- Email: [email protected]
Study on the mixing and migration behavior of micron-size particles in acoustofluidics
Chemical Engineering Journal 369 (2019) 370–375
Contents lists available at ScienceDirect
Chemical Engineering Journal journal homepage: www.elsevier.com/locate/cej
Study on the mixing and migration behavior of micron-size particles in acoustofluidics Pierre Gelina, Özlem Sardan Sukasa, Karine Hellemansb, Dominique Maesc, Wim De Malschea,
T ⁎
a
Department of Chemical Engineering, µFlow Group, Vrije Universiteit Brussel, 1050 Brussels, Belgium Diabetes Research Centre, Vrije Universiteit Brussel, 1000 Brussels, Belgium c Structural Biology Brussels, Vrije Universiteit Brussel, 1050 Brussels, Belgium b
H I GH L IG H T S
mixing in microfluidic channel using bulk acoustic waves. • Fast transfer enhancement of particles by acoustic streaming. • Mass µm particles focus on the pressure node. • 5Focusing velocity varies along width of channel. • 0.5 µm particles flow along with the flow vortices. •
A R T I C LE I N FO
A B S T R A C T
Keywords: Acoustofluidics Mixing Particle tracking Particle focusing Acoustic streaming
Slow mass transfer remains a limiting factor in microfluidic devices, as it is purely based on molecular diffusion. Hence, the completion of a targeted chemical reaction is extremely slow. Acoustic streaming is a great methodology to create lateral convection in otherwise laminar flow operated channels and offers the possibility to reduce mass transfer resistance dramatically. Moreover, the acoustic field can elegantly by used to handle micron-sized particles. Large particles (> 2 µm), experience a radiation force that is much larger than the drag force related to streaming and will rapidly focus at either the node or anti-node of the standing pressure wave in the channel. In the present contribution, the mass transfer rate in both a drag force dominated as well as in a radiation force dominated regime is studied by in situ tracking polystyrene particles of respectively 0.5 and 5 µm. The 5 µm particles focus on the pressure node experiencing a migration velocity that decreases as the node position is approached, whereas the 0.5 µm particles remain in the vortex flow, characteristic for acoustic streaming. The experimentally obtained migration rates are in accordance with numerical simulations, suggesting that operational parameters as actuator displacement have been chosen appropriately. Mixing times of parallel flows were assessed and revealed an order of magnitude enhancement for small molecules, up to four order of magnitudes for 0.5 µm diameter particles.
1. Introduction During the past decade microfluidic devices have gained increased interest for their potential in physicochemical conversions. The dominating laminar flow (Re < 1), high surface to volume ratio and well defined residence time in microfluidic devices offer great advantages for chemical reaction and the production of highly monodisperse droplets and/or particles. Formation and handling of hard particles is common in the pharmaceutical, photonic, coating, textile and cosmetic
sector. Micron-sized monodispersed poly(lactic-co-glycolic acid) (PLGA) microparticles, for example, are used as a slow drug delivery system [1,2]. Apart from the large control over the applied flows, the small dimensions offer better heat transfer and greater yields. With diffusion coefficients in the order of 10−9 m2 s−1 for small molecules and lower than 10−10 m2 s−1 for larger molecules, diffusion based mixing in microfluidic devices is inherently slow. Both the in situ formation of hard particles or completion of a chemical reaction is therefore hindered. Moreover, it has been shown that rapidly
⁎
Corresponding author at: Department of Chemical Engineering, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Elsene, Belgium. E-mail addresses: [email protected] (P. Gelin), [email protected] (Ö. Sardan Sukas), [email protected] (K. Hellemans), [email protected] (D. Maes), [email protected] (W. De Malsche). https://doi.org/10.1016/j.cej.2019.03.004 Received 18 July 2018; Received in revised form 27 February 2019; Accepted 1 March 2019 Available online 02 March 2019 1385-8947/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
focus at the pressure node or antinode, while small particles will flow along with the vortices of the acoustic streaming. The cut-off between large and small particles is typically around 1.5 µm [25]. In the present paper, BAW and the related acoustic streaming is utilized to enhance the mass transfer rate far beyond the rate of diffusion. This is shown by fast mixing of two parallel flows with a specific aim for chemical conversion and particle synthesis. Moreover, the radiation force arising from the BAW together with the Stokes drag forces are investigated in the context of particle formation and manipulation. This is done for sub- and micron-size particles. Particle trajectories and velocities are simulated in COMSOL and confirmed experimentally. These deeper insights are of great relevance to perform in situ chemical reactions or for the synthesis of microparticles.
generating a homogeneous mixture enhances the selectivity of kinetically controlled chemical reactions [3]. Fast mixing also allows to reduce the residence time drastically minimizing the production of byproducts [4–6]. Passive mixers have the possibility to enhance mass transfer in microfluidic channels. These, however, often require high velocities and associated low residence times [7–9]. Formation and handling of particles in microfluidic devices is becoming a necessity [10], but passive mixers are very prone to fouling under these conditions due to the robust structures or many turns present in the channels [11,12]. Therefore they are not realistically suited for robust and reliable applications involving particles. An increasing interest in applying acoustics for mass transfer enhancement and particle manipulation is being observed [13–15]. Apart from the lower fouling probability compared to passive mixers, the use of acoustic waves gives more control over the mixing process. Two main types of acoustic waves in the microfluidic channel exist, surface acoustic waves (SAW) and bulk acoustics waves (BAW). SAWs are generally produced by applying an appropriate electric field to a piezoelectric material, which in turn generates a propagating mechanical wave in an elastic material. A part of this travelling SAW is refracted as a longitudinal pressure wave when contacted with a liquid. These pressure waves generate a force in their propagation direction and induce flow within the confined liquid [16]. Mixing two parallel liquid streams using SAW induced acoustic streaming in less than a second has been reported [17]. Herein a mixing efficiency (η) of 0.88 was reached at an actuation voltage of 80 Vp-p and a Bodenstein number of 7.4 * 103. Zeng et al. [18] reported a mixing time of 11 ms at a Bodenstein number of 8.64 * 102 and actuation voltage of 25 Vp-p. Here the Bodenstein number is defined as:
Bo =
u. L D
2. Materials and methods 2.1. Numerical simulations In order to gain more insight in the acoustic streaming and radiation phenomena, simulations were carried out with COMSOL® Multiphysics software. The numerical model used has been described elsewhere [26] and has been adapted towards the present study. In short, the cross section of a channel with a height of 70 µm and width of 375 µm has been built as a model. Next, the first order acoustic field was calculated with the Thermoviscous Acoustic, Frequency Domain interface. Subsequently, the second order time average flow was calculated with the Laminar Flow interface. Amplitude of the channel wall displacement could be varied to obtain different acoustic pressures and related streaming velocities. For mixing experiments, the channel was divided in three equal parts to ensure an optimal position of the vortices on the concentration profile. The mixing of diluted species was visualized with the Transport of Diluted Species interface. The diffusion constant was set to 10−9 m2 s−1. Finally, the Particle Tracking interface was used to visualize particle trajectory of large and small polystyrene particles (5.0 and 0.5 µm, respectively).
[1]
with u the linear velocity (m/s), L the width of the channel (m) and D the diffusion coefficient (m2/s). In the present paper the focus lies on the use of BAW in microfluidic channels. Herein, ultrasound waves in the low MHz range are applied to the microfluidic channel, creating a standing pressure wave. Viscous attenuation of this pressure wave leads to the streaming phenomenon called acoustic streaming, which take the form of vortices perpendicular to the channel [19]. A more in depth description of acoustic streaming can be found elsewhere [20–22]. Streaming velocities in microfluidic channels are typically in the order of ten to hundreds of µm s−1 [23]. Particles in a microfluidic channel are affected by both the acoustic pressure wave and the vortex flow, respectively corresponding to the radiation force and Stokes drag force. The radiation forces arise from scattering of the pressure wave, for a spherical particle with radius a, density ρp and compressibility κp, suspended in a liquid of density ρ0, compressibility κ0, dynamic viscosity η, and momentum diffusivity υ = η/ρ0 it is given by [24]:
Frad = 4π Φa3k 0 Eac sin(2k 0 y )
2.2. Chip design and set up Rectangular microfluidic channels were patterned in a positive resist using mid-UV lithography and subsequently etched by Bosch type DRIE etching in a silicon wafer with a thickness of 525 µm. The width and depth of the main channel is 375 and 70 µm, respectively. The main channel has a length of 33 mm. The channels were sealed by an anodically bonded borosilicate glass lid. Two inlets and outlets were provided. The geometry of the acoustofluidic chip is shown in Fig. 1a. Two syringe pumps (KDS200, KD Scientific Inc., USA) were used to introduce the liquids in the channels. Nanoport assemblies (N-333, IDEX Corp.) and glass capillaries ensured the connection from the chip to the pumps. To generate a standing pressure wave, the acoustofluidic chip was connected to a piezo ceramic element (PZT) (30 mm × 20 mm × 1 mm, APC International Ltd., USA) with a resonance frequency at 2.0 MHz. Coupling of the microfluidic chip to the PZT was done by a thin glycerol layer. An in-house built PMMA holder ensured a good connection between the chip, PZT and Nanoports. A schematic representation of the acoustofluidic chip is given in Fig. 1b. The PZT was driven by a frequency generator (AFG1062, Tektronix UK Ltd., UK) around 2.0 MHz. The applied voltage was amplified by an RF power amplifier (210L, Acquitek S.A.S, France) with a power output of 10 Watt. The voltage amplitude across the PZT was monitored with an oscilloscope (TBS1104, Tektronix UK Ltd., UK).
[2]
with k0 the angular wavenumber for the first harmonic, Eac the time average acoustic energy, Φ the acoustic contrast factor. The contrast factor varies between −1 < Φ < 1 and is a measure of how ‘hard’ a particle is compared to the surrounding liquid. A ‘hard’ (0 < Φ) particle will migrate towards and focus at the pressure node while a ‘soft’ particle (Φ < 0) will migrate towards the pressure antinodes. Next to radiation forces, particles also experience Stokes drag forces due to the streaming vortices:
Fdrag = 6πηa (vstr − vp)
[3]
with vstr the acoustic streaming velocity and vp the initial particle velocity from the radiation force. At equilibrium Fdrag + Frad = 0 and the part particle velocity equals:
vp =
Frad + vstr 6πηa
2.3. Experimental procedure Visualization of all the experiments was done using an inverted fluorescent microscope (IX71, Olympus Corporation, Japan) in combination with a CCD camera (C9100-13, Hamamatsu Photonics, Japan).
[4]
Large particles are mostly affected by the radiation force and will 371
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
I 'i =
Ii − Ib n (I − I ) ∑i = 1 i n b
[6]
with Ii’ the normalized intensity on row i, calculated based on the intensity on row i (Ii) and the background intensity (Ib). I′i,∞ the normalized intensity on row i at complete mixing, I′i,0 the normalized intensity on row i at the start of the mixing experiment. It should be noted that for liquids, mixing by molecular diffusion takes place at very small scales. Since the pixel size was 1.45 µm, this leads to an averaging of the concentration in one pixel. This, however, has a negligible effect on the mixing efficiency. Doubling the pixel size changed the mixing efficiency by only 0.003% for a representative mixing condition (voltage = 10Vpp, flowrate = 5 mm/s). Mixing efficiency was evaluated as a function of the applied voltage for two different flow velocities (5 mm s−1 and 10 mm s−1). The voltage was varied between 5 Vp-p and 20 Vp-p in both cases. 2.3.2. Experimental particle tracking. Fluorescent monodispersed polystyrene microparticles (Microparticles GmbH, Germany) with a size of 0.45 µm (SD: 0.01 µm) and 5.19 µm (SD: 0.14 µm) were used for particle tracking experiments. These are further referred to as 0.5 µm and 5.0 µm particles respectively. The 5 µm and 0.5 µm particles were diluted to a solid content of 0.25 w/v% and 0.0125 w/v%, respectively. Once particles were introduced in the acoustofluidic chip, the actuator was activated and the effect on the particles was examined. Based on the obtained movies, the radiation velocity of 5.0 µm particles as a function of their position in the channel was analyzed with a customized MATLAB code, this for different voltages (5–15 Vp-p). Streaming without flow in the axial direction, was visualized at three different channel heights (top, middle and bottom). Their related streaming velocities were analyzed with a similar MATLAB code. The standard deviation was calculated on four independently recorded movies.
Fig. 1. a) Top view of the etched microfluidic chip used in all mentioned experiments. The main channel has a length, width and height of 33 mm, 375 µm and 70 µm respectively. Two inlets and two outlets are provided. b) Schematic representation of microfluidic channel: The microfluidic channel is connected to a piezo ceramic element vibrating at 1.93 MHz. Actuation is done with a frequency generator. The vibrations are transferred to the channel wall, creating a standing pressure wave (thicknesses are not to scale).
2.3.1. Mixing experiments For the mixing experiments, water containing a fluorescent salt (fluorescein sodium salt, Sigma-Aldrich, Germany) and non-fluorescent DI water were brought into contact in the microfluidic chip. Mixing of the two water streams was monitored at 16 mm from the intersection at the middle of the main channel. Mixing was initiated by actuating the piezo ceramic element at a frequency of 1.93 MHz. The recorded movies were analyzed with ImageJ software. The intensity of each pixel was related to the fluorescence intensity and eventually to the efficiency of the mixing. By averaging the pixel array along the width of the channel, the mixing efficiency or degree of mixing (η) can be deduced [17]:
η=1−
1 n 1 n
3. Results and discussion The following results show the possibility to use BAW and the related acoustic radiation/streaming to enhance mass transfer greatly in microfluidic channels as well as the manipulation of large and small particles.
n
∑i = 1 (I 'i − I 'i, ∞ )2 n
∑i = 1 (I 'i,0 − I 'i, ∞ )2
3.1. Numerical particle tracking
[5]
Theoretically the particle trajectory is given by Eq. (4), a Fig. 2. Numerical simulation of particle trajectory for large (5 µm) and small (0.5 µm) polystyrene particles in the cross section of a channel at a displacement amplitude of 0.5 nm. a) Large particles focus fast (in 0.2 s) on the pressure node. The velocity of focusing is not constant over the complete width of the channel, varying from 0 to 1600 µm s−1. b) Small particles flow along with the streaming vortices (the liquid velocity field is shown by the red arrows).
372
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
obtained simulations were validated experimentally. Mixing due to the presence of an acoustic field was visualized by a camera. Fig. 4b shows snapshots at different times at a fixed observation position (16 mm downstream the contact area) during the mixing process, overlaid with the concentration profile. In this case the flow rate was set to 10 mm s−1 and the actuation voltages to 20 Vp-p. At t = 0.30 s the channel already had a mixing efficiency of η = 0.91, but in took roughly 1 s more to accomplish complete mixing. The mixing efficiency has been measured for a range of applied voltages and flowrates (Fig. 4c). Actuation at voltages lower than 10 Vp-p does not induce any visible mixing. For a linear velocity of 5 mm s−1 a mixing of already η = 0.65 is reached at 10 Vp-p actuation. Complete mixing (η = 0.94) is reached at an actuation voltage of 15 Vp-p for both flowrates. These experimentally obtained mixing rates are in accordance with the numerical simulations, indicating that operational parameters such as actuator displacement have been chosen appropriately. To our knowledge such fast mixing in a microfluidic channel with bulk acoustic waves has not been demonstrated yet. However, SAW has been utilized to achieve even faster mixing in similar channels [17,18,27]. Fast mixing has also been achieved in small laminated channels in the context of complex chemical reactions [6,28] and has been observed in larger channels by flow destabilisation [7,29–31]. In the present paper, however, we have focused on relatively large channel dimensions wherein the formation and use of particles is manageable under realistic conditions. Important to note is the absence of passive mixing elements which often lead to fouling of the microfluidic channel once particles are introduced. The current use of BAW for fast mixing is limited by energy transfer into the fluid. Acoustic streaming is a secondary flow effect and requires a high energy input to have a significant effect. Applying higher voltages will have a direct effect on the streaming velocity and thus on the mixing efficiency. At higher voltages, however, rises in temperature must be taken into account. It is so that the resonance frequency of the acoustofluidic device depends on the temperature [32]. A change in temperature will shift the resonance frequency and less energy will be transferred into the microfluidic channel.
combination of Stokes drag forces from the vortices and the radiation forces. Fig. 2 shows a numerical simulation for the particle trajectory of both 5 µm and 0.5 µm polystyrene particles. The simulation shows migration of 5 µm particles towards the pressure node (Fig. 2a). The focusing velocity, however, is not constant over the width of the channel. A decrease in velocity is observed as the node position is reached, which is in agreement with Eqs. (2) and (4). For a numerically imposed displacement amplitude of 0.5 nm, a maximum migration velocity of 1462 µm s−1 is reached. The equilibrium position, where lateral movement ceases, is reached after 0.2 s. In contrast to the large particles, the small 0.5 µm particles do not focus on the pressure node (Fig. 2b). Instead they flow along with the vortex flow of the acoustic streaming. The particle velocity varies strongly as a function of the particle’s position in the vortex. The velocities that are reached on the outer areas of the vortices are around 100 µm s−1. In the literature radiation velocities of 44 µm s−1 and streaming velocities of 7 µm s−1 are reported [26] for a displacement amplitude of 0.1 nm. Comparable values are obtained in our case for the same displacement amplitude. In the context of fast mixing for particle maturation and chemical reactions we have chosen to show the results with the higher displacement amplitude of 0.5 nm. Experimentally obtained radiation velocities (actuation voltage of 15Vp-p) were fitted using the obtained simulation (Fig. 3). Small changes in displacement amplitude have a large influence on the radiation velocity. A satisfactory standard deviation (S = 260 µm s−1) of the residuals (vmeasured – vcurve) was obtained for a fit on the 0.5 nm displacement amplitude curve. 3.2. Acoustic mixing Mixing in microfluidic devices remains a considerable hurdle in the context of microfluidics. In (generally occurring) laminar flow conditions, mixing only relies on molecular diffusion which is inherently slow. To assess the potential of mixing improvement with respect to diffusion-based mixing, mixing of two flows in a microfluidic channel with acoustic streaming was simulated with COMSOL. The results are shown in Fig. 4a. At a displacement amplitude of 1 nm, complete mixing of the channel is observed in 1.52 s. This is about ten times faster compared with simple molecular diffusion. Because mass transport occurs by active mixing, the mixing rate is to a much lower extend dependent on the diffusion coefficient. Small particles of 0.5 µm (D = 10−13 m2 s−1) in diameter will mix as fast as a dye. Next, the
3.3. Experimental particle tracking 3.3.1. Large particles Trajectories and velocities of 5 µm polystyrene microparticles in the presence of the acoustic field were visualized with the camera. Particle migration was triggered as the acoustic field was activated. The radial migration velocity as function of the applied amplitude is shown in Fig. 5. A higher amplitude results in larger focusing velocity. This is in line with expectations, as the acoustic energy (Eac) scales with radiation force (Frad) (Eq. (2)). Velocities as high as 1600 µm s−1 were reached for an actuation of 20 Vp-p. As predicted by theory (Eq. (4)) and according to the previously performed simulation, the particle velocity is low at the channel wall, rises gradually to a maximum and eventually goes back to zero when reaching the equilibrium position. Notice that at low radiation velocity near the equilibrium position, a small inflection is observed, which is not present in the simulation data (Fig. 3). This can be attributed to a small mismatch between the pressure wavelength and the channel width leading to a small variation in the position of the pressure node along the length of the channel. Averaging the radiation velocity of all the particles, will lead to this small deviation from theory. 3.3.2. Small particles In contrast to the large particles, 0.5 µm particles follow the streaming lines of the vortex flow. Particle trajectories and velocities were measured at 3 different heights of the channel: top, middle and bottom. These obtained PIV data (Fig. 6) reveal that at the top and bottom layer, particles flow towards the channel wall, while at the middle layer particles move towards the channel center. This confirms
Fig. 3. Comparison of experimental radiation velocities (▲) with radiation velocities obtained with COMSOL for different displacement amplitudes. The best fit was obtained for a displacement amplitude of 0.5 nm. In this case the standard deviation of residuals equals 260 µm/s. 373
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
Fig. 4. a) Cross sectional view of a microfluidic channel. Shown is a numerical simulation for the mixing of two liquid streams in presence of an acoustic field. Displacement amplitude of 1 nm was set to the left and right wall. Complete mixing is reached after 1.52 s. b) Top view of microfluidic channel. Shown is the mixing at a flow rate of 10 mm s−1 and an actuation of 20Vp-p. A mixing efficiency of 97% is reached after 1.53 s. c) Mixing efficiency in function of applied voltage for two different flow rates, 5 mm s−1 (♦) and 10 mm s−1 (□). Mixing was measured at 16 mm from the point where the flows meet. At lower voltages the water streams mix better for a low flow rate of 5 mm s−1. The time needed for complete mixing is around 1 s and 2 s for flow rates of 5 mm s−1 and 10 mm s−1, respectively.
channel wall and the centre of the channel can be estimated to be 3 s, which is considerably longer that the observed 1.5 s, which indicates that diffusion between the different flow layers in the vortex represents an important contribution to mixing. The mass transfer characteristics within the vortex will be further explored in a follow-up study with a recently in-house developed 3D PIV instrument.
4. Conclusion In this article mixing of two parallel flows by acoustic streaming was characterized. We demonstrated that bulk acoustic waves (BAW) can efficiently be applied for very fast mixing of two parallel liquid streams in microfluidic devices. Complete mixing is achieved in 1.53 s at a Bodenstein number of 2.5 * 103 and actuation voltage of 20 Vp-p. Furthermore we investigated the kinetics of BAW in the context of particle manipulation. Large and small particles are affected differently by the acoustic field. Large particles are mainly affected by the radiation forces and focus towards the pressure node. Displacement velocities vary depending on the position of the particle in the channel. For 5 µm polystyrene particle velocities as high as 1600 µm s−1 were reached for an actuation voltage of 20Vp-p. Small particles do not focus and flow along with the liquid vortices. Maximum velocities of 48 µm s−1 were reached for an actuation voltage of 15 Vp-p. These deeper insights in the interactions between particles and the acoustic field are of paramount importance for the continuous production and handling of micro- or nanospheres in microfluidic devices. BAW in microfluidic devices is an excellent tool to enhance mass transfer far beyond the rate of diffusion, this is of great relevance to conduct chemical reactions in these confined environments.
Fig. 5. a) Radial focus velocity of 5 µm particles for five different voltages (see legend). At the start particles have a low velocity, this rises up to a maximum to finally decrease when the pressure node is reached. Velocities as high as 1600 µm s−1 has been reached for a voltage of 20 Vp-p. b) Top view of microfluidic channel showing the fast focusing of 5 µm polystyrene particles. At 10 Vp-p equilibrium is reached after 0.267 s.
the presence of vortices in the microfluidic channel. The average radial velocities of the particles are given in Table 1 for four different voltages. Note that the absolute values of the velocity vectors are given, velocity vectors in the central plane have an opposite direction to the vectors on the top and bottom planes. At 15 Vp-p, the travel time between the 374
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
360–369. [2] Q. Xu, et al., Preparation of monodisperse biodegradable polymer microparticles using a microfluidic flow-focusing device for controlled drug delivery, Small 5 (2009) 1575–1581. [3] S. Suga, A. Nagaki, J. Yoshida, Highly selective Friedel – crafts monoalkylation using micromixing † compounds with N-acyliminium ions has been achieved by, Chem. Commun. 354 (2003), https://doi.org/10.1007/s00170-003-1973-7. [4] J.I. Yoshida, A. Nagaki, T. Iwasaki, S. Suga, Enhancement of chemical selectivity by microreactors, Chem. Eng. Technol. 28 (2005) 259–266. [5] K.S. Elvira, X.C.I. Solvas, R.C.R. Wootton, A.J. Demello, The past, present and potential for microfluidic reactor technology in chemical synthesis, Nat. Chem. 5 (2013) 905–915. [6] T. Asai, et al., Switching reaction pathways of benzo[b ]thiophen-3-yllithium and Benzo[b ]furan-3-yllithium based on high-resolution residence-time and temperature control in a flow microreactor, Chem. Lett. 40 (2011) 393–395. [7] R. Goovaerts, W. Smits, G. Desmet, J. Denayer, W. De Malsche, Combined improved mixing and reduced energy dissipation by combining convective effects and lamination, Chem. Eng. J. 211–212 (2012) 260–269. [8] S.R.L. Gobert, S. Kuhn, L. Braeken, L.C.J. Thomassen, Characterization of milli- and microflow reactors: mixing efficiency and residence time distribution, Org. Process Res. Dev. 21 (2017) 531–542. [9] T. Vandermeersch, R. Goovaerts, J. Luyten, J.F.M. Denayer, W. De Malsche, Tracking the liquid-liquid extraction performance in mesoflow reactors, Chem. Eng. J. 279 (2015) 9–17. [10] P. Gelin, et al., Focusing of microcrystals and liquid condensates in acoustofluidics, Crystals 9 (2019) 120. [11] K. Wu, S. Kuhn, Strategies for solids handling in microreactors, Chim. Oggi/Chem. Today 32 (2014) 62–66. [12] R.L. Hartman, Managing solids in microreactors for the upstream continuous processing of fine chemicals, Org. Process Res. Dev. 16 (2012) 870–887. [13] K.D. Frampton, S.E. Martin, K. Minor, The scaling of acoustic streaming for application in micro-fluidic devices, Appl. Acoust. 64 (2003) 681–692. [14] D. Fernandez Rivas, S. Kuhn, Synergy of microfluidics and ultrasound, Top. Curr. Chem. 374 (2016) 70. [15] H. Bruus, Acoustofluidics 1: governing equations in microfluidics, Lab Chip 11 (2011) 3742. [16] X. Ding, et al., Surface acoustic wave microfluidics, Lab Chip 13 (2013) 3626. [17] T.-D. Luong, V.-N. Phan, N.-T. Nguyen, High-throughput micromixers based on acoustic streaming induced by surface acoustic wave, Microfluid. Nanofluid. 10 (2011) 619–625. [18] Q. Zeng, et al., Milliseconds mixing in microfluidic channel using focused surface acoustic wave, Sens. Actuat. B Chem. 160 (2011) 1552–1556. [19] N. Riley, Steady treaming, Annu. Rev. Fluid Mech. 33 (2001) 43–65. [20] M. Wiklund, R. Green, M. Ohlin, Acoustofluidics 14: applications of acoustic streaming in microfluidic devices, Lab Chip 12 (2012) 2438–2451. [21] P.B. Muller, H. Bruus, Numerical study of thermoviscous effects in ultrasound-induced acoustic streaming in microchannels, Phys. Rev. E – Stat. Nonlinear, Soft Matter Phys. 90 (2014) 1–12. [22] A.Y. Rednikov, S.S. Sadhal, Acoustic/steady streaming from a motionless boundary and related phenomena: generalized treatment of the inner streaming and examples, J. Fluid Mech. 667 (2011) 426–462. [23] P.B. Muller, et al., Ultrasound-induced acoustophoretic motion of microparticles in three dimensions, Phys. Rev. E – Stat. Nonlinear, Soft Matter Phys. 88 (2013) 1–12. [24] J.T. Karlsen, H. Bruus, Forces acting on a small particle in an acoustical field in a thermoviscous fluid, Phys. Rev. E – Stat. Nonlinear, Soft Matter Phys. 92 (2015). [25] R. Barnkob, P. Augustsson, T. Laurell, H. Bruus, Acoustic radiation- and streaminginduced microparticle velocities determined by microparticle image velocimetry in an ultrasound symmetry plane, Phys. Rev. E – Stat. Nonlinear, Soft Matter Phys. 86 (2012) 1–11. [26] P.B. Muller, R. Barnkob, M.J.H. Jensen, H. Bruus, A numerical study of microparticle acoustophoresis driven by acoustic radiation forces and streaming-induced drag forces, Lab Chip 12 (2012) 4617. [27] K. Sritharan, C.J. Strobl, M.F. Schneider, A. Wixforth, Z. Guttenberg, Acoustic mixing at low Reynold’s numbers, Appl. Phys. Lett. 88 (2006) 1–3. [28] N. Aoki, S. Hasebe, K. Mae, Mixing in microreactors: effectiveness of lamination segments as a form of feed on product distribution for multiple reactions, Chem. Eng. J. 101 (2004) 323–331. [29] A.P. Sudarsan, V.M. Ugaz, Multivortex micromixing, PNAS 103 (2006) 7228–7233. [30] F. Schönfeld, V. Hessel, C. Hofmann, An optimised split-and-recombine micro-mixer with uniform ‘chaotic ’ mixing, Lab Chip 4 (2004) 65–69. [31] M.A. Ansari, K. Kim, Mixing performance of unbalanced split and recombine micomixers with circular and rhombic sub-channels, Chem. Eng. J. 162 (2010) 760–767. [32] P. Augustsson, R. Barnkob, S.T. Wereley, H. Bruus, T. Laurell, Automated and temperature-controlled micro-PIV measurements enabling long-term-stable microchannel acoustophoresis characterization, Lab Chip 11 (2011) 4152–4164.
Fig. 6. Vector plot showing the trajectories of 0.5 µm particles on three different positions in the microfluidic channel. Shown is the particle trajectory at an actuation voltage of 10 Vp-p. Table 1 Average velocities of 0.5 µm particles for different actuation voltages and positions in the microfluidic channel. Voltage (Vp-p)
Bottom (µm s−1)
Middle (µm s−1)
Top (µm s−1)
Average velocity (µm s−1)
5 7 10 15
6,1 ± 0,3 12,8 ± 1,5 21,5 ± 4,7 51,2 ± 3,6
4,8 ± 0,8 14,0 ± 1,5 22,2 ± 2,8 54,6 ± 5,7
9,5 ± 1,5 15,8 ± 3,6 22,5 ± 5,1 42,1 ± 9,5
6,8 ± 0,6 14,2 ± 1,2 22,0 ± 1,3 49,3 ± 3,0
Conflicts of interest There are no conflicts to declare. Acknowledgements Wim De Malsche and Pierre Gelin greatly acknowledge the European Research Council for the support through the ERC Starting Grant EVODIS (grant number 679033EVODIS ERC-2015-STG). References [1] C. Busatto, J. Pesoa, I. Helbling, J. Luna, D. Estenoz, Effect of particle size, polydispersity and polymer degradation on progesterone release from PLGA microparticles: experimental and mathematical modeling, Int. J. Pharm. 536 (2018)
375
Contents lists available at ScienceDirect
Chemical Engineering Journal journal homepage: www.elsevier.com/locate/cej
Study on the mixing and migration behavior of micron-size particles in acoustofluidics Pierre Gelina, Özlem Sardan Sukasa, Karine Hellemansb, Dominique Maesc, Wim De Malschea,
T ⁎
a
Department of Chemical Engineering, µFlow Group, Vrije Universiteit Brussel, 1050 Brussels, Belgium Diabetes Research Centre, Vrije Universiteit Brussel, 1000 Brussels, Belgium c Structural Biology Brussels, Vrije Universiteit Brussel, 1050 Brussels, Belgium b
H I GH L IG H T S
mixing in microfluidic channel using bulk acoustic waves. • Fast transfer enhancement of particles by acoustic streaming. • Mass µm particles focus on the pressure node. • 5Focusing velocity varies along width of channel. • 0.5 µm particles flow along with the flow vortices. •
A R T I C LE I N FO
A B S T R A C T
Keywords: Acoustofluidics Mixing Particle tracking Particle focusing Acoustic streaming
Slow mass transfer remains a limiting factor in microfluidic devices, as it is purely based on molecular diffusion. Hence, the completion of a targeted chemical reaction is extremely slow. Acoustic streaming is a great methodology to create lateral convection in otherwise laminar flow operated channels and offers the possibility to reduce mass transfer resistance dramatically. Moreover, the acoustic field can elegantly by used to handle micron-sized particles. Large particles (> 2 µm), experience a radiation force that is much larger than the drag force related to streaming and will rapidly focus at either the node or anti-node of the standing pressure wave in the channel. In the present contribution, the mass transfer rate in both a drag force dominated as well as in a radiation force dominated regime is studied by in situ tracking polystyrene particles of respectively 0.5 and 5 µm. The 5 µm particles focus on the pressure node experiencing a migration velocity that decreases as the node position is approached, whereas the 0.5 µm particles remain in the vortex flow, characteristic for acoustic streaming. The experimentally obtained migration rates are in accordance with numerical simulations, suggesting that operational parameters as actuator displacement have been chosen appropriately. Mixing times of parallel flows were assessed and revealed an order of magnitude enhancement for small molecules, up to four order of magnitudes for 0.5 µm diameter particles.
1. Introduction During the past decade microfluidic devices have gained increased interest for their potential in physicochemical conversions. The dominating laminar flow (Re < 1), high surface to volume ratio and well defined residence time in microfluidic devices offer great advantages for chemical reaction and the production of highly monodisperse droplets and/or particles. Formation and handling of hard particles is common in the pharmaceutical, photonic, coating, textile and cosmetic
sector. Micron-sized monodispersed poly(lactic-co-glycolic acid) (PLGA) microparticles, for example, are used as a slow drug delivery system [1,2]. Apart from the large control over the applied flows, the small dimensions offer better heat transfer and greater yields. With diffusion coefficients in the order of 10−9 m2 s−1 for small molecules and lower than 10−10 m2 s−1 for larger molecules, diffusion based mixing in microfluidic devices is inherently slow. Both the in situ formation of hard particles or completion of a chemical reaction is therefore hindered. Moreover, it has been shown that rapidly
⁎
Corresponding author at: Department of Chemical Engineering, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Elsene, Belgium. E-mail addresses: [email protected] (P. Gelin), [email protected] (Ö. Sardan Sukas), [email protected] (K. Hellemans), [email protected] (D. Maes), [email protected] (W. De Malsche). https://doi.org/10.1016/j.cej.2019.03.004 Received 18 July 2018; Received in revised form 27 February 2019; Accepted 1 March 2019 Available online 02 March 2019 1385-8947/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
focus at the pressure node or antinode, while small particles will flow along with the vortices of the acoustic streaming. The cut-off between large and small particles is typically around 1.5 µm [25]. In the present paper, BAW and the related acoustic streaming is utilized to enhance the mass transfer rate far beyond the rate of diffusion. This is shown by fast mixing of two parallel flows with a specific aim for chemical conversion and particle synthesis. Moreover, the radiation force arising from the BAW together with the Stokes drag forces are investigated in the context of particle formation and manipulation. This is done for sub- and micron-size particles. Particle trajectories and velocities are simulated in COMSOL and confirmed experimentally. These deeper insights are of great relevance to perform in situ chemical reactions or for the synthesis of microparticles.
generating a homogeneous mixture enhances the selectivity of kinetically controlled chemical reactions [3]. Fast mixing also allows to reduce the residence time drastically minimizing the production of byproducts [4–6]. Passive mixers have the possibility to enhance mass transfer in microfluidic channels. These, however, often require high velocities and associated low residence times [7–9]. Formation and handling of particles in microfluidic devices is becoming a necessity [10], but passive mixers are very prone to fouling under these conditions due to the robust structures or many turns present in the channels [11,12]. Therefore they are not realistically suited for robust and reliable applications involving particles. An increasing interest in applying acoustics for mass transfer enhancement and particle manipulation is being observed [13–15]. Apart from the lower fouling probability compared to passive mixers, the use of acoustic waves gives more control over the mixing process. Two main types of acoustic waves in the microfluidic channel exist, surface acoustic waves (SAW) and bulk acoustics waves (BAW). SAWs are generally produced by applying an appropriate electric field to a piezoelectric material, which in turn generates a propagating mechanical wave in an elastic material. A part of this travelling SAW is refracted as a longitudinal pressure wave when contacted with a liquid. These pressure waves generate a force in their propagation direction and induce flow within the confined liquid [16]. Mixing two parallel liquid streams using SAW induced acoustic streaming in less than a second has been reported [17]. Herein a mixing efficiency (η) of 0.88 was reached at an actuation voltage of 80 Vp-p and a Bodenstein number of 7.4 * 103. Zeng et al. [18] reported a mixing time of 11 ms at a Bodenstein number of 8.64 * 102 and actuation voltage of 25 Vp-p. Here the Bodenstein number is defined as:
Bo =
u. L D
2. Materials and methods 2.1. Numerical simulations In order to gain more insight in the acoustic streaming and radiation phenomena, simulations were carried out with COMSOL® Multiphysics software. The numerical model used has been described elsewhere [26] and has been adapted towards the present study. In short, the cross section of a channel with a height of 70 µm and width of 375 µm has been built as a model. Next, the first order acoustic field was calculated with the Thermoviscous Acoustic, Frequency Domain interface. Subsequently, the second order time average flow was calculated with the Laminar Flow interface. Amplitude of the channel wall displacement could be varied to obtain different acoustic pressures and related streaming velocities. For mixing experiments, the channel was divided in three equal parts to ensure an optimal position of the vortices on the concentration profile. The mixing of diluted species was visualized with the Transport of Diluted Species interface. The diffusion constant was set to 10−9 m2 s−1. Finally, the Particle Tracking interface was used to visualize particle trajectory of large and small polystyrene particles (5.0 and 0.5 µm, respectively).
[1]
with u the linear velocity (m/s), L the width of the channel (m) and D the diffusion coefficient (m2/s). In the present paper the focus lies on the use of BAW in microfluidic channels. Herein, ultrasound waves in the low MHz range are applied to the microfluidic channel, creating a standing pressure wave. Viscous attenuation of this pressure wave leads to the streaming phenomenon called acoustic streaming, which take the form of vortices perpendicular to the channel [19]. A more in depth description of acoustic streaming can be found elsewhere [20–22]. Streaming velocities in microfluidic channels are typically in the order of ten to hundreds of µm s−1 [23]. Particles in a microfluidic channel are affected by both the acoustic pressure wave and the vortex flow, respectively corresponding to the radiation force and Stokes drag force. The radiation forces arise from scattering of the pressure wave, for a spherical particle with radius a, density ρp and compressibility κp, suspended in a liquid of density ρ0, compressibility κ0, dynamic viscosity η, and momentum diffusivity υ = η/ρ0 it is given by [24]:
Frad = 4π Φa3k 0 Eac sin(2k 0 y )
2.2. Chip design and set up Rectangular microfluidic channels were patterned in a positive resist using mid-UV lithography and subsequently etched by Bosch type DRIE etching in a silicon wafer with a thickness of 525 µm. The width and depth of the main channel is 375 and 70 µm, respectively. The main channel has a length of 33 mm. The channels were sealed by an anodically bonded borosilicate glass lid. Two inlets and outlets were provided. The geometry of the acoustofluidic chip is shown in Fig. 1a. Two syringe pumps (KDS200, KD Scientific Inc., USA) were used to introduce the liquids in the channels. Nanoport assemblies (N-333, IDEX Corp.) and glass capillaries ensured the connection from the chip to the pumps. To generate a standing pressure wave, the acoustofluidic chip was connected to a piezo ceramic element (PZT) (30 mm × 20 mm × 1 mm, APC International Ltd., USA) with a resonance frequency at 2.0 MHz. Coupling of the microfluidic chip to the PZT was done by a thin glycerol layer. An in-house built PMMA holder ensured a good connection between the chip, PZT and Nanoports. A schematic representation of the acoustofluidic chip is given in Fig. 1b. The PZT was driven by a frequency generator (AFG1062, Tektronix UK Ltd., UK) around 2.0 MHz. The applied voltage was amplified by an RF power amplifier (210L, Acquitek S.A.S, France) with a power output of 10 Watt. The voltage amplitude across the PZT was monitored with an oscilloscope (TBS1104, Tektronix UK Ltd., UK).
[2]
with k0 the angular wavenumber for the first harmonic, Eac the time average acoustic energy, Φ the acoustic contrast factor. The contrast factor varies between −1 < Φ < 1 and is a measure of how ‘hard’ a particle is compared to the surrounding liquid. A ‘hard’ (0 < Φ) particle will migrate towards and focus at the pressure node while a ‘soft’ particle (Φ < 0) will migrate towards the pressure antinodes. Next to radiation forces, particles also experience Stokes drag forces due to the streaming vortices:
Fdrag = 6πηa (vstr − vp)
[3]
with vstr the acoustic streaming velocity and vp the initial particle velocity from the radiation force. At equilibrium Fdrag + Frad = 0 and the part particle velocity equals:
vp =
Frad + vstr 6πηa
2.3. Experimental procedure Visualization of all the experiments was done using an inverted fluorescent microscope (IX71, Olympus Corporation, Japan) in combination with a CCD camera (C9100-13, Hamamatsu Photonics, Japan).
[4]
Large particles are mostly affected by the radiation force and will 371
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
I 'i =
Ii − Ib n (I − I ) ∑i = 1 i n b
[6]
with Ii’ the normalized intensity on row i, calculated based on the intensity on row i (Ii) and the background intensity (Ib). I′i,∞ the normalized intensity on row i at complete mixing, I′i,0 the normalized intensity on row i at the start of the mixing experiment. It should be noted that for liquids, mixing by molecular diffusion takes place at very small scales. Since the pixel size was 1.45 µm, this leads to an averaging of the concentration in one pixel. This, however, has a negligible effect on the mixing efficiency. Doubling the pixel size changed the mixing efficiency by only 0.003% for a representative mixing condition (voltage = 10Vpp, flowrate = 5 mm/s). Mixing efficiency was evaluated as a function of the applied voltage for two different flow velocities (5 mm s−1 and 10 mm s−1). The voltage was varied between 5 Vp-p and 20 Vp-p in both cases. 2.3.2. Experimental particle tracking. Fluorescent monodispersed polystyrene microparticles (Microparticles GmbH, Germany) with a size of 0.45 µm (SD: 0.01 µm) and 5.19 µm (SD: 0.14 µm) were used for particle tracking experiments. These are further referred to as 0.5 µm and 5.0 µm particles respectively. The 5 µm and 0.5 µm particles were diluted to a solid content of 0.25 w/v% and 0.0125 w/v%, respectively. Once particles were introduced in the acoustofluidic chip, the actuator was activated and the effect on the particles was examined. Based on the obtained movies, the radiation velocity of 5.0 µm particles as a function of their position in the channel was analyzed with a customized MATLAB code, this for different voltages (5–15 Vp-p). Streaming without flow in the axial direction, was visualized at three different channel heights (top, middle and bottom). Their related streaming velocities were analyzed with a similar MATLAB code. The standard deviation was calculated on four independently recorded movies.
Fig. 1. a) Top view of the etched microfluidic chip used in all mentioned experiments. The main channel has a length, width and height of 33 mm, 375 µm and 70 µm respectively. Two inlets and two outlets are provided. b) Schematic representation of microfluidic channel: The microfluidic channel is connected to a piezo ceramic element vibrating at 1.93 MHz. Actuation is done with a frequency generator. The vibrations are transferred to the channel wall, creating a standing pressure wave (thicknesses are not to scale).
2.3.1. Mixing experiments For the mixing experiments, water containing a fluorescent salt (fluorescein sodium salt, Sigma-Aldrich, Germany) and non-fluorescent DI water were brought into contact in the microfluidic chip. Mixing of the two water streams was monitored at 16 mm from the intersection at the middle of the main channel. Mixing was initiated by actuating the piezo ceramic element at a frequency of 1.93 MHz. The recorded movies were analyzed with ImageJ software. The intensity of each pixel was related to the fluorescence intensity and eventually to the efficiency of the mixing. By averaging the pixel array along the width of the channel, the mixing efficiency or degree of mixing (η) can be deduced [17]:
η=1−
1 n 1 n
3. Results and discussion The following results show the possibility to use BAW and the related acoustic radiation/streaming to enhance mass transfer greatly in microfluidic channels as well as the manipulation of large and small particles.
n
∑i = 1 (I 'i − I 'i, ∞ )2 n
∑i = 1 (I 'i,0 − I 'i, ∞ )2
3.1. Numerical particle tracking
[5]
Theoretically the particle trajectory is given by Eq. (4), a Fig. 2. Numerical simulation of particle trajectory for large (5 µm) and small (0.5 µm) polystyrene particles in the cross section of a channel at a displacement amplitude of 0.5 nm. a) Large particles focus fast (in 0.2 s) on the pressure node. The velocity of focusing is not constant over the complete width of the channel, varying from 0 to 1600 µm s−1. b) Small particles flow along with the streaming vortices (the liquid velocity field is shown by the red arrows).
372
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
obtained simulations were validated experimentally. Mixing due to the presence of an acoustic field was visualized by a camera. Fig. 4b shows snapshots at different times at a fixed observation position (16 mm downstream the contact area) during the mixing process, overlaid with the concentration profile. In this case the flow rate was set to 10 mm s−1 and the actuation voltages to 20 Vp-p. At t = 0.30 s the channel already had a mixing efficiency of η = 0.91, but in took roughly 1 s more to accomplish complete mixing. The mixing efficiency has been measured for a range of applied voltages and flowrates (Fig. 4c). Actuation at voltages lower than 10 Vp-p does not induce any visible mixing. For a linear velocity of 5 mm s−1 a mixing of already η = 0.65 is reached at 10 Vp-p actuation. Complete mixing (η = 0.94) is reached at an actuation voltage of 15 Vp-p for both flowrates. These experimentally obtained mixing rates are in accordance with the numerical simulations, indicating that operational parameters such as actuator displacement have been chosen appropriately. To our knowledge such fast mixing in a microfluidic channel with bulk acoustic waves has not been demonstrated yet. However, SAW has been utilized to achieve even faster mixing in similar channels [17,18,27]. Fast mixing has also been achieved in small laminated channels in the context of complex chemical reactions [6,28] and has been observed in larger channels by flow destabilisation [7,29–31]. In the present paper, however, we have focused on relatively large channel dimensions wherein the formation and use of particles is manageable under realistic conditions. Important to note is the absence of passive mixing elements which often lead to fouling of the microfluidic channel once particles are introduced. The current use of BAW for fast mixing is limited by energy transfer into the fluid. Acoustic streaming is a secondary flow effect and requires a high energy input to have a significant effect. Applying higher voltages will have a direct effect on the streaming velocity and thus on the mixing efficiency. At higher voltages, however, rises in temperature must be taken into account. It is so that the resonance frequency of the acoustofluidic device depends on the temperature [32]. A change in temperature will shift the resonance frequency and less energy will be transferred into the microfluidic channel.
combination of Stokes drag forces from the vortices and the radiation forces. Fig. 2 shows a numerical simulation for the particle trajectory of both 5 µm and 0.5 µm polystyrene particles. The simulation shows migration of 5 µm particles towards the pressure node (Fig. 2a). The focusing velocity, however, is not constant over the width of the channel. A decrease in velocity is observed as the node position is reached, which is in agreement with Eqs. (2) and (4). For a numerically imposed displacement amplitude of 0.5 nm, a maximum migration velocity of 1462 µm s−1 is reached. The equilibrium position, where lateral movement ceases, is reached after 0.2 s. In contrast to the large particles, the small 0.5 µm particles do not focus on the pressure node (Fig. 2b). Instead they flow along with the vortex flow of the acoustic streaming. The particle velocity varies strongly as a function of the particle’s position in the vortex. The velocities that are reached on the outer areas of the vortices are around 100 µm s−1. In the literature radiation velocities of 44 µm s−1 and streaming velocities of 7 µm s−1 are reported [26] for a displacement amplitude of 0.1 nm. Comparable values are obtained in our case for the same displacement amplitude. In the context of fast mixing for particle maturation and chemical reactions we have chosen to show the results with the higher displacement amplitude of 0.5 nm. Experimentally obtained radiation velocities (actuation voltage of 15Vp-p) were fitted using the obtained simulation (Fig. 3). Small changes in displacement amplitude have a large influence on the radiation velocity. A satisfactory standard deviation (S = 260 µm s−1) of the residuals (vmeasured – vcurve) was obtained for a fit on the 0.5 nm displacement amplitude curve. 3.2. Acoustic mixing Mixing in microfluidic devices remains a considerable hurdle in the context of microfluidics. In (generally occurring) laminar flow conditions, mixing only relies on molecular diffusion which is inherently slow. To assess the potential of mixing improvement with respect to diffusion-based mixing, mixing of two flows in a microfluidic channel with acoustic streaming was simulated with COMSOL. The results are shown in Fig. 4a. At a displacement amplitude of 1 nm, complete mixing of the channel is observed in 1.52 s. This is about ten times faster compared with simple molecular diffusion. Because mass transport occurs by active mixing, the mixing rate is to a much lower extend dependent on the diffusion coefficient. Small particles of 0.5 µm (D = 10−13 m2 s−1) in diameter will mix as fast as a dye. Next, the
3.3. Experimental particle tracking 3.3.1. Large particles Trajectories and velocities of 5 µm polystyrene microparticles in the presence of the acoustic field were visualized with the camera. Particle migration was triggered as the acoustic field was activated. The radial migration velocity as function of the applied amplitude is shown in Fig. 5. A higher amplitude results in larger focusing velocity. This is in line with expectations, as the acoustic energy (Eac) scales with radiation force (Frad) (Eq. (2)). Velocities as high as 1600 µm s−1 were reached for an actuation of 20 Vp-p. As predicted by theory (Eq. (4)) and according to the previously performed simulation, the particle velocity is low at the channel wall, rises gradually to a maximum and eventually goes back to zero when reaching the equilibrium position. Notice that at low radiation velocity near the equilibrium position, a small inflection is observed, which is not present in the simulation data (Fig. 3). This can be attributed to a small mismatch between the pressure wavelength and the channel width leading to a small variation in the position of the pressure node along the length of the channel. Averaging the radiation velocity of all the particles, will lead to this small deviation from theory. 3.3.2. Small particles In contrast to the large particles, 0.5 µm particles follow the streaming lines of the vortex flow. Particle trajectories and velocities were measured at 3 different heights of the channel: top, middle and bottom. These obtained PIV data (Fig. 6) reveal that at the top and bottom layer, particles flow towards the channel wall, while at the middle layer particles move towards the channel center. This confirms
Fig. 3. Comparison of experimental radiation velocities (▲) with radiation velocities obtained with COMSOL for different displacement amplitudes. The best fit was obtained for a displacement amplitude of 0.5 nm. In this case the standard deviation of residuals equals 260 µm/s. 373
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
Fig. 4. a) Cross sectional view of a microfluidic channel. Shown is a numerical simulation for the mixing of two liquid streams in presence of an acoustic field. Displacement amplitude of 1 nm was set to the left and right wall. Complete mixing is reached after 1.52 s. b) Top view of microfluidic channel. Shown is the mixing at a flow rate of 10 mm s−1 and an actuation of 20Vp-p. A mixing efficiency of 97% is reached after 1.53 s. c) Mixing efficiency in function of applied voltage for two different flow rates, 5 mm s−1 (♦) and 10 mm s−1 (□). Mixing was measured at 16 mm from the point where the flows meet. At lower voltages the water streams mix better for a low flow rate of 5 mm s−1. The time needed for complete mixing is around 1 s and 2 s for flow rates of 5 mm s−1 and 10 mm s−1, respectively.
channel wall and the centre of the channel can be estimated to be 3 s, which is considerably longer that the observed 1.5 s, which indicates that diffusion between the different flow layers in the vortex represents an important contribution to mixing. The mass transfer characteristics within the vortex will be further explored in a follow-up study with a recently in-house developed 3D PIV instrument.
4. Conclusion In this article mixing of two parallel flows by acoustic streaming was characterized. We demonstrated that bulk acoustic waves (BAW) can efficiently be applied for very fast mixing of two parallel liquid streams in microfluidic devices. Complete mixing is achieved in 1.53 s at a Bodenstein number of 2.5 * 103 and actuation voltage of 20 Vp-p. Furthermore we investigated the kinetics of BAW in the context of particle manipulation. Large and small particles are affected differently by the acoustic field. Large particles are mainly affected by the radiation forces and focus towards the pressure node. Displacement velocities vary depending on the position of the particle in the channel. For 5 µm polystyrene particle velocities as high as 1600 µm s−1 were reached for an actuation voltage of 20Vp-p. Small particles do not focus and flow along with the liquid vortices. Maximum velocities of 48 µm s−1 were reached for an actuation voltage of 15 Vp-p. These deeper insights in the interactions between particles and the acoustic field are of paramount importance for the continuous production and handling of micro- or nanospheres in microfluidic devices. BAW in microfluidic devices is an excellent tool to enhance mass transfer far beyond the rate of diffusion, this is of great relevance to conduct chemical reactions in these confined environments.
Fig. 5. a) Radial focus velocity of 5 µm particles for five different voltages (see legend). At the start particles have a low velocity, this rises up to a maximum to finally decrease when the pressure node is reached. Velocities as high as 1600 µm s−1 has been reached for a voltage of 20 Vp-p. b) Top view of microfluidic channel showing the fast focusing of 5 µm polystyrene particles. At 10 Vp-p equilibrium is reached after 0.267 s.
the presence of vortices in the microfluidic channel. The average radial velocities of the particles are given in Table 1 for four different voltages. Note that the absolute values of the velocity vectors are given, velocity vectors in the central plane have an opposite direction to the vectors on the top and bottom planes. At 15 Vp-p, the travel time between the 374
Chemical Engineering Journal 369 (2019) 370–375
P. Gelin, et al.
360–369. [2] Q. Xu, et al., Preparation of monodisperse biodegradable polymer microparticles using a microfluidic flow-focusing device for controlled drug delivery, Small 5 (2009) 1575–1581. [3] S. Suga, A. Nagaki, J. Yoshida, Highly selective Friedel – crafts monoalkylation using micromixing † compounds with N-acyliminium ions has been achieved by, Chem. Commun. 354 (2003), https://doi.org/10.1007/s00170-003-1973-7. [4] J.I. Yoshida, A. Nagaki, T. Iwasaki, S. Suga, Enhancement of chemical selectivity by microreactors, Chem. Eng. Technol. 28 (2005) 259–266. [5] K.S. Elvira, X.C.I. Solvas, R.C.R. Wootton, A.J. Demello, The past, present and potential for microfluidic reactor technology in chemical synthesis, Nat. Chem. 5 (2013) 905–915. [6] T. Asai, et al., Switching reaction pathways of benzo[b ]thiophen-3-yllithium and Benzo[b ]furan-3-yllithium based on high-resolution residence-time and temperature control in a flow microreactor, Chem. Lett. 40 (2011) 393–395. [7] R. Goovaerts, W. Smits, G. Desmet, J. Denayer, W. De Malsche, Combined improved mixing and reduced energy dissipation by combining convective effects and lamination, Chem. Eng. J. 211–212 (2012) 260–269. [8] S.R.L. Gobert, S. Kuhn, L. Braeken, L.C.J. Thomassen, Characterization of milli- and microflow reactors: mixing efficiency and residence time distribution, Org. Process Res. Dev. 21 (2017) 531–542. [9] T. Vandermeersch, R. Goovaerts, J. Luyten, J.F.M. Denayer, W. De Malsche, Tracking the liquid-liquid extraction performance in mesoflow reactors, Chem. Eng. J. 279 (2015) 9–17. [10] P. Gelin, et al., Focusing of microcrystals and liquid condensates in acoustofluidics, Crystals 9 (2019) 120. [11] K. Wu, S. Kuhn, Strategies for solids handling in microreactors, Chim. Oggi/Chem. Today 32 (2014) 62–66. [12] R.L. Hartman, Managing solids in microreactors for the upstream continuous processing of fine chemicals, Org. Process Res. Dev. 16 (2012) 870–887. [13] K.D. Frampton, S.E. Martin, K. Minor, The scaling of acoustic streaming for application in micro-fluidic devices, Appl. Acoust. 64 (2003) 681–692. [14] D. Fernandez Rivas, S. Kuhn, Synergy of microfluidics and ultrasound, Top. Curr. Chem. 374 (2016) 70. [15] H. Bruus, Acoustofluidics 1: governing equations in microfluidics, Lab Chip 11 (2011) 3742. [16] X. Ding, et al., Surface acoustic wave microfluidics, Lab Chip 13 (2013) 3626. [17] T.-D. Luong, V.-N. Phan, N.-T. Nguyen, High-throughput micromixers based on acoustic streaming induced by surface acoustic wave, Microfluid. Nanofluid. 10 (2011) 619–625. [18] Q. Zeng, et al., Milliseconds mixing in microfluidic channel using focused surface acoustic wave, Sens. Actuat. B Chem. 160 (2011) 1552–1556. [19] N. Riley, Steady treaming, Annu. Rev. Fluid Mech. 33 (2001) 43–65. [20] M. Wiklund, R. Green, M. Ohlin, Acoustofluidics 14: applications of acoustic streaming in microfluidic devices, Lab Chip 12 (2012) 2438–2451. [21] P.B. Muller, H. Bruus, Numerical study of thermoviscous effects in ultrasound-induced acoustic streaming in microchannels, Phys. Rev. E – Stat. Nonlinear, Soft Matter Phys. 90 (2014) 1–12. [22] A.Y. Rednikov, S.S. Sadhal, Acoustic/steady streaming from a motionless boundary and related phenomena: generalized treatment of the inner streaming and examples, J. Fluid Mech. 667 (2011) 426–462. [23] P.B. Muller, et al., Ultrasound-induced acoustophoretic motion of microparticles in three dimensions, Phys. Rev. E – Stat. Nonlinear, Soft Matter Phys. 88 (2013) 1–12. [24] J.T. Karlsen, H. Bruus, Forces acting on a small particle in an acoustical field in a thermoviscous fluid, Phys. Rev. E – Stat. Nonlinear, Soft Matter Phys. 92 (2015). [25] R. Barnkob, P. Augustsson, T. Laurell, H. Bruus, Acoustic radiation- and streaminginduced microparticle velocities determined by microparticle image velocimetry in an ultrasound symmetry plane, Phys. Rev. E – Stat. Nonlinear, Soft Matter Phys. 86 (2012) 1–11. [26] P.B. Muller, R. Barnkob, M.J.H. Jensen, H. Bruus, A numerical study of microparticle acoustophoresis driven by acoustic radiation forces and streaming-induced drag forces, Lab Chip 12 (2012) 4617. [27] K. Sritharan, C.J. Strobl, M.F. Schneider, A. Wixforth, Z. Guttenberg, Acoustic mixing at low Reynold’s numbers, Appl. Phys. Lett. 88 (2006) 1–3. [28] N. Aoki, S. Hasebe, K. Mae, Mixing in microreactors: effectiveness of lamination segments as a form of feed on product distribution for multiple reactions, Chem. Eng. J. 101 (2004) 323–331. [29] A.P. Sudarsan, V.M. Ugaz, Multivortex micromixing, PNAS 103 (2006) 7228–7233. [30] F. Schönfeld, V. Hessel, C. Hofmann, An optimised split-and-recombine micro-mixer with uniform ‘chaotic ’ mixing, Lab Chip 4 (2004) 65–69. [31] M.A. Ansari, K. Kim, Mixing performance of unbalanced split and recombine micomixers with circular and rhombic sub-channels, Chem. Eng. J. 162 (2010) 760–767. [32] P. Augustsson, R. Barnkob, S.T. Wereley, H. Bruus, T. Laurell, Automated and temperature-controlled micro-PIV measurements enabling long-term-stable microchannel acoustophoresis characterization, Lab Chip 11 (2011) 4152–4164.
Fig. 6. Vector plot showing the trajectories of 0.5 µm particles on three different positions in the microfluidic channel. Shown is the particle trajectory at an actuation voltage of 10 Vp-p. Table 1 Average velocities of 0.5 µm particles for different actuation voltages and positions in the microfluidic channel. Voltage (Vp-p)
Bottom (µm s−1)
Middle (µm s−1)
Top (µm s−1)
Average velocity (µm s−1)
5 7 10 15
6,1 ± 0,3 12,8 ± 1,5 21,5 ± 4,7 51,2 ± 3,6
4,8 ± 0,8 14,0 ± 1,5 22,2 ± 2,8 54,6 ± 5,7
9,5 ± 1,5 15,8 ± 3,6 22,5 ± 5,1 42,1 ± 9,5
6,8 ± 0,6 14,2 ± 1,2 22,0 ± 1,3 49,3 ± 3,0
Conflicts of interest There are no conflicts to declare. Acknowledgements Wim De Malsche and Pierre Gelin greatly acknowledge the European Research Council for the support through the ERC Starting Grant EVODIS (grant number 679033EVODIS ERC-2015-STG). References [1] C. Busatto, J. Pesoa, I. Helbling, J. Luna, D. Estenoz, Effect of particle size, polydispersity and polymer degradation on progesterone release from PLGA microparticles: experimental and mathematical modeling, Int. J. Pharm. 536 (2018)
375