JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 16, NO.

5, OCTOBER 2007

1095

An Electrolysis-Bubble-Actuated Micropump Based on the Roughness Gradient Design of Hydrophobic Surface

Chih-Ming Cheng and Cheng-Hsien Liu, Member, IEEE, Member, ASME
AbstractA novel electrolysis-bubble-actuated micropump based on the roughness gradient design in the microchannel is reported in this paper. This micropump is implemented by taking advantage of both the electrolysis actuation and the surface tension effect. The surface tension effect is controlled via the periodic generation of electrolytic bubbles and the roughness gradient design of microchannel surface, which results in the specied variation of liquid contact angle along the microchannel. Our proposed micropump could resolve the disadvantages that exist in the early reported micropumps, such as the complicated time-sequence power control, the need of long nozzle-diffuser structure, and the choking/sticking phenomena of electrolytic bubbles in a microchannel. Due to the features of large actuation force, low-power consumption, and room temperature operation, our micropump is suitable for the development of low-power consumption and compact micropumps for various applications. Experimental results show that the liquid displacement and the pumping rate could be easily and accurately controlled by adjusting the amplitude and frequency of the applied voltage. With the applied voltage of 15 V at 4.5 Hz, a maximum pumping rate of 114 nl/min is achieved for one of our micropump designs with a microchannel of 100 20 m. In this paper, we report the theoretical analysis, design, micromachining process, operating principles, characterization, and experimental demonstration of these micropumps. [2006-0174] Index TermsBubble, electrolysis, microelectromechanical systems (MEMS) micropump, roughness gradient, surface tension.

I. I NTRODUCTION ANY micropumps based on different designs and actuation mechanisms have been demonstrated over the past decade [1]. For example, diaphragm pumps actuated by piezoelectric [2][4], electromagnetic [5], thermopneumatic [6], and electrostatic [7] mechanisms have been developed to achieve a high pumping volume through a large chamber. Most of the fabrications of diaphragm-based pumps are complicated because of the many photolithographic steps involved. Fielddriven micropumps such as electro-osmotic [8], [9] and electro-

Manuscript received August 31, 2006; revised April 14, 2007. This work was supported in part by the National Science Council of Taiwan, R.O.C., under Grant NSC-94-2218-E-007-013 and in part by the Nano-Technology Research Program under Grant NSC-94-2120-M-009-015. Subject Editor A. Ricco. The authors are with the Micro-Systems and Control Laboratory, Department of Power Mechanical Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan, R.O.C. (e-mail: liuch@pme.nthu.edu.tw). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/JMEMS.2007.900880

hydrodynamic pumps [10] have also been proposed to drive the uid by applying a high voltage without the need of mechanical moving parts. Among all actuation methods, bubble-actuated valveless micropumps show the features of simple operation, miniaturized size, large actuation force, and the ability to be conformed physically to different designs of microchannels with a wide range of cross sections. Both thermal and electrolytic bubbles have been demonstrated with the function of microuidic actuation. The thermal-bubble-actuated micropumps have been developed with different mechanisms such as the traversing multiple bubbles [11], the nozzle-diffuser structure [12], and the periodic generation of a single vapor bubble relying on surface tension [13], [14] to achieve the net pumping ow. However, the thermal generation of bubbles is a highpower consumption method because of the dramatic heat loss in microscale [12], [15]. The high-temperature process might also damage the targeted biosample for biochip applications. Compared with other actuation mechanisms, the electrolysis bubble actuator has the features of simple structure, large actuation force, low-power consumption, room temperature operation, and being easy to be integrated into a lab chip. The electrolytic bubbles have been used as actuators in various designs of microvalves and micropumps. For example, a microinjector is reported by using the electrolytic bubbles, which raises the liquid pressure to push the liquid forward, and the bubble is then expelled with the liquid [16]. Micropumps actuated by electrolytic bubbles are reported and based on inating a large bubble inside a reservoir, which generates a pressure head in the uid channel during the bubble growth [17], [18]. The electrolytically actuated micropump by means of sequentially generating a series of electrolytic bubbles inside a microchannel has also been reported [19]. However, there are some disadvantages for the above pumping mechanisms, such as the complicated time-sequence power control on many pairs of electrodes, the need of large/long nozzle-diffuser structure, and the requirements of the degassing function and the wellsealed reservoir inside the uidic chip. The utilization of surface tension force promises a powerful actuation mechanism for microuidic systems because of the large force compared to the other forces in microscale. In our micropump design, the pumping principle relies on the control of the surface tension effect that is initiated via electrolysis bubble generation and tuned via the roughness gradient design of hydrophobic surface in the microchannel. Although the similar surface tension effect is used in our development, the operation

1057-7157/$25.00 2007 IEEE

1096

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 16, NO. 5, OCTOBER 2007

principle and the design in our proposed micropump are different from those used in other pumping devices [13], [14], [20] [22]. Our proposed micropump could resolve the challenging problem for most electrolysis bubble actuators on the removal of insoluble gases. The roughness gradient structure is rst introduced and utilized to design surface energy gradients to form the guide rails by Shastry et al. [23]. Droplets can move along the surface energy gradient overcoming the hysteresis by supplying energy through mechanical vibration. Lee et al. have recently reported the device that switches the roughness of a membrane to move a droplet [24]. In this paper, we present a novel electrolysis-bubble-actuated micropump utilizing the design of the roughness gradient surface to achieve the pumping function. Due to the features of large actuation force, low-power consumption, and room temperature operation, our micropump without the needs of complicated microfabrication and mechanical moving parts is suitable for the development of low-power consumption and compact micropumps for various applications. One of the long-term goals of this paper is to develop a low-power-consumption micropump in a closed loop for micro direct methanol fuel cell (DMFC) application. Meng et al. also recently proposed an electrolysis-bubble-actuated micropump for DMFC application [20]. They took advantage of the bubbles traveling from hydrophilic to hydrophobic environment in a microchannel to move liquid to generate a net pumping ow. In that report, the pressure head was measured at about 120195 Pa, and the pumping rate was characterized as 2065 nl/s in the microchannel of 600 300 m. Due to the traveling feature of electrolytic bubbles, either choking or sticking phenomenon of electrolytic bubbles might appear in the microchannel. In our micropump, the pumping mechanism does not rely on the long-distance traveling of electrolytic bubbles. Our proposed micropump could resolve disadvantages such as the complicated time-sequential power control, the need of long nozzle-diffuser structure, and the choking/sticking phenomena of electrolytic bubbles in a microchannel. The theoretical analysis, design, synthesis, microfabrication process, operating principles, and experimental results are described as follows. II. D EVICE O PERATION P RINCIPLE AND M ICROFABRICATION P ROCESS A. Operation Principle Fig. 1(a) illustrates the design concept and operation principle of our micropump, which consists of platinum electrodes, a hydrophilic microchannel, and the hydrophobic lateral breather connected to ambient air for the purpose of removing the electrolytic bubbles. The hydrophobic lateral breather, which is made of a top polydimethylsiloxane (PDMS) cover and the Teon patterns extending to the outside of the microchannel on the bottom substrate, not only degases the electrolysis bubbles but also prevents the liquid from leaking out. Because of the roughness gradient design on the lateral breather, our micropump obtains a net pumping ow after the bubbles are degassed out of the microchannel. In our micropump design, we take advantage of the hydrophobic property variation of the liquid
Fig. 1. Illustration of the design concept and the pumping principle. (a) Schematic 3-D view of our proposed micropump. A net pumping ow along the x-direction of the microchannel in one pumping cycle via three processes [(b) bubble generation, (c) degassing, and (d) liquid movement] is illustrated from the side views of the micropump. The net pumping uid is achieved over multiple electrolysis bubble actuation cycles in the microchannel. Here, Pb , PL , and PR are the pressures. r,L , r,R , and b are the contact angles.

on different rough hydrophobic surfaces to design a roughness gradient surface that is located at the lateral breather. More details about the roughness gradient design will be described in Section III. The pumping principle of our micropump relies on the surface tension effect and the generation of multiple electrolysis bubble actuation cycles. The surface tension effect is controlled via the periodic generation of electrolytic bubbles and the roughness gradient design of the microchannel surface, which results in the contact angle variation of liquid along the microchannel. Fig. 1(b) and (d) illustrates the pumping principle

CHENG AND LIU: ELECTROLYSIS-BUBBLE-ACTUATED MICROPUMP BASED ON THE ROUGHNESS GRADIENT DESIGN

1097

of our micropump in one pumping cycle from the side view of the micropump. The actuation mechanism in one pumping cycle could be roughly divided into three processes: 1) bubble generation; 2) degassing; and 3) liquid movement. First, the bubble is generated by the voltage applied on the electrodes to push the liquid both forward and backward in the microchannel, as shown in Fig. 1(b). In the process of bubble generation, there is no net pumping ow. Next, the bubble vents out of the microchannel through the hydrophobic lateral breather to the ambient air when the liquid between the two platinum electrodes turns into electrolytic bubble completely, as shown in Fig. 1(c). Due to our roughness gradient design, the apparent contact angle on the leading meniscus is larger than that on the trailing meniscus, i.e., r,R > r,L > 90 . Therefore, the pressure on the trailing meniscus of the liquid PL is larger than the pressure on the leading meniscus of the liquid PR according to the LaplaceYoungs equation [16]. Finally, both liquid meniscuses ow back with different velocities to ll in the empty section by the capillary forces during the degassing process, as shown in Fig. 1(d). As a result, one pumping cycle is completed, and a net pumping ow along the x-direction is achieved. The net pumping ow can be achieved over multiple electrolysis bubble actuation cycles in the microchannel. The net volume displacement of liquid and the pumping rate are dominated by the geometry design of the microchannel, the design of the roughness gradient surface, the frequency and amplitude of the applied voltage, and the design of the electrodes. All details will be described with the experimental results later in this paper. B. Microfabrication The result of the microfabrication process for our micropump is illustrated in Fig. 1(a). First, a standard p-type 100 silicon substrate is grown with the thermal oxide of 6000 that is lithographically patterned and etched later as the oxide mask. Then, the substrate is etched by the wet etching process in KOH solution to dene the microchannel. This silicon substrate is then grown with another 8000 thermal oxide to form the hydrophilic layer on top of the microchannel substrate. Then, a 200 /1500 titanium/platinum electrode layer is evaporated and patterned as the electrodes by a liftoff process. Next, Teon AF (Amorphous Fluoropolymer, 1%, Dupont) is spin coated on the wafer and cured at 180 C for 6 h in a high-temperature oven. The Teon lm is lithographically patterned and etched as hydrophobic regions by the O2 plasma process to serve as the bottom part of the hydrophobic lateral breather. The top PDMS cover with the roughness gradient structure is fabricated by using a mold that is fabricated on another silicon wafer with the negative-tone UV photoresist (MicroChem Corporation, SU-8 35). In this process, a curing agent and a PDMS prepolymer (Sylgard 184 Silicone Elastomer Kit) are mixed in a 1 : 10 weight ratio and stirred to ensure complete mixing. The prepolymer mixture is degassed in a vacuum oven for removing air bubbles. Then, the prepolymer mixture is poured onto the mold and cured at 65 C for 1 h in the oven. After curing, the PDMS is peeled off from the mold. Finally, the top PDMS cover is bonded onto the top of the silicon wafer.

Fig. 2. Droplet on the surface forms: (a) the complete wetting and (b) the composite wetting based on the hydrophilic and the hydrophobic properties of substrate surfaces, respectively.

III. T HEORETICAL A NALYSIS AND M ODELING A. Roughness Gradient Design of Hydrophobic Surface The contact angle has been commonly used to represent surface wettability. Surface wettability is a function of surface roughness. The latest experimental results conrm that wettability can be tuned by surface roughness [25], [26]. The earliest literatures reported that the contact angle of a droplet on a rough surface could be predicted by two main theories relating the surface structure to the apparent contact angle. The rst theory was proposed by Wenzel [27], which assumes that the liquid completely wets the solid structure, as illustrated in Fig. 2(a). The second theory was proposed by Cassie and Baxter [28], which assumes that the liquid does not wet the valleys of the structure and forms a composite surface on the rough substrate, as illustrated in Fig. 2(b). Afterwards, Bico et al. [26] fabricated the substrates with specic roughness and compared the measured contact angles with the prediction results. They claimed good agreement and proposed that Wenzels formula is valid for the hydrophilic surface and that Cassie and Baxters formula is valid for the hydrophobic surface. Shastry et al., who rst utilized the roughness gradient structure to move the droplet, have also recently presented a model for the forces acting on the droplet resting on a roughnesscontrolled contact angle gradient [29]. A regular 2-D array of square pillars creates the rough surface with the controlled roughness parameters and r, which are dened by the dimension of square pillars (a a), the pillar spacing (b), and the pillar height (c), i.e., = r= a2 (a + b)2 4bc + (a + b)2 (a + b)2 (1) (2)

where is the surface area fraction, and r is the roughness ratio of the planar surface. The parameter determines the apparent contact angle on a composite surface based on Cassie and Baxters formula. r mainly correlates the relative stability of a droplet in the composite state. The stability of a droplet in the composite state depends on the design of the parameters and r, which will be discussed later. The structural design of the surface roughness gradient in our device is illustrated in Fig. 3. The surface roughness is dominated by , which varies with the pillar patterns on the PDMS cover. For our roughness gradient design, decreases

1098

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 16, NO. 5, OCTOBER 2007

designs of pillars on the roughness gradient surface have to follow (5). More details regarding the relationship between the pressure on the meniscus of liquid and the surface area fraction will be derived based on the theoretical model addressed next. B. Electrolysis When an electric current is sent through the two noble metal electrodes (such as platinum) in water, electrolysis takes place. The minimum equilibrium potential of hydrogenoxygen electrolysis E 0 is 1.23 V. In the electrolysis reaction, the oxygen gas is produced at the anode, and the hydrogen gas is produced at the cathode, i.e.,
Fig. 3. Illustration of the roughness gradient design on the hydrophobic surface of the top PDMS cover that is made of ai ai square pillars and pillar spacing bi arranged in a regular array.

anode : 2H2 O 4H+ + 4e + O2(g) cathode : 2H2 O + 2e 2OH + H2(g) . (6)

along the x-direction of the microchannel, i.e., 1 > 2 > > n1 > n . The design of the roughness gradient surface could be modeled based on Shastrys work. The roughness gradient surface of our micropump is hydrophobic, which leads to the formation of a composite surface. Thus, the roughnesscontrolled contact angle gradient could be predicted under the CassieBaxter assumption i = a2 i (ai + bi )2 , i = 1, 2, 3, . . . , n (3) (4)

Under the assumption that all generated gases (O2 and H2 ) evolve in the form of gas bubbles, the total gas volume linearly depends on the input electrical charge [17]. The total gas volume generated by the electrolysis in the process of bubble nucleation could be estimated according to Faradays law of electrolysis and the ideal gas law [31] It zF P V = N RT N= (7) (8)

cos r,i = 1 + i (cos + 1)

where i represents the surface area fraction of region i, which is made of the rough surface on the PDMS side with ai ai square pillars and pillar spacing bi ; r,i is the apparent contact angle on the PDMS side in region i; and is the equilibrium contact angle of the liquid droplet on the at PDMS surface. The relationship between the apparent contact angle and the surface area fraction could be theoretically derived via (4). Besides, previous report from Shirtcliffe et al. has also shown that the dimension variation of square pillars allows the length of the contact perimeter per unit area to be varied without varying the contact area per unit area [30]. As a result, there is no change on the contact angles. Thus, the apparent contact angle is expected to remain constant when the surface area fraction is constant. Therefore, the surface area fraction is a critical factor for the roughness gradient design of hydrophobic surface in our proposed micropumps. Besides, the previous literature has also shown that a droplet on a rough surface might transit from composite state to complete state [29]. For the stability of the composite state, the following inequality holds: cos r < 1 r (5)

where N is the moles of produced gas, I is the applied current, z is the number of excess electrons, F is Faradays constant (9.649 104 C/mol), t is the period of electrolysis, T is temperature, P is the ambient pressure, V is the volume of the bubble, and R is the gas constant (8.314 J K1 /mol). Under the assumption of constant temperature and atmospheric pressure, the volume of produced gases is proportional to the supplied electric current [32]. The theoretical bubble growth rate could be calculated by RIT dV = . dt FPz (9)

In our micropump, the bubble volume could not be precisely estimated by Faradays law of electrolysis because this micropump might have some loss of the bubble volume during bubble expansion. However, the bubble growth rate is roughly proportional with the applied current via raising the amplitude of the applied voltage. The experimental results follow the proportional trend, which will be addressed in Section V. C. Theoretical Modeling of Micropump The electrochemical pump relies on the pressure increase via the generation of oxygen and hydrogen gases when the applied current passes through the electrolytic liquid. The theoretical models based on early literatures are addressed and summarized as follows for the purpose of roughness gradient design in our proposed micropump. The internal liquid pressures could be individually modeled for the expansion period and the venting period of the electrolytic bubble within one pumping cycle.

where r is the apparent contact angle for the case of composite state. This means the stability of a droplet in the composite state is determined by the parametric design of and r, which are the surface area fraction and the roughness ratio of the planar surface, respectively. In this paper, the theoretical derivations are based on the CassieBaxter assumption. Thus, the geometry

CHENG AND LIU: ELECTROLYSIS-BUBBLE-ACTUATED MICROPUMP BASED ON THE ROUGHNESS GRADIENT DESIGN

1099

First, while the electrolytic bubble grows, the electrolytic bubble pressure Pb , as illustrated in Fig. 1(b), is modeled according to Faradays law of electrolysis and the ideal gas law as Pb = IRT t N RT = . V zF V (10)

While the electrolytic bubble is venting, the meniscuses of the moving liquid are driven based on the surface tension effect. A schematic illustration for the driving mechanism of our proposed micropump is shown in Fig. 1(d). In the general case, one of the meniscuses of the liquid moves a forward displacement dx starting from the position x in the microchannel, on which the pressure could be modeled by using the total surface energy approach [16], [33]. The effect of gravity is neglected because of the shallow microchannel in our device. The surface energies are related to the equilibrium contact angle governed by LaplaceYoungs equation sa = sl + la cos r (11)

Fig. 4. Relationship between pressure and surface area fraction for different microchannel depths (10, 20, and 50 m). The microchannel width is xed to be 100 m in all these calculations.

where r is the apparent contact angle, and sl , sa , and la are the surface energies per unit area of solidliquid, solidair, and liquidair, respectively. The differential interfacial energy of the system dUT could be derived as dUT = sl dAsl + sa dAsa + la dAla = (sl sa )dAsl + la dAla = la (cos r dAsl + dAla ) (12)

tively. To simplify the equation derivation, the microchannel width is assumed to be the average width of the nonorthogonal microchannel, which results from the wet etching process. The shapes of meniscuses are also assumed to be constant curves. All the contact angles are assumed to be equilibrium contact angles neglecting the contact angle hysteresis. The pressure PL on the trailing meniscus of the liquid moving in region i could be derived as [33] dUT PL = dV la [wi,L (1 + cos PDMS ) = wh + w (2h + w) cos SiO2 ] .

(14)

where Asl , Asa , and Ala are the interface areas of solidliquid, solidair, and liquidair, respectively. The second equality in (12) is derived based on dAsa = dAsl because the summation of Asl and Asa is a constant. In our prototype device, micromachined PDMS pillars on top of the microchannel are employed to control the roughness of the hydrophobic surface. The bottom surface of the microchannel is coated with a hydrophilic layer of silicon. Thus, the differential interfacial energy on the trailing meniscus of the liquid could be modeled as dUT = dUcover + dUmicrochannel = la (cos r,i(PDMS) dAsl + dAla ) + la (cos r(SiO2 ) dAsl + dAla ) + la [(2h + w) cos SiO2 ] dx = la [wi cos PDMS + w(1 i ) (2h + w) cos SiO2 ] dx (13)
PDMS SiO2

The pressure PR on the leading meniscus of the liquid moving in region j could be obtained via a similar derivation PR = la [wj,R (1+cos PDMS )+ w (2h + w) cos SiO2 ] wh (15)

where V is the liquid volume. PL and PR depend on the contact angles PDMS and SiO2 , the geometry of the microchannel, and the surface area fractions i,L and j,R . Here, i,L and j,R are the surface area fractions in the roughness pattern region i and j , where the trailing and leading meniscuses of the liquid move in, respectively. The pressure difference between these two meniscuses, which affects the liquid displacement, could be derived as P = PL PR la [wj,L(1+cos PDMS )+ wj,R (1+cos PDMS )] = wh la (i,L j,R ) [1 + cos PDMS ] . (16) = h In our calculations, different microchannel depths (10, 20, and 50 m) and the same microchannel width of 100 m are assumed and used. Under the assumption of SiO2 = 20 , PDMS = 110 , and la = 72 dyn cm1 , the relationship between pressure and surface area fraction is derived and shown in Fig. 4 based on (14) and (15). The positive sign of the pressure means a forward pulling pressure. The calculation results shown in Fig. 4 provide us the information of pressure distribution versus different roughness gradient surfaces for

= la [wi cos PDMS + w(1 i )] dx

where dUcover and dUmicrochannel are the differential interfacial energies related to the top cover and the bottom microchannel substrate, respectively. r,i(PDMS) is the apparent contact angle of liquid in region i on the PDMS side. i(SiO2) is the apparent contact angle of liquid on the thermal oxide lm. PDMS and SiO2 are the equilibrium contact angles of the liquid drop on the PDMS lm and on the thermal oxide lm, respectively. h and w are the depth and width of the microchannel, respec-

1100

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 16, NO. 5, OCTOBER 2007

Fig. 5. Schematic diagram of the experimental setup. The outlet of our microchip is connected via a tube to the beaker, which could be raised by a high-precision stage. The inlet is connected to a xed beaker.

the design of our micropumps. In our design, the surface area fractions decrease along the x-direction of the microchannel, i.e., i > j for i < j . Therefore, the pressure difference is always positive P > 0 in our proposed micropump design. In other words, the micropump develops a net pressure head toward the x-direction of the microchannel under the zero-netow condition. Thus, a net pumping ow is achieved via our roughness gradient design. The pressure difference in our micropump is dominated by the roughness gradient design. max and min represent the maximum and the minimum of the surface area fraction, respectively, on the roughness gradient surface in the microchannel. The maximum pressure difference could be modeled as Pmax (max , min , h) = PL (max , h) PR (min , h). (17)

imply that the actuation pulses play a dominant role for the maximum pumping ow rate. Besides, the expansion period and the venting period of the bubble in one pumping cycle are also critical parameters to regulate the square-wave actuation pulses. The operation frequency f and the duty ratio d are dened as f= d= 1 texpand + tvent texpand texpand + tvent (18) (19)

In this paper, micropumps with different roughness gradient designs and microchannel depths are implemented and characterized to demonstrate the performance. The experimental results will verify that the maximum pressure difference of the micropump dominates the liquid displacement in one pumping cycle as well as the pumping ow rate. To simplify the derivation, the model used in the above derivations does not include the effect of contact angle hysteresis, which often results in a signicant portion of the surface tension effect for the droplet motion [34], [35]. This effect depends on surface condition of the microchannel. The calculated pressure will become less when the contact angle hysteresis is taken into account. D. Applied Voltage The pumping ow rate relies on many factors, such as the applied voltage, the duty cycle, and the driving frequency. These

where texpand and tvent are the expansion period and the venting period of the electrolytic bubble in one pumping cycle. The expansion period texpand is dominated by the period of the applied voltage in one pumping cycle. The venting period tvent is dominated by the bubble volume and the pressure magnitudes on both meniscuses. IV. E XPERIMENTAL S ETUP To characterize the performance of our micropumps with different parametric designs and operation conditions, the experimental setup is schematically represented in Fig. 5. To actuate our micropump, a Labview graphical program is implemented with the programmable DAQ card (PCI-6534, NI) to regulate the square-wave pulses and control the operation frequency and duty ratio. A direct-current power supply (E3631A, Agilent) is used to control the amplitude of the applied voltage through the electric control circuitry. A high-impedance eld-electric transistor in this electric control circuitry is used to perform the required voltage pulses. A relay is also used in this control circuitry to eliminate the alternating polarization and avoid the corrosion of noble metal electrodes. An optical microscope

CHENG AND LIU: ELECTROLYSIS-BUBBLE-ACTUATED MICROPUMP BASED ON THE ROUGHNESS GRADIENT DESIGN

1101

with a charge-coupled device (CCD) camera is utilized to observe and record the pumping process of our micropumps. The open-loop micropump is fabricated for performance characterization. Two glass tubes are attached to the through-holes at the inlet and outlet of this micropump. The inlet and outlet are connected to two beakers via soft tubes of 1.5-mm diameter. The beaker connected to the outlet could be accurately raised by a high-precision stage. Pressure transducers (Lutron) with a resolution of 20 Pa are used to monitor the pressures. The pumping pressure and the ow rate are characterized via the setup shown in Fig. 5. After the liquid of the beaker connected to outlet tube is raised to the same height as the inlet meniscus and stabilizes, the driving voltage is applied to start the pumping function. The meniscuses movement and the bubble generation/degassing during the whole process are recorded by a digital video system. For the ow rate characterization, two kinds of particles with different diameters of 3 10 m and 10 30 m (Polysciences, glass beads of 2.48 g/cm3 ) are, respectively, mixed with water that are used as the pumping liquid for two different micropumps with the microchannel depths of 20 and 50 m. The conductivity and the pH value of pumping water are characterized as 0.35 0.65 mS/cm and 6.5, respectively, by using the conductivity/pH meter (EUTECH, PC510). By tracing the particles during a time period, the ow rate is calculated based on the average speed of the moving particles. The pumping ow rates of the micropumps are measured in open microchannels (i.e., negligible back pressure) and characterized via applying the squarewave voltages of the specic voltage frequency and duty ratio. To approach the maximum ow rates, the expansion periods and the venting periods are also characterized via the recorded images under different applied voltages. For the characterization of the pressure head, the beaker connected to the outlet is lifted to adjust the pressure head. In these measurements, the pumping rates are recorded with the change of pressure head, which is monitored via the pressure meter. The maximum pressure head is recorded when the liquid stops owing forward and starts owing backward. In addition, the leakage pressure is characterized by sealing one of two inlet/outlet ports. The leakage pressure is recorded via observing the phenomenon that the liquid wets the breathing hole of the lateral breather during the lifting process of the beaker connected to the outlet. For contact angle measurements, a small-volume drop of water (5 L) is gently placed on the experimental surface. A photograph of the side prole of the liquid drop is taken by using a CCD camera with adjustable optical focus lens. The radius of the spherical cap is measured by processing the sideview image to obtain the free surface of the spherical cap. Then, we t it with a circle to get the radius of the spherical droplet. The equilibrium contact angle is calculated from the radius of the tting circle R and the height of the droplet H by using the formula = cos1 ((R H )/R). V. E XPERIMENTAL R ESULTS AND D ISCUSSION Fig. 6(a) shows a prototype device that has two micropumps with different roughness gradient design on a chip, which is

Fig. 6. Experimental demonstration for the pumping function of our micropump. (a) Prototype device with two micropumps on a chip, which is made of the PDMS cover (15 mm 11 mm 3 mm) and the silicon substrate (15 mm 15 mm 0.5 mm). The time-sequential pictures captured from the top view of the micropump show the processes of pumping ow in one pumping cycle including (b) bubble generation, (c) degassing, (d) liquid movement, and (e) net pumping ow after the electrolytic bubble is degassed completely. The cross section of microchannel is 100 m 20 m.

made of the PDMS cover (15 mm 11 mm 3 mm) and the silicon substrate (15 mm 15 mm 0.5 mm). Table I summarizes the distribution of surface area fraction (i ) for our prototype micropumps with four different structure designs. The microchannel width w is xed to be 100 m for all these micropumps. The micropumps are operated under the condition of open microchannels for pumping characterization. The timesequential pictures, which are captured from the top view of the micropump, in one pumping cycle for micropump#1 are recorded as shown in Fig. 6(b) and (e). After the microchannel is lled with the sample liquid, the electrolytic bubble is generated in the active region by applying an alternating electric current to the platinum electrodes. The liquid is then

1102

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 16, NO. 5, OCTOBER 2007

TABLE I DESIGN DETAILS FOR FOUR TYPES OF MICROPUMPS WITH DIFFERENT ROUGHNESS GRADIENT DESIGNS (A AND B ). THE MICROCHANNEL WIDTH (w) IS FIXED TO BE 100 m FOR ALL THESE MICROPUMPS

Fig. 7. Experimental characterization for the apparent contact angle r on the hydrophobic surface versus different surface area fraction . The dots ( ) and the solid line ( ) represent the experimental results and theoretical prediction based on CassieBaxters theory, respectively.

Fig. 8. Characterization of leakage pressure for ve hydrophobic lateral breathers with different spacing between adjacent pillars (5, 10, 20, 30, and 40 m).

separated into two sections and pushed toward both inlet and outlet directions, respectively, in the microchannel, as shown in Fig. 6(b). Fig. 6(c) and (d) shows the processes when the applied voltage on the platinum electrodes is turned off. The electrolysis bubble is degassed through the lateral breather to ambient air. Both meniscuses of liquid then ow back with different velocities to ll in the empty section of liquid in the microchannel. As a result, a net pumping ow with the function of degassing the electrolytic bubbles is achieved in one pumping cycle based on the roughness gradient design, as shown in Fig. 6(e). Therefore, liquid pumping is achieved over multiple electrolysis bubble actuation cycles in the microchannel. A. Characterization of Contact Angles To characterize the contact angles for various surface area fractions (), several different testing microuidic chips are fabricated. All of these microuidic chips have a design similar to our proposed electrolysis-bubble-actuated micropump except for no roughness gradient on the PDMS covers. Each PDMS cover (25 mm 25 mm 3 mm) has pillars with the same surface area fraction () along the microchannel. All of these testing chips have different surface area fractions varying from 0.1 to 0.9 via the position arrangement of pillars. Fig. 7 shows the characterized and the theoretical relationship between the apparent contact angle r and the surface area fraction . The experimental characterization agrees with the theoretical prediction from Cassie and Baxters theory. Here, the characterized data are taken from the average results of ve-

times experiments because it is difcult to precisely measure the equilibrium angles under the existence of contact angle hysteresis on these microchannel surfaces.

B. Leakage Testing The capability of the micropump to be operated against back pressures is one of the key requirements for operating a micropump in a closed-loop environment. In our micropump, the liquid is pumped forward based on the breathing function of the hydrophobic lateral breather in our micropumps. When the back pressure is higher than the leakage pressure of the lateral breather, the breathing of gases would be blocked to make our micropump lose the ability of degassing electrolytic bubbles. The leakage pressure, which represents the maximum allowed liquid pressure without leakage through the breathing holes, is dominated by the dimension of breathing hole at the lateral breather and follows LaplaceYoungs equation [36]. In our case, the leakage pressure of breather is affected by the spacing between pillars. Thus, it is important to dene the critical spacing that could support a given liquid pressure without blocking the breather. For the leakage pressure measurements, ve rough surfaces with the same pillar dimension (25 m 25 m 30 m) and different spacing between adjacent pillars (5, 10, 20, 30, and 40 m) are fabricated to form different lateral breathers. The experimental results for the leakage pressure versus the spacing between adjacent pillars are shown in Fig. 8. The solid line is obtained by a linear tting of the measured data. The maximum leakage pressure of 17.9 kPa is achieved in our prototype micropumps based on the pillar spacing of 5 m.

CHENG AND LIU: ELECTROLYSIS-BUBBLE-ACTUATED MICROPUMP BASED ON THE ROUGHNESS GRADIENT DESIGN

1103

Fig. 9. Net pumping displacement of liquid in one pumping cycle versus the calculated maximum pressure difference for the four types of micropumps. The designs and specications for the four types of micropumps with different microchannel depths and surface roughness gradients are summarized in Table I.

Fig. 10. Expansion period of the electrolysis bubble texpand to ll the roughness gradient region in the microchannel versus the applied voltage for micropump#2 and micropump#4.

C. Pressure and Flow Rate Testing According to (14)(17) and the theoretical derivation in Fig. 4, the maximum pressure difference is dominated by the roughness gradient design and the microchannel depth. Four types of micropumps with the designs of different roughness gradients (A and B ) and different microchannel depths (20 and 50 m) as summarized in Table I are used to evaluate the performance of our proposed micropumps. Fig. 9 shows the relationship between the calculated maximum pressure difference and the net pumping displacement of liquid in one pumping cycle for these four types of micropumps. Compared with other prototype micropumps, micropump#2 with the largest maximum pressure difference pumps the liquid to achieve the largest displacement in one pumping cycle. The large maximum pressure difference is achieved by the design of the shallow microchannel depth and the large surface roughness gradient. Here, the pumping displacement of the liquid is characterized via measuring the average displacement of the moving particles under the initial condition of zero ow. The data for the pressure difference shown in Fig. 9 are derived theoretically based on (14)(17) and Fig. 4. The error bars shown in Fig. 9 result from the particle displacement variation among ten measurements for each pressure difference design. For the characterization of the maximum volume ow rate, micropump#2 and micropump#4 with the same design of large surface roughness gradient (B ) and different microchannel depths (20 and 50 m) are used to observe the effect of microchannel depths. To approach the optimal pumping rates under different applied voltages, the expansion period and the venting period in one pumping cycle are characterized rst. Fig. 10 shows the expansion period of the electrolysis bubble to ll the roughness gradient region in the microchannel under different applied voltages (7, 9, 11, 13, and 15 V). The curves are obtained by simple linear tting among the measured data. The characterized venting periods are approximately 100 and 150 ms for micropump#2 and micropump#4, respectively. Based on the characterization results of the expansion period texpand and the venting period tvent , the operation frequency f and the duty ratio d of the applied voltage could be optimized to enhance the ow rate of micropumps. Fig. 11 shows the characterization results for these enhanced ow rates of micropump#2

Fig. 11. Experimental results for the enhanced pumping volume ow rate versus the operation frequency under different amplitudes of applied voltages (7, 9, 11, 13, and 15 V).

and micropump#4. The data are approximated via the curve tting of two second-order polynomial functions. The error bars shown in Fig. 11 result from the ow rate variation among ten measurements. The applied voltage signals are regulated by the operation frequency and duty ratio, which are calculated via (18) and (19). The optimal frequency of applied voltage with respect to the pumping ow rate is dominated by the expansion period and the venting period of electrolysis bubble. With the applied voltage of 15 V at 4.5 Hz, a maximum pumping rate of 114 nl/min is achieved for our micropump#2 that has a microchannel cross section of 100 m 20 m. These results indicate that the ow rate in our micropumps could be simply regulated by adjusting the voltage pulses to approach the appropriate ow rate in the microuidic chip. The characterized relationship between ow rate and pressure head is shown in Fig. 12. The ow rate is observed in the presence of different elevations between the inlet and the outlet to generate a pressure difference. The characterized data shown in Fig. 12 for micropump#2 and micropump#4 are measured under the applied voltages of 15 V at 4.5 and 2.5 Hz, respectively. When the ow rates decay to zero, the pressure heads are characterized as 1.9 and 0.8 kPa for micropump#2 and micropump#4, respectively. When the pressure head is close to zero, the maximum ow rates, which are 114 and 95 nl/min for micropump#2 and micropump#4, respectively, are achieved. Besides, the power consumption Pw and the thermodynamic efciency are also important parameters. The maximum power

1104

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 16, NO. 5, OCTOBER 2007

Fig. 12. Characterized relationship between the pumping volume ow rate and the pressure head for micropump#2 and micropump#4. The details for the structure designs of micropump#2 and micropump#4 are listed in Table I.

consumption of the micropump is measured to be 1.25 mW for an applied voltage of 15 V, which has a corresponding average current of 83 A during the pumping operation. Because of the approximately linear relationship between the ow rate and the pressure head, the thermodynamic efciency could be estimated based on the power consumption, the maximum ow rate Qmax , and the maximum pressure head (Phead )max by using the formula = (1/4)(Qmax (Phead )max /Pw ) [1]. When our micropump (micropump#2) is operated at f = 4.5 Hz and Pw = 1.25 mW, this micropump produces a Qmax of 114 nl/min and a (Phead )max of 1.9 kPa, as shown in Fig. 12. The thermodynamic efciency is about = 7.4 106 %. D. Corrosion The direct application of voltage pulse trains forces the electrodes to continuously alternate polarization during the operation. Alternating-current polarization has been reported to lead to the corrosion and the roughening of noble metal electrodes [37]. Specically, noble metals such as platinum and palladium, which readily dissolve hydrogen, exhibit rapid corrosion rates. Howe et al. reported that the corrosion of noble metal electrodes could be avoided by alternating the driving voltage between a xed operation potential and an unbiased condition during operation [21]. To enhance the micropump performance and improve the lifetime operation, the driving voltage is alternated between a xed operating potential and an unbiased voltage ground during the operation via a relay to minimize the metal corrosion. Although the long-term test has not been performed, we have applied the current to the electrodes continuously for more than 3 h without any observable electrical degradation. VI. C ONCLUSION A novel electrolysis-bubble-actuated micropump with the design of the roughness gradient on the microchannel hydrophobic surface and the lateral breather has been successfully demonstrated and reported in this paper. Theoretical models are addressed in this paper to study the pumping principle, which is dominated by the pressure difference between two

meniscuses of the liquid in our proposed micropump. The design, synthesis, micromachining process, and comprehensive characterization of our micropumps have also been presented. Furthermore, experimental results successfully demonstrate the pumping function of our micropumps with different designs of the roughness gradient and the microchannel depth. The pressure head of 1.9 kPa and the pump rate of 114 nl/min are measured for our micropump with the microchannel cross section of 100 m 20 m. Compared to other reported micropumps for DMFC application, our low-power consumption micropump has the potential to be applied in a closed loop for DMFC application. Further work to modify, optimize, and integrate this electrolysis-bubble-actuated micropump for applications like DMFC and lab chips is under going in our group. The features of our proposed micropumps on compact size, simple microfabrication, low-power consumption, and room temperature operation make it promising to be integrated with other multiple components to form microuidic systems for applications such as lab on a chip, biochip, drug delivery, and DMFC. Because of the feature of room temperature operation, this micropump is specically suitable for various bio-applications. However, the performance of this micropump critically depends on the PDMS surface maintaining its intrinsic contact angle. In some specic cases, this micropump might lose the pumping feature. For example, proteins would bind nonspecically to the PDMS surface, which would change the contact angle and even make it hydrophilic. R EFERENCES
[1] D. J. Laser and J. G. Santiago, A review of micropumps, J. Micromech. Microeng., vol. 14, no. 6, pp. R35R64, Jun. 2004. [2] H. Andersson, W. van der Wijngaart, P. Nilsson, P. Enoksson, and G. A. Stemme, A valve-less diffuser micropump for microuidic analytical systems, Sens. Actuators B, Chem., vol. 72, no. 3, pp. 259265, Feb. 2001. [3] L. Cao, S. Mantell, and D. Polla, Design and simulation of an implantable medical drug delivery system using microelectromechanical systems technology, Sens. Actuators A, Phys., vol. 94, no. 1/2, pp. 117125, Oct. 2001. [4] M. Koch, N. Harris, A. G. R. Evans, N. M. White, and A. Brunnschweiler, A novel micromachined pump based on thick-lm piezoelectric actuation, Sens. Actuators A, Phys., vol. 70, no. 1/2, pp. 98103, Oct. 1998. [5] S. Santra, P. Holloway, and C. D. Batich, Fabrication and testing of a magnetically actuated micropump, Sens. Actuators B, Chem., vol. 87, no. 2, pp. 358364, Dec. 2002. [6] O. C. Jeong and S. S. Yang, Fabrication and test of a thermopneumatic micropump with a corrugated P + diaphargm, Sens. Actuators A, Phys., vol. 83, no. 13, pp. 249255, May 2000. [7] O. Francais and I. Dufour, Dynamic simulation of an electrostatic micropump with pull-in and hysteresis phenomena, Sens. Actuators A, Phys., vol. 70, no. 1/2, pp. 5660, Oct. 1998. [8] A. D. Stroock, M. Weck, D. T. Chiu, W. T. S. Huck, P. J. A. Kenis, R. F. Ismagilov, and G. M. Whitesides, Patterning electro-osmotic ow with patterned surface charge, Phys. Rev. Lett., vol. 84, no. 15, pp. 3314 3317, Apr. 2000. [9] A. Manz, C. S. Effenhauser, N. Burggraf, D. J. Harrison, K. Seiler, and K. Fluri, Electroosmotic pumping and electrophoretic separations for miniaturized chemical analysis systems, J. Micromech. Microeng., vol. 4, no. 4, pp. 257265, Dec. 1994. [10] A. Richter, A. Plettner, K. A. Hofmann, and H. Sandmaier, A micromachined electrohydrodynamic (EHD) pump, Sens. Actuators A, Phys., vol. 29, no. 2, pp. 159168, Nov. 1991. [11] T. K. Jun and C. J. Kim, Valveless pumping using traversing vapor bubbles in microchannels, J. Appl. Phys., vol. 83, no. 11, pp. 56585664, Jun. 1998.

CHENG AND LIU: ELECTROLYSIS-BUBBLE-ACTUATED MICROPUMP BASED ON THE ROUGHNESS GRADIENT DESIGN

1105

[12] J. H. Tsai and L. Lin, A thermal-bubble-actuated micronozzle-diffuser pump, J. Microelectromech. Syst., vol. 11, no. 6, pp. 665671, Dec. 2002. [13] X. Geng, H. Yuan, H. N. Oguz, and A. Prosperetti, Bubble-based micropump for electrically conducting liquids, J. Micromech. Microeng., vol. 11, no. 3, pp. 270276, 2001. [14] Z. Yin and A. Prosperetti, A microuidic blinking bubble pump, J. Micromech. Microeng., vol. 15, no. 3, pp. 643651, Mar. 2005. [15] J. H. Tsai and L. W. Lin, Transient thermal bubble formation on polysilicon micro-resisters, J. Heat Transfer, vol. 124, no. 2, pp. 375382, Apr. 2002. [16] P. F. Man, C. H. Mastrangelo, M. A. Burns, and D. T. Burke, Microfabricated capillarity-driven stop valve and sample injector, in Proc. IEEE MEMS Workshop, 1998, pp. 4550. [17] S. Bhm, B. Timmer, W. Olthuis, and P. Bergveld, A closed-loop controlled electrochemically actuated micro-dosing system, J. Micromech. Microeng., vol. 10, no. 4, pp. 498504, 2000. [18] A. A. Deshmukh, D. Liepmann, and A. P. Pisano, Characterization of a micro-mixing, pumping, and valving system, in Proc. Solid-State Sens. Actuators (Transducers), Munich, Germany, 2001, pp. 950953. [19] D. A. Ateya, A. A. Shah, and S. Z. Hua, An electrolytically actuated micropump, Rev. Sci. Instrum., vol. 75, no. 4, pp. 915920, Apr. 2004. [20] D. S. Meng and C. J. Kim, Micropumping by directional growth and hydrophobic venting of bubbles, in Proc. IEEE MEMS Workshop, 2005, pp. 423426. [21] M. M. Maharbiz, W. J. Holtz, S. Sharifzadeh, J. D. Keasling, and R. T. Howe, A microfabricated electrochemical oxygen generator for high-density cell culture arrays, J. Microelectromech. Syst., vol. 12, no. 5, pp. 590599, Oct. 2003. [22] K.-S. Yun, I. J. Cho, J. U. Bu, C. J. Kim, and E. Yoon, A surfacetension driven micropump for low-voltage and low-power operations, J. Microelectromech. Syst., vol. 11, no. 5, pp. 454461, Oct. 2002. [23] A. Shastry, M. Case, and K. F. Bhringer, Engineering surface texture to manipulate droplets in microuidic systems, in Proc. IEEE MEMS Workshop, 2005, pp. 694697. [24] J. Lee, B. He, and N. A. Patankar, A roughness-based wettability switching membrane device for hydrophobic surfaces, J. Micromech. Microeng., vol. 15, no. 3, pp. 591600, Mar. 2005. [25] S. Shibuichi, T. Onda, N. Satoh, and K. Tsujii, Super water-repellent surfaces resulting from fractal structure, J. Phys. Chem., vol. 100, no. 50, pp. 19 51219 517, 1996. [26] J. Bico, C. Marzolin, and D. Qur, Pearl drops, Europhys. Lett., vol. 47, no. 2, pp. 220226, 1999. [27] R. N. Wenzel, Surface roughness and contact angle, J. Phys. Colloid Chem., vol. 53, no. 9, pp. 14661467, 1949. [28] A. B. D. Cassie and S. Baxter, Wettability of porous surfaces, Trans. Faraday Soc., vol. 40, pp. 546551, 1944. [29] A. Shastry, M. J. Case, and K. F. Bhringer, Directing droplets using micro-structured surfaces, Langmuir, vol. 22, no. 14, pp. 61616167, Jul. 2006. [30] N. J. Shirtcliffe, S. Aqil, C. Evans, G. McHale, M. I. Newton, C. C. Perry, and P. Roach, The use of high aspect ratio photoresist (SU-8) for super-hydrophobic pattern prototyping, J. Micromech. Microeng., vol. 14, no. 10, pp. 13841389, Oct. 2004. [31] C. G. Cameron and M. S. Freund, Reversible and efcient materialsbased actuation by electrolytic phase transformation, Chem. Eng. Technol., vol. 26, no. 9, pp. 10071011, Sep. 2003. [32] S. Bhm, W. Olthuis, and P. Bergveld, An integrated micromachined electrochemical pump and dosing system, J. Biomed. Microdev., vol. 1, no. 2, pp. 121130, Dec. 1999.

[33] J. Lee, H. Moon, J. Fowler, T. Schoellhammer, and C.-J. Kim, Electrowetting and electrowetting-on-dielectric for microscale liquid handling, Sens. Actuators A, Phys., vol. 95, no. 2/3, pp. 259268, Jan. 2002. [34] L. Gao and T. J. McCarthy, Articial lotus leaf, Langmuir, vol. 22, no. 14, pp. 59986000, Jul. 2006. [35] A. Shastry, A. Epilepsia, M. J. Case, S. Abbasi, and K. F. Bhringer, Bounds on contact angle hysteresis of textured super-hydrophobic surfaces, in Proc. MicroTAS, 2006, pp. 122124. [36] D.-S. Meng, T. Cubaud, C.-M. Ho, and C.-J. Kim, A membrane breather for micro fuel cell with high concentration methanol, in Proc. Hilton Head: A Solid State Sensor, Actuator Microsyst. Workshop, 2004, pp. 141144. [37] J. P. Hoare, Some effects of alternating current polarization on the surface of noble metal electrodes, Electrochim. Acta, vol. 9, no. 5, pp. 599605, 1964.

Chih-Ming Cheng received the B.S. degree in mechanical engineering from National Taiwan University of Science and Technology, Taipei, Taiwan, R.O.C., in 1998, and the M.S. degree in mechanical engineering from Taiwan University, Taipei, in 2000. He is currently working toward the Ph.D. degree at National Tsing Hua University, Hsinchu, Taiwan. He is currently with the Micro-Systems and Control Laboratory, National Tsing Hua University, where he focuses on the development of microuidic switches for bio-analytical devices and micropumps for micro direct methanol fuel cell (DMFC) applications.

Cheng-Hsien Liu (M02) received the B.S. degree in power mechanical engineering from National Tsing Hua University, Hsinchu, Taiwan, R.O.C., in 1987, the M.S. degree in mechanical engineering from Lehigh University, Bethlehem, PA, in 1992, and the M.S. degree in electrical engineering and the Ph.D. degree in mechanical engineering from Stanford University, Stanford, CA, in 1995 and 2000, respectively. While at Stanford, he worked with Dr. Kenny at the Stanford Microstructures and Sensors Laboratory and focused his Ph.D. work on high-performance tunneling MEMS sensors. In 19992000, he was a Senior Electrical Engineer at Halo Data Devices Inc., San Jose, CA, where he focused on the development of microdrives for portable information storage applications. Since Autumn 2000, he has been with National Tsing Hua University, where he is currently an Associate Professor in the Power Mechanical Engineering Department. He also serves as the Department Vice Chair and the Division Head of the Control Division. He currently oversees graduate students in the Micro-Systems and Control Laboratory, whose research activities cover a variety of areas such as biomimetic array chip for bio-object manipulation targeting for tissue engineering/drug screening applications, liver Labchip, advanced tunable MEMS grating, and microsystem robust control.