A SAW-driven modular acoustofluidic tweezer

Abstract

In surface acoustic wave (SAW)-driven acoustofluidic tweezers (AFTs), most setups are integrated on a piezoelectric substrate for a single purpose, limiting the reusability and versatility of devices fabricated using complex MEMS technologies. Meanwhile, prevalent devices exhibit anisotropy in SAW excitation and propagation, as well as optical birefringence and limited transmittance. This work presents a SAW-driven modular acoustofluidic tweezer consisting of up to four replaceable interdigital transducer (IDT) modules and a function module assembled on a common base. Since the IDT modules are separated, each can be fabricated using the piezoelectric substrate best suited to the requirements. For example, SAWs generated from different directions can simultaneously propagate along the X-axis of 128° Y-cut LiNbO₃, enabling highly efficient excitations. The generated SAWs couple into the function module with excellent optical properties and convert into Lamb waves, which then leak into the microfluidic domain and act on the fluid/particles. All modules are connected via standardized interfaces, eliminating potential instabilities caused by wired connections. The reliability of the setup is demonstrated via particle/cell patterning, separation, and concentration experiments, during which the replaceability and reusability of different modules, and the other advantages of the setup, e.g., simple assembly, ease of operation, and application flexibility, are proven.

---

1 Introduction

Acoustofluidic tweezers (AFTs) are highly promising tools in current lab-on-a-chip technologies.^1,2 Such devices employ the physical effects of acoustic waves, e.g., the acoustic radiation force (ARF), to manipulate particles in biological and chemical samples without the need for physical contact or labeling, enabling precise patterning,^3–6 manipulation,^7–9 separation,^10–16 and concentration^17–19 of particles. Unlike their bulk acoustic wave counterparts, surface acoustic wave (SAW)-driven AFTs do not rely on in-channel acoustic resonance and offer advantages such as selective manipulation and flexible device design.^20–22

In SAW-AFTs, straight interdigital transducers (IDTs) are usually employed to generate 1D/2D traveling/standing acoustic fields in microfluidics;^23,24 typically, multiple IDTs are fabricated on a piezoelectric substrate onto which a microfluidic device is permanently bonded. After several repeats of the experiments, it is usually necessary to refabricate the device to avoid cross-contamination between different batches of samples.^25,26 Considering that IDTs and microfluidic channels have to be fabricated using complex microfabrication technologies, non-reusability of devices results in a waste of time, cost, and labor.^27,28

Efforts to address this problem have focused on separating the microfluidic part and the substrate. Specifically, a “superstrate” that has a bottom layer of glass,^23,24 silicon,²⁹ and soft²⁵ or hard polydimethylsiloxane (PDMS)²⁶ is positioned on a SAW substrate. Wave transmission at the substrate–superstrate interface can be achieved via a couplant such as water,^23,30 uncured epoxy²⁴ or a polymer film,²⁹ or through a PDMS pillar²⁵ or the van der Waals force-induced self-adhesion between the two parts.²⁶ When the bottom layer is a hard material, e.g. glass or silicon, the transmitted vibrations excite Lamb waves^23,29 or trigger transversal resonances in it;³⁰ in case of a soft material, it acts as a waveguide.²⁵ On top of the bottom layer, a microchannel or microchamber is bonded,^24,25,29,30 or a droplet to be displaced can be placed directly.²³ The superstrate is single-use and disposable, while the SAW substrate is reusable.

However, because the frequencies, sizes, and arrangements of the multiple IDTs on the entire substrate are designed for a specific application and cannot be separated, it is difficult for a device prepared for one application to be used in another, and the relatively expensive SAW substrate remains dedicated to a single, well-defined purpose. If the requirements change even slightly, such as when the frequency or size of one of the IDTs needs to be modified, the substrate must be redesigned and remanufactured.

Problems in established SAW-AFTs can also arise from the piezoelectric substrate itself. The substrate material is generally transparent for microscopic observation of the microfluidic region and preferably has a high piezoelectric coefficient to generate SAWs efficiently. For these reasons, piezoelectric crystals, especially lithium niobate (LiNbO₃, LN), were adopted in most studies.^23–27,30 However, due to the anisotropic nature of piezoelectric crystals, the corresponding generation efficiency and propagation speed of SAWs are highly direction-dependent. Consequently, the crystal's cut type and the IDTs' orientations must be carefully chosen. For example, the 128° Y-cut LN allows for optimal efficiency in generating X-propagation SAWs,²⁷ corresponding designs are preferred in establishing 1D fields. When constructing a 2D field, it is necessary to design the orientation of the IDTs based on the slowness curve.^31,32 For example, making the SAW beams in two orthogonal directions propagate at 45° to the X-axis of the crystal can produce equal wavelengths in both directions,^4,5,7 the price is reduced SAW generation efficiency and undesired effects such as spurious modes or beam steering.²⁷

Meanwhile, most piezoelectric crystal substrates exhibit limited optical performance. For example, the inherent birefringence effect of LN substrates can cause heavy shadows in microscope observations. While polarizers can be used to mitigate this problem, they are usually incompatible with high-resolution, high-speed photography and can result in reduced light intensity. In addition, some crystals show limited transmission for certain wavelengths of light; for example, LN exhibits poor transmission of UV-light (which is commonly used in biochemical analysis).

One solution to these limitations is to switch to composite substrates. For example, by depositing ZnO film on a glass plate in the localized regions where the IDTs are to be prepared, SAW propagation becomes isotropic, and since there is no film in the microfluidic region, optical observation is not affected.^19,33 However, the preparation of such films is complicated, and the corresponding SAW generation efficiency is relatively low.³³

A modular acoustofluidic tweezer (MAFT) that enables modular design, fabrication, and assembly of SAW-driven AFTs is demonstrated here. Up to four independent IDT modules, impedance-matching modules, and a glass-bottom microfluidic device (function module) are assembled on a common base. Each IDT module is prepared with a 128° Y–X LN substrate, and it efficiently excites SAWs that couple into the glass. Appropriate modules can be selected to meet practical needs, and no wiring is required to interconnect the modules. This setup is an effective platform for standardization, cost reduction, and versatility of SAW-driven 1D and 2D AFTs, which can be easily applied in biological and testing laboratories to facilitate efficient and convenient acoustofluidic applications.

2 Methods and principles

2.1 Device design and composition

As illustrated in Fig. 1(a) and (b), MAFT consists of up to four IDT modules and corresponding impedance-matching modules, a function module, and a common base including a bus board, a motherboard, and a microscope adapter. The motherboard is the outer frame and “skeleton” of the whole setup, where different components are mounted and assembled. The Video S1† shows the detailed assembly process.

Fig. 1 Illustration of MAFT. (a) The explosion view (1 – impedance-matching module, 2 – IDT module, 3 – function module, 4 – motherboard, 5 – bus board, 6 – microscope adapter). (b) The assembled view (7 – magnet). (c) Connections between 2 – IDT modules, 1 – impedance-matching modules, and 5 – bus board. (d) Photos of three IDT modules (Type-200, Type-300, and Type-420), each with a zoomed-in view of the IDT. (e) Photos of the function modules, including type-chamber, type-channel, and type-droplet modules.

In assembling, first screw the bus board and the microscope adapter to the bottom of the motherboard to establish the common base. The bus board is a PCB board on which the electronics for connecting to the IDT busbars via pogo pins and to the impedance-matching modules via magnetic pogo pins are installed, see Fig. 1(c). The microscope adapter mounts the entire setup on a microscope stage to view the microfluidic area. Adapters can be pre-designed and manufactured for a variety of standard microscope stages.

Next, a function module, designed and manufactured on demand, is placed in the center of the motherboard, with four L-shaped limit slots defining its position. The bottom of the module is a quartz glass (QG) plate having a size of 24 (L) × 24 (W) × 0.3 (T) mm³ or 24 (L) × 24 (W) × 0.17 (T) mm³.

Then, up to four IDT modules can be installed in the limit slots on the sides of the motherboard, with their bottom-front edges in coupled contact with the top edges of the function module via silicone oil. Each IDT module is a 128° Y–X LN plate with a straight IDT fabricated on its bottom surface. As shown in Fig. 1(c) and (d), both busbars of each IDT are designed with large areas for easy connections to the bus board. A PDMS block serving as an acoustic absorber is attached behind the IDT, at the rear edge of the module, to minimize wave reflections. Modules of different designs (frequency, number of fingers, and aperture) can be prefabricated for optional use. In the current version of MAFT, the SAW substrate is rectangular and has a size of 30 (L) × 20 (W) × 1 (T) mm³.

Small magnets are installed under the IDT limit slots and at appropriate locations above the IDTs to ensure tight contact between the IDT modules and the function module, as well as stable connections between the IDT busbars and the pogo pins on the bus board, see Fig. 1(b) and S1.†

The electrical impedance of each IDT is then analyzed, and a prefabricated impedance-matching module is selected to convert its input impedance to 50 ohms. The impedance-matching modules are small circuit boards designed using traditional methods³⁴ and are connected to the bus board via magnetic pogo pins at the corners of the motherboard. External excitation signals are fed through SMA connectors.

In the subsequent demonstration experiments, three prefabricated IDT modules shown in Fig. 1(d), i.e., Type-200, Type-300, and Type-420, generate SAWs at wavelengths of 200, 300, and 420 μm, each with an aperture width of 1 cm. Meanwhile, three types of function modules displayed in Fig. 1(e) are used to demonstrate different applications. A type-droplet module, which is only a single QG plate (thickness: 0.17 mm), is used to concentrate particles/cells within an on-chip sessile droplet. A type-chamber module with a micro-chamber (size: 1800 (L) × 1800 (W) × 52 (H) μm³) bonded on the QG plate (thickness: 0.3 mm) is chosen for 2D particle/cell patterning and manipulation. A type-channel module with a straight channel (cross-sectional size: 900 (W) × 68 (H) μm) bonded to the QG plates (thickness: 0.17 mm), is employed for particle/cell separations in continuous flow; the channels are at an angle of 15° to the edge of the QG plates.

2.2 The working mechanism

When constructing 2D SAW fields, conventional SAW-AFTs exemplified in the left column of Fig. 2(a) inevitably encounter problems such as suboptimal excitation efficiency and beam steering because multiple IDTs are integrated on a single piezoelectric substrate. In MAFT shown in the right column, each IDT is prepared on a separate substrate with an optimal cut type of the crystal. The 128° Y-cut LN used here has a high electromechanical coupling coefficient, enabling efficient excitation of SAWs that propagate along the X-axis of the crystal.

Fig. 2 The working mechanism of MAFT. (a) Comparison between the conventional 2D SAW-AFT and the proposed MAFT. Each IDT module in MAFT generates X-propagation SAWs on the 128° Y-cut LN substrate. (b) Propagation of waves in MAFT. The left case shows the type-chamber/channel configuration, and the right case for the type-droplet configuration. (c) The phase velocity dispersion curves of Lamb waves in the QG plate, where the markers indicate the f·d–c_p pairs of the following three MAFT configurations. (d) Simulation results for wave propagation from different IDT modules into QG plates. The left column presents the wave fields where SAWs are coupled into the QG plate from one side. The right column shows the Fourier spectra of out-of-plane displacements at the top surface of the QG plate (the horizontal axis has been converted to the phase velocity c_p). (e) 2D simulation model of MAFT that generates standing waves in the fluid within the PDMS chamber. The results show the distributions of the in-chamber acoustic pressure amplitude and the out-of-plane displacement amplitude in the substrate. The zoomed-in view of the chamber and the LN-QG coupling region are presented in the bottom row. The green arrows indicate the ARF field acting on Rayleigh particles with Φ > 0.

Propagation of waves in the IDT module and the function module is illustrated in Fig. 2(b). As planar SAWs reach the contact region between an IDT module and the function module's bottom layer, they enter the latter through silicone oil, which has low volatility and is commonly used as couplant in ultrasonic applications.³⁵ Coupling material selection requires balancing functionality and practicality. Though polymeric/PDMS films enable effective acoustic coupling, their complex fabrication processes increase production challenges. Water-based gels suffer from evaporation-induced instability, while silicone oil provides instant, pretreatment-free coupling with non-volatility for sustained operation. At typical operating frequencies (e.g., 5–20 MHz²⁰), the speeds of longitudinal and transverse waves in QG (c_L = 6053 and c_T = 3817 m s⁻¹) induce wavelengths that are comparable to the plate thickness chosen here. As a result, Lamb waves propagate in the plate and leak into the on-chip microfluid at an angle of θ_R = sin⁻¹(c₀/c_p), where c₀ and c_p are the speed of sound in the fluid and the phase velocity of the Lamb wave mode, respectively.

Lamb waves in the plate are dispersive and may behave in different modes. The dispersion characteristics of the symmetric (S) and asymmetric (A) modes in a plate of thickness d are determined by the governing equations,³⁶

in which p and q are given byHere, ω = 2πf is the angular frequency, k is the wave number. The dispersion curves for QG plates are plotted in Fig. 2(c). At a given frequency-thickness product f·d, multiple S and A modes can exist, each propagating at a different phase velocity c_p = ω/k. Among these modes, the one with the phase velocity closest to the SAW speed (c_SAW) behaves as the dominant mode, i.e., the Lamb wave modes tend to propagate at a speed that matches c_SAW. An explanation of this phenomenon can be found in the supplementary material, where integral transforms are used to examine the generated Lamb modes.³⁷

Finite element (FE) simulations are then carried out to further explain the propagation of waves; the detailed methods are described in the supplementary material. In the left column of Fig. 2(d), three configurations using different IDT modules (Type-300, Type-420, and Type-200) coupled with QG plates (thickness: 0.3, 0.3, and 0.17 mm) are examined at the working frequencies of 13.3, 9.5, and 19.9 MHz, respectively. In the LN substrates, the vibration amplitude is much higher at the bottom than in other regions, indicating propagation of SAWs. In the first and third QG plates, the top and bottom surfaces vibrate in opposite directions, exhibiting symmetric (S) modes; the second plate shows vibrations that do not vary with thickness, indicating the presence of asymmetric (A) modes. The out-of-plane displacements at the QG plates' top surfaces are extracted for Fourier analysis, and the resulting spectra are plotted as functions of the phase velocity (c_p) in the right column of Fig. 2(d). As the dominant Lamb modes are observed to propagate at 3812, 3290, and 4111 m s⁻¹ in the three plates, the three f·d–c_p pairs are labeled in Fig. 2(c), which verify the S0, A0, and S0 modes being dominant in the three plates, respectively.

The operating mechanism of MAFT is similar to that of conventional SAW-AFTs, except that Lamb waves propagate in the QG plate instead of Rayleigh waves. When two identical IDT modules are placed opposite to each other, a standing Lamb wave (SLW) field is established. After leaking into the microfluid, the waves are capable of trapping and manipulating Rayleigh-sized particles via the ARF, which is determined as^38,39

Here, p_in and v_in are the acoustic pressure and field velocity of the incident field, a is the particle radius, f₁ and f₂ are the monopole and dipole scattering coefficients,^39,40ρ₀ and β₀ are the density and compressibility of the fluid, 〈·〉 denotes time-averaging.

In a 1D standing field, the ARF magnitude is simplified as

F_rad = −πβ₀p_a²ka³Φ sin(2kL), (5)

where p_a, Φ, and L are the amplitude of p_in, the acoustic contrast factor, and the distance between the particle and the closest wave node; k₀ = ω/c₀ is the wave number in the fluid.

Based on the first case in Fig. 2(d), a 2D simulation model is built to illustrate the physical fields in a configuration that generates standing waves in a fluid-filled chamber. Here, two Type-300 IDT modules are operated at a frequency of f = 13.3 MHz, while the type-chamber function model employs a bottom QG layer of 0.3 mm thickness. Fig. 2(e) gives the distributions of the in-chamber pressure amplitude and the out-of-plane displacement in the substrate; also presented are zoomed-in views of the chamber and the contact region between the IDT module and the function module. In the water-filled chamber, a standing wave pattern is observed, which generates an ARF field (indicated by the green arrows) pushing particles with Φ > 0 toward the wave nodes (a minority of the particles can go to the top/bottom boundaries). Vibrations in the LN-QG contact area show wave transmission at an angle θ_R1 = sin⁻¹(c_coup/c_SAW), where c_coup and c_SAW are the speed of sound in the couplant and the phase velocity of SAWs, respectively. Similar simulation results for two other MAFT configurations are presented in the supplementary Fig. S3.†

The particle can also experience the Stokes drag force (DF),

F_drag = 6πηa(v_f − v_p), (7)

where η, v_f, and v are the dynamic viscosity of the fluid, the fluid velocity, and the velocity of the particle, respectively. Small particles (whose size is below a threshold dependent on the device setup^41,42) tend to drift with the fluid, while the motions of larger particles are dominated by the ARF, and those with Φ > 0 (or Φ < 0) tend to move toward the wave nodes (or antinodes).

By placing a sessile droplet at the edge of a traveling wave field, the asymmetric exposure of the waves induces significant acoustic streaming; the DF then enables applications such as concentration of particles within the droplet.^17,18,43 Using two IDT modules on neighboring sides in MAFT makes it possible to generate a stronger streaming vortex from two orthogonal directions, and the concentration process can become more efficient.

2.3 The experimental setup

The fabrication process of the IDT modules and function modules are illustrated in the supplementary Fig. S4.† Each IDT module was fabricated through magnetron sputtering and lithography techniques to deposit IDT patterns (Au/Cr, thickness: 150/5 nm) on a 128° Y-cut rectangular LN. The type-chamber and type-channel function modules were prepared by fabricating PDMS (Sylgard 184, Dow Corning, Midland, MI, USA) microchannels or microchambers on QG plates (JGS2, Feilihua Quartz Glass, Hubei, China) using standard soft lithography and molding processes, and fluid inlets and outlets were prepared according to previous designs.^13,26 The bonding between PDMS and QG was facilitated by a plasma cleaner (Pluto-T, Plutovac, Shanghai, China).

Polystyrene (PS) spheres (Baseline, Tianjin, China) with diameters of 10 μm (red fluorescence) and 5.5 μm (green fluorescence), mouse cardiac fibroblasts (MCFs, Bluefcell, Shanghai, China), human umbilical vein endothelial cells (HUVECs, Bluefcell, Shanghai, China), human promyelocytic leukemia cells (HL-60, Procell, Hubei, China), and human red blood cells (RBCs) were used in the experiments, they are distributed in phosphate buffer saline at a concentration of approximately 8 × 10⁶ mL⁻¹. RBCs were obtained by centrifugation from whole blood samples, which were received from volunteers at the Jiangsu Provincial Hospital of Traditional Chinese Medicine after obtaining informed volunteer consent. The suspensions were injected into the function module via a syringe pump (neMESYS, CETONI GmbH, Korbussen, Germany). In examining the cell viability, calcein-AM/propidium iodide (PI) dual fluorescence staining was used, where viable cells exhibit intracellular fluorescence (calcein-AM: 2 μM) while membrane-compromised cells display nuclear red fluorescence (PI: 4 μM). Fluorescence signals were captured by confocal microscopy (10× objective) and quantified using the ImageJ software (NIH, Bethesda, MD, USA).

The electrical impedance of each IDT was measured with a network analyzer (VNA 2180, Array Solutions, Texas, USA), and appropriate impedance-matching modules were selected to install on the motherboard. The Fig. S5† presents the S11 spectra of the three IDT modules employed in the experiments.

The assembled MAFT setup was placed on an inverted microscope (IX83, Olympus, Tokyo, Japan), and the microfluidic area was recorded with a high-speed camera (FASTCAM Mini UX100, Photron, Tokyo, Japan), with the obtained images analyzed using the ImageJ software. The driving signal for each IDT module is generated using a signal generator (33622A, Keysight, Colorado Springs, CA, USA) before going through a 50 dB power amplifier (325LA, E&I, New York, NY, USA). The vibrations on the top surface of the QG plates were examined using a laser vibrometer (OFV-534, Polytech, Munich, Germany). The temperature at the chamber area was monitored using an infrared camera (FLIR E5xt, FLIR Systems, Wilsonville, OR, USA).

3 Results

3.1 Particle and cell patterning

The MAFT configuration for demonstrating particle patterning is illustrated in Fig. 3(a), where four Type-300 IDT modules and a type-chamber function module are selected. Fig. 3(b) explains that standing Lamb waves are generated in the QG plate, which load ARFs on particles (with Φ > 0) and migrate the latter to the nearby wave nodes.

Fig. 3 Achieving versatile applications with MAFT. The first row shows particle/cell patterning: (a) the device configuration, (b) the working principle, (c) 2D patterning of PS particles and (d) MCFs. The second row is particle manipulation: (e) the device configuration, (f) the working principle, and (g) trajectory of a single PS particle along ‘N’, ‘J’ and ‘U’. The third row for cell separation: (h) the device configuration, (i) the working principle, (j) separation of HL-60 cells/RBCs, (k) the proportion of cells before and after separation (the post-separation data corresponds to samples collected from the upper outlet), and (l) the viability of HL-60 cells in the SAWs OFF and SAWs ON group. The fourth row for particle/cell concentration in a sessile droplet: (m) the device configuration, (n) the working principle, and concentration of (o) PS particles and (p) HUVECs at the center of the droplet. Scale bars: 100 μm.

The sample suspension containing 10 μm fluorescent PS particles was injected into the PDMS microchamber. By applying a driving signal of V_p–p = 33 mV at the signal generator (corresponding to a power input of 24.03 dBm after being amplified) at the working frequency f = 13.05 MHz, the randomly distributed particles were efficiently captured at the grid-like node positions, forming periodical square patterns at periods of about 140 μm, see the fluorescent images captured before and after patterning in Fig. 3(c); Video S2 in the ESI† show the detailed patterning process.

One pair of Type-300 IDT modules were then replaced with Type-420 modules, and the driving signals for them were changed to the frequency f = 9.3 MHz. While everything else remains unchanged, the established SLW fields aggregated MCFs into a periodical rectangular pattern. In Fig. 3(d), the grid periods are about 140 μm and 180 μm in the two orthogonal directions; the entire process for this patterning is given in the Video S3.†

The above operation demonstrates that MAFT easily achieves 2D patterning of particles and cells in a similar way to traditional 2D AFTs. Meanwhile, the configuration can be easily modified according to requirements in one minute.

3.2 Particle manipulation

2D particle manipulation uses a similar MAFT configuration as in patterning. As shown in Fig. 3(e), by adjusting the phases of the standing waves in the two orthogonal directions (x and y), the node network in the field can be moved as a whole. For example, by changing the phases φ_x1 and φ_y1 to φ_x2 and φ_y2 (φ_x2 − φ_x1 < π and φ_y2 − φ_y1 < π), the wave nodes move from the crossings at the white lines to those at the red lines in Fig. 3(f), migrating the captured particles/cells by a distance of (φ_x2 − φ_x1)/(2k) and (φ_y2 − φ_y1)/(2k) in the x and y directions, respectively. It is then convenient to move the captured particles along predetermined trajectories by step-wise adjustment of the phases.⁴⁴

The experiment reused the four Type-300 IDT modules, the corresponding impedance-matching modules, and the type-chamber function module employed in the previous section. A 10 μm PS particle was first captured at one of the nodes, then moved along the trajectories “N”, “J” and “U”. During these processes, the step size of phase adjustment was selected as π/4, and the time step was 5 s. The driving signals were at a voltage V_p–p = 33 mV and a frequency f = 13.05 MHz at the signal generator. The particle positions at different time steps are shown in Fig. 3(g), and the manipulation process is given in the Video S4.†

3.3 Cell separation

Acoustofluidic separation of particles in long straight channels can be realized with standing SAWs (SSAWs), tilted angle SSAWs (TaSSAWs), or phase modulated SSAWs (PM-SSAWs).^15,26 Compared to SSAW devices, the separation distance between two types of particles in TaSSAWs and PM-SSAWs is not limited by a quarter of the SAW wavelength. However, particle separation in PM-SSAWs requires wider channels since the particle trajectories are sensitive to the instantaneous SSAW phase when they enter the working region.¹⁵ Therefore, tilted angle SLWs (TaSLWs) are employed for demonstrating particle separation with MAFT, the corresponding principle in Fig. 3(h) is similar to previous TaSSAW studies.^10,26 As shown in Fig. 3(i), different particles travel in the channel along different trajectories in the field. For particles of the same material, the bigger ones experience larger ARFs and move along a tilted path, while the smaller particles follow the fluid flow and drift across multiple nodal lines. The separation process can be optimized by selecting the appropriate tilting angle, IDT aperture, flow rate, and field pressure⁴⁵ with the help of an established theory.¹⁵

Two Type-200 IDT modules (and corresponding impedance-matching modules) were used here, and the type-channel function module was selected. Two sheath flows (1.2 μL min⁻¹ and 0.6 μL min⁻¹) helped focusing cells in the central sample flow (0.3 μL min⁻¹). Under asymmetric sheath flows, all cells initially left the channel from the lower outlet. With TaSLWs turned on, HL-60 cells and RBCs were effectively separated while the IDTs were driven at V_p–p = 38 mV (24.53 dBm) and f = 19.5 MHz, see Fig. 3(j). The fluorescence image (with HL-60 stained red) is overlaid onto the bright-field image to better illustrate the trajectories. The separation process is presented in the Video S5.†

The performance of MAFT is further explored by analyzing the original sample and the separated one collected from the upper outlet. As is shown in Fig. 3(k), the purity of HL-60 cells increases from 28.73% (pre-separation) to 95.07% (post-separation) in the experiments. The results in Fig. 3(l) further show the cell viability in the SAWs OFF and ON groups being 90.77% and 89.05%, respectively, indicating that TaSLWs did not significantly affect the cell viability (typical images of the stained cells are shown in the Fig. S6 and S7†).

3.4 Particle and cell concentration in a sessile droplet

In sessile droplet-based acoustofluidic applications, fabrication and bonding of microchannel or microchambers are not necessary, and the experiments do not rely on external pumps.^18,46 As illustrated in Fig. 3(m), a droplet is placed at the corner of a square field region. When exciting traveling waves from two orthogonal directions, asymmetric exposure of the droplet in the field induces vortex streaming; see Fig. 3(n) for the top and side views. In such configurations, particles and cells can be efficiently concentrated at the center of the sessile droplet.^17,18,43

A type-droplet function module was first placed at the center of MAFT, with its upper surface hydrophobized to ensure the formation of a circular water droplet; the contact angle at the water/QG interface was about 100°. Two Type-200 IDT modules used above were reused here to generate traveling Lamb waves in two orthogonal directions, both driven with a signal of voltage V_p–p = 55 mV (29.23 dBm) at the frequency of f = 19.5 MHz. Two experiments were carried out, one with 5.5 μm PS particles in a droplet of 1 μL, and the other with HUVECs in a 2 μL droplet.

Initially, particles and cells were randomly dispersed in the droplet. As the wave fields were turned on, rapid concentrations of PS particles and HUVECs inside the droplets were observed, see the results in Fig. 3(o) and (p) and the recorded processes in the Videos S6 and S7.† Therefore, particle and cell behaviors in sessile droplets achieved with MAFT are generally the same as in traditional SAW devices.

4 Discussion

4.1 The efficiency of MAFT

The efficiency of generating Lamb waves in the function module can be influenced by the couplant.^47,48 In the current design, the thickness and length of the couplant layer (d_coup and L_coup) can be adjusted by optimizing the motherboard design. This optimization is achieved via FE simulations studies, where the normalized transmission efficiency is defined as the ratio between the average out-of-plane displacement amplitudes on the LN and the QG surfaces. Fig. 4(a) shows that the couplant thickness ranging from 2 to 5 μm ensures high coupling efficiency, while near-optimal coupling efficiencies are achieved at d_coup = 3 μm and L_coup = 1500 μm. In MAFT, the magnet compression induces a couplant thickness of about 3 μm (obtained by measuring the area of the couplant for a certain volume of silicone oil), and the couplant length is approximately 1500 μm. To examine the stability of the coupling condition, experiments were conducted by repeatedly reassembling MAFT and measuring the displacements at the QG surface; the results presented in the supplementary material S3 further verify the robustness of MAFT.

Fig. 4 The other performances of MAFT. (a) Wave transmission efficiency at different thicknesses and lengths of the couplant layer. (b) Comparison between the simulated pressure amplitudes in MAFT and six LN-based AFTs. (c) Temperature variation at the PDMS chamber with the IDTs driven at different excitation voltages.

Another concern of the readers is the electric-acoustic efficiency of the proposed setup. Here, three MAFT configurations employing two IDT modules and a channel-based function module (as detailed in Fig. 2(d)), utilizing glass substrates with thicknesses of 300, 300, and 170 μm operating at 9.5, 13.3, and 19.9 MHz respectively, were evaluated against two categories of LN-based AFTs: (i) thick LN substrates (1 mm) and (ii) thin LN substrates with thicknesses matched to the MAFT glass substrates (300, 300, and 170 μm). The investigation further incorporates SAW propagation in AFTs along three orientations: X-axis, Y-axis, and 45° off the X-axis. The performance metrics, as summarized in Fig. 4(b), compare the three MAFT configurations (working at corresponding frequencies) to both thick (LN-X, LN-Y, LN-45°X) and thin AFTs (LN-X′, LN-Y′, LN-45°X′), where the simulated in-channel acoustic pressure amplitude is averaged for each configuration and normalized to the reference LN-X AFT. It is found that, MAFT generates comparable acoustic pressure to LN-X AFT while outperforming both LN-Y and LN-45°X configurations. The pressure reduction at 9.5 MHz stems from stronger phase velocity mismatching versus other frequencies, see Fig. 2(c). Notably, MAFT maintains significantly higher pressure than all thin AFTs, making it ideal for applications requiring both high resolution (via thin substrates) and energy efficiency. Unlike conventional 2D SAW-AFTs, MAFT enables optimal wave generation in orthogonal directions without beam steering. When required, its four IDT modules can incorporate different LN cuts or substrate materials.

4.2 Temperature rises in the device

In acoustofluidic devices, increasing the driving voltage can improve the manipulation efficiency by producing bigger ARFs on the targets, but the thermal effect can become non-negligible. Biological particles, such as cells, can be sensitive to high temperatures or large temperature gradients.⁴⁹ Excessive temperature elevation in the fluid can lead to cellular damage or protein denaturation, affecting the biocompatibility of the device.²⁰

Consider the 1D-SLW configuration of MAFT, in which two Type-300 IDT modules and a type-chamber function module are adopted. The temperature variation at the PDMS chamber area was measured at excitation voltages of V_p–p = 33, 45, 60 and 75 mV (corresponding to power levels of 24.03, 28.37, 29.59 and 30.97 dBm after amplification). For each case, the measured temperature averaged over the inner-bottom area of the chamber and three repeated experiments is plotted as a function of time in Fig. 4(c). At the typical driving voltage adopted in the previous experiments, i.e., V_p–p = 33 mV, the temperature rise did not exceed 1.5 °C. When the voltage is increased to as high as V_p–p = 75 mV (the power increased by a factor of 5), the averaged temperature increases from the room temperature (24.13 ± 0.15 °C) to a maximum of 32.27 ± 0.60 °C in 90 s, which is still acceptable for cells. Of course, a cooling plate can be integrated with the MAFT motherboard to achieve better biocompatibility of the device as per requirements.⁵⁰

4.3 Advantages and inspirations

The advantages of MAFT include the following. (a) Modularization and ease of change. Different designs of the functional module and IDT modules can be assembled to challenge new applications. As one or more of the modules can be easily replaced, it is convenient to change the device configuration. (b) Reusability and cost reduction. The IDT modules, which are relatively expensive, can be reused almost indefinitely as long as they are well-kept. This has already been demonstrated in the previous sections. (c) Standardized interfaces for easy connection. Without relying on silver conductive epoxy or other complex wiring methods,²⁸ the base, the IDT modules, the function module, and the impedance-matching modules are connected via pogo pins with the help of small magnets. This design helps avoid problems such as connection instability and deterioration and significantly reduces the time required for experiment preparation. (d) High efficiency for reliability. The power efficiency is near optimal in different directions and the coupling regions, enabling sufficient ARFs to be generated for purposes such as particle manipulation or cell separation. (e) Excellent optical performance. The QG plate enables high-resolution optical observation and high transmittance to UV lights. This allows coordinating MAFT with bio-analysis tools like ultraviolet-visible spectrophotometers or techniques such as UV-curing-based cell printing.⁵¹

There is still work to be done in the future. First, the impedance-matching modules can also be integrated into the base by adopting our previous online impedance analysis and matching solution.³⁴ This can eliminate the need to pre-fabricate different matching circuit modules and have them available for selection after measuring the network parameters of the IDTs. Second, it is important to study how traditional devices such as Petri dishes, multi-well plates, as well as the many commercially available microfluidic chips can be integrated with MAFT. As for the physics and mechanisms of the proposed device, it can be valuable to study how the frequency dispersion nature of Lamb waves can be further exploited to achieve multi-mode wave generations and explore corresponding applications.

5 Conclusion

A SAW-driven modular acoustofluidic tweezer is proposed, which shows the benefits of modularization, reusability, easy operation, high efficiency, and excellent optical integrability. By adopting different configurations, the proposed MAFT easily achieves the functions of a wide range of previously dedicated acoustofluidic tweezer chips; the achievable applications include but are not limited to patterning, manipulation, separation, and concentration of particles or cells. MAFT eliminates some inherent shortages of traditional acoustofluidic tweezers in both acoustics and optics, e.g., anisotropic acoustic propagation, non-identical excitation efficiency, and optical birefringence. The modular design can reduce the cost and labor in professional acoustofluidic laboratories and highly improve the convenience for users unfamiliar with the science of acoustofluidics. Therefore, MAFT has the potential to become a universal platform for acoustofluidic tweezers, forming an important step for acoustofluidic chips to move out of the acoustofluidic laboratories and become effective, commercializable tools in the hands of biologists and in laboratory medicine.

Data availability

The data supporting this article have been included as part of the ESI.† For any additional requests or queries regarding the data, please contact Dr. Xiasheng Guo at E-mail: guoxs@nju.edu.cn.

Author contributions

Conceptualization: D. S. and X. G.; methodology: D. S. and X. G.; investigation: D. S., S. D., F. T., H. Z., and Y. Z.; software: D. S. and Q. W.; data curation: D. S., F. T., H. Z., and Y. Z.; formal analysis: D. S., Q. W., and X. G.; visualization: D. S. and Q. W.; writing-original draft: D. S. and S. D.; writing – review & editing: D. S. and X. G.; resources: D. Z. and X. G.; supervision: D. Z. and X. G.; funding acquisition: D. Z. and X. G.; project administration: X. G.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (Grant No. 82427901, 11934009 and 12374437).

References

1. A. Ozcelik, J. Rufo, F. Guo, Y. Gu, P. Li, J. Lata and T. J. Huang, Nat. Methods, 2018, 15, 1021–1028.
2. Z. Zhao, S. Yang, L. Feng, L. Zhang, J. Wang, K. Chang and M. Chen, Adv. Mater. Technol., 2023, 8, 2202201.
3. J. Shi, D. Ahmed, X. Mao, S.-C. S. Lin, A. Lawit and T. J. Huang, Lab Chip, 2009, 9, 2890.
4. D. J. Collins, B. Morahan, J. Garcia-Bustos, C. Doerig, M. Plebanski and A. Neild, Nat. Commun., 2015, 6, 8686.
5. H. Pan, D. Mei, C. Xu, S. Han and Y. Wang, Lab Chip, 2023, 23, 215–228.
6. Z. Deng, V. V. Kondalkar, C. Cierpka, H. Schmidt and J. König, Lab Chip, 2023, 23, 2154–2160.
7. F. Guo, P. Li, J. B. French, Z. Mao, H. Zhao, S. Li, N. Nama, J. R. Fick, S. J. Benkovic and T. J. Huang, Proc. Natl. Acad. Sci. U. S. A., 2015, 112, 43–48.
8. G. Xu, J. Huang, Y. Zhang, L. Xie, Z. Ni, C. Huang, G. Yao, J. Tu, X. Guo and D. Zhang, Phys. Rev. Appl., 2021, 15, 044037.
9. S. Yang, Z. Tian, Z. Wang, J. Rufo, P. Li, J. Mai, J. Xia, H. Bachman, P.-H. Huang, M. Wu, C. Chen, L. P. Lee and T. J. Huang, Nat. Mater., 2022, 21, 540–546.
10. P. Li, Z. Mao, Z. Peng, L. Zhou, Y. Chen, P.-H. Huang, C. I. Truica, J. J. Drabick, W. S. El-Deiry, M. Dao, S. Suresh and T. J. Huang, Proc. Natl. Acad. Sci. U. S. A., 2015, 112, 4970–4975.
11. L. Ren, Y. Chen, P. Li, Z. Mao, P.-H. Huang, J. Rufo, F. Guo, L. Wang, J. P. McCoy, S. J. Levine and T. J. Huang, Lab Chip, 2015, 15, 3870–3879.
12. D. J. Collins, A. Neild and Y. Ai, Lab Chip, 2016, 16, 471–479.
13. M. Wu, P.-H. Huang, R. Zhang, Z. Mao, C. Chen, G. Kemeny, P. Li, A. V. Lee, R. Gyanchandani, A. J. Armstrong, M. Dao, S. Suresh and T. J. Huang, Small, 2018, 14, 1801131.
14. M. Wu, A. Ozcelik, J. Rufo, Z. Wang, R. Fang and T. Jun Huang, Microsyst. Nanoeng., 2019, 5, 32.
15. Z. Liu, G. Xu, Z. Ni, X. Chen, X. Guo, J. Tu and D. Zhang, Phys. Rev. E, 2021, 103, 033104.
16. M. Rezayati Charan and P. Augustsson, Phys. Rev. Appl., 2023, 20, 024066.
17. Y. Gu, C. Chen, Z. Mao, H. Bachman, R. Becker, J. Rufo, Z. Wang, P. Zhang, J. Mai, S. Yang, J. Zhang, S. Zhao, Y. Ouyang, D. T. W. Wong, Y. Sadovsky and T. J. Huang, Sci. Adv., 2021, 7, eabc0467.
18. T. Peng, C. Fan, M. Zhou, F. Jiang, D. Drummer and B. Jiang, Anal. Chem., 2021, 93, 13293–13301.
19. J. Huang, Z. Zhu, Y. Zhang, J. Tu, X. Guo and D. Zhang, Sens. Actuators, A, 2022, 347, 113901.
20. J. Rufo, F. Cai, J. Friend, M. Wiklund and T. J. Huang, Nat. Rev. Methods Primers, 2022, 2, 30.
21. X. Ding, P. Li, S.-C. S. Lin, Z. S. Stratton, N. Nama, F. Guo, D. Slotcavage, X. Mao, J. Shi, F. Costanzo and T. J. Huang, Lab Chip, 2013, 13, 3626.
22. L. Y. Yeo and J. R. Friend, Annu. Rev. Fluid Mech., 2014, 46, 379–406.
23. R. P. Hodgson, M. Tan, L. Yeo and J. Friend, Appl. Phys. Lett., 2009, 94, 024102.
24. F. Guo, Y. Xie, S. Li, J. Lata, L. Ren, Z. Mao, B. Ren, M. Wu, A. Ozcelik and T. J. Huang, Lab Chip, 2015, 15, 4517–4523.
25. Z. Ma, D. J. Collins and Y. Ai, Anal. Chem., 2016, 88, 5316–5323.
26. S. Zhao, M. Wu, S. Yang, Y. Wu, Y. Gu, C. Chen, J. Ye, Z. Xie, Z. Tian, H. Bachman, P.-H. Huang, J. Xia, P. Zhang, H. Zhang and T. J. Huang, Lab Chip, 2020, 20, 1298–1308.
27. W. Connacher, N. Zhang, A. Huang, J. Mei, S. Zhang, T. Gopesh and J. Friend, Lab Chip, 2018, 18, 1952–1996.
28. S. Yang, J. Rufo, R. Zhong, J. Rich, Z. Wang, L. P. Lee and T. J. Huang, Nat. Protoc., 2023, 18, 2441–2458.
29. J. Qian, J. Ren, Y. Liu, R. H. W. Lam and J. E.-Y. Lee, Analyst, 2020, 145, 7752–7758.
30. C. Witte, J. Reboud, R. Wilson, J. M. Cooper and S. L. Neale, Lab Chip, 2014, 14, 4277–4283.
31. W. Zhou, M. Chen, X. Liu, W. Zhang, F. Cai, F. Li, J. Wu, J. Wang, Y. Wang, X. Huang, Z. Lin, H. Zhou, L. Meng, L. Niu and H. Zheng, Lab Chip, 2019, 19, 3387–3396.
32. Z. Tian, S. Yang, P.-H. Huang, Z. Wang, P. Zhang, Y. Gu, H. Bachman, C. Chen, M. Wu, Y. Xie and T. J. Huang, Sci. Adv., 2019, 5, eaau6062.
33. Y. Fu, J. Luo, N. Nguyen, A. Walton, A. Flewitt, X. Zu, Y. Li, G. McHale, A. Matthews, E. Iborra, H. Du and W. Milne, Prog. Mater. Sci., 2017, 89, 31–91.
34. Z. Jin, L. Huo, T. Long, X. Guo, J. Tu and D. Zhang, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, 2019, 66, 591–599.
35. M. Baudoin, J.-L. Thomas, R. A. Sahely, J.-C. Gerbedoen, Z. Gong, A. Sivery, O. B. Matar, N. Smagin, P. Favreau and A. Vlandas, Nat. Commun., 2020, 11, 4244.
36. J. L. Rose, Ultrasonic Waves in Solid Media, Cambridge University Press, Cambridge, 1st edn, 2004.
37. I. A. Viktorov, Rayleigh and Lamb Waves: Physical Theory and Applications, Springer US, 2013.
38. H. Bruus, Lab Chip, 2012, 12, 1014.
39. M. Settnes and H. Bruus, Phys. Rev. E: Stat., Nonlinear, Soft Matter Phys., 2012, 85, 016327.
40. J. T. Karlsen and H. Bruus, Phys. Rev. E: Stat., Nonlinear, Soft Matter Phys., 2015, 92, 043010.
41. H. Bruus, Lab Chip, 2012, 12, 1578.
42. W. Qiu, H. Bruus and P. Augustsson, Phys. Rev. E, 2020, 102, 013108.
43. G. Destgeer, H. Cho, B. H. Ha, J. H. Jung, J. Park and H. J. Sung, Lab Chip, 2016, 16, 660–667.
44. Z. Liu, H. Zheng, Q. Wei, Z. Wang, Y. Zhang, D. Zhang and X. Guo, Microfluid. Nanofluid., 2024, 28, 36.
45. M. Wu, Y. Ouyang, Z. Wang, R. Zhang, P.-H. Huang, C. Chen, H. Li, P. Li, D. Quinn, M. Dao, S. Suresh, Y. Sadovsky and T. J. Huang, Proc. Natl. Acad. Sci. U. S. A., 2017, 114, 10584–10589.
46. R. M. Trujillo, G. Almanza, D. Sanchez-Saldaña, Ø. Rosand, M. Høydal, M. Fernandino and C. A. Dorao, Lab Chip, 2023, 23, 4773–4782.
47. W. N. Bodé and H. Bruus, J. Acoust. Soc. Am., 2021, 149, 3096–3105.
48. K. Kolesnik, V. Rajagopal and D. J. Collins, Ultrasonics, 2024, 137, 107202.
49. R. Weser, Z. Deng, V. V. Kondalkar, A. N. Darinskii, C. Cierpka, H. Schmidt and J. König, Lab Chip, 2022, 22, 2886–2901.
50. J. Xia, Z. Wang, R. Becker, F. Li, F. Wei, S. Yang, J. Rich, K. Li, J. Rufo, J. Qian, K. Yang, C. Chen, Y. Gu, R. Zhong, P. J. Lee, D. T. W. Wong, L. P. Lee and T. J. Huang, ACS Nano, 2024, acsnano.4c09692.
51. J. Huang, Z. Wang, R. Liu, Z. Zhu, D. Zhang and X. Guo, Microfluid. Nanofluid., 2023, 27, 34.

---

Footnotes

† Electronic supplementary information (ESI) available: The mechanism of Lamb wave excitation, the finite element modeling details, the repeatability of device performance, the assembled view of MAFT, wave transmission from LN into QG, extra results of simulated field distributions, fabrication procedure of the IDT modules and function modules, the network parameters of the IDTs, and videos for device assembling and all demonstrated applications. See DOI: https://doi.org/10.1039/d4lc00924j

‡ D. Zhang and X. Guo are Fellows at the Collaborative Innovation Center for Cardiovascular Disease Translational Medicine, Nanjing Medical University, China.

---