Rotor and Stator

CRANFIELD UNIVERSITY
N THAPAR
LIQUID-LIQUID DISPERSIONS FROM IN-LINE

ROTOR-STATOR MIXERS
SCHOOL OF MECHANICAL ENGINEERING
PhD THESIS
CRANFIELD UNIVERSITY
SCHOOL OF MECHANICAL ENGINEERING
PhD THESIS
Academic Year 1998-2001
N THAPAR
LIQUID-LIQUID DISPERSIONS FROM IN-LINE ROTOR-STATOR MIXERS
Supervisor: Prof. D. E. Brown
July 2004
This thesis is submittedin partial fulfilment of the requirementsfor the degreeof Doctor of
Philosophy.
© Cranfield University 2004. All rights reserved. No part of this publication may be
reproducedwithout the written permissionof the copyright owner.
Abstract
Experimentswere performed on an industrial scaleSilverson in-line rotor-stator mixer to
investigateits liquid-liquid dispersioncapabilities. A non-coalescingkerosene-water
systemwas usedin the testsand the effect of stator geometry,rotor-speeds,flowrates and

dispersedphaseconcentrationson the droplet size distribution was investigated.
The rotor-speed and the dispersed phase concentration were found to have the most
influence on droplet size. Drop sizes were also seen to increase with an increase in open
area. No differences were observed in the droplet sizes or power draw of the stators
with the smallest and similar open areas (Emulsor Screens and Square Hole High Shear
screen). The power supplied to the fluid was proportional to N3 indicating the mixer was
operating under turbulent conditions. d32was correlated against the rotor speed and
dispersed phase and the relationship was found to be close to that described by Chen and
Middleman (1967). This analysis suggested that inertial stresses in the viscous sub-range
were mainly responsible for drop break-up.
d32 = 2x109(1+ 20(D)(WeRe4)-'"
An estimation of the average energy dissipation rate was made in order to determine the
Kolomogov length scale. The droplet sizes ranged from below the Kolmogorov length
scale to significantly higher, suggesting that droplet break-up is due to inertial and
viscous sub-rangeeddies.
The Re could be defined in different regions within the mixer thesevalues were used
along with the nominal residencetimes in eachregion to determinewhere in the mixer

the main drop break-up was occurring. The residencetimes for each region were in the
following order: ShearGap < Stator < Inlet < Rotor < Volute. The relatively long
residence times and the magnitude of the stresses indicated that droplet break-up in a
single pass through the in-line rotor-stator mixer is predominately determined by the
viscous and inertial stressesin the rotor region.
Acknowledgements
I would like to thank the members of my support panel, in particular Professor Don
Brown for his support, guidance and patience.
Many thanks are due to PeterMatthews, Harold Rothmanand everyoneat Silverson

MachinesLtd for their unlimited supportand for supplying the in-line rotor-stator mixer,
constructingthe test rigs and also giving me the use of their Malvern Mastersizer
I would also like to thank my friends and colleagues at BHR Group Ltd, especially Barry
Johnson and the rest of the HILINE staff and members, for their assistance and guidance.
Special thanks are due to John Brown and John Knopp for their aid with the
instrumentationon the rig and to the guys in the workshops,who were ever willing to
help.
I also owe a debt of gratitude to BHR Group Ltd and also the EPSRCfor their financial
support. In addition, I would like to thank the EPSRC Loan pool for making their
Malvern Mastersizeravailable to me.
I would like to thank my friends, Tabs, Atul and Jin for continually pesteringme to finish
this thesis.
Finally, I would like to thank my Family; my mum and dad and my two brothers,this is
for you.
LIST OF FIGURES. V
»......»..».»..»..».....»..»..».»..»..»..».»..»..»..»..».»..»..»..».»..».....»..»..»..».»..»..»..»..».»..
LIST OF TABLES .......».....»..»..»..».»..»..»..»..».....»..»..».»..»..».»..»..»..».»..»..»..».....»..........».»......».....».IX
LIST OF PLATES ...»..».....»..»..»..».»..»..»............................................................................ X
NOMENCLATURE »..»..»..».»..»..»..».»..».....»..»..»..».»......»..»..».»..»..»..».»..»..».»..»..»..».»..»..»......».».XI
CHAPTER 1. INTRODUCTION ...»..»..»...........»».»..»..»..»..».»..»..»..».»..»..».»..»..»..».»..»................ 1
CHAPTER 2. LITERATURE REVIEW 3
2.1. ROTOR-STATORASR BACKGROUND.............................................................................................. 3

2.1.1. Mixer Comparison 3
.......................................................................................................................
2.1.2. Rotor-Stator mixer designs 6
.........................................................................................................
2.1.3. In-line rotor-stator mixers 9
..........................................................................................................
2.1.4. In-line rotor-stator mixer designsavailable on the market 9
....................................................
2.1.5. Operation of an in-line rotor-stator mixer 11
..............................................................................
2.1.6. Industrialapplications .............................................................................................................. 12
2.1.7. Summary 14
......................................................................................................................................
2.2. LIQUID-LIQUID THEORY 15
..................................................................................................................
2.2.1. Forces affecting drop stability 17

..................................................................................................
2.2.2. Drop break-up in turbulentfow 20
..............................................................................................
2.2.3. Mean drop size correlations 29
......................................................................................................
2.2.4. 31
The effect of dispersedphase volumefraction, (P,on turbulence ........................................
2.2.5. Summary 32
.....................................................................................................................................
2.3. ROTOR-STATOR MUCERS 34
...................................................................................................................
2.3.1. Flow patterns and Reynolds numbers in an in-line rotor-stator mixer 34
...............................
2.3.2. Drop Break-up in shear and elongatedflow fields 42
................................................................
2.3.3. Drop break-up in a rotor-stator mixer 55
.....................................................................................
2.3.4. Turbulent energy dissipation rates, c 58
......................................................................................
2.3.5. Energy dissipation and droplet size 60
...........................................................................................
2.3.6. Turbulent break-up in the sheargap 61
.........................................................................................
2.4. POWER CONSUMPTION IN A ROTOR-STATOR MIXER 62
........................................................................
2.4.1. Totalpower supplied tofluid 62
.....................................................................................................
2.4.2. Shaftpower - PsruFT 63
...................................................................................................................
2.4.3. Pumpingpower - Pp ' 65
..............................................................................................................
2.4.4. Mechanical losses- P,. 67
.,.........................................................................................................
2.4.5. Disc friction losses- PDJsc 67
.........................................................................................................
I
Z4.6. Leafage (recirculation) losses 67
-P........... »......_._.._.........»........._.........._.._...................
2.4.7. Hydraulic losses PHrzvuuuc»..... 67
- _.._.._.._...................._......................................._................
2.5. DROPLET SIZE MEASUREMENT TECHNIQUES 69
...................................................................................
2.5.1. On-line drop size measurement techniques 69
............................_.................................................
2.5.2. Off-line drop size measurement techniques 70
..............................................................................
2.5.3. Technique selection 71
......................._......................................._.................................................
2.6. COWLUStoNs 71
...................................................................................................................................
CHAPTER 3. GOALS, ALMMS
AND OBJECTIV 73
CHAPTER 4. EXPERL'%IENTALEQUIP.1MENT,11MATERIALS
M'D DMETHODS__..
_.._..____74
4.1. ExpuuMi wTAL Rio A.tii INSTRUMENTATION 74
................................................................................
4.1.1. Experimental Rig 75
..................................».........».............»..».....................................................
4.1.2. tlaterials of Construction 78
. ...............................................................
...........................................
4.1.3. Flowrate and ocontrol 78
...............................................
...............................................................
4.1.4. Rotor-stator miser desIgn. 79
...........................................
...............................................................
42. EXPERIMENTAL MATERIALS 82
.............................................. ...............................................................
4.2.1. Immiscibleliquids 82
.......................................................................................................................
4.2.2. Kerosene-I f tier Stabiltry 83
...........................................................................................................
4.3. DROPLET SUE MEASUMMawr 85
.......................................... ..............................................................
4..x.1. Malvern dlasterslzer 85
...................................................................................................................
43.2. 1idea Probe 89
...................................................................
..............................................................
433. Droplet sizedistributions 92
..............................................................
.............................................
44. SARWLtNO TEC1tti1QUE 94
....................................................... ..............................................................
4.41. Continuous Sampling 94
..............................................................
....................................................
4 S. MrAstwMr_NT OF PoW'u ses'ititn TO TI M 1W!! ) 97
......... ..............................................................
4J. 1. Thermometric method of power measurement 98
..........................................................................
4J. 2. Recycle loop and instrumentation 98
.......................................................
......................................
4333. Experimentalprocedurr 100
...........................................................................................................
4.5.4. /,&" Fr 101
.........................................................................................................................................
4JJ. P,. r 102
........................................................................................................................................
4.5.6. "MiCil 102
.............................................................................
...........................................................
4.5.7. I' L,0c 102
......................................
........................... ....................................
...................................
4J. 8. Pax and Partur4m 103
.....................................................
..........................................................
".5.9. Lossesthroughte control vahveandpipe wo,* 103
....................................................................
CHAPTERS. RE'SUL 104
S1 MEALiu MuT TEaNIQcx VALU)AnOI 104

.....................................................................................
ii
5.2. IN-LINE ROTOR-STATOR PERFORMANCE TESTS 105
.............................................................................
5.3. POWER MEASUREMENT 105
.................................................................................................................
CHAPTER 6. DISCUSSION 106

»..»..».»......»..».»..»..».»..»..»..».»..»..»..»..».»..»..».....»..»..».»..»..»..».»...
6.1. MEASUREMENT TECHNIQUE AND SAMPLING VALIDATION 106
...........................................................
6.1.1. Kerosene-WaterStability 106
.........................................................................................................
6.1.2. Comparison of measurementtechniques 107
.................................................................................
6.1.3. Sampling techniquevalidation 109
.................................................................................................
6.2. DROP SIZE DISTRIBUTIONS 112
............................................................................................................
6.2.1. Volumefrequency curves 112
.........................................................................................................
6.2.2. The relationship betweend32and d. 113
.....................................................................................
6.2.3. The effect of rotational speed, N, on droplet size 115
...................................................................
62.4. The effect of dispersed phase concentration, 0 118
.....................................................................
6.2.5. The effect offlowrate, Q, on droplet size 120
................................................................................
6.2.6. The correlation of d32with N 121
...................................................................................................
6.2.7. Correlation of d32with We and Re 124
..........................................................................................
6.2.8. Summary 127
....................................................................................................................................
6.3. POWER MEASUREMENT 129
.................................................................................................................
6.3.1. Losses across the control valve and the pipework 129
.................................................................
6.3.2. PSZUPT 130
.........................................................................................................................................
6.3.3. PPUMP 134
.........................................................................................................................................
6.3.4. PmEavand PDfsc 13 7
........................................................................................................................
6.3.5. Power supplied to the fluid by the rotor, PFLum 139
.....................................................................
6.3.6. Energy dissipation and the Kolmogorov length scale 142
............................................................
6.3.7. Summary 144
....................................................................................................................................
6.4. AN EXPLANATION OF THE TRENDS IN DROPLET SIZE BASED ON ENERGY DISSIPATION 145
...............
6.5. FLOW REGIMES IN THE ROTOR-STATOR MIXER: RE ANALYSIS 149
.....................................................
6.5.1. Inlet pipe 149
....................................................................................................................................
6.5.2. Rotor region 1.52
..............................................................................................................................
6.5.3. Shear Gap 157
..................................................................................................................................
6.5.4. Stator 163
.........................................................................................................................................
6.5.5. Volute 165
.........................................................................................................................................
6.5.6. Summary 166
....................................................................................................................................
CHAPTER 7. CONCLUSIONS 168
................................................................. »..».........»..»......»..»..
CHAPTER 8. FUTURE WORK 171

..................... .»..»..»..».»..»..»..».»..»..»..».....»..»..»..».....»..».»...
8.1. LIQUID-LIQUID DISPERSIONS 171
.........................................................................................................
iii
8.2. POWER
MEASUREMENT 172
.................................................................................................................
REFERENCES 174
.....»..»..»..».»..................... .»..»..».»..»..».._.»..»......».»..»..».....»..»......».»..»..»..»..»..
APPENDIX 1 EXPERIMENTAL RIG OPERATING PROCEDURES. 181

»..»..».»..».........................
APPENDIX 2 NAMES AND ADDRESSES OF EQUIPMENT SUPPLIERS 188
APPENDIX 3 THEORETICAL BACKGROUND TO DIFFRACTION MEASUREMENT....... 190
8.2.1. Principles of Laser Diffraction 191

................................................................................................
APPENDIX 4 DROPLET SIZE DISTRIBUTIONS 195

»..».....»..»..».....»..»..»..»......
..............................
APPENDIX 5. »..».....»..»..»..».»..»..».»..»..»..».»..»..»..»..».»..»..»..».»..»..».»......»..».»..»..».. 208

».....»..»..»..»..
LIST OF FIGURES
Figure 2.1 A comparisonof different mixers and droplet sizesthey can produce
(Cohen, 1998) 4
.....................................................................................................
Figure 2.2 In-line mixer used in conjunction with an in-tank device (Courtesy of
Silverson Machines Ltd) 7

....................................................................................
Figure 2.3 Multi-stage In-line mixer (Courtesy of Silverson Machines Ltd) 8
...................
Figure 2.4 A Silverson type mixer (Right) and a toothed mixer (Left) the stators
are at the top and the rotors are below ........................................................... 10

Figure 2.5 A standard single-stage Greerco mixer 11
..........................................................
Figure 2.6 General shape of the energy spectrum (its Fourier Transform), Keey (1967)
21
..........................................................................................................................
Figure 2.7 A Silverson type rotor and stator configuration 37
............................................
Figure 2.8 Flow pattern through a stator (Sparks 1996) 39
.................................................
Figure 2.9 Flow pattern in the volute (Sparks, 1996) 41
.....................................................
Figure 2.10 Basic types of globule deformation (Hinze 1955) 42
........................................
Figure 2.11 Flow patternsthat can causeone of the basic types of globule deformation.
43
..........................................................................................................................
Figure 2.12 Break-up of droplets in Couette flow field (Karam & Bellinger, 1968)...... 45
Figure 2.13 Effect of viscosity ratio on EB in simple shear flow field (Grace, 1982) 48
....
iv
Figure 2.14 Effect of viscosity ratio on drop draw ratio necessaryfor break-up at EB
(Grace, 1982) 50
...................................................................................................
Figure 2.15 Effect of viscosity ratio on reduced burst time at EB (Grace, 1982)............ 50
Figure 2.16 Effect of viscosity ratio on EB in hyperbolic flow field (Grace, 1982)........52
Figure 2.17 Maximum stabledrop deformation in extensionalshearfield versus
viscosity ratio (Grace, 1982)..........................................................................

53
Figure 2.18 Drop deformation for burst at critical shear in extensional shear (Grace,
1982) 54
................................................................................................................
Figure 2.19 Effect of viscosity ratio on critical burst time for extensionalshear(Grace,
1982) 55
................................................................................................................
Figure 2.20 Shaft power versusflow rate for a rangeof shaft speedsfor a Silverson.....64
Figure 2.21 Shaft power versus flow rate for a range of sheargap thicknessfor a
Silverson mixer (Sparks, 1996) 66
......................................................................
Figure 2.22 Overall power balance for a Silverson Unit (Sparks, 1996) 68
.........................
Figure 4.1 A flow diagram of the test rig 75
.........................................................................
Figure 4.2 StandardSilverson rotor and stators.Clockwise from top left; Rotor;
Emulsor screen;Generalpurposedisintegrating head;Slotted Head;
Square hole high shear screen 80
........................................................................
Figure 4.3 Stirred vessel used to create dispersions 84
........................................................
Figure 4.4 A Schematicdiagram of a Malvern Mastersizer 86
...........................................
Figure 4.5 A schematic diagram of the flow cell and Hydro 2000G arrangement 87
.......
Figure 4.6 Video Probe housing 90
.......................................................................................
Figure 4.7 Sampling arrangement. Clockwise from top left; (a) The inline
rotor-stator with samplepoint and Video probe in place; (b) Watson

Marlow pump usedfor supplying diluent; (c) Inlet to Mastersizerflow
cell and pump and stirrer unit usedfor flushing the cell; (d) Samplepoint
with sampling arrangement ............................................................................ 95
Figure 4.8 A schematicof the continuoussampling system 96
..........................................
Figure 4.9 Flow loop set-up for thermometric power measurement 99
..............................
Figure 6.1 Droplets collecting at the top of the sampleunder a microscope(5%
dispersed phase concentration, undiluted) 106
...................................................
V
Figure 6.2 Image at 3000 rpm and t= 5% 110
...................................................................
Figure 6.3 A comparison of DSD measured using the Mastersizer and the Video
Probe for a rotor speed of 3000 rpm and c= 5% at 0.81/s 111
.........................
Figure 6.4 Correlation between d32and dm. 114
.................................................................
Figure 6.5 d32versus N (RPM) for the ES and SHHSS (69.85 mm rotor) for
varying c .......................................................................................................
117
Figure 6.6 d32versus N (RPM) for four different screens (50.8 mm rotor), 0.4 Us
and (D =10% ..................................................................................................117

Figure 6.7 d32versus N for 0.4 Us and 0.8 Us (at (D = 10% & 20%) 120
............................
Figure 6.8 d32versus rotor speed for the SHHSS and the ES at different c and Q.... 122
Figure 6.9 d32/D versus (WeRe4)-1/7for different (D 125

......................................................
Figure 6.10 Measured d32versus calculated d32 126
..............................................................
Figure 6.11 Power input across the control valve versus flowrate 130
.................................
Figure 6.12 Shaft power (W) versus rotational speed (RPM) for the Emulsor Screen. 131
Figure 6.13 Shaft power (W) versus rotational speed (RPM) for the Square Hole
High ShearScreen 131

.........................................................................................
Figure 6.14 A comparisonof values for Pshaft from the ES and the SHHSS at 0.41/s
and 0.81/s........................................................................................................132
Figure 6.15 Measured PSHAFT data versus correlation 134
.....................................................
Figure 6.16 Pumping Power versus Flowrate for a range of shaft speed, 69.85 mm
rotor and Emulsor screen..............................................................................

135
Figure 6.17 A comparisonof pumping power values from both the ES and the
SHHSS at different flowrates (for increasing N) 136
........................................
Figure 6.18 Mechanical and Disc frictional losses versus rotational speed 138
..................
Figure 6.19 PF,uID versus N and the dependency of PFLUIDon N 141
..................................
Figure 6.20 P., um/pD5 versus N3 141
.....................................................................................
Figure 6.21 The Kolmogorov length scale versus impeller speed 144
.................................
Figure 6.22 A comparison of d32and Xk versus N (at 0.41/s for ES) 146
..............................
Figure 6.23 Volume DSD generated by the 67.85mm rotor with the Emulsor Screen at Q
= 0.41/s and «= 10%, for varying N (rpm) 154

................................................
vi
Figure 6.24 Comparison between E/Screen and SHHSS: 8000rpm & 0.41/s (69.85mm
rotor) ...............................................................................................................161
Figure 6.25 Schematic showing the "Lip" at the top of the Silverson rotor 162
..................
Figure A. 1 Flow diagram of the rotor-stator liquid-liquid contactor rig 182

......................
Figure A. 2 A comparison of particle size distributions (Volume and Number) generated
by the Mastersizerand the Video Probe for Latex beads
98 µm 195
.............................................................................................................
Figure A. 3 Comparisonof particle size distributions (Volume & Number)
generatedby the Mastersizerand the Video Probe for Latex beads
113 µm 196
...........................................................................................................
Figure A. 4 Volume DSD generated by the 67.85mm rotor with the Emulsor
Screen at Q=0.4 10%, for varying N (rpm)

I/s and t= 197
...........................
Screen at Q=0.4 Us and c=
20%, for varying N (rpm) 197
...........................
screen at Q=0.41/s = 30%, for varying N (rpm) ............................ 198

and'
I/s and cD= 40%, for varying N (rpm)
Screen at Q=0.4 198
...........................
Screen at Q=0.8 Us and c= 5%, for varying N (rpm) 199

.............................
Figure A. 9 Volume DSD generatedby the 69.85mmrotor with the Emulsor
Screen at Q=0.8 I/s and t= 10%, for varying N (rpm) 199
...........................
Figure A. 10 Volume DSD generatedby the 69.85mmrotor with the Emulsor
Screen at Q=0.8 Us and c= 20%, for varying N (rpm) 200
...........................
Figure A 11 Volume DSD generatedby the 69.85mmrotor with the Emulsor
Screenat Q=0.8 Usand J= 30%, for varying N (rpm) 200
...........................
Figure A. 12 Volume DSD generatedby the 69.85mmrotor with the
SQHHS Screen at Q=0.41/s and c= 10%, for varying N (rpm) 201
.............
Figure A. 13 Volume DSD generatedby the 69.85mm rotor with the SQHHS
vii
Screenat Q=0.41/s and c= 20%, for varying N (rpm) 201
...........................
Figure A. 14 Volume DSD generated by the 69.85mm rotor with the SQHHS
Us and 0= 30%, for varying N (rpm)
Screen at Q=0.4 202
...........................
Us and (D = 05%, for varying N (rpm)

Screen at Q=0.8 202
...........................
Us and cD= 10%, for varying N (rpm)
Screen at Q=0.8 203
...........................
Screenat Q=0.8 Usand (D= 20%, for varying N (rpm) 203

...........................
Screenat Q=0.41/s and c= 10%, for varying N (rpm) 204

..........................
I/s and N= 7000 rpm for varying 1(%)
Screen at Q=0.4 204
.......................
...........................
Figure A. 21 Volume DSD generated by the 50.80mm rotor with the Slotted
Head at Q=0.4 10%, for varying N (rpm)
I/s and c= 205
..............................
Figure A. 22 Volume DSD generated by the 50.80mm rotor with the SHHS
...........................
Figure A. 23 Volume DSD generated by the 50.80mm rotor with the SHHS
Screenat Q=0.4 Vs and N= 7000 rpm for varying (D(%) 206

.......................
Figure A. 24 Volume DSD generatedby the 50.80mmrotor with the GDH at
Q=0.4 Us and c=
...........................................
Figure A. 25 Volume DSD generated by the 50.80mm rotor with the GDH at
Q=0.4 Usand N= 7000 rpm for varying 0 (%) 207

.......................................
viii
LIST OF TABLES
Table 2.1 Typical Liquid-Liquid Applications that have benefited by using Rotor-
Stator Mixers (Silverson Machines, 1999) 12
....................................................
Table 2.2 A summary of theoretical correlations to predict droplet size 33
.....................
Table 2.3 Flow patternsthat causeeachdeformation mode 44
..........................................
Table 4.1 Dimensions of the rotors and stators investigated 81
.........................................
Table 4.2 Physical properties of Alcosol D70 and water 82
..............................................
Table 6.1 Latex beadstestedusing the Mastersizerand the video probe 107
...................
Table 6.2 The best fit exponents on N for varying 123
..................................................
Table 6.3 Gradients of the lines in Figures 6.11 125
..........................................................
Table 6.4 The best fit exponentson c for varying 4) 147
...................................................
Table 6.5 dmaxwith corresponding Re and y' and GB values in the Rotor Region
(69.85mm rotor, ES, 0.41/s and 4=10%) with Xk 155

.........................................
Table 6.6 Res for the ES and SHHSS (69.85 mm) 158
......................................................
Table 6.7 Critical shear rates, G8, and burst times, tB, for drop diameters equivalent to
dmaxand d32and the corresponding nominal shear rates (y') from both the ES
and SHHSS (at 0.41/s and 4=10%) 159

...............................................................
Table 6.8 Regtatorin the stator holes and corresponding ti and y' 163
..................................
Table 6.9 Re,, and corresponding values for r and y' 165
oit0 ..............................................
Table A. 1 A summary of estimated absorption values as recommended by
Malvern Instruments Ltd 194
..............................................................................
Table A. 2 Calculated Gu and tB for measured dm and d32from ES and SHHSS 208
.....
LIST OF PLATES
Plate 0.1 Photograph of test rig (Front View) 76
................................................
Plate 0.2 Photograph of test rig (side view) 77
................................................
ix
Nomenclature
Symbol Meaning Units
a Constant
-
ß Kolmogorov constant -
S Sheargap m
Mass specific local energydissipation rate W/kg
emax Maximum local energydissipation W/kg

s-g Mean energy dissipation rate W/kg
(D Dispersedphasevolume concentration -
Y Shearrate s'1
A Shaft torque (Nm)
x Wavelength light in
Xk Kolmogorov length scale µm
Dynamic viscosity of the continuous phase Ns/m2
cy Dynamic viscosity of the dispersedphase Ns/m2

,
PC The density of the continuousphase kg/m3
a Interfacial tension N/m
ze The disruptive stressdue to turbulent pressure

fluctuations N/m2
zk Kolmogorov time scale s

Resistive force per unit areadue to interfacial tension N/m2
zy Resistive force per unit areadue to viscosity N/m2
v Kinematic viscosity m2/s
A] Constant
-
a Distance from scattererto detector in
a8 Interfacial area m2
B, Energy concentrationfactor -
B2 Constant -
X
C Constant -
Cp Specific heat capacity J/kgK
d Droplet diameter gm
dmax Maximum stable drop diameter µm
d32 Sautermean diameter gm
d43 Volume mean diameter µm
D Rotor diameter m
Dc Radial length of a region in the mixer m
L-B
D Drop deformation =
L+B
EB The reducedshearrequired for break-up
E Energy J
E(k) Energy spectrum function m/s3
E(k) Energy spectrum function m/s3
I(®) Intensity of the scatteredlight (as a function of angle, O)
Io Intensity of the incident light
JI Bessel function
Wavenumber m'1
.k
kp Imaginary (absorption) part of the of the refractive index
N Rotational speed RPM
Np Complex refractive index of the particle
np Real part of the refractive index
m Mass flowrate kg/s

P Power W
PFLUID Power supplied to the fluid by the rotor W
PLEAK Leakagelosses W
PMECH Mechanical losses W
Pa Power number -
Power supplied by the motor shaft W
PPUMP Pumping power W
xi
PHYDRAULIC Hydraulic losses W
dp Differential pressure N/m2
Q Flowrate Us
Re Reynolds number p CND2

-
PC
AT Differential temperature K
ATs Temperature rise over a single pass K
t Time s
tb Reducedburst time s
Uk Kolmogorov velocity scale m/s
O (d) Mean squarevelocity difference across m2/s2
the surfaceof the drop with diameterd.
U Mean velocity m/s
V Volume m3
vt , Rotor tip speed m/s
VPDIAL Nominal radial velocity m/s
We Weber Number P_N2D3

Q
Wect Critical Weber Number
-
x Particle diameter m
X11
CHAPTER 1. INTRODUCTION
The generation of dispersionshas traditionally been performed in stirred tanks. However,
other devices are rapidly becoming more popular and are proving to be more efficient.
One such alternative is the Rotor-Stator mixer, which consistsof a high-speedrotor
to
spinning close a motionless stator. Any liquid (or solid-liquid) mixture passing
through it is subjectedto high shearforces, chopping and turbulence. The Rotor-Stator
Mixer, due to its versatility, high-energy dissipation rates and cost efficiency is increasing
in popularity and as a result there is a wide range of different designsavailable on the
market.
The main aim of this work was to quantify the behaviour of an industrial sized in-line
rotor-stator mixer with regards to its liquid-liquid performance. Radial flow devices,
such as those manufactured by Silverson Machines Ltd, were the main emphasisof this
research.
The generation of liquid-liquid dispersionsand the various parametersaffecting the

droplet size distributions were investigated. A number of different stator designswere
tested for a number of different operating conditions, (varying rotor speed,dispersed
phaseconcentration and flowrate).
Particular emphasiswas put on the droplet size distribution generatedunder different
operating conditions. This then led to the application of various mechanistic correlations
to determine the dominant break-up mechanismsin the mixer.
In order to achieve the main goals a liquid-liquid contactor rig was constructed
incorporating hygienic fittings to maintain cleanliness. The droplet size distribution is an
important parameterin characterisingliquid-liquid dispersions.The small droplet sizes
expected from the rotor-stator mixer led to the use of a laser diffraction unit (Malvern
Mastersizer) to measuredrop sizes.

The mechanistic correlations show a dependencyon droplet size with energy dissipation
rate and therefore the of

measurement power supplied to the fluid was also performed.
Following on from this introduction an assessmentof the current level of knowledge on
rotor-stator mixers, with particular emphasison liquid-liquid theories, is given (Chapter

2). The findings of the literature determinethe main goals of this researchand are
presentedin Chapter 3.
Chapter 4 then details the experimental approachadoptedto achieve the main aims. A
detailed description of the liquid-liquid contactor rig is given and the materials used in
the experiments.A description of the experimental methodology is provided including
the validation of the drop size measurementtechnique and the sampling strategy used.
The technique used for power measurementis also described.
Chapter 5 then reviews the experimentsperformed and details the location of the
corresponding results. A discussion of these results is given in Chapter 6, followed by
conclusions and recommendations for future work
2
CHAPTER 2. LITERATURE REVIEW
The mixing of two immiscible liquids is an important unit operation in the chemical and
process industry. The most common immiscible liquid dispersions contain an organic
and an aqueousphaseand are traditionally manufacturedusing stirred vessels.
However, increased competitiveness in the process industry has made mixer selection
more rigorous and analytical and consequently alternative devices, such as Rotor-Stator
mixers are rapidly growing in popularity. The rotor-stator's versatile nature, high shear
and energy dissipation rates makes it ideal for processes requiring fast disintegration,
homogenisation, dissolving, emulsification and dispersions.
Despite their popularity, there is very little literature openly available on their design.
Operators are very much dependent on their own experience and that of the
manufacturers, both of which may result in discrepancies in performance. The lack of

detailed design information has led to the requirement of laboratory testing and pilot
plant trials.
Laboratory testing is predominantly concernedwith the measurementof the mean droplet
size and the size distribution. Droplet size distribution is an important parameter in
characterising liquid-liquid dispersions as rate dependent processes are greatly influenced

by interfacial area. To be able to establish effective design guides, an understanding of
the mechanismsinvolved in the processof drop break-up is required.
2.1. Rotor-stator mixer background
2.1.1. Mixer Comparison
Selecting the correct type and size of mixer has traditionally been based on the
experience of operators and mixer manufacturers. Consequently, different mixer types
are used based on their performance on specific applications.
3
Generally, mixer manufacturers tend to consider rotor-stators to be best suited for
blending dissimilar viscosity liquids, dissolving low viscosity liquids and also high-
contact reactions as well as liquid-liquid emulsions (Dietsche, 1998). The main

performance and selection criterion of these mixers with regards to liquid-liquid
dispersionsis the droplet size distributions generatedby them.
Cohen (1998) showed a comparison of different in-line mixers based on their droplet
generating properties (Figure 2.1). Unfortunately the author does not define tank turnover
and it is not known whether the comparisons are based on the similar size mixer, similar
volumes and for what liquid systems. However, it is reasonable to assume that the
mixture was re-circulated through each mixing device until an equilibrium droplet size
had been achieved and that the mixers used were ones that the different manufacturers
would recommend for the sameduty.
-- High pressure homogenizer

20
Low-profile rotor-stator
£ Colloid mill
-A-
--f- Multistage rotor-stator
L15
-. - Single-stage rotor-stator
10
W
n5
0
123456789 10
Tank turnovers
Figure 2.1 A comparison of different mixers and droplet sizes they can
produce (Cohen, 1998)
4
It can be seen from Figure 2.1 that there are various mixers available for the production
of liquid-liquid dispersions. However, for applications that require fine droplet size
distributions the High-PressureHomogenisertends to be the preferred option. This mixer
consists of a high-pressurepump that forces a fluid (two immiscible liquids) at very high
velocities through a narrow orifice.
The Colloid mill, a rotor-stator device which consists of a conical rotor and stator, is also
an effective device for producing small droplets. The surfacesof the rotor and the stator
can be either smooth or toothed and can consist of one or more pairs of concentrically
arranged discs of various designs. The fluid is retained in the mixing head for longer
periods than a traditional rotor-stator mixer, hencemore energy is dissipated to the fluid.
It is worth noting that Cohen (1998) claims that the low-profile rotor-stator also generates
droplet sizes close to those from the High Pressure Homogeniser. This mixer is
essentially a multi-stage rotor-stator device (a number of concentrically arranged fine

toothed discs) which runs at high speeds. Although this claim is unsubstantiated,it gives
a good idea of the capabilities of a rotor-stator mixer by making changesto the operating
speedsand geometry.
The trends exhibited in Figure 2.1 agree with the findings of Schubert (1997) who
investigated the relationship between power per unit volume of various mixers and their
resultant droplet sizes and also the observations of Dietsche (1998). They both
characterisedthe High-Pressure Homogeniser and Colloid Mill with high shear-ratesand

higher power per unit volume than other designs. It was claimed that this was the reason
for their finer droplet size distributions.
This makes High-Pressure Homogenisers and Colloid Mills suited to applications that
involve fine emulsions (the pressurehomogenisercan generatesub-micron emulsions and
is often used in the homogenisation of milk), fine particle break up and in the case of the
Colloid Mills, viscous emulsions.
5
However, the High-Pressure Homogeniser is limited to use with low viscosity liquids,
typically below 0.5 Pa.s (Dietsche, 1998). Also the energy consumption and equipment
size limit these units to low flow capacities. The Colloid Mill almost always requires
pump assistanceand also has a low capacity range. Both these units are difficult to clean
in place.
In-line rotor-stator mixers incorporate relatively high power per unit volume and do not
suffer the same disadvantagesof the High-Pressure Homogeniser and the colloid mills.
Furthermore, Figure 2.1 and the investigations of Schubert (1997) reveal that the droplet
size distributions generated by these mixers can be close to those of the High-Pressure
Homogeniser and the Colloid Mill depending on the physical properties of the fluids
being mixed. An added advantageof rotor-stator devices is that their performance can be
improved greatly by changing the design of the rotor and the stator.
Further advantages of the in-line rotor-stator mixer are discussed in Section 2.13.
2.1.2. Rotor-Stator mixer designs
Rotor-stator mixers consist of a high-speed rotor that spins very close to a motionless
stator and any fluid mixture passing through it is subjectedto high shear forces, chopping
and turbulence. The clearancebetween the rotor tip and the stator wall is typically below
1 mm and is commonly referred to as the sheargap. Rotor-stator mixers can be used in a
number of different configurations, as batch (in-tank) or as continuous in-line devices.

Cohen (1998) discussedthe three most common configurations in detail.
(a) Batch single stage.

(b) Single-stagein-line systems.
(c) Multi-stage in-line systems.
The batch single stageunits refer to in-tank devices. The single stage in-line systemscan
be incorporated into a pipeline, as a once through mixer or, as is often the case, in
conjunction with an agitator in a vessel (Figure 2.2).
6
.40
Agitator for
in-tank
uniformity
In-Line
Rotor-Stator
Mixer
moll.
Figure 2.2 In-line mixer used in conjunction with an in-tank device

(Courtesy of Silverson Machines Ltd)
The objective of a multi-stage unit is to generate greater total energy input by increasing
the number of shearing events that each particle or droplet experiences, with each pass
through the mixer. Some designs involve several single-stage units lined up in series,
other designs have the same effect but as a single stage (Figure 2.3).
These single stage units are made up of multiple rotor-stator pairs arranged
concentrically. The theory behind these units is that, as the mixture moves outwards
from the centre, it undergoes a quick successionof shearing events, which generates a
shorter processing time and a smaller droplet size. However, due to the lack of
correlations relating droplet size to mixer type and size there is no definitive way of
7
knowing whether an equilibrium droplet size has been attained. Thus, regular testing of
the product is required.
It is clear that the in-line rotor-stator mixer is the more versatile of the two main
configurations, (the multi-stage can be consideredto be an adaptation on the single-stage

unit). The in-line mixer popularity can also be attributed to its ability to be incorporated
within a recirculating loop ensuring a greater percentage of the fluid (in some cases
100%) is exposedto areasof high-energy dissipation. The fact that numerous passesare
required in many applications (Figure 2.1.) leads to the assumption that during a single-
passall of the fluid may not be subjectedto areasof high energy dissipation.
. nM1
Figure 2.3 Multi-stage In-line mixer (Courtesy of Silverson Machines Ltd)
8
The focus of this research, therefore, was specific to single stage in-line rotor-stator
devices. Sparks (1996) defined an in-line rotor-stator mixer as one that can be
incorporated into a process pipeline, similar to a centrifugal pump, often contributing a
pumping effect.
2.1.3. In-line rotor-stator mixers
The main advantagesof an in-line rotor-stator mixer are:
"A small low-powered unit can handle a batch of any size by recirculating material
until the is
process completed.
" They are ideal for processlines that require frequent changeover.
" Solids and liquids can be injected directly into the mixing head, which is the area of
highest shear.
" They also offer more accurate control, easy and reliable sampling of mixed product
and better controlled recirculation, than batch systems(Cohen, 1998).
The in-line rotor-stator can pump liquids with viscosities up to 5 Pa.s, but with pump
assistance, can handle greater than 100 Pa.s (Dietsche, 1998). Cohen (1998) reported
viscosities from 10 Pa.s up to 1000 Pa.s with a supplemental pump, emphasising the
in
variance performance of different designs.
They can also cope with a wide range of flow capacities from as low as 4 litres/min to
more than 4000 litres/min. The amount of shear, milling, and particle disintegration is
determined by the high speed of the rotor, which also supplies flow. The stator design
varies with manufacturer and has an influence on the effectivenessof the mixer.
2.1.4. In-line rotor-stator mixer designs available on the market
In-line rotor-stator mixers are supplied with various designs of rotor and stator. In
general these designs be

can grouped into three categories:
9
(a) Mixers that are supplied with swept blade rotors combined with stators (basically
cylinders with perforations). These mixers are the most common design available
and will be referred to in this report as a Silverson design (Figure 2.4.).
Figure 2.4 A Silverson type mixer (Right) and a toothed mixer (Left) the stators
are at the top and the rotors are below
(b) A toothed design is one in which the rotor and stator are discs with a number of
approximately trapezoidal teeth around their circumference (Figure 2.4.). These units
are typical of ones produced by German and Swiss manufacturers.
(c) The third design is an axial flow device, an example of which is the Greerco pipeline
mixer (Figure 2.5).
10
Figure 2.5 A standard single-stage Greerco mixer
2.1.5. Operation of an in-line rotor-stator mixer
The operation of a Silverson mixer is described below. The fundamentals are true for all
in-line rotor-stator mixers. The basic operation of a rotor-stator mixer can be divided into
three main stages:
i. Suction
ii. Milling
iii. Intense hydraulic shear
The high-speed rotation of the rotor blades createssuction, which draws the material to
be mixed into the mixing head. The material is then forced, by the centrifugal force,
towards the edge of the mixing head and into the gap betweenthe tips of the rotor blade
and the stator wall, (the sheargap). The flow of material approaching the stator is
tangential. The mixture is then exposedto intensehydraulic shearas it is forced, at high
velocity, through the perforations in the stator. The flow of material out of the stator is
radial.
11
2.1.6. Industrial applications
Examples of rotor-stator applications can be found in the food, chemical, pharmaceutical
and cosmetics industries. As this is

project predominantly concerned with liquid-liquid
systems,the majority of the following examplesconcentratein this area.
Food, Cosmeticsand Toiletries Industries:

Table 2-2 gives a brief review of liquid-liquid applications that can be found in each of
these sectors.
Table 2.1 Typical Liquid-Liquid Applications that have benefited by using Rotor-
Stator Mixers (Silverson Machines, 1999)
Industry Application Manufacturing Problems

(Liquid-Liquid) Solved by Rotor-Stator mixers
Food " SaladDressing " Formation of agglomerates
" Flavour emulsions " Unstable emulsions
" Long processtimes
Cosmetics and " Creamsand Lotions " Formation of agglomerates

Toiletries " Deodorantsand " Unstable emulsions
Antiperspirants " Long processtimes
" Sun Tan Lotions " Homogenousproduct
" Reducedaeration
" Increasedshear
Other examplesthat are not liquid-liquid processesinclude:
Food Industry:
9 The preparation of isinglass finings for the brewing industry.
" Dispersion of Pectin (a polysaccharideobtained from fruit) for jams and preserves.
Cosmeticsand toiletries industry:
12
" The dilution of high activity surfactants.
The oil and gas industry

Saadevandi& Zakin (1996) investigated the use of rotor-stator mixers in the petroleum
industry. In particular they were concerned with the mixing of heavy crude oil with an
aqueous surfactant solution to form an oil/water emulsion. This enabled the easy
transportation of the otherwise heavy crude, resulting in reducedpumping costs.
Chemical & pharmaceutical

The following are not examples of liquid-liquid processes,but list other chemical and
pharmaceuticalprocessesthat have benefited from using rotor-stator mixers.
Nikiforov, et al (1990) reported the use of rotor-stator mixers in the production of
aqueouspolyacrylamide (PAA) solutions. PAA is used to enhancethe strength of paper.

The mixer was a modified centrifugal pump. This was done to investigate an alternative
method to a conventional stirred tank for producing a PAA solution. It was found that
the stirred vessel had problems attaining the ideal specific energy rate, E, (= 10 W/kg) and
also had difficulties producing a uniform PAA solution due to the formation of stagnant
areas. The modified pump not only generated a value of e= 1015 W/kg but after two
passes,formed in
a uniform solution less time than the stirred vessel.
Other examples include (Silverson, 1999):
" The production of polymer modified bitumen for road surfacing.
" The re-dispersion of filter cake.
" High-speed dispersion of filter cake.
" High-speed dissolving of viscosity index improvers in luboils
" The production of cough mixtures and pharmaceuticalsyrups.
" Manufacture of pharmaceuticaltablet coatings
13
2.1.7. Summary
There are a large number of different rotor-stator designsavailable on the market, each
with its own unique geometric variation. They can be classified into three main
categories:
Radial flow devices (Silverson design)

ii. Toothed type mixers
iii. Axial flow devices
However there is very little academicdata available on any of these designs.

The diversity of the in-line rotor-stator mixer is highlighted by the fact that they are used
in a broad range of industries that include liquid-liquid mixing applications.
Although the high-pressurehomogeniserand the colloid mills are traditionally used to
produce fine emulsions, there are disadvantages associatedwith these large energy
consuming units with high capital and operating costs. The work of Schubert (1997) and
manufacturers experience(Dietsche, 1998 & Cohen 1998) has shown that the rotor-stator
be to
mixer can made produce fine emulsions for some applications.
It is thought that the droplet generatingproperties of thesemixers is very much dependant
on the physical properties of the fluid, the mixer geometry and rotational speed. The
main advantagesof the in-line rotor-stator mixer over these other mixers are:
(a) Cheaperto run and maintain.

(b) They can handle higher throughputs (up to 4000 litres/min).
(c) They are suitable for high viscosities (up to 1000 Pa.s).
(d) They are relatively easy to incorporate into existing processes.
(e) It is possible to passthe entire fluid mixture through the regions of high-energy
dissipation by increasing the number of passes(stages).
14
2.2. Liquid-Liquid Theory
Liquid-liquid dispersionsare made up of two immiscible liquids one of which is
suspended in the other in the form of droplets. The liquid that is present in the form of
droplets is called the dispersedphaseand the liquid in which these droplets are suspended
is the continuous phase. Dispersions can exist in two forms, an oil suspendedin water
(O/W) or the reverse,water suspendedin oil (W/O). The term emulsion generally covers
dispersionswith small and relatively stable droplets, usually with the addition of a
stabilising agent (Becher, 1966).
Most stabilising agentsare surfactantswhich consist of a hydrophilic group (attracted to

the aqueousphase)and a lipophilic group (attractedto the organic phase).The lipophilic
group is attracted to the oil droplet leaving the hydrophilic group exposedto the aqueous
phase. The hydrophilic groups sticking out of the droplet repel similar groups in other
drops, thus preventing coalescence.In addition to their stabilising qualities, surfactants
also exhibit interfacial in

effects and many casesdecreasethe interfacial tension.
The process of making droplets is a simple one. In a large number of cases, simple
shaking proves to be sufficient. However, a large amount of energy is required to break

down the drops further.
The suspension of the dispersed phase in the continuous phase is caused by the direct
action of shear stressesexerted by the continuous phase on the dispersed phase. There
are two sourcesfor these shear those

stresses, due to velocity gradients, (i. e. laminar flow
conditions) and those due to inertial forces, (i. e. turbulent flow conditions).
In addition to the break-up stresses imparted to a drop there are forces that oppose
deformation. Various models have been developed to attempt to describe the scaling of
the stressesthat deform a drop againstthose that resist drop break-up. These models form
a foundation for correlating experimental data and provide a method for confidently
extrapolating experimental conditions for scale-uppurposes.
15
Whilst the process of drop break-up is taking place, at the same time drops are colliding
with each other. In turbulent conditions droplets that are smaller than the micro-scale of
turbulence are completely contained within the eddies whereas larger drops are acted
upon by the turbulent eddies. The collision of drops below the micro-scale of turbulence
is caused by two different mechanisms depending on whether the drop has a density
similar or different from that of the surrounding fluid. Droplets with similar densities
follow the motion of the fluid completely and the droplet velocity fluctuations can be
described by the fluid velocity function. Droplets with significantly different densities
are not completely entrained by the turbulent eddies and drops with differing diameters
move at different in
velocities resulting collisions.
If the drops remain together long enough, so that the separating continuous phase film
thickness reachesa minimum, then the film ruptures and the drops coalesce. There are
many factors that determine whether a collision will result in coalescence including,
surfactants, mass transfer, surface tension gradients, physical properties and Van Der
Waals forces. For coalescenceto occur, contact time must be greater than coalescence
time. A review of various models developed to describe the drop coalescencerates is
given by Tavlarides and Stamatoudis(1981).
Tackling drop break-up and coalescencetogether is far too complicated and as a result
the role of coalescencecan be limited by considering dilute dispersions, or by adding a
stabilising agent. Also, in dilute systems the continuous phase can be described using
concepts developed for a single-phase fluid (Calabrese1992).
Although this research uses dispersions with high concentrations, Calabrese (1992)
stipulated that theories developed for dilute systems can be applied to more concentrated
systems that have been stabilised. This is because the presence of a stabiliser would
inhibit coalescence. This statement, can therefore, be extended to incorporate systems
that do not readily coalesce. It has been shown both in this work (Chapter 3) and work
performed by Brown and Pitt (1971) that the kerosene-water system (used in this work)
can be described as such.
16
Both keroseneand water are liquids of relatively low viscosity (Chapter 3), therefore, the
flow in a rotor-stator mixer is expected to be predominantly turbulent. Sparks (1996),
assumedthat the power supplied by the rotor to the fluid was dissipated in a small region
of the rotor-stator surrounding the rotor teeth, stator teeth and the gap between the two.
Calculating the Reynolds number (Re) using the gap width as the characteristic length
scale and the tip speed as the characteristic velocity (see Section 2.3.1) for rotational
speedsfrom 3000rpm to 12000rpm (with water) gives values of Re ranging from 3.5 x
106to 7x 106(respectively) which would indicate turbulent conditions.
However, it has also been reported that the energy dissipation within the mixer is highly
inhomogeneousand there may be regions within the mixer that may not exhibit turbulent
characteristics, (Sparks, 1996; Pedrocchi and Widmer, 1988; Kroezen & Wassink, 1987).
Furthermore, Pedrocchi and Widmer (1988) found that the gap width had little or no
affect on the droplet size distribution. This would suggest that Spark's (1996)
assumption may not hold true and other factors may be more important for drop break in
a rotor-stator mixer. This is discussedin more detail in Section 2.3.
Most of the work on modelling dispersed phase dynamics such as droplet breakage is
basedon models developed from isotropic turbulence. The main conceptsare outlined in
the following sections.
2.2.1. Forces affecting drop stability
Models that have been developed from theories of droplet break-up are based on
balancing the disruptive force that acts on the drop with the forces that resist disruption.
It is, therefore, important to have an understandingof the forces that affect the stability of
a drop.
If the force acting on the surface of a single drop is due to turbulent pressurefluctuations
then the disruptive force per unit area, -r,, is given by Equation (2.1).
17
r, = pýU2(d ) (2.1)
Where p, is the density of the continuous phase

T is the mean squarevelocity difference acrossthe surface of the drop
with diameter d.
If the disruptive force is due to mean velocity gradients rather than turbulent pressure
fluctuations, then Newton's law of viscosity will apply (Calabrese,1992):
ie a 14Cul (2.2)
Where µ, is the viscosity of the continuous phase

U' is the characteristicmean velocity gradient in the system.
It should be noted that Equation (2.2) is only applicable when the turbulence intensity is
low and the instantaneous velocity gradient averaged over the time-scale of a single
breakageevent is well approximatedby that for mean flow.
The effect of shear is counteracted by the interfacial and viscous forces, which act to
maintain the most compact droplet shape.Equation (2.3) is an expressionfor the resistive
force per unit area,;, due to interfacial tension.
Cs a-
or (2.3)
d
Where, aris the interfacial tension.
18
The dispersedphaseviscosity, µd, provides a resistive force per unit area, r,, within the
drop and is given by:
1/2
µd Tc
CC (2.4)
V d Pd
Where the term (ic/pd)112is the characteristicvelocity inside the drop as determined by
balancing the kinetic energy on the drop's surface.
Calabrese(1992) reasonedthat the interfacial tension provides a restorative force on the

drop whilst the dispersedphaseviscosity dampensany deviations from a spherical shape.
Equation (2.1) shows that the disruptive force is dependenton the diameter of the droplet
and it can be seenfrom Equations (2.3) and (2.4) that the cohesive forces decreasewith
increasing diameter. This leadsto the realisation that there exists a maximum stable
droplet diameter, dmax"The cohesive forces on drops with diameter larger than dmaxare
insufficient to prevent droplet break-up. If the drop diameter is below dmaxthen the
hydrodynamic forces acting on the drop are insufficient to causebreak-up. The
maximum stable drop size can then be defined as the drop diameter for which disruptive
and cohesive forces are in balance(Equation (2.5)).
(dmax )= (dmax )+rv (dmax )

tc 2s (2.5)
It is important to note that breakageis a rate dependentprocessand the concept of dmax
and Equation (2.5) only apply when correlating mean drop size at equilibrium conditions.
In addition to equilibrium conditions, it is further assumedthat:
1) Both phasesexhibit Newtonian fluid behaviour
2) No masstransfer occurs betweenthe two and there is no surfactant present.
3) Dispersion is fully turbulent.
4) Density difference between the two phasesis small to avoid buoyancy effects.
19
2.2.2. Drop break-up in turbulent flow
Theoretical prediction of droplet size distributions has traditionally been based on the
assumption of homogenous isotropic turbulence, more commonly referred to as the

Kolmogorov theory (Kolmogorov, 1949). Another important assumptionof this theory is
that the role of coalescenceis unimportant.
Turbulent flow can be considered to be made up of the random flow of eddies
superimposedon the overall average flow. In fully developed turbulence the velocity at
any point within the fluid varies continually about some mean value and the amplitude of
the variations about the mean may not be small compared with the magnitude of the
mean itself. A similar random variation exists between two points within the fluid. The
instantaneouspoint velocity in space,U, is given by Equation (2.6):
U=iUx+jUy+kU. (2.6)
Where, U, Uy and UZare the scalar co-ordinatesalong the unit vectors Q, k respectively.
It is convenient to break down the instantaneousvelocity U, into the sum of the mean
value over a time interval, U, and the superimposedfluctuation u':
U=U+u' (2.7)
The magnitude of u' provides a qualitative indication of the degreeof turbulence of the
flow. The time averagevalue of this component is zero, as no net motion arises from the
eddies. However, the averageof the squaresof the fluctuations (ur2 ) will not be zero
and thus the root mean square(rms), is

u'Z , value used to calculate an average.
In order to simplify the mathematical treatment of turbulent flow the assumptionthat the
flow is homogenousand isotropic is often made. Homogenousturbulence occurs when
the time averagedvelocity componentsare independentof position (this rarely applies in
20
practice except over short distances). The flow is termed isotropic if the magnitudeof the
fluctuating velocity componentis the samein eachof the three principle directions (i. e.
uxz== uyz. uZ
Turbulent flow can be described as a spectrum of eddy sizes and each eddy has its own
energy associated with it, Figure 2.6. The energy spectral density function E(k) describes
the total turbulent kinetic energy contained within a range of eddies. (Where k is the
wave number and is the reciprocal of eddy size (1/d), such that E(k) represents the energy
contained in eddies of size k to k+ dk).
ENERGY-CONTAINING
t
EDDIES.
V
W
ik kýý
LARGE ý1NERTIAL VISCOUS DISSIPATION
EDDIES SUBRANGE
ýi k, 13
k4 I
UNIVERSAL EQUILIBRIUM RANGE 10
NUMBER -----º-
,k
Figure 2.6 Generalshapeof the energyspectrum(its Fourier Transform), Keey

(1967)
The larger eddies(which are effectedby external forces such as the impeller blade in a
stirred vessel) transfer their energyto the smaller onesuntil the smallesteddiesdissipate
kinetic energy into heat through viscous shear. The smaller eddiesbecomestatistically
21
independentfrom the larger ones. For steadystateto be maintained there must be an
input of energy, provided for example by an impeller.
There is thus a range of frequenciesthat are not excited by external forces and are instead
createdby inertial interactions of larger scalemotions. This range of eddie sizes,where

the energy is self adjusting through inertial forces, is known as the Universal Equilibrium
Range, as turbulence is uniquely determinedby the energy dissipatedper unit mass e, and
the kinematic viscosity, v. The universal equilibrium range is made up of two sub-
ranges: the inertial sub-rangewhere the energy spectrumis independentof v, and only
dependenton e, and the viscous dissipation range where the energy spectrum is
dependenton both (Figure 2.6).
If the assumptions are made that break-up is due to turbulent pressure fluctuations and
that the droplet size is small compared to the macro-scale of turbulence, L, Kolmogorov
(1949) and then later Hinze (1955) arguedthat the flow could be characterisedin terms of
the energy dissipation.
The largest scale of turbulence is the integral length scale of the velocity correlation
function (Hinze, 1955) and is generally assumedto be proportional to impeller diameter,
D, which for rotor-stators can be assumedto be analogousto the rotor diameter. The size
of the smallest eddies is called the Kolmogorov micro scale,?.k.
(V3 )Y4
a'k
= (2.8)
v Kinematic viscosity (m2/s)
Mass specific local energy dissipation rate (W/kg)
22
The characteristic velocity and time scalesin this range can be defined respectively by:
uk=(ve
Y4
and ik
(v y (2.9)
`s )
The region between eddy sizes of D and Xkis the inertial sub-rangewhere the assumption
of homogeneous isotropic turbulence can be applied (Kolmogorov, 1949). For stirred

vessels it is postulated that drop sizes, as a result of turbulent pressure fluctuations, are
much larger than Xk. Therefore, it can be said that eddies in the inertial sub-range are
predominantly responsible for drop break-up. It is worth mentioning that many workers
have reported droplet sizes from a rotor-stator mixer an order of magnitude smaller than
in a stirred vessel. Values for kk in a rotor-stator mixer will help to determine which
energy region is predominant for drop break-up in rotor-stator mixers.
In industrial applications it is difficult to determine local energy dissipation rates and

hence average energy dissipation rate, Eav,is taken over the whole volume of a vessel.
The local energy dissipation rates vary considerably throughout an in-line rotor-stator
mixer and the majority of researchershave estimated values for the highest local energy
dissipation rates, (Sparks, 1996; Pedrocchi and Widmer, 1988; Kroezen & Wassink,
1987).
(a) Drop break-up in the inertial sub-range
Kolmogorov (1949) and later Hinze (1955) developed a model of drop-break in the
inertial sub-range. The model assumes that the disruptive stress on a drop is due to
turbulent eddies in the inertial sub-range and viscous forces of the dispersed phase are
consideredto be negligible.
If the Reynolds number is high, the motion of the fluid in a very small volume are
determinedsolely by the energydissipationrate, e (i.e. local isotropy). This assumption
23
holds true if the macro-scale of turbulence (the length scale of the energy containing
eddies), D, is large comparedto the smallest scale, Xk.

If the turbulent fluctuations acrossthe drop are isotropic, then the mean square velocity
difference acrossthe surface of the drop, U2(d) can be related to the total kinetic energy
,
using the energy spectrum (Equation (2.10)).
UZ(d) = fy E(k)dk (2.10)
The scaling of the E(k) in the inertial sub-rangeis then given by:
25
E(k) =PE 3k 3 (2.11)
Where ß is commonly called the Kolmogorov constant.
Substituting Equation (2.11) into Equation (2.10) and then the result into Equation (2.1)
yields:
ic =1.5ßpc£2/3d2/3 (2.12)
Equation (2.12) describes the magnitude of the inertial stresses acting on a drop of
diameter, d and is valid for D»d» Ak. This equation can be combined with the
expressions for the resistive force Equations (2.3) and (2.4) to give a value for dmax"As
dmaxis determined by the maximum local energy dissipation, c should be replaced by

£max"
in a turbulentfield is thengiven by:

The maximumstablediameterof a drop,dmax,
24
3
Pc s£
d -2/5
max
«
max (2.13)
or
Shinnar (1961) extendedHinze's theory to stirred vesselsby relating smax

to the mean
energy dissipation rate (power draw) per unit mass, r g. In an agitated vessel it can be
assumed that the local energy dissipation rate is proportional to the averageenergy
dissipation rate, sQ,,,over the whole vessel for all sizes of mixing vessel (geometric
similarity).
Therefore Ems,can be expressedas:
£max= B,£avg (2.14)
The constant of proportionality was defined as the energy concentration factor, B1. The
concentration factor is dependanton the geometry of the vessel and impeller.
In a stirred vessel, for sufficiently long residencetimes all the drops in the dispersion will
experience the full of

range energy dissipation. Therefore, it can be safe to assumethat
dmaxwill be determined by the largest values of C. Emaxcan be assumedto be the total
power draw of the mixer. For fully turbulent conditions the power draw, P, can be
determined from drag coefficient analysis which gives:
P= PopýN3D5 (2.15)
Where Po is the power number (a dimensionless group that is a function of geometric,

kinematic and dynamic properties of the system).
Substituting Equation (2.15) into (2.14) yields:
Eavg= B2N3D5 (2.16)
25
Where B2, the constant of proportionality, incorporates the geometric factors including
the power number, Po.
Hinze (1955) suggestedthat the drop size could be related to the droplet Weber number,
We,the ratio of inertia force (or kinetic energy) acting on the surface of the drop to
opposing interfacial forces. This term has a critical value, Wecrig,

above which the drop
becomesunstable. As Eggis a function of the impeller diameter and the impeller tip
speed(5tND) the Weber be

number can expressedin terms of impeller speedand impeller
diameter for mixing vessels:
We= PCN2D3 (2.17)

Q
Equation (2.16) can be inserted into (2.13) and simplified using the definition of We,to
give the Weber number correlation:
dm
= A1We-315 (2.18)
It should be noted that this correlation is only applicable when the droplets are of a
similar density/viscosity to the continuous phaseand inertial stressescreated by eddies in
the inertial sub-rangeare responsiblefor drop break-up. Also, it is assumed that the
in is to
residencetime the vessel sufficient ensure that all the droplets in the dispersion
are exposedto the full range of energy dissipation (i. e. d.,, is determined by the largest
ax
values of E).
(b) Drop break-up in the viscous sub-range

The stress mechanism by which a drop breaks up is dependent on the drop size and
location in the mixing vessel. If the drop passes through a region of high-energy
dissipation, it is possible that drop sizes below Ak will be generated. In such cases the
definition for the functional form of E(k) would not be appropriate. Alternative
previous
definitions for E(k) were presentedby Shinnar (1961) and Chen and Middleman (1967).
26
These expressions can be substituted into Equation (2.7), which then allows a similar
analysis to be performed to yield a correlation in terms of dmaxfor the viscous sub-range

for turbulence.
Shinnar (1961) used the following functional form for E(k), (d < Ak):
E(k)oc ev-'k-3 (2.19)
Which gives Equation (2.20):
-'3
dm «Q'" p-2311'3 £ (2.20)
Equation (2.20) can then be simplified in terms of We and Re:
d« (WeRe)X (2.21)
D
The Reynolds number is the ratio of inertial forces to viscous forces and is defined as:
P, NDZ
Re = (2.22)
Chen and Middleman (1967) used the following functional form for E(k), (d «Ak):
E(k)oc E'v-4k-' (2.23)
Which gives:
2
34v 44
dmax°CQ'7 p E (2.24)
27
Equation (2.24) can then be simplified to:
dmax (,
pc ye Re4 (2.25)
D
Equation (2.21) and (2.25) both apply to the viscous sub-rangeof turbulence and assume
that break-up is causedby inertial stresses. The differences in the two can be attributed
to the definition of the functional form of E(k).
However, Shinnar (1961) suggestedthat for drops smaller than Ak, shear forces would be
more significant than inertial stresses acting on the drop. The viscous shear force
responsible for break-up was given by (d < Ak):
TD « µy' (2.26)
Where the shearrate, y 'oc (e/v)112
This then gives the following expressionfor dmaxto representthe magnitude of the forces
causing drop break-up below the Kolmogorov micro-scale:
Yvyr x
dmax«a'p (2.27)
Simplifying in terms of We and Re:
dmax
« We-'Rey (2.28)
D
It is clear that expressions for dmaxshow different dependencieson system parameters.

The definitions of the main stressesthat are responsible for dmaxdependupon the mode of
droplet deformation or the stressmechanism, which vary depending upon location in the
28
vessel and the droplet diameter. Many workers have reported that the rotor-stator mixer
exhibits inhomogeneousenergy dissipation and it is probable that the stressmechanisms
that cause break-up will vary throughout the mixer region. However, a comparison of
these models will establish which stressmechanisms are dominant in the generation of
dispersionsin a rotor-stator mixer.
Shinnar (1960) assumedthat We was dependanton the energy dissipation as describedby

Equation (2.16). As this has not been establishedfor rotor-stator mixers, (for the range of
rotational speedsto be studied) and to ensure that this assumption is valid, the energy
dissipation will be measured(Section 2.4).
The definition of Reynolds number above (Equation (2.22)) is the impeller Reynolds
number often used for stirred vessels. This definition takes the impeller diameter as the
characteristic length scale and the impeller tip speed as the characteristic velocity (to
determine the inertial forces createdby the rotor). However, Sparks (1996) observedthat
various regions in the mixer had different flow patterns (Section 2.3.1). In each of these
regions the characteristic length scale and the associatedvelocities causing turbulence
may change, these are discussed in further detail in Section 2.3.1.
2.2.3. Mean drop size correlations
The correlations presentedpreviously are all based on the maximum stable droplet size,
dmax.However, most of the correlations available in literature substitute the Sauter mean
diameter for dmax" Many previous authors have used image-capturing techniques to
determine droplet size distributions, which require tedious drop size counting. The
accurate determination of do requires the measurementof a large sample of droplets. It

ax
has been shown that the Sautermean diameter can be determined accurately for a smaller
sample size and therefore offers a practical replacement.
The Sauter mean diameter, d32,is defined as the quotient of the secondand third moment
of the size distribution:
29
" d3
d32 =n (2.29)
2dz
;_, ;
The Sauter mean diameter is important in mass transfer and reaction systems as it is
related to the interfacial areaper unit volume, as:
as =6 (2.30)
d32
Where, 1 is the dispersedphasevolume fraction.
In addition to its design value, d32gives a better representation of the small as well as
large droplets, since it takes volume and surfacearea into account.
The Sautermean diameter is often assumedto be directly proportional to dmax,which was
shown to be true for inviscid dispersedphasesby Chen and Middleman (1967) and also
for non-coalescing systems by Brown and Pitt (1971). They made the assumption that
break-up was due to a single mechanism. Calabrese(1992) generalisedtheir derivations
to state that dmaxis proportional to any convenient mean diameter, provided break-up is
due to a single mechanism.
Calabrese (1992) suggested the use of the volume mean diameter, d43,which better
represents the larger drops in the distribution. It was reasoned that the larger drops
should undergo the same stress mechanisms as dma,,

and would therefore act as sensible
replacement. It was also stressedthat more than one mean diameter should be tested for
suitability in models.
It is important to note that it is possible for two dispersions to have the same mean
diameterbut have very different droplet size distributions. Therefore,althoughmean
30
drop diameters are useful in design correlations the droplet size distribution should
always be consideredwhen assessinga dispersion.

Tavlarides and Stamatoudis (1981) provide a comprehensive summary of correlations
basedon the d32developedby various authors.
2.2.4. The effect of dispersed phase volume fraction, 1, on turbulence
A large number of papers confirm the inference that 1 has little or no effect on the mean
drop size (Middleman, 1974; Chen & Libby, 1978; and Streiff, 1977). However,
Tavlarides & Stamatoudis (1981) and El-Hamouz et al (1988) reported models that had
been developed for droplet size distributions (in stirred tanks) fitted experimental data
better by taking into account a damping effect on turbulence by the dispersedphase.
This agreed with findings by Brown and Pitt (1971) who observed that an increase in
dispersed phase concentration resulted in an increase in d32. As the system they were
using was a non-coalescing one, they attributed the increase to turbulence damping.
Brown and Pitt (1971) and later Streiff et al (1997) showed that d32could be related to 1
in the following form:
d32= C(1+ k(D) (2.31)
Where C and k are constantsdetermined from experiments.
Doulah (1975) derived an expressionthat accountedfor the increasein drop size with the
dispersedphaseconcentration.
Y5
P` 5(1+2.5i 65d32
d32 (2.32)
Pe
31
Where pe is the density of the dispersion and d'32representsthe Suater mean diameter of
drops in a dilute dispersion (as opposed to that in the concentrated dispersion, d32).
Equation (2.32) can be expressedin terms of the energy dissipation:
Pc (1 2.50x3
Ee=E + (2.33)
Pe
£edenotesthe energy dissipation available in a dispersion of high 1 and $ representsthe

energy dissipation in the dilute system. Francis (1999) suggestedthe use of Equation
(2.33) in the derivation of the expressions for dma,(or d32)described previously to
{
account for the effect of (D.
2.2.5. Summary
Traditionally the theoretical prediction of droplet size has been based on the assumption
of homogenous, isotropic turbulence. By extending the theories of Kolmogorov (1945)

and Hinze (1955) to stirred vessels, various correlations for dmaxor d32 have been
developed (Chen and Middleman, 1967; Shinnar, 1961; Tavlarides & Stamatoudis,
1981). These correlations are based on mechanistic models that balance the forces that
cause drop disruption and those that oppose it. The correlations have been simplified to
exclude the effects of coalescenceand drop viscosity and therefore only apply to inviscid,
dilute dispersions (only interfacial effects resist drop deformation). The most important
correlations are summarisedin Table 2.2.
The Sauter mean diameter, d32,has been shown to be proportional to dmaxand has a
greater design value than dmax.d32can, therefore, be substituted into the correlations for
dmaxlisted in Table 2.2.
The majority of authors report that the dispersed phase, (D, has little or no effect on
droplet size. There are a few exceptions,however,who all report a linear relationship
between mean drop size and « (Equation (2.31)).
32
Table 2.2 A summary of theoretical correlations to predict droplet size
Authors Correlation Equation Exp`-on Stress Range

N Mechanism
Kolmogorov 3S (2.13) Inertial D»d
dmax«Q p -35E_Y5 -6/5
& Hinze Sub-Range »2
d..,,
(1955) °CWe-315 (2.18)
Shinnar YSv'36-'S (2.20) -1 Viscous d<Ak

dmax«Q'S p
(1961) d max
. ax (WeRe)-'/3 (2.21)
«
D
Chen and 4, -3; (2.24) -6/7 Viscous d«
dmax« Q'7 P'S,v E
Middleman Sub-range 7ý
dmax (We ý'"
(1967) « Re° (2 25)
D .
Shinnar '/2vyE _Y2 (2.27) -3/2 Turbulent d< Ak
dmax«Qip
(1961) Viscous Shear
d (We 1
_Y2
mý`« Re (2.28)
D
33
2.3. Rotor-stator mixers
2.3.1. Flow patterns and Reynolds numbers in an in-line rotor-stator mixer
An understanding of the flow patterns within the rotor-stator mixer will give an
appreciation of the areas of high energy dissipation and the types of break-up
mechanisms that would be induced as a result. At this point it is worth clarifying the
basic geometry of an in-line rotor-stator mixer. In doing so, attention is drawn back to
Figure 2.4 which shows an example of a toothed and Silverson design of rotor and stator.
The basic principle (as outlined in Section 2.1.5.) is that the rotor spins at high speeds
within a motionless stator (with very small clearances). This creates a suction force
which draws material directly into the rotor region and is in turn forced by the pumping
action of the blades through the perforations or holes in the stator.
The rotor-stator mixer exhibits complex flow patterns that make direct visualisation very
difficult. The flow within a rotor-stator mixer has been reported to be unsteady and
periodic (Kroezen & Wassink, 1987; Pedrocchi and Widmer, 1988; Korshakov et al
1990; Sparks 1996). The fixed stator will experiencea pulsating force as the rotor spins.
A small element of fluid experiences a large acceleration as it enters the rotor region
followed by a suddendeceleration as it proceedsto hit the stator wall.
Sparks (1996) performed some flow visualisation tests on a Perspex toothed rotor-stator
mixer and an acid-basedecolourisation scheme(hydrochloric acid and sodium hydroxide

with bromo thymol blue tracer). The rotor-stator was broken down into sections and the
flow of tracer was observed through each of these sections. These sections included; (a)
the inlet pipe; (b) the rotor region; (c) the sheargap; (d) the stator region; (e) the volute.
The Reynolds number, Re, (Equation (2.22)) is the ratio of inertial forces to viscous
forces present in a system. The relevant forces are consideredover an area of flow and
this area can be simplified to a characteristic linear dimension of the system. In stirred
vesselsthe characteristic length scaleis usually chosenas the impeller diameter, D and
34
the representativevelocity of flow is taken to be the impeller tip speed,3tND. This gives
the impeller Reynolds number defined in Equation (2.22). This dimensionlessparameter
provides an indication of the level of turbulence in a system,the higher the value the
greaterthe likelihood of turbulent flow. In mixing applications the Re provides a
criterion for dynamic similarity for scale-up.
The flow visualisation work performed by Sparks(1996) showed the flow patterns to be
very different in the various regions of the rotor-stator mixer. Also, it has been widely
reported that the energy dissipation in a rotor-stator mixer is highly inhomogeneous. This
would indicate that the turbulent energy would differ greatly throughout the mixer.
This would imply that the impeller Reynolds number may not be a true reflection of the
flow characteristicsthroughout the whole of the mixer. Each section in the mixer will
have its own characteristic length scaleand characteristicvelocity. The various
characteristic length scales and velocities within the mixer will shed some light on where
in the mixer the turbulent energy dissipation is focussed. Taking this in to consideration
would allow the identification of regions that are operating in the turbulent and laminar
regimes.
The following is a review of the findings of Sparks (1996) and the various regions in the
mixer and the corresponding length scalesand velocities will be defined for each section
of the in-line rotor-stator mixer.
(a) Inlet Pipe

The inlet pipe was found to contain swirling fluid causedby the effects of the spinning
rotor drawing the fluid up the inlet pipe. There was no change in flow pattern observed
with operating condition (rotor speed,flowrate).
This swirling flow would result in a small degree of pre-mixing which might be
beneficialin the generationof dispersions.However,thereis alsothe possibility of phase
separation. The degree of pre-mixing will be determined by the turbulence characteristics
35
in the in-let pipe which can be assessedsimply by taking into account the pipe Reynolds
number, Reppe:
Re = pdu (2.34)
p,pe it
Where, the characteristic length scale is the inlet pipe diameter and the velocity is given
by the flowrate divided by the cross sectional areaof the pipe.
(b) Rotor
Fluid entering the rotor region was acceleratedtangentially and the flow pattern upstream
of the rotor teeth could be by

characterised swirl generated as the fluid followed the rotor
face. The flow pattern again showed little dependenceon the operating condition.
The tracer was seento follow the rotor teeth for a considerabledistance (along the rotor
circumference) leading to interaction with tracer plumes from consecutive feeds.

The injection of sodium chloride tracer into the rotor teeth and monitoring of the
conductivity upstreamrevealed little or no back flow into the region.
The flow in this region was consideredto be predominantly tangential, with a much
lower radial component. As the fluid was seento follow the rotor teeth the velocity of the
fluid at the rotor tip, vtp, was consideredto be the sameas the tip speedof the rotor:
ND
vp = ar, (2.35)
Where, N is the rotational speed(s'1)and D is the rotor diameter (m).
Sparks (1996) went on to describethe trajectory of the fluid through the rotor region as
which was defined as:

being indicated by v«,,and the nominal radial velocity, vRADIAL,
36
0
Q
vRADIAL= (2.36)
A
Where, A is the open areaof the rotor (m2) and Q is the flowrate (m3/s).
If VRADIAL can be calculated and the radial length, DRADIAL,that the fluid must travel in a
particular region can be determined. In the caseof the toothed rotor DRADIALin the rotor
region is the length of a rotor tooth, for a Silverson type rotor (a swept blade) this length
will be the length of one of the blades. The nominal residence time of the fluid in that
region can be estimatedusing Equation (2.37).
DRAolaL
_ (2.37)
V
RADIAL
It is worth pointing out that the open area of a toothed rotor is considerably smaller than
that of a Silverson rotor (i. e. a swept blade type), Figure 2.7. This would lead to smaller
radial velocities and greater nominal residence times in the rotor region. A similar
be to
calculation can applied other regions within the mixer (e.g. the stator).
OD°
°Xä
pO ýo
Q
Figure 2.7 A Silverson type rotor and stator configuration
37
Hence, two characteristic length scales can be determined for fluid movement in the rotor
region. That for tangential flow, where the rotor diameter, D, will be the characteristic
length and the rotor tip speed, vtp, the characteristic velocity; and that for radial flow,
where corresponding length scales and velocities will be Dradfaland VRADZAL.
The former results in the impeller Reynolds number (Equation (2.22)) and the latter is
given by:
PD,9-4DJALyaADJ
Re (2.38)
µ
(c) ShearGap
The scale over which effects in the sheargap occur (less than 1 mm) coupled with the
high rotor-speedsmade observationsof the flow pattern in the sheargap region difficult.
The characteristic length scalein this region will be the gap between the rotor and the
S,
stator, although the in
characteristicvelocity this region will remain the rotor tip speed,
v,,,. Therefore the sheargap Reynolds number will be:
Po'1ip
Re8 = (2.39)
It
The shear rate in this region is determined by 8 and vtp and the Reswill provide an
indication of the dissipation in the gap due to viscous shear.
The values for 8 and v,1 can be inserted into Equation 2.37 to estimate the nominal
residencetime in the sheargap.
(d) Stator
38
(d) Stator
Fluid dischargingfrom the rotor was abruptly turned by the upstreamface of the stator
teeth and emerges radially. A fast stream followed the leading edge of the stator and a
region of recirculation accompanied this (Figure 2.8).
The tangential kinetic energy given to the fluid by the rotor was transformed into
pumping power, skin friction losses and turbulent losses. It was suggested that the point
where the abrupt turn in the flow takesplace is where the lossesare greatest.
(]
\'ýi
Figure 2.8 Flow pattern through a stator (Sparks 1996)
A certain degreeof back-mixing may also be evident as the rotor createsan area of low
pressureon the blade following edge,which may draw liquid back into the slot/hole.
39
The various shapes of the holes/slots in the stator screens make it appropriate, for
turbulent flow, to specify a hydraulic mean diameter, Dhyd,as the characteristic length
scale in this region. The hydraulic mean diameter is defined as four times the cross-
sectional area divided by the wetted perimeter. The velocity of the fluid moving radially
through the stator holes/slots can be defined as the bulk flowrate divided by the open area
of the stator screensand will be the velocity of the fluid jets emanating from the screens,
The
vjegs. definition of Reynolds number in the stator region is given by:
PDhyd yjels
Res, = (2.40)
40)
p,o, µ .
The nominal residence time in the stator holes/slots can be estimated by using the
thickness of the screen(the depth of the hole/slot) and vje,s in Equation 2.37.
(e) Volute
Sparks (1996) divided the flow in the volute into three main sections (Figure 2.9).
Recirculating flows were observed behind the rotor (A). The flow directly out from the
stator was radial (B) and the outlet pipe contained swirling flow (C). Again no change
with operating condition was detected.
It is very difficult to characterisethe length scale in this region as the shaft and rotor and
the shapeof the volute will influence the flow characteristic. It is safe to assumethat the
bulk of the flow in this region and the resultant turbulent energy in the volute will be as a
direct result of the jets emanating from the stator holes/slots. It is also safe to assumethat
all of the fluid eventually finds its way to the outlet. Therefore, the characteristic linear
length scale for the flow in the volute can be taken as the distancebetween the volute
wall and the stator, which will be referred to as Dvorure
40
Figure 2.9 Flow pattern in the volute (Sparks, 1996)
Hence,the Reynoldsnumberin the volute can be defined as:
PDvoluteyfets `
Revo, = (2.41)
ute P
Sparks(1996), suggestedthat the overall residencetime distribution charactetisticsof the
mixer were determinedby the flows in the volute. This seemeda reasonablestatementas
he found that the residencetime distribution curveswere similar for a Silverson type
mixer and a toothed type. This is also reasonableas the volume of the volute is much
greaterthan the volume of the inlet and outlet pipes and the rotor-statorregion.
A nominal residencetime of the fluid in this region can be estimatedby dividing the
volume of the volute by the bulk flowrate.
41
2.3.2. Drop Break-up in shear and elongated flow fields
Up to this point drop break-up in turbulent flow has been reviewed, however it is
reasonable to assume that there are regions within the mixer where these theories will not
apply. It has been reported that the assumption of local isotropy is only fulfilled locally
within the mixer. For this reasonthe following sectionwill review the main theories
behind droplet break-upin highly shearedflows.
When one immiscible liquid is added to another the dispersion process can be considered
to occur in stages. Initially the two liquids are at present in bulk as the mixing process
develops the bulk fluid gets broken up into chunks which in turn get broken into smaller
parts or globules. Hinze (1955) provided a summary of the various modes a single drop
(or globules) could break-up. He explained that a globule could split (due to
hydrodynamic forces) in a number of different ways depending on the flow pattern
around them and generalised them into three types (Figure 2.10).
/ý1- ýý--ýý
i
,.,,..y. ý ýr
tý ''
i iý +ý J I
ý
`ýý
Type 1 Type 2 Type 3

LENTICULAR CIGAR-SHAPED BULGY
Figure 2.10 Basic types of globule deformation (Hinze 1955)
42
Type 1 is where the globule is flattened to form an oblate ellipsoid. How the globule is
broken dependson the magnitude of the forces causing deformation. Type 2 occurs as
the globule becomesmore and more elongatedand forms a cylindrical thread which then
breaks into drops. Type 3 occurs as a result of local deformations on the surface of the
globule resulting in small drops breaking off the main body. Hinze (1955) went onto
classify six different flow patternsthat may causedrop break-up, five of which are shown
diagrammatically in Figure 2.11.
abc
Parallel flow Plane hyperbolic Axi-symmetric
Flow hyperbolic flow
,"1
de
Couette Flow Rotating flow
Figure 2.11 Flow patternsthat can causeone of the basic types of globule deformation.
43
The sixth flow pattern that was describedwas an Irregular flow field (f) occurring in
turbulent flow.
It was assumedthat for the rotating flow pattern (e) the fluid rotates bodily. The flow
patterns were also assumedto be large compared to the globule size but still local within
the whole flow field.
It is possible to distinguish betweendeformation by inertial forces and that by viscous

forces. Table 2.3 summariseeachdeformation mode and the particular flow patterns that
may causethem. For it be

example, can seenthat Lenticular deformation can arise if the
is
globule subjectedto inertial or viscous forces produced by parallel, axi-symmetrical
and rotating flow patterns (a, c and e in Figure 2.11).
Table 2.3 Flow patternsthat causeeach deformation mode
Type 1 Type 2 Type 3
Lenticular Cigar-Shaped Bulgy
Inertial forces a, c, e b, d f
Viscous Forces a, c, e b, d -
The globule will break-up if the flow pattern is large enough to sustainthe globule and if
the globule is exposedto this flow pattern for a sufficient amount of time. The type of
deformation and break-up was found to dependon the local flow patterns around the drop
and the physical properties of the liquids.
In turbulent dispersions,inertial forces were thought to be dominant since at relatively

high Reynolds numbers, the local regions of viscous flow were assumedto be small
compared with the size of the drop. In viscous flows the likely deformation modes were
plane hyperbolic and Couette flow, whereas,in turbulent flow it was deemedlikely that
44
Couette,axi-symmetrical and plane hyperbolic flow and perhapsirregular flow were
responsiblefor drop deformation.
drop break-up in in Couette

Karam & Bellinger (1968) observed simple shear flow a
device. They found that drops rarely broke-up into two equal droplets and the majority of
the casesbreak-up resulted in a number of smaller drops. They concluded that under
simple shear flow a uniform droplet size is unlikely and that a range of sizes is more
likely. Figure 2.12 shows a summary of their observations.
DEFORMATION
DEFORMATION VISCOSITY PRIORTO BREAK-UP
PRIOR TO BREAK-UP AFTER BREAK-UP RATIOu/u De
00 03-. 3 60-. 83
. .
C=0 000 3-4 70-. 93
. .
O==O O0°0 005-3 90-97
. .
0000
a0000
epo
00000000000
00000000
0-0-0-0
0 NO BREAK-UP <4 10-. 36
.
NO BREAK-UP >. 005 78
.
Figure 2.12 Break-up of droplets in Couette flow field (Karam & Bellinger, 1968)
Typically, in Couetteflows drops form ligamentsthat break into smaller drops,the
number and size of thesesmaller drops is dependant

on the length of the ligament before
break-up (the longer the ligament the greaterthe numberof smaller drops). In
Extensional flow the drops will deform into a thin filament with dumb bells at eachend
which on break-up will form tow separatedrops and severalsatellite drops.
Taylor (1934) investigateddroplet deformation in the laminar flow regime (simple shear
field) over a rangeof viscosity ratios, }.b/µ,, (where, µdis the dispersedphaseviscosity,
45
and p, is the continuous phaseviscosity) 0.0003 to 20. Single drops approximately 1mm
in diameter were observedto deform and break-up in both Couette and hyperbolic flow.
Taylor formulated a theory of droplet deformation and break-up. In doing so he made the
following assumptions;that the drops are small so that they remain spherical when
deformed, small deformations, no slip at the drop interface and a continuous shearstress
acrossthe interface.
He found that drop break-up occurred at lower shearrates in a hyperbolic flow field
comparedwith the Couette flow. At the highest viscosity ratios no break-up was
observed in the Couette flow. For both types of flow patterns,higher shearrates were
required to deform and break-up drops at the viscosity ratios both considerably higher
and lower than one.
Karam & Bellinger (1968) discussedthe concept of internal circulation within a liquid
drop, which stabilised a droplet when deformed in a simple shearflow. Interfacial
tension acting on a drop under deformation will causethe drop to pinch off and divide.
Internal circulation preventsthis from happeningby building up pressurewithin the drop
against the dimple and pushing outward. Internal circulation was used as an explanation
for why the observeddrop size was approximately three times the calculated size (using
the theories of Taylor, 1934).
There have been numerous studiesthat have furthered the work of Taylor (1934). Karam
& Bellinger (1968) studied single drop break-up in Couette shearflow fields. However,
the most comprehensivework done in this areawas by Grace (1982) who studied the
deformation and break-up of droplets in various simulated shearflow fields, including
Couette and plane hyperbolic flow fields. A wide range of liquids were studied which
enabled µc from 5 to 300 kg/ms, arup to 0.025 N/m and µd/µcfrom 10"6to 950. The
particle Reynolds numbers were in the range from 10"2to 10"6.
46
The following is a summary of the findings of Grace (1982). A number of plots and
correlations were formulated using experimental data from simulated Couette (Rotational
Shearfield) and Extensional flow (plane hyperbolic shearfields).
For small deformations Taylor (1934) showedtheoretically that the deformation at burst,
DB, is equal to EB (referred to by Grace as the reduced shearrequired for break-up). The
drop deformation D equalsthe length of an elongateddrop minus breadth over length
L-B).
plus breadth, (_ EB is defined by Equation 2.42.
L+B
GBrd,lcf(P)
EB _ (2.42)
a
Where, GB is the shearrate (s ), rd is the drop radius,f(p) is a theoretical function of the
viscosity ratio suggestedby Taylor (1934), Equation (2.43). Values forf(p) vary from 1
to 1.2 for viscosity ratios from 0 to infinity.
µd 19p + 16
Viscosity ratio, p= -, f (p) = (2.43)
µ1 16p + 16
Couette Flow (Shear Flow)

Firstly, considering drop deformation and break-up in a uniform Couette shearfield.
Figure 2.13 shows a plot of EB,the ratio of viscous forces and interfacial forces necessary
for drop bursting versus µd/µc. This correlation allows the prediction of GB,required to
break a drop of radius rd, at known µc and cr for a viscosity ratio from 10-6to 3.5 in
uniform rotational shear.
47
00!. Lc a 45.5 poise
1000
pb µc= 502 poise
Eg= 0A jac = 2.815 poise
GgrauCf(p)/y= 400
Viscous Forces Drop breaking by
fracture
Forces 10 main op
at Brust /.
i
Drop breaking by Tip Streaming l
p, 1 tI111
{ö 7 «5 lös i6 2
iö3 1ö icy i 10
Viscosity Ratio, p=µ.
D/µc
Figure 2.13 Effect of viscosity ratio on EB in simple shear flow field (Grace, 1982)
It can be seenthat the viscosity ratio has a significant effect on EB, in Couette shearand a
minimum occurs in the range of viscosity ratios from 0.1 to 1.0. The critical shear
gradient increasessignificantly either side of this range. It was claimed that the
horizontal section below a viscosity ratio 0.1, representstip streaming (small drops that
are broken off from the tips of a drop that assumesa sigmoid shape)and does not
representactual drop break-up. It can be seenthat it is not possible to break drops in a
simple shear field if the viscosity ratio exceeds3.5.
48
Figure 2.13 is similar to that reported by Karam & Bellinger (1968), who also saw a
in
minimum a similar viscosity range. However, Grace showedthat drops could be
broken at viscosity ratios as low as 10"6,whereasKaram & Bellinger (1968) indicated
that this was not possible below 5x 10-3. They also showed a minimum value of reduced
shearof approximately 2 comparedto 0.6 of Grace (1982). Grace (1982) attributed these
differences to limitations in equipment and lower values of µ,, used by Karam and
Bellinger (1968) and also the values of interfacial tension used (Grace used the dynamic
interfacial tension between to pure components).
The data was also plotted in terms of drop draw ratio required for break-up at the critical
sheargradient. The drop draw ratio was defined as the length required for break-up over
the original drop diameter. Figure 2.14 shows a plot of drop draw ratio versus viscosity
ratio. The general correlation is similar to that shown in Figure 2.13, in that a minimum
in
occurs a viscosity ratio range of 0.2 to 1.0 with a rapid increaseeither side of this
minimum. For break-up to occur at the critical value of reduced shear,both a critical
drop draw ratio and a critical time under shearmust be attained.
Figure 2.14 shows the data in terms of reducedburst time, tb, versus viscosity ratio. The
tbü
reduced burst time is a dimensionless term and is represented as
2rdµ,
Figures 2.13 to 2.15 can be used to predict the required critical shear,drop draw ratio and
the required sheartime for break-up over a 106range of µd/µc. However, in most
dispersion applications there will be a range of droplet size distributions and Grace
(1982) discussedthe effects of exceedingthe critical shearfor larger size drops to break
the smaller drops.
49
Drop Draw-Ratio a 45.5 poisecont.- phaseviscosity
Necessaryfor 502 "" it It of
Breakat Ee A 2,815' is to is
100c-
10
105 04 1Ö3 02 10,1 10

löß
Viscosity Ratio, µp/1r.C,dimensionless
Figure 2.14 Effect of viscosity ratio on drop draw ratio necessaryfor break-up at EB
(Grace, 1982)
100 o For 45.5 poise cont.- phase viscosity

U,
to v)
A«
502 ««K«
A«2,815 «««
ý ö
E a, 10 A 8°
ý= E A!
N
i
ý AAo
m
Ao
a
N
m 0.1
10_s toy 164 los 102 10 1 io
Ratio,
Viscosity dimensionless
µD/µc,
Figure 2.15 Effect of viscosity ratio on reducedburst time at EB (Grace, 1982)
50
Grace discussedhis results in terms of the ratio of actual shearover critical shear(E/EB),
his main findings were as follows:
1. As E/EB is increasedthe drop draw ratio (LB/2rd,where LB is the length of the

drop at bursting point) increasesrapidly.
2. The number of drop fragmentsproduced is only dependanton µd/µcwhen E/EB is
equal to one. As E/EB increasesthe number of fragments increasesrapidly

reaching values in the order of 10,000 fragments at E/EB in the order of 100 and
the break-up behaviour becomesmuch more dependanton E/EB.
3. As E/EB increasesbeyond one the reducedburst time decreasesrapidly. This
decreasewas seento be less than proportional to the increasein E/EB implying
that the energy input is increasing quite rapidly.
4. Exceeding the critical sheargradient results in a broader droplet size distribution.
5. Increasing E/EB beyond approximately 4 to 10 times does not result in a
significant reduction in the ratio of fragment over original drop diameter.
The last point (5) above led to the inference that to achieve a large drop size reduction,
be
shearshould applied stage-wise,with sufficient low shear hold-up between stagesto
permit break-up of already deformed drops, in order to minimise energy consumption and
heat generation. It was suggestedthat the product of shearrate and sheartime should be
kept constant with low shearrate and long sheartime at the feed end and high shearrate
and short sheartime at the dischargeend of a dispersion device.
Extensional (Hyperbolic) Flow

Both Grace (1982) and Karam & Bellinger (1968) found a maximum viscosity ratio (3.5
4
and respectively) beyond which drops were not broken in simple shear flow. For this
reason Grace (1982) also looked at single drop break-up in a simulated extensional shear
field.
51
Figure 2.16 shows a plot of shearrequired for drop deformation and break-up, EB, in
extensional flow field versus viscosity ratio. Grace (1982) also included data points from
other workers, which showed good agreementwith his data.
Q
Y
a
d fC7 µß: lip yd =6 dyne/cm
co ETL4 -Roil Data(O
N /C= 45.5 p yd 2-6 dyne /cm
N tO FLC- 5026 p
_p yd = 2-5-5 dyne/cm
MASON4- Ro11 Data A uC= 45-50p
QG y 04.8-7.5dyne/cm
U TAYLOR 4-Roil Data V ps 90-lOOp
.O
U) y = 8-23 dyne/cm
.C 10
_U
oE
ÖO v
Opw 1.0
O
v o
ým o.!
UU
10 4 2
id 10 10 101 100 10' 102
'O
U 103
Viscosity Ratio,
µD/I C
Figure 2.16 Effect of viscosity ratio on EB in hyperbolic flow field (Grace, 1982)
It can be seenin Figure 2.16 that the curve exhibits a minimum at viscosity ratios
between 1 and S. Either side of this minimum the shearrequired to break up drops
increasesslowly with a maximum and minimum viscosity ratio in the order of 900 and
10"5.
52
The value of shear required for drop break-up at the minimum of the curve is 0.2. This
value is a third of that seen for a simple shear field (Figure 2.13) which showed a value of
0.6 at the minimum. Thus, it can be said that for equal shear rates, extensional shear is
more effective for drop break-up than simple shear field (Couette), even at low viscosity
ratios.
It is important to be able to determinethe shearstressand sheartime neededto exceeda
maximum stable drop deformation for drop-break to occur in extensional flow (Figure
2.17). It is important to note that exceedingthis maximum deformation does not ensure
drop break-up but makes it possible with the right combination of shearstressand shear
time.
Continuous Phase. Pale4; lie - 45.S poise

DispersedPhase 0 Silicone200fluids; rp " 4- 6 dynesI cm
O Polybutene20& 32; r0 "2-3 dynesI cm
LO
Max. Stable 6 ParaplexG-41 10 " 1.0 dynesI cm
Droo
Deformation, 0.8
L- BIL«B,
dimensionless 0.6
0.4
0.2
0.0L- 2
103 10 1Ö-1 100 101 102 10,
Dispersedto Continuous PhaseViscosityRatio,µp I uc, dimensionless
Figure 2.17 Maximum stable drop deformation in extensional shearfield versus
viscosity ratio (Grace, 1982).
53
Grace (1982) showed that a critical drop deformation existed below which the removal of
shearwould not result in break-up but would return the drop to its original shape. The
critical drop deformation is defined in Figure 2.18. The critical deformation values are
large comparedwith the maximum stable drop deformation (Figure 2.17).
L00
Drop 0 98
.
Deformation,
D 0.96
L-BI 1+B
0.94
0.92 9/ O
O0 Deformeddrops resumeoriginal
shapeupon removalof shear
0.90
8e Deformeddropsbreak up upon
0. gg removalof shear
0.86
0.01 0.1 1.0 10 100 1000
ViscosityRatio.u01µc
Figure 2.18 Drop deformation for burst at critical shearin extensional shear(Grace,
1982)
Figure 2.19 shows the critical burst time required to ensurebreak-up upon application
and removal of the critical extensional shear. The critical burst time for extensional shear
was shown to be similar to that for simple shear(Figure 2.15).
54
10
Reduced
Critical
BurstTime,
BY 102
2rdG
dimensionless
O Reduced
bursttimevs.viscosityratio
101 E114-ro11data:
O uc"45.5poise;1"4-6dynes ! cm
O pc"45.5poise; T"3-4dyneslcm
Pc"45.5poise, y-LOdynesIcm
100
id-3 1072 1
1Ö 100 101 102
ViscosityRatio.It. 1i4. dimensionless
Figure 2.19 Effect of viscosity ratio on critical burst time for extensional shear(Grace,
1982)
The critical shearrequired for drop break-up was shown to be lower for extensional shear
comparedto simple shear,with equal critical burst times required. This indicated that
under extensional shearfield a lower total energy was required for break-up compared to
simple shearconditions.
2.3.3. Drop break-up in a rotor-stator mixer
The models discussed previously (Section 2.2) have been tested extensively for liquid-
liquid dispersions in stirred vessels (Tavlarides & Stamatoudis, 1981). However, their
application to rotor-stator mixers has been few and far between. This section provides a
review of the limited work performed on rotor-stator mixers and discussesthe application
of the models for the design of rotor-stators for liquid-liquid systems.
55
In the past, many manufacturers and operators have assumedthat, due to the complex
flow patterns and the high rotational speeds of the rotor-stator mixer, the device is
operating within the turbulent regime. However, Levine (1994) states that this
assumption is often misguided and that the machine is often operating in the laminar
regime (when mixing highly viscous fluids). This may explain the discrepancies in
performance of these types of mixers when scaling up from laboratory trials to full-scale
production.
This reasoning highlights the importance of being able to predict whether the machine is
operating in laminar or turbulent regimes. The identification of scale-up parametersthat

may influence the transition between the two regimes is of paramount importance to
designers.
Levine (1994) went further to suggest that in turbulent flow the strong dependency of
(i.
power e. energy dissipation) on rotational speed leads to the deduction that volume
throughputs and geometric variables would not have a significant influence on the mixing
performance of the machine. Therefore, the most influential parameter when scaling-up
in the turbulent regime is the rotational speed (N). This opinion is echoed by those of
Kroezen and Wassink (1987), who investigated the use of rotor-stators as foam mixers.
They found N to be the parameterthat had the greatesteffect on bubble size distribution.
In laminar flow, viscous shear would be the more prominent break-up mechanism. The
longer a drop is exposedto the region of high shearthe greater the likelihood of the drop
elongating and splitting. Schubert (1997) found that for equal magnitudes of energy
dissipation per unit volume, viscous shear caused the greatest droplet disruption. In
viscous shear, break-up is controlled by the shearrate (Equation (2.23)) and the residence
time in a high shear zone is dependenton the geometry of the mixer. Therefore, for the
purposes of scaling up mixers that operate in the laminar regime, geometric similarity as
well as rotational speedwould have to be considered.
56
A method of determining the flow regime is to determine the dependency of the power
draw of the unit on rotational speed. If the power is found to be proportional to rotational
speed raised to the power of 3, then turbulent conditions will apply. If the power is
to
related rotational speedraised to the power of 2, then the flow will be laminar.
Averbukh et al (1988) presentedthe following correlation, to determine a mean droplet
size in an in-line rotor-stator mixer:
0.15 0
yc .7
0135 l
da, = CD ' (2.44)
Pc S o.25N0.8 08
Where, «, is the dispersed phase volume fraction, or is the interfacial tension, v,, is the
kinematic viscosity of the continuous phase, Q is the volume throughput, 8 is the shear
gap, N is the rotor speedand C is a constant determined from experiments.
The results calculated using Equation (2.31) compared favourably with experimental
results. They described the unit as a centrifugal pump with a perforated ring mounted on
the impeller and which rotated in a secondfixed perforated ring. Experimental test were
performed using three types of oil in water emulsions and also three different rotor
diameters (with variable shear gaps and number of slits, and constant slot dimensions).
This enabled them to quantify the effects of various physical and geometrical parameters
indicated in Equation (2.42). However, it was highlighted that this correlation was
applicable only within the limits of their experimental conditions. No further details of
experimental work were given.
It is worth noting that the parameters show similar dependencies to the correlation
derived by Chen and Middleman (1967) for drop break-up in the viscous sub-range.
Equation (2.31) also includes the effects of Q, c and 8 on the mean diameter.
57
2.3.4. Turbulent energy dissipation rates, c
Schubert (1997) compared different mixers based on their energy dissipation rates, he
defined the energy dissipation as the power per unit volume:
'
Flurd
(2.45)
V
Unfortunately, a definition of the volume was not provided and it is assumed that V
representsthe volume of the whole mixer.
The correlations discussed in Section 2.2 all show different dependencies of s on the
system variables. Bourne and Garcias-Rosas (1986) studied the micro-mixing
performance (diazo-coupling experiments) in an in-line X-20 Ystral mixer (a toothed

type). They found that the product distribution decreasedwith rotor speedindicating an
increasein turbulent energy dissipation rates.
They estimated energy dissipation rates of between 5 and 700 W/kg and correlated their
data with:
8=4x10-10N3 (2.46)
Sparks (1996) assumedthat the power was supplied to the fluid in a small region of the
rotor-stator (that included the rotor and stator teeth and the shear gap). He further
assumed that a proportion, a third of this power was dissipated through turbulence. This
estimation for the turbulence energy dissipation was considered a conservative guess
compared to the values quoted for static mixers (80%). However, it was deemed
reasonabledue to the extra frictional lossesthat may arise from the larger surface area in
a rotor-stator mixer.
58
Values in the order 1 kW/kg were quoted for the mixer as a whole. If the above
assumptions were made and the volume of the rotor-stator region was considered this
value increasedby 5 times.
However, values for turbulent energy dissipation rates were also estimated from Diazo-
coupling experiments and only a small proportion of the power supplied to the fluid by
the rotor in the rotor-stator region was found to be dissipated as turbulent energy. It was
suggestedthat bypassing occurred through the rotor-stator region, resulting in significant

in-homogeneity in the turbulence field.
This agrees with findings by Kroezen & Wassink (1987) who used a rotor-stator mixer
for producing foams (gas-liquid dispersions, with dispersedphaseconcentrations greater
than 50%). They reported that the shear rate was highest during rotor-stator interaction,
and for longer residencetimes, an equilibrium bubble size was obtained.
Sparks (1996) proposed two methods of ensuring that a higher proportion of fluid
experienced the higher dissipation rates. Firstly, the rotor-stator can be designed to
reduce the inhomogeneity. The Silverson design that was used had a lower nominal
radial velocity than the toothed designs, which may have reduced short-circuiting.
Secondly, the process fluid can be recycled through the device, in semi-batch mode
(Figure 2.2). A multi-stage rotor-stator mixer (discussed in Section 2.1.2) is an
alternative way of implementing the second solution.
Pedrocchi & Widmer (1989) found that up to 50 passesthrough the rotor-stator unit was
required to reach an equilibrium drop size distribution. The number of passes will be
dependenton the design of the mixer and also the physical properties of the liquid-liquid
system being used. A degree of caution is advised when operating in this mode, as
continual recirculation would result in a temperaturerise, which would effect the physical
properties of the fluids involved.
59
2.3.5. Energy dissipation and droplet size
It is desirable to be able to provide the minimum amount of energy needed to form a
stable emulsion with desired characteristics.Power is the principle operating cost for any
agitation system.
It can be observed from the models discussed previously that in the generation of
dispersions the principle parameter that can be varied to influence the final dispersion
properties is total energy input. The mean drop diameter can be related to the energy
dissipation by Equation (2.47):
d
mean
a£ (2.47)
The constant b is dependenton the mixer geometry (stator design). Values of b for rotor-
stator mixers were evaluated to be between to -0.76 (Pedrocchi & Widmer, 1988)
-0.54
and -0.2 to -0.4 in turbulent flow (Schubert, 1997).
According to the Kolmogorov-Hinze theory the exponent b should be -0.4 (Equation

2.13), which is close to the values quoted by Schubert. Unfortunately neither author
provides any further details of the stator dimensions and design or reasons as to why the
values differ from those according to the theory of Kolmogorov-Hinze.
Calabreseet al (2000) found the value for exponent b to be -0.48, which is close to -0.5
from the Shinnar model for break-up due to viscous sub-range eddies (Equation (2.25)).
They also found that their mean drop size data correlation, Equation (2.35), closely
matched Equation (2.22), (Chen and Middleman, 1967). Consequently, they suggested
that the break-up mechanisms for drop diameters near kk were due to eddies in the
inertial sub-rangeor viscous sub-range.
d32
= 0.040We-'*" (2.48)
D
60
It must be noted, however, that the work performed by Pedrocchi and Widmer (1988) and
Calabreseet al (2000) were done using a laboratory scale batch rotor-stator mixers.
Saadevandi& Zakin (1996) studied the affect of mixing intensity and time on the mean
droplet size. It was observed that for a given time, an increase in mixing intensity
produces a smaller mean droplet size. However, it was also observed that beyond a
critical energy input the mean droplet size started to increase. An explanation for this
phenomenon was not given. However, it is possible that an increase in turbulent energy
dissipation would increase the probability of drops colliding and coalescing. Another
factor to consider is that as drops become smaller, the total concentration of drops within
a dispersion increases(i. e. greater surface area), this would also increasethe likelihood of
collision.
2.3.6. Turbulent break-up in the sheargap
It is commonly believed that the power dissipated in the sheargap is responsible for drop
break-up. Pedrocchi and Widmer (1988) and later Calabrese et al (2000), found that
varying the dimension of the shear gap (width and height) had little affect on the volume
specific energy dissipation and droplet size. Both workers argued that the geometry of
the inlet and the outlet of the sheargap had a greater affect on s and drop size. Calabrese
et al (2000) postulated that the stator holes (slots) and the resulting turbulent jets were
mainly responsible for drop break-up.
It was suggestedthat only a small percentageof the total energy input was dissipated in
the sheargap and the assumptionof isotropic turbulent flow is only fulfilled locally in the
mixer.
Bourne and Garcias-Rosas(1986) looked at rotor-stator mixers for rapid micromixing.

They performed diazo-coupling experiments and found that removal of the stator had
little or no effect on the product distribution. Two possible explanations were given for
this observation. Firstly, it is possible that the turbulent energy dissipation rate in the
61
shear gap and the stator region were not significantly higher than in the rotor region.
Alternatively, the reaction may have been complete before the shear gap and stator was
reached. The findings of Calabreseet al (2000) and Pedrocchi and Widmer (1988) would
favour the former explanation.
2.4. Power consumption in a rotor-stator mixer
In-line rotor-stator mixers are analogousto centrifugal pumps and an in-line rotor-stator
mixer is often referred to as an inefficient pump. Nikiforov et al (1990), modified a

centrifugal pump by attaching two perforated rings, one fixed to the pump impeller
(simulating a rotor), the other fixed permanently in the volute to emulate a stator.
Sparks (1996) identified the main differences between a rotor-stator mixer and
centrifugal pump as follows:
(a) In a rotor-stator mixer the gap between the rotor and the stator, the sheargap, is very
narrow. The impeller in a centrifugal pump rotates with a large clearancefrom the
stationary volute wall.
(b) The pump impellers have swept blades to transmit momentum to the pumped fluid,
while minimising separationand recirculation losses(i. e. minimise turbulence

generation). The rotor in a rotor-stator mixer is not usually optimised for pumping.
(c) The volute (or impeller dischargecollection region) in a rotor-stator mixer is not
optimised. Pumps usually have a spiral volute to minimise head loss through gradual
pressurerecovery.
2.4.1. Total power supplied to fluid
Sparks (1996) presenteda power balance for pumps (Stepanoff, 1957 and Anderson,
1980):
PSHAFT
= PPUW+ PASCH
+ PDJSC
+ PJEAK
+ PHYDPAUMC (2.49)
62
The individual componentsare discussedin the following sections.
PSHAFT
2.4.2. Shaft power -
This parameter is important to the designer so that the correct motor can be specified.
Sparks (1996) statedthat this is analogousto the shaft power in a stirred tank, which for a
fully baffled stirred tank is given as (Rushton et al, 1950):
PSHAFT
= ý'oPN3D5 (2.50)
Po is the power number and differs with impeller type.
Sparks (1996) investigated the effect of rotor-stator design on the shaft power for toothed
and Silverson design in the

mixers, all cases shaft power increased with both shaft speed
and flowrate. He found the N
exponentson and D to be 2 and values for Poto be
approximately 10, for all rotor-stator geometries.
The exponentson N and D were explained by the fact that the flow visualisation showed
the fluid in the rotor region followed the rotor teeth, at a speedvt;p (=AND). Therefore,
the kinetic energy of a fluid element at the rotor tip, of massm (=pQ), could be expressed
as Equation (2.51).
1iz
Pflura=2nN D2Q (2.51)
The effect of the number of teeth on the rotor was found to have little or no effect on
shaft power. The effect of stator design was also found to have little effect on shaft
power.
63
2500
2000
E s 3000 rpm
i 2500 rpm
1500
. 2000 rpm
>1500 rpm
1000
144 e 1000 rpm
con 500
0 500 rpm
0
012345
Flow rate (litres/sec)
Figure 2.20 Shaft power versus flow rate for a range of shaft speedsfor a Silverson
mixer (Sparks, 1996)
Shaft power was determined by measuring the shaft torque and the rotational speedand
calculated using:
PSHAFT 2rNA
= (2.52)
Where, A is the shaft torque (Nm).
64
2.4.3. Pumping power - PpUMP
A designer would require the magnitude of pumping power in order to assesswhether a
supplementalpump is required.
The pumping power output is defined as:
QiP
PUMP = (2.53)
Sparks (1996) found that the Silverson mixer had a greaterpumping efficiency than the
toothed designs,as the Silverson blades resembledthe impeller blades in a centrifugal
pump more closely.
Expectedly, the pumping efficiencies obtained from the rotor-stator mixers were low
compared to that for centrifugal pumps. The rotor-stators inefficient pumping was
attributed to the fact it had not been optimised for pressurerecovery process.
In the rotor-stator, the flow straightening effect of the stator produces some pressure
If
recovery. significant pumping was required, this process does not occur without large
losses,principally to turbulence and friction losses. This is becausethe fluid being
dischargedtangentially by the rotor, impacts onto and then exits the stator in a radial
direction.
Sparks (1996) found the pressurerecovery processto be sensitive to rotor-stator

interaction as the sheargap thickness was reduced. It was suggestedthat if the pressure
recovery step were sensitive to rotor-stator interaction then the turbulence generating
processwould be as well.
65
2500
2000
F
1500 -ý- r/18/50/119.6
ö0 r/18/50/121.45
01.1000
-A r/18/50/123.34
cl
4
500
012345
Flow rate (litres/sec)
Figure 2.21 Shaft power versus flow rate for a range of sheargap thickness for
a Silverson mixer (Sparks, 1996)
The effect of stator design on pumping performance was also discussedand the results
from the toothed design suggestedthat the pumping efficiency was more dependentupon
stator open areathan rotor-stator interaction. The superior pumping efficiency of the
Silverson mixer, which had an open area of approximately 35%, further, supportedthis
deduction.
It was also reported that pressurerecovery was affected by rotor-stator interaction
significantly which also implied that turbulence and skin friction lossesmust be affected
as well. This suggeststhat stators might be designedto optimise for turbulence
production.
66
2.4.4. Mechanical losses- PMECH
The mechanical lossesare the lossesdue to friction in the sealsand bearings (Stepanoff,
1957). Sparks (1996) found that the mechanical lossesaccountedfor approximately 2.5%
of the shaft power.
2.4.5. Disc friction losses- PDISC
Disc friction lossesare causedby the frictional drag on the solid surfacesof the rotor and
are analogousto the skin friction lossesin motionless mixers. These lossestend to be
significant in pumps. Sparks (1996) used the following equation to predict disc friction
loss (Karassik et al, 1986):
P = N3D5 (2.54)
DISC
2.4.6. Leakage (recirculation) losses - PLEAK
The leakage loss can have two meanings,firstly (and most obviously) it can refer to
leakagethrough seals. Secondly it can mean loss due to the fluid not passing through in
the right direction. For example when fluid passesback from the high-pressure
(downstream) side of the impeller to the low-pressure(upstream) side.
2.4.7. Hydraulic losses- PHYDRAULrc
Centrifugal pumps are very poor mixers (Hearn, 1992) and hydraulic losses(including
turbulent losses) in
are very small relation to the energy balance at or near the best
efficiency point.
When considering the design of a rotor-stator mixer the rotor should be designedto
increasehydraulic lossesas this will increaseturbulent energy dissipation. An opposite
approach,to that of centrifugal pump design, be

should adopted (Sparks 1996).
67
Sparks (1996) did not measurehydraulic and leakagelosses,but instead assumedthat the
remainder of the power balancethat was not accountedfor was the hydraulic and leakage
loss combined.
BLEAK + PHYDR4ULIC =
PSHAFT
-
(PPUMP + PMECH + PD/SC) (2.55)
It was found that approximately 70% of the shaft power input was dissipated through
hydraulic and leakagelosses.
100
90
a 80
c
N
70
Cl. 60
v
50
0 40
N
rn
30
c
aý
H 20
d '
a pumpingpoweroutput
10
0 45
Flow rate (litres! second)
Figure 2.22 Overall power balancefor a Silverson Unit (Sparks, 1996).
The energy balance (Figure 2.22) showed that the combination of hydraulic and leakage
power was relatively insensitive to small changes in flowrate and appeared to be more
dependanton shaft speed.
68
2.5. Droplet size measurement techniques
There are a number of different droplet size measurementtechniquesreported in

literature. The following is a brief review of the onesmost commonly used. The details
of the measurementand sampling technique employed in this researchare discussedin
the next chapter.
Measurementtechniquescan be divided into two categories,in-line and off-line. The in-

line techniques are preferred for the majority of liquid-liquid systems,as off-line
techniquesrequire the need for sampling, giving rise to problems associatedwith
coalescenceand sample disruption.
2.5.1. On-line drop size measurementtechniques
The main on-line drop size measuringtechnique that has been reported throughout
literature is image capturing (Middleman, 1974; Brown & Pitt, 1974; Chen & Libby,
1978; Pacek et al, 1994; Pacek & Nienow, 1995, Calabreseet al, 2000). Despite the
popularity of this technique for use in liquid-liquid work it has a number of limitations:
0 Only suitable for low dispersedphaseconcentrations.

0 It is not suitable for opaque fluids.
" Cannot accurately measuredroplets below 20 µm, this makes it inapplicable for
measuring small droplet sizes (d< i1,ß).

" Difficulty in getting enough light to take pictures in small diameter pipelines.
" Data acquisition is time consuming
" The results are operator dependent(as the manual measurementof drop sizes is
required).
Another technique reported to have been used on-line, is the measurementof electrical
conductivity, Coulter counter, (Wachtel and La Mer, 1961; Becher, 1964, Saadevandi and
Zakin, 1996). The dispersion is forced through a small aperturebetween two electrodes
and drop sizes are measuredby changesin the conductivity of the continuous phase.
69
Drops can be measuredin the range of 10 to 250 microns. However, this technique is
only suitable for dilute systemsto prevent blockage of the apertureand also to lower the
probability of two drops passingthrough at the sametime. This technique also requires
the addition of an electrolyte to enhanceconductivity, which may effect the physical
properties of the system and will affect the likelihood of coalescenceoccurring.
2.5.2. Off-line drop size measurementtechniques
There are numerous off-line techniquesavailable. The most common are as follows:
" Light scattering (Averbukh et al, 1988): This techniques utilises the relationship
between droplet size and the degree of light that each size scatters (the smaller the
drop the greater the scatter). This method is simple to operate and there are
instruments available on the market that can make rapid measurementsover a wide
size range. However, they are not suitable for opaqueor concentratedsystems.
" Laser diffraction (E1-Hamouzet al, 1988, Pedrocchi and Widmer 1988; Streiff et al,
1997): This technique is similar to light scattering, however, the degree of diffraction
is related to droplet size. The smaller the droplet the greater the diffraction. Its
advantages are it is a rapid measurement device and can measure droplets from
1000µm down to sub-micron sizes. However, instruments that are currently available
on the market require the need for considerable dilution. There are several
commercially available instruments, the Malvern Mastersizer being the most

common.
9 Ultrasound: This is a relatively new technique that utilises the relationship between
the attenuation of sound as a function of frequency and droplet size. It works on the
principle that different sizes attenuate sound at different frequencies. This method is
suitable for high dispersedphase concentrations and handles a wide size range (0.01-
1000 µm) (Clements, 1996). However, due to it being a relative new technology,
instruments on the market (Malvern Mastersizer) tend to be very expensive and its in-
line application is still very much in the development stages.
70
2.5.3. Technique selection
It is apparent that research into liquid-liquid systems has been limited to dilute systems
by the constraints imposed by the drop size measurementtechnique. The measurement
technique for use in this researchmust incorporate the following important factors:
1. Measurementof drop sizes in the order of the Kolmogorov micro-scale and below.
2. The measurementof droplets in high dispersedphaseconcentrations.
3. The technique must show good repeatability.
4. Availability
The droplet sizes generatedby an in-line rotor-stator mixer are expected to be too small
for measurementusing video techniques. Semi-in-line methods such as laser diffraction
(Malvern Mastersizer) and electrical conductivity (Coulter Counter) are limited to low (D.
However, despite it's limitations the Malvern Mastersizer (laser diffraction) is the only
available technique that can successfully measure the small droplet sizes that are
expected from the rotor-stator mixer (<10µm). This is the most important criterion for
work with high-energy dissipation rates and the validation of mechanistic models for d<
Ak. As this technique requires considerabledilution at high 1 and is predominantly off-
line, a sample has to be extracted and diluted for measurementto be possible.
The sampling procedure can be validated against an in-line video capturing technique that
has been developed in-house at BHR group. A detailed description of the two techniques
used in this is
work given in chapter 3.
2.6. Conclusions
Rotor-stator mixers are used in a wide range of applications and are suitable for processes
requiring fast disintegration, homogenisation, solubilisation, emulsification and

dispersion. Despite their popularity the literature review reveals that there is a deficiency
71
in literature openly available specifically regarding correlations of drop size with
operating parameters.
The work that has been done discussesthe relationship between droplet size and energy
dissipation. It has been reported that the mean drop diameter was found to be a power
function of 8.
It is uncertain as to whether the assumption of isotropic turbulence is applicable in rotor-
stator mixers. A limited amount of work has been done on investigating the main break-
up mechanisms in a rotor-stator mixer. It has been suggested that more than one
mechanism is responsible for droplet size reduction. Various definitions can be

formulated for different regions in the mixer which can be used to determine where in the
rotor-stator the flow exhibits turbulent or laminar characteristics.
It is widely reported that energy dissipation in the sheargap is small compared to that due
to rotor-stator interactions. However, little is known about the distribution of energy
dissipation within the mixer. It is clear that the rotational speedof the mixer is the main
parameter that effects droplet size. This is is

expected,as e a direct function of N.
The general consensusis that scale-up should be basedon energy dissipation rather than
tip speed. The literature is mainly concerned with the small-scale laboratory devices.
There has been little work done on industrial sized unit.
Although theories describing the forces causing and resisting break-up is simplified to
inviscid dilute systems,they can be extendedto high values of 1 if the system studied is
non-coalescing (but still inviscid).
72
CHAPTER 3. Goals, Aims and Objectives
The literature review revealed that the level of understanding of liquid-liquid dispersion
formation from rotor-stator mixers is not as great as for stirred tanks or static mixers.
Therefore, the main aim of this project was to quantify the behaviour of an industrial
sized in-line rotor-stator mixer with regards to its liquid-liquid performance. This
research focused on one liquid-liquid system to isolate the influence of the geometrical
and mechanical features of the mixer on droplet size. To achieve the main goal, the
following objectives were specified:
(a) Design and construction of a liquid-liquid rig that incorporates suitable

instrumentation and hygienic fittings for easeof cleaning.
(b) Identify and validate a suitable droplet size measurementand sampling technique.
(c) Determine the effects of rotor speed (N), flowrate (Q), rotor diameter (D), stator
geometry and dispersedphaseconcentration (1) on droplet size.
(d) Ascertain the effect of stator design on the droplet size distribution
(e) Develop a technique for measuring power consumption in the rotor-stator mixer.
(f) Explore the relationship between droplet size and turbulent energy dissipation rate.
(g) Analyse the results of the experimental work and correlate the data in terms of dmean.
(h) Attempt to determine the main stress mechanism responsible for drop break-up
73
CHAPTER 4. Experimental Equipment, Materials and Methods
In order to achieve the goals of this research,it was necessaryto construct a liquid-liquid
contactor rig. This rig was used to perform the dispersion tests and enabledthe operating
conditions to be varied easily and rapidly. This meant for a single run, a number of
different conditions could be investigated.
A Malvern Mastersizer was used to measurethe droplet sizes. This technique required
sampling and considerabledilution. A continuous sampling arrangementwas developed

to overcome problems associatedwith extracting a sampleand diluting. This sampling
validation was validated against an image capturing technique.
A separaterig was set-up to perform the power measurementson the in-line rotor-stator
mixer.
Details of the equipment used including the methodologies are presentedin the following
sections.
4.1. Experimental Rig and Instrumentation
An experimental rig was constructedto perform the tests on the in-line rotor-stator mixer
(Plate 4.1). Details of the operating proceduresand the namesand addressesof suppliers
can be found in Appendix 1 and 2 respectively. The rig design enabledthe flowrates,
dispersedphaseconcentration and rotational speedof the mixer to be varied quickly and
easily to investigate their effect on the droplet size distribution.
The rig was designedby N Thapar (HILINE Consortium, BHR Group Ltd) and
constructed with the help of Silverson Machines Ltd. The rig was designedwith the
intention that it could be used to test the performance of static mixers as well as in-line
rotor-stator mixers. For this reason it had to be easily interchangeablebetween the two
types of mixers and easy to clean.
74
4.1.1. Experimental Rig
A schematicdiagram of the experimental rig is shown in Figure 4.1 and a general view
shown in Plate 4.1. An in-line rotor-stator mixer (basedon a Silverson 425 LSM unit,
Appendix 2) was fitted in a flow loop betweentwo feed tanks (aqueousand organic) and
two settling/wastetanks, eachwith a capacity of 1m3.
Figure 4.1 A flow diagram of the test rig
75
Plate 4.1 Photograph of test rig (Front View)
76
77
The organic feed was introduced to the aqueousline one pipe diameter up-stream of the
rotor-stator mixer via a manifold (a T-piece). Sampleswere extracted one pipe diameter
downstreamof the mixer (Section 4.4). The two liquids were supplied to the mixers using
centrifugal pumps, one for the continuous phaseand two (high flow and low flow) for the
dispersedphase,this enableda wide range of dispersedphaseconcentrationsto be
investigated. The pumps were controlled using a flow controller so that a number of
different c and Q to be investigated in a single run. This enabledrapid measurements
and a large quantity of drop measurements.
An inherent problem with working on industrial sized units is the waste product that is
generated. The dispersedphasewas allowed to settle in the settling tanks and then
recycled. Any waste water was pumped into a separatestoragevessel (along with waste
organic phase)which was periodically emptied and taken away for incineration. This
increasedthe cost of the operation of the rig.
4.1.2. Materials of Construction
The feed and settling tanks were constructed from stainless steel and all the pipe work
was hygienic stainless steel (Outer Diameter of 25.4 mm). The fittings were hygienic tri-
clover clamps. This enabled the rig to be easily dismantled for cleaning purposes. The
tri-clover gaskets were made out of Medium Density Ethylene Propylene Rubber
(EPDM) on the aqueous side and Viton on the organic side.
4.1.3. Flowrate and 4 control
The organic and the aqueousfeeds were pumped to the in-line mixer using centrifugal
pumps. The aqueousfeed was pumped using a Grundfos CRN2-40 (1.5 kW) which was
capable of delivering flowrates between 0.4-2.7 m3/h. A similar pump was used on the
organic feed in to
addition a smaller one, Grundfos CRN8- 40 (0.55 kW) that enabled
lower flowrates between 0.14-0.4 m3/h. This resulted in a wide range of flowrates (0.14-
2.7 m3/h) on the organic line and thus a wider range of (D.
78
The desired 4 was achieved by maintaining the flowrates of the organic and aqueous
phasesin the correct proportions. The flowrates were measuredusing Quadrina Q-flo
turbine meters (types QBL/1OB/EP1 & QEL/1OB/EP1)all of which gave an error of +/-
0.5%. The flow was controlled using a feedback loop from the flow meters to the pump
motor inverters via a Nixon Instrumentation digital flow controller 98RC that controlled
the speedof the pumps. The flow controller was usedto set the desired flowrates and to
selectthe required organic pump.
A gate valve was placed downstreamof the rotor-stator mixer to throttle its pumping
effect, thus preventing cavitation and maintaining a constant flowrate. The recycle loops
were present on the feed lines for use when commissioning the flow transducersand
testing the pumps. The pressureddifference acrossthe mixer was measuredusing a
Druck PTX 120/WL pressuretransducerconnectedto fittings located 1 pipe diameter
either side of the mixer.
The material in the settling tanks was left to separatefor at least 24 hours before
recycling the organic phase. The water phase (plus any remaining organic) was pumped
out to waste.
4.1.4. Rotor-stator mixer design
Tests were performed on a Silverson 425 LS unit that incorporated a motor specially
to
adapted produce rotational speedsfrom 3000 to 12000 rpm. A Vacon inverter (CXS
range) was used to drive the motor.
Four different stator designswere investigated,all of which were standardSilverson

designs (Figure 4.2).
79
Figure 4.2 StandardSilverson rotor and stators.Clockwise from top left;
Rotor; Emulsor screen;General purpose disintegrating head;
Slotted Head; Squarehole high shearscreen.
The SquareHole High Shear Screen(SQHHSS) and the Emulsor Screen(ES) are the two
favoured stator designs for liquid-liquid mixing. However, tests were also performed on
the Slotted Head (SH) and the General PurposeDisintegrating Head (GPDH) to observe
the effects of geometry on droplet size.
The dimensions of the various rotors and statorsstudied are summarisedin Table 4.1.
The dimensions listed all representstandardindustrial scale Silverson designs. The
80
clearanceis generally made as small as possible, within the limitations of the machinery
and material used.
Table 4.1 Dimensions of the rotors and statorsinvestigated
Rotor Size Stator Type' Stator I. D. Clearance Hole/Slot Open Area
(mm) (mm) (mm) size (mm) (%)
69.85 ES 70.80 0.475 1.40 25.9

SQHHSS 70.10 0.125 2.38 25.5
50.8 ES 51.05 0.125 1.40 31.7
SQHHSS 51.05 0.125 2.38 31.5
SH 51.05 0.125 113.24 33.0
GPDH 51.05 0.125 9.53 35.0
The volume of the volute was not changedwith eachrotor scale. It could be argued that
this does not representa true reflection on scale, as the liquid volume in the volute
decreasesas the rotor-stator size increases. However, discussionspresentedin Section 2
suggestedthat the areaof maximum energy dissipation might be in the rotor-stator
region. The differences in the volume were small and consequentlythe affect of these
differences on the flow in the volute would also be small.
81
4.2. Experimental materials
4.2.1. Immiscible liquids
Odourlesskerosene,marketed as Alcosol D70, was used as the organic (dispersed)phase.

This material is kerosene(Cl 1-C12 hydrocarbon) with all its light aromatic constituents
removed, (the name and addressof the supplier, Alcohols Limited, can be found in
Appendix 2). The main reasonsbehind this selection are that it is relatively cheap and
safe.
The sensitivity of liquid-liquid systemsto impurities dictated the quality of the
continuous phaseto be as pure as possible. For this reasonthe liquid-liquid rig utilised
de-ionised water. Table 4.2 lists the physical properties of Alcosol D70 and water.
Table 4.2 Physical properties of Alcosol D70 and water
Kerosene Water
Viscosity (kg/ms) 0.002 0.001
Density (kg/m) 800 1000
Interfacial Tension (Portingell, 1989) 0.046 N/m
Absorption (Malvern, 1997) 0.001 -

Refractive Index (Malvern, 1997) 1.435 1.33
The viscosity of the kerosenewas measuredusing a bob and cup type rheometer
(Contraves Rheomoat 115). The in Table 4.2 5s 1
value correspondsto a shearrate of at
20°C. The temperaturerise from a single passthrough an in-line rotor-stator mixer was
very small (typically in the order of °C).

10"3 The viscosity of water, from steam tables,
82
decreasesfrom 0.001 kg/ms at 20°C to 0.009 kg/ms at 25°C. Therefore it was assumed
that the instantaneouseffect on heating on the viscosity of the two liquids was negligible.
The interfacial tension betweenkeroseneand water was evaluatedusing the pendant drop
method, (Portingell, 1989). This was measuredeachtime fresh fluids were addedto the
test rig. Also, spot checks were performed on material that was recycled. The interfacial
tension did not vary more than 5% over the courseof the experiments.
4.2.2. Kerosene-WaterStability
Some initial bench top experimentswere performed to investigate the stability of the
kerosene/waterdispersion. These experimentswere performed in a glass vessel fitted
with a six blade Rushton turbine (the impeller speedbeing the equivalent of rotor speed)
using relatively small volumes of dispersion in to

order ensurea high power per unit
volume.
The dispersion was generatedusing a vessel (5.43 litre) fitted with a six bladed Rushton
turbine (Figure 4.3) with a rotational speedof 1000 rpm.
A range of dispersedphaseconcentrationswere investigated (0.05 - 5%). The
experiments were performed in a water bath and were run for 60 minutes to ensure
droplet size equilibrium had been attained.
Sampleswere extracted from the vessel and placed onto glass slides and viewed under a
microscope. Various sampling strategieswere employed to assesstheir effects on the

droplet size distribution.
A plastic 10 ml pipette with the end choppedoff to increasethe diameter of the tube was
to to
used extract samples observeany differences in droplet size distribution. A syringe
was also used with some 2mm Marprene tubing attached. The length of time the sample
83
remained in the pipette was also varied. This would give an idea of the effects of
sampling time on the droplet size distributions.
The results of these experiments were then used to develop a continuous sampling
technique on the main liquid-liquid rig. The results and observation of these tests are
discussed in Chapter 6.
Sampling Point
Volume =
5.43 litres
affle
6 Blade Rushton
T/3=0.6
0.19 m
Figure 4.3 Stirred vessel used to create dispersions
84
4.3. Droplet Size Measurement
4.3.1. Malvern Mastersizer
A brief description of the principles of laser diffraction is given in Appendix 3. In this
section the various elementsthat make up a practical unit are discussedand the general
operation is explained.
The Malvern Mastersizer was selectedto measurethe droplet size distributions from the
in-line rotor-stator mixer. A Mastersizer 2000 was used in this researchwhich was
supplied courtesy of Silverson Machines and also the EPSRC Loan Pool, (Appendix 2).
The unit from the loan pool was used extensively in the sampling validation work. The
results from this unit were then compared (for a similar run) to the Silverson unit to
ensurethat there was no significant differences in the different units. Comparisonswere
also made on standardlatex particles and betweenthe results obtained from the two units
were almost identical. The Silverson unit was then used for all of the performance tests
performed.
The main constituents of a laser diffraction device for measuring particle size are shown
in Figure 4.4:
(a) A light source;

(b) A meansof introducing the particles into the light;
(c) A method of measuringthe light at different scattering angles;
(d) A method of converting the scatterpattern to a size distribution.
The Mastersizer can be broken down into three constituents,the optical unit; sampling
unit and the computer system. The is to

optical unit used collect raw data that is then
processedby the software in the computer to give drop sizes. The sample accessoriesare
to
used prepare the sampleand transfer it to the optical unit
85
Single Unit
SampleCell
Detector
He - Ne Laser
Obscuration
Fourier
Monitor
Transform
Analyser Beam Lens
Figure 4.4 A Schematicdiagram of a Malvern Mastersizer
The optical unit incorporatesthree modules that correspondto points (a) - (c) mentioned
above:
The Transmitter
The transmitter contains the laser and electronics that generatethe laser beam. The
Mastersizer usesa Helium-Neon laser to provide a monochromatic light source. The
laser beam is passedthrough a processingunit that comprisesof a spatial filter and a
beam expanderthat producesan extendedand collimated beam. The expandedbeam is
then transmitted through the sample cell.
TheSampleCell
The sample area is located betweenthe transmitter and the receiver and is where the laser
is passedthrough the sample.The Mastersizer is supplied with various sampling
accessoriesthat allow it to measureboth wet and dry samples. The accessoriesare

designedto prepare and deliver the sampleto the optical unit for measurement.
86
The accessory that came with the Mastersizer 2000 was the Hydro 2000G. The Hydro
2000G has a1 litre vessel that contains a stirrer (to keep the sample in suspension) and a
pump that supplies the flow cell with sample. However, a continuous dilution technique,
Section 4.4, was developed that avoided the need for this accessory and the Hydro 2000G
was only used for cleaning purposes (flushing the cell) and not for introducing the sample
to the cell.
A schematic diagram of the flow cell and the Hydro 2000G arrangement is shown in
Figure 4.5. During the measurement process V2 was closed and VI was opened to allow
the diluted sample through the flow cell. After the sample had passed through the flow
it
cell was directed straight to waste by keeping V3 open.
To waste
)le line
tator
Figure 4.5 A schematic diagram of the flow cell and Hydro 2000G arrangement
87
Once the measurementwas complete, V1 and V3 were closed and V2 was openedand
the Hydro 2000G was filled with de-ionised water. The water was then recirculated
through the flow cell using the Hydro 2000G pump.
The cleanlinessof the cell was monitored using the background reading on the
Mastersizer. The cell was flushed through twice after every measurement. If the
background readings were too high even after flushing, the cell was taken apart and left
to soak in 5% detergentsolution (Decon 75) for 2 hours. After 2 hours the cell was
flushed thoroughly firstly with warm water and then with de-ionised water, to ensureall
of the detergentwas removed.
The Receiver
The receiver collects and storesthe information received from the diffracted laser light
its
and main componentsare a Fourier lens, detector and an obscuration monitor.
The diffracted light is interceptedby the Fourier lens, which maps the scatteredlight
from particles of the samesize into a common spatial position in the focal plane. The
detector is located in the focal plane and consistsof a number of photo-diode elements
arranged in a radial pattern. Each of these elementsis positioned to correspondto a

specific scattering angle. The Fourier lens enables light scattered by a particle to fall on
the samepart of the detector for as long as the particle is in the laser beam, independent
of its position in the laser beam. This makes it possible to measurethe intensity of light
diffracted by particles of the samesize and de-convolutethis data to produce a size
distribution.
To ensurethat the detector takes a representativereading of the scattering pattern the

Mastersizer takes over 2000 snapsfor each measurementwith each snaptaking 1
millisecond.
The obscuration monitor is used to measurethe amount of laser light lost due to the
introduction of the sample into the analyserbeam. If there are no particles in the laser
88
beam,all the light falls onto the obscurationdetector. As particles are introducedinto the
laserbeam,they block this light and scatterlight onto the detectorelements.The
obscurationis the ratio of the intensity of the incident light with no sampleto that with
sampleand is usedto monitor the concentrationof the particle. It is an important
parameteras it allows the operatorto control the concentrationof the sample,as too
many particleswould result in multiple scatteringand thereforeerroneoussize data.
The collected data is then sentto the computerfor analysis. The Mastersizeruses
Lorenz-Mie theory to predict the droplet size distribution from scatteringintensities. The
Mastersizercollects a scatterpattern from a sampleand then constructsa rough droplet
size distribution. Lorenz-Mie theory is then applied to convert this distribution back to a
scatteringpattern. The predictedpatternis then comparedto the original pattern and if it
is a close fit the correspondingparticle sizesare accepted. If the comparisonis not a
good one then the Mastersizerrepeatsthe processuntil the closestfit is found.
The accuracyof the fit betweenmeasuredand calculateddata is given by the residual,a

value lessthen 1% denotinga good fit. If the residualis greaterthan 1% then this is a
good indication that the correct refractive indicesand absorptionvalueshavenot been
used.
4.3.2. Vidco Probc

The samplingstrategywas validatedagainstan in-lint imagecapturingtechniquethat
incorporateda video probethat was insertedinto the outlet pipe of the rotor-statormixer.
The ability to measurethe droplet sizeson-line enableda comparisonto be madewith the

off"linc measurementtechnique. It was then possibleto determinewhetherthe sampling
and dilution techniqueaffecteddroplet size distribution. However,beforebeing able to
makethesecomparisonsan appreciationof the differencesbetweenthe two techniques
had to be gainedfirst. This was donemeasuringstandardlatex particles.
89
Probe for c
_ý
Figure 4.6 Video Probe housing
The Video probe consistedof a lens attachedto a CCD camera(Hitachi KP160) that was
housed in a purpose built stainlesssteel casing. The casing was fitted with tri-clover
fittings, which enabledthe probe to be inserted in-line (Figures 4.6 & 4.7a). The probe
housing was designed so that the cross sectional areabetweenthe probe and the strobe
was equal to that of the inlet and the outlet of the video probe housing.
The probe consisted of a lens that was attachedto a CCD camera(HITACHI KP160)
which in turn was connectedto a computer. The light sourcewas a stroboscope

(DRELOscope 3244) and was transmitted inside the pipe via a fibre optic, located on the
opposite side of the probe to provide backlighting. The strobe was operatedat 50 Hz to
correspond with the electronic shutter speedof the camera and had a flash half intensity
width of 1 µs and intensity 0.15 J/flash.
90
The dimensions of the casing and the spacerestrictions imposed by the positioning of the
probe and the fibre optic meant that images were taken 5 pipe diameters downstream of
the mixer.
Images were transferred from the CCD to a computer using a Neotech frame grabber,
which displayed images continuously on the computer monitor. Ninety-nine bitmap

imageswere captured for each run.
The system was calibrated by taking a picture of a graticule using the Video Probe, one
division on the graticule was the equivalent of 50 µm. A computer was then used to draw
a straight line on the image. This length of the line (in pm) was known from the divisions
on the graticule and the computer was then usedto calculate a calibration factor by
dividing this distance by the number of pixels that representedthe line. The calibration
factor was calculated to be 1.2 pm/pixel.
The droplets were sized using a semi-automaticprogramme developed in-house at BHR

Group Ltd using National Instruments LabView software. This involved the operator
defining the boundariesof a drop by dragging a circle around it using a mouse. The
program counted the number of whole pixels acrossthe diameter of the circle and then
used the calibration factor to determine a diameter in µm. Only drops in focus were
measured.
When all of the drops in a frame had been measuredthe software programme counted the
number of drops that had been measured and calculated the d32,
before moving onto the
next frame. The number of droplets measuredwas dependent on the value of d32,
which
was monitored throughout the process.

measurement The measurementwas ended when
the value of d32stoppedchanging significantly (± 1 micron) with each new measured
drop. The number of drops required before a stable value of d32was reachedwas
typically around 600 drops. The measureddiameterswere stored in data files, which were
then imported into a spreadsheetpackagefor analysis.
91
The images captured by the CCD camerawere made up of two interlaced pictures (i. e.
two frames taken a split secondapart). As the imageswere taken of a dynamic situation
the picture quality was not always the best and the drops would often have blurred edges.
During analysis, the software program allowed one of these imagesto be discarded. It
then interpolated between the gapsto obtain a full image. This procedure resulted in
improved image quality.
A potential problem with this technique is the degreeof accuracy displayed when
measuring smaller droplet sizes. The calibration factor for this measurementwas 1.2
microns/pixel so a true droplet diameter of 5 µm would cover 6 pixels. The size of one
pixel was approximately 250 µm, thus the diameter of the circle that would have to be
drawn around a5 µm drop would be 1500 µm (or 1.5 mm). The resolution of this
technique was 2.4 µm (as the thickness of the circle drawn around a drop was the
equivalent of 2 pixels).
In order to eliminate errors due to individual subjectivity, the same operator was used to
measure all the drop sizes.
Another inherent problem with this technique is that the larger droplets have a tendency
to obscure the smaller ones and thus make it difficult to spot smaller ones. This gives rise
to inaccuraciesin the droplet size distribution. However, this technique is useful in
identifying the larger droplet sizes,which are more likely to be affected by the sampling
arrangement.
4.3.3. Droplet size distributions
The Mastersizer collects a scatterpattern from a particular sample and usesLorenz-Mie

theory to determine the size distribution. The main assumptionin the use of Mie theory
is that all the particles are spherical and the results are expressedin terms of equivalent
spheres.The Mastersizer, therefore, determinesthe volume of the particle and relates this
92
back to a diameter of a spherewith an equivalent volume. It is important to note that the
size distribution derived by this technique is volume based.
The analyseddistribution from the Mastersizer is expressedin a set of size classes,which
are optimised to match the detector geometry and optical configuration giving the best
resolution. The standardMastersizer has 72 analysis size bandsand are createdby equal
logarithmic spacing between the limits of the instrument. All parametersare derived from
this fundamental distribution using the summation of the contributions from each size
band. The representativediameter for each band is taken to be the geometric mean and is
generally consideredto be more appropriateto the logarithmic spacing of the

fundamental size classes.
The droplet size distributions generatedby the Malvern Mastersizer are presentedas
Volume frequency curves. The frequency curve is useful when identifying where the
main peaks in the distribution lie and also when comparing results from different
measurements. This curve is calculated by differentiating the cumulative volume

frequency curve, which in turn is calculated from the initial size bands.
The Mastersizer software allows the volume distribution to be converted to a number

distribution. However, as the initial measurementis volume based,this conversion is
liable to systematicerrors.
The Mastersizer software allows the customisation of the size bands and can switch from
a logarithmic x-axis to a linear one to enable comparison of distributions from other

techniques.
The droplets that were measuredusing the video probe were analysedin an Excel
spreadsheetthat arranged the data in ascending order and then placed the diameters into
predetermined bin sizes (which matched those used by the Mastersizer software). The
distributions generatedby the spreadsheetwere both number and volume based. This
93
enabledthe droplet size distributions from the Mastersizer and the Video probe to be
plotted on the sameaxis.
It is important to note that different techniquesmeasuredifferent properties of a droplet.

For instancethe Malvern Mastersizer measuresthe volume of a drop, whereasthe video
probe measuresthe diameter. It was mentioned previously that the video probe was
to
subject an error of ±2 µm on a particle/droplet diameter. Therefore, for droplets
below 15 µm, an error greaterthan ±13% would be exhibited. Therefore, comparisons
were made using both number basedand volume baseddistributions to minimise errors
that may exist when converting from one to the other.
4.4. Sampling technique
4.4.1. Continuous Sampling
A continuous sampling and dilution systemwas developed (Figure 4.7) that eliminated
the need for the sampling accessorieson the Malvern Mastersizer. This allowed a larger
volume of sample to be measuredat higher concentrationsand also aided the processof
making rapid measurements.
Figure 4.7a shows the video probe in the outlet of the mixer and the sampling point. The
technique involved the extraction of a continuous streamof sample from the outlet of the
rotor-stator mixer. This sample was introduced into a water line, the flowrate of which
was controlled using a peristaltic pump (Figure 4.7b). The diluted stream was fed
directly into the sampling unit of the Mastersizer (Figure 4.7d).
The sample was extracted using two sectionsof 2mm stainlesssteal tubing both with one
end curved (Figure 4.7c & Figure 4.8). The angle of curvature was kept as large as
to minimise flow irregularities. The sampling inlet was positioned in the centre
possible
of the rotor-stator outlet pipe. The two tubes were connectedusing 2mm Marprene
tubing, a distance was kept between the two tubes to accommodatethe lab clamp valve.
94
Figure 4.7 Sampling arrangement. Clockwise from top left; (a) The inline
rotor-stator with sample point and Video probe in place; (b)

Watson Marlow pump used for supplying diluent; (c) Inlet to
Mastersizer flow cell and pump and stirrer unit used for flushing
the cell; (d) Sample point with sampling arrangement.
There was a danger that as the sample flows through the valve arrangement,further drop
break-up may occur giving rise to erroneousresults. The lab clamp valve arrangement
was selectedbecauseit eliminated any sharp surfacesor edgesthat would exist if a ball or
95
needle valve were used and minimised the effect of sampling on the droplet size
distribution.
aarrp Valve
Water
2mmstainless st
tubing
rpreneTubing
8mm FE To Nbly rn
From WS mixer Tubing
Figure 4.8 A schematic of the continuous sampling system.
The sensitivity of the Mastersizer to air bubbles led to de-ionised water (that had been left
to de-aerate over night) being used as the diluent. The correct dilution rate was achieved
by adjusting the laboratory clamp valve and also by varying the flowrate of the water.
The concentration (or obscuration, see Appendix 3) was monitored by the Mastersizer
software program and a measurement was made only when the correct level was
achieved. The sampling technique was validated against the Video Probe described
previously the results of which are discussed in Chapter 6.
96
4.5. Measurement of Power supplied to the fluid
The various mechanistic correlations, discussedpreviously, are dependenton the scaling
of the droplet sizes with the Kolmogorov length scale. The Kolmogorov length scale,
therefore, is important in the interpretation of the droplet size data.
Although it was not possible to monitor the power during the dispersion tests,
measurementswere carried out separatelyover the samerange of experimental

conditions. This provided an estimate of the order of magnitude of the Kolmogorov
length scale, which was then used when analysing the droplet size data.
The measurementof power was discussedin Chapter 2.4 and it was mentioned that in-
line rotor-stator mixers are analogousto centrifugal pumps. A power balance for pumps,
used by Sparks (1996), to evaluatethe power dissipatedto the fluid in an in-line rotor-
stator mixer was presented(Equation 2.36).
Traditionally, the method of determining total power in stirred vesselsis to measurethe
shaft torque. The shaft power is calculated from the torque using Equation 2.38. This
method is dependenton the accurate of

measurement the shaft torque and many
instruments, such as torque transducersand strain gaugesare available to do this.
Sparks (1996) used this technique to calculate the shaft power on an in-line rotor-stator
mixer. The motor shaft was extended to allow a torque transducer to be fitted between the
motor and the mixing head. He reported values of torque from 1 to 4 Nm at 3000 rpm,
(the highest rotational speedsinvestigated). However, it was found that these values
were at the lower limits of the torque transducerused and thus at the limits of accurate
measurement. An inherent problem with strain gaugesand torque transducersis that their
resolution decreasesas the rotational speedincreases.
Adopting a similar approachwasdifficult for this work dueto the high rotationalspeeds
of the rotor-statormixer. designed
Instrumentation for high rotationalspeeds did not
have adequateresolution to measurethe low values of torque expectedfrom the in-line
97
rotor-stator mixer. In addition, the cost and safety implications of extending the motor
to
shaft rotate at 12000 it
rpm meant was impractical to pursue this technique. As a result
alternative techniqueswere investigated and it was found that the best option available
was to use a thermometric technique.
4.5.1. Thermometric method of power measurement
The thermometric technique for measuringpumping efficiencies is one that has been used
extensively in the water industry (Brown, 1991). It is basedon the fact that the majority
of lossesin a pump are dissipatedinto the fluid as heat.
These lossesresult in a small increasein temperature,typically in the order of a few

hundredthsof a degree,betweenthe inlet and the outlet of the pump. Therefore, the
technique requires the accuratemeasurementof temperatureto within a few thousandths
of a degree.
This method was adaptedto measurepower in the in-line rotor-stator mixer. The
problems associatedwith measuring the small temperaturerises were overcome by

inserting the mixer into a recirculation loop and monitoring the temperaturerise over a
period of time. The numerous passesthrough the mixer heatedthe fluid sufficiently to
enable accuratetemperaturemeasurement.The time was measuredallowing the
calculation of the theoretical number of passesand therefore the temperature rise for one
pass.
4.5.2. Recycle loop and instrumentation
The rotor-stator mixer was fitted into a recycle loop using a 100 litre stainlesssteel vessel
with a1 inch outlet (Figure 4.9). The low volume vessel was chosen, it
as was easier to
lag and limit heat lossesthrough the vesselwall. In addition, for a lower volume, large
temperaturerises were possible in a shorter time.
98
The pipework from the vessel to the inlet of the mixer and immediately after the outlet
was 1 inch. A3 inch Endressand Hauser Promag 33 flowmeter was fitted downstream
(10 pipe diameters) of the mixer and as result the pipework was gradually enlarged from
1 inch to 3 inches. A3 inch butterfly valve was located downstream of the flow meter
to
and was used control the flowrate. The pipework was gradually reduced from 3 inch
back to 1 inch at the inlet of the vessel. The inlet to the vessel was submergedbelow the
liquid level in the vessel.
0.635 C
®
Thermocouple
Figure 4.9 Flow loop set-up for thermometric power measurement
To monitor the lossesacrossthe control valve, pressuregauges(Wika Bourdon Tube

PressureGaugesType 213.53) were fitted 1 pipe diameter either side of it. A pressure
gauge was also fitted one pipe diameter up-stream of the mixer.
99
The flow from the inlet to the vessel was used to mix the fluid and to ensurehomogeneity
the temperaturewas measuredat different locations in the vessel. Three type K fine wire
thermocoupleswere located near the top, at the mixer outlet and near the bottom of the
vessel. The temperaturewas monitored using a Fluke 50S handheld digital thermometer
connectedto a Type K 6-way bench selector switch (although only three of the
connectionswere used). The mixer was run until the temperatureof the water had risen
by 10 °C. This gave an error in measurementof ± 0.1°C.
The vessel, the inline rotor-stator mixer and all the pipework was fully lagged using 150
mm fibreglass insulation. After everyone run the temperatureof the water was monitored
with the mixer turned off. After 24 hours the temperature of the water was found to have
only decreased by approximately 1-2 °C, therefore, the heat lossesto the surroundings
were consideredto be negligible.
4.5.3. Experimental procedure
The vessel was calibrated to hold 100 litres of de-ionised water and fresh water was used
for every run. To ensurethat the volume of water was kept constant the outlet valve was
kept shut while the vessel was being filled. Once the correct volume had been reached
the valve was openedto fill the pipe work and the mixer.
The temperaturewas noted at time zero for all three thermocouples. It took a few
minutes for all three thermocouplesto read the sametemperature(± 0.1 °C), once this
had been achieved the mixer was startedat the appropriate rotational speed. The
butterfly valve was adjusted to control the flowrate. After the first run the speedand the
valve were set for the next run before the mixer was switched off.
The temperaturewas monitored until a total rise of 10 °C had been achieved. This
temperature difference was consideredhigh enough to obtain a reading within the desired
accuracy ((± I%). Typically, this temperaturerise was observe over a period of 3-6
100
hours. The initial temperatureof the water ranged from 12 °C to 14 °C and the final
temperaturewas between 22 °C to 24 °C. The density of water varies by less than 0.5%
over this temperaturerange.
The vessel and the pipework were drained using a drain valve that had been fitted into the
bottom of the mixing head.
4.5.4. PsFLFT
The power supplied to the fluid from heat dissipation can be defined as:
E=mCptT (4.1)
Where, E= Energy supplied to the fluid (J)

Cp = Specific heat capacity (J/kgK)
m= Mass flowrate (kg/s)

AT = Temperature rise (K)
The theoretical number of passes is given by:
No of Passes= mt (4.2)
Where, t= the time the mixer was run (s)

M= Total massof water (kg)
Therefore, the temperaturerise over a single pass,AT,, is given by:
101
ATM
AT, a (4.3)
mt
It can be assumedthat all of the power supplied by the shaft is converted to heat energy.
Therefore, the energy supplied to the fluid in a single pass,can be consideredto be the
equivalent of the shaft power giving Equation 4.4:
PSHAFT
AT, Po pN 3D5 (4.4)
= me p =
4.5.5. PPUMP
The flowrate and the differential pressureacrossthe mixer were measuredusing the
Endressand Hauser flowmeter and the Wika pressuregauges. Equation 2.39 was then
used to calculate the pumping power.
4.5.6. PMECH
The mechanical losseswere measuredby removing the rotor, filling the head with water
and blanking off the inlet and the outlet. The shaft was then rotated in the water until a
temperaturerise of 10°C had been achieved.The temperatureinside the head was
measuredusing the Fluke 50S handheld digital thermometer in conjunction with a
BSS1843 thermocouple probe (type K) inserted into a 3/4" compressionfitting at an
additional sampling point located on the side of the mixing head (tangential outlet).
4.5.7. PDISC
by spinning a smooth disc with the same

Sparks (1996) attemptedto measurethe PD1sc
dimensions as the rotor. The resultant torque readings were found to be low compared to
those specified for the torque transducerused. It was suggestedthat the flow patterns
would differ to those for a rotor (due to the absenceof teeth) and therefore a theoretical
calculation would be just as reliable. The theoretical power loss due to disk friction was,
therefore, calculated using eqn 2.40.
102
4.5.8. PLEAK and PHYDRAULIC
were not directly measuredbut it was assumedthat they accounted

PLEaxand PHYDRAULIC
for the remainder of the shaft power, once all the other losseshad been deducted.
BLEAK + PHYDRAULIC PSHAFT (PPUMP + PMECH + IDISC) (4.5)

= -
4.5.9. Lossesthrough the control valve and pipe work
The heat input causedby the lossesassociatedwith the valve and the pipework were also
calculated. The pressurewas monitored acrossthe valve and along the pipework using
the Wikon gauges. This enabledthe head loss acrossthe valve to be calculated.
The lossesin the pipe were expectedto be small in relation to the heat input from the
rotor-stator mixer. However, the heat contribution of the control valve would be
significant.
103
CHAPTER 5. RESULTS
The rotor-stator liquid-liquid contactor rig was used to perform a range of tests that
investigated the affects of N, Q and 4) on the droplet size. These tests were done on two
different rotor-stator sizes as well as four different stator designs.
The high rotational speedsof the rotor-stator mixer made it difficult to measurepower
during the tests. Therefore, a separaterig was set-up that utilised a thermometric
technique described in Section 4.5. Thesemeasurementswere done on the 69.85 mm
rotor with the SquareHole High ShearScreenand the Emulsor Screen,for which the
most comprehensiverange of droplet size data was collected.
The details of the experiments and the location of the results are presentedbelow and
discussedin Chapter 6.
5.1. Measurement technique validation
The sampling technique employed on the liquid-liquid contactor rig was validated using a
Video probe. This involved performing initial tests in a laboratory scale stirred vessel to
investigate the stability of the kerosene-watersystem. This was done to determine
whether sampling was a viable option.
Tests were also performed comparing the Malvern Mastersizer with the Video Probe by
measuring standardlatex. The results from these tests were then used to interpret the
results from the sampling validation.
The validationtestsinvolved insertingthe video probeat the outlet of the rotor-stator

mixer and comparing the results against the Malvern Mastersizer. The results and
observations of this work are presentedin Section 6.1.
104
5.2. In-line rotor-stator performance tests
The performance tests involved the investigating the effects of rotational speedon droplet
size at two different Q (0.4 Us and 0.81/s) and four different 1 (5%, 10%, 20%, 30%). A
small number of tests were also performed at 40%.
The main focus of these tests were on the 69.85 mm rotor with the Emulsor screen,(ES),
and the Square Hole High Shear Screen,(SHHSS). This rotor size was the standardfor
the unit supplied and these two stator designsare recommendby the manufacturer for use
with liquid-liquid systems.
A limited number of experiments were also performed using the 50.8 mm rotor. These
tests looked at the effect of rotor speedon droplet size at a constant flowrate (0.4 Us) and
dispersedphaseconcentration (10%). The effect of dispersedphaseconcentration was
also investigated for this rotor size (at 7000 rpm at 0.41/s). These tests were done to gain
some insight into the effects of stator design on the droplet size distribution.
The volume size distributions from theseresults can be found in Appendix 4. The
distributions are discussedin Section 6.2. The influence of the important parameterson
droplet size are discussedin Section
5.3. Power Measurement
The power input to the fluid was measuredseparatelyto give an idea of the Kolmogorov
length scale. The power was measuredfor the 69.85 mm rotor with the ES and SHHSS.
This was becausethe most comprehensivedroplet size data was collected for this rotor
size and two screens.
The power was measuredfor a range of flowrates, 0.4 to 1.21/s in 0.21/s increments for
rotor speedsform 5000 rpm to 11000 rpm. A discussion of these results are given in
Section 6.4..
105
CHAPTER 6. Discussion
6.1. Measurement technique and Sampling validation
6.1.1. Kerosene-Water Stability
The Kerosene-water stability was investigated in a stirred vessel described in Section
4.2.2. A dispersion was created and samples were taken and placed on a glass slide and
viewed under a microscope (Figure 6.1).
It was observed that the droplets in a globule of sample from a 5% dispersion placed on a
glass slide moved around considerably (drop sizes ranged from 10 to 100 hum). '['his
motion could be attributed to the droplets, due to their density, ascending to the highest
point of the globule. At the highest point, the greatest rate of coalescence would be
expected due to droplets getting closer together.
Figure 6.1 Droplets collecting at the top of the sample under a microscope (5%
dispersedphaseconcentration, undiluted)
106
However, it can clearly be seen,in Figure 6.1, that there appearsto be no evidencethat
the dropsare coalescing,(no bursting of interface and no further drainageof the
continuousphase).This appearedto be the caseafter observing periodically for up to an
hour.
These findings support those of Brown and Pitt (1971) that the kerosene-water system is
not a readily coalescingsystem. Droplets that would be generatedin a rotor-stator mixer

are expected to be significantly smaller than thosegenerated by a stirred tank. The
smaller droplet diameters would lead to an increasein the interfacial forces resisting drop
break-up (Equation 2.3) resulting in increased stability.
6.1.2. Comparison of measurement techniques
The Malvern Mastersizer and the Video Probe were both used to measure standard latex
beadswith known diameters. Two setsof latex standards,supplied by Bangs

Laboratories(Appendix 2), were usedwith different meandiameters(Table 6.1).
Table 6.1 Latex beadstestedusing the Mastersizerand the video probe.
Mean Diameter (µm) Standard deviation (µm)
98 90 -106
113 75 -150
Measurementswere madeusing the Mastersizersampling accessory(Hydro 2000S/G,

Figure 4.7c) that incorporateda stirrer and a pump. The inlet and the outlet of the
accessorywere connectedto the Mastersizer,simulating a continuousflow loop (closing

valves VI and V3 in Figure 4.5). The Video probe was inserted into the flow loop (in
place of V2 in Figure 4.5), so that could

measurements be madesimultaneouslywith the
Mastersizer.
107
The Latex beadswere introducedto 1 litre of water in the Hydro 2000S/G,and the stirrer
was usedto keep the in

particles suspension.The latex beadswere then passedthrough
the video probe and the measurementcell of the Mastersizer. The measurementswere
repeatedusing a fresh set of beadsand the results comparedfavourably to within 5% for
the video probe and almost identical for the Mastersizer.
The volume and number based particle size distributions for the 98µm and the 113µm
beads are shown in Figures A. 2 & A. 3 (Appendix 4) respectively.
The 98µm particles have the narrowest distributions (Figure A. 2), which is expected as
these beads exhibited the lowest standard deviation (Table 4.4). This observation is
consistent with both the volume and number distributions generated by the two
techniques. The main observation for the two sets of particles is that the distributions do
not match, with the Mastersizer favouring the smaller particle sizes. This is true for both
the volume and the number distributions.
Figure A. 2 shows that the distributions from the Mastersizer appear to have shifted
approximately 5µm to the left of those generated by the data from the Video probe, with
peaks at 96µm with the Mastersizer and 100µm with the video probe. Comparing this to
the average diameter (mode) specified by the manufacturer of the beads, 98µm, suggests
that both techniques are within ± 2% the acceptable limit.
The volume curve generated by the Mastersizer in Figure A. 2 exhibits a small secondary
peak between 75-90µm. This smaller peak is not apparent when viewing the number
distribution and the number curves from the two techniques appear to follow the same
shape. It is possible that this secondary is

peak smoothed out on the number curve as
there are a larger number of particles that make up a small percentage of the volume.
In Figure A. 3 the distributions from the video probe consistentlygive larger sizes. The
averagevalue (mode) for this set of beadsshould be 113µm. However, the Mastersizer
shows a peak at 95 µm and the Video Probe shows a peak at 103 µm.
108
Another interesting observation is that both techniques appear to be measuring particles
outside of the range of the standard deviation for the sets of beads. This highlights the
potential for variability. The manufacturer of the latex beads used a sieving technique to
measure the latex beads, which would also contain a certain degree of error. So it is
possible that the Mastersizer and the Video probe are measuring the beads more
accurately than technique employed by the manufacturer.
The importance in realising that different techniques measure different parameters is
clearly evident. One would not expect a perfect match between the two techniques, as
there are systematic errors involved in them both. The Mastersizer measures volume and
the video probe measures the diameter. However, the Mastersizer appears to have more
of a bias towards smaller sizes and the video probe appears to show more of a bias
towards larger particles.
The differences are more apparent for the larger and wider distributions (75-150µm). It
can be seen that the video probe is seeing a greater number of particles larger than
100µm. This could be due to the fact that the larger particles are easier to see during
measurement and that they also tend to obscure the smaller ones. This phenomenon
would be less obvious for a narrower size distribution.
In generalthe distributions although not identical were close enoughfor comparative

purposes. It is also believed that the sampling arrangement for which the video probe
be
would used to validate, would be more sensitiveto the larger drops. Therefore, this
technique can be used to determine whether the sampling arrangement is having an effect
on the droplet size.
6.1.3. Sampling technique validation
The sampling techniquewas validated againstthe in-line video probe. The probe was
insertedinto the outlet of the in-line rotor-stator mixer (5 pipe diametersdown stream)
and imageswere capturedsimultaneouslywith the sampling.
109
The video probe could only be used at the lowest rotor-speed (3000 rpm) and the lowest
(h (5 %) at a total throughput of 0.81/s, (Figure 6.2). Images at values higher than these
were not possible, due to the increase in smaller droplets and consequently increased light
scattering and resultant focussing problems.
4-;.. ,.
} ýt
Figure 6.2 Image at 3000 rpm and (1) = 5%
lt can be seen from Figure 6.2 that the quality of image obtained at the lower rotational
speed is at the limit of acceptability. This is due to the large number of droplets that are
below the minimum size range for this technique (20 µm), images become increasingly
out of focus below this size range (due to increased light scattering).
110
Figure 6.3 shows a comparison of the volume distributions generated by the Mastersizer
and by the Video Probe at the outlet of the rotor-stator mixer. As expected, the
distributions are broader than those obtained from the latex beads and the main peaks
from both techniques lie in approximately the same region (40-60µm).
Video Probe (in oulet of Rotor-Stator mixer) Vs Masterslzer. 3000 rpm, 0.811s& 5%
8
IN Its 05% VP
III 0.8
00.8 Us05% Mastersizer
6H
5
Em 0000OOo
>4" 00
0
e0
0
3 O" 08
pp
20 Op
00 00
0
0
0
20 40 60 80
000MO
100
OOOOOpOp
120 140 160

Droplet diameter (micron)
Figure 6.3 A comparison of DSD measured using the Mastersizer and the Video Probe
for a rotor speed of 3000 rpm and c= 5% at 0.81/s
As well as seeing smaller droplets, the Mastersizerseemsto have measureda greater
number of larger droplets. Only a small percentageof the droplets measuredby the
Video Probe appearto be at the larger scale. However, the maximum size measuredwas
130µm. This led to the conclusionthat droplets of sizes up to 140µm could be
successfullysampledusing the continuoussamplingarrangement.
111
6.2. Drop Size Distributions
Tests were performed on a Silverson in-line rotor-stator mixer investigating the effects of
rotor-speed, dispersed phase concentration, flowrate and the stator design on droplet size.
The following section reviews the trends in the droplet size distributions and the mean
diameters calculated from them. The effects of the individual parameters are then
discussed paying particular attention to the scaling of the various parameters against d32.
6.2.1. Volume frequency curves
The droplets size distributions generated by the 69.85 mm rotor with the Emulsor Screen
(ES) can be found in Figures A. 4 to A. 12. Figures A. 13 to A. 18 show the distributions
generated by the 69.85 mm rotor with the Square Hole High Shear Screen (SHHSS). The
distributions are shown as volume frequency curves for a constant Q and c for varying
N. The volume distributions from the 50.8 mm can be found in Figures A. 18 to A. 25.
The distributions are fairly broad, with drop sizes ranging from below 10 µm up to 100
µm. This is indicative of the inhomogeneous energy dissipation and by-passing within the
rotor-stator mixer. The majority of the distributions exhibit a long tail at the smaller size
end and a truncation of the drop size at the larger end of the size spectrum. The
truncation at the larger end is due to the distributions being plotted on a logarithmic axis.
The shapeof the distributions appearssimilar for eachof the rotor speedsbut the peak
shifts towards the smaller droplet sizes as the speedincreases.

This shift is less significant
at speedsgreaterthan 7000 rpm.
At some of the lower rotational speedsa bi-modal distribution is evident, with a

secondary peak at the smaller droplet sizes. This is more apparent for runs performed at
the higher (D. The secondarypeak appearsto smooth out as N is increased.
112
The main bulk of the droplet sizesare below 100µm for the runs performed at the lowest
values of c, 5% and 10%. The main peaks lie at approximately 50 µm. This peak shifts
towardsthe larger droplet sizeswith an increasein dispersedphaseconcentrations.
An additional peak, at the larger size ranges, is observed for some of the higher dispersed
phase concentrations (Figures A. 6, A. 7 and A. 13). This peak represents droplet sizes up
to 1000 µm at the maximum measurable size limit of the Malvern Mastersizer. The
peaks appear to cut off at these points suggesting that droplets exist that are too large for
measurement. However, these additional peaks disappear as the rotational speed
increases. Due to the curtailing of the size distributions the corresponding mean
diameters were not used in any of the subsequent analysis.
6.2.2. The relationship betweend32and dmax
There are many correlations available in literature that have substituted d32for dmax
(Tavlarides and Stamatoudis, 1981). The relationship between dm. and a mean diameter
was discussed in Section 2.2.3. It was stated that in addition to its greater design value,
for a small sample size d32could be determined more accurately than dme,,.
It was also suggestedthat dmax

was proportional to any convenientmean diameter,
provided break-up was due to a single break-upmechanism. Figure 6.4 showsa plot of
experimentally measuredd32versus dm,. for the full rangeof testsperformed using the
liquid-liquid contactor rig describedin Section4.1. The averagediameterswere those
determinedby the Malvern Mastersizer. The measurementtime was increasedto 10
secondsto ensurea large samplesize was measured(the Mastersizertook approximately

10,000snapshotsof the samplewith eachsnaptaking one millisecond), which meant that
the number of drops measuredwas in the order of thousands.
Although there is considerablescatterin the data,a linear trend can be seen. The best fit
through all the data points gives a constantof proportionality of 0.13. This is lower than
those reported in literature; Brown and Pitt (1972) reported a value of 0.72 for a 6-bladed
113
disc turbine, whereas Chen and Middleman (1967) reported a value of 0.6 for a Rushton
turbine; Francis (1999) using a laboratory batch rotor-stator reported a value of 0.44.
80
-- -,
70
d32 = 0.13 do,

a:
60- R2 0.88
50-
40-
30- """
="
20-
"__
10 """
"
0
0 50 100 150 200 250 300 350 400 450 500
d. (µm)
Figure 6.4 Correlation betweend32and dm.
The low constantof proportionality maybe due to the long tailing of the droplets at the
smaller size range. Although, thesedroplets be

may significant in number, in terms of
their volume they are small. Their influence on d32,therefore is also smaller than the
larger droplets.
At this point it is important to note that the theories of droplet break-up rely on the
prediction of mean drop size at equilibrium conditions. Due to the inhomogeneous

natureof the energy dissipation in an in-line rotor-statormixer (nominal residencetimes
in the sheargap, the region of highest shearrate, are in the order of 1O s) it is unlikely
114
that equilibrium drop size has beenattainedafter a single pass. Therefore,the
significance of dmaxcan be rebutted.
However, the fact that the kerosene-water system has been shown to be a non-cpalescing
system leads to the assumption that coalescence is not affecting the droplet size
distributions. Also, the sampling technique should show no affect on the droplet size
distribution.
It can, therefore,be assumedthat transient effects are negligible (as there is no further
break-upand coalescence)and that the correlation of mean drop size basedon
equilibrium drop-size can still be applied.
Also, the spacing between the size bands created by the Mastersizer is logarithmic. This
bin spacing is designed to match the detector geometry and provide the best resolution.
This implies that the resolution of dm. decreases as the droplet sizes increase, due to the
bin sizes increasing. For droplet sizes above 140µm the bin sizes ranged from 20µm to
50µm.
The Mastersizergeneratescurvesby mathematicallyfinding the size data that best fits the
light scatterdata,this almost always results in smooth curves. A test for the accuracyof
such a plot is to ensurethat the values of residual calculatedby the Mastersizersoftware

are close to the value of 1.
Therefore,the dmaX
data, although useful in determining whetherthe assumptionthat
break-up is due to a single mechanismis valid, should only be used qualitatively.
6.2.3. The effect of rotational speed, N, on droplet size
Figure A.4 shows the volume frequencycurve for a flowrate of 0.41/sand 4) of 10%.
The distribution can be seento shift towards the smaller size rangeas N increases.It is
interestingto seethat there is a significant changein the distributions between3000rpm
115
and 6000rpm, with the main peak shifting from approximately 100µm to 50µm.
However, rotational speedsabove6000rpm exhibit only a small changein droplet size.
This trend is consistentwith the majority of the testsperformed for both the ES and the
SHHSS.
Figure 6.5 shows the d32plotted againstrotor speedfor a rangeof dispersedphase
concentrations. The droplet size decreases

with rotor-speed,however they begin to level
off after 7000 rpm.
It is clear that the rotor-speedis only influential down to a certain droplet size below
which further size reduction is not evident.
A similar trend was observedfor the data collected for the 50.8 mm rotor, Figure 6.6.
The droplet sizes,although still appearingto level off (especially for the GPDH and SH)
it is not as prominent as that exhibited by the drop sizes from the 69.85 mm.
When considering the power supplied to the fluid, for a fully turbulent system, the power
should scale with N3 and should therefore increase. If the energy supplied to the fluid by
the rotor was predominantly responsible for drop break-up the droplet sizes would
continue to decrease with rotor speed. This would indicate that there are other factors
that influence drop break-up.
It is interesting to note that in Figures 6.5 and 6.6 there are no significant differences in
droplet size from the ES and the SHHSS for both rotor sizes.
Figure 6.6 shows that with the 50.8 mm rotor the ES and the SHHSS give the smallest
droplet sizesand the largestcome from the GDH. The GDH has the largest open areaout
of all the screens(35%) in addition to the largest hole size. However, the hole sizesof the
116
65
t
55
45
E
-' 35
25
15
5
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000
N (rpm)
0 SHHSS.0 41/s& 10% U ES 0 41/s& 10% A SHHSS:0.41/s& 20% 0 ES. 0.411s
& 20% X SHHSS:0.41/s& 30% 0 ES: 0 4115
& 30%
Figure 6.5 d32versus N (RPM) for the ES and SHHSS (69.85 mm rotor) for varying 4)
35
30
E
25
20
15+-
4000 5000 6000 7000 8000 9000 10000 11000
N (rpm)
f E/Screen x GDH f Slotted Head E3SHHSS
Figure 6.6 d32versus N (RPM) for four different screens (50.8 mm rotor), 0.4 Us and c_
10%
117
ES (1.4 mm) and the SHHSS (1.2 mm) are different and both thesescreensgive similar
droplet sizes. As the only variant is the open areaof the screensit can be reasonedthat
openareahas an influence on the droplet size distribution.
It is also worth noting that, in Figure 6.5, the droplet sizes increasewith dispersedphase
concentrations.
6.2.4. The effect of dispersed phase concentration, c
In Figure 6.5 the droplet size was seento increasewith (D,this effect is further evident in
the droplet size distributions (Appendix 4). The distributions can be seento shift towards
the larger droplet size spectrumas « is increased.
Figure A. 7 and Figure A. 11 show the distributions from the ES at 0.41/s ((D = 40%) and
at 0.81/s ((D= 30%) respectively. The distributions do not changesignificantly with an

increase in rotor speed.
Both Brown and Pitt (1971) and Streiff et al (1997) reported the effects of c on the
droplet size distribution. Brown and Pitt (1971) also looked at a non-coalescing
kerosene-water system, they attributed this effect to a form of turbulent damping. Both
researchers showed that d32could be related linearly with c (Equation 2.28).
In Figure A. 7 an additional peak exists at the larger droplet sizes. This peak is curtailed
due to the limits of the size measurementtechnique,however it is clear that droplets of
sizesabove 1000 µm exist.
It is unlikely that this peak is due to coalescence.It is possiblethat at the lower rotational
speedand high c there is a formation of keroseneglobules in the feed line. These
globules may be passingthrough the mixer without being dispersedeffectively in the
continuousphase. The low flowrates in the inlet pipe and also the effect of the swirling
118
flow at the inlet (seeSection 2.3.1) is not enoughto ensurethat the two phasesare
entering the mixer in the correct proportions.
As the dispersedphaseconcentrationincreases,the large globules (or a sausagelike flow)
of dispersedphaseenter the mixing headand passthrough it without coming into contact

with the continuousphase. The smaller droplet sizesmay be producedby the rotor
chopping parts of the sausage as it passesthrough the stator hole and also the effect of the
mixture being pumped through small holes (a type of jet mixing effect) into the volute.
The jets would also cause turbulence in the volute and drops may undergo further break-
up under a different mechanism. This is discussed in greater detail in Section 6.5.
It is worth noting that the dispersed phase is being injected one pipe diameter (0.0254m)
up-stream of the mixer. The pipe Re number at the lowest flowrate, 0.41/s, is
approximately 15000 (at 30% dispersed phase concentration) which can be considered to
be turbulent flow. However, the nominal residence time in the inlet pipe is 0.03s (at 0.4
Us) which would suggest that the two phases are not exposed to the turbulent region for a
significant period of time to allow any significant pre-mixing to occur.
This would point to the importance of feed strategy for high dispersed phase
concentration. In batch systems, where an in-line rotor-stator mixer is used in

conjunction with an agitator in a vessel, the agitator would create a pre-mix. Also, in
many practical applications a static mixer can be insertedup-streamof the rotor-stator

mixer to reducethe effect of feed injection.
This theory was not testedin greaterdetail for this work. However, varying the length
betweenthe mixer and the dispersedphaseinjection point and measuringthe resulting
droplet size distribution would provide further insight into the effect of feed strategy. A
further discussioninto the effects of flow regimeson the droplet size in the inlet is
presentedin Section 6.5.
119
6.2.5. The effect of flowrate, Q, on droplet size
The flowrate was not varied significantly in thesetests. However, to obtain an
appreciation of the effect of Q on the droplet size testswere performed at two different
flowrates (0.4 Usand 0.81/s).
Figure 6.7 shows d32versus N for 0.41/s and 0.81/s at two different (D. The drop sizes
for 0.81/s are smaller than for 0.41/s at the lower rotational speeds. However, the droplet
sizes appear to come together as the rotational speed is increased.
65
0
55
45
0
E
35
0
0
" '
25 . o
" p'
0
"
15
s !
"
5
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000
N (rpm)
ES: 0.41/s& 10% 0 ES: 0.41/s& 20% f ES: 0.81/s& 10% L ES: 0.81/s& 20%
Figure 6.7 d32versus N for 0.41/s and 0.8 Vs (at (D= 10% & 20%)
Two possible explanationsfor this occurrencefollow. Firstly, going back to the
commentsmade previously on feed strategy,the higher flowrate may createa better pre-
mix in the inlet pipe. This would ensurethat the two fluids are entering the mixer in the
120
correct proportions at the lower rotational speeds(when the swirling effect of the flow
was less).
Secondly, if the power draw of the mixer is proportional to N3 then the energy input to
the fluid by the rotor should increase significantly as the rotor speed is increased. It is
possible that at the lower speeds an increase in flowrate would increase the velocity of the
jets emanating from the holes in the stator. It can be assumed that the velocity of the
fluid out of the jets is the ratio of flowrate and the open area (of the stator). The energy
input of these jets in the volute coupled with the impinging fluid on the stator, maybe
greater than that of the rotor at the lower speeds.
This secondexplanation would lead to the possibility that the mixer is operating with
drowned suction (when the pumping effect of the rotor is overrun by the pump upstream
of the mixer) at the lower speeds.
Sparks (1996) reasoned that when the mixer is operating drowned the energy dissipation
would result from the lossesassociatedwith the rotor and from static lossessimilar to a
motionlessmixer.
Averbukh et al (1988) incorporated Q into their correlation for calculating mean droplet
size in an in-line rotor-stator mixer (Equation 2.31). However, the mean drop size only
showed a small dependence on Q (an exponent of -0.08). For the limited range of Q
studied, there only appears to be an effect on droplet size at the lower rotor speeds. For
this reason Q was assumed to have little or no effect on the droplet size.
6.2.6. The correlation of d32with N
Four different correlations for d32were discussedin Section2.2.3 and are summarisedin
Table 2.3. Thesecorrelations are basedon mechanisticmodels that balancethe forces
causingand opposing drop break-up.
121
The following section attemptsto determinethe dominant break-up mechanismby
determiningthe correlation of d32with the rotor-speed,We and Re.
Previousdiscussionshave highlighted that there are no significant differencesin droplet
size from the ES and the SHHSS. For this reasondata from both thesescreensis
consideredtogether in the following analysis.
Figure 6.8 shows d32versusrotor speedon a log-log plot for the different (D. A linear
regressionwas performed on eachof the data set to determinethe scaling of d32with N.
100
E
-A 10
Pi
v
1000 10000 100000

N (RPM)
Figure 6.8 d32versus rotor speedfor the SHHSS and the ES at different c and Q
122
Table 6.2 showsthat the exponenton N for the data collected bestmatch the Chen and
Middleman correlation, -0.86 (Equation 2.22). This would suggestthat inertial stresses
are responsible for droplet break-up and that the drop sizes are below the Kolmogorov
length scale.
Table 6.2 The best fit exponents on N for varying t
Q (I/s) (%) Nz R2
10 -0.88 0.92
0.4
20 -0.82 0.85
30 -0.58 0.95
05 -0.82 0.88
0.8 10 0.86
-0.79
20 -0.66 0.88
Average -0.77
Standard Deviation 0.11
The drop sizescan therefore be correlatedby:
dal ccN 7 (6.1)
From Table 2.3 it can be seenthat for this correlation to hold true the droplet sizesmust
be significantly smaller than Kolmogorov length scale. An estimation of the Kolmogorov
length scale is given in Section 6.3.6 and the comparisonwith d32is discussedin Section
6.4
123
The correlation of d32with N enablesa relationship to be determinedthat incorporatesthe
effect of (D.
6.2.7. Correlation of d32with We and Re
Equation 6.1 gives the scaling of d32with the rotor speed.The generalform of the Chen
and Middleman correlation relatesd32scaledwith rotor diameterproportionally to

(WeRe4)-1/7.The constantof proportionality would be dependenton the mixer geometry
and from Figure 6.8 the dispersedphaseconcentration.
However, for the data set shown in Figure 6.8 the rotor diameter is constant and the
geometry of the stator did not have a significant effect on the droplet sizes produced.
Therefore, it can safely be assumed that the dispersed phase concentration is the only
parameter, in addition to the rotor speed, that affects the droplet size. This allows the
following analysis to be performed.
Although, the majority of the data in table 6.2 correlate well with Equation 6.1 it is
apparent that there are some deviations at the higher values of (D, with the largest
occurring for (D = 30%. For this reason the data for c= 30% was not considered in the
subsequent analysis.
Brown and Pitt (1971) and Streiff et al (1997) showedthat c could be relatedto d32
linearly in the form shown in Equation 2.31. An attempt is now madeto show a similar
relationship for the data in Figure 6.8.
As a result of the general relationship in Equation 6.1 the d32can be plotted against
(WeRe4)-1/7for each of the dispersed phase concentrations, Figure 6.9. A straight line can
be drawn through the origin and the size data for each of the dispersedphase
(using
concentrations, a linear least squaresfit).
124
Table 6.3 lists the gradientsof the lines for the different values of (D. If the slopesare
plotted against (D then a linear relationship can be determined that takes the form of
Equation 6.2.
7. E-04
6. E-04
5.E-04
4. E-04
3. E-04
2. E-04
1.E-04
0. E+00
0 1E-14 2E-14 3E-14 4E-14 5E-14 6E-14
(We.Re4)an
Figure 6.9 d37IDversus (WeRe4)'"7for different c
Table 6.3 Gradients of the lines in Figures 6.11
CD 5 10 20
Gradient, m 4x 109 6x 109 1x 10
m= c+
4x1010 2x109 (6.2)
125
The generalcorrelation for d32for an in-line rotor-stator mixer (with a constantgeometry)
in a turbulent systemfor (Dup to 20% is given by:
d32 = 2x109(1 + 200)(We Re4)-ý (6.3)
It must be noted that this correlation does not take into account the rotor diameter or
(e.
stator geometry g open area, hole dimensions). The for
reason this is the diameter was
not varied for this data set and the screensshowedno difference in d32values.
Figure 6.10 shows the d32values determined experimentally against those calculated
using Equation 6.3. Although there appearsto be certain degreeof scatter in the data the
correlation shows a reasonablefit.
50--
45-
+20%
40-
35- (=x, R2=1
30-
-20 %
25-
20-
15-
10-
vom'
5-
0
0 5 10 15 20 25 30 35 40 45 50
Calculated d32(µm)
Figure 6.10 Measuredd32versus calculatedd32
126
The scatter appears to be more evident at the larger droplet sizes resulting from the higher
dispersed phase concentrations. This is not surprising as these values showed the greatest
deviation from Equation 6.1. However, Figure 6.10 shows that the scatter is within 20%
of the predicted value and it can therefore be said that Equation 6.3 provides a reasonable
estimate of d32values.
The scaling of d32with N"617

should only apply when drop sizesare significantly below
the Kolmogorov Length scale. This should occur in the regions of high energy
dissipation, which should be in the shear gap where skin friction and turbulent losses are
the greatest (Section 2.3.1).
In order to determinean estimateof the Kolmogorov length scaleand also the scaling of
d32with energy dissipation power measurementwere undertaken. The findings of these
resultsare presentedand an explanation of the trends in drop size basedon power

measurements is given in Section 6.4.
6.2.8. Summary
Droplets size data was collected using the Emulsor Screen, ES and the Square Hole High
Shear Screen, SHHSS, with a 69.85 mm rotor. The resultant droplet size distributions
were relatively broad which is indicative of inhomogeneous energy dissipation and also
the occurrence of by-passing.
The droplet sizesdecreasedwith N, however,the droplet size distribution did not change
significantly at speedsof 7000 rpm and above. There were no significant differences in
the droplet sizes from the ES and the SHHSS. Furthermore,the flowrate, Q, only had an
effect on droplet sizes at the lower rotational speeds. Droplet sizeswere seento increase
with dispersedphaseconcentration.
A small number of tests were performedon a 50.8 mm rotor, with four different screens
(Slotted head,SH, GeneralPurposeDisintegrating Head, GPDH in addition to the ES and
127
the SHHSS). The droplet size distributions from the SHHSS and the ES were found to be
similar. The GPDH (with the largest open area)gave the largestdroplet sizes.
The Sautermean diameter,d32,was related to dm. in the following way:
d32 = 0.13dm
The linear relationship validated the assumptionthat drop break-upwas due to a single
mechanismand enabledthe analysis of the drop size data basedon mechanistic

correlations.
As the distributions were similar d32values from the ES and SHHSS were considered
together. The d32was found to correlate with N with an exponent of approximately -0.8,
which was closest to the Chen and Middleman correlation. This implied that the
dominant break-up mechanism was due to inertial stresses in the viscous sub-range.
A linear relationship was found betweendispersedphaseconcentrationand d32and the

following correlation was developed:
d32 = 2x109(1+ 20c)(We Rea)-'"
128
The most comprehensive droplet size data was collected using the 69.85 mm rotor with
the SHHSS and the ES. For this reason power data was collected over the same
experimental conditions for both these screens. To investigate the effects of flowrate on
power draw the experimental matrix was expanded for the ES to include a wider range of
flowrates.
The following sections discuss the results from the individual components that make up
the power balance (Equation 2.49). This data is then used to calculate the total power
supplied to the fluid, which in turn is used to determine energy dissipation and the
Kolmogorov length scale.
6.3.1. Lossesacrossthe control valve and the pipework

The differential pressure was measured across the valve and also before and after each of
the pipe reducers and enlargers. This enabled the calculation of the losses in the
pipework and across the valve.
The head loss in the pipework was low, a typical value at the highest flowrate
investigated was approximately 15W. This value was less than 1% of the corresponding
value for total power input. These losses considered to be insignificant relative to the
total power input and were incorporated into the losses across the valve.
Figure 6.11 shows the power input associatedwith lossesacrossthe control valve for
varying flowrate and a range of rotor speeds. The power input to the fluid by the valve is
high and increaseswith rotor speedand also flowrate. Thesevalues were deductedfrom
the total power to give the power contributed to the fluid from the shaft.
129
900
800
700
600 " 5000

6000
500
" 7000
* 8000
CL400
x 9000
300
" 10000
200 + 11000
100
0I
0 0.2 0.4 0.6 0811.2 14
Flowate (Us)
Figure 6.11 Power input across the control valve versus flowrate.
6.3.2. PSHAFT
The results of the shaft power measurements are shown in Figures 6.12 and 6.13 for the
Emulsor Screen and the Square Hole High Shear Screen respectively. Figure 6.14 shows
a comparison of the power drawn by the two screens at similar flowrates.
As expected,the shaft power increaseswith rotational speedand there also appearsto be
greaterpower draw at the higher flowrates. The influence of flowrate on power draw
appearsto be more significant at the higher rotational speeds.
There doesnot appearto be a significant difference betweenthe power drawn by the two
different screens(Figure 6.14). There is somescatterat the highest rotational speed,
however no valid inferencescan be madeon this single data set.
130
4000
3500
3000
2500
.
2000 x "
CL
1500 "
x
a
1000 x "
500
I
V
0
0 2000 4000 6000 8000 10000 12000 14000
N (rpm)
"0.41/s f0.61/s 0.81/s f1.0Vs x1.21/s
Figure 6.12 Shaft power (W) versus rotational speed(RPM) for the Emulsor Screen
4000
3500 .
3000
U
2500
2000
a
1500
1000 "
i
500
0
0 2000 4000 6000 8000 10000 12000 14000
N (RPM)
. 0.4 us 0 0.8 us
Figure 6.13 Shaft power (W) versus rotational speed(RPM) for the SquareHole High
ShearScreen
131
4000
3500
3000
2500
J
2000
1500
1000
500
0i -ý iii
0 2000 4000 6000 8000 10000 12000
N (RPM)
f SHHSS 0.8 I/s SHHSS 04 Is " ES 0.8 I/S x ES 0.4 I/s
Figure 6.14 A comparison of values for PBhsafrom the ES and the SHHSS at 0.41/s and
0.81/s.
Figures 6.12 to 6.14 shows that shaft power increases with rotational speed and flowrate,
the rotor diameter was not varied in these tests. The similarities in the power draw for the
two screens suggest that the main parameters effecting power draw in a rotor-stator mixer
are the rotational speed and perhaps the diameter of the rotor.
It is also a reflection on the effect of sheargap on power draw, asthe clearancebetween

the rotor and the stator are markedly different (0.475 mm for the ES and 0.125 mm for
the SHHSS).This suggeststhat lossesdue to meanvelocity gradientsin the sheargap are
insignificant supporting claims madeby Pedrocchiand Widmer (1988) and Calabreseet
al (2000).
132
It is interesting to note that the open area is similar for both screens (Table 4.1). It was
mentioned that the stator performed a flow straightening effect as fluid discharged
tangentially by the rotor impacted into the stator wall. This results in a transformation of
high kinetic energy into pumping power, skin friction losses and turbulent losses. It was
suggested that these losses were greatest where the fluid turned abruptly in the stator
hole.
Increasingthe number of holes in the stator would effectively increasetheselosses. The

fact that the open areais a direct function of the number of holes suggeststhat it would
display an influence on the power draw of the mixer. This effect was not investigated
further in terms of power dissipation but the droplet size data collected for all the
different screensfor the 50.8 mm rotor seemedto favour this theory.
It can be assumedthat the power would dependon the rotational speedand the flowrate
by the following relationship:
' = PN°Qb (6.4)

1L4Fr
Where, P is analogous to the power number although this equation is not dimensionless.
The effect of rotational speed and flowrate was analysed by performing a multiple
regression on all the data collected. Values for Pn, a and b were found to be 0.008,3.0
and 0.25 respectively with an R2 value of 0.98, showing good correlation.
Sparksreportedvalues between2 and 2.5 for the exponenton N and 0.8 to 1 for the
exponent on Q. Values ranging from 14.6 to 62.5 were given for P.
The lower value of the exponenton Q could be due to the fact that the flowrates
investigatedin this work are significantly lower than those usedby Sparks(1996). Values
for P. would be dependenton a number of variables including the geometryof the mixer.
133
Figure 6.15 shows the values calculated using Equation (6.4) plotted against measured
values. It can be seen that the calculated values match the measured values well and that
Equation (6.4) provides a good prediction for the values of Psx.&yT.
400(
350(
300(
250(
200(
Üy
p! 1501
100(
50(
0
0 500 1000 1500 2000 2500 3000 3500 4000
MeasuredPhg (W)
Figure 6.15 Measured PSS data versus correlation
6.3.3. PpuMP
It has previously been statedthat in-line rotor-statormixers can be consideredas

inefficient pumps. It is important to note that although the pumping effect of these
mixers may help in reducing pumping requirements of a process they are not designed for
pumping.
134
Figure 6.16 shows the pumping power plotted against flowrate for the ES. Predictably,
the pumping power increaseswith flowrate and rotational speed.The curves for the lower
rotor speedsare flatter than that for the higher rotor Also,
speeds. the effect of rotor speed
is greaterat the higher flowrates.
800--
700-
600-
500-
400-
300-
200-
xý ý=
100-
0.
0 0.5 1 1.5
Flowrate (1/s)
-.-- 5000 rpm t 6000 rpm -*- 7000rpm -x- 8000 rpm -+-9000 rpm t 10000 rpm + 11000 rpm
Figure 6.16 Pumping Power versus Flowrate for a range of shaft speed, 69.85 mm
rotor and Emulsor screen.
The shapeof the curves for rotational speedsof 8000 rpm and above appearsto be
increasing. This suggeststhat higher flowrates are possibleand that the mixer is
operatingat the lower end of the pumping curve (i. e. the best pumping efficiency point
hasnot been reached).
135
The pumping efficiencies were calculated to be between 15% and 20%, these figures are
very low compared to centrifugal pumps, for which efficiencies close to 100% are
possible. This is an encouraging sign when considering the mixing performance, as in
this region of operation the losses useful to mixing (hydraulic and leakage) are greater.
Figure 6.17 shows a comparison of values of pumping power from both the SHHSS and
the ES. It can be seen that there is no significant difference in pumping power values for
both these screens. The corresponding values for Ppumpfrom both screens at constant
flowrate and varying N are very close to each other.
600
11000rpm x
500
X
10000 M. X
400
300
X 11000rpm
8000 rpm X
10000n.
200 x
7000 rpm
8000 rpm
100 7000 rpm

3000 rpm x
x 5000 rpm
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9
Flowrate (I/s)
X SquareHole High ShearScreen Emulsor Screen
Figure 6.17 A comparisonof pumping power values from both the ES and the SHHSS
at different flowrates (for increasingN).
136
Again, this similarity points to the rotor diameter and possibly the open area, in addition
to rotational speed, as being the main influences on power draw. Although power data
was not collected for different open areas, it is reasonable to believe that the open area
may have a significant effect on pumping power. Centrifugal pumps are designed for
optimum pressure recovery (the process of converting kinetic energy from the rotor to
pumping output), however rotor-stator mixers are not.
The pressure recovery process in a rotor-stator mixer comes from the flow straightening
effect of the stator, which does not occur without significant turbulent and frictional
losses (Sparks, 1996). This is because fluid is transmitted tangentially with high kinetic
energy by the rotor blade. It then impacts on the stator and exits in a predominantly
radial direction.
Therefore, the main losses in a rotor-stator mixer are due to the acceleration of the fluid
in the rotor and the abrupt deceleration of the fluid in the stator. A smaller open area
would increase the amount of fluid that impacts on the stator and thus reduce the power
contributed to pumping. The amount of these losses should then depend on the stator
design and open area. It would seem reasonable to believe that these losses are
advantageous in the generation of dispersions as they would contribute to the turbulent

energy dissipation.
The benefits are reflected to a certain degreeby Figure 6.6 which showedd32values from
the dispersiontests performed on the 50.8mm rotor with the four different screens.
Smaller droplet sizeswere seenfrom the SquareHole High ShearScreenand the
Emulsor Screen(with the smallestopen area),whereas,the GPDH (with the largestopen
area)producedthe largestdroplets.
6.3.4. PMECH
and PDISc
The mechanicallosseswere measuredby spinning the shaft without the rotor attached.
The mixing head was filled with water and the inlet and the outlet was blanked off. The
temperatureof the water and the time was measured.The resultanttemperaturerise of the
137
water was due to friction within the seal and the bearing and also the drag on the shaft
and spring.
It was not possibleto measurethe lossesassociatedwith disc friction and thereforean
empirical correlation (Equation 2.54) was usedto calculatethem (Karassik, 1986).
Figure 6.18 shows the measured values for PMEcHand PDISC.It can be seen that both
these losses increase with rotational speed. The highest value of PMECHis close to 40 W
(for 12000 rpm) which, compared to the equivalent shaft power, approximately 3 kW.
Therefore, it can be said that at the highest shaft speeds, where the mechanical losses are
greatest, they account for approximately 1% of the total shaft power. The corresponding
value for disc friction (_ 300 W) was approximately 10% of the total shaft power input.
350
300
250
200
0
a 150
100
50
0
0 2000 4000 6000 8000 10000 12000 14000
N (RPM)
Figure 6.18 Mechanical and Disc frictional lossesversusrotational speed.
138
6.3.5. Power supplied to the fluid by the rotor, PFLUID
The contribution of the mechanical and disc friction losses to mixing and pumping in a
rotor-stator mixer is negligible. The total power supplied to the fluid associated with
pumping and mixing, is
PFLUnD, the remainder of the power balance (i. e. energy supplied
to the fluid by Hydraulic and Leakage losses):
PFIWD= PLEAK + PPLaP= PSHAn- (PkEcH+ PDIsc)

+ PHYDRAULIC (6.5)
As values for shaft and pumping power were not significantly different between the ES
and the SHHSS,all the power data collected was consideredin the following analysis.
As there is very little information on power draw for rotor-stator mixers the following
analysiswas done in terms of power number,(similar to that for an impeller in a stirred

tank). For turbulent systems the power dissipation would be expected to be proportional
to 3
N.
Flow visualisation work done by Sparks(1996) showedthat the fluid at the tip of the
rotor blade follows the rotor at the samevelocity as the tip speed,v1; Therefore, the
p.
kinetic energy imparted to the fluid, of mass m, at the rotor tip is given by:
KE =2 myp (6.6)
The equivalent power is then:
P=2OQvp (6.7)
However, vtp can be expressed as ; rND giving:
139
PaWD=I ;r2,OQN2D2 (6.8)
When the rotor-stator mixer is operating without a supplementary pump, the flowrate can
be expressed as (Sparks, 1996):
QacND3 (6.9)
Therefore, the following expression can be obtained for power supplied to a fluid in an
in-line rotor-stator mixer:
PFLUID= F, 3D5 (6.10)
Where, the constantof proportionality Po is the power number, a dimensionlessgroup

that is a function of the geometric,kinematic and dynamic propertiesof the system.
Figure 6.21 shows a plot of PErro versus rotational speed and the corresponding results
from the regression performed on all of the data. The exponent on N is 3 (with an R2
value of 0.94 showing a good correlation) which is what would be expected for fully
turbulent conditions. Equation (6.10) can therefore be used to determine the power
number and hence calculate the power supplied to the fluid by the rotor.
From Equation (6.10), the power number is:
PFLUID
Po= (6.11)
pD5N3
As the values for Pn, UIDscale with N3, the power number can be determined from Figure
6.19.
140
3000
2500
2000
1500
1000
500
0I ri
0 2000 4000 6000 8000 10000 12000 14000
N (RPM)
Figure 6.19 PF,UIDversus N and the dependency of PFLUIDon N.
1.8E+06
1.6E+06
y=0.2081x
1.4E+06 R2 = 0.9476
,"
"
1.2E+06 "
1.0E+06
8.0E+05
aj
"I
6.0E+05
"
4.0E+05
"
2.0E+05 ""
0.0E+00
0.0E+00 1.0E+06 2.0E+06 3.0E+06 4.0E+06 5.0E+06 6.0E+06 7.0E+06 8.0E+06
Na
Figure 6.20 P D/pD5versus N3.

,u
141
Figure 6.20 shows the power number to be 0.2 which is low compared to values quoted
by Sparks (1996) who evaluated power numbers in the region of 10 for both the Toothed
design and the Silverson design. Francis (1999) reported values between 2.5 and 4.5 in a
laboratory scale batch rotor-stator with various stators (none of which were similar to the
in
ones studied this power work).
In fully turbulent conditions the power number is constant and is dependent only on the
geometry of the rotor. A possible reason for this low power number is that a valve is
used to control the flow, reducing the pumping required of the mixer. The shaft power
was seen to increase with flowrate suggesting that power draw would increase if the
mixer was left to pump by itself
Equation (6.10) defines the energydissipatedto the fluid by the rotor in the rotor
chamber.It is in this region of the mixer where it is thought that the maximum turbulent
energy will be dissipated
Figure 6.19 show that it is reasonableto assumethat the mixer is exhibiting turbulent
characteristics,at least within the rotor chamber. The energydissipation rateswill vary
throughout the rotor-stator mixer with a maximum value around the stator region (where
the fluid impinges on the stator).
The broad droplet size distributions generatedby a single-passthrough the mixer adds
further weight to this observation. This also leadsto the conclusion that the majority of
the droplets are not exposed,either at all or for long enough,to the regions of high energy
dissipation.
63.6. Energy dissipation and the Kolmogorov length scale
The energy dissipation rate, e, was discussedin Section2.3.4 and it was defined as the
power supplied to the fluid by the rotor divided by the massof the fluid. It can be
assumedthat the pumping power has little influence on mixing. It be

can also assumed
142
that the hydraulic losses would be responsible for the majority of the turbulent mixing
that occurs in a rotor-stator mixer. Therefore, referring back to the power balance
(Equation (2.49)) the power contributed to useful mixing is represented by PHYDRAULIC
+
PLEAKAGE.
Various researchers have used different definitions of the mass (or volume) of fluid
depending on the area in the mixer considered to have the maximum energy dissipation.
The rotor supplies energy to a small region of the rotor-stator mixer encompassing the
rotor, the stator holes and the shear gap. Consequently, the total volume of this region of
the mixer (0.064 litres) was used when calculating the energy dissipation rates.
Sparks(1996) estimated,from diazo-coupling experimentsthat lessthan 3% of the power
supplied to the fluid was dissipatedas turbulent energy,c. This would appearto be a
very small percentage,however he reasonedthat there was considerableby-passingand
inhomogeneityin the turbulencefield meaningthat a proportion of the fluid passing
through the mixer may not experiencethe highest energydissipation rates.
Subsequently, the values calculated for energy dissipation rates ranged from 70 W/kg at
5000 rpm to 800 W/kg at 11000 rpm. These values were then used in Equation 2.6 to
determine the Kolmogorov length scale, Xk. Figure 6.21 shows Ilk versus rotational speed
for all the power data collected.
In a stirred tank E is an undeterminedfunction of the spatial coordinatesof the tank. For
sufficiently long residencetimes all drops in a dispersionwill samplethe full spectrumof

energy dissipation. Therefore,D... will be determinedby the largest values of C. It is
worth noting that the energy dissipation rateswill vary throughout the rotor-stator mixer,
with a maximum value around the stator region (where the fluid from the rotor impinges
on the stator). The values for Xk would be smaller in this areathan those calculatedfor
the mixer volume as a whole.
143
14
12
10
8
E
0
3000 4000 5000 6000 7000 8000 9000 10000 11000 12000
N (RPM)
Figure 6.21 The Kolmogorov length scale versus impeller speed
6.3.7. Summary
A thermometric technique was employed to determine the power the in-line rotor-stator
mixer supplied to the fluid. This then enabled values for c to be estimated which in turn
to
were used estimate Xk.
Pshaft
was found to increasewith N and Q. The SHHSS and the ES showedvery similar
which
power characteristics, led to the inferencethat the effect of sheargap was
insignificant. Also, it was reasonedthat the open areaof the screenwas an important
factor in the power of the mixer.
144
This was supportedby the droplet size data collected from the 50.8 mm rotor. The
GPDH (largest open area)generatedthe largest sizesand the SHHSS and the ES
(smallestand similar open areas)createdthe smallestdroplets.
(which would be useful in sizing motors) was

A generalcorrelation to estimatePehaft
developed for flowrates ranging from 0.41/s to 1.21/s:
PSWI= 0.008xN' xQo''5
The pumping power of the rotor increasedwith Q and N and as for Pshaftthe open area
,
was also thought to be influential.
The power supplied to the fluid in terms of pumping and hydraulic losses,PFLUID was
found to scalewith N. This led to the deduction that the mixer was operating in the
turbulent regime and the power number was 0.2.
Values for e were determined by assuming a proportion (<3% as reported by Sparks,
1996) of PF,uu was converted to turbulent energy. Consequently, values ranged from 40
W/kg at 5000 rpm and 800 W/kg at 11000 rpm. The corresponding values for Xk were
then estimated to range from 12µm (at 5000 rpm) to 6µm (at 12000 rpm).
6.4. An explanation of the trends in droplet size based on energy dissipation
Figure 6.21 showsthat values of kk rangefrom approximately 12µm down to 6µm as the
rotational speed varies from 5000 rpm to 11500 rpm. Figure 6.22 shows a comparison of
thesevaluesto d32(from Figure 6.5).
Figure 6.22 showsthat values of d32are approximately twice the size of Xkat 4= 10%
and up to 5 times at 4= 30%. This suggeststhat D >d32> Xkwhich are more appropriate
conditions for drop break-up due to the inertial sub-rangeeddiesrather than the viscous
145
sub-range as implied by Equation (6.3). It would appear that d32at the lowest dispersed
phaseconcentrationsgeneratesdroplets closestto Xk. This again points to the fact that an

increasein dispersedphaseconcentrationdampensthe formation of turbulent stresses.
60
50
40
3 30
20
I0
0
4000 6000 8000 10000 12000
N (rpm)
ES: 0.4Us& 10% 0 ES: 0.4Vs& 20% f ES: 0.41/s& 30%X lk
Figure 6.22 A comparisonof d32and XkversusN (at 0.41/sfor ES)
Droplets ranging from approximately 100 µm with the main peak at around 50 µm and
the smaller range below the 10 in

µm were observed the volume frequencycurves
(Figure A. 4 and A. 12). The broad size distributions and the fact a percentageof the
distributions are below the smallestscaleof turbulence would suggestthat not all of the
droplets are passingthrough the areasof high energydissipation (that determinethe
146
Kolmogorov length scale). This would further suggestthat break-up is more than likely
causedby more than one mechanism.
The determinationof the dominant drop break-upmechanismcan also be basedon the

dependencyof d32on energydissipation rates,c, (Equations,2.10,2.17,2.21,2.24). The
results of a regressionperformed on d32plotted againste are presentedin Table 6.4.
Table 6.4 The best fit exponents on E for varying t
(%) E= R
05 -0.21 0.82
10 -0.22 0.75
20 -0.19 0.74
30 -0.13 0.80
Although, there is a certain degreeof scatter in the results (the R2values show a
significant deviation from 1) the exponentson c closely match -0.2 similar to that in the
Shinnar correlation for droplet break-up in the viscous sub-range. This also corresponds
with values reported by Schubert(1997).
When correlating the size data with energydissipation ratesthe dominant drop-break
mechanism is found to be inertial stressesin the viscous sub-range.This was similar to

observations made when correlating against rotor speed(albeit with a different form of
the energyequation). However, as highlighted previously, both thesemodels should only
hold if d32is lower than Xk.
147
It is worthwhile to draw further attention to the values of ?1kfor a possible explanation in
the discrepancies between the correlations and the drops sizes relative to Xk. The
Kolmogorov length scale was calculated using the average energy dissipation rate. The
smallest eddy sizes would be generated in the region of the mixer that exhibits the highest
energy dissipation rates.
It has previously been discussedthat the energydissipation in the rotor-stator mixer
exhibits inhomogeneous
tendencies. The number of droplets that would come into
contactwith this region, in a single passwould be small. This high localised e would
accountfor the droplet sizesbelow 10 µm. This would that
suggest d32is not a result of
the highest energy dissipation rate and may be the reasonfor the possible discrepanciesin
the models relative to the applicable size ranges.
Figure 6.22 showsthat the droplet sizes closestto Xkare generatedat the lowest dispersed
phaseconcentrations. It is possiblethat at this dispersedphaseconcentration,larger

dropletsare being more effectively broken up before they enter the region of high energy
dissipation rates (as there is more turbulent flow throughout the rotor region). This
would mean that the number of droplets will grow, increasingthe likelihood of them
coming into contact with the high e zones.
It must be emphasised that the approach used to identify the dominant break-up
is
mechanism a simplistic method of determining an understanding of the complexities of
the drop break-up process. At any given point and time in the mixer one mechanism may
be responsible for droplet break-up, however it is more than likely that more then one
mechanism is responsible for the final distribution. By comparing Xk with the droplet
size distributions it can reasoned that macro-scale eddies, inertial stressesand turbulent
viscous shear are responsible for droplet break-up.
The following section is an analysisbasedon Reynolds numbersand residencetimes (as

defined in Section 2.3.1) to determinewhich regions in the mixer are most likely to
determinethe droplet size distribution.
148
6.5. Flow regimes in the rotor-stator mixer: Re analysis
The following section discussesthe flow in the various regions of the mixer (as defined
in Section2.3.1). An attempt is madeto determinewhether the time scaleswould allow
turbulent or viscous stressesto causedrop break-up in region of the mixer.
The in-line rotor-stator mixer is divided into the following sections,as defined in detail in
Section 2.3.1; the inlet pipe, rotor region, shear gap, stator and volute. A summaryof the
calculatedresults can be found in Table A. 2, Appendix 5.
6.5.1. Inlet pipe
The effect of inlet flows on the droplet size distribution was first discussed in Section
6.2.4. It was found that the droplet size distributions showed bi-modal tendencies with
drop diameters observed above 1000 µm, which is beyond the measurement range of the
Malvern Mastersizer. A possible explanation for this occurring was that the inlet flows to
the mixer were not well mixed and therefore large globules of dispersed phase were
passing through the mixer without coming into contact with the continuous phase and the
areas of high energy dissipation.
The pipe Re number was estimated using Equation (2.34) and values in the order of
20,000 and 40,000 were calculated (at 0.41/s and 0.81/s respectively). This would
indicate the flow in the inlet pipe is fully turbulent.
In turbulent flow, if drop break-up is due to viscous shearingaction the un-deformed drop
and resulting elongateddrop be

must small comparedwith the local regions of viscous
flow (Hinze, 1955). However, when the Re is relatively high, such as in the inlet pipe,
the spatial dimensionsof such local regions would be very small comparedwith the
largestdrops in the system. Therefore it would be a reasonableassumptionthat inertial
stressesare responsiblefor the maximum droplet sizes in the inlet pipe.
149
The following is a crude analysis to determine whether it is feasible for turbulent eddies
to be the main cause of drop break-up. The assumption of homogenous local isotropy is
valid over the volume of pipe containing a turbulent liquid (Davies, 1985, Hinze, 1955).
Equation (2.18) can therefore be used in order to estimate the maximum stable drop size
that would exist in the in-let pipe before entering the mixer. (The velocity of the bulk
fluid in the pipe and the pipe diameter were used as the characteristic scales in the
calculation of We and the constant of proportionality was assumed to be 1).
Values for dmaxwere calculated to be 800µm and 400µm (at 0.41/s and 0.81/s
respectively). Predictably a2 fold increase in pipe velocity results in a2 fold decrease in

dma,,.These calculated values of dm. can be inserted into Equation (2.13) to back
calculate the energy dissipation rate required to create droplets of this size. The values
fore were calculated as 20 W/kg at 0.41/s and 150 W/kg at 0.8 Us.
The time scaleof the duration of an eddy causing drop break up can be estimatedby
assumingthat the eddy length is the samesize as dm,,,and the instantaneousvelocity
acting on the drop is the maximum in
velocity the pipe. Therefore,the time scaleof an
eddy responsiblefor drop break-up(dm/u) is in the order of 10'6s. This is significantly
smaller than the nominal residencetimes in the pipe which are in the order of 10-2s.
Hinze (1955) reasonedthat in turbulent flow the areaaround the drop, plane hyperbolic
(extensional),axi-symmetric and Couetteflow are most likely to be responsiblefor
breaking up the largest drops. The length of time a drop must be exposedin the Couette
and Extensional flows (for a Kerosene/waterviscosity ratio of 2) for break-up to occur
from Pa
can be determined the plots developedby Grace (1982). The high of 2 would
'tic
indicate that drops are less likely to break in simple shearflows (Couette flows).
Values of reducedburst time can be evaluatedfrom Figures 2.15 and 2.19, and the
calculatedvalues of dmax
can be used to back calculate the sheartimes required for break-
150
up (Table A. 2). Thus, the corresponding shear times in Couette flow and Extensional
flow are in the order of 10'3s.
Again, thesevaluesare smaller than the nominal residencetime in the pipe but
significantly larger than eddy length times. Thus, it would seemreasonableto believe
that there is not any significant stretching of the drops before break up by turbulent
eddies. Also, it can be concludedthat inertial stressesare predominantly responsiblefor
drop break up.
The value of dmaXfor a flowrate of 0.41/s was calculated to be 800µm. This is close to the
maximum measurable size limits of the Malvern Mastersizer. Droplet sizes were
observed in this size range and above at the lower rotor speeds and higher dispersed
phase concentrations (Figures A. 6; A. 7; A. 11; A. 13 & A. 14). This figure would initially
seem a reasonable estimate of dm being fed to the mixer.
However, the effect of the swirling flows should increasewith an increasein rotational
speedand as a result an increase in the suction by

created the rotor (Sparks, 1996). This
should lead to increasedturbulent energy at the inlet of the mixer and thus the globule
sizesentering the mixer should decrease.
Also, Equation (2.18) does not take into account the dispersed phase concentration.
However, it was seen that the droplet sizes increased with an increase in c, as observed
by Brown and Pitt (1971); Streiff et al (1997) and Doulah (1975). An in increase in 4 is
thought to dampen turbulence and it would seem reasonable to believe that the inlet
would contain larger globules than 800µm.
This would seemreasonableas pi is twice that of µoand as 4 increasesthe flows can no

longer be describedby the continuousphase. The inertial stressesat high 4 would
thereforebe reducedby the higher viscosity of the dispersedphase. Therefore, at higher

dispersedphaseconcentrationsdrop break-upwould occur due to both viscous and
inertial forces.
151
6.5.2. Rotor region
Sparks (1996) described the flow of fluid entering the rotor region as being accelerated
tangentially and the flow pattern upstream of the rotor teeth could be characterised by
swirl generated as the fluid followed the rotor face. The flow in this region was
considered to be predominantly tangential, with a much lower radial component. As the

fluid was seento follow the rotor teeth the velocity of the fluid at the rotor tip, v,; was
p,
consideredto be the sameas the tip speedof the rotor.
The Silverson rotor used in this work has a greateropen areathan the toothed designs
used in the flow visualisation work of Sparks (1996). This will result in a lower radial
componentof the velocity (Q/A) in this region. Also, the rotor blades on the Silverson
mixer (close in design to a pump impeller) are significantly longer than the teeth on the
unit investigatedby Sparks (1996). This will increasethe residencetime of the fluid in
this region.
The Reynolds numbers in this region, as defined by Equation (2.38) were calculated to be
4000 and 8000 (at 0.41/s and 0.81/s respectively). These, relatively low Reynolds
numbers would suggest that turbulence levels are low and the assumption of local
isotropy may not be applicable. In addition, at low Reynolds numbers the spatial
dimensions of the viscous flow will be larger and therefore drop break-up due to viscous
shearing action is possible as well as inertial forces.
However, the droplet size was seen to decrease with rotational speed up to approximately
7000rpm after which no significant change was observed. The definition of Re as given
in Equation (2.38), is only dependant on the bulk flowrate and open area of the rotor (as
=
Vradial Q/A), as well as the physical properties of the liquid system.
Therefore, if drop break-up was due to inertial stressesor viscous forces in the rotor
region, Equation (2.38) would indicate that an increasein flowrate in

would result a
decreasein droplet size. However, the affect of flowrate appearedto be insignificant
comparedto the rotor speed. Thus, it can be concludedthat the rotor speedis more likely
152
to affect drop break-up and that the definition of Re in terms of v,,,d;d is not a true
reflection of the flow in
velocities this region.
This is not totally unexpectedas Equation (2.38) is an over simplification of the flow in
the rotor, as in addition to the radial componentthere is also a tangential componentof
the velocity. In Section 2.3.1 a seconddefinition of Re in the rotor region was discussed
remainsthe characteristiclength but the rotor tip
for tangential flow, where DRADIAL
v,
speed, ip,is used as the characteristicvelocity. This then results in the impeller
Reynoldsnumber (Equation (2.22)) divided by a factor of 2, as DIALD/2.
If the two velocity components are considered to be vectors, the sum of the two vectors
gives a resultant velocity close to vtp. This is because vPADI, (0.12 msl at 0.41/s) is
significantly smaller than vtp, (10.97 ms" at 3000rpm). It would thus seem reasonable to
assume vtp is a more appropriate characteristic velocity in the rotor region. This then
results in significantly larger values for Re (4x105 at 3000rpm) which would then infer
greater turbulence and the likelihood of drop break-up by inertial stresses. It must also be
noted that these larger values of velocities would also lead to greater nominal shear rates
(from 300s' at 3000rpm up to 1300s'' at 12000rpm).
The general trend in the droplet size data (Figure 6.5) showed that droplet size decreases
with N up to 7000rpm with only small changes in droplet size distributions at rotor
speeds above this. Hence, as Re becomes greater the sizes are seen to decrease, however
once turbulence is fully generated at speeds of 7000rpm and above no significant change
in droplet size is observed. This would imply that break-up due to inertial stresses is only
prominent at speeds below 7000rpm as above this higher Re do no correspond with
significant changes to the droplet size distribution.
Further explanation into the effects of this possibleflow affect on the droplet size
distribution can be obtainedby consideringthe droplet size data more closely. The
droplet size data from the Emulsor screenat 0.41/sand 4= 10% is shown in Figure 6.24.
153
25
" 3000 rpm 1

5000 rpm
20 ---
" 6000 rpm
7000 rpm
a 8000 rpm
15 " 9000 rpm
---
° 10000 rpm
"
11000 rpm
11610 rpm!
Z1
10 ; "
.- "'"
5 _-r
ý+ "" "
LY-.rFt-x-. Q, Y 6i00ýýýýýýiý fýýA t 1i"
0 ýe aýaý a iAtlI o..
T_i
ol1 10 100 1000
Droplet Diameter (microns)
Figure 6.23 Volume DSD generated by the 67.85mm rotor with the Ernulsor Screen at
Q=0.4 1/s and cD = 10%, for varying N (rpm).
As rotor speed increases (and thus Re) values of d,,,,,, appear to decrease from
approximately 300µm at 3000rpm down to 150µm at 7000rpm. At speeds above this
there is no significant change in the droplet sizes at the larger end of the droplet size
distribution (i. e. d,,,,, appears to level off at approximately 150µm). However, rotational
speeds above 7000rpm have more of an effect on the smaller droplet sizes in the size
distribution (< 10µm).
If viscous stresses are responsible for drop break-up the values of critical shear rates
required to produce droplet sizes equivalent to dma can be determined using the plots of
Grace (1982), (Figures 2.13 and 2.16 & Equation (2.42)).
154
The nominal residencetime at 0.41/s in this region is 0.3s and nominal shearratesfor
N
varying and the correspondingRe, critical shearrates, Gg and burst times, tB (Grace,
1982) for dmax(0.41/s and 4=10%) are summarised in Table 6.5. It is worth reiterating
that 'ud =2 would indicate that drops are less likely to break in simple shearflows
(Couetteflows) and more likely to break due to extensionalflows (Grace, 1982).
Table 6.5 dmaxwith corresponding Re and y' and Gg values in the Rotor
Region (69.85mm rotor, ES, 0.41/s and ý=10%) with Xk
N Vtip y" dmax Extensional

Critical shear, Xk
(rpm) (m/s) Re (S-1) (µm) tB (s)
Ga (s') (µm
5000 18.29 1.3E+06 524 191 171442 0.037 4.70
6000 21.94 1.5E+06 628 141 232703 0.028 4.20
7000 25.60 1.8E+06 733 141 232703 0.028 3.73
8000 29.26 2.0E+06 838 141 232703 0.028 3.40
9000 32.92 2.3E+06 942 141 232703 0.028 3.07
10000 36.57 2.6E+06 1047 121 271079 0.024 2.80
11000 40.23 2.8E+06 1152 121 271079 0.024 2.53
Values for the critical extensionalshearrate are significantly higher (up to 300 times)
than the nominal shearratesin this region. It is worth noting that Grace (1982) found
that calculatedcritical shearratesunder extensionalflows were proportional to the square
root of the wall shear rate in a static mixer.
dmacappearsto level off at the higher rotational speeds,suggestingthat a maximum eddy

length scaleof ; 150µm may exist in this region. This is despitean increasein turbulent
155
energy dissipation rates. Furthermore, there is an increasein the number of smaller drops
in the order of, if not smaller, than the Kolmogorov scale of turbulence.
Thesesmaller drops could be attributed to the fact that if the drops were breaking due to
Couetteor Extensional flows they would break into more than two separatedrops (Figure
2.12). If the drops underwent large enoughdeformation they would split into more than
two drops and the resultant drops will be smaller than if they underwent smaller
deformations (Grace, 1982).
This analysisassumesthat the velocity of a drop is well defined by the tip speed.
However, the rotor can be consideredsimilar to an agitator in a stirred vessel.
Consequently,it can be reasonedthat turbulent vortices would exist at the trailing edges
of the rotor blade. The local energy dissipation rates in this region would be high. The
distancethe droplet will travel tangentially in an equivalent time, at a speedv';F, (10.97
ms-1at 3000 rpm) will therefore be approximately 3m. This would suggestthat the path
by which a droplet travels through the rotor region would be close to a spiral.
Also, the larger open area of the Silverson rotor may result in a certain amount of back
mixing as material is being drawn back into the rotor region to replacethe fluid displaced
by the rotor. This would indicate that the flow through this region is highly
inhomogeneous.
It can therefore be concludedthat drop break-upat the lower rotational speeds(up to

7000rpm) is due to inertial and viscous stresses.As the speedis increasedthe inertial
stressesis fully formed and inertial stressesthen becomemore dominant. The largest
length scales will be in the order of dm8,,.

However, it is evident that the number of
eddy
drops, in the order of or below Xk,continuesto increaseat thesehigher speeds
smaller
break-up due to viscous forces.
suggestthat also occurring
The values of Re in this region would suggestthat it is more likely that inertial stresses
for the smaller drops. However, the turbulent length scaleswill be
are responsible
156
smallerand the shearrates will be greater in the sheargap. It seemsmore likely that the
smaller droplets be in
would generated this region of the mixer. This is discussedin
further detail in the following section.
6.5.3. Shear Gap
The Emulsor Screen shear gap, (0.475 mm) is almost 4 times the size of the Square Hole
High Shear Screen shear gap (0.125). The Reynolds numbers (Equation 2.39) in this
region are low compared to those calculated in the rotor region (as S is «DRADU, ).
Consequently, the shear rates are significantly higher in this region of the mixer.
The Resfor the two screensare shown in Table 6.6.
Values for Res for the ES range from 5000 (at 3000rpm) to 20,000 (at 12000 rpm).
Whereasfor the sameN values for the SHHSSrange from 1400to 5000. This would
indicate that the viscous forces are more prominent in this region of the mixer. The
from 23,000s 1 to 90,000x1 90,000s 1
corresponding nominal shear rates range and to
300,000s' respectively.
Despite the large difference in 5 betweenthe two screens,the droplet size distributions
generated by the two screens are quite similar. This implies that the size of the shear gap
has little effect on the overall droplet size distribution.
However, the inferencesmaderegardingthe flows in the rotor region suggestedthat the

higher rotor speedshad a greateraffect on the smaller droplet sizes. For break-up of the
smaller drops to occur, either the turbulence length scaleshave to be small enough(in the
order of or smaller than the drop) or the shearrates have to be high. Also the relevant
times scaleshave to be sufficient to causebreak-up. This would suggestthat it is more
likely that break-up of the smaller dropletswould occur in the sheargap region of the
mixer.
157
Table 6.6 Res for the ES and SHHSS (69.85 mm)
N Res
(rpm) ES SHHSS
3000 5212 1372
5000 8686 2286

6000 10423 2743
7000 12161 3200
8000 13898 3657
9000 15635 4115

10000 17372 4572
11000 19110 5029
12000 20847 5486
The determinationof whether it is the shearforces or inertial stressesthat are dominant in

the sheargap follows. The values of critical shearratesand burst times for drop
diameters equivalent to dmaxand d32from both the ES and SHHSS (at 0.41/s and 4=10%)
can be determined from Figures 2.13 and 2.16 & Equation (2.42) and are summarised in
Table 6.7.
Let us first consider the break-up of drops of sizessimilar to dm in both Couetteand

Extensionalflow. In the ES the values of critical shearrates(Ge) calculated(Table 6.7)
are approximately 10 times and 4 times the nominal shearrates (y') in this region for
Couetteand Extensional flows respectively. The correspondingvalues in the SHHSS are

3 times and 1 times y'.
158
WWWWWWW
ý
vý N Gý ýD MOr '.D '.Q h ýi ' MMN
ooýcr: q tt rý MLON
-ce; C
a csrao ý. '- -NNN N0M
z merýn
I
9 C?ril C?19
NQ
x wW wW14w iah k GUW W WW W fs:
h (ýj Yl Vr 41 eD owl in '.D M
i-ý P r t- Mm M Nn hrýPtlh
M NNNN MNNNNN
r4 rl
E
ý' w ýy MMMM MM
y lH MMMM to OOOOcoo
,8 t oc4gQq £ tý fs5 W L] f:ý Lý IL
V wwwwwww
asp=OOP-N0O
----
CD
orqova U
"- bý oý r +ö ýc ýn
cn MMMMMMM
t''1 MMMM C1 C1 <? x? 990 OO
vi 000000
0
ºre ti7 Ltý X71
Qý 1ý yG G Q
hß-1 (-ý 0
O rý1 -0 00 N
Ný ýh
N '7'Cf NNN-N
Y lV [ý NNN
2-0 : ýt
L
V
ö4 öcacö4
bw 0 t d wwWUww
WwwwcnWL:
r- Z; C7. rn V) ri - -NUC
6 V: ô00 ON 00 kc 10 ýc
0 ONN
'n in in V1 qA in in tn *41 In Yl In an
0000000 00 00O Qp
20
W w(I M Mh
ß QMMlNNh
C`ý 1 ý" M
eýi Y'*i t+l tV ri fei
-NNNNNN w
a> ý
1 4) 4"1 At) in in qn tn in
0000000 V) 40 0
y W OWN
" ww*wwc+u o Q
V V OD 00 00 a a+ cy
o r) oc Ooco aý Qa Qo u
V1 Vh Vh V' lý lý Q+
U 'i ini i a 'D
.0~
bb in
OcO
m^' -i19 Mo19
S W
QO
W
N tY ý- M
w llý ý" "^ t+I cV r*1 m e'i
-NNNNMN ýö
vi
v so Z
'V' +- koL o ýo V]
000000 0
WWWWWW
LL) W ww
Li
to f! » O CýY 1 ýo U ö
Cý %0 10 An kA t- 00 r-
,, °f? NNNNN
vn v3 vi W tG r N DQ 00 N
V"ý Hfl ýp
OC 00
ßf1 V'i
c
oq G MO00+5
ýtNPOä
Üv -
tý N V' v1 M %0 N
C i4 oß ýY vS Oh
IM O fit
'O Z.ii N ý
iý
N iii
%0
Q
-
i
im
.in ° 0
Qý *41 1cN 00 a '--
F
159
The critical shearvalues for the SHHSS are close to the nominal shearrates. This would
suggestthat the shearrates in this region of the mixer are sufficiently large to determine
dma, This statementwould only hold if the exposuretimes of the drops to the critical
t.
shearratesare long enough to cause break-up.
The nominal residence times (Sr)at 0.41/s in the shear gap region are approximately 4 ms
for the ES and 1ms for the SHHSS. Comparing these values to the estimated critical
shear times (tb) shows that for ES the values for tb are twice ti in Couette flows and 6
times ti in Extensional. The equivalent values in the SHHSS were found to be 7 times
and 20 times greater than ti in Couette and Extensional flow (respectively).
Thus, it would appearthat the nominal residencetimes are not large enough for d., to be
.
determinedby shearforces in the sheargap.
Performing a similar analysis on values of d32shows values of Gg in the ES are
approximately 100 times and 40 times y' in Couette and Extensional flows
(respectively). Interestingly, the corresponding values for tb are 5 times and 2 times less
than t. In the SHHSS GB is 25 times greater than y' for Couette and 10 times in
Extensional flows. The corresponding values for tb in the Couette flow are equivalent to
T. However, for Extensional flows tb is less than r (;t;0.5 times).
Hence, with the exceptionof Extensional flows in the SHHSS,the nominal residence
times are sufficient for drops in the order of and lessthan d32to break due to viscous
forces. The effects of the viscous forces in the sheargap would increaseas the droplet
sizes below
decreased d32. tb will decreaseincreasingthe chancesfor a drop to break in
Couetteand Extensional flows. However, the fact that the droplet size distributions did
not show significant differences betweenthe two screenswould indicate that it is
turbulent energy dissipation that is in

more prominent this region of the mixer.
The areasof highest energydissipation would be expectedin this region where fluid
acceleratingat high speedsundergoes a rapid decelerationas it impacts againstthe stator
160
wall (or edges of the holes). The turbulent length scales will be very small in this region.
This would explain why larger quantities of smaller droplets are observed at the higher
rotor speeds. This effect will be dependant on tip speed and also the open area of'the
screens.
The large differences in shear rates and Res of the two screens did not result in significant
difference in the droplet size distribution. Therefore, in fully turbulent conditions the
effects of the gap size can be assumed to be negligible. However, its influence will
increase as the continuous phase viscosity increases (µßj/µc becomes smaller).
This is evident when the droplet size distributions from the two screens are compared at
different values of ý (Figure 6.24).
2s
E/S 10%
0 K SHHSS 10%!,
f E/S 20%
20
0 SHHSS 20%
E/S 30%
O
o SHHSS 30%
IS
v
E
ö Qa
i
lo j
C ""
"cQ
o"
S
QA
SEE ýýý
ýýý 3() 40 ;n 6) 70
Droplet Diameter (µm)
Figure 6.24 Comparison between E/Screen and SHHSS: 8000rpm & 0.41/s (69.85mm
rotor)
161
The droplet size is seen to increase with an increase in 4 and this was attributed to
turbulence damping. As ý increases the system can no longer be defined by the physical
parameters of the continuous phase and it is logical to believe that the viscosity of'the
resultant dispersion will increase.
Figure 6.24 shows that the droplet size distributions from the two screens are similar up
to ý =20%. However, there is significant change at 30% with the SH1HSS generating
smaller droplets than the ES.
It is worth noting that the leakage flows that were observed by Sparks (1996) were also
negligible. This can be explained by the fact that toothed mixer had an open top gap (i e
no "Lip" covering the gap). However, the Silverson rotor does have a "Lip" (Figure
6.25) which will reduce the amount of leakage that occurs.
This seems a reasonable deduction as there was no significant difference in the
performance of the screen despite the gap in the ES being 4 times that in the Sill ISS.
Lip 10
:M99
Stator
Rotor
Figure 6.25 Schematic showing the "Lip" at the top of the Silverson rotor
162
6.5.4. Stator
The Reynolds number in the holes in the stator is defined by Equation (2.40) and the
values,along with the nominal ti and y' are summarisedin Table 6.8
Table 6.8 Reethorin the stator holes and corresponding ti and y'
Dhyd Q Vim T ,Y9
m m3/s m/s Re (s) s'1

ES 0.0014 0.0004 0.42 584 0.0034 298
0.0008 0.83 1168 0.0017 596
SQIHISS 0.0024 0.0004 0.42 1010 0.0048 175
0.0008 0.84 2020 0.0024 351
The flow through the stator holes will be laminar as the values for Re in the stator holes
are very low. This also indicatesthat viscous stresses
will play a more significant role in
breaking up drops.
is determinedby Q/A) and doubling Q also

is a function of the flowrates (as vJets
Reßtator
increases,r and y' by a factor of 2. However, the flowrate was found to have a little or
no effect on the droplet size distribution. Also, the performanceof the two screenswas
similar (with regards to the droplets sizes generated by them both). This would suggest
is
that there very little or no drop break up in this region of the mixer (at least due to
viscous stresses).
Comparingthe values of ti and y' to the values of critical shearratesand burst times for
drop diametersequivalent to dm. and d32(Table A. 2). It can be determinedthat y' is
significantly smaller than values for GB, (up to 104times).
163
The equivalentvalues of tb for dm were again comparedto r and were found to be
larger. Implying that drops of size dmax
and above did not have enoughtime to break in
holesof both the ES and the SHHSS.
However, values for tb calculatedfor d32were found to be significantly smaller than 'r (up
to 6x in the SHHSS) for Couette flows and up to 2x smaller in Extensional flows. Again,
this would imply that drops in the order of and below d32could break due to viscous
forces in the flow through the stator holes.
It is possible that there is further break-up as the drops passthrough the stator holes. In
low viscosity systemsthis effect would be small and not influence the droplet size
distribution significantly (as the performanceof the two screenswas similar).
The number of holes would also affect the turbulent energy dissipation as Sparks (1996)
observed the fluid discharging from the rotor and turning abruptly by the upstream face
of the stator teeth and emerging radially. A fast stream followed the leading edge of the
stator and a region of recirculation accompanied this (Figure 2.8).
It was postulated that the tangential kinetic energy given to the fluid by the rotor was
transformed into pumping power, skin friction losses and turbulent losses. It was
suggested that the point where the abrupt turn in the flow takes place is where the losses
are greatest.
Therefore, it is reasonableto believe that the on the leading edge of the rotor (where the
pressure is positive) there is likely to be a maximum in the losses (as described by

Sparks, 1996). This is also where the streamlineflow through the stator will be greatest.
As the rotor passesthe holes, there will be a certain degreeof back-mixing, as an area of
low pressurewill exist, which may draw liquid back into the slot/hole. This would cause
the recirculating flow in the holes, until the positive pressure is applied again by the
following rotor.
164
6.5.5. Volute
The flow in the volute will be determinedby the velocity of the radial jets from the stator
holes. Sparks (1996) described the flow in this region as being disordered. This region
makesup the largest volume of the mixer and consequentlywill have the largest nominal
residencetimes.
As Dvorteand the open areaare not significantly different betweenthe two screensthe
for Revai,, (Table 6.9)
values teare similar
Table 6.9 showsthat the Reoit,, are high and the nominal shearrates in this region are
very low. Therefore, it can be assumedthat the drop break-up due to viscous stressesis
in
unlikely this region.
Table 6.9 Reoit. and correspondingvalues for r and y'
Stator Dvoi,te Q Vj T Y'

(m) (m3/s) (m/s) RevoG, (s) (s-1)
re
ES & 0.0004 0.42 11136 0.16 16
0.0267
SHHSS 0.0008 0.83 22272 0.08 31
It is reasonableto assumethat the turbulent energydissipation rate is significantly
smaller in the volute than in the rotor region of the mixer. However, due to the longer
residencetimes in this part of the mixer it is possiblethat the any larger drops that may
have by-passedthe regions of high shearratesand high energydissipation in the
distribution may undergo further break-up due to inertial stresses.
165
6.5.6. Summary
The in-line rotor-stator mixer was divided into 5 sectionsbasedon flow patternsobserved
in the visualisation work of Sparks(1996). These sectionswere the inlet, rotor chamber,
sheargap, stator and the volute. By defining the Re in eachof theseregions which made
it possibleto investigatewhich stressmechanismwould be responsiblefor drop break-up.
The Re in the pipe were calculated to be 20,000 (at 0.41/s) and 40,000 (at 0.81/s),
indicating that the flow was turbulent. It was established that the nominal residence time
in the pipe was long enough to sustain drop break-up due to both viscous and inertial
stresses. However, the high Re implied that inertial stresses would be more dominant.
In the rotor chamberit was consideredmore appropriateto define Re using the rotor tip
speedand the blade length as the characteristicvelocities and length. This was because
the droplet size was seen to decrease with rotor speed. The resultant values for Re were
calculated to be in the order of 4x105, which points to turbulent conditions.
The effective shearrates (Grace, 1982) in this region were found to be considerably
higher than the nominal shearrates. However, the critical burst times, tB, were
significantly smaller than -r suggestingthat drop break-up due to viscous stresseswas

possible.
The values for dmaxlevelled off at 150µm after approximately 7000 rpm and as Re
increasedwith N. It was assumedthat the maximum eddy length scaleswould be in the
order of dm,.. It was concluded that inertial in

stresses this region were responsiblefor
drops with diametersclose to dmax.
The values for Re in the shear gap ranged from 5000 to 20,000 with the ES and 1400 to
5000 with the SHHSS. These relatively low values of Re implied that the viscous
stresseswould be more dominant in the break-up of drops. The corresponding nominal

23000 to 90,000 s1 90,000 to 300,000s 1 for the ES
shear rates were s'1 and and the
SHHSS respectively.
166
However, despitethesedifferencesin Re and y' there was no discernabledifference in the
droplet size distribution from the two screens. The nominal residencetimes in this region
were not large enoughto sustainthe break-up of drops with sizes similar and above dmax.
However, drops similar and below d32would be broken in the short time in the sheargap.
It could then be deductedthat drops with diametersclose to and below d32most likely to
be disruptedby viscous sub-rangeeddies.
In the stator the flow was consideredto be laminar and break-upby viscous stresses
would be more likely. However, the fact that the two screensshowedno significant
differencesin performancewould lead to inferencethat there is very little break-up in this
region. It is also possible that some drops that have undergonedeformation in the shear
gap will break in

up this region.
The flows in the volute would be responsiblefor break-up of larger drops that may have
by-passedthe regions of high energyand shearrates.
In summary,under turbulent conditions, it is reasonableto assumethat drops are likely to

break due to:
a) inertial stresses in the inlet to the mixer.
b) inertial and viscous stressesin the rotor region.
c) inertial in
sub-rangeeddies the sheargap.
d) viscous stressesin the stator holes.
e) inertial stressesin the volute.
Theseregions can be put into the following order of residencetimes; ShearGap < Stator
< Inlet < Rotor < Volute. The very short residencetimes in the sheargap and the fact the
high localised energy dissipation are very small, would indicate that a large
regions of
of the drops are by-passingthis region. The relatively long residencetimes
percentage
the stresseswould point to the fact that droplet size distribution is
and the magnitudeof
determined by the viscous and inertial stresses in the rotor region.
predominantly
167
CHAPTER 7. Conclusions
The main findings of this work are as follows:
1. The kerosene-watersystemwas confirmed to be a non-coalescingsystem.

Consequently,drop break-up at high dispersedphaseconcentrationscould be
investigatedwithout having to deal with the complexities of coalescencephenomena.
This also led to the fact that samplescould be taken and droplet size measured
without coalescence occurring, thus giving a true reflection of the droplet size
distribution from the in-line rotor-stator mixer.
2. The comparisonbetweenthe Video Probe (on-line droplets sizesmeasurement)and

the Malvern Mastersizer(off-line measurement)showedthat the Mastersizer could be
used to measurethe droplets generatedfrom the rotor-stator mixer. The Mastersizer
was found to be the better of the two techniquesfor measuringdroplets below 20µm.
3. A continuous in-line sampling techniquewas developedthat enabledthe rapid

measurement of large sample sizes.
4. The liquid-liquid contactor rig was designedand built specifically for this work. It
allowed the measurementof droplet sizesgeneratedfrom a single-passthrough the in-
line rotor-stator mixer. It was then possibleto investigatethe effects of varying; rotor
speed,geometry, throughput and dispersedphaseconcentrationon the droplet sizes.
5. A single-passthrough the in-line rotor-statormixer resultedin broad droplet size

distributions with sizes ranging from below 10 µm to approximately 100 µm. This
highlighted the inhomogeneousnature of the rotor-stator mixer.
6. The performancetestsshowedthat the droplets sizes:

decreasedwith N, until a speedwas reached,(approximately 7000 rpm) above
-
which no significant size reduction was observed.
168
- increasedlinearly with dispersedphaseconcentration.
were not influenced significantly by flowrate had little or no effect on the droplet
-
size distribution.
were similar from the Emulsor screenand the SquareHole High ShearScreen.
-
- increasedwith an increasein open area. The tests performed with the 50.8 mm
rotor showedthat the GeneralPurposeDisintegrating Head resulted in the largest

droplet size, then the Slotted Head, with the Emulsor Screenand the SquareHole
High ShearScreenproducing the smallestdroplet.
were not affected by the sheargap. The sheargap in the Emulsor Screenwas
twice that in the SquareHole High ShearScreen.
7. d32was linearly proportional to dmax

which indicated that the assumptionthat drop
break-upwas due to a single mechanismis valid. This then enabledd32to be usedin
mechanisticcorrelations developedfor dmax.
S. The Chen and Middleman correlation was found to be the most appropriatemodel for
describing correlating d32.This suggeststhat viscous sub-rangeeddiesare the
dominant break-up mechanism. The drop size data could be correlatedfor « up to
20%, by:
d
32 = 2x109(1+ 20(D)(WeRe4)-'"
D
9. The dispersedphaseconcentrationhas a dampeningeffect on turbulent stresses.
10. The power supplied to the fluid by the in-line rotor-stator mixer was measuredusing a
thermometric technique. The data collected from this rig enabledthe estimation of
the energydissipation rate and the Kolmogorov length scale.
11. The Shaft power increaseswith rotational speedand flowrate. The shaft power can
be related to N and Q by the following:
169
PSS = 25
0.01N2.8Qo.
12. P,haftwas similar for both the Emulsor screenand the SquareHole High ShearScreen.
-The effects of Sheargap were not significant on Psh

.
-Open areaas an effect on Pý,
t.
13. The power supplied to the fluid, PFLUID,

is proportional to N3. The Power number for
the Emulsor Screenand the SquareHole High ShearScreenis 0.21.
14. The mixer is operating in turbulent conditions.
15. Energy dissipation within the rotor-stator mixer is highly inhomogeneousand values
of e were estimated to range from 40 W/kg to 800 W/kg.
16. d32was proportional to e-0.2,which best matchedthe Shinnarmodel for drop break-
up causedby turbulent stressesin the viscous sub-range.
17. The Kolmogorov length scales, Xk, ranged from 12µm down to 6µm.
18. Under turbulent conditions, drops are likely to breakdue to:
a) inertial in
stresses the inlet to the mixer.
b) inertial and viscous stressesin the rotor region.
c) inertial sub-rangeeddiesin the sheargap.

d) viscous stressesin the stator holes.
e) inertial in
stresses the volute.
19. The droplet size distribution from a single passthrough an in-line rotor-stator mixer is
predominantly determinedby the viscousand inertial stressesin the rotor region.
170
CHAPTER S. Future Work
8.1. Liquid-Liquid Dispersions
There is still a considerableamount of work required in the understandingof the
mechanismsof droplet break-up in in-line rotor-stator mixers. This work concentratedon

a turbulent system,at a single viscosity and interfacial tension. The majority of tests
were performedat a single scale.However, it is apparentthat the droplet size distribution
is not determinedsolely by inertial stresses.Different regions of the mixer exhibit
different flow tendencies.
It would be useful in the design of such units to be able to assessthe influence of each
of
these regions on the final droplet size distributions at varying flow conditions. The
definition of Re in the different regions of the mixer is a good indication of
whether
inertial or viscous stressesare responsible for drop break-up. The effects of
varying the
viscosity ratio (and µa) in addition to studying a wider range of throughputs (thus varying
residence times in each region) will help to identify when viscous stresses are more
dominant than inertial stresses and vice versa. This will help identify which
parameters
are useful in scale-up (rotor diameter, shear gap, open area, stator thickness etc) and
under what conditions. In this work the shear gap was found to have no effect on the
droplet size distribution, however, at higher viscosities it is possible that the
shear
stresses in the gap are mainly responsible for drop break up.
In considering turbulent flow regimesthe determinationof dm in this work was made

difficult due the amount of bypassingthat occurredin the mixing-head. This could be
overcomeby inserting the mixer into a recycle loop and increasingthe number of passes
through the mixer. However, one draw back to this type of set-up is that temperature
effects begin to come into play and will alter the physical propertiesof the system.One
possiblemethod of overcoming thesetemperatureeffects are to study a single passmulti-
stageunit (one with a number of concentrically arrangedrotor-stator teeth). Alternatively,
the mixture could be passedcompletely through the mixer from one vesselto anotherand
171
then back through the mixer into feed tank. This secondmethod would be dependanton
the stability of the resultant dispersionsand the both the vesselsshould be equipped with
agitatorsto promote homogeneity.
The resultsat the higher dispersedphaseconcentrationssuggestedthat feed strategy
might improve the for

performance a single-passmachine. This could be investigatedby
varying the location of the point where the dispersedphaseis addedto the continuous
phase. This can be further investigatedby introducing a known droplet size distribution
to the inlet of the mixer. This could be done by way of adding a static mixer to the in-let
and using the Video Probeto measuredroplet size distributions generatedby it. Smaller
droplet sizesare more likely to break in the sheargap region (smaller values of tb), which
is where the highest shearratesare likely to be.
The literature review highlighted that there is very little power draw data available for
rotor-stator mixers. It has been discussed that very often rotor-stator mixers are operating
in the laminar regime where viscous shear is more prominent. This would suggest that
the identification of the operating regime in a rotor-stator mixer will help in the
understanding of the break-up mechanisms involved.
A method of determining the flow regime is to determine the dependencyof the power
draw of the unit on rotational speed. If the power is found to be proportional to rotational
speed raised to the power of 3, then turbulent conditions will apply. If the power is
to
related rotational speedraised to the power of 2, then the flow will be laminar.
Power measurementswould also be necessaryin order to determinethe Kolmogorov

length scaleand energydissipation ratesfor the interpretationof droplet size data. It
be to
would useful monitor power draw whilst mixing the immiscible liquids, as the
effects of dispersedphaseand viscosity and interfacial tension on power draw could be
determinedin thesemixers. A method of doing this would be to measurethe shaft
172
torque. A possible method of increasingthe accuracyof measurementat high shaft
speedswould be to use a gearing device. This requires an additional shaft with a torque
transducer,connectedto the mixer shaft using a toothed belt and pulley. A larger pulley
on the secondaryshaft would reducethe speedallowing accuratemeasurementof torque.
Toothed belts typically have efficiencies of 95% and above.
173
References
Anderson, H. H., (1980), in "Centrifugal Pumps ". 3rd edition. Trade and Technical
press. ISBN 85461 076 6
Avertukh, Yu. N., Nikoforov, N.M., Kostin, N.M. & Korshakov, A.V. (1988),
"Computation of Dispersions for Emulsions Formed in a Rotor-Stator Unit". Journal of
Applied Chemistry of the USSR, vol 61, (2). 396-397.
Becher,P (1966), in "Emulsions: Theory and Practice" 2°dedition, Reinhold Publishing

Corporation (New York, USA).
Becher, P., (1964), "A particle-size analyzer for emulsion photomicrographs", Journal of
Colloid Science, 19,468-472.
Bohren, G. F. and Huffman, D. R. (1983), "Absorption and scatteringof light by small

particles" Wiley-Interscience.
Born, M. and Wolf, E. (1980), "Principle of optics, electromagnetic theory of propagation

interference and diffraction of light" Pergamon Press.
Bourne, J. R. and Garcias-Rosas,J. (1985), "Rotor-stator mixers for rapid micromixing".

Chem. Eng. Res. Des. Vol 64.
Brown, D. E. and Pitt, K., (1971), "Drop Break-up in a stirred liquid-liquid contactor",
Proc. of Chemica `70, Melbourne and Sidney,August 1970,publ. Butterworth and
IChemE., pp 83-96.
Brown, D. E. and Pitt, K., (1972), "Drop size distribution of stirred non-coalescing
liquid-liquid system", Chemical EngineeringScience,27,577-583.
174
Brown, D. E. and Pitt, K., (1974), "Effect of impeller geometry on drop break-up in a
stirred liquid-liquid contactor", Chemical Engineering Science, 29,345-348.
Brown, J., (1991), "Application of the thermometric method of pump efficiency
measurement to bore hole pumps", BHR Group Report 47020, BHRG Ltd. Cranfield.
Calabrese, R. V. (1992), "Analysis of dilute liquid-liquid dispersions: Physical
considerations,discriminating amongmechanismsand modelling approaches". FMP

Report 054, BHRG Limited, Cranfield.
Calabrese,R V., Francis, M. K., Mishra, V. P. and Phongikaroon,S. (2000)

"Measurementand analysis of drop size in batch rotor-stator mixer". Proceedings:10th
European conference on mixing, 149-156.
Chen,H. T. and Middleman, S. (1967), "Drop size distributions in agitated liquid-liquid
systems",AIChE J., 13,989-995.
Chen, S.J. and Libby, D. R. (1978), "Gas Liquid and Liquid-Liquid Dispersions in a
Kenics Mixer", 71st Annual AIChE Meeting.
Cheung (1998), "Hiline Liquid-Liquid Literature Review'. Hiline report HLLO1 BHRG
Limited Cranfield.
Clements,D. J., (1996), "Principles of ultrasonic droplet size determinationin
emulsions", Langmuir, 12,3454-3461.
Cohen,D (1998), "How to SelectRotor-StatorMixers" Chemical Engineering, August

1998,76-79.
Davies, J. T., (1985), "Drop sizes of emulsions related to turbulent energy dissipation
rates ", Chemical Engineering Science, Vol. 40,5,839-842
175
Dietsche, W (1998), "Mix or Match: Choose the Best Mixers Every time" Chemical
Engineering,August 1998,70-75.
Doulah, M. S. (1975), "An Effect of Hold-Up on Drop Sizes in Liquid-Liquid

Dispersions" Industrial Engineering Chemistry Fundamentals, Vol. 14 (2), 137.
El-Hamouz, A. M., Stewart, A. C. and Davis, G.A. (1988), "Kerosene/Water Dispersions
produced by a Lightnin In-Line Static Mixer", IChemE Symposium Series 136,457-464.
Francis,M. K. (1999) "The developmentof a novel probe for the in-situ meaurementof
particle size distributions, and application to the measurementof drop-size in rotor-stator

mixers" Ph.D Thesis,University of Maryland.
Grace, H. P. (1982) "Dispersion phenomena in high viscosity immiscible fluid systems
and application of static mixers as dispersion devices in such systems", Chem. Eng.
Communications, Vol. 14, pp 225-277.
GreercoMixers by Chemineer,Inc. (CompanyBrochure, bulletin 117,1999).
Hearn, S. (1992), "Residence Time Distribution and Pressure Gain Measurements in a
Weir Group Centrifugal Pump with an OpenImpeller". HILINE report HLP 02. BHRG
Ltd, Cranfield.
Hinze, J.O. (1955), "Fundamentalsof the Hydrodynamic Mechanism of Splitting in

Dispersion Process",AlChe Journal,Vol. 1 (3), 289-295.
IS013320-1: 1999(E),"Particle Size Analysis - Laser diffraction methods- Part 1:

GeneralPrinciples"
176
Karam, H. J. & Bellinger, J. C. (1968), "Deformation and Break-up of liquid droplets in a
simple shear field", Ind. Eng. Chem. Fundamentals, Vol. 7 (4), pp 576-581.
Karassik, I. J. Krutzsch, W. C. Fraser, W. H. & Messina, J.P. eds. (1986), in `Pump
Handbook", 2nd edition. McGraw-Hill, New York
Karbstein,H. & Schubert,H., (1995), "Developmentsin the ContinuousMechanical

Production of Oil-in-Water Macro-Emulsions". Chem. Eng. & Processing, 34, (1995)
205-211.
Keey, R. B. (1967), "Interpreting mixing with isotropic turbulencetheory". British

ChemicalEngineering, Vol. 12, No. 7,341-344.
Kolmogorov, A.N. (1949), The Break-up of Droplets in a Turbulent Stream", Dokl

Akad Nauk, 66,825-828.
Korshakov, AN, Nikiforov, AO., & Terentev, O.A. (1990), "Optimisation of the
Processof Emulsification in a Rotor-StatorApparatus". Izv Vyssh Uchebn zaved,Lesn
ZH, N. 5,1990,76-79.
Kroezen,A. B.J. & Wassink, J. Groot (1987), "Bubble Size Distribution and Energy
Dissipation in Foam Mixers". Journal of Society of Dyers Colour, Vol 103, Nov. 1987,
386-394.
Levine, L. (1994) "High shearmixers", CerealFoods World, 39 (11), 874-876.
Middleman, S. (1974), "Drop Size Distributions Producedby Turbulent Pipe Flow of

Immiscible Fluids Through a Static Mixer", Ind. Eng. Chem.ProcessDes Dev, 3,78-83.
177
Nikiforov, A. O., Korshakov, A. V., Averbukh, Yu. I. and Sergeev, E. Yu. (1990),
"Dispersing Pumps for Producing Polyacrylamide Solution". Khimicheskoe I Neftyanoe
Mashinostroenie, No. 11, pp. 14-15 November 1990.
Ottewill, R.H. (1984), in "Surfactants".ed Tadros, Th. F., Academic PressInc. (London)
Ltd, pp. 1-18.
Pacek,A. W. and Nienow, A. W., (1995), "Measurementof drop size distribution in

concentrated liquid-liquid dispersions: Video and capillary technique", The 1995 IChemE
Research Event, 1`t European conference, University of Edinburgh, 50'-6`1`January, 725-

727.
Pacek,A. W., Moore, I. P. T. and Nienow, A. W., Calabrese,R V. (1994) "Video

technique for measuring dynamics of liquid-liquid dispersion during phase inversion",
AIChE Journal, December, Vol. 40, No. 12., 1940-1948.
Pedrocchi, L. & Widmer, F. (1988), "Formation of Emulsions in a Turbulent Shearfield".

6th European Conference on Mixing, Pavia, Italy, 24-26 May, 205-213.
Pedrocchi, L. & Widmer, F. (1989), "Emulsionsherstellung im Turbulenten Scherfeld. "
Ing. -T Chem. Ech. 61, Heft 1.
Rushton,J.H. Costich, E.W. & Everett, H.J., (1950), "Power Characteristicsof Mixing
Impellers". Chemical EngineeringProgress.Vol 46, No 8.
Saadevandi,B. A. and Zakin, J., L. (1996), "A Study of Silicone Oil-in-Water

Emulsions". Chem. Eng. Comm., 1996.Vol. 156,227-246.
Schubert,H. (1997), "Advances in the mechanicalproduction of food emulsions".

ProceedingsSeventhInternational Congresson Engineeringand Food. Part 1, S. 82-87.
178
Shinnar, R. (1961), "On the behaviour of liquid dispersions in mixing vessels", J. Fluid
Mech., Vol. 10,259-275.
Silverson Applications Report April 1999 (Issue4).
Silverson High Shear In-Line Mixers (Company Brochure), 1999.
Sparks, T. G. (1996) "Fluid Mixing in Rotor-Stator Mixers" Ph.D Thesis, Cranfield
University.
Stepanoff,A.J. (1957). "Centrifugal and axial flow pumps, theory design and
application". 2nd edition. John Wiley and Sons.
Streiff, F., (1977), "In-Line Dispersion and Mass Transfer using Static Mixing
Equipment", Sulzer Technical Review 3,108-113.
Streiff, F. A., Mathys, P. And Fischser, T. U., (1997), "New Fundmentals for Liquid-
Liquid Dispersion using Static Mixers", Recents Progres en Genie des Procedes, Vol 1
(51), 307-314.
Rawle, A., "Basic principles of particle size analysis" Technical Paper, Malvern
Instruments Ltd.
Tavlarides,L. L. and Stamatoudis,M. (1981),"The analysis of inter-phasereactionsand
masstransfer in liquid-liquid dispersions",Advancesin Chemical Engineering, Vol. 11,

Academic PressInc. (ISBN 0-12-008511-9).
Taylor, G. I. (1934), "The formation of emulsionsin definable fields of flow",

Proceedingsof the Royal Society of London, A146, pp501.
179
Wachtel,R. E. and La Mer, V. K., (1962), "The preperationand size distribution of some
monodisperseemulsions", Journal of Colloid Science,17,531-564.
Walstra,Pieter (1983), "Formation of Emulsions" Encyclopaediaof Emulsion

Technology,Vol. 1, (Basic Theory), ed. Becher,P., 57-127.
Walstra,Pietra (1993), "Principles of emulsion formation", Chemical Engineering

Science,48 (2), 333-349
180
Appendix 1 Experimental rig operating procedures
A schematic of the experimental rig is shown in Figure A. 1. The liquids are pumped
from the two feed tanks (1 m3) and the organic liquid was introduced to the aqueous line
via a manifold located one pipe diameter up-stream of the rotor-stator mixer. The fluid
was passed through the mixer to the settling tanks. The recycle loops were present on the
feed lines for use when commissioning the flow transducers or testing the pumps.
Standard Operating Procedure

The operation can be divided into the following categories:
(a) Filling the feed tanks
(b) Start-up
(c) Shut-down
(d) Waste
(e) Cleaning proceduresand checks
Protective clothing: Safety glasses,safety boots, nitrile gloves.
(a) Filling the feed tanks

1. Check that both the settling tanks are empty.
2. Move the kerosenedrum into the bund. The keroseneis delivered in 200 litre drums
(the stainlesssteel tanks can hold up to 1000 litres each).
3. CloseV1 andV5.
4. Use the barrel pump (P4) to transportthe organic phasefrom the drum to the feed
tank. When filling the tanks, the sight glassesmust be monitored to prevent
overflowing. Note: Before using the barrel pump, flush through with warm water to
ensurethat it is clean.
5. The water is supplied from the mains line via a de-ionisedcylinder. Ensure the flow
is
through the cylinder not too high (to cause damage to the ion exchangeresin). The
flow can be controlled by using a valve (V17), located on the mains water supply.
181
bA
O
U
CC3
O
U
C7"
'C7
cat
cý
M
C3
'Co
3
0
w
L
b.D
w
ww
0
Figure A. 1 liquid-liquid contactorrig

Flow diagram of the r8otor-stator
Once the feed tanks are full, the valves have to be positioned correctly to allow flow
through the rig.
(b) Start-up
Aqueous line:
6. Close V3 and V4.
7. Position V2 (three-wayvalve) to allow flow to the mixer (as opposedto the recycle
line).
8. Open Vl.
Organic line:
High flowrates
9. Close V8, V9, V10 and V12
10. Open V6 and VI4
11. Position V7 (three-wayvalve) to allow flow to the mixer (as opposedto the recycle
line).
Low flowrates
12. Close V6, V10 and V12.
13. Open V9, V11 and V14
14. Position V7 to prevent flow back towards the pump (P2) and the recycle line.
15. Open V5.

16. Open V15 and V16.
17. Close V19
18. Open V17 and V18 (thesevalves should remain open when the rig is running, to
prevent the settling tanks from overflowing).
Operation of the Flow controller and Pumps:

19. Ensure both feed tanks contain liquid. "
20. Once all the valves are in the correct positions, check that the pump switches on the
flow controller are all in the off position.
183
21. Turn the power on from the isolator switch. Check that the red button on the
frequencycontroller is switched to Run and not Stop. Push the black button on the
electrical box to turn on the inverter.

22. Select the correct organic pump depending on whether a high or low flowrate is
required, (P2: 0.14-1.41/sand P3: 0.03-0.161/s). The switch is on the top of the
is
control panel and marked with the numbers 1&2. The number 1 refers to the
large organic pump (P2) and 2 to the smaller one (P3). In addition, the switch on the
power board has to be turned to the correct setting, this can be found to the right of
the main electrical box and hasthree settings, 1,2 and off.,
23. Before starting the pumps, checkthat the correct scalefactors have been inputted.
Flick the program switch to on. Press the function button on the display panel, and
continue to press until the scale factor option is displayed. If the is

value incorrect,
change by using the number buttons along the bottom of the display panel. Repeat
this process for the aqueous flow controller (P1).
The scale factors should be as following:
Organic: (I)P2: 1.27380
(2) P3: 0.16918
Aqueous: P1: 1.29391
24. Flick the program switch to off.

25. Input the desiredflowrate. Find the flowrate option by pressingthe function button
and use the number buttons on the bottom of the display panel to enter the correct
flowrate. Note: The numberon the display panel should be to three decimal places.
This can be changedusing the button labelled DP.
26. Pressthe RUN SET button to acceptthe flowrate.
27. Repeatsteps21-26 for the aqueousflow controller (P1)
28. Start P1 using the start switch on top of the flow controller.
29. When the flowrate has stabilised,the relevant organic pump (P2 or P3) can be
started.The flowrate be
can adjustedwhilst the pumps are running by following the
sameprocedure. Note: If the flow controller is showing 0 flowrate, switch off the
pumps and check for closed valves that should be open.
184
Mixer Start-up
30. Read the Silverson installation, operating and maintenance manual for the rotor-
stator mixer.
31. Open Valves VI 5 and V16.
32. Turn the isolator switch to on (at the bottom of the rotor-stator control panel).
33. Turn the power on (the greenbutton at the top of the control panel).
34. Check the rotor speed,on the VACON display panel (the speeddisplay can be found
by pressingeither the left or the right arrow).
Note: Power, Torque and other information can be obtained from the display panel
by pressingthe arrow buttons until the required display is found.
35. Set the speedusing the SpeedAdjust knob.
36. Turn on the feed pumps (see 19-29)
37. Wait a2 minutes before switching on the rotor-stator mixer using the green start
button on the rotor-stator control panel.
(c) Shut Down

38. Stop the rotor-stator mixer.
39. Switch off the organic pump and then the aqueous pump (see 46-52 for pump shut
down procedures).
40. Turn the rotor-stator power off.

41. Turn the isolator switch to off.
42. Close valves V15 and V16.
43. Drain the liquid from the volute (mixing head)from the drain valve located below it.
44. If the mixer is to be changed,turn all power off from the isolator switch on the pump
control box.
45. When changing the mixer head,refer to the Silverson installation, operating and
maintenancemanual.
Pump Shut-down
46. Stop the organic pump (using the start switcheson the flow controller)
185
47. Wait a few minutes before stopping the aqueous pump. This allows the water to
flush most of the organic liquid out of the mixing sections.
48. Turn the switch on the power board to off.

49. Turn the isolator switch off.
50. CloseVI on the aqueousline.
4
51. Closevalves V5, V6 and V9 on the organic line.
52. Close V17, V18 and V19.
Emergencyshut-down
There are three emergency stop buttons (large red buttons) on the rig. The most
important one is located on the back of the blue power board (facing towards the bund).
When pressed, it will cut off all the power to the rig, switching off both the pumps and
the rotor-stator mixer. The one that is located on the electrical box on the blue power
board will only switch off the pumps. The one on the Rotor-stator control box will only
switch off power to the rotor-statormixer.
(d) Waste
To allow the dispersionin the settling tanks to fully separateit should be left for 24
hours. The organic (less dense)phaserisesto the top of the tank creating an interface
betweenthe two phases.Pumping the aqueousphaseto waste requirestwo people, one
to monitor when the organic phaseis reachedand one to turn the pump off.
53. Close valve V18.
54. Close valve V21 (recycle line).
55. Open V19
56, Start P4 (which is a diaphragmpump and requirescompressedair to operate)
57. Stop P4 when the interface is reachedand there is only organic phaseleft in the tank.
58. Close V20.
59. Open V21.
60. Start P4.
61. Stop P4 when ST2 is empty.
62. Repeatfor STI.
186
(e) Cleaning procedures and checks
Before introducing new inventory, the whole rig should be flushed twice with de-ionised
water and the tanks should be scrubbed if necessary. If the tanks require scrubbing an
extra rinse using de-ionised is
water required before filling.
The seals(o-rings) on the tank pipefittings and the gasketson the pumps are made from
nitrile rubber. The tri-clamp seals(green) on the organic line and the mixing lines are
madefrom Viton and the sealson the aqueousline (blue) are EPDM. The sealsshould be
checkedfor wear and tear wheneverthe tanks are empty and the mixer section seals
should be checkedwhen the mixing elementsare changed.
187
Appendix 2 Names and Addresses of Equipment Suppliers
In-line Rotor-Stator Mixer
Silverson Machines Ltd.
Waterside, Chesham
Bucks, 11P51PQ,
UK
Alcosol D70 (Kerosene)

Alcohols Ltd
Charringtons House
The Causeway
Bishop's Stortford
Herts, CM23 2EW
Pumps
Cougar Industries,
19 Empire Centre,
Imperial Way,
Watford,
Herts, WD2 4YH
UK
Latex Beads
Bangs Laboratories Inc.
9025 Technology Drive,
Fishers, IN 4603 8-2886,
USA
Pressure Transducer
Druck Limited
Fir Tree Lane
Groby
188
Leicester, LE6 OFH
UK
Pressure Gauge
WIKA Instruments Limited
Station Approach
Coulsdon,
Surrey CR5 2UD

UK
Flowmeters
Endress+Hauser Ltd
Floats Road,
Manchester, M23 9NF
UK
Temperature Measurement
RS Components
Lammas Road,
Weldon Industrial Estate,
Corby,
Northamptonshire,NN17 9RS,
UK
189
Appendix 3 Theoretical background to diffraction measurement
It is apparent that researchinto liquid-liquid systems from rotor-stator mixers has been
limited to dilute systems by the constraints imposed by the drop size measurement
technique.
The measurement technique for use in this research must incorporate the following
important factors:
1. Measurement of drop sizes in the order of the Kolmogorov micro-scale and below.
2. The measurement of droplets in high dispersed phase concentrations.
3. The technique must show good repeatability.
4. Availability
The droplet sizes generatedby an in-line rotor-stator mixer are expectedto be too small
for measurementusing video techniques. Semi-in-line methods such as laser diffraction
(Malvern Mastersizer)and electrical conductivity (Coulter Counter) are limited to low t
However, despite it being an off-line technique and requires considerable dilution, the
Malvern Mastersizer (laser diffraction) was the only available technique that can
successfully measure the small droplet sizes that are expected from the rotor-stator mixer
(<10µm). This was the most important criterion for work with high-energy dissipation
rates and the validation of mechanistic models for d< 2k. As this technique required
considerable dilution at high c and was predominantly off-line, a sample had to be
extracted and diluted for a measurement to be possible.
The sampling procedure was validated against an in-line video capturing technique that
had beendevelopedin-houseat BHR group.
A degreeof caution must be exercisedwhen analysing particle size data,as different

techniquesmeasuredifferent mean diametersand calculateor derive othersfrom it. To
190
be able to correctly describethe drop size distribution (DSD) it is important to have an
understandingof the principles of the measurementtechnique.
8.2.1. Principles of Laser Diffraction
The main principle behind laser diffraction is that particles passing through a laser beam
scatter light at angles that are inversely proportional to their size. Instruments that use
this principle measure the intensity of light scattered. The subsequent intensity pattern is
then transformed by way of an appropriate optical model and mathematical procedure to
a volume distribution.
Early laser diffraction devices only considered small scattering angles and were
commonly described as Low Angle Laser Light Scattering (LALLS) or Fraunhofer

diffraction devices. However, more recently the application of Lorenz-Mie theory has
broadened the technique to include wider scattering angles.
(a) Fraunhofer Approximation

When a spherical particle is passed through a beam of light, the resultant forward
scattering pattern consists of light resulting from diffraction and scattering between the
particle and the light (reflection, refraction and absorption). If the particle is large
compared to the wavelength of light then diffracted light will be mainly responsible for
the resultant radial scatter pattern.
Fraunhoferdiffraction, therefore, relies on the following assumptions:

i. The particles are spherical
ii. All particles are larger than the wavelengthof light.
iii. Only scattering in the near-forward direction is considered(small angles of
diffraction).
The scatteringpattern of light by a single sphericalparticle using the Fraunhofer
approximation can be describedby:
191
Jas sin
I (E)) a4 (A. 0.1)
2k a2 n0
Where, I(®) is the intensity of the scatteredlight (as a function of angle, O);
lo is the intensity of the incident light;
k is the wavenumber = 2n/? ;
x is the wavelengthof the illuminating sourcein air;
a is the distancefrom scattererto detector;

a is a dimensionless size parameter = 7tx/X;
x is the diameter of the particle (µm);
Ji is the Bessel function of the first kind of order unity.
The main advantageof Equation A. 1 is that it is relatively simple and quick to calculate.
The Fraunhofer approximation does not incorporate the optical properties of the system
and its use is limited to particles of diameters at least 40 times the wavelength of light.
The Helium-Neon gas laser that the Mastersizer uses had a wavelength of 0.63µm.
Therefore, this approximation is a good one for particle sizes larger than 25µm (ISO
13320-1: 1999 (E)). However, the Mastersizer 2000 uses a blue laser (wavelength of
0.45µm) which enables it to measure sizes down to 18µm.
(b) Lorenz-Mie theory
When the particle size approachesthat of the wavelengthof light, the scatteringbecomes
a complex function with maxima and minima present.The Lorenz-Mie theory provides
to
equations account for the complete light scatteringpattern in a particulate systemby
consideringthe electromagneticfield, both inside and outside a single particle. Finite
points within this field are usedas sourcesand further the equationsusedfor predicting
how thesesourcesradiateto elementsin the far field (e.g. a detector).
192
Therefore, for any position in the far field, a summationfor a given polarisation can be
madefor all the radiation's contributions from all points within the sphere. Many
positions in the far field can be examinedand thus a more concisedescription of the
radiation pattern can be obtained.
The complex scattering pattern of a particle is a function of the scattering angle, the
difference in the refractive indices of the particle and the continuous phase, the
wavelength of the incident radiation and the diameter of the particle. Therefore, the
equations derived using the Lorenz-Mie theory are a more complex version of Equation
(A. 1) and thus are more difficult to solve (Bom & Wolf, 1980, Bohren & Huffman,
1983). The recent advent of powerful desktop computer technology enables the
analytical solution of these in

complex equations a reasonable time and allows many
instruments to use full Lorenz-Mie theory.
The main difference between Lorenz-Mie Theory and Fraunhofer is the former assumes
the volume of the particle as opposed to the latter, which predicts a projected area. It is
claimed that the use of Lorenz-Mie theory allows accurate results over a wide size range,
0.02-2000 µm, (Rawle, 2002).
In addition to its ability to predict smaller particle sizesthe other advantagesof Lorenz-
Me theory is the assumptions, with the exception of spherical particles, which applied to
the Fraunhofer Approximation, can be eliminated.
(c) Refractive Index

As mentioned earlier the scatteringpattern of a particular particle is a function of its
In
optical propertiesand size. order to apply Lorenz-Mie Theory to particle size
measurementthen requires knowledge of the optical properties of the system. These

include:
i. The complex refractive index of the particle, Np, which consistsof a real and
imaginary part. The imaginary part refersto the amountof light that the particle
absorbs(or the attenuationof light through the particle). Np can be expressedas:
193
Np
=np -ikp (A. o.2)
Where, np is the real part of the refractive index;

i is an indication for the imaginary part of the refractive index;
kp imaginary (absorption) part of the of the particles refractive index.
ii. The complex refractive index of the particle relative to that of the medium, m,
which can be expressedas:
m= Np /nm (A. 0.3)
Where, nmis the real part of the refractive index of the medium.
The real refractive indices of various compounds can be found readily in the literature or
be
can measured. However, the choice of the imaginary refractive index is a little more
difficult, as there is very little information available in literature. Malvern Instruments
recommenda rule of thumb for evaluating the correct imaginary refractive index a
summary of which is given in Table 4.3.
Table A. 1 A summary of estimated absorption values as recommended by Malvern
InstrumentsLtd.
Particle details Absorption

Spherical;Transparent,e.g. GlassBeads 0
Spherical;not completely transparent,e.g. Oil drops 0.001
Not completely spherical; Transparent,e.g. Air pockets 0.01
Particles that appeargrey under a microscope 0.1
Coloured and metal powders (appearblack under a microscope) 1.0
194
Appendix 4 Droplet Size Distributions 9
Latex Bead Tests
Malvern Vs Video Probe

Latex Beads (90.106 microns) Volume Fraction
2S
" Video Probe
f" "f f Malvern

20
A
A
15
äe f
10
As
AS
""
f"
S
A
f
A
f A
"" ff""A""
0
0 20 40 00 80 100 120 140 100
Diameter(microns)

Latex Beads (90.106 microns)
25
" Video Probe
" MaNem
20 ""
ti
al
"
15
"
Z "
10
"
ýa
YYYY' 'YYýýý. ý .. """" "ý". ý ..... .
0 20 40 60 80 100 120 140 160
Diameter(t ucrons)
Figure A. 2 A comparisonof particle size distributions; Volume % (above)and Number

% (below) generatedby the Mastersizerand the Video Probe for Latex beads98 gm.
195
Latex Beads (75-150 microns)
a
""" Video Probe
a "f Malvern
A"
A "
6
f
"
f
ae 5 "
E """"
>4
""
2
0* i
f""
"A
""
0
0 20 40 60 60 100 120 140 160 180
Diameter (microns)

Latex Beads (75-150 microns)
10
" Video Probe
a
f
AAA ff Malvern
Malvern
ff
B
A
A "
7 "
f
"
."
6 ý _"
f"
Z
4 "
AA.
""
3 f"
A
f"" .. A
,"
2
A"A
f f"
" f ff
0 . uf. uuf"f"ý. uf. f. f"uuuf. bbý. """ ._
0 20 40 60 80 100 120 140 160
Diameter(microns)
Figure A. 3 Comparisonof particle size distributions; Volume % (above)& Number %

(below) generatedby the Mastersizerand the Video Probefor Latex beads113 µm.
196
Performance Tests: Droplet Size Distributions (Volume `%º)
69.85 mm Rotor: Emulsor Screen:
25
3000 rpm
20 " 5000 rpm

A 6000 rpm
7000 rpm
8000 rpm
9000 rpm
15 ----- ."
10000 rpm
- 11000 rpm
Z)
11 11610 rpm
10 .-+
5
4
Y- f
..
iI II II i"" f+
Oa 014a ! "- 4VYV "t d! i/ YiYY
p a" w-r-r fr-fý. ý r ro a3bO, -
01 1 10 100 1000
Screen at Q=0.4 I/s and (= 10%, for varying N (rpm).
25
f 3000 rpm
0 5000 rpm
20
e 6000 rpm
x 7000 rpm
x 8000 rpm
15 f
"9000 rpm
E ö
+ 10000 rpm
n -11000 rpm
0 x"
> 10 Xe
QQn
00 f±
f +
0
01 1 10 100 1000
Screen at Q=0.4 I/s and cD= 20%, for varying N (rpm).
197
25
x
. 3000 rpm
j 5000 rpm
20
f 6000 rpm
x 7000 rpm
x 8000 rpm
"x
15 "" 9000 rpm "
i" 10000 rpm
E
ý_ 11000rpm
73
> 10 1
x
" "x
1Y f
iii, :
list a
01 . " . . s. ..
0 11 10 100 1000
Figure A. 6 Volume DSD generated by the 69.85mm rotor with the Emulsor screen
at Q=0.41/s and (D = 30%, for varying N (rpm).
25
f 3000 rpm
5000 rpm
20
f 6000 rpm
x 7000 rpm
V0 z 8000 rpm
15
lu " 9000 rpm 1. x
+ 10000 rpm x*
m +
e -11000ipm'
10
X x"
A. "X "
+x 2
5
* -03
IuI;
o I.
01 1 10
10100 -exzw--
1000
Screen at Q=0.41/s and c= 40%, for varying N (rpm).
198
25
"3000 rpm
20 5000 rpm
" 6000 rpm
x8000 rpm
15
X9000 rpm
E
O
i
10
#' 'x'
"
"
"x"
5
01
01 10 loo WOO
Screen at Q=0.8 1/s and D= 5%, for varying N (rpm).
25
f 3000 rpm
5000rpm
20
f 6000 rpm
x 7000 rpm
8000 rprn
15 ri*
t 9000 rpm
10
Ir It $$$$$Jiifffi41 *i
s
o SM-aura-*-1i-acAM,>«-llt CIM aý .1*
01 1 10 100 1000
Screen at Q=0.8 I/s and 0= 10%, for varying N (rpm).
199
25
3000rpm
."
20 U5000rpmý---.
- -- ----. _
f 6000rpm'
I
x 8000rpm'
a x 9000rpm'
v
t ý
10 -r0
MX
ýk
0 r rrrrcrrý* YCYLiýýý
at "frýýýýrtrrr afErtiE
0 11 10 100 10 00
Screen at Q=0.8 I/s and = 20%, for varying N (rpm).
25
f 3000rpm
20 "
5000rpm
"
f 6000rpm xI
15 ýx 8000rpm ",
Wk
Ex 9000rpm
0ýx i
ö
> 10
46 Xx
Xx "
"x
w 0x
"" 0x
0_,... ýýýýr.., s>wýýýýý1r: iiýýtýtl ýi$i" iýrýý: t

01 1 10 100 1000
Droplet Diameter(microns)
Figure A. l 1 Volume DSD generated by the 69.85mm rotor with the Emulsor
Screen at Q=0.8 I/s and D= 30%, for varying N (rpm).
200
Square Hole High Shear Screen (69.85 mm Rotor):
25
" 3000 rpm I

5000rpm
20
" 6000rpm
7000rpm
8000rpm
e. . 9000rpm
10000rpm
z
- 11000rpm a
41r" ý
"
.
xA
+
"x .
.
-Y.
V
ý"xý " " A

Ti !"" V-
0
ý -ý -«-ý
01 1 10 100 1000
Figure A. 12 Volume DSD generated by the 69.85mm rotor with the SQHHS Screen
at Q=0.4 I/s and D= 10%, for varying N (rpm).
25 -
" 3000rpm
5000rpm
20
" 6000rpm
x 7000rpm
15
x 8000rpm
E " 9000rpm
> + 10000rpm
10 ! " A
11000rpm x
+ia It
+
+g 0++
a 1fLl tiadd9 "" t t" it

0 "a s"ý
0 11 10 100 1000
at Q=0.4 1/s and 1= 20%, for varying N (rpm).
201
" 3000rpm
5000rpm
:'0 r- ----
" 6000rpm
x 7000rpm
8000rpm
j
. 9000rpm
ö
10 10000rpm
r. i
11000rpm -
"
x
5i
"ýe
0 a "ý aý Nw a-a s-N aai tl tl, tl raY1Y tl

-i . w a
0 11 10 100 1000
Figure A. 14 Volume DSD generated by the 69.85mm rotor with the SQHHS Screer
at Q -= 0.4 1/s and D= 30%, for varying N (rpm).
25
" 3000 rpm
20 f- _
5000 rpm
" 6000 rpm
8000 rpm
15
e
Q) 9000 rpm
E "
E
°
> " 10000 rpm ..
10
-. -. _-. __.__- -------------. -----'---- ------ ---- ---
pie r
"
x
" %`
7 "
". ab-w
0 !". s was
rs"itiiiiýi-A_i-! _ - - ------- ------ --"---t-"--ýt-"-: a-... "....
-a-. _
01 1 10 100 1000
at Q=0.8 1/s and cb = 05%, for varying N (rpm).
202
25 . .......... - _....... ....
" 5000 rpm
20 . 6000 rpm ----
x 7000 rpm
15
. 9000 rpm -- ---
" 10000 rpm

ö
x
ýýx"t
ýY
5 x
"ýx"
"xf" x
01 1 10 100 1000
at Q=0.8 I/s and = 10%, for varying N (rpm).
25
20
" 5000rpm
6000rpm
x 7000 rpm
15
r 9000rpm
d
10000rpm
ö
x
10 -x x-
X x
J2
Y.
"X
ýý x.
o F"
1 10 100 1000
0f
at Q=0.8 1/s and c= 20%, for varying N (rpm).
203
50.8mm Rotor: Emulsor Screen
5000 rpm
20
" 7000 rpm
x 9000 rpm
" 10000 rpm

15
10 4ßx
"X
:k
"
5 .x
"x
Jr_N..11_ ALlr..1r II &i

0 !am* f. a.. rý W-t1 ýt a-f f a-: -; ..
01 1 10 100 1000
Screen at Q=0.4 1/s and c= 10%, for varying N (rpm).
20
. 10%
20%
X
15 "30%I --X
X40°. 6 j"
ö"
yXX "
E 10
ö" "
"".
XX
"A
"
5I"A
""" X i<
"A
"" "X"""iii X
0
011 10 100 1000
Droplet Diameter (rricrons)
Screen at Q=0.4 1/s and N= 7000 rpm for varying (D (%).
204
25 7
5000 rpm
A 7000 rprn
20
9000 rpm
15
.
10
Y.
A
7'. ..
X ` i..
' *..
0 tr .. r. *-*-r-* ýýý[-rýI{äfSa
rr - -A-A-sx. -r-a. arrarr... ar
01 1 10 100 1000
ierated by the 50.80mm rotor with the Emulsor

Screen at Q=1.4 1/s and (D = 10%, for varying N (rpm).
50.8mm Rotor: Slotted Head
25
5000 rpm
20 " 7000 rpm
" 8000 rpm
x 9000 rpm
"
15 - -x 10000 rpm
E
2 "
*
10 ..
tý r
"A'"
llC ll( 111CY 11ý

Yý Y lL lr 1Cä ý___
0 .aýi a-s-tea a-a sal Ali-*-I f i-a aaaa a_a-a i
-ý -ara __-
0.1 1 10 100 1000
Figure A. 21 Volume DSD generated by the 50.80mm rotor with the Slotted Head at
Q=0.4 1/s and 0= 10%, for varying N (rpm).
205
50.8mm Rotor: Square Hole High Shear Screen
25
" 3000 rpm
5000 rpm
20 f 6000 rpm -
x 7000 rpm
x 8000 rpm
15 . 9000 rpm
10000 rpm
"N
N
x
0
>
xx"ý
10 Ia -: $
>r= ýx
"
I "' "1
x
R
ý
t
I 0 DY i YY Y ý Y
0 SYH A T !ý C
9Ou anamesOaY - . . . .- i . . - _ _ . IL YAAONA 4" ..
01 1 10 100 1000
Droplet Diameter (micron)
Figure A. 22 Volume DSD generated by the 50.80mm rotor with the SHHS Screen
at Q=0.4 I/s and c= 10%, for varying N (rpm)
25
100
20%
20
A30%
x40%
15 x
oI xX
vI
E
"X f
10 ffX
X
"X
"
51 fX_
X
"£ "
fX
"
X
"
"" fX
_. ttEýt .x"X
o.1 1 10 100 1000 1

Figure A. 23 Volume DSD generated by the 50.80mm rotor with the SHHS Screen
at Q=0.41/s and N= 7000 rpm for varying (D (%).
206
50.8mm Rotor: General Disintegrating Head
25
5000 rpm
20
A 7000 rpm
8000 rpm
x 9000 rpm
15 x 10000 rpm a
10 xf
:
*x
f
i x
X x "
5 xx
xx ý
Xx! A
X
yt! A
x
x"
if60!
0 s*rrrýr**rrr ra t+rý , a ' ilk ja * 0r
-
0. 1 1 10 100 1000
Figure A. 24 Volume DSD generated by the 50.80mm rotor with the GDH at Q=
0.4 1/s and 1= 10%, for varying N (rpm)
,.,..... , vww Ipill ivi val yiii, 'r v/U)
207
Appendix 5
Table A. 2 Calculated GB and tB for measureddmaxand d32from ES and SHHSS
9Y Ui g LU 4
Y YY YY YY°YY
wW ww, wW w" ww u
,
to Co ` w
.r. r2 ý> `
^ rtr N 'M Co C-4 Co Co w f2 Co N NN
.N CU U N
Y YY Ir `Y L Ui' Y Y Y
ww,. W ww W WU ww Y
Co -- ö r- o ,w ww wwwWWWWw
t-
{y[ w
'r r- ti 1D
Co CO Co . o,
oC rN. Co C CU 9 ý . - ao acr
ý 4 7 V V' Qi rr M1 SIÖ v- K+ N O ! O i Of O! h. IN
.
MM MM MMMMM M Co 'M M csM Co Co Co M Co

ica 4
w
ý+i
Wp 4
Wd tu 44
W W ýi 44
Wt+ý Ww 44
UI W 4
wN
Q 449QC
WWW WU 4 4
, , e
ýp 2 . w(y UI UI
w 2 Ö; 4i r ?2
Vv "S *l:
1t Q ul 43 P .
V NN ri MI h1 n% [V of *0 4 v9 e' N N
r . e4 NN cm
99i q
4? C) cý c" agý cý
y!
cl
41 wW wW LUJ LU www wW ,W L1 WW EL! lii LI LLt
ý+ ýr1 O- pp Ä r , Q ii CO
1 . C. C) Co aD rt7 Ca 0)
- CO Q1 CO CIP W K
Os Co Kß W Op EL1 d7
ÖO Ö d
C G] O O OOO
% ih #t i}+++ } + + w
wm. m rn wöWT Wý Mw Co M ti
W Co M
CO t0 NN MOdN. C'4 N Co N. MMM Co N. N
M ey f V' <Ì Ni V of r ý- O i Co r tV NNr N N
ý,
*0 Co t0 Co 1l) Co Co Co Co Co Co Co Co V1 Co Co to Co to 4')
1 g'2 Ör N
9 N
99
Wpf w-w"
912 Y1?
Co C
9 «' 212T 12 1? W12 W2
r " rW w ,, 1J,
wwWW
cl N- Sgg Hi-5 2 NNNN U? M
N N G" C: Co CO Co CO fti N-
V r )
_ Co cm CP ºr o( 4U N R wi ui vi ui rp ,P
eL e t3 11 UI LU LU Lu Lu Lu e UI UI Ui
i F
wW w WW ia w ui w wW w
, , . . . ,w ,, Wwwww W w
_
i
emý C
ryQý
N
J
ýO
lj e4 l7 ui
$ ia
nj ly
ui
NN
Co
Cf lý
aA
M
iý
A
N
p
Ci
a'f
`2
lýF
i6 e fA <O t0 Co +ý CO
y Co C) Q Ö
C c4 OQ G d Co Co
UI ü1 k+f "h } 4 +++i + +
ww , wL
iJ W r w' ww Lii w r w w WWwww w
so t0 o . d ao Co o
r , Co w a) w
. , w
i Z : va *0 - e- 0T, 1
, Ue 00 'f" KO MW [R lw: OR t0 ý'y Of AO r
V Q
k) (0 N. Co Co PO CO to fV M Co Co Co 4l b Y') Co ! `. 1p
ui
E
4 Mp dýQp üpf ý x A
ei , Co , r NOp` CO r N , 1+ eý r
pU paö Ö SSS
û" O O OO O O
ÖÖ C >Ö Ö ÖÖQ (21 ÖÖ C) Ö ¢O ÖÖöÖ 0 Ö
l0 U tý Of Co Co OS W M. ch M t0 V) t0 V) t0 *0 N
' fV
1 ýOoy
, , ' r r rR rr
S S S
O OO OOO O
CÖ Ö C/ C i> Ö t7 Ö Ö Ö Ö Ö
cm CS CJ CT fý G
00 N- P. ! ti Co U) Co tC to h !ýP. 1.. h P-
Öý r[1 Pr A CO M Oa QY C
W M! <C NY M) 40 1n Co f0
dM Co C) M OJ CO SÖ4OO
op pý cp Z; C1 O
GO QO h rrr 0 N r - r< ' N
r CU c ra (C) . -
xN N- K H
${ C Q CUD ('1 ýf Ufa !7 QNi r
'd V) 2
'C± öýI C PNVNN N- N v N 2"" e+l Ö N
. c4
ö ^
- p /ý. S ýO aNO'Y p c"A 2 {ý t8 4 'FMt2 S3 tI curl F! Z uni,
w `8
W c.
Ä
ar0 M
NV
1z OJ
1ý
M
~M cý5 f. 0ý Mr
M fam. oD r'te'
hrN eNV
ýö
e0 tay
M
i0
{ýi 'K ýD K N Co N- VI
(D C>
Co C) C) MH C) C)
c2 c2 C> K2 CD ^ wý- r NN NN N NN
c3
ÖÖ fl C's Ö
c:i
Ö ÖÖC CD, Ö Ö Cü C) m Ö ÖÖ öÖÖ Co C Ö ti
CO Co Co Co Co Co Co eo (0 vi 00 cd a0 Co Co CO Co Co CO CO CO vv "a' vv a
C ÖÖ4 Ö OG Q p p ÖQ G> Ö Ö ÖÖÖÖ p p
Ci CD, 42 GÖ Ci C, c2
208
NNNNNNN NNNNNNN NNNN tV NNN
9w 4ww9w 99 9w Rw 9w 9ww V4RR
LU wAAOA
OO Ol O
CD
cf
w
Om
CD wR
A
O
ww
CO CD ww924
O
CO
Co OO
OD OO CO eW
w
d
AY
O7 w4
Of M w9
O7 ww
O (2 ww
1 ýo
Il} tK N SV /ý fý
Mi CO U) M
.> U) . [i Un CO u7 x U7 CO 421 - C') CY PN
NNNNNNN NNN tV NNN t') M CS e') M NNM pY

9999999
ww wwwWw
9999999
ww Www ww
99
w w
99 9 9999
,, . ui UA ui
w pý OW mO A) m rn 91 pý AW Q ýp ýp pp pý ö9 ,,,
r0 OA1 `v C14 C) CO Oi Gi Of G1
rrrrrNr r !Vrr r4 Y1 VY *f Y - O> M
MMMMMMM MM c7 c7 Nf MM CMMMM a> MMM

00099 c7 cý o o99 99O oOo00 0oao
X N CD N0W Wy {ý W 1W 0+ CrW . w. N w.,
03 A CO rCO 1N OWi CD
Co OD
oß 46 ui +f Un .f di ad CO CO ti IN N NN eý CO eil r4 f4
w ww ww ýp 2ýV
w wwwu w ww y w
Qý www 2W
ýp yp w
O
bf w ýy cý w
2NMO
ýp cp ww
p tp w
0v Og CO '! NM2
NN- IIDI-e
. '= -w P> NNNNNN %m N {G '7 CO 4- (0 OP O>
ÖOÖ262
82PÖx22 OC6OO OOOÖ
it tMtt+ tft `h ktt fff f ff
in
wwwwwww
CO CO CO fa CO rn Co wwwwwww
Co Co co ao Co w
Co w.
ww www rn rn www.
OO2 w,,
i7
NNO (y O 07. O iy t7 QO ý`7 OO ýO NNNN OO C">
r YMr - O- r Ob 4 N7 -v ! of e N. NNN
.+w sý "-
to Ln tn pöä ä `n äöä "' Q E U)(0 (0 y0 (0

wwwwww 2 X U) g CD
w wwwwwu lw E w w, w ww UI w w
ta 1+ p h; ethAi p NÖÖQÖ pr pf ýy .w ýy .
P! t'!
lZ
fei Cif NNN
O
MNNNNNN
1-fZ
-- aj Wwwr
W92
t
'f 1! )
00
CL)
C
of Co Uli CD NN V) .! > uY C CO Co
912 ý CUee e
ww p wwwww
CO MMN wwww wwW ww ýp .www , wwww
NN
ýp glb Q Co
a YN `C OD CO O7 CM') KI
s: QD 1ý 1ý Co ED CD b !N !V (+) P'S !V ! ýI P4
wwrrr w- 't
ry
CO t0 to to CO CO <O to to CO CO (O (O t0 A Co CO CD CO to
ao öc
2 o+ oý > o9o o0 o* o o of> oý of of g o+
wwwwwww + w iu w 11 wwwU) Z w w. www fo
www. W
Co OONMA ýp 0) <O 01 Qf CD aý 01 N CD A U) c0 a{ Co
im 49 h C7 r ý[] iD 9D m cD _ N " l'1 P7 OD NA w: C') f'r) C[)
n nj c4 M bf vi C'J NNN Ö CCi f- W-W 4 ID
w w- 1C tG
CO CO Q' Co sr H7 V to lh <V C') Vt SG 90 f0 fd t0 O. in 07 0)

VQ LD V Co M1 00 QP . w CO P .a O (0 Q mmmm CO o0 N NN
CJ 61 Ov Oa ýfI nNO
C» tNn N N 1ý v1 Aa
ar
.: 1 M, ,1
8
g
O OaOQQO9 OOOOd c> 4 GG OOO C1 O
Ö CI ÖÖ co ÖG ÖÖÖÖpCQ ÖÖÖÖÖ ÖÖO
Ö ZOO :9 ptýý 4"' ýýa; 2MS, b

$(> C7
, O 47 ý ý+7 LO N ti
[O 47 N N +ý ýt o0 A g t+1 d 8ý
E _
ÖÖOÖÖÖÖ ÖÖÖÖgOC p SSrO 00 00
U $
0000oo OoCoo 0000 00
aööö 0
oö ci ci ei a Co ödo ti c> oc cS öd
. [f . [f CO u! t0 V1 CO . l! ýp EO ýp (0 (0 M (ý P. N CO OD
C1 C1 <O CD < t0 of Öo 192 99
E Co 90 t0 0: e0 tjk c 79
Äh
A. r"Ir
(
)
0 aA
uý q qvýOo ý7 r- 1- Owe YZ
NpNOpc9 N Ci t N C'> i'J C7 t'i ý7 f'i
Ný. N9 Ä Öl NNÄ Ci N Cý lft

. fl OAI C7f ON CD OD dD
NNN Z i0
M !Jm PZ C)cii r N r
c'1 t" MM a-
r ýy wO ýp tý ospýA ÖO A
9 0O ýp CO
M Y? 11 N f0 gn O9 Cý cý Co
ý» 9
!2 ýy N ýp Q, M ýy ^ fM1 ýy AN I'+ 1+ N
NS OD NM N Of NwM ýO fli ' "
NV {O Of *- N Co O -wm Co
w
r4 Cl! NNNNN t'7 CO in 91 Cf th C) OCOOO Q4O rrrrw vý w

C7 C7 ooo7o o7QÖooo 00 a> 0D l7
GCÖOÖÖÖ CI
N veavvv vva *r aq CO e 4e mw Co m co oe co Co to CO CO CO
odöö0öö 0dööö0o 00ö0öööö 0odddddö
209
N NN NN NN NNNNN NNNN NNN NN N
9 ww ww 92 999 Mf 9R RR 9a 9 4R 9
K , +
w+1Ae b
W W ti
W
M
Ö
Cý7 t91
N n c'l ui w ý
hA[!. R7 OO OO OA O
O b' j
m O
V NV vi mN
w 't NN cal cri vi U, Afi tai AYi uj U, Co , lj
ry N NN N N CD CD M CD CI) H NNN NNN NN N

22 ý9 99 9 99 99999 999 999 99 9
d W WU J, w w1.., Wrw
Ö wwwwp
n ti wpýwpýww wwW w w
i0 O> aD t0 4 ig Oý N
U n w (O
ei OA Ci N iD ui off rt +r t : 4- N rN
.t .i r+ w+
Wa w, w LLJ ýtpyJ llk 4 Jwww,,, T 'f T

A wýQ www W WQý W
/, Tt i+ rÖ h O
d ' W 07 ' M V' WO N
fq O {O 00 N OD
y Y '! iii 4
. MM V {ý1 NNNrN ri r: {p yj Oj Oj ap ý, ý
n
41 9 öö ö9 9R ü9 ö ö eý IM) ö
+9 CD cri 99
RR 9
R
w w
OV ui w
Co w
O ýu w
O w
ü
iCD u i w i<»ü U) . M Ww
W M
r h go 40 V üV V O G O
A WV
r
NM ut 10 1+ CO V> 92 Aý mO eý Nm/, _
Cy N
r % rr «- ob U, CO 5D CO NN c-ä r N] 1'ßf N NN N
g Ö 8O LU It g
i t W + W,
w w WW,. w ý Www Li w ttJ WW lL
t- to m ý ,v,
as i S :° ry ry 2.
a Na
w ýe a+ NN ý u q&q
` w r re NN C'i Ci rrre. Oi .ý `- qi r
vi Y, Ii) to U7 ti1 V] Uf tC lA to 1A U, U, Y5 W 7f $A v9 6«> U, Hf

13
41
Q
wppw w0p wopw wQfw
(4 ww
CD M1 wp P w tu ww e
40 www
ÄO .w w.
C7 CD AD Iti Ol N C/ V3 --N.
V f+ C..j ')
N) e+) Nf ("f lV V U, AÄ tÖ I+ h. O) CDCD tV CD [V ! +f N !VN l+f
N W W w w N W ww W w www Ww w
v <0 C)
NV1 7 V
eT
OA
tý
4
N
M
mo
hVW ra so W
M N. GO
ý
rr 'r r, tV tý Pf lei dý ED r iD fý tý CQ Oi
CC
.
TN
p g ýp 22 82 822s6 99 9
6 O
t 4
OO28 O99
d i 4t i+4f 4 + it +
j ý! J ( j]
3 J, ýtJ
n z, 2Z 9O r_ ä ,
2O,,, Z W
ONO r^ (V I Dr- f0 KO
CV N C') 'R l'/ m 1". NN NN
V ui CO Ir 4V CD cV
Co V- ZE % '» N N9 Oi 6? vo EOV 1V ^ 1
. f2 Co,
O IO-A kti v V' V o0 p
E O p O p
, ' C_V
t3 p ýj p 1` A
OO OO __ _N
Q Op OOO OO OO
G? C> t7 CD G> 9 4T Ö CG ÖG Ö
C7 Ci C] SS C? Ci c, CD 0
T (4 cli W 44 M P, C. > U'3 MM CC CM h cock (pp
CC) o0) cq V n. N2N n] MM Ö N
e V ýý Ö ý m
tt [ +O P .ý pv 1 >n Q 47 v
_ _ p o g fUO i 1_
_ p _ _ , NI N e
8 OO OO UOOOO OCOO QOO OO O
ö d cS öö +ö odd ö 6öö
x:b o cS CD od 00 ö
N N ýp ý
'j N Of tG Aý ýh qnp Alf. t[= Of phi R 00f Orb
N 4P '?
CO N mp
Zýý333'3 N
cýyV rp !h LS G CD of a> a0. ob
r N- O NN N MN q o ý!y N
N N C') . f') C') C') c7
e 4 9 8q s3 gt V. 222 9
O2 O ANA OZC i+9 I. O; O
TS cm (D ý ppC
r^ ýr CD t0 O to r, c'! (C, C) Ö C') CD V ýT
N 0) V <'i CD NNN e- VWV i7 x`1 mx
Ö ti
O OD MO ON tVý, i0+ O H'ý VO 1ý
N N^
w Cy CA NNV m sa
ý-ýy w pý ýp _N pýtý cp ONE pp ep ý
w0- s"f ýD NN X11 G»D » 'tV u) 10 N AA Co
Ottýý Ocgý
Z p, b t=s O 4C7 O tl OO O OO O O dO O OO OÖ
Cýt pCS OO
N NNN NN CV N r er rrrrw Ö fi7 NNMNN c'? a+7 r7 C'1 A+! th

O C i OO OOOC? 16 42 ÖCO i7 Ö ÖÖÖÖOÖ ÖÖp Gf Ö ÖÖ
Co, Co Co e em
a9 Co Co «4, 4er ýt r -Z -e
O ÖÖÖ Ö ÖQG V, 000ÖÖ c:, c> Ci O R? Ö C> C! CJ ä Ö Ci
öO
210
211

Rotor and Stator

Uploaded by

Copyright:

Available Formats

Rotor and Stator

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rotor and Stator

Uploaded by

Copyright:

Available Formats

CRANFIELD UNIVERSITY

LIQUID-LIQUID DISPERSIONS FROM IN-LINE

SCHOOL OF MECHANICAL ENGINEERING

SCHOOL OF MECHANICAL ENGINEERING

Academic Year 1998-2001

LIQUID-LIQUID DISPERSIONS FROM IN-LINE ROTOR-STATOR MIXERS

Supervisor: Prof. D. E. Brown

systemwas usedin the testsand the effect of stator geometry,rotor-speeds,flowrates and

were mainly responsible for drop break-up.

d32 = 2x109(1+ 20(D)(WeRe4)-'"

along with the nominal residencetimes in eachregion to determinewhere in the mixer

Many thanks are due to PeterMatthews, Harold Rothmanand everyoneat Silverson

LIST OF TABLES .......».....»..»..»..».»..»..»..»..».....»..»..».»..»..».»..»..»..».»..»..»..».....»..........».»......».....».IX

LIST OF PLATES ...»..».....»..»..»..».»..»..»............................................................................ X

CHAPTER 1. INTRODUCTION ...»..»..»...........»».»..»..»..»..».»..»..»..».»..»..».»..»..»..».»..»................ 1

CHAPTER 2. LITERATURE REVIEW 3

2.1. ROTOR-STATORASR BACKGROUND.............................................................................................. 3

2.2.1. Forces affecting drop stability 17

S1 MEALiu MuT TEaNIQcx VALU)AnOI 104

CHAPTER 6. DISCUSSION 106

CHAPTER 8. FUTURE WORK 171

APPENDIX 1 EXPERIMENTAL RIG OPERATING PROCEDURES. 181

APPENDIX 2 NAMES AND ADDRESSES OF EQUIPMENT SUPPLIERS 188

APPENDIX 3 THEORETICAL BACKGROUND TO DIFFRACTION MEASUREMENT....... 190

8.2.1. Principles of Laser Diffraction 191

APPENDIX 4 DROPLET SIZE DISTRIBUTIONS 195

APPENDIX 5. »..».....»..»..»..».»..»..».»..»..»..».»..»..»..»..».»..»..»..».»..»..».»......»..».»..»..».. 208

Silverson Machines Ltd) 7

are at the top and the rotors are below ........................................................... 10

viscosity ratio (Grace, 1982)..........................................................................

rotor-stator with samplepoint and Video probe in place; (b) Watson

and (D =10% ..................................................................................................117

Figure 6.9 d32/D versus (WeRe4)-1/7for different (D 125

High ShearScreen 131

rotor and Emulsor screen..............................................................................

= 0.41/s and «= 10%, for varying N (rpm) 154

Figure A. 1 Flow diagram of the rotor-stator liquid-liquid contactor rig 182

Screen at Q=0.4 10%, for varying N (rpm)

screen at Q=0.41/s = 30%, for varying N (rpm) ............................ 198

Screen at Q=0.8 Us and c= 5%, for varying N (rpm) 199

Us and (D = 05%, for varying N (rpm)

Screenat Q=0.8 Usand (D= 20%, for varying N (rpm) 203

Screenat Q=0.41/s and c= 10%, for varying N (rpm) 204

Screenat Q=0.4 Vs and N= 7000 rpm for varying (D(%) 206

Q=0.4 Usand N= 7000 rpm for varying 0 (%) 207

(69.85mm rotor, ES, 0.41/s and 4=10%) with Xk 155

and SHHSS (at 0.41/s and 4=10%) 159

emax Maximum local energydissipation W/kg

cy Dynamic viscosity of the dispersedphase Ns/m2

a Interfacial tension N/m

ze The disruptive stressdue to turbulent pressure

zk Kolmogorov time scale s

zy Resistive force per unit areadue to viscosity N/m2

v Kinematic viscosity m2/s

Np Complex refractive index of the particle

np Real part of the refractive index

m Mass flowrate kg/s

Re Reynolds number p CND2

ATs Temperature rise over a single pass K

We Weber Number P_N2D3