Microfluidic flow switching via localized acoustic streaming controlled by surface acoustic waves

We propose an acoustic flow switching device that utilizes high-frequency surface acoustic waves (SAWs) produced by a slanted-finger interdigitated transducer. As the acoustic field induced by the SAWs was attenuated in the fluid, it produced an acoustic streaming flow in the form of a pair of symmetrical microvortices, which induced flow switching between two fluid streams in a controlled manner. The microfluidic device was composed of a piezoelectric substrate attached to a polydimethylsiloxane (PDMS) microchannel having an H-shaped junction that connected two fluid streams in the middle. The two immiscible fluids, separated by the PDMS wall, flowed in parallel, briefly came in contact at the junction, and separated again into the downstream microchannels. The acoustic streaming flow induced by the SAWs rotated the fluid streams within the microchannel cross-section, thereby altering the respective positions of the two fluids and directing them into the opposite flow paths. The characteristics of the flow switching mechanism were investigated by tuning the input voltage and the flowrates. On-demand acoustic flow switching was successfully achieved without additional moving parts inside the microchannel. This technique may be useful for fundamental studies that integrate complex experimental platforms into a single chip.


Introduction
Miniaturizing benchtop experimental setups in palm-size microuidic platforms has been the focus of recent research in micro-total analysis systems (mTAS). [1][2][3][4] The development and integration of fundamental ow control modules, such as pumpless ow actuation and ow-directing valves, in micro-uidic devices remains challenging due to a lack of intermodule compatibility. [5][6][7][8][9] On-demand control over the uid ow direction in a microuidic channel is essential to the design of complex integrated experimental mTAS platforms that simultaneously perform distinct chemical reactions. The realization of all-encompassing microuidic platforms capable of handling multiple biochemical assays on a single chip using polydimethylsiloxane (PDMS)-based pneumatic valves, which harness the elastic properties of the polymer material, has been difficult. 2,10,11 Multiple PDMS layers form overlaying uidic channels separated by a thin PDMS membrane that can be easily deformed to block liquid passage in one channel by applying an external air pressure to another channel. By integrating an additional air pressure control layer, complex ow networks may be built to enable uid ow switching operation. The design and fabrication, with reasonable repeatability, of multi-layered microchannels in large quantities is difficult because the deformation rate of a PDMS wall depends strongly on the material elasticity, which cannot be readily controlled. The need for additional pneumatic pumps, which are typically larger than the microuidic system, renders the experimental setup bulky and resistant to miniaturization.
Previous studies integrated a bubble gate into a singlelayered PDMS microchannel to directly block a specic liquid ow path using an air bubble that did not require stacks of multiple PDMS layers. 9 However, the liquid and gas pressures needed to be controlled individually, which meant that an additional pressurized module was required. Flow switching experiments using external forces have been introduced to build uidic networks on a microchip. A thermocapillary force 12 and surface acoustic waves (SAWs) 13 were used in segmented ow control to switch dispersed liquid plugs in a microchannel. These methods were easily integrated into PDMS microchips; however, these and other efforts have not yet addressed the need to control the direction of a continuous ow in a microchannel.
Acoustouidic miniaturized devices using SAW-based actuation were recently investigated for the dexterous handling of suspended micro-objects [14][15][16][17][18] and uids at the microscale. 3,[19][20][21][22][23][24][25] The use of an acoustic streaming ow produced by highfrequency SAWs is one of the most efficient methods of inducing localized high-speed uid motions using two symmetric microvortices inside a microchannel. 26,27 SAWs turn into a compressional leaky wave as they interact with the uid medium in the microchannel positioned atop the piezoelectric substrate. The SAWs refract at the substrate surface as they radiate energy into the uid, and the leaky SAWs are attenuated during propagation through the uid medium. The acoustic waves' attenuation produces a pressure gradient in the direction of wave propagation that generates a time-averaged body force (F B ) on the uid, thereby producing a streaming ow in the form of micro-vortices. Microvortices produced by SAWs have been widely used in microuidic platforms to mix uids, 26 sort particles, 27,28 manipulate cells, 29,30 drive SAW propulsion devices, 31,32 manipulate droplets, [33][34][35] and pump uids using acoustic ows. 36 In this study, we took advantage of localized micro-vortices to realize a ow switching system that could interchangeably direct the ows of two immiscible uids, such as co-owing aqueous and non-aqueous phases in parallel streams, into two separate outlet ports. The acoustic streaming effect, which has been used previously to drive segmented ows, was utilized here to uniquely control the uid-uid interface in a continuous fashion. Previously, we used a slanted-nger interdigitated transducer (SF-IDT) with an effective aperture size that was comparable to the microchannel width to generate a very narrow beam ($10 2 mm) of acoustic waves that deformed the uid-uid surface and split a droplet into two. 37 A narrow acoustic wave beam was essential for achieving ow switching between two immiscible uids at the microchannel junction, in which the uids contacted one another before being separated by a thin PDMS wall downstream. We categorized four different ow-switching regimes during uid-uid interface actuation by acoustic waves of various amplitudes. The experimental observations and the mechanism underlying the ow switching behaviour could be explained in terms of the strong acoustic streaming ow generated at the microchannel junction.

Experimental section
A schematic diagram of the acoustic ow switching system is shown in Fig. 1. The device was composed of a PDMS microchannel attached to a piezoelectric substrate (lithium niobate (LN), LiNbO 3 , 128 Y-cut, MTI Korea) patterned with bimetallic interdigitated electrode nger pairs (Cr/Au, 300 A/1000 A, Ebeam evaporation process). The bonding between the PDMS channel and the LN substrate was enhanced by coating the LN substrate with a SiO 2 layer. The number of nger pairs was 40, and the SF-IDT aperture was 1 mm. The pitch of the nger pairs (l) ranged from 28 to 36 mm, which corresponded to SAW frequencies (f SAW ¼ c s /l) between 109 and 141 MHz. The effective aperture of the SF-IDT could be estimated as 38 where f w is the working frequency, N is the number of SF-IDT ngers, f H and f L are the highest and lowest frequencies from the SF-IDT, and A 0 is the total aperture. The effective aperture of the SF-IDT was around 85 mm, smaller than the uid-uid interface at the junction. An AC electrical signal was provided by an RF signal generator (N5171B, Keysight Technologies), and the signal was amplied by an amplier (UP-3015, Unicorn Tech.). The SAW voltage was measured using an oscilloscope (DSO-X 2022A, Keysight Technologies). The PDMS microchannel was fabricated using so lithography techniques and was attached to the LN substrate by oxygen plasma bonding. The microchannel width and height were 300 and 130 mm, respectively. The microchannel was composed of two inlets and two outlets, and the two uids were separated by a 50 mm thick PDMS wall. In the acoustic actuation region at the microchannel junction, the wall was removed so that the SAWs could generate an acoustic streaming ow over the uid-uid interface to ip the uid streams. The width of the open area at the junction of the uid-uid interface was 200 mm.
Isopropyl alcohol (IPA, Sigma Aldrich) and FC-40 (3M) were used as two immiscible uids. A syringe pump (neMESYS, CETONI GmbH) with four independent units was used to control the ow rates of the two incoming uids via pumping and the two outgoing uids via ow suction. Suction of the outgoing uid was important for stabilizing the uid ow in the microchannel, especially at the junction. Experimental images were recorded using a high-speed camera (pco.1200 hs PCO camera) attached to an inverted microscope (Olympus IX71).
The interfacial tension between the two uids was measured with the pendant drop method using a tensiometer (Biolin Scientic). The FC-40 pendant drop was formed at the tip of a stainless steel needle (OD ¼ 0.72 mm) immersed in the IPA solution to enable drop-shape analysis. The interfacial tension was measured as g ¼ 5.2 mN m À1 that prevented segmented droplet production during ow switching at the microchannel junction.

Results and discussion
Experimental images of the acoustic ow switching action are shown in Fig. 2. The IPA and FC-40 uids owed in parallel from the bottom to the top of the frame. The ow rate of both uids was 200 mL h À1 each. The SAWs were directed toward the uid-uid interface. As the SAWs propagated into the microchannel and interacted with the uid-uid interface, the free surface began to deform between t ¼ 0-150 ms. The recorded images were captured using an inverted microscope focused at the bottom of the microchannel to reveal changes in the interface geometry. The IPA (right) uid was displaced toward the bottom-le of the microchannel cross-section (looking in the direction of ow) (t ¼ 150 ms). At 300 ms, the IPA stream penetrated the le microchannel and replaced some of the volumetric ow of the FC-40 uid, and a similar portion of the FC-40 moved toward the right microchannel. As the IPA stream blocked the le ow path and the FC-40 stream was deected into the right side, the IPA stream inside the right channel broke entirely (at 450 ms), and the positions of the two uids were completely altered (at 600 ms). The IPA and FC-40 streams were switched as they crossed each other at the microchannel junction when exposed to the SAW.
The mechanism underlying the acoustic ow switching phenomenon was investigated by characterizing the uid stream cross-sectional prole at the uid crossover position. Fig. 3 shows an experimental image (a, top view) and a schematic diagram (b, side view) of the acoustic ow switching. The experimental conditions were same as those described in Fig. 2. The FC-40 and IPA uids owed in parallel from the bottom to the top of the image frame prior to encountering the acoustic eld at the microchannel junction. The experimental image (a, top view) was collected using an inverted microscope. The IPA stream was positioned at the bottom of the microchannel in the b 2 -b 2 0 cross-sectional diagram. The schematic diagrams shown in cross-sectional view in Fig. 3(b) illustrate the ow switching mechanism. The IPA and FC-40 uids were parallel in the microchannel in the b 1 -b 1 0 cross-section. As the uid streams entered the SAW actuation region, the IPA stream bent toward the le ow path while the FC-40 stream entered the right ow path (b 2 -b 2 0 ). The experimental results suggested that the uid streams rotated in the clockwise direction within the microchannel cross-section which is consistent with the direction of the acoustic streaming ow eld produced by a SAW. 39 As the uid streams owed from the b 2 -b 2 0 to the b 3 -b 3 0 congurations, the IPA and FC-40 streams owed into the le and right ow paths, respectively. The two uid streams were completely altered and clearly separated from one another. Their trajectories remained stable aer cross-over. These results suggested that the uid streams rotated continuously in the clockwise direction until they passed the SAW actuation region. The b 3 -b 3 0 cross-section shows that the two uids were again separated by the thin PDMS wall, and no additional uid stream translation occurred.
The clockwise direction of the uid stream rotation could be understood in terms of the acoustic streaming effect. When the SAWs met the uid medium inside the microchannel, the SAWs turned into a leaky wave that propagated at the Rayleigh angle q R from the surface normal direction, described according to where c f and c s are the sound speeds in the uid medium and the substrate, respectively. The typical SAW frequencies used in the microchannel ranged from MHz to GHz. The highfrequency SAWs quickly attenuated in the uid medium as they propagated, while the amount of energy transferred to the uid medium was proportional to the SAW frequency. The high spatial gradient of the acoustic eld intensity was required for direct acoustic streaming generation. Thus, the acoustic streaming was easily generated when the acoustic eld was highly localized and the SAW wavelength was small. The acoustic gradient was generated using focused interdigitated transducers (FIDT) 27 or SF-IDT. The time-averaged body force within a microchannel, induced by acoustic streaming, has been thoroughly studied. 27,40 The body force in the direction of the Rayleigh wave could be expressed as where r is the liquid density, b is the attenuation coefficient in the uid, and v is the displacement velocity. The attenuation coefficient between the piezoelectric surface-uid interface (a) and the uid medium (b) could be expressed as 41 where l is the SAW wavelength at the LN substrate, and u is the corresponding angular frequency. m and m 0 are the dynamic and bulk viscosities of the uid medium. The inverse values of the attenuation coefficients were the characteristic length of the SAW propagation, such that higher attenuation coefficients corresponded to larger acoustic streaming. Collins et al. evaluated the particle displacement magnitude induced by the SAWs by considering the reections at the microchannel boundaries. 27 The rst reection at the microchannel roof was considered to estimate the particle velocity according to v ¼ ux 0 À e ÀaðxÀy tan q R Þ e Àbðy sec q R Þ þ R e ÀaðxÀðhÀyÞtan q R Þ e ÀbððhÀyÞsec q R Þ Á ; where x 0 is the magnitude of the particle displacement at the le corner of the microchannel, h is the microchannel height, R is the acoustic wave reection coefficient, and Z 1 and Z 2 are the acoustic impedance values of the uid medium and PDMS, respectively. The acoustic streaming trends were observed by assuming that the uid medium was FC-40. The time-averaged body force distribution due to the acoustic streaming effects is shown in the upper right corner in Fig. 3(b). The horizontal and vertical axes of the graphs represent the lateral width and vertical height of the microchannel. The magnitude of the body force was normalized to 1. The direction of the acoustic body force was aligned with the Rayleigh angle at which the SAWs penetrated the uid medium. The body force induced by the acoustic streaming was stronger at the bottom-le corner of the microchannel. The body force decreased as the SAWs propagated along the uid medium, inducing ow displacement in the clockwise direction within the microchannel cross-section. The uid velocity was stronger on the le side of the channel and weaker on the right side. The uid streams can be easily displaced; however, to induce a complete switching of the uid streams, it is essential to rotate them 180 in the clockwise direction.
In order to verify the acoustic streaming ow effect, a PIV experiment was conducted by using 1 mm diameter polymer particles suspended in the IPA solution. The SAW voltage was set to a lower value of 4.2 V pp to generate a low velocity . The mechanism by which the fluid streams rotated at the microchannel cross-section was investigated. The time-averaged body force due to the acoustic streaming effect is plotted in (b). The direction of the acoustic body force was equal to the Rayleigh angle at which the SAWs penetrated the fluid medium. (c) The velocity field image measured from the PIV experiment. The bulk velocity of the micro channel was stopped for measuring the velocity field induced by the acoustic field only. For the measurement, 1 mm polymer particles were used while the SAW voltage was 4.2 V pp and the images were taken at 2500 fps. The maximum acoustic streaming velocity was measured as 1.13 mm s À1 . Fig. 4 Experimental images of the acoustic flow switching as a function of the SAW voltage. The flow rates of the IPA and FC-40 stream were equal and set to 200 mL h À1 each while the SAW voltage ranged from 7.2 to 12.6 V pp . The experimental results revealed four distinct acoustic flow switching regimes. At 7.2 V pp , the acoustic streaming effect was not sufficiently strong to push the fluid streams into the opposite channels; however, the IPA stream (right side) assumed a concave shape at the SAW actuation region. As the SAW voltage increased, the IPA stream partially flowed into the left flow path (V SAW ¼ 7.92 and 8.64 V pp .) Stable acoustic flow switching was achieved when the SAW voltage was increased to 10.44 V pp . In this regime, the IPA and FC-40 fluid streams crossed one another at the microchannel junction under the effects of the SAW beam and the acoustic streaming effect. At higher SAW voltages (V SAW ¼ 12.6 V pp ), the acoustic flow switching was still achieved; however, the flow streams sporadically broke under the much stronger streaming flow microvortices in the xz plane. An experimental video illustrating acoustic flow switching by varying the SAW voltage is provided in the ESI I. † streaming ow due to the image acquisition limitation (2500 fps) while the bulk uid ow was stopped. The velocity eld produced by the SAW is shown in Fig. 3(c). The maximum velocity of the micro vortex was measured as 1.13 mm s À1 . As the acoustic streaming ow velocity is proportional to the square of the SAW voltage, we estimated the maximum streaming ow velocity of this system to be 3.32 to 10.16 mm s À1 corresponding to the input voltages of 7.2 to 12.6 V pp (Fig. 4). The bulk uid ow rate used in the system ranged from 300 to 1200 mL h À1 which corresponds to the mean ow velocity of 2.1 to 8.5 mm s À1 . For a given owrate, the input voltage is adjusted to induce a sufficiently strong acoustic streaming ow velocity to rotate the uid streams at the micro channel cross-section before the uid owed past the SAW actuation zone. For example, when a net owrate of 800 mL h À1 (5.68 mm s À1 ) of the two co-owing streams of immiscible uids was interrupted by streaming ow with a maximum velocity of 3.32 mm s À1 at 7.2 V pp input voltage, the streaming velocity was not strong enough to induce a switching ow regime (see Fig. 5). As the input voltage was increased to 9.0 V pp , the maximum streaming velocity of 5.20 mm s À1 approached the average bulk velocity of 5.68 mm s À1 , which resulted in a transition regime. However, the switching regime was only realized when the input voltage was further increased to a higher value of 10.44 V pp resulting in a maximum streaming velocity of 7.00 mm s À1 greater than 5.68 mm s À1 .
The experimental acoustic ow switching results obtained by varying the SAW voltage are shown in Fig. 4. The ow rates of the IPA and FC-40 streams were equal, 200 mL h À1 , and the SAW voltage ranged from 7.2 to 12.6 V pp . When the SAW voltage was low (V SAW ¼ 7.2 V pp ), the acoustic streaming was insufficiently strong to rotate the uid stream at the microchannel crosssection. As a result, the IPA stream deformed to form a convex shape in the SAW actuation region. As the SAW voltage increased (V SAW ¼ 7.92 and 8.64 V pp ), the IPA stream began to deect toward the le ow path; however, the acoustic streaming velocity was insufficient to displace the whole IPA stream to the le side, and the IPA stream was divided among both ow paths. The IPA stream was not fully rotated in the microchannel cross-section (see Fig. 3(b)) before the uid stream passed the acoustic actuation zone. In this regime (transition regime), each uid stream was split amongst the two ow paths. A SAW voltage of 10.44 V pp fully switched the IPA and FC-40 uid streams to the opposite ow paths. Within the SAW working region in the microchannel cross-sectional area, the uid streams were rotated 180 . The ow switching effect remained active when the power was on, and the uid stream restored its ow direction as the power was turned off. An input voltage of 12.6 V pp rendered the uid stream unstable, although the acoustic ow switching effect remained valid. The horizontal microvortex was strong enough to break the uid stream, and the IPA stream continued to break and reconnect. In this regime, the acoustic ow switching worked normally, but the excessive acoustic energy could break the uid stream or make satellite droplets. A video collected from an experimental study is provided in the ESI I. † The acoustic ow-switching trend was investigated by generating a regime diagram, as shown in Fig. 5, by controlling the ow rates and SAW powers. The net ow rate ranged from 300 to 1200 mL h À1 , and the SAW voltage ranged from 7.2 to 12.6 V pp . During the experiment, the two inlet and two outlet ow rates remained unchanged. The outlet ow rates were controlled by applying a negative pressure using the syringe pump. The symbols 'x', 'D', and 'o' represent the normal, transition, and switching regimes, respectively. The sky blue symbol 'o' indicates the unstable ow-switching regime. In the normal regime, the streams did not cross one another, and only displayed a concave shape in the SAW actuation region (Fig. 4. V SAW ¼ 7.2 V pp ). In this regime, the displaced uid streams in the microchannel cross-sectional image were restored until the ow passed the SAW working zone. In the transition regime, the two uid streams were divided into the two ow path and formed a Y-shaped ow pattern. As the SAW voltage increased, IPA stream volume that switched to the le ow path increased  (see Fig. 4, case V SAW ¼ 7.92 and 8.64 V pp , ESI I †). In the switching regime, the uid streams stably crossed to follow the opposite ow path. The transition ow lines were estimated using second-order polynomials (normal-transition line: Q ¼ 17.2V SAW 2 À 371.5V SAW + 2239.2, transition-switching line: Q ¼ À10.2V SAW 2 À 527.1V SAW À 5691.7).
The acoustic ow-switching rise time was measured as shown in Fig. 6. The ow switching speed was determined by the ow switching rising time. At the moment of ow switching, the IPA uid stream disconnected from its original uid stream and bent toward the le ow path. The rising time was dened as the time delay between the moments of the SAW operation tuned ON to the disconnection of the IPA uid stream. The bars and error bars indicate the average and standard deviation of the rising time. The inset gure shows the rising time measured at various ow rates and SAW voltages. The rising time decreased as the SAW voltage increased. The ow rate was not clearly correlated with the rising time, as shown in the inset. These results indicated that the ow switching speed depended solely on the acoustic streaming speed, although the feasibility of ow switching was a function of the acoustic streaming and the ow rate. The average rising time of the ow switching raged from 108 to 1235 ms. In the acoustouidic device, the rising time was limited to 100 ms, and the value may decrease as the microchannel geometry and SAW frequency are tuned.

Conclusions
In summary, we demonstrated an acoustic ow switching device that altered parallel ow streams using the acoustic streaming effect. The high-frequency localized SAWs in the uid medium directly generated a microvortex inside the microchannel. The microchannel used in the device included two parallel ow paths and an open window (junction) that enabled SAW interaction with the uid-uid interface. At the SAW actuation region, the acoustic streaming rotated the uid stream in the microchannel cross-sectional view. As a result, the uid streams were rotated by 180 to alter the positions, and the streams completely crossed to follow the opposite ow path. The mechanism underlying the acoustic ow switching was investigated by considering the acoustic streaming theory in the microchannel. The characteristics of the acoustic ow switching were investigated by tuning the SAW voltage and ow rate. The ow switching phenomena were categorized as the ow switching, transition, and normal regimes. In the normal regime, the acoustic streaming deected the uid stream shape along a convex arc in the SAW working zone. In the transition regime, the uid streams were partially switched and were divided among the two ow paths. On the other hand, in the switching regime, the uid streams were completely crossed in the microchannel cross-section, and their positions were stably switched within one second. The rising time of the acoustic ow switching was measured as a device characteristic. The results revealed that the rising time depended on the acoustic streaming power, with a low correlation with the bulk uid ow rate. The experimental results successfully demonstrated the ow switching operation without the use of additional moving parts in the microchannel. This technique may be useful for fundamental studies of microchannel applications in which complex experimental platforms are integrated into a single chip.

Conflicts of interest
There are no conicts to declare.