massoud thesis

Numerical Design Study of Chaotic

Mixing of Magnetic Particles

A thesis submitted for the degree of

Master of Philosophy (MPhil)

by

Massoud Zolgharni

School of Engineering & Design

Brunel University

December 2006

i

Abstract

Finding an effective mixing mechanism has been a challenge in the design

and fabrication of microfluidic platforms and so-called Lab-on-a-Chip devices due

to the fact that these devices usually operate under low Reynolds number flows

where it is not possible to create turbulence to enhance mixing. In such cases,

diffusion at the molecular scale becomes the sole mechanism of mixing. To

promote the rate of diffusion, various methods of diffusion interface stretching

have been proposed. Here, a numerical design and simulation of an active chaotic

micro-mixer for superparamagnetic particle-based mixing is presented.

Magnetophoresis force is utilized as an external force to manipulate the particles.

The aim of the design is to spread the micro-particles in order to tag the suspending

target biological entities in a bio-fluidic buffer. After the mixing process, tagged

entities can be separated from their native environment and other non-target

molecules in a separation process. Separated entities can be used in other

downstream protocols such as PCR (Polymerase Chain Reaction) or detection

algorithms in diagnostic micro-systems.

In the proposed concept, a straight channel with two embedded serpentine

conductors beneath the channel is utilized to produce the chaotic pattern in the

motion of particles and intensify the capturing of biological cells. Two flows; bio-

cells suspension and the particle laden buffer, are introduced into the channel and

manipulated by a pressure-driven flow. While the bio-cells follow exactly the

mainstream, the motion of the magnetic particles is affected by both surrounding

flow field and localized time-dependent magnetic field generated by periodic

activation of two conductor arrays.

ii

A two-dimensional numerical simulation of the mixing process is

performed in order to characterize the efficiency of the micro-mixer. In the present

study, the main focus is on the effect of two driving parameters (i.e., the fluid

velocity and frequency of magnetic activation) on the mixing quality. Two criteria,

which are dependent on the performance of the system are investigated for a wide

range of driving parameters. Lyapunov exponent (first criterion) as an index of

chaotic advection is found to be highly dependent on the driving parameters and

the maximum chaotic strength is realized corresponding to the Lyapunov exponent

of 0.36.

Moreover, it is found that capturing efficiency (second criterion) in the

mixer cannot be used as a stand alone index, which might suggest operating

conditions that are not practical. Therefore, both indices need to be taken into

account while characterizing the device. Maximum capturing efficiency is found to

be 67%, which means that more than half of the existing cells are labelled with the

magnetic particles during the mixing process.

iii

To my parents

for their love and sacrifices…

iv

Acknowledgements

First and foremost, I would like to thank my principal research advisor, Prof.

Wamadeva Balachandran. I am grateful for his frequent discussions, supports and

encouragements throughout my research which extended my abilities in research.

I would also like to thank Dr Predrag Slijepcevic for all discussions on the

biological aspects of this research.

I would also like to acknowledge Prof. Tony Anson for all his support. Throughout

my entire graduate studies at Brunel, I have been fortunate to enjoy the benefits of

working with him as my friend, supervisor and employer.

Special thanks goes to my friends. Thanks to Amir and Arash for being my best

friends who I shared the moments of happiness and sadness with. I know I can

always count on you. Thanks to Mohamad for always being by my side since the

first step of this journey. Thanks to Reza for his kind helps through all research

and life challenges over here.

I would like to express my deepest thanks to my family for their support of all my

life choices and their love, which is a constant source of strength for everything I

do. My parents, my brother and sisters, you have always been there to support me

in every endeavour I have undertaken. Your faith in me is cherished and I can

never thank you enough.

Last but not least, I would like to thank God for always hearing my voice and giving

me what I need.

Thank you.

v

Abbreviations

2D Two Dimensional

3D Three Dimensional

µ-TAS Micro Total Analysis Systems

CE Cell Capturing Efficiency

CFD Computational Fluid Dynamics

DEP Dielectrophoresis

DI Deionized (Water)

DM Diamagnetic

DNA Deoxyribonucleic Acid

EHD Electro-hydrodynamic

FM Ferromagnetic

LE Lyapunov Exponents

LOC Lab-on-a-Chip

LPCVD Low Pressure Chemical Vapour Deposition

MHD Magneto-hydrodynamic

MEMS Micro-Electro-Mechanical Systems

N-S Navier-Stokes

NdFeB Neodymium-Iron-Boron Magnet

PACVD Plasma Assisted Chemical Vapour Deposition

PDMS Polydimethylsiloxane

PIV Particle Image Velocimetory

PM Paramagnetic

PTV Particle Tracking Velocimetory

PZT Lead Zirconate Titanate

Re Reynolds Number

RNA Ribonucleic Acid

SAR Split-and-Recombine

SEM Scanning Electron Microscope

vi

SHM Staggered Herringbone Mixer

SPM Super Paramagnetic

St Strouhal Number

UV Ultra-Violet

vii

List of Figures

Chapter 1

Figure 1.1 Conceptual diagram of a typical LOC 2

Figure 1.2 (a) system level block diagram of the LOC, (b) block diagram

of sample preparation sub-system

3

Figure 1.3 Magnetic micro-particles 4

Figure 1.4 Conceptual diagram of magnetic cell sorting 5

Figure 1.6 CD4-T cell isolated with Dyna-beads 6

Chapter 2

Figure 2.1 Velocity profile of a steady pressure-driven laminar flow with

no-slip boundary conditions developed in a micro-channel

14

Figure 2.2 Basic designs in parallel lamination; (a) T-mixer, (b) Y-mixer 18

Figure 2.3 The observations of the mixing process at the junction of micro

T-mixer at different applied pressures

19

Figure 2.4 Multi-lamination 20

Figure 2.5 Vortex micro-mixers 20

Figure 2.6 Parallel lamination 21

Figure 2.7 SAR configuration 22

Figure 2.8 SAR lamination 23

Figure 2.9 Basic idea of the micro-injection mixer 24

Figure 2.10 Mixing through winding microfluidic channels shown 24

Figure 2.11 Various designs for producing chaotic advection at high

Reynolds numbers

26

Figure 2.12 Micro-mixer designs for mixing with chaotic advection at

intermediate Reynolds numbers

28

Figure 2.13 Micro-mixer designs for mixing with chaotic advection at low

Reynolds numbers

31

viii

Figure 2.14 Micro-mixer designs for mixing with chaotic advection at low

Reynolds numbers

32

Figure 2.15 Activation by moving parts 33

Figure 2.16 Active micro-mixer based on pressure field disturbance 34

Figure 2.17 Numerical simulation results 35

Figure 2.18 Schematic illustration of the micro-chamber for mixing of red

blood cells by ultrasound irradiation

36

Figure 2.19 Snapshots showing multi-bubble induced acoustic mixing in a

chamber at time

37

Figure 2.20 Active micro-mixer based on the EHD disturbance 38

Figure 2.21 Active micro-mixer based on the MHD disturbance 40

Figure 2.22 Active micro-mixer based on switching electroosmotic

disturbance

41

Figure 2.23 Electroosmotic micro-mixer 42

Figure 2.24 Active micro-mixer based on electroosmotic disturbance 43

Figure 2.25 Active micro-mixer based on dielectrophoretic disturbance 44

Figure 2.26 Active micro-mixer based on magnetophoretic disturbance 45

Figure 2.27 Mixer based on magnetophoretic disturbance 46

Figure 2.28 Schematic diagram of the realized three-dimensional

microfluidic mixer with embedded permalloy parts

47

Figure 2.29 Experimental results 47

Chapter 3

Figure 3.1 Magnetic responses associated with diamagnetic materials (left)

and paramagnetic materials (right)

54

Figure 3.2 Magnetization (M) versus magnetic field (H) 55

Figure 3.3 Schematic depiction of spin arrangements in different types of

magnetic materials

56

Figure 3.4 Coercivity as a function of particle size 57

Figure 3.5 Domain structures observed in magnetic particles: (a) super

paramagnetic; (b) single domain particle; (c) multi-domain

particle

57

ix

Figure 3.6 Magnetic response of ferromagnetic particles 58

Figure 3.7 Illustration of the concept of super-paramagnetism 58

Figure 3.8 Sphere of radius R and permeability μ2 immersed in a media of

permeability μ1 and subjected to a uniform magnetic field of

magnitude H0.

60

Figure 3.9 Magnetic flux and magnetic field for a uniformly magnetized

particle

62

Chapter 4

Figure 4.1 Channel with straight embedded conductor beneath it 66

Figure 4.2 Generated magnetic field inside the channel 67

Figure 4.3 Applied magnetic forces on the particles in the channel.

Colour-map represents the magnitude of the force and arrows

show its direction

68

Figure 4.4 Sensitive dependency on initial conditions 70

Figure 4.5 Producing stretching and folding using positive and negative

DEP forces generated by different frequencies

73

Figure 4.6 Burst-view of the proposed mixer 74

Figure 4.7 Top view of the proposed mixer illustrating one mixing unit 74

Figure 4.8 Magnetic field near the tip of one tooth in one mixing unit

during a single phase of activation

76

Figure 4.9 Magnetic forces exerted on one single particle at different

heights above the conductor in one mixing unit during a single

phase of activation

78

Figure 4.10 Conceptual diagram of one single circular tip in (a) three-

dimensional and (b) two-dimensional models

80

Figure 4.11 Variation of the force against width (H=20 μm) 81

Figure 4.12 Variation of the force against thickness (W=20 μm) 81

Figure 4.13 Concept of the model illustrating the thermal boundary

conditions

83

Figure 4.14 Temperature distribution on the boundaries 84

Figure 4.15 Maximum temperature rises in the domain for different cross-

sectional areas

84

x

Chapter 5

Figure 5.1 Schematic diagram of coupling 87

Figure 5.2 Free-body force diagram of the particle for a one-way coupled

problem

88

Figure 5.3 Boundaries of one simulated (extended) and one mixing unit 92

Figure 5.4 Phase shift control signal 93

Figure 5.5 Developed parabolic fluid field velocity inside the channel.

Arrows show the direction and magnitude of the velocity field

94

Figure 5.6 Advection of cells and particles within three and a half mixing

units when no external perturbation is applied

95

Figure 5.7 Advection of cells and particles within three and a half mixing

units with magnetic perturbation (St=0.4, V=40 μm/s)

97

Figure 5.8 Consecutive stretching and folding in trajectories which results

in chaotic advection (St=0.2, V=45 μm/s)

98

Chapter 6

Figure 6.1 Block diagram of the various techniques employed for

characterizing the micro-mixers

103

Figure 6.2 Schematic illustration showing that the λ-map 105

Figure 6.3 Schematic illustration for calculating the largest Lyapunov

exponent

105

Figure 6.4 Convergence of the largest Lyapunov exponent for one particle 107

Figure 6.5 Concept of the collision where cell is tagged by the magnetic

particle

108

Figure 6.6 Initial positions of the magnetic particle for computation of the

largest Lyapunov exponent

109

Figure 6.7 Variation of characterizing indices versus different system

operating conditions: (a) cell capturing efficiency, (b) largest

Lyapunov exponent

110

Figure 6.8 Trajectories of particles at St=0.8 and V=40 μm/s; rectangles

indicate the location of trapped particles

111

xi

List of Tables

Table 2.1 Main characteristics of water at 20 ºC and 1 atm 12

Table 4.1 Properties of the reference magnetic particle 67

Table 4.2 Electrical and thermal properties of the glass and copper 82

xii

Table of Contents

Chapter 1 – Introduction 1.1. Lab-on-a-Chip Systems 1

1.2. Magnetic Cell Sorting and Isolation 3

1.3. Advantages of Magnetic Cell Sorting 5

1.4. Motivation and Objectives of the Research 7

1.5. Contribution to Knowledge 8

1.6. Outline of the Thesis 8

Chapter 2 – Literature Survey 2.1. Microfluidics-Basic Concepts and Definitions 10

2.1.1. Newtonian fluid 10

2.1.2. Flow regime 11

2.1.3. Reynolds number 11

2.1.4. Incompressible flow 12

2.1.5. Navier-Stokes equations 12

2.1.6. Steady flow 13

2.1.7. No-slip condition 13

2.1.8 Pressure driven flow (Poiseuille flow) 14

2.2. Mixing in Microfluidics 14

2.2.1. Diffusion 15

2.2.2. Chaotic advection 17

2.3. Passive Micro-Mixers 18

2.3.1. Basic T-mixer and Y-mixer 18

2.3.2. Passive micro-mixers based on multi-lamination (parallel lamination) 19

2.3.3. Passive micro-mixers based on SAR configurations (serial lamination) 21

2.3.4. Injection micro-mixers 23

2.3.5. Droplet micro-mixers 24

xiii

2.3.6. Passive micro-mixers based on chaotic advection 25

2.3.6.1. Chaotic advection at high Reynolds numbers (Re>100) 25

2.3.6.2. Chaotic advection at intermediate Reynolds numbers (10<Re<100) 26

2.3.6.3. Chaotic advection at low Reynolds numbers (Re<10) 29

2.4. Active Micro-Mixers 32

2.4.1. Micro-impellers 32

2.4.2. Pressure field disturbance 33

2.4.3. Acoustic/ultrasonic disturbance 36

2.4.4. Thermal disturbance 37

2.4.5. Electrokinetic disturbance 38

2.4.5.1. Electro-hydrodynamic (EHD) disturbance 38

2.4.5.2. Magneto-hydrodynamic (MHD) disturbance 39

2.4.5.3. Electroosmotic disturbance 40

2.4.5.4. Dielectrophoretic disturbance 43

2.4.5.5. Magnetophoretic disturbance 44

2.5. Discussion 48

2.5.1. Fabrication 48

2.5.2. Performance 48

2.5.3. Application 49

Chapter 3 – Magnetophoresis 3.1. Introduction 51

3.2. Magnetic Field and Magnetic Materials 51

3.2.1. Diamagnetic materials 53

3.2.2. Paramagnetic materials 53

3.2.3. Ferromagnetic, ferrimagnetic and anti-ferromagnetic materials 54

3.2.4. Super-paramagnetism and magnetic nano-particles 56

3.3. Force on a Magnetized Particle in a Magnetic Field 59

Chapter 4 – Basic Design of the Micro-Mixer 4.1. Sources of Magnetic Field 65

4.2. Magnetic Force due to Current Carrying Conductors 66

4.3. Chaotic Mixing 69

4.3.1. Chaos theory 69

4.3.2. Chaos in laminar flows 70

xiv

4.4. Basic Design 72

4.5. Scaling Effects 78

4.5.1. Magnetic forces 79

4.5.2. Electro-Thermal analysis (Joule heating) 82

4.5.3. Conductor size 85

Chapter 5 – Numerical Simulations and Results 5.1. Multiphase Flows 86

5.1.1. Phase coupling 86

5.1.2. Motion of a single particle in a viscous fluid 88

5.2. Numerical Simulations 90

5.2.1. Simulation procedure 90

5.2.2. Simulation parameters 92

5.3. Simulation Results 93

5.3.1. Advection of the cells and particles 94

5.3.2. Basis of chaotic advection in particles 98

Chapter 6 – Characterization of the Micro-Mixer 6.1. Mixing Assessment 100

6.1.1. Experimental techniques 101

6.1.2. Numerical techniques 101

6.2. Characterization Methods Used in this Study 103

6.2.1. Lyapunov exponents 104

6.2.2. Cell capturing efficiency 106

6.2.3. Results and discussion 108

Chapter 7 – Concluding Remarks and Future Work 7.1. Conclusions 112

7.2. Recommendations for Future Research 114

7.2.1. Modified particle properties 114

7.2.2. Three-dimensional mixing 114

7.2.3. Coupled simulations 115

7.2.4. Experiments 116

xv

References 118

Appendix A - COMSOL Multiphysics Simulation 130

Appendix B - Calculation of the largest Lyapunov exponent 142

Appendix C - PUBLICATIONS 144

Chapter 1

Introduction

1.1. Lab-on-a-Chip Systems

Over the past decade, the advent of Micro-Electro-Mechanical Systems (MEMS)

which is based on the miniaturization of mechanical components and their

integration with micro-electrical systems, has created the potential to fabricate

various structures and devices on the order of micrometers. This technology takes

advantage of almost the same fabrication techniques, equipment, and materials

that were developed by semi-conductor industries. The range of MEMS

applications is growing significantly and is mainly in the area of micro-sensors

and micro-actuators. In recent years, miniaturization and integration of bio-

chemical analysis systems to MEMS devices has been of great interest which has

led to invention of Micro Total Analysis Systems (μ-TAS) or Lab-on-a-Chip

(LOC) systems. Since the majority of chemical reactions occur in liquid

environments, the development of µ-TAS is essentially connected to the design of

liquid handling micro-devices (e.g., micro-pumps, micro-valves, micro-flow

sensors, micro-filters, micro-separators and micro-mixers). In fact, microfluidic

platforms are utilized to add an analytical functionality to the system in addition to

its electrical function(s).

New techniques to interface analytical systems with electro-mechanical

components are continuously being developed and offer the design and fabrication

of μ-TAS with a wide range of applications including drug delivery systems,

monitoring devices, nucleic acid-based analysis and automatic point-of-care

diagnostic micro-chips. In diagnostic applications, it is possible to perform all

Chapter 1 – Introduction 2

traditional bench-top protocols such as sample introduction, handling, extraction,

purification or isolation, amplification, filtering and detection.

The main advantages of μ-TAS over traditional devices lie in lower fabrication

costs, improvement of analytical performance regarding quality and operation

time, small size, disposability, precise detection, minimal human interference and

lower power consumption. Moreover, the problem of rare chemicals and samples

which restrain the application of genetic typing and other molecular analyses has

been resolved by employment of μ-TAS.

Figure 1.1 shows a conceptual diagram of a typical LOC system which includes

different microfluidic components for extraction, amplification and detection of

DNA molecules from whole human blood.

Figure 1.1 Conceptual diagram of a typical LOC. It consists of fluidic and electronic components

for human DNA analysis (© Brunel University).

Figure 1.2 illustrates block diagram of the whole system (a) and sample

preparation sub-system (b) in micro-chip under development at Brunel University.

This project focuses on the mixing process as a major step in sample preparation.


Figure 1.2 (a) system level block diagram of the LOC, (b) block diagram of sample preparation

sub-system.

1.2. Magnetic Cell Sorting and Isolation

In bio-medicine it is often essential to separate specific biological entities or cells

out of their native environment in order that concentrated samples may be

prepared for subsequent analysis in downstream or other applications [1].

Generally, there are two types of magnetic sorting when working with cells. In the

first type, cells to be isolated demonstrate adequate intrinsic magnetic property so

that magnetic separations can be performed without any modification. There are

solely two types of such cells in the nature, namely red blood cells (erythrocytes)

containing high concentrations of paramagnetic hemoglobin, and magnetotactic

bacteria containing small magnetic particles within their cells [2]. In the second

type, non-magnetic (diamagnetic) target entities have to be tagged by a magnetic

label to achieve the required contrast in magnetic susceptibility between the cell

and the solution. Through employment of a magnetic label and a proper magnetic

field, a five-order-of magnitude difference in magnetic susceptibility between a

labeled and unlabeled cell may be obtained [3]. These labels are often known as

magnetic micro-particles or micro-beads.


The tagging is made possible by modification of the surface of the particles in a

way, which leads to chemical binding between target entities and particles. In this

technique, usually polymer particles are used and their surface is chemically

functionalized through a coating process, thereby providing a link between the

particle and the target site on a cell or molecule. This coating is a specific bio-

compatible substance and can be an antibody or an m-RNA string but the

possibilities are unlimited. Figure 1.3a and 1.3b show 1 µm diameter magnetic

particles and particles with different functional groups attached to their surface,

respectively.

Figure 1.3 Magnetic micro-particles (a) 1µm Dyna-beads [4]; (b) schematic diagram of surface

functionalized magnetic particles [5].

If magnetic particles are coated with an antibody and then mixed into a solution

containing the target antigen along with other materials only the target antigens

will bind to the antibodies and thus to the magnetic particles. If the magnetic

particles can be subsequently separated from the solution the target antigens will

also be separated from the solution in this way. The separation step is made

possible through utilizing magnetic properties of the particles. The particles used

for this purpose are mostly polymer particles doped with magnetite (Fe3O4) or its

oxidized form maghemite (γ-Fe2O3) and are magnetized in an external magnetic

field. Such external field, generated by a permanent magnet or an electromagnet,

may be used to manipulate these particles through magnetophoresis phenomenon

(i.e., migration of magnetic particles in liquids).

By virtue of their small size; ranging from 100 μm down to 5 nm, particles lose

their magnetic properties when the external magnetic field is removed, exhibiting


superparamagnetic characteristics, which means they have neither coercivity nor

remanence. If the fluid mixture containing magnetically labeled cells are passed

through a region where there is magnetic field, particles and therefore tagged cells

will be immobilized while rest of the fluid is washed away. In fact, magnetic

particles are used as a label for actuation. In the next step, magnetic field is

removed and particle-cell complex is free to flow and be collected for further

analysis in downstream. Figure 1.4 illustrates the concept of magnetic cell sorting.

Figure 1.4 Conceptual diagram of magnetic cell sorting.

Prior to separation of the cell-particle complex from contaminants, magnetic

particles should be distributed throughout the bio-fluidic solution which contains

target cells. This is done by a mixing process which helps to tag the target with

particles. However, mixing remains as the main challenge throughout the whole

cell sorting process and has a significant impact on the efficiency of the protocol.

Therefore, the objective of this research is to propose a practical method for the

mixing process which is discussed in section 1.4.

1.3. Advantages of Magnetic Cell Sorting

Magnetic particle-based manipulation, in particular sorting, in microfluidic

systems is a technique, which offers to simplify and integrate the isolation and

rinsing procedures for extremely small samples of biological materials. In

molecular biology studies, this technique has been widely used for the purification

of specific target bio-molecules, e.g., cell, DNA, RNA, protein, or other macro-


molecules, out of the heterogeneous suspension [6]. Compared to other traditional

and bench-top techniques, magnetic sorting of the cells offers several advantages.

It is relatively simple and fast and allows the target cells to be isolated directly

from crude samples such as blood, bone marrow, tissue homogenates, cultivation

media, food, water etc. Those cells isolated by magnetic method are normally pure

and viable [2].

A conventional isolation method such as centrifugation applies a large shear force

on biological entities. Use of magnetic method can eliminate mechanical force

and therefore, prevent possible damages to bio-cells. With the advent of micro-

fabrication techniques, the miniaturization and system integration of the magnetic

sorting protocol onto the chip will have a significant impact on reducing the

amount of rare samples and expensive reagents [7]. However, in such micro-

systems, the inherent problem of mixing will arise.

On the other hand, the applied magnetic field does not interfere with the

movement of ions and charged solutes in aqueous solutions (at low flow rates) as

does the electric field. Moreover, the large differences between magnetic

permeablities of the magnetic and non-magnetic materials may be exploited in

developing highly selective sorting protocols [2]. Moreover as required in some

applications, it is possible to remove the magnetic label from the isolated cells in

an elution process in order to make it ready for downstream applications and

analyses [8]. Figure 1.5 shows one isolated CD4-T cell with Dyna-beads [9].

Figure 1.5 CD4-T cell isolated with Dyna-beads [10].


1.4. Motivation and Objectives of the Research

While in bench-top protocols mixing is carried out through mechanical

phenomena, in micro-scale devices mixing remains a challenging task. Mixing

stage has a crucial effect on the efficiency of the whole sorting protocol. Without

a proper mixing, a low percentage of the target cells will be tagged by the

particles and consequently, lower number of the samples can be isolated from the

mixture later in downstream even in case of a good separation step.

However, in micro-scale devices where the Reynolds number is often less than 1,

mixing is not a trivial task due to the absence of turbulence. In this scale, laminar

pattern is a dominant feature of the micro-flows and inertial forces are dismissed.

Particularly in the case of dealing with particle laden fluids, some of the standard

micro-mixing methods such as lamination techniques are not practical due to high

probability of clogging in narrow micro-channels. Therefore, development of a

proper technique is required in order to achieve an efficient mixing and also

prevent the clogging of the channel.

The primary object of this research is to design a micro-mixer for chaotic mixing

of magnetic particles into a bio-fluidic solution containing target cells in order to

enhance the attachment of cells to particles. Moreover, a numerical model of the

mixer will be developed to characterize the design. Two indices will be utilized;

firstly calculation of Lyapunov exponents which is a common definition for

inspection of induced chaotic extent in the system, which in turn evaluates the

quantity of mixing. Secondly, the ability of the system to capture target cells will

be evaluated as a supplemental index.

Affecting parameters on the performance of the system comprise structural (i.e.,

geometrical dimensions and type of the components) and operating parameters.

Based on preliminary results obtained from the developed model, reasonably

optimized structural parameters will be adopted for the design. Subsequently, the

effect of the variation of system operating parameters will be investigated against


the above mentioned indices in order to obtain the optimum practical range of

operating parameters for the mixer.

1.5. Contribution to knowledge

The following contributions to knowledge are claimed:

• Designing a microfluidic mixer for chaotic mixing of two separate

solutions; one laden with magnetic particles and the other containing target

cells. This mixer is a part of an integrated device which will be utilized for

extraction of specific molecules out of their native environment.

• Developing a multiphysics model of the mixer combining magnetic,

fluidic and particle dynamics models. This was done using the finite

element analysis package COMSOL. Outcome of the model is the particle

velocity fields which will be used for tracing the particles.

• Developing Matlab codes for calculation of Lyapunov exponents and

therefore, inspection of chaotic extent and also evaluation of the cell

capturing efficiency in the system.

• During the course of MPhil project, the research outcome and

achievements were presented through one journal paper, two conference

papers and two posters. In addition, a design patent application was filed

for the mixer developed in this research. Details are given in Appendix C.

1.6. Outline of the thesis

This thesis comprises seven chapters and two appendices. In the first section of

every chapter, there is an opening introduction on the subject of that chapter and a

technical explanation, followed by the main body.

In the first chapter, a brief description of Lab-on-a-Chip systems and concept of

magnetic isolation is presented. An overview of the research is added at the end of

this chapter.


Chapter 2 covers a comprehensive literature review on existing techniques of

micro-mixing. The review process is based on the reported mechanism of mixing

and includes two major divisions; passive and active types. Also main

characteristics of microfluidic regimes are introduced in this chapter.

Chapter 3 provides a brief description of technical background of magnetic

materials and fields. Underlying physics of super-paramagnetic micro-particles

and the origin of magnetophoretic forces are also addressed in details.

Chapter 4 describes the design of the micro-mixer and proposed mechanism to

create chaotic advection in the motion of the particles. Geometrical structure and

dimensions of the micro-channel and micro-conductors are introduced. An

estimation of generated magnetic field by micro-conductors and injected

magnetophoretic forces on the particles is presented.

Chapter 5 presents detailed simulation procedure of the system. Utilized software

packages and mathematical methods are explained. Simulation results for

different operating parameters are given subsequently, together with a through

discussion.

In chapter 6 the mixing process is evaluated by inspection of chaotic regimes.

Moreover, ability of the system to capture target cells, which is the main goal of

the design, is used as a supplemental index to characterize the device.

Finally in chapter 7, main conclusions are drawn and some suggestion for the

future research in order to further investigate the performance of the mixer and

optimize the design, are presented.

Chapter 2

Literature Survey

Microfluidics deals with the behaviour, precise control and manipulation of

micro-litre and nano-litre volumes of fluids. It is a multi-disciplinary field

comprising physics, chemistry, engineering and bio-technology, with practical

applications to the design of systems in which such small volumes of fluids will

be used. Ascribed to the micron dimensions, microfluidics has some special

characteristics such as high surface-to-volume ratio, high mass-heat transfer rate,

high shear-extension rate, and low Reynolds number. Therefore, in order to

understand the behaviour of micromixers, a reasonable knowledge of the theory of

microfluidics is necessary. In this chapter a brief introduction to microfluidics and

some of the key definitions is presented and, subsequently, a through and

comprehensive literature review on micro-mixing and different reported methods

is provided.

2.1. Microfluidics - basic concepts and definitions

2.1.1. Newtonian fluid

A fluid is called Newtonian when the shear stress induced by the viscosity of the

fluid is directly proportional to the strain gradient:

dudy

τ = μ (2.1)

The constant of proportionality µ, is the dynamic viscosity coefficient of the fluid.

Water, the fluids of interest in this research, is a Newtonian fluid.

Chapter 2 – Literature Survey 11

2.1.2. Flow regime

Laminar flow, also known as streamline flow, occurs when a fluid flows in

parallel streamlines, with no disturbance between the lines. In fluid dynamics,

laminar flow is a flow regime associated with high momentum diffusion, low

momentum convection, and velocity and pressure independence from time. On the

contrary, turbulence or turbulent flow is a flow regime characterized by chaotic,

stochastic property changes. This implies lower momentum diffusion, higher

momentum convection, and quick variations of velocity and pressure in time and

space. Viscous forces dominate in a laminar flow regime, while inertial forces

dominate in a turbulent flow regime.

2.1.3. Reynolds number

In fluid mechanics, the Reynolds number is a dimensionless parameter obtained

from dimensional analysis. It has an important physical meaning, since it is the

ratio of inertial forces (ρv) to viscous forces (µ/L). Reynolds number is used for

determining whether a flow will be laminar or turbulent and is defined as:

Intertial forces vL vLReViscous forces

ρ= = =

μ ν (2.2)

where:

• v is the characteristic velocity of the flow

• L is the characteristic length of the geometry

• ρ is the fluid density

• µ is the dynamic (absolute) fluid viscosity

• ν is the kinematic viscosity of the fluid, ν=µ/ρ

At sufficiently high Reynolds numbers, the flow becomes unstable and a turbulent

regime develops. However, at lower Reynolds numbers as that in micro-channels,

viscous forces dominate over inertial forces, and flow disturbance quickly gets

damped. In fact, in micro-scale turbulent phenomena are practically not


encountered. As will be discussed, such characteristics make the mixing process

difficult in microfluidic devices where the goal of the mixing is to distribute the

molecules by a random, non-reversible process. At the macroscopic scale where

inertial forces dominate, local instability in the form of turbulence can be easily

created to enhance the mixing [1]. On the contrary, when the Reynolds number is

small, any assay to induce turbulent disturbance is quickly damped by viscous

forces. Thus diffusion becomes the only viable process of the mixing for low

Reynolds number flow.

2.1.4. Incompressible flow

Certain fluids undergo very little change in density despite the existence of large

pressures. In such circumstances when density variation in a problem is

inconsequential, the fluid is called incompressible and the density is treated as a

constant value in computations. Water is an incompressible fluid and table 2.1

lists the main characteristics of water in standard conditions of pressure and

temperature.

Table 2.1 Main characteristics of water at 20 ºC and 1 atm.

Density ρ Dynamic viscosity

µ

Kinematic

viscosity ν

998 kg/m3 1.0 ×10-3 kg/(m.s) 1.01 ×10-6 m2/s

2.1.5. Navier-Stokes equations

The Navier-Stokes (N-S) equations are a set of fundamental differential equations

that explain the motion of the fluid substances such as liquids and gases. These

equations are derived from conservation principles (i.e., conservation of mass,

momentum and energy) and are the governing constitutive equations of

conventional flows. The vectorial form of the N-S equations for an incompressible

Newtonian flow is:


2d ( P )dt

ρ = ρ + −∇ + μ∇V B V (2.3)

where:

• V is the velocity

• B is the body force per unit mass

• P is the pressure

In Computational Fluid Dynamics (CFD), the N-S equations are the most

common equations to solve fluid dynamics problems with finite elements analysis

methods.

2.1.6. Steady flow

A flow is called steady when flow characteristics (e.g., velocity components) and

thermodynamic properties at each position in space are invariant with time.

Individual fluid particles may move, but at any particular position in domain, such

particle behaves just like as any other particle when it was at that point. There is

no time dependency in parameters for steady flow equations (d/dt=0).

2.1.7. No-slip condition

When a fluid flow is bounded by a solid surface, molecular interactions cause the

fluid in contact with the surface to seek momentum and energy equilibrium with

that surface. All liquids essentially are in equilibrium with the surface they contact

[2]. Then, all fluids at a point of contact with a solid take on the velocity of that

surface which means the fluid relative velocity at all liquid-solid boundaries is

zero (Vfluid=Vwall). In other words, the outermost molecule of a fluid sticks to

surfaces past which it flows. This is called the no-slip condition and serves as the

boundary condition for analysis of the fluid flow past a solid surface. The no-slip

condition gives rise to the velocity profiles of a flowing fluid as it is shown in the

following section.


2.1.8. Pressure-driven flow (Poiseuille flow)

Pressure-driven flow is commonly found in fluid handling systems, including

microfluidic devices. For a pressure-driven flow in rectangular cross-sectional

channels with no-slip boundary conditions, velocity profiles are parabolic. The N-

S equations with restriction of two-dimensional steady flow simplifies to:

2

2p ux y∂ ∂

= μ∂ ∂

(2.4)

The solution to this differential equation can be found with the boundary

conditions (the velocity is zero at the channel wall (y=d/2)):

= −μ

221 dp du(y) ( y )

2 dx 4 (2.5)

where u is the velocity, µ is the dynamic viscosity of the fluid and p is the

pressure. Figure 2.1 illustrates the developed parabolic velocity profile with the

highest velocity along central streamline.

Figure 2.1 Velocity profile of a steady pressure-driven laminar flow with no-slip boundary

conditions developed in a micro-channel.

2.2. Mixing in microfluidics

Mixing is a fundamental step in most of the microfluidic systems used in

biochemistry analysis where biological processes such as enzyme reactions often

engage reactions that require mixing of reactants. Mixing is also essential in LOC


platforms for tagging of specific entities by some labels such as magnetic particles

which are used for actuation. Micro-mixers can be integrated in a microfluidic

platform or utilized as a stand-alone device. However, mixing several fluids at the

micro-scale is not as easy as it might seem at first glance. As discussed earlier, the

Reynolds number at these dimensions is usually quite small and no turbulence

takes place. Therefore, flow streamlines do not interfere with each other which

results in zero mixing. Nevertheless, over small distances mixing can be

performed by diffusion phenomenon. Alternatively, mixing may be enhanced by

chaotic patterns, which can be induced by various schemes.

Micro-mixers can be generally categorized as passive and active mixers. In

passive micro-mixers where no external energy is required, the mixing process

can rely on diffusion or chaotic advection. Passive mixers can be further

categorized by their arrangement for the mixed phases such as lamination,

injection, chaotic advection and droplet. In active micro-mixers an external field is

used to generate disturbance to enhance the mixing process. Therefore, active

mixers can be categorized by their type of external sources such as pressure,

temperature, electrokinetics, and acoustics. Almost in all active mixers the basis

of mixing is the chaotic advection of the flows. In the following sections, a brief

introduction on the diffusion and chaotic advection phenomena is given and,

subsequently, the review considers various types of passive and active micro-

mixers.

2.2.1. Diffusion

Diffusion is the instinctive spreading of matter (particles), heat, or momentum and

represents one type of transport phenomenon. It is the movement of entities from

regions with higher chemical potential to lower chemical potential. One type of

diffusion is the molecular diffusion (Brownian motion) in which we are dealing

with transfer of the matter. Here, chemical potential can be interpreted as the

concentration of molecules or particles. In fact, Brownian motion is an entropy-

minimizing process occurring in the presence of a non-uniform distribution of


molecules. In microfluidic systems, the molecular diffusion is the dominant

mechanism of mixing of mass species unless some external perturbation is

applied. It is, however, mostly too slow and thus impractical in many cases,

especially for large molecules. Let us estimate the characteristic time of diffusion.

The reason for the diffusion is the large gradient of the concentration of the fluid

molecules (or suspended particles) which exists when two different liquids have a

common interface. The mathematical model of diffusion can be described by

Fick’s second law [3]:

= ⋅ = ⋅∇2dc D div grad c D cdt

(2.6)

where c is the concentration for a particular fluid molecule type and D is the

solute diffusion constant. For steady state diffusion (when the concentration

within the diffusion volume does not change with respect to time) the equation

(2.6) is reduced to Fick’s first law, which gives the flux of the diffusing species as

a function of the change in concentration in space (distance):

cJ Dx∂

= −∂

(2.7)

where J is the diffusive mass flux per unit of area (area perpendicular to x) and x

is the position. D, diffusion coefficient or diffusivity, is defined as:

κ= =

πμT driving potentialD

6 r resis tance (2.8)

where

• κ is the Boltzmann’s constant (=1.35054×10-23 [J/K])

• T is the absolute temperature of the fluid

• r is the molecular radius of the solute

• µ is the dynamic viscosity of the fluid

Temperature dependency of the diffusion coefficient is associated with this fact

that the Brownian motion of the particles is due to the applied forces from small


liquid molecules which are excited by the temperature. The average time for the

suspended entity to diffuse over a given distance is directly proportional to the

square of the distance:

τ ∝2L

D (2.9)

where L is the characteristic mixing length (e.g., channel width) and τ is the time

of mixing. τ can be up to the order of 105 seconds for particles with 1 μm diameter

dispersed in water solution diffusing a distance of 100 μm. Obviously, such a

diffusion time is not realistic and microfluidic devices that employ natural

diffusion as their sole mixing mechanism will not be able to satisfy the rapid-

mixing requirement in bio-chemical analyses. Therefore, an innovative method of

mixing is essential to enhance the process. As equation (2.9) suggests, the rate of

diffusion is dependent on diffusion coefficient, and the mixing length. Both

viscosity and diameter are intrinsic properties of the solution and the chosen

species, and thus the only remaining possibility of enhancing diffusion is to

increase the contact surface and decrease the diffusion path (see equation (2.7)).

2.2.2. Chaotic advection

In addition to diffusion, advection is another important form of mass transfer in

flows. Advection is normally parallel to the main flow direction, and is not

functional for the transversal mixing process. However, the so-called chaotic

advection can enhance the mixing in microfluidic devices significantly. Mixing in

these devices generally involves two steps; at first, a heterogeneous mixture of

homogeneous domains of the two fluids is created by advection and,

subsequently, diffusion between adjacent domains leads to a homogeneous

mixture at the molecular level [4]. In the context of micro-mixers, the question

arises on how the principle of chaotic advection can be implemented, as macro-

scale techniques such as employment of stirrers are not available. Chaotic

advection can generally be produced by special geometries and three-dimensional

structures in the mixing channel or induced by an external force in passive and


active micro-mixers, respectively. The basis of chaotic mixing will be addressed

in details later in chapter 4.

2.3. Passive micro-mixers

Because of their simple concept, passive mixers were one of the first microfluidic

devices reported. Here we review the passive mixers based on their arrangement

for the mixed phases.

2.3.1. Basic T-mixer and Y-mixer

As discussed earlier, fast diffusion mixing can be accomplished by decreasing the

mixing path and increasing the contact surface between two liquid phases.

Lamination separates the inlet streams into “n” sub-streams and then joins them

into one stream. The most simple design is a channel with merely two inlets (n =

2); known as the T-mixer or the Y-mixer [5,6]. Figures 2.2a and 2.2b illustrate the

design of a typical T-mixer and Y-mixer, respectively [7].

Figure 2.2 Basic designs in parallel lamination; (a) T-mixer, (b) Y-mixer.

Since the basic T-mixer depends solely on molecular diffusion, a long mixing

channel is required to accomplish the process. Nevertheless, efficient mixing may

be achieved in a short mixing length at the expense of increasing the Reynolds

number [8,9]. A chaotic regime can be induced at these high Reynolds numbers.

Wong et al [9] reported a T-mixer which utilizes Reynolds numbers up to 500,

where flow velocity is as high as 7.60 m/s at a pressure of up to 7 bar (figure 2.3).

However, in such micro-mixers, the high velocities on the order of 1 m/s or even


higher require high supply pressures. The high pressure may be a crucial

challenge for bonding and inter-connection techniques.

Figure 2.3 The observations of the mixing process at the junction of micro T-mixer at different

applied pressures: (a) 1.12 bar; (b) 1.88 bar; (c) 2.11 bar; (d) 2.48 bar; (e) 2.77 bar, (f) 4.27 bar [9].

The design of a T-mixer may be enhanced by roughening [10] or throttling [11]

the channel wall and entrance, respectively. At rather high Reynolds numbers the

basic T-mixer can be further modified by implementation of some obstacles in the

channel, which generate vortices and chaotic advection. These types are discussed

in section 2.3.6.

2.3.2. Passive micro-mixers based on multi-lamination (parallel lamination)

When the number of sub-streams is greater than 2, the concept of multi-

lamination is realized. Multi-laminating flow configurations can be realized by

different types of feed arrangements. As explained, lamination is based on the

concept of reducing the mixing path by making narrow channels [12-15]. Another

method to make narrow paths is implementation of interdigital structures in the

channel [16]. The flow is usually driven by pressure, but can also be generated by

electrokinetic forces [17-19]. Figure 2.4 illustrates the concept of multi-lamination

and three reported mixers based on this principle.


Figure 2.4 Multi-lamination; (a) concept, (b) principle of the lateral micro-mixer, (c) mixing of red

and green ink at p= 7.8 kPa [15], (d) photograph of a micro-mixer consisting of a mixing device

with an interdigital channel structure [16].

Vortex (cyclone) mixers are a further type of multi-laminating mixers where fast

vortices are generated to enhance mixing with multiple inlet streams focused in a

circular chamber [20-21]. In the work by Hardt et al [22], a numerical model was

used to analyse the streamline distributions of the three-dimensional vortex and to

predict the mixing performance of the micromixer. It was found that when the

Reynolds number is higher than a critical value of 2.32, a self-rotation effect is

induced in the circular micro-chamber, which in turn generates a three-

dimensional vortex. Respective flow patterns were confirmed by microscopy

analysis and resemble the prediction made by CFD analysis (figure 2.5).

Figure 2.5 Vortex micro-mixers; (a) concept of the mixer, (b) cross-section of vortex chamber

showing mixing process (green indicates complete mixing) [20].


An alternative concept to reduce the mixing path for multi-lamination micro-

mixers is hydrodynamic focusing. The basic design for hydrodynamic focusing is

a relatively long channel with three inlets. The middle inlet is dedicated to the

sample flow, while the solvent streams join through two encompassing inlets and

act as the sheath flows (figure 2.6a).

Hydrodynamic focusing technique was first developed to enable a fast mixing

process in less than one second. It reduces the stream width and, consequently, the

mixing path. Knight et al [23] reported a prototype with a narrow mixing channel

of 10 μm×10 μm in section. The sample fluid may be focused to a specific width

by adjusting the pressure ratio between the sample flow and the sheath flows. In

this way, diffusion distances are significantly reduced by compressing the fluid

layer to a few micrometers, resulting in a mixing in the milliseconds range [24].

Walker et al [25] reported a micro-mixer based on hydrodynamic focusing used

for cell infection (figure 2.6b).

Figure 2.6 Parallel lamination; (a) concept of the hydrodynamic focusing, (b) an image of the blue

food colouring stream in the middle and water streams on either side [25].

2.3.3. Passive micro-mixers based on Split-and-Recombine configurations

(serial lamination)

Split-and-Recombine (SAR) micro-mixers can improve the mixing by splitting

and later joining the streams, creating sequentially multi-laminating patterns

(figure 2.7a). For instance, the inlet streams may be first joined horizontally and

then in the next stage vertically. SAR mixing commonly relies on a multi-step

procedure. The basic operations are: splitting of a bi- or multi-layered stream

perpendicular to the main orientation into sub-streams, re-direction or re-


alignment of the sub-streams, and the recombination of these. These basic steps

are usually accompanied by one or more re-shaping steps [26]. After m splitting

and joining stages, 2m liquid layers can be laminated. The process leads to a 4m-1

times improvement in the mixing time [7].

Figure 2.7 SAR configuration; (a) concept of join-split-join [7], (b)-(d) cross-sectional views of

layer configurations within a SAR step: b) splitting; c) re-arrangement of the sub-streams and

recombination; d) reshaping [27].

Branebjerg et al. [28] and Schwesinger et al. [29] were among the first who

considered a micron-sized implementation of the SAR approach. Since then,

several kinds of micro-mixers have been realized utilizing some kind of multi-step

SAR approach. The designs of SAR mixers differ in the exact geometry by which

they actually achieve the fluidic arrangement. Ramp-like [28], fork-like [29], and

curved architectures [27] with and without splitting plane were reported. In

context with micro-technological applications, the SAR concept is especially

appealing, since it allows achieving fine multi-lamination with moderate pressure

drops and without severe fabrication constraints.

The concept of the SAR lamination micro-mixing may also be utilized for

electrokinetic flows [30]. Using electroosmosis flows between the multiple

intersecting channels, mixing was considerably enhanced (see figure 2.8). Figure

2.8c shows the flow configuration in the mixer. As it can be seen, diffusive

mixing is enhanced and convective mixing also takes place, which would not

occur in open channels. Melin et al [31] later reported a similar design for a


pressure-driven flow. However, this design works only for discrete liquid

samples.

Figure 2.8 SAR lamination; (a) concept, (b) fabricated mixer, (c) how mixing might occur [30].

2.3.4. Injection micro-mixers

The basis of the injection mixer is similar to the SAR lamination mixer. However,

instead of splitting both inlet flows, the mixer solely splits the solute flow into

many sub-streams and injects them into the solvent flow. On top of one stream is

an array of nozzles, which create a number of micro-plumes of the solute. These

plumes enlarge the contact surface and decrease the mixing path, thereby

improving the mixing efficiency [7].

Miyake et al [32] presented an injection micro-mixer with 400 nozzles arranged in

a square array. The mixer has an area for mixing, which is very flat and thin with

micro-nozzles provided at the bottom of the mixing chamber. First, the mixing

area is filled with one liquid, and the other liquid is injected into the area through

the micro-nozzles, making many micro-plumes. The nozzles are positioned very

closely in rows, 10-100 µm apart, in order that the plumes may quickly diffuse for

this distance. Thus, effective mixing will be performed without any additional

driving. Figure 2.9 shows the concept of the injection mixro-mixer. Similar

technique for the mixing with different nozzle shapes was reported by other

researchers [33-35].


Figure 2.9 Basic idea of the micro-injection mixer [32].

2.3.5. Droplet micro-mixers

An alternative method for reducing the mixing path is to form droplets of the

mixed liquids. The movement of a droplet may lead to an internal flow field and

make the mixing inside the droplet feasible. If mixing is achieved by droplet

movement only, this is passive mixing due to convection. Droplets may be

generated and manipulated individually using pressure [36] or capillary effects

such as thermo-capillary [37] and electro-wetting [38]. Moreover, droplets may be

generated by virtue of the large difference of surface forces in a narrow channel

with multiple immiscible phases such as oil-water or water-gas [39]. In this type,

by using a carrier liquid such as oil, droplets of the aqueous samples can be

formed. While moving through the channel, the shear force between the carrier

liquid and the sample accelerates the mixing process in the droplet (figure 2.10).

Figure 2.10 Mixing by winding microfluidic channels shown, (a) experimentally (left: a scheme of

the microfluidic network. right: photograph of plugs) and (b) schematically [39].


2.3.6. Passive micro-mixers based on chaotic advection

Chaos cannot occur in steady two-dimensional flows, but only in three-

dimensional and two-dimensional time-dependent flows. In two-dimensional

flows, time-dependency may be considered as an added third dimension. Time-

dependency may be induced by external forces, which is the principle of active

mixing class and will be dealt with in section 2.4. In passive micro-mixers the

basic idea is to modify the configuration and shape of the channel in a way that

leads to splitting, stretching, and folding of the flow. Here, we classify the passive

chaotic mixers based on the range of flow Reynolds number; high, intermediate

and low. However, it is not always possible to dedicate a particular design to a

specific range of Reynolds number.

2.3.6.1. Chaotic advection at high Reynolds numbers (Re>100)

A simple method is to insert obstacle structures in the mixing micro-channel in

order to induce the chaotic advection. Various configurations and arrangements

have been reported. Lin et al [40] used seven cylinders of 10 μm diameter placed

in a narrow channel (50 μm × 100 μm × 100 μm) to enhance mixing. The mixing

was performed with Reynolds numbers ranging from 200 to 2000 and a reaction

time of 50 μs. Wang et al [41] reported a mixer using the same type of obstacles

with different arrangements and carried out a numerical investigation of the

mixing at high Reynolds numbers. The mixing channel was 300 μm in width, 100

μm in depth and 1.2-2 mm in length, and the diameter of the obstacle was 60 μm

(see figures 2.11a and 2.11d). It was revealed that obstacles in a channel at low

Reynolds numbers cannot generate eddies or re-circulations. However, simulation

results showed that obstacles could enhance the mixing performance at high

Reynolds numbers. As shown in figures 2.11b and 2.11e, obstacles may also be

placed on the channel’s walls [7,42].


Figure 2.11 Various designs for producing chaotic advection at high Reynolds numbers; (a)&(d)

obstacles placed in the channel [41], (b)&(e) obstacles placed on the wall [42], (c)&(f) zigzag

channel [43].

An alternative method to generate chaotic advection is utilizing zigzag channels to

produce re-circulation around the turns in the channel. Mengeaud et al [43] used a

micro-channel with a width of 100 μm, a depth of 48 μm and a length of 2 mm

(see figures 2.11c and 2.11f). In conducting a numerical investigation, they

adopted the periodic steps of the zigzag shape as the main optimization parameter.

Reynolds number was varied ranging from 0.26 to 267 and a critical Reynolds

number of 80 was found. Below this number the mixing process relied entirely on

diffusion whereas as at higher Reynolds numbers, mixing was performed by the

generated re-circulations at the turns along the channel. The re-circulations could

induce a transversal component of the velocity, which enhances the mixing

process.

2.3.6.2. Chaotic advection at intermediate Reynolds numbers (10<Re<100)

Most of the micro-mixers in this category are based on the modified three-

dimensional twisted channels, but there may be some exceptions as well. For

instance, Hong et al [44] presented an in-plane micro-mixer with two-dimensional

modified Tesla structures (figures 2.12a to 2.12d). The Coanda effect in this


structure leads to chaotic advection and enhances mixing noticeably. The mixer

performs well at Reynolds numbers higher than 5.

Liu et al [45] reported a three-dimensional serpentine mixing channel comprised

of a series of C-shaped segments placed in perpendicular planes (figures 2.12e to

2.12g). The micro-mixer has two inlet channels joined in a T-junction and a

sequence of six mixing segments. It was observed that the mixer is that the mixing

time is short at higher Reynolds numbers; chaotic advection only occurred at

Reynolds numbers ranging from 25 to 70.

Chen and Meiners [46] reported a complex mixing unit which consists of two

connected out-of-plane L-shapes (figures 2.12h to 2.12j). A single mixing unit

measures about 400 μm × 300 μm and the mixer is composed of a series of such

mixing units. Effective mixing could be obtained on short length scales with a

purely laminar flow with the Reynolds number of Re= 0.1-2. This concept was

called as ‘flow-folding topological structure’ by the authors.

Park et al [47] presented the results for mixing two fluids in a three-dimensional

passive rotation micro-mixers using the break-up process (figures 2.12k to

2.12m). The complex channel rotates and separates the two fluids by partitioning

walls, and consequently, generates smaller blobs exponentially. In practical

experiments, over 70% mixing was achieved at Re=1, 10 and 50, only after

passing through a 4 mm long channel.

Vijayendran et al [48] reported a three-dimensional serpentine mixing channel

where the channel was designed as a series of L-shaped segments in perpendicular

planes (see figure 2.12n). The mixer was experimentally tested at Reynolds

numbers of 1, 5 and 20. The results indicated that better mixing was achieved at

higher Reynolds numbers. Jen et al [49] proposed various designs of twisted

micro-channel providing a third degree of freedom for chaotic advection. Mixing

of methanol and oxygen was numerically investigated at different velocities (0.5-

2.5 m/s). Figures 2.12o to 2.12q illustrate the concept of the twisted channels.


Figure 2.12 Micro-mixer designs for mixing with chaotic advection at intermediate Reynolds

numbers: (a) modified Tesla structure, (b)-(d) experimental results for the Tesla structure, (e) C-

shape concept, (f)-(g) photographs of side-by-side experiments at Reynolds numbers of 12 and 70,

respectively, (h) connected out-of-plane L-shapes, (i) topologic structure, (j) mixing of two

fluorescently labelled protein solutions in a six-stage mixer at Reynolds number of 0.1. (top).

mixing of the same dyes in an aqueous 54% glycerol solution with ten fold higher viscosity at

Reynolds number of 0.1 (bottom), (k) twisted micro-channel concept, (l) schematic diagrams of

twisted mixer segment and flows in the channel re-circulating along the walls, (m) images of

mixer obtained using a confocal scanning microscope; one fluid is propagating into the other fluid

along the 4 mm long channel, (n) L-shape concept, (o)-(q) other designs of the twisted channel.


2.3.6.3. Chaotic advection at low Reynolds numbers (Re<10)

One of the most promising types of the passive micro-mixers falls in this category

which works based on the idea of placing micro-structured objects within the flow

passage on one side of the channels. Johnson et al [50] and Stroock et al [51] were

the first to investigate this concept and since then, much effort has been dedicated

to improve their proposed mixers. In Johnson et al [50], a T-mixer was modified

with a pulsed UV excimer laser to ablate a series of slanted wells at the junction

(figure 2.13a). This structure allowed inducing a high degree of lateral transport

across the channel in either electroosmotic or pressure-driven flows. The

performance of the slanted-well design was evaluated over a range of effective

electroosmotic flow rates. The captured fluorescence microscopy images for two

flow rates are shown in figures 2.13c and 2.13d for slanted-well design (left) and

T-channel without the ablated wells (right).

Stroock and his colleagues [51-53] pointed out another way of creating secondary

re-circulating flows in a channel. They considered geometries with grooved

channel walls, such that at least one of the walls contains ridges standing at a

tilted angle with the main flow direction. Two different groove patterns were

considered; obliquely oriented and staggered ridges. Corresponding schematic

designs are shown in figures 2.13e and 2.13i, respectively. They referred to later

one as the staggered herringbone mixer (SHM).

One way to induce a chaotic pattern is to subject volumes of fluid to a repeated

sequence of rotational and extensional local flows. This sequence of local flows in

the SHM may be obtained by varying the shape of the grooves as a function of

axial position in the channel: The alteration in the orientation of the herringbones

between half cycles exchanges the positions of the centres of rotation and the up-

and down-wellings in the transverse flow. When a pressure-driven fluid flows

over such a surface, the grooves can be viewed as if they induce a slip flow in a

particular direction. Confined to a channel, the flow develops re-circulation

patterns, which leads to an exponential increase of specific interface, therefore to


fast mixing. The SHM mixing is superior to similar channels without inserted

structures or with straight ridges only. SHM can work well at a Reynolds numbers

ranging from 1 to 100. Confocal micrographs of vertical cross-sections of the

channel in both designs are shown in figures 2.13h and 2.13k.

The effect of chaotic advection in a channel with grooves was numerically

investigated by Wang et al [54] and Schonfeld and Hardt [55] for channels with

straight ridges and Aubin et al. [56] for both patterns using CFD methods. They

showed that an exponential stretching of the fluid interface occurs where with

simple linear grooves (straight ridges), the interface area increases more slowly.

Also ablation of grooves on the PDMS substrate by a laser was investigated by

Lim et al [57].

Kim et al [58] improved the design of straight ridges with embedded barriers

parallel to the flow direction. The embedded barrier changes the original elliptic

mixing pattern (developed in SHM) to a hyperbolic pattern as shown in figure

2.14a. Cross-sectional confocal microscope images of the mixing patterns at

several positions are shown in figure 2.14b.

A miniaturized version of a conventional large-scale static mixer with helical

elements was presented by Bertsch et al [59]. This concept modifies the three-

dimensional inner surface of a cylindrical mixing channel. Two kinds of

geometries were studied. The first type was composed of a series of stationary

rigid elements that form intersecting channels to split, rearrange and combine

component streams. The second type is comprised of a series of short helix

elements arranged in pairs, each pair composed of a right-handed and left-handed

element arranged alternately in a pipe (figure 2.14c). To numerically characterize

the efficiency of the mixer, they injected 65000 evenly distributed particles in the

half section of the pipe just before the mixer inlet. Figure 2.14d shows the location

of the particles at regularly spaced axial locations along both types of the mixers.

Moreover, experimental results revealed that the mixing efficiency of the mixer

made of intersecting channels is better than the mixer made of helical elements.


Figure 2.13 Micro-mixer designs for mixing with chaotic advection at low Reynolds numbers: (a)

concept of slanted ribs, (b)-(d) fluorescence images of experiments with and without ablated wells,

(e) slanted grooves, (f) schematic diagram of the channel with ridges, (g) optical micrograph

showing a top view of a red stream and a green stream flowing on either side of a clear stream in

the channel, (h) fluorescent confocal micrographs of vertical cross-sections of the channel, (i)

staggered herringbone grooves (SHM), (j) schematic diagram of one-and-a-half cycles of the

SHM, (k) confocal micrographs of vertical cross-sections of the channel.


Figure 2.14 Micro-mixer designs for mixing with chaotic advection at low Reynolds numbers: (a)

schematic view of the barrier embedded micro-mixer, (b) cross-sectional confocal microscope

images of the mixing pattern, (c) cut-of view of the micro-mixer structures made of intersecting

channels and helical elements, (d) plots of the locations of particles at regularly spaced axial

locations along the micro-mixers (from left to right); left plot corresponds to the beginning of the

mixers. The top line corresponds to the mixer made of intersecting channels, whereas the bottom

line corresponds to the mixer made of helical elements.

2.4. Active micro-mixers

As discussed earlier, in active micro-mixers an external field is used to generate

disturbance to enhance the mixing process. Most of the active mixers rely on the

chaotic regime induced by virtue of the induced periodic perturbation. In the

following, various active mixers classified by the type of employed external

sources are presented.


2.4.1. Micro-impellers

Traditionally, stirring with impellers is the most common way to perform mixing

of large volumes. However, several miniaturised stirrers have been developed for

mixing of the liquids in micro-scale [60-62]. In macroscopic stirrers, the stir-bar

or propeller rotation causes turbulence by increasing the local velocity. In micro-

scale, the stir-bar helps mixing by providing more interfacial area rather than

inducing turbulence. Claimed advantages of such mixers are the possibility to

match the impeller diameter to the mixing volume, carry out large-area mixing,

undergo mixing on-demand (switch on/off), and the flexibility of the mixing

approach regarding the choice of liquids [63]. A micro-stir-bar with a span of 400

µm was fabricated and placed at the interface between two liquids in a PDMS

channel by Ryu et al [61]. An external magnetic field provided by a rotating

magnet in a hotplate/stirrer drives the stirrer remotely (figures 2.15a and 2.15b).

Experimental results proved that nearly complete mixing is achieved instantly.

Figure 2.15c shows the sequential shots of the mixing operation of a micro stir-bar

at different time lapses.

Figure 2.15 Activation by moving parts: (a) schematic view, (b) fabricated magnetic stir-bar, (c)

sequential shots of the mixing operation of a micro stir-bar at different time lapses [61].

2.4.2. Pressure field disturbance

Pressure disturbance was one of the earliest methods used in active micro-mixers.

Deshmukh et al [64] presented a T-mixer with pressure disturbance where an


integrated micropump drives and stops the flow in the channel to divide the mixed

liquids into multiple serial segments and make the mixing process independent of

convection. The performance of this mixer was later evaluated and mixing was

found to proceed quickly in the mixing channel [65]. Figure 2.16 shows the

mixing process at different stages. The pressure disturbance may also be

generated using external embedded micropumps [66].

Figure 2.16 Active micro-mixer based on pressure field disturbance: (a) schematic view, (b) steady

flow, (c) the top stream is stopped for 1/6 seconds, (d) the bottom stream is stopped, (e) mixed

fluid after many pulses [64].

Another method to achieve pressure disturbance is the generation of pulsing

velocity by alternating switches of the flows from a high to a low flow rate,

periodically. In this way, a pulsation of the whole stream is achieved promoting

axial mixing. Glasgow and Aubry [67] reported a simple T-mixer and detailed

CFD simulations with a pulsed side flow at a small Reynolds number of about 0.3.

When both inlets have constant flow rates, the mixing zone is confined to a

narrow band around the horizontal interface (figure2.17a). Time pulsing of one

inlet flow rate distorts the interface to an asymmetrically curved shape which

changes with time. Therefore, liquid transport is promoted and mixing is

improved (figure 2.17b). The degree of mixing was 22%, being 79% larger than

for constant flows. The periodicity and the number of pulsing streams have a

significant effect on the mixing efficiency. The best results were obtained for two

pulsed inlet flows having a phase difference of 180º with the same amplitude and

frequency. CFD simulations showed the bending of the fluid interface along the


channel cross-section and associated stretching and folding in the direction of the

flow. The corresponding degree of mixing was considerably increased to 59%.

Figures 2.17c and 2.17d show the simulation results for two pulsed inlet flows

with phase differences of 90º and 180º, respectively.

Figure 2.17 Numerical simulation results, (a) constant mean velocity in both inlets, (b) , pulsed

flow from the perpendicular inlet (c) and (d) two inlet flows pulsed at a 90 and 180 degree phase

difference, respectively [67].


Same concept was extended to the multiple pulsing injection of flows into one

channel, thereby generating chaotic advection [68]. However, such devices

require a complex computer controlled source-sink system. A further modelling

work on pressure disturbance was reported by Okkels and Tabeling [69].

2.4.3. Acoustic/ultrasonic disturbance

Acoustic (ultrasonic) actuation may be utilized to stir the fluids in active micro-

mixers [70-72]. However, ultrasonic mixing may be a challenging issue in

applications for biological analysis owing to the temperature rise due to acoustic

energy. Many biological fluids are sensitive to high temperatures. Moreover,

ultrasonic waves around 50 kHz are harmful to biological samples by virtue of the

possible cavitations. The non-destructive ultrasonic mixer reported by Yasuda

[73] used loosely focused acoustic waves to induce stirring movements where the

wave was generated by a piezoelectric zinc oxide thin film (figure 2.18a). The

actuator was driven by a programmable function generator providing a 500

kHz/3.5 MHz sine waves and programmed waveforms corresponding to the

thickness-mode resonance of the piezoelectric film. The mixer performed without

any consequential temperature increase and could be used for fluids sensitive to

the temperature. Figures 2.18b to 2.18d show the mixing of red blood cells by

ultrasound irradiation.

Figure 2.18 Schematic illustration of the micro-chamber for mixing of red blood cells by

ultrasound irradiation: (a) longitudinal section (b) area A in the chamber, (c) micrograph before

ultrasound irradiation, (d) micrograph during ultrasound irradiation [73].


An air bubble in a liquid can perform as an actuator, when it is energised by an

acoustic field. The bubble surface behaves like a vibrating membrane and this

type of actuation is mainly dependent on the bubble resonance characteristics.

Bubble vibration due to a sound field generates friction forces at the air/liquid

interface which leads to a bulk fluid flow around the air bubble (known as

cavitation or acoustic micro-streaming). Liu et al [75-75] used acoustic streaming

around an air bubble for mixing where streaming was induced by the field

generated by an integrated PZT actuator. Fluidic movements led to the global

convection flows with “Tornado” pattern in the vicinity of the bubbles. The time

required to fully mix the whole chamber was approximately 45 s. Figure 2.19

shows snapshots of multi-bubble induced acoustic mixing in a chamber at

different time stages. Further acoustic devices for mixing water and ethanol [76]

as well as water and uranine [77] were reported. Yaralioglu et al [78] also used

acoustic streaming to perturb the flow in a conventional Y-mixer.

Figure 2.19 Snapshots showing multi-bubble induced acoustic mixing in a chamber at time (a) 0 s;

(b) 28 s; (c) 1 min 7 s; (d) 1 min 46 s. [74].

2.4.4. Thermal disturbance

According to equation (2.8), diffusion coefficient is highly dependent on

temperature. Therefore, thermal energy may also be utilized to enhance the

mixing. Mao et al [79] generated a linear temperature gradient across a number of


parallel channels in order to examine the temperature dependence of fluorescent

dyes. Also a micro-mixer with a gas bubble filter activated by a thermal bubble

actuated micropump was successfully demonstrated by Tsai and Lin [80]. The

generated oscillatory flow could induce disturbance and wavy interface to

increase the contact area of fluids and accelerate the mixing process.

2.4.5. Electrokinetic disturbance

Electrokinetics is the study of the motion of bulk fluids or selected particles

embedded in fluids when they are subjected to electric or magnetic fields.

Electrokinetic forces can be utilized to manipulate liquid and/or particles in

micro-mixers as an alternative to pressure-driven flow. In the following, various

active mixers classified regarding the employed electrokinetic forces are

presented.

2.4.5.1. Electro-hydrodynamic (EHD) disturbance

Electro-hydrodynamic effect has been used to generate chaotic flows in micro-

mixers [81-82]. A simple geometry mixer was proposed, which works based on

the EHD force when the fluids to be mixed have different electrical properties and

are subjected to an electric field [81]. The electrodes are arranged so that the

electric field is perpendicular to the interface between the two fluids, creating a

transversal flow. Figure 2.20a illustrates the concept of the EHD mixer.

Figure 2.20 Active micro-mixer based on the EHD disturbance: (a) schematic view of the mixer,

(b)-(d) visualization of the flow.


Two fluids of identical viscosity and density, but with different electrical

conductivities and permittivities were used for experiments. Each fluid enters the

microfluidic chamber in its own inlet channel. As soon as they meet, a jump in

electrical conductivity and/or permittivity is generated at the interface between the

two fluids, which has no effect as long as the electric field is absent. However, as

the fluids enter the electric field influence zone close to a pair of facing electrodes,

they are subjected to an electrical force, which creates a transversal secondary

flow across the interface between the two fluids, therefore destabilizing the

interface and enhancing the mixing process. By alternating the voltage and

frequency on the electrodes, efficient mixing was obtained in less than 0.1s at a

low Reynolds number of 0.02. Figures 2.20b to 2.20d show the photographs of the

experiment.

2.4.5.2. Magneto-hydrodynamic (MHD) disturbance

The magneto-hydrodynamic force has been utilized in an active micro-mixers

reported by Bau et al [83]. This mixer uses the arrays of electrodes deposited on a

conduit’s wall as shown in figure 2.21a. By applying alternating potential

differences across pairs of electrodes, currents are induced in various directions in

the solution. In the presence of a magnetic field, the coupling between the

magnetic and electric fields induces body (Lorentz) forces in the fluid which in

turn produce mixing movement in the chamber. The Lorentz force can roll and

fold the liquids in a mixing chamber.

Figures 2.21b to 2.21f illustrate the deformation of a line of dye resulting from the

application of the Lorentz forces. After each time unit (a few seconds), the

polarity of the electrodes and the direction of the Lorentz force are reverse and the

dye returns to its previous initial position. After several reversals, dye continues to

deform in opposite directions and eddies are formed. These concepts work only

with an electrolyte solution. Since the electrodes can be patterned in various ways;

relatively complex flow fields can be generated.


Figure 2.21 Active micro-mixer based on the MHD disturbance: (a) schematic (top-view)

depiction of the mixer, (b) top-view of the fabricated mixer at beginning of the experiment when a

thin line of dye is laid across the cavity, (c)-(f) deformation of the dye line.

2.4.5.3. Electroosmotic disturbance

Lin et al [84-85] reported a T-form micro-mixer using alternatively switching

electroosmotic flow. A switching DC field is utilized to generate an

electroosmotic force which concurrently drives and mixes the electrolytic fluid

samples (figure 2.22a). It was shown that a mixing performance as high as 97%

can be obtained within a mixing distance of 1 mm downstream from the T-

junction when a 6 kV/m driving voltage and a 2 Hz switching frequency are

applied. Figure 2.22b presents the flow contours for optimized operating

conditions in the cases of low and high driving voltages.

Design and fabrication of a ring electroosmotic chaotic micro-mixer with

integrated electrodes was reported by Zhang et al [86] and numerical investigation

of the same mixer was later carried out by Chen et al [87]. Figure 2.22c shows the

SEM picture of the mixer. It takes two fluids from different inlets and combines

them into a single channel where the fluids enter the central loop in downstream.

Four microelectrodes are positioned on the outer wall of the central loop with an

angular distance of 45º. These microelectrodes impose a spatially varying electric

field, and the fluids are manipulated via the electroosmotic slip boundary

condition before they enter the outlet channel. Electric potentials on the


microelectrodes are time-dependent, which adds the third dimension necessary for

chaotic mixing. Generated electroosmosis agitates the low Reynolds number flow.

Figure 2.22d shows the streamlines at t=25 s obtained from simulations and figure

2.22e illustrates the induced stretching and folding of a small volume of fluid. Red

and blue curves are particle trajectories starting from the upper and lower half of

the inlet, respectively.

Figure 2.22 Active micro-mixer based on switching electroosmotic disturbance: (a) schematic

representation of the mixer, (b) flow contours for two optimized operating conditions at low and

high driving voltages, (c) SEM picture of the ring micro-mixer, (d) streamlines at t=25 s, (e)

stretching and folding of a small volume of fluid.

Sasaki et al [88] presented a mixer based on AC electroosmotic flow, which is

induced by applying an AC voltage to a pair of coplanar meandering electrodes

configured in parallel to the channel. The mixing time was 0.18 s, which was 20-

fold faster than that of diffusional mixing without an additional mixing

mechanism.

Tang et al [89] also utilized an electroosmotic flow to improve mixing where

switching on or off the voltage supplied to the flow generates fluid segments in

the mixing channel. This flow modulation scheme was capable of injecting

reproducible and stable fluid segments into microchannels at a frequency between

0.01 Hz and 1 Hz.


A mixer based on periodical field-effect control to dynamically manipulate local

flow field in the micro-channel was demonstrated by Wu and Liu [90]. The

proposed mixing mechanism combines temporal modulation (periodical out-of-

phase AC radial voltage control) with spatial modulation (asymmetric

herringbone-electrode feature) on the ζ-potential of the channel walls to induce

complex flow field for mixing enhancement (figure 2.23a). Numerical and

experimental results showed that good mixing efficiency of over 90% can be

achieved within a 5 mm long micro-channel (figures 2.23b and 2.23c).

Figure 2.23 Electroosmotic micro-mixer: (a) T-shape mixer with embedded electrodes, (b)

simulation results for the transverse velocity vectors at different cross-sections (y–z planes) along

the channel, (c) photograph of an electroosmotic flow with uranine dye.

In another case, oscillating electroosmotic flow in a mixing channel/chamber is

caused by an AC voltage [91]. The pressure-driven flow becomes unstable in a

mixing channel and the rapid stretching and folding of material lines associated

with this instability can be used to stir fluid streams with Reynolds numbers of

order of unity. Figure 2.24 shows schematic concepts of the mixers and also time-

stamped images showing an initially stable interface and its development after the

onset of the instability in both channel and chamber concepts.


Figure 2.24 Active micro-mixer based on electroosmotic disturbance: (a)-(d) schematic view of

mixing channel and chamber, (e) & (f) time-stamped images obtained from mixer with channel

and chamber configurations, respectively.

2.4.5.4. Dielectrophoretic disturbance

When a polarizable particle is exposed to an electric field, a dipole is induced in

the particle. If the electric field is non-uniform, the particle experiences a force

that can move it towards the high or low-electric field region, depending on the

particle polarizability compared with the surrounding medium. This phenomenon

is known as dielectrophoresis (DEP). If the polarizability of the particle is higher

than the solution, the force is towards the high field strength region (positive

DEP). Otherwise, the force is towards the lower field region (negative DEP) [92].


DEP has been utilized in active chaotic micro-mixers by Deval et al [93] and Lee

et al [94]. Figure 2.25a and 2.25b show a schematic view and the fabricated

mixer, respectively. Chaotic advection was generated by embedded polystyrene

particles with a combination of electrical actuation and local geometry channel

variation. Figure 2.25c shows the evolution of an interface as it advects through

the chamber. Where the electric field is constantly set to zero, the interface

remains flat as it travels across the chamber. However, when it is periodically

switched on and off, stretching and folding can take place, resulting in a

favourable situation for mixing. The yellow line indicates the evolution of the

interface between particle solution (lower part) and DI water (upper part).

Figure 2.25 Active micro-mixer based on dielectrophoretic disturbance: (a) schematic view, (b)

fabricated mixer, (c) stretching and folding as dielectrophoretic force is applied.

2.4.5.5. Magnetophoretic disturbance

The magnetic field-induced migration of particles in liquids is known as

Magnetophoresis which is dealt with in details in next chapter. Recently, in

addition to separation, magnetophoretic forces are exploited to enhance the

mixing of the particles in a solution in micro-scale devices.

A magnetic force driven chaotic micro-mixer was reported, in which magnetic

particles are stirred by the local time-dependent magnetic field to enhance the

attachment of magnetic particles onto biological molecules suspended in the

medium [95-98]. A serpentine channel geometry with the perpendicular electrodes


arrangement was used to create the stretching and folding of material lines. It is

claimed that good mixing was achieved in a short time (convective time of less

than 10 s) and distance (mixer length of 1.3 mm). Figure 2.26a shows the

fabricated magnetic mixer with a serpentine shaped channel. Magnetic particles

do not mix without an external disturbance. In figure 2.26b, particles are dispersed

over the entire channel at downstream due to magnetophoretic perturbation.

Figure 2.26 Active micro-mixer based on magnetophoretic disturbance: (a) fabricated mixer, (b)

dispersed particles over the entire channel at downstream.

Another micro-mixer is presented by Rong et al. [99] using magnetic micro-tips

for active mixing of magnetic particles or bio-cells. Mixing is achieved by a

combined rotational/vibrational force exerted on the particles as the magnetic tips

are sequentially excited to produce a rotating magnetic field. Mixer is driven by

three magnetic pole pairs excited with electromagnets coupled to magnetic pole

tips using through-hole vias as shown in Figure 2.27a and 2.27b.

Before excitation, particles are randomly distributed in the centre junction region

(figure 2.27c). When a sequential driving signal is applied, the magnetic particles

move around in the junction region of the channels with the applied magnetic

field. Figure 2.27d shows the mixing action before and after applying drive signal

to the three pair tips. Two liquids with and without magnetic particles are

introduced via separate channels. In the absence of excitation, two fluids will flow

separately according to laminar flow theory. If sequential actuation is applied to

the pole tips, particles in the junction region will be agitated by both rotating and

vibrating motion, which will then produce a rapid mixing action. However,

manufacturing the proposed mixer requires the utilization of complex micro-

fabrication techniques.


Figure 2.27 Mixer based on magnetophoretic disturbance: (a) & (b) schematic view and fabricated

mixer, repectively, (c) working principle, (d) mixing action before and after applying derive signal.

In another report, active fluid mixing was demonstrated in micro-channels where

mixing was based on the manipulation by a local alternating magnetic field of

self-assembled porous structures of magnetic micro-particles that are placed over

the section of the channel [100]. The mixing is the result of the chaotic splitting of

the fluid streams by the structures. In fact, they have followed the approach of

placing obstacles in the micro-channel in order to create stirring (convective)

effects by forcing one fluid stream into another. Another factor is the possibility to

induce a rotational motion of the magnetic micro-particles by using an AC

magnetic field.

Figure 2.28 is a schematic diagram of the realized three-dimensional and

monolithic microfluidic chip with embedded permalloy parts. Magnetic field is

generated by an external electromagnet, brought in mechanical contact with the

permalloy parts. When placed in the field, particles start interacting by means of

the magnetic dipole interaction. This interaction induces a spontaneous clustering

of the particles into larger structures. Using a sinusoidally varying magnetic field

(1 Hz<f<100 Hz), a rotational motion of the particles was induced, thereby

enhancing the fluid perfusion by the structure that behaved as a dynamic random

porous medium.


Figure 2.28 Schematic diagram of the realized three-dimensional microfluidic mixer with

embedded permalloy parts [100-101].

Fig. 2.29 shows the experimental results where according to the authors, a 70%

(static field) and 95% (AC field) mixing efficiency over a channel length as small

as the channel width (200 µm) and at velocity of 5×10-3 m/s was obtained.

However, the most important drawback of the mentioned system is its complexity

from the fabrication point of view.

Figure 2.29 Experimental results [101].


2.5. Discussion

Each of the investigated mixers has its own specific advantages and drawbacks

and there is not any particular type as the best general candidate for the mixing

process in micro-scale. Therefore, one must decide on an appropriate mixer type

considering various parameters such as desired functionality, fabrication costs,

disposability, and operating conditions.

2.5.1. Fabrication

Generally speaking, passive micro-mixers are more preferable as no external

source is required to drive these devices. Integrating actuation mechanisms such

as heaters, micro-conductors, power generators and controllers to provide the

required external energy in active mixers, calls for employment of sophisticated

fabrication techniques, which in turn adds an extra cost to the manufacturing

process. This may be a challenging issue particularly for disposable devices.

However, there are some exceptions in passive mixers where fabrication of micro-

channels with three-dimensional configurations such as Tesla structure, staggered

herringbone parts and obstacles is as complex as active mixers. Perhaps, most

convenient mixers from fabrication point of view are passive mixers, which rely

on lamination techniques and no complex structure or component is required to

operate them.

2.5.2. Performance

Performance of the micro-mixer can be a crucial factor in determining the proper

type of mixing mechanism for a particular application. Extent of the mixing of

micro-particles in bio-fluid, for instance, has a significant effect on the quality of

whole magnetic isolation process. Therefore, a mixing technique with sufficient

capability must be adopted for this protocol. Efficiency may also be interpreted as

the mixing time or the space required (e.g., channel length) to achieve the full

extent of the mixing as in most of the integrated systems, a considerable effort is


dedicated to minimizing these factors. In fact, one often needs to reach a

compromise between different parameters regarded as the efficiency of the mixer.

Moreover, controllability of the mixer must be factored in. While active mixers

can be activated on-demand (switch on/ff), in a passive mixer there is not any

chance to operate the device in particular ranges of time or space.

2.5.3. Application

Micro-mixers are widely used in chemical, biological and medical analysis

applications where one deals with variety of fluidic environments. Each type of

fluids has its own intrinsic properties such as viscosity, density, electrical

properties, etc. Therefore, based on the working fluid, a proper type of the mixing

technique must be adopted as some of the mixers are designed to work with

particular liquids. For instance, in most active mixers where the driving force is

electrokinetic, the possibilities for two mixing phases are limited; MHD mixers

work solely with electrolyte solutions, in EHD mixers two fluids are expected to

have distinct different electrical properties such as conductivity and permittivity,

electroosmotic mixers are highly dependent upon pH and the concentration of the

different ion species in the solution, and finally in dielectrophoretic and

magnetophoretic mixers, presence of some polarisable elements in mixing phases

is essential.

On the other hand, another major limiting factor for mixing phases must be taken

into account for almost all passive mixers, which rely on lamination methods; if a

particle laden fluid is passed through narrow channels the probability of clogging

is very high. Moreover, in those mixers where embedded conductors are utilized

to supply necessary electric or magnetic field for actuation, heat generation can be

a challenging issue for buffers sensitive to high temperatures. The same problem

is observed in acoustic micro-mixers.


In addition to the type of mixing liquids, operating conditions such as pressure

and bulk fluid velocity (Reynolds number) may be a crucial parameter in choosing

the suitable micro-mixing mechanism. For instance, as discussed earlier, a passive

micro-mixer with inserted obstacles which relies on the chaotic advection is not

an appropriate candidate for mixing of flows with low velocities.

Having considered the properties of buffer containing magnetic particles and the

presence of particles themselves, in this research it was decided to employ

magnetophoretic forces to perform the mixing as the same type of force is used

for separation stage. Besides this view to ultimately integrate the mixer to the

magnetic isolation chip as its particular application, it was intended to propose a

mixer with flexibility of the mixing approach regarding the choice of liquids.

Magnetic particles can be loaded into most fluids and be utilized as a label for

actuation. After the mixing, particles can be easily separated in downstream.

Chapter 3

Magnetophoresis

3.1. Introduction

Migration of magnetic particles in a fluid due to an inhomogeneous magnetic field

is known as magnetophoresis, which is the magnetic analogue of

dielectrophoresis. Magnetophoresis finds its application in separation and mixing

processes where some entity of interest is either magnetic itself or attached to, for

example, a magnetic bead. Since most materials show negligible para or

diamagnetism, it is usually necessary to introduce a magnetic ‘handle’ or ‘label’

such as the mentioned magnetic beads. In the following sections, general

principles of the magnetophoresis will be addressed.

3.2. Magnetic field and magnetic materials

In order to investigate the behaviour of magnetic particles under the influence of

magnetic forces, it is essential to know the theory of magnetic field and magnetic

materials, which is the origin of desired forces for manipulation of particles. This

section reviews the magnetic field theory in brief. Further details can be found in

one of many textbooks on magnetism (e.g. [1-3]).

A magnetic field intensity H is produced whenever there is electrical charge in

motion (e.g., an electrical current flowing in a conductor). Magnetic field intensity

is measured in the unit of amperes per meter [A/m]. In free space, magnetic flux

density B in tesla [T] is a linear function of H and we can write:

Chapter 3 – Magnetophoresis 52

0= μB H (3.1)

where constant µ0 (=4π×10-7 [H/m]) is the permeability of free space. In other

media B is no longer a linear function of H. Nevertheless, they are still related by

the permeability of the medium which is not necessarily a constant.

When a specimen is placed in a magnetic field arising from an external source,

there is a field inside the specimen and its atoms or molecules are magnetized.

The field B is made up of two contributions. One is the original field B0 present

when the specimen was absent (here it is assumed that the external field will

remain unaffected by magnetization of the specimen) and other contribution is the

field Bm owing to the magnetization of the specimen. The total field B is the sum

of the fields from two sources:

00 mB = B + B = (H+ M)μ (3.2)

Magnetization M is defined as the magnetic moment per unit volume and m is the

magnetic moment on a volume V of the material:

VmM = (3.3)

All materials are magnetic to some extent, with their response depending on their

atomic structure and temperature [4]. They may be conveniently classified in

terms of their volumetric magnetic susceptibility, χ, where

= χM H (3.4)

describes the magnetization induced in a material by H. By combining equations

(3.2) and (3.4) we can write:


0 0 r o rμ χ μ (1+ χ μ μ μ 1+ χ= + = = =B (H H) )H H , (3.5)

where µr is relative permeability and is related to susceptibility of the material.

The difference between B and B0 in equation (3.2) is dependent on the magnetized

material. Most of the magnetic materials can be classified into four categories

which are discussed hereafter.

3.2.1. Diamagnetic Materials

When a magnetic field is applied to a piece of material the orbital motion of the

electrons in the atoms will be affected, and a very small magnetic moment is

induced, which is opposite in direction and proportional to the applied magnetic

field. This phenomenon is termed diamagnetism, and materials in which this is the

dominant magnetic effect are termed diamagnetic materials. A diamagnetic

material is repelled towards field-free regions, but so weakly that sensitive

apparatus is required to measure the repulsive force. For diamagnetic materials, χ

is negative and falls in the range -10-6 to -10-3. M is opposite the H, hence B<B0.

3.2.2. Paramagnetic Materials

If the atoms of the material possess a permanent dipole moment due to unpaired

electron spins (the orientation of the dipole moments of individual molecules is

random in the absence of a magnetic field), these dipole moments will tend to

align themselves to an externally applied magnetic field and thus they will

enhance the field. Such materials are called paramagnetic materials. The

molecules also acquire induced magnetic dipole moments, but this diamagnetic

effect is usually smaller than the paramagnetism due to the permanent moments.

A paramagnetic material is pulled into the field. Although the attraction is usually

weak, it can sometimes be strong enough to observe in a simple way. For

paramagnetic materials, χ is positive and falls in the range 10-6 to 10-1. M is in the


same direction as H, hence B>B0. Figure 3.1 shows the magnetic response of a

typical diamagnetic and paramagnetic material subjected to an external field.

Figure 3.1 Magnetic responses associated with diamagnetic materials (left) and paramagnetic

materials (right) [4].

3.2.3. Ferromagnetic, Ferrimagnetic and Anti-ferromagnetic Materials

In some materials the quantum mechanical exchange energy of the atoms is so

large that they interact with the surrounding atoms. Below a certain temperature

called the Curie temperature, magnetic moments will tend to align to each other

even in the absence of a magnetic field. This also means that these materials can

support a permanent magnetization when no externally applied field is present. A

volume of a ferromagnetic material in which all the atomic moments are aligned

to each other is called a magnetic domain. For an un-magnetized material these

domains will cancel each other and the net magnetization will be zero. When a

magnetic field is applied, the domains that are parallel to the applied field will

grow and the others will shrink thus giving rise to a net magnetization. When the

parallel domains have grown to fill the entire piece of material the material is said

to be saturated, and the material is said to have reached its saturation

magnetization.

However, when the field is removed the parallel domains will not shrink enough

to remove the magnetization completely. There will be a remanent magnetic field

which is known as the remanence BR due to the remanent magnetization MR. If a


magnetic field is now applied in the opposite direction it is possible to remove the

magnetization. The H-field necessary to do so is called the coercive field HC. If

the applied field is further increased it is possible to saturate the material in the

other direction. If this field is removed the magnetization will follow a new path

going through the negative remanence. If one traces the magnetic field back and

forth between the positive and negative saturation points, the hysteresis loop for

the magnetic material will be obtained. Figure 3.2 illustrates these concepts.

Figure 3.2 Magnetization (M) versus magnetic field (H) where MS is the saturation magnetization,

MR is the remanence magnetization and HC is the coercivity [5].

For a ferromagnetic material it is obvious that the magnetization is not linearly

dependent on the applied field and thus the susceptibility and relative permeability

now depends on the applied H-field and the history of the magnetic material. The

effect of ferromagnetism is very large compared to diamagnetism and

paramagnetism. Therefore, equation (3.4) does not suffice to predict the behaviour

of the material since its previous state affects the magnetization (see figure 3.2). In

these materials B>>B0.

A local order, somewhat similar to ferromagnetic materials, may also be found,

for example, in certain compounds containing ferromagnetic elements and in

some non-ferromagnetic metals. This order results in adjacent dipole moments

being equal in size and parallel but reversed in direction. These materials are


called anti-ferromagnetic and have very small magnetism. Anti-ferromagnetic

materials in which adjacent dipoles are of unequal size are called ferrimagnetic

and have characteristic similar to ferromagnetic but more weakly appeared [2].

Figure 3.3 shows a summary of five different basic types of magnetic materials.

Figure 3.3 Schematic depiction of spin arrangements in different types of magnetic materials [6].

3.2.4. Super-paramagnetism and magnetic nano-particles

Super-paramagnetism is a phenomenon by which magnetic materials may exhibit

a behaviour similar to paramagnetism at temperatures below the Curie or the Neel

temperature. The width of a domain wall is a function of the magneto-crystalline

anisotropy, the exchange energy and lattice spacing of the crystal structure [5].

The domain wall is approximately a few hundred angstroms thick. As the particle

size decreases, the number of magnetic domains per particle decreases down to

the limit where it is energetically unfavourable for a domain wall to exist. In the

presence of an applied magnetic field, the spin’s orientation and subsequent

magnetic saturation is achieved with lower field strengths than with the analogous

bulk materials. When the field is decreased, demagnetization is dependent on

coherent rotation of the spins, which results in large coercive forces. The large


coercive force in single domain particles is due to magneto-crystalline and shape

anisotropies for non-spherical particles [6]. The coercive force is also dependent

on particle size as shown in figure 3.4. Theoretical prediction for the single

domain size is Ds=14 nm [5].

Figure 3.4 Coercivity as a function of particle size (~diameter) .Dsp is the super-paramagnetic size

and Ds is the single domain particle size. [5].

The magnetic anisotropy, which keeps a particle magnetized in specific direction,

is generally proportional to the volume of a particle. As the size of the particle

decreases, the energy associated with the uniaxial anisotropy decreases until

thermal energy is sufficient to overcome any preferential orientation of the

moment in the particle. A single domain particle that reaches magnetization

equilibrium at experimental temperatures in short times relative to the

measurement time is commonly referred to as super-paramagnetic. Figures 3.5

and 3.6 show domain structures and magnetic response of the particles,

respectively.

Figure 3.5 Domain structures observed in magnetic particles: a) super-paramagnetic; b) single

domain particle; c) multi-domain particle [7].


Figure 3.6 Magnetic response of ferromagnetic particles where response can be either multi-

domain (- - - - in FM diagram), single-domain (—— in FM diagram) or super-paramagnetic

(SPM), depending on the size of the particle [4].

It is important to recognize that observations of super-paramagnetism are

implicitly dependent not just on temperature, but also on the measurement time τm

of the experimental technique being used (see figure 3.7). If τ<<τm the flipping is

fast relative to the experimental time window and the particles appear to be

paramagnetic (PM); while if τ>>τm the flipping is slow and quasi-static properties

are observed (the so-called ‘blocked’ state of the system). A ‘blocking

temperature’ TB is defined as the mid-point between these two states, where τ=τm.

In typical experiments τm can range from the slow to medium time-scales of 100 s

for DC magnetization and 10-1 to 10-5 s for AC susceptibility [4].

Figure 3.7 Illustration of the concept of super-paramagnetism, where the circles depict three

magnetic particles and the arrows represent the net magnetization direction in those particles. In

case (a), at temperatures well below the blocking temperature TB of the particles or for relaxation

times τ (the time between moment reversals) much longer than the characteristic measurement

time τm, the net moments are quasi-static. In case (b), at temperature well above TB or for τ much

shorter than τm, the moment reversals are so rapid that in zero external field the time-averaged net

moment on the particles is zero [4].


3.3. Force on a magnetized particle in a magnetic field

In order to investigate the motion of the magnetic particles, it is essential to

evaluate the exerted forces on them. Generally, there are two types of forces

generated due to the motion of a magnetized particle in the fluid advecting

influenced by a magnetic field (i.e., namely hydrodynamic and magnetic forces).

Hydrodynamic forces are induced owing to the motion of any object in the fluid.

Magnetic forces are generated due to the presence of the magnetized object(s) in a

magnetic field. While hydrodynamic forces are dealt with in chapter 5, magnetic

forces are the subject of this section.

When a magnetized particle is placed in a magnetic field, two distinct types of

magnetic forces may be identified in the domain: ‘imposed field’ and ‘mutual

particle’ interactions [8]. Imposed field interactions are formed when a single

particle, or an ensemble of non-interacting particles, is influenced by an externally

imposed field. Here, it is assumed that the particle(s) does not influence the

external field. Mutual particle interactions occur where particles are so closely

spaced that the local field of a particle influences its neighbours. This interaction

comes about in two ways; the field due to the magnetization of one particle

induces an additional magnetic moment in the neighbouring particle and also

gives rise to the in-homogeneity of the field at the position of that particle. Both

effects can exert an extra magnetic force on the subject particle.

However, in particle laden fluids when the concentration of the suspended

particles is low, assuming that the particles are uniformly distributed in the media,

the distance between neighbouring particles is large. Consequently, mutual

particle interactions are too small to be considered in force equilibrium diagram of

the particles. There is not any discrete value for the concentration below which the

particle interactions become negligible as it depends on various parameters such

as the size of the particles themselves. Nevertheless, for concentration equal to or

less than 1015 particles/m3 it is reasonably justified to assume that the only major

magnetic force on the particles is the force exerted by the external field [9]. In this


research, the concentration of 1015 particles/m3 is considered for the particles in

the buffer in the course of simulations. This concentration can be provided by

most of the micro-particle suppliers. Therefore, the motion of the particles is

treated as if they are moving individually in isolation.

There are two approaches to estimate the induced magnetic force on a single

particle due to the external field: moment-energy [8] and thermodynamic [9]

methods. In this section based on the former approach, first the induced effective

magnetic moment in particles is obtained, and subsequently, the magnetic force

expression is derived.

In the beginning, we consider the dipole identification problem in general where

particle may have permanent magnetization when it is immersed in a linear media.

Then it is possible to modify extracted expressions for specific situations like

when particle is super-paramagnetic (does not include any permanent

magnetization) or when it is considered in vacuum. Let us imagine a

homogeneous sphere with radius R, permeability µ2 and net magnetic polarization

M2 which includes permanent magnetization plus any linear or non-linear

function of H0. The sphere is immersed in a magnetically linear media of

permeability µ1 and subjected to an almost uniform magnetic field H0. It is

assumed that M2 is parallel to H0 (M2H0) where H0 is the externally imposed

magnetic field vector. Figure 3.8 illustrates the sphere and direction of external

magnetic field.

Figure 3.8 Sphere of radius R and permeability µ2 immersed in a media of permeability µ1 and

subjected to a uniform magnetic field of magnitude H0.


As there is no electric current flowing in the considered space, the expression

∇×H=0 is valid everywhere and problem can be solved using a scalar potential ψ

where:

ψ= −∇H (3.6)

Assumed solutions for ψ1 and ψ2 outside and inside the sphere, respectively, have

the following forms [8]:

2

1 0ψ (r, ) H r cos Xr cos , r R−θ = − θ+ θ (3.7)

2ψ (r, ) Yr cos , r Rθ = − θ ≺ (3.8)

Here X and Y are constants which will be determined by applying two boundary

conditions. In equation (3.7) first term is the contribution of external magnetic

field and second term is the contribution of the magnetized sphere (dipole) to

magnetic potential outside the sphere. First, the magnetic potential must be

continuous across boundary between the sphere and surrounding media. Hence,

we can write:

1 2ψ (r, ) ψ (r, ) , r R,θ = θ = θ (3.9)

Second, the normal magnetic flux density must be continuous across the particle-

media interface where we can write:

1 r1 0 r2 r2H (H M ) , r Rμ = μ + = (3.10)

where 1r1

ψHr

∂= −

∂ and 2

r2ψHr

∂= −

∂ are the normal components of the magnetic

field in the media and the sphere, respectively. Using equations (3.7)-(3.10) the

constants X and Y are determined:


3 30 1 00 2

0 1 0 1

X R H R M2 2

μ −μ μ= +μ + μ μ + μ

(3.11)

010 2

0 1 0 1

3Y H M2 2

μμ= +μ + μ μ + μ

(3.12)

Figure 3.9 shows the magnetic flux and magnetic field of a magnetized particle

where the contribution of externally applied field is removed.

Figure 3.9 Magnetic flux and magnetic field for a uniformly magnetized particle.

Here Y is the magnitude of magnetic field H2 inside the sphere. In order to derive

the magnetic force on the sphere (particle) it is appropriate to assume the sphere

as a dipole where it has same effects when placed in location of the sphere. Hence,

next step is to find the magnetic moment of mentioned equal dipole. Magnetic

potential of a dipole can be estimated by [3]:

m 3 2

mr m cosψ4 r 4 r

θ= =

π π (3.13)

By comparing equation (3.13) to second term of equation (3.7), effective magnetic

moment is:

3 0 1 0

0 1 0 1

4 X 4 R2 2eff 0 2

-m ( H M )μ μ μ= π = π +

μ + μ μ + μ (3.14)


First term is the contribution of the media displaced by sphere, while second term

is owing to any magnetization (which is not necessarily proportional to H0) of the

sphere itself. Now let us consider a situation when a particle is magnetically linear

where we have:

22 0 2 r

0

1 12M H , μ= χ χ = − = μ −

μ (3.15)

where χ2 and µr are susceptibility and relative permeability of the particle,

respectively. Effective magnetic moment in obtained by substituting equation

(3.15) into equation (3.14):

3 2 1 2 1

2 1 2 1

4 R 3V2 2eff 0 0m H Hμ −μ μ −μ

= π =μ + μ μ + μ

(3.16)

where V is volume of the particle. From equation (3.16) it is straightforward to

calculate the energy of the induced dipole and the force on the particle using the

standard formula for energy and forces on point-like dipoles:

3 22 10 eff 0

2 1

U m 4 R20 0m .B .H Hμ −μ

= − = −μ = πμμ + μ

(3.17)

3 22 10 0

2 1

U 2 R2m eff 0 0F (m .B) (m .H ) Hμ −μ

= −∇ = ∇ = μ ∇ = πμ ∇μ + μ

(3.18)

It is worth noting that:

• Force is a function of volume of the particle and intensity of the magnetic

field (through induced magnetic moment).

• Force is proportional to gradient of the magnetic field intensity.

• Force is directed along the gradient of the magnetic field intensity.

• If the particle is paramagnetic (µ2>µ1), it will be attracted towards higher

magnetic field regions (positive magnetophoresis). On the contrary, the

particle is repelled by the field (negative magnetophoresis). However, this


is unlikely to occur where a medium with a considerable high relative

permeability is required.

Assuming that particle is in vacuum or de-ionized water where µ1=µ0, equation

(3.18) can be simplified to:

3 2r0

r

12 R2m 0F Hμ −

= πμ ∇μ +

(3.19)

Equation (3.19) is the final expression, which will be utilized to estimate the

exerted magnetic forces on the particles. Magnetic force has a major contribution

to the differential equation of motion of the particles, which must be solved in

order to obtain the Lagrangian trajectories of particles in the mixer domain. This

is dealt with in chapter five where the simulation procedure is addressed.

Chapter 4

Basic Design of the Micro-Mixer

4.1. Sources of magnetic field

Generally, there are two sources of magnetic field that can be integrated into

MEMS devices, namely permanents magnets and current-fed conductors.

Permanent magnets can generate strong fields on the order of 0.5-1 T compared to

the field generated by conductors, which is in the range of 10 mT. One advantage

in exploiting permanent magnets over conductors is that the actuation is not

involved in the Joule heating, which is a serious challenge in Bio-MEMS

applications. Various magnetic materials (e.g., NdFeB films) have been reported

by some researchers which can be fabricated in micro-scale. Fabrication

techniques comprise micro-machining of bulk magnets [1-2], screen printing/

bonding/mould injection [3-8], mechanical deformation [9-10], plasma spraying

[11-13], electro-deposition [14-18], sputtering [19-23] and pulsed laser deposition

[24-26]. However, it is not trivial to integrate the fabricated permanent magnet

with other components as it may be incompatible with the overall processing of

the micro-system. For instance, high temperature processes, such as LPCVD (low

pressure chemical vapour deposition) and thermal oxidation, are not compatible

with most magnetic materials.

On the other hand, various approaches have been taken to fabricate electromagnet

for a variety of applications such as magnetic actuators or sensors [27-30]. In

order to compensate for the reduced magnetic field strength, a current with

relatively large magnitude must be fed into the conductors. Therefore, the cross-

sectional area of the conductor has to be large to decrease the power consumption

Chapter 4 – Basic Design of the Micro-Mixer 66

and Joule heating. Common metal deposition techniques, such as sputtering and e-

beam deposition, are not the proper methods to make thick metal layer due to their

low deposition rates. In most cases, electroplating of high-conductive metals, such

as copper and gold, has been used to fabricate thick conductors utilizing

photoresist lithography as a mold [27-28]. In this way, high-aspect ratio (large

cross-sectional area) on the order of 10-100 µm thick can be fabricated.

In addition to the fabrication issue, there is one important advantage in integrating

electromagnets as the magnetic field source. The magnetic field generated by an

electromagnet is dynamic and can be used to provide time-varying and on/off

magnetic fields. This extra functionality is crucial in operation of some MEMS

devices such as mixers and separators.

4.2. Magnetic force due to current carrying conductors

In order to evaluate the magnitude and direction of magnetic forces injected on

micro-particles due to the field generated by a micro-conductor, a simple concept

is investigated as following. Let us assume a micro-channel with one straight

conductor embedded beneath the channel. The concept and dimensions are shown

in figure 4.1 (all dimensions are in microns). A uniform current density of

J=6×108 A/m2 (current of I=240 mA) is injected into the conductor. Direction of

the current is perpendicular to the xy-plane along the conductor. It is assumed that

the conductor in z-direction is unlimited; therefore a 2D model will suffice to

investigate the model.

Figure 4.1 Channel with straight embedded conductor beneath it.


Figure 4.2 illustrates the generated magnetic field due to the injected current. The

field inside the channel domain where particles exist is of interest; therefore the

field outside the channel is not shown. Colour-map represents the magnitude of

the field and arrows show its direction. As it can be observed, the field is

relatively stronger near the conductor where the maximum of 1800 A/m is

generated. Moving away from the conductor, the field decays quickly down to

400 A/m in the top corners of the channel.

Figure 4.2 Generated magnetic field inside the channel. Colour-map represents the intensity of the

field in A/m.

Once the magnetic field is obtained, the next step is to calculate the force applied

on the particles. The magnetic particle used in this study is M-280, Dyna-beads

(Dynal, Oslo, Norway) which is commercially available. Magnetic properties of

the reference particles are shown in table 4.1 [31].

Table 4.1 Properties of the reference magnetic particle.

Diameter, d 2.83 µm

Density, ρ 1.4 g/cm3

Relative Permeability, µr 1.76

Mass Sat. Magnetization 10.8 A.m2/kg

Saturation Magnetization 15120 A/m

Before saturation, particles are linearly magnetized with their magnetic moment

magnitude increasing in the direction of the external field. Beyond the saturation

point, magnitude of the moment tends to a constant value. Given the magnetic


field due the current in the conductor (see figure 4.2), particle will never reach the

saturation point and will remain in the linear area. Therefore, the equation (3.19)

in previous chapter is valid for the entire domain in the channel. By expanding

this equation in two-dimensional space, the expression for magnetic force is

obtained:

3 2 2 2r0 x y

r

y yx xx y x y

3 r0

r

12 R K H H2

H HH HK H H i H H jx x y y

1K 2 R2

μ −= πμ ∇ = ∇ +

μ +

∂ ∂∂ ∂= + + +

∂ ∂ ∂ ∂μ −

= πμμ +

mF H . ( )

.[ ( . . ) ( . . ) ] (4.1)

For the selected particle, coefficient K is 4.5×10-24. Figure 4.3 shows the magnetic

forces which would be applied on a particle if it was placed in the channel at

various positions in the domain. Colour-map represents the magnitude of

magnetic force in pN and arrows show the direction of the resultant force. As it

can be observed, the particle from any position in the channel will be attracted

towards the high strength magnetic field region close to the conductor. Also, it is

worth noting that moving away from the intense region, the force declines more

quickly than the magnetic field itself. This is due to the fact that the force is a

function of both magnetic field and its gradient and therefore, is influenced by a

double effect. A maximum force of 0.68 pN is obtained at the bottom of the

channel along its central line.

Figure 4.3 Applied magnetic forces on the particles in the channel. Colour-map represents the

magnitude of the force (pN) and arrows show its direction.


An appropriate concept must be adopted to use such forces to induce a chaotic

advection in the particles. In order to reach a proper design, a good understanding

of the chaos phenomenon and the mechanism to induce chaotic regimes, is

essential. Therefore, a brief introduction to chaos in given in the following

section, and subsequently, the basis of the micro-mixer design is discussed.

4.3. Chaotic mixing

4.3.1. Chaos theory

Chaos is a technical term for a particular type of irregular motion induced by a

deterministic system. Chaos is a long term phenomenon and therefore when

energy dissipation occurs, for instance in mechanical systems in the presence of

friction, a continuous input of energy is essential to maintain the chaotic response,

otherwise any observed irregularity will be transient. Chaos has now been viewed

in a wide variety of physical systems including mechanical, fluid, electronic,

chemical and even biological experiments [32].

In the context of fluid dynamics, two branches can be identified; one

corresponding to volume-contracting or dissipative systems, and the other to

Hamiltonian systems or volume-preserving systems. While chaos associated with

dissipative systems is called Eulerian or turbulent chaos, the latter is often called

Lagrangian or non-turbulent chaos, which is of interest in this study.

A state of Lagrangian chaos can be identified when the solution of the trajectory

equations has a sensitive dependency on initial conditions, and initially nearby

trajectories diverge exponentially. Since the system is frictionless or Hamiltonian,

the phase space is conservative. The possibility of having Lagrangian (laminar)

chaos without Eulerian (turbulent) chaos implies that mixing enhancement is

possible in the flows with very low Reynolds numbers.


Sensitivity to initial conditions means that each point in the system is arbitrarily

closely approximated by other points with considerably different future

trajectories. Therefore, an arbitrarily small disturbance in the current trajectory

may result in significantly different future behaviour. Sensitivity to initial

conditions is usually known as the "butterfly effect". This phrase refers to the idea

that a butterfly's wings might create very small changes in the atmosphere that

eventually cause a tornado to appear (or prevent a tornado from appearing). The

flapping wing represents a small change in the initial condition of the system,

which leads to a chain of events causing large-scale phenomena. Had the butterfly

not flapped its wings, the trajectory of the system might have been extensively

different. Figure 4.4 illustrates the sensitive dependency on initial condition in a

system. As it can be observed, some trajectories are evolved into different

attractors despite their close initial positions. On the contrary, trajectories with

different initial conditions can be evolved into the same attractor.

Figure 4.4 Sensitive dependency on initial conditions, blue squares represent initial state and black

squares represent equilibria.

4.3.2. Chaos in laminar flows

It has recently been generally realized that high Reynolds number turbulent flow

is not necessary for complex particle trajectories in fluid dynamics. Laminar flow,

once thought to have simple dynamics, can give rise to chaotic behaviour of

Lagrangian particle trajectories [33]. Mixing in laminar flows can be enhanced


through chaotic advection, the phenomenon in which particles advected by a

periodic velocity field show chaotic trajectories.

Laminar flow can be one-, two-, or three-dimensional, and may be steady (time-

independent) or unsteady (time-dependent) [33]. However, chaos cannot occur in

steady two-dimensional flows, but only in three-dimensional and two-dimensional

time-dependent flows. In a two-dimensional flow, time-dependency may be

considered as an added third dimension. Since a two-dimensional simulation is

carried out in this study (this is discussed in the next section), the kinematics of

Lagrangian chaos in such flows is investigated as follows:

The starting point for the chaotic advection is the Lagrangian representation of the

fluid as:

x y zdx dy dzu (x,y,z, t) , u (x,y,z, t) , u (x,y,z, t)dt dt dt

= = = (4.2)

where ux, uy and uz are the Cartesian components of the velocity field u(ux,uy,uz).

Here we are dealing with laminar flows where the fluid is incompressible (i.e.,

∇·u=0). In the case of a two-dimensional flow, the incompressibility of the flow

implies that:

∂∂

− =∂ ∂

yx uu 0x y

(4.3)

Therefore, there exists an exact differential dψ so that:

dx dy,dt y dt x

∂ψ ∂ψ= =∂ ∂

(4.4)

where ψ is known as stream function and is equivalent to Hamilton’s equations of

motion in classical mechanics. In fact, two-dimensional kinematics of advection

by an incompressible flow is equivalent to the Hamiltonian dynamics of a one-

degree-of-freedom system (ψ=ψ(x,y)). It can be proven that the motion in two-


dimensional static flow is always regular and integrable [35-36]. In this case, the

fluid elements (or particles) flow along the streamlines and cannot produce

chaotic regimes. Two-dimensional unsteady incompressible flows, on the other

hand, have the dynamics of one-degree-of-freedom time-dependent Hamiltonians

(ψ=ψ(x,y,t)), which are generically non-integrable. In this case, it is possible for

the system to exhibit chaotic particle trajectories as it is non-integrable. The time-

dependency of ψ can be caused by some simple external variation of the flow

system. The two space dimensions and time together provide the minimum of

three dimensions required for chaos to take place.

Although time-dependency as the third dimension is usually referred to the fluid

velocity components, the same strategy can be utilized for particles’ velocity

components. Solid particles suspended in a fluid, follow the flow streamlines and

therefore, will have the same velocity field. If an oscillating component is added

to their motion through using external forces, chaotic advection can be produced.

Relative to integrable advection, chaotic advection enhances stretching and

folding of material interfaces. This deformation of fluid–fluid boundaries

increases the interfacial area across which diffusion occurs, which typically leads

to significantly more rapid mixing [34]. In the case of mixing of the suspended

particles which here is of interest, this consecutive deformation leads to an

increase in attachment of biological cells onto the particles. Therefore, the key to

effective mixing lies in producing stretching and folding of lines in two-

dimensional (surfaces in three-dimensional) flows which can be equated with

chaos. In rough terms, a necessary condition for chaos is the crossing of

streamlines, which must occur at different times.

4.4. Basic design

Now the question arises on how to employ magnetic forces generated by current-

fed conductors to produce chaotic regimes in motion of the magnetic particles. To

answer this question, let us start with investigating the dielectrophoretic chaotic


mixer mentioned in chapter 2 [37]. In this mixer attractive or repulsive forces on

polystyrene particles in aqueous medium are induced by changing the frequency

of the signal. High frequencies (10 MHz) generate a negative DEP force, where

particles are repelled from electrode edges, while at lower frequencies (100 KHz)

positive DEP forces move the particles to the electrode edges, as they correspond

to highest field gradients (see figure 4.5). The combination of the two opposite

motions generates stretching and folding required for chaotic advection.

Figure 4.5 Producing stretching and folding using positive and negative DEP forces generated by

different frequencies [37].

However, this concept cannot be used for magnetophoretic forces due to the major

limitation of these forces, i.e., merely attractive forces can be generated by

magnetophoresis. Therefore, opposite force, which is required to repel the

particles attracted to low fluid velocity field region cannot be generated. Even if

the particles are exposed to an AC magnetic field, they will be attracted towards

the source. This is due to this fact that an alternating magnetic field can only

change the direction of polarity in the particles considering the relaxation time

(time between moment reversals) of the particles. Hence, another strategy must be

adopted to utilize magnetophoretic forces.

In this investigation, a straight channel with two embedded serpentine conductors

beneath the channel is used to produce chaotic patterns in the motion of particles


and intensify the capturing of bio-cells. The burst-view of the mixer is depicted in

figure 4.6. Two flows; target cells suspension and the particle laden buffer, are

introduced into the channel and manipulated by pressure-driven flow (see figure

4.7). While the cells follow the mainstream in upper half section of the channel

(they are transported by convection of the suspending bio-fluid), the motion of

magnetic particles is affected by both the surrounding flow field and the localized

time-dependent magnetic field generated by sequential activation of two

serpentine conductors (here we call the advection due to bulk flow field passive

and due to magnetophoretic forces active for the sake of distinction of these

phenomena).

Figure 4.6 Burst-view of the proposed mixer.

Figure 4.7 Top view of the proposed mixer illustrating one mixing unit (all dimensions are in

microns).


As discussed in section 4.2, magnetic field declines quickly as we move away

from the conductors. To overcome this problem and inject particles from various

positions across the channel, geometry of the conductors must be adopted in a

way that generates larger forces. Circular form of the tip of each tooth in

conductors provides an intensified magnetic field in the centre where it acts as a

sink for particles. Particles from various positions in the upstream and

downstream sides move across the streamlines and are attracted towards the centre

of nearest activated tip where the maximum magnetic field exists. By using a

proper periodical current density and structural geometry, chaotic patterns can be

produced in the particles’ motion, which leads to their better mixing into the bio-

fluid suspension.

Dimensions are shown in figure 4.7. Channel is 150 µm wide and 50 µm deep.

Conductors are 50 µm high and 25 µm wide in the section and distances between

centre of circular tips of the conductors are 100 µm and 65 µm in x and y

directions, respectively. Each row of upper and lower conductors is connected to

the power supply alternately. Employment of two rows of conductors at both sides

of the channel allows pulling particles from each side, therefore, compensating for

the limitation of magnetic forces not being bi-directional. The mixing operation

cycle consists of two phases. In the first half cycle, one of the conductor arrays in

switched on while the other one is off. In the next half cycle, the status of

conductor arrays is reversed. Each mixing unit consists of two adjacent teeth from

opposite conductor arrays and the mixer is composed of a series of such mixing

units which are connected together. Boundaries of one mixing unit are illustrated

in figure 4.7.

Figure 4.8a shows one mixing unit with its magnetic field generated near the

circular tip of the conductor when a current of 750 mA is injected into one

conductor array and is turned off in the opposite array during a half cycle of

activation. The colour-map represents variations in the magnetic field intensity at

a plane 10 µm above the surface of the conductor where the maximum magnitude

of the field is about 6000 A/m at the centre of the circular tip (point P). Figures


4.8b and 4.8c show the magnitude of the total magnetic field (H=(Hx2+Hy

2)1/2) in

x-y plane along two lines A-A and B-B, respectively. Graphs show the field at

different heights above the conductor and as expected, the closer to the conductor,

the stronger the magnetic field can be observed.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1x 10

-4

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

L (m)

Mag

netic

Fie

ld (A

/m)

5 µm20 µm2 µm15 µm10 µm

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1x 10

-4

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

L (m)

Mag

netic

Fie

ld (A

/m)


(a)

(b)

(c)

Figure 4.8 Magnetic field near the tip of one tooth in one mixing unit during a single phase of

activation, (a) colour-map of the field (A/m), (b) & (c) magnitude of the field at different heights

above the conductor along lines A-A and B-B, respectively.


Magnetic forces exerted on one single particle along lines A-A and B-B are

plotted in figure 4.9 showing the x and y components of the force and also the

total resultant force at different heights above the conductor. As expected, the

maximum force (~6 pN) is applied on particles near the conductor and inside the

circle of its tip where the intensity of magnetic field is at its maximum value.

Although the magnetic field is maxima at the centre point P, the force on particles

is relatively small at this point. This is due to the fact that the magnetic force is

proportional to the gradient of the field which is almost constant in the

neighbourhood of the point P (see figures 4.8b and 4.8c). In moving away from

the conductor, the force drops significantly due to a dramatic decrease in the

magnetic field, which in turn affects the magnetic moment.

It is worth noting that the magnetic force is three-dimensional and the z

component of the force is downward, which together with gravity, pull the

particles towards the bottom of the channel and restrict their motion to a two-

dimensional pattern. In fact, this component has no contribution to the chaotic

motion of the particles and is assumed not to be influential on the process of

mixing. Therefore, a two-dimensional simulation is conducted in this study and

planar forces at 10 µm above the surface of the conductors are of interest as closer

layers might not be practical due to fabrication restrictions.


0 1 2 3 4 5 6 7 8 9x 10

-5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 x 10-11

L (m)

F (N

)


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1x 10

-4

-8

-6

-4

-2

0

2

4

6

8 x 10-12

L (m)

Fx (N

)


1 2 3 4 5 6 7 8 9x 10

-5

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1 x 10-11

L (m)

Fx (N

)


0 1 2 3 4 5 6 7 8 9 10x 10

-5

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1 x 10-11

L (m)

Fy (N

)


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1x 10

-4

-8

-6

-4

-2

0

2

4

6

8 x 10-12

L (m)

Fy (N

)


1 2 3 4 5 6 7 8 9x 10

-5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 x 10-11

L (m)

F (N

)


(a) (b)

(c) (d)

(e) (f)

Figure 4.9 Magnetic forces exerted on one single particle along lines A-A and B-B at different

heights above the conductor in one mixing unit during a single phase of activation, (a) & (b) x-

components, (c) & (d) y-components, (e) & (f) resultant force.

4.5. Scaling effects

One way to produce larger magnetic forces is to generate stronger magnetic fields

by increasing the current in the conductors. However, this approach is limited as


the maximum current density cannot exceed certain values. Otherwise, problem of

Joule heating arises. Current density J, is a function of current I, and the cross-

sectional area of the conductor A (J=I/A). Therefore, one solution is to use a

conductor with large cross-sectional area which in turn allows increasing the

current while the current density is held in the permissible range. Nevertheless, as

discussed in section 4.1, utilization of thick conductors is restricted by fabrication

techniques. On the other hand, employment of large conductors can diminish the

magnitude field gradient which in turn declines the force at points of interest.

Another solution is to optimize the conductor geometry through modifying its

cross-sectional area. In this way, it is possible to generate intensified fields with

high gradients at points of interest. In this study, conductors with rectangle

sections are adopted and the process of optimizing the cross-sectional area is

carried out through determining the thickness and width of the conductor and their

ratio. The scaling effects are investigated considering two crucial issues, namely

magnitude of the magnetic forces and Joule heating phenomenon.

4.5.1. Magnetic forces

Since the circular tips of the conductors generate the major part of the forces, a

circle is assumed to resmble a single tip of the conductor in order to evaluate the

effect of the conductor size. This assumption allows conducting an axisymmetric

two-dimensional simulation which reduces the consumption of computer

resources and increases the accuracy of the model significantly. The concept is

shown in figure 4.10a. The cross-section of the conductor and channel in plane A-

A is depicted in figure 4.10b illustrating all geometrical parameters. The inner

radius (Ri) of the conductor and distance between the channel and conductor (h)

are 15 µm and 10 µm, respectively. Thickness (H) and width (W) are varied

between 20 µm and 50 µm with 5 µm step. A current of 750 mA is fed into the

conductor and the radial component of the generated force is evaluated along line

B-B which represents the resultant of the x and y components of the force in

Cartesian system in three-dimensional model.


Figure 4.10 Conceptual diagram of one single circular tip in (a) three-dimensional and (b) two-

dimensional models.

Figure 4.11 shows the variation of the force against width where the thickness is

20 µm. As it can be observed, increasing the conductor width decreases the

maximum magnitude of the force (here it occurs at about 20 µm from the central

axis). However, the force generated by wider conductors extends more which

means a wider area is influenced by the force. In narrow conductors, the force is

stronger in the vicinity of the centre and diminishes hastily outside the tip.


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1x 10-4

-2

-1.8

-1.6

-1.4

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0 x 10-12

L (m)

Fr (N

)

20253035404550

Figure 4.11 Variation of the force against width (H=20 µm).

Figure 4.12 shows the variation of the force against the thickness where the width

is 20 µm. The thinner the conductor, the stronger the magnetic force. The

maximum generated force in H=20 µm is decreased down to almost 25% when

the thickness of H=50 µm is used. In fact, thick conductors extend the field along

the y direction below the channel which is not useful for the mixing. However, the

extension of the force in x direction is not affected by the thickness of the

conductor. Therefore, use of sections with small thickness down to the point

where Joule heating is not an issue, is of interest.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1x 10-4

-2

-1.8

-1.6

-1.4

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0 x 10-12

L (m)

Fr (N

)

20253035404550

Figure 4.12 Variation of the force against thickness (W=20 µm).


4.5.2. Electro-Thermal analysis (Joule heating)

Joule heating (also known as ohmic heating or resistive heating) refers to the

increase in temperature of a conductor as a result of resistance to an electrical

current flowing through it. At an atomic level, Joule heating is the result of

moving electrons colliding with atoms in a conductor, where momentum is

transferred to the atom, increasing its kinetic energy. This phenomenon is

governed by the Joule's law, relating the amount of heat released from an

electrical resistor to its resistance and the charge passed through it:

2 LQ I R t , RA⋅ ρ

= ⋅ ⋅ = (4.5)

where Q is the heat generated by a constant current I flowing through a conductor

of electrical resistance R, for a time t. Terms L, ρ and A are length, electrical

resistivity of the material and cross-sectional area of the conductor, respectively.

Here, the effect of the conductor geometry on Joule heating is through its cross-

sectional size which is inversely related to the electrical resistance. Copper is

considered as the bulk material for the conductors.

In order to evaluate the temperature rise in the mixer, a simplified model of one

conductor tip with surrounding substrate (here is the Glass) which represents half

a mixing unit is considered as shown in figure 4.13. Three cross-sectional area

(highlighted in the figure) are considered; small (25 µm × 25 µm), medium (35

µm × 35 µm) and large (45 µm × 45 µm). A constant current of 120 mA is passed

through the conductor and the generated heat (temperature rise) is investigated

against the size. Thermal boundary conditions for the model are shown in figure

where free convective cooling is considered for the sides exposed to air with a

temperature of Ta=290 K and heat transfer coefficient of h=50 W/m2K. For the

two lateral faces where the other mixing units are attached, thermal insulation is

considered assuming that the same quantity of heat is generated in lateral mixing

units and therefore, no conduction occurs through mentioned boundaries.


Figure 4.13 Concept of the model illustrating the thermal boundary conditions. Surrounding area is

air with h=50 W/m2K at 290 K.

Electrical resistivity ρ for copper and glass is the reciprocal of the temperature-

dependent electrical conductivity σ:

σ =ρ + α −0 0

1(T)(1 (T T ))

(4.6)

where ρ0 is the resistivity at the reference temperature and α is the temperature

compensation slope of the material. Electrical and thermal properties and density

for the both materials are given in table 4.2.

Table 4.2 Electrical and thermal properties of the glass and copper.

a Resistivity at reference temperature of Ta=293 K.

Figure 4.14 shows the temperature rise in a conductor with cross-sectional area of

A=35 µm× 35 µm=1225 µm2 where the colour-map represents the temperature

Property / Material Glass Copper

Thermal conductivity K (W/m.K) 1.38 400

Heat capacity Cp (J/kg.K) 703 385

Resistivitya ρ0 (Ωm) 1010 1.72×10-8

Temperature coefficient α (1/K) 0.55×10-6 17×10-6

Density (kg/m3) 2203 8700


distribution on the boundaries. The maximum temperature of 300.18 K is deduced

on the copper face while the lowest temperature (300 K) is observed on outside

boundaries of the substrate where the material is directly exposed to convective

cooling of the air. Shown in figure 4.15, a graph of the maximum temperature rise

versus the cross-sectional area of the conductor; the larger the area, the less the

temperature rise. Temperature rise in smallest area (~19 K) is almost twice as that

in the largest area (~8 K). This is due to this fact that as the area increases, the

electrical resistance, and consequently, the generated heat deceases. Therefore

from the heat generation point of view, it is preferable to employ wider and

thicker conductors.

Figure 4.14 Temperature distribution on the boundaries (cross-sectional area A=1225 µm2).

400 600 800 1000 1200 1400 1600 1800 2000 2200

296

298

300

302

304

306

308

310

312

Cross-sectional area (micron2)

Tem

pera

ture

(K)

A=25*25

A=35*35

A=45*45

Figure 4.15 Maximum temperature rises in the domain for different cross-sectional areas. The

current is constant in all models.


4.5.3. Conductor size

Based on the results obtained from evaluations of the effect of conductor size on

the magnetic force and temperature rise, it is deduced that a compromise between

both issues must be reached. Therefore, a thickness and width of 35 µm is

considered for preliminary simulations. The adopted width allows applying

required forces on the particle in a relatively large area. Moreover, the moderate

thickness can compensate for the Joule heating owing to the larger current

densities.

Chapter 5

Numerical Simulations and Results

Processes for the separation and mixing of particles often depend on the behaviour

of the particles when they are subjected to the action of a moving fluid and

external forces. Most of the methods for the determination of particles’ path in the

domain involve relative motion between the particles and a fluid. In this chapter, a

brief introduction on the multiphase flows and phase coupling is presented first,

followed by a discussion of the flow of a spherical particle relative to the fluid.

Subsequently, the simulation procedure and parameters are explained and finally,

the results are presented.

5.1. Multiphase flows

A phase refers to the solid, liquid, or vapour state of matter. A multiphase flow is

the flow of a mixture of phases such as gases (bubbles) in a liquid, or liquid

(droplets) in gases, and so on. As in the case of this study, liquid-solid flows

consist of flows in which solid particles are carried by the liquid.

5.1.1. Phase coupling

An important concept in the analysis of multiphase flows is coupling. If the flow

of one phase affects the other while there is no reverse effect, the flow is said to

be one-way-coupled. If there is a mutual effect between the flows of both phases,

then the flow is two-way-coupled. Coupling can take place through mass,

momentum, and energy transfer between phases. Mass coupling is the addition of

Chapter 5 – Numerical Simulations and Results 87

mass through evaporation or the removal of mass from the carrier stream by

condensation. Momentum coupling is the result of an interaction force, such as a

drag force, between the dispersed and continuous phase. Energy coupling occurs

through heat transfer between phases [1]. Here, the momentum coupling which

explains the hydrodynamic interactions is of interest and a schematic diagram of

this type of coupling is shown in Figure 5.1.

Figure 5.1 Schematic diagram of coupling [1].

A further definition of coupling includes four-way coupling in which

hydrodynamic interactions extends beyond the two-phase interactions. Four-way

coupling addresses the situation where, in addition to ‘discrete phase’-‘carrier

phase’ (particle-liquid) interaction, particle-particle collisions also affect the

multiphase motion. However, two-way and four-way couplings effects become

important when particle fraction exceeds certain values which is the situation in

dense flows. Strictly speaking, mutual couplings do not apply to particle laden

flows with a concentration lower than 1015 particles/m3 [2]. Therefore,

considering the mentioned concentration, one-way coupling is assumed to be

valid in this study and as in the case of magnetic interaction which was discussed

in chapter 3, motion of the particles is treated as if they are moving individually.


5.1.2. Motion of a single particle in a viscous fluid

Consider the motion of a spherical particle inside a viscous fluid which flows

slowly enough for the condition Re<1 to be satisfied. No-slip condition applies on

the surface of the sphere and the centre of coordinates is coincident with the

centre of the sphere. The flow domain is much larger than the diameter of the

sphere (this is referred to sometimes as an “infinite domain”). The multiphase

flow in one-way coupled and there is no magnetic interaction between particles.

Hydrodynamic and magnetic forces that may act on the particle are shown in

figure 5.2 illustrating the free-body force diagram of the particle. Net force acting

on the particle will be used to extract the velocity expression which can yield the

Lagrangian equation of motion.

Figure 5.2 Free-body force diagram of the particle for a one-way-coupled problem.

Term Fm is the magnetic force which was obtained in chapter 3. Term Fd refers to

the parasitic drag force which is the force that resists the movement of the particle

through surrounding fluid. The drag force is generated in parallel to the relative

motion of the particle to the flow and in opposite direction. The magnitude of the

drag force depends on both particle and flow properties. Generally, conditions of

flow relative to a spherical particle are characterised by the particle Reynolds

number:

pudRe ρ

=μ

(5.1)

where

ρ is the density of the fluid

µ is the viscosity of the fluid

d is the diameter of the sphere

u is the velocity of the particle relative to the fluid


For the case of creeping flow (i.e., flow at very low velocities relative to the

particle), the drag force Fd on the particle was obtained by Stokes [3] who solved

the hydrodynamic equations of motion, the Navier–Stokes equations, to give:

dF 3 du= πμ (5.2)

Drag is made up of friction forces, which act in a direction parallel to the particle's

surface (primarily along its sides, as friction forces at the front and back cancel

themselves out), plus pressure forces, which act in a direction perpendicular to the

particle's surface (primarily at the front and back, as pressure forces at the sides

cancel themselves out). Skin friction constitutes two-thirds of the total drag on the

particle as given by equation (5.2). Therefore, the total force Fd is made up of two

components:

(i) skin friction: 2πμdu

(ii) form drag: πμdu

While several textbooks make this distinction between the two parts of the drag

force, here drag force is considered as a single entity that arises from the

interactions between the fluid and the particle and not two different forces.

Equation (5.2), which is known as Stokes’ law is applicable only at very low

values of the particle Reynolds number where Rep<0.2 which is encountered in

micro-flows [4]. The velocity of the particle due to the magnetic and drag forces

can be described by Newton’s second law:

p mum F 3 dut

∂= − πμ

∂ (5.3)

where mp is the particle mass. At steady state, the two opposing forces are equal

in magnitude and the spheres move at constant velocity. When magnetic force is

exerted, particle accelerates and reaches the terminal velocity Vm as follows:


mm

FV3 d

=πμ

(5.4)

The particle reaches this velocity in a very short time which is known as particle

relaxation time and is estimated by:

2

p d18ρ

τ =μ

(5.5)

where ρp is the particle density. Given the particle properties (table 4.1) and

viscosity of 0.001 kg/ms (characteristic of water at room temperature), particle

relaxation time is less than 100 ns. During this time, particle moves a minute

distance which can be neglected in tracing the particle position. Hence, ignoring

the acceleration phase, we assume particles react to magnetic forces with no delay

and total velocity of the particle at each moment (Vp) is the sum of velocity due to

fluid field (Vf; passive advection) and velocity due to the magnetic field (Vm;

active advection):

mp f m f

FV V V V3 d

= + = +πμ

(5.6)

Equation (5.6) will be used in the following section to obtain the Lagrangian

trajectories of the particles.

5.2. Numerical simulations

5.2.1. Simulation procedure

A two-dimensional numerical simulation is carried out assuming that particles are

neutrally buoyant and their motion in z direction is either zero or negligible as

discussed in chapter 4. Procedure consists of two steps: first, the steady-state

velocity field of an incompressible Newtonian fluid (water) and time-dependent

magnetic field are computed using commercial multiphysics finite element

package Comsol (COMSOL, UK) and velocities of the particles due to the fluid


and magnetic fields are extracted (see Appendix A). Then trajectories of particles

are evaluated by integrating the sum of velocities using Euler integration method

and through developed Matlab codes:

mp p f

Fr V dt (V ) dt3 d

= ⋅ = + ⋅πμ∫ ∫ (5.7)

This method of integration is adopted because the equation of motion of the

particles is highly stiff due to quick changes in the magnetic field and,

consequently, in magnetic forces when the signal phase change occurs. Therefore,

some commonly used algorithms, such as Runge-Kutta, for computing the

solution to differential equations method can take a relatively long time to solve

such stiff problems. In order to obtain accurate results, a small discrete time-step

(10 ms) is considered for particle tracking procedure and, where necessary, the

fluid velocity and the magnetic intensity is linearly interpolated between two

adjacent grids. Neutral diffusion (Brownian motion) of magnetic particles of this

size is insignificant. Diffusion rate for such a particle in water would be in the

order of 10-13 m/s. Advection due to magnetic forces is approximately 105-106

times greater than the diffusion fluxes. Therefore, diffusion is neglected. In fact,

particles are small enough not to agitate the flow, but large enough not to get

involved with Brownian motion, moving only with the surrounding flow itself.

Although the trajectories of the particles will suffice to numerically evaluate the

induced chaotic regimes in the mixer, trajectories of the biological cells are also

obtained. These trajectories will be later utilized to quantify the efficiency of the

mixer. Trajectories of the cells are obtained using the same method as for

magnetic particles, with the exception that cells are magnetically inactive and

simply follow the mainstream in the fluid flow field. The size of biological

entities may vary from a few nano-meters (such as proteins) to several micro-

meters (such as cells). In this study, cells are considered to be spheres of 1 μm

diameter.


For reduction of the computational domain, the smallest possible mixing unit with

periodic boundary conditions must be used. For a periodic mixer like the one

proposed, the flow field solution is also periodic and remains invariant in each

mixing unit. This indicates that every single mixing unit contains all the

information of the flow in the whole mixer system. However, the same cannot be

said for the magnetic field as the generated field by neighbouring teeth affects the

field in each unit. In other words, one mixer unit as shown in figure 5.3 cannot

comprise the field of outer teeth if it is simulated as a stand-alone domain. Hence,

one extended unit which includes adjacent teeth at both sides is used (see figure

5.3 for the boundaries of this extended unit) and field solution is extracted for

inner unit. Other teeth are too far to influence the field in the inner unit and their

effect is not taken into account.

Figure 5.3 Boundaries of one simulated (extended) and one mixing unit.

5.2.2. Simulation parameters

In order to characterize the mixing efficiency and to optimize the design, effects

of a large number of different parameters can be investigated. These parameters

include geometry (e.g., channel dimensions, size of the conductors and spacing

between them), bulk flow rate, particle characteristics, particle concentration, and

magnitude and frequency of the current. Nevertheless, it seems impractical to

consider the effect of variation of all influential parameters simultaneously due to

computational restrictions. Therefore, based on preliminary calculations, a

reasonably optimized geometry for the channel and conductors (as discussed in

the previous chapter) and a current magnitude of 750 mA, are adopted and the


effect of variation of two driving parameters; namely the bulk flow velocity and

frequency of the current, has been investigated against mixing efficiency. The

frequency is the reciprocal of the period which consists of two half-cycles of

pulsing signal as shown in figure 5.4. T is the length of period and Tp is the length

of a half-cycle or phase.

p

1 1fT 2T

= = (5.8)

Figure 5.4 Phase shift control signal.

The ratio of these driving parameters is defined as a dimensionless number St

(Strouhal number):

f LStV

= (5.9)

where f is the frequency, L is the characteristic length (here, distance between two

adjacent teeth), and V is the bulk mean velocity of the fluid.

5.3. Simulation results

Figure 5.5 illustrates the developed parabolic fluid field velocity inside the

channel for a steady flow with a bulk velocity of 40 µm/s. Colour map represent

the pressure and arrows show the direction and magnitude of the velocity. As it

can be observed, the velocity profile is uniform along the channel length and


merely pressure is changing which is not of interest in investigating the mixing

process. The pressure drop along a 1 mm long channel is about 3.5×10-3 Pa.

Density and viscosity of the liquid remain almost constant for such range of

pressure variation. Therefore, the premise that each mixing unit contains all

necessary information for the velocity field is valid.

The width of the channel is 150 µm and considering the fluidic properties of water

(ρ=103 kg/m3, μ=10-3 kg/ms), the Reynolds number would be in order of 10-3

emphasizing the flow is absolutely laminar.

Figure 5.5 Developed parabolic fluid field velocity inside the channel. Arrows show the direction

and magnitude of the velocity field.

5.3.1. Advection of the cells and particles

Figure 5.6 illustrates the position of the particles and cells while advecting within

three and a half mixing units. Bio-cells (red dots) and magnetic particles (blue

dots) enter the first mixing unit (across line A-A) from the left in upper and lower

halves of the section, respectively, and with the same concentration. When there is

no magnetic actuation, both cells and particles remain in their initial section and

simply follow the streamlines of the parabolic velocity profile in Poiseuille flow.

In this situation, tagging might occur only in the middle of the channel along the

interface between two halves. All dimensions in the figure are normalized to the

characteristic length (200 µm).


Figure 5.6 Advection of cells and particles within three and a half mixing units when no external

perturbation is applied. Dimensions are normalized to the characteristic length.

Fig. 5.7 illustrates a typical effect of magnetic actuation (St=0.4, V=40 μm/s)

within the same mixing units at different snapshots. When the external field is

applied, particles travel across the streamlines as well as the interface. Therefore,

they find the opportunity to spread in upper section where they can meet and

collect cells. Magnetically inactive cells will have the same behaviour as previous

situation when no perturbation was applied. As it can be observed, some particles

far from the central line of the channel remain in the lower section as the magnetic

forces in these regions are not strong enough to attract them during the lower

array activated half-cycle. In order to agitate more particles, one solution is to

inject higher magnitudes of the current into the conductors. An alternative

solution can be driving the flow at lower velocities and use of longer activation

periods, whereupon these particles will have the chance to be advected towards

the central region before any phase change occurs. The effect of operating

parameters on the advection of the particles is examined in next chapter.


(for the caption see next page)


Figure 5.7 Advection of cells and particles within three and a half mixing units with magnetic

perturbation (St=0.4, V=40 μm/s). Dimensions are normalized to the characteristic length.


5.3.2. Basis of chaotic advection in particles

In order to explain the basis for chaotic advection in the proposed micro-mixer,

trajectories of four particles (particles 1-4 in figure 5.8) are considered as typical

trajectories in the mixer. Particles are released in the first mixing unit uniformly

with the spacing of 10 µm when St=0.2 and V=45 µm/s. During the first half-

cycle, first array (conductor I) is on and second array (conductor II) is off. Particle

1 feels a strong magnetic force in y direction and tends to move in this direction

while it is advected by the mainstream in x direction. Note that depending on its

location in the channel which determines both drag force in the Poiseuille flow

and magnetic force, particle 1 can have a positive or negative velocity in x

direction according to equation (5.6). Particle 2 is further from the conductor I and

does not find any chance to be attracted upwards completely during the first half-

cycle. Therefore, two initially nearby particles diverge inducing the mechanism of

stretching, which is marked with a rectangle. In this phase particle 1 is exposed to

the target cells across different streamlines and captures them in case on any

collision.

Figure 5.8 Consecutive stretching and folding in trajectories which results in chaotic advection

(St=0.2, V=45 μm/s).

In the following half cycle, electric current is injected into the conductor II and

turned off in conductor I. In this phase, particle 1 is free to move from the

previous location and is further advected by the mainstream until it approaches a

region of strong enough magnetic force and, consequently, is pulled towards the

centre of conductor II. Particle 2 is subject to a small magnitude of magnetic force

in y direction (see figure 4.9c in chapter 4) but tends to move faster than the


mainstream by virtue of magnetic force in x direction. In this phase, particle 2

approaches and tags the target cells, if any, along one streamline. Folding is

achieved where two distant trajectories converge and even in some operating

conditions cross each other. Consecutive stretching and folding can be produced

in this way which is the basis of chaos.

Particles 3 and 4 which are too far from the conductor I to be attracted, are

dragged downstream by the fluid and gradually move towards the upper half

section of the channel. After passing a few mixing units, almost all particles

penetrate to cells’ region and fluctuate in a chaotic regime confined to the tips of

two conductor arrays.

Chapter 6

Characterization of the Micro-Mixer

Characterization of the micro-mixers can be regarded as evaluation of several

aspects of the device, including mixing time, mixing length and quality of the

mixing process. The quantification of the extent of mixing (which is the subject of

this chapter) is crucial for evaluation of performance as well as design

optimization of micro-mixers. Despite the numerous recent works on micro-

mixers, characterization of micro-mixers still remains a challenging issue. In the

following, a brief description on different techniques for the characterizing the

mixers is presented, followed by detailed assessment of the proposed concept in

this study.

6.1. Mixing assessment

While the performance of some micro-components such as pumps, or extractors

can be characterized by comparison of standard parameters (e.g. pump frequency,

or volumetric mass-transfer coefficient), a similar protocol for the analysis of

mixing is more complicated. The best solution for mixing characterization would

be a locally in-line measurement of the concentration profiles along the flow axis.

However, sensors of such small size and fast response time are not available at

present [1]. Therefore, the quality of today’s micro-mixers has to be characterized

by indirect means. Here, various techniques are classified into two categories;

namely experimental and numerical approaches. Since this study is a numerical

design study, the main focus will be on the later approach.

Chapter 6 – Characterization of the Micro-Mixer 101

6.1.1. Experimental techniques

Experimental techniques primarily rely on optical inspection of the fluids

transported through the mixer. The most common quantification technique is

using dilution of a tracer dye to determine the extent of mixing (dilution-type

experiments) which is done by the aid of microscopic-, photo- or video- cameras.

Usually, fluorescent dye streams are observed, followed by the evaluation of the

corresponding recorded intensity image. Since the concentration of the dye is

proportional to the intensity of the recorded image, the uniformity of the

concentration image can be quantified by determining the standard deviation of

the pixel intensity values [2]. In cases where the imaging direction is

perpendicular to the fluid layers, the two layers, even at the channel entrance,

appear to be completely mixed. In such cases, an imaging system with a confocal

microscope is required for the three-dimensional spatial distribution of the

concentration field [3-4].

Another quantification method is measuring the fluorescent product of a chemical

reaction (reaction-type experiments). The simplest outcome of a reaction is the

formation of a coloured species and therefore, the intensity of the product is a

direct measure of the extent of mixing. Typically, this process is an acid–base

reaction with a dye having a fluorescence quantum yield that is pH-sensitive [5].

6.1.2. Numerical techniques

Most numerical methods are focused on solving a mass transport equation in order

to visualize and evaluate the mixing performance of micro-mixers. For this

purpose, massless non-interacting virtual particles or tracers are placed in the flow

field and their trajectories are computed using a Lagrangian method. This particle

tracing algorithm can be used to both visualize and evaluate the mixing

performance of the mixers. If the virtual tracers are distinguished by different


colours, interpreting the results of particle advection can lead to formation of

Poincare maps or used in other methods, which are based on plotting the position

of the tracers at various snapshots or downstream distances revealing the

evolution of the mixing patterns. Poincare sections, which relate information on

the chaotic nature of the flow, are generated by tracking the tracers through the

flow and recording their cross-sectional positions at different downstream

coordinates [5]. The cross-sectional positions are then superimposed to form a

two-dimensional plot. The disposition of the points in the Poincare map can be

used to study the chaotic nature of the flow. Figure 5.7 in the previous chapter,

which illustrated the advection of the particles and cells with two different

assigned colours, falls in this category as a simple example.

Quantifying the performance of the mixer, trajectories of the particles can also be

used to calculate Lyapunov exponents. In this case, the underlying assumption in

the interpretation of the results is that mixing can be achieved efficiently only in

chaotic flows and therefore, this method is appropriate for mixers based on the

chaotic advection. In fact Lyapunov exponents are used primarily to evaluate the

chaotic behaviour of the system and the strength of chaos can be related to the

mixing performance. Computation of the Lyapunov exponents is one of the

employed techniques in this study, which is dealt with in the next section.

An alternative approach is to measure the mixing using statistical analyses of the

concentration samples in the mixture such as information entropy or Shannon

entropy. The Shannon entropy (S) is the rigorous measure of the mixing and it has

been used in many different scientific areas [6]. The Shannon entropy is

determined from statistical properties where: (I) it depends on the probability

distribution p only; (II) the lowest entropy (S = 0) corresponds to one of the p

being 1 and the rest being zero (i.e., perfect order, complete segregation); (III) the

largest value for the entropy is achieved when all ps are equal to each other (i.e.,

complete disorder, perfect mixing). However, in order to obtain more accurate

results, a large number of particles needs to be tracked, which would require an


extremely long computational time. Figure 6.1 shows a diagram of various

methods including both experimental and numerical methods.

Figure 6.1 Block diagram of the various techniques employed for characterizing the micro-mixers.

6.2. Characterization methods used in this study

In order to quantitatively evaluate the degree of mixing, two criteria are computed

for the investigated range of simulation parameters. A common definition of

mixing quality is based on the inspection of chaotic regimes developed in the

mixer and calculating the Lyapunov exponent is a standard method of

investigating chaos. However, as discussed earlier, Lyapunov exponents are used

primarily to evaluate the chaotic behaviour of the system, but not the mixing

performance directly. Therefore, as a supplemental index, the ability of the system

to capture the target cells is introduced, which can represent the performance of


the mixer. Since the buffer is intrinsically particle laden and in fact, this is the

mixing of these particles which is of interest, the trajectories of magnetic particles

are used for calculation of characterizing indices.

6.2.1. Lyapunov exponents

As explained in chapter 4, sensitivity to initial conditions is an indication of chaos

related to fluid mixing. In chaotic systems, time evolution of two initially nearby

particles shows exponential divergence. Lyapunov exponent (or Lyapunov

characteristic exponent) is the average exponential rate of divergence or

convergence of initially neighbouring orbits in the phase space and is defined as:

t

d(t)1lim lnt d(0)→∞

λ = (6.1)

where d(t) and d(0) are distance between two orbits at time t and initial time,

respectively. Calculation of the Lyapunov exponent can be used to detect the

incidence of chaos, measure its extent, and investigate the relationships between

various affecting parameters and chaos. For chaotic mixing problems, Lyapunov

exponent reflects the dispersion rate of the fluid particles and in this study, it is

used to quantify the chaotic advection of magnetic particles.

Lyapunov exponents are defined as a spectrum with n components in an n-

dimensional phase space (the rate of divergence or convergence can be different

for different orientations of initial separation vector). However, normally it

suffices to consider its largest component to describe the system because it

determines the predictability of a dynamical system. A positive value is the

signature of chaos, while zero indicates stable properties. A negative value of λ is

an indication of dissipative systems. Whereas the (global) Lyapunov exponent

gives a measure for the total mixing of a system, it is sometimes interesting to

estimate the local behaviour around a specific point in the domain (these values

are usually called local Lyapunov exponents). When λ is positive in part of flow


regions (labelled as region ‘a’ in figure 6.2), while in other regions it is zero/near

zero (region ‘b’), the mixer is only partially chaotic and the mixing is incomplete.

Figure 6.2 Schematic illustration showing that the λ-map.

Generally the calculation of Lyapunov exponents cannot be carried out

analytically, and in most cases one must resort to numerical techniques. Here

Sprott’s method [7] is used to calculate the largest Lyapunov exponent (hereafter

λl). This method utilizes the general idea of tracking two initially close particles,

and calculates average logarithmic rate of separation of the two particles. Further

description of the Sprott’s method is given in Appendix B.

Figure 6.3 shows the convergence of λl versus time for one particle with two

different driving parameters and also in the absence of the external field. Without

magnetic perturbation, λl convergences to zero indicating a steady flow. At

Strouhal number of St=0.2 and bulk fluid velocity of V=35 μm/s, it convergences

to a constant value of about 0.45 whereas at velocity of 40 μm/s and the same

Strouhal number, this value is about 0.19. Hence, it can be deduced that in the

former operating condition, the system exhibits a stronger chaotic behaviour.


Close observation of the graphs reveals that there are some points where λl

declines quickly (marked with circles in the figure). Considering the time of these

incidents, it can be inferred that near the end of each phase, some particles enter

the centre of tips and become trapped in these regions yielding a zero value for λl

(for that specific time) and reducing the overall λl. When the driving current is

switched and system is in next half-cycle, particles are free to flow with the

mainstream up to the point where they are re-attracted by nearby active tooth

leading to an increase in λl. Examination of λl for various particles reveals that

regardless of system operating parameters, λl approaches its converged value

generally after a period of 20s of activation.

0 5 10 15 20 25 30 35 40 45 50-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time (s)

Lyap

unov

Exp

onen

t

St=0.2 , V=35 µm/sSt=0.2 , V=40 µm/sNo magnetic forceFrequent re-injection

Figure 6.3 Convergence of the largest Lyapunov exponent for one particle.

6.2.2. Cell capturing efficiency

In this section, a new supplemental index for characterization is introduced, which

can be implemented for this particular concept. Up to now, solely the chaotic

behaviour of magnetic particles has been discussed. Since the ultimate goal of the

system is to improve the attachment of cells to magnetic particles, efficiency of


capturing of the target cells can be another criterion for characterizing the mixer.

It offers the ability to examine the efficiency of the mixer versus chaotic advection

and also analyze the effect of driving parameters on the cell capturing ratio. This

method uses the idea of monitoring the trajectories of cells and particles to predict

their collision (if any) in the mixer domain.

Both particles and cells with a uniform distribution and the same concentration

(1015 particles/m3) enter the lower and upper halves of the channel, respectively. It

is assumed that collision happens when the distance between the centre of circular

particle and cell becomes smaller than the sum of their radii (as shown in figure

6.4) and then cell is attached to the particle due to chemical binding. Low flow

velocities in the order of 10 µm/s allow the cells to be in the proximity of the

particles long enough for the chemical binding to occur. Since reference particles

are larger than the cells in this study, more than one cell may be attached to a

single particle. In such occasions, once a cell is attached, it will be counted and

then assumed as removed, leaving the particle ready for another collision.

However, after every collision the trajectories of the particles must be re-

calculated using new free-body force diagram. Although the driving force is the

same for the cell/particle complex (magnetic force is applied merely on the

particles), the drag coefficients need to be modified according to the number of

the attached cells. Subsequently, Capturing Efficiency (CE), i.e., ratio of the

captured cells to the total number of entered cells, is calculated after a period of

20s of mixing.

Figure 6.4 Concept of the collision where cell is tagged by the magnetic particle. Rp and Rc are the

radius of the particle and cell, respectively.


6.2.3. Results and discussion

Both criteria are calculated for a wide range of simulation parameters. For

computation of the largest Lyapunov exponent, 21 particles are uniformly

distributed in upper half of the first mixing unit as the initial positions and λl is

calculated for each individual particle using the method explained in section 6.2.1.

The initial positions of the particles are shown in figure 6.5. The time period is

20s when the particles approach their constant value of λl. In order to quantify the

extent of chaos over the entire domain in the upper section (where cells exist), the

average of λls of 21 particles is taken.

Figure 6.5 Initial positions of the magnetic particle for computation of the largest Lyapunov

exponent.

Figure 6.6a illustrates variation of CE (capturing efficiency) for different driving

parameters (St=0.2-1) where each graph represents the values of CE for a constant

fluid velocity (V=30-50 μm/s). Results for λl calculated over the same range of

driving parameters are shown in figure 6.6b. When no external agitation is

applied, no penetration of particles into the cells’ region occurs and collision

happens merely at the interface between upper and lower halves of the flows.

Application of external forces leads to an increase in both indices. The global


variations of λl is almost identical for different bulk flow velocities; the maximum

chaos happens at St=0.4, while the minimum occurs at St=0.8. CE exhibits a

similar behaviour at Strouhal numbers less than 0.6, which means that an increase

in chaos leads to an increase in captured cells. Maximum values for λl and CE are

realized at St=0.4, which are 0.36 and 67%, respectively.

Figure 6.6 Variation of characterizing indices versus different system operating conditions: (a) cell

capturing efficiency, (b) largest Lyapunov exponent.

At higher Strouhal numbers (namely 0.8), two indices show different variations.

Although at high bulk flow velocities (larger than 40 μm/s) a good agreement

between two indices can still be observed, in the case of lower velocities they

show contradicting behaviours. At low velocities of the bulk flow, some particles


are advected until they are attracted to the centre of one tip in the upper conductor.

In the vicinity of the channel wall, flow velocity is much less than the central

region as the parabolic velocity profile in Poiseuille flow is developed in the

channel. Since the magnetic forces are significantly large in the centre of the

conductor, these particles will be retained in this area. Even after the current is

switched to the opposite array, due to low fluid velocity particles will not have the

opportunity to escape from the previous conductor and come close enough to the

opposite conductor. Therefore, in the next period of activation, particles are again

pulled back towards the same region quickly and become trapped again. For such

particles, mixer is partially acting like an asymptotically stable system which

results in a decrease in the Lyapunov exponent of the whole domain.

However, trapped particles play the role of nearly fixed posts which collect

multiple cells when they meet them, thereby increasing the value of CE. Although

the efficiency is relatively high, in practice, it is a challenging issue where trapped

particles can clog the channel. Figure 6.7 illustrates the trajectories of such

particles at Strouhal number of St=0.8 and bulk velocity of V=40 μm/s. Five

particles are released in the first unit and after flowing along three and a half units,

three particles (particles 1-3) are trapped and never exit the channel. Trajectories

of these particles are plotted with dotted lines and locations of traps are marked

with rectangles. In such operating conditions, the mixer is only partially chaotic,

and the mixing is incomplete. However, when the Strouhal number is low, i.e., in

case of longer time periods, particles have the chance to move away from these

attractors, even though the velocity is low.

Figure 6.7 Trajectories of particles at St=0.8 and V=40 µm/s; rectangles indicate the location of

trapped particles.


Some particles (here particles 4-5) seem to have nearly similar and close

trajectories. There are two possible scenarios. First, two particles may flow along

the same Lagrangian path but at different times. In this case these particles will

capture different cells if any collision happens. The other scenario occurs when

operating conditions force the particles to take almost the same trajectories at the

same time, despite their different initial positions.

Finally, cell capturing ratio is calculated for a mixing time of 30s which improves

the maximum efficiency up to 75%. However, longer mixing times require longer

channels. In fact, mixing length is a function of bulk fluid velocity as well as

mixing time. For instance, 30s of mixing process at bulk velocity of 45 μm/s can

take place through 5.5 mixing units (corresponding to CE=66%) while solely

three units are required for 20s of mixing at bulk velocity of 35 μm/s

(corresponding to CE=57%). Therefore, one needs to reach the best compromise

between efficiency and size of the design.

Chapter 7

Concluding Remarks and Future Work

7.1. Conclusions

In this thesis, a micro-mixing device for magnetic particles is designed and it is

shown that a straight channel with two embedded serpentine conductors beneath

the channel can be utilized to produce the chaotic pattern in the motion of

particles and intensify the capturing of biological cells. Two flows; bio-

cells/molecules suspension and the particle laden buffer, are introduced into the

channel and manipulated by pressure-driven flow. While the cells follow the

mainstream, the motion of magnetic particles is affected by both the surrounding

flow field and the localized time-dependent magnetic field generated by

sequential activation of two serpentine conductors.

Prior to numerical simulations, preliminary modelling is carried out to reach a

reasonably optimized geometrical structure for the conductors. It was found that a

compromise between the magnitude of the applied forces and the amount of

generated heat must be reached in determining the dimensions of the cross-

sectional area in the conductors. Subsequently, a two-dimensional numerical

study of the mixing process is performed in order to characterize the efficiency of

the micro-mixer. Although employed simulation techniques and developed codes

allow the evaluation of the effect of various geometry configurations and particle

characteristics, this study focuses on the effect of two driving parameters (i.e., the

fluid velocity and frequency of magnetic activation) on the mixing quality.

Chapter 7 – Concluding Remarks and Future Work 113

Outline of the simulation procedure is as follows:

Steady-state velocity field of an incompressible Newtonian fluid (water)

and time-dependent magnetic field were computed using commercial

multiphysics finite element package COMSOL.

Passive and active advection of the particles and cells (cells have only

passive advection) were extracted from the model in order to investigate

the motion of them in the mixing domain.

Trajectories of particles were evaluated using developed codes in Matlab.

Two indices which are highly dependent on the performance of the system

were investigated for a wide range of driving parameters (namely the bulk

flow velocity and frequency of the current), thereby characterizing the

mixer. Trajectories of the particles were used in order to:

o Detect chaos in their motion and quantify its extent by calculating

the Lyapunov exponents.

o Examine the capability of the system to capture target bio-cells (as

a supplemental index).

Based on the simulation results, optimum driving parameters were

concluded.

In the present configuration, the stretching and folding mechanism which

consecutively arises along the mixing channel is considered to be a major cause of

the chaotic behaviour. Lyapunov exponent as an index of the chaotic advection is

found to be highly dependent on the Strouhal number where the maximum chaotic

strength is realized in Strouhal numbers close to 0.4, which corresponds to the

Lyapunov exponent of 0.36. It is shown that capturing efficiency in the mixer

cannot be used as a stand alone index, which might suggest operating conditions

that are not practical. Therefore, both indices need to be taken into account while

characterizing the device. Maximum capturing efficiency is found to be 67%,

which means that more than half of the existing cells can be separated out of the


medium; although this could be further increased at the cost of longer mixing time

and channel length.

7.2. Recommendations for future research

This project was originally initiated with the interest whether the magnetic

particles can be arbitrary manipulated and mixed in a two-dimensional micro-flow

channel. Therefore, the main interest has been focused on the physical phenomena

and behaviour of magnetic particles. However, when one tries to use the present

design in the real applications, there are some issues which might be worth

considering.

7.2.1. Modified particle properties

One serious challenge in the proposed mixer can be the problem of Joule heating

which is directly influenced by the magnitude of the current injection. In order to

diminish the generated heat, polymeric particles with higher concentration of the

magnetic materials as the core can be used (i.e., particles with higher

permeability). This way, larger magnetic moments and subsequently, stronger

magnetic forces may be generated. Therefore keeping the force at the same level,

magnetic field can be minimized which in turn, reduces the required current

density. Moreover, it is an important issue from power consumption point of

view.

7.2.2. Three-dimensional mixing

This study investigated the mixing of the magnetic particles in a two-dimensional

design. In order to increase the output of the device, a three-dimensional mixing

can be a great step forward. However, as explained in the basic design section,

merely attractive forces can be applied through magnetic field and a repulsive

force which is required to move the particles in z-direction is lacking. One


possible option is to embed identical conductors in the top layer (cover layer) in

order to drag particles from top as well as bottom. Nevertheless, it is likely that

magnetic forces generated by the top conductors not to be strong enough to attract

the particles from a large distance at the bottom of the channel. In that case, one

can use soft magnetic materials such as permalloy which can be embedded in the

same way as the conductors. Utilizing a soft magnetic material outside the mixing

domain, it performs as a shield which can protect the field from being lost.

Therefore, magnetic field can be concentrated in the required space with a

significantly increased magnitude. Some basic numerical examinations reveal a 5-

fold increase in the field in case of an added permalloy.

7.2.3. Coupled simulations

As discussed in chapter 5 (simulation procedure), the investigation is based on the

premise that there is no hydrodynamic interactions between particles and

surrounding fluid and the problem is faced with by using one-way coupled

analysis. In fact, in this study the mixing of particles into the bio-fluid has been of

interest. However, if the buffers contain a high concentration of the particles, the

mutual interactions must be taken into account. Use of higher concentrations may

have two advantages. First, increased number of the particles in the same volume

can improve the ratio of the tagged target entities and therefore, enhance the

efficiency of the mixer. Second, in case of two- and four-coupled circumstances

the mixing of the buffers can also be attained. In fact, it would be possible to

utilize the magnetic particles as a mediator element to perform liquid-liquid

mixing whether there is any target cell to be collected or not. It is worth noting

that in some applications such extracting DNA molecules from the whole human

blood, the process of the liquid-liquid mixing is essential. In bench-top protocols,

a lysis buffer suspended with magnetic particles is mixed with the blood. As the

first step, the mixed lysis buffer will lyse the white blood cells and release DNA

molecules. In the following step, released DNA molecules can be tagged with the

magnetic particles.


As mentioned earlier, even if collection of the cells is not the aim of the mixer,

concentrated particles can be used for liquid-liquid mixing solely. Prior to mixing,

they can be loaded in any buffer easily and separated using simple magnetic

forces later in downstream. However, simulation of such coupled problems calls

for implementation of sophisticated mathematical methods and computational

facilities.

7.2.4. Experiments

Needless to say, the best method for the evaluation of the proposed mixer is

fabricating and conducting practical experiments. Standard micro-fabrication

techniques can be utilized to manufacture different layers of the device. The

fabrication involves micro-patterning of the serpentine conductors in thin films

and microfluidic channels, surface treatment to facilitate flow of particle laden

buffers in the micro-channels, packaging of the microfluidic chip and introduction

of fluid inlet/outlet. Conductors can be fabricated using micro-photolithography

and electroplating as other film deposition processes such as sputtering are

employed for deposition of thin films. For insulation of the conductors, PACVD

(Plasma Assisted Chemical Vapour Deposition) can be used to cover the surface

with silicon oxide. The initial materials for the microfluidic chip can be silicon

and glass so that anodic bonding can be employed for chip packaging. However,

the cover layer needs to be made of a transparent material, thereby allowing the

optical experiments to be conducted. Other materials such as plastic and PDMS

may be investigated using soft lithography and hot embossing techniques as

microfluidic chips based on polymer materials are low cost and potential for

disposable devices.

Once fabricated, the device can be evaluated through various techniques discussed

in the previous chapter for a reasonable range of operating parameters. One

interesting experiment may be recording the motion of the particles using PIV

(Particle Image Velocimetory) or PTV (Particle Tracking Velocimetory)

techniques. Particularly in chaotic systems, these experimentally obtained


trajectories of the particles can be used for computing analytical indices such as

Lyapunov exponents and performing visualization methods such as Poincare

maps in order to investigate the behaviour of the system. This way, the validity of

the models can be evaluated by comparing the numerical and experimental results.

References:

Chapter 1

[1] Pankhurst QA, Connolly J, Jones SK, Dobson J: Applications of magnetic

nanoparticles in biomedicine. J Phys D: Appl Phys 2003, 36:R167-R181.

[2] Safarik, I., and Safarikova, M. 1999. Use of magnetic techniques for the

isolation of cells. J. Chromatogr. B 722:33-53.

[3] Chalmers JJ, Zborowski M, Mandal S, Fang B. Theoretical analysis of cell

separation based on cell surface marker density. Biotechnol Bioeng

1998;59:10–20.

[4] Dynal Biotech, www.invitrogen.com

[5] Magnetism and microfluidics, Nicole Pamme, Lab Chip, 2006, 6, 24–38

[6] Hoyos M., Moore LR, McCloskey KE, Margel S, Zuberi M, Chalmers JJ,

Zborowski M. 2000. Study of magnetic particles pulse-injected into an

annular SPLITT-like channel inside a quadrupole magnetic field. J

Chromatogr A 903:99–116.

[7] PhD thesis - Hiroaki Suzuki, Development of Chaotic Micro-Mixer Using

Magnetic Beads, March 2003

[8] B. Sinclair, “To Bead or Not To Bead: Applications of Magnetic Bead

Technology,” The Scientist, Vol. 12, No. 13 (1998).

[9] Cell Separation - Basic Research Products, Dynal Biotech

Chapter 2

[1] PhD thesis, Pong Yu (Peter) Huang, Electro-Osmotic Mixing in

Microchannels, September, 2002, Brown University

[2] Frank M. White, Fluid Mechanics, Fourth Edition, , McGraw-Hill

[3] Michael Koch, Alan Evans, Arthur Brunnschweiler, Microfluidic Technology

and Applications, Taylor & Francis Group

[4] Jen C P et al 2003 Design and simulation of the micromixer with chaotic

advection in twisted microchannels Lab on a Chip 3 77–81

References 119

[5] A. E. KAMHOLZ, B. H. WEIGL, B. A. FINLAYSON, P. YAGER,

Quantitative Analysis of Molecular Interaction in a Microfluidic Channel:

The T-Sensor, Analytical Chemistry, Vol. 71, No. 23, pp. 5340-5347, 1999.

[6] RF Ismagilov, AD Stroock, PJA Kenis, G Whitesides 2000 Experimental and

theoretical scaling laws for transverse diffusive broadening in two-phase

laminar flows in microchannels Appl. Phys. Lett. 76 2376–78

[7] Nam-Trung Nguyen and Zhigang Wu, Micromixers-a review, J. Micromech.

Microeng. 15 (2005) R1–R16

[8] Yi M and Bau H H 2003 The kinematics of bend-induced mixing in micro-

conduits Int. J. Heat Fluid Flow 24 645–56

[9] Wong S H, Ward M C L and Wharton C W 2004 Micro T-mixer as a rapid

mixing micromixer Sensors Actuators B 100 365–85

[10] Wong S H et al 2003 Investigation of mixing in a cross-shaped micromixer

with static mixing elements for reaction kinetics studies Sensors Actuators B

95 414–24

[11] Gobby D, Angeli P and Gavriilidis A 2001 Mixing characteristics of T-type

microfluidic mixers J. Micromech. Microeng. 11 126–32

[12] Veenstra T T 1999 Characterization method for a new diffusion mixer

applicable in micro flow injection analysis systems J. Micromech. Microeng. 9

199–202

[13] Jackman R J et al 2001 Microfluidic systems with on-line UV detection

fabricated in photo-defineable epoxy J. Micromech. Microeng. 11 263–9

[14] Mobius et al 1995 A sensor controlled process in chemical microreactors Int.

Conf. on Solid-State Sensors and Actuators (Stockholm, Sweden) pp 775–8

[15] Koch M et al 1999 Improved characterization technique for micromixers J.

Micromech. Microeng. 9 156–8

[16] Haverkamp V et al 1999 The potential of micromixers for contacting of

disperse liquid phases Fresenius’ J. Anal. Chem. 364 617–24

[17] Fluri K et al 1996 Integrated capillary electrophoresis devices with an

efficient postcolumn reactor in planar quartz and glass chips Anal. Chem. 68

4285–90

[18] Hadd A G et al 1997 Microchip device for performing enzyme assays Anal.

Chem. 69 3407–12

[19] Jacobson S C, McKnight T E and Ramsey J M 1999 Microfluidic devices for

electrokinematically driven parallel and serial mixing Anal. Chem. 71 4455–9

References 120

[20] Bohm S, K. Greiner, S. Schlautmann, S. de Vries and A. van den Berg, A rapid

vortex micromixer for studying high-speed chemical reactions Technical

Proc. Micro Total Analysis Systems MicroTAS 2001 (Monterey, CA, USA) 25–7

[21] Che-Hsin Lin, Chien-Hsiung Tsai and Lung-Ming Fu, A rapid three-

dimensional vortex micromixer utilizing self-rotation effects under low

Reynolds number conditions, J. Micromech. Microeng. 15 (2005) 935–943

[22] Hardt S, Dietrich T, Freitag A, Hessel V, Lowe H, Hofmann C, Oroskar A,

Schonfeld F and Vanden B K 2002 Radial and tangential injection of

liquid/liquid and gas/liquid streams and focusing thereof in a special cyclone

mixer Proc. 6th Int. Conf. on Microreaction Technology pp 329–44

[23] Knight J B, Vishwanath A, Brody J P and Austin R H 1998 Hydrodynamic

focusing on a silicon chip: mixing nanoliters in microseconds Phys. Rev. Lett.

80 3863–6

[24] Jensen K 1998 Chemical kinetics: smaller, faster chemistry Nature 393 735–6

[25] Walker G M, Ozers M S and Beebe D J 2004 Cell infection within a

microfluidic device using virus gradients Sensors Actuators B 98 347–55

[26] S Hardt, KS Drese, V Hessel, F Schonfeld, Passive micromixers for

applications in the microreactor and μTAS fields, Microfluidics and

Nanofluidics, 2005 - Springer

[27] F. Schonfeld, V. Hessel and C. Hofmann, An optimised split-and recombine

micro-mixer with uniform ‘chaotic’ mixing, Lab Chip , 2004, 4, 65-69

[28] Branebjerg J et al 1996 Fast mixing by lamination Proc. MEMS’96, 9th IEEE

Int. Workshop Micro Electromechanical System (San Diego, CA) pp 441–6

[29] Schwesinger N et al 1996 A modular microfluid system with an integrated

micromixer J. Micromech. Microeng. 6 99–102

[30] Bing He, Brian J. Burke, Xiang Zhang, Roujian Zhang, and Fred E. Regnier 2001

A picoliter–volume mixer for microfluidic analytical systems Anal. Chem. 73

1942–7

[31] Jessica Melin, Guillem Giménez, Niclas Roxhed, Wouter van der Wijngaart and

Göran Stemme 2004 A fast passive and planar liquid sample micromixer Lab

on a Chip, in press

[32] Miyake R et al 1993 Micro mixer with fast diffusion Proc. MEMS’93, 6th IEEE

Int. Workshop Micro Electromechanical System (San Diego, CA) pp 248–53

References 121

[33] Larsen U D, Rong W and Telleman P 1999 Design of rapid micromixers using

CFD Proc. Transducers’99, 10th Int. Conf. on Solid-State Sensors and Actuators

(Sendai, Japan) pp 200–3

[34] Seidel R U et al 1999 Capilary force mixing device as sampling module for

chemical analysis Proc. Transducers’99, 10th Int. Conf. on Solid-State Sensors

and Actuators (Sendai, Japan) pp 438–41

[35] Voldman J, Gray M L and Schmidt M A 2000 An integrated liquid mixer/valve

J. Microelectromech. Syst. 9 295–302

[36] Hosokawa K, Fujii T and Endo I 1999 Droplet-based nano/picoliter mixer

using hydrophobic microcapillary vent Proc the IEEE International Workshop

Micro Electromechanical System (Piscataway, NJ, USA) pp 388–93

[37] Handique K and Burns M A 2001 Mathematical modeling of drop mixing in a

slit-type microchannel J. Micromech. Microeng. 11 548–54

[38] Paik P, Pamula V K and Fair R B 2003 Rapid droplet mixers for digital

microfluidic systems Lab on a Chip 3 253–9

[39] Song H et al 2003 Experimental test of scaling of mixing by chaotic advection

in droplets moving through microfluidic channels Appl. Phys. Lett. 83 4664–6

[40] Lin Y et al 2003 Ultrafast microfluidic mixer and freeze–quenching device

Anal. Chem. 75 5381–6

[41] Wang H et al 2002 Optimizing layout of obstacles for enhanced mixing in

microchannels Smart Mater. Struct. 11 662–7

[42] Wong S H et al 2003 Investigation of mixing in a cross-shaped micromixer

with static mixing elements for reaction kinetics studies Sensors Actuators B

95 414–24

[43] Mengeaud V, Josserand J and Girault H H 2002 Mixing processes in a zigzag

microchannel: finite element simulation and optical study Anal. Chem. 74

4279–86

[44] Hong C C, Choi J W and Ahn C H 2004 A novel in-plane microfluidic mixer

with modified tesla structures Lab on a Chip 4 109–13

[45] Liu R H et al 2000 Passive mixing in a three-dimensional serpentine

microchannel J. Microelectromech. Syst. 9 190–7

[46] Chen H and Meiners J C 2004 Topologic mixing on a microfluidic chip Appl.

Phys. Lett. 84 2193–5

[47] Park S J et al 2004 Rapid three-dimensional passive rotation micromixer

using the breakup process J. Micromech. Microeng. 14 6-14

References 122

[48] Vijiayendran et al 2003 Evaluation of a three-dimensional micromixer in a

surface-based biosensor Langmuir 19 1824–28

[49] Jen C P et al 2003 Design and simulation of the micromixer with chaotic

advection in twisted microchannels Lab on a Chip 3 77–81

[50] Johnson T J, Ross D and Locascio L E 2002 Rapid microfluidic mixing Anal.

Chem. 74 45–51

[51] Stroock A D et al 2002 Chaotic mixer for microchannels Science 295 647–51

[52] Stroock A D and Whitesides G M 2003 Controlling flows in microchannels

with patterned surface charge and topography Acc. Chem. Res. 36 597–604

[53] Stroock AD, Dertinger SKW, Whitesides GM, Ajdari A (2002) Patterning flows

using grooved surfaces. Anal Chem 74:5306– 5312

[54] Wang H et al 2003 Numerical investigation of mixing in microchannels with

patterned grooves J. Micromech. Microeng. 13 801–8

[55] Schonfeld F, Hardt S (2004) Simulation of helical flows in microchannels.

AIChE J 50: 771–778

[56] Aubin J, Fletcher DF, Bertrand J, Xuereb C (2003) Characterization of the

mixing quality in micromixers. Chem Eng Technol 26:1262–1270

[57] Lim D et al 2003 Fabrication of microfluidic mixers and artificial

vasculatures using a high-brightness diode-pumped Nd:YAG laser direct

write method Lab on a Chip 3 318–23

[58] Kim D S et al 2004 A barrier embedded chaotic micromixer, J. Micromech.

Microeng. 14 798–805

[59] Bertsch A et al 2001 Static micromixers based on large-scale industrial mixer

geometry Lab on a Chip 1 56–60

[60] Lu, Kee Suk Ryu, and Chang Liu, Liang-Hsuan, A Magnetic Microstirrer and

Array for Microfluidic Mixing, JOURNAL OF

MICROELECTROMECHANICAL SYSTEMS, VOL. 11, NO. 5, OCTOBER

2002

[61] K. S. Ryu, K. Shaikh, E. Goluch, Z. Fan, and C. Liu, “Micro magnetic stir-bar

mixer integrated with parylene microfluidic channels,” Lab on a Chip, vol. 4,

no. 6, pp. 608–613, 2004.

[62] G. A. Mensing, T. M. Pearce, M. D. Graham, and D. J. Beebe, “An externally

driven magnetic microstirrer,” Phil. Trans.: Math., Phys. Eng. Sci., vol. 362,

pp. 1059–1068, Mar. 11, 2004.

References 123

[63] Hessel, Holger Lowe, Friedhelm Schonfeld Volker, Micromixers-a review on

passive and active mixing principles, Chemical Engineering Science 60 (2005)

2479 – 2501

[64] Deshmukh A A, Liepmann D and Pisano A P 2000 Continuous micromixer

with pulsatile micropumps Technical Digest of the IEEE Solid State Sensor and

Actuator Workshop (Hilton Head Island, SC) pp 73–6

[65] Deshmukh A A, Liepmann D and Pisano A P 2001 Characterization of a

micro-mixing, pumping, and valving system Proc. Transducers’01, 11th Int.

Conf. on Solid-State Sensors and Actuators (Munich, Germany) pp 779–82

[66] Fujii et al 2003 A plug and play microfluidic device Lab on a Chip 3 193–7

[67] Glasgow I and Aubry N 2003 Enhancement of microfluidic mixing using time

pulsing Lab on a Chip 3 114–20

[68] Niu X Z and Lee Y K 2003 Efficient spatial–temporal chaotic mixing in

microchannels J. Micromech. Microeng. 13 454–62

[69] Okkels F and Tabeling P 2004 Spatiotemporal resonances in mixing of open

viscous fluids Phys. Rev. Lett. 92 038301

[70] Moroney R M, White R M and Howe R T 1991 Ultrasonically induced

microtransport Proc. MEMS’91, 3th IEEE Int. Workshop Micro

Electromechanical System (Nara, Japan) pp 277–82

[71] Rife J C et al 2000 Miniature valveless ultrasonic pumps and mixers Sensors

Actuators A 86 135–40

[72] Vivek V, Zeng Y and Kim E S 2000 Novel acoustic-wave micromixer Proc.

MEMS’00, 13th IEEE Int. Workshop Micro Electromechanical System

(Miyazaci, Japan) pp 668–73

[73] Yasuda K 2000 Non-destructive, non-contact handling method for

biomaterials in micro-chamber by ultrasound Sensors Actuators B 64 128–35

[74] Liu R H et al 2002 Bubble-induced acoustic micromixing Lab on a Chip 2

151–7

[75] Liu R H et al 2003 Hybridization enhancement using cavitation

microstreaming Anal. Chem. 75 1911–7

[76] Yang Z et al 2000 Active micromixer for microfluidic systems using lead-

zirconate-ttitanate (PZT)-generated ultrasonic vibration Electrophoresis 21

116–9

[77] Yang Z et al 2001 Ultrasonic micromixer for microfluidic systems Sensors

Actuators A 93 266–72

References 124

[78] Yaralioglu G G et al 2004 Ultrasonic mixing in microfluidic channels using

integrated transducers Anal. Chem. 76 3694–8

[79] Mao H, Yang T and Cremer P S 2002 A microfluidic device with a linear

temperature gradient for parallel and combinatorial measurements J. Am.

Chem. Soc. 124 4432–5

[80] Tsai J H and Lin L 2002 Active microfluidic mixer and gas bubble filter

driven by thermal bubble pump Sensors Actuators A 97–98 665–71

[81] Ahmed Ould El Moctar, Nadine Aubry and John Batton, Electro-hydrodynamic

micro-fluidic mixer, Lab Chip, 2003, 3, 273–280

[82] C. Tsouris, C. T. Culbertson, D. W. DePaoli, S. C. Jacobson, V. F. de Almeida,

and J. M. Ramsey, Electrohydrodynamic Mixing in Microchannels, AICHE J.,

49 (8), 2003

[83] Bau H H, Zhong J and Yi M 2001 A minute magneto hydro dynamic (MHD)

mixer Sensors Actuators B 79 207–15

[84] Che-Hsin Lin, Lung-Ming Fu, and Yu-Sheng Chien, Microfluidic T-Form

Mixer Utilizing Switching Electroosmotic Flow, Anal. Chem. 2004, 76, 5265-

5272

[85] Lung-Ming Fu, Ruey-Jen Yang, Che-Hsin Lin, Yu-Sheng Chien, A novel

microfluidic mixer utilizing electrokinetic driving forces under low switching

frequency, Electrophoresis 2005, 5, 1814–1824

[86] Yanting Zhang, David Follman, Han Chen, Igor Mezic, Carl Meinhart, Linda R.

Petzold and Noel C. MacDonald, DESIGN AND FABRICATION OF A RING

ELECTROKINETIC CHAOTIC MICROMIXER (RECM) WITH

INTEGRATED ELECTRODES, Proceedings of IMECE: 2003 ASME

INTERNATIONAL MECHANICAL ENGINEERING CONGRESS AND

RD&D EXPO, Washington, D.C., USA, November 15-21, 2003

[87] Han Chen and Yanting Zhang, NUMERICAL SIMULATION OF AN

ELECTROOSMOTIC MICROMIXER, Proceedings of IMECE: 2003 ASME

INTERNATIONAL MECHANICAL ENGINEERING CONGRESS AND

RD&D EXPO, Washington, D.C., USA, November 15-21, 2003

[88] Naoki Sasaki, Takehiko Kitamori and Haeng-Boo Kim, AC electroosmotic

micromixer for chemical processing in a microchannel, Lab Chip, 2006, 6,

550–554

References 125

[89] Zhongliang Tang, Seungbae Hong, Djordje Djukic,Vijay Modi, Alan CWest,

JamesYardley and Richard M Osgood, 2002 Electrokinetic flow control for

composition modulation in a microchannel J. Micromech. Microeng. 12 870–7

[90] Hsin-Yu Wu and Cheng-Hsien Liu, A novel electrokinetic micromixer, Sensors

and Actuators A 118 (2005) 107–115

[91] Oddy M H, Santiago J G and Mikkelsen J C 2001 Electrokinetic instability

micromixing Anal. Chem. 73 5822–32

[92] Wong, P. K.; Wang, T. H.; Deval, J. H.; Ho, C. M., Electrokinetics in Micro

Devices for Biotechnology Applications. IEEE/ASME Trans. Mechantronic

2004, 9, 366-376.

[93] Deval J, Tabeling P and Ho C M 2002 A dielectrophoretic chaotic mixer Proc.

MEMS’02, 15th IEEE Int. Workshop Micro Electromechanical System (Las

Vegas, Nevada) pp 36–9

[94] Lee Y K et al 2001 Chaotic mixing in electrokinetically and pressure driven

micro flows Proc. MEMS’01, 14th IEEE Int. Workshop Micro

Electromechanical System (Interlaken, Switzerland) pp 483–6

[95] Suzuki H and Ho C M 2002 A magnetic force driven chaotic

micro-mixer Proc. MEMS’02, 15th IEEE Int. Workshop Micro

Electromechanical System (Las Vegas, Nevada) pp 40–3

[96] Hiroaki Suzuki, Nobuhide Kasagi, CHAOTIC MIXING OF MAGNETIC

BEADS IN MICRO CELL SEPARATOR, , Proc. 3rd Int. Symp. Turbulence

and Shear Flow Phenomena Sendai, Japan, June 24-27, 2003

[97] Hiroaki SUZUKI, Masanori NAKANO, Nobuhide KASAGI, Chih-Ming HO,

PARTICLE TRACKING VELOCIMETRY MEASUREMENT OF

CHAOTIC MIXING IN A MICRO MIXER, , , Int. Symp. Micro-Mechanical

Engineering (ISMME2003), Tsuchiura, 1-3 December 2003

[98] Hiroaki Suzuki, Chih-Ming Ho, and Nobuhide Kasagi, A Chaotic Mixer for

Magnetic Bead-Based Micro Cell Sorter, JOURNAL OF

MICROELECTROMECHANICAL SYSTEMS, VOL. 13, NO. 5, OCTOBER

2004

[99] R. Rong, J. W. Choi, and C. H. Ahn, A NOVEL MAGNETIC CHAOTIC

MIXER FOR IN-FLOW MIXING OF MAGNETIC BEADS, 7th

International Conference on Miniaturized Chemical and Biochemical Analysts

Systems October 5-9, 2003, Squaw Valley, California USA

References 126

[100] A. Rida, T. Lehnert and M. A. M. Gijs, Microfluidic mixer using magnetic

beads, 7th International Conference on Miniaturized Chemical and Biochemical

Analysts Systems October 5-9, 2003, Squaw Valley, California USA

[101] Martin A. M. Gijs, Magnetic bead handling on-chip: new opportunities for

analytical applications, Microfluid Nanofluid (2004) 1: 22–40

Chapter 3

[1] B. I. Bleaney, B. Bleaney 1976 Electricity and Magnetism, Oxford University

Press

[2] I. S. Grant, W. R. Phillips 1994 Electromagnetism, John Wiley & Sons Ltd

[3] J.D. Jackson 1999 Classical Electrodynamics, John Wiley & Sons Ltd

[4] Pankhurst QA, Connolly J, Jones SK, Dobson J: Applications of magnetic

nanoparticles in biomedicine. J Phys D: Appl Phys 2003, 36:R167-R181.

[5] C. M. Sorensen, Nanoscale Materials in Chemistry (Ed.: K. J. Klabunde), John

Wiley and Sons, Inc., New York, 2001, pp. 169.

[6] L. A. Harris, Polymer Stabilized Magnetite Nanoparticles and Poly(propylene

oxide) Modified Styrene-Dimethacrylate Networks, Ph.D. thesis, Virginia

Polytechnic Institute and State University, 2002

[7] M. Ozaki, in Fine Particles: Synthesis, Characterization, and Mechanisms of

Growth, Vol. 92 (Ed.: T. Sugimoto), Marcel Dekker, Inc., New York, 2000.

[8] T. B. Jones 1995 Electromechanics of Particles, Cambridge University Press

[9] C. I. Mikkelsen. Magnetic Separation and Hydrodynamic Interactions in

Microfluidic Systems. Ph.D. thesis, Technical University of Denmark, 2005.

Chapter 4

[1] B. Wagner and W. Benecke, Microsystem Technologies 90, Berlin, p. 838, H

Reichl (Ed.), Springer Verlag, 1990.

[2] C. Yan, X. Zhao, G. Ding, C. Zhang, and B. Cai, J. Micromech. Microeng. 11

(2001) 113

[3] T. Christenson, T. Garino, E. Venturini, and D. Berry, TRANSDUCERS '99

(1999) 98

[4] T. Christenson, T. Garino and E. Venturini, HARMST, Kisarazu, Japan (1999)

82

References 127

[5] B. Dutoit, P.A. Besse, H. Blanchard, L. Guérin, and R.S. Popovic, Sensors &

Actuators A77 (1999) 178

[6] E. Kallenbach et al, ACTUATOR 2000 (2000) 611

[7] Z.C. Yuan, A.J. Williams, T.C. Shields, C.B. Ponton, J.S. Abell, and I.R. Harris,

8th International Conference on Ferrites, Kyoto, Japan (2000) 35

[8] B. Pawlowski, A. Rahmig, and J. Töpfer, 15th International Workshop on Rare

Earth Magnets and their Applications, Dresden, Germany (1998) 1045

[9] N.H. Hai, N.M. Dempsey, M. Veron, M. Verdier and D. Givord, J. Magn. Magn.

Mater. 257 (2003) L139

[10] F. Ingwiller, N.M. Dempsey, D. Givord, M. Veron, M. Verdier, O. Gutfleisch

and K. Khlopkov, These Proceedings

[11] R.A. Overfelt, C.D. Anderson and W.F. Flanagan, Appl. Phys. Lett. 49 (1986)

1799

[12] J. Wyslocki, J. Magn. Mag. Mat. 104-107 (1992) 363

[13] G. Rieger et al, J. Appl. Phys. 87-9 (2000) 5329

[14] H.J. Cho and C.H. Ahn, Electro-Chemical Soc.Proc. ECS'00 (2000) 586

[15] P. Evans, I. Zana, and G. Zangari, Electro-Chemical Soc.Proc. ECS'00 (2000)

610

[16] T.M. Liakopoulos, W. Zhang, and C.H. Ahn, MEMS (1996) 79

[17] G. Zangari, P. Bucher, N. Lecis, P.L. Cavallotti, L. Callegaro, and E. Puppin, J.

Mag. Mag. Mat. 157/158 (1996) 256

[18] Zangari and I. Zana, These Proceedings

[19] .A. Kapitanov, N.V. Kornilov, Ya.L Linetsky, and V.Yu. Tsvetkov, J. Magn.

Magn. Mater. 127 (1993) 289

[20] a.L. Linetsky and N.V. Kornilov, J. of Materials Engineering and Performance

4(2) (1995) 188-195

[21] V. Kornilov, Ph.D thesis: “The structure and magnetic properties of sputtered

magnetically-anisotropic Nd-Fe-B films”, Moscow Institute of Steel and Alloys

(1993) (in Russian)

[22] Yamashita, J. Yamasaki, M. Ikeda and N. Iwabuchi, J. Appl. Phys. 70 (1991)

6627

[23] adieu, 17th International Workshop on Rare Earth Magnets and their

Applications, Delaware, USA (2002) 416

[24] .J. Cadieu, R. Rani, X.R. Qian, and Li Chen, J. Appl. Phys. 83-11 (1998) 6247

[25] .J. Yang, S.W. Kim, and J.S. Kang, J. Appl. Phys. 83 (1998) 6620

References 128

[26] . Nakano et al., 17th International Workshop on Rare Earth Magnets and their

Applications, Delaware, USA (2002) 457

[27] Jin-Woo Choi, Chong H. Ahn, Shekhar Bhansali, H. Thurman Henderson, A new

magnetic bead-based, filterless bio-separator with planar electromagnet

surfaces for integrated bio-detection systems, Sensors and Actuators B 68

2000.34–39

[28] Kristian Smistrupa, Peter T. Tang, Ole Hansen, Mikkel F. Hansen,

Microelectromagnet for magnetic manipulation in lab-on-a-chip systems,

Journal of Magnetism and Magnetic Materials 300 (2006) 418–426

[29] Qasem Ramadan, Victor Samper, Daniel Poenar, Chen Yu, Magnetic-based

microfluidic platform for biomolecular separation, Biomed Microdevices

(2006) 8: 151–158

[30] Qasem Ramadana, Victor Samperb, Daniel Poenar, Chen Yu, On-chip micro-

electromagnets for magnetic-based bio-molecules separation, Journal of

Magnetism and Magnetic Materials 281 (2004) 150–172

[31] Fonnum G, Johansson C, Molteberg A, Morup S, Aksnes E (2005)

Characterisation of Dynabeads by magnetization measurements and

Mossbauer spectroscopy. J Magn Magn Mater 293:41–47

[32] T. Kapitaniak, S. R. Bishop, The Illustrated Dictionary of Nonlinear

Dynamics and Chaos, John Wiley & Sons, 1998, Surrey, UK

[33] Julyan H. E. Cartwright, Mario Feingold, and Oreste Piro, An Introduction to

Chaotic Advection, in mxing: chaos and turbulence, Ed. By H. Chate and E.

Villermaux (Kluwer, Dordrecht, 1999)

[34] Mark A. Stremler, F. R. Haselton and Hassan Aref, Designing for chaos:

applications of chaotic advection at the microscale, Phil. Trans. R. Soc. Lond.

A (2004) 362, 1019–1036

[35] Aref, H., 1984, “Stirring by Chaotic Advection,” J. Fluid Mech., 143, pp. 1-21.

[36] Ottino, J. M., 1989, “The Kinematics of Mixing: Stretching, Chaos, and

Transport,” Cambridge Univ. Press.

[37] Deval J, Tabeling P and Ho C M 2002 A dielectrophoretic chaotic mixer Proc.

MEMS’02, 15th IEEE Int. Workshop Micro Electromechanical System (Las

Vegas, Nevada) pp 36–9

References 129

Chapter 5

[1] Clayton T. Crowe, Multiphase Flow Handbook, , CRC Press, Boca Raton, 2006

[2] Mikkelsen C, Bruus H (2005) Microfluidic capturing-dynamics of

paramagnetic bead suspensions. Lab Chip 5:1293–1297

[3] Stokes, G. G., On the effect of the internal friction of fluids on the motion of

pendulums, Trans. Cam. Phil. Soc. 9 (1851) 8

[4] Coulson JM, Richardson JF, Backhurst JR, Harker JH (1991) Chemical

Engineering: Particle Technology and Separation Processes vol 2.

Pergammon Press Ltd, Oxford

Chapter 6

[1] Wolfgang Ehrfeld, Klaus Golbig, Volker Hessel, Holger Lo1we, and Thomas

Richter, Characterization of Mixing in Micromixers by a Test Reaction:

Single Mixing Units and Mixer Arrays, Ind. Eng. Chem. Res. 1999, 38, 1075-

1082,

[2] Liu R H et al 2000 Passive mixing in a three-dimensional serpentine

microchannel J. Microelectromech. Syst. 9 190–7

[3] Ismagilov R F et al 2000 Experimental and theoretical scaling laws for

transverse diffusive broadening in two-phase laminar flows in

microchannels Appl. Phys. Lett. 76 2376–78

[4] Knight J B, Vishwanath A, Brody J P and Austin R H 1998 Hydrodynamic

focusing on a silicon chip: mixing nanoliters in microseconds Phys. Rev. Lett.

80 3863–6

[5] Aubin, J., D. F. Fletcher, and C. Xuereb, “Characterization of Mixing Quality

in Micromixers,” Chem. Eng. and Technol., 26(12), 1262 (2003).

[6] Marco Camesasca, Ica Manas-Zloczower and Miron Kaufman, Entropic

characterization of mixing in microchannels, J. Micromech. Microeng. 15

(2005) 2038–2044

[7] Sprott JC (2003) Chaos and Time-Series Analysis. Oxford University Press,

Oxford

Appendix A

COMSOL Multiphysics Simulation

This appendix explains the COMSOL Multiphysics package for simulations and

how to use it for both electromagnetic and fluid field problems. When creating a

model in COMSOL Multiphysics, the typical modelling steps include:

1. Creating or importing the geometry

2. Meshing the geometry

3. Defining the physics on the domains and at the boundaries

4. Solving the model

5. Post-processing the solution

Among the above list, solving the model is one of the most crucial steps.

Therefore, the issue of solving a model in the COMSOL is explained in the

following (the main part of this appendix is based on the COMSOL 3.2

documentation).

A.1. Solver Overview

The COMSOL Multiphysics Solvers

COMSOL Multiphysics includes a set of solvers for PDE-based problems and

depending on the type of the problem, the proper set must be chosen. Table A.1

summarizes the available types:

Appendix A- COMSOL Multiphysics Simulation 131

Table A.1 Available types of solvers in COMSOL.

SOLVER USAGE

Stationary linear solver

For linear or linearized stationary PDE problems

Stationary nonlinear solver

For nonlinear stationary PDE problems

Time-dependent solver

For time-dependent PDE problems (linear or nonlinear)

Eigenvalue solver For eigenvalue PDE problems

Parametric linear solver

For linear stationary PDE problems depending on a parameter

Parametric nonlinear solver

For nonlinear stationary PDE problems depending on a parameter

Adaptive solver For stationary (linear or nonlinear) or eigenvalue PDE problems using adaptive mesh refinement

Selecting an Analysis Type

Many application modes suggest a set of analysis types as an application mode

property. The possible analysis types vary with the application area. Typical types

you can expect to see include stationary, eigenfrequency, transient, time-

dependent, and parametric analyses.

You can choose the analysis type when selecting most application modes in the

Model Navigator. To change the analysis type later on, choose Properties from the

Physics menu to open the Application Mode Properties dialog box. The analysis

types often set up various equations in the application mode, and the type also

suggests a default solver. To override this default, go to the Solver Parameters

dialog box.

For multiphysics models, the ruling application mode determines the analysis type

and it also suggests a default solver. When you change the ruling application

mode or its analysis type in the Model Navigator, the program changes certain

solver settings accordingly.


In addition to the analysis type, other application mode properties include the

default element type, and weak constraints. The weak constraint property enables

more-accurate reaction-force computations by introducing an extra equation,

which in turn means extra work for the solver.

Selecting a Solver

The analysis type generally selects an appropriate solver, so normally it is not

necessary to select one yourself. If you do prefer to make a selection, the first

question to ask is whether the problem is stationary or time-dependent. Most real-

world phenomena develop in time, but you might know that a given solution

approaches a stationary value.

A.2. Modelling Electromagnetics

This section explains the application modes in COMSOL Multiphysics for

electromagnetics and how to use them for electromagnetic field simulations.

Fundamentals of Electromagnetics

The problem of electromagnetic analysis is that of solving Maxwell’s equations

subject to certain boundary conditions. Maxwell’s equations are a set of

equations, written in differential or integral form, stating the relationships between

the fundamental electromagnetic quantities. These quantities are:

• The electric field intensity, E

• The electric displacement or electric flux density, D

• The magnetic field intensity, H

• The magnetic flux density, B

• The current density, J

• The electric charge density, ρ


You can formulate the equations in differential or integral form. This discussion

presents them in differential form because it leads to differential equations that the

finite element method can be handle. For general time-varying fields, Maxwell’s

equations are:

The first two equations are also referred to as Maxwell-Ampère’s law and

Faraday’s law, respectively. The last two are forms of Gauss’ law in the electric

and magnetic form, respectively. Another fundamental relationship is the equation

of continuity:

Out of these five equations only three are independent. The first two combined

with either the electric form of Gauss’ law or the equation of continuity form an

independent system.

Constitutive Relationships

To obtain a closed system, you need the constitutive relationships describing the

macroscopic properties of the medium. They are:

where ε0 is the permittivity of vacuum, μ0 is the permeability of vacuum, and σ is

the electrical conductivity. In the SI system the permeability of a vacuum is

4π×10-7 H/m.


Potentials

Under certain circumstances it can be helpful to formulate a problem in terms of

the electric scalar potential V and magnetic vector potential A. They are given by

the equalities:

which are direct consequences of the magnetic case of Gauss’ law and Faraday’s

law, respectively.

Material Properties

This discussion has so far only formally introduced the constitutive relationships.

These seemingly simple relationships can be quite complicated at times. In fact,

these relationships require some special considerations when working with four

main groups of materials:

Inhomogeneous materials

Anisotropic materials

Nonlinear materials

A material can belong to one or more of these groups. Inhomogeneous materials

are the least complicated. An inhomogeneous medium is one in which the

constitutive parameters vary with the space coordinates so that different field

properties prevail at different parts of the material structure.

For anisotropic materials the field relationships at any point differ for different

directions of propagation. This means that a 3x3 tensor is necessary to properly

define the constitutive relationships. If this tensor is symmetric, the material is

often referred to as reciprocal. In these cases you can rotate the coordinate system


such that a diagonal matrix results. If two of the diagonal entries are equal, the

material is uni-axially anisotropic; if none of the elements have the same value,

the material is bi-axially anisotropic.

In some nonlinear materials the permittivity or permeability depend on the

intensity of the electromagnetic field. Nonlinearity also includes hysteresis effects

where not only the existing field intensities influence a material’s physical

properties but the history of the field distribution also plays a role.

Boundary and Interface Conditions

To get a full description of an electromagnetic problem, you must also specify

boundary conditions at material interfaces and physical boundaries. At interfaces

between two media, you can mathematically express the boundary conditions as:

where and denote the surface charge density and surface current density,

respectively, and n2 is the outward normal from medium 2. Of these four

equations, only two are independent. This is an overdetermined system of

equations, so you would like to reduce it. First select either equation one or

equation four. The select either equation two or equation three. Together these

selections form a set of two independent conditions. From these relationships, you

can derive the interface condition for the current density,

Electromagnetic Quantities

The table below shows the symbol and SI unit for most of the physical quantities

that appear in the Electromagnetics Module. Although COMSOL Multiphysics


supports other unit systems, the equations in the Electromagnetics Module are

written for SI units.

Table A.2 Electromagnetic Quantities in COMSOL.

QUANTITY SYMBOL UNIT ABBREVIATION

Angular frequency ω radian/second rad/s

Attenuation constant α meter-1 m-1

Capacitance C farad F

Charge q coulomb C

Charge density (surface) ρs coulomb/meter2 C/m2

Charge density (volume) ρ coulomb/meter3 C/m3

Current I ampere A

Current density (surface) Js ampere/meter A/m

Current density (volume) J ampere/meter2 A/m2

Electric displacement D coulomb/meter2 C/m2

Electric field E volt/meter V/m

Electric potential V volt V

Electric susceptibility χe (dimensionless) -

Electrical conductivity σ siemens/meter S/m

Energy density W joule/meter3 J/m3

Force F newton N

Frequency ν hertz Hz

Impedance Z, η ohm Ω Inductance L henry H

Magnetic field H ampere/meter A/m

Magnetic flux F weber Wb

Magnetic flux density B tesla T

Magnetic potential (scalar) Vm ampere A

Magnetic potential (vector) A weber/meter Wb/m

Magnetic susceptibility χm (dimensionless) - Magnetization M ampere/meter A/m

Permeability μ henry/meter H/m

Permittivity ε farad/meter F/m

Polarization P coulomb/meter2 C/m2

Poynting vector S watt/meter2 W/m2


Propagation constant β radian/meter rad/m

Reactance X ohm Ω

Relative permeability μr (dimensionless) -

Relative permittivity εr (dimensionless) -

Resistance R ohm Ω Resistive loss Q watt/meter3 W/m3

Torque T newton-meter N·m

Velocity v meter/second m/s

Wavelength λ meter m

Wave number k radian/meter rad/m

A.3. Fluid Mechanics

This section explains how to use the Incompressible Navier-Stokes application

mode for the modeling and simulation of fluid mechanics and fluid statics.

Navier-Stokes Application Mode

When studying liquid flows, it is often safe to assume that the material’s density is

constant or almost constant. You then have an incompressible fluid flow. Using

the Incompressible Navier-Stokes application mode you can solve transient and

steady-state models of incompressible fluid dynamics.

Variables and Space Dimension

The Incompressible Navier-Stokes application mode solves for the pressure p and

the velocity vector components. It is available for 2D, 2D axisymmetric, and 3D

geometries.

PDE Formulation and Equations

Use the Incompressible Navier-Stokes application mode to model incompressible

flow in fluids. The Navier-Stokes equations for fluid flow,


are deduced for incompressible Newtonian flow. However, for both Cartesian and

axisymmetric coordinates, COMSOL Multiphysics uses a generalized version of

the Navier-Stokes equations to allow for variable viscosity (non-Newtonian

fluids). Starting with the momentum balance in terms of stresses, the generalized

equations in terms of transport properties and velocity gradients are:

The first equation is the momentum balance, and the second is the equation of

continuity for incompressible fluids. The following variables and parameters

appear in the equations:

• η is the dynamic viscosity.

• ρ is the density.

• u is the velocity field.

• p is the pressure.

• F is a volume force field such as gravity.

These application modes are general enough to account for all types of

incompressible flow. In practice, though, successful analysis of turbulent flows

requires simplifications of the description of transport of momentum.

Subdomain Settings

The subdomain quantities are listed in table A.3.


Table A.3 Subdomain quantities.

PARAMETER VARIABLE DESCRIPTION

ρ rho Density

η eta Dynamic viscosity

F F Volume force

Density

This material property specifies the fluid density.

Dynamic Viscosity

This term describes the relationship between the shear stresses in a fluid and the

shear rate. Intuitively, water and air have a low viscosity, and substances often

described as thick, such as oil, have a higher viscosity. You can describe a non-

Newtonian fluid by defining a shear-rate dependent viscosity.

Boundary Conditions

The boundary conditions for the Incompressible Navier-Stokes application mode

are:

Table A.4 Boundary conditions, Navier-Stokes equations..

BOUNDARY CONDITION DESCRIPTION

u = u0 = (u0, v0, w0) Inflow/Outflow velocity

T = -p0n Outflow/Pressure (total stress tensor)

p = p0, K = 0 Outflow/Pressure (viscous stress tensor)

Slip/Symmetry, 2D

Slip/Symmetry, 3D

u = 0 No slip


Normal flow/Pressure, 2D (total stress tensor)

Normal flow/Pressure, 2D (viscous stress tensor)

Normal flow/Pressure, 3D (total stress tensor)

Normal flow/Pressure, 3D (viscous stress tensor)

T = 0 Neutral (total stress tensor)

K = 0 Neutral (viscous stress tensor)

u = 0, K = 0 Axial symmetry

Inflow/Outflow velocity

At an inflow or outflow boundary you can specify the fluid’s velocity field as

in 3D; in the 2D case, drop the last component.

Outflow or Pressure

Using the total stress tensor form, this boundary condition states that the total

force on the boundary is a pressure force defined by p0:

Using the viscous stress tensor form, the pressure is set to p0, and the viscous

force is zero:


This means that the implementation of the outflow condition for the total stress

sensor form uses a Neumann boundary condition, which provides a better-posed

problem than the Dirichlet boundary condition in the viscous stress tensor form.

Appendix B

Calculation of the largest Lyapunov exponent

Sprott’s method utilizes the general idea of tracking two initially close particles,

and calculates average logarithmic rate of separation of the two particles. The

numerical procedure is shown in figure.

Schematic illustration for calculating the largest Lyapunov exponent.

For any arbitrary particle, a virtual particle is considered with a minute distance of

d(0) from the chosen particle and trajectories of these particles are tracked. At the

end of each time-step, the new distance, d(t), between real and virtual particles

and also the value of ln⎪d(t)/d(0)⎪ are calculated. The virtual particle is then

placed at distance d(0) along its connecting line to the real particle (procedure

known as Gram-Schmit Reorthonormalization, GSR). After repeating this process

for several time-steps, λl will be converged and is evaluated by:

→∞ =λ =

Δ ∑n

i1

n i 1

d (t)1lim lnn t d(0)

where Δt is the duration of one time-step and n is the number of steps. By using

this algorithm, the orientation of the orbit is kept along the direction of maximum

Appendix A- Calculation of the largest Lyapunov exponent 143

expansion and therefore, the largest Lyapunov exponent is achieved. An

appropriate choice of d(0) is one that is about 1000 times larger than the precision

of the floating point numbers that are being used. Therefore, a value of 0.01 μm

will suffice for initial distance. This algorithm, however, is shown to be robust to

any choice of d(0) and the frequency of normalization.

Appendix C

PUBLICATIONS

1. M. Zolgharni, S. M. Azimi, M. R. Bahmanyar, W. Balachandran

A numerical design study of chaotic mixing of magnetic particles in a microfluidic bio-separator, Journal of Microfluidics and Nanofluidics (2006) in-press

2. M. Zolgharni, S. M. Azimi, H. Ayers, W. Balachandran

Labelling of Biological Cells with Magnetic Particles in a Chaotic Microfluidic Mixer, 2nd Annual IEEE Int. Conf. on Nano/Micro Engineered and Molecular Systems (IEEE-NEMS), Bangkok, Thailand, January 16-19 (2007) in-press

3. M. Zolgharni, B. J. Jones, R. Bulpett, A. W. Anson, J. Franks

Tribological behaviour and surface properties of diamond-like carbon for efficiency improvements of coated drill bits, International Journal of Machine Tools and Manufacture (2006) submitted

4. M. Zolgharni, S. M. Azimi, M. R. Bahmanyar, W. Balachandran

A Microfluidic Mixer for Chaotic Mixing of Magnetic Particles, 10th Annual NSTI Nanotechnology Conference, Santa Clara, California, May 20-24 (2007) accepted

5. S. M. Azimi, M. R. Bahmanyar, M. Zolgharni, W. Balachandran

Numerical Investigation of a Magnetic Sensor for DNA Hybridization Detection Using Planar Transformers, Electronics Letters (IEE) (2006) submitted


An Inductance-based Sensor for DNA Hybridization Detection, 2nd Annual IEEE Int. Conf. on Nano/Micro Engineered and Molecular Systems (IEEE-NEMS), Bangkok, Thailand, January 16-19 (2007) in-press


Using Spiral Inductors for Detecting Hybridization of DNAs Labeled with Magnetic Beads, 10th Annual NSTI Nanotechnology Conference, Santa Clara, California, May 20-24 (2007) accepted

PATENT

M. Zolgharni, W. Balachandran “ Microfluidic device for extraction of biological cells” U.S Patent Disclosure, Brunel University, Jan. 2007. (Patent Pending)

massoud thesis

News & Politics