Ultrasonic separation of suspended particles

Reprint Proc. of the 2001 IEEE International Ultrasonics Symposium, a Joint Meeting with the World Congress on Ultrasonics, Atlanta, Georgia, USA, Oct. 7-10, 2001 ULTRASONIC SEPARATION OF SUSPENDED PARTICLES E. Benes, M. Gröschl, H. Nowotny, F. Trampler1, T. Keijzer1, H. Böhm 2, S. Radel3, L. Gherardini3, J.J. Hawkes4, R. König4, Ch. Delouvroy5 Vienna University of Technology, Institut für Allgemeine Physik and Institut für Theoretische Physik, Wiedner Hauptstr. 8, A-1040 Vienna, Austria 1Applikon Dependable Instr. B.V., AE Schiedam, The Netherlands 2University of Nottingham, School of Biological Sciences, Nottingham, UK 3University College Dublin, Industrial Microbiology, Dublin, Ireland 4School of Biosciences, Cardiff University, Cardiff, UK 5Institut für Entsorgung und Umwelttechnik GmbH, Iserlohn, Germany Abstract: The forces on suspended particles in acoustic fields are reviewed briefly and the theoretical modelling of ultrasonic separators based on piezoelectrically excited layered resonators is described. Two flow-through resonator chamber concepts for ultrasonic particle (bio-cell) separation are investigated: a) the coagulation or sedimentation approach, b) the so-called h-shaped ultrasonic separator. The h-shaped ultrasonic separator is analysed by combining for the first time the mathematical modelling of the laminar flow with the acoustic force based velocity field of the particles relative to the suspension medium. This allows a complete modelling of the resonator's particle separation performance. Examples for separation chamber designs optimized by use of the mathematical model are presented and the calculated particle traces in the h-resonator are shown and compared with experimental results. For direct comparison of different ultrasonic flow through separator concepts a separation performance figure is introduced and its value is given for the two investigated separator concepts for the sample suspensions of polystyrene spheres, yeast and spirulina cells in (salt) water. The presented results are of importance for the state of the art design of acoustic cell filters for perfusion type bioreactors, as recently launched at the biotechnology market, as well as for the ultrasonic separation of plant (algae) cells under low gravity conditions, where the sedimentation concept fails. INTRODUCTION - THE PHENOMENON The use of acoustic standing waves to concentrate initially homogeneously suspended particles at acoustic pressure nodal, or antinodal, planes within a fluid was first described by Kundt und Lehmann [1]. The effect was originally used only to make ultrasonic fields visible. However, the interaction of standing ultrasonic waves with particles dispersed in a fluid produces forces on the particles which can be utilized for the separation of the dispersed particles from the fluid as well [2-15]. Fig.1 shows the effect of a standing wave on a suspen- 0-7803-7177-1/01/$10.00 © 2001 IEEE sion consisting of small Pyrex glass spheres with diameter d ≈ 100 µm in water. The standing wave of frequency f = 0.67 MHz is generated in an acrylic glass tube with an inside diameter of 40 mm, whereby the amplitude of the velocity field shows a symmetric distribution in lateral direction around the axis of symmetry of the tube. Fig.1. Photograph of the pattern of particle positions in a vertical ultrasonic standing wave field. The example shows spherical Pyrex glass particles with diameter d ≈ 0.1 mm, excitation frequency is 670 kHz corresponding to a half-wavelength λ/2 = 1.1 mm, viewing window 1 cm x 1 cm. The theoretical explanation of the obeyed migration of the initially homogeneously dispersed glass spheres is given in Section 2, here it is only qualitatively described by means of Fig.1. At the beginning of irradiation, almost instantaneously the spheres are driven towards the acoustic pressure nodes, whereby the average distance between the particles considerably diminishes. Then the particles trapped within the planes migrate closer together, whereby coagulation and even coalescence may be triggered. The phenomenon is applicable to all kinds of dispersions. Table 1 shows the common characterization of dispersions according to the physical state of the particles and the dispersion medium. The particles can be gaseous, liquid, solid, or even biological cells. The dispersion fluid can be gaseous or liquid. The most important practical examples are particles of all kinds in air (aerosoles) or in water (hydrosoles). The overview about all possible kinds of dispersions given in Table 1 reviews the established characterising terms and indicates the great potential of the ultrasonic 2001 IEEE ULTRASONICS SYMPOSIUM — 649 Initial configuration separation technology. Nevertheless, the phenomenon has not yet gained widespread industrial application as the process can be highly sensitive to disturbances and involves acoustic forces that have to be compared with the separation speed limiting viscous drag forces. Final configuration Antinode Table 1. Kinds of Dispersions Medium → ↓ Particle Liquid Gas Solid Suspension Smoke Liquid Emulsion Fog, Mist Gas λ /4 Disp. Bubbles Although an extensive literature on the theory of the interaction of ultrasonic waves with particles already exists, there is still a gap between interesting research projects and an efficient acoustical separation technology of practical importance. However, with nowadays available highly advanced piezoelectric transducers and driving electronics, some commercial applications of this separation effect became visible [13, 16, 17], and the mentioned gap has been reduced significantly by the progress made in a joint European effort, the EUSS network, within the European Commission's Training and Mobility (TMR) Program for young researchers [14, 18]. This paper is focussing entirely on the application of the ultrasonic separation technology to hydrosoles including bio-cell suspensions, where most recent progress was made. ACOUSTIC RADIATION FORCES ON SUSPENDED PARTICLES Interaction between single particle and acoustic field Acoustic mean forces on suspended particles can be subdivided into forces on an individual particle (single particle and acoustic field interaction), i.e., the primary radiation force, and into the interparticle force between two or more particles in a sound field, i.e., the secondary radiation force. At the beginning of irradiation, the strong axial component FA of the primary radiation force drives the spheres towards the displacement velocity antinodes, by which the average distance between the particles considerably diminishes. Then, due to the transversal component F T of the primary radiation force and the secondary radiation force FI (Interparticle Force), the particles come closer together, whereby coagulation and even coalescence may be triggered (see Fig.2). All the forces mentioned above can be derived as special limit cases from one common physical principle, the momentum flow analysis. This analysis of the total momentum flow passing through the regarded particle was first described by Gor'kov [19]. According to Gor'kov, the radiation force caused by a standing wave is usually several orders of magnitude larger than that caused by a travelling wave of the same amplitude. Node FT FI FA λ /4 F Antinode Lateral displacement amplitude distribution Fig.2. Acoustic forces causing migration of dispersed spheres in a standing wave with a lateral amplitude variation. Acoustic radiation forces drive initially homogeneously suspended glass particles into the anti-node planes of the acoustic displacement amplitude and within these planes towards the regions of local amplitude maxima. The primary ("sound radiation") force in the propagation direction of the standing sound wave is the strongest force, which almost instantaneously drives the particles towards the antinode or node planes (dependent upon the acoustic properties of the dispersion) of the applied alternating acoustic displacement velocity field r r r rr FA (r p ) = 4πka 3 Eac K (ρ f , ρ p , c f , c p ) sin(2kr p ) , (1) r r whereby r p is the locus vector, k the wave vector, a the particle radius, Eac the time averaged acoustic energy density, K (ρ f , ρ p , c f , c p ) the acoustic contrast factor, ρ f , ρ p density of fluid and particle, c f , c p sound speed in fluid and particle, respectively. The derivation of this equation is based on the assumption that the particle radius is much smaller than the wavelength. From this equation it can be learned that the force is proportional to the third power of the particle radius, and thus is proportional to the particle volume. Further the acoustic radiation force is proportional to the time averaged acoustic energy density in the liquid. The spatial 650 Reprint Proc. of the 2001 IEEE International Ultrasonics Symposium, a Joint Meeting with the World Congress on Ultrasonics, Atlanta, Georgia, USA, Oct. 7-10, 2001 dependence has twice the periodicity of the standing wave, the direction of the force to or from the displacement velocity antinodal planes is determined by the sign of the acoustic contrast factor that depends on the ratio of the densities and sound speeds in the liquid and the particle material. As a general rule, solid particles are driven towards the displacement velocity antinode planes, while gas bubbles are driven towards the node planes. Interaction between two particles in an acoustic field The forces between two particles situated in an acoustic standing wave field have first comprehensevely been described by Bjerknes [21]. Within the displacement antinode planes, acoustic interparticle forces increase with the fourth power of the reciprocal distance between the particles and are therefore negligible for low particle concentrations. Since the concentration of the particles within the planes is much higher than in the initially homogenous dispersion, the average distance is much smaller and the chance for agglomerations due to interparticle forces increases. Nonetheless, agglomeration caused by these secondary forces typically needs several minutes. Besides that, the application of the effect to particle separation is restricted to particles which tend to coalesce and flocculate. This property is needed since the final macroscopic particle separation in the conventional coagulation approach is performed by gravity forces which are effective only on the bigger flocculated clusters. These gravity forces are rather weak and, after switching off the acoustic standing wave field, again need typically several minutes to sediment the flocculated particles on the bottom of the resonance cell. Hence, the use of conventional acoustic standing wave fields dramatically restricts the application to a small share of the numerous technically important suspensions and needs relatively long periods of sound treatment. A more comprehensive review of quantitative relations for the various types of forces acting on spherical obstacles freely suspended in a fluid medium has been given recently by Gröschl [22]. 0-7803-7177-1/01/$10.00 © 2001 IEEE For obtaining a well defined acoustic plane standing wave field, composite piezoelectric resonators are used that can be schematically described by a layered structure according to Fig.3. The electric excitation of the acoustic field is usually performed by PZT (Lead-Zirconate-Titanate) piezoceramic plates vibrating primarily in thickness extensional modes. PZT Piezoceramics Glass carrier B Glass reflector Agglomerated particles Suspension container A C D E l/2 y Further concentration of the particles within the antinode or node planes of the displacement velocity amplitude occurs by the transversal component of the primary radiation force that is proportional to the gradient of the acoustic energy density. Within the antinode planes suspended solid particles are therefore driven by forces perpendicular to the direction of the sound wave propagation and pointing towards the lateral displacement velocity amplitude maxima. [20] Layered piezoelectric resonator model v(y) Instead of the commonly used term "acoustic particle velocity" (in the german language "Schallschnelle") here and throughout this paper the term "acoustic displacement velocity" is used to avoid at one hand any confusion with the velocity of the dispersed particle and at the other hand with the propagation velocity of the acoustic wave. RESONATOR ANALYSIS V Fig.3. Scheme of a layered resonator. The outermost interface planes to the surrounding air function as almost totally reflecting acoustic mirrors for the acoustic wave excited (free surface boundary conditions). The glass plates function simultaneously as walls of the vessel for the liquid medium. One-dimensional theory The transfer matrix multilayer resonator model The transfer matrix model of piezoelectric multilayer resonators used in this work for calculation of filter properties was developed by Nowotny and Benes [23, 24]. Application of the model to piezoelectric resonators for particle separation was described first by Gröschl [22]. The model is based on the fundamental equations of piezoelectricity [25] that relate the coupled electro-acoustic field quantities, acoustic displacement u, mechanical stress T, electric potential ϕ (quasi-static approximation), and dielectric displacement D. The model is generally restricted to harmonic time-dependence and to the one-dimensional case, that is, all considered quantities are assumed to show spacedependence in only one direction (direction of sound propagation, thickness direction of the layers). Furthermore, in some situations, the displacement of the sound wave may be restricted to this direction as well (longitudinal waves only). This treatment is justified here, because the piezoceramic plate transducers used essentially permit electrical excitation of longitudinal waves only. As a consequence of this restriction to a single displacement direction, all material constants, which are tensor quantities in the general case, are reduced to scalars, whereby the medium (layer) under consideration is described by its relevant elastic stiffness constant c, piezoelectric constant e, and dielectric constant ε relevant for the regarded exten- 2001 IEEE ULTRASONICS SYMPOSIUM — 651 Practical application of the model requires knowledge of all material parameters involved, these are for each layer: thickness d, mass density ρ, elastic stiffness constant Liquid Glass Borofloat Schott Germany Glass Degassed water Piezo- Glue ceramic Borofloat Schott Germany For a given multilayer resonator, typically comprising an active piezoelectric layer followed by several passive solid and/or liquid layers, in a first step, the transfer matrix model allows calculation of the electrical admittance (real and imaginary parts) between the electrodes of the active layer as a function of frequency. In a second step, the spatial progress of the primary electro-acoustic field quantities u, T, ϕ, D within each layer, can be determined for a fixed frequency and a given voltage amplitude applied across the electrodes. This is accomplished by stepwise evaluation of the transfer matrix of the corresponding layer, whereby the layer thickness d is replaced by a variable thickness running from one boundary to the next. From the primary field quantities the spatial progress of other quantities of interest, like displacement velocity or acoustic energy density in the considered layer, can be derived easily [22]. Layer Araldite AV118 Ciba Geigy A general way of considering losses (viscoelastic damping, dielectric and piezoelectric losses) is the introduction of complex material constants [26]. Consequently, the transfer matrix elements and the acoustic and electric field quantities become complex, too. Specifically, viscoelastic and dielectric losses can be expressed by an acoustic quality factor Q and a dielectric loss angle δ, that form the imaginary parts of the material constants c and ε, respectively. (Piezoelectric losses can be neglected in most practical cases.) Other loss mechanisms relevant for acoustic particle filters, like sound attenuation due to bubbles or dispersed particles, losses due to divergence of the sound beam, etc., can be accounted for in a global way by using reduced Q-values (effective quality factors) for each material layer, as described by Gröschl [22]. Table 2. Input parameter values for the calculations PIC 181 PI Ceramic Germany According to the model, a homogeneous material layer is represented by a transfer matrix relating the boundary values of u, T, ϕ , and D at adjacent surfaces. The transfer matrix elements depend only on the material constants ρ (mass density), c, e, ε, the layer thickness d, and the angular frequency ω. The explicit expressions for the matrix elements were given by Nowotny and Benes [23]. For multiple layers, the total transfer matrix of the layered structure can be obtained by multiplication of the transfer matrices of each single layer. This follows from the continuity conditions that apply at interfacing surfaces. The electrodes of the piezoelectric layer have to be treated as separate films, but may be regarded as massless if they are sufficiently thin. On the outer free surfaces of the total sandwich arrangement, stress must be zero and dielectric displacement is assumed to be zero. This general free surface boundary condition establishes an expression for the electrical admittance Y between the electrodes. The explicit result for Y as a function of angular frequency ω and electrode area A was given by Nowotny and Benes [23]. c, effective quality factor Qeff . Furthermore, if the considered layer is located between electrodes: dielectric constant ε and dielectric loss angle δ ; and, in addition for piezoelectric layers: piezoelectric constant e (or electromechanical coupling factor). The parameter values used for the calculations presented in this work are listed in Table 2. For the mathematical expressions used see Gröschl [22] and Nowotny [23]. Thickness 1 [mm] Mass den- 7850 sity [kg/m3] Sound speed 4720 [m/s] Electrom. coupling 0.5 factor [-] Dielectric -9 constant 5·10 [As/Vm] Dielectric loss angle 0.004 [-] Electrode 570 area [mm2 ] Effective acoustic Q - 400 factor [-] 0.005 2.45 10 2.35 1100 2200 1000 2200 2400 5400 1490 5400 - - - - - - - - - - - - - - - - 2 200 3000 200 Material Manufacturer sional mode. "Relevant" quantity in this context means the parameter values valid for the regarded wave. Electrical admittance Fig.4 shows the electrical admittance spectrum (absolute values) of the multilayer-resonator section of the hshape separator (see Fig.14) without water filling, measured with a specialised computer-controlled electrical admittance measurement system [27], compared to the values predicted by the transfer matrix model described in the previous section. The calculation is based on the layered structure of the PZT/glass composite transducer, comprising a piezoceramic layer A and a glass layer C (compare Fig.3). 652 Reprint Proc. of the 2001 IEEE International Ultrasonics Symposium, a Joint Meeting with the World Congress on Ultrasonics, Atlanta, Georgia, USA, Oct. 7-10, 2001 10 the resonance at f = 2.2 MHz therefore corresponds to a total phase shift between the outer boundaries of 3 π. Additional resonances can be seen clearly in Fig.4, corresponding to a total phase shift of 2π (f = 1.5 MHz), and 4π (f = 2.8 MHz). Admittance [mhO] Measured Calculated 1 0.1 0.01 0.001 1 1.5 2 2.5 3 Frequency [MHz] Fig.4. Measured and calculated admittance versus frequency curves of the PZT-ceramic/glass composite transducer without consideration of the very thin bonding (glue) layer. All parameter values of the investigated composite transducer are listed in Table 2. Each peak shown in the admittance spectra represents a quasi-harmonic resonance of the two-layer structure (composite resonator). 10 Measured Calculated Correct modelling of chamber characteristics with respect to resonance peak heights and quality factors requires consideration of a glue layer between the piezoceramic and the glass carrier, as can be seen from Fig.5. Modelling without a glue layer results in an over-representation of resonance peaks neighbouring the fundamental at 2.2 MHz (see Fig.4). Introduction of a glue layer reduces peak heights of the neighboured resonances, but not of the 2.2 MHz fundamental, thus leading to excellent agreement with the measured admittance curve (Fig.5). The effect of the glue can be explained as follows: Around the piezoceramic's fundamental at 2.2 MHz, the displacement amplitude of the acoustic wave is a maximum at the ceramic surfaces. (Exactly one half-wavelength fits into the ceramic layer.) Therefore, in this frequency range the glue layer lies in a region of maximum amplitude and vanishing stress (pressure node) and has almost no effect on the resonator. At other resonances the glue layer is not in a pressure node, thus being stressed significantly and because of its low Q-value (Q eff = 2 was assumed for calculations) causing strong resonance degradation. 10 1 Admittance [mhO] Admittance [mhO] f 0.1 0.01 Measured Calculated 1 1 f 2 0.1 0.01 0.001 1 1.5 2 2.5 3 Frequency [MHz] Fig.5. Measured and calculated admittance versus frequency curves of the PZT-ceramic/glass composite transducer with consideration of the bonding (glue) layer. Each resonance of the multi-layer structure is characterized by a total phase shift of the acoustic wave between the outer boundaries equal to a multiple of π. For the fundamental resonance frequency (f = 2.2 MHz) the thickness of layer A corresponds to one half wavelength (λ/2) of the acoustic wave in the piezoceramic and the thickness of layer C corresponds to 2λ/2 in the glass. (The thin glue layer B can be neglected in this consideration.) Thus, at 2.2 MHz the phase shift of the acoustic wave is π between the boundaries of layer A and 2π between the boundaries of layer C. For the empty chamber (layers A and C only) 0-7803-7177-1/01/$10.00 © 2001 IEEE 0.001 1.6 1.8 2 2.2 2.4 2.6 Frequency [MHz] Fig.6. Measured and calculated admittance versus frequency curves of the resonator filled with water (PZT-ceramic/glass/water/glass composite resonator). Fig.6 shows the measured and calculated frequency spectra of the composite resonator filled with water. Two characteristic frequencies are marked: f1 coinciding with the fundamental resonance frequency of the piezoceramic (this coincidence is a consequence of chosen dimensions of layers A and C) and which is electrically strongest pronounced, and f2 , which is electrically much weaker pronounced and nonetheless will be identified as one of the frequencies of optimum filter performance (see below). 2001 IEEE ULTRASONICS SYMPOSIUM — 653 Spatial dependences in axial direction Fig.7 shows the calculated spatial progress of displacement velocity amplitude and stored acoustic energy density, respectively, along the axial direction within the layers A, B, C, D, E of the h-shape separator (see also Fig.3). Displacement velocity amplitude [m/s] Piezoceramic A Glass C Water D Glass E Glue B 0.7 f1 = 2.13 MHz f2 = 1.86 MHz 0.4 Spatial amplitude distribution in lateral direction The very good agreement of the electrical admittance spectra obtained from the one-dimensional model with that obtained from the measurements of the actual composite resonator in Figs.4-6 indicates that couplings to lateral modes are only spuriously affecting the electrical behaviour. Nonetheless, a remaining important question is, how the coupling with these lateral modes influences the lateral displacement amplitude distribution, since the homogenity of the plane standing wave in the water volume is essential for any type of ultrasonic flow-through separator. Therefore the lateral amplitude variations of a two-layer PZT/glass composite sample transducer have been investigated. 0.6 0.5 cies of the piezoceramic (layer A) or the ceramic-glass composite transducer (empty chamber, layers A, B, C), as already shown by Gröschl for ultrasonic filters based on the ultrasonically enhanced sedimentation principle [22]. For the calculations shown in Fig.7, an electric driving power of 4 Wrms has been assumed, according to the experimental conditions applied. 0.3 0.2 0.1 0 Acoustic energy density [J/m3] 120 PZTCeramic 100 f1 = 2.13 MHz f2 = 1.86 MHz 80 a) b) 60 Reference corner 40 Glass 20 0 0 2 4 6 8 10 12 14 Position [mm] Fig.7. Spatial course of acoustic displacement velocity amplitude and acoustic energy density in axial direction of the PZT-ceramic/glass/water/glass composite resonator. The graphs represent the results for the two selected frequencies f1 and f2 (see Fig.6). Surprisingly enough, the maximal displacement velocity amplitudes as well as the maximal energy densities in the water layer do not appear at the electrically strongly pronounced piezoceramic's fundamental resonance f1 but at the electrically much less pronounced composite system's resonance f2. With respect to filtration efficiency, for a given electric energy supplied, maximum stored acoustic energy in the water (liquid) layer is desired. Thus, optimum system performance is achieved at resonance frequencies not coinciding with eigenfrequen- Fig.8. Two-layer PZT/glass composite sample transducer. a) and b) indicate the two view directions for the out-of-plane vibration amplitude distribution measurements. A composite transducer consisting of a 25 mm x 25 mm x 1 mm PZT plate bonded on a 75 mm x 55 mm x 3.41 mm glass plate was chosen as the sample. For the measurement of the out-of-plane amplitude distribution at the PZT-plate and the glass-plate surfaces, a specially developed computer-controlled scanning Laser Doppler Velocimeter (LDV) system has been used [28, 29]. It has been shown by Böhm et al. [30] that the amplitude distribution on the glass side directly determines the lateral amplitude variation in the adjecent liquid volume of an acoustic separator. Fig.9 shows a characteristic amplitude pattern being not suitable and Fig.10 shows one being advantageous for ultrasonic flow-through separators. 654 Reprint Proc. of the 2001 IEEE International Ultrasonics Symposium, a Joint Meeting with the World Congress on Ultrasonics, Atlanta, Georgia, USA, Oct. 7-10, 2001 a) [nm] b) [nm] n=1 Fig.9. Vibration amplitude distribution for excitation of the first quasi-harmonic (n = 1) at frequency f1 = 486 744 Hz. Two-layer PZT/glass composite transducer sample according to Fig.8. a) PZT side, b) glass side surface. The edges of the PZT-plate are indicated by the dotted line, the reference corner is marked in the same way as in Fig.8. At the glass side the scan-area was chosen slightly larger than the area of the PZT-plate. a) [nm] b) [nm] n=2 Fig.10. Vibration amplitude distribution for composite transducer sample as in Fig.8, quasi-harmonic number n = 2, f 2 = 1 164 788 Hz. a) PZT side, b) glass side surface. 0-7803-7177-1/01/$10.00 © 2001 IEEE 2001 IEEE ULTRASONICS SYMPOSIUM — 655 COMPARISON OF TWO SEPARATOR CONCEPTS Ultrasonically enhanced sedimentation Acoustic bio-cell filters sands of hours of continuous operation. Such ultrasonic cell filters are already deliverable in three different sizes with 10 L/day (see Fig.11), 50 L/day, and 250 L/day (see Fig.12) separation capacities, respectively.1 Fig.11 shows a typical ultrasonic cell filter situated on top of the headplate of a 10 liters laboratory size bioreactor. The cell filter can be regarded as a composite resonator consisting of layers of different materials. These "layers" are from left to right a PZT piezo-ceramic plate, a carrier glass, the liquid medium and an acoustic glass reflector. The glass plates function simultaneously as walls of the vessel for the medium. The ultrasonic standing wave axis is in horizontal direction. On the right hand side of Fig.11 a view through a glass window of the active volume (liquid layer) is shown. In the liquid layer the cells are trapped in vertical planes due to the primary acoustic radiation force and, because of gravity forces, slip downwards against the upward medium stream. This ultrasonically enforced sedimentation principle is based on the fact that the surface to volume ratio of the cell agglomerates is drastically reduced compared to that of the homogeneously suspended single particles. As a consequence, for the agglomerates the gravity force exceeds the Stoke's drag force and the cells slip downwards back into the bio-reactor. Although these acoustic cell retention systems work continuously over periods of several months, there is no influence on cell viability by the high frequency ultrasonic field, if the resonator is well designed and the field energy density is kept below the cavitation threshold and below the level where acoustic streaming starts to cause particle disordering [14, 31]. Fig.12. Photograph of the large 250 Liter/day production size bio-cell filter. In Fig.13 the measured separation efficiency of the 250 L/day production size ultrasonic filter is shown for yeast cell suspensions. Yeast cells were used as model particles instead of mammalian cells since yeast is easy to obtain and safely to handle. 100 90 85 80 75 70 65 50 Fig.11. Photograph of a small 10 Liters/day laboratory size biocell filter. In the enlarged window view at the right hand side the agglomeration of the cells in the node planes of the acoustic pressure amplitude of the applied ultrasonic standing wave field can be seen. In contrast to other cell separation techniques, the acoustic radiation force field created within the ultrasonic cell filter constitutes “virtual”, thus non-contact, non-fouling, non-moving filtration means allowing for up to thou- 60 100 150 200 55 250 low rate [L/day] 300 Separation efficiency [%] 95 50 350 0,5 0,9 1,9 5 10 20 ell concentration [g/L] Fig.13. Separation efficiency versus flow rate and cell concentration of the 250 L/d.bio-cell filter of Fig.12 for yeast cell suspensions. The measured separation efficiency SE is defined as the difference between the concentration C s of cells in the suspension (in the reactor volume) and the concentration 656 Reprint Proc. of the 2001 IEEE International Ultrasonics Symposium, a Joint Meeting with the World Congress on Ultrasonics, Atlanta, Georgia, USA, Oct. 7-10, 2001 C h of cells in the upper clarified liquid outlet (harvest stream) over the concentration of cells in the suspension: SE = Cs − Ch C =1− h . Cs Cs (2) Cell concentration was determined by measuring optical turbidity. The turbidity reading was calibrated by a cell counter. Ultrasonic h-shape separator In Fig.14 the scheme of the h-shape resonator is introduced. In contrast to the ultrasonically enhanced sedimentation based separators, the h-shape resonator utilizes the acoustic radiation forces directly for separating the liquid flow lines into the cleaned outlet from the particle traces into the particle enriched outlet. Therefore the direct ultrasonic separation concept of the h-resonator is not relying on gravity. A Separation chamber C B I O1 D O2 E Particle tracking pressure node planes Fig.14. Scheme of the h-shape separator. On the left side the suspension is fed into inlet I of the separator, on the right side the upper outlet O1 is for the separated clean liquid and the lower outlet O2 for the particle enriched suspension. Fig.15. h-Separator performance simulation: a) color coded absolute value of the flow velocity in m/s; b) direction (unit flow velocity vector); c) streamlines of the liquid flow; and d) particle traces. Polystyrene particles a = 20 µm; VI = 3 L/h (liter per hour); VO1 / VO 2 = 1. Flow and force field simulation Calculation of the particle traces was performed in two steps. First the velocity and the streamlines of the fluid was determined by solving the Navier-Stokes equations with the software Package FLUENT® (see Fig.15 a, b, c). Input data are the h-separator geometry (Fig.14) analytically given by the graphic boundary condition defining software GAMBIT®, the material data mass density and viscosity of the fluid, the volume flow rate VI in the inlet I and the output volume flow ratio VO1 / VO2 . In the second step Newton's equation of motion r r r r r r 4πa 3 d2 r FA (r p ) + FSt (r p ) + Fg − Fb = ρ p 2 rp 3 dt (3) for the suspended particles were r rsolved to obtain the partir cle traces r p (t ) , whereby FA (r p ) is the primary acoustic radiation force in axial (vertical) direction given by Eq. (1), r r FSt (r p ) is the Stokes force r r r r FSt (r p ) = −6πηa (v p − v f ) , (4) 0-7803-7177-1/01/$10.00 © 2001 IEEE Fig.16. h-Separator performance simulation: d>) particle traces for overcritical VI = 4 L/h; d) for critical VI = 3 L/h (see also Fig.15); and d<) for undercritical input flow rate VI = 2.5 L/h; Other parameters same as in Fig.15. r η the viscosity of the liquid, a the particle radius, v p and r v f are the velocities of the particle and the fluid, respec- 2001 IEEE ULTRASONICS SYMPOSIUM — 657 r r tively, Fg is the gravity and Fb is the buoyancy force and ρ p the mass density of the particle. The results for the particle traces are given in Fig.15 d and in Fig.16. The cross section of the separation chamber was 10 mm in height and 19 mm in width. h-shape separator sample experiments Fig.17 shows a photograph of the h-shape resonator used in the experiments. critical input flow rate). The chosen concentration of the particles was the limit value where the assumption of the simulation of single particle versus acoustic field interaction was still justified. As further sample suspensions yeast cell cultures of various concentrations have been investigated. The h-separator has also been tested successfully with spirulina platensis algaes suspensions under microgravity conditions in the recent ESA 29th zero g parabolic flight campaign within the frame of the ESA Melissa project [32]. Performance comparison For the comparison of different ultrasonic separator concepts the ratio of volume flow [L/h] over the needed true electrical power [W] is defined as the performance figure of the separator concept under investigation [ ] SP = V / Prms = L / Wh . (5) The presently most used term "separation capacity" in liters per day is, of course, directly dependent on the size of the separator and the electrical power supply, while the reduced quantitiy suggested as performance figure allows a comparison of the concepts itself independent on the upscaling grade. In addition, of course, the meaning of the figure is dependent on the sample (application), especially on its concentration Fig.17. Photograph of the (tilted) h-shape separator. On the left side the input silicone tube for the suspension and on the right side the two outlet silicone tubes for the separated clean water and for the particle enriched suspension,, respectively, can be seen. The black cable on top connects the electrodes of the PZT plate with the drive electronics. Fig.18. Frame picture of the h-separator performance. The output flow ratio VO1 / VO 2 = 1 corresponds to a respective ratio of the mean flow velocities of 2:1, because the cross sectional area of output O1 was chosen half the value of the cross sectional area of output O2. In Fig.18 a typical frame picture of the DV (digital video) movie documentation of the performance of the hshape separator is shown for polystyrene particles with radius a = 20 µm, particle concentration C I = 0.314 %, f = 1.96 MHz, P = 3 Wrms, VO1 / VO2 = 1, VI = 2.75 L/d ( = SUMMARY The most recent progress in the ultrasonic separation technology is based on the availability of powerful new tools described in this paper: The rigorous one-dimensional model of the layered resonator, a highly specialized electrical admittance measurement system, a LDV vibration amplitude distribution measurement system, highly specialized drive electronics with automatic resonance frequency and true input power control, and the mathematical model combining for the first time the hydrodynamic flow and the acoustic field influence on the traces of the suspended particles. For comparing the separation performance of different resonator concepts the acoustic separation performance figure has been introduced. The basic concepts of ultrasonically enhanced sedimentation and the h-shape separator have been analysed and compared. For higher particle concentrations, the ultrasonically enhanced sedimentation is the state of the art bio-cell filter concept. For very dilute suspension, the h-separator exhibits a significantly higher separation performance figure and has been proven to work also under zero g conditions. AKNOWLEDGEMENTS Work supported by the European Commission's TMR Program, Contract No. ERBFMRXCT97-0156, EuroUltraSonoSep, by Applikon Dependable Instruments AB, Schiedam NL, by SonoSep Inc., Canada, and by Anton Paar GmbH, Graz, AT. 658 Reprint Proc. of the 2001 IEEE International Ultrasonics Symposium, a Joint Meeting with the World Congress on Ultrasonics, Atlanta, Georgia, USA, Oct. 7-10, 2001 LITERATURE [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] A. Kundt and O. Lehmann, "Longitudinal vibrations and acoustic figures in cylindrical columns of liquids," Annalen der Physik und Chemie (Poggendorff's Annalen), vol. 153, pp. 1, 1874. L. A. Crum, "Acoustic force on a liquid droplet in an acoustic stationary wave," J. Acoust. Soc. Am., vol. 50, pp. 157163, 1971. R. E. Apfel, "Acoustic Radiation Pressure - Principles and Application to Separation Science," in Fortschritte der Akustik - DAGA '90, vol. Teil A. Bad Honnef, Germany: DPG-GmbH, 1990, pp. 19-36. C. J. Schram, "Manipulation of particles in an acoustic field," in Advances in Sonochemistry, vol. Vol. 2, T. J. Mason, Ed. London: JAI Press Ltd., 1991, pp. 293-322. T. L. Tolt and D. L. Feke, "Separation of dispersed phases from liquids in acoustically driven chambers," Chem. Eng. Sci., vol. 48, pp. 527-540, 1993. Z. Mandralis, W. Bolek, W. Burger, E. Benes, and D. L. Feke, "Enhanced synchronized ultrasonic and flow-field fractionation of suspensions," Ultrasonics, vol. 32, pp. 113121, 1994. O. Doblhoff-Dier, T. Gaida, H. Katinger, W. Burger, M. Gröschl, and E. Benes, "A novel ultrasonic resonance field device for the retention of animal cells," Biotechnol. Prog., vol. 10, pp. 428-432, 1994. W. T. Coakley, G. Whitworth, M. A. Grundy, R. K. Gould, and R. Allman, "Ultrasonic manipulation of particles and cells," Bioseparation, vol. 4, pp. 73-83, 1994. P. W. S. Pui, F. Trampler, S. A. Sonderhoff, M. Gröschl, D. G. Kilburn, and J. M. Piret, "Batch and semicontinuous aggregation and sedimentation of hybridoma cells by acoustic resonance fields," Biotechnol. Prog., vol. 11, pp. 146-152, 1995. K. Yasuda, S. Umemura, and K. Takeda, "Concentration and fractionation of small particles in liquid by ultrasound," Jpn. J. Appl. Phys. (Part 1), vol. 34, pp. 2715-2720, 1995. M. Gröschl, "Ultrasonic separation of suspended particles Part II: Design and operation of separation devices," Acustica - acta acustica, vol. 84, pp. 632-642, 1998. C. M. Cousins, P. Holownia, J. J. Hawkes, C. P. Price, P. Keay, and W. T. Coakley, "Clarification of Plasma from Whole Human Blood using Ultrasound," Ultrasonics, vol. 38, pp. 654-656, 2000. E. Riera-Franco de Sarabia, J. A. Gallego-Juárez, G. Rodríguez-Corral, L. Elvira-Segura, and I. GonzálezGómez, "Application of high-power ultrasound to enhance fluid/solid particle separation processes," Ultrasonics, vol. 38, pp. 642-646, 2000. S. Radel, A. J. McLoughlin, L. Gherardini, O. DoblhoffDier, and E. Benes, "Viability of yeast cells in well controlled propagating and standing ultrasonic plane waves," Ultrasonics, vol. 38, pp. 633-637, 2000. J. J. Hawkes and W. T. Coakley, "Force field particle filter, combining ultrasound standing waves and laminar flow," Sensors and Actuators B, vol. 75, pp. 213-222, 2001. T. Gaida, O. Doblhoff-Dier, K. Strutzenberger, H. Katinger, W. Burger, M. Gröschl, B. Handl, and E. Benes, "Selective Retention of Viable Cells in Ultrasonic Resonance Field Devices," Biotechnol. Prog., vol. 12, pp. 73-76, 1996. F. Trampler, S. A. Sonderhoff, P. W. S. Pui, D. G. Kilburn, and J. M. Piret, "Acoustic cell filter for high density perfusion culture of hybridoma cells," Bio/Technology, vol. 12, pp. 281-284, 1994. 0-7803-7177-1/01/$10.00 © 2001 IEEE [18] M. Maitz, F. Trampler, M. Gröschl, A. da Câmara Machado, and M. Laimer da Câmara Machado, "Use of an ultrasound cell retention system for the size fractionation of somatic embryos of woody species," Plant Cell Reports, vol. 19, pp. 1057-1063, 2000. [19] L. P. Gor'kov, "On the forces acting on a small particle in an acoustical field in an ideal fluid," Sov. Phys. Dokl., vol. 6, pp. 773-775, 1962. [20] S. M. Woodside, J. M. Piret, M. Gröschl, E. Benes, and B. D. Bowen, "Acoustic force distribution in resonators for ultrasonic particle separation," AIChE Journal, vol. 44, pp. 1976-1984, 1998. [21] V. F. K. Bjerknes, Die Kraftfelder. Braunschweig, Germany: Vieweg und Sohn, 1909. [22] M. Gröschl, "Ultrasonic separation of suspended particles Part I: Fundamentals," Acustica - acta acustica, vol. 84, pp. 432-447, 1998. [23] H. Nowotny and E. Benes, "General one-dimensional treatment of the layered piezoelectric resonator with two electrodes," J. Acoust. Soc. Am., vol. 82, pp. 513-521, 1987. [24] H. Nowotny, E. Benes, and M. Schmid, "Layered piezoelectric resonators with an arbitrary number of electrodes (general one-dimensional treatment)," J. Acoust. Soc. Am., vol. 90, pp. 1238-1245, 1991. [25] B. A. Auld, Acoustic fields and waves in solids, vol. I. New York: John Wiley & Sons, 1973. [26] R. Holland, "Representation of dielectric, elastic, and piezoelectric losses by complex coefficients," IEEE Trans. Sonics Ultrason., vol. 14, pp. 18-20, 1967. [27] M. Schmid, E. Benes, and R. Sedlaczek, "A computer-controlled system for the measurement of complete admittance spectra of piezoelectric resonators," Meas. Sci. Technol., vol. 1, pp. 970-975, 1990. [28] E. Benes, M. Schmid, and V. Kravchenko, "Vibration modes of mass-loaded planoconvex quartz crystal resonators," J. Acoust. Soc. Am., vol. 90, pp. 700-706, 1991. [29] E. Benes, A. Frank, H. Böhm, M. Gröschl, F. Trampler, W. Burger, and H. Nowotny, "Vibration amplitude distribution measurements on piezoelectric PZT/glass composite transducers for ultrasonic bio-cell filters," presented at ISAF 2000, 12th IEEE Int. Symp. on Applications of Ferroelectrics, Honolulu, Hawaii, USA, 2001. [30] H. Böhm, F. Trampler, L. G. Briarty, K. C. Lowe, J. B. Power, E. Benes, and M. R. Davey, "Lateral displacement amplitude distribution of water-filled ultrasonic bio-separation resonators with laser-interferometry and thermochromatic foils," presented at ISAF 2000, 12th IEEE Int. Symp. on Applications of Ferroelectrics, Honolulu, Hawaii, USA, 2001. [31] M. Gröschl, W. Burger, B. Handl, O. Doblhoff-Dier, T. Gaida, and C. Schmatz, "Ultrasonic separation of suspended particles - Part III: Application in biotechnology," Acustica - acta acustica, vol. 84, pp. 815822, 1998. [32] C. Lasseur and I. Fedele, "MELISSA Final Report for 1999," ESA ESA/EWP-2092, 2000. 1 Applikon Dependable Instruments BV, De Brauwweg 13, NL-3125 AE Schiedam, The Netherlands For further recent application literature and reports see: http://www.iap.tuwien.ac.at/www/euss/ 2001 IEEE ULTRASONICS SYMPOSIUM — 659