouin, H.; Bowen, B. D.; Piret, J. M. Biotechnol. Bioeng. 2005, 90, 746–753.

Analytical Chemistry, Vol. 80, No. 22, November 15, 2008

8447

standing waves within the microchannel. These standing waves produce acoustic radiation force fields that direct microparticles toward the nodes (i.e., pressure minimums) or the antinodes (i.e., pressure maximums) of the standing waves depending on the relative compressibility and density between the particle and the suspending liquid.24 For particles larger than 2 µm, the transverse velocities induced in response to these force fields are sufficient to achieve continuous, high-throughput separation. In this work, we develop a microfluidic particle filter that uses acoustic focusing to remove particles above a selected size by adjusting the driving conditions of the transducer. Using the theory described in the following section, we developed a finite element modeling tool to predict the two-dimensional acoustic radiation force field perpendicular to the flow direction in microfluidic devices. We compare results from this model with experimental parametric studies including variations of the PZT driving frequencies and voltages as well as multiple particle sizes. As a demonstration of this rapid and robust filtering technique, we separated eukaryotic cells (Saccharomyces cerevisiae) from viruses (MS2 bacteriophage) with operating conditions optimized via our experimental parametric studies. These results help demonstrate the capability to extend acoustic focusing to more complex samples. THEORY AND MODELING Typical solutions for the acoustic radiation force in a microfluidic channel use one-dimensional approximations for the acoustic standing waves.25 Microfluidic channel geometries, however, tend to be on the order of the structural wavelength (∼1 mm) for the operational frequencies (∼1 MHz) in these devices. Modeling approaches that consider two- and three-dimensional behavior for the fluid-to-structure interaction of the system are therefore necessary to accurately capture the dynamic behavior of the particles. To predict the motion of particles, we apply a method described by Fisher and Miles24 that combines a finite-difference approximation for the acoustic radiation force with a finite-element model of the pressure field within the fluid. This 2-D analysis is a significant improvement over existing 1-D solutions16,25 for predicting device performance. This approach facilitates prediction of operational frequencies based on acoustic energy density and corresponding focusing location within the 2-D acoustic radiation force field. We used the multiphysics finite element package ATILA (Magsoft, Ballston Spa, NY) to model a 2-D cross section of the fluid, silicon, glass, and piezoelectric. The forces on the particles are first obtained by calculating the acoustic pressure field in the fluid generated by the piezoelectric transducer. The finite-element model provides the solution for the pressure field in the fluid and the displacement field in the elastic and piezoelectric structure. To determine the acoustic radiation force in the fluid from the pressure field solution, we used the theoretical acoustic radiation force on a sphere first described by King26 and extended by Gor’kov25 and by Nyborg.27 These solutions apply to spherical particles assuming the radius of the particle, a, is small compared to k ) 2π/λf, where λf is the acoustic wavelength in the fluid. (24) Fisher, K.; Miles, R. J. Acoust. Soc. Am. 2008, 123, 1100–1104. (25) Gor’kov, L. P. Soviet Phys. Dokl. 1962, 6, 773–775. (26) King, L. V. Proc. R. Soc. London, Ser. A: Math. Phys. Sci. 1934, 147, 212– 240. (27) Nyborg, W. L. J. Acoust. Soc. Am. 1967, 42, 947–952.

8448

Analytical Chemistry, Vol. 80, No. 22, November 15, 2008

Figure 1. Schematic of the microfluidic chip. The microchannel is designed as an H-filter with two inlets and two outlets. The crosssectional width and depth of the main separation channel are respectively 500 and 200 µm and the length of the main channel is 31 mm. The PZT is bonded to the backside of silicon layer.

For ka , 1, the generalized force field on a spherical particle due to the acoustic standing wave pattern in the fluid is27 4 F ) πa3{B(∆〈KEa 〉) - (1 - γ)(∆〈PEa 〉)} 3

(1)

The force, F, is a combination of the gradients of the timeaveraged kinetic energy, 〈KEa〉, and time-averaged potential energy, 〈PEa〉. The contribution of the energies to the overall force is weighted by B ) 3(F - Fo)/(2F + Fo), a ratio of the fluid density F and particle density Fo, and γ ) β/βo, where β and βo are the bulk compressibility of the fluid and particle, respectively. Note that the force is proportional to the particle volume, 4/3πa3. From the FEM solution for the pressure field, we solve the force field using finite difference approximations for the kinetic and potential energy in terms of the pressure field. (See ref 24 for additional details.) Assuming negligible inertial effects (i.e., Reynolds number ,1), particles within the acoustic force field reach terminal velocity nearly instantaneously. The force field therefore directly predicts particle motion and the resulting focusing location. The balance between the acoustic radiation force, Fa, and the viscous drag, Fd ) 6πµUa, determines the particle velocity. Therefore, the resultant velocity of a particle is given by U ) Fa/6πµa, where µ is the viscosity of the liquid. EXPERIMENTAL SECTION Our setup for the acoustic particle filtering experiments includes a microfluidic chip as well as supporting benchtop instrumentation to drive the acoustics, manipulate the fluids, and optically detect the results. Two figures of merit, focusing efficiency and separation efficiency, are introduced to quantitatively analyze the device performance. Microfluidic Chip. The microfluidic chip has three layers consisting of 500-µm-thick borosilicate glass, 525-µm-thick silicon wafer, and a ceramic piezoelectric transducer. We patterned the microchannels as an H-filter28 (Figure 1) with two inlets and two outlets using standard photolithography techniques and etched them into the silicon to depths of 200 µm using DRIE (ASE (28) Yager, P.; Weigl, B. H.; Brody, J. P.; Holl, M. R. University of Washington, 1998.

system; Surface Technology Systems). The width and depth of the main separation channel were respectively 500 and 200 µm for a channel volume of ∼3 µL. We etched the through-ports from the wafer backside through to the front-side channels using DRIE. An anodic bond between the borosilicate glass and the etched silicon sealed the microfluidic channels. Using cyanoacrylate glue, we mounted the PZT (T120-A4E-602; Piezo Systems, Cambridge, MA) to the backside of the silicon layer to drive the acoustic structure. Instrumentation and Materials. The acoustics instrumentation consists of a waveform generator (model 33220A; Agilent Technologies, Santa Clara, CA) and an rf power amplifier (model 325LA; EIN) to drive the piezoelectric transducer. The fluidics instrumentation includes multiposition valves and two-position valves (Valco Instruments, Houston, TX) for automated sequential injection analysis capability, large-volume (1 mL) syringe pumps (Cavro pump; Tecan, San Jose, CA) for priming and decontamination, and an additional syringe pump (PHD 2000; Harvard Apparatus, Holliston, MA) for accurate flow control during filtering. The detection system includes an epifluorescence microscope (Eclipse 80i; Nikon, Melville, NY) fitted with a wide field-of-view objective (4×, NA of 0.1) and a CCD camera (CoolSNAP fx; Photometrics, Tucson, AZ). We control all equipment using an in-house Labview program to enable repeatable, automated experiments. We used red (emission peak 605 nm) and green (emission peak 515 nm) polystyrene fluorescent microparticles (Invitrogen, Carlsbad, CA). The microparticles were diluted in deionized water to concentrations of 2.2 × 1013, 2.9 × 109, 4.5 × 108, 1.1 × 108, 1.2 × 108, and 2.9 × 107 particles/mL for, respectively, microparticles with diameters of 0.5, 1.0, 2.0, 3.1, and 5.0 µm. Using a BacLight Red bacterial stain kit (Invitrogen, Carlsbad, CA), we labeled commercial S. cerevisiae (baker’s yeast). The MS2 bacteriophage was obtained from the Critical Reagents Program,29 propagated in Escherichia coli K12 bacteria, and purified through centrifugation. The S. cerevisiae has a typical cell size of 4-6 µm depending on the cell growth stage as measured using a Coulter counter. MS2 bacteriophage has a diameter of ∼30 nm.30 The concentration of S. cerevisiae and MS2 used in the acoustic filter experiment was ∼107 cells/mL and ∼106 pfu/mL, respectively. Experimental Protocol. For each experiment, we pump 5-10 µL of sample through the main separation channel of the acoustic filter at a flow rate of 20 µL/min. For performance characterization experiments, we pump sample through both inlet channels filling the entire channel with sample. To characterize the acoustic separation efficiency, we pump sample through only one inlet channel and deionized water through the other inlet setting up an interface along the center line of the main separation channel (Figure 1). For all experiments, a CCD camera acquired fluorescence images, normalized by flat-field and dark-field images of the microchannel. The images correspond to a viewing area of 1.0 by 0.5 mm in the object pane at a point 1 mm upstream from the outlet bifurcation. We used an ensemble average of 20 sequential (29) Hawkes, J. J.; Limaye, M. S.; Coakley, W. T., IEEE Ultrasonics Symposium, 1995, vol. 2, 1069-1072. (30) Anobom, C. D.; Albuquerque, S. C.; Albernaz, F. P.; Oliveira, A. C.; Silva, J. L.; Peabody, D. S.; Valente, A. P.; Almeida, F. C. L. Biophys. J. 2003, 84, 3894–3903.

Figure 2. Typical acoustic focusing image for microparticles in our device and the resulting intensity distribution. (a) Fluorescence image of focused microparticles (diameter, 2.0 µm) with a driving frequency of 1.459 MHz and voltage of 6.60 V. The particle-laden solution initially fills the entire width of the channel (y-direction) as the particles flow from left to right at 20 µL/min. The microparticles tightly focus (full width half-maximum ∼35 µm) in less than 20 s. (b) Intensity profile for the focusing example in (a). The intensity is calculated across the microchannel width (y-direction) by averaging intensity along x-direction.

images (∆t ) 100 ms) to increase the signal-to-noise ratio (Figure 2a) and also averaged the image pixels along the microchannel direction (x-direction) to obtain the fluorescence intensity profile across the microchannel (y-direction). Figures of Merit in Acoustic Focusing. We define the focusing efficiency of our acoustic system as the ratio of the relative width of the focused sample zone (Wp) compared to the width of microchannel (Wc) such that the nondimensional characteristic peak width is δ1. The focused sample zone width, Wp, is the distance between two points along the y-axis, symmetric across the center of the microchannel, within which the integral of intensity (i.e., concentration) is equal to half the total intensity value. We calculate the value for Wp by expanding the integration bounds in the integral

∫

I (y) Wp p

dy ) 0.5Itot

(2)

until the equality is satisfied. In eq 2, Ip(y) is the fluorescence intensity distribution, and Itot is the total intensity value of the sample integrated across the entire channel. δ1 is unity for a uniformly distributed sample (e.g., no focusing) and is near zero for a tightly focused sample zone in a single node at the center of the microchannel. Figure 2a shows a typical image of focused microparticles (diameter, 2.0 µm) in a 500-µm-wide microchannel, and Figure 2b shows the corresponding intensity distribution and characteristic width. The width of the focused sample in the example is 70 µm and the characteristic peak width is 0.28. We introduce a second figure of merit, η80, to quantify the efficiency of separation using our acoustic focusing technique. The separation efficiency, η80, is the percentage of small particles that Analytical Chemistry, Vol. 80, No. 22, November 15, 2008

8449

Figure 3. Two-dimensional numerical estimate of the acoustic pressure field and the acoustic radiation force field shown for a cross section of the microchannel. The dark regions and the light regions represent nodes (low acoustic pressure) and antinodes (high acoustic pressure) in the acoustic pressure field, respectively. The driving frequency was set to 1.4809 MHz for the geometry described in Figure 1. The acoustic radiation force field (vector plot) predicts strong acoustic focusing at the center of the microchannel.

are not within the width W80. Similar to Wp, we define the width W80 as the distance between two points along the y-axis, symmetric across the center of the microchannel, within which the integral of intensity (i.e., concentration) is equal to 80% the total intensity value for the large particles in the system. Therefore, high η80 refers to high sample purity (i.e., good separation) of small particles versus the large filtered particles. RESULTS AND DISCUSSION We simulated the microchannel geometry and operating conditions from the experiments (described above) and compared these vector plots with the experimental focusing results. We also analyzed our experimental parametric studies and calculated the resulting focusing efficiency to show the optimal operating conditions for our system and validate our modeling. Using these optimization results, we determined the separation efficiency for our system by separating virus from yeast. The surface plot in Figure 3 shows the simulation results for the two-dimensional pressure field orthogonal to the flow direction. The dark regions and the light regions represent the nodes and antinodes of the acoustic pressure field, respectively. The superimposed vector plot shows the acoustic radiation force in this plane. The asymmetry of force vectors in the z-direction is specific to the selected operating frequency and is due to the different material properties (e.g., Young’s modulus) of the microchannel wall materials (Si and glass). The corresponding force field predicts acoustic focusing at the center of the microchannel, which is confirmed by the experimental results shown in Figure 2a. For the experimental parametric studies, we mainly focused on three key parameters: driving frequency, driving voltage, and particle size. The acoustic focusing reached steady state within 20 s after activating the PZT. After achieving steady state, we recorded three sets of 20 images for each operating condition with an exposure time and frame rate of 10 ms and 10 fps, respectively. We varied the driving frequency from 1.009 to 1.999 MHz with a 10-kHz discretization, the driving voltage from 0 to 9.24 V with steps of 0.44 V, and used 5 different sizes of microparticles (0.5, 1.0, 2.0, 3.1, and 5.0 µm in diameter). Figure 4a shows acoustic focusing efficiency (1 - δ1) as a function of PZT driving frequency for a 3.1-µm particle size. The 8450

Analytical Chemistry, Vol. 80, No. 22, November 15, 2008

Figure 4. (a) Acoustic focusing efficiency as a function of driving frequency for 3.1-µm particles. The driving voltage applied to the PZT was 4.76 V, and the flow rate was fixed at 20 µL/min. (b) Acoustic focusing efficiency as a function of driving voltage. Root-mean-square voltage applied to the PZT (Vrms) was varied from 0 to 9.25 V. The symbols represent data for microbeads with a diameter of 0.5 (white circle), 1.0 (gray triangle), 2.0 (white square), 3.1 (gray circle), and 5.0 µm (white triangle) as well as Baclight-labeled yeast (black diamond). The driving frequency was 1.459 MHz, and flow rate was 20 µL/min.

error bars reflect 90% confidence intervals as determined from three independent realizations of each condition. High acoustic focusing efficiency (1 - δ1), thereby, represents a tightly focused sample zone. The acoustic focusing efficiency in Figure 4a has a maximum value of 0.88 at 1.459 MHz; selected as the optimal operating frequency for this system. The second maximum occurs at 1.409 MHz, but the acoustic focusing efficiency is limited to 0.75. Distinct focusing (1 - δ1 < 0.5) occurred only at a frequency range between 1.389 and 1.489 MHz for a single-node focusing. While the current 2-D model predicts focusing to a single node over this range, it does not provide a quantitative measure of the focusing efficiency. To provide even greater predictive analysis in the future, we are currently developing more accurate models,

Figure 5. Sample separation efficiency for various microparticle combinations and a mixture of yeast and MS2. Separation efficiency (η80) represents the percentage of background (smaller) particles outside of focused zone of target (larger) particles. Driving voltages applied to a PZT were varied from 1.02 to 3.83 V, and driving frequency was fixed at 1.459 MHz.

which include three-dimensional acoustic pressure and acoustic radiation force field calculations. Figure 4b shows the effect of driving voltage and particle size on the acoustic focusing efficiency at a frequency of 1.459 MHz. δ1 generally decreases as the driving voltage increases but plateaus at larger driving voltages for particles with a diameter 2.0 µm or greater. The plateau values are different for each diameter of particle (large particles have lower terminal δ1). We attribute this to the balance between diffusive dispersion and acoustic radiation force. The 4-6-µm yeast cells showed focusing trends comparable to polystyrene microparticles with diameter between 2 and 3 µm due to the lower compressibility of yeast cells despite similar density to the polystyrene microparticles. The results shown in Figure 4b suggest that a size cutoff lies between 1 and 2 µm, where particles do not tightly focus (δ1 < 0.4) even for high driving voltages. The optimal driving voltages (i.e., the lowest voltages at which maximum focusing occurs) also vary for different particle size. Therefore, our system may be used as a high-throughput, clogless, tunable size filter. For example, 5.0-µm microparticles can be separated from 2.0-µm microparticles with a driving voltage of 1.93 V. Figure 5 shows separation efficiency (η80) as a function of driving voltage for various microparticle mixtures and a mixture of yeast and MS2. To demonstrate particle separations, we used microparticle mixtures with six size combinations including the following: 5 µm and 20 nm, 5 and 0.5 µm, 5 and 2.0 µm, 3.1 µm and 20 nm, 3.1 and 0.5 µm, and 3.1 and 2.0 µm. In each pair, the small and large microbeads were respectively red (emission peak 580 nm) and green (emission peak 515 nm) fluorescent particles. We varied the driving voltages from 1.02 to 3.83 V and fixed the driving frequency at 1.459 MHz. For the three sets of 5-µm mixtures (5 µm and 20 nm, 5 and 0.5 µm, and 5 and 2.0 µm), the ratio in diameters between the 5-µm particles targeted by the acoustic filter and smaller background particles is a factor of 2.5 or larger. These combinations show good separation (80% removal of large particles with greater than 80% recovery of small particles) at 2.89 V. The mixture of 3.1 µm and 20 nm, and the mixture of 3.1 and 0.5 µm, also show moderate separation (80% removal of large particles with ∼70%

Figure 6. Separations of a mixture of yeast cells and MS2 viruses. Concentration profiles (normalized intensity) across the microchannel width (y-direction) are shown for the driving voltages from 1.96 to 4.76 V. The driving frequency and flow rate were 1.459 MHz and 20 µL/min, respectively. Samples were initially introduced into only half of the microchannel and initially flow parallel to deionized water in the other half of the channel.

recovery of small particles) at 3.83 V. However, separation efficiency of a mixture with the smallest size difference (3.1 and 2.0 µm, diameter ratio of 1.55) is limited to 51% at 2.89 V and does not improve with higher acoustic focusing power. This result demonstrates the separation limit for our system. Improvement may be possible using different channel geometries.29 In Figure 6, we demonstrate the separation of S. cerevisiae from MS2 bacteriophage. We prepared a mixture of S. cerevisiae and MS2 and injected the sample into one inlet of the microchip and deionized water into the other so that the sample filled initially half of the microchannel. We varied driving voltages from 1.96 to 4.76 V, while fixing the driving frequency at 1.459 MHz and flow rate at 20 µL/min. We selectively imaged each particle type by switching between elastic scattered light microscopy (yeast) and fluorescence microscopy (MS2). Each of the lines in Figure 6 represents the average signal intensity profile for the yeast cells and viruses over three realizations. The acoustic radiation force did not affect the MS2 viruses, and their concentration profile remained unchanged. Increased driving voltages enhanced the acoustic focusing of the yeast cells, thereby achieving good separation. We are able to achieve yields of >80% and sample purities of >90% in this continuous-flow sample preparation device. The separation efficiency (η80) is 90% at a driving voltage of 3.86 V. SUMMARY Acoustic focusing provides rapid and robust manipulation of suspended particles and can be easily integrated with other µTAS functionalities. Our two-dimensional numerical simulation using FEM yields a first-order estimate of operation conditions for a PZT and predicts the location of focused particles in the microchannel. We performed an experimental parametric study focused on the variation of particle size, driving voltage, and driving frequency. Consistent with the simulation, we found that the polystyrene microparticles and yeast cells focus at the nodes of acoustic standing waves located at the center of the microchannel. We also Analytical Chemistry, Vol. 80, No. 22, November 15, 2008

8451

found that particles larger than 2 µm in diameter focused effectively, demonstrating the application of our device as a clogless filter for large particles (e.g., cells, debris) from complex samples containing smaller biological particles or molecules of interest. For fixed channel dimensions and resonant frequency, the focusing efficiency of particles depends on the driving voltage as well as particle size. These results give insight into a filter with an adjustable effective pore size where the effective pore size can be easily varied by adjusting the driving voltage of PZT. We demonstrated the separation of yeast cells from MS2 viruses with maximum recovery of 90% for the MS2 with 80% removal of yeast cells. We also analyzed the separation of mixtures of microparticles by adjusting the driving voltage, and the results show good separation of cell-sized particles (5 µm) against viruslike (20 nm) and bacteria-like (0.5 µm) particles with effective size discrimination for diameter ratios greater than 2.5. Future work will include the development and validation of more comprehensive models for the acoustic focusing and further

8452

Analytical Chemistry, Vol. 80, No. 22, November 15, 2008

applications of our acoustic focusing protocol to real-world biological samples. ACKNOWLEDGMENT This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. The authors thank Brian Harrel and Shanavaz Nasarabadi for the preparation of S. cerevisiae and MS2 virus, Julie Hamilton for the fabrication of the microchip, and William Benett for the design of the microfluidic packaging. R.P.M. thanks Drs. Martin Gro¨schl and Greg Kaduchak for kindly providing information and advice on this topic.

Received for review June 10, 2008. Accepted August 23, 2008. AC8011768