1

Enhanced creaming of milk fat globules in milk emulsions by the application of 1 ultrasound and detection by means of optical methods 2 3 P. Juliano1, * , A. Kutter1,2, L.J. Cheng1, P. Swiergon1, R. Mawson1, and M. A. 4

Augustin1 5

1CSIRO Food and Nutritional Sciences, 671 Sneydes Road, Werribee VIC, 3030, Australia 6

2Institute of Fluid Mechanics and Erlangen Graduate School in Advanced Optical Technologies, 7

University of Erlangen-Nuremberg, Cauerstrasse 4, 91058 Erlangen, Germany 8

Abstract 9

The effects of application of ultrasonic waves to recombined milk emulsions (3.5% fat, 10

7% total solids) and raw milk on fat destabilization and creaming were examined. 11

Coarse and fine recombined emulsions (D[4,3] = 2.7 µm and 9.3 µm) and raw milk 12

(D[4,3] = 4.9 µm) were subject to ultrasound for 5 min at 35°C and 400 kHz or 13

1.6 MHz (using a single transducer) or 400kHz (where the emulsion was sandwiched 14

between two transducers). Creaming, as confirmed by Turbiscan measurements, was 15

more evident in the coarse recombined emulsion and raw milk compared to that of the 16

recombined fine emulsion. Particle size analysis, supported by microscopic images, 17

showed that both flocculation and particle coalescence occurred upon ultrasound 18

treatment, the extent of which was dependent on the milk sample and ultrasound 19

conditions. These results imply that the ultrasound has potential to pre-dispose fat 20

particles in milk emulsions to creaming due to flocculation and coalescence, not only 21

in standing wave systems, but in systems with inhomogeneous sound distributions. 22

23

Keywords 24

Ultrasound, separation, milk, standing waves, milk fat globule, emulsion, coalescence 25

* Corresponding author. Tel.: +61 397313276; fax: + 397313201. E-mail address: pablo.juliano@csiro.au (P. Juliano).

2

1. Introduction 26

Ultrasonics has been proposed as a technique to separate particles under a standing 27

wave field. A number of publications concentrate on the removal of solid particles 28

(mainly polystyrene and polyamide beads) in batch (Tolt and Feke, 1993; Whitworth 29

et al., 1991) and continuous processes at laboratory scale (Johnson and Feke, 1995; 30

Pangu and Feke, 2004; Hawkes and Coakley, 2001; Mandralis and Feke, 1993; 31

Kapishnikov et al., 2006; Groeschl, 1998a; Groeschl, 1998b). In the food area, 32

ultrasound has been used to split canola oil emulsions by enhancing particle 33

flocculation and coalescence (Nii et al., 2009). 34

Particles in a standing pressure wave field experience the so called primary acoustic 35

radiation force, which can be expressed analytically for a compressible sphere 36

(Yosioka and Kawasima, 1955): 37

( )34sin 2

3ac acF R k E kxπ= − Φ (1) 38

where R denotes the particle radius, k the wave number of the ultrasound (=2π/λ, with 39

λ being the wavelength), Eac the average acoustic energy density, Φ the acoustic 40

contrast factor and x the distance from a nodal point of the standing wave. When 41

particles with a non-zero acoustic contrast factor (ACF) are subjected to a standing 42

ultrasonic wave the acoustic radiation force drives them either to the nodes or 43

antinodes for particles with positive and negative ACF values respectively. The 44

spatial distance between successive bands is therefore λ\2. The acoustic contrast 45

factor is given by 46

5 2

2M P M

M P P

ρ ρ βρ ρ β

−Φ = −+ (2) 47

3

where ρM, ρP and βM, βP are the density and the compressibility of the continuous 48

medium and the dispersed particles respectively. 49

In a pressure field, the container or chamber walls act as pressure antinodes as they 50

are the points where the pressure differences are built up. As can be derived from Eq, 51

2, fat particles in an aqueous environment show a negative contrast factor and 52

therefore have the tendency to move towards the walls. In addition, Eq. 2 shows that 53

even particles with zero density difference with respect to their surrounding medium 54

can be manipulated by ultrasound, as long as their compressibilities differ. After the 55

particles form bands of high concentration at the pressure nodes or antinodes the so 56

called “second acoustic force” or Bjerknes force (Apfel, 1988) drives them together, 57

leading to clusters. The second acoustic force originates from interactions of the 58

particles with the fluid, e.g. reflections of the ultrasound field at the particles’ surface. 59

A review dealing with the effects of ultrasound on proteins in dairy products was 60

recently published by Ashokkumar et al. (2010). However, the information on the use 61

of ultrasound for separation of fat in milk is scarce. A few references dealing with the 62

separation of milk fat globules are found in the field of biotechnology where fat 63

particles are separated from red blood cells using standing waves (Petersson et al., 64

2004; Petersson et al., 2005). Miles et al. (1995) determined the threshold levels for 65

the separation of micro-organisms from suspensions followed in a cuvette resonance 66

chamber set up and further suggests that the application of ultrasound standing waves 67

(1-3 MHz) can enhance not only bacteria but also cream separation. However, a 68

systematic evaluation of the effects of ultrasound standing waves on the creaming 69

properties of milk has not been done to date. The present contribution evaluates the 70

formation of milk fat globule banding, flocculation and coalescence, in coarse and 71

fine emulsions of recombined milks and in raw milk. 72

4

2. Materials and methods 73

Sample preparation 74

To study the effect of ultrasound on the creaming of milk fat globules three model 75

systems were used. These comprised of two models made out of recombined milk 76

emulsions (a coarse emulsion and a fine emulsion) and freshly collected raw milk. 77

The recombined milk emulsion was prepared by dispersion of skim milk powder 78

(SMP, Tatura Milk Industries Limited, Victoria, Australia) in water at 50ºC followed 79

by the addition of anhydrous milk fat (AMF) (Murray Goulburn Co-operative 80

Company Limited, Victoria Australia) to obtain 7% total solids emulsion with 3.5% 81

fat. The AMF was previously dyed by adding 0.01% oil-red-O colorant (BDH 82

Laboratory suppliers, Poole, England). An Ultra-turrax operated at 20,000 rpm for 83

3 min (ultra-turrax T25, IKA-Labortechnik, Staufen, Germany,) was utilised to form a 84

coarse recombined milk emulsion (hereafter termed “coarse emulsion”). To reduce 85

the particle size, the coarse emulsion was treated with a laboratory homogenizer after 86

two passes (FOSS NIRSystems Inc., Laurel, USA, Type D head, equivalent to a 87

treatment of approximately 140 bar). This homogenized emulsion of recombined milk 88

emulsion was hereafter termed a “fine emulsion”. All trials including raw milk used 89

fresh milk directly obtained from Bega Cheese (Victoria, Australia) on the day of the 90

experiment. 91

92

Resonance chamber configuration and sample treatment 93

The sonication of the samples (7 mL) was carried out in a Turbiscan tube 94

(Formulaction, Toulouse, France) which was sealed with a metal plate at the bottom, 95

acting as a sound transmitter into the fluid. Fig. 1 and Fig. 2 illustrate the 96

experimental set up for the one transducer plate (1) and two transducer plate (2) 97

5

configuration. Ultrasound frequencies used were 400 kHz and 1.6 MHz with the 98

single transducer set-up and only at 400 kHz with the two transducer plate set-up. 99

The sample was placed in a water bath with 35°C in order to minimize the 100

temperature increase due to the energy dissipation. In Fig. 2 the tube walls acted as a 101

sound transmitter as the set up had to be submerged in the water bath. Therefore, the 102

sound distribution in the tube in this case is complex and not predictable, as the glass 103

walls transmit and interfere with the external sound field of the transducers. In all 104

experiments the ultrasound treatment time was set to 5 min. All ultrasonication runs 105

were carried out as triplicates. 106

Two different ultrasound transducer systems were used with 400 kHz (Submersible 107

Transducers, Sonosys Ultraschallsysteme GmbH, Neuenbuerg, Germany) and 108

1.6 MHz (Nebulizer, APC International Inc., Mackayville, Pennsylvania, USA). To 109

estimate the power input into the liquid a slightly modified set up compared to Fig. 1 110

served as model system. A defined mass of water (10 g ± 0.1 g) was filled into the 111

tube, which was not submerged but in contact to air (metal cap still in the water bath). 112

By measuring the temperature increase after a defined period of time (120 s) the 113

power input into the liquid could be estimated. For transferring these results to the 114

dairy system this approach assumes that the differences in the physical properties of 115

water and milk are negligible or do not influence the power input. It is also assumed 116

that the heat conduction into the ambient through the glass walls is negligible 117

compared to the energy input through the dissipation of sound energy. The power 118

input values could be determined to be 1.6 W ± 0.2 W for the 400 kHz and 119

0.35 W ± 0.07 W for the 1.6 MHz unit, respectively. No acoustic streaming was 120

present at all times, which would lead to a remixing of the particles in the dairy 121

systems due to turbulent structures. 122

6

The sound intensity in the tube was estimated by using a hydrophone (TC 4038, 123

Reson A/S, Slangerup, Denmark). It was inserted into the set up at different positions 124

to get an approximation of the different sound levels throughout the tube. For the 125

described process parameters this sound pressure amplitude ranged around 350 to 126

500 kPa, which corresponds to an average acoustic energy density Eac of 10 to 25 J/m3. 127

128

Turbiscan measurements 129

In order to get an estimation of the creaming in the sample a Turbiscan MA 2000 130

(Formulaction, Toulouse, France) was employed, which scans the whole sample in the 131

tube. The Turbiscan measures backscattered and transmitted light as a function of the 132

axial tube coordinate and time. Due to the increase of backscattered light at the layers 133

with a higher number of dispersed particles, it is possible to measure high cream 134

concentration regions. 135

Scans were performed over a 10 minute period right after removing the samples from 136

the ultrasonic set up. For each model system a non-sonicated control was treated 137

exactly in the same way as the sonicated samples. They were filled into the Turbiscan 138

tube, put into the water bath and after 5 min they were placed into the Turbiscan and 139

measured for further 10 min. 140

141

Creaming and recovery rates 142

To compare the different creaming extents the ratio of the area of the peak 143

corresponding to the separated cream phase was calculated with respect to the total 144

backscattering measured in the sample (Eq. 3). Fig. 3 illustrates the computation of 145

the “creaming extent” (CE). The cream phase was defined as the curve area, starting 146

where the backscattering value is higher than 1.5 % in comparison to the baseline 147

7

value representing the bulk liquid. The 1.5% value represents the measurement error 148

of backscattering determined for the Turbiscan equipment. For comparison between 149

different measurements (e.g. initial particle concentration) the area under the cream 150

phase curve was related to the area under the whole curve. Analytically this can be 151

expressed as 152

( )

( )

creamphase0

0

creaming extent =

L

L

BS z dz

BS z dz

∫

∫ (3) 153

where z represents the axial position in the tube (origin at the bottom), L the free 154

surface and BS(z) the backscattering at position z. 155

This approach assumes that the “backscattering potential” and hence number of 156

particles, which equals concentration, is shifted when a cream phase is formed. To 157

assure that this procedure yields valid results, the so called “recovery rate” (RR) was 158

computed with the following equation 159

( )

( )

sample0

control0

recovery rate=

L

L

BS z dz

BS z dz

∫

∫ (4) 160

where BS(z)sample stands for the backscattering of the sample and BS(z)control for the 161

corresponding non-ultrasonic treated control at the same time. If this value is close to 162

unity this means that the overall backscattering of the sample and the control match. 163

For instance, assuming no change in particle sizes, a recovery rate of one can be 164

attributed to the fact that all particles are moved to another position in the tube but do 165

not flocculate or coalesce. It also guarantees that no significant non-linear effects are 166

8

introduced due to high particle concentrations and the backscattering is directly 167

proportional to the square root of the concentration. 168

169

Particle sizing and microscopy 170

All particle size measurements were conducted with a Galai CIS-1 (Galai Production 171

LTD, Haemek, Israel), which provides particle size ranges allowing the detection of 172

changes in size. All measurements were repeated twice per sample treatment. Samples 173

were adjusted to a detectable dilution range with distilled water. To differentiate 174

between flocculation (which is a reversible phenomenon in which the integrity of the 175

interfacial layer of the initial fat droplet is maintained) and coalescence (where there 176

disruption of the interfacial layer and the formation of a larger oil droplet) the particle 177

size increase for the same set of described experiments was conducted a second time. 178

After sonication the cream phase was remixed by shaking the Turbiscan tube gently to 179

break up particle clusters and determine if particle coalescence occurred. After the 180

Turbiscan measurements microscopy images (400x lens,Olympus BH-2 light 181

microscope, soft imaging system Color View IIIu, Japan) and particle size 182

distributions of the top cream layer (1 ml extracted from the surface with a pipette) 183

were recorded to observe potential changes in the globule structure and concentration. 184

185

Statistics 186

Changes in particle size were evaluated by using the General Linear Model in the 187

Minitab Statistical package to a 95% level of significance (Minitab ® Release 14.11, 188

2003, Coventry, UK). The Tukey method was applied at 95% confidence level for 189

pairwise comparisons. 190

9

3. Results and Discussion 191

192 3.1 Particle size of initial emulsions 193 194 195

Fig. 4 shows the number distribution of all samples, where a wider distribution can be 196

found in the case of raw milk. The calculated volume mean diameters D[4,3] for the 197

samples were 4.9 µm for raw milk, 2.7 µm for the fine emulsion and 9.3 µm for the 198

coarse emulsion. 199

200 3.2 Effect of ultrasonication 201 202

3.2.1 Creaming behavior 203

The creaming behavior expressed as the creaming extent from Turbiscan readings as 204

well as particle size results are presented after sonication runs. The extent of creaming 205

was dependent on the milk sample used, the frequency applied and the transducer 206

geometry. Differences in creaming were obtained with the use of a single transducer 207

at the two frequencies (400kHz or 1.6 MHz) and also at fixed frequency (400 KHz) 208

when different transducer set-ups were used (single or two transducer geometry). 209

210

3.2.1.1. Single transducer configuration at various frequencies 211

By applying one transducer and a reflector (Fig. 1) and adjusting the distance between 212

them a planar standing wave could be achieved. Under these conditions the force on a 213

particle is maximized. This effect is shown as an example in Fig. 5 for the coarse 214

emulsion of recombined milk. It can be seen that the ultrasound leads to the formation 215

of bands with higher fat concentration. After switching off the ultrasound field 216

floccules start rising immediately and re-disperse at the same time. If the sonicated 217

sample is stirred all floccules completely re-disperse back into the solution. This re-218

10

dispersing suggests that the fat globules clustered in bands made up of discrete single 219

droplets. Miles et al. (1995) have identified the threshold amplitudes to form bands in 220

suspensions of latex microspheres at particle size ranges between 1 and 5 µm at 1 221

MHz and higher frequencies. The minimum average sound pressure determined for 222

our systems was 350 kPa, which exceeds the threshold values reported for banding of 223

all particles sizes tested at 1 MHz. 224

The enhanced creaming extent caused by ultrasonics in the coarse emulsion, in 225

comparison to the untreated control, could be observed with the naked eye. The top 226

layer exhibited a more intense red color which is due to the higher concentration of fat 227

particles. In these experiments with the Turbiscan tubes, the Turbiscan readings (Fig. 228

6) reveal that hardly any natural creaming could be observed for the coarse emulsion 229

of recombined milk without ultrasound after 10 min. The results for the single 230

transducer for the two frequencies are summarized in Table 1 and Table 2. Comparing 231

the results for the least stable coarse emulsion at 400 kHz and 1.6 MHz it becomes 232

obvious that the higher frequency leads to a two-fold increase in creaming extent. The 233

increase with higher frequency was also observed with the raw milk. In this case a 234

weak level of creaming could be measured at 400 kHz after 10 min, in contrast to 235

1.6 MHz. The more prominent separation observed in the coarse emulsion could be 236

due to the presence of larger particles in the 9 to 15 µm range, which were not present 237

in milk. 238

The fine emulsion showed no creaming, regardless of the frequency and settling time 239

during the Turbiscan trials. The smaller particle sizes of the fine emulsion compared 240

to the coarse emulsion and the raw milk contributed to the resistance for the fine 241

emulsion to creaming. Table 1 and Table 2 reveal that all the observed separations 242

occurred during the sonication for the coarse emulsion at 400 KHz and 1.6 MHz and 243

11

for the raw milk at 1.6 MHz. Thus, these experiments clearly reveal that ultrasound 244

can be used to accumulate and separate milk fat in its native environment. 245

246

3.1.2. Particle size changes 247

During particle flocculation the probability of coalescence rises due to the increased 248

local concentration. Both processes, pure flocculation and coalescence, differ from the 249

point of view of their physics. Both phenomena result in an increase in particle size. 250

In the case of flocculation where two or more fat droplets aggregate while each 251

maintaining their interfacial layer, there is an apparent hydrodynamic increase. In 252

flocculation, the clustered fat globules may be re-dispersed. In the case of coalescence, 253

there is disruption of the original interfacial layer of droplets, leading to an 254

irreversible increase in particle size of the oil droplet. Both flocculation and 255

coalescence result in enhanced creaming. This has been shown in previous studies 256

which mention these mechanisms operating for the ultrasonic enhanced separation of 257

algae and vegetable oil from the continuous medium (Bosma et al., 2003; Nii et al., 258

2009). 259

Fig. 7 shows that no significant particle size (P>0.05) increase was detected for the 260

coarse emulsion in the 400 KHz single transducer set up. The same was found for raw 261

milk and the fine emulsion for both single transducer geometries at 400 kHz and 1.6 262

MHz. However, the coarse emulsion showed that the number of particle sizes in the 2-263

3 µm range increased when using the 1.6 MHz unit. 264

These observations imply that the primary and secondary acoustic forces drive the 265

particles together so they form floccules or clusters, which redisperse once ultrasonic 266

radiation is turned off. This was verified by remixing the fat phase and finding no 267

increased particle size. Extending the sonication time from 5 min to 30 min (data not 268

12

shown) did not result in further coalescence in all samples for the single transducers at 269

400 kHz and 1.6 MHz. 270

271

3.2. Two-transducer configuration at 400 KHz 272

3.2.1. Creaming behavior 273

Major differences were found when comparing the coarse emulsion after sonication in 274

the one and two-transducer set ups, which were clearly observable with the naked eye. 275

Firstly, there were large clusters forming at the tube walls in the two-transducer 276

resonance chamber (Fig. 8). The clusters formed in the two transducer geometry were 277

visually larger in size than those obtained in the one transducer geometry forming a 278

band. After switching off the ultrasound the clusters destabilized, rising to the top of 279

the tube, and redispersed; but not as quickly as in the case of the single transducer. 280

The destabilized fat clusters obtained upon sonication in the two transducer geometry 281

could still be measured minutes later with the Turbiscan (which was not the case in 282

the single transducer geometry). Where sonication was applied with two transducers, 283

the fat particles rose to the top forming a more distinct cream layer, which means that 284

more particles flocculated or coalesced, and as a consequence more fat was 285

accumulated. This effect can be noticed in Fig. 9 for the coarse emulsion and in Fig. 286

10 for the raw milk. Both the coarse emulsion and raw milk samples showed higher 287

creaming extents in contrast to the fine emulsion where no creaming was observed 288

(see Table 3). 289

In addition, the treatment with the two transducer configuration results in very high 290

concentrations of milk fat in the top (i.e., creamed) layer of raw milk after sonication 291

(Fig. 11d and Fig. 12d). These images illustrate that sonication with 1.6 MHz (Fig. 292

11b and Fig. 12b) drives bigger particles to the top right after ultrasonication, which 293

13

could also be confirmed with particle size analysis (data not shown). It is worth noting 294

that the raw milk was not influenced by the 400 kHz sonication in the single as it was 295

in the two transducer set up. The microscopy images in Fig. 11 and Fig. 12 show 296

excellent agreement with the creaming extents in tables 1 to 3. 297

Although the sound distribution in the tube is not predictable with the two transducer 298

geometry, the sound intensity seems to be higher and the particles are driven together 299

with a bigger force, leading to these characteristic floccules which tend to cluster at 300

the tube walls. The fact that these floccules rose slower than in the single transducer 301

geometry (Fig. 9 and Fig. 10) implies that the fat particles clustered more strongly to 302

the walls due to the higher intensity of the sound field. These higher forces may lead 303

to changes in the particle size, and possibly interfacial membranes, which is addressed 304

in the next section. 305

The recovery rate can be used to demonstrate the ability of the clustered fat droplets to 306

re-disperse after treatment in the two transducer set up at 400 kHz. To account for the 307

peaks appearing in the Turbiscan readings as a result of the large floccules distributed 308

along the tube walls in the two transducer set up (Fig. 9 and Fig. 10), the recovery rate 309

(Eq. 4) is calculated by using the integral of the curve obtained after 10 min in the 310

denominator. Instead of using the reading for the control sample, the 10 min reading 311

is used as a baseline for comparison between recovery rates. The single peak 10 min 312

reading mainly occurs when all floccules have risen to the top of the tube and 313

therefore allows the evaluation of the extent of re-dispersion of large floccules in 314

relative to this point in time. 315

Thus, all recovery rates measured through time in the tube show values higher than 316

one as a result of the additional peaks encountered by the Turbiscan optic sensor. 317

While the formed clusters disassembled into floccules or particulates rising towards 318

14

the top cream layer the recovery rates decrease with time (Fig. 13), which indicated 319

that some globules forming the clusters partially re-dispersed. 320

321

3.2.2. Particle size changes 322

A comparison of particle sizes of the unsonicated milk (control) with milk sonicated 323

in the two transducer set up shows no significant increase in particle sizes in raw milk 324

and in the fine emulsion. For the coarse emulsion, however, a significant increase can 325

be observed for particles in the size range greater than 6 µm (P>0.05), which is 326

demonstrated by Fig. 7. This means that the ultrasonic forces in the coarse emulsion 327

were enough as to promote fat destabilization probably by means of particle 328

coalescence in addition to flocculation. 329

On the other hand, even though flocculation was achieved, the raw milk matrix has 330

shown to require greater forces to achieve particle coalescence. This may be due to 331

differences in the interfacial properties of the different fat globules and the presence 332

of larger sized droplets in the 9 to 15 µm range in the coarse emulsion, which were 333

not present in milk. In addition, in fresh raw milk, the fat droplet is surrounded by the 334

natural milk fat globule membrane (MFGM), which is rich in phospholipids, whereas 335

in recombined emulsions, proteins form the interfacial layer. 336

Only a few literature references deal with the phenomenon of coalescence in standing 337

ultrasound fields, as solid particles are often used as dispersed phase, which usually 338

do not exhibit coalescence. In particular some work has shown instantaneous 339

coalescence of particles in vegetable oil emulsions in standing ultrasonic wave fields 340

(Pangu and Feke, 2007; Pangu and Feke, 2009). 341

342

15

4. Conclusion 343

This contribution shows for the first time that the application of ultrasound at 344

frequencies lower than 1 MHz leads to enhanced creaming in a coarse recombined 345

milk emulsion formulated with skim milk powder and milk fat and in raw milk. This 346

was supported by microscopy imaging as well as Turbiscan readings, which showed 347

higher particle concentration in the top cream layer as a result of ultrasound. Particle 348

flocculation and clustering was detected in both the coarse emulsion and raw milk, 349

where in the coarse emulsion particle coalescence was observed in the two transducer 350

system at 400 kHz. However, no coalescence was observed in milk in the two 351

transducer set up. These findings suggest that a non standing wave system with two 352

transducers can enhance particle coalescence. 353

Separation of milk fat in dairy streams can benefit from the increased coalescence 354

phenomena by pre-treating milk with high frequency ultrasound. Further research is 355

required to determine the kinetics of flocculation, coalescence and creaming of milk 356

fat globules in high frequency ultrasonic systems. Parameters from kinetic models 357

will enable the design of ultrasonic based unit operations applicable to the dairy 358

industry. 359

360

Acknowledgements 361

The authors gratefully acknowledge the support by the Erlangen Graduate School in 362

Advanced Optical Technologies (SAOT) by the German National Science Foundation 363

(DFG) in the framework of the excellence initiative. The authors would also like to 364

acknowledge Dr. Kai Knoerzer and Ms. Elodie Cotte for their contribution in this 365

project. 366

367

16

References 368

Apfel, R.E. 1988. Acoustically induced square law forces and some speculations 369 about gravitation. American Journal of Physics 56:726-29. 370

Ashokkumar, M., Bhaskaracharya, R., Kentish, S., Lee, J., Palmer, M., and Zisu, B. 371 2010. The ultrasonic processing of dairy products - An overview. Dairy 372 Sci.Technol. 90:147-68. 373

Bosma, R., van Spronsen, W.A., Tramper, J., and Wijffels, R.H. 2003. Ultrasound, a 374 new separation technique to harvest microalgae. Journal of Applied Phycology 375 15:143-53. 376

Groeschl, M. 1998a. Ultrasonic Separation of Suspended Particles - Part I: 377 Fundamentals. Acta Acustica united with Acustica 84:432-47. 378

Groeschl, M. 1998b. Ultrasonic Separation of Suspended Particles - Part II: Design 379 and Operation of Separation Devices. Acta Acustica united with Acustica 380 84:632-42. 381

Hawkes, J.J. and Coakley, W.T. 2001. Force field particle filter, combining ultrasound 382 standing waves and laminar flow. Sensors and Actuators B-Chemical 75:213-383 22. 384

Johnson, D.A. and Feke, D.L. 1995. Methodology for fractionating suspended 385 particles using ultrasonic standing wave and divided flow fields. Separations 386 Technology 5:251-58. 387

Kapishnikov, S., Kantsler, V., and Steinberg V. 2006. Continuous particle size 388 separation and size sorting using ultrasound in a microchannel. Journal of 389 Statistical Mechanics: Theory and Experiment 2-15. 390

Mandralis, Z.I. and Feke, D.L. 1993. Continuous suspension fractionation using 391 acoustic and divided-flow fields. Chemical Engineering Science 48:3897-905. 392

Miles, C.A., Morley, M.J., Hudson, W.R., and Mackey, B.M. 1995. Principles of 393 separating micro-organisms from suspensions using ultrasound. Journal of 394 Applied Bacteriology 78:47-54. 395

Nii, S., Kikumoto, S., and Tokuyama, H. 2009. Quantitative approach to ultrasonic 396 emulsion separation. Ultrasonics Sonochemistry 16:145-49. 397

Pangu, G.D. and Feke, D.L. 2004. Acoustically aided separation of oil droplets from 398 aqueous emulsions. Chemical Engineering Science 59:3183-93. 399

Pangu, G.D. and Feke, D.L. 2007. Droplet transport and coalescence kinetics in 400 emulsions subjected to acoustic fields. Ultrasonics 46:289-302. 401

Pangu, G.D. and Feke, D.L. 2009. Kinetics of ultrasonically induced coalescence 402 within oil/water emulsions: Modeling and experimental studies. Chemical 403 Engineering Science 64:1445-54. 404

17

Petersson, F., Nilsson, A., Holm, C., Jonsson, H., and Laurell, T. 2005. Continuous 405 separation of lipid particles from erythrocytes by means of laminar flow and 406 acoustic standing wave forces. Lab on A Chip 5:20-22. 407

408

Tolt, T.L. and Feke, D.L. 1993. Separation of dispersed phases from liquids in 409 acoustically driven chambers. Chemical Engineering Science 48:527-40. 410

Whitworth, G., Grundy, M.A., and Coakley, W.T. 1991. Transport and Harvesting of 411 Suspended Particles Using Modulated Ultrasound. Ultrasonics 29:439-44. 412

Yosioka, K. and Kawasima, Y. 1955. Acoustic Radiation Pressure On A 413 Compressible Sphere. Acustica 5:167-73. 414

415 416

417

18

Figures 418 419

420 Fig. 1. Experimental set up for the one transducer geometry (400 kHz and 1.6 MHz) 421 422 423

424

Fig. 2. Experimental set up for the two transducer geometry. In this set up sound is 425 also transferred into the fluid through the tube walls (only 400 kHz) 426 427

19

428

429

Fig. 3. Example for the computation of the creaming extent. This figure shows the 430

backscattering of the coarse emulsion control (no ultrasonic treatment) in comparison 431

with the sonicated sample (measurement 1 min after switching of ultrasound). The 432

total area under the graphs is equal for all times 433

434

20

435

436

0

10

20

30

40

50

60

70

80

90

100

0-1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 8-9 9-10 10-20 20-30

particle size [µm]

cum

ulat

ive

num

ber

dist

ribut

ion

[%]

raw milkcoarse emulsion fine emulsion

437

Fig. 4. Cumulative number distribution for the raw milk, coarse emulsion and fine 438 emulsion. The coarse emulsion shows more similarities regarding the particle sizes 439 and can therefore be used as a milk model, with the advantage of using dye to 440 visualize the particle concentration 441

442

21

443

444

Fig. 5. a) Dyed coarse emulsion under sonication (400 kHz) in an open vessel (no 445

reflector). The distinct bands can clearly be observed. b) After switching of the 446

ultrasound field the formed floccules start to rise immediately. The formation of the 447

streaks leads to the conclusion that the floccules comprise of single particles 448

22

449

Fig. 6. Example of the Turbiscan reading for the control samples. Here the results for 450

the coarse emulsion are used, as it represents the most unstable of all three dispersions. 451

In neither of the samples any natural creaming could be observed. Only the range 452

between 8 mm and 63 mm where the backscattering shows the constant value of 53 % 453

can be considered to be inside the tube. The region below 8 mm marks the metal base 454

and the strong decay of backscattering above 63 mm the beginning of the free surface 455

of the sample. This holds true for all following Turbiscan readings 456

457

23

458

Fig. 7. Particle size distribution after the sonication of the coarse emulsion at 400 kHz 459

(one and two transducer geometry) and 1.6 MHz at 35°C and gentle remixing of the 460

cream and the “skimmed” phase. Different letters represent significant differences per 461

treatment (P<0.05) 462

463

464

24

465

466

Fig. 8. Coarse emulsion after sonication in the sandwich two transducer geometry. 467

Large particle clusters were formed at the tube wall showing a greater lag time for re-468

dispersion than bands observed in a single transducer set up. 469

470

25

471

Fig. 9. Turbiscan reading after the sonication of the coarse emulsion in the one and 472

two transducer geometry. For the two transducers the formed clusters persist longer 473

than for the single transducer 474

475

26

476

Fig. 10. Turbiscan reading after the sonication of raw milk in the one and two 477

transducer geometry. Like in the coarse emulsion the particle clusters persist longer 478

than for the single transducer 479

480

27

481

Fig. 11. Microscopy image of the top layer 10 min after sonication for the different 482

frequencies for the coarse emulsion. The particle concentration correlates very well 483

with the computed creaming extents in the tables 1 to 3 484

485

28

486

Fig. 12. Microscopy image of the top layer 10 min after sonication for the different 487

frequencies for the raw milk. The particle concentration correlates very well with the 488

computed creaming extents in the tables 1 to 3 489

490

29

491

Fig. 13. Example for the recovery rate as a function of time for raw milk in the two 492

transducer geometry. The recovery rate is greater than one, showing the higher 493

integral backscattering compared to the steady state 10 min after sonication. The 494

decrease can be assigned to reversible re-dispersion of particles out of the floccules. 495

496

30

Tables 497

498

Table 1 499 Creaming extents (CE) and recovery rates (RR) for the experiments with the single 500 transducer geometry at a frequency of 400 kHz for the coarse emulsion (coarse), fine 501 emulsion (fine) and raw milk 502 Time

[min]

CE

coarse

RR

coarse

CE

fine

RR

fine

CE

raw milk

RR

raw milk

0* 0.086 99% 0 100% 0 100%

1 0.100 99% 0 100% 0 100%

2 0.108 99% 0 100% 0 100%

5 0.104 99% 0 100% 0 100%

10 0.104 99% 0 100% 0.0185 100%

*1 min after sonication 503

504 Table 2 505 Creaming extents (CE) and recovery rates (RR) for the experiments with the single 506 transducer geometry at 1.6 MHz for the coarse emulsion (coarse), fine emulsion (fine) 507 and raw milk 508

Time

[min]

CE

coarse

RR

coarse

CE

fine

RR

fine

CE

raw milk

RR

raw milk

0* 0.265 99% 0 100% 0.116 99%

3 0.236 98% 0 100% 0.119 99%

5 0.237 98% 0 100% 0.119 99%

7 0.239 98% 0 100% 0.119 99%

10 0.238 99% 0 100% 0.119 99%

*1 min after sonication 509

510 Table 3 511 Comparison between the creaming extents (CE) and recovery rates (RR) for the 512 experiments with the one and two transducer geometry at 400 kHz for the coarse 513 emulsion (coarse), fine emulsion (fine) and raw milk (10 min after sonication, as only 514 then all particle clusters had risen to the top) 515

CE

coarse

RR

coarse

CE

fine

RR

fine

CE

raw milk

RR

raw milk

one transducer

0.104 99% 0 100% 0.0185 100%

two transducer

0.551 91% 0 100% 0.643 85%

516