Strategies for single particle manipulation using acoustic and flow fields

Ultrasonics 50 (2010) 247–257

Contents lists available at ScienceDirect

Ultrasonics journal homepage: www.elsevier.com/locate/ultras

Strategies for single particle manipulation using acoustic and flow fields S. Oberti a,*, D. Möller a, A. Neild c, J. Dual a, F. Beyeler b, B.J. Nelson b, S. Gutmann d,1 a

Institute of Mechanical Systems, Dept. of Mechanical and Process Eng., ETH Zurich, CH-8092 Zurich, Switzerland Institute of Robotics and Intelligent Systems, ETH Zurich, 8092 Zurich, Switzerland c Dept. of Mechanical and Aerospace Engineering, Monash University, VIC 3800, Australia d Swiss Light Source at Paul Scherrer Institute, CH-5232 Villigen, Switzerland b

a r t i c l e

i n f o

Article history: Received 16 July 2009 Received in revised form 31 August 2009 Accepted 11 September 2009 Available online 23 September 2009 Keywords: Ultrasound Cell manipulation Particle manipulation Crystal manipulation Single particle manipulation

a b s t r a c t Acoustic radiation forces have often been used for the manipulation of large amounts of micrometer sized suspended particles. The nature of acoustic standing wave fields is such that they are present throughout the whole fluidic volume; this means they are well suited to such operations, with all suspended particles reacting at the same time upon exposure. Here, this simultaneous positioning capability is exploited to pre-align particles along the centerline of channels, so that they can successively be removed by means of an external tool for further analysis. This permits a certain degree of automation in single particle manipulation processes to be achieved as initial identification of particles’ location is no longer necessary, rather predetermined. Two research fields in which applications are found have been identified. First, the manipulation of copolymer beads and cells using a microgripper is presented. Then, sample preparation for crystallographic analysis by positioning crystals into a loop using acoustic manipulation and a laminar flow will be presented. Ó 2009 Elsevier B.V. All rights reserved.

1. Introduction Primary acoustic radiation forces have been successfully exploited for the handling of large numbers of suspended particles of micrometer dimensions, such as beads and biological cells. When an acoustic wave interacts with the particles it gives rise to these forces as a non-linear effect. The wave is emitted into the fluid by vibration of the system containing the suspension [1]. Typically frequencies are sought at which standing wave pressure fields are created, because they permit the collecting of particles at a known location (within the periodic standing wave pattern) with sufficient force amplitudes. For typical excitation frequencies used in such positioning systems (ranging from upper kHz to lower MHz) the periodicity is of a few hundreds of micrometers to a few millimetres, a fact which makes exploitation of such techniques interesting for microfluidic platforms for high-throughput applications. A variety of micromachined devices for separation, fractionation, trapping and positioning of particles and cells, working both in batch [2,3] or fluid flow [4,5] mode, have been proposed in the last few years. On the other hand, in life sciences an increasing interest has been witnessed in the recent past in the study of phenomena tak* Corresponding author. Address: IMES-Mechanik, Tannenstrasse 3 – CLA H23.2, ETH Zurich, 8092 Zurich, Switzerland. Fax: +41 44 632 1145. E-mail address: [email protected] (S. Oberti). 1 Present address: Novartis Institutes for Biomedical Research, Novartis Pharma AG, 4056 Basel, Switzerland. 0041-624X/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.ultras.2009.09.004

ing place at the cellular level [6]. Such experiments require either the formation of aggregates of a known number of cells (i.e. relies on the ability of tracking the position of each cell) or they are aimed at manipulating single cells with special tools, such as microgrippers and pipettes [7]. In a similar way, the sample preparation of protein crystals prior to crystallographic analysis also includes the handling of single units [8]. Crystals (e.g. insulin crystals with sizes up to 200 lm) are individually removed from the solution in which they have been growing and mounted onto nylon loops. As these analysis processes are typically repeated for large numbers of individual particles, for instance, in order to obtain enough statistical data, there is an interest in automating them. A few groups active in the field of acoustic manipulation have recently started exploring this new direction. Svennebring et al. [9] reported on a confocal resonator, which can be operated to create cell aggregates at its focal position, by switching between two modes so that the operator can decide if the incoming cell should be added to the clump at the centre or flow around it and leave the perfusion chamber through the outlet at the opposite side. Furthermore, Manneberg et al. [10] described the creation of levitated aggregates in a small chamber by the use of three dimensional acoustic fields. In both cases aggregates of known number of cells are formed. In the context of single particle manipulation, GlynneJones et al. [11] recently anticipated a system where a particle trapped in an acoustic field is used to exert forces on a DNA strand attached to its surface on one end and fixed on a substrate at the other extremity.

248

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

Here, strategies for the automation of the manipulation of particles using a range of special tools previously presented will be compared [12–15]. Acoustic manipulation will be shown to be capable of being used as a technique for pre-positioning particles in such a way that they can be individually accessed by the external tool, dispensing thus with the need of localization. In order for this to be achieved, all the particles must be made accessible to the tool in a known location and an interface must be provided. The manipulation might be assisted by microfluidics in order to bring the particles into the region where acoustic forces have the highest magnitude or to move them from there towards the location where they are accessed. This presents a novel challenge, in that open access points must be present for the tools which introduce air/fluid interfaces, whilst most ultrasonic systems can be considered as consisting of closed fluidic chambers with the only external openings being ports used to inject or remove the fluid. For this reason extensive modeling has been performed. Two systems are discussed, in Section 2 a system for the semi-automated manipulation of particles and cells using a microgripper is introduced. It will be shown how automation using only a quasi one dimensional standing pressure field is not possible. An extension of this system is therefore presented in Section 3, which uses the addition of a controlled laminar flow created along the channel to enable the control of the longitudinal position of particles. A similar principle has been applied for the manipulation of protein crystals with the device discussed in Section 4. 2. Acoustics assisted manipulation of particles using a microgripper In order for a microgripper to efficiently sequentially grasp single suspended particles, we investigate the pre-positioning of these particles along the centerline of the channel containing the fluid suspension. In this way, the first particle can be accessed by inserting the microgripper along the length axis of the channel with its fingers parallel to the bottom surface of the channel. This is done through an air/water meniscus at the abrupt end of the channel. The meniscus is oriented 90° with respect to the line of particles formed by the ultrasound. Such alignment of particles using ultrasound affords a few advantages: The geometry of the gripper presented here means particles must be in front of it to allow capture, and furthermore, they must not be adjacent to chamber side walls, both these criteria are eased by ultrasonic pre-positioning. Moreover, the handling of subsequent particles requires the return to a fixed location, easing automation procedures. The vertical position is determined by the geometry itself as particles lie on the channel’s bottom surface, pulled there under the effect of gravitation. The microgripper has a capacitively actuated finger which can be pressed against the other one (fixed) and so performs grasping by applying a DC voltage. Its construction and operation is described by Beyeler et al. in [12]. In the first part of this section the system is described, while in the second part the lateral positioning of particles will be investigated by means of the results of a two dimensional finite element simulation of the acoustic field (COMSOL Multiphysics, v.3.4). From the comparison with experimental observations the limitations of this manipulation method as well as of this modeling approach will be highlighted.

on the top of the channel using two component epoxy adhesiveseals it along its entire length, leaving exposed a round reservoir at one extremity. At the opposite end the channel has been cut straight leaving a rectangular water meniscus as boundary, held in place by surface tension forces. This allows access for the microgripper to enter the fluid volume. Excitation to vibration of the solid structure and in turn setting up of an acoustic standing wave field in the fluid (when the resonant condition is fulfilled) is achieved by means of a 500 lm thick piezoelectric plate glued with conductive epoxy on the backside of the silicon wafer covering the whole width of the device, having the same length as the channel, and cut at two locations in correspondence to the channel walls. The original aim of such cuts was that of decoupling the vibration of the piezoelectric confined within them and the rest of the plate, to concentrate all the energy in the channel. From a finite element simulation reported in [13] it is known that this is possible only to a certain extent. A peculiar electrode configuration, introduced previously by the authors in [16], has been adopted here. The time harmonic electrical signal has been applied only to a narrow electrode strip defined on the lower surface of the piezoelectric element, whereas the adjacent area has been grounded together with the electrode on the opposite side of the plate. The resulting actuation is antisymmetric with respect to the y–z plane, passing along the centre of the fluid cavity allowing generation of pressure fields which are either symmetric or antisymmetric. As it will be shown in the following section, antisymmetric fields correspond to an odd number of trapping planes, with always one plane running along the centerline of the channel, as required in order for the microgripper to be able to reach each particle (space is thus provided on either side of the particles for each finger). Furthermore, keeping the active piezoelectric part to a minimum size, together with the fact that the transducer was operated below its (thickness) resonant frequency and considering the fact that in some of the experiments it was kept active only for short periods, leads to a minimization of heat generation. It turned out that the magnitude of the acoustic radiation forces is still large enough to manipulate particles without the need of increasing the applied voltage. During operation the acoustic device was placed under a microscope (Olympus ZSH), so that the manipulation process and the motion of the microgripper could be monitored from above through the glass plate using a CCD camera (Sony CCD-Iris, Model SSC-M370CE, frame grabber for data acquisition Newport, PCIMAQ 1408). The signal driving the piezoelectric transducer was produced by a Stanford Research DS345 function generator amplified by an ENI 2100L. The microgripper was mounted on a xyzstage (Newport MM4006), operated manually with a joystick through a LabView interface used to define its speed. In the experiments 74 lm diameter copolymer particles (Duke Scientific) suspended in de-ionized water and MCF10A cells suspended in buffer solution were loaded into the channel, by depositing a droplet of suspension in the reservoir and waiting until the channel is filled by the effect of capillary forces. Cells were grown in DMEM/F12 medium (Invitrogen AG) supplemented with 20 ngml1 epidermal growth factor, 100 ng ml1 cholera toxin, 10 ngml1 insulin, 500 ng ml1 hydrocortisone and 5% fetal horse serum at 37 °C and 5% CO2. Before use cells were trypsinized to obtain a single cell suspension. Cells in suspension were washed once by centrifugation at 1000g for 5 min and re-suspended in an appropriate volume of phosphate buffered saline (PBS).

2.1. System description 2.2. Prediction of the particles’ lateral position Pre-positioning of particles and cells has been achieved along the centerline of a 5 mm long channel with a 0.2  1 mm rectangular cross section, fabricated by dry etching a 500 lm thick silicon wafer to a depth of 200 lm (Fig. 1). A 1 mm thick glass plate – fixed

In order to determine the particle’s lateral position, the locations of vanishing acoustic radiation force are determined. This is done using Gor’kov approach [17], which requires the calculation

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

249

Fig. 1. Schematics of the device, top view (a), bottom view (b) and close up of the front side (c). It consists of a 1 mm wide channel (1) etched in a 300 lm thick silicon wafer (2), ending in a reservoir (3) at one extremity and cut straight at the other one (4), leaving a liquid meniscus in the x–z plane. Through this interface the microgripper (lying in the x–y plane) enters the channel to remove the first particle in a line along the central channel’s axis. The channel is sealed on the top with a 1 mm thick glass plate (5). A piezoelectric transducer (6) located on the bottom side of the device provides the excitation. It has been cut at two locations (7), parallel to the channel walls. The harmonic signal is applied only to one of the two strip electrodes defined on the lower surface (8). Copper wires (9), attached with a droplet of glue (10) to the silicon, are used to connect the transducer to the signal generator.

of the pressure and velocity distribution in the fluid first. As demonstrated previously in [18], the response to excitation of such an acoustic system can be accurately predicted only by taking into account the whole geometry and not by considering the fluid volume alone (i.e. assuming the fluid to be resonating between rigid walls). In particular, the phase speed of the acoustic standing wave set up in the x-direction results from the coupling between the displacement of the solid structure and that of the fluid. Such a fluid-structure interaction is included in the model. As further boundary condition for the system, all free boundaries have been assumed stress free. Damping has been included by using complex stiffness parameters for the glass and complex speed of sound in the water as given in [19]. Further details about the modeling can be found in [18]. A plot of the pressure distribution across a cross section at 780 kHz and 2140 kHz is shown in Fig. 2a and b, respectively. It can be seen that at the lower frequency a standing wave of approximately half a wavelength is set up across the channel, with the pressure vanishing in correspondence of the centerline (light grey or green in the color version, marked with the dashed line). At 2140 kHz the pressure vanishes at two other locations as well, distant half of wavelength from the centerline. It can also be observed that the pressure amplitude remains constant over the vertical direction, since the channel depth does not exceed half of the wavelength in the fluid, so that no resonance can be set up in that direction. Once the pressure field is calculated, the trapping locations can be determined. They correspond to the stable locations where the primary acoustic radiation forces vanish. As given by Gor’kov in

[17] for a compressible sphere of radius rS (rSk  1, with k being the wave number of the acoustic wave, i.e. for a particle much smaller than the acoustic wavelength) suspended in an ideal fluid in an arbitrary pressure field2 the time-averaged force acting on it is described as:

   1 hp2 i 1 2 F ¼ rU ¼ r 2pr3S qF f  v if h 1 2 3 q2S c2F 2

ð1Þ

with f1 ¼ 1  qF c2F =ðqS c2S Þ and f2 ¼ 2ðqS  qF Þ=ð2qS þ qF Þ, where q is the density and c the speed of sound in the particle (index S) and in the fluid (index F). The rhp2 i-term and the rhv 2 i represent the spatial gradients of the time-averaged pressure and fluid velocity fluctuation and can be interpreted as spatial gradients of the timeaveraged kinetic and potential energy densities, and U is a force potential. The force vanishes where the potential energy is minimal (i.e. p = 0, pressure nodes) and the kinetic energy maximal (velocity is 90°-phase shifted in space with respect to the pressure), i.e. where the overall force potential U has a minimum (the second case in which the gradients/force vanish is at potential maxima, but these are unstable locations). The velocity can be derived from the pressure using the simplified Euler equation @/=@t ¼ p=q where

2 Eq. (1) requires [17] that the field is different from a plane progressive wave – a condition which is fulfilled here – and that the particle is not close to a wall, as multiple scattering is not included (only the incident field and the one scattered from the particle are considered). This last condition is not fulfilled in a micromachined system where particles lie on the bottom cavity surface. Nevertheless it seems reasonable to assume that the change affects mainly the magnitude rather than the shape of the overall force field.

250

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

Fig. 2. Acoustic pressure p (a and b) and force potential U (b and c) set up in a channel cross section (x–y plane) by excitation at 780 kHz (a) and 2140 kHz (b). In the color coding, red indicates the maximum value, blue the minimum. According to Eq. (1) particles collect at the location of minimum force potential, thus in a plane running along the centerline at 780 kHz and three planes (one along the centerline, two distant quarter of wavelength from the centerline) at 2140 kHz. In this one dimensional standing pressure field these locations correspond to the pressure nodes. At the higher frequency the gradient of the force potential across the channel @U=@x is steeper, corresponding to a higher acoustic radiation force amplitude. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

/ is a velocity potential v ¼ r/. It has to be said that the model neglects two major phenomena often occurring in acoustic fields: acoustic streaming and inter-particle forces (also known as secondary radiation forces). As streaming forces scale with the radius of the particles, for the sizes of interest, the motion of these particles is dominated mainly by the acoustic radiation forces (these latter being proportional to the particles’ volume). Secondly, inter-particle forces only become effective when two particles are very close to each other and act to define the exact position of a particle in a clump. However, especially for large particles, the overall shape of the clump is mainly determined by the potential well (i.e. by the primary acoustic radiation force). It has to be added that the glue layers are not included in the model and the way the device is clamped can only be approximated. Fig. 2c and d shows the force potential U for the two frequencies already considered above. It appears evident that the minima of the force potential do indeed correspond to the pressure nodes, as mentioned before. Thus particles are collected in a plane along the channel’s centerline at 780 kHz and in three planes at 2140 kHz, respectively. As there are no significant pressure gradients over the channel depth, no acoustic forces are acting on the particles in the vertical direction and hence particles are expected to be found on the lower surface. Furthermore, since the lateral gradient of the force potential @U=@x is higher at 2140 kHz a force of higher magnitude is exerted on the particles. This frequency can therefore be used to improve the alignment of particles roughly positioned along the centerline by exciting the system at 780 kHz. In Fig. 3 a pick and place experiment is illustrated where three 74 lm particles are successively removed from the channel and deposited on a glass surface placed in front of it, at a lower location. Firstly, as the system has been excited at 780 kHz (b) the particles, at the start randomly distributed on the bottom surface (a), have been gathered in the central region, in a rather crude arrangement. Subsequently in (c), the frequency was switched directly to 2080 kHz, in order to create a tighter line of single particles. As too many particles are present in the potential well, some are seen to jump on the top of each other (these are marked with arrows and

are slightly out of focus) or filling the next level of minimum force potential. In preparation for the microgripper to enter the channel, the excitation is stopped, as this might set the microgripper into vibration as well, causing side effects, such as attraction of particles towards the microgripper fingers (for a detailed description see [13]). The end of excitation can be recognized in (d) as the two marked particles in (c) on top of other ones have fallen on the bottom surface, sliding besides the lower ones. Next, the microgripper is entered into the channel 20 lm above the lower surface, is moved at a velocity of 0.2 mm/s to the position of the first particle and the fingers are actuated by applying 90 V (e). The grasped particle is then removed from the fluid and is released on a glass plate, placed in front of the channel in a lower location, this being visible by the refocusing in (f). The fluid surrounding the particle when it is removed revealed to be sufficient to make it adhere to the substrate releasing it from the fingers. In order to demonstrate that this process is repeatable - an essential point in demonstrating the automation possibilities of the combination of acoustic positioning and external gripping the process has been repeated for two further particles (not shown in the sequence). Before the microgripper reentered the channel the acoustic field has been activated again to tighten the line. In this way, any lateral misalignment due to motion of the microgripper or to an excessive amount of particles as seen in (e) could be eliminated. It has been observed that there was no need to re-adjust the microgripper lateral position in order to pick up the next particle in the row. From a detailed investigation it turned out that the lateral position uncertainty of the acoustic positioning process amounts to about 14.0 lm and has been determined by taking the average value for the position measured when the microgripper is in position to grasp the particle, but the fingers still were not actuated. This variation is smaller than the particle positional error that the microgripper can tolerate in order to guarantee that the particle still lies within its unactuated fingers microgripper. In fact the empty space left between the fingers and the trapped particle amounts to 26 lm. In contrast, the positioning uncertainty along the channel, amounting to 37 lm, turned out to be the limiting factor for auto-

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

251

Fig. 3. Sequence showing a pick and place experiment of three 74 lm copolymer particles at the beginning randomly distributed on the bottom channel’s surface (a). Particles are first roughly gathered along the centerline by exciting the system at 780 kHz, so that they are all accessible to the microgripper (b). Then the frequency is switched to 2.08 MHz, in order to form a chain of (almost) single particles (c). Before the microgripper is entered in the liquid through the air–water interface (e), the acoustic field is turned off (d). Next, the first particle in the line is grasped by applying 90 V to the microgripper fingers and removed from the fluid. The microgripper is lowered on a glass plate positioned in front of the channel, in a lower location. At this stage the particle is released (f).

mation. The accuracy is limited by two factors. First, a gradient in the acoustic field, created by the pressure release boundary at the air–water interface and at the reservoir, causes an uncontrolled movement of particles along the channel axis. This is shown in Fig. 3b by the arrows marking the displacement of two particles in time and cannot be predicted by the two dimensional model. Secondly, as seen previously, when more particles are present than there is space along the line e.g. as in (c), by reactivating the field some might rearrange and squeeze between two particles on the lower level, moving the whole line to the right or to the left. Furthermore, the magnitude of the acoustic radiation forces decays towards zero in proximity to the interface, making the handling of particles in that region impossible. Therefore, there is an interest in reducing this ‘‘dead” volume. This experiment shows that full automation cannot be achieved by the exclusive use of quasi one dimensional standing pressure fields. A method to control the particles’ position along the channel is necessary. This is presented in Section 3, after having presented another use of acoustic radiation forces and discussed the limitations of this approach. 2.3. Further use of acoustic radiation forces Acoustic radiation forces have been successfully exploited by the authors in [13] for the removal of particles attached to the manipulation tool. In particular when cells are handled, it might happen that they remain stuck on the fingers of the microgripper.

By slightly displacing the finger off-axis (and with it the centre of the cell) a force is exerted on the cell when the field is excited. Its magnitude has been seen to be sufficient to push the cells to the trapping plane and hence release it from the microgripper. The manipulation process of cells is shown in Fig. 4. Three lines of particles have been formed and then the first cell in the middle row was accessed, this being the line shown in the figure. The procedure is similar to the one described previously, the only difference being the fact that the cells were only displaced within the channel and not removed from it. After turning the acoustic field off, the microgripper was inserted and moved to the first cell in the row (a). Then its fingers are closed (b) and the cell is displaced to another position (c). To release the cell it is brought close to (but not on) the middle trapping plane (d). The field is shortly activated again, so that the cell is attracted towards the channel centerline and released from its fingers. 2.4. Limitations of the current manipulation approach The manipulation of cells has been reported here because it indicates the limitations of this combined manipulation technique, which are reached when small particles – in particular present at relatively high concentrations – are handled. Since the trapping site is never perfectly coinciding with the channel centerline (this being possibly due to manufacturing imprecisions, to the exact shape of the air–water interface and the droplet in the reservoir) the lateral position of the particles might slightly vary along the

252

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

Fig. 4. Manipulation of MCF10A cells. The sequence shows cells collected along the channel’s centerline (particles have been collected in the two external lines as well). After having been positioned (a), the first cell is grasped (b). In order to illustrate the use of acoustic radiation forces to release stuck cells, the microgripper fingers are first opened, after having displaced the cell in x and y direction (c), corresponding to the horizontal and vertical direction in the figure. As it can be seen, the cell remains stuck on the finger. To release it the acoustic field is shortly activated, so that a force is exerted on the cell pulling it towards the trapping plane (d). Furthermore, it can be seen that the positioning accuracy is the major factor limiting the automation possibilities. Under the effect of a non perfect one dimensional field and possibly as a result of secondary forces cells are not located exactly along the channel centerline and tend to merge into small groups.

axis. While this effect is less visible for large particles, it might become critical for small ones. Furthermore, due to the pressure gradient particles have the tendency to form (a) clump(s).

tubes. Schematics of the new design are shown in Fig. 5, as top (a) and bottom (b) view. 3.2. Prediction of the lateral and longitudinal particles’ position

3. Acoustics and fluidics assisted manipulation 3.1. System description In an extension of the device [14] described in the previous section, the main channel has been supplemented with two side channels used to move liquid from the reservoir towards the interface, dragging with it the suspended particles. The side channels have a width of 150 lm and leave the channel at a right angle 250 lm away from the interface and then continued parallel to it. Each side channel ends in a socket on the backside of the silicon wafer, where Teflon tubes are plugged in and fixed with epoxy glue. By connecting the other ends of the tubes to a syringe pump (kDScientific, Model 270) operated in withdrawal mode, liquid can be moved along the main channel. The main channel in this device is 7 mm long and the glass plate (a glass wafer of 500 lm thickness) has been anodically bonded, rather than adhered with epoxy glue, in order to prevent clogging of the small feature sizes. The piezoelectric transducer has been reduced in size and now has the same width as the main channel, in order to leave place for the Teflon

In order to predict the lateral and longitudinal position, the finite element model of the channel has been extended to three dimensions so that the interfaces at the extremities of the main channel could be taken into consideration. Both the air–water interface and the interface to the droplet in the reservoir have been modeled as pressure release boundaries. In Fig. 6 lines of constant force potential U are plotted at 727.5 kHz (in the color version blue marking the minimum value, red the maximum) in the x–y plane (no substantial variations of the acoustic field over the z-direction). It can be seen from the picture that particles are collected at the minima of the potential, in an area defined at the centre of the channel both in x and y direction. i.e. in the area of zero acoustic pressure and within it at the locations where the velocity is maximal, as predicted by Eq. (1). The longitudinal drift is created by the velocity term, in particular the component in x-direction ffi 0 and v z ¼  i1x @p ffi 0 as the pressure gradients in (v y ¼  i1x @p @y @z these two directions are almost vanishing). This component varies along the channel, i.e. v = v(x, y, t). The simulated motion of particles is shown in the figure, as black lines, the black dot represent-

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

253

Fig. 5. Extended device for the flow assisted manipulation of particles, (a) top view, (b) bottom view. The design shown in Fig. 1 has been supplemented with two side channels (3), leaving the main channel (1) 250 lm away from the air–water interface (2). By connecting their ends via two Teflon tubes (6) connected to a syringe pump, liquid can be moved from the reservoir (4) towards the interface, dragging with it the suspended particles. The piezoelectric element’s width (5) has been reduced to the width of the channel. The Teflon Tubes are fixed to the silicon wafer using two-components epoxy (7) and are used to clamp the device in a plastic plate (8) used as an holder.

Fig. 6. Force potential field (plotted as lines of constant potential) set up in the channel at 727.5 kHz and 750 kHz. Particles are collected at one and two distinct locations along the channel’s axis, respectively. This is caused by the presence of two pressure release boundaries at both ends of the channel (air–water interface on the right-end side and interface to droplet on the left-end side, modeled as straight interface as well). The path traveled by the particles when the acoustic field is excited is shown in both plots as well, as black lines. The dot marks the end position of a particle.

ing the final position. Interestingly, the simulation predicts a second frequency (750 kHz), at which particles are positioned along the channel’s centerline, at two distinct locations, though. Depending on the position of a particle with respect to the centre of these force potential minima the drift might be towards or away from the interface. In fact, such a force field probably corresponds to the field excited in the channel shown in Fig. 4. It has to be added that the two frequencies 727.5 kHz and 750 kHz are the only two frequencies for which particle alignment solely along the centerline is expected. This fact leads to the conclusion that in such a sys-

tem design characterized by pressure release boundaries it is not possible to control the particles’ position by means of a quasi one dimensional field alone, as a drift is typically present. In order to bypass this drawback two solutions can be suggested. The first one is based on transport of particles by means of a controlled bulk flow along the channel, which drags the particles towards the interface. This is the solution implemented here. The second method is based on the exploitation of acoustic radiation forces. The use of acoustics to transport particles has been previously exploited in macroscopic systems for instance by Haake et al. [20] and Saito et

254

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

Fig. 7. In (a) a pressure field is set up at 735 kHz and kept active for the whole duration of the experiment. After particles have been gathered along the centerline (b, c) fluid has been removed in steps of 0.05 ll form each side channel by activating the syringe pump in withdrawal mode. To illustrate the net movement of the particles when fluid has been applied a second series of figures (d)–(f) has been taken after 8.04 s from the beginning of the experiment. A further, single activation step of the pump is shown in (e) and (f).

al. [21]. By raising the frequency they have both been able to displace the particles. In [20] the frequency was continuously raised causing a net movement of particles in one direction. However, the distance between the lines was also reduced, so that all the particles were eventually concentrated at one extremity of the fluid volume. In this process the number of trapping planes increased with the frequency and particles moved in one or the other direction towards the nearest nodal plane. Therefore clumps of particles were seen to split or merge into new ones with the resulting transfer of particles from one clump into another one. In contrast in the approach by Saito et al. [21] the frequency was raised in discrete steps so that a new mode with a higher number of nodal planes was excited. By a clever choice of the frequencies it is basically possible to move single particles (or group of particles) in one preferential direction. Haake et al. [20] also described a method where particles could be transported by varying the voltage of two oppositely arranged piezoelectric transducers. Manneberg et al. [3] recently reported on motion of clumps of particles and cells along a microchannel without bulk flow rather using acoustic radiation forces only instead to finally load them into a small chamber at the end of the channel. The principle was based on a slight change of the nodes’ position when the frequency is linearly ramped in a small range so that at the end of the ramp the particle’s position was closer to the ‘‘node in front” for the frequency at the beginning of the ramp rather than the original node and so fell into this new node. In Manneberg’s experiments particles mostly remained in the same clump. These methods, if applicable at all to the present geometry, would require the ability to load each ‘‘moving node” with a single particle and hence require a more complex system and operation. In order to move particles towards the interface and at the same time keep control of their lateral position, fluid has been removed from the side channels whilst the acoustic field was constantly present in the channel. A sequence illustrating this technique is

shown in Fig. 7. Liquid has been withdrawn again in steps of 0.05 ll using a syringe pump from each side channel. It can be seen that in spite of the predicted drift a net motion towards the air– water interface is successfully achieved. If operated with low concentration of particles, in order for full automation to be obtained, the acoustic device should be extended with a control loop with optical feedback to know the exact position of the particles. 4. Acoustics and fluidics assisted manipulation of crystals In this last section it will be discussed how acoustic radiation forces can aid the automation of the sample preparation process for crystallographic analysis of protein crystals (e.g. insulin). The main difference in this approach lies in the mechanical manipulation process, which requires a different design of the acoustic system, in particular in terms of how the particles (here crystals) are accessed. In fact, in contrast to a microgripper which can actively grasp single particles, a nylon loop attached to a metal pin serves as a crystal manipulation tool. This tool is used to harvest a crystal from the drop in which it grew and to mount it onto a goniometer for X-ray diffraction analysis. A picture of it is shown in Fig. 8. The loop can be considered as a ‘‘passive” tool. It works like a butterfly net or a spoon, in so much that it traps everything that comes into it, without being able to discern between two particles. Furthermore, the crystal has to be approached from above or below and not from the side. 4.1. System design For reasons related to the manipulation tool’s geometry, as mentioned above, the air–water interface has been created on the bottom surface of a channel and the crystals are consequently removed from below. Such an interface – called hereafter the ori-

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

255

Fig. 8. Schematics of the device used for the manipulation of crystals and picture of the nylon loop. Crystals and buffer are inserted separately in the 1 mm wide, 11 mm long channel through two inlets (2 and 3) defined in the 1 mm thick glass plate (6). Crystals are then brought into the region of highest acoustic radiation forces above the piezoelectric transducer (7) by applying a flow by means of a syringe pump connected with the other extremity of the channel (1). Once positioned along the channel’s centerline crystals are further moved towards an orifice (4) defined on the lower surface. Flow is stopped immediately after one crystal has fallen into it. From there, the crystal is removed using the loop, as illustrated in the sequence (a)–(c) in a cross section view.

fice – also provides a way to define the particles’ position, as a crystal is accessed with the loop only after it has fallen into this orifice (Fig. 8). On the other hand, spatial separation is mostly indirectly achieved by the low crystals concentration. In fact, in most of the experiments performed it was possible to have a single crystal in the orifice. A further difference is in the way crystals are injected in the system. In order to avoid any loss of this precious sample (such as crystals which remain in the reservoir or stuck on the walls at the entrance), buffer solution is loaded into a reservoir at one extremity of a channel whilst suspended crystals are inserted separately by means of a pipette downstream so that they are all exposed to the velocity field. Transport into the region of minimum force potential (according to the findings of the three dimensional finite element simulation reported in the previous section, a minimum confined between the pressure release interfaces – specimen inlet and orifice – is expected here as well) and away from it into the orifice is done by withdrawing liquid from the channel using a syringe pump connected to the other end of the channel. The flow is stopped when a crystal has fallen into

the orifice. During application of flow the fluid meniscus remained pinned along the orifice boundary, preventing air to enter the channel. The channel is 1 mm wide, 11 mm long. The orifice has an oval shape and extends for 700 lm in channel direction and 300 lm across. The glass plate sealing it is 1 mm thick and has been attached using two-components epoxy. The holes (buffer solution reservoir and specimen inlet) have been manually drilled. The 500 lm thick piezoelectric plate is located between the orifice and the specimen inlet, is 3 mm wide and is provided by a strip electrode configuration. The removal process is shown schematically in Fig. 8 as well. Once a crystal has fallen into the orifice, the loop is inserted from the opposite side (a), moved above it (b) and pressed down against the water meniscus. When its deflection is too large and the surface tension is no longer sufficient to withstand the force exerted by the loop, the crystal is transferred into the loop and trapped there in a droplet. Fig. 9 shows the manipulation process. Crystals have already been injected in the system and brought into the region of higher acoustic radiation force, above the transducer, by applying a flow

256

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

Fig. 9. Sequence showing the manipulation process of crystals, starting with a random distribution on the bottom surface of the channel (a). By exciting the system at 751 kHz crystals are collected along the centerline (b and c). Due to the presence of pressure release boundaries (orifice, specimen inlet) a minimum of force potential is created along the channel causing a drift of the crystals, which can be seen by comparing their position in (b) and (c). By activating the syringe pump connected to the extremity of the channel at the top of the picture (d) crystals are moved until one falls into the orifice (e). Then, the loop is inserted from the opposite side of the orifice, from beneath (f), moved above the crystal (g) and pressed down against the meniscus. This can be seen in (h) by the different light reflection created on the liquid surface. Finally, when the surface tension is overcome, the crystal is removed (i) from the channel and remains suspended in a droplet of buffer solution trapped in the loop.

of 0.05 ml/s (a). Then the transducer is activated at 751 kHz, so that the crystals are gathered along the channel’s centerline (b and c). As seen previously the pressure release boundaries cause the creation of a force potential minimum along the channel, leading to a drift of particles. This has been observed in this system as well, as can be seen by comparing the position of the crystals, by aid of the dashed line, in (b) and (c). Then flow is applied again until one crystal has fallen into the orifice (e). Then the loop is inserted into the orifice (f), moved above the crystal (g) and pressed against the liquid meniscus (h). Once the surface tension is overcome the crystal is removed from the channel (i) and remains suspended in a droplet of buffer solution trapped in the loop. If a crystal should loose its lateral position during motion, ultrasound can be shortly applied again. This step was not necessary in this experiment. During this process no detrimental effects to the crystals could be observed, as recently reported in [15]. 5. Conclusions The intrinsic nature of standing wave acoustic fields, which allows the simultaneous positioning of large amounts of particles,

has been exploited here to achieve a certain degree of automation in single particle manipulation processes. In the applications presented here the single particle manipulation is not achieved by use of ultrasound alone, as in not currently able to discern between single particles (as optical tweezers in contrast are). Instead acoustic manipulation represents a step in a more complex strategy, where the role of ultrasound is that of pre-positioning particles at known locations, where they can be individually accessed in a sequential fashion dispensing with the need of prior localization. The strategy reported for the first time by the authors and reviewed here consisted of gathering all the particles in a row, distant from the side walls and the sticking interactions these cause, and accessing the first particle in the row through an air– water interface. The location and shape of the interface depends on the geometry and operation principle of the tool used. While the microgripper (used for the manipulation of copolymer particles and cells) needs to grasp particles from the horizontal plane, the nylon loop (used in the manipulation of crystals) requires access from below (or above). In this last case spatial separation is also necessary, as the loop cannot actively discriminate between single and multiple crystals. It has been shown that successful semi-auto-

S. Oberti et al. / Ultrasonics 50 (2010) 247–257

mated manipulation can be achieved by aiding the manipulation process by the use of laminar flow to move the particles along the channel. In fact, this counteracts the drift caused by the air– water interface(s), which is responsible for a poor longitudinal accuracy of the particles’ position. Either by adding two side channels or removing fluid from an inline channel a net movement could be observed. Optical control is needed only to verify when a crystal has fallen into the orifice or, when worked with low concentration suspension, to know when a particle has reached the interface. The approach presented reaches its limits when small particles are manipulated, though this may be overcome by the use of smaller chambers and correspondingly higher acoustic frequencies. However, in general, in the acoustic fields established in these devices the trapping site never perfectly coincides with the channel’s centerline, due mainly to manufacturing imprecisions. Deviations in the particles’ position from this ideal site have therefore to be expected. In contrast, as shown here, when larger particles are handled this effect is less detrimental, so that the semi-automated manipulation could be shown. Acknowledgments The authors thank KTI/CTI Switzerland (Top Nano 21 grant, project number 6643.1 and 6989.1) for funding the work presented in Sections 2 and 3 and Dr. G. Radziwill from the Institute of Medical Virology, University of Zurich (Switzerland) for the experiment with MCF10A cells. Furthermore the authors thank the NCCR (National Center of Competence in Research) Structural Biology (Swiss National Science Foundation, SNSF) and the PX beamlines team at the Paul Scherrer Institute for the support of the work presented in Section 4. References [1] M. Groschl, Ultrasonic separation of suspended particles – part i: fundamentals, Acustica 84 (1998) 432–447. [2] S. Oberti, A. Neild, J. Dual, Manipulation of micrometer sized particles within a micromachined fluidic device to form two-dimensional patterns using ultrasound, Journal of the Acoustical Society of America 121 (2007) 778–785. [3] O. Manneberg, B. Vanherberghen, B. Önfelt, M. Wiklund, Flow-free transport of cells in microchannels by frequency-modulated ultrasound, Lab on a Chip 9 (2009) 833–837.

257

[4] T. Laurell, F. Petersson, A. Nilsson, Chip integrated strategies for acoustic separation and manipulation of cells and particles, Chemical Society Reviews 36 (2007) 492–506. [5] R.J. Townsend, M. Hill, N.R. Harris, N.M. White, Investigation of twodimensional acoustic resonant modes in a particle separator, Ultrasonics (2005) E467–E471. [6] H. Andersson, A. van den Berg, Microfluidic devices for cellomics: a review, Sensors and Actuators: B—Chemical 92 (2003) 315–325. [7] M. Honnatti, G. Highes, Directed cellular manipulation using polymer microgrippers, Zyvex Application Note 9720, 2006. . [8] J.R. Helliwell, Macromolecular Crystallography with Synchrotron Radiation, Cambridge University Press, 2005. [9] J. Svennebring, O. Manneberg, P. Skafte-Pedersen, H. Bruus, M. Wiklund, Selective bioparticle retention and characterization in a chip-integrated confocal ultrasonic cavity, Biotechnology and Bioengineering 103 (2009) 323–328. [10] O. Manneberg, B. Vanherberghen, J. Svennebring, H.M. Hertz, B. Önfelt, M. Wiklund, A three.dimensional ultrasonic cage for characterization of individual cells, Applied Physics Letters 93 (2008). [11] P. Glynne-Jones, R.J. Boltryk, M. Hill, F. Zhang, L.Q. Dong, J.S. Wilkinson, T. Melvin, N.R. Harris, T. Brown, Flexible acoustic particle manipulation device with integrated optical waveguide for enhanced microbead assays, Analytical Sciences 25 (2009) 285–291. [12] F. Beyeler, A. Neild, S. Oberti, D.J. Bell, Y. Sun, J. Dual, B.J. Nelson, Monolithically fabricated microgripper with integrated force sensor for manipulating microobjects and biological cells aligned in an ultrasonic field, Journal of Microelectromechanical Systems 16 (2007) 7–14. [13] A. Neild, S. Oberti, F. Beyeler, J. Dual, B.J. Nelson, A micro-particle positioning technique combining an ultrasonic manipulator and a microgripper, Journal of Micromechanics and Microengineering 16 (2006) 1562–1570. [14] S. Oberti, A. Neild, D. Moller, J. Dual, Towards the automation of micron sized particle handling by use of acoustic manipulation assisted by microfluidics, Ultrasonics 48 (2008) 529–536. [15] S. Oberti, D. Möller, S. Gutmann, A. Neild, J. Dual, Novel sample preparation technique for protein crystal X-ray crystallographic analysis combining microfluidics and acoustic manipulation, Journal of Applied Crystallography 42 (2009) 636–641. [16] A. Neild, S. Oberti, A. Haake, J. Dual, Finite element modeling of a microparticle manipulator, Ultrasonics (2005) E455–E460. [17] L.P. Gor’kov, Forces acting on a small particle in an acoustic field within an ideal fluid, Doklady Akademii Nauk Sssr 140 (1961) 88–92. [18] A. Neild, S. Oberti, J. Dual, Design, modeling and characterization of microfluidic devices for ultrasonic manipulation, Sensors and Actuators: B— Chemical 121 (2007) 452–461. [19] M. Groschl, Ultrasonic separation of suspended particles – part ii: design and operation of separation devices, Acustica 84 (1998) 632–642. [20] A. Haake, A. Neild, G. Radziwill, J. Dual, Positioning, displacement, and localization of cells using ultrasonic forces, Biotechnology and Bioengineering 92 (2005) 8–14. [21] M. Saito, N. Kitamura, M. Terauchi, Ultrasonic manipulation of locomotive microorganisms and evaluation of their activity, Journal of Applied Physics 92 (2002) 7581–7586.