Contactless acoustic micro/nano manipulation: a paradigm for next generation applications in life sciences

Acoustic actuation techniques offer a promising tool for contactless manipulation of both synthetic and biological micro/nano agents that encompass different length scales. The traditional usage of sound waves has steadily progressed from mid-air manipulation of salt grains to sophisticated techniques that employ nanoparticle flow in microfluidic networks. State-of-the-art in microfabrication and instrumentation have further expanded the outreach of these actuation techniques to autonomous propulsion of micro-agents. In this review article, we provide a universal perspective of the known acoustic micromanipulation technologies in terms of their applications and governing physics. Hereby, we survey these technologies and classify them with regards to passive and active manipulation of agents. These manipulation methods account for both intelligent devices adept at dexterous non-contact handling of micro-agents, and acoustically induced mechanisms for self-propulsion of micro-robots. Moreover, owing to the clinical compliance of ultrasound, we provide future considerations of acoustic manipulation techniques to be fruitfully employed in biological applications that range from label-free drug testing to minimally invasive clinical interventions.


SM, 0000-0002-7423-3611
Acoustic actuation techniques offer a promising tool for contactless manipulation of both synthetic and biological micro/nano agents that encompass different length scales. The traditional usage of sound waves has steadily progressed from midair manipulation of salt grains to sophisticated techniques that employ nanoparticle flow in microfluidic networks. State-of-the-art in microfabrication and instrumentation have further expanded the outreach of these actuation techniques to autonomous propulsion of micro-agents. In this review article, we provide a universal perspective of the known acoustic micromanipulation technologies in terms of their applications and governing physics. Hereby, we survey these technologies and classify them with regards to passive and active manipulation of agents. These manipulation methods account for both intelligent devices adept at dexterous noncontact handling of micro-agents, and acoustically induced mechanisms for self-propulsion of microrobots. Moreover, owing to the clinical compliance of ultrasound, we provide future considerations of acoustic manipulation techniques to be fruitfully employed in biological applications that range from label-free drug testing to minimally invasive clinical interventions.

Introduction
Non-contact manipulation methods form an integral part of modern-day microsystem technology and provide prospects for diverse biomedical applications from tissue engineering to clinical diagnostics. Many of these applications require the manipulation methods to achieve safe, low-power and precise handling of sensitive micro-agents such as cells, bio-molecules and bioorganisms [1]. In order to ascertain such low-power and reliable actuation, these manipulation methods judiciously harvest energy provided by indirect physical forces to enable locomotion of target micro-agents. Various physical mechanisms based on optics [2], magnetism [3], plasmonics [4] and acoustics [5] have been exploited for remote actuation of synthetic and biological agents. Optical tweezers have been the most primitive technology that employs focused laser beams to trap agents across a wide range of sizes (100 nm-1 mm) [2,6]. Besides the costly and complex optical hardware required in optical tweezers, the high-power operation of lasers may damage living or otherwise sensitive biological specimens. Plasmonic tweezers on the other hand, are low power variant of optical tweezers capable of much finer nanomanipulation. However, they require a nano-structured substrate to control light beams for high-resolution traps and also suffer from heat dissipation into the surrounding environment [1,4]. Magnetic tweezers are ubiquitous in DNA manipulation though they provide low resolution for video inspection via microscopy [7]. Moreover, the requirement for high magnetic field or gradient burdens the instrumentation of magnetic tweezers. These trade-offs necessitate a low-power actuation tool that offers a broad range of target size for micromanipulation.
Contactless actuation using acoustics has emerged as a low power, biocompatible and versatile technique that overcomes these aforementioned limitations of other manipulation methods [1,5,[8][9][10]. The techniques employing acoustic manipulation exist in a broad operating frequency range of 100 Hz-100 MHz. Consequently, acoustic actuation techniques can manipulate agents across several orders of magnitude in size, from clusters of nanoparticles up to a millimetresized biological worms [5]. Furthermore, this frequency range complements the operating range of ultrasound imaging devices which makes acoustic manipulation techniques suitable for clinical applications. Importantly, most of the acoustic manipulation techniques use sound intensities of less than 10 W cm −2 on the target agents which makes them safe for sensitive label-free handling of biological agents such as embryos and cells [1].
In terms of versatility, acoustic manipulation techniques can be employed to trap and manoeuver particles as acoustic tweezers [11] or enable fast fluid transport to pump and mix different liquids [8]. Acoustic tweezers are devices that use focused sound waves to trap and manipulate agents. These are typically bulk acoustic wave (BAW) tweezers for three-dimensional (3-D) manipulation in air [12], and surface acoustic wave (SAW) tweezers for two-dimensional (2-D) manipulation in fluid [5]. In case of BAW tweezers, sound waves propagate through the bulk of a medium whereas for SAW tweezers, sound waves propagate along an elastic interface and decay into the surround medium, generally fluids. SAW devices extend from tweezers in confined fluidic reservoirs to devices for flow-assisted patterning of agents in microchannels [9]. Additionally, many acoustofluidic devices employ sound-fluid interaction to indirectly exert forces on target agents. These devices enable localized fluid oscillations to generate an induced flow for long-range transport of both fluid and agents [13]. Overall, the methods described here do not involve any active participation of the agent for their actuation. These technologies benefit from the intelligence, dexterity and efficiency of the actuation process, and the instrumentation behind it. We classify such technologies that function on off-board powering mechanisms as passive actuation techniques.
Various other remote actuation strategies exploit the self-propulsion ability of the agent triggered by their acoustic excitation to specific resonant frequencies [14][15][16]. These agents actively participate in their actuation process by means of an on-board powering unit that mobilizes them remotely. The most prolific of these are agents containing air-cavities or flagella that vibrate to generate localized fluid oscillations which leads to their directional propulsion [16,17]. Design  techniques, material composition or chemical functionalization of the agents post-synthesis. Such actuation strategies contribute towards microrobotic technologies that play a pivotal role in next generation surgical interventions [18,19]. Such technologies are classified here as active actuation techniques. A gamut of previous literature surveys provide a comprehensive assessment of acoustic tweezers based on BAW [10,12], SAW [8,9,20] or both [1,5]. However, a few reports address acoustically actuated autonomous agents and are typically limited to only a single type of agent [15,21]. In this survey, we present a global perspective of the known methodologies in acoustic manipulation technology (figure 1). This survey is organized as follows. Firstly, we discuss the fundamental theory behind different actuation strategies in §2. In § §3 and 4, we cover the aforementioned passive and active actuation mechanisms, and their applications, respectively. Additionally, we describe hybrid actuation strategies in §4 where acoustic manipulation techniques work in tandem with other physical mechanisms, notably magnetic actuation. Thereafter, in §5, we provide our insight over improvements and future directions of the different actuation strategies for their prospective biomedical applications. Finally, we summarize the actuation strategies and our recommendations for their respective technologies that may advance the field of acoustic manipulation.

Acoustic forces acting on a target agent
The propagation of sound through air or liquid medium may generate forces on immersed or suspended agents, known as radiation forces. These forces arise due to the difference in the acoustic properties of the agents with respect to that of the medium [ Figure 2. Acoustic radiation forces acting on agents of size (a) when exposed to sound waves of wavelength (λ) where, k = 2π/λ: (a) scenario ka 1: gradient forces acting on agents under standing waves that results into a potential field, U; (b) scenario ka > 1: scattered forces acting on larger agents under travelling wave propagation; (c) scenario ka < 1: streaming induced drag forces acting on agents, (d) microchannel protrusions showing streaming patterns under vibration. Microstreaming effects around self-propelling agents under acoustic excitation: (e) asymmetrically designed flagellated microswimmer and (f ) bubble-powered micro-propeller. (Online version in colour.) waves may also put the surrounding medium into motion owing to a phenomena known as acoustic streaming, and thereby indirectly influence the agents present in the medium. Nearly, all the analytical models describing the acoustic radiation forces are premised upon the overall largest dimension (a) of the target agent, relative to the acoustic wavelength (λ) [5,9,11]. The general formulation of acoustic radiation force is derived from first-order acoustic pressure acting on the agents, that comprises of gradient and scattered forces. The streaming effects result from second-order acoustic pressure and nonlinearities in the governing sound-fluid interaction [22,23]. As a result, the streaming-induced forces act indirectly on the target agent owing to the localized oscillations in the surrounding fluid. Hence, the overall force on the agent can be described in terms of these components (denoted by respective subscripts) as Hereon, the dominance of above-mentioned force contributions is discussed based on the relation between sound wavelength and size of the agent (figure 2). Specifically, the first term (F grad ) in the right-hand side (r.h.s.) of (2.1) scales as ka 3 , while the remaining terms in r.h.s. scale as ka 6 (where k = 2π/λ) [5,10]. Given this dependency, the gradient forces are dominant in Rayleigh regime (ka 1), while the scattered forces are dominant in the Mie regime (ka > 1). The streaminginduced forces typically apply to agents when ka < 1 while their relative magnitude depends on the workspace dimensions [9,11].

(a) Radiation forces on target agents
When target agents are much smaller than the acoustic wavelength, they can be treated as compressible spheres of diameter (a) to propagating sound waves [10]. In this case, these agents can be manipulated with the application of gradient forces that can be described in terms of an acoustic potential field (U), as [24]: . In (2.2), |p 1 | 2 and |v 1 | 2 refer to the mean value of gradient pressure and particle velocity of the first order, over a time period of sound propagation, respectively. Furthermore, κ * and ρ * describe the compressibility and density for the agent ( * = p) and the fluid medium ( * = o), respectively. In (2.2), κ o and ρ o are related with the speed of sound c as κ o = 1/ρ o c 2 . The gradient forces are prominent in case of standing waves where counter-propagating waves exert these forces on the agent from both the sides. As a result, the targeted agents can be trapped in pressure nodes of minimum potential (figure 2a). This strategy is capitalized in acoustic tweezers based on superimposition of two or more sound waves ( §3a). Furthermore, the contrast in material properties of the agent determines its trapping position in a standing wave field. Case in point, the agents with ρ p > 0.4ρ o move towards the pressure nodes of the standing wave field whereas those with ρ p < 0.4ρ o , navigate towards the pressure antinodes [5]. On other hand, for bigger size of target agent (i.e. ka > 1), the scattered component of radiation forces (F scatter ) overcome the gradient forces (F grad ) acting on it due to their higherorder dependency on size [10]. Importantly, many microfluidic tweezers ( §3b) operate in this regime where agents can either absorb or reflect incoming sound waves (figure 2b). As a result, the travelling waves-based tweezers exploit scattered forces to provide a focused flow of target agents. For comparison, standing waves of kHz order can trap a target agent of 10 µm, while can be made to translate along travelling waves in MHz range.

(b) Streaming effects in microchannels
Besides the radiation forces, acoustic tweezers discussed previously also encounter streaming effects usually dominant near oscillating solid-fluid boundaries. Although as microfluidic workspaces typically have a high surface area compared to overall volume of fluid, streamingdriven forces operate well within the bulk of bounded medium. In several instances streaminginduced drag forces enable out-of-plane manoeuvering of agents for 2-D acoustic traps [25] and vibrating Chladni plates [26]. Fundamentally, acoustic streaming results from the viscous absorption of sound waves as they propagate from a vibrating solid boundary into the fluid medium [5,11,22,23]. As a result, the response of fluid to harmonic sound propagation yields a first-order component (v 1 ) and a dominant second-order term (v 2 ) among other higher order components. This second-order component manifests into counter-flowing fluid fluxes near the vibrating boundary that forms an overall steady flow over a periodic cycle of wave propagation i.e. v 2 . The streaming forces originate from a Reynold's pressure of the form (ρ o (v 1 .∇)v 1 ) and a second-order acoustic pressure field (p 2 ) acting on the fluid. These components arise from the governing acoustofluidics theory (based on Navier-Stokes and continuity equations) that can be simplified to represent acoustic transmission (v 1 ) and resultant fluid flow ( v 2 ) on different time scales [22,23]. Consequently, a boundary-driven streaming originates within a critical depth (δ v ), near the vibrating boundary which further gives rise to a bulk streaming in a bounded fluid channel. Given the dimension w of workspace in direction of sound propagation, the influence of boundary-driven streaming is pronounced when where μ is the viscosity and ω is the frequency of the sound wave. This resultant bulk streaming can induce acoustophoretic motion of agents at MHz range of frequencies (figure 2c). For this case, Muller et al. describes the regime where streaming forces dominate over scattered forces for agents below a critical size relative to workspace considerations (2.3) [22]. A comprehensive overview of various streaming phenomena and their respective regimes has been described by Wiklund et al. [11].
Typically, streaming-induced fluxes can locally capture and manoeuver agents in the vicinity of (δ v ). Although for microchannels with high aspect ratio, these fluxes can lead to a large circulation of agents in the bulk of the channel. This boundary-driven streaming can however be enhanced with the provision of various indented structures along the boundary such as sharp protrusions [23] or air-bubble pockets [27] (figure 2d). Vibration of these structures produce counter-flowing vortices in their near field which results in a steady flow ( v 2 ) in the far-field of the channel. Application of this method involves design and modelling of these protruding structures on the boundaries of targeted workspace [23]. This strategy has been used for directional fluid flow and long-range transport of target agents ( §3c).

(c) Streaming-induced forces on target agents
Going beyond workspace geometry, intentionally designing the agents with vibrational units can enable their autonomous propulsion based on streaming forces. These self-propelling agents comprise of either an asymmetric profile [14] or trapped air bubbles [16] that vibrate to acoustic frequencies to produce microstreaming around them (figure 2e and 2f ). For such agents, the radiation forces are calculated by integrating the second-order pressure and a term analogous to Reynold's pressure, around a boundary enclosing the agent as where n is an outward normal to the surface of the agent and S defines the boundary enclosing the agent [10]. The first term in the r.h.s. of (2.4) describes the microstreaming component whereas the second term describes the vibrations induced in extended agents that can be remotely actuated as swimmers [28] ( §4a). Autonomous agents with trapped air bubbles encounter radiation forces acting on the bubbles, known as Bjerknes forces, that translate them in the direction of sound propagation [29]. In their case, stable oscillations of bubble at high frequencies induce a microstreaming flow ( v 2 ) around them analogous to the case of extended swimmers [11]. Agents with one side of the bubble exposed to the fluid produce asymmetric microstreaming around them which imparts a propulsive force in the direction opposite to the flow [30]. These bubblepowered agents propel at their resonant frequencies (f o ) dependent on the predefined geometry of their air-filled cylindrical cavities, and can be expressed as where P o is the ambient pressure in the liquid, γ is the adiabatic index, L b is the length of the trapped bubble and L is the total length of the cylindrical cavity [17]. Lastly, the sound scattered from nearby bubbles or particles can exert a secondary radiation force on the agent which enables them to carry other objects ( §4b). Hereon, we investigate different acoustic manipulation strategies with regards to the aforementioned theory of radiation and streaming-induced forces.

Passive agents and actuation strategies
Passive actuation strategies comprise of off-board powering methods to energize the target agent that either remains inactive or does not contribute to their actuation process. These strategies either create pressure disturbances in the medium to trap the agents, or exert forces on them based their relative acoustic contrast to the medium. The target agents usually comprise of microparticles, cells and biological specimens like Zebrafish embryos, pollen grains and immobilized Caenorhabditis Elegans. Since these agents lack the autonomy in their physical manipulation, the burden of manipulation entirely lies on the actuation system. Hence, we explore passive acoustic manipulation strategies where the innovation revolves around the following factors: -Instrumentation, system integration and programmability of acoustic actuation units; -Design and synthesis of the target workspace for manipulation; -Numerical modelling of sound-fluid interactions occurring at interfaces.
(a) Acoustic tweezing in enclosed workspace The most primitive idea to exploit acoustic energy for manipulation is to suspend particles [31] and droplets [34] above an acoustic transduction platform and manoeuver them as they levitate in the air. The principle of operation here is to apply gradient forces to trap the target agents in a bounded workspace, which can be both air and liquid medium. Given that gradients forces are dominant in Rayleigh regime, i.e. ka 1, agents in the size range 1 µm-1 mm can be manipulated with systems operable at frequencies greater than 40 kHz. Based on the agent size and target application, the manipulation techniques can be broadly classified as BAW and SAW-I tweezing, respectively. Generally, BAW tweezers employ standing waves or focused travelling waves from multiple sound sources to levitate objects in air [35,36,41]. In addition, there exist BAW-based resonant fluid chambers where standing waves are used for bulk patterning of micro-agents [33]. SAW-I tweezers on the other hand, operate on bounded fluid reservoir whereby incoming sound waves are coupled from one or multiple sound sources into the fluid via a shared substrate [25,[46][47][48]. In both techniques, the design and distribution of acoustic transducers play a crucial role to ascertain controlled manipulation of the target agents. BAW devices generally employ array of commercially available cm-scale piezoelectric transmitters [31], ceramic sheets [33] or metallic transducers [32,92] for long-range transport of mm-sized agents. SAW-I devices often require planar micromachined transducers synthesized using photolithography, and are aimed at tissue engineering and laboratory-on-a-chip applications [9]. The distribution of transducers varies from rectangular [25,31] to circular [40,42,50] morphology to trap particles at desired focal points like an optical lens. Owing to this functional similarity with optical tweezers, this class of manipulation devices are referred as acoustic tweezers [1,25,41]. Acoustic tweezers have been employed in numerous applications like crystallography [93], cell cultures [92], contactless micro-mixing of droplets [34] and bio-printing [25] to name a few. Hereon, we discuss acoustic tweezing methods employing BAW and SAW that are confined to manipulation of agents in a enclosed workspace, which consist of an air-filled space or a bounded fluid reservoir. We focus here on the requisite instrumentation in terms of different configurations of transducers and the desired pressure field available for manipulation.
(i) Bulk acoustic manipulation BAW tweezers are devices that are designed to manipulate 100 µm-1 mm sized particles, typically in air, with the operating frequencies of the transducers in the range of 500 Hz-100 kHz [12,34], with the exception of BAW resonators that operate at MHz frequencies [33]. BAW tweezers hold the ability for long distance transport (up to 100a) and assembly of macro-scale payload using multi-transducer arrays [31,45]. The central idea with the instrumentation here is to develop reconfigurable devices that provide a dynamic 3-D pressure field with the ability for spatial modulation of acoustic trap. Such a dynamic field enables real-time spatial manoeuvering of acoustic traps based on frequency modulation [38], phase modulation [39] or a combination of both [41]. The most ubiquitous of these systems are the ultrasonic phased arrays which can levitate and manoeuvre trapped agents across different focal points by varying phase difference between the transducers [31,34,45]. In terms of their configuration, the transducer arrays can be single-sided or double-sided, with planar or spherical morphology of transducer distribution. The double-sided configuration is the most traditionally followed setup where a transducer and  reflecting substrate [34,37] or mutually orthogonal transducers [31] form pressure nodes between them (figure 3a). On the other hand, the one-sided configuration typically consists of a planar, spherical or cylindrical arrangement of transducers [41] (figure 3b) and Bessel beam manipulators [40] with a focal point providing greater accessibility to the workspace. With the advent of rapid prototyping and fabrication techniques, the one-sided Bessel manipulators have matured into more intricately designed monolithic transducers that function as acoustic vortex traps [42] and holographic lenses [43]. The double-sided transducer configuration offers a confined volume which presents a bottleneck for the targeted specimen to fit in the limited available space for manipulation. Whereas the one-sided configuration has the advantage of a greater open volume for biomedical applications. Overcoming this limitation, a double-sided cylindrical array of transducers has been recently demonstrated as an end-effector of a robot to accomplish pick-and-place tasks opening possibilities for a mobile and remote actuation [45]. For both the configurations, the spherical morphology of transducer array offers a higher acoustic trapping force as compared to the planar array [35].
Furthermore, the optimization strategies for phase adjustments in mapping the acoustic fields led to the evolution of acoustic tweezers as holographic devices that offer additional degrees of manipulation [36,41]. With this speciality acoustic tweezers possess the ability to levitate, and orient multiple agents simultaneously. Similar to the working principle of their optical counterpart, the holographic acoustic tweezers offer a more promising future for biological applications as they provide higher radiation force while consuming less power [36].
(ii) Surface acoustic manipulation SAW-I devices operate at a higher and more extensive frequency range compared to BAW devices i.e. 10 kHz to 50 MHz. Thus, they are able to create a finer grid of pressure nodes to trap agents down to nanoparticles. They comprise of planar pairs or array of transducers that typically enclose a bounded medium over a shared substrate. This bounded medium may comprise of a fluidic reservoir like PDMS microchannels [25,[46][47][48]50], or simply a clamped metal substrate such as Chladni plate [26,52]. The most prevalent SAW-I devices are fluidic tweezers that are typically composed of a pair of inter-digital transducers (IDTs) deposited over a piezoelectric substrate like LiNbO 3 (figure 4a). Akin to BAW devices, SAW-I tweezers also exist as Bessel beam manipulators [47,50], and provide dynamic fields governed by phase or frequency modulation of different IDTs [46,48]. SAW-I tweezers are essential for microbiology and tissue engineering   applications premised upon contactless handling of cell cultures [25,46,47] and fragile entities like embryos [48].
In addition to gradient forces, SAW-I tweezers also capitalize on streaming effects conditional to workspace geometry and acoustic wavelength as discussed earlier. For instance, micro-particles on a vibrating Chladni plate experience both in-plane acoustophoresis and an out-of-plane motion tendency owing to streaming-induced drag forces [26]. Interestingly, these out-of-plane streaming forces can levitate the target agents thus enabling 2-D tweezers to perform 3-D manipulation [25]. Furthermore, this phenomenon of streaming-induced acoustic traps have led to the development of digital acoustofluidic devices whereby a carrier droplet can be manoeuvered in a 2-D grid driven by hydrodynamic acoustic traps [48,51]. A variety of SAW-I tweezers for droplet manipulation has been discussed elsewhere [20].
Recently, more programmable configurations of SAW-I devices have been reported that enable precise path planning of target agents based on multiplexing actuators [46,48]. With advancements in photolithography, such a multi-nodal inter-digital transducer-based network has been demonstrated as an acoustofluidic micro-pump (figure 4b and 4c). As SAW devices overlap with microfluidic tweezers, we treat them as a separate sections where a continuous sheath flow in microchannels replenishes the target agents for manipulation.

(b) Microfluidic tweezers
SAW-II devices integrate microfluidic technologies with acoustic actuators where an external fluid flow enables continuous batch processing of micro-agents dissolved in solvents. These devices can be further classified into standing surface acoustic wave-based (SSAW) and travelling surface acoustic wave-based tweezers (TSAW) based on their ability to provide gradient (ka 1) or scattered radiation forces (ka > 1), respectively. Similar in working principle to SAW-I devices, SSAW tweezers consist of double-sided IDTs enclosing a microchannel, bonded together on a substrate (figure 5a). Their operating frequencies are in the range of 100 kHz-100 MHz [53]. TSAW tweezers on the other hand, can operate with a one-sided IDT, usually with a curved morphology, adjacent to a microchannel (figure 5b) [54]. Their curved transducers provides the ability to focus sound waves on to a target agent as they propagate, and hence require a higher operational frequency of 100 MHz-1 GHz [55]. A detailed account of various IDT morphology and their functionalities is presented elsewhere [56]. The targeted applications of SAW-II devices include cell sorting [  The principle of operation for both SSAW and TSAW is premised upon the acoustic contrast of the target agents with respect to the fluid. Specifically, SSAW devices provide patterned segregation of agents based on differences in density in a multi-component fluid in the presence of a sheath flow [53,58,60]. Likewise, TSAW devices are employed as a more focused cell sorting technique in a sheathless manner, thus preventing any undesired dilution of the carrier fluid [59,73]. Additionally, TSAW devices have also been proven to be an effective tool to generate and channelize more compressible agents like droplets [62]. In comparison, SSAW devices possess excellent segregation ability while the TSAW devices provide a natural flow focusing ability to manipulate both the agents and the fluid [5]. Hence, various multi-stage SAW-II have been pursued as a two-step solution for high throughput agent segregation [59,66,74]. These SAW-II devices either consist of two stages each of SSAW [74] and TSAW tweezers [59], or a combination of both the device types (figure 5c).
In recent years, many SAW-II devices have been commercialized into industrial products like Ascent Bio-Nano TM [70] and AcouSort TM [66,71]. A commonly reported configuration of AcouSort TM device enables segregation of bio-analytes from blood plasma in a multi-step fashion [66,67,71]. Recently, a variant of this aforementioned device has been reported that compensates for undesired streaming occurring in microchannels by optimizing its co-flow during the segregation [71]. On the other hand, various other SAW-II devices also capitalize on streaming effects [69,75,76]. They generate acoustically induced vortices for manipulation using focused IDTs (figure 5d). Next, we discuss acoustofluidic streaming-driven devices that explicitly generate streaming-induced vortices for flow-assisted micromanipulation.

(c) Streaming-driven acoustic tweezers
Many of the previously discussed devices harness streaming phenomenon as a secondary source of actuation for micromanipulation. Some prominent examples include droplet microfluidics [48], Chladni plates [26], focused TSAW tweezers [61,75,94] and acoustophoresis in microchannels [22]. Streaming-driven acoustofluidic tweezers are devices that are designed to exploit acoustic streaming to manipulate both the surrounding medium and agents immersed in it. These devices typically comprise of a microfluidic workspace with protruding vibrational structures on its boundaries, bonded on a shared substrate with a transducer. These vibrating structures produce the streaming-induced vortices that create an oscillatory fluid flow in bulk of the microchannels [23]. In contrast to SAW devices, streaming-driven tweezers rely on the workspace design more than that of the transducers. As a result, the scope of innovation in these devices shifts from sophisticated IDTs to construction of these specialized microchannels. Overall, these devices employ commercially available piezoelectric transducers that operate in the range of less than 100kHz to manipulate nm-µm size particles. Important applications of streaming-driven tweezers include transport of cells [77], pollen grains [80], manipulation of bio-organisms [78,79], micro-mixing [27,81], pumping [13,[82][83][84] and enrichment of fluid mixtures [85,86]. The vibrational structures in streaming-driven devices comprise of either sharp protrusions [13,[86][87][88] or air-filled cavities [79,80,[83][84][85] indented at the boundaries of the microchannel. These structures can be incorporated in the bulk [27,81,89] or side-walls [78,79] of the channels to generate the requisite streaming effects. Broadly, streaming-driven devices exist in two major variants. The first generation of devices incorporate symmetric vibrational structures (figure 6a). These devices are primarily intended for localized mixing of fluids [27,86] or studying morphology of bio-species [78][79][80]. Here, the symmetrically indented structures to the microchannel walls provide counter-flowing vortices with no net directional flow in the channel. Whereas the second generation streaming-driven devices introduce an asymmetry in the design and distribution of vibrational structures to enable long-range transport of agents and fluids (6b). These devices contain tilted side-wall protrusions [13,87,88] and bubble columns [84,85] that offer a more directional fluid transport for micro-pumping applications. Furthermore, frequencyselective vibration of these tilted structures enable bi-directional steering of target agents inside these channels [83,88].
Recently, streaming-driven devices have incorporated long microfluidic trails as an extensive workspace [84][85][86], and use of active tethered components inside the channel for higher volumetric resolution [82,90]. Many of these devices have been synthesized with periodic cyclic microchannel trails that facilitate multi-stage treatment of fluid samples for their enrichment [84,85] and purification [86]. On the other hand, a sophisticated assembly of IDTs within the workspace streaming-driven tweezers has been reported to achieve a pumping precision of nL s −1 [82]. In addition to these developments, streaming-driven devices have also been engineered with dynamic components such as acoustofluidic gears or rotors [89,91]. The fabrication strategy behind these acoustically powered gears later formed the basis for untethered micromachines based on acoustic streaming principles.

Active agents and actuation strategies
Previously described actuation methodologies apply on passive target agents that do not possess any autonomy for self-propulsion. Additionally, these passive agents lack the ability to cope with unpredictable conditions that are likely to be encountered in in vivo operations like fluctuations in pressure or fluid flow in their surrounding workspace. As a result, the collective burden of manipulation falls upon the accuracy of the actuation system and pre-existing knowledge of the workspace. This limitation makes the actuation process less compatible to in vivo studies and can be overcome with either designing agents with an on-board actuation unit or devising actuation strategy to be agent specific. Addressing such concerns, this section describes autonomous propulsion of active micro-agents, namely -Self-propelling micro-agents based on standalone acoustic vibratory units [95][96][97][98][99][100][101][102][103][104][105][106][107][108].
The challenges associated with actuating these micro-agents shift from instrumentation and workspace design, towards synthesis and functionalization of the agents. Specifically, the photolithography techniques used in synthesis of IDTs mentioned earlier, have now been capitalized to design and develop multi-component artificial agents. Hereon, we survey different designs of active micro-agents with focus on their fabrication process and propulsion mechanism.
(a) Self-propelling acoustic micro-agents Self-propelling micro-agents are specifically designed with geometrical shapes to enable their acoustic steering or functional units that vibrate to assist their locomotion. These microagents exist in diverse size and morphology, from nm-sized cylindrical rods to millimetre-sized assembly of vibratory micro-cavities. We classify these agents based on the underlying physical mechanisms, a majority of which are based on the streaming-induced forces ( figure 2e and 2f ). Specifically, they comprise of metallic nanorods [95][96][97][98][99][100][101][102][103], flagellated microswimmers [14,28,104] and bubble-powered micro-propellers [16,17,[105][106][107][108]124]. Generally, these agents are actuated using commercially available piezoelectric transducers attached to a shared substrate that leaks sound waves into the medium in which they are immersed. Herein, we first describe the nanorods that rely partly on MHz range passive acoustic trapping while their directional self-propulsion is aided by their shape anisotropy. Thereafter, we subsequently delve into the other two micro-agents that contain self-actuating vibratory components tunable to kHz range frequencies.

(i) Metallic micro and nanorods
One of the first self-propelling agents capitalized on autonomous motion of axis-symmetric metallic micro-and nano-rods while they are acoustically levitated [95][96][97][98][99][100][101]. These micro-agents are made up of a single [96] or two dissimilar metals [95,97]. The actuation strategy applied here traps these nanorods in a nodal plane, while the in-plane acoustic streaming effects contribute to their propulsion [102,103]. The in-plane forces acting on these nanorods, arise owing to a heterogeneity in their profile, either in terms of density of materials [95,97]  (Online version in colour.) [97,98]. Furthermore, the secondary forces acting on these rods can also enable their dynamic self-organization in linear or circular arrangements [95] (figure 7a). These micro-agents are synthesized using template-based electrodeposition as cylindrically grown rods. Following their synthesis, they can be loaded with enzymes [100], proteins [99], cells [101] and dyes [99,101], which also enables their real-time fluorescent monitoring. Owing to this complimentarity with funtionalization, nanorods have found a gamut of biomedical applications like intra-cellular delivery [96,100], micro-RNA sensing [99] and toxin removal in blood samples [101]. Despite their promising in vivo applications, metallic nanorods often require them to be accurately positioned at the standing wave nodes. This poses difficulty for the nanorods to be employed in in vivo studies as they are less autonomous compared to other active micro-agents.

(ii) Flagellar acoustic microswimmers
Flagellar microswimmers have been designed to reciprocate the streaming effects of indented protrusions like in the case of streaming-driven tweezers. These swimmers comprise of an acoustically resonant flagellum that mimicks the flagellar dynamics of various biological and magnetic microswimmers during their motion [104]. In addition to acoustic streaming, the propulsion mechanism for these swimmers relies on induced structural vibrations in their flagella in response to incoming sound waves. These vibrations are attributed by the heterogeneous construction of the swimmers, which further amplify the microstreaming effects around the tip of their flagella [28]. Akin to metallic nanorods, the heterogeneity in these agents can be introduced with dissimilar material composition [28] or asymmetric shape of the flagellum [14]. These microswimmers can be synthesized using electrodeposition [28] or traditional UV lithography [14]. Several design variants of the aforementioned swimmers have shown their steerability (figure 7b) [14]. However, a single acoustic microswimmer capable of both translation and rotation is yet to be realized. Moreover, as these swimmers rely on effective coupling of surface acoustic waves to induce structural vibrations, their applications have been limited to microchannels.

(iii) Bubble-powered micro-propellers
Similar to vibrating protrusions, the microstreaming effects around oscillating bubbles described in §2, have been exploited to design self-propelling agents [16]. In this case, the microagents are powered by bubble-trapped columns, which are designed as open cavities of predetermined geometry, that resonate to specific acoustic frequencies. Analogous to flagellar acoustic swimmers, the shape asymmetry in terms of the position and size of cavity in these micro-propeller attributes to steering capability [16,105]. Additionally, actuation of bubblepowered micro-propellers can be selective to multiple frequencies owing to the presence of multiple cavities. As a result, the bubble-powered propellers can effectively perform both translation and rotational motion [105,106] (figure 7c). Alternatively, some instances of bubblebased propellers have been steered with the application of radiation forces using focused ultrasound transducers [124].
The traditional protocol for synthesis of acoustic micro-propellers consists of a standard 2-D photolithography followed by their development to obtain the micro-agents of a predefined cavity size [16,105,106,124]. This is followed by the surface treatment of agents to nucleate air bubbles inside their cavities once they are immersed in the fluid for operation. With the development of maskless 3-D nanoprinting, more complex and intricate geometries in the propeller designs have been reported [124]. Subsequently, larger prototypes of propellers came into existence with multiple micromachined cavities that offer higher degrees of mobility [106][107][108]. As these large propellers could be driven at kHz frequencies (refer (2.5)), they can be visualized using medical ultrasound (US) systems that operate at MHz frequencies. Recently, one such propeller has been reported to be both acoustically actuated and tracked under ultrasound guidance [106].
While many bubble-based propellers have achieved frequency selective steering, standalone acoustic actuation does not always provide adequate frequency range for both translation and rotation. Despite a large variance in bubble column length in many such propellers, only a narrow range of distinctly operable acoustic frequencies are achieved [105,106]. This calls for hybrid actuation approaches that enable independent directional control for these propellers, as will be discussion in the next section.

(i) Magneto-acoustic actuation
Magnetic actuation has been the most ubiquitously employed technique for contactless manipulation across a wide size range of target agents. Likewise, acoustic manipulation methods under magnetic guidance find diverse strategies like assembly of magnetic agents [109], selfpropelling acoustic agents [110][111][112][113][114][115][116], bio-hybrid agents [117] and collective behaviour of microparticles [118]. The principle of actuation in all these strategies employs acoustic forces primarily to either levitate or propel the agents in the presence of an applied magnetic field. With regards to levitation, a BAW tweezer has been combined with a Helmholtz magnetic coil set-up to demonstrate multi-component assembly of mm-sized objects and droplets [109]. This combined actuation set-up (8a), supplies radiation forces to levitate magnetic agents, which can also orient themselves under magnetic guidance. Furthermore, the previously discussed self-propelling agents like metallic nanorods [110][111][112] and bubble-powered micro-propellers [113][114][115][116] have been re-engineered to accommodate magnetic components during their synthesis. In the former case, tri-metallic nanorods [110,111] and nanoshells [112] have been developed which use the streaming due to their anisotropic profile along with magnetic guidance for directional motion. While on the other hand, the bubble-powered magneto-acoustic propellers evolved to provide advanced modes of locomotion aided by magnetic steering [113][114][115][116]. Specifically, they have been demonstrated to navigate complex trajectories [113] and climb along surfaces [114,115] with promising potential for bulk actuation (figure 8b). In addition, they can also perform pick-and-place operations capitalizing the secondary radiation forces to trap and eject particles [114].
In terms of biological applications, many of the magneto-acoustic micro-agents have been used for transport of cells [111,114], bacteria [110] and drug delivery to cancerous cells [112]. An interesting bio-hybrid actuation strategy involves functionalized magnetic Red Blood Cells, that can navigate through various biological media in the presence of both magnetic and acoustic fields [117]. Another bioinspired approach mimicks collective behaviour of neutrophil cells in form of micro-particles that agglomerate under magnetic field, and can be manoeuvered in vasculatures aided by streaming effects [118]. When the magnetic field is switched off, this magnetically triggered formation can disembark to release an encapsulated drug at the target site. Recently, a bubble-powered multi-cavity propeller has also shown its ability for targeted drug delivery under combined magneto-acoustic guidance. Aside from the bubble columns for acoustic propulsion, this propeller also consists of a drug-filled cavity that can be triggered to release its contents at specific frequencies [116].
(ii) Chemo-acoustic actuation Many prevalent strategies of chemotaxis have been combined with acoustics either as a triggering mechanism for propulsion [119,120], or a guiding force for navigation and organization of agents [121][122][123]. The first technique utilizes acoustic excitation of agents to catalyse ejection of chemical fuel as a microscale ballistic tool [119,120]. In this strategy, the agents are synthesized with an encapsulated emulsion that vaporizes on application of focused ultrasound pulse and releases the carrier particles (figure 8c). Whereas the second technique utilizes the acoustic selfpropulsion of metallic nanorods, to achieve bi-directional motion induced by their chemical reaction in the surrounding medium [121,123]. The nanorods used in this approach act as catalyst to the surrounding fluid, whereby their chemical reaction yields an electro-osmotic flow which moves them forward [123]. Alternatively, the self-propulsion of chemical nanorods has been demonstrated with reversible swarm control triggered by acoustic radiation forces [122]. A more detailed account of chemically fueled acoustic agents has been reported previously [15]. However, limited clinical applications of chemo-acoustic agents exist as their dependence on a chemical medium or fuel does not allow their invasive operation.

Future considerations on emerging technologies
So far we have surveyed different contactless acoustic actuation strategies and described the recent advances in the state-of-the-art techniques. The importance of acoustic micromanipulation in life sciences can be deciphered from its versatile applications in biology and medicine. These technologies have the potential to serve as lab-on-chip based diagnostic platforms, and to be used for minimally invasive targeted interventions. With their dual advantage of both particle and fluid manipulation, acoustic tweezers exhibit diverse configurations for label-free handling of both sensitive biological agents and solvents. Acoustically triggered self-propulsion of agents provide autonomous means of actuation, thus paving way for targeted microrobotic applications. Although specific to their application, both passive and active actuation technologies can be vastly enhanced to extend their outreach towards translational clinical studies [1,5,12,21]. Firstly, acoustic tweezers have limited ability to transmit acoustic forces across tissues and in heterogeneous media like the blood plasma. In addition, while a majority of tweezers can accurately deploy multiple agents at their desired location, they seldom have the ability to trigger any functional response from the agents. These limitations necessitate improvements in the instrumentation and application strategy of tweezers. Acoustic tweezers can be employed with specialized agents which can be both manoeuvered to a target site, and triggered to carry out additional functions such as drug delivery and diagnosis. Furthermore, the tweezers may also act on autonomous agents and deploy them closer to the target site from where they can selfpropel more reliably. On the other hand, while actuation of autonomous agents offers intelligent and agent-selective strategies, their standalone operation is less effective. Therefore, the collective action and 3-D actuation of these agents are desirable for targeted drug delivery. Besides these factors, visualization of these micro-agents under in vivo conditions calls for clinically compliant imaging techniques, most notably medical USA. In this section, we evaluate the aforementioned criteria and propose research directions that could precipitate additional innovations in the existing technologies.

(a) Instrumentation and operational efficacy of tweezers
Among acoustic tweezers, BAW devices are typically air-borne systems that manipulate larger agents (mm-scale) compared to SAW devices that operate on micro-particles mostly in fluids. BAW tweezers offer tremendous scope for integration with robotic platforms for navigation and visualization of agents across a large workspace [45]. Single-sided BAW tweezers may be interfaced to a programmable robotic [45] or haptic assembly [125] and manoeuvered around a surgical site to perform targeted operations. Furthermore, the instrumentation of BAW tweezers has evolved from static devices to holographic transducer elements for generation of dynamic high-resolution acoustic traps. Besides the miniaturization of transducers, research directions on use of holographic spatial sound modulators (SSM) can substantially enhance resolution of acoustic traps [126,127]. This technology holds the potential for simultaneous amplitude and phase modulation of acoustic beams using masks with sub-wavelength features.
On the other hand, performance of SAW tweezers depends on the design of IDTs and their ability to couple acoustic energy via piezoelectric substrates. The methods of generating surface acoustic traps have evolved to their phase-and amplitude-controlled re-configurable distributions using multiple concentric IDTs [46]. In this regard, a prospective research direction capitalizes on deep learning methods to generate spatially non-uniform acoustic field distribution for any given workspace design [128]. These approaches broaden the scope for SAW devices to manipulate target agents in dynamic patterns conformable to non-conventional microchannel geometries. Importantly, miniaturization and integration of SAW devices with flexible substrates and electronics can vastly extend their outreach towards in vitro applications [125]. In several instances, SAW technologies have been translated to wearable electrodes that enables deep penetration of acoustic waves to impact neurological studies in animal trials [129].

(b) Design and functionalization of autonomous agents
In case of autonomous agents, the reliable operation and ability to carry payload play a major role for their successful deployment in biological applications. Among these agents, bubble-powered propellers became popular due to their ease of actuation and convenient fabrication protocol. These micro-agents gained a huge impetus due to the developments in 3-D nanoprinting technology which led to a wave of eclectic propeller designs [106,114,124]. The next generation of these micro-propellers may consist of larger arrays of acoustically resonant cavities that can offer multiple degrees-of-motion. Alternatively, a different process flow can be pursued to achieve batch fabrication of much smaller nano-propellers that are equally adept at multi-modal locomotion [130]. Furthermore, in order to overcome bubble instability due to high-frequency oscillations, innovative measures are required in both the design and surface treatment of the propeller cavities. In this regard, secondary structures in the propeller designs can contribute additional surface tension at the bubble interface and impart longevity to their operation [115].
Aside from their synthesis, micro-agents can be functionalized to carry and release drugs or other agents for targeted therapy. In this regard, hybrid actuation strategies can play a crucial role for otherwise non-functional biological agents, as their acoustophoretic motion can be more directional under magnetic or chemical guidance [131]. Various chemo-and bio-functionalized metallic nanorods have been pursued for localized delivery of cells, drugs and RNA strands. Similar functionalities can be attributed to bubble-powered propellers, which can also use secondary radiation forces to carry and release a much heavier payload [114,130]. In addition to microstreaming around bubble columns, other alternative means to harvest acoustic forces on bubbles may be exploited as drug release or actuation mechanisms. Firstly, vaporization of carrier droplets triggered by pulsed ultrasound can serve as a method to release drugs encapsulated inside the propellers [116,132]. Secondly, mimicking US contrast agents, non-spherical modes of bubble oscillations may be exploited to facilitate directional release of drugs coated on the bubble interface of these propellers [133]. Although various such drug-coated micro-agents have been

(c) Three-dimensional micromanipulation of target agents
Various clinical scenarios require the actuated agents to manoeuver through obstacles, cluttered environments and uneven surfaces like arteries which demand their 3-D manipulation. Primarily, BAW tweezers are the only qualified tools that are explicitly designed for bulk actuation of agents. Although SAW tweezers rely on surface transmission of acoustic energy, they can be smartly adapted to perform 3-D or upstream actuation of agents against gravity. Conventional SAW-based strategies may incorporate a more realistic workspace architecture that emulate tubular biological vasculatures [134,135]. A previously suggested method of designing flexible transducers or piezoelectric substrate may allow SAW devices to be attached to these tubular microchannels [125]. Moreover, next generation of streaming-driven tweezers may be envisioned with 3-D topographical features that allow streaming-induced migration of agents along sophisticated contours that imitate side-walls of blood capillaries [136]. Similar to passive agents, many autonomous propellers have been shown to ascend obstacles and curved surfaces. However, in most cases their 3-D actuation of agents is partly aided by hybrid actuation strategies [114,115,130]. A majority of the reported strategies of self-propulsion show minimal dependency on instrumentation. This leaves enormous scope for collaborative studies employing acoustic tweezers for more controlled two-step actuation of autonomous agents. In this regard, these agents can be initially steered to deep-seated locations in artificial phantoms or tissues using focused acoustic beams [133]. Thereafter, the agents can exploit their self-propulsion to reach nearby targets in an overall semi-autonomous fashion.

(d) Ultrasound-based imaging of micro-agents
Another important aspect imperative to clinical application is imaging of these agents in environments opaque to camera-based vision. Most of the actuation strategies reported so far are limited to traditional microscopy and are less suitable for non-invasive studies. With prospects of homogeneous actuation and surveillance, US imaging (MHz-GHz range) offers a suitable alternative for tracking micro-and nano-agents. The transducer arrays in BAW tweezers can be combined with commercially available US imaging probes for synchronized actuation and imaging of agents in biological tissues [137]. We envision the next generation of BAW devices to inherit features of US probes in order to transmit steerable acoustic beams inside the human body. Moreover, employing such specialized US probes on shared basis for actuation and tracking agents can open up opportunities for closed-loop studies in biological specimens. Alternatively, some of the reported technologies used for BAW tweezers may also serve as imaging elements for acoustic holographic displays [127,138]. Among autonomous agents, acoustic micro-propellers complement USA as coated gas bubbles are already in practice as contrast agents. In this context, these bubble-powered propellers may be designed to resonate at acoustic frequencies noninterfering with that required for US imaging. As a result, synchronized use of imaging probes and actuators can also be employed here to achieve ultrasound guided closed-loop servoing of these agents. However, the challenges that undermine US-based servoing of micro-agents is the resolution limit of the imaging probes. Although US probes (GHz range) may even detect up to nanoparticles in a noise-free environment, the tracking accuracy is noise-prone and thus compromised during a real-time clinical operation. One of the ways to overcome this limitation is to deploy groups of multiple agents that can easily be detected as a collective entity. convenient clinical detection. With the ability to handle multiple agents, various acoustic tweezers can pattern or segregate agents with different acoustic properties. Yet many of these tweezers either cannot differentiate between similar agents, or are unable trigger a coordinated response from a specific type of agents. Here, advanced control strategies like iterative algorithms [36] and reinforcement learning [52] can be useful for agent-selective actuation. Nonetheless, the absence of external stimulus from the participating agents can make these advanced strategies less effective. Consequently, the passive collective actuation techniques may either lead to an overactuated hardware or demand high computational power. Again, this provides opportunity for tweezers to collaborate with autonomous agents for a more controlled collective manipulation. Several instances of metallic nanorods and magnetic nanoparticles have shown swarming ability when triggered by USA. More interesting strategies for swarm manipulation may be devised based on secondary acoustic forces acting on bubble-powered propellers. Specifically, a single propeller can potentially capitalize on secondary radiation forces to attract or repel multiple neighbouring agents [130]. Prospective mechanisms may involve large formation of agents based on these secondary forces, that can be further manoeuvered as a swarm using other hybrid mechanisms [135]. Emerging strategies may extend to well-characterized agentagent interactions that enable complex formations of autonomous micro-agents. Such dynamic formations might mimick the state-of-the-art magnetic swarms that can organize themselves as clusters and disembark into isolated agents upon removal of acoustic forces [139].

Conclusion
Our survey presents a panoramic view of acoustic micromanipulation methods targeting agents that range from mm-size size bio-organisms, microrobots and up to nanoparticles. Herein, we entail an eclectic set of devices that enable sound waves to trap these agents or transport them along microchannels, and different autonomous agents with self-propulsion ability. This report describes the governing physics, instrumentation, design, synthesis and control aspects for both the actuation platform and the target micro-agents. We finally aim to promote greater utility of acoustics in life science applications and stimulate collaborative research across different biomedical technologies premised on acoustic actuation.
We envision acoustic tweezers as diagnostic equipment, and self-propelling agents as microrobotic tools for prospective clinical applications. Notably, BAW tweezers are adept at automated manipulation and offer tremendous scope for integration with conventional robotic or haptic platforms. On the other hand, SAW tweezers provide a much higher resolution for localization and are important for cell patterning and tissue cultures. Various SAW devices have the potential to be commercialized into flexible diagnostic platforms that may be transformed into wearable devices. A collaboration between different SAW and streaming-driven tweezers can push the boundaries of their traditional usage in microfluidics, to design of acoustically driven surgical devices [140]. Lastly, we discuss self-propelling micro-agents with on-board vibrational units designed to resonate at specific acoustic frequencies. Besides their acoustic propulsion, these agents can join forces with other physical mechanisms like magnetism to allow for hybrid actuation. Offering prospects of swarm control and detection under medical USA, these agents are a promising tool for minimally invasive interventions. Nevertheless, the full potential of clinically relevant micro-agents has not yet been realized with combined studies on acoustic actuation and imaging of these agents under in vivo conditions. A futuristic surgical suite may combine several of these techniques, where autonomous micro-agents are deployed through a lumen using a streaming-driven tweezers at the target site while their swarms are manoeuvered using BAW tweezers.