Microfluidic Actuation Using Electrochemically Generated Bubbles

Anal. Chem. 2002, 74, 6392-6396 Technical Notes Microfluidic Actuation Using Electrochemically Generated Bubbles Susan Z. Hua,*, Frederick Sachs, David X. Yang, and Harsh Deep Chopra Center for Bio-MEMS, School of Medicine, State University of New York at Buffalo, Buffalo, New York 14260, Department of Physiology & Biophysics, 320 Cary Hall, State University of New York at Buffalo, Buffalo, New York 14214, and Thin Films & Nanosynthesis Laboratory, Materials Program, Mechanical and Aerospace Engineering Department, State University of New York at Buffalo, Buffalo, New York 14260 Bubble-based actuation in microfluidic applications is attractive owing to elementary microfabrication requirements. In the present study, the mechanical and chemical characteristics of electrochemically generated bubble valves were studied. By generating electrochemical bubbles as valves directly inside the channel, valves could be closed and opened in milliseconds. Whereas bubble inflation (or valve closing) rate increases with applied voltage, small microfluidic dimensions accelerate bubble deflation rates. It is found that bubbles need not collapse fully to restore full flow, and the channel opens when its hydraulic resistance equals that between the bubble and the wallsa process requiring only milliseconds. Since only picomoles of salt are needed to generate bubbles, ph gradients that are invariably associated with electrochemical reactions were readily suppressed by using a small amount of buffer, as visualized by a ph-sensitive fluorescent dye. A range of common laboratory reagents and electrolytes in varying concentrations, including weak to strong acids and bases, as well as nonaqueous/aqueous mixtures were successfully tested. Using such bubble valves, an eight-way multiplexer was fabricated and tested. * Corresponding author. Phone: (716) 645-2593 ext. 2358. Fax: (716) 645-3875. E-mail: zhua@eng.buffalo.edu. Center for Bio-MEMS, School of Medicine. Department of Physiology & Biophysics. Thin Films & Nanosynthesis Laboratory, Materials Program, Mechanical and Aerospace Engineering Department. (1) Burns, M. A.; Johnson, B. N.; Brahmasandra, S. N.; Handique, K.; Webster, J. R.; Krishnan, M.; Sammarco, T. S.; Man, P. M.; Jones, D.; Heldsinger, D.; Mastrangelo, C. H.; Burke, D. T. Science 1998, 282, 484-487. (2) Zhao, B.; Moore, J. S.; Beebe, D. J. Science 2001, 291, 1023-1026. (3) Gallardo, B. S.; Gupta, V. K.; Eagerton, F. D.; Jong, L. I.; Craig, V. S.; Shah, R. R.; Abbott, N. L. Science 1999, 283, 57-60. (4) Hatch, A.; Kamholz, A. E.; Hawkins, K. R.; Munson, M. S.; Schilling, E. A.; Weigl, B. H.; Yager, P. Nat. Biotechnol. 2001,19, 461-465. Microelectromechanical systems continue to spawn new technological applications in addition to being catalysts for key scientific discoveries. Intense efforts are currently underway to develop multifunctional microfluidic chips, a technology commonly referred to as lab-on-a-chip. 1-5 Applications range from combinatorial and analytical chemistry and drug discovery, to microbiology, biotechnology, and drug delivery. To regulate fluid flow through labyrinthine microfluidic channels using pumps and valves, various actuation mechanisms, based on piezoelectricity, 6 electrostatics, 7 thermopneumatics, 8,9 and electromagnetism 10 have been developed, along with advances in microfabrication, for example, soft lithography. 11-14 Bubble-based actuators are of interest since they are simple to fabricate and the bubbles have an ability to readily conform physically to different channel cross sections. Both thermal and electrochemically generated bubbles have been used. 15-19 While electrochemical bubbles require low power in the microwatt range, and the bubble inflation rates are comparable to thermal bubbles, 20-23 their use has been limited by slow bubble deflation rates, since the dissolution of gas into the fluid is kinetics-limited. In the present study, it was found that, by generating bubbles as (5) Gau, H.; Herminghaus, S.; Lenz, P.; Lipowsky, R. Science 1999, 283, 46-49. (6) Andersson, H.; van der Wijngaart, W.; Nilsson, P.; Enoksson, P.; Stemme, G. Sens. Actuators, B 2001, 72, 259-265. (7) Francais, O.; Dufour, I. Sens. Actuators, A 1998, 70, 56-60. (8) Van de Pol, F. C. M.; Van Lintel, H. T. G.; Elwenspoek, M.; Fluitman, J. H. J. Sens. Actuators, A 1990, 21, 198-202. (9) Kataoka, D. E.; Troian, S. M. Nature 1999, 402, 794-797. (10) Feustel, A.; Krusemark, O.; Lehmann, U.; Müller, J.; Sperling, T. Actuator 96, Bremen, Germany, 1996; p 76. (11) Unger, M. A.; Chou, H. P.; Thorsen, T.; Scherer, A.; Quake, S. R. Science 2000, 288, 113-116. (12) Quake, S. R.; Scherer, A. Science 2000, 290, 1536-1539. (13) Whitesides, G. M.; Stroock, A. D. Phys. Today 2001, 42, 41. (14) Beebe, D. J.; Moore, J. S.; Bauer, J. M.; Yu, Q.; Liu, R. H.; Devadoss, C.; Jo, B. H. Nature 2000, 404, 588-590. (15) Tsai, J. H.; Lin, L. Sens. Actuators, A 2002, 1-7, 3325-3331. (16) Neagu, C.; Gardeniers, J. G. E.; Elwenspoek, M.; Kelly, J. J. Micromech. Syst. 1996, 5, 2-9. (17) Bohm, S.; Olthuis, W.; Bergveld, P. Proc. 10th Int. Conf. Micro Electro Mech. Syst. 2000, 92-95. (18) Papavasiliou, A. P.; Pisano, A. P.; Liepmann, D. Proc. 11th Int. Conf. Solid State Sens. Actuators 2001, 940-943. (19) Deshmukh, A. A., Liepmann, D.; Pisano, A. P. Proc. 11th Int. Conf. Solid State Sens. Actuators 2001, 950-953. (20) Geng, X.; Yuan, H.; Oguz, H. N.; Prosperetti, A. J. Micromech. Microeng. 2001, 11, 270-276. (21) Lin, L.; Pisano, A. P.; Carey, V. P. J. Heat Transfer Tran. ASME 1998, 120, 735-742. (22) Neagu, C. R.; Gardeniers, J. G. E.; Elwenspoek, M.; Kelly, J. J. Electrochim. Acta 1997, 42, 3367-3373. (23) Neagu, C.; Jansen, H.; Gardeniers, H.; Elwenspoek, M. Mechatronics 2000, 10, 571-581. 6392 Analytical Chemistry, Vol. 74, No. 24, December 15, 2002 10.1021/ac0259818 CCC: $22.00 2002 American Chemical Society Published on Web 11/13/2002

Figure 1. Micrographs A-D (run counterclockwise). (A) SEM micrograph of an electrochemical bubble valve chip. The channel connecting the reservoirs is 25 µm wide. (B 1-6) Optical micrographs showing bubble inflation and deflation. The flow direction is indicated by an arrow. The dark edges near the channel are an optical artifact. (C 1-6) Fluorescent microscopy images corresponding to micrographs B 1-6 showing valve closing and opening. In C 1-6, the interaction between the bubble and the flow was visualized using polystyrene fluorescent microspheres as tracers of flow (0.02-µm diameter, Nile Red F-8784, Molecular Probes). (D) Profile of the applied voltage pulse and the measured current through the electrodes. valves directly inside the microchannels, the effective deflation rates are increased. Here a bubble as a valve is first generated inside the channel to stop the fluid flow. This is achieved by an applied voltage pulse of arbitrarily small durations, milliseconds or shorter, and without the need for any moving mechanical parts. To restore flow, the bubble need not collapse fully since the only requirement to open the channel is that the hydraulic resistance of the fluid between the bubble and the channel wall be less than that of the open channel. Experimentally this has been found to occur at bubble diameters only slightly smaller than the channel cross section, a deflation process requiring only milliseconds. The small channel dimensions characteristic of a microfluidic chip are therefore well suited and enhance deflation rates because (1) the surface-to-volume ratio of the bubble increases with reduced dimensions, and (2) for a given interfacial tension, the internal bubble pressure increases, with decreasing channel dimensions. To characterize the performance of bubble valves, prototype microfluidic chips were built on silicon. Such valves were further used to construct an eight-way multiplexer to demonstrate the feasibility of this approach to build more complex chips. EXPERIMENTAL SECTION The microfluidic chips used to test the mechanical and chemical characteristics of bubble valves consisted of a fluid channel connecting an inlet and an outlet reservoir and anode/ cathode electrode pairs perpendicular to the channel to generate bubble valves at different locations. Figure 1A shows a scanning electron micrograph (SEM) of a portion of the channel (the inlet and outlet reservoirs are not shown) showing two sets of electrode pairs along the length of the channel. The system was micromachined on a silicon wafer using standard microfabrication techniques. The channel was 25 µm square in cross section and 5.2 mm long. Near one pair of electrodes, a 15-µm-wide neck was introduced to create a back pressure, although from experiments, it was subsequently found that surface forces alone were adequate and the neck was not needed to prevent the bubble from blowing away. Following photolithography, the channel was first etched to 25 µm in depth using deep reactive ion etching. Platinum electrodes were then deposited by e-beam deposition using a liftoff technique. The Pt electrodes were 300 nm thick and 25 µm wide. Finally, a poly(dimethylsiloxane) (PDMS) film (using Sylgard-184, Analytical Chemistry, Vol. 74, No. 24, December 15, 2002 6393

Corning) was used to cover and seal the etched channel. Silicone tubing with 0.3-mm inner diameter was placed within the PDMS film during the curing process, and this tubing was subsequently aligned on top of the inlet/outlet reservoirs. A syringe pump connected to a pressure reservoir perfused the channel with the electrolyte. The flow rate was adjusted by changing the inlet pressure while the outlet was kept at atmospheric. Whereas the experimental results in the following pertain to 1.0 M NaCl in distilled water (ph ) 6.4), various other common and useful laboratory reagents were also successfully tested. These included NaCl (0.1-1.0 M), weak acids (1.0 M acetic acid and oxalic acid), strong acids (0.1-1.0 M hydrochloric acid and sulfuric acid), and bases (10-5 -1.0 M NaOH), as well as nonaqueous/aqueous mixtures such as ethanol/water and acetonitrile/water in varying proportions (using analytical grade reagents). In particular, electrolysis of aqueous NaCl releases H 2 gas at the cathode (2H 2 O + 2e - f H 2 (g) + 2OH -, E )-0.83 V) and Cl 2 gas at the anode (2Cl - f Cl 2 (g) + 2e -, standard reduction potential E )+1.36 V), with an overall cell E )-2.19 V. 24,25 Note that water is always reduced in preference to Na + ions at the cathode since water accepts electrons more readily (E for Na + + e - being -2.71 V). A secondary reaction can occur as a small amount of Cl 2 gas reacts with water to form HOCl (hypochlorous acid) and HCl: Cl 2 (g) + H 2 O T H + + OCl - + H + + Cl -. 24,25 However, note that this secondary reaction is not an electrochemical reaction and does not rely on charge transfer occurring at the electrodes. Moreover the effect of this reaction is expected to be negligible because (1) it is very slow and (2) bubbles of useful sizes can be easily formed even in very dilute solutions and in a wide range of solution chemistries. In the present study, the voltage required to generate bubbles was 3.3 V instead of 2.19 V. The need for a slightly higher voltage (referred to as overvoltage in electrochemistry) is well known and is commonly observed due to nonequilibrium kinetics of electron transfer, especially when a gas phase is present. Hydrostatic pressures above 1 atm also favor the dissolution of gas into water, thereby increasing the overvoltage for the formation of Cl 2 gas. Bubble valves were characterized for voltages from 3.3 to 8.0 V. An arbitrary waveform function generator was used to apply a square voltage pulse across anode/cathode. The bubble valves characteristics were simultaneously observed with an optical microscope and video-recorded for later image analysis. An examination of video recordings of thousands of triggered bubbles at the Pt electrodes (on a given chip) showed no discernible degradation or damage to the electrodes (initially, the microfluidic chips were made using gold, which was found to dissolve away rapidly during bubble generation). Fluorescent microspheres of 0.02-µm diameter (polystyrene fluorescent microspheres, Nile Red F-8784) from Molecular Probes (Seattle, WA) were used to visualize the interaction of fluid flow with the bubble valves and also to measure the flow velocities. RESULTS AND DISCUSSION Prior to quantitatively discussing the mechanical characteristics of electrolytic bubble valves, first consider the physical interaction (24) Pourbaix, M. Atlas of Electrochemical Equilibria in Aqueous Solutions; Pergamon Press: Oxford, 1966. (25) Brett, C. M. A.; Brett, A. M. O. Electrochemistry Principles, Methods And Applications; Oxford Science Publications: Oxford, 1993. of the bubble with the fluid during valve closing and opening. Panels 1-6 in Figure 1B show a sequence of bright-field optical images of bubble growth and deflation whereas panels 1-6 in Figure 1C show the corresponding fluid flow images as visualized by fluorescent microscopy; the streak length of fluorescent microspheres in Figure 1C 1-6 is an indicator of fluid velocity. Figure 1B 1 (or Figure 1C 1 ) shows fluid flow in the channel prior to triggering an electrochemical bubble (measured open flow rate 16 mm/s, inlet pressure 103 kpa). Upon triggering a voltage, an electrolytic bubble is nucleated, as shown in Figure 1B 2. Importantly, the corresponding fluorescent image in Figure 1C 2 shows no measurable reduction in flow rate (as seen from unchanged streak length with respect to Figure 1C 1 ) even though the bubble is more than half the channel width; the flow trajectories of beads can be seen curving over the bubble in Figure 1C 2. Figure 1B 3 or C 3 shows that only when the bubble grows to a sufficiently large size does the flow rate begins to show a marked decrease, as seen from shorter fluorescent streaks in Figure 1C 3. Finally, when the bubble grows to completely block the channel, Figure 1B 4, it fully stops the flow and the beads become static, subject only to Brownian motion. In Figure 1C 4, the fluorescent beads are visible as spots with circular halos. Note that the fluorescent microspheres are only 0.02 µm in diameter and by themselves cannot be resolved in the optical micrographs in Figures 1C 1-6. Instead, it is the halo associated with them that manifests as bright streaks when the fluorescent beads are in motion, as in panels 1-3, 5 and 6 in Figure 1C. When the flow is stopped on closing the valve, Figure 1C 4, the microspheres appear as bright spherical halos. The observed variation in the size of spherical halos in Figure 1C 4 is mainly due to the limited depth of focus of an optical microscope. Figure 1B 5 shows the bubble as it begins to deflate. Significantly, its corresponding image in Figure 1C 5 shows that full flow is already restored at this stage. Thus, the valve opening does not require full bubble collapse (as in Figure 1B 6 or C 6 ). Instead, the restoration of full flow requires only a slight collapse of the bubble, which occurs in a short period of time of the order of milliseconds (quantitatively characterized in the following). It is also clear from Figure 1C 4 that the shut valve does not leak despite the fact that the channel has a square cross section. It is this ability of the bubble to conform to arbitrary geometries (in the present case, square cross section) that makes bubble valves robust. Figure 1D shows the current response to a voltage pulse (4.6 V, 50 ms) applied to a pair of electrodes. Between the two current spikes in Figure 1D, the current decays with time, and this decay is associated with the nucleation and growth of bubbles at the electrodes in the microchannel. It shows that the energy required to generate the bubble is very small and of the order of 10 µj. Parts A-C of Figure 2 show the valve closing characteristics for three different voltages, viz., 3.8, 4.2, and 4.6 V, respectively. For each driving voltage, the valve closing was characterized at four different flow rates, viz., 5.6 (inlet pressure 102 kpa), 16.4 (inlet pressure 103 kpa), 23.8 (inlet pressure 104 kpa), and 26.6 mm/s (inlet pressure 105 kpa). The flow rates are given in units of velocity instead of flux to enable comparison with literature citing channels with different cross-sectional areas. Figure 2A shows that the 3.8 V applied voltage is capable of shutting all flow with moderate applied pressures (up to 104 kpa). Shutting off the 6394 Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

Figure 2. (A-C) Valve-closing characteristics for four different flow rates at three different voltages. (D) Valve-opening characteristics for two different flow rates. The corresponding valve closing curves are also shown on the left in (D) to enable comparison between opening and closing rates. flow at higher pressures simply required a slightly higher voltage, as shown in Figure 2B,C. In other words, the flow can be regulated simply by tuning the voltage to suit a given flow rate and channel cross section. Although low camera light intensity at high shutter speeds to record fast-moving fluorescent beads prevented data recording at higher flow rates, flow regulation was successfully tested (by visual observation) up to inlet pressures as high as 110 kpa. Also note that the ability of a bubble valve to withstand high pressures is related to the design of the fluid channel and the surface conditions. For example, experimental and theoretical calculations show that, by making the channel width at the neck region smaller, the bubble can be made to withstand even higher inlet pressures. 26 Also note from Figure 2A-C that the valve closing rate (slope of the curves) becomes steeper with higher applied voltage. The valve opening response is shown in Figure 2D for two different flow rates. In Figure 2D, the corresponding valve-closing curves are also shown on the left portion of the graph to enable a comparison between valve-closing and -opening rates. Also note that the time for which the valve is desired to stay fully closed in Figure 2D (time between valve opening and closing) can simply be changed by keeping the applied voltage to any desired length of time. As seen from Figure 2D, both opening and closing can be completed within 30 ms. While the valveclosing rate increase with higher voltage (Figure 2A-C), the valveopening rates depend on the rate of bubble collapse. As shown above, full collapse of the bubble is not required to open the valve, since the valve opens when hydraulic resistance of the region (26) Leung Ki, Y.-S.; Kharouf, M.; Lintel, H. T. G. van.; Haller, M.; Renaud, Ph. 1st Annual International IEEE-EMBS Spec. Top. Conf. Microtechnol. Med. Biol. 2000, 390-393. Figure 3. Spatial extent of solution chemistry variation delineated by a ph-sensitive fluorescent dye (Oregon Green, Molecular Probes). (A) Schematic of channel and electrochemical reactions occurring at anode and cathode. The secondary chemical reaction at the anode (+) is shown below the primary Cl - ion reduction reaction. (B) Uniform fluorescence in the channel with no voltage applied. (C) Brighter (basic)/darker (acidic) fluorescence at the cathode/anode during the electrochemical reaction. (D) Same electrochemical reaction as in (C) with HEPES buffering added to the solution whose presence suppresses any ph gradients. The dark areas in (D) in the vicinity of the electrodes are electrolysis bubbles. containing the bubble becomes comparable to that of channel (as seen in Figure 1C 5 ). The rate of collapse depends on the rate of gas dissolution into the liquid, which in turn depends on the surface-to-volume ratio of the bubble and the surface tension of the interface. For spherical bubbles of radius r at a fixed hydrostatic pressure, the rate of collapse is proportional to 3RTφ/ 4r, where φ is the permeability of the gas-liquid interface and RT is the gas constant times the temperature. Thus, smaller bubbles tend to collapse faster than large ones so that microfluidic channels are better suited as their dimensions are further reduced. As a closed valve begins to open, the liquid flow also washes away the dissolved gas (not the bubble) at a higher rate and favors further collapse. Since less than 1 pmol of salt is needed to generate a bubble of useful size, ph gradients that are invariably associated with electrochemical reactions can readily be suppressed by mild buffering. The electrolysis effect on solution ph with and without buffer was studied by using a ph-sensitive fluorescent dye. Figure 3A schematically shows the half-cell reactions at the anode and the cathode; the secondary reaction at the anode is shown in smaller lettering. Figure 3B shows a fluorescent image in nonbuffered solution flowing in the channel. In accordance with the above-described electrochemistry, upon applying a voltage, the fluorescence around the cathode becomes brighter (representing an increase in ph) and that near the anode becomes darker (decrease in ph), Figure 3C. To illustrate that a buffer can indeed easily and readily suppress ph gradients generated during electrochemical process, 50 mm HEPES buffer (N-2-hydroxyeth- Analytical Chemistry, Vol. 74, No. 24, December 15, 2002 6395

Traditionally, an ability to make large-scale manifolds has been difficult to achieve in microfluidic systems using moving mechanical parts due to complex micromachining and assembly. Moreover, moving mechanical parts suffer from imperfect sealing and unwanted adhesion. In contrast, making complex fluid logic arrays becomes simpler with the present approach. This ability to make large-scale serial and parallel manifolds controlled by low-power standard electronic logic can be applied to high-throughput screening and sorting of cells and organelles, as well as to soluble analytes. Furthermore, reaction kinetics of compounds in the fluid can be manipulated by including heaters, owing to simple basic design. Sensors can also be built into the chip during fabrication; the simplest of these would be electrodes for voltammetry, conductivity, and thermocouples/heaters for heat capacity. Figure 4. (A) Optical micrograph of an eight-way multiplexer based on electrochemical bubble valves. (B) Fluorescent micrograph showing fluid being directed to output channel 6. (C) Fluorescent micrograph showing fluid being directed to output channel 5. ylpeperazine-n -2-ethanesulfonic acid, C 8 H 17 N 2 NaO 4 S) was added to the 1 M NaCl solution, and the electrochemistry was repeated under the same condition. HEPES is a common buffer that is used in a wide range of biological applications, including those involving electrochemical processes. As shown in Figure 3D, the buffer readily renders any ph gradients negligible that are generated during electrolysis. Finally, an eight-way prototype multiplexer was built to further study the feasibility of the method to make more complex microfluidic chips. An optical micrograph of the multiplexer is shown in Figure 4A; it required the same steps as those needed to make the above-described valves. The distribution channels in Figure 4A have a 25-µm square cross section. The multiplexer chip in Figure 4A consisted of one inlet channel and 2 n (n ) 3) outlet channels such that the fluid can be distributed to any of the 2 n () 8) channels by closing just n () 3) valves. Panels B and C of Figure 4 show the fluorescent optical micrographs of the flow being switched alternately between output channels 5 and 6, respectively. In Figure 4B, the fluorescent light streaks in the channel show the flow being directed to channel 6 by keeping the valves V 1,V 6, and V 11 closed, whereas in Figure 4C, the flow is shown directed to channel 5 by keeping the valves V 1,V 6, and V 12 closed. CONCLUSIONS In summary, the mechanical and chemical characteristics of electrochemical bubble valves have been studied. By generating bubble valves directly inside the channel, valves can be closed and opened in milliseconds. Whereas higher applied voltages increase valve-closing rates, bubble deflation is promoted by a small surface-to-volume ratio of the bubble, and due to the fact that for a given interfacial tension, the internal bubble pressure increases, with decreasing bubble size. Fluorescent microscopy has shown that the valves are completely shut owing to the ability of the bubble to take the channel cross-sectional shape. Only picomoles of salt were needed to generate bubbles. Moreover, visualization of chemistry associated with bubble generation using ph-sensitive dye shows that ph gradients can be readily suppressed by mild buffering. A number of commonly occurring electrolytes in varying concentrations were successfully tested. Using such bubble valves, an eight-way multiplexer was fabricated and tested. Combined with low power in the microwatt range and robust bubble behavior, this approach may offer complex microfluidic systems to be realized. While we made our prototypes in silicon, they may also be made in glass or plastic. ACKNOWLEDGMENT This work was supported by the National Science Foundation, Grants NSF-CMS-0201293 (S.Z.H.), NSF-DMR-97-31733 (H.D.C.), and NIH (F.S.), and the support is gratefully acknowledged. The authors thank Dr. Lev Deresh for useful discussions and M. Dewitt for help in microfabrication. This work was performed, in part, at the Cornell Nanofabrication Facility (a member of the National Nanofabrication Users Network), which is supported by the National Science Foundation Grant ECS-9731293, Cornell University and industrial affiliates. Received for review July 24, 2002. Accepted October 8, 2002. AC0259818 6396 Analytical Chemistry, Vol. 74, No. 24, December 15, 2002